From 2d198f640ed4627cf6f12d03589ccc8f16b2722e Mon Sep 17 00:00:00 2001
From: fenglv <fenglv15@mails.ucas.ac.cn>
Date: Wed, 6 Nov 2019 18:34:13 +0800
Subject: [PATCH 001/504] add simhash and minhash

---
 dbms/src/Functions/ExtractString.h            | 187 ++++++
 dbms/src/Functions/FunctionsStringHash.cpp    | 585 ++++++++++++++++++
 dbms/src/Functions/FunctionsStringHash.h      | 124 ++++
 .../Functions/FunctionsStringSimilarity.cpp   | 158 +----
 4 files changed, 919 insertions(+), 135 deletions(-)
 create mode 100644 dbms/src/Functions/ExtractString.h
 create mode 100644 dbms/src/Functions/FunctionsStringHash.cpp
 create mode 100644 dbms/src/Functions/FunctionsStringHash.h

diff --git a/dbms/src/Functions/ExtractString.h b/dbms/src/Functions/ExtractString.h
new file mode 100644
index 00000000000..05566496cba
--- /dev/null
+++ b/dbms/src/Functions/ExtractString.h
@@ -0,0 +1,187 @@
+#include <Common/UTF8Helpers.h>
+
+#include <algorithm>
+#include <climits>
+#include <cstring>
+#include <memory>
+#include <utility>
+
+#ifdef __SSE4_2__
+#    include <nmmintrin.h>
+#endif
+
+namespace DB
+{
+//used by FunctionsStringSimilarity and FunctionsStringHash
+//includes exacting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
+template <size_t N, bool CaseInsensitive>
+struct ExtractStringImpl
+{
+    static constexpr size_t default_padding = 16;
+
+    static ALWAYS_INLINE size_t readASCIICodePoints(UInt8 * code_points, const char *& pos, const char * end)
+    {
+        /// Offset before which we copy some data.
+        constexpr size_t padding_offset = default_padding - N + 1;
+        /// We have an array like this for ASCII (N == 4, other cases are similar)
+        /// |a0|a1|a2|a3|a4|a5|a6|a7|a8|a9|a10|a11|a12|a13|a14|a15|a16|a17|a18|
+        /// And we copy                                ^^^^^^^^^^^^^^^ these bytes to the start
+        /// Actually it is enough to copy 3 bytes, but memcpy for 4 bytes translates into 1 instruction
+        memcpy(code_points, code_points + padding_offset, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt8));
+        /// Now we have an array
+        /// |a13|a14|a15|a16|a4|a5|a6|a7|a8|a9|a10|a11|a12|a13|a14|a15|a16|a17|a18|
+        ///              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+        /// Doing unaligned read of 16 bytes and copy them like above
+        /// 16 is also chosen to do two `movups`.
+        /// Such copying allow us to have 3 codepoints from the previous read to produce the 4-grams with them.
+        memcpy(code_points + (N - 1), pos, default_padding * sizeof(UInt8));
+
+        if constexpr (CaseInsensitive)
+        {
+            /// We really need template lambdas with C++20 to do it inline
+            unrollLowering<N - 1>(code_points, std::make_index_sequence<padding_offset>());
+        }
+        pos += padding_offset;
+        if (pos > end)
+            return default_padding - (pos - end);
+        return default_padding;
+    }
+
+    //used by FunctionsStringHash
+    //it's not easy to add padding for ColumnString, so we need safety check each memcpy
+    static ALWAYS_INLINE size_t readASCIICodePointsNoPadding(UInt8 * code_points, const char *& pos, const char * end)
+    {
+        constexpr size_t padding_offset = default_padding - N + 1;
+        memcpy(code_points, code_points + padding_offset, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt8));
+
+        //safety check
+        size_t cpy_size = (pos + padding_offset > end) ? end - pos : padding_offset;
+
+        memcpy(code_points + (N - 1), pos, cpy_size * sizeof(UInt8));
+
+        if constexpr (CaseInsensitive)
+        {
+            unrollLowering<N - 1>(code_points, std::make_index_sequence<padding_offset>());
+        }
+        pos += padding_offset;
+        if (pos > end)
+            return default_padding - (pos - end);
+        return default_padding;
+    }
+
+    //read a ASCII word from pos to word
+    //if the word size exceeds max_word_size, only read max_word_size byte
+    //in  FuntionsStringHash, the default value of max_word_size is 128
+    static ALWAYS_INLINE inline size_t readOneASCIIWord(UInt8 * word, const char *& pos, const char * end, const size_t & max_word_size)
+    {
+        //jump seperators
+        while (pos < end && !isAlphaNum(*pos))
+            ++pos;
+
+        // word start from here
+        const char * word_start = pos;
+        while (pos < end && isAlphaNum(*pos))
+            ++pos;
+
+        size_t word_size = (static_cast<size_t>(pos - word_start) <= max_word_size) ? pos - word_start : max_word_size;
+
+        memcpy(word, word_start, word_size);
+        if (CaseInsensitive)
+        {
+            std::transform(word, word + word_size, word, [](UInt8 c) { return std::tolower(c); });
+        }
+        return word_size;
+    }
+
+    static ALWAYS_INLINE inline size_t readUTF8CodePoints(UInt32 * code_points, const char *& pos, const char * end)
+    {
+        memcpy(code_points, code_points + default_padding - N + 1, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt32));
+
+        size_t num = N - 1;
+        while (num < default_padding && pos < end)
+        {
+            code_points[num++] = readOneUTF8Code(pos, end);
+        }
+        return num;
+    }
+
+    //read one UTF8 word from pos to word
+    //also, we assume that one word size cann't exceed max_word_size with default value 128
+    static ALWAYS_INLINE inline size_t readOneUTF8Word(UInt32 * word, const char *& pos, const char * end, const size_t & max_word_size)
+    {
+        // jump UTF8 seperator
+        while (pos < end && isUTF8Sep(*pos))
+            ++pos;
+        //UTF8 word's character number
+        size_t num = 0;
+        while (pos < end && num < max_word_size && !isUTF8Sep(*pos))
+        {
+            word[num++] = readOneUTF8Code(pos, end);
+        }
+        return num;
+    }
+
+private:
+    static ALWAYS_INLINE inline bool isAlphaNum(const UInt8 c)
+    {
+        return (c >= 48 && c <= 57) || (c >= 65 && c <= 90) || (c >= 97 && c <= 122);
+    }
+
+    template <size_t Offset, typename Container, size_t... I>
+    static ALWAYS_INLINE inline void unrollLowering(Container & cont, const std::index_sequence<I...> &)
+    {
+        ((cont[Offset + I] = std::tolower(cont[Offset + I])), ...);
+    }
+
+    //we use ASCII non-alphanum character as UTF8 seperator
+    static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNum(c); }
+
+    // read one UTF8 character and return it
+    static ALWAYS_INLINE inline UInt32 readOneUTF8Code(const char *& pos, const char * end)
+    {
+        size_t length = UTF8::seqLength(*pos);
+
+        if (pos + length > end)
+            length = end - pos;
+        UInt32 res;
+        switch (length)
+        {
+            case 1:
+                res = 0;
+                memcpy(&res, pos, 1);
+                break;
+            case 2:
+                res = 0;
+                memcpy(&res, pos, 2);
+                break;
+            case 3:
+                res = 0;
+                memcpy(&res, pos, 3);
+                break;
+            default:
+                memcpy(&res, pos, 4);
+        }
+
+        if constexpr (CaseInsensitive)
+        {
+            switch (length)
+            {
+                case 4:
+                    res &= ~(1u << (5 + 3 * CHAR_BIT));
+                    [[fallthrough]];
+                case 3:
+                    res &= ~(1u << (5 + 2 * CHAR_BIT));
+                    [[fallthrough]];
+                case 2:
+                    res &= ~(1u);
+                    res &= ~(1u << (5 + CHAR_BIT));
+                    [[fallthrough]];
+                default:
+                    res &= ~(1u << 5);
+            }
+        }
+        pos += length;
+        return res;
+    }
+};
+}
diff --git a/dbms/src/Functions/FunctionsStringHash.cpp b/dbms/src/Functions/FunctionsStringHash.cpp
new file mode 100644
index 00000000000..797d7d30078
--- /dev/null
+++ b/dbms/src/Functions/FunctionsStringHash.cpp
@@ -0,0 +1,585 @@
+#include <Functions/FunctionsStringHash.h>
+
+#include <Functions/ExtractString.h>
+#include <Functions/FunctionFactory.h>
+#include <Functions/FunctionsHashing.h>
+#include <Common/HashTable/ClearableHashMap.h>
+#include <Common/HashTable/Hash.h>
+
+#include <Core/Defines.h>
+
+#include <bitset>
+#include <tuple>
+#include <common/unaligned.h>
+
+namespace DB
+{
+struct Hash
+{
+    static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points)
+    {
+        return intHashCRC32(unalignedLoad<UInt32>(code_points));
+    }
+
+    static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points)
+    {
+        UInt64 combined = (static_cast<UInt64>(code_points[0]) << 32) | code_points[1];
+#ifdef __SSE4_2__
+        return _mm_crc32_u64(code_points[2], combined);
+#else
+        return (intHashCRC32(combined) ^ intHashCRC32(code_points[2]));
+#endif
+    }
+
+    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, const size_t & size, const size_t & offset)
+    {
+        UInt64 res = 0;
+        UInt8 flag = 0;
+        for (size_t i = offset; i < size; ++i)
+        {
+            if (flag)
+                res &= intHashCRC32(hashes[i]);
+            else
+                res |= intHashCRC32(hashes[i]);
+            flag = (flag + 1) % 2;
+        }
+        for (size_t i = 0; i < offset; ++i)
+        {
+            if (flag)
+                res &= intHashCRC32(hashes[i]);
+            else
+                res |= intHashCRC32(hashes[i]);
+            flag = (flag + 1) % 2;
+        }
+        return res;
+    }
+
+    template <typename CodePoint>
+    static ALWAYS_INLINE inline UInt64 hashSum(const CodePoint * hashes, const size_t & K)
+    {
+        UInt64 even = 0;
+        UInt64 odd = 0;
+        size_t i = 0;
+        for (; i + 1 < K; i += 2)
+        {
+            even |= intHashCRC32(hashes[i]);
+            odd |= intHashCRC32(hashes[i + 1]);
+        }
+        if (i < K)
+            even |= intHashCRC32(hashes[K - 1]);
+#ifdef __SSE4_2__
+        return _mm_crc32_u64(even, odd);
+#else
+        return (intHashCRC32(even) ^ intHashCRC32(odd));
+#endif
+    }
+};
+
+//Sinhash String -> UInt64
+template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
+struct SimhashImpl
+{
+    using ResultType = UInt64;
+    using StrOp = ExtractStringImpl<N, CaseInsensitive>;
+    // we made an assumption that the size of one word cann't exceed 128, which may not true
+    // if some word's size exceed 128, it would be cut up to several word
+    static constexpr size_t max_word_size = 1u << 7;
+    static constexpr size_t max_string_size = 1u << 15;
+    static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
+
+    // Simhash ngram calculate function: String ->UInt64
+    // this function extracting ngram from input string, and maintain a 64-dimensions vector
+    // for each ngram, calculate a 64 bit hash value, and update the vector according the hash value
+    // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
+    static ALWAYS_INLINE inline UInt64 ngramCalculateHashValue(
+        const char * data,
+        const size_t size,
+        size_t (*read_code_points)(CodePoint *, const char *&, const char *),
+        UInt64 (*hash_functor)(const CodePoint *))
+    {
+        const char * start = data;
+        const char * end = data + size;
+        // fingerprint vector, all dimensions initialized to zero at the first
+        Int64 finger_vec[64] = {};
+        CodePoint cp[simultaneously_codepoints_num] = {};
+
+        size_t found = read_code_points(cp, start, end);
+        size_t iter = N - 1;
+
+        do
+        {
+            for (; iter + N <= found; ++iter)
+            {
+                // for each ngram, we can calculate an 64 bit hash
+                // then update finger_vec according to this hash value
+                // if the i'th bit is 1, finger_vec[i] plus 1, otherwise minus 1
+                UInt64 hash_value = hash_functor(cp + iter);
+                std::bitset<64> bits(hash_value);
+                for (size_t i = 0; i < 64; ++i)
+                {
+                    finger_vec[i] += ((bits.test(i)) ? 1 : -1);
+                }
+            }
+            iter = 0;
+        } while (start < end && (found = read_code_points(cp, start, end)));
+
+        //finally, we return a 64 bit value according to finger_vec
+        //if finger_vec[i] > 0, the i'th bit of the value is 1, otherwise 0
+        std::bitset<64> res_bit(0u);
+        for (size_t i = 0; i < 64; ++i)
+        {
+            if (finger_vec[i] > 0)
+                res_bit.set(i);
+        }
+        return res_bit.to_ullong();
+    }
+
+    // Simhash word shingle calculate funtion: String -> UInt64
+    // this function extracting n word shingle from input string, and maintain a 64-dimensions vector as well
+    // for each word shingle, calculate a 64 bit hash value, and update the vector according the hash value
+    // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
+    //
+    // word shingle hash value calculate:
+    // 1. at the first, extracts N word shingles and calculate N hash values, store into an array, use this N hash values
+    // to calculate the first word shingle hash value
+    // 2. next, we extrac one word each time, and calculate a new hash value of the new word,then use the latest N hash
+    // values to caculate the next word shingle hash value
+    static ALWAYS_INLINE inline UInt64 wordShinglesCalculateHashValue(
+        const char * data,
+        const size_t size,
+        size_t (*read_one_word)(CodePoint *, const char *&, const char *, const size_t &),
+        UInt64 (*hash_functor)(const UInt64 *, const size_t &, const size_t &))
+    {
+        const char * start = data;
+        const char * end = data + size;
+
+        // Also, a 64 bit vector initialized to zero
+        Int64 finger_vec[64] = {};
+        // a array to store N word hash values
+        UInt64 nwordHashes[N] = {};
+        // word buffer to store one word
+        CodePoint word_buf[max_word_size] = {};
+        size_t word_size;
+        //get first word shingle
+        for (size_t i = 0; i < N && start < end; ++i)
+        {
+            word_size = read_one_word(word_buf, start, end, max_word_size);
+            if (word_size)
+            {
+                // for each word, calculate a hash value and stored into the array
+                nwordHashes[i++] = Hash::hashSum(word_buf, word_size);
+            }
+        }
+
+        // calculate the first word shingle hash value
+        UInt64 hash_value = hash_functor(nwordHashes, N, 0);
+        std::bitset<64> bits_(hash_value);
+        for (size_t i = 0; i < 64; ++i)
+        {
+            finger_vec[i] += ((bits_.test(i)) ? 1 : -1);
+        }
+
+        size_t offset = 0;
+        while (start < end && (word_size = read_one_word(word_buf, start, end, max_word_size)))
+        {
+            // we need to store the new word hash value to the oldest location.
+            // for example, N = 5, array |a0|a1|a2|a3|a4|, now , a0 is the oldest location,
+            // so we need to store new word hash into location of a0, then ,this array become
+            // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
+            // word hash value into locaion of a1, then array become |a5|a6|a2|a3|a4|
+            nwordHashes[offset] = Hash::hashSum(word_buf, word_size);
+            offset = (offset + 1) % N;
+            //according to the word hash storation way, in order to not lose the word shingle's
+            //sequence information, when calculation word shingle hash value, we need provide the offset
+            //inforation, which is the offset of the first word's hash value of the word shingle
+            hash_value = hash_functor(nwordHashes, N, offset);
+            std::bitset<64> bits(hash_value);
+            for (size_t i = 0; i < 64; ++i)
+            {
+                finger_vec[i] += ((bits.test(i)) ? 1 : -1);
+            }
+        }
+
+        std::bitset<64> res_bit(0u);
+        for (size_t i = 0; i < 64; ++i)
+        {
+            if (finger_vec[i] > 0)
+                res_bit.set(i);
+        }
+        return res_bit.to_ullong();
+    }
+
+    template <typename CalcFunc, typename... Args>
+    static ALWAYS_INLINE inline auto dispatch(CalcFunc calc_func, Args &&... args)
+    {
+        if constexpr (Ngram)
+        {
+            if constexpr (!UTF8)
+                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePointsNoPadding, Hash::ngramASCIIHash);
+            else
+                return calc_func(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
+        }
+        else
+        {
+            if constexpr (!UTF8)
+                return calc_func(std::forward<Args>(args)..., StrOp::readOneASCIIWord, Hash::wordShinglesHash);
+            else
+                return calc_func(std::forward<Args>(args)..., StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+        }
+    }
+
+    // constant string
+    static inline void constant(const String data, UInt64 & res)
+    {
+        if constexpr (Ngram)
+            res = dispatch(ngramCalculateHashValue, data.data(), data.size());
+        else
+            res = dispatch(wordShinglesCalculateHashValue, data.data(), data.size());
+    }
+
+    //non-constant string
+    static void vector(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, PaddedPODArray<UInt64> & res)
+    {
+        for (size_t i = 0; i < offsets.size(); ++i)
+        {
+            const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
+            const size_t data_size = offsets[i] - offsets[i - 1] - 1;
+            if (data_size <= max_string_size)
+            {
+                if constexpr (Ngram)
+                    res[i] = dispatch(ngramCalculateHashValue, one_data, data_size);
+                else
+                    res[i] = dispatch(wordShinglesCalculateHashValue, one_data, data_size);
+            }
+        }
+    }
+};
+
+//Minhash: String -> Tuple(UInt64, UInt64)
+//for each string, we extract ngram or word shingle,
+//for each ngram or word shingle, calculate a hash value,
+//then we take the K minimum hash values to calculate a hashsum,
+//and take the K maximum hash values to calculate another hashsum,
+//return this two hashsum: Tuple(hashsum1, hashsum2)
+template <size_t N, size_t K, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
+struct MinhashImpl
+{
+    using ResultType = UInt64;
+    using StrOp = ExtractStringImpl<N, CaseInsensitive>;
+    static constexpr size_t max_word_size = 1u << 7;
+    static constexpr size_t max_string_size = 1u << 15;
+    static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
+
+    // insert a new value into K minimum hash array if this value
+    // is smaller than the greatest value in the array
+    static ALWAYS_INLINE inline void insert_minValue(UInt64 * hashes, UInt64 v)
+    {
+        size_t i = 0;
+        for (; i < K && hashes[i] <= v; ++i)
+            ;
+        if (i == K)
+            return;
+        for (size_t j = K - 2; j >= i; --j)
+            hashes[j + 1] = hashes[j];
+        hashes[i] = v;
+    }
+
+    // insert a new value into K maximum hash array if this value
+    // is greater than the smallest value in the array
+    static ALWAYS_INLINE inline void insert_maxValue(UInt64 * hashes, UInt64 v)
+    {
+        int i = K - 1;
+        for (; i >= 0 && hashes[i] >= v; --i)
+            ;
+        if (i < 0)
+            return;
+        for (int j = 1; j <= i; ++j)
+            hashes[j - 1] = hashes[j];
+        hashes[i] = v;
+    }
+
+    //Minhash ngram calculate function, String -> Tuple(UInt64, UInt64)
+    //we extract ngram from input string, and calculate a hash value for each ngram
+    //then we take the K minimum hash values to calculate a hashsum,
+    //and take the K maximum hash values to calculate another hashsum,
+    //return this two hashsum: Tuple(hashsum1, hashsum2)
+    static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> ngramCalculateHashValue(
+        const char * data,
+        const size_t size,
+        size_t (*read_code_points)(CodePoint *, const char *&, const char *),
+        UInt64 (*hash_functor)(const CodePoint *))
+    {
+        const char * start = data;
+        const char * end = data + size;
+        // we just maintain the K minimu and K maximum hash values
+        UInt64 k_minimum[K] = {};
+        UInt64 k_maxinum[K] = {};
+        CodePoint cp[simultaneously_codepoints_num] = {};
+
+        size_t found = read_code_points(cp, start, end);
+        size_t iter = N - 1;
+
+        do
+        {
+            for (; iter + N <= found; ++iter)
+            {
+                auto new_hash = hash_functor(cp + iter);
+                // insert the new hash value into array used to store K minimum value
+                // and K maximum value
+                insert_minValue(k_minimum, new_hash);
+                insert_maxValue(k_maxinum, new_hash);
+            }
+            iter = 0;
+        } while (start < end && (found = read_code_points(cp, start, end)));
+
+        // calculate hashsum of the K minimum hash values and K maximum hash values
+        UInt64 res1 = Hash::hashSum(k_maxinum, K);
+        UInt64 res2 = Hash::hashSum(k_maxinum, K);
+        return std::make_tuple(res1, res2);
+    }
+
+    // Minhash word shingle hash value calculate function: String ->Tuple(UInt64, UInt64)
+    //for each word shingle, we calculate a hash value, but in fact, we just maintain the
+    //K minimum and K maximum hash value
+    static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> wordShinglesCalculateHashValue(
+        const char * data,
+        const size_t size,
+        size_t (*read_one_word)(CodePoint *, const char *&, const char *, const size_t &),
+        UInt64 (*hash_functor)(const UInt64 *, const size_t &, const size_t &))
+    {
+        const char * start = data;
+        const char * end = start + size;
+        //also we just store the K minimu and K maximum hash values
+        UInt64 k_minimum[K] = {};
+        UInt64 k_maxinum[K] = {};
+        // array to store n word hashes
+        UInt64 nwordHashes[N] = {};
+        // word buffer to store one word
+        CodePoint word_buf[max_word_size] = {};
+        size_t word_size;
+        //how word shingle hash value calculation and word hash storation is same as we
+        //have descripted in Simhash wordShinglesCalculateHashValue function
+        for (size_t i = 0; i < N && start < end; ++i)
+        {
+            word_size = read_one_word(word_buf, start, end, max_word_size);
+            if (word_size)
+            {
+                nwordHashes[i++] = Hash::hashSum(word_buf, word_size);
+            }
+        }
+
+        auto new_hash = hash_functor(nwordHashes, N, 0);
+        insert_minValue(k_minimum, new_hash);
+        insert_maxValue(k_maxinum, new_hash);
+
+        size_t offset = 0;
+        while (start < end && (word_size = read_one_word(word_buf, start, end, max_word_size)))
+        {
+            nwordHashes[offset] = Hash::hashSum(word_buf, word_size);
+            offset = (offset + 1) % N;
+            new_hash = hash_functor(nwordHashes, N, offset);
+            insert_minValue(k_minimum, new_hash);
+            insert_maxValue(k_maxinum, new_hash);
+        }
+
+        // calculate hashsum
+        UInt64 res1 = Hash::hashSum(k_minimum, K);
+        UInt64 res2 = Hash::hashSum(k_maxinum, K);
+        return std::make_tuple(res1, res2);
+    }
+
+    template <typename CalcFunc, typename... Args>
+    static ALWAYS_INLINE inline auto dispatch(CalcFunc calc_func, Args &&... args)
+    {
+        if constexpr (Ngram)
+        {
+            if constexpr (!UTF8)
+                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePointsNoPadding, Hash::ngramASCIIHash);
+            else
+                return calc_func(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
+        }
+        else
+        {
+            if constexpr (!UTF8)
+                return calc_func(std::forward<Args>(args)..., StrOp::readOneASCIIWord, Hash::wordShinglesHash);
+            else
+                return calc_func(std::forward<Args>(args)..., StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+        }
+    }
+
+    // constant string
+    static void constant(const String data, UInt64 & res1, UInt64 & res2)
+    {
+        if constexpr (Ngram)
+            std::tie(res1, res2) = dispatch(ngramCalculateHashValue, data.data(), data.size());
+        else
+            std::tie(res1, res2) = dispatch(wordShinglesCalculateHashValue, data.data(), data.size());
+    }
+
+    //non-constant string
+    static void vector(
+        const ColumnString::Chars & data,
+        const ColumnString::Offsets & offsets,
+        PaddedPODArray<UInt64> & res1,
+        PaddedPODArray<UInt64> & res2)
+    {
+        for (size_t i = 0; i < offsets.size(); ++i)
+        {
+            const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
+            const size_t data_size = offsets[i] - offsets[i - 1] - 1;
+            if (data_size <= max_string_size)
+            {
+                if constexpr (Ngram)
+                    std::tie(res1[i], res2[i]) = dispatch(ngramCalculateHashValue, one_data, data_size);
+                else
+                    std::tie(res1[i], res2[i]) = dispatch(wordShinglesCalculateHashValue, one_data, data_size);
+            }
+        }
+    }
+};
+
+struct NameNgramSimhash
+{
+    static constexpr auto name = "ngramSimhash";
+};
+
+struct NameNgramSimhashCaseInsensitive
+{
+    static constexpr auto name = "ngramSimhashCaseInsensitive";
+};
+
+struct NameNgramSimhashUTF8
+{
+    static constexpr auto name = "ngramSimhashUTF8";
+};
+
+struct NameNgramSimhashCaseInsensitiveUTF8
+{
+    static constexpr auto name = "ngramSimhashCaseInsensitiveUTF8";
+};
+
+struct NameWordShingleSimhash
+{
+    static constexpr auto name = "wordShingleSimhash";
+};
+
+struct NameWordShingleSimhashCaseInsensitive
+{
+    static constexpr auto name = "wordShingleSimhashCaseInsensitive";
+};
+
+struct NameWordShingleSimhashUTF8
+{
+    static constexpr auto name = "wordShingleSimhashUTF8";
+};
+
+struct NameWordShingleSimhashCaseInsensitiveUTF8
+{
+    static constexpr auto name = "wordShingleSimhashCaseInsensitiveUTF8";
+};
+
+struct NameNgramMinhash
+{
+    static constexpr auto name = "ngramMinhash";
+};
+
+struct NameNgramMinhashCaseInsensitive
+{
+    static constexpr auto name = "ngramMinhashCaseInsensitive";
+};
+
+struct NameNgramMinhashUTF8
+{
+    static constexpr auto name = "ngramMinhashUTF8";
+};
+
+struct NameNgramMinhashCaseInsensitiveUTF8
+{
+    static constexpr auto name = "ngramMinhashCaseInsensitiveUTF8";
+};
+
+struct NameWordShingleMinhash
+{
+    static constexpr auto name = "wordShingleMinhash";
+};
+
+struct NameWordShingleMinhashCaseInsensitive
+{
+    static constexpr auto name = "wordShingleMinhashCaseInsensitive";
+};
+
+struct NameWordShingleMinhashUTF8
+{
+    static constexpr auto name = "wordShingleMinhashUTF8";
+};
+
+struct NameWordShingleMinhashCaseInsensitiveUTF8
+{
+    static constexpr auto name = "wordShingleMinhashCaseInsensitiveUTF8";
+};
+
+//Simhash
+using FunctionNgramSimhash = FunctionsStringHash<SimhashImpl<4, UInt8, false, true, false>, NameNgramSimhash, true>;
+
+using FunctionNgramSimhashCaseInsensitive
+    = FunctionsStringHash<SimhashImpl<4, UInt8, false, true, true>, NameNgramSimhashCaseInsensitive, true>;
+
+using FunctionNgramSimhashUTF8 = FunctionsStringHash<SimhashImpl<3, UInt32, true, true, false>, NameNgramSimhashUTF8, true>;
+
+using FunctionNgramSimhashCaseInsensitiveUTF8
+    = FunctionsStringHash<SimhashImpl<3, UInt32, true, true, true>, NameNgramSimhashCaseInsensitiveUTF8, true>;
+
+using FunctionWordShingleSimhash = FunctionsStringHash<SimhashImpl<3, UInt8, false, false, false>, NameWordShingleSimhash, true>;
+
+using FunctionWordShingleSimhashCaseInsensitive
+    = FunctionsStringHash<SimhashImpl<3, UInt8, false, false, true>, NameWordShingleSimhashCaseInsensitive, true>;
+
+using FunctionWordShingleSimhashUTF8 = FunctionsStringHash<SimhashImpl<3, UInt32, true, false, false>, NameWordShingleSimhashUTF8, true>;
+
+using FunctionWordShingleSimhashCaseInsensitiveUTF8
+    = FunctionsStringHash<SimhashImpl<3, UInt32, true, false, true>, NameWordShingleSimhashCaseInsensitiveUTF8, true>;
+
+//Minhash
+using FunctionNgramMinhash = FunctionsStringHash<MinhashImpl<4, 6, UInt8, false, true, false>, NameNgramMinhash, false>;
+
+using FunctionNgramMinhashCaseInsensitive
+    = FunctionsStringHash<MinhashImpl<4, 6, UInt8, false, true, true>, NameNgramMinhashCaseInsensitive, false>;
+
+using FunctionNgramMinhashUTF8 = FunctionsStringHash<MinhashImpl<4, 6, UInt32, true, true, false>, NameNgramMinhashUTF8, false>;
+
+using FunctionNgramMinhashCaseInsensitiveUTF8
+    = FunctionsStringHash<MinhashImpl<4, 6, UInt32, true, true, true>, NameNgramMinhashCaseInsensitiveUTF8, false>;
+
+using FunctionWordShingleMinhash = FunctionsStringHash<MinhashImpl<3, 6, UInt8, false, false, false>, NameWordShingleMinhash, false>;
+
+using FunctionWordShingleMinhashCaseInsensitive
+    = FunctionsStringHash<MinhashImpl<3, 6, UInt8, false, false, true>, NameWordShingleMinhashCaseInsensitive, false>;
+
+using FunctionWordShingleMinhashUTF8
+    = FunctionsStringHash<MinhashImpl<3, 6, UInt32, true, false, false>, NameWordShingleMinhashUTF8, false>;
+
+using FunctionWordShingleMinhashCaseInsensitiveUTF8
+    = FunctionsStringHash<MinhashImpl<3, 6, UInt32, true, false, true>, NameWordShingleMinhashCaseInsensitiveUTF8, false>;
+
+void registerFunctionsStringHash(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionNgramSimhash>();
+    factory.registerFunction<FunctionNgramSimhashCaseInsensitive>();
+    factory.registerFunction<FunctionNgramSimhashUTF8>();
+    factory.registerFunction<FunctionNgramSimhashCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionWordShingleSimhash>();
+    factory.registerFunction<FunctionWordShingleSimhashCaseInsensitive>();
+    factory.registerFunction<FunctionWordShingleSimhashUTF8>();
+    factory.registerFunction<FunctionWordShingleSimhashCaseInsensitiveUTF8>();
+
+    factory.registerFunction<FunctionNgramMinhash>();
+    factory.registerFunction<FunctionNgramMinhashCaseInsensitive>();
+    factory.registerFunction<FunctionNgramMinhashUTF8>();
+    factory.registerFunction<FunctionNgramMinhashCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionWordShingleMinhash>();
+    factory.registerFunction<FunctionWordShingleMinhashCaseInsensitive>();
+    factory.registerFunction<FunctionWordShingleMinhashUTF8>();
+    factory.registerFunction<FunctionWordShingleMinhashCaseInsensitiveUTF8>();
+}
+}
+
diff --git a/dbms/src/Functions/FunctionsStringHash.h b/dbms/src/Functions/FunctionsStringHash.h
new file mode 100644
index 00000000000..185097ade99
--- /dev/null
+++ b/dbms/src/Functions/FunctionsStringHash.h
@@ -0,0 +1,124 @@
+#pragma once
+
+#include <Columns/ColumnConst.h>
+#include <Columns/ColumnString.h>
+#include <Columns/ColumnTuple.h>
+#include <Columns/ColumnVector.h>
+#include <DataTypes/DataTypeFactory.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Functions/FunctionHelpers.h>
+#include <Functions/IFunction.h>
+
+namespace DB
+{
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+    extern const int ILLEGAL_COLUMN;
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+    extern const int TOO_LARGE_STRING_SIZE;
+}
+
+//FunctionStringHash
+//Simhash: String -> UInt64
+//Minhash: String -> (UInt64, UInt64)
+template <typename Impl, typename Name, bool IsSimhash>
+class FunctionsStringHash : public IFunction
+{
+public:
+    static constexpr auto name = Name::name;
+
+    static FunctionPtr create(const Context &) { return std::make_shared<FunctionsStringHash>(); }
+
+    String getName() const override { return name; }
+
+    size_t getNumberOfArguments() const override { return 1; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (!isString(arguments[0]))
+            throw Exception(
+                "Illegal type " + arguments[0]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        if (IsSimhash)
+            return std::make_shared<DataTypeNumber<typename Impl::ResultType>>();
+        auto element = DataTypeFactory::instance().get("UInt64");
+        return std::make_shared<DataTypeTuple>(DataTypes{element, element});
+    }
+
+    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
+    {
+        const ColumnPtr & column = block.getByPosition(arguments[0]).column;
+        const ColumnConst * col_const = typeid_cast<const ColumnConst *>(&*column);
+        using ResultType = typename Impl::ResultType;
+        if constexpr (IsSimhash)
+        {
+            if (col_const)
+            {
+                ResultType res{};
+                const String & str_data = col_const->getValue<String>();
+                if (str_data.size() > Impl::max_string_size)
+                {
+                    throw Exception(
+                        "String size is too big for function " + getName() + ". Should be at most " + std::to_string(Impl::max_string_size),
+                        ErrorCodes::TOO_LARGE_STRING_SIZE);
+                }
+                Impl::constant(str_data, res);
+                block.getByPosition(result).column = block.getByPosition(result).type->createColumnConst(1, toField(res));
+            }
+            else
+            {
+                // non const string
+                auto col_res = ColumnVector<ResultType>::create();
+                typename ColumnVector<ResultType>::Container & vec_res = col_res->getData();
+                vec_res.resize(column->size());
+                const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
+                Impl::vector(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_res);
+                block.getByPosition(result).column = std::move(col_res);
+            }
+        }
+        else // Min hash
+        {
+            if (col_const)
+            {
+                ResultType h1, h2;
+                const String & str_data = col_const->getValue<String>();
+                if (str_data.size() > Impl::max_string_size)
+                {
+                    throw Exception(
+                        "String size is too big for function " + getName() + ". Should be at most " + std::to_string(Impl::max_string_size),
+                        ErrorCodes::TOO_LARGE_STRING_SIZE);
+                }
+                Impl::constant(str_data, h1, h2);
+                auto h1_col = ColumnVector<ResultType>::create(1);
+                auto h2_col = ColumnVector<ResultType>::create(1);
+                typename ColumnVector<ResultType>::Container & h1_data = h1_col->getData();
+                typename ColumnVector<ResultType>::Container & h2_data = h2_col->getData();
+                h1_data[0] = h1;
+                h2_data[0] = h2;
+                MutableColumns tuple_columns;
+                tuple_columns.emplace_back(std::move(h1_col));
+                tuple_columns.emplace_back(std::move(h2_col));
+                block.getByPosition(result).column = ColumnTuple::create(std::move(tuple_columns));
+            }
+            else
+            {
+                //non const string
+                auto col_h1 = ColumnVector<ResultType>::create();
+                auto col_h2 = ColumnVector<ResultType>::create();
+                typename ColumnVector<ResultType>::Container & vec_h1 = col_h1->getData();
+                typename ColumnVector<ResultType>::Container & vec_h2 = col_h2->getData();
+                vec_h1.resize(column->size());
+                vec_h2.resize(column->size());
+                const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
+                Impl::vector(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_h1, vec_h2);
+                MutableColumns tuple_columns;
+                tuple_columns.emplace_back(std::move(col_h1));
+                tuple_columns.emplace_back(std::move(col_h2));
+                block.getByPosition(result).column = ColumnTuple::create(std::move(tuple_columns));
+            }
+        }
+    }
+};
+}
+
diff --git a/dbms/src/Functions/FunctionsStringSimilarity.cpp b/dbms/src/Functions/FunctionsStringSimilarity.cpp
index 9dda521cd29..c6327ad59b4 100644
--- a/dbms/src/Functions/FunctionsStringSimilarity.cpp
+++ b/dbms/src/Functions/FunctionsStringSimilarity.cpp
@@ -1,5 +1,6 @@
 #include <Functions/FunctionsStringSimilarity.h>
 
+#include <Functions/ExtractString.h>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionsHashing.h>
 #include <Common/HashTable/ClearableHashMap.h>
@@ -10,13 +11,6 @@
 
 #include <common/unaligned.h>
 
-#include <algorithm>
-#include <climits>
-#include <cstring>
-#include <limits>
-#include <memory>
-#include <utility>
-
 #ifdef __SSE4_2__
 #    include <nmmintrin.h>
 #endif
@@ -36,6 +30,7 @@ template <size_t N, class CodePoint, bool UTF8, bool case_insensitive, bool symm
 struct NgramDistanceImpl
 {
     using ResultType = Float32;
+    using StrOp = ExtractStringImpl<N, case_insensitive>;
 
     /// map_size for ngram difference.
     static constexpr size_t map_size = 1u << 16;
@@ -44,7 +39,7 @@ struct NgramDistanceImpl
     static constexpr size_t max_string_size = 1u << 15;
 
     /// Default padding to read safely.
-    static constexpr size_t default_padding = 16;
+    static constexpr size_t default_padding = StrOp::default_padding;
 
     /// Max codepoints to store at once. 16 is for batching usage and PODArray has this padding.
     static constexpr size_t simultaneously_codepoints_num = default_padding + N - 1;
@@ -70,102 +65,6 @@ struct NgramDistanceImpl
 #endif
     }
 
-    template <size_t Offset, class Container, size_t... I>
-    static ALWAYS_INLINE inline void unrollLowering(Container & cont, const std::index_sequence<I...> &)
-    {
-        ((cont[Offset + I] = std::tolower(cont[Offset + I])), ...);
-    }
-
-    static ALWAYS_INLINE size_t readASCIICodePoints(CodePoint * code_points, const char *& pos, const char * end)
-    {
-        /// Offset before which we copy some data.
-        constexpr size_t padding_offset = default_padding - N + 1;
-        /// We have an array like this for ASCII (N == 4, other cases are similar)
-        /// |a0|a1|a2|a3|a4|a5|a6|a7|a8|a9|a10|a11|a12|a13|a14|a15|a16|a17|a18|
-        /// And we copy                                ^^^^^^^^^^^^^^^ these bytes to the start
-        /// Actually it is enough to copy 3 bytes, but memcpy for 4 bytes translates into 1 instruction
-        memcpy(code_points, code_points + padding_offset, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(CodePoint));
-        /// Now we have an array
-        /// |a13|a14|a15|a16|a4|a5|a6|a7|a8|a9|a10|a11|a12|a13|a14|a15|a16|a17|a18|
-        ///              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-        /// Doing unaligned read of 16 bytes and copy them like above
-        /// 16 is also chosen to do two `movups`.
-        /// Such copying allow us to have 3 codepoints from the previous read to produce the 4-grams with them.
-        memcpy(code_points + (N - 1), pos, default_padding * sizeof(CodePoint));
-
-        if constexpr (case_insensitive)
-        {
-            /// We really need template lambdas with C++20 to do it inline
-            unrollLowering<N - 1>(code_points, std::make_index_sequence<padding_offset>());
-        }
-        pos += padding_offset;
-        if (pos > end)
-            return default_padding - (pos - end);
-        return default_padding;
-    }
-
-    static ALWAYS_INLINE size_t readUTF8CodePoints(CodePoint * code_points, const char *& pos, const char * end)
-    {
-        /// The same copying as described in the function above.
-        memcpy(code_points, code_points + default_padding - N + 1, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(CodePoint));
-
-        size_t num = N - 1;
-        while (num < default_padding && pos < end)
-        {
-            size_t length = UTF8::seqLength(*pos);
-
-            if (pos + length > end)
-                length = end - pos;
-
-            CodePoint res;
-            /// This is faster than just memcpy because of compiler optimizations with moving bytes.
-            switch (length)
-            {
-                case 1:
-                    res = 0;
-                    memcpy(&res, pos, 1);
-                    break;
-                case 2:
-                    res = 0;
-                    memcpy(&res, pos, 2);
-                    break;
-                case 3:
-                    res = 0;
-                    memcpy(&res, pos, 3);
-                    break;
-                default:
-                    memcpy(&res, pos, 4);
-            }
-
-            /// This is not a really true case insensitive utf8. We zero the 5-th bit of every byte.
-            /// And first bit of first byte if there are two bytes.
-            /// For ASCII it works https://catonmat.net/ascii-case-conversion-trick. For most cyrrilic letters also does.
-            /// For others, we don't care now. Lowering UTF is not a cheap operation.
-            if constexpr (case_insensitive)
-            {
-                switch (length)
-                {
-                    case 4:
-                        res &= ~(1u << (5 + 3 * CHAR_BIT));
-                        [[fallthrough]];
-                    case 3:
-                        res &= ~(1u << (5 + 2 * CHAR_BIT));
-                        [[fallthrough]];
-                    case 2:
-                        res &= ~(1u);
-                        res &= ~(1u << (5 + CHAR_BIT));
-                        [[fallthrough]];
-                    default:
-                        res &= ~(1u << 5);
-                }
-            }
-
-            pos += length;
-            code_points[num++] = res;
-        }
-        return num;
-    }
-
     template <bool save_ngrams>
     static ALWAYS_INLINE inline size_t calculateNeedleStats(
         const char * data,
@@ -250,9 +149,9 @@ struct NgramDistanceImpl
     static inline auto dispatchSearcher(Callback callback, Args &&... args)
     {
         if constexpr (!UTF8)
-            return callback(std::forward<Args>(args)..., readASCIICodePoints, ASCIIHash);
+            return callback(std::forward<Args>(args)..., StrOp::readASCIICodePoints, ASCIIHash);
         else
-            return callback(std::forward<Args>(args)..., readUTF8CodePoints, UTF8Hash);
+            return callback(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, UTF8Hash);
     }
 
     static void constant_constant(std::string data, std::string needle, Float32 & res)
@@ -269,7 +168,8 @@ struct NgramDistanceImpl
         size_t distance = second_size;
         if (data_size <= max_string_size)
         {
-            size_t first_size = dispatchSearcher(calculateHaystackStatsAndMetric<false>, data.data(), data_size, common_stats, distance, nullptr);
+            size_t first_size
+                = dispatchSearcher(calculateHaystackStatsAndMetric<false>, data.data(), data_size, common_stats, distance, nullptr);
             /// For !symmetric version we should not use first_size.
             if constexpr (symmetric)
                 res = distance * 1.f / std::max(first_size + second_size, size_t(1));
@@ -313,23 +213,14 @@ struct NgramDistanceImpl
             if (needle_size <= max_string_size && haystack_size <= max_string_size)
             {
                 /// Get needle stats.
-                const size_t needle_stats_size = dispatchSearcher(
-                    calculateNeedleStats<true>,
-                    needle,
-                    needle_size,
-                    common_stats,
-                    needle_ngram_storage.get());
+                const size_t needle_stats_size
+                    = dispatchSearcher(calculateNeedleStats<true>, needle, needle_size, common_stats, needle_ngram_storage.get());
 
                 size_t distance = needle_stats_size;
 
                 /// Combine with haystack stats, return to initial needle stats.
                 const size_t haystack_stats_size = dispatchSearcher(
-                    calculateHaystackStatsAndMetric<true>,
-                    haystack,
-                    haystack_size,
-                    common_stats,
-                    distance,
-                    haystack_ngram_storage.get());
+                    calculateHaystackStatsAndMetric<true>, haystack, haystack_size, common_stats, distance, haystack_ngram_storage.get());
 
                 /// Return to zero array stats.
                 for (size_t j = 0; j < needle_stats_size; ++j)
@@ -391,12 +282,8 @@ struct NgramDistanceImpl
 
                 if (needle_size <= max_string_size && haystack_size <= max_string_size)
                 {
-                    const size_t needle_stats_size = dispatchSearcher(
-                        calculateNeedleStats<true>,
-                        needle,
-                        needle_size,
-                        common_stats,
-                        needle_ngram_storage.get());
+                    const size_t needle_stats_size
+                        = dispatchSearcher(calculateNeedleStats<true>, needle, needle_size, common_stats, needle_ngram_storage.get());
 
                     size_t distance = needle_stats_size;
 
@@ -420,15 +307,11 @@ struct NgramDistanceImpl
 
                 prev_offset = needle_offsets[i];
             }
-
         }
     }
 
     static void vector_constant(
-        const ColumnString::Chars & data,
-        const ColumnString::Offsets & offsets,
-        std::string needle,
-        PaddedPODArray<Float32> & res)
+        const ColumnString::Chars & data, const ColumnString::Offsets & offsets, std::string needle, PaddedPODArray<Float32> & res)
     {
         /// zeroing our map
         NgramStats common_stats = {};
@@ -454,7 +337,8 @@ struct NgramDistanceImpl
                 size_t haystack_stats_size = dispatchSearcher(
                     calculateHaystackStatsAndMetric<true>,
                     reinterpret_cast<const char *>(haystack),
-                    haystack_size, common_stats,
+                    haystack_size,
+                    common_stats,
                     distance,
                     ngram_storage.get());
                 /// For !symmetric version we should not use haystack_stats_size.
@@ -516,14 +400,18 @@ struct NameNgramSearchUTF8CaseInsensitive
 };
 
 using FunctionNgramDistance = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, false, true>, NameNgramDistance>;
-using FunctionNgramDistanceCaseInsensitive = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, true>, NameNgramDistanceCaseInsensitive>;
+using FunctionNgramDistanceCaseInsensitive
+    = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, true>, NameNgramDistanceCaseInsensitive>;
 using FunctionNgramDistanceUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, false, true>, NameNgramDistanceUTF8>;
-using FunctionNgramDistanceCaseInsensitiveUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, true>, NameNgramDistanceUTF8CaseInsensitive>;
+using FunctionNgramDistanceCaseInsensitiveUTF8
+    = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, true>, NameNgramDistanceUTF8CaseInsensitive>;
 
 using FunctionNgramSearch = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, false, false>, NameNgramSearch>;
-using FunctionNgramSearchCaseInsensitive = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, false>, NameNgramSearchCaseInsensitive>;
+using FunctionNgramSearchCaseInsensitive
+    = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, false>, NameNgramSearchCaseInsensitive>;
 using FunctionNgramSearchUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, false, false>, NameNgramSearchUTF8>;
-using FunctionNgramSearchCaseInsensitiveUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, false>, NameNgramSearchUTF8CaseInsensitive>;
+using FunctionNgramSearchCaseInsensitiveUTF8
+    = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, false>, NameNgramSearchUTF8CaseInsensitive>;
 
 
 void registerFunctionsStringSimilarity(FunctionFactory & factory)

From e0cf07e958c77cf3b7f1faeb727ba3541ae00f18 Mon Sep 17 00:00:00 2001
From: fenglv <fenglv15@mails.ucas.ac.cn>
Date: Wed, 6 Nov 2019 18:35:23 +0800
Subject: [PATCH 002/504] add hammingdistance function

---
 dbms/src/Functions/bitHammingDistance.cpp     | 174 ++++++++++++++
 dbms/src/Functions/registerFunctions.cpp      |   2 +
 .../Functions/registerFunctionsArithmetic.cpp |   4 +
 dbms/src/Functions/tupleHammingDistance.cpp   | 224 ++++++++++++++++++
 4 files changed, 404 insertions(+)
 create mode 100644 dbms/src/Functions/bitHammingDistance.cpp
 create mode 100644 dbms/src/Functions/tupleHammingDistance.cpp

diff --git a/dbms/src/Functions/bitHammingDistance.cpp b/dbms/src/Functions/bitHammingDistance.cpp
new file mode 100644
index 00000000000..2572720bb4e
--- /dev/null
+++ b/dbms/src/Functions/bitHammingDistance.cpp
@@ -0,0 +1,174 @@
+#include <Columns/ColumnVector.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Functions/FunctionFactory.h>
+#include <Functions/FunctionHelpers.h>
+#include <Functions/IFunction.h>
+#include <Functions/castTypeToEither.h>
+
+namespace DB
+{
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_COLUMN;
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+}
+
+
+template <typename A, typename B>
+struct BitHammingDistanceImpl
+{
+    using ResultType = UInt8;
+
+    static void NO_INLINE vector_vector(const PaddedPODArray<A> & a, const PaddedPODArray<B> & b, PaddedPODArray<ResultType> & c)
+    {
+        size_t size = a.size();
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(a[i], b[i]);
+    }
+
+    static void NO_INLINE vector_constant(const PaddedPODArray<A> & a, B b, PaddedPODArray<ResultType> & c)
+    {
+        size_t size = a.size();
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(a[i], b);
+    }
+
+    static void NO_INLINE constant_vector(A a, const PaddedPODArray<B> & b, PaddedPODArray<ResultType> & c)
+    {
+        size_t size = b.size();
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(a, b[i]);
+    }
+
+    static ResultType constant_constant(A a, B b) { return apply(a, b); }
+
+private:
+    static UInt8 pop_cnt(UInt64 res)
+    {
+        UInt8 count = 0;
+        for (; res; res >>= 1)
+            count += res & 1u;
+        return count;
+    }
+
+    static inline UInt8 apply(UInt64 a, UInt64 b)
+    {
+        UInt64 res = a ^ b;
+        return pop_cnt(res);
+    }
+};
+
+template <typename F>
+bool castType(const IDataType * type, F && f)
+{
+    return castTypeToEither<
+        DataTypeInt8,
+        DataTypeInt16,
+        DataTypeInt32,
+        DataTypeInt64,
+        DataTypeUInt8,
+        DataTypeUInt16,
+        DataTypeUInt32,
+        DataTypeUInt64>(type, std::forward<F>(f));
+}
+
+template <typename F>
+static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
+{
+    return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+}
+
+//bitHammingDistance function: (Integer, Integer) -> UInt8
+class FunctionBitHammingDistance : public IFunction
+{
+public:
+    static constexpr auto name = "bitHammingDistance";
+    using ResultType = UInt8;
+    static FunctionPtr create(const Context &) { return std::make_shared<FunctionBitHammingDistance>(); }
+
+    String getName() const override { return name; }
+
+    size_t getNumberOfArguments() const override { return 2; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (!isInteger(arguments[0]))
+            throw Exception(
+                "Illegal type " + arguments[0]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        if (!isInteger(arguments[1]))
+            throw Exception(
+                "Illegal type " + arguments[1]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        return std::make_shared<DataTypeUInt8>();
+    }
+
+    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
+    {
+        auto * left_generic = block.getByPosition(arguments[0]).type.get();
+        auto * right_generic = block.getByPosition(arguments[1]).type.get();
+        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
+            using LeftDataType = std::decay_t<decltype(left)>;
+            using RightDataType = std::decay_t<decltype(right)>;
+            using T0 = typename LeftDataType::FieldType;
+            using T1 = typename RightDataType::FieldType;
+            using ColVecT0 = ColumnVector<T0>;
+            using ColVecT1 = ColumnVector<T1>;
+            using ColVecResult = ColumnVector<ResultType>;
+
+            using OpImpl = BitHammingDistanceImpl<T0, T1>;
+
+            auto col_left_raw = block.getByPosition(arguments[0]).column.get();
+            auto col_right_raw = block.getByPosition(arguments[1]).column.get();
+            if (auto col_left = checkAndGetColumnConst<ColVecT0>(col_left_raw))
+            {
+                if (auto col_right = checkAndGetColumnConst<ColVecT1>(col_right_raw))
+                {
+                    //constant integer - constant integer
+                    auto res = OpImpl::constant_constant(col_left->template getValue<T0>(), col_right->template getValue<T1>());
+                    block.getByPosition(result).column = DataTypeUInt8().createColumnConst(col_left->size(), toField(res));
+                    return true;
+                }
+            }
+
+            typename ColVecResult::MutablePtr col_res = nullptr;
+            col_res = ColVecResult::create();
+
+            auto & vec_res = col_res->getData();
+            vec_res.resize(block.rows());
+
+            if (auto col_left_const = checkAndGetColumnConst<ColVecT0>(col_left_raw))
+            {
+                if (auto col_right = checkAndGetColumn<ColVecT1>(col_right_raw))
+                {
+                    // constant integer - non-constant integer
+                    OpImpl::constant_vector(col_left_const->template getValue<T0>(), col_right->getData(), vec_res);
+                }
+                else
+                    return false;
+            }
+            else if (auto col_left = checkAndGetColumn<ColVecT0>(col_left_raw))
+            {
+                if (auto col_right = checkAndGetColumn<ColVecT1>(col_right_raw))
+                    //non-constant integer - non-constant integer
+                    OpImpl::vector_vector(col_left->getData(), col_right->getData(), vec_res);
+                else if (auto col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw))
+                    //non-constant integer - constant integer
+                    OpImpl::vector_constant(col_left->getData(), col_right_const->template getValue<T1>(), vec_res);
+                else
+                    return false;
+            }
+            else
+                return false;
+
+            block.getByPosition(result).column = std::move(col_res);
+            return true;
+        });
+        if (!valid)
+            throw Exception(getName() + "'s arguments do not match the expected data types", ErrorCodes::ILLEGAL_COLUMN);
+    }
+};
+
+void registerFunctionBitHammingDistance(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionBitHammingDistance>();
+}
+}
diff --git a/dbms/src/Functions/registerFunctions.cpp b/dbms/src/Functions/registerFunctions.cpp
index 501f8e7f90a..09000a1dadd 100644
--- a/dbms/src/Functions/registerFunctions.cpp
+++ b/dbms/src/Functions/registerFunctions.cpp
@@ -41,6 +41,7 @@ void registerFunctionsFindCluster(FunctionFactory &);
 void registerFunctionsJSON(FunctionFactory &);
 void registerFunctionsIntrospection(FunctionFactory &);
 void registerFunctionsConsistentHashing(FunctionFactory & factory);
+void registerFunctionsStringHash(FunctionFactory & factory);
 
 void registerFunctions()
 {
@@ -80,6 +81,7 @@ void registerFunctions()
     registerFunctionsJSON(factory);
     registerFunctionsIntrospection(factory);
     registerFunctionsConsistentHashing(factory);
+    registerFunctionsStringHash(factory);
 }
 
 }
diff --git a/dbms/src/Functions/registerFunctionsArithmetic.cpp b/dbms/src/Functions/registerFunctionsArithmetic.cpp
index 1faa28e395e..a03058c37e9 100644
--- a/dbms/src/Functions/registerFunctionsArithmetic.cpp
+++ b/dbms/src/Functions/registerFunctionsArithmetic.cpp
@@ -32,6 +32,8 @@ void registerFunctionIntExp10(FunctionFactory & factory);
 void registerFunctionRoundToExp2(FunctionFactory & factory);
 void registerFunctionRoundDuration(FunctionFactory & factory);
 void registerFunctionRoundAge(FunctionFactory & factory);
+void registerFunctionBitHammingDistance(FunctionFactory & factory);
+void registerFunctionTupleHammingDistance(FunctionFactory & factory);
 
 void registerFunctionBitBoolMaskOr(FunctionFactory & factory);
 void registerFunctionBitBoolMaskAnd(FunctionFactory & factory);
@@ -69,6 +71,8 @@ void registerFunctionsArithmetic(FunctionFactory & factory)
     registerFunctionRoundToExp2(factory);
     registerFunctionRoundDuration(factory);
     registerFunctionRoundAge(factory);
+    registerFunctionBitHammingDistance(factory);
+    registerFunctionTupleHammingDistance(factory);
 
     /// Not for external use.
     registerFunctionBitBoolMaskOr(factory);
diff --git a/dbms/src/Functions/tupleHammingDistance.cpp b/dbms/src/Functions/tupleHammingDistance.cpp
new file mode 100644
index 00000000000..4a727aef59a
--- /dev/null
+++ b/dbms/src/Functions/tupleHammingDistance.cpp
@@ -0,0 +1,224 @@
+#include <Columns/ColumnTuple.h>
+#include <Columns/ColumnVector.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Functions/FunctionFactory.h>
+#include <Functions/FunctionHelpers.h>
+#include <Functions/IFunction.h>
+#include <Functions/castTypeToEither.h>
+
+namespace DB
+{
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_COLUMN;
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+}
+
+template <typename A, typename B>
+struct TupleHammingDistanceImpl
+{
+    using ResultType = UInt8;
+
+    static void NO_INLINE vector_vector(
+        const PaddedPODArray<A> & a1,
+        const PaddedPODArray<A> & b1,
+        const PaddedPODArray<B> & a2,
+        const PaddedPODArray<B> & b2,
+        PaddedPODArray<ResultType> & c)
+    {
+        size_t size = a1.size();
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(a1[i], a2[i]) + apply(b1[i], b2[i]);
+    }
+
+    static void NO_INLINE
+    vector_constant(const PaddedPODArray<A> & a1, const PaddedPODArray<A> & b1, UInt64 a2, UInt64 b2, PaddedPODArray<ResultType> & c)
+    {
+        size_t size = a1.size();
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(a1[i], a2) + apply(b1[i], b2);
+    }
+
+    static void NO_INLINE
+    constant_vector(UInt64 a1, UInt64 b1, const PaddedPODArray<B> & a2, const PaddedPODArray<B> & b2, PaddedPODArray<ResultType> & c)
+    {
+        size_t size = a2.size();
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(a1, a2[i]) + apply(b1, b2[i]);
+    }
+
+    static ResultType constant_constant(UInt64 a1, UInt64 b1, UInt64 a2, UInt64 b2) { return apply(a1, a2) + apply(b1, b2); }
+
+private:
+    static UInt8 pop_cnt(UInt64 res)
+    {
+        UInt8 count = 0;
+        for (; res; res >>= 1)
+            count += res & 1u;
+        return count;
+    }
+
+    static inline UInt8 apply(UInt64 a, UInt64 b)
+    {
+        UInt64 res = a ^ b;
+        return pop_cnt(res);
+    }
+};
+
+template <typename F>
+bool castType(const IDataType * type, F && f)
+{
+    return castTypeToEither<
+        DataTypeInt8,
+        DataTypeInt16,
+        DataTypeInt32,
+        DataTypeInt64,
+        DataTypeUInt8,
+        DataTypeUInt16,
+        DataTypeUInt32,
+        DataTypeUInt64>(type, std::forward<F>(f));
+}
+
+template <typename F>
+static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
+{
+    return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+}
+
+//tupleHammingDistance function: (Tuple(Integer, Integer), Tuple(Integer, Integer))->UInt8
+//in order to avoid code bloating, for non-constant tuple, we make sure that the elements
+//in the tuple should have same data type, and for constant tuple, elements can be any integer
+//data type, we cast all of them into UInt64
+class FunctionTupleHammingDistance : public IFunction
+{
+public:
+    static constexpr auto name = "tupleHammingDistance";
+    using ResultType = UInt8;
+    static FunctionPtr create(const Context &) { return std::make_shared<FunctionTupleHammingDistance>(); }
+
+    String getName() const override { return name; }
+
+    size_t getNumberOfArguments() const override { return 2; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (!isTuple(arguments[0]))
+            throw Exception(
+                "Illegal type " + arguments[0]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        if (!isTuple(arguments[1]))
+            throw Exception(
+                "Illegal type " + arguments[1]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        return std::make_shared<DataTypeUInt8>();
+    }
+
+    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
+    {
+        const ColumnWithTypeAndName & arg1 = block.getByPosition(arguments[0]);
+        const ColumnWithTypeAndName & arg2 = block.getByPosition(arguments[1]);
+        const DataTypeTuple & type1 = static_cast<const DataTypeTuple &>(*arg1.type);
+        const DataTypeTuple & type2 = static_cast<const DataTypeTuple &>(*arg2.type);
+        auto & left_elems = type1.getElements();
+        auto & right_elems = type2.getElements();
+        if (left_elems.size() != 2 || right_elems.size() != 2)
+            throw Exception(
+                "Illegal column of arguments of function " + getName() + ", tuple should have exactly two elements.",
+                ErrorCodes::ILLEGAL_COLUMN);
+        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right) {
+            using LeftDataType = std::decay_t<decltype(left)>;
+            using RightDataType = std::decay_t<decltype(right)>;
+            using T0 = typename LeftDataType::FieldType;
+            using T1 = typename RightDataType::FieldType;
+            using ColVecT0 = ColumnVector<T0>;
+            using ColVecT1 = ColumnVector<T1>;
+            using ColVecResult = ColumnVector<ResultType>;
+
+            using OpImpl = TupleHammingDistanceImpl<T0, T1>;
+
+            // constant tuple - constant tuple
+            if (const ColumnConst * const_col_left = checkAndGetColumnConst<ColumnTuple>(arg1.column.get()))
+            {
+                if (const ColumnConst * const_col_right = checkAndGetColumnConst<ColumnTuple>(arg2.column.get()))
+                {
+                    auto cols1 = convertConstTupleToConstantElements(*const_col_left);
+                    auto cols2 = convertConstTupleToConstantElements(*const_col_right);
+                    Field a1, b1, a2, b2;
+                    cols1[0]->get(0, a1);
+                    cols1[1]->get(0, b1);
+                    cols2[0]->get(0, a2);
+                    cols2[1]->get(0, b2);
+                    auto res = OpImpl::constant_constant(a1.get<UInt64>(), b1.get<UInt64>(), a2.get<UInt64>(), b2.get<UInt64>());
+                    block.getByPosition(result).column = DataTypeUInt8().createColumnConst(const_col_left->size(), toField(res));
+                    return true;
+                }
+            }
+
+            typename ColVecResult::MutablePtr col_res = nullptr;
+            col_res = ColVecResult::create();
+            auto & vec_res = col_res->getData();
+            vec_res.resize(block.rows());
+            // constant tuple - non-constant tuple
+            if (const ColumnConst * const_col_left = checkAndGetColumnConst<ColumnTuple>(arg1.column.get()))
+            {
+                if (const ColumnTuple * col_right = typeid_cast<const ColumnTuple *>(arg2.column.get()))
+                {
+                    auto const_cols = convertConstTupleToConstantElements(*const_col_left);
+                    Field a1, b1;
+                    const_cols[0]->get(0, a1);
+                    const_cols[1]->get(0, b1);
+                    auto col_r1 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(0));
+                    auto col_r2 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(1));
+                    if (col_r1 && col_r2)
+                        OpImpl::constant_vector(a1.get<UInt64>(), b1.get<UInt64>(), col_r1->getData(), col_r2->getData(), vec_res);
+                    else
+                        return false;
+                }
+                else
+                    return false;
+            }
+            else if (const ColumnTuple * col_left = typeid_cast<const ColumnTuple *>(arg1.column.get()))
+            {
+                auto col_l1 = checkAndGetColumn<ColVecT0>(&col_left->getColumn(0));
+                auto col_l2 = checkAndGetColumn<ColVecT0>(&col_left->getColumn(1));
+                if (col_l1 && col_l2)
+                {
+                    // non-constant tuple - constant tuple
+                    if (const ColumnConst * const_col_right = checkAndGetColumnConst<ColumnTuple>(arg2.column.get()))
+                    {
+                        auto const_cols = convertConstTupleToConstantElements(*const_col_right);
+                        Field a2, b2;
+                        const_cols[0]->get(0, a2);
+                        const_cols[1]->get(0, b2);
+                        OpImpl::vector_constant(col_l1->getData(), col_l2->getData(), a2.get<UInt64>(), a2.get<UInt64>(), vec_res);
+                    }
+                    // non-constant tuple - non-constant tuple
+                    else if (const ColumnTuple * col_right = typeid_cast<const ColumnTuple *>(arg2.column.get()))
+                    {
+                        auto col_r1 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(0));
+                        auto col_r2 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(1));
+                        if (col_r1 && col_r2)
+                            OpImpl::vector_vector(col_l1->getData(), col_l2->getData(), col_r1->getData(), col_r2->getData(), vec_res);
+                        else
+                            return false;
+                    }
+                    else
+                        return false;
+                }
+                else
+                    return false;
+            }
+            else
+                return false;
+            block.getByPosition(result).column = std::move(col_res);
+            return true;
+        });
+        if (!valid)
+            throw Exception(getName() + "'s arguments do not match the expected data types", ErrorCodes::ILLEGAL_COLUMN);
+    }
+};
+
+void registerFunctionTupleHammingDistance(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionTupleHammingDistance>();
+}
+}

From 9403dd1520dfd7a887a159ab0af0da699747e2ec Mon Sep 17 00:00:00 2001
From: fenglv <fenglv15@mails.ucas.ac.cn>
Date: Wed, 6 Nov 2019 18:35:55 +0800
Subject: [PATCH 003/504] add test

fix comment style

fix lambda function style
---
 dbms/src/Functions/ExtractString.h            | 26 ++++-----
 dbms/src/Functions/FunctionsStringHash.cpp    | 54 +++++++++----------
 dbms/src/Functions/FunctionsStringHash.h      | 10 ++--
 dbms/src/Functions/bitHammingDistance.cpp     | 15 +++---
 dbms/src/Functions/tupleHammingDistance.cpp   | 15 +++---
 .../01016_simhash_minhash.reference           | 50 +++++++++++++++++
 .../0_stateless/01016_simhash_minhash.sql     | 47 ++++++++++++++++
 .../01017_bithamming_distance.reference       | 15 ++++++
 .../0_stateless/01017_bithamming_distance.sql | 20 +++++++
 .../01017_tuplehamming_distance.reference     | 15 ++++++
 .../01017_tuplehamming_distance.sql           | 19 +++++++
 11 files changed, 228 insertions(+), 58 deletions(-)
 create mode 100644 dbms/tests/queries/0_stateless/01016_simhash_minhash.reference
 create mode 100644 dbms/tests/queries/0_stateless/01016_simhash_minhash.sql
 create mode 100644 dbms/tests/queries/0_stateless/01017_bithamming_distance.reference
 create mode 100644 dbms/tests/queries/0_stateless/01017_bithamming_distance.sql
 create mode 100644 dbms/tests/queries/0_stateless/01017_tuplehamming_distance.reference
 create mode 100644 dbms/tests/queries/0_stateless/01017_tuplehamming_distance.sql

diff --git a/dbms/src/Functions/ExtractString.h b/dbms/src/Functions/ExtractString.h
index 05566496cba..c74b5175ea6 100644
--- a/dbms/src/Functions/ExtractString.h
+++ b/dbms/src/Functions/ExtractString.h
@@ -12,8 +12,8 @@
 
 namespace DB
 {
-//used by FunctionsStringSimilarity and FunctionsStringHash
-//includes exacting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
+// used by FunctionsStringSimilarity and FunctionsStringHash
+// includes exacting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
 template <size_t N, bool CaseInsensitive>
 struct ExtractStringImpl
 {
@@ -47,14 +47,14 @@ struct ExtractStringImpl
         return default_padding;
     }
 
-    //used by FunctionsStringHash
-    //it's not easy to add padding for ColumnString, so we need safety check each memcpy
+    // used by FunctionsStringHash
+    // it's not easy to add padding for ColumnString, so we need safety check each memcpy
     static ALWAYS_INLINE size_t readASCIICodePointsNoPadding(UInt8 * code_points, const char *& pos, const char * end)
     {
         constexpr size_t padding_offset = default_padding - N + 1;
         memcpy(code_points, code_points + padding_offset, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt8));
 
-        //safety check
+        // safety check
         size_t cpy_size = (pos + padding_offset > end) ? end - pos : padding_offset;
 
         memcpy(code_points + (N - 1), pos, cpy_size * sizeof(UInt8));
@@ -69,12 +69,12 @@ struct ExtractStringImpl
         return default_padding;
     }
 
-    //read a ASCII word from pos to word
-    //if the word size exceeds max_word_size, only read max_word_size byte
-    //in  FuntionsStringHash, the default value of max_word_size is 128
+    // read a ASCII word from pos to word
+    // if the word size exceeds max_word_size, only read max_word_size byte
+    // in  FuntionsStringHash, the default value of max_word_size is 128
     static ALWAYS_INLINE inline size_t readOneASCIIWord(UInt8 * word, const char *& pos, const char * end, const size_t & max_word_size)
     {
-        //jump seperators
+        // jump seperators
         while (pos < end && !isAlphaNum(*pos))
             ++pos;
 
@@ -105,14 +105,14 @@ struct ExtractStringImpl
         return num;
     }
 
-    //read one UTF8 word from pos to word
-    //also, we assume that one word size cann't exceed max_word_size with default value 128
+    // read one UTF8 word from pos to word
+    // also, we assume that one word size cann't exceed max_word_size with default value 128
     static ALWAYS_INLINE inline size_t readOneUTF8Word(UInt32 * word, const char *& pos, const char * end, const size_t & max_word_size)
     {
         // jump UTF8 seperator
         while (pos < end && isUTF8Sep(*pos))
             ++pos;
-        //UTF8 word's character number
+        // UTF8 word's character number
         size_t num = 0;
         while (pos < end && num < max_word_size && !isUTF8Sep(*pos))
         {
@@ -133,7 +133,7 @@ private:
         ((cont[Offset + I] = std::tolower(cont[Offset + I])), ...);
     }
 
-    //we use ASCII non-alphanum character as UTF8 seperator
+    // we use ASCII non-alphanum character as UTF8 seperator
     static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNum(c); }
 
     // read one UTF8 character and return it
diff --git a/dbms/src/Functions/FunctionsStringHash.cpp b/dbms/src/Functions/FunctionsStringHash.cpp
index 797d7d30078..215d49544cb 100644
--- a/dbms/src/Functions/FunctionsStringHash.cpp
+++ b/dbms/src/Functions/FunctionsStringHash.cpp
@@ -75,7 +75,7 @@ struct Hash
     }
 };
 
-//Sinhash String -> UInt64
+// Sinhash String -> UInt64
 template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimhashImpl
 {
@@ -123,8 +123,8 @@ struct SimhashImpl
             iter = 0;
         } while (start < end && (found = read_code_points(cp, start, end)));
 
-        //finally, we return a 64 bit value according to finger_vec
-        //if finger_vec[i] > 0, the i'th bit of the value is 1, otherwise 0
+        // finally, we return a 64 bit value according to finger_vec
+        // if finger_vec[i] > 0, the i'th bit of the value is 1, otherwise 0
         std::bitset<64> res_bit(0u);
         for (size_t i = 0; i < 64; ++i)
         {
@@ -160,7 +160,7 @@ struct SimhashImpl
         // word buffer to store one word
         CodePoint word_buf[max_word_size] = {};
         size_t word_size;
-        //get first word shingle
+        // get first word shingle
         for (size_t i = 0; i < N && start < end; ++i)
         {
             word_size = read_one_word(word_buf, start, end, max_word_size);
@@ -189,9 +189,9 @@ struct SimhashImpl
             // word hash value into locaion of a1, then array become |a5|a6|a2|a3|a4|
             nwordHashes[offset] = Hash::hashSum(word_buf, word_size);
             offset = (offset + 1) % N;
-            //according to the word hash storation way, in order to not lose the word shingle's
-            //sequence information, when calculation word shingle hash value, we need provide the offset
-            //inforation, which is the offset of the first word's hash value of the word shingle
+            // according to the word hash storation way, in order to not lose the word shingle's
+            // sequence information, when calculation word shingle hash value, we need provide the offset
+            // inforation, which is the offset of the first word's hash value of the word shingle
             hash_value = hash_functor(nwordHashes, N, offset);
             std::bitset<64> bits(hash_value);
             for (size_t i = 0; i < 64; ++i)
@@ -237,7 +237,7 @@ struct SimhashImpl
             res = dispatch(wordShinglesCalculateHashValue, data.data(), data.size());
     }
 
-    //non-constant string
+    // non-constant string
     static void vector(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, PaddedPODArray<UInt64> & res)
     {
         for (size_t i = 0; i < offsets.size(); ++i)
@@ -255,12 +255,12 @@ struct SimhashImpl
     }
 };
 
-//Minhash: String -> Tuple(UInt64, UInt64)
-//for each string, we extract ngram or word shingle,
-//for each ngram or word shingle, calculate a hash value,
-//then we take the K minimum hash values to calculate a hashsum,
-//and take the K maximum hash values to calculate another hashsum,
-//return this two hashsum: Tuple(hashsum1, hashsum2)
+// Minhash: String -> Tuple(UInt64, UInt64)
+// for each string, we extract ngram or word shingle,
+// for each ngram or word shingle, calculate a hash value,
+// then we take the K minimum hash values to calculate a hashsum,
+// and take the K maximum hash values to calculate another hashsum,
+// return this two hashsum: Tuple(hashsum1, hashsum2)
 template <size_t N, size_t K, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct MinhashImpl
 {
@@ -298,11 +298,11 @@ struct MinhashImpl
         hashes[i] = v;
     }
 
-    //Minhash ngram calculate function, String -> Tuple(UInt64, UInt64)
-    //we extract ngram from input string, and calculate a hash value for each ngram
-    //then we take the K minimum hash values to calculate a hashsum,
-    //and take the K maximum hash values to calculate another hashsum,
-    //return this two hashsum: Tuple(hashsum1, hashsum2)
+    // Minhash ngram calculate function, String -> Tuple(UInt64, UInt64)
+    // we extract ngram from input string, and calculate a hash value for each ngram
+    // then we take the K minimum hash values to calculate a hashsum,
+    // and take the K maximum hash values to calculate another hashsum,
+    // return this two hashsum: Tuple(hashsum1, hashsum2)
     static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> ngramCalculateHashValue(
         const char * data,
         const size_t size,
@@ -339,8 +339,8 @@ struct MinhashImpl
     }
 
     // Minhash word shingle hash value calculate function: String ->Tuple(UInt64, UInt64)
-    //for each word shingle, we calculate a hash value, but in fact, we just maintain the
-    //K minimum and K maximum hash value
+    // for each word shingle, we calculate a hash value, but in fact, we just maintain the
+    // K minimum and K maximum hash value
     static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> wordShinglesCalculateHashValue(
         const char * data,
         const size_t size,
@@ -349,7 +349,7 @@ struct MinhashImpl
     {
         const char * start = data;
         const char * end = start + size;
-        //also we just store the K minimu and K maximum hash values
+        // also we just store the K minimu and K maximum hash values
         UInt64 k_minimum[K] = {};
         UInt64 k_maxinum[K] = {};
         // array to store n word hashes
@@ -357,8 +357,8 @@ struct MinhashImpl
         // word buffer to store one word
         CodePoint word_buf[max_word_size] = {};
         size_t word_size;
-        //how word shingle hash value calculation and word hash storation is same as we
-        //have descripted in Simhash wordShinglesCalculateHashValue function
+        // how word shingle hash value calculation and word hash storation is same as we
+        // have descripted in Simhash wordShinglesCalculateHashValue function
         for (size_t i = 0; i < N && start < end; ++i)
         {
             word_size = read_one_word(word_buf, start, end, max_word_size);
@@ -416,7 +416,7 @@ struct MinhashImpl
             std::tie(res1, res2) = dispatch(wordShinglesCalculateHashValue, data.data(), data.size());
     }
 
-    //non-constant string
+    // non-constant string
     static void vector(
         const ColumnString::Chars & data,
         const ColumnString::Offsets & offsets,
@@ -518,7 +518,7 @@ struct NameWordShingleMinhashCaseInsensitiveUTF8
     static constexpr auto name = "wordShingleMinhashCaseInsensitiveUTF8";
 };
 
-//Simhash
+// Simhash
 using FunctionNgramSimhash = FunctionsStringHash<SimhashImpl<4, UInt8, false, true, false>, NameNgramSimhash, true>;
 
 using FunctionNgramSimhashCaseInsensitive
@@ -539,7 +539,7 @@ using FunctionWordShingleSimhashUTF8 = FunctionsStringHash<SimhashImpl<3, UInt32
 using FunctionWordShingleSimhashCaseInsensitiveUTF8
     = FunctionsStringHash<SimhashImpl<3, UInt32, true, false, true>, NameWordShingleSimhashCaseInsensitiveUTF8, true>;
 
-//Minhash
+// Minhash
 using FunctionNgramMinhash = FunctionsStringHash<MinhashImpl<4, 6, UInt8, false, true, false>, NameNgramMinhash, false>;
 
 using FunctionNgramMinhashCaseInsensitive
diff --git a/dbms/src/Functions/FunctionsStringHash.h b/dbms/src/Functions/FunctionsStringHash.h
index 185097ade99..bb1e42ab5fa 100644
--- a/dbms/src/Functions/FunctionsStringHash.h
+++ b/dbms/src/Functions/FunctionsStringHash.h
@@ -15,14 +15,12 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int ILLEGAL_TYPE_OF_ARGUMENT;
-    extern const int ILLEGAL_COLUMN;
-    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
     extern const int TOO_LARGE_STRING_SIZE;
 }
 
-//FunctionStringHash
-//Simhash: String -> UInt64
-//Minhash: String -> (UInt64, UInt64)
+// FunctionStringHash
+// Simhash: String -> UInt64
+// Minhash: String -> (UInt64, UInt64)
 template <typename Impl, typename Name, bool IsSimhash>
 class FunctionsStringHash : public IFunction
 {
@@ -103,7 +101,7 @@ public:
             }
             else
             {
-                //non const string
+                // non const string
                 auto col_h1 = ColumnVector<ResultType>::create();
                 auto col_h2 = ColumnVector<ResultType>::create();
                 typename ColumnVector<ResultType>::Container & vec_h1 = col_h1->getData();
diff --git a/dbms/src/Functions/bitHammingDistance.cpp b/dbms/src/Functions/bitHammingDistance.cpp
index 2572720bb4e..fdef72d4c43 100644
--- a/dbms/src/Functions/bitHammingDistance.cpp
+++ b/dbms/src/Functions/bitHammingDistance.cpp
@@ -75,10 +75,12 @@ bool castType(const IDataType * type, F && f)
 template <typename F>
 static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
 {
-    return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+    return castType(left, [&](const auto & left_) {
+        return castType(right, [&](const auto & right_) { return f(left_, right_); });
+    });
 }
 
-//bitHammingDistance function: (Integer, Integer) -> UInt8
+// bitHammingDistance function: (Integer, Integer) -> UInt8
 class FunctionBitHammingDistance : public IFunction
 {
 public:
@@ -105,7 +107,8 @@ public:
     {
         auto * left_generic = block.getByPosition(arguments[0]).type.get();
         auto * right_generic = block.getByPosition(arguments[1]).type.get();
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
+        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+        {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
@@ -122,7 +125,7 @@ public:
             {
                 if (auto col_right = checkAndGetColumnConst<ColVecT1>(col_right_raw))
                 {
-                    //constant integer - constant integer
+                    // constant integer - constant integer
                     auto res = OpImpl::constant_constant(col_left->template getValue<T0>(), col_right->template getValue<T1>());
                     block.getByPosition(result).column = DataTypeUInt8().createColumnConst(col_left->size(), toField(res));
                     return true;
@@ -148,10 +151,10 @@ public:
             else if (auto col_left = checkAndGetColumn<ColVecT0>(col_left_raw))
             {
                 if (auto col_right = checkAndGetColumn<ColVecT1>(col_right_raw))
-                    //non-constant integer - non-constant integer
+                    // non-constant integer - non-constant integer
                     OpImpl::vector_vector(col_left->getData(), col_right->getData(), vec_res);
                 else if (auto col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw))
-                    //non-constant integer - constant integer
+                    // non-constant integer - constant integer
                     OpImpl::vector_constant(col_left->getData(), col_right_const->template getValue<T1>(), vec_res);
                 else
                     return false;
diff --git a/dbms/src/Functions/tupleHammingDistance.cpp b/dbms/src/Functions/tupleHammingDistance.cpp
index 4a727aef59a..45c113edad4 100644
--- a/dbms/src/Functions/tupleHammingDistance.cpp
+++ b/dbms/src/Functions/tupleHammingDistance.cpp
@@ -83,13 +83,15 @@ bool castType(const IDataType * type, F && f)
 template <typename F>
 static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
 {
-    return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+    return castType(left, [&](const auto & left_) {
+        return castType(right, [&](const auto & right_) { return f(left_, right_); });
+    });
 }
 
-//tupleHammingDistance function: (Tuple(Integer, Integer), Tuple(Integer, Integer))->UInt8
-//in order to avoid code bloating, for non-constant tuple, we make sure that the elements
-//in the tuple should have same data type, and for constant tuple, elements can be any integer
-//data type, we cast all of them into UInt64
+// tupleHammingDistance function: (Tuple(Integer, Integer), Tuple(Integer, Integer))->UInt8
+// in order to avoid code bloating, for non-constant tuple, we make sure that the elements
+// in the tuple should have same data type, and for constant tuple, elements can be any integer
+// data type, we cast all of them into UInt64
 class FunctionTupleHammingDistance : public IFunction
 {
 public:
@@ -124,7 +126,8 @@ public:
             throw Exception(
                 "Illegal column of arguments of function " + getName() + ", tuple should have exactly two elements.",
                 ErrorCodes::ILLEGAL_COLUMN);
-        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right) {
+        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right)
+        {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
diff --git a/dbms/tests/queries/0_stateless/01016_simhash_minhash.reference b/dbms/tests/queries/0_stateless/01016_simhash_minhash.reference
new file mode 100644
index 00000000000..fa62adde45c
--- /dev/null
+++ b/dbms/tests/queries/0_stateless/01016_simhash_minhash.reference
@@ -0,0 +1,50 @@
+0
+2718169299
+2718169299
+3333471646
+26585365
+4151513063
+4151513063
+4151513063
+3150464485
+(0,0)
+(2736268688,2736268688)
+(2736268688,2736268688)
+(916562399,916562399)
+(3436376151,3436376151)
+(0,3423682776)
+(0,3423682776)
+(0,3423682776)
+(0,2393737641)
+2548869326
+2548869326
+401385678
+401385710
+4258739090
+4260836242
+718415633
+718681881
+4026448893
+4026449917
+4026466301
+4026466301
+4026448893
+4026449917
+3957325823
+4217372671
+(3946088007,3946088007)
+(3946088007,3946088007)
+(2332295796,2332295796)
+(535012010,535012010)
+(3696559901,3696559901)
+(3696559901,3696559901)
+(169287209,169287209)
+(169287209,169287209)
+(0,1509393235)
+(0,1509393235)
+(0,1509393235)
+(0,1509393235)
+(0,1509393235)
+(0,1509393235)
+(0,1975937193)
+(0,1975937193)
diff --git a/dbms/tests/queries/0_stateless/01016_simhash_minhash.sql b/dbms/tests/queries/0_stateless/01016_simhash_minhash.sql
new file mode 100644
index 00000000000..9e87216d26f
--- /dev/null
+++ b/dbms/tests/queries/0_stateless/01016_simhash_minhash.sql
@@ -0,0 +1,47 @@
+SELECT ngramSimhash('');
+SELECT ngramSimhash('what a cute cat.');
+SELECT ngramSimhashCaseInsensitive('what a cute cat.');
+SELECT ngramSimhashUTF8('what a cute cat.');
+SELECT ngramSimhashCaseInsensitiveUTF8('what a cute cat.');
+SELECT wordShingleSimhash('what a cute cat.');
+SELECT wordShingleSimhashCaseInsensitive('what a cute cat.');
+SELECT wordShingleSimhashUTF8('what a cute cat.');
+SELECT wordShingleSimhashCaseInsensitiveUTF8('what a cute cat.');
+
+SELECT ngramMinhash('');
+SELECT ngramMinhash('what a cute cat.');
+SELECT ngramMinhashCaseInsensitive('what a cute cat.');
+SELECT ngramMinhashUTF8('what a cute cat.');
+SELECT ngramMinhashCaseInsensitiveUTF8('what a cute cat.');
+SELECT wordShingleMinhash('what a cute cat.');
+SELECT wordShingleMinhashCaseInsensitive('what a cute cat.');
+SELECT wordShingleMinhashUTF8('what a cute cat.');
+SELECT wordShingleMinhashCaseInsensitiveUTF8('what a cute cat.');
+
+DROP TABLE IF EXISTS defaults;
+CREATE TABLE defaults
+(
+   s String
+)ENGINE = Memory();
+
+INSERT INTO defaults values ('It is the latest occurrence of the Southeast European haze, the issue that occurs in constant intensity during every wet season. It has mainly been caused by forest fires resulting from illegal slash-and-burn clearing performed on behalf of the palm oil industry in Kazakhstan, principally on the islands, which then spread quickly in the dry season.') ('It is the latest occurrence of the Southeast Asian haze, the issue that occurs in constant intensity during every wet season. It has mainly been caused by forest fires resulting from illegal slash-and-burn clearing performed on behalf of the palm oil industry in Kazakhstan, principally on the islands, which then spread quickly in the dry season.');
+
+SELECT ngramSimhash(s) FROM defaults;
+SELECT ngramSimhashCaseInsensitive(s) FROM defaults;
+SELECT ngramSimhashUTF8(s) FROM defaults;
+SELECT ngramSimhashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT wordShingleSimhash(s) FROM defaults;
+SELECT wordShingleSimhashCaseInsensitive(s) FROM defaults;
+SELECT wordShingleSimhashUTF8(s) FROM defaults;
+SELECT wordShingleSimhashCaseInsensitiveUTF8(s) FROM defaults;
+
+SELECT ngramMinhash(s) FROM defaults;
+SELECT ngramMinhashCaseInsensitive(s) FROM defaults;
+SELECT ngramMinhashUTF8(s) FROM defaults;
+SELECT ngramMinhashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT wordShingleMinhash(s) FROM defaults;
+SELECT wordShingleMinhashCaseInsensitive(s) FROM defaults;
+SELECT wordShingleMinhashUTF8(s) FROM defaults;
+SELECT wordShingleMinhashCaseInsensitiveUTF8(s) FROM defaults;
+
+DROP TABLE defaults;
diff --git a/dbms/tests/queries/0_stateless/01017_bithamming_distance.reference b/dbms/tests/queries/0_stateless/01017_bithamming_distance.reference
new file mode 100644
index 00000000000..cc2d4f39154
--- /dev/null
+++ b/dbms/tests/queries/0_stateless/01017_bithamming_distance.reference
@@ -0,0 +1,15 @@
+1
+7
+63
+2
+1
+3
+5
+4
+6
+6
+6
+3
+5
+9
+9
diff --git a/dbms/tests/queries/0_stateless/01017_bithamming_distance.sql b/dbms/tests/queries/0_stateless/01017_bithamming_distance.sql
new file mode 100644
index 00000000000..4b36894b97c
--- /dev/null
+++ b/dbms/tests/queries/0_stateless/01017_bithamming_distance.sql
@@ -0,0 +1,20 @@
+SELECT bitHammingDistance(1, 5);
+SELECT bitHammingDistance(100, 100000);
+SELECT bitHammingDistance(-1, 1);
+
+DROP TABLE IF EXISTS defaults;
+CREATE TABLE defaults
+(
+	n1 UInt8,
+	n2 UInt16,
+	n3 UInt32,
+	n4 UInt64
+)ENGINE = Memory();
+
+INSERT INTO defaults VALUES (1, 2, 3, 4) (12, 4345, 435, 1233) (45, 675, 32343, 54566) (90, 784, 9034, 778752);
+
+SELECT bitHammingDistance(4, n1) FROM defaults;
+SELECT bitHammingDistance(n2, 100) FROM defaults;
+SELECT bitHammingDistance(n3, n4) FROM defaults;
+
+DROP TABLE defaults;
diff --git a/dbms/tests/queries/0_stateless/01017_tuplehamming_distance.reference b/dbms/tests/queries/0_stateless/01017_tuplehamming_distance.reference
new file mode 100644
index 00000000000..eee1a7eee3b
--- /dev/null
+++ b/dbms/tests/queries/0_stateless/01017_tuplehamming_distance.reference
@@ -0,0 +1,15 @@
+3
+5
+60
+5
+3
+10
+10
+114
+119
+111
+104
+69
+13
+65
+25
diff --git a/dbms/tests/queries/0_stateless/01017_tuplehamming_distance.sql b/dbms/tests/queries/0_stateless/01017_tuplehamming_distance.sql
new file mode 100644
index 00000000000..0db73232bb3
--- /dev/null
+++ b/dbms/tests/queries/0_stateless/01017_tuplehamming_distance.sql
@@ -0,0 +1,19 @@
+SELECT tupleHammingDistance((1, 2), (3, 4));
+SELECT tupleHammingDistance((120, 2434), (123, 434));
+SELECT tupleHammingDistance((-12, 434), (987, 432));
+
+DROP TABLE IF EXISTS defaults;
+CREATE TABLE defaults
+(
+	t1 Tuple(UInt16, UInt16),
+	t2 Tuple(UInt32, UInt32),
+	t3 Tuple(Int64, Int64)
+)ENGINE = Memory();
+
+INSERT INTO defaults VALUES ((12, 43), (12312, 43453) ,(-10, 32)) ((1, 4), (546, 12345), (123, 456)) ((90, 9875), (43456, 234203), (1231, -123)) ((87, 987), (545645, 768354634), (9123, 909));
+
+SELECT tupleHammingDistance((1, 3), t1) FROM defaults;
+SELECT tupleHammingDistance(t2, (-1, 1)) FROM defaults;
+SELECT tupleHammingDistance(t2, t3) FROM defaults;
+
+DROP TABLE defaults;

From ced7fe59dbe48f261abb3fec427eadbc50ba7c5f Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Thu, 5 Dec 2019 06:48:40 +0300
Subject: [PATCH 004/504] Update ExtractString.h

---
 dbms/src/Functions/ExtractString.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dbms/src/Functions/ExtractString.h b/dbms/src/Functions/ExtractString.h
index c74b5175ea6..040e62d9580 100644
--- a/dbms/src/Functions/ExtractString.h
+++ b/dbms/src/Functions/ExtractString.h
@@ -13,7 +13,7 @@
 namespace DB
 {
 // used by FunctionsStringSimilarity and FunctionsStringHash
-// includes exacting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
+// includes extracting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
 template <size_t N, bool CaseInsensitive>
 struct ExtractStringImpl
 {

From 241fd556576fc7833174c5346568732e1742a8d8 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Thu, 5 Dec 2019 07:08:35 +0300
Subject: [PATCH 005/504] Update FunctionsStringHash.cpp

---
 dbms/src/Functions/FunctionsStringHash.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dbms/src/Functions/FunctionsStringHash.cpp b/dbms/src/Functions/FunctionsStringHash.cpp
index 215d49544cb..d7277fcb98b 100644
--- a/dbms/src/Functions/FunctionsStringHash.cpp
+++ b/dbms/src/Functions/FunctionsStringHash.cpp
@@ -75,7 +75,7 @@ struct Hash
     }
 };
 
-// Sinhash String -> UInt64
+// Simhash String -> UInt64
 template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimhashImpl
 {

From 83c0807b43d7ca5587b16c43a577fec6ee51ec75 Mon Sep 17 00:00:00 2001
From: feng lv <fenglv15@mails.ucas.ac.cn>
Date: Fri, 22 May 2020 21:23:49 +0800
Subject: [PATCH 006/504] update

update name
---
 src/Functions/ExtractString.h                 |  14 +--
 src/Functions/FunctionsStringHash.cpp         | 106 ++++++++----------
 src/Functions/FunctionsStringHash.h           |   8 +-
 src/Functions/bitHammingDistance.cpp          |  31 +----
 src/Functions/registerFunctions.cpp           |   7 +-
 src/Functions/tupleHammingDistance.cpp        |  27 ++---
 .../01016_simhash_minhash.reference           |  59 ++++++++++
 .../0_stateless/01016_simhash_minhash.sql     |  47 ++++++++
 .../01017_bithamming_distance.reference       |  15 +++
 .../0_stateless/01017_bithamming_distance.sql |  20 ++++
 .../01017_tuplehamming_distance.reference     |  15 +++
 .../01017_tuplehamming_distance.sql           |  19 ++++
 12 files changed, 251 insertions(+), 117 deletions(-)
 create mode 100644 tests/queries/0_stateless/01016_simhash_minhash.reference
 create mode 100644 tests/queries/0_stateless/01016_simhash_minhash.sql
 create mode 100644 tests/queries/0_stateless/01017_bithamming_distance.reference
 create mode 100644 tests/queries/0_stateless/01017_bithamming_distance.sql
 create mode 100644 tests/queries/0_stateless/01017_tuplehamming_distance.reference
 create mode 100644 tests/queries/0_stateless/01017_tuplehamming_distance.sql

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index 040e62d9580..f6a7394a9fc 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -72,18 +72,18 @@ struct ExtractStringImpl
     // read a ASCII word from pos to word
     // if the word size exceeds max_word_size, only read max_word_size byte
     // in  FuntionsStringHash, the default value of max_word_size is 128
-    static ALWAYS_INLINE inline size_t readOneASCIIWord(UInt8 * word, const char *& pos, const char * end, const size_t & max_word_size)
+    static ALWAYS_INLINE inline size_t readOneASCIIWord(UInt8 * word, const char *& pos, const char * end, size_t max_word_size)
     {
         // jump seperators
-        while (pos < end && !isAlphaNum(*pos))
+        while (pos < end && !isAlphaNumericASCII(*pos))
             ++pos;
 
         // word start from here
         const char * word_start = pos;
-        while (pos < end && isAlphaNum(*pos))
+        while (pos < end && isAlphaNumericASCII(*pos))
             ++pos;
 
-        size_t word_size = (static_cast<size_t>(pos - word_start) <= max_word_size) ? pos - word_start : max_word_size;
+        size_t word_size = std::min<size_t>(pos - word_start, max_word_size);
 
         memcpy(word, word_start, word_size);
         if (CaseInsensitive)
@@ -107,7 +107,7 @@ struct ExtractStringImpl
 
     // read one UTF8 word from pos to word
     // also, we assume that one word size cann't exceed max_word_size with default value 128
-    static ALWAYS_INLINE inline size_t readOneUTF8Word(UInt32 * word, const char *& pos, const char * end, const size_t & max_word_size)
+    static ALWAYS_INLINE inline size_t readOneUTF8Word(UInt32 * word, const char *& pos, const char * end, size_t max_word_size)
     {
         // jump UTF8 seperator
         while (pos < end && isUTF8Sep(*pos))
@@ -122,7 +122,7 @@ struct ExtractStringImpl
     }
 
 private:
-    static ALWAYS_INLINE inline bool isAlphaNum(const UInt8 c)
+    static ALWAYS_INLINE inline bool isAlphaNumericASCII(const UInt8 c)
     {
         return (c >= 48 && c <= 57) || (c >= 65 && c <= 90) || (c >= 97 && c <= 122);
     }
@@ -134,7 +134,7 @@ private:
     }
 
     // we use ASCII non-alphanum character as UTF8 seperator
-    static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNum(c); }
+    static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNumericASCII(c); }
 
     // read one UTF8 character and return it
     static ALWAYS_INLINE inline UInt32 readOneUTF8Code(const char *& pos, const char * end)
diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index d7277fcb98b..2195ff7c703 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -31,47 +31,35 @@ struct Hash
 #endif
     }
 
-    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, const size_t & size, const size_t & offset)
+    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
     {
-        UInt64 res = 0;
-        UInt8 flag = 0;
+        UInt64 crc = -1ULL;
+#ifdef __SSE4_2__
         for (size_t i = offset; i < size; ++i)
-        {
-            if (flag)
-                res &= intHashCRC32(hashes[i]);
-            else
-                res |= intHashCRC32(hashes[i]);
-            flag = (flag + 1) % 2;
-        }
+            crc = _mm_crc32_u64(crc, hashes[i]);
         for (size_t i = 0; i < offset; ++i)
-        {
-            if (flag)
-                res &= intHashCRC32(hashes[i]);
-            else
-                res |= intHashCRC32(hashes[i]);
-            flag = (flag + 1) % 2;
-        }
-        return res;
+            crc = _mm_crc32_u64(crc, hashes[i]);
+#else
+        for (size_t i = offset; i < size; ++i)
+            crc = intHashCRC32(crc) ^ intHashCRC32(hashes[i]);
+        for (size_t i = 0; i < offset; ++i)
+            crc = intHashCRC32(crc) ^ intHashCRC32(hashes[i]);
+#endif
+        return crc;
     }
 
     template <typename CodePoint>
-    static ALWAYS_INLINE inline UInt64 hashSum(const CodePoint * hashes, const size_t & K)
+    static ALWAYS_INLINE inline UInt64 hashSum(const CodePoint * hashes, size_t K)
     {
-        UInt64 even = 0;
-        UInt64 odd = 0;
-        size_t i = 0;
-        for (; i + 1 < K; i += 2)
-        {
-            even |= intHashCRC32(hashes[i]);
-            odd |= intHashCRC32(hashes[i + 1]);
-        }
-        if (i < K)
-            even |= intHashCRC32(hashes[K - 1]);
+        UInt64 crc = -1ULL;
 #ifdef __SSE4_2__
-        return _mm_crc32_u64(even, odd);
+        for (size_t i = 0; i < K; ++i)
+            crc = _mm_crc32_u64(crc, hashes[i]);
 #else
-        return (intHashCRC32(even) ^ intHashCRC32(odd));
+        for (size_t i = 0; i < K; ++i)
+            crc = intHashCRC32(crc) ^ intHashCRC32(hashes[i]);
 #endif
+        return crc;
     }
 };
 
@@ -93,7 +81,7 @@ struct SimhashImpl
     // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
     static ALWAYS_INLINE inline UInt64 ngramCalculateHashValue(
         const char * data,
-        const size_t size,
+        size_t size,
         size_t (*read_code_points)(CodePoint *, const char *&, const char *),
         UInt64 (*hash_functor)(const CodePoint *))
     {
@@ -146,9 +134,9 @@ struct SimhashImpl
     // values to caculate the next word shingle hash value
     static ALWAYS_INLINE inline UInt64 wordShinglesCalculateHashValue(
         const char * data,
-        const size_t size,
-        size_t (*read_one_word)(CodePoint *, const char *&, const char *, const size_t &),
-        UInt64 (*hash_functor)(const UInt64 *, const size_t &, const size_t &))
+        size_t size,
+        size_t (*read_one_word)(CodePoint *, const char *&, const char *, size_t),
+        UInt64 (*hash_functor)(const UInt64 *, size_t, size_t))
     {
         const char * start = data;
         const char * end = data + size;
@@ -156,7 +144,7 @@ struct SimhashImpl
         // Also, a 64 bit vector initialized to zero
         Int64 finger_vec[64] = {};
         // a array to store N word hash values
-        UInt64 nwordHashes[N] = {};
+        UInt64 nword_hashes[N] = {};
         // word buffer to store one word
         CodePoint word_buf[max_word_size] = {};
         size_t word_size;
@@ -167,16 +155,16 @@ struct SimhashImpl
             if (word_size)
             {
                 // for each word, calculate a hash value and stored into the array
-                nwordHashes[i++] = Hash::hashSum(word_buf, word_size);
+                nword_hashes[i++] = Hash::hashSum(word_buf, word_size);
             }
         }
 
         // calculate the first word shingle hash value
-        UInt64 hash_value = hash_functor(nwordHashes, N, 0);
-        std::bitset<64> bits_(hash_value);
+        UInt64 hash_value = hash_functor(nword_hashes, N, 0);
+        std::bitset<64> first_bits(hash_value);
         for (size_t i = 0; i < 64; ++i)
         {
-            finger_vec[i] += ((bits_.test(i)) ? 1 : -1);
+            finger_vec[i] += ((first_bits.test(i)) ? 1 : -1);
         }
 
         size_t offset = 0;
@@ -187,12 +175,12 @@ struct SimhashImpl
             // so we need to store new word hash into location of a0, then ,this array become
             // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
             // word hash value into locaion of a1, then array become |a5|a6|a2|a3|a4|
-            nwordHashes[offset] = Hash::hashSum(word_buf, word_size);
+            nword_hashes[offset] = Hash::hashSum(word_buf, word_size);
             offset = (offset + 1) % N;
             // according to the word hash storation way, in order to not lose the word shingle's
             // sequence information, when calculation word shingle hash value, we need provide the offset
             // inforation, which is the offset of the first word's hash value of the word shingle
-            hash_value = hash_functor(nwordHashes, N, offset);
+            hash_value = hash_functor(nword_hashes, N, offset);
             std::bitset<64> bits(hash_value);
             for (size_t i = 0; i < 64; ++i)
             {
@@ -272,7 +260,7 @@ struct MinhashImpl
 
     // insert a new value into K minimum hash array if this value
     // is smaller than the greatest value in the array
-    static ALWAYS_INLINE inline void insert_minValue(UInt64 * hashes, UInt64 v)
+    static ALWAYS_INLINE inline void insertMinValue(UInt64 * hashes, UInt64 v)
     {
         size_t i = 0;
         for (; i < K && hashes[i] <= v; ++i)
@@ -286,7 +274,7 @@ struct MinhashImpl
 
     // insert a new value into K maximum hash array if this value
     // is greater than the smallest value in the array
-    static ALWAYS_INLINE inline void insert_maxValue(UInt64 * hashes, UInt64 v)
+    static ALWAYS_INLINE inline void insertMaxValue(UInt64 * hashes, UInt64 v)
     {
         int i = K - 1;
         for (; i >= 0 && hashes[i] >= v; --i)
@@ -305,7 +293,7 @@ struct MinhashImpl
     // return this two hashsum: Tuple(hashsum1, hashsum2)
     static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> ngramCalculateHashValue(
         const char * data,
-        const size_t size,
+        size_t size,
         size_t (*read_code_points)(CodePoint *, const char *&, const char *),
         UInt64 (*hash_functor)(const CodePoint *))
     {
@@ -326,8 +314,8 @@ struct MinhashImpl
                 auto new_hash = hash_functor(cp + iter);
                 // insert the new hash value into array used to store K minimum value
                 // and K maximum value
-                insert_minValue(k_minimum, new_hash);
-                insert_maxValue(k_maxinum, new_hash);
+                insertMinValue(k_minimum, new_hash);
+                insertMaxValue(k_maxinum, new_hash);
             }
             iter = 0;
         } while (start < end && (found = read_code_points(cp, start, end)));
@@ -343,9 +331,9 @@ struct MinhashImpl
     // K minimum and K maximum hash value
     static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> wordShinglesCalculateHashValue(
         const char * data,
-        const size_t size,
-        size_t (*read_one_word)(CodePoint *, const char *&, const char *, const size_t &),
-        UInt64 (*hash_functor)(const UInt64 *, const size_t &, const size_t &))
+        size_t size,
+        size_t (*read_one_word)(CodePoint *, const char *&, const char *, size_t),
+        UInt64 (*hash_functor)(const UInt64 *, size_t, size_t))
     {
         const char * start = data;
         const char * end = start + size;
@@ -353,7 +341,7 @@ struct MinhashImpl
         UInt64 k_minimum[K] = {};
         UInt64 k_maxinum[K] = {};
         // array to store n word hashes
-        UInt64 nwordHashes[N] = {};
+        UInt64 nword_hashes[N] = {};
         // word buffer to store one word
         CodePoint word_buf[max_word_size] = {};
         size_t word_size;
@@ -364,22 +352,22 @@ struct MinhashImpl
             word_size = read_one_word(word_buf, start, end, max_word_size);
             if (word_size)
             {
-                nwordHashes[i++] = Hash::hashSum(word_buf, word_size);
+                nword_hashes[i++] = Hash::hashSum(word_buf, word_size);
             }
         }
 
-        auto new_hash = hash_functor(nwordHashes, N, 0);
-        insert_minValue(k_minimum, new_hash);
-        insert_maxValue(k_maxinum, new_hash);
+        auto new_hash = hash_functor(nword_hashes, N, 0);
+        insertMinValue(k_minimum, new_hash);
+        insertMaxValue(k_maxinum, new_hash);
 
         size_t offset = 0;
         while (start < end && (word_size = read_one_word(word_buf, start, end, max_word_size)))
         {
-            nwordHashes[offset] = Hash::hashSum(word_buf, word_size);
+            nword_hashes[offset] = Hash::hashSum(word_buf, word_size);
             offset = (offset + 1) % N;
-            new_hash = hash_functor(nwordHashes, N, offset);
-            insert_minValue(k_minimum, new_hash);
-            insert_maxValue(k_maxinum, new_hash);
+            new_hash = hash_functor(nword_hashes, N, offset);
+            insertMinValue(k_minimum, new_hash);
+            insertMaxValue(k_maxinum, new_hash);
         }
 
         // calculate hashsum
diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index bb1e42ab5fa..bada7490288 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -8,7 +8,7 @@
 #include <DataTypes/DataTypeTuple.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <Functions/FunctionHelpers.h>
-#include <Functions/IFunction.h>
+#include <Functions/IFunctionImpl.h>
 
 namespace DB
 {
@@ -21,7 +21,7 @@ namespace ErrorCodes
 // FunctionStringHash
 // Simhash: String -> UInt64
 // Minhash: String -> (UInt64, UInt64)
-template <typename Impl, typename Name, bool IsSimhash>
+template <typename Impl, typename Name, bool is_simhash>
 class FunctionsStringHash : public IFunction
 {
 public:
@@ -38,7 +38,7 @@ public:
         if (!isString(arguments[0]))
             throw Exception(
                 "Illegal type " + arguments[0]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-        if (IsSimhash)
+        if constexpr (is_simhash)
             return std::make_shared<DataTypeNumber<typename Impl::ResultType>>();
         auto element = DataTypeFactory::instance().get("UInt64");
         return std::make_shared<DataTypeTuple>(DataTypes{element, element});
@@ -49,7 +49,7 @@ public:
         const ColumnPtr & column = block.getByPosition(arguments[0]).column;
         const ColumnConst * col_const = typeid_cast<const ColumnConst *>(&*column);
         using ResultType = typename Impl::ResultType;
-        if constexpr (IsSimhash)
+        if constexpr (is_simhash)
         {
             if (col_const)
             {
diff --git a/src/Functions/bitHammingDistance.cpp b/src/Functions/bitHammingDistance.cpp
index fdef72d4c43..5c13a57c426 100644
--- a/src/Functions/bitHammingDistance.cpp
+++ b/src/Functions/bitHammingDistance.cpp
@@ -40,21 +40,11 @@ struct BitHammingDistanceImpl
             c[i] = apply(a, b[i]);
     }
 
-    static ResultType constant_constant(A a, B b) { return apply(a, b); }
-
 private:
-    static UInt8 pop_cnt(UInt64 res)
-    {
-        UInt8 count = 0;
-        for (; res; res >>= 1)
-            count += res & 1u;
-        return count;
-    }
-
     static inline UInt8 apply(UInt64 a, UInt64 b)
     {
         UInt64 res = a ^ b;
-        return pop_cnt(res);
+        return __builtin_popcountll(res);
     }
 };
 
@@ -75,9 +65,7 @@ bool castType(const IDataType * type, F && f)
 template <typename F>
 static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
 {
-    return castType(left, [&](const auto & left_) {
-        return castType(right, [&](const auto & right_) { return f(left_, right_); });
-    });
+    return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
 }
 
 // bitHammingDistance function: (Integer, Integer) -> UInt8
@@ -103,12 +91,13 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
+    bool useDefaultImplementationForConstants() const override { return true; }
+
     void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
     {
         auto * left_generic = block.getByPosition(arguments[0]).type.get();
         auto * right_generic = block.getByPosition(arguments[1]).type.get();
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
-        {
+        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
@@ -121,16 +110,6 @@ public:
 
             auto col_left_raw = block.getByPosition(arguments[0]).column.get();
             auto col_right_raw = block.getByPosition(arguments[1]).column.get();
-            if (auto col_left = checkAndGetColumnConst<ColVecT0>(col_left_raw))
-            {
-                if (auto col_right = checkAndGetColumnConst<ColVecT1>(col_right_raw))
-                {
-                    // constant integer - constant integer
-                    auto res = OpImpl::constant_constant(col_left->template getValue<T0>(), col_right->template getValue<T1>());
-                    block.getByPosition(result).column = DataTypeUInt8().createColumnConst(col_left->size(), toField(res));
-                    return true;
-                }
-            }
 
             typename ColVecResult::MutablePtr col_res = nullptr;
             col_res = ColVecResult::create();
diff --git a/src/Functions/registerFunctions.cpp b/src/Functions/registerFunctions.cpp
index 02013e33d16..f3e2883a179 100644
--- a/src/Functions/registerFunctions.cpp
+++ b/src/Functions/registerFunctions.cpp
@@ -3,7 +3,6 @@
 
 namespace DB
 {
-
 void registerFunctionsArithmetic(FunctionFactory &);
 void registerFunctionsArray(FunctionFactory &);
 void registerFunctionsTuple(FunctionFactory &);
@@ -37,6 +36,9 @@ void registerFunctionsIntrospection(FunctionFactory &);
 void registerFunctionsNull(FunctionFactory &);
 void registerFunctionsJSON(FunctionFactory &);
 void registerFunctionsConsistentHashing(FunctionFactory & factory);
+void registerFunctionBitHammingDistance(FunctionFactory & factory);
+void registerFunctionTupleHammingDistance(FunctionFactory & factory);
+void registerFunctionsStringHash(FunctionFactory & factory);
 
 
 void registerFunctions()
@@ -78,6 +80,9 @@ void registerFunctions()
     registerFunctionsJSON(factory);
     registerFunctionsIntrospection(factory);
     registerFunctionsConsistentHashing(factory);
+    registerFunctionBitHammingDistance(factory);
+    registerFunctionTupleHammingDistance(factory);
+    registerFunctionsStringHash(factory);
 }
 
 }
diff --git a/src/Functions/tupleHammingDistance.cpp b/src/Functions/tupleHammingDistance.cpp
index 45c113edad4..8b3f9a696aa 100644
--- a/src/Functions/tupleHammingDistance.cpp
+++ b/src/Functions/tupleHammingDistance.cpp
@@ -51,19 +51,7 @@ struct TupleHammingDistanceImpl
     static ResultType constant_constant(UInt64 a1, UInt64 b1, UInt64 a2, UInt64 b2) { return apply(a1, a2) + apply(b1, b2); }
 
 private:
-    static UInt8 pop_cnt(UInt64 res)
-    {
-        UInt8 count = 0;
-        for (; res; res >>= 1)
-            count += res & 1u;
-        return count;
-    }
-
-    static inline UInt8 apply(UInt64 a, UInt64 b)
-    {
-        UInt64 res = a ^ b;
-        return pop_cnt(res);
-    }
+    static inline UInt8 apply(UInt64 a, UInt64 b) { return a != b; }
 };
 
 template <typename F>
@@ -83,12 +71,10 @@ bool castType(const IDataType * type, F && f)
 template <typename F>
 static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
 {
-    return castType(left, [&](const auto & left_) {
-        return castType(right, [&](const auto & right_) { return f(left_, right_); });
-    });
+    return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
 }
 
-// tupleHammingDistance function: (Tuple(Integer, Integer), Tuple(Integer, Integer))->UInt8
+// tupleHammingDistance function: (Tuple(Integer, Integer), Tuple(Integer, Integer))->0/1/2
 // in order to avoid code bloating, for non-constant tuple, we make sure that the elements
 // in the tuple should have same data type, and for constant tuple, elements can be any integer
 // data type, we cast all of them into UInt64
@@ -126,8 +112,7 @@ public:
             throw Exception(
                 "Illegal column of arguments of function " + getName() + ", tuple should have exactly two elements.",
                 ErrorCodes::ILLEGAL_COLUMN);
-        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right)
-        {
+        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right) {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
@@ -138,7 +123,9 @@ public:
 
             using OpImpl = TupleHammingDistanceImpl<T0, T1>;
 
-            // constant tuple - constant tuple
+            // we can not useDefaultImplementationForConstants,
+            // because with that, tupleHammingDistance((10, 300), (10, 20)) does not work,
+            // since 10 has data type UInt8, and 300 has data type UInt16
             if (const ColumnConst * const_col_left = checkAndGetColumnConst<ColumnTuple>(arg1.column.get()))
             {
                 if (const ColumnConst * const_col_right = checkAndGetColumnConst<ColumnTuple>(arg2.column.get()))
diff --git a/tests/queries/0_stateless/01016_simhash_minhash.reference b/tests/queries/0_stateless/01016_simhash_minhash.reference
new file mode 100644
index 00000000000..7fa70b343a4
--- /dev/null
+++ b/tests/queries/0_stateless/01016_simhash_minhash.reference
@@ -0,0 +1,59 @@
+0
+2718169299
+2718169299
+3333471646
+26585365
+4124079607
+4124079607
+4124079607
+979945684
+(3614688582,3614688582)
+(3614688582,3614688582)
+(765622645,765622645)
+(765622645,765622645)
+(765622645,765622645)
+(765622645,765622645)
+(3573094983,3573094983)
+(3573094983,3573094983)
+(3604768422,3604768422)
+(3604768422,3604768422)
+(3614688582,1599892600)
+(3614688582,1599892600)
+(3614688582,1599892600)
+(3614688582,1599892600)
+(3614688582,1599892600)
+(3614688582,1599892600)
+(3614688582,996508363)
+(3614688582,996508363)
+2548869326
+2548869326
+401385678
+401385710
+4258739090
+4260836242
+718415633
+718681881
+2314703251
+1238864275
+3900085650
+3907425682
+2314703251
+1238864275
+3569207545
+3568143609
+(1436198067,1436198067)
+(1436198067,1436198067)
+(3846780865,3846780865)
+(1956854492,1956854492)
+(2929435161,2929435161)
+(2929435161,2929435161)
+(3310088565,3310088565)
+(3310088565,3310088565)
+(3614688582,1294895121)
+(3614688582,1294895121)
+(3614688582,1138551650)
+(3614688582,1138551650)
+(3614688582,1294895121)
+(3614688582,1294895121)
+(3614688582,2840007763)
+(3614688582,929186815)
diff --git a/tests/queries/0_stateless/01016_simhash_minhash.sql b/tests/queries/0_stateless/01016_simhash_minhash.sql
new file mode 100644
index 00000000000..9e87216d26f
--- /dev/null
+++ b/tests/queries/0_stateless/01016_simhash_minhash.sql
@@ -0,0 +1,47 @@
+SELECT ngramSimhash('');
+SELECT ngramSimhash('what a cute cat.');
+SELECT ngramSimhashCaseInsensitive('what a cute cat.');
+SELECT ngramSimhashUTF8('what a cute cat.');
+SELECT ngramSimhashCaseInsensitiveUTF8('what a cute cat.');
+SELECT wordShingleSimhash('what a cute cat.');
+SELECT wordShingleSimhashCaseInsensitive('what a cute cat.');
+SELECT wordShingleSimhashUTF8('what a cute cat.');
+SELECT wordShingleSimhashCaseInsensitiveUTF8('what a cute cat.');
+
+SELECT ngramMinhash('');
+SELECT ngramMinhash('what a cute cat.');
+SELECT ngramMinhashCaseInsensitive('what a cute cat.');
+SELECT ngramMinhashUTF8('what a cute cat.');
+SELECT ngramMinhashCaseInsensitiveUTF8('what a cute cat.');
+SELECT wordShingleMinhash('what a cute cat.');
+SELECT wordShingleMinhashCaseInsensitive('what a cute cat.');
+SELECT wordShingleMinhashUTF8('what a cute cat.');
+SELECT wordShingleMinhashCaseInsensitiveUTF8('what a cute cat.');
+
+DROP TABLE IF EXISTS defaults;
+CREATE TABLE defaults
+(
+   s String
+)ENGINE = Memory();
+
+INSERT INTO defaults values ('It is the latest occurrence of the Southeast European haze, the issue that occurs in constant intensity during every wet season. It has mainly been caused by forest fires resulting from illegal slash-and-burn clearing performed on behalf of the palm oil industry in Kazakhstan, principally on the islands, which then spread quickly in the dry season.') ('It is the latest occurrence of the Southeast Asian haze, the issue that occurs in constant intensity during every wet season. It has mainly been caused by forest fires resulting from illegal slash-and-burn clearing performed on behalf of the palm oil industry in Kazakhstan, principally on the islands, which then spread quickly in the dry season.');
+
+SELECT ngramSimhash(s) FROM defaults;
+SELECT ngramSimhashCaseInsensitive(s) FROM defaults;
+SELECT ngramSimhashUTF8(s) FROM defaults;
+SELECT ngramSimhashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT wordShingleSimhash(s) FROM defaults;
+SELECT wordShingleSimhashCaseInsensitive(s) FROM defaults;
+SELECT wordShingleSimhashUTF8(s) FROM defaults;
+SELECT wordShingleSimhashCaseInsensitiveUTF8(s) FROM defaults;
+
+SELECT ngramMinhash(s) FROM defaults;
+SELECT ngramMinhashCaseInsensitive(s) FROM defaults;
+SELECT ngramMinhashUTF8(s) FROM defaults;
+SELECT ngramMinhashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT wordShingleMinhash(s) FROM defaults;
+SELECT wordShingleMinhashCaseInsensitive(s) FROM defaults;
+SELECT wordShingleMinhashUTF8(s) FROM defaults;
+SELECT wordShingleMinhashCaseInsensitiveUTF8(s) FROM defaults;
+
+DROP TABLE defaults;
diff --git a/tests/queries/0_stateless/01017_bithamming_distance.reference b/tests/queries/0_stateless/01017_bithamming_distance.reference
new file mode 100644
index 00000000000..cc2d4f39154
--- /dev/null
+++ b/tests/queries/0_stateless/01017_bithamming_distance.reference
@@ -0,0 +1,15 @@
+1
+7
+63
+2
+1
+3
+5
+4
+6
+6
+6
+3
+5
+9
+9
diff --git a/tests/queries/0_stateless/01017_bithamming_distance.sql b/tests/queries/0_stateless/01017_bithamming_distance.sql
new file mode 100644
index 00000000000..4b36894b97c
--- /dev/null
+++ b/tests/queries/0_stateless/01017_bithamming_distance.sql
@@ -0,0 +1,20 @@
+SELECT bitHammingDistance(1, 5);
+SELECT bitHammingDistance(100, 100000);
+SELECT bitHammingDistance(-1, 1);
+
+DROP TABLE IF EXISTS defaults;
+CREATE TABLE defaults
+(
+	n1 UInt8,
+	n2 UInt16,
+	n3 UInt32,
+	n4 UInt64
+)ENGINE = Memory();
+
+INSERT INTO defaults VALUES (1, 2, 3, 4) (12, 4345, 435, 1233) (45, 675, 32343, 54566) (90, 784, 9034, 778752);
+
+SELECT bitHammingDistance(4, n1) FROM defaults;
+SELECT bitHammingDistance(n2, 100) FROM defaults;
+SELECT bitHammingDistance(n3, n4) FROM defaults;
+
+DROP TABLE defaults;
diff --git a/tests/queries/0_stateless/01017_tuplehamming_distance.reference b/tests/queries/0_stateless/01017_tuplehamming_distance.reference
new file mode 100644
index 00000000000..017ffb0cd33
--- /dev/null
+++ b/tests/queries/0_stateless/01017_tuplehamming_distance.reference
@@ -0,0 +1,15 @@
+2
+1
+1
+0
+2
+2
+2
+2
+1
+2
+2
+2
+0
+2
+2
diff --git a/tests/queries/0_stateless/01017_tuplehamming_distance.sql b/tests/queries/0_stateless/01017_tuplehamming_distance.sql
new file mode 100644
index 00000000000..d0ed1cee096
--- /dev/null
+++ b/tests/queries/0_stateless/01017_tuplehamming_distance.sql
@@ -0,0 +1,19 @@
+SELECT tupleHammingDistance((1, 2), (3, 4));
+SELECT tupleHammingDistance((120, 243), (120, 434));
+SELECT tupleHammingDistance((-12, 434), (434, 434));
+
+DROP TABLE IF EXISTS defaults;
+CREATE TABLE defaults
+(
+	t1 Tuple(UInt16, UInt16),
+	t2 Tuple(UInt32, UInt32),
+	t3 Tuple(Int64, Int64)
+)ENGINE = Memory();
+
+INSERT INTO defaults VALUES ((12, 43), (12312, 43453) ,(-10, 32)) ((1, 4), (546, 12345), (546, 12345)) ((90, 9875), (43456, 234203), (1231, -123)) ((87, 987), (545645, 768354634), (9123, 909));
+
+SELECT tupleHammingDistance((12, 43), t1) FROM defaults;
+SELECT tupleHammingDistance(t2, (546, 456)) FROM defaults;
+SELECT tupleHammingDistance(t2, t3) FROM defaults;
+
+DROP TABLE defaults;

From 7b4fc7300c85b38a272a276ff860e226f33a578a Mon Sep 17 00:00:00 2001
From: feng lv <fenglv15@mails.ucas.ac.cn>
Date: Wed, 10 Jun 2020 23:02:58 +0800
Subject: [PATCH 007/504] update

fix

fix

fix
---
 src/Functions/ExtractString.h                 |  57 ++----
 src/Functions/FunctionsStringHash.cpp         | 167 +++++++++---------
 src/Functions/FunctionsStringHash.h           |  84 +++------
 .../01016_simhash_minhash.reference           |  59 +++----
 4 files changed, 145 insertions(+), 222 deletions(-)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index f6a7394a9fc..51d6f17380c 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -1,3 +1,5 @@
+#include <Common/PODArray.h>
+#include <Common/StringUtils/StringUtils.h>
 #include <Common/UTF8Helpers.h>
 
 #include <algorithm>
@@ -19,6 +21,9 @@ struct ExtractStringImpl
 {
     static constexpr size_t default_padding = 16;
 
+    // the length of code_points = default_padding + N -1
+    // pos: the current beginning location that we want to copy data
+    // end: the end loction of the string
     static ALWAYS_INLINE size_t readASCIICodePoints(UInt8 * code_points, const char *& pos, const char * end)
     {
         /// Offset before which we copy some data.
@@ -47,32 +52,8 @@ struct ExtractStringImpl
         return default_padding;
     }
 
-    // used by FunctionsStringHash
-    // it's not easy to add padding for ColumnString, so we need safety check each memcpy
-    static ALWAYS_INLINE size_t readASCIICodePointsNoPadding(UInt8 * code_points, const char *& pos, const char * end)
-    {
-        constexpr size_t padding_offset = default_padding - N + 1;
-        memcpy(code_points, code_points + padding_offset, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt8));
-
-        // safety check
-        size_t cpy_size = (pos + padding_offset > end) ? end - pos : padding_offset;
-
-        memcpy(code_points + (N - 1), pos, cpy_size * sizeof(UInt8));
-
-        if constexpr (CaseInsensitive)
-        {
-            unrollLowering<N - 1>(code_points, std::make_index_sequence<padding_offset>());
-        }
-        pos += padding_offset;
-        if (pos > end)
-            return default_padding - (pos - end);
-        return default_padding;
-    }
-
-    // read a ASCII word from pos to word
-    // if the word size exceeds max_word_size, only read max_word_size byte
-    // in  FuntionsStringHash, the default value of max_word_size is 128
-    static ALWAYS_INLINE inline size_t readOneASCIIWord(UInt8 * word, const char *& pos, const char * end, size_t max_word_size)
+    // read a ASCII word
+    static ALWAYS_INLINE inline size_t readOneASCIIWord(PaddedPODArray<UInt8> & word_buf, const char *& pos, const char * end)
     {
         // jump seperators
         while (pos < end && !isAlphaNumericASCII(*pos))
@@ -83,14 +64,12 @@ struct ExtractStringImpl
         while (pos < end && isAlphaNumericASCII(*pos))
             ++pos;
 
-        size_t word_size = std::min<size_t>(pos - word_start, max_word_size);
-
-        memcpy(word, word_start, word_size);
+        word_buf.assign(word_start, pos);
         if (CaseInsensitive)
         {
-            std::transform(word, word + word_size, word, [](UInt8 c) { return std::tolower(c); });
+            std::transform(word_buf.begin(), word_buf.end(), word_buf.begin(), [](UInt8 c) { return std::tolower(c); });
         }
-        return word_size;
+        return word_buf.size();
     }
 
     static ALWAYS_INLINE inline size_t readUTF8CodePoints(UInt32 * code_points, const char *& pos, const char * end)
@@ -106,27 +85,21 @@ struct ExtractStringImpl
     }
 
     // read one UTF8 word from pos to word
-    // also, we assume that one word size cann't exceed max_word_size with default value 128
-    static ALWAYS_INLINE inline size_t readOneUTF8Word(UInt32 * word, const char *& pos, const char * end, size_t max_word_size)
+    static ALWAYS_INLINE inline size_t readOneUTF8Word(PaddedPODArray<UInt32> & word_buf, const char *& pos, const char * end)
     {
         // jump UTF8 seperator
         while (pos < end && isUTF8Sep(*pos))
             ++pos;
+        word_buf.clear();
         // UTF8 word's character number
-        size_t num = 0;
-        while (pos < end && num < max_word_size && !isUTF8Sep(*pos))
+        while (pos < end && !isUTF8Sep(*pos))
         {
-            word[num++] = readOneUTF8Code(pos, end);
+            word_buf.push_back(readOneUTF8Code(pos, end));
         }
-        return num;
+        return word_buf.size();
     }
 
 private:
-    static ALWAYS_INLINE inline bool isAlphaNumericASCII(const UInt8 c)
-    {
-        return (c >= 48 && c <= 57) || (c >= 65 && c <= 90) || (c >= 97 && c <= 122);
-    }
-
     template <size_t Offset, typename Container, size_t... I>
     static ALWAYS_INLINE inline void unrollLowering(Container & cont, const std::index_sequence<I...> &)
     {
diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 2195ff7c703..f8c78a808b3 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -5,11 +5,15 @@
 #include <Functions/FunctionsHashing.h>
 #include <Common/HashTable/ClearableHashMap.h>
 #include <Common/HashTable/Hash.h>
+#include <Common/PODArray.h>
 
 #include <Core/Defines.h>
 
 #include <bitset>
+#include <functional>
+#include <memory>
 #include <tuple>
+#include <vector>
 #include <common/unaligned.h>
 
 namespace DB
@@ -64,6 +68,11 @@ struct Hash
 };
 
 // Simhash String -> UInt64
+// N: the length of ngram or words shingles
+// CodePoint: UInt8(ASCII) or UInt32(UTF8)
+// UTF8: means ASCII or UTF8, these two parameters CodePoint and UTF8 can only be (UInt8, false) or (UInt32, true)
+// Ngram: means ngram(true) or words shingles(false)
+// CaseInsensitive: means should we consider about letter case or not
 template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimhashImpl
 {
@@ -71,7 +80,6 @@ struct SimhashImpl
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
     // we made an assumption that the size of one word cann't exceed 128, which may not true
     // if some word's size exceed 128, it would be cut up to several word
-    static constexpr size_t max_word_size = 1u << 7;
     static constexpr size_t max_string_size = 1u << 15;
     static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
 
@@ -135,7 +143,7 @@ struct SimhashImpl
     static ALWAYS_INLINE inline UInt64 wordShinglesCalculateHashValue(
         const char * data,
         size_t size,
-        size_t (*read_one_word)(CodePoint *, const char *&, const char *, size_t),
+        size_t (*read_one_word)(PaddedPODArray<CodePoint> &, const char *&, const char *),
         UInt64 (*hash_functor)(const UInt64 *, size_t, size_t))
     {
         const char * start = data;
@@ -146,16 +154,15 @@ struct SimhashImpl
         // a array to store N word hash values
         UInt64 nword_hashes[N] = {};
         // word buffer to store one word
-        CodePoint word_buf[max_word_size] = {};
-        size_t word_size;
+        PaddedPODArray<CodePoint> word_buf;
         // get first word shingle
         for (size_t i = 0; i < N && start < end; ++i)
         {
-            word_size = read_one_word(word_buf, start, end, max_word_size);
-            if (word_size)
+            read_one_word(word_buf, start, end);
+            if (!word_buf.empty())
             {
                 // for each word, calculate a hash value and stored into the array
-                nword_hashes[i++] = Hash::hashSum(word_buf, word_size);
+                nword_hashes[i++] = Hash::hashSum(word_buf.data(), word_buf.size());
             }
         }
 
@@ -168,14 +175,14 @@ struct SimhashImpl
         }
 
         size_t offset = 0;
-        while (start < end && (word_size = read_one_word(word_buf, start, end, max_word_size)))
+        while (start < end && read_one_word(word_buf, start, end))
         {
             // we need to store the new word hash value to the oldest location.
             // for example, N = 5, array |a0|a1|a2|a3|a4|, now , a0 is the oldest location,
             // so we need to store new word hash into location of a0, then ,this array become
             // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
             // word hash value into locaion of a1, then array become |a5|a6|a2|a3|a4|
-            nword_hashes[offset] = Hash::hashSum(word_buf, word_size);
+            nword_hashes[offset] = Hash::hashSum(word_buf.data(), word_buf.size());
             offset = (offset + 1) % N;
             // according to the word hash storation way, in order to not lose the word shingle's
             // sequence information, when calculation word shingle hash value, we need provide the offset
@@ -203,7 +210,7 @@ struct SimhashImpl
         if constexpr (Ngram)
         {
             if constexpr (!UTF8)
-                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePointsNoPadding, Hash::ngramASCIIHash);
+                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
             else
                 return calc_func(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
         }
@@ -216,17 +223,7 @@ struct SimhashImpl
         }
     }
 
-    // constant string
-    static inline void constant(const String data, UInt64 & res)
-    {
-        if constexpr (Ngram)
-            res = dispatch(ngramCalculateHashValue, data.data(), data.size());
-        else
-            res = dispatch(wordShinglesCalculateHashValue, data.data(), data.size());
-    }
-
-    // non-constant string
-    static void vector(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, PaddedPODArray<UInt64> & res)
+    static void apply(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, PaddedPODArray<UInt64> & res)
     {
         for (size_t i = 0; i < offsets.size(); ++i)
         {
@@ -239,53 +236,64 @@ struct SimhashImpl
                 else
                     res[i] = dispatch(wordShinglesCalculateHashValue, one_data, data_size);
             }
+            else
+                res[i] = -1ull;
         }
     }
 };
 
+template <typename F, size_t K, size_t v>
+class FixedHeap
+{
+public:
+    FixedHeap() = delete;
+
+    explicit FixedHeap(F f_) : f(f_), data_t(std::make_shared<std::vector<size_t>>(K, v))
+    {
+        std::make_heap(data_t->begin(), data_t->end(), f);
+    }
+
+    void insertAndReplace(size_t new_v)
+    {
+        data_t->push_back(new_v);
+        std::push_heap(data_t->begin(), data_t->end(), f);
+        std::pop_heap(data_t->begin(), data_t->end(), f);
+        data_t->pop_back();
+    }
+
+    const size_t * data() { return data_t->data(); }
+
+private:
+    F f;
+    std::shared_ptr<std::vector<size_t>> data_t;
+};
+
+
 // Minhash: String -> Tuple(UInt64, UInt64)
 // for each string, we extract ngram or word shingle,
 // for each ngram or word shingle, calculate a hash value,
 // then we take the K minimum hash values to calculate a hashsum,
 // and take the K maximum hash values to calculate another hashsum,
 // return this two hashsum: Tuple(hashsum1, hashsum2)
+//
+// N: the length of ngram or words shingles
+// K: the number of minimum hashes and maximum hashes that we keep
+// CodePoint: UInt8(ASCII) or UInt32(UTF8)
+// UTF8: means ASCII or UTF8, these two parameters CodePoint and UTF8 can only be (UInt8, false) or (UInt32, true)
+// Ngram: means ngram(true) or words shingles(false)
+// CaseInsensitive: means should we consider about letter case or not
 template <size_t N, size_t K, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct MinhashImpl
 {
+    using Less = std::less<size_t>;
+    using Greater = std::greater<size_t>;
+    using MaxHeap = FixedHeap<std::less<size_t>, K, -1ULL>;
+    using MinHeap = FixedHeap<std::greater<size_t>, K, 0>;
     using ResultType = UInt64;
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    static constexpr size_t max_word_size = 1u << 7;
     static constexpr size_t max_string_size = 1u << 15;
     static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
 
-    // insert a new value into K minimum hash array if this value
-    // is smaller than the greatest value in the array
-    static ALWAYS_INLINE inline void insertMinValue(UInt64 * hashes, UInt64 v)
-    {
-        size_t i = 0;
-        for (; i < K && hashes[i] <= v; ++i)
-            ;
-        if (i == K)
-            return;
-        for (size_t j = K - 2; j >= i; --j)
-            hashes[j + 1] = hashes[j];
-        hashes[i] = v;
-    }
-
-    // insert a new value into K maximum hash array if this value
-    // is greater than the smallest value in the array
-    static ALWAYS_INLINE inline void insertMaxValue(UInt64 * hashes, UInt64 v)
-    {
-        int i = K - 1;
-        for (; i >= 0 && hashes[i] >= v; --i)
-            ;
-        if (i < 0)
-            return;
-        for (int j = 1; j <= i; ++j)
-            hashes[j - 1] = hashes[j];
-        hashes[i] = v;
-    }
-
     // Minhash ngram calculate function, String -> Tuple(UInt64, UInt64)
     // we extract ngram from input string, and calculate a hash value for each ngram
     // then we take the K minimum hash values to calculate a hashsum,
@@ -300,8 +308,8 @@ struct MinhashImpl
         const char * start = data;
         const char * end = data + size;
         // we just maintain the K minimu and K maximum hash values
-        UInt64 k_minimum[K] = {};
-        UInt64 k_maxinum[K] = {};
+        MaxHeap k_minimum_hashes(Less{});
+        MinHeap k_maximum_hashes(Greater{});
         CodePoint cp[simultaneously_codepoints_num] = {};
 
         size_t found = read_code_points(cp, start, end);
@@ -314,15 +322,15 @@ struct MinhashImpl
                 auto new_hash = hash_functor(cp + iter);
                 // insert the new hash value into array used to store K minimum value
                 // and K maximum value
-                insertMinValue(k_minimum, new_hash);
-                insertMaxValue(k_maxinum, new_hash);
+                k_minimum_hashes.insertAndReplace(new_hash);
+                k_maximum_hashes.insertAndReplace(new_hash);
             }
             iter = 0;
         } while (start < end && (found = read_code_points(cp, start, end)));
 
         // calculate hashsum of the K minimum hash values and K maximum hash values
-        UInt64 res1 = Hash::hashSum(k_maxinum, K);
-        UInt64 res2 = Hash::hashSum(k_maxinum, K);
+        UInt64 res1 = Hash::hashSum(k_minimum_hashes.data(), K);
+        UInt64 res2 = Hash::hashSum(k_maximum_hashes.data(), K);
         return std::make_tuple(res1, res2);
     }
 
@@ -332,47 +340,46 @@ struct MinhashImpl
     static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> wordShinglesCalculateHashValue(
         const char * data,
         size_t size,
-        size_t (*read_one_word)(CodePoint *, const char *&, const char *, size_t),
+        size_t (*read_one_word)(PaddedPODArray<CodePoint> &, const char *&, const char *),
         UInt64 (*hash_functor)(const UInt64 *, size_t, size_t))
     {
         const char * start = data;
         const char * end = start + size;
         // also we just store the K minimu and K maximum hash values
-        UInt64 k_minimum[K] = {};
-        UInt64 k_maxinum[K] = {};
+        MaxHeap k_minimum_hashes(Less{});
+        MinHeap k_maximum_hashes(Greater{});
         // array to store n word hashes
         UInt64 nword_hashes[N] = {};
         // word buffer to store one word
-        CodePoint word_buf[max_word_size] = {};
-        size_t word_size;
+        PaddedPODArray<CodePoint> word_buf;
         // how word shingle hash value calculation and word hash storation is same as we
         // have descripted in Simhash wordShinglesCalculateHashValue function
         for (size_t i = 0; i < N && start < end; ++i)
         {
-            word_size = read_one_word(word_buf, start, end, max_word_size);
-            if (word_size)
+            read_one_word(word_buf, start, end);
+            if (!word_buf.empty())
             {
-                nword_hashes[i++] = Hash::hashSum(word_buf, word_size);
+                nword_hashes[i++] = Hash::hashSum(word_buf.data(), word_buf.size());
             }
         }
 
         auto new_hash = hash_functor(nword_hashes, N, 0);
-        insertMinValue(k_minimum, new_hash);
-        insertMaxValue(k_maxinum, new_hash);
+        k_minimum_hashes.insertAndReplace(new_hash);
+        k_maximum_hashes.insertAndReplace(new_hash);
 
         size_t offset = 0;
-        while (start < end && (word_size = read_one_word(word_buf, start, end, max_word_size)))
+        while (start < end && read_one_word(word_buf, start, end))
         {
-            nword_hashes[offset] = Hash::hashSum(word_buf, word_size);
+            nword_hashes[offset] = Hash::hashSum(word_buf.data(), word_buf.size());
             offset = (offset + 1) % N;
             new_hash = hash_functor(nword_hashes, N, offset);
-            insertMinValue(k_minimum, new_hash);
-            insertMaxValue(k_maxinum, new_hash);
+            k_minimum_hashes.insertAndReplace(new_hash);
+            k_maximum_hashes.insertAndReplace(new_hash);
         }
 
         // calculate hashsum
-        UInt64 res1 = Hash::hashSum(k_minimum, K);
-        UInt64 res2 = Hash::hashSum(k_maxinum, K);
+        UInt64 res1 = Hash::hashSum(k_minimum_hashes.data(), K);
+        UInt64 res2 = Hash::hashSum(k_maximum_hashes.data(), K);
         return std::make_tuple(res1, res2);
     }
 
@@ -382,7 +389,7 @@ struct MinhashImpl
         if constexpr (Ngram)
         {
             if constexpr (!UTF8)
-                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePointsNoPadding, Hash::ngramASCIIHash);
+                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
             else
                 return calc_func(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
         }
@@ -395,17 +402,7 @@ struct MinhashImpl
         }
     }
 
-    // constant string
-    static void constant(const String data, UInt64 & res1, UInt64 & res2)
-    {
-        if constexpr (Ngram)
-            std::tie(res1, res2) = dispatch(ngramCalculateHashValue, data.data(), data.size());
-        else
-            std::tie(res1, res2) = dispatch(wordShinglesCalculateHashValue, data.data(), data.size());
-    }
-
-    // non-constant string
-    static void vector(
+    static void apply(
         const ColumnString::Chars & data,
         const ColumnString::Offsets & offsets,
         PaddedPODArray<UInt64> & res1,
@@ -422,6 +419,8 @@ struct MinhashImpl
                 else
                     std::tie(res1[i], res2[i]) = dispatch(wordShinglesCalculateHashValue, one_data, data_size);
             }
+            else
+                std::tie(res1[i], res2[i]) = std::make_tuple(-1ull, -1ull);
         }
     }
 };
diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index bada7490288..23c6db51e8e 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -44,77 +44,37 @@ public:
         return std::make_shared<DataTypeTuple>(DataTypes{element, element});
     }
 
+    bool useDefaultImplementationForConstants() const override { return true; }
+
     void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
     {
         const ColumnPtr & column = block.getByPosition(arguments[0]).column;
-        const ColumnConst * col_const = typeid_cast<const ColumnConst *>(&*column);
         using ResultType = typename Impl::ResultType;
         if constexpr (is_simhash)
         {
-            if (col_const)
-            {
-                ResultType res{};
-                const String & str_data = col_const->getValue<String>();
-                if (str_data.size() > Impl::max_string_size)
-                {
-                    throw Exception(
-                        "String size is too big for function " + getName() + ". Should be at most " + std::to_string(Impl::max_string_size),
-                        ErrorCodes::TOO_LARGE_STRING_SIZE);
-                }
-                Impl::constant(str_data, res);
-                block.getByPosition(result).column = block.getByPosition(result).type->createColumnConst(1, toField(res));
-            }
-            else
-            {
-                // non const string
-                auto col_res = ColumnVector<ResultType>::create();
-                typename ColumnVector<ResultType>::Container & vec_res = col_res->getData();
-                vec_res.resize(column->size());
-                const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
-                Impl::vector(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_res);
-                block.getByPosition(result).column = std::move(col_res);
-            }
+            // non const string, const case is handled by useDefaultImplementationForConstants.
+            auto col_res = ColumnVector<ResultType>::create();
+            typename ColumnVector<ResultType>::Container & vec_res = col_res->getData();
+            vec_res.resize(column->size());
+            const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
+            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_res);
+            block.getByPosition(result).column = std::move(col_res);
         }
         else // Min hash
         {
-            if (col_const)
-            {
-                ResultType h1, h2;
-                const String & str_data = col_const->getValue<String>();
-                if (str_data.size() > Impl::max_string_size)
-                {
-                    throw Exception(
-                        "String size is too big for function " + getName() + ". Should be at most " + std::to_string(Impl::max_string_size),
-                        ErrorCodes::TOO_LARGE_STRING_SIZE);
-                }
-                Impl::constant(str_data, h1, h2);
-                auto h1_col = ColumnVector<ResultType>::create(1);
-                auto h2_col = ColumnVector<ResultType>::create(1);
-                typename ColumnVector<ResultType>::Container & h1_data = h1_col->getData();
-                typename ColumnVector<ResultType>::Container & h2_data = h2_col->getData();
-                h1_data[0] = h1;
-                h2_data[0] = h2;
-                MutableColumns tuple_columns;
-                tuple_columns.emplace_back(std::move(h1_col));
-                tuple_columns.emplace_back(std::move(h2_col));
-                block.getByPosition(result).column = ColumnTuple::create(std::move(tuple_columns));
-            }
-            else
-            {
-                // non const string
-                auto col_h1 = ColumnVector<ResultType>::create();
-                auto col_h2 = ColumnVector<ResultType>::create();
-                typename ColumnVector<ResultType>::Container & vec_h1 = col_h1->getData();
-                typename ColumnVector<ResultType>::Container & vec_h2 = col_h2->getData();
-                vec_h1.resize(column->size());
-                vec_h2.resize(column->size());
-                const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
-                Impl::vector(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_h1, vec_h2);
-                MutableColumns tuple_columns;
-                tuple_columns.emplace_back(std::move(col_h1));
-                tuple_columns.emplace_back(std::move(col_h2));
-                block.getByPosition(result).column = ColumnTuple::create(std::move(tuple_columns));
-            }
+            // non const string
+            auto col_h1 = ColumnVector<ResultType>::create();
+            auto col_h2 = ColumnVector<ResultType>::create();
+            typename ColumnVector<ResultType>::Container & vec_h1 = col_h1->getData();
+            typename ColumnVector<ResultType>::Container & vec_h2 = col_h2->getData();
+            vec_h1.resize(column->size());
+            vec_h2.resize(column->size());
+            const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
+            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_h1, vec_h2);
+            MutableColumns tuple_columns;
+            tuple_columns.emplace_back(std::move(col_h1));
+            tuple_columns.emplace_back(std::move(col_h2));
+            block.getByPosition(result).column = ColumnTuple::create(std::move(tuple_columns));
         }
     }
 };
diff --git a/tests/queries/0_stateless/01016_simhash_minhash.reference b/tests/queries/0_stateless/01016_simhash_minhash.reference
index 7fa70b343a4..2ababa29d1e 100644
--- a/tests/queries/0_stateless/01016_simhash_minhash.reference
+++ b/tests/queries/0_stateless/01016_simhash_minhash.reference
@@ -7,24 +7,15 @@
 4124079607
 4124079607
 979945684
-(3614688582,3614688582)
-(3614688582,3614688582)
-(765622645,765622645)
-(765622645,765622645)
-(765622645,765622645)
-(765622645,765622645)
-(3573094983,3573094983)
-(3573094983,3573094983)
-(3604768422,3604768422)
-(3604768422,3604768422)
-(3614688582,1599892600)
-(3614688582,1599892600)
-(3614688582,1599892600)
-(3614688582,1599892600)
-(3614688582,1599892600)
-(3614688582,1599892600)
-(3614688582,996508363)
-(3614688582,996508363)
+(3700739653,3614688582)
+(2594676265,556335836)
+(2594676265,556335836)
+(3157724679,410999184)
+(1378962320,1336242123)
+(3277652371,1284714580)
+(3277652371,1284714580)
+(3277652371,1284714580)
+(3140472415,3787127930)
 2548869326
 2548869326
 401385678
@@ -41,19 +32,19 @@
 1238864275
 3569207545
 3568143609
-(1436198067,1436198067)
-(1436198067,1436198067)
-(3846780865,3846780865)
-(1956854492,1956854492)
-(2929435161,2929435161)
-(2929435161,2929435161)
-(3310088565,3310088565)
-(3310088565,3310088565)
-(3614688582,1294895121)
-(3614688582,1294895121)
-(3614688582,1138551650)
-(3614688582,1138551650)
-(3614688582,1294895121)
-(3614688582,1294895121)
-(3614688582,2840007763)
-(3614688582,929186815)
+(1525603924,509999509)
+(1525603924,3764233597)
+(1525603924,2706466536)
+(1525603924,1315689278)
+(3824755630,2122451089)
+(946380879,2122451089)
+(3295904092,4129673330)
+(3295904092,4129673330)
+(138351420,974287950)
+(824220170,974287950)
+(3300081739,2402902535)
+(3300081739,3993394872)
+(138351420,974287950)
+(824220170,974287950)
+(3083836461,957058619)
+(4120380459,90533100)

From 61817b30fc5474599b48b16456f0d2f55f756b59 Mon Sep 17 00:00:00 2001
From: feng lv <fenglv15@mails.ucas.ac.cn>
Date: Wed, 24 Jun 2020 00:28:17 +0800
Subject: [PATCH 008/504] fix

---
 src/Functions/FunctionsStringSimilarity.cpp | 49 +++++++++++++--------
 1 file changed, 30 insertions(+), 19 deletions(-)

diff --git a/src/Functions/FunctionsStringSimilarity.cpp b/src/Functions/FunctionsStringSimilarity.cpp
index 81adb1de26f..cf9d4d6e42a 100644
--- a/src/Functions/FunctionsStringSimilarity.cpp
+++ b/src/Functions/FunctionsStringSimilarity.cpp
@@ -1,6 +1,6 @@
+#include <Functions/FunctionsStringSimilarity.h>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionsHashing.h>
-#include <Functions/FunctionsStringSimilarity.h>
 #include <Common/HashTable/ClearableHashMap.h>
 #include <Common/HashTable/Hash.h>
 #include <Common/UTF8Helpers.h>
@@ -268,8 +268,7 @@ struct NgramDistanceImpl
         size_t distance = second_size;
         if (data_size <= max_string_size)
         {
-            size_t first_size
-                = dispatchSearcher(calculateHaystackStatsAndMetric<false>, data.data(), data_size, common_stats, distance, nullptr);
+            size_t first_size = dispatchSearcher(calculateHaystackStatsAndMetric<false>, data.data(), data_size, common_stats, distance, nullptr);
             /// For !symmetric version we should not use first_size.
             if constexpr (symmetric)
                 res = distance * 1.f / std::max(first_size + second_size, size_t(1));
@@ -313,14 +312,23 @@ struct NgramDistanceImpl
             if (needle_size <= max_string_size && haystack_size <= max_string_size)
             {
                 /// Get needle stats.
-                const size_t needle_stats_size
-                    = dispatchSearcher(calculateNeedleStats<true>, needle, needle_size, common_stats, needle_ngram_storage.get());
+                const size_t needle_stats_size = dispatchSearcher(
+                    calculateNeedleStats<true>,
+                    needle,
+                    needle_size,
+                    common_stats,
+                    needle_ngram_storage.get());
 
                 size_t distance = needle_stats_size;
 
                 /// Combine with haystack stats, return to initial needle stats.
                 const size_t haystack_stats_size = dispatchSearcher(
-                    calculateHaystackStatsAndMetric<true>, haystack, haystack_size, common_stats, distance, haystack_ngram_storage.get());
+                    calculateHaystackStatsAndMetric<true>,
+                    haystack,
+                    haystack_size,
+                    common_stats,
+                    distance,
+                    haystack_ngram_storage.get());
 
                 /// Return to zero array stats.
                 for (size_t j = 0; j < needle_stats_size; ++j)
@@ -382,8 +390,12 @@ struct NgramDistanceImpl
 
                 if (needle_size <= max_string_size && haystack_size <= max_string_size)
                 {
-                    const size_t needle_stats_size
-                        = dispatchSearcher(calculateNeedleStats<true>, needle, needle_size, common_stats, needle_ngram_storage.get());
+                    const size_t needle_stats_size = dispatchSearcher(
+                        calculateNeedleStats<true>,
+                        needle,
+                        needle_size,
+                        common_stats,
+                        needle_ngram_storage.get());
 
                     size_t distance = needle_stats_size;
 
@@ -407,11 +419,15 @@ struct NgramDistanceImpl
 
                 prev_offset = needle_offsets[i];
             }
+
         }
     }
 
     static void vectorConstant(
-        const ColumnString::Chars & data, const ColumnString::Offsets & offsets, std::string needle, PaddedPODArray<Float32> & res)
+        const ColumnString::Chars & data,
+        const ColumnString::Offsets & offsets,
+        std::string needle,
+        PaddedPODArray<Float32> & res)
     {
         /// zeroing our map
         NgramStats common_stats = {};
@@ -437,8 +453,7 @@ struct NgramDistanceImpl
                 size_t haystack_stats_size = dispatchSearcher(
                     calculateHaystackStatsAndMetric<true>,
                     reinterpret_cast<const char *>(haystack),
-                    haystack_size,
-                    common_stats,
+                    haystack_size, common_stats,
                     distance,
                     ngram_storage.get());
                 /// For !symmetric version we should not use haystack_stats_size.
@@ -500,18 +515,14 @@ struct NameNgramSearchUTF8CaseInsensitive
 };
 
 using FunctionNgramDistance = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, false, true>, NameNgramDistance>;
-using FunctionNgramDistanceCaseInsensitive
-    = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, true>, NameNgramDistanceCaseInsensitive>;
+using FunctionNgramDistanceCaseInsensitive = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, true>, NameNgramDistanceCaseInsensitive>;
 using FunctionNgramDistanceUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, false, true>, NameNgramDistanceUTF8>;
-using FunctionNgramDistanceCaseInsensitiveUTF8
-    = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, true>, NameNgramDistanceUTF8CaseInsensitive>;
+using FunctionNgramDistanceCaseInsensitiveUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, true>, NameNgramDistanceUTF8CaseInsensitive>;
 
 using FunctionNgramSearch = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, false, false>, NameNgramSearch>;
-using FunctionNgramSearchCaseInsensitive
-    = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, false>, NameNgramSearchCaseInsensitive>;
+using FunctionNgramSearchCaseInsensitive = FunctionsStringSimilarity<NgramDistanceImpl<4, UInt8, false, true, false>, NameNgramSearchCaseInsensitive>;
 using FunctionNgramSearchUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, false, false>, NameNgramSearchUTF8>;
-using FunctionNgramSearchCaseInsensitiveUTF8
-    = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, false>, NameNgramSearchUTF8CaseInsensitive>;
+using FunctionNgramSearchCaseInsensitiveUTF8 = FunctionsStringSimilarity<NgramDistanceImpl<3, UInt32, true, true, false>, NameNgramSearchUTF8CaseInsensitive>;
 
 
 void registerFunctionsStringSimilarity(FunctionFactory & factory)

From 07b5f9a58f1546a2afe1a65ef084d359b1c3dbf4 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 7 Aug 2020 18:04:51 +0300
Subject: [PATCH 009/504] Fix build.

---
 src/Functions/FunctionsMiscellaneous.h | 2 ++
 src/Functions/FunctionsStringHash.h    | 2 +-
 src/Functions/bitHammingDistance.cpp   | 2 +-
 src/Functions/tupleHammingDistance.cpp | 2 +-
 4 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/src/Functions/FunctionsMiscellaneous.h b/src/Functions/FunctionsMiscellaneous.h
index 5703f72ce2a..6cd11b12bd9 100644
--- a/src/Functions/FunctionsMiscellaneous.h
+++ b/src/Functions/FunctionsMiscellaneous.h
@@ -210,6 +210,8 @@ public:
             if (action.type == ExpressionAction::Type::JOIN || action.type == ExpressionAction::Type::ARRAY_JOIN)
                 throw Exception("Expression with arrayJoin or other unusual action cannot be captured", ErrorCodes::BAD_ARGUMENTS);
 
+std::cerr << "=============== FunctionCaptureOverloadResolver expr " << expression_actions->dumpActions() << std::endl;
+
         std::unordered_map<std::string, DataTypePtr> arguments_map;
 
         const auto & all_arguments = expression_actions->getRequiredColumnsWithTypes();
diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index 23c6db51e8e..64ee7f9fe59 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -46,7 +46,7 @@ public:
 
     bool useDefaultImplementationForConstants() const override { return true; }
 
-    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
+    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) const override
     {
         const ColumnPtr & column = block.getByPosition(arguments[0]).column;
         using ResultType = typename Impl::ResultType;
diff --git a/src/Functions/bitHammingDistance.cpp b/src/Functions/bitHammingDistance.cpp
index 5c13a57c426..21d4aa2c69c 100644
--- a/src/Functions/bitHammingDistance.cpp
+++ b/src/Functions/bitHammingDistance.cpp
@@ -93,7 +93,7 @@ public:
 
     bool useDefaultImplementationForConstants() const override { return true; }
 
-    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
+    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) const override
     {
         auto * left_generic = block.getByPosition(arguments[0]).type.get();
         auto * right_generic = block.getByPosition(arguments[1]).type.get();
diff --git a/src/Functions/tupleHammingDistance.cpp b/src/Functions/tupleHammingDistance.cpp
index 8b3f9a696aa..a0dc938ab17 100644
--- a/src/Functions/tupleHammingDistance.cpp
+++ b/src/Functions/tupleHammingDistance.cpp
@@ -100,7 +100,7 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
-    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
+    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) const override
     {
         const ColumnWithTypeAndName & arg1 = block.getByPosition(arguments[0]);
         const ColumnWithTypeAndName & arg2 = block.getByPosition(arguments[1]);

From 3b8020168f5bea44c5a1429cddd9ada9159017c0 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Sat, 8 Aug 2020 07:44:04 +0300
Subject: [PATCH 010/504] Add simple watchdog

---
 programs/server/Server.cpp | 56 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 56 insertions(+)

diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index ddc5ec080fb..c19a537795a 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -4,6 +4,7 @@
 #include <sys/resource.h>
 #include <sys/stat.h>
 #include <sys/types.h>
+#include <sys/wait.h>
 #include <errno.h>
 #include <pwd.h>
 #include <unistd.h>
@@ -263,6 +264,7 @@ void checkForUsersNotInMainConfig(
 int Server::main(const std::vector<std::string> & /*args*/)
 {
     Poco::Logger * log = &logger();
+
     UseSSL use_ssl;
 
     ThreadStatus thread_status;
@@ -1178,8 +1180,62 @@ int Server::main(const std::vector<std::string> & /*args*/)
 #pragma GCC diagnostic ignored "-Wunused-function"
 #pragma GCC diagnostic ignored "-Wmissing-declarations"
 
+
+void forkAndWatch(char * process_name)
+{
+    std::string original_process_name = process_name;
+
+    memset(process_name, 0, original_process_name.size());
+    strncpy(process_name, "clickhouse-watchdog", original_process_name.size());
+
+    setThreadName("clckhouse-watch");   /// 15 characters
+
+    while (true)
+    {
+        pid_t pid = fork();
+
+        if (-1 == pid)
+        {
+            std::cerr << "Cannot fork\n";
+            exit(1);
+        }
+
+        if (0 == pid)
+        {
+            strncpy(process_name, original_process_name.data(), original_process_name.size());
+            setThreadName("clickhouse-serv");
+            return;
+        }
+
+        int status = 0;
+        if (-1 == waitpid(pid, &status, 0))
+        {
+            std::cerr << "Cannot waitpid\n";
+            exit(2);
+        }
+
+        if (WIFEXITED(status))
+        {
+            std::cerr << fmt::format("Child process exited normally with code {}.\n", WEXITSTATUS(status));
+            exit(status);
+        }
+
+        if (WIFSIGNALED(status))
+            std::cerr << fmt::format("Child process was terminated by signal {}.\n", WTERMSIG(status));
+        else if (WIFSTOPPED(status))
+            std::cerr << fmt::format("Child process was stopped by signal {}.\n", WSTOPSIG(status));
+        else
+            std::cerr << "Child process was not exited normally by unknown reason.\n";
+
+        std::cerr << "Will restart.\n";
+    }
+}
+
+
 int mainEntryClickHouseServer(int argc, char ** argv)
 {
+    forkAndWatch(argv[0]);
+
     DB::Server app;
     try
     {

From ae716e13e0a804d4a7d570287f0a17d00754deb7 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Sat, 8 Aug 2020 07:52:09 +0300
Subject: [PATCH 011/504] Watchdog (experimental)

---
 programs/server/Server.cpp | 108 +++++++++++++++++++------------------
 1 file changed, 55 insertions(+), 53 deletions(-)

diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index c19a537795a..57d8956ff69 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -93,6 +93,8 @@ namespace CurrentMetrics
 namespace
 {
 
+char * process_name; /// argv[0]
+
 void setupTmpPath(Poco::Logger * log, const std::string & path)
 {
     LOG_DEBUG(log, "Setting up {} to store temporary data in it", path);
@@ -261,9 +263,61 @@ void checkForUsersNotInMainConfig(
 }
 
 
+void forkAndWatch(Poco::Logger * log)
+{
+    std::string original_process_name = process_name;
+
+    memset(process_name, 0, original_process_name.size());
+    strncpy(process_name, "clickhouse-watchdog", original_process_name.size());
+
+    setThreadName("clckhouse-watch");   /// 15 characters
+
+    while (true)
+    {
+        pid_t pid = fork();
+
+        if (-1 == pid)
+        {
+            LOG_FATAL(log, "Cannot fork");
+            exit(1);
+        }
+
+        if (0 == pid)
+        {
+            strncpy(process_name, original_process_name.data(), original_process_name.size());
+            setThreadName("clickhouse-serv");
+            return;
+        }
+
+        int status = 0;
+        if (-1 == waitpid(pid, &status, 0))
+        {
+            LOG_FATAL(log, "Cannot waitpid");
+            exit(2);
+        }
+
+        if (WIFEXITED(status))
+        {
+            LOG_INFO(log, "Child process exited normally with code {}.", WEXITSTATUS(status));
+            exit(status);
+        }
+
+        if (WIFSIGNALED(status))
+            LOG_FATAL(log, "Child process was terminated by signal {}.\n", WTERMSIG(status));
+        else if (WIFSTOPPED(status))
+            LOG_FATAL(log, "Child process was stopped by signal {}.\n", WSTOPSIG(status));
+        else
+            LOG_FATAL(log, "Child process was not exited normally by unknown reason.");
+
+        LOG_INFO(log, "Will restart.");
+    }
+}
+
+
 int Server::main(const std::vector<std::string> & /*args*/)
 {
     Poco::Logger * log = &logger();
+    forkAndWatch(log);
 
     UseSSL use_ssl;
 
@@ -1181,61 +1235,9 @@ int Server::main(const std::vector<std::string> & /*args*/)
 #pragma GCC diagnostic ignored "-Wmissing-declarations"
 
 
-void forkAndWatch(char * process_name)
-{
-    std::string original_process_name = process_name;
-
-    memset(process_name, 0, original_process_name.size());
-    strncpy(process_name, "clickhouse-watchdog", original_process_name.size());
-
-    setThreadName("clckhouse-watch");   /// 15 characters
-
-    while (true)
-    {
-        pid_t pid = fork();
-
-        if (-1 == pid)
-        {
-            std::cerr << "Cannot fork\n";
-            exit(1);
-        }
-
-        if (0 == pid)
-        {
-            strncpy(process_name, original_process_name.data(), original_process_name.size());
-            setThreadName("clickhouse-serv");
-            return;
-        }
-
-        int status = 0;
-        if (-1 == waitpid(pid, &status, 0))
-        {
-            std::cerr << "Cannot waitpid\n";
-            exit(2);
-        }
-
-        if (WIFEXITED(status))
-        {
-            std::cerr << fmt::format("Child process exited normally with code {}.\n", WEXITSTATUS(status));
-            exit(status);
-        }
-
-        if (WIFSIGNALED(status))
-            std::cerr << fmt::format("Child process was terminated by signal {}.\n", WTERMSIG(status));
-        else if (WIFSTOPPED(status))
-            std::cerr << fmt::format("Child process was stopped by signal {}.\n", WSTOPSIG(status));
-        else
-            std::cerr << "Child process was not exited normally by unknown reason.\n";
-
-        std::cerr << "Will restart.\n";
-    }
-}
-
-
 int mainEntryClickHouseServer(int argc, char ** argv)
 {
-    forkAndWatch(argv[0]);
-
+    process_name = argv[0];
     DB::Server app;
     try
     {

From 11966f62576e527e26d2f751f7ecf1cdc1cde14b Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <KochetovNicolai@users.noreply.github.com>
Date: Mon, 10 Aug 2020 18:22:08 +0300
Subject: [PATCH 012/504] Update FunctionsMiscellaneous.h

---
 src/Functions/FunctionsMiscellaneous.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/src/Functions/FunctionsMiscellaneous.h b/src/Functions/FunctionsMiscellaneous.h
index 6cd11b12bd9..5703f72ce2a 100644
--- a/src/Functions/FunctionsMiscellaneous.h
+++ b/src/Functions/FunctionsMiscellaneous.h
@@ -210,8 +210,6 @@ public:
             if (action.type == ExpressionAction::Type::JOIN || action.type == ExpressionAction::Type::ARRAY_JOIN)
                 throw Exception("Expression with arrayJoin or other unusual action cannot be captured", ErrorCodes::BAD_ARGUMENTS);
 
-std::cerr << "=============== FunctionCaptureOverloadResolver expr " << expression_actions->dumpActions() << std::endl;
-
         std::unordered_map<std::string, DataTypePtr> arguments_map;
 
         const auto & all_arguments = expression_actions->getRequiredColumnsWithTypes();

From 2067501ead62322c51ec4de0bea469c7e758d8b9 Mon Sep 17 00:00:00 2001
From: feng lv <fenglv15@mails.ucas.ac.cn>
Date: Sun, 16 Aug 2020 15:42:35 +0800
Subject: [PATCH 013/504] fix

---
 src/Functions/bitHammingDistance.cpp   | 23 ++++++++++++-----------
 src/Functions/tupleHammingDistance.cpp | 19 ++++++++++---------
 2 files changed, 22 insertions(+), 20 deletions(-)

diff --git a/src/Functions/bitHammingDistance.cpp b/src/Functions/bitHammingDistance.cpp
index 5c13a57c426..08678689a15 100644
--- a/src/Functions/bitHammingDistance.cpp
+++ b/src/Functions/bitHammingDistance.cpp
@@ -19,21 +19,21 @@ struct BitHammingDistanceImpl
 {
     using ResultType = UInt8;
 
-    static void NO_INLINE vector_vector(const PaddedPODArray<A> & a, const PaddedPODArray<B> & b, PaddedPODArray<ResultType> & c)
+    static void NO_INLINE vectorVector(const PaddedPODArray<A> & a, const PaddedPODArray<B> & b, PaddedPODArray<ResultType> & c)
     {
         size_t size = a.size();
         for (size_t i = 0; i < size; ++i)
             c[i] = apply(a[i], b[i]);
     }
 
-    static void NO_INLINE vector_constant(const PaddedPODArray<A> & a, B b, PaddedPODArray<ResultType> & c)
+    static void NO_INLINE vectorConstant(const PaddedPODArray<A> & a, B b, PaddedPODArray<ResultType> & c)
     {
         size_t size = a.size();
         for (size_t i = 0; i < size; ++i)
             c[i] = apply(a[i], b);
     }
 
-    static void NO_INLINE constant_vector(A a, const PaddedPODArray<B> & b, PaddedPODArray<ResultType> & c)
+    static void NO_INLINE constantVector(A a, const PaddedPODArray<B> & b, PaddedPODArray<ResultType> & c)
     {
         size_t size = b.size();
         for (size_t i = 0; i < size; ++i)
@@ -95,9 +95,10 @@ public:
 
     void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) override
     {
-        auto * left_generic = block.getByPosition(arguments[0]).type.get();
-        auto * right_generic = block.getByPosition(arguments[1]).type.get();
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
+        const auto * left_generic = block.getByPosition(arguments[0]).type.get();
+        const auto * right_generic = block.getByPosition(arguments[1]).type.get();
+        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+        {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
@@ -108,8 +109,8 @@ public:
 
             using OpImpl = BitHammingDistanceImpl<T0, T1>;
 
-            auto col_left_raw = block.getByPosition(arguments[0]).column.get();
-            auto col_right_raw = block.getByPosition(arguments[1]).column.get();
+            const auto col_left_raw = block.getByPosition(arguments[0]).column.get();
+            const auto col_right_raw = block.getByPosition(arguments[1]).column.get();
 
             typename ColVecResult::MutablePtr col_res = nullptr;
             col_res = ColVecResult::create();
@@ -122,7 +123,7 @@ public:
                 if (auto col_right = checkAndGetColumn<ColVecT1>(col_right_raw))
                 {
                     // constant integer - non-constant integer
-                    OpImpl::constant_vector(col_left_const->template getValue<T0>(), col_right->getData(), vec_res);
+                    OpImpl::constantVector(col_left_const->template getValue<T0>(), col_right->getData(), vec_res);
                 }
                 else
                     return false;
@@ -131,10 +132,10 @@ public:
             {
                 if (auto col_right = checkAndGetColumn<ColVecT1>(col_right_raw))
                     // non-constant integer - non-constant integer
-                    OpImpl::vector_vector(col_left->getData(), col_right->getData(), vec_res);
+                    OpImpl::vectorVector(col_left->getData(), col_right->getData(), vec_res);
                 else if (auto col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw))
                     // non-constant integer - constant integer
-                    OpImpl::vector_constant(col_left->getData(), col_right_const->template getValue<T1>(), vec_res);
+                    OpImpl::vectorConstant(col_left->getData(), col_right_const->template getValue<T1>(), vec_res);
                 else
                     return false;
             }
diff --git a/src/Functions/tupleHammingDistance.cpp b/src/Functions/tupleHammingDistance.cpp
index 8b3f9a696aa..c2d0ae66875 100644
--- a/src/Functions/tupleHammingDistance.cpp
+++ b/src/Functions/tupleHammingDistance.cpp
@@ -20,7 +20,7 @@ struct TupleHammingDistanceImpl
 {
     using ResultType = UInt8;
 
-    static void NO_INLINE vector_vector(
+    static void NO_INLINE vectorVector(
         const PaddedPODArray<A> & a1,
         const PaddedPODArray<A> & b1,
         const PaddedPODArray<B> & a2,
@@ -33,7 +33,7 @@ struct TupleHammingDistanceImpl
     }
 
     static void NO_INLINE
-    vector_constant(const PaddedPODArray<A> & a1, const PaddedPODArray<A> & b1, UInt64 a2, UInt64 b2, PaddedPODArray<ResultType> & c)
+    vectorConstant(const PaddedPODArray<A> & a1, const PaddedPODArray<A> & b1, UInt64 a2, UInt64 b2, PaddedPODArray<ResultType> & c)
     {
         size_t size = a1.size();
         for (size_t i = 0; i < size; ++i)
@@ -41,14 +41,14 @@ struct TupleHammingDistanceImpl
     }
 
     static void NO_INLINE
-    constant_vector(UInt64 a1, UInt64 b1, const PaddedPODArray<B> & a2, const PaddedPODArray<B> & b2, PaddedPODArray<ResultType> & c)
+    constantVector(UInt64 a1, UInt64 b1, const PaddedPODArray<B> & a2, const PaddedPODArray<B> & b2, PaddedPODArray<ResultType> & c)
     {
         size_t size = a2.size();
         for (size_t i = 0; i < size; ++i)
             c[i] = apply(a1, a2[i]) + apply(b1, b2[i]);
     }
 
-    static ResultType constant_constant(UInt64 a1, UInt64 b1, UInt64 a2, UInt64 b2) { return apply(a1, a2) + apply(b1, b2); }
+    static ResultType constantConstant(UInt64 a1, UInt64 b1, UInt64 a2, UInt64 b2) { return apply(a1, a2) + apply(b1, b2); }
 
 private:
     static inline UInt8 apply(UInt64 a, UInt64 b) { return a != b; }
@@ -112,7 +112,8 @@ public:
             throw Exception(
                 "Illegal column of arguments of function " + getName() + ", tuple should have exactly two elements.",
                 ErrorCodes::ILLEGAL_COLUMN);
-        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right) {
+        bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right)
+        {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
@@ -137,7 +138,7 @@ public:
                     cols1[1]->get(0, b1);
                     cols2[0]->get(0, a2);
                     cols2[1]->get(0, b2);
-                    auto res = OpImpl::constant_constant(a1.get<UInt64>(), b1.get<UInt64>(), a2.get<UInt64>(), b2.get<UInt64>());
+                    auto res = OpImpl::constantConstant(a1.get<UInt64>(), b1.get<UInt64>(), a2.get<UInt64>(), b2.get<UInt64>());
                     block.getByPosition(result).column = DataTypeUInt8().createColumnConst(const_col_left->size(), toField(res));
                     return true;
                 }
@@ -159,7 +160,7 @@ public:
                     auto col_r1 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(0));
                     auto col_r2 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(1));
                     if (col_r1 && col_r2)
-                        OpImpl::constant_vector(a1.get<UInt64>(), b1.get<UInt64>(), col_r1->getData(), col_r2->getData(), vec_res);
+                        OpImpl::constantVector(a1.get<UInt64>(), b1.get<UInt64>(), col_r1->getData(), col_r2->getData(), vec_res);
                     else
                         return false;
                 }
@@ -179,7 +180,7 @@ public:
                         Field a2, b2;
                         const_cols[0]->get(0, a2);
                         const_cols[1]->get(0, b2);
-                        OpImpl::vector_constant(col_l1->getData(), col_l2->getData(), a2.get<UInt64>(), a2.get<UInt64>(), vec_res);
+                        OpImpl::vectorConstant(col_l1->getData(), col_l2->getData(), a2.get<UInt64>(), a2.get<UInt64>(), vec_res);
                     }
                     // non-constant tuple - non-constant tuple
                     else if (const ColumnTuple * col_right = typeid_cast<const ColumnTuple *>(arg2.column.get()))
@@ -187,7 +188,7 @@ public:
                         auto col_r1 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(0));
                         auto col_r2 = checkAndGetColumn<ColVecT1>(&col_right->getColumn(1));
                         if (col_r1 && col_r2)
-                            OpImpl::vector_vector(col_l1->getData(), col_l2->getData(), col_r1->getData(), col_r2->getData(), vec_res);
+                            OpImpl::vectorVector(col_l1->getData(), col_l2->getData(), col_r1->getData(), col_r2->getData(), vec_res);
                         else
                             return false;
                     }

From 6bc8da633c9121875679b0fa659213476f2b6849 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Sun, 16 Aug 2020 14:52:55 +0300
Subject: [PATCH 014/504] Minor modification

---
 programs/server/Server.cpp | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index 05b05934d35..04d116af3fa 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -309,7 +309,8 @@ void forkAndWatch(Poco::Logger * log)
         else
             LOG_FATAL(log, "Child process was not exited normally by unknown reason.");
 
-        LOG_INFO(log, "Will restart.");
+        exit(status);
+        // LOG_INFO(log, "Will restart.");
     }
 }
 
@@ -317,7 +318,9 @@ void forkAndWatch(Poco::Logger * log)
 int Server::main(const std::vector<std::string> & /*args*/)
 {
     Poco::Logger * log = &logger();
-    forkAndWatch(log);
+
+    if (!isInteractive())
+        forkAndWatch(log);
 
     UseSSL use_ssl;
 

From 9a370a03ef19e6e306e479cd5f7ac445a67fad75 Mon Sep 17 00:00:00 2001
From: feng lv <fenglv15@mails.ucas.ac.cn>
Date: Thu, 10 Sep 2020 15:36:38 +0800
Subject: [PATCH 015/504] fix

fix
---
 src/Functions/FunctionsStringHash.h    | 1 -
 src/Functions/bitHammingDistance.cpp   | 7 +++----
 src/Functions/tupleHammingDistance.cpp | 4 ++--
 3 files changed, 5 insertions(+), 7 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index 64ee7f9fe59..19fea2d4fc6 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -15,7 +15,6 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int ILLEGAL_TYPE_OF_ARGUMENT;
-    extern const int TOO_LARGE_STRING_SIZE;
 }
 
 // FunctionStringHash
diff --git a/src/Functions/bitHammingDistance.cpp b/src/Functions/bitHammingDistance.cpp
index cb79b498aa6..cb34634b00d 100644
--- a/src/Functions/bitHammingDistance.cpp
+++ b/src/Functions/bitHammingDistance.cpp
@@ -97,8 +97,7 @@ public:
     {
         const auto * left_generic = block.getByPosition(arguments[0]).type.get();
         const auto * right_generic = block.getByPosition(arguments[1]).type.get();
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
-        {
+        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
@@ -109,8 +108,8 @@ public:
 
             using OpImpl = BitHammingDistanceImpl<T0, T1>;
 
-            const auto col_left_raw = block.getByPosition(arguments[0]).column.get();
-            const auto col_right_raw = block.getByPosition(arguments[1]).column.get();
+            const auto * const col_left_raw = block.getByPosition(arguments[0]).column.get();
+            const auto * const col_right_raw = block.getByPosition(arguments[1]).column.get();
 
             typename ColVecResult::MutablePtr col_res = nullptr;
             col_res = ColVecResult::create();
diff --git a/src/Functions/tupleHammingDistance.cpp b/src/Functions/tupleHammingDistance.cpp
index 2f0475f3a6c..aa38426d228 100644
--- a/src/Functions/tupleHammingDistance.cpp
+++ b/src/Functions/tupleHammingDistance.cpp
@@ -106,8 +106,8 @@ public:
         const ColumnWithTypeAndName & arg2 = block.getByPosition(arguments[1]);
         const DataTypeTuple & type1 = static_cast<const DataTypeTuple &>(*arg1.type);
         const DataTypeTuple & type2 = static_cast<const DataTypeTuple &>(*arg2.type);
-        auto & left_elems = type1.getElements();
-        auto & right_elems = type2.getElements();
+        const auto & left_elems = type1.getElements();
+        const auto & right_elems = type2.getElements();
         if (left_elems.size() != 2 || right_elems.size() != 2)
             throw Exception(
                 "Illegal column of arguments of function " + getName() + ", tuple should have exactly two elements.",

From 5884b1e79b0eb40d9c39b019d345d9dbc3c45640 Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Fri, 30 Oct 2020 04:23:38 +0300
Subject: [PATCH 016/504] my changes to gitignore

---
 .gitignore                  | 3 +++
 contrib/AMQP-CPP            | 2 +-
 contrib/cyrus-sasl          | 2 +-
 contrib/grpc                | 2 +-
 contrib/jemalloc            | 2 +-
 contrib/libhdfs3            | 2 +-
 contrib/mariadb-connector-c | 2 +-
 contrib/openssl             | 2 +-
 contrib/poco                | 2 +-
 contrib/replxx              | 2 +-
 10 files changed, 12 insertions(+), 9 deletions(-)

diff --git a/.gitignore b/.gitignore
index 52d58e68cb6..6a88dbf59e9 100644
--- a/.gitignore
+++ b/.gitignore
@@ -124,3 +124,6 @@ website/package-lock.json
 
 # Toolchains
 /cmake/toolchain/*
+
+# My changes
+/contrib/*
diff --git a/contrib/AMQP-CPP b/contrib/AMQP-CPP
index d63e1f01658..1c08399ab0a 160000
--- a/contrib/AMQP-CPP
+++ b/contrib/AMQP-CPP
@@ -1 +1 @@
-Subproject commit d63e1f016582e9faaaf279aa24513087a07bc6e7
+Subproject commit 1c08399ab0ab9e4042ef8e2bbe9e208e5dcbc13b
diff --git a/contrib/cyrus-sasl b/contrib/cyrus-sasl
index 9995bf9d8e1..6054630889f 160000
--- a/contrib/cyrus-sasl
+++ b/contrib/cyrus-sasl
@@ -1 +1 @@
-Subproject commit 9995bf9d8e14f58934d9313ac64f13780d6dd3c9
+Subproject commit 6054630889fd1cd8d0659573d69badcee1e23a00
diff --git a/contrib/grpc b/contrib/grpc
index a6570b863cf..8aea4e168e7 160000
--- a/contrib/grpc
+++ b/contrib/grpc
@@ -1 +1 @@
-Subproject commit a6570b863cf76c9699580ba51c7827d5bffaac43
+Subproject commit 8aea4e168e78f3eb9828080740fc8cb73d53bf79
diff --git a/contrib/jemalloc b/contrib/jemalloc
index 93e27e435ca..026764f1999 160000
--- a/contrib/jemalloc
+++ b/contrib/jemalloc
@@ -1 +1 @@
-Subproject commit 93e27e435cac846028da20cd9b0841fbc9110bd2
+Subproject commit 026764f19995c53583ab25a3b9c06a2fd74e4689
diff --git a/contrib/libhdfs3 b/contrib/libhdfs3
index 30552ac527f..1b666578c85 160000
--- a/contrib/libhdfs3
+++ b/contrib/libhdfs3
@@ -1 +1 @@
-Subproject commit 30552ac527f2c14070d834e171493b2e7f662375
+Subproject commit 1b666578c85094306b061352078022f6350bfab8
diff --git a/contrib/mariadb-connector-c b/contrib/mariadb-connector-c
index 1485b0de3ea..3f512fedf0b 160000
--- a/contrib/mariadb-connector-c
+++ b/contrib/mariadb-connector-c
@@ -1 +1 @@
-Subproject commit 1485b0de3eaa1508dfe49a5ba1e4aa2a71fd8335
+Subproject commit 3f512fedf0ba0f769a1b4852b4bac542d92c5b20
diff --git a/contrib/openssl b/contrib/openssl
index 237260dd6a4..07e96230645 160000
--- a/contrib/openssl
+++ b/contrib/openssl
@@ -1 +1 @@
-Subproject commit 237260dd6a4bca5cb5a321d366a8a9c807957455
+Subproject commit 07e9623064508d15dd61367f960ebe7fc9aecd77
diff --git a/contrib/poco b/contrib/poco
index 757d947235b..297fc905e16 160000
--- a/contrib/poco
+++ b/contrib/poco
@@ -1 +1 @@
-Subproject commit 757d947235b307675cff964f29b19d388140a9eb
+Subproject commit 297fc905e166392156f83b96aaa5f44e8a6a35c4
diff --git a/contrib/replxx b/contrib/replxx
index 8cf626c04e9..94b1f568d16 160000
--- a/contrib/replxx
+++ b/contrib/replxx
@@ -1 +1 @@
-Subproject commit 8cf626c04e9a74313fb0b474cdbe2297c0f3cdc8
+Subproject commit 94b1f568d16183214d26c7c0e9ce69a4ce407f65

From 078a52ae62c014ad0a2a37bdd8b94679301c30b4 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Thu, 5 Nov 2020 04:35:44 +0300
Subject: [PATCH 017/504] Updated description

---
 docs/en/sql-reference/functions/type-conversion-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index 5c8dc5fd272..c8a88545e97 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -325,7 +325,7 @@ This function accepts a number or date or date with time, and returns a FixedStr
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-This function accepts FixedString, and returns UUID. Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
+This function accepts big-endian `FixedString`, and returns `UUID`. Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
 
 ## CAST(x, T) {#type_conversion_function-cast}
 

From b9f287ed76be6505daced918a6cb66b8b49862c5 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 6 Nov 2020 22:22:04 +0300
Subject: [PATCH 018/504] Updated description

---
 .../functions/type-conversion-functions.md            | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index c8a88545e97..f9506606d92 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -325,7 +325,16 @@ This function accepts a number or date or date with time, and returns a FixedStr
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-This function accepts big-endian `FixedString`, and returns `UUID`. Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
+This function accepts big-endian [FixedString](../../sql-reference/data-types/fixedstring), and returns [UUID](../../sql-reference/data-types/uuid). Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
+
+**Syntax**
+
+``` sql
+reinterpretAsUUID(fixed_string)
+```
+**Returned value**
+
+-   `UUID`.
 
 ## CAST(x, T) {#type_conversion_function-cast}
 

From 80f9ba3e5b002a6a86fdbc7a94b4513121e45824 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 6 Nov 2020 22:22:59 +0300
Subject: [PATCH 019/504] Added translation

---
 .../functions/type-conversion-functions.md          | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/docs/ru/sql-reference/functions/type-conversion-functions.md b/docs/ru/sql-reference/functions/type-conversion-functions.md
index 773850b65ce..19d7e68b4e4 100644
--- a/docs/ru/sql-reference/functions/type-conversion-functions.md
+++ b/docs/ru/sql-reference/functions/type-conversion-functions.md
@@ -319,6 +319,19 @@ SELECT toFixedString('foo\0bar', 8) AS s, toStringCutToZero(s) AS s_cut
 
 Функция принимает число или дату или дату-с-временем и возвращает строку, содержащую байты, представляющие соответствующее значение в host order (little endian). При этом, отбрасываются нулевые байты с конца. Например, значение 255 типа UInt32 будет строкой длины 1 байт.
 
+## reinterpretAsUUID {#reinterpretasuuid}
+
+Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
+
+**Syntax**
+
+``` sql
+reinterpretAsUUID(fixed_string)
+```
+**Returned value**
+
+-   `UUID`.
+
 ## CAST(x, T) {#type_conversion_function-cast}
 
 Преобразует x в тип данных t.

From 5d1f67a5e2e4d909cc0c06d485c7eab5896c94b9 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 6 Nov 2020 22:46:45 +0300
Subject: [PATCH 020/504] Fixed links

---
 docs/en/sql-reference/functions/type-conversion-functions.md | 2 +-
 docs/ru/sql-reference/functions/type-conversion-functions.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index f9506606d92..e865abd141d 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -325,7 +325,7 @@ This function accepts a number or date or date with time, and returns a FixedStr
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-This function accepts big-endian [FixedString](../../sql-reference/data-types/fixedstring), and returns [UUID](../../sql-reference/data-types/uuid). Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
+This function accepts big-endian [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring), and returns [UUID](../../sql-reference/data-types/uuid.md#uuid). Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
 
 **Syntax**
 
diff --git a/docs/ru/sql-reference/functions/type-conversion-functions.md b/docs/ru/sql-reference/functions/type-conversion-functions.md
index 19d7e68b4e4..132ff5adb7c 100644
--- a/docs/ru/sql-reference/functions/type-conversion-functions.md
+++ b/docs/ru/sql-reference/functions/type-conversion-functions.md
@@ -321,7 +321,7 @@ SELECT toFixedString('foo\0bar', 8) AS s, toStringCutToZero(s) AS s_cut
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
+Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid.md#uuid). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
 
 **Syntax**
 

From efd24ac0181c07a44c4f2a37c0b9482d70880bf6 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 6 Nov 2020 23:07:31 +0300
Subject: [PATCH 021/504] Fixed links 2

---
 docs/en/sql-reference/functions/type-conversion-functions.md | 2 +-
 docs/ru/sql-reference/functions/type-conversion-functions.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index e865abd141d..71802ad08bf 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -325,7 +325,7 @@ This function accepts a number or date or date with time, and returns a FixedStr
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-This function accepts big-endian [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring), and returns [UUID](../../sql-reference/data-types/uuid.md#uuid). Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
+This function accepts big-endian [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring), and returns [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type). Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
 
 **Syntax**
 
diff --git a/docs/ru/sql-reference/functions/type-conversion-functions.md b/docs/ru/sql-reference/functions/type-conversion-functions.md
index 132ff5adb7c..d874520e22b 100644
--- a/docs/ru/sql-reference/functions/type-conversion-functions.md
+++ b/docs/ru/sql-reference/functions/type-conversion-functions.md
@@ -321,7 +321,7 @@ SELECT toFixedString('foo\0bar', 8) AS s, toStringCutToZero(s) AS s_cut
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid.md#uuid). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
+Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
 
 **Syntax**
 

From 24497801655fc2a4e6caef745ac000a978ba7927 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 6 Nov 2020 23:19:26 +0300
Subject: [PATCH 022/504] Fixed mistakes in translation

---
 docs/ru/sql-reference/functions/type-conversion-functions.md | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/docs/ru/sql-reference/functions/type-conversion-functions.md b/docs/ru/sql-reference/functions/type-conversion-functions.md
index d874520e22b..7ec37bda741 100644
--- a/docs/ru/sql-reference/functions/type-conversion-functions.md
+++ b/docs/ru/sql-reference/functions/type-conversion-functions.md
@@ -323,12 +323,13 @@ SELECT toFixedString('foo\0bar', 8) AS s, toStringCutToZero(s) AS s_cut
 
 Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
 
-**Syntax**
+**Синтаксис**
 
 ``` sql
 reinterpretAsUUID(fixed_string)
 ```
-**Returned value**
+
+**Возвращаемое значение**
 
 -   `UUID`.
 

From b480c3c2d32d0995c036161854c8b53960ddb686 Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Sun, 8 Nov 2020 08:56:56 +0300
Subject: [PATCH 023/504] Docs for the Regexp input format (en)

---
 docs/en/interfaces/formats.md | 107 ++++++++++++++++++++++++++--------
 1 file changed, 84 insertions(+), 23 deletions(-)

diff --git a/docs/en/interfaces/formats.md b/docs/en/interfaces/formats.md
index d310705d1c1..071c0d72d27 100644
--- a/docs/en/interfaces/formats.md
+++ b/docs/en/interfaces/formats.md
@@ -57,6 +57,7 @@ The supported formats are:
 | [XML](#xml)                                                                             | ✗     | ✔      |
 | [CapnProto](#capnproto)                                                                 | ✔     | ✗      |
 | [LineAsString](#lineasstring)                                                           | ✔     | ✗      |
+| [Regexp](#data-format-regexp)                                                           | ✔     | ✗      |
 
 You can control some format processing parameters with the ClickHouse settings. For more information read the [Settings](../operations/settings/settings.md) section.
 
@@ -1290,6 +1291,89 @@ $ cat filename.orc | clickhouse-client --query="INSERT INTO some_table FORMAT OR
 
 To exchange data with Hadoop, you can use [HDFS table engine](../engines/table-engines/integrations/hdfs.md).
 
+## LineAsString {#lineasstring}
+
+In this format, a sequence of string objects separated by a newline character is interpreted as a single value. This format can only be parsed for table with a single field of type [String](../sql-reference/data-types/string.md). The remaining columns must be set to  [DEFAULT](../sql-reference/statements/create/table.md#default) or [MATERIALIZED](../sql-reference/statements/create/table.md#materialized), or omitted.
+
+**Example**
+
+Query:
+
+``` sql
+DROP TABLE IF EXISTS line_as_string;
+CREATE TABLE line_as_string (field String) ENGINE = Memory;
+INSERT INTO line_as_string FORMAT LineAsString "I love apple", "I love banana", "I love orange";
+SELECT * FROM line_as_string;
+```
+
+Result:
+
+``` text
+┌─field─────────────────────────────────────────────┐
+│ "I love apple", "I love banana", "I love orange"; │
+└───────────────────────────────────────────────────┘
+```
+
+## Regexp {#data-format-regexp}
+
+When working with the `Regexp` format, you can use the following settings:
+
+- `format_regexp` — [String](../sql-reference/data-types/string.md). Contains regular expression in the [re2](https://github.com/google/re2/wiki/Syntax) format.
+- `format_regexp_escaping_rule` — [String](../sql-reference/data-types/string.md). The following escaping rules are supported:
+    - CSV (similarly to [CSV](#csv))
+    - JSON (similarly to [JSONEachRow](#jsoneachrow))
+    - Escaped (similarly to [TSV](#tabseparated))
+    - Quoted (similarly to [Values](#data-format-values))
+    - Raw (extracts subpatterns as a whole, no escaping rules)
+- `format_regexp_skip_unmatched` — [UInt8](../sql-reference/data-types/int-uint.md). Defines the need to throw an exeption in case the `format_regexp` expression does not match the imported data. Can be set to `0` or `1`. 
+
+**Usage** 
+
+The regular expression from `format_regexp` setting is applied to every line of imported data. The number of subpatterns in the regular expression must be equal to the number of columns in imported dataset. 
+
+Lines of the imported data must be separated by newline character `'\n'` or DOS-style newline `"\r\n"` (except the `Raw` format, which does not support any escaping characters). 
+
+The content of every matched subpattern is parsed with the method of corresponding data type, according to `format_regexp_escaping_rule` setting. 
+
+If the regular expression does not match the line and `format_regexp_skip_unmatched` is set to 1, the line is silently skipped. If `format_regexp_skip_unmatched` is set to 0, exception is thrown.
+
+**Example**
+
+Consider the file data.tsv:
+
+```text
+id: 1 array: [1,2,3] string: str1 date: 2020-01-01
+id: 2 array: [1,2,3] string: str2 date: 2020-01-02
+id: 3 array: [1,2,3] string: str3 date: 2020-01-03
+```
+and the table:
+
+```sql
+CREATE TABLE imp_regex_table (id UInt32, array Array(UInt32), string String, date Date) ENGINE = Memory;
+```
+
+Import command:
+
+```bash
+$ cat data.tsv | clickhouse-client  --query "INSERT INTO imp_regex_table FORMAT Regexp SETTINGS format_regexp='id: (.+?) array: (.+?) string: (.+?) date: (.+?)', format_regexp_escaping_rule='Escaped', format_regexp_skip_unmatched=0;"
+```
+
+Query:
+
+```sql
+SELECT * FROM imp_regex_table;
+```
+
+Result:
+
+```txt
+┌─id─┬─array───┬─string─┬───────date─┐
+│  1 │ [1,2,3] │ str1   │ 2020-01-01 │
+│  2 │ [1,2,3] │ str2   │ 2020-01-02 │
+│  3 │ [1,2,3] │ str3   │ 2020-01-03 │
+└────┴─────────┴────────┴────────────┘
+```
+
 ## Format Schema {#formatschema}
 
 The file name containing the format schema is set by the setting `format_schema`.
@@ -1315,27 +1399,4 @@ Limitations:
 - In case of parsing error `JSONEachRow` skips all data until the new line (or EOF), so rows must be delimited by `\n` to count errors correctly.
 - `Template` and `CustomSeparated` use delimiter after the last column and delimiter between rows to find the beginning of next row, so skipping errors works only if at least one of them is not empty.
 
-## LineAsString {#lineasstring}
-
-In this format, a sequence of string objects separated by a newline character is interpreted as a single value. This format can only be parsed for table with a single field of type [String](../sql-reference/data-types/string.md). The remaining columns must be set to  [DEFAULT](../sql-reference/statements/create/table.md#default) or [MATERIALIZED](../sql-reference/statements/create/table.md#materialized), or omitted.
-
-**Example**
-
-Query:
-
-``` sql
-DROP TABLE IF EXISTS line_as_string;
-CREATE TABLE line_as_string (field String) ENGINE = Memory;
-INSERT INTO line_as_string FORMAT LineAsString "I love apple", "I love banana", "I love orange";
-SELECT * FROM line_as_string;
-```
-
-Result:
-
-``` text
-┌─field─────────────────────────────────────────────┐
-│ "I love apple", "I love banana", "I love orange"; │
-└───────────────────────────────────────────────────┘
-```
-
 [Original article](https://clickhouse.tech/docs/en/interfaces/formats/) <!--hide-->

From fd93d31950c95c15dd489aedd8641416f3db1321 Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Sun, 8 Nov 2020 22:33:42 +0300
Subject: [PATCH 024/504] Revert "my changes to gitignore"

This reverts commit 5884b1e79b0eb40d9c39b019d345d9dbc3c45640.
---
 .gitignore                  | 3 ---
 contrib/AMQP-CPP            | 2 +-
 contrib/cyrus-sasl          | 2 +-
 contrib/grpc                | 2 +-
 contrib/jemalloc            | 2 +-
 contrib/libhdfs3            | 2 +-
 contrib/mariadb-connector-c | 2 +-
 contrib/openssl             | 2 +-
 contrib/poco                | 2 +-
 contrib/replxx              | 2 +-
 10 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/.gitignore b/.gitignore
index 6a88dbf59e9..52d58e68cb6 100644
--- a/.gitignore
+++ b/.gitignore
@@ -124,6 +124,3 @@ website/package-lock.json
 
 # Toolchains
 /cmake/toolchain/*
-
-# My changes
-/contrib/*
diff --git a/contrib/AMQP-CPP b/contrib/AMQP-CPP
index 1c08399ab0a..d63e1f01658 160000
--- a/contrib/AMQP-CPP
+++ b/contrib/AMQP-CPP
@@ -1 +1 @@
-Subproject commit 1c08399ab0ab9e4042ef8e2bbe9e208e5dcbc13b
+Subproject commit d63e1f016582e9faaaf279aa24513087a07bc6e7
diff --git a/contrib/cyrus-sasl b/contrib/cyrus-sasl
index 6054630889f..9995bf9d8e1 160000
--- a/contrib/cyrus-sasl
+++ b/contrib/cyrus-sasl
@@ -1 +1 @@
-Subproject commit 6054630889fd1cd8d0659573d69badcee1e23a00
+Subproject commit 9995bf9d8e14f58934d9313ac64f13780d6dd3c9
diff --git a/contrib/grpc b/contrib/grpc
index 8aea4e168e7..a6570b863cf 160000
--- a/contrib/grpc
+++ b/contrib/grpc
@@ -1 +1 @@
-Subproject commit 8aea4e168e78f3eb9828080740fc8cb73d53bf79
+Subproject commit a6570b863cf76c9699580ba51c7827d5bffaac43
diff --git a/contrib/jemalloc b/contrib/jemalloc
index 026764f1999..93e27e435ca 160000
--- a/contrib/jemalloc
+++ b/contrib/jemalloc
@@ -1 +1 @@
-Subproject commit 026764f19995c53583ab25a3b9c06a2fd74e4689
+Subproject commit 93e27e435cac846028da20cd9b0841fbc9110bd2
diff --git a/contrib/libhdfs3 b/contrib/libhdfs3
index 1b666578c85..30552ac527f 160000
--- a/contrib/libhdfs3
+++ b/contrib/libhdfs3
@@ -1 +1 @@
-Subproject commit 1b666578c85094306b061352078022f6350bfab8
+Subproject commit 30552ac527f2c14070d834e171493b2e7f662375
diff --git a/contrib/mariadb-connector-c b/contrib/mariadb-connector-c
index 3f512fedf0b..1485b0de3ea 160000
--- a/contrib/mariadb-connector-c
+++ b/contrib/mariadb-connector-c
@@ -1 +1 @@
-Subproject commit 3f512fedf0ba0f769a1b4852b4bac542d92c5b20
+Subproject commit 1485b0de3eaa1508dfe49a5ba1e4aa2a71fd8335
diff --git a/contrib/openssl b/contrib/openssl
index 07e96230645..237260dd6a4 160000
--- a/contrib/openssl
+++ b/contrib/openssl
@@ -1 +1 @@
-Subproject commit 07e9623064508d15dd61367f960ebe7fc9aecd77
+Subproject commit 237260dd6a4bca5cb5a321d366a8a9c807957455
diff --git a/contrib/poco b/contrib/poco
index 297fc905e16..757d947235b 160000
--- a/contrib/poco
+++ b/contrib/poco
@@ -1 +1 @@
-Subproject commit 297fc905e166392156f83b96aaa5f44e8a6a35c4
+Subproject commit 757d947235b307675cff964f29b19d388140a9eb
diff --git a/contrib/replxx b/contrib/replxx
index 94b1f568d16..8cf626c04e9 160000
--- a/contrib/replxx
+++ b/contrib/replxx
@@ -1 +1 @@
-Subproject commit 94b1f568d16183214d26c7c0e9ce69a4ce407f65
+Subproject commit 8cf626c04e9a74313fb0b474cdbe2297c0f3cdc8

From 1721341fe8b4339aee32abce7710f5934e108cff Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Mon, 9 Nov 2020 17:19:12 +0300
Subject: [PATCH 025/504] Improved description and added examples

---
 .../functions/type-conversion-functions.md    | 47 ++++++++++++++++++-
 .../functions/type-conversion-functions.md    | 46 +++++++++++++++++-
 2 files changed, 89 insertions(+), 4 deletions(-)

diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index 71802ad08bf..d55d5a0fecb 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -325,16 +325,59 @@ This function accepts a number or date or date with time, and returns a FixedStr
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-This function accepts big-endian [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring), and returns [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type). Takes 16 bytes string. If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
+This function accepts 16 bytes string, and returns UUID containing bytes representing the corresponding value in network byte order (big-endian). If the string isn't long enough, the functions work as if the string is padded with the necessary number of null bytes to the end. If the string longer than 16 bytes, the extra bytes at the end are ignored. 
 
 **Syntax**
 
 ``` sql
 reinterpretAsUUID(fixed_string)
 ```
+
+**Parameters**
+
+-   `fixed_string` — Big-endian byte string. [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring).
+
 **Returned value**
 
--   `UUID`.
+-   The UUID type value. [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type).
+
+**Examples**
+
+String to UUID.
+
+Query:
+
+``` sql
+SELECT reinterpretAsUUID(reverse(unhex('000102030405060708090a0b0c0d0e0f')))
+```
+
+Result:
+
+``` text
+┌─reinterpretAsUUID(reverse(unhex('000102030405060708090a0b0c0d0e0f')))─┐
+│                                  08090a0b-0c0d-0e0f-0001-020304050607 │
+└───────────────────────────────────────────────────────────────────────┘
+```
+
+Going back and forth from String to UUID.
+
+Query:
+
+``` sql
+WITH
+    generateUUIDv4() AS uuid,
+    identity(lower(hex(reverse(reinterpretAsString(uuid))))) AS str,
+    reinterpretAsUUID(reverse(unhex(str))) AS uuid2
+SELECT uuid = uuid2;
+```
+
+Result:
+
+``` text
+┌─equals(uuid, uuid2)─┐
+│                   1 │
+└─────────────────────┘
+```
 
 ## CAST(x, T) {#type_conversion_function-cast}
 
diff --git a/docs/ru/sql-reference/functions/type-conversion-functions.md b/docs/ru/sql-reference/functions/type-conversion-functions.md
index 7ec37bda741..62b045543d4 100644
--- a/docs/ru/sql-reference/functions/type-conversion-functions.md
+++ b/docs/ru/sql-reference/functions/type-conversion-functions.md
@@ -321,7 +321,7 @@ SELECT toFixedString('foo\0bar', 8) AS s, toStringCutToZero(s) AS s_cut
 
 ## reinterpretAsUUID {#reinterpretasuuid}
 
-Функция принимает шестнадцатибайтную big-endian строку типа [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring) и возвращает [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
+Функция принимает шестнадцатибайтную строку и интерпретирует ее байты в network order (big-endian). Если строка имеет недостаточную длину, то функция работает так, как будто строка дополнена необходимым количетсвом нулевых байт с конца. Если строка длиннее, чем шестнадцать байт, то игнорируются лишние байты с конца.
 
 **Синтаксис**
 
@@ -329,9 +329,51 @@ SELECT toFixedString('foo\0bar', 8) AS s, toStringCutToZero(s) AS s_cut
 reinterpretAsUUID(fixed_string)
 ```
 
+**Параметры**
+
+-   `fixed_string` — Строка с big-endian порядком байтов. [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring).
+
 **Возвращаемое значение**
 
--   `UUID`.
+-   Значение типа UUID. [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type).
+
+**Примеры**
+
+Интерпретация строки как UUID.
+
+Запрос:
+
+``` sql
+SELECT reinterpretAsUUID(reverse(unhex('000102030405060708090a0b0c0d0e0f')))
+```
+
+Результат:
+
+``` text
+┌─reinterpretAsUUID(reverse(unhex('000102030405060708090a0b0c0d0e0f')))─┐
+│                                  08090a0b-0c0d-0e0f-0001-020304050607 │
+└───────────────────────────────────────────────────────────────────────┘
+```
+
+Переход в UUID и обратно.
+
+Запрос:
+
+``` sql
+WITH
+    generateUUIDv4() AS uuid,
+    identity(lower(hex(reverse(reinterpretAsString(uuid))))) AS str,
+    reinterpretAsUUID(reverse(unhex(str))) AS uuid2
+SELECT uuid = uuid2;
+```
+
+Результат:
+
+``` text
+┌─equals(uuid, uuid2)─┐
+│                   1 │
+└─────────────────────┘
+```
 
 ## CAST(x, T) {#type_conversion_function-cast}
 

From 683960725a0646609d4fe4619762758ba6d913ea Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Mon, 9 Nov 2020 17:30:26 +0300
Subject: [PATCH 026/504] Minor fixes

---
 docs/ru/sql-reference/functions/type-conversion-functions.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/ru/sql-reference/functions/type-conversion-functions.md b/docs/ru/sql-reference/functions/type-conversion-functions.md
index 62b045543d4..e80157d70fb 100644
--- a/docs/ru/sql-reference/functions/type-conversion-functions.md
+++ b/docs/ru/sql-reference/functions/type-conversion-functions.md
@@ -331,11 +331,11 @@ reinterpretAsUUID(fixed_string)
 
 **Параметры**
 
--   `fixed_string` — Строка с big-endian порядком байтов. [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring).
+-   `fixed_string` — cтрока с big-endian порядком байтов. [FixedString](../../sql-reference/data-types/fixedstring.md#fixedstring).
 
 **Возвращаемое значение**
 
--   Значение типа UUID. [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type).
+-   Значение типа [UUID](../../sql-reference/data-types/uuid.md#uuid-data-type).
 
 **Примеры**
 

From 4b17188de1b7a1b73ba637baac277d7c1b1bde5f Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Mon, 9 Nov 2020 22:40:32 +0300
Subject: [PATCH 027/504] Minor fix in en. translated to ru. Plus template
 updates (administrative)..

---
 .../template-system-table.md                  |  2 +-
 docs/en/interfaces/formats.md                 |  4 +-
 docs/ru/interfaces/formats.md                 | 98 ++++++++++++++++---
 3 files changed, 89 insertions(+), 15 deletions(-)

diff --git a/docs/_description_templates/template-system-table.md b/docs/_description_templates/template-system-table.md
index 137766a34b6..3fdf9788d79 100644
--- a/docs/_description_templates/template-system-table.md
+++ b/docs/_description_templates/template-system-table.md
@@ -1,4 +1,4 @@
-## system.table_name {#system-tables_table-name}
+# system.table_name {#system-tables_table-name}
 
 Description.
 
diff --git a/docs/en/interfaces/formats.md b/docs/en/interfaces/formats.md
index 071c0d72d27..70cd4d57600 100644
--- a/docs/en/interfaces/formats.md
+++ b/docs/en/interfaces/formats.md
@@ -1293,7 +1293,7 @@ To exchange data with Hadoop, you can use [HDFS table engine](../engines/table-e
 
 ## LineAsString {#lineasstring}
 
-In this format, a sequence of string objects separated by a newline character is interpreted as a single value. This format can only be parsed for table with a single field of type [String](../sql-reference/data-types/string.md). The remaining columns must be set to  [DEFAULT](../sql-reference/statements/create/table.md#default) or [MATERIALIZED](../sql-reference/statements/create/table.md#materialized), or omitted.
+In this format, every line of input data is interpreted as a single string value. This format can only be parsed for table with a single field of type [String](../sql-reference/data-types/string.md). The remaining columns must be set to [DEFAULT](../sql-reference/statements/create/table.md#default) or [MATERIALIZED](../sql-reference/statements/create/table.md#materialized), or omitted.
 
 **Example**
 
@@ -1316,6 +1316,8 @@ Result:
 
 ## Regexp {#data-format-regexp}
 
+Each line of imported data is parsed according to the regular expression.
+
 When working with the `Regexp` format, you can use the following settings:
 
 - `format_regexp` — [String](../sql-reference/data-types/string.md). Contains regular expression in the [re2](https://github.com/google/re2/wiki/Syntax) format.
diff --git a/docs/ru/interfaces/formats.md b/docs/ru/interfaces/formats.md
index 042c62e310c..0d16e999488 100644
--- a/docs/ru/interfaces/formats.md
+++ b/docs/ru/interfaces/formats.md
@@ -1209,22 +1209,10 @@ $ cat filename.orc | clickhouse-client --query="INSERT INTO some_table FORMAT OR
 
 Для обмена данных с Hadoop можно использовать [движок таблиц HDFS](../engines/table-engines/integrations/hdfs.md).
 
-## Схема формата {#formatschema}
-
-Имя файла со схемой записывается в настройке `format_schema`. При использовании форматов `Cap'n Proto` и `Protobuf` требуется указать схему.
-Схема представляет собой имя файла и имя типа в этом файле, разделенные двоеточием, например `schemafile.proto:MessageType`.
-Если файл имеет стандартное расширение для данного формата (например `.proto` для `Protobuf`),
-то можно его не указывать и записывать схему так `schemafile:MessageType`.
-
-Если для ввода/вывода данных используется [клиент](../interfaces/cli.md) в [интерактивном режиме](../interfaces/cli.md#cli_usage), то при записи схемы можно использовать абсолютный путь или записывать путь
-относительно текущей директории на клиенте. Если клиент используется в [batch режиме](../interfaces/cli.md#cli_usage), то в записи схемы допускается только относительный путь, из соображений безопасности.
-
-Если для ввода/вывода данных используется [HTTP-интерфейс](../interfaces/http.md), то файл со схемой должен располагаться на сервере в каталоге,
-указанном в параметре [format_schema_path](../operations/server-configuration-parameters/settings.md#server_configuration_parameters-format_schema_path) конфигурации сервера.
 
 ## LineAsString {#lineasstring}
 
- В этом формате последовательность строковых объектов, разделенных символом новой строки, интерпретируется как одно значение. Парситься может только таблица с единственным полем типа [String](../sql-reference/data-types/string.md). Остальные столбцы должны быть заданы как [DEFAULT](../sql-reference/statements/create/table.md#create-default-values) или [MATERIALIZED](../sql-reference/statements/create/table.md#create-default-values), либо отсутствовать.
+ В этом формате каждая строка импортируемых данных интерпретируется как одно строковое значение. Парситься может только таблица с единственным полем типа [String](../sql-reference/data-types/string.md). Остальные столбцы должны быть заданы как [DEFAULT](../sql-reference/statements/create/table.md#create-default-values) или [MATERIALIZED](../sql-reference/statements/create/table.md#create-default-values), либо отсутствовать.
 
 **Пример**
 
@@ -1245,4 +1233,88 @@ SELECT * FROM line_as_string;
 └───────────────────────────────────────────────────┘
 ```
 
+## Regexp {#data-format-regexp}
+
+Каждая строка импортируемых данных разбирается в соответствии с регулярным выражением. 
+
+При работе с форматом `Regexp` можно использовать следующие параметры:
+
+- `format_regexp` — [String](../sql-reference/data-types/string.md). Строка с регулярным выражением в формате [re2](https://github.com/google/re2/wiki/Syntax).
+- `format_regexp_escaping_rule` — [String](../sql-reference/data-types/string.md). Правило сериализации. Поддерживаются следующие правила:
+    - CSV (как в [CSV](#csv))
+    - JSON (как в [JSONEachRow](#jsoneachrow))
+    - Escaped (как в [TSV](#tabseparated))
+    - Quoted (как в [Values](#data-format-values))
+    - Raw (данные импортируются как есть, без сериализации)
+- `format_regexp_skip_unmatched` — [UInt8](../sql-reference/data-types/int-uint.md). Признак, будет ли генерироваться исключение в случае, если импортируемые данные не соответствуют регулярному выражению `format_regexp`. Может принимать значение `0` или `1`. 
+
+**Использование** 
+
+Регулярное выражение (шаблон) из параметра `format_regexp` применяется к каждой строке импортируемых данных. Количество частей в шаблоне (подшаблонов) должно соответствовать количеству колонок в импортируемых данных. 
+
+Строки импортируемых данных должны разделяться символом новой строки `'\n'` или символами `"\r\n"` (перенос строки в формате DOS), за исключением формата `Raw`, который не поддерживает сериализацию. 
+
+Данные, выделенные по подшаблонам, интерпретируются в соответствии с типом, указанным в параметре `format_regexp_escaping_rule`. 
+
+Если строка импортируемых данных не соответствует регулярному выражению и параметр `format_regexp_skip_unmatched` равен 1, строка просто игнорируется. Если же параметр `format_regexp_skip_unmatched` равен 0, генерируется исключение.
+
+**Пример**
+
+Рассмотрим файл data.tsv:
+
+```text
+id: 1 array: [1,2,3] string: str1 date: 2020-01-01
+id: 2 array: [1,2,3] string: str2 date: 2020-01-02
+id: 3 array: [1,2,3] string: str3 date: 2020-01-03
+```
+и таблицу:
+
+```sql
+CREATE TABLE imp_regex_table (id UInt32, array Array(UInt32), string String, date Date) ENGINE = Memory;
+```
+
+Команда импорта:
+
+```bash
+$ cat data.tsv | clickhouse-client  --query "INSERT INTO imp_regex_table FORMAT Regexp SETTINGS format_regexp='id: (.+?) array: (.+?) string: (.+?) date: (.+?)', format_regexp_escaping_rule='Escaped', format_regexp_skip_unmatched=0;"
+```
+
+Запрос:
+
+```sql
+SELECT * FROM imp_regex_table;
+```
+
+Результат:
+
+```txt
+┌─id─┬─array───┬─string─┬───────date─┐
+│  1 │ [1,2,3] │ str1   │ 2020-01-01 │
+│  2 │ [1,2,3] │ str2   │ 2020-01-02 │
+│  3 │ [1,2,3] │ str3   │ 2020-01-03 │
+└────┴─────────┴────────┴────────────┘
+```
+
+
+## Схема формата {#formatschema}
+
+Имя файла со схемой записывается в настройке `format_schema`. При использовании форматов `Cap'n Proto` и `Protobuf` требуется указать схему.
+Схема представляет собой имя файла и имя типа в этом файле, разделенные двоеточием, например `schemafile.proto:MessageType`.
+Если файл имеет стандартное расширение для данного формата (например `.proto` для `Protobuf`),
+то можно его не указывать и записывать схему так `schemafile:MessageType`.
+
+Если для ввода/вывода данных используется [клиент](../interfaces/cli.md) в [интерактивном режиме](../interfaces/cli.md#cli_usage), то при записи схемы можно использовать абсолютный путь или записывать путь
+относительно текущей директории на клиенте. Если клиент используется в [batch режиме](../interfaces/cli.md#cli_usage), то в записи схемы допускается только относительный путь, из соображений безопасности.
+
+Если для ввода/вывода данных используется [HTTP-интерфейс](../interfaces/http.md), то файл со схемой должен располагаться на сервере в каталоге,
+указанном в параметре [format_schema_path](../operations/server-configuration-parameters/settings.md#server_configuration_parameters-format_schema_path) конфигурации сервера.
+
+## Игнорирование ошибок {#skippingerrors}
+
+Некоторые форматы, такие как `CSV`, `TabSeparated`, `TSKV`, `JSONEachRow`, `Template`, `CustomSeparated` и `Protobuf`, могут игнорировать строки, которые не соответствуют правилам и  разбор которых может вызвать ошибку. При этом обработка импортируемых данных продолжается со следующей строки. См. настройки [input_format_allow_errors_num](../operations/settings/settings.md#settings-input_format_allow_errors_num) и
+[input_format_allow_errors_ratio](../operations/settings/settings.md#settings-input_format_allow_errors_ratio).
+Ограничения:
+- В формате `JSONEachRow` в случае ошибки игнорируются все данные до конца текущей строки (или до конца файла). Поэтому строки должны быть разделены символом `\n`, чтобы ошибки обрабатывались корректно.
+- Форматы `Template` и `CustomSeparated` используют разделитель после последней колонки и разделитель между строками. Поэтому игнорирование ошибок работает только если хотя бы одна из строк не пустая.
+
 [Оригинальная статья](https://clickhouse.tech/docs/ru/interfaces/formats/) <!--hide-->

From 713d809096f07ab6119dd2122e1a99559aa4c8ee Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Mon, 9 Nov 2020 22:50:43 +0300
Subject: [PATCH 028/504] Links fixed (ru)

---
 docs/ru/interfaces/formats.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/ru/interfaces/formats.md b/docs/ru/interfaces/formats.md
index 0d16e999488..dd75927db7a 100644
--- a/docs/ru/interfaces/formats.md
+++ b/docs/ru/interfaces/formats.md
@@ -1311,8 +1311,8 @@ SELECT * FROM imp_regex_table;
 
 ## Игнорирование ошибок {#skippingerrors}
 
-Некоторые форматы, такие как `CSV`, `TabSeparated`, `TSKV`, `JSONEachRow`, `Template`, `CustomSeparated` и `Protobuf`, могут игнорировать строки, которые не соответствуют правилам и  разбор которых может вызвать ошибку. При этом обработка импортируемых данных продолжается со следующей строки. См. настройки [input_format_allow_errors_num](../operations/settings/settings.md#settings-input_format_allow_errors_num) и
-[input_format_allow_errors_ratio](../operations/settings/settings.md#settings-input_format_allow_errors_ratio).
+Некоторые форматы, такие как `CSV`, `TabSeparated`, `TSKV`, `JSONEachRow`, `Template`, `CustomSeparated` и `Protobuf`, могут игнорировать строки, которые не соответствуют правилам и  разбор которых может вызвать ошибку. При этом обработка импортируемых данных продолжается со следующей строки. См. настройки [input_format_allow_errors_num](../operations/settings/settings.md#input_format_allow_errors_num) и
+[input_format_allow_errors_ratio](../operations/settings/settings.md#input_format_allow_errors_ratio).
 Ограничения:
 - В формате `JSONEachRow` в случае ошибки игнорируются все данные до конца текущей строки (или до конца файла). Поэтому строки должны быть разделены символом `\n`, чтобы ошибки обрабатывались корректно.
 - Форматы `Template` и `CustomSeparated` используют разделитель после последней колонки и разделитель между строками. Поэтому игнорирование ошибок работает только если хотя бы одна из строк не пустая.

From fc4410ed5c55e1a938dbb829a7d560ac6372d270 Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Mon, 9 Nov 2020 22:59:34 +0300
Subject: [PATCH 029/504] Now really

---
 docs/ru/interfaces/formats.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/ru/interfaces/formats.md b/docs/ru/interfaces/formats.md
index dd75927db7a..287ae142bdd 100644
--- a/docs/ru/interfaces/formats.md
+++ b/docs/ru/interfaces/formats.md
@@ -1311,8 +1311,8 @@ SELECT * FROM imp_regex_table;
 
 ## Игнорирование ошибок {#skippingerrors}
 
-Некоторые форматы, такие как `CSV`, `TabSeparated`, `TSKV`, `JSONEachRow`, `Template`, `CustomSeparated` и `Protobuf`, могут игнорировать строки, которые не соответствуют правилам и  разбор которых может вызвать ошибку. При этом обработка импортируемых данных продолжается со следующей строки. См. настройки [input_format_allow_errors_num](../operations/settings/settings.md#input_format_allow_errors_num) и
-[input_format_allow_errors_ratio](../operations/settings/settings.md#input_format_allow_errors_ratio).
+Некоторые форматы, такие как `CSV`, `TabSeparated`, `TSKV`, `JSONEachRow`, `Template`, `CustomSeparated` и `Protobuf`, могут игнорировать строки, которые не соответствуют правилам и  разбор которых может вызвать ошибку. При этом обработка импортируемых данных продолжается со следующей строки. См. настройки [input_format_allow_errors_num](../operations/settings/settings.md#input-format-allow-errors-num) и
+[input_format_allow_errors_ratio](../operations/settings/settings.md#input-format-allow-errors-ratio).
 Ограничения:
 - В формате `JSONEachRow` в случае ошибки игнорируются все данные до конца текущей строки (или до конца файла). Поэтому строки должны быть разделены символом `\n`, чтобы ошибки обрабатывались корректно.
 - Форматы `Template` и `CustomSeparated` используют разделитель после последней колонки и разделитель между строками. Поэтому игнорирование ошибок работает только если хотя бы одна из строк не пустая.

From d3d78286424d32c8bc08be779fb785eb2688c6b7 Mon Sep 17 00:00:00 2001
From: Olga Revyakina <revolg@yandex-team.ru>
Date: Mon, 9 Nov 2020 23:07:34 +0300
Subject: [PATCH 030/504] Minor formatting error fixed

---
 docs/en/interfaces/formats.md | 2 +-
 docs/ru/interfaces/formats.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/interfaces/formats.md b/docs/en/interfaces/formats.md
index 70cd4d57600..6c57fb62939 100644
--- a/docs/en/interfaces/formats.md
+++ b/docs/en/interfaces/formats.md
@@ -1368,7 +1368,7 @@ SELECT * FROM imp_regex_table;
 
 Result:
 
-```txt
+```text
 ┌─id─┬─array───┬─string─┬───────date─┐
 │  1 │ [1,2,3] │ str1   │ 2020-01-01 │
 │  2 │ [1,2,3] │ str2   │ 2020-01-02 │
diff --git a/docs/ru/interfaces/formats.md b/docs/ru/interfaces/formats.md
index 287ae142bdd..153653d8cff 100644
--- a/docs/ru/interfaces/formats.md
+++ b/docs/ru/interfaces/formats.md
@@ -1287,7 +1287,7 @@ SELECT * FROM imp_regex_table;
 
 Результат:
 
-```txt
+```text
 ┌─id─┬─array───┬─string─┬───────date─┐
 │  1 │ [1,2,3] │ str1   │ 2020-01-01 │
 │  2 │ [1,2,3] │ str2   │ 2020-01-02 │

From ad9a0c6144c228a8e4c0ade2652fe483be3934bc Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 15:43:37 +0300
Subject: [PATCH 031/504] Update poco

---
 contrib/poco | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/poco b/contrib/poco
index f49c6ab8d3a..4d06db3947a 160000
--- a/contrib/poco
+++ b/contrib/poco
@@ -1 +1 @@
-Subproject commit f49c6ab8d3aa71828bd1b411485c21722e8c9d82
+Subproject commit 4d06db3947ac2343133220a5cd5d9b35bc89814c

From fa9814921c758fb1e5b439843832f51a1fe22ca5 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 17:57:31 +0300
Subject: [PATCH 032/504] Updae boost.

---
 contrib/boost | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/boost b/contrib/boost
index a04e72c0464..d2da6db25de 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit a04e72c0464f0c31d3384f18f0c0db36a05538e0
+Subproject commit d2da6db25deff1a927b166ce62cf8967a40273c9

From 0f293e60e1e411cfcee762625ff5522b119d287d Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 18:22:34 +0300
Subject: [PATCH 033/504] Update CMakeLists.txt for boost

---
 contrib/boost-cmake/CMakeLists.txt | 92 ++++++++++++++++++++++++++++++
 1 file changed, 92 insertions(+)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index fd860c9f9b0..b3bf3b38f8b 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -11,6 +11,7 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         iostreams
         program_options
         regex
+        context
     )
 
     if(Boost_INCLUDE_DIR AND Boost_FILESYSTEM_LIBRARY AND Boost_FILESYSTEM_LIBRARY AND
@@ -27,18 +28,21 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         add_library (_boost_program_options INTERFACE)
         add_library (_boost_regex INTERFACE)
         add_library (_boost_system INTERFACE)
+        add_library (_boost_context INTERFACE)
 
         target_link_libraries (_boost_filesystem INTERFACE ${Boost_FILESYSTEM_LIBRARY})
         target_link_libraries (_boost_iostreams INTERFACE ${Boost_IOSTREAMS_LIBRARY})
         target_link_libraries (_boost_program_options INTERFACE ${Boost_PROGRAM_OPTIONS_LIBRARY})
         target_link_libraries (_boost_regex INTERFACE ${Boost_REGEX_LIBRARY})
         target_link_libraries (_boost_system INTERFACE ${Boost_SYSTEM_LIBRARY})
+        target_link_libraries (_boost_context INTERFACE ${Boost_CONTEXT_LIBRARY})
 
         add_library (boost::filesystem ALIAS _boost_filesystem)
         add_library (boost::iostreams ALIAS _boost_iostreams)
         add_library (boost::program_options ALIAS _boost_program_options)
         add_library (boost::regex ALIAS _boost_regex)
         add_library (boost::system ALIAS _boost_system)
+        add_library (boost::context ALIAS _boost_context)
     else()
         set(EXTERNAL_BOOST_FOUND 0)
         message (${RECONFIGURE_MESSAGE_LEVEL} "Can't find system boost")
@@ -142,4 +146,92 @@ if (NOT EXTERNAL_BOOST_FOUND)
     add_library (_boost_system ${SRCS_SYSTEM})
     add_library (boost::system ALIAS _boost_system)
     target_include_directories (_boost_system PRIVATE ${LIBRARY_DIR})
+
+    # context
+
+    set (SRCS_CONTEXT
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_pe_armasm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_combined_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_gas.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_masm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_x86_64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips32_o32_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips64_n64_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_ppc64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_xcoff_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_xcoff_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_gas.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_masm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_pe_armasm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/make_combined_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_gas.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_masm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_x86_64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_mips32_o32_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_mips64_n64_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_ppc64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_xcoff_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_xcoff_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_gas.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_masm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_pe_armasm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_combined_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_gas.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_masm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_x86_64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips32_o32_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips64_n64_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_ppc64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_xcoff_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_xcoff_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_gas.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_masm.asm
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_elf_gas.S
+        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_macho_gas.S
+        ${LIBRARY_DIR}/libs/context/src/continuation.cpp
+        ${LIBRARY_DIR}/libs/context/src/dummy.cpp
+        ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
+        ${LIBRARY_DIR}/libs/context/src/fiber.cpp
+        ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
+        ${LIBRARY_DIR}/libs/context/src/untested.cpp
+        ${LIBRARY_DIR}/libs/context/src/windows/stack_traits.cpp
+    )
+
+    add_library (_boost_context ${SRCS_CONTEXT})
+    add_library (boost::context ALIAS _boost_context)
+    target_include_directories (_boost_context PRIVATE ${LIBRARY_DIR})
 endif ()

From 4f442cd8f396cc0297052b17b4a50bcb3984d53d Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 18:32:01 +0300
Subject: [PATCH 034/504] Update CMakeLists.txt for boost

---
 contrib/boost-cmake/CMakeLists.txt | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index b3bf3b38f8b..59deea802cf 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -148,6 +148,8 @@ if (NOT EXTERNAL_BOOST_FOUND)
     target_include_directories (_boost_system PRIVATE ${LIBRARY_DIR})
 
     # context
+    enable_language(ASM)
+    SET(ASM_OPTIONS "-x assembler-with-cpp")
 
     set (SRCS_CONTEXT
         ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S

From d50a0e63e6189b98581309f0ed2c3eb4e780b6ef Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 18:46:58 +0300
Subject: [PATCH 035/504] Added example from boost.

---
 src/Interpreters/tests/CMakeLists.txt |  3 +++
 src/Interpreters/tests/context.cpp    | 26 ++++++++++++++++++++++++++
 2 files changed, 29 insertions(+)
 create mode 100644 src/Interpreters/tests/context.cpp

diff --git a/src/Interpreters/tests/CMakeLists.txt b/src/Interpreters/tests/CMakeLists.txt
index 20aa73166fb..9bdedc7b76b 100644
--- a/src/Interpreters/tests/CMakeLists.txt
+++ b/src/Interpreters/tests/CMakeLists.txt
@@ -29,6 +29,9 @@ target_link_libraries (string_hash_map PRIVATE dbms)
 add_executable (string_hash_map_aggregation string_hash_map.cpp)
 target_link_libraries (string_hash_map_aggregation PRIVATE dbms)
 
+add_executable (context context.cpp)
+target_link_libraries (context PRIVATE dbms)
+
 add_executable (two_level_hash_map two_level_hash_map.cpp)
 target_include_directories (two_level_hash_map SYSTEM BEFORE PRIVATE ${SPARSEHASH_INCLUDE_DIR})
 target_link_libraries (two_level_hash_map PRIVATE dbms)
diff --git a/src/Interpreters/tests/context.cpp b/src/Interpreters/tests/context.cpp
new file mode 100644
index 00000000000..eed7ca60790
--- /dev/null
+++ b/src/Interpreters/tests/context.cpp
@@ -0,0 +1,26 @@
+#include <iostream>
+#include <boost/context/fiber.hpp>
+
+int main(int, char **)
+{
+    namespace ctx=boost::context;
+    int a;
+    ctx::fiber source{[&a](ctx::fiber&& sink)
+    {
+        a=0;
+        int b=1;
+        while (true)
+        {
+            sink=std::move(sink).resume();
+            int next=a+b;
+            a=b;
+            b=next;
+        }
+        return std::move(sink);
+    }};
+    for (int j=0;j<10;++j)
+    {
+        source=std::move(source).resume();
+        std::cout << a << " ";
+    }
+}

From 0e043202598010d9ffab947ca475244e66370054 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 19:10:18 +0300
Subject: [PATCH 036/504] Update CMakeLists.txt

---
 contrib/boost-cmake/CMakeLists.txt    | 138 +++++++++++++-------------
 src/Interpreters/tests/CMakeLists.txt |   2 +-
 2 files changed, 70 insertions(+), 70 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 59deea802cf..712aa7cdac0 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -152,78 +152,78 @@ if (NOT EXTERNAL_BOOST_FOUND)
     SET(ASM_OPTIONS "-x assembler-with-cpp")
 
     set (SRCS_CONTEXT
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_pe_armasm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_combined_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_gas.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_masm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_x86_64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips32_o32_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips64_n64_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_ppc64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_xcoff_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_xcoff_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_gas.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_masm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_pe_armasm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_combined_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_gas.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_masm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_x86_64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips32_o32_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips64_n64_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_ppc64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_xcoff_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_xcoff_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_gas.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_masm.asm
         ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_pe_armasm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/make_combined_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_gas.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_masm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_x86_64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_mips32_o32_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_mips64_n64_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_ppc64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_xcoff_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_xcoff_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_gas.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_masm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_pe_armasm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_combined_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_gas.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_masm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_x86_64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_mips32_o32_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_mips64_n64_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_ppc64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_xcoff_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_xcoff_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_gas.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_masm.asm
         ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_pe_armasm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_combined_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_gas.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_masm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_x86_64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips32_o32_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips64_n64_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_ppc64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_xcoff_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_xcoff_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_gas.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_masm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_pe_armasm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_combined_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_gas.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_masm.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_x86_64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips32_o32_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips64_n64_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_ppc64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_xcoff_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_elf_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_xcoff_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_gas.asm
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_masm.asm
         ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_elf_gas.S
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_macho_gas.S
+#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_macho_gas.S
         ${LIBRARY_DIR}/libs/context/src/continuation.cpp
         ${LIBRARY_DIR}/libs/context/src/dummy.cpp
         ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
diff --git a/src/Interpreters/tests/CMakeLists.txt b/src/Interpreters/tests/CMakeLists.txt
index 9bdedc7b76b..f8fa33d8d2e 100644
--- a/src/Interpreters/tests/CMakeLists.txt
+++ b/src/Interpreters/tests/CMakeLists.txt
@@ -30,7 +30,7 @@ add_executable (string_hash_map_aggregation string_hash_map.cpp)
 target_link_libraries (string_hash_map_aggregation PRIVATE dbms)
 
 add_executable (context context.cpp)
-target_link_libraries (context PRIVATE dbms)
+target_link_libraries (context PRIVATE dbms _boost_context)
 
 add_executable (two_level_hash_map two_level_hash_map.cpp)
 target_include_directories (two_level_hash_map SYSTEM BEFORE PRIVATE ${SPARSEHASH_INCLUDE_DIR})

From 319d36a3b76cc7b9000d05cc5b4236e1e4aa6739 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 30 Nov 2020 19:11:56 +0300
Subject: [PATCH 037/504] Update CMakeLists.txt

---
 contrib/boost-cmake/CMakeLists.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 712aa7cdac0..5bf09ca29bd 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -229,8 +229,8 @@ if (NOT EXTERNAL_BOOST_FOUND)
         ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
         ${LIBRARY_DIR}/libs/context/src/fiber.cpp
         ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
-        ${LIBRARY_DIR}/libs/context/src/untested.cpp
-        ${LIBRARY_DIR}/libs/context/src/windows/stack_traits.cpp
+#        ${LIBRARY_DIR}/libs/context/src/untested.cpp
+#        ${LIBRARY_DIR}/libs/context/src/windows/stack_traits.cpp
     )
 
     add_library (_boost_context ${SRCS_CONTEXT})

From 0fae325d76abb49c4a74e73ebe4bcccfcde9f88f Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 2 Dec 2020 14:18:46 +0300
Subject: [PATCH 038/504] Add FiberStack

---
 src/Common/FiberStack.h            | 45 +++++++++++++++++++
 src/Interpreters/tests/context.cpp | 72 ++++++++++++++++++++++++++++--
 2 files changed, 113 insertions(+), 4 deletions(-)
 create mode 100644 src/Common/FiberStack.h

diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
new file mode 100644
index 00000000000..c133cb6250c
--- /dev/null
+++ b/src/Common/FiberStack.h
@@ -0,0 +1,45 @@
+#include <boost/context/stack_context.hpp>
+#include <Common/Allocator.h>
+
+#if defined(BOOST_USE_VALGRIND)
+#include <valgrind/valgrind.h>
+#endif
+
+/// This is an implementation of allocator for fiber stack.
+/// It uses internal allocator, so we track memory usage. It is the main reason why this class is needed.
+/// The reference implementations are pooled_fixedsize_stack and protected_fixedsize_stack from boost::context.
+template <typename TAllocator = Allocator<false, false>>
+class FiberStack
+{
+private:
+    size_t stack_size;
+public:
+    /// 8MB of memory per fiber stack may seem too expensive. It is indeed.
+    /// The reason is that current (patched) libunwind needs > 4MB of stack memory to unwind stack.
+    /// If we allocate less memory, any thrown exception inside fiber will cause segfault.
+    static constexpr size_t default_stack_size = 8 * 1024 * 1024;
+
+    explicit FiberStack(size_t stack_size_ = default_stack_size) : stack_size(stack_size_) {}
+
+    boost::context::stack_context allocate()
+    {
+        void * vp = TAllocator().alloc(stack_size);
+
+        boost::context::stack_context sctx;
+        sctx.size = stack_size;
+        sctx.sp = static_cast< char * >(vp) + sctx.size;
+#if defined(BOOST_USE_VALGRIND)
+        sctx.valgrind_stack_id = VALGRIND_STACK_REGISTER(sctx.sp, vp);
+#endif
+        return sctx;
+    }
+
+    void deallocate(boost::context::stack_context & sctx)
+    {
+#if defined(BOOST_USE_VALGRIND)
+        VALGRIND_STACK_DEREGISTER( sctx.valgrind_stack_id);
+#endif
+        void * vp = static_cast< char * >(sctx.sp) - sctx.size;
+        TAllocator().free(vp, stack_size);
+    }
+};
diff --git a/src/Interpreters/tests/context.cpp b/src/Interpreters/tests/context.cpp
index eed7ca60790..407a7a9ba3b 100644
--- a/src/Interpreters/tests/context.cpp
+++ b/src/Interpreters/tests/context.cpp
@@ -1,26 +1,90 @@
 #include <iostream>
+/// #define BOOST_USE_UCONTEXT
 #include <boost/context/fiber.hpp>
+#include <boost/context/pooled_fixedsize_stack.hpp>
+#include <boost/context/segmented_stack.hpp>
+#include <Common/Exception.h>
+#include <Common/FiberStack.h>
+
+void __attribute__((__noinline__)) foo(std::exception_ptr exception)
+{
+    if (exception)
+        std::rethrow_exception(exception);
+}
+
+void __attribute__((__noinline__)) bar(int a)
+{
+    std::cout << StackTrace().toString() << std::endl;
+
+    if (a > 0)
+        throw DB::Exception(0, "hello");
+}
+
+void __attribute__((__noinline__)) gar(int a)
+{
+    char buf[1024];
+    buf[1023] = a & 255;
+    if (a > 2)
+        return gar(a - 1);
+    else
+        bar(a);
+}
 
 int main(int, char **)
-{
+try {
     namespace ctx=boost::context;
     int a;
-    ctx::fiber source{[&a](ctx::fiber&& sink)
+    std::exception_ptr exception;
+    // ctx::protected_fixedsize allocator
+    // ctx::pooled_fixedsize_stack(1024 * 64 + 2 * 2 * 1024 * 1024 * 16, 1)
+    ctx::fiber source{std::allocator_arg_t(), FiberStack(), [&](ctx::fiber&& sink)
     {
         a=0;
         int b=1;
-        while (true)
+        for (size_t i = 0; i < 9; ++i)
         {
             sink=std::move(sink).resume();
             int next=a+b;
             a=b;
             b=next;
         }
+        try
+        {
+            gar(1024);
+        }
+        catch (...)
+        {
+            std::cout << "Saving exception\n";
+            exception = std::current_exception();
+        }
         return std::move(sink);
     }};
+
     for (int j=0;j<10;++j)
     {
-        source=std::move(source).resume();
+        try
+        {
+            source=std::move(source).resume();
+        }
+        catch (DB::Exception & e)
+        {
+            std::cout << "Caught exception in resume " << e.getStackTraceString() << std::endl;
+        }
         std::cout << a << " ";
     }
+
+    std::cout << std::endl;
+
+    try
+    {
+        foo(exception);
+    }
+    catch (const DB::Exception & e)
+    {
+        std::cout << e.getStackTraceString() << std::endl;
+    }
+}
+catch (...)
+{
+    std::cerr << "Uncaught exception\n";
 }

From e3946bc2b54069b3549f3eb9dcccdc910674604d Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 2 Dec 2020 20:02:14 +0300
Subject: [PATCH 039/504] Add async read to RemoteQueryExecutor.

---
 src/Client/Connection.h                 |   5 +-
 src/Client/MultiplexedConnections.h     |   2 +
 src/DataStreams/RemoteQueryExecutor.cpp | 184 ++++++++++++++++--------
 src/DataStreams/RemoteQueryExecutor.h   |  24 ++++
 src/IO/ReadBufferFromPocoSocket.cpp     |  16 ++-
 src/IO/ReadBufferFromPocoSocket.h       |  13 ++
 6 files changed, 186 insertions(+), 58 deletions(-)

diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index f4c25001f3e..bc4decb67be 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -16,6 +16,7 @@
 #include <DataStreams/BlockStreamProfileInfo.h>
 
 #include <IO/ConnectionTimeouts.h>
+#include <IO/ReadBufferFromPocoSocket.h>
 
 #include <Core/Settings.h>
 #include <Interpreters/TablesStatus.h>
@@ -188,6 +189,8 @@ public:
     size_t outBytesCount() const { return out ? out->count() : 0; }
     size_t inBytesCount() const { return in ? in->count() : 0; }
 
+    void setFiber(ReadBufferFromPocoSocket::Fiber * fiber) { in->setFiber(fiber); }
+
 private:
     String host;
     UInt16 port;
@@ -224,7 +227,7 @@ private:
     String server_display_name;
 
     std::unique_ptr<Poco::Net::StreamSocket> socket;
-    std::shared_ptr<ReadBuffer> in;
+    std::shared_ptr<ReadBufferFromPocoSocket> in;
     std::shared_ptr<WriteBuffer> out;
     std::optional<UInt64> last_input_packet_type;
 
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index eaec7f744bc..1ec424593b8 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -67,6 +67,8 @@ public:
     /// Without locking, because sendCancel() does not change the state of the replicas.
     bool hasActiveConnections() const { return active_connection_count > 0; }
 
+    void setFiber(ReadBufferFromPocoSocket::Fiber * fiber) { current_connection->setFiber(fiber); }
+
 private:
     /// Internal version of `receivePacket` function without locking.
     Packet receivePacketUnlocked();
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index a7fe9d99688..e4cbf21066d 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -9,6 +9,7 @@
 #include <Interpreters/castColumn.h>
 #include <Interpreters/Cluster.h>
 #include <Interpreters/InternalTextLogsQueue.h>
+#include <Common/FiberStack.h>
 
 namespace DB
 {
@@ -199,65 +200,136 @@ Block RemoteQueryExecutor::read()
 
         Packet packet = multiplexed_connections->receivePacket();
 
-        switch (packet.type)
+        if (auto block = processPacket(std::move(packet)))
+            return *block;
+    }
+}
+
+void RemoteQueryExecutor::read(ReadContext & read_context)
+{
+    if (!sent_query)
+    {
+        sendQuery();
+
+        if (context.getSettingsRef().skip_unavailable_shards && (0 == multiplexed_connections->size()))
         {
-            case Protocol::Server::Data:
-                /// If the block is not empty and is not a header block
-                if (packet.block && (packet.block.rows() > 0))
-                    return adaptBlockStructure(packet.block, header);
-                break;  /// If the block is empty - we will receive other packets before EndOfStream.
-
-            case Protocol::Server::Exception:
-                got_exception_from_replica = true;
-                packet.exception->rethrow();
-                break;
-
-            case Protocol::Server::EndOfStream:
-                if (!multiplexed_connections->hasActiveConnections())
-                {
-                    finished = true;
-                    return Block();
-                }
-                break;
-
-            case Protocol::Server::Progress:
-                /** We use the progress from a remote server.
-                  * We also include in ProcessList,
-                  * and we use it to check
-                  * constraints (for example, the minimum speed of query execution)
-                  * and quotas (for example, the number of lines to read).
-                  */
-                if (progress_callback)
-                    progress_callback(packet.progress);
-                break;
-
-            case Protocol::Server::ProfileInfo:
-                /// Use own (client-side) info about read bytes, it is more correct info than server-side one.
-                if (profile_info_callback)
-                    profile_info_callback(packet.profile_info);
-                break;
-
-            case Protocol::Server::Totals:
-                totals = packet.block;
-                break;
-
-            case Protocol::Server::Extremes:
-                extremes = packet.block;
-                break;
-
-            case Protocol::Server::Log:
-                /// Pass logs from remote server to client
-                if (auto log_queue = CurrentThread::getInternalTextLogsQueue())
-                    log_queue->pushBlock(std::move(packet.block));
-                break;
-
-            default:
-                got_unknown_packet_from_replica = true;
-                throw Exception(ErrorCodes::UNKNOWN_PACKET_FROM_SERVER, "Unknown packet {} from one of the following replicas: {}",
-                    toString(packet.type),
-                    multiplexed_connections->dumpAddresses());
+            read_context.is_read_in_progress = false;
+            read_context.result.clear();
+            return;
         }
     }
+
+    do
+    {
+        if (!read_context.is_read_in_progress)
+        {
+            auto routine = [&read_context, this](boost::context::fiber && sink)
+            {
+                read_context.fiber_context.fiber = std::move(sink);
+
+                try
+                {
+                    multiplexed_connections->setFiber(&read_context.fiber_context);
+                    read_context.packet = multiplexed_connections->receivePacket();
+                    multiplexed_connections->setFiber(nullptr);
+                }
+                catch (...)
+                {
+                    read_context.exception = std::current_exception();
+                }
+
+                return std::move(read_context.fiber_context.fiber);
+            };
+
+            read_context.fiber = boost::context::fiber(std::allocator_arg_t(), read_context.stack, std::move(routine));
+        }
+
+        read_context.fiber = std::move(read_context.fiber).resume();
+
+        if (read_context.exception)
+            std::rethrow_exception(std::move(read_context.exception));
+
+        if (read_context.fiber)
+        {
+            read_context.is_read_in_progress = true;
+            read_context.fd = read_context.fiber_context.fd;
+            return;
+        }
+        else
+        {
+            read_context.is_read_in_progress = false;
+            if (auto data = processPacket(std::move(read_context.packet)))
+            {
+                read_context.result = std::move(*data);
+                return;
+            }
+        }
+    }
+    while (true);
+}
+
+std::optional<Block> RemoteQueryExecutor::processPacket(Packet packet)
+{
+    switch (packet.type)
+    {
+        case Protocol::Server::Data:
+            /// If the block is not empty and is not a header block
+            if (packet.block && (packet.block.rows() > 0))
+                return adaptBlockStructure(packet.block, header);
+            break;  /// If the block is empty - we will receive other packets before EndOfStream.
+
+        case Protocol::Server::Exception:
+            got_exception_from_replica = true;
+            packet.exception->rethrow();
+            break;
+
+        case Protocol::Server::EndOfStream:
+            if (!multiplexed_connections->hasActiveConnections())
+            {
+                finished = true;
+                return Block();
+            }
+            break;
+
+        case Protocol::Server::Progress:
+            /** We use the progress from a remote server.
+              * We also include in ProcessList,
+              * and we use it to check
+              * constraints (for example, the minimum speed of query execution)
+              * and quotas (for example, the number of lines to read).
+              */
+            if (progress_callback)
+                progress_callback(packet.progress);
+            break;
+
+        case Protocol::Server::ProfileInfo:
+            /// Use own (client-side) info about read bytes, it is more correct info than server-side one.
+            if (profile_info_callback)
+                profile_info_callback(packet.profile_info);
+            break;
+
+        case Protocol::Server::Totals:
+            totals = packet.block;
+            break;
+
+        case Protocol::Server::Extremes:
+            extremes = packet.block;
+            break;
+
+        case Protocol::Server::Log:
+            /// Pass logs from remote server to client
+            if (auto log_queue = CurrentThread::getInternalTextLogsQueue())
+                log_queue->pushBlock(std::move(packet.block));
+            break;
+
+        default:
+            got_unknown_packet_from_replica = true;
+            throw Exception(ErrorCodes::UNKNOWN_PACKET_FROM_SERVER, "Unknown packet {} from one of the following replicas: {}",
+                toString(packet.type),
+                multiplexed_connections->dumpAddresses());
+    }
+
+    return {};
 }
 
 void RemoteQueryExecutor::finish()
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 0db0e0218be..633d442ffc9 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -3,6 +3,7 @@
 #include <Interpreters/Context.h>
 #include <Client/ConnectionPool.h>
 #include <Client/MultiplexedConnections.h>
+#include <Common/FiberStack.h>
 
 namespace DB
 {
@@ -46,11 +47,31 @@ public:
 
     ~RemoteQueryExecutor();
 
+    struct ReadContext
+    {
+        bool is_read_in_progress = false;
+
+        /// If is_read_in_progress, use this fd to poll
+        int fd;
+
+        /// If not is_read_in_progress, result block is set.
+        Block result;
+
+        /// Internal data
+
+        boost::context::fiber fiber;
+        Packet packet;
+        std::exception_ptr exception;
+        FiberStack<> stack;
+        ReadBufferFromPocoSocket::Fiber fiber_context;
+    };
+
     /// Create connection and send query, external tables and scalars.
     void sendQuery();
 
     /// Read next block of data. Returns empty block if query is finished.
     Block read();
+    void read(ReadContext & read_context);
 
     /// Receive all remain packets and finish query.
     /// It should be cancelled after read returned empty block.
@@ -159,6 +180,9 @@ private:
 
     /// Returns true if exception was thrown
     bool hasThrownException() const;
+
+    /// Process packet for read and return data block if possible.
+    std::optional<Block> processPacket(Packet packet);
 };
 
 }
diff --git a/src/IO/ReadBufferFromPocoSocket.cpp b/src/IO/ReadBufferFromPocoSocket.cpp
index 5c66c3209f6..85249002c51 100644
--- a/src/IO/ReadBufferFromPocoSocket.cpp
+++ b/src/IO/ReadBufferFromPocoSocket.cpp
@@ -28,10 +28,24 @@ bool ReadBufferFromPocoSocket::nextImpl()
     ssize_t bytes_read = 0;
     Stopwatch watch;
 
+    int flags = 0;
+    if (fiber)
+        flags |= MSG_DONTWAIT;
+
     /// Add more details to exceptions.
     try
     {
-        bytes_read = socket.impl()->receiveBytes(internal_buffer.begin(), internal_buffer.size());
+        bytes_read = socket.impl()->receiveBytes(internal_buffer.begin(), internal_buffer.size(), flags);
+
+        /// If fiber is specified, and read is blocking, run fiber and try again later.
+        /// It is expected that file descriptor may be polled externally.
+        /// Note that receive timeout is not checked here. External code should check it while polling.
+        while (bytes_read < 0 && fiber && (errno == POCO_EAGAIN || errno == POCO_EWOULDBLOCK))
+        {
+            fiber->fd = socket.impl()->sockfd();
+            fiber->fiber = std::move(fiber->fiber).resume();
+            bytes_read = socket.impl()->receiveBytes(internal_buffer.begin(), internal_buffer.size(), flags);
+        }
     }
     catch (const Poco::Net::NetException & e)
     {
diff --git a/src/IO/ReadBufferFromPocoSocket.h b/src/IO/ReadBufferFromPocoSocket.h
index f328b89d99c..790a350a4de 100644
--- a/src/IO/ReadBufferFromPocoSocket.h
+++ b/src/IO/ReadBufferFromPocoSocket.h
@@ -5,6 +5,8 @@
 #include <IO/ReadBuffer.h>
 #include <IO/BufferWithOwnMemory.h>
 
+#include <boost/context/fiber.hpp>
+
 
 namespace DB
 {
@@ -28,6 +30,17 @@ public:
     ReadBufferFromPocoSocket(Poco::Net::Socket & socket_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE);
 
     bool poll(size_t timeout_microseconds);
+
+    struct Fiber
+    {
+        boost::context::fiber fiber;
+        int fd;
+    };
+
+    void setFiber(Fiber * fiber_) { fiber = fiber_; }
+
+private:
+    Fiber * fiber;
 };
 
 }

From 8ee86e35d24e969c6b3f34b8ac7fa290c38715de Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Wed, 2 Dec 2020 21:16:31 +0300
Subject: [PATCH 040/504] debug

---
 .../AggregateFunctionSum.cpp                  |   1 +
 .../registerFunctionsMiscellaneous.cpp        |   2 +
 src/Functions/runningAccumulate.cpp           | 132 ++++++++++++++++++
 src/Interpreters/ExpressionAnalyzer.cpp       |  68 +++++++++
 src/Interpreters/ExpressionAnalyzer.h         |   2 +
 .../Transforms/ExpressionTransform.h          |   6 +-
 src/Storages/IStorage.cpp                     |  39 ++++++
 src/Storages/SelectQueryInfo.h                |   4 +
 8 files changed, 253 insertions(+), 1 deletion(-)

diff --git a/src/AggregateFunctions/AggregateFunctionSum.cpp b/src/AggregateFunctions/AggregateFunctionSum.cpp
index 6afae98ef2d..e937769f3a3 100644
--- a/src/AggregateFunctions/AggregateFunctionSum.cpp
+++ b/src/AggregateFunctions/AggregateFunctionSum.cpp
@@ -52,6 +52,7 @@ template <typename T> using AggregateFunctionSumKahan =
 template <template <typename> class Function>
 AggregateFunctionPtr createAggregateFunctionSum(const std::string & name, const DataTypes & argument_types, const Array & parameters)
 {
+    fmt::print(stderr, "create sum at \n{}\n", StackTrace().toString());
     assertNoParameters(name, parameters);
     assertUnary(name, argument_types);
 
diff --git a/src/Functions/registerFunctionsMiscellaneous.cpp b/src/Functions/registerFunctionsMiscellaneous.cpp
index 2c0baa7bba4..199567680e7 100644
--- a/src/Functions/registerFunctionsMiscellaneous.cpp
+++ b/src/Functions/registerFunctionsMiscellaneous.cpp
@@ -43,6 +43,7 @@ void registerFunctionBuildId(FunctionFactory &);
 void registerFunctionUptime(FunctionFactory &);
 void registerFunctionTimeZone(FunctionFactory &);
 void registerFunctionRunningAccumulate(FunctionFactory &);
+void registerFunctionWindow(FunctionFactory &);
 void registerFunctionRunningDifference(FunctionFactory &);
 void registerFunctionRunningDifferenceStartingWithFirstValue(FunctionFactory &);
 void registerFunctionFinalizeAggregation(FunctionFactory &);
@@ -109,6 +110,7 @@ void registerFunctionsMiscellaneous(FunctionFactory & factory)
     registerFunctionUptime(factory);
     registerFunctionTimeZone(factory);
     registerFunctionRunningAccumulate(factory);
+    registerFunctionWindow(factory);
     registerFunctionRunningDifference(factory);
     registerFunctionRunningDifferenceStartingWithFirstValue(factory);
     registerFunctionFinalizeAggregation(factory);
diff --git a/src/Functions/runningAccumulate.cpp b/src/Functions/runningAccumulate.cpp
index 1a01b88e22c..9125597fff9 100644
--- a/src/Functions/runningAccumulate.cpp
+++ b/src/Functions/runningAccumulate.cpp
@@ -75,6 +75,8 @@ public:
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
     {
+        fmt::print(stderr, "executeImpl at \n{}\n", StackTrace().toString());
+
         const ColumnAggregateFunction * column_with_states
             = typeid_cast<const ColumnAggregateFunction *>(&*arguments.at(0).column);
 
@@ -136,6 +138,136 @@ public:
 
 }
 
+namespace
+{
+
+class FunctionWindow : public IFunction
+{
+    mutable ColumnsWithTypeAndName prev_columns = {};
+    mutable int prev_index = -1;
+
+    mutable AlignedBuffer place;
+    mutable std::unique_ptr<Arena> arena;
+
+public:
+    static constexpr auto name = "window";
+    static FunctionPtr create(const Context &)
+    {
+        return std::make_shared<FunctionWindow>();
+    }
+
+    String getName() const override
+    {
+        return name;
+    }
+
+    bool isStateful() const override
+    {
+        return true;
+    }
+
+    bool isVariadic() const override { return true; }
+
+    size_t getNumberOfArguments() const override { return 0; }
+
+    bool isDeterministic() const override { return false; }
+
+    bool isDeterministicInScopeOfQuery() const override
+    {
+        return false;
+    }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        fmt::print(stderr, "getReturnType for {} at \n{}\n",
+            static_cast<const void *>(this), StackTrace().toString());
+
+        if (arguments.empty() || arguments.size() > 3)
+            throw Exception("Incorrect number of arguments of function " + getName() + ". Must be up to 3 (function, PARTITION BY, ORDER BY).",
+                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+        const DataTypeAggregateFunction * type = checkAndGetDataType<DataTypeAggregateFunction>(arguments[0].get());
+        if (!type)
+            throw Exception("Argument for function " + getName() + " must have type AggregateFunction - state of aggregate function.",
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+        const auto & agg_func = *type->getFunction();
+        place.reset(agg_func.sizeOfData(), agg_func.alignOfData());
+
+        /// Will pass empty arena if agg_func does not allocate memory in arena
+        arena = agg_func.allocatesMemoryInArena() ? std::make_unique<Arena>() : nullptr;
+
+
+        return type->getReturnType();
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
+    {
+        fmt::print(stderr, "executeImpl for {} at \n{}\n",
+            static_cast<const void *>(this), StackTrace().toString());
+
+        const ColumnAggregateFunction * column_with_states
+            = typeid_cast<const ColumnAggregateFunction *>(&*arguments.at(0).column);
+
+        if (!column_with_states)
+            throw Exception("Illegal column " + arguments.at(0).column->getName()
+                    + " of first argument of function "
+                    + getName(),
+                ErrorCodes::ILLEGAL_COLUMN);
+
+        ColumnPtr column_with_groups;
+
+        if (arguments.size() >= 2)
+            column_with_groups = arguments[1].column;
+
+        AggregateFunctionPtr aggregate_function_ptr = column_with_states->getAggregateFunction();
+        const IAggregateFunction & agg_func = *aggregate_function_ptr;
+
+
+        auto result_column_ptr = agg_func.getReturnType()->createColumn();
+        IColumn & result_column = *result_column_ptr;
+        result_column.reserve(column_with_states->size());
+
+        const auto & states = column_with_states->getData();
+
+        bool state_created = false;
+        SCOPE_EXIT({
+            if (state_created)
+                agg_func.destroy(place.data());
+        });
+
+        size_t row_number = 0;
+        for (const auto & state_to_add : states)
+        {
+            if (row_number == 0 || (column_with_groups && column_with_groups->compareAt(row_number, row_number - 1, *column_with_groups, 1) != 0))
+            {
+                if (state_created)
+                {
+                    agg_func.destroy(place.data());
+                    state_created = false;
+                }
+
+                agg_func.create(place.data());
+                state_created = true;
+            }
+
+            agg_func.merge(place.data(), state_to_add, arena.get());
+            agg_func.insertResultInto(place.data(), result_column, arena.get());
+
+            ++row_number;
+        }
+
+        return result_column_ptr;
+    }
+};
+
+}
+
+void registerFunctionWindow(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionWindow>();
+}
+
 void registerFunctionRunningAccumulate(FunctionFactory & factory)
 {
     factory.registerFunction<FunctionRunningAccumulate>();
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 7c9bb79dd24..4ac337c81e2 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1241,6 +1241,8 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
     removeExtraColumns();
 
     checkActions();
+
+    fmt::print(stderr, "ExpressionAnalysisResult: \n{}\n", dump());
 }
 
 void ExpressionAnalysisResult::finalize(const ExpressionActionsChain & chain, size_t where_step_num)
@@ -1314,4 +1316,70 @@ void ExpressionAnalysisResult::checkActions() const
     }
 }
 
+std::string ExpressionAnalysisResult::dump() const
+{
+    std::stringstream ss;
+
+    ss << "ExpressionAnalysisResult\n";
+    ss << "need_aggregate " << need_aggregate << "\n";
+    ss << "has_order_by " << has_order_by << "\n";
+
+    if (before_array_join)
+    {
+        ss << "before_array_join " << before_array_join->dumpDAG() << "\n";
+    }
+
+    if (array_join)
+    {
+        ss << "array_join " << "FIXME doesn't have dump" << "\n";
+    }
+
+    if (before_join)
+    {
+        ss << "before_join " << before_join->dumpDAG() << "\n";
+    }
+
+    if (before_where)
+    {
+        ss << "before_where " << before_where->dumpDAG() << "\n";
+    }
+
+    if (prewhere_info)
+    {
+        ss << "prewhere_info " << prewhere_info->dump() << "\n";
+    }
+
+    if (filter_info)
+    {
+        ss << "filter_info " << filter_info->dump() << "\n";
+    }
+
+    if (before_aggregation)
+    {
+        ss << "before_aggregation " << before_aggregation->dumpDAG() << "\n";
+    }
+
+    if (before_having)
+    {
+        ss << "before_having " << before_having->dumpDAG() << "\n";
+    }
+
+    if (before_order_and_select)
+    {
+        ss << "before_order_and_select " << before_order_and_select->dumpDAG() << "\n";
+    }
+
+    if (before_limit_by)
+    {
+        ss << "before_limit_by " << before_limit_by->dumpDAG() << "\n";
+    }
+
+    if (final_projection)
+    {
+        ss << "final_projection " << final_projection->dumpDAG() << "\n";
+    }
+
+    return ss.str();
+}
+
 }
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index 14fd601732e..0ea7b317ebe 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -172,6 +172,8 @@ class SelectQueryExpressionAnalyzer;
 /// Result of SelectQueryExpressionAnalyzer: expressions for InterpreterSelectQuery
 struct ExpressionAnalysisResult
 {
+    std::string dump() const;
+
     /// Do I need to perform the first part of the pipeline - running on remote servers during distributed processing.
     bool first_stage = false;
     /// Do I need to execute the second part of the pipeline - running on the initiating server during distributed processing.
diff --git a/src/Processors/Transforms/ExpressionTransform.h b/src/Processors/Transforms/ExpressionTransform.h
index 525273bad81..eed8bf089cd 100644
--- a/src/Processors/Transforms/ExpressionTransform.h
+++ b/src/Processors/Transforms/ExpressionTransform.h
@@ -19,7 +19,11 @@ public:
             const Block & header_,
             ExpressionActionsPtr expression_);
 
-    String getName() const override { return "ExpressionTransform"; }
+    String getName() const override
+    {
+        fmt::print(stderr, "et getname at \n{}\n", StackTrace().toString());
+        return "ExpressionTransform";
+    }
 
     static Block transformHeader(Block header, const ExpressionActionsPtr & expression);
 
diff --git a/src/Storages/IStorage.cpp b/src/Storages/IStorage.cpp
index 94f37254b91..76c975c4654 100644
--- a/src/Storages/IStorage.cpp
+++ b/src/Storages/IStorage.cpp
@@ -168,4 +168,43 @@ NamesAndTypesList IStorage::getVirtuals() const
     return {};
 }
 
+std::string PrewhereDAGInfo::dump() const
+{
+    std::stringstream ss;
+    ss << "PrewhereDagInfo\n";
+
+    if (alias_actions)
+    {
+        ss << "alias_actions " << alias_actions->dumpDAG() << "\n";
+    }
+
+    if (prewhere_actions)
+    {
+        ss << "prewhere_actions " << prewhere_actions->dumpDAG() << "\n";
+    }
+
+    if (remove_columns_actions)
+    {
+        ss << "remove_columns_actions " << remove_columns_actions->dumpDAG() << "\n";
+    }
+
+    ss << "remove_prewhere_column " << remove_prewhere_column
+       << ", need_filter " << need_filter << "\n";
+
+    return ss.str();
+}
+
+std::string FilterInfo::dump() const
+{
+    std::stringstream ss;
+    ss << "FilterInfo for column '" << column_name <<"', do_remove_column "
+       << do_remove_column << "\n";
+    if (actions_dag)
+    {
+        ss << "actions_dag " << actions_dag->dumpDAG() << "\n";
+    }
+
+    return ss.str();
+}
+
 }
diff --git a/src/Storages/SelectQueryInfo.h b/src/Storages/SelectQueryInfo.h
index 683b2463d1f..5a3ada6288b 100644
--- a/src/Storages/SelectQueryInfo.h
+++ b/src/Storages/SelectQueryInfo.h
@@ -45,6 +45,8 @@ struct PrewhereDAGInfo
     PrewhereDAGInfo() = default;
     explicit PrewhereDAGInfo(ActionsDAGPtr prewhere_actions_, String prewhere_column_name_)
             : prewhere_actions(std::move(prewhere_actions_)), prewhere_column_name(std::move(prewhere_column_name_)) {}
+
+    std::string dump() const;
 };
 
 /// Helper struct to store all the information about the filter expression.
@@ -53,6 +55,8 @@ struct FilterInfo
     ActionsDAGPtr actions_dag;
     String column_name;
     bool do_remove_column = false;
+
+    std::string dump() const;
 };
 
 struct InputOrderInfo

From 082a4963641dea9c5a2508f57950af58ad1079ef Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 3 Dec 2020 15:21:10 +0300
Subject: [PATCH 041/504] Add async read to RemoteQueryExecutor

---
 src/Client/Connection.h                 |   3 +-
 src/Client/MultiplexedConnections.h     |   3 +-
 src/Common/TimerDescriptor.cpp          |  82 +++++++++
 src/Common/TimerDescriptor.h            |  28 +++
 src/DataStreams/RemoteQueryExecutor.cpp | 219 +++++++++++++++++++-----
 src/DataStreams/RemoteQueryExecutor.h   |  27 +--
 src/IO/ReadBufferFromPocoSocket.cpp     |   5 +-
 src/IO/ReadBufferFromPocoSocket.h       |  11 +-
 8 files changed, 302 insertions(+), 76 deletions(-)
 create mode 100644 src/Common/TimerDescriptor.cpp
 create mode 100644 src/Common/TimerDescriptor.h

diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index bc4decb67be..c891f77d1a7 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -189,7 +189,8 @@ public:
     size_t outBytesCount() const { return out ? out->count() : 0; }
     size_t inBytesCount() const { return in ? in->count() : 0; }
 
-    void setFiber(ReadBufferFromPocoSocket::Fiber * fiber) { in->setFiber(fiber); }
+    void setFiber(boost::context::fiber * fiber) { in->setFiber(fiber); }
+    Poco::Net::Socket & getSocket() { return in->getSocket(); }
 
 private:
     String host;
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index 1ec424593b8..9573dfe4c93 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -67,7 +67,8 @@ public:
     /// Without locking, because sendCancel() does not change the state of the replicas.
     bool hasActiveConnections() const { return active_connection_count > 0; }
 
-    void setFiber(ReadBufferFromPocoSocket::Fiber * fiber) { current_connection->setFiber(fiber); }
+    void setFiber(boost::context::fiber * fiber) { current_connection->setFiber(fiber); }
+    Poco::Net::Socket & getSocket() { return current_connection->getSocket(); }
 
 private:
     /// Internal version of `receivePacket` function without locking.
diff --git a/src/Common/TimerDescriptor.cpp b/src/Common/TimerDescriptor.cpp
new file mode 100644
index 00000000000..406ce53ca49
--- /dev/null
+++ b/src/Common/TimerDescriptor.cpp
@@ -0,0 +1,82 @@
+#include <Common/TimerDescriptor.h>
+#include <Common/Exception.h>
+
+#include <sys/timerfd.h>
+#include <sys/ioctl.h>
+#include <unistd.h>
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int CANNOT_CREATE_TIMER;
+    extern const int CANNOT_SET_TIMER_PERIOD;
+    extern const int CANNOT_FCNTL;
+    extern const int CANNOT_READ_FROM_SOCKET;
+}
+
+TimerDescriptor::TimerDescriptor(int clockid, int flags)
+{
+    timer_fd = timerfd_create(clockid, flags);
+    if (timer_fd == -1)
+        throw Exception(ErrorCodes::CANNOT_CREATE_TIMER, "Cannot create timer_fd descriptor");
+}
+
+TimerDescriptor::~TimerDescriptor()
+{
+    /// Do not check for result cause cannot throw exception.
+    close(timer_fd);
+}
+
+void TimerDescriptor::reset() const
+{
+    itimerspec spec;
+    spec.it_interval.tv_nsec = 0;
+    spec.it_interval.tv_sec = 0;
+    spec.it_value.tv_sec = 0;
+    spec.it_value.tv_nsec = 0;
+
+    if (-1 == timerfd_settime(timer_fd, 0 /*relative timer */, &spec, nullptr))
+        throwFromErrno("Cannot reset timer_fd", ErrorCodes::CANNOT_SET_TIMER_PERIOD);
+
+    /// Drain socket.
+    /// It may be possible that alarm happened and socket is readable.
+    drain();
+}
+
+void TimerDescriptor::drain() const
+{
+    int num_bytes_to_read = 0;
+    if (-1 == ioctl(timer_fd, FIONREAD, &num_bytes_to_read))
+        throwFromErrno("Cannot get available bytes form timer_fd", ErrorCodes::CANNOT_FCNTL);
+
+    if (num_bytes_to_read)
+    {
+        /// It is expected that socket returns 8 bytes when readable.
+        /// Read in loop anyway cause signal may interrupt read call.
+        uint64_t buf;
+        while (num_bytes_to_read)
+        {
+            ssize_t res = ::read(timer_fd, &buf, sizeof(buf));
+            if (res < 0 && errno != EINTR)
+                throwFromErrno("Cannot drain timer_fd", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+            else
+                num_bytes_to_read -= res;
+        }
+    }
+}
+
+void TimerDescriptor::setRelative(const Poco::Timespan & timespan) const
+{
+    itimerspec spec;
+    spec.it_interval.tv_nsec = 0;
+    spec.it_interval.tv_sec = 0;
+    spec.it_value.tv_sec = timespan.totalSeconds();
+    spec.it_value.tv_nsec = timespan.useconds();
+
+    if (-1 == timerfd_settime(timer_fd, 0 /*relative timer */, &spec, nullptr))
+        throwFromErrno("Cannot set time for timer_fd", ErrorCodes::CANNOT_SET_TIMER_PERIOD);
+}
+
+}
diff --git a/src/Common/TimerDescriptor.h b/src/Common/TimerDescriptor.h
new file mode 100644
index 00000000000..04c592ebae6
--- /dev/null
+++ b/src/Common/TimerDescriptor.h
@@ -0,0 +1,28 @@
+#pragma once
+#include <Poco/Timespan.h>
+
+namespace DB
+{
+
+class TimerDescriptor
+{
+private:
+    int timer_fd;
+
+public:
+    explicit TimerDescriptor(int clockid, int flags);
+    ~TimerDescriptor();
+
+    TimerDescriptor(const TimerDescriptor &) = delete;
+    TimerDescriptor & operator=(const TimerDescriptor &) = delete;
+    TimerDescriptor(TimerDescriptor &&) = default;
+    TimerDescriptor & operator=(TimerDescriptor &&) = default;
+
+    int getDescriptor() const { return timer_fd; }
+
+    void reset() const;
+    void drain() const;
+    void setRelative(const Poco::Timespan & timespan) const;
+};
+
+}
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index e4cbf21066d..a902541b875 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -11,14 +11,174 @@
 #include <Interpreters/InternalTextLogsQueue.h>
 #include <Common/FiberStack.h>
 
+#include <sys/epoll.h>
+
 namespace DB
 {
 
 namespace ErrorCodes
 {
     extern const int UNKNOWN_PACKET_FROM_SERVER;
+    extern const int CANNOT_FCNTL;
+    extern const int CANNOT_READ_FROM_SOCKET;
+    extern const int CANNOT_OPEN_FILE;
+    extern const int SOCKET_TIMEOUT;
 }
 
+struct RemoteQueryExecutor::ReadContext
+{
+    bool is_read_in_progress = false;
+    Packet packet;
+
+    std::exception_ptr exception;
+    FiberStack<> stack;
+    boost::context::fiber fiber;
+
+    Poco::Timespan receive_timeout;
+    MultiplexedConnections & connections;
+
+    TimerDescriptor timer{CLOCK_MONOTONIC, 0};
+    int socket_fd;
+    int epoll_df;
+
+    explicit ReadContext(MultiplexedConnections & connections_) : connections(connections_)
+    {
+        auto & socket = connections.getSocket();
+        socket_fd = socket.impl()->sockfd();
+        receive_timeout = socket.impl()->getReceiveTimeout();
+
+        if (-1 == epoll_create(2))
+            throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
+
+        {
+            epoll_event socket_event;
+            socket_event.events = EPOLLIN | EPOLLPRI;
+            socket_event.data.fd = socket_fd;
+
+            if (-1 == epoll_ctl(epoll_df, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
+                throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+        }
+
+        {
+            epoll_event timer_event;
+            timer_event.events = EPOLLIN | EPOLLPRI;
+            timer_event.data.fd = timer.getDescriptor();
+
+            if (-1 == epoll_ctl(epoll_df, EPOLL_CTL_ADD, timer_event.data.fd, &timer_event))
+                throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+        }
+
+        fiber = boost::context::fiber(std::allocator_arg_t(), stack, Routine{connections, *this});
+    }
+
+    static void initialize(std::unique_ptr<ReadContext> & read_context, MultiplexedConnections & connections)
+    {
+        try
+        {
+            read_context = std::make_unique<ReadContext>(connections);
+        }
+        catch (DB::Exception & e)
+        {
+            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+        }
+    }
+
+    void checkTimeout() const
+    {
+        try
+        {
+            checkTimeoutImpl();
+        }
+        catch (DB::Exception & e)
+        {
+            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+        }
+    }
+
+    void checkTimeoutImpl() const
+    {
+        epoll_event events[2];
+
+        /// Wait for epoll_fd will not block if it was polled externally.
+        int num_events = epoll_wait(epoll_df, events, 2, 0);
+        if (num_events == -1)
+            throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+
+        bool is_socket_ready = false;
+        bool has_timer_alarm = false;
+
+        for (int i = 0; i < num_events; ++i)
+        {
+            if (events[i].data.fd == socket_fd)
+                is_socket_ready = true;
+            if (events[i].data.fd == timer.getDescriptor())
+                has_timer_alarm = true;
+        }
+
+        if (has_timer_alarm && !is_socket_ready)
+        {
+            /// Socket receive timeout. Drain it in case or error, or it may be hide by timeout exception.
+            timer.drain();
+            throw NetException("Timeout exceeded", ErrorCodes::SOCKET_TIMEOUT);
+        }
+    }
+
+    void setTimer() const
+    {
+        /// Did not get packet yet. Init timeout for the next async reading.
+        timer.reset();
+
+        if (receive_timeout.totalMicroseconds())
+            timer.setRelative(receive_timeout);
+    }
+
+    void resumeRoutine()
+    {
+        fiber = std::move(fiber).resume();
+
+        if (exception)
+            std::rethrow_exception(std::move(exception));
+    }
+
+    ~ReadContext()
+    {
+        /// socket_fd is closed by Poco::Net::Socket
+        /// timer_fd is closed by TimerDescriptor
+        close(epoll_df);
+    }
+
+    struct Routine
+    {
+        MultiplexedConnections & connections;
+        ReadContext & read_context;
+
+        boost::context::fiber operator()(boost::context::fiber && sink) const
+        {
+            try
+            {
+                while (true)
+                {
+                    read_context.is_read_in_progress = true;
+                    connections.setFiber(&sink);
+
+                    read_context.packet = connections.receivePacket();
+
+                    read_context.is_read_in_progress = false;
+                    connections.setFiber(nullptr);
+
+                    sink = std::move(sink).resume();
+                }
+            }
+            catch (...)
+            {
+                read_context.exception = std::current_exception();
+            }
+
+            return std::move(sink);
+        };
+    };
+};
+
 RemoteQueryExecutor::RemoteQueryExecutor(
     Connection & connection,
     const String & query_, const Block & header_, const Context & context_, const Settings * settings,
@@ -205,64 +365,35 @@ Block RemoteQueryExecutor::read()
     }
 }
 
-void RemoteQueryExecutor::read(ReadContext & read_context)
+std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext> & read_context)
 {
     if (!sent_query)
     {
         sendQuery();
 
         if (context.getSettingsRef().skip_unavailable_shards && (0 == multiplexed_connections->size()))
-        {
-            read_context.is_read_in_progress = false;
-            read_context.result.clear();
-            return;
-        }
+            return Block();
     }
 
+    if (!read_context)
+        ReadContext::initialize(read_context, *multiplexed_connections);
+
+    if (read_context->is_read_in_progress)
+        read_context->checkTimeout();
+
     do
     {
-        if (!read_context.is_read_in_progress)
+        read_context->resumeRoutine();
+
+        if (read_context->is_read_in_progress)
         {
-            auto routine = [&read_context, this](boost::context::fiber && sink)
-            {
-                read_context.fiber_context.fiber = std::move(sink);
-
-                try
-                {
-                    multiplexed_connections->setFiber(&read_context.fiber_context);
-                    read_context.packet = multiplexed_connections->receivePacket();
-                    multiplexed_connections->setFiber(nullptr);
-                }
-                catch (...)
-                {
-                    read_context.exception = std::current_exception();
-                }
-
-                return std::move(read_context.fiber_context.fiber);
-            };
-
-            read_context.fiber = boost::context::fiber(std::allocator_arg_t(), read_context.stack, std::move(routine));
-        }
-
-        read_context.fiber = std::move(read_context.fiber).resume();
-
-        if (read_context.exception)
-            std::rethrow_exception(std::move(read_context.exception));
-
-        if (read_context.fiber)
-        {
-            read_context.is_read_in_progress = true;
-            read_context.fd = read_context.fiber_context.fd;
-            return;
+            read_context->setTimer();
+            return read_context->epoll_df;
         }
         else
         {
-            read_context.is_read_in_progress = false;
-            if (auto data = processPacket(std::move(read_context.packet)))
-            {
-                read_context.result = std::move(*data);
-                return;
-            }
+            if (auto data = processPacket(std::move(read_context->packet)))
+                return std::move(*data);
         }
     }
     while (true);
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 633d442ffc9..6de5847015a 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -4,6 +4,8 @@
 #include <Client/ConnectionPool.h>
 #include <Client/MultiplexedConnections.h>
 #include <Common/FiberStack.h>
+#include <Common/TimerDescriptor.h>
+#include <variant>
 
 namespace DB
 {
@@ -47,31 +49,16 @@ public:
 
     ~RemoteQueryExecutor();
 
-    struct ReadContext
-    {
-        bool is_read_in_progress = false;
-
-        /// If is_read_in_progress, use this fd to poll
-        int fd;
-
-        /// If not is_read_in_progress, result block is set.
-        Block result;
-
-        /// Internal data
-
-        boost::context::fiber fiber;
-        Packet packet;
-        std::exception_ptr exception;
-        FiberStack<> stack;
-        ReadBufferFromPocoSocket::Fiber fiber_context;
-    };
-
     /// Create connection and send query, external tables and scalars.
     void sendQuery();
 
     /// Read next block of data. Returns empty block if query is finished.
     Block read();
-    void read(ReadContext & read_context);
+
+    /// Async variant of read. Returns ready block or file descriptor which may be used for polling.
+    /// ReadContext is an internal read state. Pass empty ptr first time, reuse created one for every call.
+    struct ReadContext;
+    std::variant<Block, int> read(std::unique_ptr<ReadContext> & read_context);
 
     /// Receive all remain packets and finish query.
     /// It should be cancelled after read returned empty block.
diff --git a/src/IO/ReadBufferFromPocoSocket.cpp b/src/IO/ReadBufferFromPocoSocket.cpp
index 85249002c51..dd6cd4817df 100644
--- a/src/IO/ReadBufferFromPocoSocket.cpp
+++ b/src/IO/ReadBufferFromPocoSocket.cpp
@@ -42,8 +42,9 @@ bool ReadBufferFromPocoSocket::nextImpl()
         /// Note that receive timeout is not checked here. External code should check it while polling.
         while (bytes_read < 0 && fiber && (errno == POCO_EAGAIN || errno == POCO_EWOULDBLOCK))
         {
-            fiber->fd = socket.impl()->sockfd();
-            fiber->fiber = std::move(fiber->fiber).resume();
+            //fiber->fd = socket.impl()->sockfd();
+            //fiber->timeout = socket.impl()->getReceiveTimeout();
+            *fiber = std::move(*fiber).resume();
             bytes_read = socket.impl()->receiveBytes(internal_buffer.begin(), internal_buffer.size(), flags);
         }
     }
diff --git a/src/IO/ReadBufferFromPocoSocket.h b/src/IO/ReadBufferFromPocoSocket.h
index 790a350a4de..0c744ae8fe3 100644
--- a/src/IO/ReadBufferFromPocoSocket.h
+++ b/src/IO/ReadBufferFromPocoSocket.h
@@ -31,16 +31,11 @@ public:
 
     bool poll(size_t timeout_microseconds);
 
-    struct Fiber
-    {
-        boost::context::fiber fiber;
-        int fd;
-    };
-
-    void setFiber(Fiber * fiber_) { fiber = fiber_; }
+    void setFiber(boost::context::fiber * fiber_) { fiber = fiber_; }
+    Poco::Net::Socket & getSocket() { return socket; }
 
 private:
-    Fiber * fiber;
+    boost::context::fiber * fiber;
 };
 
 }

From 92d60a7f38a359a9b717a6f2c5ee8c5a1613c7ea Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 3 Dec 2020 20:03:13 +0300
Subject: [PATCH 042/504] Add AsyncTaskQueue.

---
 src/DataStreams/RemoteQueryExecutor.cpp       | 15 ++--
 src/Processors/Executors/AsyncTaskQueue.cpp   | 73 +++++++++++++++++++
 src/Processors/Executors/AsyncTaskQueue.h     | 35 +++++++++
 src/Processors/Executors/PipelineExecutor.cpp |  4 -
 src/Processors/IProcessor.h                   |  9 +--
 5 files changed, 119 insertions(+), 17 deletions(-)
 create mode 100644 src/Processors/Executors/AsyncTaskQueue.cpp
 create mode 100644 src/Processors/Executors/AsyncTaskQueue.h

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index a902541b875..100f20b4123 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -39,7 +39,7 @@ struct RemoteQueryExecutor::ReadContext
 
     TimerDescriptor timer{CLOCK_MONOTONIC, 0};
     int socket_fd;
-    int epoll_df;
+    int epoll_fd;
 
     explicit ReadContext(MultiplexedConnections & connections_) : connections(connections_)
     {
@@ -47,7 +47,8 @@ struct RemoteQueryExecutor::ReadContext
         socket_fd = socket.impl()->sockfd();
         receive_timeout = socket.impl()->getReceiveTimeout();
 
-        if (-1 == epoll_create(2))
+        socket_fd = epoll_create(2);
+        if (-1 == socket_fd)
             throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
 
         {
@@ -55,7 +56,7 @@ struct RemoteQueryExecutor::ReadContext
             socket_event.events = EPOLLIN | EPOLLPRI;
             socket_event.data.fd = socket_fd;
 
-            if (-1 == epoll_ctl(epoll_df, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
+            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
                 throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
         }
 
@@ -64,7 +65,7 @@ struct RemoteQueryExecutor::ReadContext
             timer_event.events = EPOLLIN | EPOLLPRI;
             timer_event.data.fd = timer.getDescriptor();
 
-            if (-1 == epoll_ctl(epoll_df, EPOLL_CTL_ADD, timer_event.data.fd, &timer_event))
+            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, timer_event.data.fd, &timer_event))
                 throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
         }
 
@@ -100,7 +101,7 @@ struct RemoteQueryExecutor::ReadContext
         epoll_event events[2];
 
         /// Wait for epoll_fd will not block if it was polled externally.
-        int num_events = epoll_wait(epoll_df, events, 2, 0);
+        int num_events = epoll_wait(epoll_fd, events, 2, 0);
         if (num_events == -1)
             throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
 
@@ -144,7 +145,7 @@ struct RemoteQueryExecutor::ReadContext
     {
         /// socket_fd is closed by Poco::Net::Socket
         /// timer_fd is closed by TimerDescriptor
-        close(epoll_df);
+        close(epoll_fd);
     }
 
     struct Routine
@@ -388,7 +389,7 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
         if (read_context->is_read_in_progress)
         {
             read_context->setTimer();
-            return read_context->epoll_df;
+            return read_context->epoll_fd;
         }
         else
         {
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
new file mode 100644
index 00000000000..417b3ce25f1
--- /dev/null
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -0,0 +1,73 @@
+#include <Processors/Executors/AsyncTaskQueue.h>
+#include <Common/Exception.h>
+#include <sys/epoll.h>
+#include <unistd.h>
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int CANNOT_OPEN_FILE;
+    extern const int CANNOT_READ_FROM_SOCKET;
+}
+
+
+AsyncTaskQueue::AsyncTaskQueue()
+{
+    epoll_fd = epoll_create(1);
+    if (-1 == epoll_fd)
+        throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
+}
+
+AsyncTaskQueue::~AsyncTaskQueue()
+{
+    close(epoll_fd);
+}
+
+void AsyncTaskQueue::addTask(void * data, int fd)
+{
+    epoll_event socket_event;
+    socket_event.events = EPOLLIN | EPOLLPRI;
+    socket_event.data.fd = fd;
+    socket_event.data.ptr = data;
+
+    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
+        throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+
+    ++num_tasks;
+    if (num_tasks == 1)
+        condvar.notify_one();
+}
+
+void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> lock)
+{
+    condvar.wait(lock, [&] { return !empty() || is_finished; });
+
+    if (is_finished)
+        return nullptr;
+
+    lock.unlock();
+
+    epoll_event event;
+    int num_events = 0;
+
+    while (num_events == 0)
+    {
+        num_events = epoll_wait(epoll_fd, &event, 1, 0);
+        if (num_events == -1)
+            throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+    }
+
+    lock.lock();
+    --num_tasks;
+    return event.data.ptr;
+}
+
+void AsyncTaskQueue::finish()
+{
+    is_finished = true;
+    condvar.notify_one();
+}
+
+}
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
new file mode 100644
index 00000000000..b66fe932d46
--- /dev/null
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -0,0 +1,35 @@
+#include <cstddef>
+#include <mutex>
+#include <atomic>
+
+namespace DB
+{
+
+class AsyncTaskQueue
+{
+private:
+    int epoll_fd;
+    size_t num_tasks;
+    std::atomic_bool is_finished = false;
+    std::condition_variable condvar;
+
+public:
+    AsyncTaskQueue();
+    ~AsyncTaskQueue();
+
+    size_t size() const { return num_tasks; }
+    bool empty() const { return num_tasks == 0; }
+
+    /// Add new task to queue.
+    void addTask(void * data, int fd);
+
+    /// Wait for any descriptor. If no descriptors in queue, blocks.
+    /// Returns ptr which was inserted into queue or nullptr if finished was called.
+    /// Lock is used to wait on condvar.
+    void * wait(std::unique_lock<std::mutex> lock);
+
+    /// Interrupt waiting.
+    void finish();
+};
+
+}
diff --git a/src/Processors/Executors/PipelineExecutor.cpp b/src/Processors/Executors/PipelineExecutor.cpp
index 517e07a3ba4..a9f6562693e 100644
--- a/src/Processors/Executors/PipelineExecutor.cpp
+++ b/src/Processors/Executors/PipelineExecutor.cpp
@@ -254,10 +254,6 @@ bool PipelineExecutor::prepareProcessor(UInt64 pid, size_t thread_number, Queue
 //            addAsyncJob(pid);
 //            break;
             }
-            case IProcessor::Status::Wait:
-            {
-                throw Exception("Wait is temporary not supported.", ErrorCodes::LOGICAL_ERROR);
-            }
             case IProcessor::Status::ExpandPipeline:
             {
                 need_expand_pipeline = true;
diff --git a/src/Processors/IProcessor.h b/src/Processors/IProcessor.h
index c774b43a9b2..08adc151424 100644
--- a/src/Processors/IProcessor.h
+++ b/src/Processors/IProcessor.h
@@ -146,13 +146,10 @@ public:
         /// You may call 'work' method and processor will do some work synchronously.
         Ready,
 
-        /// You may call 'schedule' method and processor will initiate some background work.
+        /// You may call 'schedule' method and processor will return descriptor.
+        /// You need to poll this descriptor and call work() afterwards.
         Async,
 
-        /// Processor is doing some work in background.
-        /// You may wait for next event or do something else and then you should call 'prepare' again.
-        Wait,
-
         /// Processor wants to add other processors to pipeline.
         /// New processors must be obtained by expandPipeline() call.
         ExpandPipeline,
@@ -207,7 +204,7 @@ public:
       * Note that it can fire many events in EventCounter while doing its job,
       *  and you have to wait for next event (or do something else) every time when 'prepare' returned Wait.
       */
-    virtual void schedule(EventCounter & /*watch*/)
+    virtual int schedule()
     {
         throw Exception("Method 'schedule' is not implemented for " + getName() + " processor", ErrorCodes::NOT_IMPLEMENTED);
     }

From 00492eec67f115cbbef24729622f82a5234931cd Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 4 Dec 2020 12:55:27 +0300
Subject: [PATCH 043/504] Support of Async status for PipelineExecutor.

---
 src/Processors/Executors/AsyncTaskQueue.cpp   |  2 +-
 src/Processors/Executors/AsyncTaskQueue.h     |  2 +-
 src/Processors/Executors/PipelineExecutor.cpp | 70 ++++++++++++++-----
 src/Processors/Executors/PipelineExecutor.h   | 10 ++-
 4 files changed, 64 insertions(+), 20 deletions(-)

diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index 417b3ce25f1..d0a63a87165 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -40,7 +40,7 @@ void AsyncTaskQueue::addTask(void * data, int fd)
         condvar.notify_one();
 }
 
-void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> lock)
+void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
 {
     condvar.wait(lock, [&] { return !empty() || is_finished; });
 
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index b66fe932d46..2f3cdf8ab82 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -26,7 +26,7 @@ public:
     /// Wait for any descriptor. If no descriptors in queue, blocks.
     /// Returns ptr which was inserted into queue or nullptr if finished was called.
     /// Lock is used to wait on condvar.
-    void * wait(std::unique_lock<std::mutex> lock);
+    void * wait(std::unique_lock<std::mutex> & lock);
 
     /// Interrupt waiting.
     void finish();
diff --git a/src/Processors/Executors/PipelineExecutor.cpp b/src/Processors/Executors/PipelineExecutor.cpp
index a9f6562693e..dce21067557 100644
--- a/src/Processors/Executors/PipelineExecutor.cpp
+++ b/src/Processors/Executors/PipelineExecutor.cpp
@@ -164,7 +164,7 @@ bool PipelineExecutor::expandPipeline(Stack & stack, UInt64 pid)
     return true;
 }
 
-bool PipelineExecutor::tryAddProcessorToStackIfUpdated(ExecutingGraph::Edge & edge, Queue & queue, size_t thread_number)
+bool PipelineExecutor::tryAddProcessorToStackIfUpdated(ExecutingGraph::Edge & edge, Queue & queue, Queue & async_queue, size_t thread_number)
 {
     /// In this method we have ownership on edge, but node can be concurrently accessed.
 
@@ -185,7 +185,7 @@ bool PipelineExecutor::tryAddProcessorToStackIfUpdated(ExecutingGraph::Edge & ed
     if (status == ExecutingGraph::ExecStatus::Idle)
     {
         node.status = ExecutingGraph::ExecStatus::Preparing;
-        return prepareProcessor(edge.to, thread_number, queue, std::move(lock));
+        return prepareProcessor(edge.to, thread_number, queue, async_queue, std::move(lock));
     }
     else
         graph->nodes[edge.to]->processor->onUpdatePorts();
@@ -193,7 +193,7 @@ bool PipelineExecutor::tryAddProcessorToStackIfUpdated(ExecutingGraph::Edge & ed
     return true;
 }
 
-bool PipelineExecutor::prepareProcessor(UInt64 pid, size_t thread_number, Queue & queue, std::unique_lock<std::mutex> node_lock)
+bool PipelineExecutor::prepareProcessor(UInt64 pid, size_t thread_number, Queue & queue, Queue & async_queue, std::unique_lock<std::mutex> node_lock)
 {
     /// In this method we have ownership on node.
     auto & node = *graph->nodes[pid];
@@ -248,11 +248,9 @@ bool PipelineExecutor::prepareProcessor(UInt64 pid, size_t thread_number, Queue
             }
             case IProcessor::Status::Async:
             {
-                throw Exception("Async is temporary not supported.", ErrorCodes::LOGICAL_ERROR);
-
-//            node.status = ExecStatus::Executing;
-//            addAsyncJob(pid);
-//            break;
+                node.status = ExecutingGraph::ExecStatus::Executing;
+                async_queue.push(&node);
+                break;
             }
             case IProcessor::Status::ExpandPipeline:
             {
@@ -284,13 +282,13 @@ bool PipelineExecutor::prepareProcessor(UInt64 pid, size_t thread_number, Queue
     {
         for (auto & edge : updated_direct_edges)
         {
-            if (!tryAddProcessorToStackIfUpdated(*edge, queue, thread_number))
+            if (!tryAddProcessorToStackIfUpdated(*edge, queue, async_queue, thread_number))
                 return false;
         }
 
         for (auto & edge : updated_back_edges)
         {
-            if (!tryAddProcessorToStackIfUpdated(*edge, queue, thread_number))
+            if (!tryAddProcessorToStackIfUpdated(*edge, queue, async_queue, thread_number))
                 return false;
         }
     }
@@ -321,7 +319,7 @@ bool PipelineExecutor::prepareProcessor(UInt64 pid, size_t thread_number, Queue
         while (!stack.empty())
         {
             auto item = stack.top();
-            if (!prepareProcessor(item, thread_number, queue, std::unique_lock<std::mutex>(graph->nodes[item]->status_mutex)))
+            if (!prepareProcessor(item, thread_number, queue, async_queue, std::unique_lock<std::mutex>(graph->nodes[item]->status_mutex)))
                 return false;
 
             stack.pop();
@@ -374,6 +372,7 @@ void PipelineExecutor::finish()
     {
         std::lock_guard lock(task_queue_mutex);
         finished = true;
+        async_task_queue.finish();
     }
 
     std::lock_guard guard(executor_contexts_mutex);
@@ -518,13 +517,24 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                     break;
                 }
 
-                if (threads_queue.size() + 1 == num_threads)
+                if (threads_queue.size() + 1 == num_threads && async_task_queue.empty())
                 {
                     lock.unlock();
                     finish();
                     break;
                 }
 
+                if (num_threads == 1)
+                {
+                    /// If we execute in single thread, wait for async tasks here.
+                    void * res = async_task_queue.wait(lock);
+                    if (!res)
+                        throw Exception("Empty task was returned from async task queue", ErrorCodes::LOGICAL_ERROR);
+
+                    node = static_cast<ExecutingGraph::Node *>(res);
+                    continue;
+                }
+
                 threads_queue.push(thread_num);
             }
 
@@ -575,6 +585,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
             /// Try to execute neighbour processor.
             {
                 Queue queue;
+                Queue async_queue;
 
                 ++num_processing_executors;
                 while (auto * task = expand_pipeline_task.load())
@@ -583,7 +594,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                 /// Prepare processor after execution.
                 {
                     auto lock = std::unique_lock<std::mutex>(node->status_mutex);
-                    if (!prepareProcessor(node->processors_id, thread_num, queue, std::move(lock)))
+                    if (!prepareProcessor(node->processors_id, thread_num, queue, async_queue, std::move(lock)))
                         finish();
                 }
 
@@ -597,17 +608,23 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                 }
 
                 /// Push other tasks to global queue.
-                if (!queue.empty())
+                if (!queue.empty() || !async_queue.empty())
                 {
                     std::unique_lock lock(task_queue_mutex);
 
+                    while (!async_queue.empty() && !finished)
+                    {
+                        async_task_queue.addTask(async_queue.front(), async_queue.front()->processor->schedule());
+                        async_queue.pop();
+                    }
+
                     while (!queue.empty() && !finished)
                     {
                         task_queue.push(queue.front(), thread_num);
                         queue.pop();
                     }
 
-                    if (!threads_queue.empty() && !finished /* && task_queue.quota() > threads_queue.size()*/)
+                    if (!threads_queue.empty() && !finished)
                     {
                         auto thread_to_wake = task_queue.getAnyThreadWithTasks(thread_num + 1 == num_threads ? 0 : (thread_num + 1));
 
@@ -665,6 +682,7 @@ void PipelineExecutor::initializeExecution(size_t num_threads)
         std::lock_guard lock(task_queue_mutex);
 
         Queue queue;
+        Queue async_queue;
         size_t next_thread = 0;
 
         while (!stack.empty())
@@ -672,7 +690,7 @@ void PipelineExecutor::initializeExecution(size_t num_threads)
             UInt64 proc = stack.top();
             stack.pop();
 
-            prepareProcessor(proc, 0, queue, std::unique_lock<std::mutex>(graph->nodes[proc]->status_mutex));
+            prepareProcessor(proc, 0, queue, async_queue, std::unique_lock<std::mutex>(graph->nodes[proc]->status_mutex));
 
             while (!queue.empty())
             {
@@ -683,6 +701,12 @@ void PipelineExecutor::initializeExecution(size_t num_threads)
                 if (next_thread >= num_threads)
                     next_thread = 0;
             }
+
+            while (!async_queue.empty())
+            {
+                async_task_queue.addTask(async_queue.front(), async_queue.front()->processor->schedule());
+                async_queue.pop();
+            }
         }
     }
 }
@@ -743,6 +767,20 @@ void PipelineExecutor::executeImpl(size_t num_threads)
             });
         }
 
+        {
+            /// Wait for async tasks.
+            std::unique_lock lock(task_queue_mutex);
+            size_t next_thread = 0;
+            while (void * task = async_task_queue.wait(lock))
+            {
+                task_queue.push(static_cast<ExecutingGraph::Node *>(task), next_thread);
+
+                ++next_thread;
+                if (next_thread >= num_threads)
+                    next_thread = 0;
+            }
+        }
+
         for (auto & thread : threads)
             if (thread.joinable())
                 thread.join();
diff --git a/src/Processors/Executors/PipelineExecutor.h b/src/Processors/Executors/PipelineExecutor.h
index b457cca34b1..34339a34608 100644
--- a/src/Processors/Executors/PipelineExecutor.h
+++ b/src/Processors/Executors/PipelineExecutor.h
@@ -1,6 +1,7 @@
 #pragma once
 
 #include <Processors/IProcessor.h>
+#include <Processors/Executors/AsyncTaskQueue.h>
 #include <Processors/Executors/ThreadsQueue.h>
 #include <Processors/Executors/TasksQueue.h>
 #include <Processors/Executors/ExecutingGraph.h>
@@ -57,6 +58,11 @@ private:
     /// Stores processors need to be prepared. Preparing status is already set for them.
     TaskQueue<ExecutingGraph::Node> task_queue;
 
+    /// Queue which stores tasks where processors returned Async status after prepare.
+    /// If multiple threads are using, main thread will wait for async tasks.
+    /// For single thread, will wait for async tasks only when task_queue is empty.
+    AsyncTaskQueue async_task_queue;
+
     ThreadsQueue threads_queue;
     std::mutex task_queue_mutex;
 
@@ -126,14 +132,14 @@ private:
 
     /// Pipeline execution related methods.
     void addChildlessProcessorsToStack(Stack & stack);
-    bool tryAddProcessorToStackIfUpdated(ExecutingGraph::Edge & edge, Queue & queue, size_t thread_number);
+    bool tryAddProcessorToStackIfUpdated(ExecutingGraph::Edge & edge, Queue & queue, Queue & async_queue, size_t thread_number);
     static void addJob(ExecutingGraph::Node * execution_state);
     // TODO: void addAsyncJob(UInt64 pid);
 
     /// Prepare processor with pid number.
     /// Check parents and children of current processor and push them to stacks if they also need to be prepared.
     /// If processor wants to be expanded, ExpandPipelineTask from thread_number's execution context will be used.
-    bool prepareProcessor(UInt64 pid, size_t thread_number, Queue & queue, std::unique_lock<std::mutex> node_lock);
+    bool prepareProcessor(UInt64 pid, size_t thread_number, Queue & queue, Queue & async_queue, std::unique_lock<std::mutex> node_lock);
     bool doExpandPipeline(ExpandPipelineTask * task, bool processing);
 
     /// Continue executor (in case there are tasks in queue).

From 9ca837d5355c417a9fc9a2ac0c86b9a56e184d12 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 4 Dec 2020 13:52:57 +0300
Subject: [PATCH 044/504] Add async status to RemoteSource.

---
 src/DataStreams/RemoteQueryExecutor.cpp | 15 ++++++-----
 src/DataStreams/RemoteQueryExecutor.h   |  5 +++-
 src/Processors/ISource.cpp              | 33 ++++++++++++++++++++++---
 src/Processors/ISource.h                |  3 ++-
 src/Processors/Sources/RemoteSource.cpp | 17 +++++++++++--
 src/Processors/Sources/RemoteSource.h   | 10 +++++++-
 6 files changed, 68 insertions(+), 15 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 100f20b4123..56c8dc128fe 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -25,8 +25,11 @@ namespace ErrorCodes
     extern const int SOCKET_TIMEOUT;
 }
 
-struct RemoteQueryExecutor::ReadContext
+class RemoteQueryExecutorReadContext
 {
+public:
+    using Self = RemoteQueryExecutorReadContext;
+
     bool is_read_in_progress = false;
     Packet packet;
 
@@ -41,7 +44,7 @@ struct RemoteQueryExecutor::ReadContext
     int socket_fd;
     int epoll_fd;
 
-    explicit ReadContext(MultiplexedConnections & connections_) : connections(connections_)
+    explicit RemoteQueryExecutorReadContext(MultiplexedConnections & connections_) : connections(connections_)
     {
         auto & socket = connections.getSocket();
         socket_fd = socket.impl()->sockfd();
@@ -72,11 +75,11 @@ struct RemoteQueryExecutor::ReadContext
         fiber = boost::context::fiber(std::allocator_arg_t(), stack, Routine{connections, *this});
     }
 
-    static void initialize(std::unique_ptr<ReadContext> & read_context, MultiplexedConnections & connections)
+    static void initialize(std::unique_ptr<Self> & read_context, MultiplexedConnections & connections)
     {
         try
         {
-            read_context = std::make_unique<ReadContext>(connections);
+            read_context = std::make_unique<Self>(connections);
         }
         catch (DB::Exception & e)
         {
@@ -141,7 +144,7 @@ struct RemoteQueryExecutor::ReadContext
             std::rethrow_exception(std::move(exception));
     }
 
-    ~ReadContext()
+    ~RemoteQueryExecutorReadContext()
     {
         /// socket_fd is closed by Poco::Net::Socket
         /// timer_fd is closed by TimerDescriptor
@@ -151,7 +154,7 @@ struct RemoteQueryExecutor::ReadContext
     struct Routine
     {
         MultiplexedConnections & connections;
-        ReadContext & read_context;
+        Self & read_context;
 
         boost::context::fiber operator()(boost::context::fiber && sink) const
         {
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 6de5847015a..2ca13ffd9b7 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -19,10 +19,14 @@ using ProgressCallback = std::function<void(const Progress & progress)>;
 struct BlockStreamProfileInfo;
 using ProfileInfoCallback = std::function<void(const BlockStreamProfileInfo & info)>;
 
+class RemoteQueryExecutorReadContext;
+
 /// This class allows one to launch queries on remote replicas of one shard and get results
 class RemoteQueryExecutor
 {
 public:
+    using ReadContext = RemoteQueryExecutorReadContext;
+
     /// Takes already set connection.
     /// If `settings` is nullptr, settings will be taken from context.
     RemoteQueryExecutor(
@@ -57,7 +61,6 @@ public:
 
     /// Async variant of read. Returns ready block or file descriptor which may be used for polling.
     /// ReadContext is an internal read state. Pass empty ptr first time, reuse created one for every call.
-    struct ReadContext;
     std::variant<Block, int> read(std::unique_ptr<ReadContext> & read_context);
 
     /// Receive all remain packets and finish query.
diff --git a/src/Processors/ISource.cpp b/src/Processors/ISource.cpp
index 90f3962b83e..7ae988f7cdb 100644
--- a/src/Processors/ISource.cpp
+++ b/src/Processors/ISource.cpp
@@ -4,6 +4,11 @@
 namespace DB
 {
 
+namespace ErrorCodes
+{
+    extern const int NOT_IMPLEMENTED;
+}
+
 ISource::ISource(Block header)
     : IProcessor({}, {std::move(header)}), output(outputs.front())
 {
@@ -45,11 +50,17 @@ void ISource::work()
 {
     try
     {
-        current_chunk.chunk = generate();
-        if (!current_chunk.chunk || isCancelled())
-            finished = true;
+        if (auto chunk = tryGenerate())
+        {
+            current_chunk.chunk = std::move(*chunk);
+            if (current_chunk.chunk)
+                has_input = true;
+        }
         else
-            has_input = true;
+            finished = true;
+
+        if (isCancelled())
+            finished = true;
     }
     catch (...)
     {
@@ -58,5 +69,19 @@ void ISource::work()
     }
 }
 
+Chunk ISource::generate()
+{
+    throw Exception(ErrorCodes::NOT_IMPLEMENTED, "generate is not implemented for {}", getName());
+}
+
+std::optional<Chunk> ISource::tryGenerate()
+{
+    auto chunk = generate();
+    if (!chunk)
+        return {};
+
+    return chunk;
+}
+
 }
 
diff --git a/src/Processors/ISource.h b/src/Processors/ISource.h
index 9be21c3a398..b7e2b5dce8e 100644
--- a/src/Processors/ISource.h
+++ b/src/Processors/ISource.h
@@ -15,7 +15,8 @@ protected:
     bool got_exception = false;
     Port::Data current_chunk;
 
-    virtual Chunk generate() = 0;
+    virtual Chunk generate();
+    virtual std::optional<Chunk> tryGenerate();
 
 public:
     ISource(Block header);
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index 2efbf3ed7c2..a277a794b36 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -21,6 +21,9 @@ RemoteSource::~RemoteSource() = default;
 
 ISource::Status RemoteSource::prepare()
 {
+    if (is_async_state)
+        return Status::Async;
+
     Status status = SourceWithProgress::prepare();
     /// To avoid resetting the connection (because of "unfinished" query) in the
     /// RemoteQueryExecutor it should be finished explicitly.
@@ -29,7 +32,7 @@ ISource::Status RemoteSource::prepare()
     return status;
 }
 
-Chunk RemoteSource::generate()
+std::optional<Chunk> RemoteSource::tryGenerate()
 {
     /// onCancel() will do the cancel if the query was sent.
     if (was_query_canceled)
@@ -52,7 +55,17 @@ Chunk RemoteSource::generate()
         was_query_sent = true;
     }
 
-    auto block = query_executor->read();
+    auto res = query_executor->read(read_context);
+    if (std::holds_alternative<int>(res))
+    {
+        fd = std::get<int>(res);
+        is_async_state = true;
+        return Chunk();
+    }
+
+    is_async_state = false;
+
+    auto block = std::get<Block>(std::move(res));
 
     if (!block)
     {
diff --git a/src/Processors/Sources/RemoteSource.h b/src/Processors/Sources/RemoteSource.h
index 7b537023306..0a04f72d14a 100644
--- a/src/Processors/Sources/RemoteSource.h
+++ b/src/Processors/Sources/RemoteSource.h
@@ -11,6 +11,8 @@ namespace DB
 class RemoteQueryExecutor;
 using RemoteQueryExecutorPtr = std::shared_ptr<RemoteQueryExecutor>;
 
+class RemoteQueryExecutorReadContext;
+
 /// Source from RemoteQueryExecutor. Executes remote query and returns query result chunks.
 class RemoteSource : public SourceWithProgress
 {
@@ -33,8 +35,10 @@ public:
             cancel();
     }
 
+    int schedule() override { return fd; }
+
 protected:
-    Chunk generate() override;
+    std::optional<Chunk> tryGenerate() override;
     void onCancel() override;
 
 private:
@@ -43,6 +47,10 @@ private:
     bool add_aggregation_info = false;
     RemoteQueryExecutorPtr query_executor;
     RowsBeforeLimitCounterPtr rows_before_limit;
+
+    bool is_async_state = false;
+    std::unique_ptr<RemoteQueryExecutorReadContext> read_context;
+    int fd;
 };
 
 /// Totals source from RemoteQueryExecutor.

From f31610a2e1ae503868d64164d8d9e7ae3a7ebacb Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 4 Dec 2020 13:55:01 +0300
Subject: [PATCH 045/504] Remove Wait status.

---
 src/Processors/Formats/InputStreamFromInputFormat.h | 1 -
 src/Storages/RabbitMQ/RabbitMQBlockInputStream.cpp  | 1 -
 2 files changed, 2 deletions(-)

diff --git a/src/Processors/Formats/InputStreamFromInputFormat.h b/src/Processors/Formats/InputStreamFromInputFormat.h
index 792d2e45f7f..4369287d39e 100644
--- a/src/Processors/Formats/InputStreamFromInputFormat.h
+++ b/src/Processors/Formats/InputStreamFromInputFormat.h
@@ -56,7 +56,6 @@ protected:
 
                 case IProcessor::Status::NeedData:
                 case IProcessor::Status::Async:
-                case IProcessor::Status::Wait:
                 case IProcessor::Status::ExpandPipeline:
                     throw Exception("Source processor returned status " + IProcessor::statusToName(status), ErrorCodes::LOGICAL_ERROR);
             }
diff --git a/src/Storages/RabbitMQ/RabbitMQBlockInputStream.cpp b/src/Storages/RabbitMQ/RabbitMQBlockInputStream.cpp
index 4e6de4d4715..9a47a218128 100644
--- a/src/Storages/RabbitMQ/RabbitMQBlockInputStream.cpp
+++ b/src/Storages/RabbitMQ/RabbitMQBlockInputStream.cpp
@@ -137,7 +137,6 @@ Block RabbitMQBlockInputStream::readImpl()
                 }
                 case IProcessor::Status::NeedData:
                 case IProcessor::Status::Async:
-                case IProcessor::Status::Wait:
                 case IProcessor::Status::ExpandPipeline:
                     throw Exception("Source processor returned status " + IProcessor::statusToName(status), ErrorCodes::LOGICAL_ERROR);
             }

From eb0c817bf2887f5a5fd25ffc1ae9b3314435798f Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Wed, 9 Dec 2020 14:14:40 +0300
Subject: [PATCH 046/504] expression analyzer part

---
 .../AggregateFunctionFactory.cpp              |   3 +
 src/Interpreters/ActionsVisitor.cpp           |   8 +
 src/Interpreters/AggregateDescription.h       |  35 ++-
 src/Interpreters/ExpressionAnalyzer.cpp       | 237 +++++++++++++++++-
 src/Interpreters/ExpressionAnalyzer.h         |  12 +
 src/Interpreters/GetAggregatesVisitor.h       |  51 +++-
 src/Interpreters/TreeRewriter.cpp             |  36 +++
 src/Interpreters/TreeRewriter.h               |   2 +
 src/Parsers/IAST.cpp                          |   7 +
 src/Parsers/IAST.h                            |   1 +
 src/Processors/Transforms/FilterTransform.cpp |   3 +
 11 files changed, 377 insertions(+), 18 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionFactory.cpp b/src/AggregateFunctions/AggregateFunctionFactory.cpp
index 1c66650be45..2fdf3c18a68 100644
--- a/src/AggregateFunctions/AggregateFunctionFactory.cpp
+++ b/src/AggregateFunctions/AggregateFunctionFactory.cpp
@@ -58,6 +58,9 @@ static DataTypes convertLowCardinalityTypesToNested(const DataTypes & types)
 AggregateFunctionPtr AggregateFunctionFactory::get(
     const String & name, const DataTypes & argument_types, const Array & parameters, AggregateFunctionProperties & out_properties) const
 {
+    fmt::print(stderr, "get aggregate function {} at \n{}\n",
+        name, StackTrace().toString());
+
     auto type_without_low_cardinality = convertLowCardinalityTypesToNested(argument_types);
 
     /// If one of the types is Nullable, we apply aggregate function combinator "Null".
diff --git a/src/Interpreters/ActionsVisitor.cpp b/src/Interpreters/ActionsVisitor.cpp
index a44dba84aa6..5df50808a59 100644
--- a/src/Interpreters/ActionsVisitor.cpp
+++ b/src/Interpreters/ActionsVisitor.cpp
@@ -738,6 +738,14 @@ void ActionsMatcher::visit(const ASTFunction & node, const ASTPtr & ast, Data &
     if (AggregateFunctionFactory::instance().isAggregateFunctionName(node.name))
         return;
 
+    /// FIXME need proper grammar for window functions. For now, ignore it --
+    /// the resulting column is added in ExpressionAnalyzer, similar to the
+    /// aggregate functions.
+    if (node.name == "window")
+    {
+        return;
+    }
+
     FunctionOverloadResolverPtr function_builder;
     try
     {
diff --git a/src/Interpreters/AggregateDescription.h b/src/Interpreters/AggregateDescription.h
index 396a62c446a..528b38e86f7 100644
--- a/src/Interpreters/AggregateDescription.h
+++ b/src/Interpreters/AggregateDescription.h
@@ -1,13 +1,18 @@
 #pragma once
 
+#include <AggregateFunctions/IAggregateFunction.h>
+#include <DataTypes/IDataType.h>
 #include <Core/ColumnNumbers.h>
 #include <Core/Names.h>
-#include <AggregateFunctions/IAggregateFunction.h>
+#include <Core/SortDescription.h>
+#include <Parsers/IAST_fwd.h>
 
 
 namespace DB
 {
 
+class ASTFunction;
+
 struct AggregateDescription
 {
     AggregateFunctionPtr function;
@@ -21,4 +26,32 @@ struct AggregateDescription
 
 using AggregateDescriptions = std::vector<AggregateDescription>;
 
+
+struct WindowFunctionDescription
+{
+    std::string window_name;
+    std::string column_name;
+    const IAST * wrapper_node;
+    const ASTFunction * function_node;
+    AggregateFunctionPtr aggregate_function;
+    Array function_parameters;
+    DataTypes argument_types;
+    Names argument_names;
+
+    std::string dump() const;
+};
+
+struct WindowDescription
+{
+    std::string window_name;
+    // Always ASC for now.
+    std::vector<std::string> partition_by;
+    std::vector<std::string> order_by;
+    // No frame info as of yet.
+};
+
+using WindowFunctionDescriptions = std::vector<WindowFunctionDescription>;
+
+using WindowDescriptions = std::unordered_map<std::string, WindowDescription>;
+
 }
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 4ac337c81e2..60c534c82ad 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1,5 +1,7 @@
 #include <Core/Block.h>
 
+#include <iostream>
+
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ASTIdentifier.h>
 #include <Parsers/ASTLiteral.h>
@@ -57,12 +59,13 @@ using LogAST = DebugASTLog<false>; /// set to true to enable logs
 
 namespace ErrorCodes
 {
-    extern const int UNKNOWN_TYPE_OF_AST_NODE;
-    extern const int UNKNOWN_IDENTIFIER;
+    extern const int BAD_ARGUMENTS;
     extern const int ILLEGAL_PREWHERE;
-    extern const int LOGICAL_ERROR;
     extern const int ILLEGAL_TYPE_OF_ARGUMENT;
     extern const int ILLEGAL_TYPE_OF_COLUMN_FOR_FILTER;
+    extern const int LOGICAL_ERROR;
+    extern const int UNKNOWN_IDENTIFIER;
+    extern const int UNKNOWN_TYPE_OF_AST_NODE;
 }
 
 namespace
@@ -277,6 +280,8 @@ void ExpressionAnalyzer::analyzeAggregation()
     {
         aggregated_columns = temp_actions->getNamesAndTypesList();
     }
+
+    has_window = makeWindowDescriptions(temp_actions);
 }
 
 
@@ -438,7 +443,11 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
 
             auto it = index.find(name);
             if (it == index.end())
-                throw Exception(ErrorCodes::UNKNOWN_IDENTIFIER, "Unknown identifier (in aggregate function '{}'): {}", node->name, name);
+            {
+                throw Exception(ErrorCodes::UNKNOWN_IDENTIFIER,
+                    "Unknown identifier '{}' in aggregate function '{}'",
+                    name, node->formatForErrorMessage());
+            }
 
             types[i] = (*it)->result_type;
             aggregate.argument_names[i] = name;
@@ -455,6 +464,125 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
 }
 
 
+bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
+{
+    for (const ASTFunction * wrapper_node : windowFunctions())
+    {
+        fmt::print(stderr, "window function ast: {}\n", wrapper_node->dumpTree());
+
+        // Not sure why NoMakeSet, copied from aggregate functions.
+        getRootActionsNoMakeSet(wrapper_node->arguments, true /* no subqueries */,
+            actions);
+
+        // FIXME not thread-safe, should use a per-query counter.
+        static int window_index = 1;
+
+        WindowDescription window_description;
+        window_description.window_name = fmt::format("window_{}", window_index++);
+
+        const auto * elist = wrapper_node->arguments
+            ? wrapper_node->arguments->as<const ASTExpressionList>()
+            : nullptr;
+        if (elist)
+        {
+            if (elist->children.size() >= 2)
+            {
+                const auto partition_by_ast = elist->children[1];
+                fmt::print(stderr, "partition by ast {}\n",
+                    partition_by_ast->dumpTree());
+                if (const auto * as_tuple = partition_by_ast->as<ASTFunction>();
+                    as_tuple
+                    && as_tuple->name == "tuple"
+                    && as_tuple->arguments)
+                {
+                    // untuple it
+                    for (const auto & element_ast
+                        : as_tuple->arguments->children)
+                    {
+                        const auto * with_alias = dynamic_cast<
+                            const ASTWithAlias *>(element_ast.get());
+                        if (!with_alias)
+                        {
+                            throw Exception(ErrorCodes::BAD_ARGUMENTS,
+                                "(1) Expected column in PARTITION BY"
+                                " for window '{}', got '{}'",
+                                window_description.window_name,
+                                element_ast->formatForErrorMessage());
+                        }
+                        window_description.partition_by.push_back(
+                            with_alias->getColumnName());
+                    }
+                }
+                else if (const auto * with_alias
+                    = dynamic_cast<const ASTWithAlias *>(partition_by_ast.get()))
+                {
+                    window_description.partition_by.push_back(
+                        with_alias->getColumnName());
+                }
+                else
+                {
+                    throw Exception(ErrorCodes::BAD_ARGUMENTS,
+                        "(2) Expected tuple or column in PARTITION BY"
+                        " for window '{}', got '{}'",
+                        window_description.window_name,
+                        partition_by_ast->formatForErrorMessage());
+                }
+            }
+        }
+
+        WindowFunctionDescription window_function;
+        window_function.window_name = window_description.window_name;
+        window_function.wrapper_node = wrapper_node;
+        window_function.function_node
+            = &elist->children.at(0)->as<ASTFunction &>();
+        window_function.column_name
+            = window_function.function_node->getColumnName();
+        window_function.function_parameters
+            = window_function.function_node->parameters
+                ? getAggregateFunctionParametersArray(
+                    window_function.function_node->parameters)
+                : Array();
+
+        const ASTs & arguments
+            = window_function.function_node->arguments->children;
+        window_function.argument_types.resize(arguments.size());
+        window_function.argument_names.resize(arguments.size());
+        const auto & index = actions->getIndex();
+        for (size_t i = 0; i < arguments.size(); ++i)
+        {
+            const std::string & name = arguments[i]->getColumnName();
+
+            auto it = index.find(name);
+            if (it == index.end())
+            {
+                throw Exception(ErrorCodes::UNKNOWN_IDENTIFIER,
+                    "Unknown identifier '{}' in window function '{}'",
+                    name, window_function.function_node->formatForErrorMessage());
+            }
+
+            window_function.argument_types[i] = (*it)->result_type;
+            window_function.argument_names[i] = name;
+        }
+
+
+        AggregateFunctionProperties properties;
+        window_function.aggregate_function
+            = AggregateFunctionFactory::instance().get(
+                window_function.function_node->name,
+                window_function.argument_types,
+                window_function.function_parameters, properties);
+
+        window_descriptions.insert({window_description.window_name,
+            window_description});
+        window_functions.push_back(window_function);
+
+        fmt::print(stderr, "{}\n", window_function.dump());
+    }
+
+    return !windowFunctions().empty();
+}
+
+
 const ASTSelectQuery * ExpressionAnalyzer::getSelectQuery() const
 {
     const auto * select_query = query->as<ASTSelectQuery>();
@@ -822,6 +950,37 @@ void SelectQueryExpressionAnalyzer::appendAggregateFunctionsArguments(Expression
     for (const ASTFunction * node : data.aggregates)
         for (auto & argument : node->arguments->children)
             getRootActions(argument, only_types, step.actions());
+
+
+    fmt::print(stderr, "actions after appendAggregateFunctionsArguments: \n{} at \n{}\n", chain.dumpChain(), StackTrace().toString());
+}
+
+void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
+    ExpressionActionsChain & chain, bool only_types)
+{
+    fmt::print(stderr, "actions before window: {}\n", chain.dumpChain());
+    const auto * select_query = getSelectQuery();
+
+    ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
+
+    /*
+    for (const auto & desc : aggregate_descriptions)
+        for (const auto & name : desc.argument_names)
+            step.required_output.emplace_back(name);
+    */
+
+    /// Collect aggregates removing duplicates by node.getColumnName()
+    /// It's not clear why we recollect aggregates (for query parts) while we're able to use previously collected ones (for entire query)
+    /// @note The original recollection logic didn't remove duplicates.
+    GetAggregatesVisitor::Data data;
+    GetAggregatesVisitor(data).visit(select_query->select());
+
+    /// TODO: data.aggregates -> aggregates()
+    for (const ASTFunction * node : data.window_functions)
+        for (auto & argument : node->arguments->children)
+            getRootActions(argument, only_types, step.actions());
+
+    fmt::print(stderr, "actions after window: {}\n", chain.dumpChain());
 }
 
 bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain, bool only_types)
@@ -831,6 +990,8 @@ bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain,
     if (!select_query->having())
         return false;
 
+    fmt::print(stderr, "has having\n");
+
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
     getRootActionsForHaving(select_query->having(), only_types, step.actions());
@@ -848,7 +1009,16 @@ void SelectQueryExpressionAnalyzer::appendSelect(ExpressionActionsChain & chain,
     getRootActions(select_query->select(), only_types, step.actions());
 
     for (const auto & child : select_query->select()->children)
+    {
+        /// FIXME add proper grammar for window functions
+        if (const auto * as_function = child->as<ASTFunction>();
+            as_function && as_function->name == "window")
+        {
+            continue;
+        }
+
         step.required_output.push_back(child->getColumnName());
+    }
 }
 
 ActionsDAGPtr SelectQueryExpressionAnalyzer::appendOrderBy(ExpressionActionsChain & chain, bool only_types, bool optimize_read_in_order,
@@ -943,6 +1113,13 @@ ActionsDAGPtr SelectQueryExpressionAnalyzer::appendProjectResult(ExpressionActio
     ASTs asts = select_query->select()->children;
     for (const auto & ast : asts)
     {
+        /// FIXME add proper grammar for window functions
+        if (const auto * as_function = ast->as<ASTFunction>();
+            as_function && as_function->name == "window")
+        {
+            continue;
+        }
+
         String result_name = ast->getAliasOrColumnName();
         if (required_result_columns.empty() || required_result_columns.count(result_name))
         {
@@ -1069,6 +1246,7 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
     : first_stage(first_stage_)
     , second_stage(second_stage_)
     , need_aggregate(query_analyzer.hasAggregation())
+    , has_window(query_analyzer.hasWindow())
 {
     /// first_stage: Do I need to perform the first part of the pipeline - running on remote servers during distributed processing.
     /// second_stage: Do I need to execute the second part of the pipeline - running on the initiating server during distributed processing.
@@ -1181,6 +1359,8 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
             chain.addStep();
         }
 
+        fmt::print(stderr, "chain before aggregate: {}\n", chain.dumpChain());
+
         if (need_aggregate)
         {
             /// TODO correct conditions
@@ -1189,18 +1369,29 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
                     && storage && query.groupBy();
 
             query_analyzer.appendGroupBy(chain, only_types || !first_stage, optimize_aggregation_in_order, group_by_elements_actions);
+
+            fmt::print(stderr, "chain after appendGroupBy: {}\n", chain.dumpChain());
+
             query_analyzer.appendAggregateFunctionsArguments(chain, only_types || !first_stage);
+
+            fmt::print(stderr, "chain after appendAggregateFunctionsArguments: {}\n", chain.dumpChain());
+
             before_aggregation = chain.getLastActions();
 
             finalize_chain(chain);
 
+            fmt::print(stderr, "chain after finalize_chain: {}\n", chain.dumpChain());
+
             if (query_analyzer.appendHaving(chain, only_types || !second_stage))
             {
+                fmt::print(stderr, "chain after appendHaving: {}\n", chain.dumpChain());
                 before_having = chain.getLastActions();
                 chain.addStep();
             }
         }
 
+        fmt::print(stderr, "chain after aggregate: {}\n", chain.dumpChain());
+
         bool join_allow_read_in_order = true;
         if (hasJoin())
         {
@@ -1216,8 +1407,18 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
             && !query.final()
             && join_allow_read_in_order;
 
+        if (has_window)
+        {
+            query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
+            before_window = chain.getLastActions();
+
+            finalize_chain(chain);
+        }
+
         /// If there is aggregation, we execute expressions in SELECT and ORDER BY on the initiating server, otherwise on the source servers.
         query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
+        fmt::print(stderr, "chain after select: {}\n", chain.dumpChain());
+
         selected_columns = chain.getLastStep().required_output;
         has_order_by = query.orderBy() != nullptr;
         before_order_and_select = query_analyzer.appendOrderBy(
@@ -1226,6 +1427,10 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
                 optimize_read_in_order,
                 order_by_elements_actions);
 
+        fmt::print(stderr, "chain after order by: {}\n", chain.dumpChain());
+
+        //if (h
+
         if (query_analyzer.appendLimitBy(chain, only_types || !second_stage))
         {
             before_limit_by = chain.getLastActions();
@@ -1242,6 +1447,8 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
 
     checkActions();
 
+    fmt::print(stderr, "ExpressionAnalysisResult created at \n{}\n",
+        StackTrace().toString());
     fmt::print(stderr, "ExpressionAnalysisResult: \n{}\n", dump());
 }
 
@@ -1320,9 +1527,9 @@ std::string ExpressionAnalysisResult::dump() const
 {
     std::stringstream ss;
 
-    ss << "ExpressionAnalysisResult\n";
     ss << "need_aggregate " << need_aggregate << "\n";
     ss << "has_order_by " << has_order_by << "\n";
+    ss << "has_window " << has_window << "\n";
 
     if (before_array_join)
     {
@@ -1364,6 +1571,11 @@ std::string ExpressionAnalysisResult::dump() const
         ss << "before_having " << before_having->dumpDAG() << "\n";
     }
 
+    if (before_window)
+    {
+        ss << "before_window " << before_window->dumpDAG() << "\n";
+    }
+
     if (before_order_and_select)
     {
         ss << "before_order_and_select " << before_order_and_select->dumpDAG() << "\n";
@@ -1382,4 +1594,19 @@ std::string ExpressionAnalysisResult::dump() const
     return ss.str();
 }
 
+std::string WindowFunctionDescription::dump() const
+{
+    std::stringstream ss;
+    ss << "window function '" << column_name << "' over '" << window_name <<"\n";
+    ss << "wrapper node " << wrapper_node->dumpTree() << "\n";
+    ss << "function node " << function_node->dumpTree() << "\n";
+    ss << "aggregate function '" << aggregate_function->getName() << "'\n";
+    if (function_parameters.size())
+    {
+        ss << "parameters " << toString(function_parameters) << "\n";
+    }
+
+    return ss.str();
+}
+
 }
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index 0ea7b317ebe..b1d7bfba495 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -60,6 +60,11 @@ struct ExpressionAnalyzerData
     NamesAndTypesList aggregation_keys;
     AggregateDescriptions aggregate_descriptions;
 
+    bool has_window = false;
+    WindowDescriptions window_descriptions;
+    WindowFunctionDescriptions window_functions;
+    NamesAndTypesList window_columns;
+
     bool has_global_subqueries = false;
 
     /// All new temporary tables obtained by performing the GLOBAL IN/JOIN subqueries.
@@ -139,6 +144,7 @@ protected:
     const TableJoin & analyzedJoin() const { return *syntax->analyzed_join; }
     const NamesAndTypesList & sourceColumns() const { return syntax->required_source_columns; }
     const std::vector<const ASTFunction *> & aggregates() const { return syntax->aggregates; }
+    const std::vector<const ASTFunction *> & windowFunctions() const { return syntax->window_functions; }
     /// Find global subqueries in the GLOBAL IN/JOIN sections. Fills in external_tables.
     void initGlobalSubqueriesAndExternalTables(bool do_global);
 
@@ -162,6 +168,8 @@ protected:
     void analyzeAggregation();
     bool makeAggregateDescriptions(ActionsDAGPtr & actions);
 
+    bool makeWindowDescriptions(ActionsDAGPtr & actions);
+
     const ASTSelectQuery * getSelectQuery() const;
 
     bool isRemoteStorage() const { return syntax->is_remote_storage; }
@@ -181,6 +189,7 @@ struct ExpressionAnalysisResult
 
     bool need_aggregate = false;
     bool has_order_by   = false;
+    bool has_window = false;
 
     bool remove_where_filter = false;
     bool optimize_read_in_order = false;
@@ -194,6 +203,7 @@ struct ExpressionAnalysisResult
     ActionsDAGPtr before_where;
     ActionsDAGPtr before_aggregation;
     ActionsDAGPtr before_having;
+    ActionsDAGPtr before_window;
     ActionsDAGPtr before_order_and_select;
     ActionsDAGPtr before_limit_by;
     ActionsDAGPtr final_projection;
@@ -261,6 +271,7 @@ public:
 
     /// Does the expression have aggregate functions or a GROUP BY or HAVING section.
     bool hasAggregation() const { return has_aggregation; }
+    bool hasWindow() const { return has_window; }
     bool hasGlobalSubqueries() { return has_global_subqueries; }
     bool hasTableJoin() const { return syntax->ast_join; }
 
@@ -331,6 +342,7 @@ private:
     bool appendWhere(ExpressionActionsChain & chain, bool only_types);
     bool appendGroupBy(ExpressionActionsChain & chain, bool only_types, bool optimize_aggregation_in_order, ManyExpressionActions &);
     void appendAggregateFunctionsArguments(ExpressionActionsChain & chain, bool only_types);
+    void appendWindowFunctionsArguments(ExpressionActionsChain & chain, bool only_types);
 
     /// After aggregation:
     bool appendHaving(ExpressionActionsChain & chain, bool only_types);
diff --git a/src/Interpreters/GetAggregatesVisitor.h b/src/Interpreters/GetAggregatesVisitor.h
index ba1501fc624..966b301a00b 100644
--- a/src/Interpreters/GetAggregatesVisitor.h
+++ b/src/Interpreters/GetAggregatesVisitor.h
@@ -19,8 +19,10 @@ public:
     struct Data
     {
         const char * assert_no_aggregates = nullptr;
+        const char * assert_no_windows = nullptr;
         std::unordered_set<String> uniq_names;
         std::vector<const ASTFunction *> aggregates;
+        std::vector<const ASTFunction *> window_functions;
     };
 
     static bool needChildVisit(const ASTPtr & node, const ASTPtr & child)
@@ -28,8 +30,14 @@ public:
         if (child->as<ASTSubquery>() || child->as<ASTSelectQuery>())
             return false;
         if (auto * func = node->as<ASTFunction>())
-            if (isAggregateFunction(func->name))
+        {
+            if (isAggregateFunction(func->name)
+                // FIXME temporary hack while we don't have grammar
+                || func->name == "window")
+            {
                 return false;
+            }
+        }
         return true;
     }
 
@@ -42,19 +50,32 @@ public:
 private:
     static void visit(const ASTFunction & node, const ASTPtr &, Data & data)
     {
-        if (!isAggregateFunction(node.name))
-            return;
+        if (isAggregateFunction(node.name))
+        {
+            if (data.assert_no_aggregates)
+                throw Exception("Aggregate function " + node.getColumnName()  + " is found " + String(data.assert_no_aggregates) + " in query",
+                                ErrorCodes::ILLEGAL_AGGREGATION);
 
-        if (data.assert_no_aggregates)
-            throw Exception("Aggregate function " + node.getColumnName()  + " is found " + String(data.assert_no_aggregates) + " in query",
-                            ErrorCodes::ILLEGAL_AGGREGATION);
+            String column_name = node.getColumnName();
+            if (data.uniq_names.count(column_name))
+                return;
 
-        String column_name = node.getColumnName();
-        if (data.uniq_names.count(column_name))
-            return;
+            data.uniq_names.insert(column_name);
+            data.aggregates.push_back(&node);
+        }
+        else if (node.name == "window")
+        {
+            if (data.assert_no_windows)
+                throw Exception("Window function " + node.getColumnName()  + " is found " + String(data.assert_no_windows) + " in query",
+                                ErrorCodes::ILLEGAL_AGGREGATION);
 
-        data.uniq_names.insert(column_name);
-        data.aggregates.push_back(&node);
+            String column_name = node.getColumnName();
+            if (data.uniq_names.count(column_name))
+                return;
+
+            data.uniq_names.insert(column_name);
+            data.window_functions.push_back(&node);
+        }
     }
 
     static bool isAggregateFunction(const String & name)
@@ -66,9 +87,15 @@ private:
 using GetAggregatesVisitor = GetAggregatesMatcher::Visitor;
 
 
+inline void assertNoWindows(const ASTPtr & ast, const char * description)
+{
+    GetAggregatesVisitor::Data data{.assert_no_windows = description};
+    GetAggregatesVisitor(data).visit(ast);
+}
+
 inline void assertNoAggregates(const ASTPtr & ast, const char * description)
 {
-    GetAggregatesVisitor::Data data{description, {}, {}};
+    GetAggregatesVisitor::Data data{.assert_no_aggregates = description};
     GetAggregatesVisitor(data).visit(ast);
 }
 
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 5f1dc950814..e505d968f58 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -380,11 +380,44 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
 
     /// There can not be other aggregate functions within the aggregate functions.
     for (const ASTFunction * node : data.aggregates)
+    {
         for (auto & arg : node->arguments->children)
+        {
             assertNoAggregates(arg, "inside another aggregate function");
+            assertNoWindows(arg, "inside another window function");
+        }
+    }
     return data.aggregates;
 }
 
+std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSelectQuery & select_query)
+{
+    /// There can not be aggregate functions inside the WHERE and PREWHERE.
+    if (select_query.where())
+        assertNoWindows(select_query.where(), "in WHERE");
+    if (select_query.prewhere())
+        assertNoWindows(select_query.prewhere(), "in PREWHERE");
+
+    GetAggregatesVisitor::Data data;
+    GetAggregatesVisitor(data).visit(query);
+
+    /// There can not be other aggregate functions within the aggregate functions.
+    for (const ASTFunction * node : data.window_functions)
+    {
+        for (auto & arg : node->arguments->children)
+        {
+            //assertNoAggregates(arg, "inside another aggregate function");
+            assertNoWindows(arg, "inside another window function");
+        }
+    }
+
+    fmt::print(stderr, "getWindowFunctions ({}) for \n{}\n at \n{}\n",
+        data.window_functions.size(), query->formatForErrorMessage(),
+        StackTrace().toString());
+
+    return data.window_functions;
+}
+
 }
 
 TreeRewriterResult::TreeRewriterResult(
@@ -398,6 +431,8 @@ TreeRewriterResult::TreeRewriterResult(
 {
     collectSourceColumns(add_special);
     is_remote_storage = storage && storage->isRemote();
+
+    fmt::print(stderr, "TreeRewriterResult created at \n{}\n", StackTrace().toString());
 }
 
 /// Add columns from storage to source_columns list. Deduplicate resulted list.
@@ -668,6 +703,7 @@ TreeRewriterResultPtr TreeRewriter::analyzeSelect(
     collectJoinedColumns(*result.analyzed_join, *select_query, tables_with_columns, result.aliases);
 
     result.aggregates = getAggregates(query, *select_query);
+    result.window_functions = getWindowFunctions(query, *select_query);
     result.collectUsedColumns(query, true);
     result.ast_join = select_query->join();
 
diff --git a/src/Interpreters/TreeRewriter.h b/src/Interpreters/TreeRewriter.h
index 4075be4326d..7d801ef9d83 100644
--- a/src/Interpreters/TreeRewriter.h
+++ b/src/Interpreters/TreeRewriter.h
@@ -35,6 +35,8 @@ struct TreeRewriterResult
     Aliases aliases;
     std::vector<const ASTFunction *> aggregates;
 
+    std::vector<const ASTFunction *> window_functions;
+
     /// Which column is needed to be ARRAY-JOIN'ed to get the specified.
     /// For example, for `SELECT s.v ... ARRAY JOIN a AS s` will get "s.v" -> "a.v".
     NameToNameMap array_join_result_to_source;
diff --git a/src/Parsers/IAST.cpp b/src/Parsers/IAST.cpp
index bc00e2afe77..9d8c8997a53 100644
--- a/src/Parsers/IAST.cpp
+++ b/src/Parsers/IAST.cpp
@@ -157,4 +157,11 @@ void IAST::dumpTree(WriteBuffer & ostr, size_t indent) const
         child->dumpTree(ostr, indent + 1);
 }
 
+std::string IAST::dumpTree(size_t indent) const
+{
+    WriteBufferFromOwnString wb;
+    dumpTree(wb, indent);
+    return wb.str();
+}
+
 }
diff --git a/src/Parsers/IAST.h b/src/Parsers/IAST.h
index 01ce4971c45..54eaf088797 100644
--- a/src/Parsers/IAST.h
+++ b/src/Parsers/IAST.h
@@ -77,6 +77,7 @@ public:
     virtual void updateTreeHashImpl(SipHash & hash_state) const;
 
     void dumpTree(WriteBuffer & ostr, size_t indent = 0) const;
+    std::string dumpTree(size_t indent = 0) const;
 
     /** Check the depth of the tree.
       * If max_depth is specified and the depth is greater - throw an exception.
diff --git a/src/Processors/Transforms/FilterTransform.cpp b/src/Processors/Transforms/FilterTransform.cpp
index 23b1bf8a984..d16fbef6905 100644
--- a/src/Processors/Transforms/FilterTransform.cpp
+++ b/src/Processors/Transforms/FilterTransform.cpp
@@ -63,6 +63,9 @@ FilterTransform::FilterTransform(
     auto & column = transformed_header.getByPosition(filter_column_position).column;
     if (column)
         constant_filter_description = ConstantFilterDescription(*column);
+
+    fmt::print(stderr, "FilterTransform created at \n{}\n",
+        StackTrace().toString());
 }
 
 IProcessor::Status FilterTransform::prepare()

From 088c12883275fb57a88ee2e0dfbeffe863e44299 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 4 Dec 2020 14:00:32 +0300
Subject: [PATCH 047/504] Fixing build and tests.

---
 src/Common/FiberStack.h                       |   1 +
 src/DataStreams/RemoteQueryExecutor.cpp       | 163 +----------------
 .../RemoteQueryExecutorReadContext.h          | 172 ++++++++++++++++++
 src/Processors/IProcessor.cpp                 |   2 -
 src/Processors/Sources/RemoteSource.cpp       |   1 +
 src/Storages/Kafka/KafkaBlockInputStream.cpp  |   1 -
 6 files changed, 175 insertions(+), 165 deletions(-)
 create mode 100644 src/DataStreams/RemoteQueryExecutorReadContext.h

diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index c133cb6250c..f490b509443 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -1,3 +1,4 @@
+#pragma once
 #include <boost/context/stack_context.hpp>
 #include <Common/Allocator.h>
 
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 56c8dc128fe..5cf225b8c0f 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -1,4 +1,5 @@
 #include <DataStreams/RemoteQueryExecutor.h>
+#include <DataStreams/RemoteQueryExecutorReadContext.h>
 
 #include <Columns/ColumnConst.h>
 #include <Common/CurrentThread.h>
@@ -19,170 +20,8 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int UNKNOWN_PACKET_FROM_SERVER;
-    extern const int CANNOT_FCNTL;
-    extern const int CANNOT_READ_FROM_SOCKET;
-    extern const int CANNOT_OPEN_FILE;
-    extern const int SOCKET_TIMEOUT;
 }
 
-class RemoteQueryExecutorReadContext
-{
-public:
-    using Self = RemoteQueryExecutorReadContext;
-
-    bool is_read_in_progress = false;
-    Packet packet;
-
-    std::exception_ptr exception;
-    FiberStack<> stack;
-    boost::context::fiber fiber;
-
-    Poco::Timespan receive_timeout;
-    MultiplexedConnections & connections;
-
-    TimerDescriptor timer{CLOCK_MONOTONIC, 0};
-    int socket_fd;
-    int epoll_fd;
-
-    explicit RemoteQueryExecutorReadContext(MultiplexedConnections & connections_) : connections(connections_)
-    {
-        auto & socket = connections.getSocket();
-        socket_fd = socket.impl()->sockfd();
-        receive_timeout = socket.impl()->getReceiveTimeout();
-
-        socket_fd = epoll_create(2);
-        if (-1 == socket_fd)
-            throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
-
-        {
-            epoll_event socket_event;
-            socket_event.events = EPOLLIN | EPOLLPRI;
-            socket_event.data.fd = socket_fd;
-
-            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
-                throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
-        }
-
-        {
-            epoll_event timer_event;
-            timer_event.events = EPOLLIN | EPOLLPRI;
-            timer_event.data.fd = timer.getDescriptor();
-
-            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, timer_event.data.fd, &timer_event))
-                throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
-        }
-
-        fiber = boost::context::fiber(std::allocator_arg_t(), stack, Routine{connections, *this});
-    }
-
-    static void initialize(std::unique_ptr<Self> & read_context, MultiplexedConnections & connections)
-    {
-        try
-        {
-            read_context = std::make_unique<Self>(connections);
-        }
-        catch (DB::Exception & e)
-        {
-            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
-        }
-    }
-
-    void checkTimeout() const
-    {
-        try
-        {
-            checkTimeoutImpl();
-        }
-        catch (DB::Exception & e)
-        {
-            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
-        }
-    }
-
-    void checkTimeoutImpl() const
-    {
-        epoll_event events[2];
-
-        /// Wait for epoll_fd will not block if it was polled externally.
-        int num_events = epoll_wait(epoll_fd, events, 2, 0);
-        if (num_events == -1)
-            throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
-
-        bool is_socket_ready = false;
-        bool has_timer_alarm = false;
-
-        for (int i = 0; i < num_events; ++i)
-        {
-            if (events[i].data.fd == socket_fd)
-                is_socket_ready = true;
-            if (events[i].data.fd == timer.getDescriptor())
-                has_timer_alarm = true;
-        }
-
-        if (has_timer_alarm && !is_socket_ready)
-        {
-            /// Socket receive timeout. Drain it in case or error, or it may be hide by timeout exception.
-            timer.drain();
-            throw NetException("Timeout exceeded", ErrorCodes::SOCKET_TIMEOUT);
-        }
-    }
-
-    void setTimer() const
-    {
-        /// Did not get packet yet. Init timeout for the next async reading.
-        timer.reset();
-
-        if (receive_timeout.totalMicroseconds())
-            timer.setRelative(receive_timeout);
-    }
-
-    void resumeRoutine()
-    {
-        fiber = std::move(fiber).resume();
-
-        if (exception)
-            std::rethrow_exception(std::move(exception));
-    }
-
-    ~RemoteQueryExecutorReadContext()
-    {
-        /// socket_fd is closed by Poco::Net::Socket
-        /// timer_fd is closed by TimerDescriptor
-        close(epoll_fd);
-    }
-
-    struct Routine
-    {
-        MultiplexedConnections & connections;
-        Self & read_context;
-
-        boost::context::fiber operator()(boost::context::fiber && sink) const
-        {
-            try
-            {
-                while (true)
-                {
-                    read_context.is_read_in_progress = true;
-                    connections.setFiber(&sink);
-
-                    read_context.packet = connections.receivePacket();
-
-                    read_context.is_read_in_progress = false;
-                    connections.setFiber(nullptr);
-
-                    sink = std::move(sink).resume();
-                }
-            }
-            catch (...)
-            {
-                read_context.exception = std::current_exception();
-            }
-
-            return std::move(sink);
-        };
-    };
-};
-
 RemoteQueryExecutor::RemoteQueryExecutor(
     Connection & connection,
     const String & query_, const Block & header_, const Context & context_, const Settings * settings,
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
new file mode 100644
index 00000000000..7a78e460803
--- /dev/null
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -0,0 +1,172 @@
+#pragma once
+#include <sys/epoll.h>
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int CANNOT_READ_FROM_SOCKET;
+    extern const int CANNOT_OPEN_FILE;
+    extern const int SOCKET_TIMEOUT;
+}
+
+class RemoteQueryExecutorReadContext
+{
+public:
+    using Self = RemoteQueryExecutorReadContext;
+
+    bool is_read_in_progress = false;
+    Packet packet;
+
+    std::exception_ptr exception;
+    FiberStack<> stack;
+    boost::context::fiber fiber;
+
+    Poco::Timespan receive_timeout;
+    MultiplexedConnections & connections;
+
+    TimerDescriptor timer{CLOCK_MONOTONIC, 0};
+    int socket_fd;
+    int epoll_fd;
+
+    explicit RemoteQueryExecutorReadContext(MultiplexedConnections & connections_) : connections(connections_)
+    {
+        auto & socket = connections.getSocket();
+        socket_fd = socket.impl()->sockfd();
+        receive_timeout = socket.impl()->getReceiveTimeout();
+
+        socket_fd = epoll_create(2);
+        if (-1 == socket_fd)
+            throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
+
+        {
+            epoll_event socket_event;
+            socket_event.events = EPOLLIN | EPOLLPRI;
+            socket_event.data.fd = socket_fd;
+
+            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
+                throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+        }
+
+        {
+            epoll_event timer_event;
+            timer_event.events = EPOLLIN | EPOLLPRI;
+            timer_event.data.fd = timer.getDescriptor();
+
+            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, timer_event.data.fd, &timer_event))
+                throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+        }
+
+        fiber = boost::context::fiber(std::allocator_arg_t(), stack, Routine{connections, *this});
+    }
+
+    static void initialize(std::unique_ptr<Self> & read_context, MultiplexedConnections & connections)
+    {
+        try
+        {
+            read_context = std::make_unique<Self>(connections);
+        }
+        catch (DB::Exception & e)
+        {
+            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+        }
+    }
+
+    void checkTimeout() const
+    {
+        try
+        {
+            checkTimeoutImpl();
+        }
+        catch (DB::Exception & e)
+        {
+            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+        }
+    }
+
+    void checkTimeoutImpl() const
+    {
+        epoll_event events[2];
+
+        /// Wait for epoll_fd will not block if it was polled externally.
+        int num_events = epoll_wait(epoll_fd, events, 2, 0);
+        if (num_events == -1)
+            throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+
+        bool is_socket_ready = false;
+        bool has_timer_alarm = false;
+
+        for (int i = 0; i < num_events; ++i)
+        {
+            if (events[i].data.fd == socket_fd)
+                is_socket_ready = true;
+            if (events[i].data.fd == timer.getDescriptor())
+                has_timer_alarm = true;
+        }
+
+        if (has_timer_alarm && !is_socket_ready)
+        {
+            /// Socket receive timeout. Drain it in case or error, or it may be hide by timeout exception.
+            timer.drain();
+            throw NetException("Timeout exceeded", ErrorCodes::SOCKET_TIMEOUT);
+        }
+    }
+
+    void setTimer() const
+    {
+        /// Did not get packet yet. Init timeout for the next async reading.
+        timer.reset();
+
+        if (receive_timeout.totalMicroseconds())
+            timer.setRelative(receive_timeout);
+    }
+
+    void resumeRoutine()
+    {
+        fiber = std::move(fiber).resume();
+
+        if (exception)
+            std::rethrow_exception(std::move(exception));
+    }
+
+    ~RemoteQueryExecutorReadContext()
+    {
+        /// socket_fd is closed by Poco::Net::Socket
+        /// timer_fd is closed by TimerDescriptor
+        close(epoll_fd);
+    }
+
+    struct Routine
+    {
+        MultiplexedConnections & connections;
+        Self & read_context;
+
+        boost::context::fiber operator()(boost::context::fiber && sink) const
+        {
+            try
+            {
+                while (true)
+                {
+                    read_context.is_read_in_progress = true;
+                    connections.setFiber(&sink);
+
+                    read_context.packet = connections.receivePacket();
+
+                    read_context.is_read_in_progress = false;
+                    connections.setFiber(nullptr);
+
+                    sink = std::move(sink).resume();
+                }
+            }
+            catch (...)
+            {
+                read_context.exception = std::current_exception();
+            }
+
+            return std::move(sink);
+        }
+    };
+};
+
+}
diff --git a/src/Processors/IProcessor.cpp b/src/Processors/IProcessor.cpp
index 36beeea8476..a2533ee4c8c 100644
--- a/src/Processors/IProcessor.cpp
+++ b/src/Processors/IProcessor.cpp
@@ -33,8 +33,6 @@ std::string IProcessor::statusToName(Status status)
             return "Ready";
         case Status::Async:
             return "Async";
-        case Status::Wait:
-            return "Wait";
         case Status::ExpandPipeline:
             return "ExpandPipeline";
     }
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index a277a794b36..c4f12b2e23d 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -1,5 +1,6 @@
 #include <Processors/Sources/RemoteSource.h>
 #include <DataStreams/RemoteQueryExecutor.h>
+#include <DataStreams/RemoteQueryExecutorReadContext.h>
 #include <Processors/Transforms/AggregatingTransform.h>
 #include <DataTypes/DataTypeAggregateFunction.h>
 
diff --git a/src/Storages/Kafka/KafkaBlockInputStream.cpp b/src/Storages/Kafka/KafkaBlockInputStream.cpp
index 6a137bd4b8b..bf985902b4d 100644
--- a/src/Storages/Kafka/KafkaBlockInputStream.cpp
+++ b/src/Storages/Kafka/KafkaBlockInputStream.cpp
@@ -125,7 +125,6 @@ Block KafkaBlockInputStream::readImpl()
                 }
                 case IProcessor::Status::NeedData:
                 case IProcessor::Status::Async:
-                case IProcessor::Status::Wait:
                 case IProcessor::Status::ExpandPipeline:
                     throw Exception("Source processor returned status " + IProcessor::statusToName(status), ErrorCodes::LOGICAL_ERROR);
             }

From f9b45b83a971da4d15aee9fcdc5a4079b856712b Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 4 Dec 2020 15:43:05 +0300
Subject: [PATCH 048/504] Try fix CMakeLists.txt

---
 src/CMakeLists.txt                    | 2 ++
 src/Interpreters/tests/CMakeLists.txt | 2 +-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 0863675f876..1cd05790179 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -422,6 +422,8 @@ if (USE_ROCKSDB)
     dbms_target_include_directories(SYSTEM BEFORE PUBLIC ${ROCKSDB_INCLUDE_DIR})
 endif()
 
+dbms_target_link_libraries(PRIVATE _boost_context)
+
 if (ENABLE_TESTS AND USE_GTEST)
     macro (grep_gtest_sources BASE_DIR DST_VAR)
         # Cold match files that are not in tests/ directories
diff --git a/src/Interpreters/tests/CMakeLists.txt b/src/Interpreters/tests/CMakeLists.txt
index f8fa33d8d2e..9bdedc7b76b 100644
--- a/src/Interpreters/tests/CMakeLists.txt
+++ b/src/Interpreters/tests/CMakeLists.txt
@@ -30,7 +30,7 @@ add_executable (string_hash_map_aggregation string_hash_map.cpp)
 target_link_libraries (string_hash_map_aggregation PRIVATE dbms)
 
 add_executable (context context.cpp)
-target_link_libraries (context PRIVATE dbms _boost_context)
+target_link_libraries (context PRIVATE dbms)
 
 add_executable (two_level_hash_map two_level_hash_map.cpp)
 target_include_directories (two_level_hash_map SYSTEM BEFORE PRIVATE ${SPARSEHASH_INCLUDE_DIR})

From 156f44808fd65ed0dde5174f6f309528c63126f6 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 4 Dec 2020 16:35:24 +0300
Subject: [PATCH 049/504] Fixing crash.

---
 src/Client/MultiplexedConnections.cpp         |  6 ++
 src/Client/MultiplexedConnections.h           |  4 +-
 src/Common/TimerDescriptor.cpp                | 28 +++---
 src/DataStreams/RemoteQueryExecutor.cpp       | 13 +--
 src/DataStreams/RemoteQueryExecutor.h         |  2 +-
 .../RemoteQueryExecutorReadContext.h          | 90 +++++++++++++------
 src/IO/ReadBufferFromPocoSocket.h             |  2 +-
 src/Processors/Executors/AsyncTaskQueue.cpp   | 12 ++-
 src/Processors/Executors/AsyncTaskQueue.h     |  3 +
 src/Processors/Executors/PipelineExecutor.cpp |  5 +-
 src/Processors/Sources/RemoteSource.cpp       |  2 +-
 11 files changed, 112 insertions(+), 55 deletions(-)

diff --git a/src/Client/MultiplexedConnections.cpp b/src/Client/MultiplexedConnections.cpp
index a99b0f9d7cc..3f0c5a29946 100644
--- a/src/Client/MultiplexedConnections.cpp
+++ b/src/Client/MultiplexedConnections.cpp
@@ -249,8 +249,14 @@ Packet MultiplexedConnections::receivePacketUnlocked()
     if (current_connection == nullptr)
         throw Exception("Logical error: no available replica", ErrorCodes::NO_AVAILABLE_REPLICA);
 
+    if (fiber)
+        current_connection->setFiber(fiber);
+
     Packet packet = current_connection->receivePacket();
 
+    fiber = nullptr;
+    current_connection->setFiber(fiber);
+
     switch (packet.type)
     {
         case Protocol::Server::Data:
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index 9573dfe4c93..84283c063e9 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -67,7 +67,7 @@ public:
     /// Without locking, because sendCancel() does not change the state of the replicas.
     bool hasActiveConnections() const { return active_connection_count > 0; }
 
-    void setFiber(boost::context::fiber * fiber) { current_connection->setFiber(fiber); }
+    void setFiber(boost::context::fiber * fiber_) { fiber = fiber_; }
     Poco::Net::Socket & getSocket() { return current_connection->getSocket(); }
 
 private:
@@ -108,6 +108,8 @@ private:
     /// A mutex for the sendCancel function to execute safely
     /// in separate thread.
     mutable std::mutex cancel_mutex;
+
+    boost::context::fiber * fiber = nullptr;
 };
 
 }
diff --git a/src/Common/TimerDescriptor.cpp b/src/Common/TimerDescriptor.cpp
index 406ce53ca49..a77a9d24ed0 100644
--- a/src/Common/TimerDescriptor.cpp
+++ b/src/Common/TimerDescriptor.cpp
@@ -2,7 +2,7 @@
 #include <Common/Exception.h>
 
 #include <sys/timerfd.h>
-#include <sys/ioctl.h>
+#include <fcntl.h>
 #include <unistd.h>
 
 namespace DB
@@ -21,6 +21,9 @@ TimerDescriptor::TimerDescriptor(int clockid, int flags)
     timer_fd = timerfd_create(clockid, flags);
     if (timer_fd == -1)
         throw Exception(ErrorCodes::CANNOT_CREATE_TIMER, "Cannot create timer_fd descriptor");
+
+    if (-1 == fcntl(timer_fd, F_SETFL, O_NONBLOCK))
+        throwFromErrno("Cannot set O_NONBLOCK for timer_fd", ErrorCodes::CANNOT_FCNTL);
 }
 
 TimerDescriptor::~TimerDescriptor()
@@ -47,22 +50,19 @@ void TimerDescriptor::reset() const
 
 void TimerDescriptor::drain() const
 {
-    int num_bytes_to_read = 0;
-    if (-1 == ioctl(timer_fd, FIONREAD, &num_bytes_to_read))
-        throwFromErrno("Cannot get available bytes form timer_fd", ErrorCodes::CANNOT_FCNTL);
-
-    if (num_bytes_to_read)
+    /// It is expected that socket returns 8 bytes when readable.
+    /// Read in loop anyway cause signal may interrupt read call.
+    uint64_t buf;
+    while (true)
     {
-        /// It is expected that socket returns 8 bytes when readable.
-        /// Read in loop anyway cause signal may interrupt read call.
-        uint64_t buf;
-        while (num_bytes_to_read)
+        ssize_t res = ::read(timer_fd, &buf, sizeof(buf));
+        if (res < 0)
         {
-            ssize_t res = ::read(timer_fd, &buf, sizeof(buf));
-            if (res < 0 && errno != EINTR)
+            if (errno == EAGAIN)
+                break;
+
+            if (errno != EINTR)
                 throwFromErrno("Cannot drain timer_fd", ErrorCodes::CANNOT_READ_FROM_SOCKET);
-            else
-                num_bytes_to_read -= res;
         }
     }
 }
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 5cf225b8c0f..a473d66a85c 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -219,14 +219,12 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
     }
 
     if (!read_context)
-        ReadContext::initialize(read_context, *multiplexed_connections);
-
-    if (read_context->is_read_in_progress)
-        read_context->checkTimeout();
+        read_context = std::make_unique<ReadContext>(*multiplexed_connections);
 
     do
     {
-        read_context->resumeRoutine();
+        if (!read_context->resumeRoutine())
+            return Block();
 
         if (read_context->is_read_in_progress)
         {
@@ -352,7 +350,7 @@ void RemoteQueryExecutor::finish()
     }
 }
 
-void RemoteQueryExecutor::cancel()
+void RemoteQueryExecutor::cancel(std::unique_ptr<ReadContext> * read_context)
 {
     {
         std::lock_guard lock(external_tables_mutex);
@@ -366,6 +364,9 @@ void RemoteQueryExecutor::cancel()
     if (!isQueryPending() || hasThrownException())
         return;
 
+    if (read_context && *read_context)
+        (*read_context)->cancel();
+
     tryCancel("Cancelling query");
 }
 
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 2ca13ffd9b7..f0ba74a8b65 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -69,7 +69,7 @@ public:
 
     /// Cancel query execution. Sends Cancel packet and ignore others.
     /// This method may be called from separate thread.
-    void cancel();
+    void cancel(std::unique_ptr<ReadContext> * read_context = nullptr);
 
     /// Get totals and extremes if any.
     Block getTotals() { return std::move(totals); }
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 7a78e460803..17a0f8c5057 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -21,34 +21,22 @@ public:
 
     std::exception_ptr exception;
     FiberStack<> stack;
+    std::mutex fiber_mutex;
     boost::context::fiber fiber;
 
     Poco::Timespan receive_timeout;
     MultiplexedConnections & connections;
 
     TimerDescriptor timer{CLOCK_MONOTONIC, 0};
-    int socket_fd;
+    int socket_fd = -1;
     int epoll_fd;
 
     explicit RemoteQueryExecutorReadContext(MultiplexedConnections & connections_) : connections(connections_)
     {
-        auto & socket = connections.getSocket();
-        socket_fd = socket.impl()->sockfd();
-        receive_timeout = socket.impl()->getReceiveTimeout();
-
-        socket_fd = epoll_create(2);
-        if (-1 == socket_fd)
+        epoll_fd = epoll_create(2);
+        if (-1 == epoll_fd)
             throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
 
-        {
-            epoll_event socket_event;
-            socket_event.events = EPOLLIN | EPOLLPRI;
-            socket_event.data.fd = socket_fd;
-
-            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
-                throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
-        }
-
         {
             epoll_event timer_event;
             timer_event.events = EPOLLIN | EPOLLPRI;
@@ -61,16 +49,28 @@ public:
         fiber = boost::context::fiber(std::allocator_arg_t(), stack, Routine{connections, *this});
     }
 
-    static void initialize(std::unique_ptr<Self> & read_context, MultiplexedConnections & connections)
+    void setSocket(Poco::Net::Socket & socket)
     {
-        try
+        int fd = socket.impl()->sockfd();
+        if (fd == socket_fd)
+            return;
+
+        epoll_event socket_event;
+        socket_event.events = EPOLLIN | EPOLLPRI;
+        socket_event.data.fd = fd;
+
+        if (socket_fd != -1)
         {
-            read_context = std::make_unique<Self>(connections);
-        }
-        catch (DB::Exception & e)
-        {
-            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, socket_fd, &socket_event))
+                throwFromErrno("Cannot remove socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
         }
+
+        socket_fd = fd;
+
+        if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_fd, &socket_event))
+            throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+
+        receive_timeout = socket.impl()->getReceiveTimeout();
     }
 
     void checkTimeout() const
@@ -82,6 +82,7 @@ public:
         catch (DB::Exception & e)
         {
             e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+            throw;
         }
     }
 
@@ -122,12 +123,43 @@ public:
             timer.setRelative(receive_timeout);
     }
 
-    void resumeRoutine()
+    bool resumeRoutine()
     {
-        fiber = std::move(fiber).resume();
+        if (is_read_in_progress)
+            checkTimeout();
+
+        {
+            std::lock_guard guard(fiber_mutex);
+            if (!fiber)
+                return false;
+
+            fiber = std::move(fiber).resume();
+        }
 
         if (exception)
             std::rethrow_exception(std::move(exception));
+
+        if (is_read_in_progress)
+        {
+            auto & socket = connections.getSocket();
+            try
+            {
+                setSocket(socket);
+            }
+            catch (DB::Exception & e)
+            {
+                e.addMessage(" while reading from socket ({})", socket.peerAddress().toString());
+                throw;
+            }
+        }
+
+        return true;
+    }
+
+    void cancel()
+    {
+        std::lock_guard guard(fiber_mutex);
+        boost::context::fiber to_destroy = std::move(fiber);
     }
 
     ~RemoteQueryExecutorReadContext()
@@ -148,17 +180,19 @@ public:
             {
                 while (true)
                 {
-                    read_context.is_read_in_progress = true;
                     connections.setFiber(&sink);
 
+                    read_context.is_read_in_progress = true;
                     read_context.packet = connections.receivePacket();
-
                     read_context.is_read_in_progress = false;
-                    connections.setFiber(nullptr);
 
                     sink = std::move(sink).resume();
                 }
             }
+            catch (const boost::context::detail::forced_unwind &)
+            {
+                throw;
+            }
             catch (...)
             {
                 read_context.exception = std::current_exception();
diff --git a/src/IO/ReadBufferFromPocoSocket.h b/src/IO/ReadBufferFromPocoSocket.h
index 0c744ae8fe3..d5d1c071210 100644
--- a/src/IO/ReadBufferFromPocoSocket.h
+++ b/src/IO/ReadBufferFromPocoSocket.h
@@ -35,7 +35,7 @@ public:
     Poco::Net::Socket & getSocket() { return socket; }
 
 private:
-    boost::context::fiber * fiber;
+    boost::context::fiber * fiber = nullptr;
 };
 
 }
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index d0a63a87165..020a79ad809 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -29,10 +29,9 @@ void AsyncTaskQueue::addTask(void * data, int fd)
 {
     epoll_event socket_event;
     socket_event.events = EPOLLIN | EPOLLPRI;
-    socket_event.data.fd = fd;
     socket_event.data.ptr = data;
 
-    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, socket_event.data.fd, &socket_event))
+    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fd, &socket_event))
         throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
 
     ++num_tasks;
@@ -40,6 +39,13 @@ void AsyncTaskQueue::addTask(void * data, int fd)
         condvar.notify_one();
 }
 
+void AsyncTaskQueue::removeTask(int fd) const
+{
+    epoll_event event;
+    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, fd, &event))
+        throwFromErrno("Cannot remove socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+}
+
 void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
 {
     condvar.wait(lock, [&] { return !empty() || is_finished; });
@@ -57,6 +63,7 @@ void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
         num_events = epoll_wait(epoll_fd, &event, 1, 0);
         if (num_events == -1)
             throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+
     }
 
     lock.lock();
@@ -67,6 +74,7 @@ void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
 void AsyncTaskQueue::finish()
 {
     is_finished = true;
+    num_tasks = 0;
     condvar.notify_one();
 }
 
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index 2f3cdf8ab82..52aeeed3446 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -23,6 +23,9 @@ public:
     /// Add new task to queue.
     void addTask(void * data, int fd);
 
+    /// Remove task.
+    void removeTask(int fd) const;
+
     /// Wait for any descriptor. If no descriptors in queue, blocks.
     /// Returns ptr which was inserted into queue or nullptr if finished was called.
     /// Lock is used to wait on condvar.
diff --git a/src/Processors/Executors/PipelineExecutor.cpp b/src/Processors/Executors/PipelineExecutor.cpp
index dce21067557..3ab1ddee9b4 100644
--- a/src/Processors/Executors/PipelineExecutor.cpp
+++ b/src/Processors/Executors/PipelineExecutor.cpp
@@ -532,6 +532,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                         throw Exception("Empty task was returned from async task queue", ErrorCodes::LOGICAL_ERROR);
 
                     node = static_cast<ExecutingGraph::Node *>(res);
+                    async_task_queue.removeTask(node->processor->schedule());
                     continue;
                 }
 
@@ -773,7 +774,9 @@ void PipelineExecutor::executeImpl(size_t num_threads)
             size_t next_thread = 0;
             while (void * task = async_task_queue.wait(lock))
             {
-                task_queue.push(static_cast<ExecutingGraph::Node *>(task), next_thread);
+                auto * node = static_cast<ExecutingGraph::Node *>(task);
+                async_task_queue.removeTask(node->processor->schedule());
+                task_queue.push(node, next_thread);
 
                 ++next_thread;
                 if (next_thread >= num_threads)
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index c4f12b2e23d..0c22e8af0f5 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -91,7 +91,7 @@ std::optional<Chunk> RemoteSource::tryGenerate()
 void RemoteSource::onCancel()
 {
     was_query_canceled = true;
-    query_executor->cancel();
+    query_executor->cancel(&read_context);
 }
 
 

From 6a8384e6a483a6b51db39b5551a712a30d222eed Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 7 Dec 2020 14:16:23 +0300
Subject: [PATCH 050/504] Use same thread for async task continuation.

---
 src/Processors/Executors/AsyncTaskQueue.cpp   | 35 +++++++-------
 src/Processors/Executors/AsyncTaskQueue.h     | 27 +++++++----
 src/Processors/Executors/PipelineExecutor.cpp | 47 ++++++++++---------
 src/Processors/Executors/PipelineExecutor.h   |  3 ++
 4 files changed, 66 insertions(+), 46 deletions(-)

diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index 020a79ad809..9d73a669be6 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -25,33 +25,28 @@ AsyncTaskQueue::~AsyncTaskQueue()
     close(epoll_fd);
 }
 
-void AsyncTaskQueue::addTask(void * data, int fd)
+void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
 {
+    auto it = tasks.insert(tasks.end(), TaskData{thread_number, data, fd, {}});
+    it->self = it;
+
     epoll_event socket_event;
     socket_event.events = EPOLLIN | EPOLLPRI;
-    socket_event.data.ptr = data;
+    socket_event.data.ptr = &(*it);
 
     if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fd, &socket_event))
         throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
 
-    ++num_tasks;
-    if (num_tasks == 1)
+    if (size() == 1)
         condvar.notify_one();
 }
 
-void AsyncTaskQueue::removeTask(int fd) const
-{
-    epoll_event event;
-    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, fd, &event))
-        throwFromErrno("Cannot remove socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
-}
-
-void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
+AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
 {
     condvar.wait(lock, [&] { return !empty() || is_finished; });
 
     if (is_finished)
-        return nullptr;
+        return {};
 
     lock.unlock();
 
@@ -63,18 +58,24 @@ void * AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
         num_events = epoll_wait(epoll_fd, &event, 1, 0);
         if (num_events == -1)
             throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
-
     }
 
     lock.lock();
-    --num_tasks;
-    return event.data.ptr;
+
+    auto it = static_cast<TaskData *>(event.data.ptr)->self;
+
+    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, it->fd, &event))
+        throwFromErrno("Cannot remove socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+
+    auto res = *it;
+    tasks.erase(it);
+    return res;
 }
 
 void AsyncTaskQueue::finish()
 {
     is_finished = true;
-    num_tasks = 0;
+    tasks.clear();
     condvar.notify_one();
 }
 
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index 52aeeed3446..cd1a1ca0d0d 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -1,35 +1,46 @@
 #include <cstddef>
 #include <mutex>
 #include <atomic>
+#include <list>
 
 namespace DB
 {
 
 class AsyncTaskQueue
 {
+public:
+    struct TaskData
+    {
+        size_t thread_num;
+
+        void * data = nullptr;
+        int fd = -1;
+
+        std::list<TaskData>::iterator self;
+
+        explicit operator bool() const { return data; }
+    };
+
 private:
     int epoll_fd;
-    size_t num_tasks;
     std::atomic_bool is_finished = false;
     std::condition_variable condvar;
+    std::list<TaskData> tasks;
 
 public:
     AsyncTaskQueue();
     ~AsyncTaskQueue();
 
-    size_t size() const { return num_tasks; }
-    bool empty() const { return num_tasks == 0; }
+    size_t size() const { return tasks.size(); }
+    bool empty() const { return tasks.empty(); }
 
     /// Add new task to queue.
-    void addTask(void * data, int fd);
-
-    /// Remove task.
-    void removeTask(int fd) const;
+    void addTask(size_t thread_number, void * data, int fd);
 
     /// Wait for any descriptor. If no descriptors in queue, blocks.
     /// Returns ptr which was inserted into queue or nullptr if finished was called.
     /// Lock is used to wait on condvar.
-    void * wait(std::unique_lock<std::mutex> & lock);
+    TaskData wait(std::unique_lock<std::mutex> & lock);
 
     /// Interrupt waiting.
     void finish();
diff --git a/src/Processors/Executors/PipelineExecutor.cpp b/src/Processors/Executors/PipelineExecutor.cpp
index 3ab1ddee9b4..c717265248c 100644
--- a/src/Processors/Executors/PipelineExecutor.cpp
+++ b/src/Processors/Executors/PipelineExecutor.cpp
@@ -497,11 +497,18 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
             {
                 std::unique_lock lock(task_queue_mutex);
 
-                if (!task_queue.empty())
+                if (!context->async_tasks.empty())
                 {
+                    node = context->async_tasks.front();
+                    context->async_tasks.pop();
+                    --num_waiting_async_tasks;
+                }
+                else if (!task_queue.empty())
                     node = task_queue.pop(thread_num);
 
-                    if (!task_queue.empty() && !threads_queue.empty() /*&& task_queue.quota() > threads_queue.size()*/)
+                if (node)
+                {
+                    if (!task_queue.empty() && !threads_queue.empty())
                     {
                         auto thread_to_wake = task_queue.getAnyThreadWithTasks(thread_num + 1 == num_threads ? 0 : (thread_num + 1));
 
@@ -517,7 +524,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                     break;
                 }
 
-                if (threads_queue.size() + 1 == num_threads && async_task_queue.empty())
+                if (threads_queue.size() + 1 == num_threads && async_task_queue.empty() && num_waiting_async_tasks == 0)
                 {
                     lock.unlock();
                     finish();
@@ -527,13 +534,12 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                 if (num_threads == 1)
                 {
                     /// If we execute in single thread, wait for async tasks here.
-                    void * res = async_task_queue.wait(lock);
+                    auto res = async_task_queue.wait(lock);
                     if (!res)
                         throw Exception("Empty task was returned from async task queue", ErrorCodes::LOGICAL_ERROR);
 
-                    node = static_cast<ExecutingGraph::Node *>(res);
-                    async_task_queue.removeTask(node->processor->schedule());
-                    continue;
+                    node = static_cast<ExecutingGraph::Node *>(res.data);
+                    break;
                 }
 
                 threads_queue.push(thread_num);
@@ -615,7 +621,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
 
                     while (!async_queue.empty() && !finished)
                     {
-                        async_task_queue.addTask(async_queue.front(), async_queue.front()->processor->schedule());
+                        async_task_queue.addTask(thread_num, async_queue.front(), async_queue.front()->processor->schedule());
                         async_queue.pop();
                     }
 
@@ -625,7 +631,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                         queue.pop();
                     }
 
-                    if (!threads_queue.empty() && !finished)
+                    if (!threads_queue.empty() && !task_queue.empty() && !finished)
                     {
                         auto thread_to_wake = task_queue.getAnyThreadWithTasks(thread_num + 1 == num_threads ? 0 : (thread_num + 1));
 
@@ -704,10 +710,8 @@ void PipelineExecutor::initializeExecution(size_t num_threads)
             }
 
             while (!async_queue.empty())
-            {
-                async_task_queue.addTask(async_queue.front(), async_queue.front()->processor->schedule());
-                async_queue.pop();
-            }
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "Async is only possible after work() call. Processor {}",
+                                async_queue.front()->processor->getName());
         }
     }
 }
@@ -771,16 +775,17 @@ void PipelineExecutor::executeImpl(size_t num_threads)
         {
             /// Wait for async tasks.
             std::unique_lock lock(task_queue_mutex);
-            size_t next_thread = 0;
-            while (void * task = async_task_queue.wait(lock))
+            while (auto task = async_task_queue.wait(lock))
             {
-                auto * node = static_cast<ExecutingGraph::Node *>(task);
-                async_task_queue.removeTask(node->processor->schedule());
-                task_queue.push(node, next_thread);
+                auto * node = static_cast<ExecutingGraph::Node *>(task.data);
+                executor_contexts[task.thread_num]->async_tasks.push(node);
+                ++num_waiting_async_tasks;
 
-                ++next_thread;
-                if (next_thread >= num_threads)
-                    next_thread = 0;
+                if (threads_queue.has(task.thread_num))
+                {
+                    threads_queue.pop(task.thread_num);
+                    wakeUpExecutor(task.thread_num);
+                }
             }
         }
 
diff --git a/src/Processors/Executors/PipelineExecutor.h b/src/Processors/Executors/PipelineExecutor.h
index 34339a34608..1ddb6cbe8b9 100644
--- a/src/Processors/Executors/PipelineExecutor.h
+++ b/src/Processors/Executors/PipelineExecutor.h
@@ -62,6 +62,7 @@ private:
     /// If multiple threads are using, main thread will wait for async tasks.
     /// For single thread, will wait for async tasks only when task_queue is empty.
     AsyncTaskQueue async_task_queue;
+    size_t num_waiting_async_tasks = 0;
 
     ThreadsQueue threads_queue;
     std::mutex task_queue_mutex;
@@ -96,6 +97,8 @@ private:
         /// This can be solved by using atomic shard ptr.
         std::list<ExpandPipelineTask> task_list;
 
+        std::queue<ExecutingGraph::Node *> async_tasks;
+
         std::condition_variable condvar;
         std::mutex mutex;
         bool wake_flag = false;

From effc94daafde521b0ef75b2145111c00db82fc77 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 7 Dec 2020 16:35:30 +0300
Subject: [PATCH 051/504] Added perftest.

---
 tests/performance/async_remote_read.xml | 11 +++++++++++
 1 file changed, 11 insertions(+)
 create mode 100644 tests/performance/async_remote_read.xml

diff --git a/tests/performance/async_remote_read.xml b/tests/performance/async_remote_read.xml
new file mode 100644
index 00000000000..7f0ee6473ab
--- /dev/null
+++ b/tests/performance/async_remote_read.xml
@@ -0,0 +1,11 @@
+<test>
+    <query>
+        SELECT sum(x)
+        FROM
+        (
+            SELECT sipHash64(sipHash64(sipHash64(number))) AS x
+            FROM remote('127.0.0.{{2,3,4,5}}', numbers(10000000))
+        )
+        SETTINGS max_threads = 2, max_distributed_connections = 2
+    </query>
+</test>

From e8667bad45bc9e30986a2a2804fa823a6c6ef7f2 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 7 Dec 2020 16:47:11 +0300
Subject: [PATCH 052/504] Fix build and tests.

---
 contrib/boost-cmake/CMakeLists.txt          |  4 +--
 src/DataStreams/RemoteQueryExecutor.cpp     |  5 +++-
 src/DataStreams/RemoteQueryExecutor.h       |  2 +-
 src/Processors/Executors/AsyncTaskQueue.cpp | 32 +++++++++++++++++----
 src/Processors/Executors/AsyncTaskQueue.h   |  2 +-
 src/Processors/Sources/RemoteSource.cpp     |  8 ++++--
 6 files changed, 40 insertions(+), 13 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 5bf09ca29bd..714dfd4f011 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -224,10 +224,10 @@ if (NOT EXTERNAL_BOOST_FOUND)
 #        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_masm.asm
         ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_elf_gas.S
 #        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_macho_gas.S
-        ${LIBRARY_DIR}/libs/context/src/continuation.cpp
+#        ${LIBRARY_DIR}/libs/context/src/continuation.cpp
         ${LIBRARY_DIR}/libs/context/src/dummy.cpp
         ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
-        ${LIBRARY_DIR}/libs/context/src/fiber.cpp
+#        ${LIBRARY_DIR}/libs/context/src/fiber.cpp
         ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
 #        ${LIBRARY_DIR}/libs/context/src/untested.cpp
 #        ${LIBRARY_DIR}/libs/context/src/windows/stack_traits.cpp
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index a473d66a85c..a315f621164 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -304,7 +304,7 @@ std::optional<Block> RemoteQueryExecutor::processPacket(Packet packet)
     return {};
 }
 
-void RemoteQueryExecutor::finish()
+void RemoteQueryExecutor::finish(std::unique_ptr<ReadContext> * read_context)
 {
     /** If one of:
       * - nothing started to do;
@@ -316,6 +316,9 @@ void RemoteQueryExecutor::finish()
     if (!isQueryPending() || hasThrownException())
         return;
 
+    if (read_context && *read_context)
+        (*read_context)->cancel();
+
     /** If you have not read all the data yet, but they are no longer needed.
       * This may be due to the fact that the data is sufficient (for example, when using LIMIT).
       */
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index f0ba74a8b65..bd49e0d66b9 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -65,7 +65,7 @@ public:
 
     /// Receive all remain packets and finish query.
     /// It should be cancelled after read returned empty block.
-    void finish();
+    void finish(std::unique_ptr<ReadContext> * read_context = nullptr);
 
     /// Cancel query execution. Sends Cancel packet and ignore others.
     /// This method may be called from separate thread.
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index 9d73a669be6..f0fd2f33dcb 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -2,6 +2,7 @@
 #include <Common/Exception.h>
 #include <sys/epoll.h>
 #include <unistd.h>
+#include <fcntl.h>
 
 namespace DB
 {
@@ -18,11 +19,23 @@ AsyncTaskQueue::AsyncTaskQueue()
     epoll_fd = epoll_create(1);
     if (-1 == epoll_fd)
         throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
+
+    if (-1 == pipe2(pipe_fd, O_NONBLOCK))
+        throwFromErrno("Cannot create pipe", ErrorCodes::CANNOT_OPEN_FILE);
+
+    epoll_event socket_event;
+    socket_event.events = EPOLLIN | EPOLLPRI;
+    socket_event.data.ptr = pipe_fd;
+
+    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, pipe_fd[0], &socket_event))
+        throwFromErrno("Cannot add pipe descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
 }
 
 AsyncTaskQueue::~AsyncTaskQueue()
 {
     close(epoll_fd);
+    close(pipe_fd[0]);
+    close(pipe_fd[1]);
 }
 
 void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
@@ -36,15 +49,10 @@ void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
 
     if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fd, &socket_event))
         throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
-
-    if (size() == 1)
-        condvar.notify_one();
 }
 
 AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
 {
-    condvar.wait(lock, [&] { return !empty() || is_finished; });
-
     if (is_finished)
         return {};
 
@@ -62,6 +70,9 @@ AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & loc
 
     lock.lock();
 
+    if (event.data.ptr == pipe_fd)
+        return {};
+
     auto it = static_cast<TaskData *>(event.data.ptr)->self;
 
     if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, it->fd, &event))
@@ -76,7 +87,16 @@ void AsyncTaskQueue::finish()
 {
     is_finished = true;
     tasks.clear();
-    condvar.notify_one();
+
+    uint64_t buf = 0;
+    while (-1 == write(pipe_fd[1], &buf, sizeof(buf)))
+    {
+        if (errno == EAGAIN || errno == EWOULDBLOCK)
+            break;
+
+        if (errno != EINTR)
+            throwFromErrno("Cannot write to pipe", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+    }
 }
 
 }
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index cd1a1ca0d0d..d2a37c94370 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -23,8 +23,8 @@ public:
 
 private:
     int epoll_fd;
+    int pipe_fd[2];
     std::atomic_bool is_finished = false;
-    std::condition_variable condvar;
     std::list<TaskData> tasks;
 
 public:
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index 0c22e8af0f5..3fc1b713c71 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -29,7 +29,10 @@ ISource::Status RemoteSource::prepare()
     /// To avoid resetting the connection (because of "unfinished" query) in the
     /// RemoteQueryExecutor it should be finished explicitly.
     if (status == Status::Finished)
-        query_executor->finish();
+    {
+        query_executor->finish(&read_context);
+        is_async_state = false;
+    }
     return status;
 }
 
@@ -70,7 +73,7 @@ std::optional<Chunk> RemoteSource::tryGenerate()
 
     if (!block)
     {
-        query_executor->finish();
+        query_executor->finish(&read_context);
         return {};
     }
 
@@ -92,6 +95,7 @@ void RemoteSource::onCancel()
 {
     was_query_canceled = true;
     query_executor->cancel(&read_context);
+    is_async_state = false;
 }
 
 

From a1d4d92b540a739728a3da4a64ffea1d1694b920 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 7 Dec 2020 21:19:22 +0300
Subject: [PATCH 053/504] Use poco from master. Fixing tests.

---
 contrib/poco                              | 2 +-
 src/Common/FiberStack.h                   | 2 +-
 src/Common/ya.make                        | 1 +
 src/Processors/Executors/AsyncTaskQueue.h | 1 +
 src/Processors/Sources/RemoteSource.cpp   | 7 +++++++
 src/Processors/ya.make                    | 1 +
 6 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/contrib/poco b/contrib/poco
index 4d06db3947a..b5523bb9b4b 160000
--- a/contrib/poco
+++ b/contrib/poco
@@ -1 +1 @@
-Subproject commit 4d06db3947ac2343133220a5cd5d9b35bc89814c
+Subproject commit b5523bb9b4bc4239640cbfec4d734be8b8585639
diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index f490b509443..28a38cb7a5e 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -38,7 +38,7 @@ public:
     void deallocate(boost::context::stack_context & sctx)
     {
 #if defined(BOOST_USE_VALGRIND)
-        VALGRIND_STACK_DEREGISTER( sctx.valgrind_stack_id);
+        VALGRIND_STACK_DEREGISTER(sctx.valgrind_stack_id);
 #endif
         void * vp = static_cast< char * >(sctx.sp) - sctx.size;
         TAllocator().free(vp, stack_size);
diff --git a/src/Common/ya.make b/src/Common/ya.make
index 71c0edaea95..ba95158f054 100644
--- a/src/Common/ya.make
+++ b/src/Common/ya.make
@@ -74,6 +74,7 @@ SRCS(
     ThreadPool.cpp
     ThreadProfileEvents.cpp
     ThreadStatus.cpp
+    TimerDescriptor.cpp
     TraceCollector.cpp
     UTF8Helpers.cpp
     UnicodeBar.cpp
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index d2a37c94370..068bdcbd09c 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -1,3 +1,4 @@
+#pragma once
 #include <cstddef>
 #include <mutex>
 #include <atomic>
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index 3fc1b713c71..b34a06b3790 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -22,6 +22,13 @@ RemoteSource::~RemoteSource() = default;
 
 ISource::Status RemoteSource::prepare()
 {
+    /// Check if query was cancelled before returning Async status. Otherwise it may lead to infinite loop.
+    if (was_query_canceled)
+    {
+        getPort().finish();
+        return Status::Finished;
+    }
+
     if (is_async_state)
         return Status::Async;
 
diff --git a/src/Processors/ya.make b/src/Processors/ya.make
index eacc0c44d24..bb99cabb1dd 100644
--- a/src/Processors/ya.make
+++ b/src/Processors/ya.make
@@ -14,6 +14,7 @@ SRCS(
     Chunk.cpp
     ConcatProcessor.cpp
     DelayedPortsProcessor.cpp
+    Executors/AsyncTaskQueue.cpp
     Executors/ExecutingGraph.cpp
     Executors/PipelineExecutingBlockInputStream.cpp
     Executors/PipelineExecutor.cpp

From faa5b71efc1f4978488c5d1c9a7114a101a66dc2 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 9 Dec 2020 17:00:50 +0300
Subject: [PATCH 054/504] Update boost submodule

---
 contrib/boost | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/boost b/contrib/boost
index d2da6db25de..a7ceabe4747 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit d2da6db25deff1a927b166ce62cf8967a40273c9
+Subproject commit a7ceabe4747ecc3309dd3dcd9de4b29660dfd298

From 0e44a22f43d0c237d643ac58bbdbce3fff349d91 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 9 Dec 2020 19:17:59 +0300
Subject: [PATCH 055/504] Fixing build.

---
 src/Common/Config/CMakeLists.txt      | 1 +
 src/IO/ReadBufferFromPocoSocket.cpp   | 2 +-
 src/Interpreters/tests/context.cpp    | 4 ++--
 src/Processors/Sources/RemoteSource.h | 2 +-
 4 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/src/Common/Config/CMakeLists.txt b/src/Common/Config/CMakeLists.txt
index a7914fb17ec..b7a7f8ecda4 100644
--- a/src/Common/Config/CMakeLists.txt
+++ b/src/Common/Config/CMakeLists.txt
@@ -14,4 +14,5 @@ target_link_libraries(clickhouse_common_config
         Poco::XML
     PRIVATE
         string_utils
+        _boost_context
 )
diff --git a/src/IO/ReadBufferFromPocoSocket.cpp b/src/IO/ReadBufferFromPocoSocket.cpp
index dd6cd4817df..b0bfadcf57d 100644
--- a/src/IO/ReadBufferFromPocoSocket.cpp
+++ b/src/IO/ReadBufferFromPocoSocket.cpp
@@ -40,7 +40,7 @@ bool ReadBufferFromPocoSocket::nextImpl()
         /// If fiber is specified, and read is blocking, run fiber and try again later.
         /// It is expected that file descriptor may be polled externally.
         /// Note that receive timeout is not checked here. External code should check it while polling.
-        while (bytes_read < 0 && fiber && (errno == POCO_EAGAIN || errno == POCO_EWOULDBLOCK))
+        while (bytes_read < 0 && fiber && (errno == EAGAIN || errno == EWOULDBLOCK))
         {
             //fiber->fd = socket.impl()->sockfd();
             //fiber->timeout = socket.impl()->getReceiveTimeout();
diff --git a/src/Interpreters/tests/context.cpp b/src/Interpreters/tests/context.cpp
index 407a7a9ba3b..655f93a57e1 100644
--- a/src/Interpreters/tests/context.cpp
+++ b/src/Interpreters/tests/context.cpp
@@ -1,8 +1,8 @@
 #include <iostream>
 /// #define BOOST_USE_UCONTEXT
 #include <boost/context/fiber.hpp>
-#include <boost/context/pooled_fixedsize_stack.hpp>
-#include <boost/context/segmented_stack.hpp>
+// #include <boost/context/pooled_fixedsize_stack.hpp>
+// #include <boost/context/segmented_stack.hpp>
 #include <Common/Exception.h>
 #include <Common/FiberStack.h>
 
diff --git a/src/Processors/Sources/RemoteSource.h b/src/Processors/Sources/RemoteSource.h
index 0a04f72d14a..1d9282ff570 100644
--- a/src/Processors/Sources/RemoteSource.h
+++ b/src/Processors/Sources/RemoteSource.h
@@ -50,7 +50,7 @@ private:
 
     bool is_async_state = false;
     std::unique_ptr<RemoteQueryExecutorReadContext> read_context;
-    int fd;
+    int fd = -1;
 };
 
 /// Totals source from RemoteQueryExecutor.

From ac8b896f36960c12501a379a71474f837ebc40a1 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 10 Dec 2020 22:06:52 +0300
Subject: [PATCH 056/504] returning constant column as window function

---
 src/Core/SortDescription.cpp                |   7 +
 src/Core/SortDescription.h                  |   2 +
 src/Interpreters/AggregateDescription.h     |   9 +-
 src/Interpreters/ExpressionAnalyzer.cpp     | 237 +++++++++++++++-----
 src/Interpreters/ExpressionAnalyzer.h       |   2 +-
 src/Interpreters/InterpreterSelectQuery.cpp | 121 ++++++++--
 src/Interpreters/InterpreterSelectQuery.h   |   2 +
 7 files changed, 295 insertions(+), 85 deletions(-)

diff --git a/src/Core/SortDescription.cpp b/src/Core/SortDescription.cpp
index 4a5952c3bc2..cb7378cf096 100644
--- a/src/Core/SortDescription.cpp
+++ b/src/Core/SortDescription.cpp
@@ -37,5 +37,12 @@ void dumpSortDescription(const SortDescription & description, const Block & head
     }
 }
 
+std::string dumpSortDescription(const SortDescription & description)
+{
+    WriteBufferFromOwnString wb;
+    dumpSortDescription(description, Block{}, wb);
+    return wb.str();
+}
+
 }
 
diff --git a/src/Core/SortDescription.h b/src/Core/SortDescription.h
index 79ac7ddf142..1450393ebd8 100644
--- a/src/Core/SortDescription.h
+++ b/src/Core/SortDescription.h
@@ -72,4 +72,6 @@ class Block;
 /// Outputs user-readable description into `out`.
 void dumpSortDescription(const SortDescription & description, const Block & header, WriteBuffer & out);
 
+std::string dumpSortDescription(const SortDescription & description);
+
 }
diff --git a/src/Interpreters/AggregateDescription.h b/src/Interpreters/AggregateDescription.h
index 528b38e86f7..f4ac0cf3898 100644
--- a/src/Interpreters/AggregateDescription.h
+++ b/src/Interpreters/AggregateDescription.h
@@ -31,7 +31,7 @@ struct WindowFunctionDescription
 {
     std::string window_name;
     std::string column_name;
-    const IAST * wrapper_node;
+    const ASTFunction * wrapper_node;
     const ASTFunction * function_node;
     AggregateFunctionPtr aggregate_function;
     Array function_parameters;
@@ -45,9 +45,12 @@ struct WindowDescription
 {
     std::string window_name;
     // Always ASC for now.
-    std::vector<std::string> partition_by;
-    std::vector<std::string> order_by;
+    SortDescription partition_by;
+    SortDescription order_by;
+    SortDescription full_sort_description;
     // No frame info as of yet.
+
+    std::string dump() const;
 };
 
 using WindowFunctionDescriptions = std::vector<WindowFunctionDescription>;
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 60c534c82ad..cbc056bcaee 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -463,6 +463,57 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
     return !aggregates().empty();
 }
 
+// Parses order by & partition by from window() wrapper function.
+// Remove this when we have proper grammar.
+static SortDescription windowArgumentToSortDescription(IAST* ast, const WindowDescription & w)
+{
+    SortDescription result;
+    if (const auto * as_tuple = ast->as<ASTFunction>();
+        as_tuple
+        && as_tuple->name == "tuple"
+        && as_tuple->arguments)
+    {
+        // untuple it
+        for (const auto & element_ast
+            : as_tuple->arguments->children)
+        {
+            const auto * with_alias = dynamic_cast<
+                const ASTWithAlias *>(element_ast.get());
+            if (!with_alias)
+            {
+                throw Exception(ErrorCodes::BAD_ARGUMENTS,
+                    "(1) Expected column in PARTITION BY"
+                    " for window '{}', got '{}'",
+                    w.window_name,
+                    element_ast->formatForErrorMessage());
+            }
+            result.push_back(
+                SortColumnDescription(
+                    with_alias->getColumnName(),
+                    1 /* direction */,
+                    1 /* nulls_direction */));
+        }
+    }
+    else if (const auto * with_alias
+        = dynamic_cast<const ASTWithAlias *>(ast))
+    {
+        result.push_back(
+            SortColumnDescription(
+                with_alias->getColumnName(),
+                1 /* direction */,
+                1 /* nulls_direction */));
+    }
+    else
+    {
+        throw Exception(ErrorCodes::BAD_ARGUMENTS,
+            "(2) Expected tuple or column in PARTITION BY"
+            " for window '{}', got '{}'",
+            w.window_name,
+            ast->formatForErrorMessage());
+    }
+
+    return result;
+}
 
 bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 {
@@ -470,10 +521,6 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
     {
         fmt::print(stderr, "window function ast: {}\n", wrapper_node->dumpTree());
 
-        // Not sure why NoMakeSet, copied from aggregate functions.
-        getRootActionsNoMakeSet(wrapper_node->arguments, true /* no subqueries */,
-            actions);
-
         // FIXME not thread-safe, should use a per-query counter.
         static int window_index = 1;
 
@@ -490,45 +537,33 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
                 const auto partition_by_ast = elist->children[1];
                 fmt::print(stderr, "partition by ast {}\n",
                     partition_by_ast->dumpTree());
-                if (const auto * as_tuple = partition_by_ast->as<ASTFunction>();
-                    as_tuple
-                    && as_tuple->name == "tuple"
-                    && as_tuple->arguments)
-                {
-                    // untuple it
-                    for (const auto & element_ast
-                        : as_tuple->arguments->children)
-                    {
-                        const auto * with_alias = dynamic_cast<
-                            const ASTWithAlias *>(element_ast.get());
-                        if (!with_alias)
-                        {
-                            throw Exception(ErrorCodes::BAD_ARGUMENTS,
-                                "(1) Expected column in PARTITION BY"
-                                " for window '{}', got '{}'",
-                                window_description.window_name,
-                                element_ast->formatForErrorMessage());
-                        }
-                        window_description.partition_by.push_back(
-                            with_alias->getColumnName());
-                    }
-                }
-                else if (const auto * with_alias
-                    = dynamic_cast<const ASTWithAlias *>(partition_by_ast.get()))
-                {
-                    window_description.partition_by.push_back(
-                        with_alias->getColumnName());
-                }
-                else
-                {
-                    throw Exception(ErrorCodes::BAD_ARGUMENTS,
-                        "(2) Expected tuple or column in PARTITION BY"
-                        " for window '{}', got '{}'",
-                        window_description.window_name,
-                        partition_by_ast->formatForErrorMessage());
-                }
+
+                window_description.partition_by = windowArgumentToSortDescription(
+                    partition_by_ast.get(), window_description);
+            }
+
+            if (elist->children.size() == 3)
+            {
+                const auto order_by_ast = elist->children[2];
+                fmt::print(stderr, "order by ast {}\n",
+                    order_by_ast->dumpTree());
+
+                window_description.order_by = windowArgumentToSortDescription(
+                    order_by_ast.get(), window_description);
+            }
+
+            if (elist->children.size() > 3)
+            {
+                throw Exception(ErrorCodes::BAD_ARGUMENTS,
+                    "Too many arguments to window function '{}'",
+                    wrapper_node->formatForErrorMessage());
             }
         }
+        window_description.full_sort_description = window_description.partition_by;
+        window_description.full_sort_description.insert(
+            window_description.full_sort_description.end(),
+            window_description.order_by.begin(),
+            window_description.order_by.end());
 
         WindowFunctionDescription window_function;
         window_function.window_name = window_description.window_name;
@@ -543,6 +578,18 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
                     window_function.function_node->parameters)
                 : Array();
 
+        // We have to fill actions for window function arguments, so that we are
+        // then able to find their argumen types. The `actions` passed to this
+        // functions are temporary and are discarded.
+        // The same calculation is done in appendWindowFunctionsArguments.
+        getRootActionsNoMakeSet(wrapper_node->arguments, true /* no subqueries */,
+            actions);
+        // We have to separately get actions for the arguments of the aggregate
+        // function we calculate over window, because the ActionsVisitor does
+        // not descend into aggregate functions.
+        getRootActionsNoMakeSet(window_function.function_node->arguments,
+            true /* no subqueries */, actions);
+
         const ASTs & arguments
             = window_function.function_node->arguments->children;
         window_function.argument_types.resize(arguments.size());
@@ -956,29 +1003,77 @@ void SelectQueryExpressionAnalyzer::appendAggregateFunctionsArguments(Expression
 }
 
 void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
-    ExpressionActionsChain & chain, bool only_types)
+    ExpressionActionsChain & chain, bool /* only_types */)
 {
     fmt::print(stderr, "actions before window: {}\n", chain.dumpChain());
-    const auto * select_query = getSelectQuery();
 
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
-    /*
-    for (const auto & desc : aggregate_descriptions)
-        for (const auto & name : desc.argument_names)
-            step.required_output.emplace_back(name);
-    */
+    for (const auto & f : window_functions)
+    {
+        // Not sure why NoMakeSet, copied from aggregate functions.
+        getRootActionsNoMakeSet(f.wrapper_node->arguments,
+            true /* no subqueries */, step.actions());
 
-    /// Collect aggregates removing duplicates by node.getColumnName()
-    /// It's not clear why we recollect aggregates (for query parts) while we're able to use previously collected ones (for entire query)
-    /// @note The original recollection logic didn't remove duplicates.
-    GetAggregatesVisitor::Data data;
-    GetAggregatesVisitor(data).visit(select_query->select());
+        // We have to separately get actions for the arguments of the aggregate
+        // function we calculate over window, because the ActionsVisitor does
+        // not descend into aggregate functions.
+        // Not sure why NoMakeSet, copied from aggregate functions.
+        getRootActionsNoMakeSet(f.function_node->arguments,
+            true /* no subqueries */, step.actions());
 
-    /// TODO: data.aggregates -> aggregates()
-    for (const ASTFunction * node : data.window_functions)
-        for (auto & argument : node->arguments->children)
-            getRootActions(argument, only_types, step.actions());
+        // Add column with window function name and value "1".
+        // It is an aggregate function, so it won't be added by getRootActions.
+        ColumnWithTypeAndName col;
+        col.type = std::make_shared<DataTypeInt64>();
+        col.column = col.type->createColumnConst(1 /* size */, UInt64(1) /* field */);
+        col.name = f.column_name;
+
+        step.actions()->addColumn(col);
+    }
+
+    // /*
+    // for (const auto & desc : aggregate_descriptions)
+    //     for (const auto & name : desc.argument_names)
+    //         step.required_output.emplace_back(name);
+    // */
+
+    // const auto * select_query = getSelectQuery();
+    // /// Collect aggregates removing duplicates by node.getColumnName()
+    // /// It's not clear why we recollect aggregates (for query parts) while we're able to use previously collected ones (for entire query)
+    // /// @note The original recollection logic didn't remove duplicates.
+    // GetAggregatesVisitor::Data data;
+    // GetAggregatesVisitor(data).visit(select_query->select());
+
+    // // 1) just add everything there is in the AST,
+    // for (const ASTFunction * node : data.window_functions)
+    // {
+    //     for (auto & argument : node->arguments->children)
+    //     {
+    //         getRootActions(argument, only_types, step.actions());
+    //     }
+    // }
+
+    // 2) mark the columns that are really required:
+    for (const auto & f : window_functions)
+    {
+        for (const auto & a : f.function_node->arguments->children)
+        {
+            // 2.1) function arguments,
+            step.required_output.push_back(a->getColumnName());
+        }
+        // 2.2) function result,
+        step.required_output.push_back(f.column_name);
+    }
+
+    // 2.3) PARTITION BY and ORDER BY columns.
+    for (const auto & [_, w] : window_descriptions)
+    {
+        for (const auto & c : w.full_sort_description)
+        {
+            step.required_output.push_back(c.column_name);
+        }
+    }
 
     fmt::print(stderr, "actions after window: {}\n", chain.dumpChain());
 }
@@ -1155,6 +1250,18 @@ ActionsDAGPtr SelectQueryExpressionAnalyzer::appendProjectResult(ExpressionActio
         }
     }
 
+    for (const auto & f : window_functions)
+    {
+        if (required_result_columns.empty()
+            || required_result_columns.count(f.column_name))
+        {
+            result_columns.emplace_back(f.column_name, f.column_name);
+            step.required_output.push_back(f.column_name);
+        }
+    }
+
+    fmt::print(stderr, "chain before last projection: {}\n",
+        chain.dumpChain());
     auto actions = chain.getLastActions();
     actions->project(result_columns);
     return actions;
@@ -1407,6 +1514,7 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
             && !query.final()
             && join_allow_read_in_order;
 
+        /*
         if (has_window)
         {
             query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
@@ -1414,11 +1522,15 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
 
             finalize_chain(chain);
         }
+        */
 
         /// If there is aggregation, we execute expressions in SELECT and ORDER BY on the initiating server, otherwise on the source servers.
         query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
         fmt::print(stderr, "chain after select: {}\n", chain.dumpChain());
 
+        query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
+        fmt::print(stderr, "chain after window: {}\n", chain.dumpChain());
+
         selected_columns = chain.getLastStep().required_output;
         has_order_by = query.orderBy() != nullptr;
         before_order_and_select = query_analyzer.appendOrderBy(
@@ -1609,4 +1721,15 @@ std::string WindowFunctionDescription::dump() const
     return ss.str();
 }
 
+std::string WindowDescription::dump() const
+{
+    std::stringstream ss;
+    ss << "window '" << window_name << "'\n";
+    ss << "partition_by " << dumpSortDescription(partition_by) << "\n";
+    ss << "order_by " << dumpSortDescription(order_by) << "\n";
+    ss << "full_sort_description " << dumpSortDescription(full_sort_description) << "\n";
+
+    return ss.str();
+}
+
 }
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index b1d7bfba495..dd9a76c1f5c 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -76,7 +76,7 @@ struct ExpressionAnalyzerData
   *
   * NOTE: if `ast` is a SELECT query from a table, the structure of this table should not change during the lifetime of ExpressionAnalyzer.
   */
-class ExpressionAnalyzer : protected ExpressionAnalyzerData, private boost::noncopyable
+class ExpressionAnalyzer : public ExpressionAnalyzerData, private boost::noncopyable
 {
 private:
     /// Extracts settings to enlight which are used (and avoid copy of others).
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 43c5102fa32..1fc414c482d 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -33,36 +33,37 @@
 #include <Interpreters/QueryAliasesVisitor.h>
 
 #include <Processors/Pipe.h>
-#include <Processors/Sources/SourceFromInputStream.h>
-#include <Processors/Sources/NullSource.h>
-#include <Processors/Transforms/ExpressionTransform.h>
-#include <Processors/Transforms/JoiningTransform.h>
-#include <Processors/Transforms/AggregatingTransform.h>
-#include <Processors/Transforms/FilterTransform.h>
-#include <Processors/QueryPlan/ArrayJoinStep.h>
-#include <Processors/QueryPlan/SettingQuotaAndLimitsStep.h>
-#include <Processors/QueryPlan/ExpressionStep.h>
-#include <Processors/QueryPlan/FilterStep.h>
-#include <Processors/QueryPlan/ReadNothingStep.h>
-#include <Processors/QueryPlan/ReadFromPreparedSource.h>
-#include <Processors/QueryPlan/PartialSortingStep.h>
-#include <Processors/QueryPlan/MergeSortingStep.h>
-#include <Processors/QueryPlan/MergingSortedStep.h>
-#include <Processors/QueryPlan/DistinctStep.h>
-#include <Processors/QueryPlan/LimitByStep.h>
-#include <Processors/QueryPlan/LimitStep.h>
-#include <Processors/QueryPlan/MergingAggregatedStep.h>
 #include <Processors/QueryPlan/AddingDelayedSourceStep.h>
 #include <Processors/QueryPlan/AggregatingStep.h>
+#include <Processors/QueryPlan/ArrayJoinStep.h>
 #include <Processors/QueryPlan/CreatingSetsStep.h>
-#include <Processors/QueryPlan/TotalsHavingStep.h>
-#include <Processors/QueryPlan/RollupStep.h>
 #include <Processors/QueryPlan/CubeStep.h>
-#include <Processors/QueryPlan/FillingStep.h>
+#include <Processors/QueryPlan/DistinctStep.h>
+#include <Processors/QueryPlan/ExpressionStep.h>
 #include <Processors/QueryPlan/ExtremesStep.h>
-#include <Processors/QueryPlan/OffsetStep.h>
+#include <Processors/QueryPlan/FillingStep.h>
+#include <Processors/QueryPlan/FilterStep.h>
 #include <Processors/QueryPlan/FinishSortingStep.h>
+#include <Processors/QueryPlan/LimitByStep.h>
+#include <Processors/QueryPlan/LimitStep.h>
+#include <Processors/QueryPlan/MergeSortingStep.h>
+#include <Processors/QueryPlan/MergingAggregatedStep.h>
+#include <Processors/QueryPlan/MergingSortedStep.h>
+#include <Processors/QueryPlan/OffsetStep.h>
+#include <Processors/QueryPlan/PartialSortingStep.h>
 #include <Processors/QueryPlan/QueryPlan.h>
+#include <Processors/QueryPlan/ReadFromPreparedSource.h>
+#include <Processors/QueryPlan/ReadNothingStep.h>
+#include <Processors/QueryPlan/RollupStep.h>
+#include <Processors/QueryPlan/SettingQuotaAndLimitsStep.h>
+#include <Processors/QueryPlan/TotalsHavingStep.h>
+#include <Processors/QueryPlan/WindowStep.h>
+#include <Processors/Sources/NullSource.h>
+#include <Processors/Sources/SourceFromInputStream.h>
+#include <Processors/Transforms/AggregatingTransform.h>
+#include <Processors/Transforms/ExpressionTransform.h>
+#include <Processors/Transforms/FilterTransform.h>
+#include <Processors/Transforms/JoiningTransform.h>
 
 #include <Storages/MergeTree/MergeTreeData.h>
 #include <Storages/MergeTree/MergeTreeWhereOptimizer.h>
@@ -949,13 +950,15 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
 
             if (expressions.need_aggregate)
             {
-                executeAggregation(query_plan, expressions.before_aggregation, aggregate_overflow_row, aggregate_final, query_info.input_order_info);
+               executeAggregation(query_plan, expressions.before_aggregation, aggregate_overflow_row, aggregate_final, query_info.input_order_info);
                 /// We need to reset input order info, so that executeOrder can't use  it
                 query_info.input_order_info.reset();
             }
             else
             {
+                /// FIXME calculate windows here
                 executeExpression(query_plan, expressions.before_order_and_select, "Before ORDER BY and SELECT");
+                executeWindow(query_plan);
                 executeDistinct(query_plan, true, expressions.selected_columns, true);
             }
 
@@ -1001,7 +1004,9 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
                 else if (expressions.hasHaving())
                     executeHaving(query_plan, expressions.before_having);
 
+                /// FIXME calculate windows here
                 executeExpression(query_plan, expressions.before_order_and_select, "Before ORDER BY and SELECT");
+                executeWindow(query_plan);
                 executeDistinct(query_plan, true, expressions.selected_columns, true);
 
             }
@@ -1735,6 +1740,74 @@ void InterpreterSelectQuery::executeExpression(QueryPlan & query_plan, const Act
 }
 
 
+void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
+{
+    for (const auto & f : query_analyzer->window_functions)
+    {
+        const auto & w = query_analyzer->window_descriptions[f.window_name];
+
+        fmt::print(stderr, "window function '{}' over window '{}'\n",
+            f.column_name, f.window_name);
+        fmt::print(stderr, "{}\n{}\n", f.dump(), w.dump());
+
+        const Settings & settings = context->getSettingsRef();
+
+        auto partial_sorting = std::make_unique<PartialSortingStep>(
+            query_plan.getCurrentDataStream(),
+            w.full_sort_description,
+            0 /* LIMIT */,
+            SizeLimits(settings.max_rows_to_sort, settings.max_bytes_to_sort,
+                settings.sort_overflow_mode));
+        partial_sorting->setStepDescription("Sort each block for window '"
+            + w.window_name + "'");
+        query_plan.addStep(std::move(partial_sorting));
+
+        auto merge_sorting_step = std::make_unique<MergeSortingStep>(
+            query_plan.getCurrentDataStream(),
+            w.full_sort_description,
+            settings.max_block_size,
+            0 /* LIMIT */,
+            settings.max_bytes_before_remerge_sort,
+            settings.max_bytes_before_external_sort,
+            context->getTemporaryVolume(),
+            settings.min_free_disk_space_for_temporary_data);
+        merge_sorting_step->setStepDescription("Merge sorted blocks for window '"
+            + w.window_name + "'");
+        query_plan.addStep(std::move(merge_sorting_step));
+
+
+        // First MergeSorted, now MergingSorted......
+        auto merging_sorted = std::make_unique<MergingSortedStep>(
+            query_plan.getCurrentDataStream(),
+            w.full_sort_description,
+            settings.max_block_size,
+            0 /* LIMIT */);
+        merging_sorted->setStepDescription("Merge sorted streams for window '"
+            + w.window_name + "'");
+        query_plan.addStep(std::move(merging_sorted));
+
+        // Add column with window function name and value "1".
+        ColumnWithTypeAndName col;
+        col.type = std::make_shared<DataTypeInt64>();
+        col.column = col.type->createColumnConst(1 /* size */, UInt64(1) /* field */);
+        col.name = f.column_name;
+
+        ActionsDAGPtr window_dag = std::make_shared<ActionsDAG>();
+        window_dag->addColumn(col);
+
+        fmt::print(stderr, "window dag: {}\n", window_dag->dumpDAG());
+
+        auto window_step = std::make_unique<WindowStep>(
+            query_plan.getCurrentDataStream(),
+            window_dag);
+        window_step->setStepDescription("Window step for function '"
+            + f.column_name + "'");
+
+        query_plan.addStep(std::move(window_step));
+    }
+}
+
+
 void InterpreterSelectQuery::executeOrderOptimized(QueryPlan & query_plan, InputOrderInfoPtr input_sorting_info, UInt64 limit, SortDescription & output_order_descr)
 {
     const Settings & settings = context->getSettingsRef();
diff --git a/src/Interpreters/InterpreterSelectQuery.h b/src/Interpreters/InterpreterSelectQuery.h
index d82ea4e1187..1fff316e1d4 100644
--- a/src/Interpreters/InterpreterSelectQuery.h
+++ b/src/Interpreters/InterpreterSelectQuery.h
@@ -120,6 +120,8 @@ private:
     void executeTotalsAndHaving(QueryPlan & query_plan, bool has_having, const ActionsDAGPtr & expression, bool overflow_row, bool final);
     void executeHaving(QueryPlan & query_plan, const ActionsDAGPtr & expression);
     static void executeExpression(QueryPlan & query_plan, const ActionsDAGPtr & expression, const std::string & description);
+    /// FIXME should go through ActionsDAG to behave as a proper function
+    void executeWindow(QueryPlan & query_plan);
     void executeOrder(QueryPlan & query_plan, InputOrderInfoPtr sorting_info);
     void executeOrderOptimized(QueryPlan & query_plan, InputOrderInfoPtr sorting_info, UInt64 limit, SortDescription & output_order_descr);
     void executeWithFill(QueryPlan & query_plan);

From 6f991725c38d8d9489e628777bd4621a7f4b62eb Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 11 Dec 2020 01:16:58 +0300
Subject: [PATCH 057/504] forgot new files

---
 src/Processors/QueryPlan/WindowStep.cpp       | 110 ++++++++++++++++++
 src/Processors/QueryPlan/WindowStep.h         |  32 +++++
 src/Processors/Transforms/WindowTransform.cpp |  33 ++++++
 src/Processors/Transforms/WindowTransform.h   |  36 ++++++
 4 files changed, 211 insertions(+)
 create mode 100644 src/Processors/QueryPlan/WindowStep.cpp
 create mode 100644 src/Processors/QueryPlan/WindowStep.h
 create mode 100644 src/Processors/Transforms/WindowTransform.cpp
 create mode 100644 src/Processors/Transforms/WindowTransform.h

diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
new file mode 100644
index 00000000000..2db6bd3d416
--- /dev/null
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -0,0 +1,110 @@
+#include <Processors/QueryPlan/WindowStep.h>
+
+#include <Processors/Transforms/WindowTransform.h>
+#include <Processors/QueryPipeline.h>
+#include <Interpreters/ExpressionActions.h>
+#include <IO/Operators.h>
+
+namespace DB
+{
+
+static ITransformingStep::Traits getTraits(const ActionsDAGPtr & actions)
+{
+    return ITransformingStep::Traits
+    {
+        {
+            .preserves_distinct_columns = !actions->hasArrayJoin(),
+            .returns_single_stream = false,
+            .preserves_number_of_streams = true,
+            .preserves_sorting = !actions->hasArrayJoin(),
+        },
+        {
+            .preserves_number_of_rows = !actions->hasArrayJoin(),
+        }
+    };
+}
+
+static ITransformingStep::Traits getJoinTraits()
+{
+    return ITransformingStep::Traits
+    {
+        {
+            .preserves_distinct_columns = false,
+            .returns_single_stream = false,
+            .preserves_number_of_streams = true,
+            .preserves_sorting = false,
+        },
+        {
+            .preserves_number_of_rows = false,
+        }
+    };
+}
+
+WindowStep::WindowStep(const DataStream & input_stream_,
+        ActionsDAGPtr actions_dag_)
+    : ITransformingStep(
+        input_stream_,
+        Transform::transformHeader(input_stream_.header,
+            std::make_shared<ExpressionActions>(actions_dag_)),
+        getTraits(actions_dag_))
+    , actions_dag(std::move(actions_dag_))
+{
+    /// Some columns may be removed by expression.
+    updateDistinctColumns(output_stream->header, output_stream->distinct_columns);
+}
+
+void WindowStep::updateInputStream(DataStream input_stream, bool keep_header)
+{
+    Block out_header = keep_header
+        ? std::move(output_stream->header)
+        : Transform::transformHeader(input_stream.header,
+            std::make_shared<ExpressionActions>(actions_dag));
+
+    output_stream = createOutputStream(
+            input_stream,
+            std::move(out_header),
+            getDataStreamTraits());
+
+    input_streams.clear();
+    input_streams.emplace_back(std::move(input_stream));
+}
+
+void WindowStep::transformPipeline(QueryPipeline & pipeline)
+{
+    auto expression = std::make_shared<ExpressionActions>(actions_dag);
+    pipeline.addSimpleTransform([&](const Block & header)
+    {
+        return std::make_shared<Transform>(header, expression);
+    });
+
+    if (!blocksHaveEqualStructure(pipeline.getHeader(), output_stream->header))
+    {
+        auto convert_actions_dag = ActionsDAG::makeConvertingActions(
+                pipeline.getHeader().getColumnsWithTypeAndName(),
+                output_stream->header.getColumnsWithTypeAndName(),
+                ActionsDAG::MatchColumnsMode::Name);
+        auto convert_actions = std::make_shared<ExpressionActions>(convert_actions_dag);
+
+        pipeline.addSimpleTransform([&](const Block & header)
+        {
+            return std::make_shared<ExpressionTransform>(header, convert_actions);
+        });
+    }
+}
+
+void WindowStep::describeActions(FormatSettings & settings) const
+{
+    String prefix(settings.offset, ' ');
+    bool first = true;
+
+    auto expression = std::make_shared<ExpressionActions>(actions_dag);
+    for (const auto & action : expression->getActions())
+    {
+        settings.out << prefix << (first ? "Actions: "
+                                         : "         ");
+        first = false;
+        settings.out << action.toString() << '\n';
+    }
+}
+
+}
diff --git a/src/Processors/QueryPlan/WindowStep.h b/src/Processors/QueryPlan/WindowStep.h
new file mode 100644
index 00000000000..4b67e9563db
--- /dev/null
+++ b/src/Processors/QueryPlan/WindowStep.h
@@ -0,0 +1,32 @@
+#pragma once
+#include <Processors/QueryPlan/ITransformingStep.h>
+
+namespace DB
+{
+
+class ActionsDAG;
+using ActionsDAGPtr = std::shared_ptr<ActionsDAG>;
+
+class WindowTransform;
+
+class WindowStep : public ITransformingStep
+{
+public:
+    using Transform = WindowTransform;
+
+    explicit WindowStep(const DataStream & input_stream_, ActionsDAGPtr actions_dag_);
+    String getName() const override { return "Expression"; }
+
+    void transformPipeline(QueryPipeline & pipeline) override;
+
+    void updateInputStream(DataStream input_stream, bool keep_header);
+
+    void describeActions(FormatSettings & settings) const override;
+
+    const ActionsDAGPtr & getExpression() const { return actions_dag; }
+
+private:
+    ActionsDAGPtr actions_dag;
+};
+
+}
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
new file mode 100644
index 00000000000..f4bc660b940
--- /dev/null
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -0,0 +1,33 @@
+#include <Processors/Transforms/WindowTransform.h>
+
+#include <Interpreters/ExpressionActions.h>
+
+namespace DB
+{
+
+Block WindowTransform::transformHeader(Block header, const ExpressionActionsPtr & expression)
+{
+    size_t num_rows = header.rows();
+    expression->execute(header, num_rows, true);
+    return header;
+}
+
+
+WindowTransform::WindowTransform(const Block & header_,
+        ExpressionActionsPtr expression_)
+    : ISimpleTransform(header_, transformHeader(header_, expression_), false)
+    , expression(std::move(expression_))
+{
+}
+
+void WindowTransform::transform(Chunk & chunk)
+{
+    size_t num_rows = chunk.getNumRows();
+    auto block = getInputPort().getHeader().cloneWithColumns(chunk.detachColumns());
+
+    expression->execute(block, num_rows);
+
+    chunk.setColumns(block.getColumns(), num_rows);
+}
+
+}
diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
new file mode 100644
index 00000000000..08648905eda
--- /dev/null
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -0,0 +1,36 @@
+#pragma once
+#include <Processors/ISimpleTransform.h>
+
+namespace DB
+{
+
+class ExpressionActions;
+using ExpressionActionsPtr = std::shared_ptr<ExpressionActions>;
+
+/** Executes a certain expression over the block.
+  * The expression consists of column identifiers from the block, constants, common functions.
+  * For example: hits * 2 + 3, url LIKE '%yandex%'
+  * The expression processes each row independently of the others.
+  */
+class WindowTransform : public ISimpleTransform
+{
+public:
+    WindowTransform(
+            const Block & header_,
+            ExpressionActionsPtr expression_);
+
+    String getName() const override
+    {
+        return "WindowTransform";
+    }
+
+    static Block transformHeader(Block header, const ExpressionActionsPtr & expression);
+
+protected:
+    void transform(Chunk & chunk) override;
+
+private:
+    ExpressionActionsPtr expression;
+};
+
+}

From dea446be2f01b2c7f59030463149d669270169b1 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 11 Dec 2020 02:56:57 +0300
Subject: [PATCH 058/504] Remove some redundant includes to speed up build

---
 .../AggregateFunctionSequenceMatch.h                  |  2 +-
 src/Columns/ColumnLowCardinality.h                    |  1 -
 src/Common/ColumnsHashingImpl.h                       |  1 +
 src/DataTypes/DataTypeLowCardinality.cpp              |  1 +
 src/Interpreters/ExpressionAnalyzer.cpp               |  1 +
 src/Interpreters/HashJoin.h                           |  1 +
 src/Interpreters/IJoin.h                              |  1 +
 src/Interpreters/InterpreterCreateQuery.h             |  2 ++
 src/Interpreters/InterpreterRenameQuery.h             |  1 +
 src/Interpreters/InterpreterSelectQuery.cpp           |  2 ++
 src/Interpreters/MySQL/InterpretersMySQLDDLQuery.cpp  |  2 ++
 src/Interpreters/TextLog.cpp                          | 11 +++++++----
 src/Storages/ConstraintsDescription.cpp               |  3 +++
 src/Storages/ConstraintsDescription.h                 |  2 +-
 src/Storages/LiveView/StorageLiveView.cpp             |  6 ++++++
 src/Storages/LiveView/StorageLiveView.h               |  2 +-
 src/Storages/MergeTree/MergeTreeData.h                |  1 +
 src/Storages/ReadInOrderOptimizer.cpp                 |  6 ++++--
 src/Storages/StorageMaterializeMySQL.cpp              |  4 ++++
 .../System/StorageSystemDistributionQueue.cpp         |  1 +
 src/Storages/System/StorageSystemTables.cpp           |  1 +
 src/Storages/System/StorageSystemZooKeeper.cpp        |  1 +
 src/Storages/System/attachSystemTablesImpl.h          |  2 ++
 23 files changed, 45 insertions(+), 10 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionSequenceMatch.h b/src/AggregateFunctions/AggregateFunctionSequenceMatch.h
index c7e2a700dd7..1a3da20e347 100644
--- a/src/AggregateFunctions/AggregateFunctionSequenceMatch.h
+++ b/src/AggregateFunctions/AggregateFunctionSequenceMatch.h
@@ -590,7 +590,7 @@ public:
 
     void insertResultInto(AggregateDataPtr place, IColumn & to, Arena *) const override
     {
-        const_cast<Data &>(this->data(place)).sort();
+        this->data(place).sort();
         assert_cast<ColumnUInt64 &>(to).getData().push_back(count(place));
     }
 
diff --git a/src/Columns/ColumnLowCardinality.h b/src/Columns/ColumnLowCardinality.h
index 0874f619b8a..5ace0a5d64f 100644
--- a/src/Columns/ColumnLowCardinality.h
+++ b/src/Columns/ColumnLowCardinality.h
@@ -3,7 +3,6 @@
 #include <Columns/IColumnUnique.h>
 #include <Common/typeid_cast.h>
 #include <Common/assert_cast.h>
-#include <AggregateFunctions/AggregateFunctionCount.h>
 #include "ColumnsNumber.h"
 
 
diff --git a/src/Common/ColumnsHashingImpl.h b/src/Common/ColumnsHashingImpl.h
index 826e8d31ee9..cdd8dd20dd0 100644
--- a/src/Common/ColumnsHashingImpl.h
+++ b/src/Common/ColumnsHashingImpl.h
@@ -1,6 +1,7 @@
 #pragma once
 
 #include <Columns/IColumn.h>
+#include <Columns/ColumnNullable.h>
 #include <Common/assert_cast.h>
 #include <Common/HashTable/HashTableKeyHolder.h>
 #include <Interpreters/AggregationCommon.h>
diff --git a/src/DataTypes/DataTypeLowCardinality.cpp b/src/DataTypes/DataTypeLowCardinality.cpp
index 6b1166dc26a..8f4b2bf7636 100644
--- a/src/DataTypes/DataTypeLowCardinality.cpp
+++ b/src/DataTypes/DataTypeLowCardinality.cpp
@@ -10,6 +10,7 @@
 #include <DataTypes/DataTypeFactory.h>
 #include <DataTypes/DataTypeLowCardinality.h>
 #include <DataTypes/DataTypeNullable.h>
+#include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypeDate.h>
 #include <DataTypes/DataTypeDateTime.h>
 #include <Parsers/IAST.h>
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index e8fa1401700..0d54f75c588 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -12,6 +12,7 @@
 #include <Parsers/DumpASTNode.h>
 
 #include <DataTypes/DataTypeNullable.h>
+#include <DataTypes/DataTypesNumber.h>
 #include <Columns/IColumn.h>
 
 #include <Interpreters/ExpressionAnalyzer.h>
diff --git a/src/Interpreters/HashJoin.h b/src/Interpreters/HashJoin.h
index 986b4474473..ceddc9e4c7c 100644
--- a/src/Interpreters/HashJoin.h
+++ b/src/Interpreters/HashJoin.h
@@ -22,6 +22,7 @@
 #include <DataStreams/SizeLimits.h>
 #include <DataStreams/IBlockStream_fwd.h>
 
+#include <Core/Block.h>
 
 namespace DB
 {
diff --git a/src/Interpreters/IJoin.h b/src/Interpreters/IJoin.h
index b9f88b57b37..3d313adcb6c 100644
--- a/src/Interpreters/IJoin.h
+++ b/src/Interpreters/IJoin.h
@@ -12,6 +12,7 @@ namespace DB
 
 class Block;
 struct ExtraBlock;
+using ExtraBlockPtr = std::shared_ptr<ExtraBlock>;
 
 class IJoin
 {
diff --git a/src/Interpreters/InterpreterCreateQuery.h b/src/Interpreters/InterpreterCreateQuery.h
index 07fca5f3910..735cc2122fa 100644
--- a/src/Interpreters/InterpreterCreateQuery.h
+++ b/src/Interpreters/InterpreterCreateQuery.h
@@ -16,6 +16,8 @@ class Context;
 class ASTCreateQuery;
 class ASTExpressionList;
 class ASTConstraintDeclaration;
+class IDatabase;
+using DatabasePtr = std::shared_ptr<IDatabase>;
 
 
 /** Allows to create new table or database,
diff --git a/src/Interpreters/InterpreterRenameQuery.h b/src/Interpreters/InterpreterRenameQuery.h
index f2c08f22602..6fb0803d038 100644
--- a/src/Interpreters/InterpreterRenameQuery.h
+++ b/src/Interpreters/InterpreterRenameQuery.h
@@ -9,6 +9,7 @@ namespace DB
 
 class Context;
 class AccessRightsElements;
+class DDLGuard;
 
 /// To avoid deadlocks, we must acquire locks for tables in same order in any different RENAMES.
 struct UniqueTableName
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 1fc414c482d..2d6cb64bb7e 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -14,6 +14,8 @@
 
 #include <Access/AccessFlags.h>
 
+#include <AggregateFunctions/AggregateFunctionCount.h>
+
 #include <Interpreters/ApplyWithAliasVisitor.h>
 #include <Interpreters/ApplyWithSubqueryVisitor.h>
 #include <Interpreters/InterpreterSelectQuery.h>
diff --git a/src/Interpreters/MySQL/InterpretersMySQLDDLQuery.cpp b/src/Interpreters/MySQL/InterpretersMySQLDDLQuery.cpp
index 37714c1a83e..2f8a544103e 100644
--- a/src/Interpreters/MySQL/InterpretersMySQLDDLQuery.cpp
+++ b/src/Interpreters/MySQL/InterpretersMySQLDDLQuery.cpp
@@ -21,6 +21,8 @@
 #include <Common/assert_cast.h>
 #include <Interpreters/Context.h>
 #include <Interpreters/InterpreterCreateQuery.h>
+#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/TreeRewriter.h>
 #include <Storages/IStorage.h>
 
 namespace DB
diff --git a/src/Interpreters/TextLog.cpp b/src/Interpreters/TextLog.cpp
index d1cc6a052e8..f60b6acae6f 100644
--- a/src/Interpreters/TextLog.cpp
+++ b/src/Interpreters/TextLog.cpp
@@ -1,11 +1,14 @@
 #include <Interpreters/TextLog.h>
-#include <DataTypes/DataTypeEnum.h>
-#include <DataTypes/DataTypesNumber.h>
+
+#include <Common/ClickHouseRevision.h>
+#include <DataTypes/DataTypeDate.h>
 #include <DataTypes/DataTypeDateTime.h>
 #include <DataTypes/DataTypeDateTime64.h>
-#include <DataTypes/DataTypeDate.h>
+#include <DataTypes/DataTypeEnum.h>
+#include <DataTypes/DataTypeLowCardinality.h>
 #include <DataTypes/DataTypeString.h>
-#include <Common/ClickHouseRevision.h>
+#include <DataTypes/DataTypesNumber.h>
+
 #include <array>
 
 namespace DB
diff --git a/src/Storages/ConstraintsDescription.cpp b/src/Storages/ConstraintsDescription.cpp
index cff0b975a1d..e6315872a66 100644
--- a/src/Storages/ConstraintsDescription.cpp
+++ b/src/Storages/ConstraintsDescription.cpp
@@ -1,5 +1,8 @@
 #include <Storages/ConstraintsDescription.h>
 
+#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/TreeRewriter.h>
+
 #include <Parsers/formatAST.h>
 #include <Parsers/ParserCreateQuery.h>
 #include <Parsers/parseQuery.h>
diff --git a/src/Storages/ConstraintsDescription.h b/src/Storages/ConstraintsDescription.h
index ddc9597f813..d6d2baefbd2 100644
--- a/src/Storages/ConstraintsDescription.h
+++ b/src/Storages/ConstraintsDescription.h
@@ -1,7 +1,7 @@
 #pragma once
 
 #include <Parsers/ASTConstraintDeclaration.h>
-#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/ExpressionActions.h>
 
 namespace DB
 {
diff --git a/src/Storages/LiveView/StorageLiveView.cpp b/src/Storages/LiveView/StorageLiveView.cpp
index bdf82307b07..0803b3a9c9c 100644
--- a/src/Storages/LiveView/StorageLiveView.cpp
+++ b/src/Storages/LiveView/StorageLiveView.cpp
@@ -35,6 +35,7 @@ limitations under the License. */
 #include <Parsers/ASTSubquery.h>
 #include <Parsers/queryToString.h>
 #include <Interpreters/DatabaseAndTableWithAlias.h>
+#include <Interpreters/DatabaseCatalog.h>
 #include <Interpreters/getTableExpressions.h>
 #include <Interpreters/AddDefaultDatabaseVisitor.h>
 #include <Access/AccessFlags.h>
@@ -303,6 +304,11 @@ Block StorageLiveView::getHeader() const
     return sample_block;
 }
 
+StoragePtr StorageLiveView::getParentStorage() const
+{
+    return DatabaseCatalog::instance().getTable(select_table_id, global_context);
+}
+
 ASTPtr StorageLiveView::getInnerBlocksQuery()
 {
     std::lock_guard lock(sample_block_lock);
diff --git a/src/Storages/LiveView/StorageLiveView.h b/src/Storages/LiveView/StorageLiveView.h
index 9c843dc71de..4a03e29342a 100644
--- a/src/Storages/LiveView/StorageLiveView.h
+++ b/src/Storages/LiveView/StorageLiveView.h
@@ -57,7 +57,7 @@ public:
     {
         return getStorageID().table_name + "_blocks";
     }
-    StoragePtr getParentStorage() const { return DatabaseCatalog::instance().getTable(select_table_id, global_context); }
+    StoragePtr getParentStorage() const;
 
     ASTPtr getInnerQuery() const { return inner_query->clone(); }
     ASTPtr getInnerSubQuery() const
diff --git a/src/Storages/MergeTree/MergeTreeData.h b/src/Storages/MergeTree/MergeTreeData.h
index 1f9ca97abaf..38e6c34e232 100644
--- a/src/Storages/MergeTree/MergeTreeData.h
+++ b/src/Storages/MergeTree/MergeTreeData.h
@@ -43,6 +43,7 @@ struct JobAndPool;
 
 class ExpressionActions;
 using ExpressionActionsPtr = std::shared_ptr<ExpressionActions>;
+using ManyExpressionActions = std::vector<ExpressionActionsPtr>;
 
 namespace ErrorCodes
 {
diff --git a/src/Storages/ReadInOrderOptimizer.cpp b/src/Storages/ReadInOrderOptimizer.cpp
index 3613fbff5cf..ae49433dbc6 100644
--- a/src/Storages/ReadInOrderOptimizer.cpp
+++ b/src/Storages/ReadInOrderOptimizer.cpp
@@ -1,8 +1,10 @@
 #include <Storages/ReadInOrderOptimizer.h>
+
+#include <Functions/IFunction.h>
+#include <Interpreters/TableJoin.h>
+#include <Interpreters/TreeRewriter.h>
 #include <Storages/MergeTree/MergeTreeData.h>
 #include <Storages/MergeTree/StorageFromMergeTreeDataPart.h>
-#include <Interpreters/TableJoin.h>
-#include <Functions/IFunction.h>
 
 namespace DB
 {
diff --git a/src/Storages/StorageMaterializeMySQL.cpp b/src/Storages/StorageMaterializeMySQL.cpp
index 7c9ec35f3c5..9be285adc69 100644
--- a/src/Storages/StorageMaterializeMySQL.cpp
+++ b/src/Storages/StorageMaterializeMySQL.cpp
@@ -8,6 +8,8 @@
 
 #include <Core/Settings.h>
 #include <Interpreters/Context.h>
+#include <Interpreters/ExpressionAnalyzer.h>
+#include <Interpreters/TreeRewriter.h>
 
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ASTSelectQuery.h>
@@ -19,6 +21,8 @@
 #include <Processors/Pipe.h>
 #include <Processors/Transforms/FilterTransform.h>
 
+#include <Storages/SelectQueryInfo.h>
+
 namespace DB
 {
 
diff --git a/src/Storages/System/StorageSystemDistributionQueue.cpp b/src/Storages/System/StorageSystemDistributionQueue.cpp
index c8d8c88ec08..edba7c13b1c 100644
--- a/src/Storages/System/StorageSystemDistributionQueue.cpp
+++ b/src/Storages/System/StorageSystemDistributionQueue.cpp
@@ -4,6 +4,7 @@
 #include <DataTypes/DataTypesNumber.h>
 #include <Storages/System/StorageSystemDistributionQueue.h>
 #include <Storages/Distributed/DirectoryMonitor.h>
+#include <Storages/SelectQueryInfo.h>
 #include <Storages/StorageDistributed.h>
 #include <Storages/VirtualColumnUtils.h>
 #include <Access/ContextAccess.h>
diff --git a/src/Storages/System/StorageSystemTables.cpp b/src/Storages/System/StorageSystemTables.cpp
index 02aa7d8302a..c9b3223ee38 100644
--- a/src/Storages/System/StorageSystemTables.cpp
+++ b/src/Storages/System/StorageSystemTables.cpp
@@ -4,6 +4,7 @@
 #include <DataTypes/DataTypeDateTime.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <Storages/System/StorageSystemTables.h>
+#include <Storages/SelectQueryInfo.h>
 #include <Storages/VirtualColumnUtils.h>
 #include <Databases/IDatabase.h>
 #include <Access/ContextAccess.h>
diff --git a/src/Storages/System/StorageSystemZooKeeper.cpp b/src/Storages/System/StorageSystemZooKeeper.cpp
index a3660cf2dec..287650ef86c 100644
--- a/src/Storages/System/StorageSystemZooKeeper.cpp
+++ b/src/Storages/System/StorageSystemZooKeeper.cpp
@@ -2,6 +2,7 @@
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypeDateTime.h>
 #include <Storages/System/StorageSystemZooKeeper.h>
+#include <Storages/SelectQueryInfo.h>
 #include <Parsers/ASTSelectQuery.h>
 #include <Parsers/ASTIdentifier.h>
 #include <Parsers/ASTLiteral.h>
diff --git a/src/Storages/System/attachSystemTablesImpl.h b/src/Storages/System/attachSystemTablesImpl.h
index 0dc34c3c116..78e944f7d27 100644
--- a/src/Storages/System/attachSystemTablesImpl.h
+++ b/src/Storages/System/attachSystemTablesImpl.h
@@ -1,6 +1,8 @@
 #pragma once
 #include <Databases/IDatabase.h>
 
+#include <Interpreters/DatabaseCatalog.h>
+
 namespace DB
 {
 

From 4809b5975a16efd36f3678446b59d31e35c39fd1 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Mon, 30 Nov 2020 16:53:07 +0800
Subject: [PATCH 059/504] Fix system.stack_trace in daemon

---
 src/Storages/System/StorageSystemStackTrace.cpp | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index 9edcb1ede47..e822ac26744 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -33,7 +33,8 @@ namespace ErrorCodes
 
 namespace
 {
-    const pid_t expected_pid = getpid();
+    // Initialized in StorageSystemStackTrace's ctor and used in signalHandler.
+    pid_t expected_pid;
     const int sig = SIGRTMIN;
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.
@@ -133,6 +134,7 @@ StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
 
     /// Setup signal handler.
 
+    expected_pid = getpid();
     struct sigaction sa{};
     sa.sa_sigaction = signalHandler;
     sa.sa_flags = SA_SIGINFO;
@@ -214,7 +216,11 @@ void StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Conte
             res_columns[2]->insertDefault();
         }
 
-        ++sequence_num; /// FYI: For signed Integral types, arithmetic is defined to use two’s complement representation. There are no undefined results.
+        /// Signed integer overflow is undefined behavior in both C and C++. However, according to
+        /// C++ standard, Atomic signed integer arithmetic is defined to use two's complement; there
+        /// are no undefined results. See https://en.cppreference.com/w/cpp/atomic/atomic and
+        /// http://eel.is/c++draft/atomics.types.generic#atomics.types.int-8
+        ++sequence_num;
     }
 }
 

From a0f4dfd94816603db22324edce631f72caa0d249 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Mon, 30 Nov 2020 21:53:04 +0800
Subject: [PATCH 060/504] get back tests

---
 tests/queries/0_stateless/01051_system_stack_trace.reference | 1 +
 tests/queries/0_stateless/01051_system_stack_trace.sql       | 2 ++
 2 files changed, 3 insertions(+)
 create mode 100644 tests/queries/0_stateless/01051_system_stack_trace.reference
 create mode 100644 tests/queries/0_stateless/01051_system_stack_trace.sql

diff --git a/tests/queries/0_stateless/01051_system_stack_trace.reference b/tests/queries/0_stateless/01051_system_stack_trace.reference
new file mode 100644
index 00000000000..d00491fd7e5
--- /dev/null
+++ b/tests/queries/0_stateless/01051_system_stack_trace.reference
@@ -0,0 +1 @@
+1
diff --git a/tests/queries/0_stateless/01051_system_stack_trace.sql b/tests/queries/0_stateless/01051_system_stack_trace.sql
new file mode 100644
index 00000000000..32d344fce7e
--- /dev/null
+++ b/tests/queries/0_stateless/01051_system_stack_trace.sql
@@ -0,0 +1,2 @@
+-- at least this query should be present
+SELECT count() > 0 FROM system.stack_trace WHERE query_id != '';

From d0044886ab1b41f30cd8ec9255411897f625ecac Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Tue, 1 Dec 2020 11:45:10 +0800
Subject: [PATCH 061/504] fix TSan

---
 src/Storages/System/StorageSystemStackTrace.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index e822ac26744..a72fbeb2d5d 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -34,7 +34,7 @@ namespace ErrorCodes
 namespace
 {
     // Initialized in StorageSystemStackTrace's ctor and used in signalHandler.
-    pid_t expected_pid;
+    std::atomic<pid_t> expected_pid;
     const int sig = SIGRTMIN;
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.

From c4306b5788a8f4b327e5c73cc3a0e397f8afeb4d Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Tue, 1 Dec 2020 16:37:30 +0800
Subject: [PATCH 062/504] Another try of TSan fix

---
 src/Storages/System/StorageSystemStackTrace.cpp | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index a72fbeb2d5d..ef4bf023933 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -34,7 +34,6 @@ namespace ErrorCodes
 namespace
 {
     // Initialized in StorageSystemStackTrace's ctor and used in signalHandler.
-    std::atomic<pid_t> expected_pid;
     const int sig = SIGRTMIN;
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.
@@ -49,6 +48,8 @@ namespace
 
     void signalHandler(int, siginfo_t * info, void * context)
     {
+        // getpid() is an async-signal-safe function required by POSIX.1
+        static const pid_t expected_pid = getpid();
         auto saved_errno = errno;   /// We must restore previous value of errno in signal handler.
 
         /// In case malicious user is sending signals manually (for unknown reason).
@@ -133,8 +134,6 @@ StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
     notification_pipe.open();
 
     /// Setup signal handler.
-
-    expected_pid = getpid();
     struct sigaction sa{};
     sa.sa_sigaction = signalHandler;
     sa.sa_flags = SA_SIGINFO;

From 5e71f4e810a7005f489f6a2cb6a4df509ad2eca1 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Tue, 1 Dec 2020 20:33:42 +0800
Subject: [PATCH 063/504] Have to resort to NO_SANITIZE_THREAD

---
 src/Storages/System/StorageSystemStackTrace.cpp | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index ef4bf023933..4f2ade17900 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -34,6 +34,7 @@ namespace ErrorCodes
 namespace
 {
     // Initialized in StorageSystemStackTrace's ctor and used in signalHandler.
+    std::atomic<pid_t> expected_pid;
     const int sig = SIGRTMIN;
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.
@@ -46,10 +47,9 @@ namespace
 
     LazyPipeFDs notification_pipe;
 
-    void signalHandler(int, siginfo_t * info, void * context)
+    // TSan complains about expected_pid being accessed unsafely, which is not the case.
+    void NO_SANITIZE_THREAD signalHandler(int, siginfo_t * info, void * context)
     {
-        // getpid() is an async-signal-safe function required by POSIX.1
-        static const pid_t expected_pid = getpid();
         auto saved_errno = errno;   /// We must restore previous value of errno in signal handler.
 
         /// In case malicious user is sending signals manually (for unknown reason).
@@ -134,6 +134,7 @@ StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
     notification_pipe.open();
 
     /// Setup signal handler.
+    expected_pid = getpid();
     struct sigaction sa{};
     sa.sa_sigaction = signalHandler;
     sa.sa_flags = SA_SIGINFO;

From b513eeee7f2ea5148b16118ffe9bb9030109a940 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Tue, 1 Dec 2020 22:20:02 +0300
Subject: [PATCH 064/504] Another try...

---
 src/Storages/System/StorageSystemStackTrace.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index 4f2ade17900..954318fa7ac 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -161,7 +161,7 @@ NamesAndTypesList StorageSystemStackTrace::getNamesAndTypes()
 }
 
 
-void StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
+void NO_SANITIZE_THREAD StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
 {
     /// It shouldn't be possible to do concurrent reads from this table.
     std::lock_guard lock(mutex);

From cfeaff6aa7ac7f1ac94e9166e60aba8b14d8b1b5 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Tue, 1 Dec 2020 22:20:49 +0300
Subject: [PATCH 065/504] Update StorageSystemStackTrace.cpp

---
 src/Storages/System/StorageSystemStackTrace.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index 954318fa7ac..6b6612e611b 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -47,7 +47,7 @@ namespace
 
     LazyPipeFDs notification_pipe;
 
-    // TSan complains about expected_pid being accessed unsafely, which is not the case.
+    // TSan complains about stack_trace being accessed unsafely, which is not the case.
     void NO_SANITIZE_THREAD signalHandler(int, siginfo_t * info, void * context)
     {
         auto saved_errno = errno;   /// We must restore previous value of errno in signal handler.

From 0aa75e6023577c10eb73cb41afaf120cb44bf66a Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Tue, 8 Dec 2020 01:48:41 +0800
Subject: [PATCH 066/504] Another try of TSan fix

---
 src/Storages/System/CMakeLists.txt              | 2 ++
 src/Storages/System/StorageSystemStackTrace.cpp | 5 ++---
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/src/Storages/System/CMakeLists.txt b/src/Storages/System/CMakeLists.txt
index a1eb525dceb..f6d443b9aa3 100644
--- a/src/Storages/System/CMakeLists.txt
+++ b/src/Storages/System/CMakeLists.txt
@@ -15,6 +15,8 @@ include(${ClickHouse_SOURCE_DIR}/cmake/dbms_glob_sources.cmake)
 add_headers_and_sources(storages_system .)
 list (APPEND storages_system_sources ${CONFIG_BUILD})
 
+set_source_files_properties(StorageSystemStackTrace.cpp PROPERTIES COMPILE_FLAGS -fno-sanitize=thread)
+
 add_custom_target(generate-contributors
     ./StorageSystemContributors.sh
     SOURCES StorageSystemContributors.sh
diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index 6b6612e611b..5cfad1f5bdc 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -47,8 +47,7 @@ namespace
 
     LazyPipeFDs notification_pipe;
 
-    // TSan complains about stack_trace being accessed unsafely, which is not the case.
-    void NO_SANITIZE_THREAD signalHandler(int, siginfo_t * info, void * context)
+    void signalHandler(int, siginfo_t * info, void * context)
     {
         auto saved_errno = errno;   /// We must restore previous value of errno in signal handler.
 
@@ -161,7 +160,7 @@ NamesAndTypesList StorageSystemStackTrace::getNamesAndTypes()
 }
 
 
-void NO_SANITIZE_THREAD StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
+void StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
 {
     /// It shouldn't be possible to do concurrent reads from this table.
     std::lock_guard lock(mutex);

From 04128c33e1f99730935edcf2ef1e766dd916e29d Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Sat, 12 Dec 2020 14:17:48 +0800
Subject: [PATCH 067/504] disable TSAN for real

---
 src/Storages/System/CMakeLists.txt              | 2 --
 src/Storages/System/StorageSystemStackTrace.cpp | 8 ++++----
 2 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/src/Storages/System/CMakeLists.txt b/src/Storages/System/CMakeLists.txt
index f6d443b9aa3..a1eb525dceb 100644
--- a/src/Storages/System/CMakeLists.txt
+++ b/src/Storages/System/CMakeLists.txt
@@ -15,8 +15,6 @@ include(${ClickHouse_SOURCE_DIR}/cmake/dbms_glob_sources.cmake)
 add_headers_and_sources(storages_system .)
 list (APPEND storages_system_sources ${CONFIG_BUILD})
 
-set_source_files_properties(StorageSystemStackTrace.cpp PROPERTIES COMPILE_FLAGS -fno-sanitize=thread)
-
 add_custom_target(generate-contributors
     ./StorageSystemContributors.sh
     SOURCES StorageSystemContributors.sh
diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index 5cfad1f5bdc..92d12c19986 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -47,7 +47,7 @@ namespace
 
     LazyPipeFDs notification_pipe;
 
-    void signalHandler(int, siginfo_t * info, void * context)
+    void NO_SANITIZE_THREAD signalHandler(int, siginfo_t * info, void * context)
     {
         auto saved_errno = errno;   /// We must restore previous value of errno in signal handler.
 
@@ -79,7 +79,7 @@ namespace
     }
 
     /// Wait for data in pipe and read it.
-    bool wait(int timeout_ms)
+    bool NO_SANITIZE_THREAD wait(int timeout_ms)
     {
         while (true)
         {
@@ -127,7 +127,7 @@ namespace
 }
 
 
-StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
+NO_SANITIZE_THREAD StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
     : IStorageSystemOneBlock<StorageSystemStackTrace>(table_id_)
 {
     notification_pipe.open();
@@ -160,7 +160,7 @@ NamesAndTypesList StorageSystemStackTrace::getNamesAndTypes()
 }
 
 
-void StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
+void NO_SANITIZE_THREAD StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
 {
     /// It shouldn't be possible to do concurrent reads from this table.
     std::lock_guard lock(mutex);

From 0d4b0410958a5aeb7c1de4194266989d8ed63662 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Sat, 12 Dec 2020 21:03:41 +0800
Subject: [PATCH 068/504] try harder

---
 src/Common/StackTrace.cpp | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/Common/StackTrace.cpp b/src/Common/StackTrace.cpp
index b285a45bdc5..38a2235d516 100644
--- a/src/Common/StackTrace.cpp
+++ b/src/Common/StackTrace.cpp
@@ -252,12 +252,12 @@ void StackTrace::symbolize(const StackTrace::FramePointers & frame_pointers, siz
 #endif
 }
 
-StackTrace::StackTrace()
+NO_SANITIZE_THREAD StackTrace::StackTrace()
 {
     tryCapture();
 }
 
-StackTrace::StackTrace(const ucontext_t & signal_context)
+NO_SANITIZE_THREAD StackTrace::StackTrace(const ucontext_t & signal_context)
 {
     tryCapture();
 
@@ -295,17 +295,17 @@ void StackTrace::tryCapture()
 #endif
 }
 
-size_t StackTrace::getSize() const
+size_t NO_SANITIZE_THREAD StackTrace::getSize() const
 {
     return size;
 }
 
-size_t StackTrace::getOffset() const
+size_t NO_SANITIZE_THREAD StackTrace::getOffset() const
 {
     return offset;
 }
 
-const StackTrace::FramePointers & StackTrace::getFramePointers() const
+const NO_SANITIZE_THREAD StackTrace::FramePointers & StackTrace::getFramePointers() const
 {
     return frame_pointers;
 }

From 116bed2cfdd16e8e434bfb16253895e55d76357e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 9 Dec 2020 21:11:10 +0300
Subject: [PATCH 069/504] Fixing build.

---
 contrib/boost-cmake/CMakeLists.txt  | 109 +++++++---------------------
 src/Common/ZooKeeper/CMakeLists.txt |   2 +-
 src/IO/ReadBufferFromPocoSocket.cpp |   2 +-
 3 files changed, 30 insertions(+), 83 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 714dfd4f011..cfa41a90b5c 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -151,87 +151,34 @@ if (NOT EXTERNAL_BOOST_FOUND)
     enable_language(ASM)
     SET(ASM_OPTIONS "-x assembler-with-cpp")
 
-    set (SRCS_CONTEXT
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_arm_aapcs_pe_armasm.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_combined_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_gas.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_ms_pe_masm.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_i386_x86_64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips32_o32_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_mips64_n64_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_ppc64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc32_sysv_xcoff_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_ppc64_sysv_xcoff_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_gas.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_ms_pe_masm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_arm_aapcs_pe_armasm.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_combined_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_gas.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_ms_pe_masm.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_i386_x86_64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_mips32_o32_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_mips64_n64_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_ppc64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc32_sysv_xcoff_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_ppc64_sysv_xcoff_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_gas.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_ms_pe_masm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm_aapcs_pe_armasm.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_combined_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_gas.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_ms_pe_masm.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_i386_x86_64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips32_o32_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_mips64_n64_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_ppc64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc32_sysv_xcoff_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_ppc64_sysv_xcoff_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_gas.asm
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_ms_pe_masm.asm
-        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_elf_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_macho_gas.S
-#        ${LIBRARY_DIR}/libs/context/src/continuation.cpp
-        ${LIBRARY_DIR}/libs/context/src/dummy.cpp
-        ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
-#        ${LIBRARY_DIR}/libs/context/src/fiber.cpp
-        ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
-#        ${LIBRARY_DIR}/libs/context/src/untested.cpp
-#        ${LIBRARY_DIR}/libs/context/src/windows/stack_traits.cpp
-    )
+    if (ARCH_ARM)
+        set (SRCS_CONTEXT
+            ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S
+            ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_elf_gas.S
+            ${LIBRARY_DIR}/libs/context/src/asm/ontop_arm64_aapcs_elf_gas.S
+            ${LIBRARY_DIR}/libs/context/src/dummy.cpp
+            ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
+            ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
+        )
+    elseif(OS_DARWIN)
+        set (SRCS_CONTEXT
+            ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_macho_gas.S
+            ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_macho_gas.S
+            ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_macho_gas.S
+            ${LIBRARY_DIR}/libs/context/src/dummy.cpp
+            ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
+            ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
+        )
+    else()
+        set (SRCS_CONTEXT
+            ${LIBRARY_DIR}/libs/context/src/asm/jump_x86_64_sysv_elf_gas.S
+            ${LIBRARY_DIR}/libs/context/src/asm/make_x86_64_sysv_elf_gas.S
+            ${LIBRARY_DIR}/libs/context/src/asm/ontop_x86_64_sysv_elf_gas.S
+            ${LIBRARY_DIR}/libs/context/src/dummy.cpp
+            ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
+            ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
+        )
+    endif()
 
     add_library (_boost_context ${SRCS_CONTEXT})
     add_library (boost::context ALIAS _boost_context)
diff --git a/src/Common/ZooKeeper/CMakeLists.txt b/src/Common/ZooKeeper/CMakeLists.txt
index ef32d9266c0..7012f263d00 100644
--- a/src/Common/ZooKeeper/CMakeLists.txt
+++ b/src/Common/ZooKeeper/CMakeLists.txt
@@ -4,7 +4,7 @@ add_headers_and_sources(clickhouse_common_zookeeper .)
 
 add_library(clickhouse_common_zookeeper ${clickhouse_common_zookeeper_headers} ${clickhouse_common_zookeeper_sources})
 
-target_link_libraries (clickhouse_common_zookeeper PUBLIC clickhouse_common_io common PRIVATE string_utils)
+target_link_libraries (clickhouse_common_zookeeper PUBLIC clickhouse_common_io common PRIVATE string_utils _boost_context)
 
 if (ENABLE_TESTS)
     add_subdirectory (tests)
diff --git a/src/IO/ReadBufferFromPocoSocket.cpp b/src/IO/ReadBufferFromPocoSocket.cpp
index b0bfadcf57d..7498fb869e0 100644
--- a/src/IO/ReadBufferFromPocoSocket.cpp
+++ b/src/IO/ReadBufferFromPocoSocket.cpp
@@ -40,7 +40,7 @@ bool ReadBufferFromPocoSocket::nextImpl()
         /// If fiber is specified, and read is blocking, run fiber and try again later.
         /// It is expected that file descriptor may be polled externally.
         /// Note that receive timeout is not checked here. External code should check it while polling.
-        while (bytes_read < 0 && fiber && (errno == EAGAIN || errno == EWOULDBLOCK))
+        while (bytes_read < 0 && fiber && errno == EAGAIN)
         {
             //fiber->fd = socket.impl()->sockfd();
             //fiber->timeout = socket.impl()->getReceiveTimeout();

From dd6b5baa598e4abf784d6f5ac7bf5f2cad0a2d82 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 9 Dec 2020 23:11:25 +0300
Subject: [PATCH 070/504] Add timerfd from musl

---
 base/glibc-compatibility/musl/timerfd.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 base/glibc-compatibility/musl/timerfd.c

diff --git a/base/glibc-compatibility/musl/timerfd.c b/base/glibc-compatibility/musl/timerfd.c
new file mode 100644
index 00000000000..0f9adb54389
--- /dev/null
+++ b/base/glibc-compatibility/musl/timerfd.c
@@ -0,0 +1,17 @@
+#include <sys/timerfd.h>
+#include "syscall.h"
+
+int timerfd_create(int clockid, int flags)
+{
+    return syscall(SYS_timerfd_create, clockid, flags);
+}
+
+int timerfd_settime(int fd, int flags, const struct itimerspec *new, struct itimerspec *old)
+{
+    return syscall(SYS_timerfd_settime, fd, flags, new, old);
+}
+
+int timerfd_gettime(int fd, struct itimerspec *cur)
+{
+    return syscall(SYS_timerfd_gettime, fd, cur);
+}

From e295dfe6e337be67556d3579895c223d4a13676a Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 10 Dec 2020 12:30:43 +0300
Subject: [PATCH 071/504] Use ucontext for asan

---
 base/common/defines.h                            |  5 +++++
 contrib/boost-cmake/CMakeLists.txt               | 13 ++++++++++++-
 src/Client/Connection.h                          |  2 +-
 src/Client/MultiplexedConnections.h              |  2 +-
 src/Common/Fiber.h                               |  5 +++++
 src/Common/FiberStack.h                          |  1 +
 src/DataStreams/RemoteQueryExecutorReadContext.h |  5 ++++-
 src/IO/ReadBufferFromPocoSocket.h                |  8 +++-----
 src/Interpreters/tests/context.cpp               |  2 +-
 src/Processors/Executors/AsyncTaskQueue.cpp      |  1 +
 10 files changed, 34 insertions(+), 10 deletions(-)
 create mode 100644 src/Common/Fiber.h

diff --git a/base/common/defines.h b/base/common/defines.h
index e4c456796d3..f6d1d5805c3 100644
--- a/base/common/defines.h
+++ b/base/common/defines.h
@@ -61,6 +61,11 @@
 #    endif
 #endif
 
+#if defined(ADDRESS_SANITIZER)
+#    define BOOST_USE_ASAN 1
+#    define BOOST_USE_UCONTEXT 1
+#endif
+
 /// TODO: Strange enough, there is no way to detect UB sanitizer.
 
 /// Explicitly allow undefined behaviour for certain functions. Use it as a function attribute.
diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index cfa41a90b5c..83731c933f7 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -151,7 +151,18 @@ if (NOT EXTERNAL_BOOST_FOUND)
     enable_language(ASM)
     SET(ASM_OPTIONS "-x assembler-with-cpp")
 
-    if (ARCH_ARM)
+    if (SANITIZE)
+        add_compile_definitions(BOOST_USE_UCONTEXT)
+        add_compile_definitions(BOOST_USE_ASAN)
+
+        set (SRCS_CONTEXT
+                ${LIBRARY_DIR}/libs/context/src/fiber.cpp
+                ${LIBRARY_DIR}/libs/context/src/continuation.cpp
+                ${LIBRARY_DIR}/libs/context/src/dummy.cpp
+                ${LIBRARY_DIR}/libs/context/src/execution_context.cpp
+                ${LIBRARY_DIR}/libs/context/src/posix/stack_traits.cpp
+        )
+    elseif (ARCH_ARM)
         set (SRCS_CONTEXT
             ${LIBRARY_DIR}/libs/context/src/asm/jump_arm64_aapcs_elf_gas.S
             ${LIBRARY_DIR}/libs/context/src/asm/make_arm64_aapcs_elf_gas.S
diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index c891f77d1a7..6801c7c54b4 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -189,7 +189,7 @@ public:
     size_t outBytesCount() const { return out ? out->count() : 0; }
     size_t inBytesCount() const { return in ? in->count() : 0; }
 
-    void setFiber(boost::context::fiber * fiber) { in->setFiber(fiber); }
+    void setFiber(Fiber * fiber) { in->setFiber(fiber); }
     Poco::Net::Socket & getSocket() { return in->getSocket(); }
 
 private:
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index 84283c063e9..ab1a2d6ab91 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -67,7 +67,7 @@ public:
     /// Without locking, because sendCancel() does not change the state of the replicas.
     bool hasActiveConnections() const { return active_connection_count > 0; }
 
-    void setFiber(boost::context::fiber * fiber_) { fiber = fiber_; }
+    void setFiber(Fiber * fiber_) { fiber = fiber_; }
     Poco::Net::Socket & getSocket() { return current_connection->getSocket(); }
 
 private:
diff --git a/src/Common/Fiber.h b/src/Common/Fiber.h
new file mode 100644
index 00000000000..e4efc0bdb6a
--- /dev/null
+++ b/src/Common/Fiber.h
@@ -0,0 +1,5 @@
+#pragma once
+#include <common/defines.h>
+#include <boost/context/fiber.hpp>
+
+using Fiber = boost::context::fiber;
diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index 28a38cb7a5e..bd27616a36f 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -1,4 +1,5 @@
 #pragma once
+#include <common/defines.h>
 #include <boost/context/stack_context.hpp>
 #include <Common/Allocator.h>
 
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 17a0f8c5057..026147f0eee 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -1,5 +1,8 @@
 #pragma once
 #include <sys/epoll.h>
+#include <Common/Fiber.h>
+#include <Common/FiberStack.h>
+#include <Common/TimerDescriptor.h>
 
 namespace DB
 {
@@ -174,7 +177,7 @@ public:
         MultiplexedConnections & connections;
         Self & read_context;
 
-        boost::context::fiber operator()(boost::context::fiber && sink) const
+        Fiber operator()(Fiber && sink) const
         {
             try
             {
diff --git a/src/IO/ReadBufferFromPocoSocket.h b/src/IO/ReadBufferFromPocoSocket.h
index d5d1c071210..2311d20cced 100644
--- a/src/IO/ReadBufferFromPocoSocket.h
+++ b/src/IO/ReadBufferFromPocoSocket.h
@@ -4,9 +4,7 @@
 
 #include <IO/ReadBuffer.h>
 #include <IO/BufferWithOwnMemory.h>
-
-#include <boost/context/fiber.hpp>
-
+#include <Common/Fiber.h>
 
 namespace DB
 {
@@ -31,11 +29,11 @@ public:
 
     bool poll(size_t timeout_microseconds);
 
-    void setFiber(boost::context::fiber * fiber_) { fiber = fiber_; }
+    void setFiber(Fiber * fiber_) { fiber = fiber_; }
     Poco::Net::Socket & getSocket() { return socket; }
 
 private:
-    boost::context::fiber * fiber = nullptr;
+    Fiber * fiber = nullptr;
 };
 
 }
diff --git a/src/Interpreters/tests/context.cpp b/src/Interpreters/tests/context.cpp
index 655f93a57e1..9b908e26248 100644
--- a/src/Interpreters/tests/context.cpp
+++ b/src/Interpreters/tests/context.cpp
@@ -1,6 +1,6 @@
 #include <iostream>
 /// #define BOOST_USE_UCONTEXT
-#include <boost/context/fiber.hpp>
+#include <Common/Fiber.h>
 // #include <boost/context/pooled_fixedsize_stack.hpp>
 // #include <boost/context/segmented_stack.hpp>
 #include <Common/Exception.h>
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index f0fd2f33dcb..48db4485912 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -59,6 +59,7 @@ AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & loc
     lock.unlock();
 
     epoll_event event;
+    event.data.ptr = nullptr;
     int num_events = 0;
 
     while (num_events == 0)

From 75ac87c24160fffd60d5f3a395f98f1bfcbb325a Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 10 Dec 2020 15:20:18 +0300
Subject: [PATCH 072/504] Fixing build.

---
 contrib/boost-cmake/CMakeLists.txt            |  4 +-
 docker/packager/unbundled/Dockerfile          |  1 +
 src/Client/MultiplexedConnections.h           |  2 +
 src/DataStreams/RemoteQueryExecutor.cpp       |  6 +-
 .../RemoteQueryExecutorReadContext.h          | 57 ++++++++++++++-----
 src/Processors/Executors/AsyncTaskQueue.cpp   |  2 +-
 src/Processors/Sources/RemoteSource.cpp       | 10 ++++
 src/Processors/Sources/RemoteSource.h         |  6 +-
 8 files changed, 64 insertions(+), 24 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 83731c933f7..d0b3d7f925c 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -15,7 +15,7 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
     )
 
     if(Boost_INCLUDE_DIR AND Boost_FILESYSTEM_LIBRARY AND Boost_FILESYSTEM_LIBRARY AND
-       Boost_PROGRAM_OPTIONS_LIBRARY AND Boost_REGEX_LIBRARY AND Boost_SYSTEM_LIBRARY)
+       Boost_PROGRAM_OPTIONS_LIBRARY AND Boost_REGEX_LIBRARY AND Boost_SYSTEM_LIBRARY AND Boost_CONTEXT_LIBRARY)
 
         set(EXTERNAL_BOOST_FOUND 1)
 
@@ -151,7 +151,7 @@ if (NOT EXTERNAL_BOOST_FOUND)
     enable_language(ASM)
     SET(ASM_OPTIONS "-x assembler-with-cpp")
 
-    if (SANITIZE)
+    if (SANITIZE AND SANITIZE STREQUAL "address")
         add_compile_definitions(BOOST_USE_UCONTEXT)
         add_compile_definitions(BOOST_USE_ASAN)
 
diff --git a/docker/packager/unbundled/Dockerfile b/docker/packager/unbundled/Dockerfile
index 2f501f76e68..c6ebe95d44a 100644
--- a/docker/packager/unbundled/Dockerfile
+++ b/docker/packager/unbundled/Dockerfile
@@ -21,6 +21,7 @@ RUN apt-get update \
         libboost-thread-dev \
         libboost-iostreams-dev \
         libboost-regex-dev \
+        libboost-context-dev \
         zlib1g-dev \
         liblz4-dev \
         libdouble-conversion-dev \
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index ab1a2d6ab91..5f8500ebe6f 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -110,6 +110,8 @@ private:
     mutable std::mutex cancel_mutex;
 
     boost::context::fiber * fiber = nullptr;
+
+    friend class RemoteQueryExecutorReadContext;
 };
 
 }
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 44c6f680e71..247caf8676c 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -223,7 +223,7 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
 
     do
     {
-        if (!read_context->resumeRoutine())
+        if (!read_context->resumeRoutine(was_cancelled_mutex))
             return Block();
 
         if (read_context->is_read_in_progress)
@@ -317,7 +317,7 @@ void RemoteQueryExecutor::finish(std::unique_ptr<ReadContext> * read_context)
         return;
 
     if (read_context && *read_context)
-        (*read_context)->cancel();
+        (*read_context)->cancel(was_cancelled_mutex);
 
     /** If you have not read all the data yet, but they are no longer needed.
       * This may be due to the fact that the data is sufficient (for example, when using LIMIT).
@@ -368,7 +368,7 @@ void RemoteQueryExecutor::cancel(std::unique_ptr<ReadContext> * read_context)
         return;
 
     if (read_context && *read_context)
-        (*read_context)->cancel();
+        (*read_context)->cancel(was_cancelled_mutex);
 
     tryCancel("Cancelling query");
 }
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 026147f0eee..dd135a0be6e 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -24,7 +24,6 @@ public:
 
     std::exception_ptr exception;
     FiberStack<> stack;
-    std::mutex fiber_mutex;
     boost::context::fiber fiber;
 
     Poco::Timespan receive_timeout;
@@ -33,6 +32,7 @@ public:
     TimerDescriptor timer{CLOCK_MONOTONIC, 0};
     int socket_fd = -1;
     int epoll_fd;
+    int pipe_fd[2];
 
     explicit RemoteQueryExecutorReadContext(MultiplexedConnections & connections_) : connections(connections_)
     {
@@ -40,6 +40,18 @@ public:
         if (-1 == epoll_fd)
             throwFromErrno("Cannot create epoll descriptor", ErrorCodes::CANNOT_OPEN_FILE);
 
+        if (-1 == pipe2(pipe_fd, O_NONBLOCK))
+            throwFromErrno("Cannot create pipe", ErrorCodes::CANNOT_OPEN_FILE);
+
+        {
+            epoll_event socket_event;
+            socket_event.events = EPOLLIN | EPOLLPRI;
+            socket_event.data.fd = pipe_fd[0];
+
+            if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, pipe_fd[0], &socket_event))
+                throwFromErrno("Cannot add pipe descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
+        }
+
         {
             epoll_event timer_event;
             timer_event.events = EPOLLIN | EPOLLPRI;
@@ -76,11 +88,11 @@ public:
         receive_timeout = socket.impl()->getReceiveTimeout();
     }
 
-    void checkTimeout() const
+    bool checkTimeout() const
     {
         try
         {
-            checkTimeoutImpl();
+            return checkTimeoutImpl();
         }
         catch (DB::Exception & e)
         {
@@ -89,16 +101,18 @@ public:
         }
     }
 
-    void checkTimeoutImpl() const
+    bool checkTimeoutImpl() const
     {
-        epoll_event events[2];
+        epoll_event events[3];
+        events[0].data.fd = events[1].data.fd = events[2].data.fd = -1;
 
         /// Wait for epoll_fd will not block if it was polled externally.
-        int num_events = epoll_wait(epoll_fd, events, 2, 0);
+        int num_events = epoll_wait(epoll_fd, events, 3, 0);
         if (num_events == -1)
             throwFromErrno("Failed to epoll_wait", ErrorCodes::CANNOT_READ_FROM_SOCKET);
 
         bool is_socket_ready = false;
+        bool is_pipe_alarmed = false;
         bool has_timer_alarm = false;
 
         for (int i = 0; i < num_events; ++i)
@@ -107,14 +121,21 @@ public:
                 is_socket_ready = true;
             if (events[i].data.fd == timer.getDescriptor())
                 has_timer_alarm = true;
+            if (events[i].data.fd == pipe_fd[0])
+                is_pipe_alarmed = true;
         }
 
+        if (is_pipe_alarmed)
+            return false;
+
         if (has_timer_alarm && !is_socket_ready)
         {
             /// Socket receive timeout. Drain it in case or error, or it may be hide by timeout exception.
             timer.drain();
             throw NetException("Timeout exceeded", ErrorCodes::SOCKET_TIMEOUT);
         }
+
+        return true;
     }
 
     void setTimer() const
@@ -126,13 +147,13 @@ public:
             timer.setRelative(receive_timeout);
     }
 
-    bool resumeRoutine()
+    bool resumeRoutine(std::mutex & mutex)
     {
-        if (is_read_in_progress)
-            checkTimeout();
+        if (is_read_in_progress && !checkTimeout())
+            return false;
 
         {
-            std::lock_guard guard(fiber_mutex);
+            std::lock_guard guard(mutex);
             if (!fiber)
                 return false;
 
@@ -159,10 +180,20 @@ public:
         return true;
     }
 
-    void cancel()
+    void cancel(std::mutex & mutex)
     {
-        std::lock_guard guard(fiber_mutex);
+        std::lock_guard guard(mutex);
         boost::context::fiber to_destroy = std::move(fiber);
+
+        uint64_t buf = 0;
+        while (-1 == write(pipe_fd[1], &buf, sizeof(buf)))
+        {
+            if (errno == EAGAIN)
+                break;
+
+            if (errno != EINTR)
+                throwFromErrno("Cannot write to pipe", ErrorCodes::CANNOT_READ_FROM_SOCKET);
+        }
     }
 
     ~RemoteQueryExecutorReadContext()
@@ -186,7 +217,7 @@ public:
                     connections.setFiber(&sink);
 
                     read_context.is_read_in_progress = true;
-                    read_context.packet = connections.receivePacket();
+                    read_context.packet = connections.receivePacketUnlocked();
                     read_context.is_read_in_progress = false;
 
                     sink = std::move(sink).resume();
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index 48db4485912..a760d16c3fc 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -92,7 +92,7 @@ void AsyncTaskQueue::finish()
     uint64_t buf = 0;
     while (-1 == write(pipe_fd[1], &buf, sizeof(buf)))
     {
-        if (errno == EAGAIN || errno == EWOULDBLOCK)
+        if (errno == EAGAIN)
             break;
 
         if (errno != EINTR)
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index b34a06b3790..e7b1872ee66 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -105,6 +105,16 @@ void RemoteSource::onCancel()
     is_async_state = false;
 }
 
+void RemoteSource::onUpdatePorts()
+{
+    if (getPort().isFinished())
+    {
+        was_query_canceled = true;
+        query_executor->finish(&read_context);
+        is_async_state = false;
+    }
+}
+
 
 RemoteTotalsSource::RemoteTotalsSource(RemoteQueryExecutorPtr executor)
     : ISource(executor->getHeader())
diff --git a/src/Processors/Sources/RemoteSource.h b/src/Processors/Sources/RemoteSource.h
index 1d9282ff570..79bf750f897 100644
--- a/src/Processors/Sources/RemoteSource.h
+++ b/src/Processors/Sources/RemoteSource.h
@@ -29,11 +29,7 @@ public:
     void setRowsBeforeLimitCounter(RowsBeforeLimitCounterPtr counter) { rows_before_limit.swap(counter); }
 
     /// Stop reading from stream if output port is finished.
-    void onUpdatePorts() override
-    {
-        if (getPort().isFinished())
-            cancel();
-    }
+    void onUpdatePorts() override;
 
     int schedule() override { return fd; }
 

From db9ad80168936c7f53edaca8775782d07f166e67 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 14 Dec 2020 19:16:08 +0300
Subject: [PATCH 073/504] Fixing tests.

---
 src/DataStreams/RemoteQueryExecutor.cpp       |  2 ++
 src/DataStreams/RemoteQueryExecutor.h         |  3 +-
 .../RemoteQueryExecutorReadContext.h          | 32 ++++++++++++++++---
 src/Processors/Executors/AsyncTaskQueue.cpp   |  4 +++
 src/Processors/Executors/AsyncTaskQueue.h     |  8 +++++
 src/Processors/Executors/PipelineExecutor.cpp |  6 ++++
 src/Processors/Sources/RemoteSource.cpp       |  5 ++-
 7 files changed, 53 insertions(+), 7 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 90e650415bc..e25a69a34cb 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -201,6 +201,7 @@ Block RemoteQueryExecutor::read()
     }
 }
 
+#if defined(OS_LINUX)
 std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext> & read_context)
 {
     if (!sent_query)
@@ -232,6 +233,7 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
     }
     while (true);
 }
+#endif
 
 std::optional<Block> RemoteQueryExecutor::processPacket(Packet packet)
 {
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 51f78816f01..6c668f280a1 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -59,10 +59,11 @@ public:
     /// Read next block of data. Returns empty block if query is finished.
     Block read();
 
+#if defined(OS_LINUX)
     /// Async variant of read. Returns ready block or file descriptor which may be used for polling.
     /// ReadContext is an internal read state. Pass empty ptr first time, reuse created one for every call.
     std::variant<Block, int> read(std::unique_ptr<ReadContext> & read_context);
-
+#endif
     /// Receive all remain packets and finish query.
     /// It should be cancelled after read returned empty block.
     void finish(std::unique_ptr<ReadContext> * read_context = nullptr);
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index dd135a0be6e..18078330f4e 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -1,4 +1,7 @@
 #pragma once
+
+#if defined(OS_LINUX)
+
 #include <sys/epoll.h>
 #include <Common/Fiber.h>
 #include <Common/FiberStack.h>
@@ -25,6 +28,8 @@ public:
     std::exception_ptr exception;
     FiberStack<> stack;
     boost::context::fiber fiber;
+    std::mutex fiber_lock;
+    std::unique_lock<std::mutex> * connection_lock;
 
     Poco::Timespan receive_timeout;
     MultiplexedConnections & connections;
@@ -61,7 +66,9 @@ public:
                 throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
         }
 
-        fiber = boost::context::fiber(std::allocator_arg_t(), stack, Routine{connections, *this});
+        auto routine = Routine{connections, *this, std::unique_lock(connections.cancel_mutex, std::defer_lock)};
+        connection_lock = &routine.connection_lock;
+        fiber = boost::context::fiber(std::allocator_arg_t(), stack, std::move(routine));
     }
 
     void setSocket(Poco::Net::Socket & socket)
@@ -147,17 +154,23 @@ public:
             timer.setRelative(receive_timeout);
     }
 
-    bool resumeRoutine(std::mutex & mutex)
+    bool resumeRoutine()
     {
         if (is_read_in_progress && !checkTimeout())
             return false;
 
         {
-            std::lock_guard guard(mutex);
+            std::lock_guard guard(fiber_lock);
             if (!fiber)
                 return false;
 
+            if (!connection_lock->owns_lock())
+                connection_lock->lock();
+
             fiber = std::move(fiber).resume();
+
+            if (!is_read_in_progress)
+                connection_lock->unlock();
         }
 
         if (exception)
@@ -180,9 +193,9 @@ public:
         return true;
     }
 
-    void cancel(std::mutex & mutex)
+    void cancel()
     {
-        std::lock_guard guard(mutex);
+        std::lock_guard guard(fiber_lock);
         boost::context::fiber to_destroy = std::move(fiber);
 
         uint64_t buf = 0;
@@ -207,6 +220,7 @@ public:
     {
         MultiplexedConnections & connections;
         Self & read_context;
+        std::unique_lock<std::mutex> connection_lock;
 
         Fiber operator()(Fiber && sink) const
         {
@@ -236,5 +250,13 @@ public:
         }
     };
 };
+}
+#else
+namespace DB
+{
+class RemoteQueryExecutorReadContext
+{
+};
 
 }
+#endif
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index a760d16c3fc..eeb7603022d 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -1,4 +1,7 @@
 #include <Processors/Executors/AsyncTaskQueue.h>
+
+#if defined(OS_LINUX)
+
 #include <Common/Exception.h>
 #include <sys/epoll.h>
 #include <unistd.h>
@@ -101,3 +104,4 @@ void AsyncTaskQueue::finish()
 }
 
 }
+#endif
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index 068bdcbd09c..8b6e986cb9c 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -7,6 +7,7 @@
 namespace DB
 {
 
+#if defined(OS_LINUX)
 class AsyncTaskQueue
 {
 public:
@@ -46,5 +47,12 @@ public:
     /// Interrupt waiting.
     void finish();
 };
+#else
+class AsyncTaskQueue
+{
+    bool empty() { return true; }
+    void finish() {}
+};
+#endif
 
 }
diff --git a/src/Processors/Executors/PipelineExecutor.cpp b/src/Processors/Executors/PipelineExecutor.cpp
index c717265248c..dc89821eaa0 100644
--- a/src/Processors/Executors/PipelineExecutor.cpp
+++ b/src/Processors/Executors/PipelineExecutor.cpp
@@ -531,6 +531,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                     break;
                 }
 
+#if defined(OS_LINUX)
                 if (num_threads == 1)
                 {
                     /// If we execute in single thread, wait for async tasks here.
@@ -541,6 +542,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                     node = static_cast<ExecutingGraph::Node *>(res.data);
                     break;
                 }
+#endif
 
                 threads_queue.push(thread_num);
             }
@@ -619,11 +621,13 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                 {
                     std::unique_lock lock(task_queue_mutex);
 
+#if defined(OS_LINUX)
                     while (!async_queue.empty() && !finished)
                     {
                         async_task_queue.addTask(thread_num, async_queue.front(), async_queue.front()->processor->schedule());
                         async_queue.pop();
                     }
+#endif
 
                     while (!queue.empty() && !finished)
                     {
@@ -772,6 +776,7 @@ void PipelineExecutor::executeImpl(size_t num_threads)
             });
         }
 
+#if defined(OS_LINUX)
         {
             /// Wait for async tasks.
             std::unique_lock lock(task_queue_mutex);
@@ -788,6 +793,7 @@ void PipelineExecutor::executeImpl(size_t num_threads)
                 }
             }
         }
+#endif
 
         for (auto & thread : threads)
             if (thread.joinable())
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index e7b1872ee66..5c59c552f69 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -65,7 +65,7 @@ std::optional<Chunk> RemoteSource::tryGenerate()
 
         was_query_sent = true;
     }
-
+#if defined(OS_LINUX)
     auto res = query_executor->read(read_context);
     if (std::holds_alternative<int>(res))
     {
@@ -77,6 +77,9 @@ std::optional<Chunk> RemoteSource::tryGenerate()
     is_async_state = false;
 
     auto block = std::get<Block>(std::move(res));
+#else
+    auto block = query_executor->read();
+#endif
 
     if (!block)
     {

From 3af47d2bda3f4bd025a778ca562651310ce2f8ff Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 14 Dec 2020 19:18:12 +0300
Subject: [PATCH 074/504] Fixing tests.

---
 src/DataStreams/RemoteQueryExecutor.cpp | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index e25a69a34cb..45e50b1a1bd 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -217,7 +217,7 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
 
     do
     {
-        if (!read_context->resumeRoutine(was_cancelled_mutex))
+        if (!read_context->resumeRoutine())
             return Block();
 
         if (read_context->is_read_in_progress)
@@ -312,7 +312,7 @@ void RemoteQueryExecutor::finish(std::unique_ptr<ReadContext> * read_context)
         return;
 
     if (read_context && *read_context)
-        (*read_context)->cancel(was_cancelled_mutex);
+        (*read_context)->cancel();
 
     /** If you have not read all the data yet, but they are no longer needed.
       * This may be due to the fact that the data is sufficient (for example, when using LIMIT).
@@ -363,7 +363,7 @@ void RemoteQueryExecutor::cancel(std::unique_ptr<ReadContext> * read_context)
         return;
 
     if (read_context && *read_context)
-        (*read_context)->cancel(was_cancelled_mutex);
+        (*read_context)->cancel();
 
     tryCancel("Cancelling query");
 }

From f26c0aafc20179a4d7c0a38d5c0f7071d5d8c4d5 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 14 Dec 2020 20:46:27 +0300
Subject: [PATCH 075/504] Fixing tests.

---
 src/DataStreams/RemoteQueryExecutorReadContext.h | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 18078330f4e..e5b973b9d19 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -66,8 +66,9 @@ public:
                 throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
         }
 
-        auto routine = Routine{connections, *this, std::unique_lock(connections.cancel_mutex, std::defer_lock)};
-        connection_lock = &routine.connection_lock;
+        auto lock = std::make_unique<std::unique_lock<std::mutex>>(connections.cancel_mutex, std::defer_lock);
+        auto routine = Routine{connections, *this, std::move(lock)};
+        connection_lock = routine.connection_lock.get();
         fiber = boost::context::fiber(std::allocator_arg_t(), stack, std::move(routine));
     }
 
@@ -220,7 +221,7 @@ public:
     {
         MultiplexedConnections & connections;
         Self & read_context;
-        std::unique_lock<std::mutex> connection_lock;
+        std::unique_ptr<std::unique_lock<std::mutex>> connection_lock;
 
         Fiber operator()(Fiber && sink) const
         {

From 90cc37ae561f90085601f3bf437a543d85f155e3 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Mon, 14 Dec 2020 23:23:29 +0300
Subject: [PATCH 076/504] Document
 execute_merges_on_single_replica_time_threshold setting
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Задокументировал настройку execute_merges_on_single_replica_time_threshold.
---
 .../mergetree-family/replication.md              |  2 ++
 docs/en/operations/settings/settings.md          | 16 +++++++++++++++-
 2 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/docs/en/engines/table-engines/mergetree-family/replication.md b/docs/en/engines/table-engines/mergetree-family/replication.md
index 625869a3cb8..4758ff321e8 100644
--- a/docs/en/engines/table-engines/mergetree-family/replication.md
+++ b/docs/en/engines/table-engines/mergetree-family/replication.md
@@ -113,6 +113,8 @@ You can have any number of replicas of the same data. Yandex.Metrica uses double
 
 The system monitors data synchronicity on replicas and is able to recover after a failure. Failover is automatic (for small differences in data) or semi-automatic (when data differs too much, which may indicate a configuration error).
 
+You can use a special logic for performing merges (for more information, see the documentation for [execute_merges_on_single_replica_time_threshold](../../../operations/settings/settings.md#execute-merges-on-single-replica-time-threshold) setting). 
+
 ## Creating Replicated Tables {#creating-replicated-tables}
 
 The `Replicated` prefix is added to the table engine name. For example:`ReplicatedMergeTree`.
diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index c5e44d4c464..3b2c9a67999 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2410,7 +2410,6 @@ Result:
 {"number":"1"}
 {"number":"2"}
 ```
-
 =======
 ## allow_nullable_key {#allow-nullable-key}
 
@@ -2423,4 +2422,19 @@ Possible values:
 
 Default value: `0`.
 
+## execute_merges_on_single_replica_time_threshold {#execute-merges-on-single-replica-time-threshold}
+
+Enables special logic to perform merges on replicas.
+
+Possible values:
+
+-   Positive integer (in seconds).
+-   0 — Special merge logic is not executed.
+
+Default value: `0`.
+
+**Special logic to perform merges**
+
+Selects one replica to perform the merge on. Sets the time threshold from the start of the merge. Other replicas wait for the merge to finish, then download the result. If the time threshold passes and the selected replica does not perform merges, then the merge is performed manually.
+
 [Original article](https://clickhouse.tech/docs/en/operations/settings/settings/) <!-- hide -->

From d9a1b05910ebd73cabd2711cc43aabdca71483cf Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Mon, 14 Dec 2020 23:37:03 +0300
Subject: [PATCH 077/504] Update settings.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Убрал лишние символы в заголовке настройки allow_nullable_key.
---
 docs/en/operations/settings/settings.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 3b2c9a67999..c60cd2c2c52 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2410,7 +2410,7 @@ Result:
 {"number":"1"}
 {"number":"2"}
 ```
-=======
+
 ## allow_nullable_key {#allow-nullable-key}
 
 Allows using of the [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable)-typed values in a sorting and a primary key for [MergeTree](../../engines/table-engines/mergetree-family/mergetree.md#table_engines-mergetree) tables.

From fc89078f81fc23c1fe5653f4766085da3a374d6e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 00:03:11 +0300
Subject: [PATCH 078/504] Fixing tests.

---
 src/DataStreams/RemoteQueryExecutorReadContext.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index e5b973b9d19..4f777c23bbc 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -197,6 +197,7 @@ public:
     void cancel()
     {
         std::lock_guard guard(fiber_lock);
+        connections.setFiber(nullptr);
         boost::context::fiber to_destroy = std::move(fiber);
 
         uint64_t buf = 0;

From 00de8abb0a0ca8886169b0dc950bd250130473fe Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 00:07:10 +0300
Subject: [PATCH 079/504] Fixing tests.

---
 src/Client/MultiplexedConnections.cpp            | 12 +++++++++---
 src/DataStreams/RemoteQueryExecutorReadContext.h |  1 -
 2 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/src/Client/MultiplexedConnections.cpp b/src/Client/MultiplexedConnections.cpp
index 3f0c5a29946..30311f40512 100644
--- a/src/Client/MultiplexedConnections.cpp
+++ b/src/Client/MultiplexedConnections.cpp
@@ -252,10 +252,16 @@ Packet MultiplexedConnections::receivePacketUnlocked()
     if (fiber)
         current_connection->setFiber(fiber);
 
-    Packet packet = current_connection->receivePacket();
+    Packet packet;
+    {
+        SCOPE_EXIT(
+        {
+           fiber = nullptr;
+           current_connection->setFiber(fiber);
+        });
 
-    fiber = nullptr;
-    current_connection->setFiber(fiber);
+        packet = current_connection->receivePacket();
+    }
 
     switch (packet.type)
     {
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 4f777c23bbc..e5b973b9d19 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -197,7 +197,6 @@ public:
     void cancel()
     {
         std::lock_guard guard(fiber_lock);
-        connections.setFiber(nullptr);
         boost::context::fiber to_destroy = std::move(fiber);
 
         uint64_t buf = 0;

From 6d5b23de672f8ddc3e0857ab13f82d8bc0c96f46 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 15 Dec 2020 03:36:03 +0300
Subject: [PATCH 080/504] something works

---
 src/Core/Block.cpp                            |  16 ++
 src/Interpreters/ActionsDAG.cpp               |   2 +-
 src/Interpreters/ExpressionActions.cpp        |   4 +
 src/Interpreters/ExpressionAnalyzer.cpp       |   6 +-
 src/Interpreters/InterpreterSelectQuery.cpp   |  14 +-
 src/Processors/QueryPlan/WindowStep.cpp       | 102 ++++++------
 src/Processors/QueryPlan/WindowStep.h         |  15 +-
 .../Transforms/ExpressionTransform.h          |   1 -
 src/Processors/Transforms/WindowTransform.cpp | 149 ++++++++++++++++--
 src/Processors/Transforms/WindowTransform.h   |  36 ++++-
 .../RocksDB/StorageEmbeddedRocksDB.cpp        |   2 +
 11 files changed, 258 insertions(+), 89 deletions(-)

diff --git a/src/Core/Block.cpp b/src/Core/Block.cpp
index cd2855739e2..014fa001588 100644
--- a/src/Core/Block.cpp
+++ b/src/Core/Block.cpp
@@ -56,6 +56,10 @@ void Block::insert(size_t position, const ColumnWithTypeAndName & elem)
 
     index_by_name.emplace(elem.name, position);
     data.emplace(data.begin() + position, elem);
+
+//    fmt::print(stderr, "block: {}\n", dumpStructure());
+//    fmt::print(stderr, "(1) insert column {} at \n{}\n", elem.name,
+//        StackTrace().toString());
 }
 
 void Block::insert(size_t position, ColumnWithTypeAndName && elem)
@@ -70,17 +74,29 @@ void Block::insert(size_t position, ColumnWithTypeAndName && elem)
 
     index_by_name.emplace(elem.name, position);
     data.emplace(data.begin() + position, std::move(elem));
+
+//    fmt::print(stderr, "block: {}\n", dumpStructure());
+//    fmt::print(stderr, "(2) insert column {} at \n{}\n", elem.name,
+//        StackTrace().toString());
 }
 
 
 void Block::insert(const ColumnWithTypeAndName & elem)
 {
+//    fmt::print(stderr, "block: {}\n", dumpStructure());
+//    fmt::print(stderr, "(3) insert column {} at \n{}\n", elem.name,
+//        StackTrace().toString());
+
     index_by_name.emplace(elem.name, data.size());
     data.emplace_back(elem);
 }
 
 void Block::insert(ColumnWithTypeAndName && elem)
 {
+//    fmt::print(stderr, "block: {}\n", dumpStructure());
+//    fmt::print(stderr, "(4) insert column {} at \n{}\n", elem.name,
+//        StackTrace().toString());
+
     index_by_name.emplace(elem.name, data.size());
     data.emplace_back(std::move(elem));
 }
diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index 75915bdad1a..cd6fab40aaf 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -624,7 +624,7 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
             {
                 auto & input = inputs[res_elem.name];
                 if (input.empty())
-                    throw Exception("Cannot find column " + backQuoteIfNeed(res_elem.name) + " in source stream",
+                    throw Exception("Cannot find column '" + backQuoteIfNeed(res_elem.name) + "' in source stream",
                                     ErrorCodes::THERE_IS_NO_COLUMN);
 
                 src_node = actions_dag->inputs[input.front()];
diff --git a/src/Interpreters/ExpressionActions.cpp b/src/Interpreters/ExpressionActions.cpp
index f0dcf830c82..4c2ec0e0dc8 100644
--- a/src/Interpreters/ExpressionActions.cpp
+++ b/src/Interpreters/ExpressionActions.cpp
@@ -373,6 +373,10 @@ static void executeAction(const ExpressionActions::Action & action, ExecutionCon
             res_column.column = action.node->column->cloneResized(num_rows);
             res_column.type = action.node->result_type;
             res_column.name = action.node->result_name;
+
+
+            fmt::print(stderr, "execute column action {} at\n{}\n",
+                action.toString(), StackTrace().toString());
             break;
         }
 
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 0d54f75c588..802004b7ad6 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1024,14 +1024,14 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
         getRootActionsNoMakeSet(f.function_node->arguments,
             true /* no subqueries */, step.actions());
 
-        // Add column with window function name and value "1".
+        // Add empty INPUT with window function name.
         // It is an aggregate function, so it won't be added by getRootActions.
         ColumnWithTypeAndName col;
         col.type = std::make_shared<DataTypeInt64>();
-        col.column = col.type->createColumnConst(1 /* size */, UInt64(1) /* field */);
+        col.column = col.type->createColumn();
         col.name = f.column_name;
 
-        step.actions()->addColumn(col);
+        step.actions()->addInput(col);
     }
 
     // /*
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 2d6cb64bb7e..446072602e0 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1788,20 +1788,10 @@ void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
             + w.window_name + "'");
         query_plan.addStep(std::move(merging_sorted));
 
-        // Add column with window function name and value "1".
-        ColumnWithTypeAndName col;
-        col.type = std::make_shared<DataTypeInt64>();
-        col.column = col.type->createColumnConst(1 /* size */, UInt64(1) /* field */);
-        col.name = f.column_name;
-
-        ActionsDAGPtr window_dag = std::make_shared<ActionsDAG>();
-        window_dag->addColumn(col);
-
-        fmt::print(stderr, "window dag: {}\n", window_dag->dumpDAG());
-
         auto window_step = std::make_unique<WindowStep>(
             query_plan.getCurrentDataStream(),
-            window_dag);
+            w,
+            std::vector<WindowFunctionDescription>(1, f));
         window_step->setStepDescription("Window step for function '"
             + f.column_name + "'");
 
diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
index 2db6bd3d416..495513c1a2a 100644
--- a/src/Processors/QueryPlan/WindowStep.cpp
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -1,6 +1,7 @@
 #include <Processors/QueryPlan/WindowStep.h>
 
 #include <Processors/Transforms/WindowTransform.h>
+#include <Processors/Transforms/ExpressionTransform.h>
 #include <Processors/QueryPipeline.h>
 #include <Interpreters/ExpressionActions.h>
 #include <IO/Operators.h>
@@ -8,18 +9,18 @@
 namespace DB
 {
 
-static ITransformingStep::Traits getTraits(const ActionsDAGPtr & actions)
+static ITransformingStep::Traits getTraits()
 {
     return ITransformingStep::Traits
     {
         {
-            .preserves_distinct_columns = !actions->hasArrayJoin(),
+            .preserves_distinct_columns = true,
             .returns_single_stream = false,
             .preserves_number_of_streams = true,
-            .preserves_sorting = !actions->hasArrayJoin(),
+            .preserves_sorting = true,
         },
         {
-            .preserves_number_of_rows = !actions->hasArrayJoin(),
+            .preserves_number_of_rows = true
         }
     };
 }
@@ -40,71 +41,72 @@ static ITransformingStep::Traits getJoinTraits()
     };
 }
 
+static Block addWindowFunctionColumns(const Block & block,
+    std::vector<WindowFunctionDescription> window_functions)
+{
+    fmt::print(stderr, "input header: {}\n", block.dumpStructure());
+
+    //auto result = block.cloneWithoutColumns();
+    auto result = block;
+
+    fmt::print(stderr, "header after clone: {}\n", result.dumpStructure());
+
+    for (const auto & f : window_functions)
+    {
+        ColumnWithTypeAndName column_with_type;
+        column_with_type.name = f.column_name;
+        column_with_type.type = f.aggregate_function->getReturnType();
+        column_with_type.column = column_with_type.type->createColumn();
+
+        result.insert(column_with_type);
+    }
+
+    fmt::print(stderr, "header after insert: {}\n", result.dumpStructure());
+
+    return result;
+}
+
 WindowStep::WindowStep(const DataStream & input_stream_,
-        ActionsDAGPtr actions_dag_)
+        const WindowDescription & window_description_,
+        const std::vector<WindowFunctionDescription> & window_functions_)
     : ITransformingStep(
         input_stream_,
-        Transform::transformHeader(input_stream_.header,
-            std::make_shared<ExpressionActions>(actions_dag_)),
-        getTraits(actions_dag_))
-    , actions_dag(std::move(actions_dag_))
+            addWindowFunctionColumns(input_stream_.header, window_functions_),
+        getTraits())
+    , window_description(window_description_)
+    , window_functions(window_functions_)
+    , input_header(input_stream_.header)
 {
     /// Some columns may be removed by expression.
     updateDistinctColumns(output_stream->header, output_stream->distinct_columns);
 }
 
-void WindowStep::updateInputStream(DataStream input_stream, bool keep_header)
-{
-    Block out_header = keep_header
-        ? std::move(output_stream->header)
-        : Transform::transformHeader(input_stream.header,
-            std::make_shared<ExpressionActions>(actions_dag));
-
-    output_stream = createOutputStream(
-            input_stream,
-            std::move(out_header),
-            getDataStreamTraits());
-
-    input_streams.clear();
-    input_streams.emplace_back(std::move(input_stream));
-}
-
 void WindowStep::transformPipeline(QueryPipeline & pipeline)
 {
-    auto expression = std::make_shared<ExpressionActions>(actions_dag);
-    pipeline.addSimpleTransform([&](const Block & header)
+    pipeline.addSimpleTransform([&](const Block & /*header*/)
     {
-        return std::make_shared<Transform>(header, expression);
+        return std::make_shared<Transform>(input_header,
+            output_stream->header, window_description, window_functions);
     });
 
-    if (!blocksHaveEqualStructure(pipeline.getHeader(), output_stream->header))
-    {
-        auto convert_actions_dag = ActionsDAG::makeConvertingActions(
-                pipeline.getHeader().getColumnsWithTypeAndName(),
-                output_stream->header.getColumnsWithTypeAndName(),
-                ActionsDAG::MatchColumnsMode::Name);
-        auto convert_actions = std::make_shared<ExpressionActions>(convert_actions_dag);
-
-        pipeline.addSimpleTransform([&](const Block & header)
-        {
-            return std::make_shared<ExpressionTransform>(header, convert_actions);
-        });
-    }
+    assertBlocksHaveEqualStructure(pipeline.getHeader(), output_stream->header,
+        "WindowStep transform for '" + window_description.window_name + "'");
 }
 
 void WindowStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
-    bool first = true;
+    (void) prefix;
+    //bool first = true;
 
-    auto expression = std::make_shared<ExpressionActions>(actions_dag);
-    for (const auto & action : expression->getActions())
-    {
-        settings.out << prefix << (first ? "Actions: "
-                                         : "         ");
-        first = false;
-        settings.out << action.toString() << '\n';
-    }
+    //auto expression = std::make_shared<ExpressionActions>(actions_dag);
+    //for (const auto & action : expression->getActions())
+    //{
+    //    settings.out << prefix << (first ? "Actions: "
+    //                                     : "         ");
+    //    first = false;
+    //    settings.out << action.toString() << '\n';
+    //}
 }
 
 }
diff --git a/src/Processors/QueryPlan/WindowStep.h b/src/Processors/QueryPlan/WindowStep.h
index 4b67e9563db..5bdb14ae5fe 100644
--- a/src/Processors/QueryPlan/WindowStep.h
+++ b/src/Processors/QueryPlan/WindowStep.h
@@ -1,6 +1,8 @@
 #pragma once
 #include <Processors/QueryPlan/ITransformingStep.h>
 
+#include <Interpreters/AggregateDescription.h>
+
 namespace DB
 {
 
@@ -14,19 +16,20 @@ class WindowStep : public ITransformingStep
 public:
     using Transform = WindowTransform;
 
-    explicit WindowStep(const DataStream & input_stream_, ActionsDAGPtr actions_dag_);
+    explicit WindowStep(const DataStream & input_stream_,
+            const WindowDescription & window_description_,
+            const std::vector<WindowFunctionDescription> & window_functions_);
+
     String getName() const override { return "Expression"; }
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void updateInputStream(DataStream input_stream, bool keep_header);
-
     void describeActions(FormatSettings & settings) const override;
 
-    const ActionsDAGPtr & getExpression() const { return actions_dag; }
-
 private:
-    ActionsDAGPtr actions_dag;
+    WindowDescription window_description;
+    std::vector<WindowFunctionDescription> window_functions;
+    Block input_header;
 };
 
 }
diff --git a/src/Processors/Transforms/ExpressionTransform.h b/src/Processors/Transforms/ExpressionTransform.h
index eed8bf089cd..c0c0a63a6a8 100644
--- a/src/Processors/Transforms/ExpressionTransform.h
+++ b/src/Processors/Transforms/ExpressionTransform.h
@@ -21,7 +21,6 @@ public:
 
     String getName() const override
     {
-        fmt::print(stderr, "et getname at \n{}\n", StackTrace().toString());
         return "ExpressionTransform";
     }
 
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index f4bc660b940..1e166fe1aff 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -2,30 +2,157 @@
 
 #include <Interpreters/ExpressionActions.h>
 
+#include <Common/Arena.h>
+
 namespace DB
 {
 
-Block WindowTransform::transformHeader(Block header, const ExpressionActionsPtr & expression)
+WindowTransform::WindowTransform(const Block & input_header_,
+        const Block & output_header_,
+        const WindowDescription & window_description_,
+        const std::vector<WindowFunctionDescription> & window_function_descriptions
+        )
+    // FIXME this is where the output column is added
+    : ISimpleTransform(input_header_, output_header_,
+        false /* skip_empty_chunks */)
+    , input_header(input_header_)
+    , window_description(window_description_)
 {
-    size_t num_rows = header.rows();
-    expression->execute(header, num_rows, true);
-    return header;
+    fmt::print(stderr, "input header {}\n", input_header.dumpStructure());
+
+    workspaces.reserve(window_function_descriptions.size());
+    for (size_t i = 0; i < window_function_descriptions.size(); ++i)
+    {
+        WindowFunctionWorkspace workspace;
+        workspace.window_function = window_function_descriptions[i];
+
+        const auto & aggregate_function
+            = workspace.window_function.aggregate_function;
+        if (!arena && aggregate_function->allocatesMemoryInArena())
+        {
+            arena = std::make_unique<Arena>();
+        }
+
+        workspace.argument_column_indices.reserve(
+            workspace.window_function.argument_names.size());
+        workspace.argument_columns.reserve(
+            workspace.window_function.argument_names.size());
+        for (const auto & argument_name : workspace.window_function.argument_names)
+        {
+            workspace.argument_column_indices.push_back(
+                input_header.getPositionByName(argument_name));
+
+            fmt::print(stderr,
+                "window function '{}' argument column '{}' at '{}'\n",
+                workspace.window_function.column_name,
+                argument_name,
+                workspace.argument_column_indices.back());
+
+        }
+
+        workspace.aggregate_function_state.reset(aggregate_function->sizeOfData(),
+            aggregate_function->alignOfData());
+        aggregate_function->create(workspace.aggregate_function_state.data());
+
+        workspaces.push_back(std::move(workspace));
+    }
+
+    partition_by_indices.reserve(window_description.partition_by.size());
+    for (const auto & column : window_description.partition_by)
+    {
+        partition_by_indices.push_back(
+            input_header.getPositionByName(column.column_name));
+    }
 }
 
-
-WindowTransform::WindowTransform(const Block & header_,
-        ExpressionActionsPtr expression_)
-    : ISimpleTransform(header_, transformHeader(header_, expression_), false)
-    , expression(std::move(expression_))
+WindowTransform::~WindowTransform()
 {
+    // Some states may be not created yet if the creation failed.
+    for (size_t i = 0; i < workspaces.size(); i++)
+    {
+        workspaces[i].window_function.aggregate_function->destroy(
+            workspaces[i].aggregate_function_state.data());
+    }
 }
 
 void WindowTransform::transform(Chunk & chunk)
 {
-    size_t num_rows = chunk.getNumRows();
+    const size_t num_rows = chunk.getNumRows();
     auto block = getInputPort().getHeader().cloneWithColumns(chunk.detachColumns());
 
-    expression->execute(block, num_rows);
+    fmt::print(stderr, "block before window transform:\n{}\n", block.dumpStructure());
+
+    for (auto & workspace : workspaces)
+    {
+        workspace.argument_columns.clear();
+        for (const auto column_index : workspace.argument_column_indices)
+        {
+            fmt::print(stderr, "argument column index '{}'\n", column_index);
+            workspace.argument_columns.push_back(
+                block.getColumns()[column_index].get());
+        }
+    }
+
+    for (auto & ws : workspaces)
+    {
+        const auto & f = ws.window_function;
+        const auto * a = f.aggregate_function.get();
+
+        //*
+        ColumnWithTypeAndName column_with_type;
+        column_with_type.name = f.column_name;
+        column_with_type.type = a->getReturnType();
+        auto c = column_with_type.type->createColumn();
+        column_with_type.column.reset(c.get());
+
+        size_t partition_start = 0;
+        for (size_t row = 0; row < num_rows; row++)
+        {
+            // Check whether the new partition has started and reinitialize the
+            // aggregate function states.
+            if (row > 0)
+            {
+                for (const size_t column_index : partition_by_indices)
+                {
+                    const auto * column = block.getColumns()[column_index].get();
+                    if (column->compareAt(row, row - 1, *column,
+                        1 /* nan_direction_hint */) != 0)
+                    {
+                        ws.window_function.aggregate_function->destroy(
+                            ws.aggregate_function_state.data());
+                        ws.window_function.aggregate_function->create(
+                            ws.aggregate_function_state.data());
+                        break;
+                    }
+                }
+            }
+
+            // Update the aggregate function state and save the result.
+            a->add(ws.aggregate_function_state.data(),
+                ws.argument_columns.data(),
+                row,
+                arena.get());
+
+            a->insertResultInto(ws.aggregate_function_state.data(),
+                *c,
+                arena.get());
+        }
+
+        block.insert(column_with_type);
+        /*/
+        auto & column_with_type = block.getByName(f.column_name);
+        auto c = IColumn::mutate(std::move(column_with_type.column));
+
+        for (size_t i = 0; i < num_rows; i++)
+        {
+            c->insert(UInt64(i));
+        }
+
+        column_with_type.column.reset(c.get());
+        //*/
+    }
+
+    fmt::print(stderr, "block after window transform:\n{}\n", block.dumpStructure());
 
     chunk.setColumns(block.getColumns(), num_rows);
 }
diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
index 08648905eda..58b1ed13ae0 100644
--- a/src/Processors/Transforms/WindowTransform.h
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -1,12 +1,27 @@
 #pragma once
 #include <Processors/ISimpleTransform.h>
 
+#include <Interpreters/AggregateDescription.h>
+
+#include <Common/AlignedBuffer.h>
+
 namespace DB
 {
 
 class ExpressionActions;
 using ExpressionActionsPtr = std::shared_ptr<ExpressionActions>;
 
+class Arena;
+
+struct WindowFunctionWorkspace
+{
+    WindowFunctionDescription window_function;
+    AlignedBuffer aggregate_function_state;
+    std::vector<size_t> argument_column_indices;
+    // Be careful, this is per-chunk.
+    std::vector<const IColumn *> argument_columns;
+};
+
 /** Executes a certain expression over the block.
   * The expression consists of column identifiers from the block, constants, common functions.
   * For example: hits * 2 + 3, url LIKE '%yandex%'
@@ -16,8 +31,12 @@ class WindowTransform : public ISimpleTransform
 {
 public:
     WindowTransform(
-            const Block & header_,
-            ExpressionActionsPtr expression_);
+            const Block & input_header,
+            const Block & output_header,
+            const WindowDescription & window_description,
+            const std::vector<WindowFunctionDescription> & window_functions);
+
+    ~WindowTransform() override;
 
     String getName() const override
     {
@@ -26,11 +45,18 @@ public:
 
     static Block transformHeader(Block header, const ExpressionActionsPtr & expression);
 
-protected:
     void transform(Chunk & chunk) override;
 
-private:
-    ExpressionActionsPtr expression;
+public:
+    Block input_header;
+
+    WindowDescription window_description;
+    std::vector<size_t> partition_by_indices;
+
+    std::vector<WindowFunctionWorkspace> workspaces;
+
+    std::unique_ptr<Arena> arena;
+    std::vector<AlignedBuffer> aggregate_function_data;
 };
 
 }
diff --git a/src/Storages/RocksDB/StorageEmbeddedRocksDB.cpp b/src/Storages/RocksDB/StorageEmbeddedRocksDB.cpp
index ae28205d07b..80b25793806 100644
--- a/src/Storages/RocksDB/StorageEmbeddedRocksDB.cpp
+++ b/src/Storages/RocksDB/StorageEmbeddedRocksDB.cpp
@@ -3,6 +3,8 @@
 #include <Storages/RocksDB/EmbeddedRocksDBBlockInputStream.h>
 
 #include <DataTypes/DataTypesNumber.h>
+
+#include <Storages/SelectQueryInfo.h>
 #include <Storages/StorageFactory.h>
 
 #include <Parsers/ASTSelectQuery.h>

From 35dda5851c9c5f80d6b620b0ffc81144300a3cb6 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 10:02:15 +0300
Subject: [PATCH 081/504] Try remove connection lock.

---
 src/DataStreams/RemoteQueryExecutorReadContext.h | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index e5b973b9d19..5a72f4b5963 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -29,7 +29,7 @@ public:
     FiberStack<> stack;
     boost::context::fiber fiber;
     std::mutex fiber_lock;
-    std::unique_lock<std::mutex> * connection_lock;
+    // std::unique_lock<std::mutex> * connection_lock;
 
     Poco::Timespan receive_timeout;
     MultiplexedConnections & connections;
@@ -68,7 +68,7 @@ public:
 
         auto lock = std::make_unique<std::unique_lock<std::mutex>>(connections.cancel_mutex, std::defer_lock);
         auto routine = Routine{connections, *this, std::move(lock)};
-        connection_lock = routine.connection_lock.get();
+        // connection_lock = routine.connection_lock.get();
         fiber = boost::context::fiber(std::allocator_arg_t(), stack, std::move(routine));
     }
 
@@ -165,13 +165,13 @@ public:
             if (!fiber)
                 return false;
 
-            if (!connection_lock->owns_lock())
-                connection_lock->lock();
+//            if (!connection_lock->owns_lock())
+//                connection_lock->lock();
 
             fiber = std::move(fiber).resume();
 
-            if (!is_read_in_progress)
-                connection_lock->unlock();
+//            if (!is_read_in_progress)
+//                connection_lock->unlock();
         }
 
         if (exception)

From c4e80ccfc588cb6d347bd3834f455b5c434302e2 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 11:27:42 +0300
Subject: [PATCH 082/504] Try fix special builds.

---
 src/CMakeLists.txt                      | 1 +
 src/DataStreams/RemoteQueryExecutor.cpp | 2 --
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 4771afd4c24..f3afbeb61dc 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -428,6 +428,7 @@ if (USE_ROCKSDB)
     dbms_target_include_directories(SYSTEM BEFORE PUBLIC ${ROCKSDB_INCLUDE_DIR})
 endif()
 
+target_link_libraries (clickhouse_common_io PRIVATE _boost_context)
 dbms_target_link_libraries(PRIVATE _boost_context)
 
 if (ENABLE_TESTS AND USE_GTEST)
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 45e50b1a1bd..b31b41a36f2 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -14,8 +14,6 @@
 #include <IO/ConnectionTimeoutsContext.h>
 #include <Common/FiberStack.h>
 
-#include <sys/epoll.h>
-
 namespace DB
 {
 

From d070920c72b69a5aa177d5cc76a5d61d73033589 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 12:22:14 +0300
Subject: [PATCH 083/504] Update FiberStack

---
 src/Common/FiberStack.h | 50 +++++++++++++++++++++++++++++++----------
 1 file changed, 38 insertions(+), 12 deletions(-)

diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index bd27616a36f..d73270c23d0 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -1,34 +1,57 @@
 #pragma once
 #include <common/defines.h>
 #include <boost/context/stack_context.hpp>
-#include <Common/Allocator.h>
+#include <Common/formatReadable.h>
+
+#include <sys/time.h>
+#include <sys/resource.h>
+#include <sys/mman.h>
 
 #if defined(BOOST_USE_VALGRIND)
 #include <valgrind/valgrind.h>
 #endif
 
+namespace DB::ErrorCodes
+{
+    extern const int CANNOT_ALLOCATE_MEMORY;
+}
+
 /// This is an implementation of allocator for fiber stack.
-/// It uses internal allocator, so we track memory usage. It is the main reason why this class is needed.
-/// The reference implementations are pooled_fixedsize_stack and protected_fixedsize_stack from boost::context.
-template <typename TAllocator = Allocator<false, false>>
+/// The reference implementation is protected_fixedsize_stack from boost::context.
+/// This implementation additionally track memory usage. It is the main reason why it is needed.
 class FiberStack
 {
 private:
     size_t stack_size;
+    size_t page_size = 0;
 public:
-    /// 8MB of memory per fiber stack may seem too expensive. It is indeed.
-    /// The reason is that current (patched) libunwind needs > 4MB of stack memory to unwind stack.
-    /// If we allocate less memory, any thrown exception inside fiber will cause segfault.
-    static constexpr size_t default_stack_size = 8 * 1024 * 1024;
+    static constexpr size_t default_stack_size = 64 * 1024;
 
-    explicit FiberStack(size_t stack_size_ = default_stack_size) : stack_size(stack_size_) {}
+    explicit FiberStack(size_t stack_size_ = default_stack_size) : stack_size(stack_size_)
+    {
+        page_size = ::sysconf( _SC_PAGESIZE);
+    }
 
     boost::context::stack_context allocate()
     {
-        void * vp = TAllocator().alloc(stack_size);
+        size_t num_pages = 1 + (stack_size - 1) / page_size;
+        size_t num_bytes = (num_pages + 1) * page_size; /// Add one page at bottom that will be used as guard-page
+
+        void * vp = ::mmap( nullptr, num_bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+        if (MAP_FAILED == vp)
+            DB::throwFromErrno(fmt::format("FiberStack: Cannot mmap {}.", ReadableSize(num_bytes)), DB::ErrorCodes::CANNOT_ALLOCATE_MEMORY);
+
+        if (-1 == ::mprotect( vp, page_size, PROT_NONE))
+        {
+            ::munmap(vp, num_bytes);
+            DB::throwFromErrno("FiberStack: cannot protect guard page", DB::ErrorCodes::CANNOT_ALLOCATE_MEMORY);
+        }
+
+        /// Do not count guard page in memory usage.
+        CurrentMemoryTracker::alloc(num_pages * page_size);
 
         boost::context::stack_context sctx;
-        sctx.size = stack_size;
+        sctx.size = num_bytes;
         sctx.sp = static_cast< char * >(vp) + sctx.size;
 #if defined(BOOST_USE_VALGRIND)
         sctx.valgrind_stack_id = VALGRIND_STACK_REGISTER(sctx.sp, vp);
@@ -42,6 +65,9 @@ public:
         VALGRIND_STACK_DEREGISTER(sctx.valgrind_stack_id);
 #endif
         void * vp = static_cast< char * >(sctx.sp) - sctx.size;
-        TAllocator().free(vp, stack_size);
+        ::munmap( vp, sctx.size);
+
+        /// Do not count guard page in memory usage.
+        CurrentMemoryTracker::free(sctx.size - page_size);
     }
 };

From 716e0b472b8e0f2b4d2192a853a91999160b1599 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 13:08:13 +0300
Subject: [PATCH 084/504] Update FiberStack

---
 src/DataStreams/RemoteQueryExecutorReadContext.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 5a72f4b5963..cedfc805ee1 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -26,7 +26,7 @@ public:
     Packet packet;
 
     std::exception_ptr exception;
-    FiberStack<> stack;
+    FiberStack stack;
     boost::context::fiber fiber;
     std::mutex fiber_lock;
     // std::unique_lock<std::mutex> * connection_lock;
@@ -257,6 +257,8 @@ namespace DB
 {
 class RemoteQueryExecutorReadContext
 {
+public:
+    void cancel() {}
 };
 
 }

From 24c390eb49e31dda31949cd1f2557966171e8bc7 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 15:07:31 +0300
Subject: [PATCH 085/504] Fixing tests.

---
 src/Processors/Executors/PipelineExecutor.cpp | 6 +++++-
 src/Processors/Executors/PipelineExecutor.h   | 1 +
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/src/Processors/Executors/PipelineExecutor.cpp b/src/Processors/Executors/PipelineExecutor.cpp
index dc89821eaa0..1646a1a01fe 100644
--- a/src/Processors/Executors/PipelineExecutor.cpp
+++ b/src/Processors/Executors/PipelineExecutor.cpp
@@ -502,6 +502,9 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                     node = context->async_tasks.front();
                     context->async_tasks.pop();
                     --num_waiting_async_tasks;
+
+                    if (context->async_tasks.empty())
+                        context->has_async_tasks = false;
                 }
                 else if (!task_queue.empty())
                     node = task_queue.pop(thread_num);
@@ -610,7 +613,7 @@ void PipelineExecutor::executeStepImpl(size_t thread_num, size_t num_threads, st
                 node = nullptr;
 
                 /// Take local task from queue if has one.
-                if (!queue.empty())
+                if (!queue.empty() && !context->has_async_tasks)
                 {
                     node = queue.front();
                     queue.pop();
@@ -784,6 +787,7 @@ void PipelineExecutor::executeImpl(size_t num_threads)
             {
                 auto * node = static_cast<ExecutingGraph::Node *>(task.data);
                 executor_contexts[task.thread_num]->async_tasks.push(node);
+                executor_contexts[task.thread_num]->has_async_tasks = true;
                 ++num_waiting_async_tasks;
 
                 if (threads_queue.has(task.thread_num))
diff --git a/src/Processors/Executors/PipelineExecutor.h b/src/Processors/Executors/PipelineExecutor.h
index 1ddb6cbe8b9..16ace7b40bb 100644
--- a/src/Processors/Executors/PipelineExecutor.h
+++ b/src/Processors/Executors/PipelineExecutor.h
@@ -98,6 +98,7 @@ private:
         std::list<ExpandPipelineTask> task_list;
 
         std::queue<ExecutingGraph::Node *> async_tasks;
+        std::atomic_bool has_async_tasks = false;
 
         std::condition_variable condvar;
         std::mutex mutex;

From e7bce63d3a7c41736b8994ef1fcb2bee8d02e1a9 Mon Sep 17 00:00:00 2001
From: fastio <pengjian.uestc@gmail.com>
Date: Mon, 7 Dec 2020 17:30:47 +0800
Subject: [PATCH 086/504] hints for column names

---
 src/Interpreters/TreeRewriter.cpp | 11 +++---
 src/Storages/IStorage.cpp         | 61 +++++++++++++++++--------------
 src/Storages/IStorage.h           |  5 ++-
 3 files changed, 42 insertions(+), 35 deletions(-)

diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 1ad22efc3fc..2dc9a4ee818 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -624,14 +624,13 @@ void TreeRewriterResult::collectUsedColumns(const ASTPtr & query, bool is_select
         for (const auto & name : columns_context.requiredColumns())
             ss << " '" << name << "'";
 
-        if (!source_column_names.empty())
+        ss << ", maybe you meant: ";
+        for (const auto & name : columns_context.requiredColumns())
         {
-            ss << ", source columns:";
-            for (const auto & name : source_column_names)
-                ss << " '" << name << "'";
+            auto hints = storage->getHints(name);
+            if (!hints.empty())
+                ss << " '" << toString(hints) << " '";
         }
-        else
-            ss << ", no source columns";
 
         if (columns_context.has_table_join)
         {
diff --git a/src/Storages/IStorage.cpp b/src/Storages/IStorage.cpp
index 94f37254b91..e0863094f5f 100644
--- a/src/Storages/IStorage.cpp
+++ b/src/Storages/IStorage.cpp
@@ -17,7 +17,6 @@
 
 namespace DB
 {
-
 namespace ErrorCodes
 {
     extern const int TABLE_IS_DROPPED;
@@ -32,17 +31,18 @@ bool IStorage::isVirtualColumn(const String & column_name, const StorageMetadata
 }
 
 RWLockImpl::LockHolder IStorage::tryLockTimed(
-        const RWLock & rwlock, RWLockImpl::Type type, const String & query_id, const std::chrono::milliseconds & acquire_timeout) const
+    const RWLock & rwlock, RWLockImpl::Type type, const String & query_id, const std::chrono::milliseconds & acquire_timeout) const
 {
     auto lock_holder = rwlock->getLock(type, query_id, acquire_timeout);
     if (!lock_holder)
     {
         const String type_str = type == RWLockImpl::Type::Read ? "READ" : "WRITE";
         throw Exception(
-                type_str + " locking attempt on \"" + getStorageID().getFullTableName() +
-                "\" has timed out! (" + std::to_string(acquire_timeout.count()) + "ms) "
-                "Possible deadlock avoided. Client should retry.",
-                ErrorCodes::DEADLOCK_AVOIDED);
+            type_str + " locking attempt on \"" + getStorageID().getFullTableName() + "\" has timed out! ("
+                + std::to_string(acquire_timeout.count())
+                + "ms) "
+                  "Possible deadlock avoided. Client should retry.",
+            ErrorCodes::DEADLOCK_AVOIDED);
     }
     return lock_holder;
 }
@@ -82,26 +82,26 @@ TableExclusiveLockHolder IStorage::lockExclusively(const String & query_id, cons
 }
 
 Pipe IStorage::read(
-        const Names & /*column_names*/,
-        const StorageMetadataPtr & /*metadata_snapshot*/,
-        SelectQueryInfo & /*query_info*/,
-        const Context & /*context*/,
-        QueryProcessingStage::Enum /*processed_stage*/,
-        size_t /*max_block_size*/,
-        unsigned /*num_streams*/)
+    const Names & /*column_names*/,
+    const StorageMetadataPtr & /*metadata_snapshot*/,
+    SelectQueryInfo & /*query_info*/,
+    const Context & /*context*/,
+    QueryProcessingStage::Enum /*processed_stage*/,
+    size_t /*max_block_size*/,
+    unsigned /*num_streams*/)
 {
     throw Exception("Method read is not supported by storage " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 }
 
 void IStorage::read(
-        QueryPlan & query_plan,
-        const Names & column_names,
-        const StorageMetadataPtr & metadata_snapshot,
-        SelectQueryInfo & query_info,
-        const Context & context,
-        QueryProcessingStage::Enum processed_stage,
-        size_t max_block_size,
-        unsigned num_streams)
+    QueryPlan & query_plan,
+    const Names & column_names,
+    const StorageMetadataPtr & metadata_snapshot,
+    SelectQueryInfo & query_info,
+    const Context & context,
+    QueryProcessingStage::Enum processed_stage,
+    size_t max_block_size,
+    unsigned num_streams)
 {
     auto pipe = read(column_names, metadata_snapshot, query_info, context, processed_stage, max_block_size, num_streams);
     if (pipe.empty())
@@ -117,15 +117,12 @@ void IStorage::read(
 }
 
 Pipe IStorage::alterPartition(
-    const StorageMetadataPtr & /* metadata_snapshot */,
-    const PartitionCommands & /* commands */,
-    const Context & /* context */)
+    const StorageMetadataPtr & /* metadata_snapshot */, const PartitionCommands & /* commands */, const Context & /* context */)
 {
     throw Exception("Partition operations are not supported by storage " + getName(), ErrorCodes::NOT_IMPLEMENTED);
 }
 
-void IStorage::alter(
-    const AlterCommands & params, const Context & context, TableLockHolder &)
+void IStorage::alter(const AlterCommands & params, const Context & context, TableLockHolder &)
 {
     auto table_id = getStorageID();
     StorageInMemoryMetadata new_metadata = getInMemoryMetadata();
@@ -146,7 +143,8 @@ void IStorage::checkAlterIsPossible(const AlterCommands & commands, const Settin
     }
 }
 
-void IStorage::checkAlterPartitionIsPossible(const PartitionCommands & /*commands*/, const StorageMetadataPtr & /*metadata_snapshot*/, const Settings & /*settings*/) const
+void IStorage::checkAlterPartitionIsPossible(
+    const PartitionCommands & /*commands*/, const StorageMetadataPtr & /*metadata_snapshot*/, const Settings & /*settings*/) const
 {
     throw Exception("Table engine " + getName() + " doesn't support partitioning", ErrorCodes::NOT_IMPLEMENTED);
 }
@@ -168,4 +166,13 @@ NamesAndTypesList IStorage::getVirtuals() const
     return {};
 }
 
+Names IStorage::getAllRegisteredNames() const
+{
+    Names result;
+    auto getter = [](const auto & column) { return column.name; };
+    const NamesAndTypesList & available_columns = getInMemoryMetadata().getColumns().getAllPhysical();
+    std::transform(available_columns.begin(), available_columns.end(), std::back_inserter(result), getter);
+    return result;
+}
+
 }
diff --git a/src/Storages/IStorage.h b/src/Storages/IStorage.h
index 61ce207a09d..9d5a1fe85a9 100644
--- a/src/Storages/IStorage.h
+++ b/src/Storages/IStorage.h
@@ -78,7 +78,7 @@ struct ColumnSize
   * - data storage structure (compression, etc.)
   * - concurrent access to data (locks, etc.)
   */
-class IStorage : public std::enable_shared_from_this<IStorage>, public TypePromotion<IStorage>
+class IStorage : public std::enable_shared_from_this<IStorage>, public TypePromotion<IStorage>, public IHints<1, IStorage>
 {
 public:
     IStorage() = delete;
@@ -87,7 +87,7 @@ public:
         : storage_id(std::move(storage_id_))
         , metadata(std::make_unique<StorageInMemoryMetadata>()) {} //-V730
 
-    virtual ~IStorage() = default;
+    virtual ~IStorage() override = default;
     IStorage(const IStorage &) = delete;
     IStorage & operator=(const IStorage &) = delete;
 
@@ -169,6 +169,7 @@ public:
     /// By default return empty list of columns.
     virtual NamesAndTypesList getVirtuals() const;
 
+    Names getAllRegisteredNames() const override;
 protected:
 
     /// Returns whether the column is virtual - by default all columns are real.

From 5cbe47301d12d9fda9e17a6d4362f468dc385cb5 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 15:56:14 +0300
Subject: [PATCH 087/504] Fix style

---
 src/Common/FiberStack.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index d73270c23d0..c2a6008e8d6 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -29,7 +29,7 @@ public:
 
     explicit FiberStack(size_t stack_size_ = default_stack_size) : stack_size(stack_size_)
     {
-        page_size = ::sysconf( _SC_PAGESIZE);
+        page_size = ::sysconf(_SC_PAGESIZE);
     }
 
     boost::context::stack_context allocate()
@@ -37,11 +37,11 @@ public:
         size_t num_pages = 1 + (stack_size - 1) / page_size;
         size_t num_bytes = (num_pages + 1) * page_size; /// Add one page at bottom that will be used as guard-page
 
-        void * vp = ::mmap( nullptr, num_bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+        void * vp = ::mmap(nullptr, num_bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
         if (MAP_FAILED == vp)
             DB::throwFromErrno(fmt::format("FiberStack: Cannot mmap {}.", ReadableSize(num_bytes)), DB::ErrorCodes::CANNOT_ALLOCATE_MEMORY);
 
-        if (-1 == ::mprotect( vp, page_size, PROT_NONE))
+        if (-1 == ::mprotect(vp, page_size, PROT_NONE))
         {
             ::munmap(vp, num_bytes);
             DB::throwFromErrno("FiberStack: cannot protect guard page", DB::ErrorCodes::CANNOT_ALLOCATE_MEMORY);
@@ -65,7 +65,7 @@ public:
         VALGRIND_STACK_DEREGISTER(sctx.valgrind_stack_id);
 #endif
         void * vp = static_cast< char * >(sctx.sp) - sctx.size;
-        ::munmap( vp, sctx.size);
+        ::munmap(vp, sctx.size);
 
         /// Do not count guard page in memory usage.
         CurrentMemoryTracker::free(sctx.size - page_size);

From 5277b565a81371978ef2f65b9b2234d423b19bc8 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 17:47:31 +0300
Subject: [PATCH 088/504] Fixing build.

---
 src/Common/FiberStack.h                   | 1 +
 src/Processors/Executors/AsyncTaskQueue.h | 1 +
 2 files changed, 2 insertions(+)

diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index c2a6008e8d6..f26ee55ce2b 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -2,6 +2,7 @@
 #include <common/defines.h>
 #include <boost/context/stack_context.hpp>
 #include <Common/formatReadable.h>
+#include <Common/MemoryTracker.h>
 
 #include <sys/time.h>
 #include <sys/resource.h>
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index 8b6e986cb9c..af616ca48be 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -50,6 +50,7 @@ public:
 #else
 class AsyncTaskQueue
 {
+public:
     bool empty() { return true; }
     void finish() {}
 };

From f3319cd06a2895c0a211d08a68ec48614dbb2822 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 18:18:08 +0300
Subject: [PATCH 089/504] Enlarge FiberStack size.

---
 src/Common/FiberStack.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Common/FiberStack.h b/src/Common/FiberStack.h
index f26ee55ce2b..2917a64a692 100644
--- a/src/Common/FiberStack.h
+++ b/src/Common/FiberStack.h
@@ -26,7 +26,7 @@ private:
     size_t stack_size;
     size_t page_size = 0;
 public:
-    static constexpr size_t default_stack_size = 64 * 1024;
+    static constexpr size_t default_stack_size = 128 * 1024; /// 64KB was not enough for tests
 
     explicit FiberStack(size_t stack_size_ = default_stack_size) : stack_size(stack_size_)
     {

From a23a9fd9474d2d9003debc4ed79905748f22f06b Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 15 Dec 2020 20:16:01 +0300
Subject: [PATCH 090/504] fixes after merge

---
 src/Interpreters/InterpreterSelectQuery.cpp | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 4b786163bb1..837ea50df01 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1770,6 +1770,7 @@ void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
             settings.max_block_size,
             0 /* LIMIT */,
             settings.max_bytes_before_remerge_sort,
+            settings.remerge_sort_lowered_memory_bytes_ratio,
             settings.max_bytes_before_external_sort,
             context->getTemporaryVolume(),
             settings.min_free_disk_space_for_temporary_data);
@@ -1846,9 +1847,13 @@ void InterpreterSelectQuery::executeOrder(QueryPlan & query_plan, InputOrderInfo
     /// Merge the sorted blocks.
     auto merge_sorting_step = std::make_unique<MergeSortingStep>(
             query_plan.getCurrentDataStream(),
-            output_order_descr, settings.max_block_size, limit,
-            settings.max_bytes_before_remerge_sort, settings.remerge_sort_lowered_memory_bytes_ratio,
-            settings.max_bytes_before_external_sort, context->getTemporaryVolume(),
+            output_order_descr,
+            settings.max_block_size,
+            limit,
+            settings.max_bytes_before_remerge_sort,
+            settings.remerge_sort_lowered_memory_bytes_ratio,
+            settings.max_bytes_before_external_sort,
+            context->getTemporaryVolume(),
             settings.min_free_disk_space_for_temporary_data);
 
     merge_sorting_step->setStepDescription("Merge sorted blocks for ORDER BY");

From fabb37d3c98a6c3893bd1cce94da7dca2ef65ff5 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 15 Dec 2020 20:39:58 +0300
Subject: [PATCH 091/504] cleanup

---
 .../AggregateFunctionFactory.cpp              |   3 -
 .../AggregateFunctionSum.cpp                  |   1 -
 src/Core/Block.cpp                            |  16 ---
 .../registerFunctionsMiscellaneous.cpp        |   2 -
 src/Functions/runningAccumulate.cpp           | 132 ------------------
 src/Interpreters/ActionsDAG.cpp               |   6 +-
 src/Interpreters/ExpressionActions.cpp        |   4 -
 src/Interpreters/ExpressionAnalyzer.cpp       |  66 ---------
 src/Interpreters/InterpreterSelectQuery.cpp   |   6 -
 src/Interpreters/TreeRewriter.cpp             |  10 +-
 src/Processors/QueryPlan/WindowStep.cpp       |  18 +--
 src/Processors/Transforms/FilterTransform.cpp |   3 -
 src/Processors/Transforms/WindowTransform.cpp |  15 --
 13 files changed, 6 insertions(+), 276 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionFactory.cpp b/src/AggregateFunctions/AggregateFunctionFactory.cpp
index 2fdf3c18a68..1c66650be45 100644
--- a/src/AggregateFunctions/AggregateFunctionFactory.cpp
+++ b/src/AggregateFunctions/AggregateFunctionFactory.cpp
@@ -58,9 +58,6 @@ static DataTypes convertLowCardinalityTypesToNested(const DataTypes & types)
 AggregateFunctionPtr AggregateFunctionFactory::get(
     const String & name, const DataTypes & argument_types, const Array & parameters, AggregateFunctionProperties & out_properties) const
 {
-    fmt::print(stderr, "get aggregate function {} at \n{}\n",
-        name, StackTrace().toString());
-
     auto type_without_low_cardinality = convertLowCardinalityTypesToNested(argument_types);
 
     /// If one of the types is Nullable, we apply aggregate function combinator "Null".
diff --git a/src/AggregateFunctions/AggregateFunctionSum.cpp b/src/AggregateFunctions/AggregateFunctionSum.cpp
index e937769f3a3..6afae98ef2d 100644
--- a/src/AggregateFunctions/AggregateFunctionSum.cpp
+++ b/src/AggregateFunctions/AggregateFunctionSum.cpp
@@ -52,7 +52,6 @@ template <typename T> using AggregateFunctionSumKahan =
 template <template <typename> class Function>
 AggregateFunctionPtr createAggregateFunctionSum(const std::string & name, const DataTypes & argument_types, const Array & parameters)
 {
-    fmt::print(stderr, "create sum at \n{}\n", StackTrace().toString());
     assertNoParameters(name, parameters);
     assertUnary(name, argument_types);
 
diff --git a/src/Core/Block.cpp b/src/Core/Block.cpp
index 014fa001588..cd2855739e2 100644
--- a/src/Core/Block.cpp
+++ b/src/Core/Block.cpp
@@ -56,10 +56,6 @@ void Block::insert(size_t position, const ColumnWithTypeAndName & elem)
 
     index_by_name.emplace(elem.name, position);
     data.emplace(data.begin() + position, elem);
-
-//    fmt::print(stderr, "block: {}\n", dumpStructure());
-//    fmt::print(stderr, "(1) insert column {} at \n{}\n", elem.name,
-//        StackTrace().toString());
 }
 
 void Block::insert(size_t position, ColumnWithTypeAndName && elem)
@@ -74,29 +70,17 @@ void Block::insert(size_t position, ColumnWithTypeAndName && elem)
 
     index_by_name.emplace(elem.name, position);
     data.emplace(data.begin() + position, std::move(elem));
-
-//    fmt::print(stderr, "block: {}\n", dumpStructure());
-//    fmt::print(stderr, "(2) insert column {} at \n{}\n", elem.name,
-//        StackTrace().toString());
 }
 
 
 void Block::insert(const ColumnWithTypeAndName & elem)
 {
-//    fmt::print(stderr, "block: {}\n", dumpStructure());
-//    fmt::print(stderr, "(3) insert column {} at \n{}\n", elem.name,
-//        StackTrace().toString());
-
     index_by_name.emplace(elem.name, data.size());
     data.emplace_back(elem);
 }
 
 void Block::insert(ColumnWithTypeAndName && elem)
 {
-//    fmt::print(stderr, "block: {}\n", dumpStructure());
-//    fmt::print(stderr, "(4) insert column {} at \n{}\n", elem.name,
-//        StackTrace().toString());
-
     index_by_name.emplace(elem.name, data.size());
     data.emplace_back(std::move(elem));
 }
diff --git a/src/Functions/registerFunctionsMiscellaneous.cpp b/src/Functions/registerFunctionsMiscellaneous.cpp
index 199567680e7..2c0baa7bba4 100644
--- a/src/Functions/registerFunctionsMiscellaneous.cpp
+++ b/src/Functions/registerFunctionsMiscellaneous.cpp
@@ -43,7 +43,6 @@ void registerFunctionBuildId(FunctionFactory &);
 void registerFunctionUptime(FunctionFactory &);
 void registerFunctionTimeZone(FunctionFactory &);
 void registerFunctionRunningAccumulate(FunctionFactory &);
-void registerFunctionWindow(FunctionFactory &);
 void registerFunctionRunningDifference(FunctionFactory &);
 void registerFunctionRunningDifferenceStartingWithFirstValue(FunctionFactory &);
 void registerFunctionFinalizeAggregation(FunctionFactory &);
@@ -110,7 +109,6 @@ void registerFunctionsMiscellaneous(FunctionFactory & factory)
     registerFunctionUptime(factory);
     registerFunctionTimeZone(factory);
     registerFunctionRunningAccumulate(factory);
-    registerFunctionWindow(factory);
     registerFunctionRunningDifference(factory);
     registerFunctionRunningDifferenceStartingWithFirstValue(factory);
     registerFunctionFinalizeAggregation(factory);
diff --git a/src/Functions/runningAccumulate.cpp b/src/Functions/runningAccumulate.cpp
index 9125597fff9..1a01b88e22c 100644
--- a/src/Functions/runningAccumulate.cpp
+++ b/src/Functions/runningAccumulate.cpp
@@ -75,8 +75,6 @@ public:
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
     {
-        fmt::print(stderr, "executeImpl at \n{}\n", StackTrace().toString());
-
         const ColumnAggregateFunction * column_with_states
             = typeid_cast<const ColumnAggregateFunction *>(&*arguments.at(0).column);
 
@@ -138,136 +136,6 @@ public:
 
 }
 
-namespace
-{
-
-class FunctionWindow : public IFunction
-{
-    mutable ColumnsWithTypeAndName prev_columns = {};
-    mutable int prev_index = -1;
-
-    mutable AlignedBuffer place;
-    mutable std::unique_ptr<Arena> arena;
-
-public:
-    static constexpr auto name = "window";
-    static FunctionPtr create(const Context &)
-    {
-        return std::make_shared<FunctionWindow>();
-    }
-
-    String getName() const override
-    {
-        return name;
-    }
-
-    bool isStateful() const override
-    {
-        return true;
-    }
-
-    bool isVariadic() const override { return true; }
-
-    size_t getNumberOfArguments() const override { return 0; }
-
-    bool isDeterministic() const override { return false; }
-
-    bool isDeterministicInScopeOfQuery() const override
-    {
-        return false;
-    }
-
-    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
-    {
-        fmt::print(stderr, "getReturnType for {} at \n{}\n",
-            static_cast<const void *>(this), StackTrace().toString());
-
-        if (arguments.empty() || arguments.size() > 3)
-            throw Exception("Incorrect number of arguments of function " + getName() + ". Must be up to 3 (function, PARTITION BY, ORDER BY).",
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
-
-        const DataTypeAggregateFunction * type = checkAndGetDataType<DataTypeAggregateFunction>(arguments[0].get());
-        if (!type)
-            throw Exception("Argument for function " + getName() + " must have type AggregateFunction - state of aggregate function.",
-                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-
-        const auto & agg_func = *type->getFunction();
-        place.reset(agg_func.sizeOfData(), agg_func.alignOfData());
-
-        /// Will pass empty arena if agg_func does not allocate memory in arena
-        arena = agg_func.allocatesMemoryInArena() ? std::make_unique<Arena>() : nullptr;
-
-
-        return type->getReturnType();
-    }
-
-    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
-    {
-        fmt::print(stderr, "executeImpl for {} at \n{}\n",
-            static_cast<const void *>(this), StackTrace().toString());
-
-        const ColumnAggregateFunction * column_with_states
-            = typeid_cast<const ColumnAggregateFunction *>(&*arguments.at(0).column);
-
-        if (!column_with_states)
-            throw Exception("Illegal column " + arguments.at(0).column->getName()
-                    + " of first argument of function "
-                    + getName(),
-                ErrorCodes::ILLEGAL_COLUMN);
-
-        ColumnPtr column_with_groups;
-
-        if (arguments.size() >= 2)
-            column_with_groups = arguments[1].column;
-
-        AggregateFunctionPtr aggregate_function_ptr = column_with_states->getAggregateFunction();
-        const IAggregateFunction & agg_func = *aggregate_function_ptr;
-
-
-        auto result_column_ptr = agg_func.getReturnType()->createColumn();
-        IColumn & result_column = *result_column_ptr;
-        result_column.reserve(column_with_states->size());
-
-        const auto & states = column_with_states->getData();
-
-        bool state_created = false;
-        SCOPE_EXIT({
-            if (state_created)
-                agg_func.destroy(place.data());
-        });
-
-        size_t row_number = 0;
-        for (const auto & state_to_add : states)
-        {
-            if (row_number == 0 || (column_with_groups && column_with_groups->compareAt(row_number, row_number - 1, *column_with_groups, 1) != 0))
-            {
-                if (state_created)
-                {
-                    agg_func.destroy(place.data());
-                    state_created = false;
-                }
-
-                agg_func.create(place.data());
-                state_created = true;
-            }
-
-            agg_func.merge(place.data(), state_to_add, arena.get());
-            agg_func.insertResultInto(place.data(), result_column, arena.get());
-
-            ++row_number;
-        }
-
-        return result_column_ptr;
-    }
-};
-
-}
-
-void registerFunctionWindow(FunctionFactory & factory)
-{
-    factory.registerFunction<FunctionWindow>();
-}
-
 void registerFunctionRunningAccumulate(FunctionFactory & factory)
 {
     factory.registerFunction<FunctionRunningAccumulate>();
diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index cd6fab40aaf..57616c49f8a 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -624,7 +624,7 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
             {
                 auto & input = inputs[res_elem.name];
                 if (input.empty())
-                    throw Exception("Cannot find column '" + backQuoteIfNeed(res_elem.name) + "' in source stream",
+                    throw Exception("Cannot find column '" + backQuote(res_elem.name) + "' in source stream",
                                     ErrorCodes::THERE_IS_NO_COLUMN);
 
                 src_node = actions_dag->inputs[input.front()];
@@ -641,12 +641,12 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
                 if (ignore_constant_values)
                    src_node = const_cast<Node *>(&actions_dag->addColumn(res_elem, true));
                 else if (res_const->getField() != src_const->getField())
-                    throw Exception("Cannot convert column " + backQuoteIfNeed(res_elem.name) + " because "
+                    throw Exception("Cannot convert column " + backQuote(res_elem.name) + " because "
                                     "it is constant but values of constants are different in source and result",
                                     ErrorCodes::ILLEGAL_COLUMN);
             }
             else
-                throw Exception("Cannot convert column " + backQuoteIfNeed(res_elem.name) + " because "
+                throw Exception("Cannot convert column " + backQuote(res_elem.name) + " because "
                                 "it is non constant in source stream but must be constant in result",
                                 ErrorCodes::ILLEGAL_COLUMN);
         }
diff --git a/src/Interpreters/ExpressionActions.cpp b/src/Interpreters/ExpressionActions.cpp
index 4c2ec0e0dc8..f0dcf830c82 100644
--- a/src/Interpreters/ExpressionActions.cpp
+++ b/src/Interpreters/ExpressionActions.cpp
@@ -373,10 +373,6 @@ static void executeAction(const ExpressionActions::Action & action, ExecutionCon
             res_column.column = action.node->column->cloneResized(num_rows);
             res_column.type = action.node->result_type;
             res_column.name = action.node->result_name;
-
-
-            fmt::print(stderr, "execute column action {} at\n{}\n",
-                action.toString(), StackTrace().toString());
             break;
         }
 
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 2ae55b9d5bc..09a4855f74f 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1004,16 +1004,11 @@ void SelectQueryExpressionAnalyzer::appendAggregateFunctionsArguments(Expression
         if (node->arguments)
             for (auto & argument : node->arguments->children)
                 getRootActions(argument, only_types, step.actions());
-
-
-    fmt::print(stderr, "actions after appendAggregateFunctionsArguments: \n{} at \n{}\n", chain.dumpChain(), StackTrace().toString());
 }
 
 void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
     ExpressionActionsChain & chain, bool /* only_types */)
 {
-    fmt::print(stderr, "actions before window: {}\n", chain.dumpChain());
-
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
     for (const auto & f : window_functions)
@@ -1039,28 +1034,6 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
         step.actions()->addInput(col);
     }
 
-    // /*
-    // for (const auto & desc : aggregate_descriptions)
-    //     for (const auto & name : desc.argument_names)
-    //         step.required_output.emplace_back(name);
-    // */
-
-    // const auto * select_query = getSelectQuery();
-    // /// Collect aggregates removing duplicates by node.getColumnName()
-    // /// It's not clear why we recollect aggregates (for query parts) while we're able to use previously collected ones (for entire query)
-    // /// @note The original recollection logic didn't remove duplicates.
-    // GetAggregatesVisitor::Data data;
-    // GetAggregatesVisitor(data).visit(select_query->select());
-
-    // // 1) just add everything there is in the AST,
-    // for (const ASTFunction * node : data.window_functions)
-    // {
-    //     for (auto & argument : node->arguments->children)
-    //     {
-    //         getRootActions(argument, only_types, step.actions());
-    //     }
-    // }
-
     // 2) mark the columns that are really required:
     for (const auto & f : window_functions)
     {
@@ -1081,8 +1054,6 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
             step.required_output.push_back(c.column_name);
         }
     }
-
-    fmt::print(stderr, "actions after window: {}\n", chain.dumpChain());
 }
 
 bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain, bool only_types)
@@ -1092,8 +1063,6 @@ bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain,
     if (!select_query->having())
         return false;
 
-    fmt::print(stderr, "has having\n");
-
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
     getRootActionsForHaving(select_query->having(), only_types, step.actions());
@@ -1267,8 +1236,6 @@ ActionsDAGPtr SelectQueryExpressionAnalyzer::appendProjectResult(ExpressionActio
         }
     }
 
-    fmt::print(stderr, "chain before last projection: {}\n",
-        chain.dumpChain());
     auto actions = chain.getLastActions();
     actions->project(result_columns);
     return actions;
@@ -1473,8 +1440,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
             chain.addStep();
         }
 
-        fmt::print(stderr, "chain before aggregate: {}\n", chain.dumpChain());
-
         if (need_aggregate)
         {
             /// TODO correct conditions
@@ -1483,29 +1448,18 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
                     && storage && query.groupBy();
 
             query_analyzer.appendGroupBy(chain, only_types || !first_stage, optimize_aggregation_in_order, group_by_elements_actions);
-
-            fmt::print(stderr, "chain after appendGroupBy: {}\n", chain.dumpChain());
-
             query_analyzer.appendAggregateFunctionsArguments(chain, only_types || !first_stage);
-
-            fmt::print(stderr, "chain after appendAggregateFunctionsArguments: {}\n", chain.dumpChain());
-
             before_aggregation = chain.getLastActions();
 
             finalize_chain(chain);
 
-            fmt::print(stderr, "chain after finalize_chain: {}\n", chain.dumpChain());
-
             if (query_analyzer.appendHaving(chain, only_types || !second_stage))
             {
-                fmt::print(stderr, "chain after appendHaving: {}\n", chain.dumpChain());
                 before_having = chain.getLastActions();
                 chain.addStep();
             }
         }
 
-        fmt::print(stderr, "chain after aggregate: {}\n", chain.dumpChain());
-
         bool join_allow_read_in_order = true;
         if (hasJoin())
         {
@@ -1521,22 +1475,10 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
             && !query.final()
             && join_allow_read_in_order;
 
-        /*
-        if (has_window)
-        {
-            query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
-            before_window = chain.getLastActions();
-
-            finalize_chain(chain);
-        }
-        */
-
         /// If there is aggregation, we execute expressions in SELECT and ORDER BY on the initiating server, otherwise on the source servers.
         query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
-        fmt::print(stderr, "chain after select: {}\n", chain.dumpChain());
 
         query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
-        fmt::print(stderr, "chain after window: {}\n", chain.dumpChain());
 
         selected_columns = chain.getLastStep().required_output;
         has_order_by = query.orderBy() != nullptr;
@@ -1546,10 +1488,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
                 optimize_read_in_order,
                 order_by_elements_actions);
 
-        fmt::print(stderr, "chain after order by: {}\n", chain.dumpChain());
-
-        //if (h
-
         if (query_analyzer.appendLimitBy(chain, only_types || !second_stage))
         {
             before_limit_by = chain.getLastActions();
@@ -1565,10 +1503,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
     removeExtraColumns();
 
     checkActions();
-
-    fmt::print(stderr, "ExpressionAnalysisResult created at \n{}\n",
-        StackTrace().toString());
-    fmt::print(stderr, "ExpressionAnalysisResult: \n{}\n", dump());
 }
 
 void ExpressionAnalysisResult::finalize(const ExpressionActionsChain & chain, size_t where_step_num)
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 837ea50df01..a9d2dcec821 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -958,7 +958,6 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
             }
             else
             {
-                /// FIXME calculate windows here
                 executeExpression(query_plan, expressions.before_order_and_select, "Before ORDER BY and SELECT");
                 executeWindow(query_plan);
                 executeDistinct(query_plan, true, expressions.selected_columns, true);
@@ -1006,7 +1005,6 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
                 else if (expressions.hasHaving())
                     executeHaving(query_plan, expressions.before_having);
 
-                /// FIXME calculate windows here
                 executeExpression(query_plan, expressions.before_order_and_select, "Before ORDER BY and SELECT");
                 executeWindow(query_plan);
                 executeDistinct(query_plan, true, expressions.selected_columns, true);
@@ -1748,10 +1746,6 @@ void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
     {
         const auto & w = query_analyzer->window_descriptions[f.window_name];
 
-        fmt::print(stderr, "window function '{}' over window '{}'\n",
-            f.column_name, f.window_name);
-        fmt::print(stderr, "{}\n{}\n", f.dump(), w.dump());
-
         const Settings & settings = context->getSettingsRef();
 
         auto partial_sorting = std::make_unique<PartialSortingStep>(
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index c6e69f1c06c..befae896d2f 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -443,7 +443,7 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
 
 std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSelectQuery & select_query)
 {
-    /// There can not be aggregate functions inside the WHERE and PREWHERE.
+    /// There can not be window functions inside the WHERE and PREWHERE.
     if (select_query.where())
         assertNoWindows(select_query.where(), "in WHERE");
     if (select_query.prewhere())
@@ -452,7 +452,7 @@ std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSel
     GetAggregatesVisitor::Data data;
     GetAggregatesVisitor(data).visit(query);
 
-    /// There can not be other aggregate functions within the aggregate functions.
+    /// There can not be other window functions within the aggregate functions.
     for (const ASTFunction * node : data.window_functions)
     {
         if (node->arguments)
@@ -465,10 +465,6 @@ std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSel
         }
     }
 
-    fmt::print(stderr, "getWindowFunctions ({}) for \n{}\n at \n{}\n",
-        data.window_functions.size(), query->formatForErrorMessage(),
-        StackTrace().toString());
-
     return data.window_functions;
 }
 
@@ -485,8 +481,6 @@ TreeRewriterResult::TreeRewriterResult(
 {
     collectSourceColumns(add_special);
     is_remote_storage = storage && storage->isRemote();
-
-    fmt::print(stderr, "TreeRewriterResult created at \n{}\n", StackTrace().toString());
 }
 
 /// Add columns from storage to source_columns list. Deduplicate resulted list.
diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
index 495513c1a2a..3eccffc5828 100644
--- a/src/Processors/QueryPlan/WindowStep.cpp
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -44,13 +44,8 @@ static ITransformingStep::Traits getJoinTraits()
 static Block addWindowFunctionColumns(const Block & block,
     std::vector<WindowFunctionDescription> window_functions)
 {
-    fmt::print(stderr, "input header: {}\n", block.dumpStructure());
-
-    //auto result = block.cloneWithoutColumns();
     auto result = block;
 
-    fmt::print(stderr, "header after clone: {}\n", result.dumpStructure());
-
     for (const auto & f : window_functions)
     {
         ColumnWithTypeAndName column_with_type;
@@ -61,8 +56,6 @@ static Block addWindowFunctionColumns(const Block & block,
         result.insert(column_with_type);
     }
 
-    fmt::print(stderr, "header after insert: {}\n", result.dumpStructure());
-
     return result;
 }
 
@@ -97,16 +90,7 @@ void WindowStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     (void) prefix;
-    //bool first = true;
-
-    //auto expression = std::make_shared<ExpressionActions>(actions_dag);
-    //for (const auto & action : expression->getActions())
-    //{
-    //    settings.out << prefix << (first ? "Actions: "
-    //                                     : "         ");
-    //    first = false;
-    //    settings.out << action.toString() << '\n';
-    //}
+    /// FIXME add some printing
 }
 
 }
diff --git a/src/Processors/Transforms/FilterTransform.cpp b/src/Processors/Transforms/FilterTransform.cpp
index d16fbef6905..23b1bf8a984 100644
--- a/src/Processors/Transforms/FilterTransform.cpp
+++ b/src/Processors/Transforms/FilterTransform.cpp
@@ -63,9 +63,6 @@ FilterTransform::FilterTransform(
     auto & column = transformed_header.getByPosition(filter_column_position).column;
     if (column)
         constant_filter_description = ConstantFilterDescription(*column);
-
-    fmt::print(stderr, "FilterTransform created at \n{}\n",
-        StackTrace().toString());
 }
 
 IProcessor::Status FilterTransform::prepare()
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index 1e166fe1aff..1084d96cddd 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -12,14 +12,11 @@ WindowTransform::WindowTransform(const Block & input_header_,
         const WindowDescription & window_description_,
         const std::vector<WindowFunctionDescription> & window_function_descriptions
         )
-    // FIXME this is where the output column is added
     : ISimpleTransform(input_header_, output_header_,
         false /* skip_empty_chunks */)
     , input_header(input_header_)
     , window_description(window_description_)
 {
-    fmt::print(stderr, "input header {}\n", input_header.dumpStructure());
-
     workspaces.reserve(window_function_descriptions.size());
     for (size_t i = 0; i < window_function_descriptions.size(); ++i)
     {
@@ -42,12 +39,6 @@ WindowTransform::WindowTransform(const Block & input_header_,
             workspace.argument_column_indices.push_back(
                 input_header.getPositionByName(argument_name));
 
-            fmt::print(stderr,
-                "window function '{}' argument column '{}' at '{}'\n",
-                workspace.window_function.column_name,
-                argument_name,
-                workspace.argument_column_indices.back());
-
         }
 
         workspace.aggregate_function_state.reset(aggregate_function->sizeOfData(),
@@ -80,14 +71,11 @@ void WindowTransform::transform(Chunk & chunk)
     const size_t num_rows = chunk.getNumRows();
     auto block = getInputPort().getHeader().cloneWithColumns(chunk.detachColumns());
 
-    fmt::print(stderr, "block before window transform:\n{}\n", block.dumpStructure());
-
     for (auto & workspace : workspaces)
     {
         workspace.argument_columns.clear();
         for (const auto column_index : workspace.argument_column_indices)
         {
-            fmt::print(stderr, "argument column index '{}'\n", column_index);
             workspace.argument_columns.push_back(
                 block.getColumns()[column_index].get());
         }
@@ -105,7 +93,6 @@ void WindowTransform::transform(Chunk & chunk)
         auto c = column_with_type.type->createColumn();
         column_with_type.column.reset(c.get());
 
-        size_t partition_start = 0;
         for (size_t row = 0; row < num_rows; row++)
         {
             // Check whether the new partition has started and reinitialize the
@@ -152,8 +139,6 @@ void WindowTransform::transform(Chunk & chunk)
         //*/
     }
 
-    fmt::print(stderr, "block after window transform:\n{}\n", block.dumpStructure());
-
     chunk.setColumns(block.getColumns(), num_rows);
 }
 

From 54efd830e63ee5ad88e80eb5bbb74ed488434874 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 21:04:12 +0300
Subject: [PATCH 092/504] Fixing build

---
 src/Common/TimerDescriptor.cpp | 2 ++
 src/Common/TimerDescriptor.h   | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/src/Common/TimerDescriptor.cpp b/src/Common/TimerDescriptor.cpp
index a77a9d24ed0..f4c3ec35588 100644
--- a/src/Common/TimerDescriptor.cpp
+++ b/src/Common/TimerDescriptor.cpp
@@ -1,3 +1,4 @@
+#if defined(OS_LINUX)
 #include <Common/TimerDescriptor.h>
 #include <Common/Exception.h>
 
@@ -80,3 +81,4 @@ void TimerDescriptor::setRelative(const Poco::Timespan & timespan) const
 }
 
 }
+#endif
diff --git a/src/Common/TimerDescriptor.h b/src/Common/TimerDescriptor.h
index 04c592ebae6..0e24c78696c 100644
--- a/src/Common/TimerDescriptor.h
+++ b/src/Common/TimerDescriptor.h
@@ -1,4 +1,5 @@
 #pragma once
+#if defined(OS_LINUX)
 #include <Poco/Timespan.h>
 
 namespace DB
@@ -26,3 +27,4 @@ public:
 };
 
 }
+#endif

From 3d6ace5890d626235afe5208b20e9085bef1181e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 21:20:14 +0300
Subject: [PATCH 093/504] Fixing tests.

---
 src/Client/Connection.cpp             |  5 ++++-
 src/Client/Connection.h               |  3 +--
 src/Client/MultiplexedConnections.cpp | 16 ++--------------
 src/Client/MultiplexedConnections.h   |  5 +----
 4 files changed, 8 insertions(+), 21 deletions(-)

diff --git a/src/Client/Connection.cpp b/src/Client/Connection.cpp
index 8f4a64766cd..16f0786c70b 100644
--- a/src/Client/Connection.cpp
+++ b/src/Client/Connection.cpp
@@ -742,8 +742,11 @@ std::optional<UInt64> Connection::checkPacket(size_t timeout_microseconds)
 }
 
 
-Packet Connection::receivePacket()
+Packet Connection::receivePacket(Fiber * fiber)
 {
+    in->setFiber(fiber);
+    SCOPE_EXIT(in->setFiber(nullptr));
+
     try
     {
         Packet res;
diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index 29a9274dca4..05763f83578 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -171,7 +171,7 @@ public:
     std::optional<UInt64> checkPacket(size_t timeout_microseconds = 0);
 
     /// Receive packet from server.
-    Packet receivePacket();
+    Packet receivePacket(Fiber * fiber);
 
     /// If not connected yet, or if connection is broken - then connect. If cannot connect - throw an exception.
     void forceConnected(const ConnectionTimeouts & timeouts);
@@ -190,7 +190,6 @@ public:
     size_t outBytesCount() const { return out ? out->count() : 0; }
     size_t inBytesCount() const { return in ? in->count() : 0; }
 
-    void setFiber(Fiber * fiber) { in->setFiber(fiber); }
     Poco::Net::Socket & getSocket() { return in->getSocket(); }
 
 private:
diff --git a/src/Client/MultiplexedConnections.cpp b/src/Client/MultiplexedConnections.cpp
index 30311f40512..87f8ff34c2f 100644
--- a/src/Client/MultiplexedConnections.cpp
+++ b/src/Client/MultiplexedConnections.cpp
@@ -237,7 +237,7 @@ std::string MultiplexedConnections::dumpAddressesUnlocked() const
     return buf.str();
 }
 
-Packet MultiplexedConnections::receivePacketUnlocked()
+Packet MultiplexedConnections::receivePacketUnlocked(Fiber * fiber)
 {
     if (!sent_query)
         throw Exception("Cannot receive packets: no query sent.", ErrorCodes::LOGICAL_ERROR);
@@ -249,19 +249,7 @@ Packet MultiplexedConnections::receivePacketUnlocked()
     if (current_connection == nullptr)
         throw Exception("Logical error: no available replica", ErrorCodes::NO_AVAILABLE_REPLICA);
 
-    if (fiber)
-        current_connection->setFiber(fiber);
-
-    Packet packet;
-    {
-        SCOPE_EXIT(
-        {
-           fiber = nullptr;
-           current_connection->setFiber(fiber);
-        });
-
-        packet = current_connection->receivePacket();
-    }
+    Packet packet = current_connection->receivePacket(fiber);
 
     switch (packet.type)
     {
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index 5f8500ebe6f..42d86086ddf 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -67,12 +67,11 @@ public:
     /// Without locking, because sendCancel() does not change the state of the replicas.
     bool hasActiveConnections() const { return active_connection_count > 0; }
 
-    void setFiber(Fiber * fiber_) { fiber = fiber_; }
     Poco::Net::Socket & getSocket() { return current_connection->getSocket(); }
 
 private:
     /// Internal version of `receivePacket` function without locking.
-    Packet receivePacketUnlocked();
+    Packet receivePacketUnlocked(Fiber * fiber = nullptr);
 
     /// Internal version of `dumpAddresses` function without locking.
     std::string dumpAddressesUnlocked() const;
@@ -109,8 +108,6 @@ private:
     /// in separate thread.
     mutable std::mutex cancel_mutex;
 
-    boost::context::fiber * fiber = nullptr;
-
     friend class RemoteQueryExecutorReadContext;
 };
 

From 4c290996dc5dc608455dd29061ccc7692af9e08e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 21:22:20 +0300
Subject: [PATCH 094/504] Fixing tests.

---
 src/Client/Connection.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index 05763f83578..d5d202d9ac2 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -171,7 +171,7 @@ public:
     std::optional<UInt64> checkPacket(size_t timeout_microseconds = 0);
 
     /// Receive packet from server.
-    Packet receivePacket(Fiber * fiber);
+    Packet receivePacket(Fiber * fiber = nullptr);
 
     /// If not connected yet, or if connection is broken - then connect. If cannot connect - throw an exception.
     void forceConnected(const ConnectionTimeouts & timeouts);

From 645e0d2c0c89895e4f0ac605f048053a33715bf0 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 21:23:41 +0300
Subject: [PATCH 095/504] Fixing tests.

---
 src/DataStreams/RemoteQueryExecutorReadContext.h | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index cedfc805ee1..9765367b647 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -229,10 +229,8 @@ public:
             {
                 while (true)
                 {
-                    connections.setFiber(&sink);
-
                     read_context.is_read_in_progress = true;
-                    read_context.packet = connections.receivePacketUnlocked();
+                    read_context.packet = connections.receivePacketUnlocked(&sink);
                     read_context.is_read_in_progress = false;
 
                     sink = std::move(sink).resume();

From f57b7bc249304ead0bfa3011ad1c53ff1f8e0a9c Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 15 Dec 2020 23:23:30 +0300
Subject: [PATCH 096/504] Try fix data race.

---
 src/Processors/Sources/RemoteSource.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index 5c59c552f69..b4f5d9272f2 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -105,7 +105,7 @@ void RemoteSource::onCancel()
 {
     was_query_canceled = true;
     query_executor->cancel(&read_context);
-    is_async_state = false;
+    // is_async_state = false;
 }
 
 void RemoteSource::onUpdatePorts()
@@ -114,7 +114,7 @@ void RemoteSource::onUpdatePorts()
     {
         was_query_canceled = true;
         query_executor->finish(&read_context);
-        is_async_state = false;
+        // is_async_state = false;
     }
 }
 

From 7ca86f176c1587a1129bb3833aa6db3de7964e06 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Wed, 16 Dec 2020 01:43:07 +0300
Subject: [PATCH 097/504] Fix TSan warning in system.stack_trace

---
 .../System/StorageSystemStackTrace.cpp        | 39 ++++++++++++-------
 1 file changed, 26 insertions(+), 13 deletions(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index 92d12c19986..dd21495ed4b 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -38,8 +38,18 @@ namespace
     const int sig = SIGRTMIN;
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.
+    std::atomic<int> data_ready_num = 0;
 
-    std::optional<StackTrace> stack_trace;
+    /** Notes:
+      * Only one query from the table can be processed at the moment of time.
+      * This is ensured by the mutex in fillData function.
+      * We obtain information about threads by sending signal and receiving info from the signal handler.
+      * Information is passed via global variables and pipe is used for signaling.
+      * Actually we can send all information via pipe, but we read from it with timeout just in case,
+      * so it's convenient to use is only for signaling.
+      */
+
+    StackTrace stack_trace{NoCapture{}};
 
     constexpr size_t max_query_id_size = 128;
     char query_id_data[max_query_id_size];
@@ -47,7 +57,7 @@ namespace
 
     LazyPipeFDs notification_pipe;
 
-    void NO_SANITIZE_THREAD signalHandler(int, siginfo_t * info, void * context)
+    void signalHandler(int, siginfo_t * info, void * context)
     {
         auto saved_errno = errno;   /// We must restore previous value of errno in signal handler.
 
@@ -57,19 +67,22 @@ namespace
             return;
 
         /// Signal received too late.
-        if (info->si_value.sival_int != sequence_num.load(std::memory_order_relaxed))
+        int notification_num = info->si_value.sival_int;
+        if (notification_num != sequence_num.load(std::memory_order_acquire))
             return;
 
         /// All these methods are signal-safe.
         const ucontext_t signal_context = *reinterpret_cast<ucontext_t *>(context);
-        stack_trace.emplace(signal_context);
+        stack_trace = StackTrace(signal_context);
 
         StringRef query_id = CurrentThread::getQueryId();
         query_id_size = std::min(query_id.size, max_query_id_size);
         if (query_id.data && query_id.size)
             memcpy(query_id_data, query_id.data, query_id_size);
 
-        int notification_num = info->si_value.sival_int;
+        /// This is unneeded (because we synchronize through pipe) but makes TSan happy.
+        data_ready_num.store(sequence_num, std::memory_order_release);
+
         ssize_t res = ::write(notification_pipe.fds_rw[1], &notification_num, sizeof(notification_num));
 
         /// We cannot do anything if write failed.
@@ -79,7 +92,7 @@ namespace
     }
 
     /// Wait for data in pipe and read it.
-    bool NO_SANITIZE_THREAD wait(int timeout_ms)
+    bool wait(int timeout_ms)
     {
         while (true)
         {
@@ -127,7 +140,7 @@ namespace
 }
 
 
-NO_SANITIZE_THREAD StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
+StorageSystemStackTrace::StorageSystemStackTrace(const StorageID & table_id_)
     : IStorageSystemOneBlock<StorageSystemStackTrace>(table_id_)
 {
     notification_pipe.open();
@@ -160,7 +173,7 @@ NamesAndTypesList StorageSystemStackTrace::getNamesAndTypes()
 }
 
 
-void NO_SANITIZE_THREAD StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
+void StorageSystemStackTrace::fillData(MutableColumns & res_columns, const Context &, const SelectQueryInfo &) const
 {
     /// It shouldn't be possible to do concurrent reads from this table.
     std::lock_guard lock(mutex);
@@ -180,7 +193,7 @@ void NO_SANITIZE_THREAD StorageSystemStackTrace::fillData(MutableColumns & res_c
         pid_t tid = parse<pid_t>(it->path().filename());
 
         sigval sig_value{};
-        sig_value.sival_int = sequence_num.load(std::memory_order_relaxed);
+        sig_value.sival_int = sequence_num.load(std::memory_order_acquire);
         if (0 != ::sigqueue(tid, sig, sig_value))
         {
             /// The thread may has been already finished.
@@ -192,15 +205,15 @@ void NO_SANITIZE_THREAD StorageSystemStackTrace::fillData(MutableColumns & res_c
 
         /// Just in case we will wait for pipe with timeout. In case signal didn't get processed.
 
-        if (wait(100))
+        if (wait(100) && sig_value.sival_int == data_ready_num.load(std::memory_order_acquire))
         {
-            size_t stack_trace_size = stack_trace->getSize();
-            size_t stack_trace_offset = stack_trace->getOffset();
+            size_t stack_trace_size = stack_trace.getSize();
+            size_t stack_trace_offset = stack_trace.getOffset();
 
             Array arr;
             arr.reserve(stack_trace_size - stack_trace_offset);
             for (size_t i = stack_trace_offset; i < stack_trace_size; ++i)
-                arr.emplace_back(reinterpret_cast<intptr_t>(stack_trace->getFramePointers()[i]));
+                arr.emplace_back(reinterpret_cast<intptr_t>(stack_trace.getFramePointers()[i]));
 
             res_columns[0]->insert(tid);
             res_columns[1]->insertData(query_id_data, query_id_size);

From c906b1ca769679af8bf8f0ed1db8fd80877fa416 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 16 Dec 2020 15:21:23 +0300
Subject: [PATCH 098/504] added the initial solution

---
 src/Functions/FunctionBinaryArithmetic.h | 29 ++++++++++++++++++++----
 src/Functions/multiply.cpp               |  2 +-
 2 files changed, 25 insertions(+), 6 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 8c684d3578f..ea6169194e4 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -3,6 +3,8 @@
 // Include this first, because `#define _asan_poison_address` from
 // llvm/Support/Compiler.h conflicts with its forward declaration in
 // sanitizer/asan_interface.h
+#include <memory>
+#include <type_traits>
 #include <Common/Arena.h>
 
 #include <DataTypes/DataTypesNumber.h>
@@ -467,6 +469,11 @@ public:
         Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>, RightDataType>,
         Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
+
+        /// e.g Decimal * Float64 = Float64
+        Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+        Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>,
+
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
         Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
@@ -790,10 +797,12 @@ public:
         }
 
         DataTypePtr type_res;
-        bool valid = castBothTypes(arguments[0].get(), arguments[1].get(), [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(arguments[0].get(), arguments[1].get(), [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
@@ -810,6 +819,7 @@ public:
             else
             {
                 using ResultDataType = typename BinaryOperationTraits<Op, LeftDataType, RightDataType>::ResultDataType;
+
                 if constexpr (!std::is_same_v<ResultDataType, InvalidType>)
                 {
                     if constexpr (IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>)
@@ -817,6 +827,10 @@ public:
                         ResultDataType result_type = decimalResultType<is_multiply, is_division>(left, right);
                         type_res = std::make_shared<ResultDataType>(result_type.getPrecision(), result_type.getScale());
                     }
+                    else if constexpr ((IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>) ||
+                        (IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>))
+                        type_res = std::make_shared<std::conditional_t<IsFloatingPoint<LeftDataType>,
+                            LeftDataType, RightDataType>>();
                     else if constexpr (IsDataTypeDecimal<LeftDataType>)
                         type_res = std::make_shared<LeftDataType>(left.getPrecision(), left.getScale());
                     else if constexpr (IsDataTypeDecimal<RightDataType>)
@@ -841,10 +855,15 @@ public:
             }
             return false;
         });
-        if (!valid)
-            throw Exception("Illegal types " + arguments[0]->getName() + " and " + arguments[1]->getName() + " of arguments of function " + String(name),
-                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-        return type_res;
+
+        if (valid)
+            return type_res;
+
+        throw Exception(
+            "Illegal types " + arguments[0]->getName() +
+            " and " + arguments[1]->getName() +
+            " of arguments of function " + String(name),
+            ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
     }
 
     ColumnPtr executeFixedString(const ColumnsWithTypeAndName & arguments) const
diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index 7552af7dbf1..18c1927b4a5 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -25,7 +25,7 @@ struct MultiplyImpl
             return static_cast<Result>(a) * b;
     }
 
-    /// Apply operation and check overflow. It's used for Deciamal operations. @returns true if overflowed, false otherwise.
+    /// Apply operation and check overflow. It's used for Decimal operations. @returns true if overflowed, false otherwise.
     template <typename Result = ResultType>
     static inline bool apply(A a, B b, Result & c)
     {

From c63c7137abd0d9d213efb0b777564c807bef511d Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 16 Dec 2020 15:35:01 +0300
Subject: [PATCH 099/504] Try use clang-10 for tsan.

---
 tests/ci/ci_config.json | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/tests/ci/ci_config.json b/tests/ci/ci_config.json
index 4e5f36a5f5d..2d788b0a598 100644
--- a/tests/ci/ci_config.json
+++ b/tests/ci/ci_config.json
@@ -52,7 +52,7 @@
             "with_coverage": false
         },
         {
-            "compiler": "clang-11",
+            "compiler": "clang-10",
             "build-type": "",
             "sanitizer": "thread",
             "package-type": "deb",
@@ -179,7 +179,7 @@
         },
         "Functional stateful tests (thread)": {
             "required_build_properties": {
-                "compiler": "clang-11",
+                "compiler": "clang-10",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -263,7 +263,7 @@
         },
         "Functional stateless tests (thread)": {
             "required_build_properties": {
-                "compiler": "clang-11",
+                "compiler": "clang-10",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -371,7 +371,7 @@
         },
         "Stress test (thread)": {
             "required_build_properties": {
-                "compiler": "clang-11",
+                "compiler": "clang-10",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -419,7 +419,7 @@
         },
         "Integration tests (thread)": {
             "required_build_properties": {
-                "compiler": "clang-11",
+                "compiler": "clang-10",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -539,7 +539,7 @@
         },
         "Unit tests TSAN": {
             "required_build_properties": {
-                "compiler": "clang-11",
+                "compiler": "clang-10",
                 "package_type": "binary",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -551,7 +551,7 @@
         },
         "Unit tests UBSAN": {
             "required_build_properties": {
-                "compiler": "clang-11",
+                "compiler": "clang-10",
                 "package_type": "binary",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",

From e335074c39ce8be0cba7769c9106648eaa84df66 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Wed, 16 Dec 2020 15:57:47 +0300
Subject: [PATCH 100/504] works across the blocks

---
 src/Processors/QueryPlan/WindowStep.cpp       | 16 ------
 src/Processors/Transforms/WindowTransform.cpp | 55 ++++++++++++++++---
 src/Processors/Transforms/WindowTransform.h   | 11 +++-
 3 files changed, 57 insertions(+), 25 deletions(-)

diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
index 3eccffc5828..8b235eca39e 100644
--- a/src/Processors/QueryPlan/WindowStep.cpp
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -25,22 +25,6 @@ static ITransformingStep::Traits getTraits()
     };
 }
 
-static ITransformingStep::Traits getJoinTraits()
-{
-    return ITransformingStep::Traits
-    {
-        {
-            .preserves_distinct_columns = false,
-            .returns_single_stream = false,
-            .preserves_number_of_streams = true,
-            .preserves_sorting = false,
-        },
-        {
-            .preserves_number_of_rows = false,
-        }
-    };
-}
-
 static Block addWindowFunctionColumns(const Block & block,
     std::vector<WindowFunctionDescription> window_functions)
 {
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index 1084d96cddd..07eb8118a3b 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -54,6 +54,8 @@ WindowTransform::WindowTransform(const Block & input_header_,
         partition_by_indices.push_back(
             input_header.getPositionByName(column.column_name));
     }
+    partition_start_columns.resize(partition_by_indices.size(), nullptr);
+    partition_start_row = 0;
 }
 
 WindowTransform::~WindowTransform()
@@ -87,6 +89,7 @@ void WindowTransform::transform(Chunk & chunk)
         const auto * a = f.aggregate_function.get();
 
         //*
+        // Create the resulting column.
         ColumnWithTypeAndName column_with_type;
         column_with_type.name = f.column_name;
         column_with_type.type = a->getReturnType();
@@ -97,21 +100,57 @@ void WindowTransform::transform(Chunk & chunk)
         {
             // Check whether the new partition has started and reinitialize the
             // aggregate function states.
-            if (row > 0)
+            assert(partition_start_columns.size() == partition_by_indices.size());
+            if (partition_start_columns.size() == 0)
             {
-                for (const size_t column_index : partition_by_indices)
+                // No PARTITION BY at all, do nothing.
+            }
+            else if (partition_start_columns[0] == nullptr)
+            {
+                // This is the first partition.
+                partition_start_columns.clear();
+                for (const auto i : partition_by_indices)
                 {
-                    const auto * column = block.getColumns()[column_index].get();
-                    if (column->compareAt(row, row - 1, *column,
+                    partition_start_columns.push_back(block.getColumns()[i]);
+                }
+                partition_start_row = row;
+            }
+            else
+            {
+                // Check whether the new partition started, by comparing all the
+                // PARTITION BY columns.
+                size_t first_inequal_column = 0;
+                for ( ; first_inequal_column < partition_start_columns.size();
+                      ++first_inequal_column)
+                {
+                    const auto * current_column = block.getColumns()[
+                        partition_by_indices[first_inequal_column]].get();
+
+                    if (current_column->compareAt(row, partition_start_row,
+                        *partition_start_columns[first_inequal_column],
                         1 /* nan_direction_hint */) != 0)
                     {
-                        ws.window_function.aggregate_function->destroy(
-                            ws.aggregate_function_state.data());
-                        ws.window_function.aggregate_function->create(
-                            ws.aggregate_function_state.data());
                         break;
                     }
                 }
+
+                if (first_inequal_column < partition_start_columns.size())
+                {
+                    // The new partition has started.
+                    // 1) Remember it.
+                    partition_start_columns.clear();
+                    for (const auto i : partition_by_indices)
+                    {
+                        partition_start_columns.push_back(block.getColumns()[i]);
+                    }
+                    partition_start_row = row;
+
+                    // 2) Reset the aggregate function states.
+                    ws.window_function.aggregate_function->destroy(
+                        ws.aggregate_function_state.data());
+                    ws.window_function.aggregate_function->create(
+                        ws.aggregate_function_state.data());
+                }
             }
 
             // Update the aggregate function state and save the result.
diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
index 58b1ed13ae0..5e1444b9949 100644
--- a/src/Processors/Transforms/WindowTransform.h
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -51,12 +51,21 @@ public:
     Block input_header;
 
     WindowDescription window_description;
+    // Indices of the PARTITION BY columns in block.
     std::vector<size_t> partition_by_indices;
+    // The columns for PARTITION BY and the row in these columns where the
+    // current partition started. They might be in some of the previous blocks,
+    // so we have to keep the shared ownership of the columns. We don't keep the
+    // entire block to save memory, only the needed columns, in the same order
+    // as the partition_by_indices array.
+    // Can be empty if there is no PARTITION BY.
+    // Columns are nullptr when it is the first partition.
+    std::vector<ColumnPtr> partition_start_columns;
+    size_t partition_start_row = 0;
 
     std::vector<WindowFunctionWorkspace> workspaces;
 
     std::unique_ptr<Arena> arena;
-    std::vector<AlignedBuffer> aggregate_function_data;
 };
 
 }

From d7145a5ee15fa81b64c3204a75e66d40ed70267b Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 16 Dec 2020 18:01:18 +0300
Subject: [PATCH 101/504] working

---
 src/Functions/FunctionBinaryArithmetic.h | 26 +++++++++++++++++++++---
 1 file changed, 23 insertions(+), 3 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index ea6169194e4..76ae6301843 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -974,7 +974,15 @@ public:
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
 
-                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
+                ResultDataType type;
+
+                if constexpr (IsFloatingPoint<NativeResultType>)
+                {
+                    if constexpr (IsFloatingPoint<LeftDataType>)
+                        type = ...
+
+                } else
+                    type = decimalResultType<is_multiply, is_division>(left, right);
 
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;
@@ -987,8 +995,20 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    NativeResultType const_a;
+                    NativeResultType const_b;
+
+                    if constexpr (IsFloatingPoint<NativeResultType> && dec_a)
+                        const_a = DecimalUtils::convertTo<NativeResultType>(
+                            col_left_const->template getValue<T0>(), scale_a);
+                    else
+                        const_a = col_left_const->template getValue<T0>();
+
+                    if constexpr (IsFloatingPoint<NativeResultType> && dec_b)
+                        const_b = DecimalUtils::convertTo<NativeResultType>(
+                            col_right_const->template getValue<T1>(), scale_b);
+                    else
+                        const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From f143a37693027395c873eab0e53085569508c5e1 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 16 Dec 2020 18:11:26 +0300
Subject: [PATCH 102/504] simplified case to multiply only

---
 src/Functions/FunctionBinaryArithmetic.h | 25 +++++++++++-------------
 1 file changed, 11 insertions(+), 14 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 76ae6301843..73fd40db454 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -470,30 +470,36 @@ public:
         Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
 
-        /// e.g Decimal * Float64 = Float64
-        Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
-        Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>,
-
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
         Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
+
         /// number <op> number -> see corresponding impl
         Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
             DataTypeFromFieldType<typename Op::ResultType>>,
+
+        /// e.g Decimal * Float64 = Float64
+        Case<IsOperation<Operation>::multiply, Switch<
+            Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+            Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>>>,
+
         /// Date + Integral -> Date
         /// Integral + Date -> Date
         Case<IsOperation<Operation>::plus, Switch<
             Case<IsIntegral<RightDataType>, LeftDataType>,
             Case<IsIntegral<LeftDataType>, RightDataType>>>,
+
         /// Date - Date     -> Int32
         /// Date - Integral -> Date
         Case<IsOperation<Operation>::minus, Switch<
             Case<std::is_same_v<LeftDataType, RightDataType>, DataTypeInt32>,
             Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, LeftDataType>>>,
+
         /// least(Date, Date) -> Date
         /// greatest(Date, Date) -> Date
         Case<std::is_same_v<LeftDataType, RightDataType> && (IsOperation<Operation>::least || IsOperation<Operation>::greatest),
             LeftDataType>,
+
         /// Date % Int32 -> Int32
         /// Date % Float -> Float64
         Case<IsOperation<Operation>::modulo, Switch<
@@ -501,7 +507,6 @@ public:
             Case<IsDateOrDateTime<LeftDataType> && IsFloatingPoint<RightDataType>, DataTypeFloat64>>>>;
 };
 
-
 template <template <typename, typename> class Op, typename Name, bool valid_on_default_arguments = true>
 class FunctionBinaryArithmetic : public IFunction
 {
@@ -974,15 +979,7 @@ public:
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
 
-                ResultDataType type;
-
-                if constexpr (IsFloatingPoint<NativeResultType>)
-                {
-                    if constexpr (IsFloatingPoint<LeftDataType>)
-                        type = ...
-
-                } else
-                    type = decimalResultType<is_multiply, is_division>(left, right);
+                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
 
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;

From 02baaddc13b5cb40f77b44bcc104421030963012 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 16 Dec 2020 18:20:41 +0300
Subject: [PATCH 103/504] removed some casts

---
 src/Functions/FunctionBinaryArithmetic.h | 16 ++--------------
 1 file changed, 2 insertions(+), 14 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 73fd40db454..ef54c059b47 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -992,20 +992,8 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a;
-                    NativeResultType const_b;
-
-                    if constexpr (IsFloatingPoint<NativeResultType> && dec_a)
-                        const_a = DecimalUtils::convertTo<NativeResultType>(
-                            col_left_const->template getValue<T0>(), scale_a);
-                    else
-                        const_a = col_left_const->template getValue<T0>();
-
-                    if constexpr (IsFloatingPoint<NativeResultType> && dec_b)
-                        const_b = DecimalUtils::convertTo<NativeResultType>(
-                            col_right_const->template getValue<T1>(), scale_b);
-                    else
-                        const_b = col_right_const->template getValue<T1>();
+                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    NativeResultType const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From 17d996054ecd77d59246a26b48ff3f2436c17d3a Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 16 Dec 2020 18:47:24 +0300
Subject: [PATCH 104/504] style fix

---
 src/Functions/FunctionBinaryArithmetic.h | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index ef54c059b47..db68f07603c 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -548,7 +548,13 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+        return castType(left, [&](const auto & left_)
+        {
+            return castType(right, [&](const auto & right_)
+            {
+                return f(left_, right_);
+            });
+        });
     }
 
     static FunctionOverloadResolverPtr
@@ -1097,10 +1103,12 @@ public:
         const auto * left_generic = left_argument.type.get();
         const auto * right_generic = right_argument.type.get();
         ColumnPtr res;
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)

From 05c621c16ca47122e95c5c3f709284c82e25adbb Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 16 Dec 2020 19:49:02 +0300
Subject: [PATCH 105/504] Revert "Try use clang-10 for tsan."

This reverts commit c63c7137abd0d9d213efb0b777564c807bef511d.
---
 tests/ci/ci_config.json | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/tests/ci/ci_config.json b/tests/ci/ci_config.json
index 2d788b0a598..4e5f36a5f5d 100644
--- a/tests/ci/ci_config.json
+++ b/tests/ci/ci_config.json
@@ -52,7 +52,7 @@
             "with_coverage": false
         },
         {
-            "compiler": "clang-10",
+            "compiler": "clang-11",
             "build-type": "",
             "sanitizer": "thread",
             "package-type": "deb",
@@ -179,7 +179,7 @@
         },
         "Functional stateful tests (thread)": {
             "required_build_properties": {
-                "compiler": "clang-10",
+                "compiler": "clang-11",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -263,7 +263,7 @@
         },
         "Functional stateless tests (thread)": {
             "required_build_properties": {
-                "compiler": "clang-10",
+                "compiler": "clang-11",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -371,7 +371,7 @@
         },
         "Stress test (thread)": {
             "required_build_properties": {
-                "compiler": "clang-10",
+                "compiler": "clang-11",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -419,7 +419,7 @@
         },
         "Integration tests (thread)": {
             "required_build_properties": {
-                "compiler": "clang-10",
+                "compiler": "clang-11",
                 "package_type": "deb",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -539,7 +539,7 @@
         },
         "Unit tests TSAN": {
             "required_build_properties": {
-                "compiler": "clang-10",
+                "compiler": "clang-11",
                 "package_type": "binary",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",
@@ -551,7 +551,7 @@
         },
         "Unit tests UBSAN": {
             "required_build_properties": {
-                "compiler": "clang-10",
+                "compiler": "clang-11",
                 "package_type": "binary",
                 "build_type": "relwithdebuginfo",
                 "sanitizer": "thread",

From a6b4dabab6de4340300c977517a4f89bdcc7f646 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 16 Dec 2020 19:51:58 +0300
Subject: [PATCH 106/504] Use ucontext for tsan.

---
 contrib/boost-cmake/CMakeLists.txt | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index d0b3d7f925c..f7c1ce22e90 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -151,9 +151,14 @@ if (NOT EXTERNAL_BOOST_FOUND)
     enable_language(ASM)
     SET(ASM_OPTIONS "-x assembler-with-cpp")
 
-    if (SANITIZE AND SANITIZE STREQUAL "address")
+    if (SANITIZE AND (SANITIZE STREQUAL "address" OR SANITIZE STREQUAL "thread"))
         add_compile_definitions(BOOST_USE_UCONTEXT)
-        add_compile_definitions(BOOST_USE_ASAN)
+
+        if (SANITIZE STREQUAL "address")
+            add_compile_definitions(BOOST_USE_ASAN)
+        elseif (SANITIZE STREQUAL "thread")
+            add_compile_definitions(BOOST_USE_TSAN)
+        endif()
 
         set (SRCS_CONTEXT
                 ${LIBRARY_DIR}/libs/context/src/fiber.cpp

From 5a19e3068e1141533d83d34c8512a2976c8bad6f Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Wed, 16 Dec 2020 19:52:15 +0300
Subject: [PATCH 107/504] Update settings.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Внес поправки в описание настройки.
---
 docs/en/operations/settings/settings.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index c60cd2c2c52..f74e8af3401 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2429,12 +2429,12 @@ Enables special logic to perform merges on replicas.
 Possible values:
 
 -   Positive integer (in seconds).
--   0 — Special merge logic is not executed.
+-   0 — Special merges logic is not executed.
 
 Default value: `0`.
 
 **Special logic to perform merges**
 
-Selects one replica to perform the merge on. Sets the time threshold from the start of the merge. Other replicas wait for the merge to finish, then download the result. If the time threshold passes and the selected replica does not perform merges, then the merge is performed manually.
+Selects one replica to perform the merge on. Sets the time threshold from the start of the merge. Other replicas wait for the merge to finish, then download the result. If the time threshold passes, and the selected replica does not perform merge, then the merge is performed manually.
 
 [Original article](https://clickhouse.tech/docs/en/operations/settings/settings/) <!-- hide -->

From a85d6e2fbc1c640e3942cfe4acadd8b046be104d Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 16 Dec 2020 19:52:29 +0300
Subject: [PATCH 108/504] Try use tsan annotated boost fibre.

---
 contrib/boost | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/boost b/contrib/boost
index a7ceabe4747..4db0de68baa 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit a7ceabe4747ecc3309dd3dcd9de4b29660dfd298
+Subproject commit 4db0de68baa345515bb0468d3d1ced619d2e2c4e

From af83d82a846d59a3325dcfa1e9947ef634f60a21 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 16 Dec 2020 20:06:04 +0300
Subject: [PATCH 109/504] Try use tsan annotated boost fibre.

---
 base/common/defines.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/base/common/defines.h b/base/common/defines.h
index f6d1d5805c3..28490f9731c 100644
--- a/base/common/defines.h
+++ b/base/common/defines.h
@@ -66,6 +66,11 @@
 #    define BOOST_USE_UCONTEXT 1
 #endif
 
+#if defined(THREAD_SANITIZER)
+#    define BOOST_USE_TSAN 1
+#    define BOOST_USE_UCONTEXT 1
+#endif
+
 /// TODO: Strange enough, there is no way to detect UB sanitizer.
 
 /// Explicitly allow undefined behaviour for certain functions. Use it as a function attribute.

From 263aa88cd9a05e872e15a32445e431e2246f02b0 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 16 Dec 2020 23:27:31 +0300
Subject: [PATCH 110/504] Fixing tests.

---
 src/DataStreams/RemoteQueryExecutor.cpp     | 14 ++++++++++----
 src/Processors/Executors/AsyncTaskQueue.cpp | 20 +++++++++++++-------
 src/Processors/Executors/AsyncTaskQueue.h   |  5 ++---
 3 files changed, 25 insertions(+), 14 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index b31b41a36f2..7dacfca8b60 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -210,8 +210,11 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
             return Block();
     }
 
-    if (!read_context)
-        read_context = std::make_unique<ReadContext>(*multiplexed_connections);
+    {
+        std::lock_guard lock(was_cancelled_mutex);
+        if (!read_context)
+            read_context = std::make_unique<ReadContext>(*multiplexed_connections);
+    }
 
     do
     {
@@ -360,8 +363,11 @@ void RemoteQueryExecutor::cancel(std::unique_ptr<ReadContext> * read_context)
     if (!isQueryPending() || hasThrownException())
         return;
 
-    if (read_context && *read_context)
-        (*read_context)->cancel();
+    {
+        std::lock_guard lock(was_cancelled_mutex);
+        if (read_context && *read_context)
+            (*read_context)->cancel();
+    }
 
     tryCancel("Cancelling query");
 }
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index eeb7603022d..2ffc424cb86 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -14,6 +14,7 @@ namespace ErrorCodes
 {
     extern const int CANNOT_OPEN_FILE;
     extern const int CANNOT_READ_FROM_SOCKET;
+    extern const int LOGICAL_ERROR;
 }
 
 
@@ -43,12 +44,14 @@ AsyncTaskQueue::~AsyncTaskQueue()
 
 void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
 {
-    auto it = tasks.insert(tasks.end(), TaskData{thread_number, data, fd, {}});
-    it->self = it;
+    if (tasks.count(data))
+        throw Exception("Task was already added to task queue", ErrorCodes::LOGICAL_ERROR);
+
+    tasks[data] = TaskData{thread_number, data, fd};
 
     epoll_event socket_event;
     socket_event.events = EPOLLIN | EPOLLPRI;
-    socket_event.data.ptr = &(*it);
+    socket_event.data.ptr = data;
 
     if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fd, &socket_event))
         throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
@@ -77,13 +80,16 @@ AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & loc
     if (event.data.ptr == pipe_fd)
         return {};
 
-    auto it = static_cast<TaskData *>(event.data.ptr)->self;
+    auto it = tasks.find(event.data.ptr);
+    if (it == tasks.end())
+        throw Exception("Task was not found in task queue", ErrorCodes::LOGICAL_ERROR);
 
-    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, it->fd, &event))
+    auto res = it->second;
+    tasks.erase(it);
+
+    if (-1 == epoll_ctl(epoll_fd, EPOLL_CTL_DEL, res.fd, &event))
         throwFromErrno("Cannot remove socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
 
-    auto res = *it;
-    tasks.erase(it);
     return res;
 }
 
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index af616ca48be..c8eb418955c 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -2,7 +2,7 @@
 #include <cstddef>
 #include <mutex>
 #include <atomic>
-#include <list>
+#include <unordered_map>
 
 namespace DB
 {
@@ -18,7 +18,6 @@ public:
         void * data = nullptr;
         int fd = -1;
 
-        std::list<TaskData>::iterator self;
 
         explicit operator bool() const { return data; }
     };
@@ -27,7 +26,7 @@ private:
     int epoll_fd;
     int pipe_fd[2];
     std::atomic_bool is_finished = false;
-    std::list<TaskData> tasks;
+    std::unordered_map<void *, TaskData> tasks;
 
 public:
     AsyncTaskQueue();

From e53281e1763e77016fb463f07916a588afb51b41 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Wed, 16 Dec 2020 23:38:28 +0300
Subject: [PATCH 111/504] Update distribution_queue.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Исправил this на these.
---
 docs/en/operations/system-tables/distribution_queue.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/en/operations/system-tables/distribution_queue.md b/docs/en/operations/system-tables/distribution_queue.md
index 9068e7e860c..fdc6a134da2 100644
--- a/docs/en/operations/system-tables/distribution_queue.md
+++ b/docs/en/operations/system-tables/distribution_queue.md
@@ -1,6 +1,6 @@
 # system.distribution_queue {#system_tables-distribution_queue}
 
-Contains information about local files that are in the queue to be sent to the shards. This local files contain new parts that are created by inserting new data into the Distributed table in asynchronous mode.
+Contains information about local files that are in the queue to be sent to the shards. These local files contain new parts that are created by inserting new data into the Distributed table in asynchronous mode.
 
 Columns:
 

From fef24c81e3b3199b76085ae415cdbe46c2a2b2ef Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 17 Dec 2020 00:44:05 +0300
Subject: [PATCH 112/504] some grammar

---
 src/Interpreters/ActionsVisitor.cpp           |  22 ++-
 src/Interpreters/AggregateDescription.h       |   1 -
 src/Interpreters/ExpressionAnalyzer.cpp       | 176 +++++-------------
 .../ExtractExpressionInfoVisitor.cpp          |   8 +-
 src/Interpreters/GetAggregatesVisitor.h       |  17 +-
 src/Interpreters/MonotonicityCheckVisitor.h   |   9 +-
 .../RewriteAnyFunctionVisitor.cpp             |   6 +-
 src/Parsers/ASTFunction.cpp                   |  40 +++-
 src/Parsers/ASTFunction.h                     |   7 +
 src/Parsers/ASTIndexDeclaration.cpp           |   2 +
 src/Parsers/ASTIndexDeclaration.h             |   4 +-
 src/Parsers/ExpressionElementParsers.cpp      |  87 +++++++++
 src/Parsers/ExpressionElementParsers.h        |   9 +
 src/Parsers/IAST.h                            |   9 +-
 src/Parsers/New/AST/TableElementExpr.cpp      |   1 +
 src/Storages/IndicesDescription.cpp           |   1 +
 src/Storages/MergeTree/MergeTreeIndices.h     |   1 -
 17 files changed, 234 insertions(+), 166 deletions(-)

diff --git a/src/Interpreters/ActionsVisitor.cpp b/src/Interpreters/ActionsVisitor.cpp
index be22af57f32..f3066e6dc2a 100644
--- a/src/Interpreters/ActionsVisitor.cpp
+++ b/src/Interpreters/ActionsVisitor.cpp
@@ -735,17 +735,25 @@ void ActionsMatcher::visit(const ASTFunction & node, const ASTPtr & ast, Data &
         }
     }
 
-    if (AggregateFunctionFactory::instance().isAggregateFunctionName(node.name))
-        return;
-
-    /// FIXME need proper grammar for window functions. For now, ignore it --
-    /// the resulting column is added in ExpressionAnalyzer, similar to the
-    /// aggregate functions.
-    if (node.name == "window")
+    if (node.is_window_function)
     {
+        // Also add columns from PARTITION BY and ORDER BY of window functions.
+        // Requiring a constant reference to a shared pointer to non-const AST
+        // doesn't really look sane, but the visitor does indeed require it.
+        visit(node.window_partition_by->clone(), data);
+        visit(node.window_order_by->clone(), data);
+
+        // Don't need to do anything more for window functions here -- the
+        // resulting column is added in ExpressionAnalyzer, similar to the
+        // aggregate functions.
         return;
     }
 
+    // An aggregate function can also be calculated as a window function, but we
+    // checked for it above, so no need to do anything more.
+    if (AggregateFunctionFactory::instance().isAggregateFunctionName(node.name))
+        return;
+
     FunctionOverloadResolverPtr function_builder;
     try
     {
diff --git a/src/Interpreters/AggregateDescription.h b/src/Interpreters/AggregateDescription.h
index f4ac0cf3898..61e684a1458 100644
--- a/src/Interpreters/AggregateDescription.h
+++ b/src/Interpreters/AggregateDescription.h
@@ -31,7 +31,6 @@ struct WindowFunctionDescription
 {
     std::string window_name;
     std::string column_name;
-    const ASTFunction * wrapper_node;
     const ASTFunction * function_node;
     AggregateFunctionPtr aggregate_function;
     Array function_parameters;
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 09a4855f74f..346c287f239 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -469,63 +469,14 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
     return !aggregates().empty();
 }
 
-// Parses order by & partition by from window() wrapper function.
-// Remove this when we have proper grammar.
-static SortDescription windowArgumentToSortDescription(IAST* ast, const WindowDescription & w)
-{
-    SortDescription result;
-    if (const auto * as_tuple = ast->as<ASTFunction>();
-        as_tuple
-        && as_tuple->name == "tuple"
-        && as_tuple->arguments)
-    {
-        // untuple it
-        for (const auto & element_ast
-            : as_tuple->arguments->children)
-        {
-            const auto * with_alias = dynamic_cast<
-                const ASTWithAlias *>(element_ast.get());
-            if (!with_alias)
-            {
-                throw Exception(ErrorCodes::BAD_ARGUMENTS,
-                    "(1) Expected column in PARTITION BY"
-                    " for window '{}', got '{}'",
-                    w.window_name,
-                    element_ast->formatForErrorMessage());
-            }
-            result.push_back(
-                SortColumnDescription(
-                    with_alias->getColumnName(),
-                    1 /* direction */,
-                    1 /* nulls_direction */));
-        }
-    }
-    else if (const auto * with_alias
-        = dynamic_cast<const ASTWithAlias *>(ast))
-    {
-        result.push_back(
-            SortColumnDescription(
-                with_alias->getColumnName(),
-                1 /* direction */,
-                1 /* nulls_direction */));
-    }
-    else
-    {
-        throw Exception(ErrorCodes::BAD_ARGUMENTS,
-            "(2) Expected tuple or column in PARTITION BY"
-            " for window '{}', got '{}'",
-            w.window_name,
-            ast->formatForErrorMessage());
-    }
-
-    return result;
-}
 
 bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 {
-    for (const ASTFunction * wrapper_node : windowFunctions())
+    for (const ASTFunction * function_node : windowFunctions())
     {
-        fmt::print(stderr, "window function ast: {}\n", wrapper_node->dumpTree());
+        fmt::print(stderr, "window function ast: {}\n", function_node->dumpTree());
+
+        assert(function_node->is_window_function);
 
         // FIXME not thread-safe, should use a per-query counter.
         static int window_index = 1;
@@ -533,38 +484,44 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
         WindowDescription window_description;
         window_description.window_name = fmt::format("window_{}", window_index++);
 
-        const auto * elist = wrapper_node->arguments
-            ? wrapper_node->arguments->as<const ASTExpressionList>()
-            : nullptr;
-        if (elist)
+        if (function_node->window_partition_by)
         {
-            if (elist->children.size() >= 2)
+            for (const auto & column_ast
+                : function_node->window_partition_by->children)
             {
-                const auto partition_by_ast = elist->children[1];
-                fmt::print(stderr, "partition by ast {}\n",
-                    partition_by_ast->dumpTree());
-
-                window_description.partition_by = windowArgumentToSortDescription(
-                    partition_by_ast.get(), window_description);
-            }
-
-            if (elist->children.size() == 3)
-            {
-                const auto order_by_ast = elist->children[2];
-                fmt::print(stderr, "order by ast {}\n",
-                    order_by_ast->dumpTree());
-
-                window_description.order_by = windowArgumentToSortDescription(
-                    order_by_ast.get(), window_description);
-            }
-
-            if (elist->children.size() > 3)
-            {
-                throw Exception(ErrorCodes::BAD_ARGUMENTS,
-                    "Too many arguments to window function '{}'",
-                    wrapper_node->formatForErrorMessage());
+                const auto * with_alias = dynamic_cast<const ASTWithAlias *>(
+                    column_ast.get());
+                if (!with_alias)
+                {
+                    throw Exception(ErrorCodes::BAD_ARGUMENTS,
+                        "Expected a column in PARTITION BY for window '{}',"
+                        " got '{}'", window_description.window_name,
+                        column_ast->formatForErrorMessage());
+                }
+                window_description.partition_by.push_back(
+                    SortColumnDescription(
+                        with_alias->getColumnName(), 1 /* direction */,
+                        1 /* nulls_direction */));
             }
         }
+
+        if (function_node->window_order_by)
+        {
+            for (const auto & column_ast
+                : function_node->window_order_by->children)
+            {
+                // Parser should have checked that we have a proper element here.
+                const auto & order_by_element
+                    = column_ast->as<ASTOrderByElement &>();
+                // Ignore collation for now.
+                window_description.order_by.push_back(
+                    SortColumnDescription(
+                        order_by_element.children.front()->getColumnName(),
+                        order_by_element.direction,
+                        order_by_element.nulls_direction));
+            }
+        }
+
         window_description.full_sort_description = window_description.partition_by;
         window_description.full_sort_description.insert(
             window_description.full_sort_description.end(),
@@ -573,9 +530,7 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 
         WindowFunctionDescription window_function;
         window_function.window_name = window_description.window_name;
-        window_function.wrapper_node = wrapper_node;
-        window_function.function_node
-            = &elist->children.at(0)->as<ASTFunction &>();
+        window_function.function_node = function_node;
         window_function.column_name
             = window_function.function_node->getColumnName();
         window_function.function_parameters
@@ -584,17 +539,10 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
                     window_function.function_node->parameters)
                 : Array();
 
-        // We have to fill actions for window function arguments, so that we are
-        // then able to find their argumen types. The `actions` passed to this
-        // functions are temporary and are discarded.
-        // The same calculation is done in appendWindowFunctionsArguments.
-        getRootActionsNoMakeSet(wrapper_node->arguments, true /* no subqueries */,
-            actions);
-        // We have to separately get actions for the arguments of the aggregate
-        // function we calculate over window, because the ActionsVisitor does
-        // not descend into aggregate functions.
-        getRootActionsNoMakeSet(window_function.function_node->arguments,
-            true /* no subqueries */, actions);
+        // Requiring a constant reference to a shared pointer to non-const AST
+        // doesn't really look sane, but the visitor does indeed require it.
+        getRootActionsNoMakeSet(window_function.function_node->clone(),
+            true, actions);
 
         const ASTs & arguments
             = window_function.function_node->arguments->children;
@@ -1013,16 +961,9 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
 
     for (const auto & f : window_functions)
     {
-        // Not sure why NoMakeSet, copied from aggregate functions.
-        getRootActionsNoMakeSet(f.wrapper_node->arguments,
-            true /* no subqueries */, step.actions());
-
-        // We have to separately get actions for the arguments of the aggregate
-        // function we calculate over window, because the ActionsVisitor does
-        // not descend into aggregate functions.
-        // Not sure why NoMakeSet, copied from aggregate functions.
-        getRootActionsNoMakeSet(f.function_node->arguments,
-            true /* no subqueries */, step.actions());
+        // Requiring a constant reference to a shared pointer to non-const AST
+        // doesn't really look sane, but the visitor does indeed require it.
+        getRootActionsNoMakeSet(f.function_node->clone(), true, step.actions());
 
         // Add empty INPUT with window function name.
         // It is an aggregate function, so it won't be added by getRootActions.
@@ -1081,13 +1022,6 @@ void SelectQueryExpressionAnalyzer::appendSelect(ExpressionActionsChain & chain,
 
     for (const auto & child : select_query->select()->children)
     {
-        /// FIXME add proper grammar for window functions
-        if (const auto * as_function = child->as<ASTFunction>();
-            as_function && as_function->name == "window")
-        {
-            continue;
-        }
-
         step.required_output.push_back(child->getColumnName());
     }
 }
@@ -1184,13 +1118,6 @@ ActionsDAGPtr SelectQueryExpressionAnalyzer::appendProjectResult(ExpressionActio
     ASTs asts = select_query->select()->children;
     for (const auto & ast : asts)
     {
-        /// FIXME add proper grammar for window functions
-        if (const auto * as_function = ast->as<ASTFunction>();
-            as_function && as_function->name == "window")
-        {
-            continue;
-        }
-
         String result_name = ast->getAliasOrColumnName();
         if (required_result_columns.empty() || required_result_columns.count(result_name))
         {
@@ -1226,16 +1153,6 @@ ActionsDAGPtr SelectQueryExpressionAnalyzer::appendProjectResult(ExpressionActio
         }
     }
 
-    for (const auto & f : window_functions)
-    {
-        if (required_result_columns.empty()
-            || required_result_columns.count(f.column_name))
-        {
-            result_columns.emplace_back(f.column_name, f.column_name);
-            step.required_output.push_back(f.column_name);
-        }
-    }
-
     auto actions = chain.getLastActions();
     actions->project(result_columns);
     return actions;
@@ -1651,7 +1568,6 @@ std::string WindowFunctionDescription::dump() const
 {
     std::stringstream ss;
     ss << "window function '" << column_name << "' over '" << window_name <<"\n";
-    ss << "wrapper node " << wrapper_node->dumpTree() << "\n";
     ss << "function node " << function_node->dumpTree() << "\n";
     ss << "aggregate function '" << aggregate_function->getName() << "'\n";
     if (function_parameters.size())
diff --git a/src/Interpreters/ExtractExpressionInfoVisitor.cpp b/src/Interpreters/ExtractExpressionInfoVisitor.cpp
index c730f49fe90..579fd13642f 100644
--- a/src/Interpreters/ExtractExpressionInfoVisitor.cpp
+++ b/src/Interpreters/ExtractExpressionInfoVisitor.cpp
@@ -19,9 +19,15 @@ void ExpressionInfoMatcher::visit(const ASTPtr & ast, Data & data)
 void ExpressionInfoMatcher::visit(const ASTFunction & ast_function, const ASTPtr &, Data & data)
 {
     if (ast_function.name == "arrayJoin")
+    {
         data.is_array_join = true;
-    else if (AggregateFunctionFactory::instance().isAggregateFunctionName(ast_function.name))
+    }
+    else if (!ast_function.is_window_function
+        && AggregateFunctionFactory::instance().isAggregateFunctionName(
+            ast_function.name))
+    {
         data.is_aggregate_function = true;
+    }
     else
     {
         const auto & function = FunctionFactory::instance().tryGet(ast_function.name, data.context);
diff --git a/src/Interpreters/GetAggregatesVisitor.h b/src/Interpreters/GetAggregatesVisitor.h
index 966b301a00b..94510f0aaa2 100644
--- a/src/Interpreters/GetAggregatesVisitor.h
+++ b/src/Interpreters/GetAggregatesVisitor.h
@@ -31,9 +31,8 @@ public:
             return false;
         if (auto * func = node->as<ASTFunction>())
         {
-            if (isAggregateFunction(func->name)
-                // FIXME temporary hack while we don't have grammar
-                || func->name == "window")
+            if (isAggregateFunction(*func)
+                || func->is_window_function)
             {
                 return false;
             }
@@ -50,7 +49,7 @@ public:
 private:
     static void visit(const ASTFunction & node, const ASTPtr &, Data & data)
     {
-        if (isAggregateFunction(node.name))
+        if (isAggregateFunction(node))
         {
             if (data.assert_no_aggregates)
                 throw Exception("Aggregate function " + node.getColumnName()  + " is found " + String(data.assert_no_aggregates) + " in query",
@@ -63,7 +62,7 @@ private:
             data.uniq_names.insert(column_name);
             data.aggregates.push_back(&node);
         }
-        else if (node.name == "window")
+        else if (node.is_window_function)
         {
             if (data.assert_no_windows)
                 throw Exception("Window function " + node.getColumnName()  + " is found " + String(data.assert_no_windows) + " in query",
@@ -78,9 +77,13 @@ private:
         }
     }
 
-    static bool isAggregateFunction(const String & name)
+    static bool isAggregateFunction(const ASTFunction & node)
     {
-        return AggregateFunctionFactory::instance().isAggregateFunctionName(name);
+        // Aggregate functions can also be calculated as window functions, but
+        // here we are interested in aggregate functions calculated in GROUP BY.
+        return !node.is_window_function
+            && AggregateFunctionFactory::instance().isAggregateFunctionName(
+                node.name);
     }
 };
 
diff --git a/src/Interpreters/MonotonicityCheckVisitor.h b/src/Interpreters/MonotonicityCheckVisitor.h
index 137f8d25b4a..87571a44eb0 100644
--- a/src/Interpreters/MonotonicityCheckVisitor.h
+++ b/src/Interpreters/MonotonicityCheckVisitor.h
@@ -43,9 +43,14 @@ public:
             if (group_by_function_hashes.count(key))
                 return false;
 
-            /// if ORDER BY contains aggregate function it shouldn't be optimized
-            if (AggregateFunctionFactory::instance().isAggregateFunctionName(ast_function.name))
+            /// if ORDER BY contains aggregate function or window functions, it
+            /// shouldn't be optimized
+            if (ast_function.is_window_function
+                || AggregateFunctionFactory::instance().isAggregateFunctionName(
+                    ast_function.name))
+            {
                 return false;
+            }
 
             return true;
         }
diff --git a/src/Interpreters/RewriteAnyFunctionVisitor.cpp b/src/Interpreters/RewriteAnyFunctionVisitor.cpp
index 7c3f1bf31b2..3e5d0b693cf 100644
--- a/src/Interpreters/RewriteAnyFunctionVisitor.cpp
+++ b/src/Interpreters/RewriteAnyFunctionVisitor.cpp
@@ -38,8 +38,12 @@ bool extractIdentifiers(const ASTFunction & func, std::unordered_set<ASTPtr *> &
             if (arg_func->name == "lambda")
                 return false;
 
-            if (AggregateFunctionFactory::instance().isAggregateFunctionName(arg_func->name))
+            if ( arg_func->is_window_function
+                || AggregateFunctionFactory::instance().isAggregateFunctionName(
+                    arg_func->name))
+            {
                 return false;
+            }
 
             if (!extractIdentifiers(*arg_func, identifiers))
                 return false;
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index 4cca8ad0807..667fde1941a 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -1,13 +1,14 @@
-#include <Common/typeid_cast.h>
-#include <Parsers/ASTLiteral.h>
 #include <Parsers/ASTFunction.h>
-#include <Parsers/ASTWithAlias.h>
-#include <Parsers/ASTSubquery.h>
-#include <IO/WriteHelpers.h>
-#include <IO/WriteBufferFromString.h>
-#include <Common/SipHash.h>
-#include <IO/Operators.h>
 
+#include <Common/SipHash.h>
+#include <Common/typeid_cast.h>
+#include <IO/Operators.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/WriteHelpers.h>
+#include <Parsers/ASTIdentifier.h>
+#include <Parsers/ASTLiteral.h>
+#include <Parsers/ASTSubquery.h>
+#include <Parsers/ASTWithAlias.h>
 
 namespace DB
 {
@@ -37,6 +38,14 @@ void ASTFunction::appendColumnNameImpl(WriteBuffer & ostr) const
             (*it)->appendColumnName(ostr);
         }
     writeChar(')', ostr);
+
+    /*
+    if (window_name)
+    {
+        writeCString(" over ", ostr);
+        window_name->appendColumnName(ostr);
+    }
+    */
 }
 
 /** Get the text that identifies this element. */
@@ -53,6 +62,21 @@ ASTPtr ASTFunction::clone() const
     if (arguments) { res->arguments = arguments->clone(); res->children.push_back(res->arguments); }
     if (parameters) { res->parameters = parameters->clone(); res->children.push_back(res->parameters); }
 
+    if (window_name)
+    {
+        res->set(res->window_name, window_name->clone());
+    }
+
+    if (window_partition_by)
+    {
+        res->set(res->window_partition_by, window_partition_by->clone());
+    }
+
+    if (window_order_by)
+    {
+        res->set(res->window_order_by, window_order_by->clone());
+    }
+
     return res;
 }
 
diff --git a/src/Parsers/ASTFunction.h b/src/Parsers/ASTFunction.h
index a6e3834ac1a..02f6f46f020 100644
--- a/src/Parsers/ASTFunction.h
+++ b/src/Parsers/ASTFunction.h
@@ -8,6 +8,8 @@
 namespace DB
 {
 
+class ASTIdentifier;
+
 /** AST for function application or operator.
   */
 class ASTFunction : public ASTWithAlias
@@ -18,6 +20,11 @@ public:
     /// parameters - for parametric aggregate function. Example: quantile(0.9)(x) - what in first parens are 'parameters'.
     ASTPtr parameters;
 
+    bool is_window_function = false;
+    ASTIdentifier * window_name;
+    ASTExpressionList * window_partition_by;
+    ASTExpressionList * window_order_by;
+
     /// do not print empty parentheses if there are no args - compatibility with new AST for data types and engine names.
     bool no_empty_args = false;
 
diff --git a/src/Parsers/ASTIndexDeclaration.cpp b/src/Parsers/ASTIndexDeclaration.cpp
index 0e8f0d0f7e8..d8ebf825674 100644
--- a/src/Parsers/ASTIndexDeclaration.cpp
+++ b/src/Parsers/ASTIndexDeclaration.cpp
@@ -1,6 +1,8 @@
 #include <Parsers/ASTIndexDeclaration.h>
+
 #include <Common/quoteString.h>
 #include <IO/Operators.h>
+#include <Parsers/ASTFunction.h>
 
 
 namespace DB
diff --git a/src/Parsers/ASTIndexDeclaration.h b/src/Parsers/ASTIndexDeclaration.h
index 64ef6eb2db1..8416ec6b0a6 100644
--- a/src/Parsers/ASTIndexDeclaration.h
+++ b/src/Parsers/ASTIndexDeclaration.h
@@ -1,12 +1,12 @@
 #pragma once
 
-#include <Parsers/ASTFunction.h>
 #include <Parsers/IAST.h>
 
-
 namespace DB
 {
 
+class ASTFunction;
+
 /** name BY expr TYPE typename(args) GRANULARITY int in create query
   */
 class ASTIndexDeclaration : public IAST
diff --git a/src/Parsers/ExpressionElementParsers.cpp b/src/Parsers/ExpressionElementParsers.cpp
index 6fbe9abc96e..0ce2c38969b 100644
--- a/src/Parsers/ExpressionElementParsers.cpp
+++ b/src/Parsers/ExpressionElementParsers.cpp
@@ -263,6 +263,7 @@ bool ParserFunction::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
     ParserKeyword distinct("DISTINCT");
     ParserExpressionList contents(false);
     ParserSelectWithUnionQuery select;
+    ParserKeyword over("OVER");
 
     bool has_distinct_modifier = false;
 
@@ -382,10 +383,96 @@ bool ParserFunction::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
         function_node->children.push_back(function_node->parameters);
     }
 
+    if (over.ignore(pos, expected))
+    {
+        function_node->is_window_function = true;
+
+        // We are slightly breaking the parser interface by parsing the window
+        // definition into an existing ASTFunction, hence this weirdness.
+        ASTPtr function_node_as_iast = function_node;
+
+        ParserWindowDefinition window_definition;
+        if (!window_definition.parse(pos, function_node_as_iast, expected))
+        {
+            return false;
+        }
+
+        fmt::print(stderr, "window function ast {}\n", function_node->dumpTree());
+    }
+
     node = function_node;
     return true;
 }
 
+bool ParserWindowDefinition::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
+{
+    ASTFunction * function = dynamic_cast<ASTFunction *>(node.get());
+
+    // Variant 1:
+    // function_name ( * ) OVER window_name
+    // FIXME doesn't work anyway for now -- never used anywhere, window names
+    // can't be defined, and TreeRewriter thinks the window name is a column so
+    // the query fails.
+    if (pos->type != TokenType::OpeningRoundBracket)
+    {
+        ASTPtr window_name_ast;
+        ParserIdentifier window_name_parser;
+        if (window_name_parser.parse(pos, window_name_ast, expected))
+        {
+            function->set(function->window_name, window_name_ast);
+            return true;
+        }
+        else
+        {
+            return false;
+        }
+    }
+    ++pos;
+
+    // Variant 2:
+    // function_name ( * ) OVER ( window_definition )
+    ParserKeyword keyword_partition_by("PARTITION BY");
+    ParserNotEmptyExpressionList columns_partition_by(
+        false /* we don't allow declaring aliases here*/);
+    ParserKeyword keyword_order_by("ORDER BY");
+    ParserOrderByExpressionList columns_order_by;
+
+    if (keyword_partition_by.ignore(pos, expected))
+    {
+        ASTPtr partition_by_ast;
+        if (columns_partition_by.parse(pos, partition_by_ast, expected))
+        {
+            function->set(function->window_partition_by, partition_by_ast);
+        }
+        else
+        {
+            return false;
+        }
+    }
+
+    if (keyword_order_by.ignore(pos, expected))
+    {
+        ASTPtr order_by_ast;
+        if (columns_order_by.parse(pos, order_by_ast, expected))
+        {
+            function->set(function->window_order_by, order_by_ast);
+        }
+        else
+        {
+            return false;
+        }
+    }
+
+    if (pos->type != TokenType::ClosingRoundBracket)
+    {
+        expected.add(pos, "')'");
+        return false;
+    }
+    ++pos;
+
+    return true;
+}
+
 bool ParserCodecDeclarationList::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
 {
     return ParserList(std::make_unique<ParserIdentifierWithOptionalParameters>(),
diff --git a/src/Parsers/ExpressionElementParsers.h b/src/Parsers/ExpressionElementParsers.h
index 4713a7ea8a6..81db411a802 100644
--- a/src/Parsers/ExpressionElementParsers.h
+++ b/src/Parsers/ExpressionElementParsers.h
@@ -124,6 +124,15 @@ protected:
     bool allow_function_parameters;
 };
 
+struct ParserWindowDefinition : public IParserBase
+{
+
+protected:
+    const char * getName() const override { return "window definition"; }
+    bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override;
+
+};
+
 class ParserCodecDeclarationList : public IParserBase
 {
 protected:
diff --git a/src/Parsers/IAST.h b/src/Parsers/IAST.h
index 54eaf088797..47cbcfde4d3 100644
--- a/src/Parsers/IAST.h
+++ b/src/Parsers/IAST.h
@@ -19,9 +19,6 @@ namespace DB
 
 namespace ErrorCodes
 {
-    extern const int NOT_A_COLUMN;
-    extern const int UNKNOWN_TYPE_OF_AST_NODE;
-    extern const int UNKNOWN_ELEMENT_IN_AST;
     extern const int LOGICAL_ERROR;
 }
 
@@ -46,7 +43,7 @@ public:
     String getColumnName() const;
     virtual void appendColumnName(WriteBuffer &) const
     {
-        throw Exception("Trying to get name of not a column: " + getID(), ErrorCodes::NOT_A_COLUMN);
+        throw Exception("Trying to get name of not a column: " + getID(), ErrorCodes::LOGICAL_ERROR);
     }
 
     /** Get the alias, if any, or the canonical name of the column, if it is not. */
@@ -58,7 +55,7 @@ public:
     /** Set the alias. */
     virtual void setAlias(const String & /*to*/)
     {
-        throw Exception("Can't set alias of " + getColumnName(), ErrorCodes::UNKNOWN_TYPE_OF_AST_NODE);
+        throw Exception("Can't set alias of " + getColumnName(), ErrorCodes::LOGICAL_ERROR);
     }
 
     /** Get the text that identifies this element. */
@@ -209,7 +206,7 @@ public:
 
     virtual void formatImpl(const FormatSettings & /*settings*/, FormatState & /*state*/, FormatStateStacked /*frame*/) const
     {
-        throw Exception("Unknown element in AST: " + getID(), ErrorCodes::UNKNOWN_ELEMENT_IN_AST);
+        throw Exception("Unknown element in AST: " + getID(), ErrorCodes::LOGICAL_ERROR);
     }
 
     // A simple way to add some user-readable context to an error message.
diff --git a/src/Parsers/New/AST/TableElementExpr.cpp b/src/Parsers/New/AST/TableElementExpr.cpp
index 1336ea06b27..d994ea7eef6 100644
--- a/src/Parsers/New/AST/TableElementExpr.cpp
+++ b/src/Parsers/New/AST/TableElementExpr.cpp
@@ -2,6 +2,7 @@
 
 #include <Parsers/ASTColumnDeclaration.h>
 #include <Parsers/ASTConstraintDeclaration.h>
+#include <Parsers/ASTFunction.h>
 #include <Parsers/ASTIndexDeclaration.h>
 #include <Parsers/New/AST/ColumnExpr.h>
 #include <Parsers/New/AST/ColumnTypeExpr.h>
diff --git a/src/Storages/IndicesDescription.cpp b/src/Storages/IndicesDescription.cpp
index 8adf2be1bd4..dbc95615383 100644
--- a/src/Storages/IndicesDescription.cpp
+++ b/src/Storages/IndicesDescription.cpp
@@ -2,6 +2,7 @@
 #include <Interpreters/TreeRewriter.h>
 #include <Storages/IndicesDescription.h>
 
+#include <Parsers/ASTFunction.h>
 #include <Parsers/ASTIndexDeclaration.h>
 #include <Parsers/formatAST.h>
 #include <Parsers/ParserCreateQuery.h>
diff --git a/src/Storages/MergeTree/MergeTreeIndices.h b/src/Storages/MergeTree/MergeTreeIndices.h
index 28795ae46b5..c7b9dfb123e 100644
--- a/src/Storages/MergeTree/MergeTreeIndices.h
+++ b/src/Storages/MergeTree/MergeTreeIndices.h
@@ -10,7 +10,6 @@
 #include <Storages/SelectQueryInfo.h>
 #include <Storages/MergeTree/MarkRange.h>
 #include <Interpreters/ExpressionActions.h>
-#include <Parsers/ASTIndexDeclaration.h>
 #include <DataTypes/DataTypeLowCardinality.h>
 
 constexpr auto INDEX_FILE_PREFIX = "skp_idx_";

From bf9c945ce24497c530952f466ec491b5cf33c47d Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 17 Dec 2020 10:27:54 +0300
Subject: [PATCH 113/504] Try to fix ubsan crash.

---
 src/Processors/Executors/AsyncTaskQueue.cpp | 8 +++++---
 src/Processors/Executors/AsyncTaskQueue.h   | 3 ++-
 2 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/AsyncTaskQueue.cpp
index 2ffc424cb86..621f526d80b 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/AsyncTaskQueue.cpp
@@ -44,10 +44,11 @@ AsyncTaskQueue::~AsyncTaskQueue()
 
 void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
 {
-    if (tasks.count(data))
+    std::uintptr_t key = reinterpret_cast<uintptr_t>(data);
+    if (tasks.count(key))
         throw Exception("Task was already added to task queue", ErrorCodes::LOGICAL_ERROR);
 
-    tasks[data] = TaskData{thread_number, data, fd};
+    tasks[key] = TaskData{thread_number, data, fd};
 
     epoll_event socket_event;
     socket_event.events = EPOLLIN | EPOLLPRI;
@@ -80,7 +81,8 @@ AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & loc
     if (event.data.ptr == pipe_fd)
         return {};
 
-    auto it = tasks.find(event.data.ptr);
+    std::uintptr_t key = reinterpret_cast<uintptr_t>(event.data.ptr);
+    auto it = tasks.find(key);
     if (it == tasks.end())
         throw Exception("Task was not found in task queue", ErrorCodes::LOGICAL_ERROR);
 
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/AsyncTaskQueue.h
index c8eb418955c..1a4bdec5dd3 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/AsyncTaskQueue.h
@@ -1,5 +1,6 @@
 #pragma once
 #include <cstddef>
+#include <cstdint>
 #include <mutex>
 #include <atomic>
 #include <unordered_map>
@@ -26,7 +27,7 @@ private:
     int epoll_fd;
     int pipe_fd[2];
     std::atomic_bool is_finished = false;
-    std::unordered_map<void *, TaskData> tasks;
+    std::unordered_map<std::uintptr_t, TaskData> tasks;
 
 public:
     AsyncTaskQueue();

From 53d5012a20e549d06635c17ece0b11451ea40a9e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 17 Dec 2020 10:32:51 +0300
Subject: [PATCH 114/504] Update boost.

---
 contrib/boost | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/boost b/contrib/boost
index 4db0de68baa..0b98b443aa7 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit 4db0de68baa345515bb0468d3d1ced619d2e2c4e
+Subproject commit 0b98b443aa7bb77d65efd7b23b3b8c8a0ab5f1f3

From a860f128f5d07b022e66b01f31aae4eefd96b917 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 17 Dec 2020 13:07:28 +0300
Subject: [PATCH 115/504] Try fix race in RemoteQueryExecutor.

---
 src/DataStreams/RemoteQueryExecutor.cpp | 26 ++++++++++++-------------
 src/DataStreams/RemoteQueryExecutor.h   |  2 +-
 2 files changed, 13 insertions(+), 15 deletions(-)

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 7dacfca8b60..06d8b0b7e37 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -210,10 +210,13 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
             return Block();
     }
 
+    if (!read_context)
     {
         std::lock_guard lock(was_cancelled_mutex);
-        if (!read_context)
-            read_context = std::make_unique<ReadContext>(*multiplexed_connections);
+        if (was_cancelled)
+            return Block();
+
+        read_context = std::make_unique<ReadContext>(*multiplexed_connections);
     }
 
     do
@@ -312,15 +315,12 @@ void RemoteQueryExecutor::finish(std::unique_ptr<ReadContext> * read_context)
     if (!isQueryPending() || hasThrownException())
         return;
 
-    if (read_context && *read_context)
-        (*read_context)->cancel();
-
     /** If you have not read all the data yet, but they are no longer needed.
       * This may be due to the fact that the data is sufficient (for example, when using LIMIT).
       */
 
     /// Send the request to abort the execution of the request, if not already sent.
-    tryCancel("Cancelling query because enough data has been read");
+    tryCancel("Cancelling query because enough data has been read", read_context);
 
     /// Get the remaining packets so that there is no out of sync in the connections to the replicas.
     Packet packet = multiplexed_connections->drain();
@@ -363,13 +363,7 @@ void RemoteQueryExecutor::cancel(std::unique_ptr<ReadContext> * read_context)
     if (!isQueryPending() || hasThrownException())
         return;
 
-    {
-        std::lock_guard lock(was_cancelled_mutex);
-        if (read_context && *read_context)
-            (*read_context)->cancel();
-    }
-
-    tryCancel("Cancelling query");
+    tryCancel("Cancelling query", read_context);
 }
 
 void RemoteQueryExecutor::sendScalars()
@@ -421,7 +415,7 @@ void RemoteQueryExecutor::sendExternalTables()
     multiplexed_connections->sendExternalTablesData(external_tables_data);
 }
 
-void RemoteQueryExecutor::tryCancel(const char * reason)
+void RemoteQueryExecutor::tryCancel(const char * reason, std::unique_ptr<ReadContext> * read_context)
 {
     {
         /// Flag was_cancelled is atomic because it is checked in read().
@@ -431,6 +425,10 @@ void RemoteQueryExecutor::tryCancel(const char * reason)
             return;
 
         was_cancelled = true;
+
+        if (read_context && *read_context)
+            (*read_context)->cancel();
+
         multiplexed_connections->sendCancel();
     }
 
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 6c668f280a1..d15bd7a2aa0 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -164,7 +164,7 @@ private:
     void sendExternalTables();
 
     /// If wasn't sent yet, send request to cancel all connections to replicas
-    void tryCancel(const char * reason);
+    void tryCancel(const char * reason, std::unique_ptr<ReadContext> * read_context);
 
     /// Returns true if query was sent
     bool isQueryPending() const;

From 9d5998fd10812ae4cf6b786dbe11fb947b6ec35b Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 18:46:01 +0300
Subject: [PATCH 116/504] added debug info

---
 src/Functions/FunctionBinaryArithmetic.h | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index db68f07603c..d97b4dff06d 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -45,6 +45,8 @@
 #    pragma GCC diagnostic pop
 #endif
 
+#include <cassert>
+
 
 namespace DB
 {
@@ -548,6 +550,18 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
+        const bool l = castType(left, [](const auto& c){
+            assert(std::is_same_v<decltype(c), Decimal32>);
+        });
+
+        assert(l);
+
+        const bool r = castType(right, [](const auto& c){
+            assert(std::is_same_v<decltype(c), Float32>);
+        });
+
+        assert(r);
+
         return castType(left, [&](const auto & left_)
         {
             return castType(right, [&](const auto & right_)

From 0d319c8c612285fc249a6aea9382e9aad3b8e7c4 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 18:47:14 +0300
Subject: [PATCH 117/504] fix

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index d97b4dff06d..ae1c344f507 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -551,13 +551,13 @@ class FunctionBinaryArithmetic : public IFunction
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
         const bool l = castType(left, [](const auto& c){
-            assert(std::is_same_v<decltype(c), Decimal32>);
+            assert((std::is_same_v<decltype(c), Decimal32>));
         });
 
         assert(l);
 
         const bool r = castType(right, [](const auto& c){
-            assert(std::is_same_v<decltype(c), Float32>);
+            assert((std::is_same_v<decltype(c), Float32>));
         });
 
         assert(r);

From 49f3b71b5f1c731c98eb0d90421bef968a675e64 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 18:48:22 +0300
Subject: [PATCH 118/504] another fix

---
 src/Functions/FunctionBinaryArithmetic.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index ae1c344f507..bd287bb66cc 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -552,12 +552,14 @@ class FunctionBinaryArithmetic : public IFunction
     {
         const bool l = castType(left, [](const auto& c){
             assert((std::is_same_v<decltype(c), Decimal32>));
+            return true;
         });
 
         assert(l);
 
         const bool r = castType(right, [](const auto& c){
             assert((std::is_same_v<decltype(c), Float32>));
+            return true;
         });
 
         assert(r);

From 3a723f77be0c818fdd082c10fcf88ab3fb8762cf Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 17 Dec 2020 19:09:54 +0300
Subject: [PATCH 119/504] Try fix submodules

---
 .gitmodules | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/.gitmodules b/.gitmodules
index 3b6ff83099e..23deab0eed0 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -188,6 +188,10 @@
 	path = contrib/cyrus-sasl
 	url = https://github.com/cyrusimap/cyrus-sasl
 	branch = cyrus-sasl-2.1
+[submodule "contrib/croaring"]
+	path = contrib/croaring
+	url = https://github.com/RoaringBitmap/CRoaring
+	branch = v0.2.66
 [submodule "contrib/miniselect"]
 	path = contrib/miniselect
 	url = https://github.com/danlark1/miniselect

From 1287ef607a568ddfbcdeb3981653758753630ee2 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 17 Dec 2020 19:26:31 +0300
Subject: [PATCH 120/504] Fix other adaprots

---
 contrib/croaring | 1 +
 1 file changed, 1 insertion(+)
 create mode 160000 contrib/croaring

diff --git a/contrib/croaring b/contrib/croaring
new file mode 160000
index 00000000000..5f20740ec0d
--- /dev/null
+++ b/contrib/croaring
@@ -0,0 +1 @@
+Subproject commit 5f20740ec0de5e153e8f4cb2ab91814e8b291a14

From fcacf5c510c1ea34f3980863c92b34a191523f7d Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 19:31:54 +0300
Subject: [PATCH 121/504] another try

---
 src/Functions/FunctionBinaryArithmetic.h | 12 ++----------
 1 file changed, 2 insertions(+), 10 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index bd287bb66cc..34a0ad1618b 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -550,18 +550,10 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        const bool l = castType(left, [](const auto& c){
-            assert((std::is_same_v<decltype(c), Decimal32>));
-            return true;
-        });
+        const bool l = castType(left, [](const auto& c){ return std::is_same_v<decltype(c), DataTypeDecimal<Decimal32>>; });
+        const bool r = castType(right, [](const auto& c){ return std::is_same_v<decltype(c), DataTypeFloat32>; });
 
         assert(l);
-
-        const bool r = castType(right, [](const auto& c){
-            assert((std::is_same_v<decltype(c), Float32>));
-            return true;
-        });
-
         assert(r);
 
         return castType(left, [&](const auto & left_)

From 737d4fc22fd2565e5551b15f6cd96aa419a271cd Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 19:43:29 +0300
Subject: [PATCH 122/504] debug fix

---
 src/Functions/FunctionBinaryArithmetic.h | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 34a0ad1618b..05989dad0a7 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -550,8 +550,10 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        const bool l = castType(left, [](const auto& c){ return std::is_same_v<decltype(c), DataTypeDecimal<Decimal32>>; });
-        const bool r = castType(right, [](const auto& c){ return std::is_same_v<decltype(c), DataTypeFloat32>; });
+        const bool l = castType(left, [](const auto& c){ 
+            return std::is_same_v<decltype(c), DataTypeDecimal<Decimal32>>; });
+        const bool r = castType(right, [](const auto& c){ 
+            return std::is_same_v<decltype(c), DataTypeFloat32>; });
 
         assert(l);
         assert(r);

From af8a72a45f2162f69c3cc0034af607cc17d288f9 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 20:16:37 +0300
Subject: [PATCH 123/504] fix

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 05989dad0a7..0147ff1de86 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -551,9 +551,9 @@ class FunctionBinaryArithmetic : public IFunction
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
         const bool l = castType(left, [](const auto& c){ 
-            return std::is_same_v<decltype(c), DataTypeDecimal<Decimal32>>; });
+            return std::is_same_v<std::decay_t<decltype(c)>, DataTypeDecimal<Decimal32>>; });
         const bool r = castType(right, [](const auto& c){ 
-            return std::is_same_v<decltype(c), DataTypeFloat32>; });
+            return std::is_same_v<std::decay_t<decltype(c)>, DataTypeFloat32>; });
 
         assert(l);
         assert(r);

From fa6a1d3b4a70f3217df426934b14de4c6a166e6d Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 20:55:35 +0300
Subject: [PATCH 124/504] fix

---
 src/Functions/FunctionBinaryArithmetic.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 0147ff1de86..3d1c53f62b0 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -553,7 +553,7 @@ class FunctionBinaryArithmetic : public IFunction
         const bool l = castType(left, [](const auto& c){ 
             return std::is_same_v<std::decay_t<decltype(c)>, DataTypeDecimal<Decimal32>>; });
         const bool r = castType(right, [](const auto& c){ 
-            return std::is_same_v<std::decay_t<decltype(c)>, DataTypeFloat32>; });
+            return std::is_same_v<std::decay_t<decltype(c)>, DataTypeFloat64>; });
 
         assert(l);
         assert(r);

From 9f217c5df57e6fe8d8ca5d823d4889e39830ffc1 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:11:39 +0300
Subject: [PATCH 125/504] case fix

---
 src/Functions/FunctionBinaryArithmetic.h | 8 --------
 1 file changed, 8 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 3d1c53f62b0..7986b6f0a24 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -550,14 +550,6 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        const bool l = castType(left, [](const auto& c){ 
-            return std::is_same_v<std::decay_t<decltype(c)>, DataTypeDecimal<Decimal32>>; });
-        const bool r = castType(right, [](const auto& c){ 
-            return std::is_same_v<std::decay_t<decltype(c)>, DataTypeFloat64>; });
-
-        assert(l);
-        assert(r);
-
         return castType(left, [&](const auto & left_)
         {
             return castType(right, [&](const auto & right_)

From d24fc58c94e270f48834451f4807aba10fe330fc Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:12:24 +0300
Subject: [PATCH 126/504] case fix 2

---
 src/Functions/FunctionBinaryArithmetic.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 7986b6f0a24..a5e706c75a8 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -472,6 +472,11 @@ public:
         Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
 
+        /// e.g Decimal * Float64 = Float64
+        Case<IsOperation<Operation>::multiply, Switch<
+            Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+            Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>>>,
+
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
         Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
@@ -480,11 +485,6 @@ public:
         Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
             DataTypeFromFieldType<typename Op::ResultType>>,
 
-        /// e.g Decimal * Float64 = Float64
-        Case<IsOperation<Operation>::multiply, Switch<
-            Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
-            Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>>>,
-
         /// Date + Integral -> Date
         /// Integral + Date -> Date
         Case<IsOperation<Operation>::plus, Switch<

From 494e9066c33c9949d2063f8a63ce1f31120c8805 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:17:38 +0300
Subject: [PATCH 127/504] fixing first compile bug

---
 src/Functions/FunctionBinaryArithmetic.h | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index a5e706c75a8..cc030d76537 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1000,8 +1000,20 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    NativeResultType const_a;
+                    NativeResultType const_b;
+
+                    if constexpr (IsFloatingPoint<ResultDataType> && dec_a)
+                        const_a = DecimalUtils::convertTo<NativeResultType>(
+                            col_left_const->template getValue<T0>(), scale_a);
+                    else
+                        const_a = col_left_const->template getValue<T0>();
+
+                    if constexpr (IsFloatingPoint<ResultDataType> && dec_b)
+                        const_b = DecimalUtils::convertTo<NativeResultType>(
+                            col_right_const->template getValue<T1>(), scale_b);
+                    else
+                        const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From 1f67b0ce8d68693262a6e714d5c9f2f40b74c9a1 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:28:51 +0300
Subject: [PATCH 128/504] reverting if constexpr

---
 src/Functions/FunctionBinaryArithmetic.h | 29 ++++++++++++------------
 1 file changed, 15 insertions(+), 14 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index cc030d76537..05cef1ce0be 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -987,11 +987,11 @@ public:
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
 
-                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
-
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;
 
+                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
+
                 typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
                 typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
@@ -1000,20 +1000,21 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a;
-                    NativeResultType const_b;
+                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    NativeResultType const_b = col_right_const->template getValue<T1>();
 
-                    if constexpr (IsFloatingPoint<ResultDataType> && dec_a)
-                        const_a = DecimalUtils::convertTo<NativeResultType>(
-                            col_left_const->template getValue<T0>(), scale_a);
-                    else
-                        const_a = col_left_const->template getValue<T0>();
+                    //if constexpr (IsFloatingPoint<ResultDataType> && dec_a)
+                    //    const_a = DecimalUtils::convertTo<NativeResultType>(
+                    //        col_left_const->template getValue<T0>(), scale_a);
+                    //else
+                    //    const_a = col_left_const->template getValue<T0>();
+
+                    //if constexpr (IsFloatingPoint<ResultDataType> && dec_b)
+                    //    const_b = DecimalUtils::convertTo<NativeResultType>(
+                    //        col_right_const->template getValue<T1>(), scale_b);
+                    //else
+                    //    const_b = col_right_const->template getValue<T1>();
 
-                    if constexpr (IsFloatingPoint<ResultDataType> && dec_b)
-                        const_b = DecimalUtils::convertTo<NativeResultType>(
-                            col_right_const->template getValue<T1>(), scale_b);
-                    else
-                        const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From ea2da764f6bd4e1500adaebfc9794879bf414526 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:33:51 +0300
Subject: [PATCH 129/504] fixed result data type init

---
 src/Functions/FunctionBinaryArithmetic.h | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 05cef1ce0be..c43b52ec71f 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -990,7 +990,14 @@ public:
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;
 
-                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
+                ResultDataType type;
+
+                if constexpr(dec_a && IsFloatingPoint<RightDataType>)
+                    type = RightDataType();
+                else if constexpr(dec_b && IsFloatingPoint<LeftDataType>)
+                    type = LeftDataType();
+                else
+                    type = decimalResultType<is_multiply, is_division>(left, right);
 
                 typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
                 typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);

From 3e9428515c28aa6b9f73fa1a9ff02f7115f913a8 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:36:15 +0300
Subject: [PATCH 130/504] fix deleted constuctor

---
 src/Functions/FunctionBinaryArithmetic.h | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index c43b52ec71f..d515872d381 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -990,14 +990,14 @@ public:
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;
 
-                ResultDataType type;
-
-                if constexpr(dec_a && IsFloatingPoint<RightDataType>)
-                    type = RightDataType();
-                else if constexpr(dec_b && IsFloatingPoint<LeftDataType>)
-                    type = LeftDataType();
-                else
-                    type = decimalResultType<is_multiply, is_division>(left, right);
+                const ResultDataType type = [&] {
+                    if constexpr(dec_a && IsFloatingPoint<RightDataType>)
+                        return RightDataType();
+                    else if constexpr(dec_b && IsFloatingPoint<LeftDataType>)
+                        return LeftDataType();
+                    else
+                        return decimalResultType<is_multiply, is_division>(left, right);
+                }();
 
                 typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
                 typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);

From 54887dacbf2a1243e1dc70dee259787141592ab2 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:38:36 +0300
Subject: [PATCH 131/504] fix scale init

---
 src/Functions/FunctionBinaryArithmetic.h | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index d515872d381..2f0f5e41f4f 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -999,8 +999,11 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
-                typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
+                typename ResultDataType::FieldType scale_a = dec_a ? type.scaleFactorFor(left, is_multiply) : 0;
+                typename ResultDataType::FieldType scale_b = dec_b
+                    ? type.scaleFactorFor(right, is_multiply || is_division)
+                    : 0;
+
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     scale_a = right.getScaleMultiplier();
 

From 39c82fe50cb3652ae1f426650634bb89468023f5 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 21:52:19 +0300
Subject: [PATCH 132/504] fixed result type

---
 src/Functions/FunctionBinaryArithmetic.h | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 2f0f5e41f4f..e3876c93817 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -999,10 +999,8 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                typename ResultDataType::FieldType scale_a = dec_a ? type.scaleFactorFor(left, is_multiply) : 0;
-                typename ResultDataType::FieldType scale_b = dec_b
-                    ? type.scaleFactorFor(right, is_multiply || is_division)
-                    : 0;
+                typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
+                typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     scale_a = right.getScaleMultiplier();
@@ -1025,13 +1023,15 @@ public:
                     //else
                     //    const_b = col_right_const->template getValue<T1>();
 
-
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
                         OpImpl::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b);
 
-                    return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
+                    if constexpr (IsDataTypeDecimal<ResultDataType>)
+                        return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
                             col_left_const->size(), toField(res, type.getScale()));
+                    else
+                         return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
                 col_res = ColVecResult::create(0, type.getScale());

From b428f8fcd88c8bcc29fe3aac86361df672d5d0e4 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:01:04 +0300
Subject: [PATCH 133/504] fixed scales

---
 src/Functions/FunctionBinaryArithmetic.h | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index e3876c93817..9bb2fb3e824 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -986,9 +986,11 @@ public:
                 using NativeResultType = typename NativeType<ResultType>::Type;
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
+                using FieldType = typename ResultDataType::FieldType;
 
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;
+                static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
 
                 const ResultDataType type = [&] {
                     if constexpr(dec_a && IsFloatingPoint<RightDataType>)
@@ -999,11 +1001,16 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
-                typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
+                FieldType scale_a;
+                FieldType scale_b;
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     scale_a = right.getScaleMultiplier();
+                else if constexpr(result_is_decimal)
+                    scale_a = type.scaleFactorFor(left, is_multiply);
+
+                if constexpr(result_is_decimal)
+                    scale_b = type.scaleFactorFor(right, is_multiply || is_division);
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
@@ -1027,7 +1034,7 @@ public:
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
                         OpImpl::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b);
 
-                    if constexpr (IsDataTypeDecimal<ResultDataType>)
+                    if constexpr (result_is_decimal)
                         return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
                             col_left_const->size(), toField(res, type.getScale()));
                     else

From fd1d24c928acdcf8d0009ba975209fc2876ebfd0 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:10:33 +0300
Subject: [PATCH 134/504] fixed overflow check

---
 src/Functions/FunctionBinaryArithmetic.h | 6 +++++-
 src/Functions/multiply.cpp               | 8 +++++++-
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 9bb2fb3e824..b58ea9b9e05 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1041,7 +1041,11 @@ public:
                          return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
-                col_res = ColVecResult::create(0, type.getScale());
+                if constexpr(IsDecimalNumber<ResultType>)
+                    col_res = ColVecResult::create(0, type.getScale());
+                else
+                    col_res = ColVecResult::create(0);
+
                 auto & vec_res = col_res->getData();
                 vec_res.resize(col_left_raw->size());
 
diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index 18c1927b4a5..d3171a48825 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -1,3 +1,4 @@
+#include <type_traits>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionBinaryArithmetic.h>
 #include <common/arithmeticOverflow.h>
@@ -29,7 +30,12 @@ struct MultiplyImpl
     template <typename Result = ResultType>
     static inline bool apply(A a, B b, Result & c)
     {
-        return common::mulOverflow(static_cast<Result>(a), b, c);
+        if constexpr(std::is_same_v<Result, float> || std::is_same_v<Result, double>)
+        {
+            c = static_cast<Result>(a) * b;
+            return false;
+        } else
+            return common::mulOverflow(static_cast<Result>(a), b, c);
     }
 
 #if USE_EMBEDDED_COMPILER

From c61af32360f95b5cffa8827b77e603fcb7c01740 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:12:12 +0300
Subject: [PATCH 135/504] fixed const column branch

---
 src/Functions/FunctionBinaryArithmetic.h | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index b58ea9b9e05..b04e07833ba 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1015,20 +1015,20 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    NativeResultType const_a;
+                    NativeResultType const_b;
 
-                    //if constexpr (IsFloatingPoint<ResultDataType> && dec_a)
-                    //    const_a = DecimalUtils::convertTo<NativeResultType>(
-                    //        col_left_const->template getValue<T0>(), scale_a);
-                    //else
-                    //    const_a = col_left_const->template getValue<T0>();
+                    if constexpr (IsFloatingPoint<ResultDataType> && dec_a)
+                        const_a = DecimalUtils::convertTo<NativeResultType>(
+                            col_left_const->template getValue<T0>(), scale_a);
+                    else
+                        const_a = col_left_const->template getValue<T0>();
 
-                    //if constexpr (IsFloatingPoint<ResultDataType> && dec_b)
-                    //    const_b = DecimalUtils::convertTo<NativeResultType>(
-                    //        col_right_const->template getValue<T1>(), scale_b);
-                    //else
-                    //    const_b = col_right_const->template getValue<T1>();
+                    if constexpr (IsFloatingPoint<ResultDataType> && dec_b)
+                        const_b = DecimalUtils::convertTo<NativeResultType>(
+                            col_right_const->template getValue<T1>(), scale_b);
+                    else
+                        const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From 654b8528b25718f933b026225a0d5ab88b0ff2d0 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 17 Dec 2020 22:14:01 +0300
Subject: [PATCH 136/504] Try fix build

---
 src/Functions/FunctionsStringHash.h    |  8 ++++----
 src/Functions/bitHammingDistance.cpp   | 17 ++++++++++-------
 src/Functions/tupleHammingDistance.cpp | 17 +++++++++++------
 3 files changed, 25 insertions(+), 17 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index 19fea2d4fc6..93795d1eaab 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -45,9 +45,9 @@ public:
 
     bool useDefaultImplementationForConstants() const override { return true; }
 
-    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) const override
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t) const override
     {
-        const ColumnPtr & column = block.getByPosition(arguments[0]).column;
+        const ColumnPtr & column = arguments[0].column;
         using ResultType = typename Impl::ResultType;
         if constexpr (is_simhash)
         {
@@ -57,7 +57,7 @@ public:
             vec_res.resize(column->size());
             const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
             Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_res);
-            block.getByPosition(result).column = std::move(col_res);
+            return std::move(col_res);
         }
         else // Min hash
         {
@@ -73,7 +73,7 @@ public:
             MutableColumns tuple_columns;
             tuple_columns.emplace_back(std::move(col_h1));
             tuple_columns.emplace_back(std::move(col_h2));
-            block.getByPosition(result).column = ColumnTuple::create(std::move(tuple_columns));
+            return ColumnTuple::create(std::move(tuple_columns));
         }
     }
 };
diff --git a/src/Functions/bitHammingDistance.cpp b/src/Functions/bitHammingDistance.cpp
index cb34634b00d..6e07ca0ad7c 100644
--- a/src/Functions/bitHammingDistance.cpp
+++ b/src/Functions/bitHammingDistance.cpp
@@ -93,10 +93,11 @@ public:
 
     bool useDefaultImplementationForConstants() const override { return true; }
 
-    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) const override
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t input_rows_count) const override
     {
-        const auto * left_generic = block.getByPosition(arguments[0]).type.get();
-        const auto * right_generic = block.getByPosition(arguments[1]).type.get();
+        const auto * left_generic = arguments[0].type.get();
+        const auto * right_generic = arguments[1].type.get();
+        ColumnPtr result_column;
         bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
@@ -108,14 +109,14 @@ public:
 
             using OpImpl = BitHammingDistanceImpl<T0, T1>;
 
-            const auto * const col_left_raw = block.getByPosition(arguments[0]).column.get();
-            const auto * const col_right_raw = block.getByPosition(arguments[1]).column.get();
+            const auto * const col_left_raw = arguments[0].column.get();
+            const auto * const col_right_raw = arguments[1].column.get();
 
             typename ColVecResult::MutablePtr col_res = nullptr;
             col_res = ColVecResult::create();
 
             auto & vec_res = col_res->getData();
-            vec_res.resize(block.rows());
+            vec_res.resize(input_rows_count);
 
             if (auto col_left_const = checkAndGetColumnConst<ColVecT0>(col_left_raw))
             {
@@ -141,11 +142,13 @@ public:
             else
                 return false;
 
-            block.getByPosition(result).column = std::move(col_res);
+            result_column = std::move(col_res);
             return true;
         });
         if (!valid)
             throw Exception(getName() + "'s arguments do not match the expected data types", ErrorCodes::ILLEGAL_COLUMN);
+
+        return result_column;
     }
 };
 
diff --git a/src/Functions/tupleHammingDistance.cpp b/src/Functions/tupleHammingDistance.cpp
index aa38426d228..67d5f73065b 100644
--- a/src/Functions/tupleHammingDistance.cpp
+++ b/src/Functions/tupleHammingDistance.cpp
@@ -100,10 +100,10 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
-    void executeImpl(Block & block, const ColumnNumbers & arguments, size_t result, size_t) const override
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t input_rows_count) const override
     {
-        const ColumnWithTypeAndName & arg1 = block.getByPosition(arguments[0]);
-        const ColumnWithTypeAndName & arg2 = block.getByPosition(arguments[1]);
+        const ColumnWithTypeAndName & arg1 = arguments[0];
+        const ColumnWithTypeAndName & arg2 = arguments[1];
         const DataTypeTuple & type1 = static_cast<const DataTypeTuple &>(*arg1.type);
         const DataTypeTuple & type2 = static_cast<const DataTypeTuple &>(*arg2.type);
         const auto & left_elems = type1.getElements();
@@ -112,6 +112,9 @@ public:
             throw Exception(
                 "Illegal column of arguments of function " + getName() + ", tuple should have exactly two elements.",
                 ErrorCodes::ILLEGAL_COLUMN);
+
+        ColumnPtr result_column;
+
         bool valid = castBothTypes(left_elems[0].get(), right_elems[0].get(), [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
@@ -139,7 +142,7 @@ public:
                     cols2[0]->get(0, a2);
                     cols2[1]->get(0, b2);
                     auto res = OpImpl::constantConstant(a1.get<UInt64>(), b1.get<UInt64>(), a2.get<UInt64>(), b2.get<UInt64>());
-                    block.getByPosition(result).column = DataTypeUInt8().createColumnConst(const_col_left->size(), toField(res));
+                    result_column = DataTypeUInt8().createColumnConst(const_col_left->size(), toField(res));
                     return true;
                 }
             }
@@ -147,7 +150,7 @@ public:
             typename ColVecResult::MutablePtr col_res = nullptr;
             col_res = ColVecResult::create();
             auto & vec_res = col_res->getData();
-            vec_res.resize(block.rows());
+            vec_res.resize(input_rows_count);
             // constant tuple - non-constant tuple
             if (const ColumnConst * const_col_left = checkAndGetColumnConst<ColumnTuple>(arg1.column.get()))
             {
@@ -200,11 +203,13 @@ public:
             }
             else
                 return false;
-            block.getByPosition(result).column = std::move(col_res);
+            result_column = std::move(col_res);
             return true;
         });
         if (!valid)
             throw Exception(getName() + "'s arguments do not match the expected data types", ErrorCodes::ILLEGAL_COLUMN);
+
+        return result_column;
     }
 };
 

From 1c52842e835a661a9fb36693a45ff70a1be96633 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:21:23 +0300
Subject: [PATCH 137/504] simplified the condition

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index b04e07833ba..ad4be84d337 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1018,13 +1018,13 @@ public:
                     NativeResultType const_a;
                     NativeResultType const_b;
 
-                    if constexpr (IsFloatingPoint<ResultDataType> && dec_a)
+                    if constexpr (!result_is_decimal && dec_a)
                         const_a = DecimalUtils::convertTo<NativeResultType>(
                             col_left_const->template getValue<T0>(), scale_a);
                     else
                         const_a = col_left_const->template getValue<T0>();
 
-                    if constexpr (IsFloatingPoint<ResultDataType> && dec_b)
+                    if constexpr (!result_is_decimal && dec_b)
                         const_b = DecimalUtils::convertTo<NativeResultType>(
                             col_right_const->template getValue<T1>(), scale_b);
                     else

From 613f4e0f04405a9ce5901f36f82935e32c1ca03b Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:24:09 +0300
Subject: [PATCH 138/504] removed explicit types

---
 src/Functions/FunctionBinaryArithmetic.h | 16 ++--------------
 1 file changed, 2 insertions(+), 14 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index ad4be84d337..c99f12a99b0 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1015,20 +1015,8 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a;
-                    NativeResultType const_b;
-
-                    if constexpr (!result_is_decimal && dec_a)
-                        const_a = DecimalUtils::convertTo<NativeResultType>(
-                            col_left_const->template getValue<T0>(), scale_a);
-                    else
-                        const_a = col_left_const->template getValue<T0>();
-
-                    if constexpr (!result_is_decimal && dec_b)
-                        const_b = DecimalUtils::convertTo<NativeResultType>(
-                            col_right_const->template getValue<T1>(), scale_b);
-                    else
-                        const_b = col_right_const->template getValue<T1>();
+                    auto const_a = col_left_const->template getValue<T0>();
+                    auto const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From 31e3ca16703e06b5ac5cd3bc39631d0dfdaa12cc Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:26:15 +0300
Subject: [PATCH 139/504] restored type signature

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index c99f12a99b0..0f6fbad3065 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1015,8 +1015,8 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    auto const_a = col_left_const->template getValue<T0>();
-                    auto const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_a = col_left_const->template getValue<T0>();
+                    const NativeResultType const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From 6eecef35a6b09f1d83f5e1c75db8200d4a438fe5 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:35:46 +0300
Subject: [PATCH 140/504] fixed result type trait

---
 src/Core/DecimalFunctions.h              | 6 +++++-
 src/Functions/FunctionBinaryArithmetic.h | 9 ++++++---
 2 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h
index b82cfd88e98..a0a51651353 100644
--- a/src/Core/DecimalFunctions.h
+++ b/src/Core/DecimalFunctions.h
@@ -6,6 +6,7 @@
 #include <common/arithmeticOverflow.h>
 
 #include <limits>
+#include <type_traits>
 
 
 namespace DB
@@ -205,10 +206,13 @@ inline typename DecimalType::NativeType getFractionalPart(const DecimalType & de
     return getFractionalPartWithScaleMultiplier(decimal, scaleMultiplier<typename DecimalType::NativeType>(scale));
 }
 
-/// Decimal to integer/float conversion
+/// Decimal to integer/float conversion with the ability to convert into itself (returns the original value);
 template <typename To, typename DecimalType>
 To convertTo(const DecimalType & decimal, size_t scale)
 {
+    if constexpr (std::is_same_v<To, DecimalType>)
+        return decimal;
+
     using NativeT = typename DecimalType::NativeType;
 
     if constexpr (std::is_floating_point_v<To>)
diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 0f6fbad3065..c871cfd914d 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -22,6 +22,7 @@
 #include <Columns/ColumnFixedString.h>
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnAggregateFunction.h>
+#include "Core/DecimalFunctions.h"
 #include "IFunctionImpl.h"
 #include "FunctionHelpers.h"
 #include "IsOperation.h"
@@ -209,7 +210,9 @@ struct DecimalBinaryOperation
         DivideIntegralImpl<NativeResultType, NativeResultType>, /// substitute divide by intDiv (throw on division by zero)
         Operation<NativeResultType, NativeResultType>>;
 
-    using ArrayC = typename ColumnDecimal<ResultType>::Container;
+    using ArrayC = typename std::conditional_t<IsDecimalNumber<ResultType>,
+        ColumnDecimal<ResultType>,
+        ColumnVector<ResultType>>::Container;
 
     template <bool is_decimal_a, bool is_decimal_b, typename ArrayA, typename ArrayB>
     static void NO_INLINE vectorVector(const ArrayA & a, const ArrayB & b, ArrayC & c,
@@ -1015,8 +1018,8 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = col_left_const->template getValue<T0>();
-                    const NativeResultType const_b = col_right_const->template getValue<T1>();
+                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    NativeResultType const_b = col_right_const->template getValue<T1>();
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :

From 5974e784c29b54c59b1d84a1cecd1049cbc5e5a5 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 22:52:52 +0300
Subject: [PATCH 141/504] added the conversion func

---
 src/Functions/FunctionBinaryArithmetic.h | 16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index c871cfd914d..5223dc6cb4a 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1015,11 +1015,18 @@ public:
                 if constexpr(result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);
 
+                const auto get_value_or_convert = []<class T>(const auto& col, const auto & scale) {
+                    if constexpr(IsFloatingPoint<ResultType> && IsDecimalNumber<T>)
+                        return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
+                    else
+                        return col->template getValue<T>();
+                };
+
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_a = get_value_or_convert<T0>(col_left_const, scale_a);
+                    const NativeResultType const_b = get_value_or_convert<T1>(col_right_const, scale_b);
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
@@ -1040,6 +1047,7 @@ public:
                 auto & vec_res = col_res->getData();
                 vec_res.resize(col_left_raw->size());
 
+
                 if (col_left && col_right)
                 {
                     if (check_decimal_overflow)
@@ -1049,7 +1057,7 @@ public:
                 }
                 else if (col_left_const && col_right)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    const NativeResultType const_a = get_value_or_convert<T0>(col_left_const, scale_a);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
@@ -1058,7 +1066,7 @@ public:
                 }
                 else if (col_left && col_right_const)
                 {
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_b = get_value_or_convert<T1>(col_right_const, scale_b);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);

From f93385dcf01697a3cd5e77d8ef7f5b69aa8c90ac Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:01:29 +0300
Subject: [PATCH 142/504] replaced lambda with helper function

---
 src/Functions/FunctionBinaryArithmetic.h | 23 ++++++++++++-----------
 1 file changed, 12 insertions(+), 11 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 5223dc6cb4a..75ce3bcd7bd 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -957,6 +957,14 @@ public:
         return nullptr;
     }
 
+    template <class T, class ResultType, class NativeResultType>
+    static auto helperGetOrConvert(const auto& col, const auto& scale) {
+        if constexpr(IsFloatingPoint<ResultType> && IsDecimalNumber<T>)
+            return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
+        else
+            return col->template getValue<T>();
+    }
+
     template <typename A, typename B>
     ColumnPtr executeNumeric(const ColumnsWithTypeAndName & arguments, const A & left, const B & right) const
     {
@@ -1015,18 +1023,11 @@ public:
                 if constexpr(result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);
 
-                const auto get_value_or_convert = []<class T>(const auto& col, const auto & scale) {
-                    if constexpr(IsFloatingPoint<ResultType> && IsDecimalNumber<T>)
-                        return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
-                    else
-                        return col->template getValue<T>();
-                };
-
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = get_value_or_convert<T0>(col_left_const, scale_a);
-                    const NativeResultType const_b = get_value_or_convert<T1>(col_right_const, scale_b);
+                    const NativeResultType const_a = helperGetOrConvert<T0>(col_left_const, scale_a);
+                    const NativeResultType const_b = helperGetOrConvert<T1>(col_right_const, scale_b);
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
@@ -1057,7 +1058,7 @@ public:
                 }
                 else if (col_left_const && col_right)
                 {
-                    const NativeResultType const_a = get_value_or_convert<T0>(col_left_const, scale_a);
+                    const NativeResultType const_a = helperGetOrConvert<T0>(col_left_const, scale_a);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
@@ -1066,7 +1067,7 @@ public:
                 }
                 else if (col_left && col_right_const)
                 {
-                    const NativeResultType const_b = get_value_or_convert<T1>(col_right_const, scale_b);
+                    const NativeResultType const_b = helperGetOrConvert<T1>(col_right_const, scale_b);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);

From e50af17545827eb9ef657ff7d17c319afb17428e Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:05:10 +0300
Subject: [PATCH 143/504] fixed template args1

---
 src/Functions/FunctionBinaryArithmetic.h | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 75ce3bcd7bd..8dadfe5b398 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1026,8 +1026,10 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0>(col_left_const, scale_a);
-                    const NativeResultType const_b = helperGetOrConvert<T1>(col_right_const, scale_b);
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultType, NativeResultType>(
+                        col_left_const, scale_a);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultType, NativeResultType>(
+                        col_right_const, scale_b);
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
@@ -1058,7 +1060,8 @@ public:
                 }
                 else if (col_left_const && col_right)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0>(col_left_const, scale_a);
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultType, NativeResultType>(
+                        col_left_const, scale_a);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
@@ -1067,7 +1070,8 @@ public:
                 }
                 else if (col_left && col_right_const)
                 {
-                    const NativeResultType const_b = helperGetOrConvert<T1>(col_right_const, scale_b);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultType, NativeResultType>(
+                        col_right_const, scale_b);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);

From 488f4ef42e90988667f77ec5b6d7e25fef9a3a03 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:09:33 +0300
Subject: [PATCH 144/504] fixed compile condition

---
 src/Functions/FunctionBinaryArithmetic.h | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 8dadfe5b398..72b5dbdcecb 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -957,9 +957,12 @@ public:
         return nullptr;
     }
 
-    template <class T, class ResultType, class NativeResultType>
+    template <class T, class ResultDataType>
     static auto helperGetOrConvert(const auto& col, const auto& scale) {
-        if constexpr(IsFloatingPoint<ResultType> && IsDecimalNumber<T>)
+        using ResultType = typename ResultDataType::FieldType;
+        using NativeResultType = typename NativeType<ResultType>::Type;
+
+        if constexpr(IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
             return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
         else
             return col->template getValue<T>();
@@ -1026,9 +1029,9 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultType, NativeResultType>(
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
                         col_left_const, scale_a);
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultType, NativeResultType>(
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
                         col_right_const, scale_b);
 
                     auto res = check_decimal_overflow ?
@@ -1060,7 +1063,7 @@ public:
                 }
                 else if (col_left_const && col_right)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultType, NativeResultType>(
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
                         col_left_const, scale_a);
 
                     if (check_decimal_overflow)
@@ -1070,7 +1073,7 @@ public:
                 }
                 else if (col_left && col_right_const)
                 {
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultType, NativeResultType>(
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
                         col_right_const, scale_b);
 
                     if (check_decimal_overflow)

From 73f3195fe17610f814453b27c8f4bbd15a54820d Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:12:41 +0300
Subject: [PATCH 145/504] fixed variable init

---
 src/Functions/FunctionBinaryArithmetic.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 72b5dbdcecb..35fdd8f0230 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1022,6 +1022,8 @@ public:
                     scale_a = right.getScaleMultiplier();
                 else if constexpr(result_is_decimal)
                     scale_a = type.scaleFactorFor(left, is_multiply);
+                else
+                    scale_a = 0.0; //won't be used
 
                 if constexpr(result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);

From 61d9d213e91f7fbc0381a2ef81b1e941d7fd2506 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:14:02 +0300
Subject: [PATCH 146/504] fixed another variable init

---
 src/Functions/FunctionBinaryArithmetic.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 35fdd8f0230..68b761c4904 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1023,10 +1023,12 @@ public:
                 else if constexpr(result_is_decimal)
                     scale_a = type.scaleFactorFor(left, is_multiply);
                 else
-                    scale_a = 0.0; //won't be used
+                    scale_a = 0.0; //won't be used as the target column is not decimal
 
                 if constexpr(result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);
+                else
+                    scale_b = 0.0; //won't be used, same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)

From 4f7d019a9ebd22b06b3607895a2160fb78c41f4a Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:19:57 +0300
Subject: [PATCH 147/504] Squashed commit of the following:

commit 61d9d213e91f7fbc0381a2ef81b1e941d7fd2506
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:14:02 2020 +0300

    fixed another variable init

commit 73f3195fe17610f814453b27c8f4bbd15a54820d
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:12:41 2020 +0300

    fixed variable init

commit 488f4ef42e90988667f77ec5b6d7e25fef9a3a03
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:09:33 2020 +0300

    fixed compile condition

commit e50af17545827eb9ef657ff7d17c319afb17428e
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:05:10 2020 +0300

    fixed template args1

commit f93385dcf01697a3cd5e77d8ef7f5b69aa8c90ac
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:01:29 2020 +0300

    replaced lambda with helper function

commit 5974e784c29b54c59b1d84a1cecd1049cbc5e5a5
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:52:52 2020 +0300

    added the conversion func

commit 6eecef35a6b09f1d83f5e1c75db8200d4a438fe5
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:35:46 2020 +0300

    fixed result type trait

commit 31e3ca16703e06b5ac5cd3bc39631d0dfdaa12cc
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:26:15 2020 +0300

    restored type signature

commit 613f4e0f04405a9ce5901f36f82935e32c1ca03b
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:24:09 2020 +0300

    removed explicit types

commit 1c52842e835a661a9fb36693a45ff70a1be96633
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:21:23 2020 +0300

    simplified the condition

commit c61af32360f95b5cffa8827b77e603fcb7c01740
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:12:12 2020 +0300

    fixed const column branch

commit fd1d24c928acdcf8d0009ba975209fc2876ebfd0
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:10:33 2020 +0300

    fixed overflow check

commit b428f8fcd88c8bcc29fe3aac86361df672d5d0e4
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:01:04 2020 +0300

    fixed scales

commit 39c82fe50cb3652ae1f426650634bb89468023f5
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:52:19 2020 +0300

    fixed result type

commit 54887dacbf2a1243e1dc70dee259787141592ab2
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:38:36 2020 +0300

    fix scale init

commit 3e9428515c28aa6b9f73fa1a9ff02f7115f913a8
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:36:15 2020 +0300

    fix deleted constuctor

commit ea2da764f6bd4e1500adaebfc9794879bf414526
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:33:51 2020 +0300

    fixed result data type init

commit 1f67b0ce8d68693262a6e714d5c9f2f40b74c9a1
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:28:51 2020 +0300

    reverting if constexpr

commit 494e9066c33c9949d2063f8a63ce1f31120c8805
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:17:38 2020 +0300

    fixing first compile bug

commit d24fc58c94e270f48834451f4807aba10fe330fc
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:12:24 2020 +0300

    case fix 2

commit 9f217c5df57e6fe8d8ca5d823d4889e39830ffc1
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:11:39 2020 +0300

    case fix

commit fa6a1d3b4a70f3217df426934b14de4c6a166e6d
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 20:55:35 2020 +0300

    fix

commit af8a72a45f2162f69c3cc0034af607cc17d288f9
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 20:16:37 2020 +0300

    fix

commit 737d4fc22fd2565e5551b15f6cd96aa419a271cd
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 19:43:29 2020 +0300

    debug fix

commit fcacf5c510c1ea34f3980863c92b34a191523f7d
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 19:31:54 2020 +0300

    another try

commit 49f3b71b5f1c731c98eb0d90421bef968a675e64
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 18:48:22 2020 +0300

    another fix

commit 0d319c8c612285fc249a6aea9382e9aad3b8e7c4
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 18:47:14 2020 +0300

    fix

commit 9d5998fd10812ae4cf6b786dbe11fb947b6ec35b
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 18:46:01 2020 +0300

    added debug info

commit 17d996054ecd77d59246a26b48ff3f2436c17d3a
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 18:47:24 2020 +0300

    style fix

commit 02baaddc13b5cb40f77b44bcc104421030963012
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 18:20:41 2020 +0300

    removed some casts
---
 src/Core/DecimalFunctions.h              |   6 +-
 src/Functions/FunctionBinaryArithmetic.h | 103 ++++++++++++++++-------
 src/Functions/multiply.cpp               |   8 +-
 3 files changed, 85 insertions(+), 32 deletions(-)

diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h
index b82cfd88e98..a0a51651353 100644
--- a/src/Core/DecimalFunctions.h
+++ b/src/Core/DecimalFunctions.h
@@ -6,6 +6,7 @@
 #include <common/arithmeticOverflow.h>
 
 #include <limits>
+#include <type_traits>
 
 
 namespace DB
@@ -205,10 +206,13 @@ inline typename DecimalType::NativeType getFractionalPart(const DecimalType & de
     return getFractionalPartWithScaleMultiplier(decimal, scaleMultiplier<typename DecimalType::NativeType>(scale));
 }
 
-/// Decimal to integer/float conversion
+/// Decimal to integer/float conversion with the ability to convert into itself (returns the original value);
 template <typename To, typename DecimalType>
 To convertTo(const DecimalType & decimal, size_t scale)
 {
+    if constexpr (std::is_same_v<To, DecimalType>)
+        return decimal;
+
     using NativeT = typename DecimalType::NativeType;
 
     if constexpr (std::is_floating_point_v<To>)
diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 73fd40db454..68b761c4904 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -22,6 +22,7 @@
 #include <Columns/ColumnFixedString.h>
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnAggregateFunction.h>
+#include "Core/DecimalFunctions.h"
 #include "IFunctionImpl.h"
 #include "FunctionHelpers.h"
 #include "IsOperation.h"
@@ -45,6 +46,8 @@
 #    pragma GCC diagnostic pop
 #endif
 
+#include <cassert>
+
 
 namespace DB
 {
@@ -207,7 +210,9 @@ struct DecimalBinaryOperation
         DivideIntegralImpl<NativeResultType, NativeResultType>, /// substitute divide by intDiv (throw on division by zero)
         Operation<NativeResultType, NativeResultType>>;
 
-    using ArrayC = typename ColumnDecimal<ResultType>::Container;
+    using ArrayC = typename std::conditional_t<IsDecimalNumber<ResultType>,
+        ColumnDecimal<ResultType>,
+        ColumnVector<ResultType>>::Container;
 
     template <bool is_decimal_a, bool is_decimal_b, typename ArrayA, typename ArrayB>
     static void NO_INLINE vectorVector(const ArrayA & a, const ArrayB & b, ArrayC & c,
@@ -470,6 +475,11 @@ public:
         Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
 
+        /// e.g Decimal * Float64 = Float64
+        Case<IsOperation<Operation>::multiply, Switch<
+            Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+            Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>>>,
+
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
         Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
@@ -478,11 +488,6 @@ public:
         Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
             DataTypeFromFieldType<typename Op::ResultType>>,
 
-        /// e.g Decimal * Float64 = Float64
-        Case<IsOperation<Operation>::multiply, Switch<
-            Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
-            Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>>>,
-
         /// Date + Integral -> Date
         /// Integral + Date -> Date
         Case<IsOperation<Operation>::plus, Switch<
@@ -548,7 +553,13 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+        return castType(left, [&](const auto & left_)
+        {
+            return castType(right, [&](const auto & right_)
+            {
+                return f(left_, right_);
+            });
+        });
     }
 
     static FunctionOverloadResolverPtr
@@ -946,6 +957,17 @@ public:
         return nullptr;
     }
 
+    template <class T, class ResultDataType>
+    static auto helperGetOrConvert(const auto& col, const auto& scale) {
+        using ResultType = typename ResultDataType::FieldType;
+        using NativeResultType = typename NativeType<ResultType>::Type;
+
+        if constexpr(IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
+            return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
+        else
+            return col->template getValue<T>();
+    }
+
     template <typename A, typename B>
     ColumnPtr executeNumeric(const ColumnsWithTypeAndName & arguments, const A & left, const B & right) const
     {
@@ -978,47 +1000,64 @@ public:
                 using NativeResultType = typename NativeType<ResultType>::Type;
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
-
-                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
+                using FieldType = typename ResultDataType::FieldType;
 
                 static constexpr const bool dec_a = IsDecimalNumber<T0>;
                 static constexpr const bool dec_b = IsDecimalNumber<T1>;
+                static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
+
+                const ResultDataType type = [&] {
+                    if constexpr(dec_a && IsFloatingPoint<RightDataType>)
+                        return RightDataType();
+                    else if constexpr(dec_b && IsFloatingPoint<LeftDataType>)
+                        return LeftDataType();
+                    else
+                        return decimalResultType<is_multiply, is_division>(left, right);
+                }();
+
+                FieldType scale_a;
+                FieldType scale_b;
 
-                typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
-                typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     scale_a = right.getScaleMultiplier();
+                else if constexpr(result_is_decimal)
+                    scale_a = type.scaleFactorFor(left, is_multiply);
+                else
+                    scale_a = 0.0; //won't be used as the target column is not decimal
+
+                if constexpr(result_is_decimal)
+                    scale_b = type.scaleFactorFor(right, is_multiply || is_division);
+                else
+                    scale_b = 0.0; //won't be used, same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a;
-                    NativeResultType const_b;
-
-                    if constexpr (IsFloatingPoint<NativeResultType> && dec_a)
-                        const_a = DecimalUtils::convertTo<NativeResultType>(
-                            col_left_const->template getValue<T0>(), scale_a);
-                    else
-                        const_a = col_left_const->template getValue<T0>();
-
-                    if constexpr (IsFloatingPoint<NativeResultType> && dec_b)
-                        const_b = DecimalUtils::convertTo<NativeResultType>(
-                            col_right_const->template getValue<T1>(), scale_b);
-                    else
-                        const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
+                        col_left_const, scale_a);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
+                        col_right_const, scale_b);
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
                         OpImpl::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b);
 
-                    return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
+                    if constexpr (result_is_decimal)
+                        return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
                             col_left_const->size(), toField(res, type.getScale()));
+                    else
+                         return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
-                col_res = ColVecResult::create(0, type.getScale());
+                if constexpr(IsDecimalNumber<ResultType>)
+                    col_res = ColVecResult::create(0, type.getScale());
+                else
+                    col_res = ColVecResult::create(0);
+
                 auto & vec_res = col_res->getData();
                 vec_res.resize(col_left_raw->size());
 
+
                 if (col_left && col_right)
                 {
                     if (check_decimal_overflow)
@@ -1028,7 +1067,8 @@ public:
                 }
                 else if (col_left_const && col_right)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
+                        col_left_const, scale_a);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
@@ -1037,7 +1077,8 @@ public:
                 }
                 else if (col_left && col_right_const)
                 {
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
+                        col_right_const, scale_b);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
@@ -1109,10 +1150,12 @@ public:
         const auto * left_generic = left_argument.type.get();
         const auto * right_generic = right_argument.type.get();
         ColumnPtr res;
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index 18c1927b4a5..d3171a48825 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -1,3 +1,4 @@
+#include <type_traits>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionBinaryArithmetic.h>
 #include <common/arithmeticOverflow.h>
@@ -29,7 +30,12 @@ struct MultiplyImpl
     template <typename Result = ResultType>
     static inline bool apply(A a, B b, Result & c)
     {
-        return common::mulOverflow(static_cast<Result>(a), b, c);
+        if constexpr(std::is_same_v<Result, float> || std::is_same_v<Result, double>)
+        {
+            c = static_cast<Result>(a) * b;
+            return false;
+        } else
+            return common::mulOverflow(static_cast<Result>(a), b, c);
     }
 
 #if USE_EMBEDDED_COMPILER

From 205db4fa92a3f9086fc154997a9dc283719aefbb Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:50:59 +0300
Subject: [PATCH 148/504] fixed the scale bug, added the test (wo reference)

---
 src/Functions/FunctionBinaryArithmetic.h        |  8 ++++----
 .../0_stateless/01603_decimal_mult_float.sql    | 17 +++++++++++++++++
 2 files changed, 21 insertions(+), 4 deletions(-)
 create mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.sql

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 68b761c4904..0d62bc2bb96 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1022,13 +1022,13 @@ public:
                     scale_a = right.getScaleMultiplier();
                 else if constexpr(result_is_decimal)
                     scale_a = type.scaleFactorFor(left, is_multiply);
-                else
-                    scale_a = 0.0; //won't be used as the target column is not decimal
+                else // The result is a floating point, so we need to get the scale explicitly
+                    scale_a = left.getScale();
 
                 if constexpr(result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);
-                else
-                    scale_b = 0.0; //won't be used, same
+                else // Same
+                    scale_b = right.getScale();
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
new file mode 100644
index 00000000000..166be0faf5a
--- /dev/null
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -0,0 +1,17 @@
+SELECT toDecimal32(2, 2) * 1.2;
+SELECT toDecimal64(0.5, 2) * 20.33;
+SELECT 0.00001 * toDecimal32(12, 2);
+SELECT 30.033 * toDecimal32(5, 1);
+
+CREATE TABLE test01603 IF NOT EXISTS(
+    f64 Float64,
+    d Decimal64(3) DEFAULT toDecimal32(f64, 3),
+    f32 Float32 DEFAULT f64
+);
+
+INSERT INTO test01603(f64) SELECT 1 / num FROM system.numbers LIMIT 1000;
+
+SELECT sum(d * 1.1) FROM test01603;
+SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
+
+DROP TABLE test01603 IF EXISTS;

From 7db7a9ef5b2727196f037af3e5d4f0834fbdc3cf Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 17 Dec 2020 23:55:59 +0300
Subject: [PATCH 149/504] added another constexpr branch

---
 src/Functions/FunctionBinaryArithmetic.h | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 0d62bc2bb96..24f3072e0b3 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1022,13 +1022,17 @@ public:
                     scale_a = right.getScaleMultiplier();
                 else if constexpr(result_is_decimal)
                     scale_a = type.scaleFactorFor(left, is_multiply);
-                else // The result is a floating point, so we need to get the scale explicitly
+                else if constexpr(dec_a)
                     scale_a = left.getScale();
+                else
+                    scale_a = 0.0; //won't be used, just to silence the warning
 
                 if constexpr(result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);
-                else // Same
+                else if constexpr(dec_b)
                     scale_b = right.getScale();
+                else
+                    scale_b = 0.0; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)

From bf5e836414f68f49d363708c8d32b0a9f0d576cc Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 00:03:45 +0300
Subject: [PATCH 150/504] updated the test

---
 tests/queries/0_stateless/01603_decimal_mult_float.sql | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
index 166be0faf5a..47a09418ebf 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.sql
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -12,6 +12,8 @@ CREATE TABLE test01603 IF NOT EXISTS(
 INSERT INTO test01603(f64) SELECT 1 / num FROM system.numbers LIMIT 1000;
 
 SELECT sum(d * 1.1) FROM test01603;
+SELECT sum(8.01 * d) FROM test01603;
 SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
+SELECT sum(toDecimal64(0.4, 2) * f32) FROM test01603;
 
 DROP TABLE test01603 IF EXISTS;

From fe1b39625664c7397d7c0f18f650a1a605be2f35 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 00:14:06 +0300
Subject: [PATCH 151/504] added test reference

---
 .../0_stateless/01603_decimal_mult_float.reference        | 8 ++++++++
 tests/queries/0_stateless/01603_decimal_mult_float.sql    | 8 ++++----
 2 files changed, 12 insertions(+), 4 deletions(-)
 create mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.reference

diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
new file mode 100644
index 00000000000..22650604ff3
--- /dev/null
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -0,0 +1,8 @@
+2.4
+10.165
+0.00012000000000000002
+150.16500000000002
+15551.800000000001
+113245.37999999999
+1197.6753376880554
+5.988376864581369
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
index 47a09418ebf..0c722c1899b 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.sql
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -3,17 +3,17 @@ SELECT toDecimal64(0.5, 2) * 20.33;
 SELECT 0.00001 * toDecimal32(12, 2);
 SELECT 30.033 * toDecimal32(5, 1);
 
-CREATE TABLE test01603 IF NOT EXISTS(
+CREATE TABLE IF NOT EXISTS test01603 (
     f64 Float64,
     d Decimal64(3) DEFAULT toDecimal32(f64, 3),
     f32 Float32 DEFAULT f64
-);
+) ENGINE=Memory() ORDER BY f32;
 
-INSERT INTO test01603(f64) SELECT 1 / num FROM system.numbers LIMIT 1000;
+INSERT INTO test01603(f64) SELECT 1 / (number + 1) FROM system.numbers LIMIT 1000;
 
 SELECT sum(d * 1.1) FROM test01603;
 SELECT sum(8.01 * d) FROM test01603;
 SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
 SELECT sum(toDecimal64(0.4, 2) * f32) FROM test01603;
 
-DROP TABLE test01603 IF EXISTS;
+DROP TABLE IF EXISTS test01603;

From 40c702c54a4d676fef6ce22b583a4523c287af2f Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Fri, 18 Dec 2020 00:56:42 +0300
Subject: [PATCH 152/504] Update multiply.cpp

---
 src/Functions/multiply.cpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index d3171a48825..8d86e1fbdd7 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -34,7 +34,8 @@ struct MultiplyImpl
         {
             c = static_cast<Result>(a) * b;
             return false;
-        } else
+        }
+        else
             return common::mulOverflow(static_cast<Result>(a), b, c);
     }
 

From 5a78eafb0547b53df329c9a2b4fb7d3d1d8d7b6b Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Fri, 18 Dec 2020 00:59:39 +0300
Subject: [PATCH 153/504] Update FunctionBinaryArithmetic.h

---
 src/Functions/FunctionBinaryArithmetic.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 24f3072e0b3..bb78cf5fa7e 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -958,7 +958,7 @@ public:
     }
 
     template <class T, class ResultDataType>
-    static auto helperGetOrConvert(const auto& col, const auto& scale) {
+    static auto helperGetOrConvert(const auto & col, const auto & scale) {
         using ResultType = typename ResultDataType::FieldType;
         using NativeResultType = typename NativeType<ResultType>::Type;
 

From 0462340d8a6c02a455b254cd12a6beb4c759c3d0 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Fri, 18 Dec 2020 01:03:25 +0300
Subject: [PATCH 154/504] Update FunctionBinaryArithmetic.h

---
 src/Functions/FunctionBinaryArithmetic.h | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index bb78cf5fa7e..93f7fa112d4 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -962,7 +962,7 @@ public:
         using ResultType = typename ResultDataType::FieldType;
         using NativeResultType = typename NativeType<ResultType>::Type;
 
-        if constexpr(IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
+        if constexpr (IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
             return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
         else
             return col->template getValue<T>();
@@ -1007,9 +1007,9 @@ public:
                 static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
 
                 const ResultDataType type = [&] {
-                    if constexpr(dec_a && IsFloatingPoint<RightDataType>)
+                    if constexpr (dec_a && IsFloatingPoint<RightDataType>)
                         return RightDataType();
-                    else if constexpr(dec_b && IsFloatingPoint<LeftDataType>)
+                    else if constexpr (dec_b && IsFloatingPoint<LeftDataType>)
                         return LeftDataType();
                     else
                         return decimalResultType<is_multiply, is_division>(left, right);
@@ -1020,16 +1020,16 @@ public:
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     scale_a = right.getScaleMultiplier();
-                else if constexpr(result_is_decimal)
+                else if constexpr (result_is_decimal)
                     scale_a = type.scaleFactorFor(left, is_multiply);
-                else if constexpr(dec_a)
+                else if constexpr (dec_a)
                     scale_a = left.getScale();
                 else
                     scale_a = 0.0; //won't be used, just to silence the warning
 
-                if constexpr(result_is_decimal)
+                if constexpr (result_is_decimal)
                     scale_b = type.scaleFactorFor(right, is_multiply || is_division);
-                else if constexpr(dec_b)
+                else if constexpr (dec_b)
                     scale_b = right.getScale();
                 else
                     scale_b = 0.0; //same
@@ -1053,7 +1053,7 @@ public:
                          return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
-                if constexpr(IsDecimalNumber<ResultType>)
+                if constexpr (IsDecimalNumber<ResultType>)
                     col_res = ColVecResult::create(0, type.getScale());
                 else
                     col_res = ColVecResult::create(0);

From 43348bd7087b7f46b36ec60d4890287d4de78144 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 01:25:30 +0300
Subject: [PATCH 155/504] fixed the possible values case

---
 src/Functions/FunctionBinaryArithmetic.h | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 24f3072e0b3..d768f7ec72b 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -476,9 +476,8 @@ public:
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
 
         /// e.g Decimal * Float64 = Float64
-        Case<IsOperation<Operation>::multiply, Switch<
-            Case<IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
-            Case<IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>>>,
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>,
 
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,

From 6a9397a64bc7c3d71520d515d917a13ceab476f3 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Fri, 18 Dec 2020 01:35:37 +0300
Subject: [PATCH 156/504] Update FunctionBinaryArithmetic.h

---
 src/Functions/FunctionBinaryArithmetic.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 3543fb80520..bf245497025 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -957,7 +957,8 @@ public:
     }
 
     template <class T, class ResultDataType>
-    static auto helperGetOrConvert(const auto & col, const auto & scale) {
+    static auto helperGetOrConvert(const auto & col, const auto & scale)
+    {
         using ResultType = typename ResultDataType::FieldType;
         using NativeResultType = typename NativeType<ResultType>::Type;
 

From fc426807a8f1e92cb5d8117e2d3f1801adb3b655 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 18 Dec 2020 03:21:23 +0300
Subject: [PATCH 157/504] formatting & fuzzer support

---
 programs/client/QueryFuzzer.cpp     | 98 +++++++++++++++++++++++++++--
 programs/client/QueryFuzzer.h       |  7 ++-
 src/Interpreters/ActionsVisitor.cpp | 10 ++-
 src/Parsers/ASTFunction.cpp         | 98 ++++++++++++++++++-----------
 4 files changed, 167 insertions(+), 46 deletions(-)

diff --git a/programs/client/QueryFuzzer.cpp b/programs/client/QueryFuzzer.cpp
index a8e32d87db5..99b6a977826 100644
--- a/programs/client/QueryFuzzer.cpp
+++ b/programs/client/QueryFuzzer.cpp
@@ -14,6 +14,7 @@
 #include <Parsers/ASTIdentifier.h>
 #include <Parsers/ASTInsertQuery.h>
 #include <Parsers/ASTLiteral.h>
+#include <Parsers/ASTOrderByElement.h>
 #include <Parsers/ASTQueryWithOutput.h>
 #include <Parsers/ASTSelectQuery.h>
 #include <Parsers/ASTSelectWithUnionQuery.h>
@@ -205,14 +206,88 @@ void QueryFuzzer::replaceWithTableLike(ASTPtr & ast)
     ast = new_ast;
 }
 
-void QueryFuzzer::fuzzColumnLikeExpressionList(ASTPtr ast)
+void QueryFuzzer::fuzzOrderByElement(ASTOrderByElement * elem)
+{
+    switch (fuzz_rand() % 10)
+    {
+        case 0:
+            elem->direction = -1;
+            break;
+        case 1:
+            elem->direction = 1;
+            break;
+        case 2:
+            elem->nulls_direction = -1;
+            elem->nulls_direction_was_explicitly_specified = true;
+            break;
+        case 3:
+            elem->nulls_direction = 1;
+            elem->nulls_direction_was_explicitly_specified = true;
+            break;
+        case 4:
+            elem->nulls_direction = elem->direction;
+            elem->nulls_direction_was_explicitly_specified = false;
+            break;
+        default:
+            // do nothing
+            break;
+    }
+}
+
+void QueryFuzzer::fuzzOrderByList(IAST * ast)
 {
     if (!ast)
     {
         return;
     }
 
-    auto * impl = assert_cast<ASTExpressionList *>(ast.get());
+    auto * list = assert_cast<ASTExpressionList *>(ast);
+
+    // Remove element
+    if (fuzz_rand() % 50 == 0 && list->children.size() > 1)
+    {
+        // Don't remove last element -- this leads to questionable
+        // constructs such as empty select.
+        list->children.erase(list->children.begin()
+                             + fuzz_rand() % list->children.size());
+    }
+
+    // Add element
+    if (fuzz_rand() % 50 == 0)
+    {
+        auto pos = list->children.empty()
+                ? list->children.begin()
+                : list->children.begin() + fuzz_rand() % list->children.size();
+        auto col = getRandomColumnLike();
+        if (col)
+        {
+            auto elem = std::make_shared<ASTOrderByElement>();
+            elem->children.push_back(col);
+            elem->direction = 1;
+            elem->nulls_direction = 1;
+            elem->nulls_direction_was_explicitly_specified = false;
+            elem->with_fill = false;
+
+            list->children.insert(pos, elem);
+        }
+        else
+        {
+            fprintf(stderr, "no random col!\n");
+        }
+    }
+
+    // We don't have to recurse here to fuzz the children, this is handled by
+    // the generic recursion into IAST.children.
+}
+
+void QueryFuzzer::fuzzColumnLikeExpressionList(IAST * ast)
+{
+    if (!ast)
+    {
+        return;
+    }
+
+    auto * impl = assert_cast<ASTExpressionList *>(ast);
 
     // Remove element
     if (fuzz_rand() % 50 == 0 && impl->children.size() > 1)
@@ -281,17 +356,28 @@ void QueryFuzzer::fuzz(ASTPtr & ast)
     {
         fuzz(expr_list->children);
     }
+    else if (auto * order_by_element = typeid_cast<ASTOrderByElement *>(ast.get()))
+    {
+        fuzzOrderByElement(order_by_element);
+    }
     else if (auto * fn = typeid_cast<ASTFunction *>(ast.get()))
     {
-        fuzzColumnLikeExpressionList(fn->arguments);
-        fuzzColumnLikeExpressionList(fn->parameters);
+        fuzzColumnLikeExpressionList(fn->arguments.get());
+        fuzzColumnLikeExpressionList(fn->parameters.get());
 
         fuzz(fn->children);
+
+        if (fn->is_window_function)
+        {
+            fuzzColumnLikeExpressionList(fn->window_partition_by);
+            fuzzOrderByList(fn->window_order_by);
+        }
     }
     else if (auto * select = typeid_cast<ASTSelectQuery *>(ast.get()))
     {
-        fuzzColumnLikeExpressionList(select->select());
-        fuzzColumnLikeExpressionList(select->groupBy());
+        fuzzColumnLikeExpressionList(select->select().get());
+        fuzzColumnLikeExpressionList(select->groupBy().get());
+        fuzzOrderByList(select->orderBy().get());
 
         fuzz(select->children);
     }
diff --git a/programs/client/QueryFuzzer.h b/programs/client/QueryFuzzer.h
index 0c7cec8dc84..ffec09c3883 100644
--- a/programs/client/QueryFuzzer.h
+++ b/programs/client/QueryFuzzer.h
@@ -12,6 +12,9 @@
 namespace DB
 {
 
+class ASTExpressionList;
+class ASTOrderByElement;
+
 /*
  * This is an AST-based query fuzzer that makes random modifications to query
  * AST, changing numbers, list of columns, functions, etc. It remembers part of
@@ -46,7 +49,9 @@ struct QueryFuzzer
     ASTPtr getRandomColumnLike();
     void replaceWithColumnLike(ASTPtr & ast);
     void replaceWithTableLike(ASTPtr & ast);
-    void fuzzColumnLikeExpressionList(ASTPtr ast);
+    void fuzzOrderByElement(ASTOrderByElement * elem);
+    void fuzzOrderByList(IAST * ast);
+    void fuzzColumnLikeExpressionList(IAST * ast);
     void fuzz(ASTs & asts);
     void fuzz(ASTPtr & ast);
     void collectFuzzInfoMain(const ASTPtr ast);
diff --git a/src/Interpreters/ActionsVisitor.cpp b/src/Interpreters/ActionsVisitor.cpp
index f3066e6dc2a..c7bbc019518 100644
--- a/src/Interpreters/ActionsVisitor.cpp
+++ b/src/Interpreters/ActionsVisitor.cpp
@@ -740,8 +740,14 @@ void ActionsMatcher::visit(const ASTFunction & node, const ASTPtr & ast, Data &
         // Also add columns from PARTITION BY and ORDER BY of window functions.
         // Requiring a constant reference to a shared pointer to non-const AST
         // doesn't really look sane, but the visitor does indeed require it.
-        visit(node.window_partition_by->clone(), data);
-        visit(node.window_order_by->clone(), data);
+        if (node.window_partition_by)
+        {
+            visit(node.window_partition_by->clone(), data);
+        }
+        if (node.window_order_by)
+        {
+            visit(node.window_order_by->clone(), data);
+        }
 
         // Don't need to do anything more for window functions here -- the
         // resulting column is added in ExpressionAnalyzer, similar to the
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index 06e4332a862..e2939b2bd89 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -435,45 +435,69 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
         }
     }
 
-    if (!written)
+    if (written)
     {
-        settings.ostr << (settings.hilite ? hilite_function : "") << name;
-
-        if (parameters)
-        {
-            settings.ostr << '(' << (settings.hilite ? hilite_none : "");
-            parameters->formatImpl(settings, state, nested_dont_need_parens);
-            settings.ostr << (settings.hilite ? hilite_function : "") << ')';
-        }
-
-        if ((arguments && !arguments->children.empty()) || !no_empty_args)
-            settings.ostr << '(' << (settings.hilite ? hilite_none : "");
-
-        if (arguments)
-        {
-            bool special_hilite_regexp = settings.hilite
-                && (name == "match" || name == "extract" || name == "extractAll" || name == "replaceRegexpOne"
-                    || name == "replaceRegexpAll");
-
-            for (size_t i = 0, size = arguments->children.size(); i < size; ++i)
-            {
-                if (i != 0)
-                    settings.ostr << ", ";
-
-                bool special_hilite = false;
-                if (i == 1 && special_hilite_regexp)
-                    special_hilite = highlightStringLiteralWithMetacharacters(arguments->children[i], settings, "|()^$.[]?*+{:-");
-
-                if (!special_hilite)
-                    arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
-            }
-        }
-
-        if ((arguments && !arguments->children.empty()) || !no_empty_args)
-            settings.ostr << (settings.hilite ? hilite_function : "") << ')';
-
-        settings.ostr << (settings.hilite ? hilite_none : "");
+        return;
     }
+
+    settings.ostr << (settings.hilite ? hilite_function : "") << name;
+
+    if (parameters)
+    {
+        settings.ostr << '(' << (settings.hilite ? hilite_none : "");
+        parameters->formatImpl(settings, state, nested_dont_need_parens);
+        settings.ostr << (settings.hilite ? hilite_function : "") << ')';
+    }
+
+    if ((arguments && !arguments->children.empty()) || !no_empty_args)
+        settings.ostr << '(' << (settings.hilite ? hilite_none : "");
+
+    if (arguments)
+    {
+        bool special_hilite_regexp = settings.hilite
+            && (name == "match" || name == "extract" || name == "extractAll" || name == "replaceRegexpOne"
+                || name == "replaceRegexpAll");
+
+        for (size_t i = 0, size = arguments->children.size(); i < size; ++i)
+        {
+            if (i != 0)
+                settings.ostr << ", ";
+
+            bool special_hilite = false;
+            if (i == 1 && special_hilite_regexp)
+                special_hilite = highlightStringLiteralWithMetacharacters(arguments->children[i], settings, "|()^$.[]?*+{:-");
+
+            if (!special_hilite)
+                arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
+        }
+    }
+
+    if ((arguments && !arguments->children.empty()) || !no_empty_args)
+        settings.ostr << (settings.hilite ? hilite_function : "") << ')';
+
+    settings.ostr << (settings.hilite ? hilite_none : "");
+
+    if (!is_window_function)
+    {
+        return;
+    }
+
+    settings.ostr << " OVER (";
+    if (window_partition_by)
+    {
+        settings.ostr << "PARTITION BY ";
+        window_partition_by->formatImpl(settings, state, nested_dont_need_parens);
+    }
+    if (window_partition_by && window_order_by)
+    {
+        settings.ostr << " ";
+    }
+    if (window_order_by)
+    {
+        settings.ostr << "ORDER BY ";
+        window_order_by->formatImpl(settings, state, nested_dont_need_parens);
+    }
+    settings.ostr << ")";
 }
 
 }

From 157fe3e8b4e7b2b5cd0ebbcf693381bf9c888412 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 18 Dec 2020 03:49:18 +0300
Subject: [PATCH 158/504] some tests + ability to echo the queries

---
 programs/client/Client.cpp                    |  5 +
 programs/client/TestHint.h                    |  5 +
 src/Interpreters/ExpressionAnalyzer.cpp       |  2 -
 src/Parsers/ExpressionElementParsers.cpp      |  2 -
 .../01591_window_functions.reference          | 92 +++++++++++++++++++
 .../0_stateless/01591_window_functions.sql    | 27 ++++++
 6 files changed, 129 insertions(+), 4 deletions(-)
 create mode 100644 tests/queries/0_stateless/01591_window_functions.reference
 create mode 100644 tests/queries/0_stateless/01591_window_functions.sql

diff --git a/programs/client/Client.cpp b/programs/client/Client.cpp
index 92ae37123fa..a0c1e24b386 100644
--- a/programs/client/Client.cpp
+++ b/programs/client/Client.cpp
@@ -949,6 +949,11 @@ private:
             TestHint test_hint(test_mode, all_queries_text);
             if (test_hint.clientError() || test_hint.serverError())
                 processTextAsSingleQuery("SET send_logs_level = 'none'");
+
+            // Echo all queries if asked; makes for a more readable reference
+            // file.
+            if (test_hint.echoQueries())
+                echo_queries = true;
         }
 
         /// Several queries separated by ';'.
diff --git a/programs/client/TestHint.h b/programs/client/TestHint.h
index 65666f4304c..f1998588261 100644
--- a/programs/client/TestHint.h
+++ b/programs/client/TestHint.h
@@ -19,6 +19,7 @@ namespace ErrorCodes
 
 /// Checks expected server and client error codes in testmode.
 /// To enable it add special comment after the query: "-- { serverError 60 }" or "-- { clientError 20 }".
+/// Also you can enable echoing all queries by writing "-- { echo }".
 class TestHint
 {
 public:
@@ -84,12 +85,14 @@ public:
 
     int serverError() const { return server_error; }
     int clientError() const { return client_error; }
+    bool echoQueries() const { return echo; }
 
 private:
     bool enabled = false;
     const String & query;
     int server_error = 0;
     int client_error = 0;
+    bool echo = false;
 
     void parse(const String & hint)
     {
@@ -107,6 +110,8 @@ private:
                 ss >> server_error;
             else if (item == "clientError")
                 ss >> client_error;
+            else if (item == "echo")
+                echo = true;
         }
     }
 
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 346c287f239..8e410c5bebd 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -474,8 +474,6 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 {
     for (const ASTFunction * function_node : windowFunctions())
     {
-        fmt::print(stderr, "window function ast: {}\n", function_node->dumpTree());
-
         assert(function_node->is_window_function);
 
         // FIXME not thread-safe, should use a per-query counter.
diff --git a/src/Parsers/ExpressionElementParsers.cpp b/src/Parsers/ExpressionElementParsers.cpp
index 575d0331b08..cd9d8e921ef 100644
--- a/src/Parsers/ExpressionElementParsers.cpp
+++ b/src/Parsers/ExpressionElementParsers.cpp
@@ -396,8 +396,6 @@ bool ParserFunction::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
         {
             return false;
         }
-
-        fmt::print(stderr, "window function ast {}\n", function_node->dumpTree());
     }
 
     node = function_node;
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
new file mode 100644
index 00000000000..32574813f1e
--- /dev/null
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -0,0 +1,92 @@
+select number, count() over (partition by intDiv(number, 3) order by number) from numbers(10);
+
+-- proper calculation across blocks
+
+0	1
+1	2
+2	3
+3	1
+4	2
+5	3
+6	1
+7	2
+8	3
+9	1
+select number, max(number) over (partition by intDiv(number, 3) order by number desc) from numbers(10) settings max_block_size = 2;
+
+-- not a window function
+
+2	2
+1	2
+0	2
+5	5
+4	5
+3	5
+8	8
+7	8
+6	8
+9	9
+select number, abs(number) over (partition by toString(intDiv(number, 3))) from numbers(10); -- { serverError 63 }
+
+-- no partition by
+
+select number, avg(number) over (order by number) from numbers(10);
+
+-- no order by
+
+0	0
+1	0
+2	1
+3	1
+4	2
+5	2
+6	3
+7	3
+8	4
+9	4
+select number, quantileExact(number) over (partition by intDiv(number, 3)) from numbers(10);
+
+-- can add an alias after window spec
+
+0	0
+1	1
+2	1
+3	3
+4	4
+5	4
+6	6
+7	7
+8	7
+9	9
+select number, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10);
+
+-- can't reference it yet -- the window functions are calculated at the
+-- last stage of select, after all other functions.
+
+0	0
+1	1
+2	1
+3	3
+4	4
+5	4
+6	6
+7	7
+8	7
+9	9
+select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
+
+-- should work in ORDER BY though
+
+select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+
+
+9	9
+6	8
+7	8
+8	8
+3	5
+4	5
+5	5
+0	2
+1	2
+2	2
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
new file mode 100644
index 00000000000..4a23eb9fa5c
--- /dev/null
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -0,0 +1,27 @@
+-- { echo }
+
+-- just something basic
+select number, count() over (partition by intDiv(number, 3) order by number) from numbers(10);
+
+-- proper calculation across blocks
+select number, max(number) over (partition by intDiv(number, 3) order by number desc) from numbers(10) settings max_block_size = 2;
+
+-- not a window function
+select number, abs(number) over (partition by toString(intDiv(number, 3))) from numbers(10); -- { serverError 63 }
+
+-- no partition by
+select number, avg(number) over (order by number) from numbers(10);
+
+-- no order by
+select number, quantileExact(number) over (partition by intDiv(number, 3)) from numbers(10);
+
+-- can add an alias after window spec
+select number, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10);
+
+-- can't reference it yet -- the window functions are calculated at the
+-- last stage of select, after all other functions.
+select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
+
+-- should work in ORDER BY though
+select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+

From 4bfcb2fc16119a7606c66263c0e065977eddb73f Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 18 Dec 2020 04:25:21 +0300
Subject: [PATCH 159/504] a setting

---
 src/Core/Settings.h                                  | 1 +
 src/Interpreters/ExpressionAnalyzer.cpp              | 9 +++++++++
 tests/queries/0_stateless/01591_window_functions.sql | 2 ++
 3 files changed, 12 insertions(+)

diff --git a/src/Core/Settings.h b/src/Core/Settings.h
index 4e6fdddc8c6..cd549038097 100644
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@@ -405,6 +405,7 @@ class IColumn;
     M(Bool, optimize_skip_merged_partitions, false, "Skip partitions with one part with level > 0 in optimize final", 0) \
     M(Bool, optimize_on_insert, true, "Do the same transformation for inserted block of data as if merge was done on this block.", 0) \
     M(Bool, allow_experimental_map_type, false, "Allow data type Map", 0) \
+    M(Bool, allow_experimental_window_functions, false, "Allow experimental window functions", 0) \
     \
     M(Bool, use_antlr_parser, false, "Parse incoming queries using ANTLR-generated parser", 0) \
     \
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 8e410c5bebd..61b8dc86350 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -472,6 +472,15 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
 
 bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 {
+    // Convenient to check here because at least we have the Context.
+    if (windowFunctions().size() > 0 &&
+        !context.getSettingsRef().allow_experimental_window_functions)
+    {
+        throw Exception(ErrorCodes::NOT_IMPLEMENTED,
+            "Window functions are not implemented (while processing '{}')",
+            windowFunctions()[0]->formatForErrorMessage());
+    }
+
     for (const ASTFunction * function_node : windowFunctions())
     {
         assert(function_node->is_window_function);
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index 4a23eb9fa5c..fcb3da896dc 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -1,5 +1,7 @@
 -- { echo }
 
+set allow_experimental_window_functions = 1;
+
 -- just something basic
 select number, count() over (partition by intDiv(number, 3) order by number) from numbers(10);
 

From 5b6143bf2f3281a9e57f9316cf7c8c313f1b8ee6 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 15:42:12 +0300
Subject: [PATCH 160/504] fix style error

---
 src/Functions/multiply.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index 8d86e1fbdd7..54695a1087d 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -30,7 +30,7 @@ struct MultiplyImpl
     template <typename Result = ResultType>
     static inline bool apply(A a, B b, Result & c)
     {
-        if constexpr(std::is_same_v<Result, float> || std::is_same_v<Result, double>)
+        if constexpr (std::is_same_v<Result, float> || std::is_same_v<Result, double>)
         {
             c = static_cast<Result>(a) * b;
             return false;

From d4f8c46210fce2491cd4ad49e41f9e1bf82d0170 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 16:04:19 +0300
Subject: [PATCH 161/504] fixed the test

---
 tests/queries/0_stateless/01603_decimal_mult_float.sql | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
index 0c722c1899b..057e3e95071 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.sql
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -7,7 +7,7 @@ CREATE TABLE IF NOT EXISTS test01603 (
     f64 Float64,
     d Decimal64(3) DEFAULT toDecimal32(f64, 3),
     f32 Float32 DEFAULT f64
-) ENGINE=Memory() ORDER BY f32;
+) ENGINE=MergeTree() ORDER BY f32;
 
 INSERT INTO test01603(f64) SELECT 1 / (number + 1) FROM system.numbers LIMIT 1000;
 

From c7ef57c6fd8e640bcd46b1848f118600e62cbdb6 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 18 Dec 2020 16:15:03 +0300
Subject: [PATCH 162/504] Add setting async_socket_for_remote

---
 src/Core/Settings.h                           |  2 +
 src/DataStreams/RemoteQueryExecutor.cpp       |  7 +++-
 src/DataStreams/RemoteQueryExecutor.h         |  3 +-
 .../ClusterProxy/SelectStreamFactory.cpp      |  7 ++--
 src/Processors/Sources/RemoteSource.cpp       | 37 +++++++++++--------
 src/Processors/Sources/RemoteSource.h         |  5 ++-
 6 files changed, 36 insertions(+), 25 deletions(-)

diff --git a/src/Core/Settings.h b/src/Core/Settings.h
index 4a50dd09a89..789d0bb99e1 100644
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@@ -405,6 +405,8 @@ class IColumn;
     \
     M(Bool, use_antlr_parser, false, "Parse incoming queries using ANTLR-generated parser", 0) \
     \
+    M(Bool, async_socket_for_remote, true, "Asynchronously read from socket executing remote query", 0) \
+    \
     /** Obsolete settings that do nothing but left for compatibility reasons. Remove each one after half a year of obsolescence. */ \
     \
     M(UInt64, max_memory_usage_for_all_queries, 0, "Obsolete. Will be removed after 2020-10-20", 0) \
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 06d8b0b7e37..5123c763218 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -199,9 +199,10 @@ Block RemoteQueryExecutor::read()
     }
 }
 
-#if defined(OS_LINUX)
 std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext> & read_context)
 {
+
+#if defined(OS_LINUX)
     if (!sent_query)
     {
         sendQuery();
@@ -236,8 +237,10 @@ std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext>
         }
     }
     while (true);
-}
+#else
+    return read();
 #endif
+}
 
 std::optional<Block> RemoteQueryExecutor::processPacket(Packet packet)
 {
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index d15bd7a2aa0..754437b22e0 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -59,11 +59,10 @@ public:
     /// Read next block of data. Returns empty block if query is finished.
     Block read();
 
-#if defined(OS_LINUX)
     /// Async variant of read. Returns ready block or file descriptor which may be used for polling.
     /// ReadContext is an internal read state. Pass empty ptr first time, reuse created one for every call.
     std::variant<Block, int> read(std::unique_ptr<ReadContext> & read_context);
-#endif
+
     /// Receive all remain packets and finish query.
     /// It should be cancelled after read returned empty block.
     void finish(std::unique_ptr<ReadContext> * read_context = nullptr);
diff --git a/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp b/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp
index e2a7c5b55dc..47726e49d50 100644
--- a/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp
+++ b/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp
@@ -126,6 +126,7 @@ void SelectStreamFactory::createForShard(
     bool add_agg_info = processed_stage == QueryProcessingStage::WithMergeableState;
     bool add_totals = false;
     bool add_extremes = false;
+    bool async_read = context_ptr->getSettingsRef().async_socket_for_remote;
     if (processed_stage == QueryProcessingStage::Complete)
     {
         add_totals = query_ast->as<ASTSelectQuery &>().group_by_with_totals;
@@ -153,7 +154,7 @@ void SelectStreamFactory::createForShard(
         if (!table_func_ptr)
             remote_query_executor->setMainTable(main_table);
 
-        remote_pipes.emplace_back(createRemoteSourcePipe(remote_query_executor, add_agg_info, add_totals, add_extremes));
+        remote_pipes.emplace_back(createRemoteSourcePipe(remote_query_executor, add_agg_info, add_totals, add_extremes, async_read));
         remote_pipes.back().addInterpreterContext(context_ptr);
     };
 
@@ -249,7 +250,7 @@ void SelectStreamFactory::createForShard(
                 pool = shard_info.pool, shard_num = shard_info.shard_num, modified_query, header = header, modified_query_ast,
                 &context, context_ptr, throttler,
                 main_table = main_table, table_func_ptr = table_func_ptr, scalars = scalars, external_tables = external_tables,
-                stage = processed_stage, local_delay, add_agg_info, add_totals, add_extremes]()
+                stage = processed_stage, local_delay, add_agg_info, add_totals, add_extremes, async_read]()
             -> Pipe
         {
             auto current_settings = context.getSettingsRef();
@@ -295,7 +296,7 @@ void SelectStreamFactory::createForShard(
                 auto remote_query_executor = std::make_shared<RemoteQueryExecutor>(
                     std::move(connections), modified_query, header, context, throttler, scalars, external_tables, stage);
 
-                return createRemoteSourcePipe(remote_query_executor, add_agg_info, add_totals, add_extremes);
+                return createRemoteSourcePipe(remote_query_executor, add_agg_info, add_totals, add_extremes, async_read);
             }
         };
 
diff --git a/src/Processors/Sources/RemoteSource.cpp b/src/Processors/Sources/RemoteSource.cpp
index b4f5d9272f2..bf3ef32214d 100644
--- a/src/Processors/Sources/RemoteSource.cpp
+++ b/src/Processors/Sources/RemoteSource.cpp
@@ -7,9 +7,10 @@
 namespace DB
 {
 
-RemoteSource::RemoteSource(RemoteQueryExecutorPtr executor, bool add_aggregation_info_)
+RemoteSource::RemoteSource(RemoteQueryExecutorPtr executor, bool add_aggregation_info_, bool async_read_)
     : SourceWithProgress(executor->getHeader(), false)
     , add_aggregation_info(add_aggregation_info_), query_executor(std::move(executor))
+    , async_read(async_read_)
 {
     /// Add AggregatedChunkInfo if we expect DataTypeAggregateFunction as a result.
     const auto & sample = getPort().getHeader();
@@ -65,21 +66,25 @@ std::optional<Chunk> RemoteSource::tryGenerate()
 
         was_query_sent = true;
     }
-#if defined(OS_LINUX)
-    auto res = query_executor->read(read_context);
-    if (std::holds_alternative<int>(res))
+
+    Block block;
+
+    if (async_read)
     {
-        fd = std::get<int>(res);
-        is_async_state = true;
-        return Chunk();
+        auto res = query_executor->read(read_context);
+        if (std::holds_alternative<int>(res))
+        {
+            fd = std::get<int>(res);
+            is_async_state = true;
+            return Chunk();
+        }
+
+        is_async_state = false;
+
+        block = std::get<Block>(std::move(res));
     }
-
-    is_async_state = false;
-
-    auto block = std::get<Block>(std::move(res));
-#else
-    auto block = query_executor->read();
-#endif
+    else
+        block = query_executor->read();
 
     if (!block)
     {
@@ -161,9 +166,9 @@ Chunk RemoteExtremesSource::generate()
 
 Pipe createRemoteSourcePipe(
     RemoteQueryExecutorPtr query_executor,
-    bool add_aggregation_info, bool add_totals, bool add_extremes)
+    bool add_aggregation_info, bool add_totals, bool add_extremes, bool async_read)
 {
-    Pipe pipe(std::make_shared<RemoteSource>(query_executor, add_aggregation_info));
+    Pipe pipe(std::make_shared<RemoteSource>(query_executor, add_aggregation_info, async_read));
 
     if (add_totals)
         pipe.addTotalsSource(std::make_shared<RemoteTotalsSource>(query_executor));
diff --git a/src/Processors/Sources/RemoteSource.h b/src/Processors/Sources/RemoteSource.h
index 79bf750f897..2109cb5eba4 100644
--- a/src/Processors/Sources/RemoteSource.h
+++ b/src/Processors/Sources/RemoteSource.h
@@ -20,7 +20,7 @@ public:
     /// Flag add_aggregation_info tells if AggregatedChunkInfo should be added to result chunk.
     /// AggregatedChunkInfo stores the bucket number used for two-level aggregation.
     /// This flag should be typically enabled for queries with GROUP BY which are executed till WithMergeableState.
-    RemoteSource(RemoteQueryExecutorPtr executor, bool add_aggregation_info_);
+    RemoteSource(RemoteQueryExecutorPtr executor, bool add_aggregation_info_, bool async_read_);
     ~RemoteSource() override;
 
     Status prepare() override;
@@ -44,6 +44,7 @@ private:
     RemoteQueryExecutorPtr query_executor;
     RowsBeforeLimitCounterPtr rows_before_limit;
 
+    const bool async_read;
     bool is_async_state = false;
     std::unique_ptr<RemoteQueryExecutorReadContext> read_context;
     int fd = -1;
@@ -84,6 +85,6 @@ private:
 /// Create pipe with remote sources.
 Pipe createRemoteSourcePipe(
     RemoteQueryExecutorPtr query_executor,
-    bool add_aggregation_info, bool add_totals, bool add_extremes);
+    bool add_aggregation_info, bool add_totals, bool add_extremes, bool async_read);
 
 }

From f34c0500dbd9b3d9af75be1151238cd3668e9b1e Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 16:29:53 +0300
Subject: [PATCH 163/504] enlarged the test, updated scales calc

---
 src/Functions/FunctionBinaryArithmetic.h      | 64 ++++++++++---------
 .../01603_decimal_mult_float.reference        |  8 +--
 .../0_stateless/01603_decimal_mult_float.sql  |  5 ++
 3 files changed, 43 insertions(+), 34 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index bf245497025..f1ed828f101 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1002,49 +1002,54 @@ public:
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
                 using FieldType = typename ResultDataType::FieldType;
 
-                static constexpr const bool dec_a = IsDecimalNumber<T0>;
-                static constexpr const bool dec_b = IsDecimalNumber<T1>;
+                static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
+                static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
                 static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
 
                 const ResultDataType type = [&] {
-                    if constexpr (dec_a && IsFloatingPoint<RightDataType>)
+                    if constexpr (left_is_decimal && IsFloatingPoint<RightDataType>)
                         return RightDataType();
-                    else if constexpr (dec_b && IsFloatingPoint<LeftDataType>)
+                    else if constexpr (right_is_decimal && IsFloatingPoint<LeftDataType>)
                         return LeftDataType();
                     else
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                FieldType scale_a;
-                FieldType scale_b;
+                FieldType left_scale; // not really a Decimal scale, depends on the latter branches.
+                FieldType right_scale;
+
+                const bool both_columns_are_const = col_left_const && col_right_const;
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                    scale_a = right.getScaleMultiplier();
+                    left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
-                    scale_a = type.scaleFactorFor(left, is_multiply);
-                else if constexpr (dec_a)
-                    scale_a = left.getScale();
+                    left_scale = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
+                else if constexpr (left_is_decimal)
+                    // While multiplying constants, we need the Decimal scales as the conversion function from
+                    // Decimal utils takes it. When one of the columns is non-const, we need the scale multiplier
+                    // instead as the OpImpl::constantConstant function uses the multiplier.
+                    left_scale = both_columns_are_const ? left.getScale() : left.getScaleMultiplier();
                 else
-                    scale_a = 0.0; //won't be used, just to silence the warning
+                    left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
-                    scale_b = type.scaleFactorFor(right, is_multiply || is_division);
-                else if constexpr (dec_b)
-                    scale_b = right.getScale();
+                    right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+                else if constexpr (right_is_decimal)
+                    right_scale = both_columns_are_const ? right.getScale() : right.getScaleMultiplier();
                 else
-                    scale_b = 0.0; //same
+                    right_scale = 0.0; //same
 
                 /// non-vector result
-                if (col_left_const && col_right_const)
+                if (both_columns_are_const)
                 {
                     const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
-                        col_left_const, scale_a);
+                        col_left_const, left_scale);
                     const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
-                        col_right_const, scale_b);
+                        col_right_const, right_scale);
 
                     auto res = check_decimal_overflow ?
-                        OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
-                        OpImpl::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b);
+                        OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(const_a, const_b, left_scale, right_scale) :
+                        OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(const_a, const_b, left_scale, right_scale);
 
                     if constexpr (result_is_decimal)
                         return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
@@ -1053,7 +1058,7 @@ public:
                          return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
-                if constexpr (IsDecimalNumber<ResultType>)
+                if constexpr (result_is_decimal)
                     col_res = ColVecResult::create(0, type.getScale());
                 else
                     col_res = ColVecResult::create(0);
@@ -1061,33 +1066,32 @@ public:
                 auto & vec_res = col_res->getData();
                 vec_res.resize(col_left_raw->size());
 
-
                 if (col_left && col_right)
                 {
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorVector<dec_a, dec_b>(col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template vectorVector<dec_a, dec_b>(col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left_const && col_right)
                 {
                     const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
-                        col_left_const, scale_a);
+                        col_left_const, left_scale);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(const_a, col_right->getData(), vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImpl::template constantVector<left_is_decimal, right_is_decimal>(const_a, col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left && col_right_const)
                 {
                     const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
-                        col_right_const, scale_b);
+                        col_right_const, right_scale);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                        OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(col_left->getData(), const_b, vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                        OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(col_left->getData(), const_b, vec_res, left_scale, right_scale);
                 }
                 else
                     return nullptr;
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
index 22650604ff3..318564b2d1a 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -2,7 +2,7 @@
 10.165
 0.00012000000000000002
 150.16500000000002
-15551.800000000001
-113245.37999999999
-1197.6753376880554
-5.988376864581369
+7775.900000000001
+56622.689999999995
+598.8376688440277
+2.9941884322906844
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
index 057e3e95071..8de35f0b3f5 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.sql
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -16,4 +16,9 @@ SELECT sum(8.01 * d) FROM test01603;
 SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
 SELECT sum(toDecimal64(0.4, 2) * f32) FROM test01603;
 
+SELECT sum(f64 * d) FROM test01603;
+SELECT sum(d * f64) FROM test01603;
+SELECT sum(f32 * d) FROM test01603;
+SELECT sum(d * f32) FROM test01603;
+
 DROP TABLE IF EXISTS test01603;

From 84013d253dff75b1f522ac1756fb8c966a51f88c Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 16:33:00 +0300
Subject: [PATCH 164/504] fix: ternary if

---
 src/Functions/FunctionBinaryArithmetic.h | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index f1ed828f101..91d90694844 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1028,14 +1028,18 @@ public:
                     // While multiplying constants, we need the Decimal scales as the conversion function from
                     // Decimal utils takes it. When one of the columns is non-const, we need the scale multiplier
                     // instead as the OpImpl::constantConstant function uses the multiplier.
-                    left_scale = both_columns_are_const ? left.getScale() : left.getScaleMultiplier();
+                    left_scale = both_columns_are_const
+                        ? left.getScale()
+                        : left.template getScaleMultiplier<FieldType>();
                 else
                     left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
                 else if constexpr (right_is_decimal)
-                    right_scale = both_columns_are_const ? right.getScale() : right.getScaleMultiplier();
+                    right_scale = both_columns_are_const
+                        ? right.getScale()
+                        : right.template getScaleMultiplier<FieldType>();
                 else
                     right_scale = 0.0; //same
 

From 15f0915ad32913a0032b497acad3367937cf2416 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 16:35:37 +0300
Subject: [PATCH 165/504] replaced ternary if

---
 src/Functions/FunctionBinaryArithmetic.h | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 91d90694844..1720a984545 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1025,21 +1025,27 @@ public:
                 else if constexpr (result_is_decimal)
                     left_scale = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
                 else if constexpr (left_is_decimal)
+                {
                     // While multiplying constants, we need the Decimal scales as the conversion function from
                     // Decimal utils takes it. When one of the columns is non-const, we need the scale multiplier
                     // instead as the OpImpl::constantConstant function uses the multiplier.
-                    left_scale = both_columns_are_const
-                        ? left.getScale()
-                        : left.template getScaleMultiplier<FieldType>();
+                    if (both_columns_are_const)
+                        left_scale = left.getScale();
+                    else
+                        left_scale = left.getScaleMultiplier();
+                }
                 else
                     left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
                 else if constexpr (right_is_decimal)
-                    right_scale = both_columns_are_const
-                        ? right.getScale()
-                        : right.template getScaleMultiplier<FieldType>();
+                {
+                    if (both_columns_are_const)
+                        right_scale = right.getScale();
+                    else
+                        right_scale = right.getScaleMultiplier();
+                }
                 else
                     right_scale = 0.0; //same
 

From bbd34b544adbcf4edfe26d4ff7bcfcf42778b9f7 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 18 Dec 2020 16:37:21 +0300
Subject: [PATCH 166/504] Try fix fluppy test.

---
 .../01244_optimize_distributed_group_by_sharding_key.sql        | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/queries/0_stateless/01244_optimize_distributed_group_by_sharding_key.sql b/tests/queries/0_stateless/01244_optimize_distributed_group_by_sharding_key.sql
index 9bc50ae2fc7..d152326091b 100644
--- a/tests/queries/0_stateless/01244_optimize_distributed_group_by_sharding_key.sql
+++ b/tests/queries/0_stateless/01244_optimize_distributed_group_by_sharding_key.sql
@@ -68,7 +68,7 @@ select 'OFFSET';
 select count(), * from dist_01247 group by number offset 1;
 -- this will emulate different data on for different shards
 select 'WHERE LIMIT OFFSET';
-select count(), * from dist_01247 where number = _shard_num-1 group by number limit 1 offset 1;
+select count(), * from dist_01247 where number = _shard_num-1 group by number order by number limit 1 offset 1;
 
 select 'LIMIT BY 1';
 select count(), * from dist_01247 group by number order by number limit 1 by number;

From c4ea410a4da0872035f3ec0a0f57b87aa66999a3 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 16:45:44 +0300
Subject: [PATCH 167/504] updated the comments and branches

---
 src/Functions/FunctionBinaryArithmetic.h | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 1720a984545..352f194cf2f 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1000,7 +1000,6 @@ public:
                 using NativeResultType = typename NativeType<ResultType>::Type;
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
-                using FieldType = typename ResultDataType::FieldType;
 
                 static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
                 static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
@@ -1015,15 +1014,18 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                FieldType left_scale; // not really a Decimal scale, depends on the latter branches.
-                FieldType right_scale;
+                /// We need to be able to assign either the integer scale or its multiplier if
+                /// only one column is decimal (if both are, the resulting decimal will be the decimalResultType).
+                using FieldType = typename ResultDataType::FieldType;
+                std::conditional_t<left_is_decimal && !result_is_decimal, T0, FieldType> left_scale;
+                std::conditional_t<right_is_decimal && !result_is_decimal, T1, FieldType> right_scale;
 
                 const bool both_columns_are_const = col_left_const && col_right_const;
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
-                    left_scale = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
+                    left_scale = type.scaleFactorFor(left, is_multiply);
                 else if constexpr (left_is_decimal)
                 {
                     // While multiplying constants, we need the Decimal scales as the conversion function from

From 07bc3a2a4d26b6be743e511aac4dfddfb343ee97 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 16:49:00 +0300
Subject: [PATCH 168/504] reverted the overengineering

---
 src/Functions/FunctionBinaryArithmetic.h | 31 ++++++------------------
 1 file changed, 7 insertions(+), 24 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 352f194cf2f..4b1018b4ab6 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1000,6 +1000,7 @@ public:
                 using NativeResultType = typename NativeType<ResultType>::Type;
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
+                using FieldType = typename ResultDataType::FieldType;
 
                 static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
                 static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
@@ -1014,45 +1015,27 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                /// We need to be able to assign either the integer scale or its multiplier if
-                /// only one column is decimal (if both are, the resulting decimal will be the decimalResultType).
-                using FieldType = typename ResultDataType::FieldType;
-                std::conditional_t<left_is_decimal && !result_is_decimal, T0, FieldType> left_scale;
-                std::conditional_t<right_is_decimal && !result_is_decimal, T1, FieldType> right_scale;
-
-                const bool both_columns_are_const = col_left_const && col_right_const;
+                FieldType left_scale; // not really a Decimal scale, depends on the latter branches.
+                FieldType right_scale;
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                     left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
-                    left_scale = type.scaleFactorFor(left, is_multiply);
+                    left_scale = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
                 else if constexpr (left_is_decimal)
-                {
-                    // While multiplying constants, we need the Decimal scales as the conversion function from
-                    // Decimal utils takes it. When one of the columns is non-const, we need the scale multiplier
-                    // instead as the OpImpl::constantConstant function uses the multiplier.
-                    if (both_columns_are_const)
-                        left_scale = left.getScale();
-                    else
-                        left_scale = left.getScaleMultiplier();
-                }
+                    left_scale = left.getScale();
                 else
                     left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
                 else if constexpr (right_is_decimal)
-                {
-                    if (both_columns_are_const)
-                        right_scale = right.getScale();
-                    else
-                        right_scale = right.getScaleMultiplier();
-                }
+                    right_scale = right.getScale();
                 else
                     right_scale = 0.0; //same
 
                 /// non-vector result
-                if (both_columns_are_const)
+                if (col_left_const && col_right_const)
                 {
                     const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
                         col_left_const, left_scale);

From c1e7d5d3c10a6e0789de3c4eb512256d719a9f1e Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 18 Dec 2020 16:49:45 +0300
Subject: [PATCH 169/504] Working code

---
 .../MergeTreeDataPartWriterOnDisk.cpp         | 15 ++---
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 65 +++++++++++++++++--
 .../MergeTree/MergeTreeDataPartWriterWide.h   |  4 ++
 3 files changed, 69 insertions(+), 15 deletions(-)

diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
index 5190b439809..150bdad55f2 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
@@ -218,6 +218,12 @@ void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block
         auto & stream = *skip_indices_streams[i];
         for (const auto & granule : granules_to_write)
         {
+            if (skip_index_accumulated_marks[i] == index_helper->index.granularity)
+            {
+                skip_indices_aggregators[i]->getGranuleAndReset()->serializeBinary(stream.compressed);
+                skip_index_accumulated_marks[i] = 0;
+            }
+
             if (skip_indices_aggregators[i]->empty() && granule.mark_on_start)
             {
                 skip_indices_aggregators[i] = index_helper->createIndexAggregator();
@@ -236,16 +242,7 @@ void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block
             size_t pos = granule.start_row;
             skip_indices_aggregators[i]->update(skip_indexes_block, &pos, granule.granularity_rows);
             if (granule.isCompleted())
-            {
                 ++skip_index_accumulated_marks[i];
-
-                /// write index if it is filled
-                if (skip_index_accumulated_marks[i] == index_helper->index.granularity)
-                {
-                    skip_indices_aggregators[i]->getGranuleAndReset()->serializeBinary(stream.compressed);
-                    skip_index_accumulated_marks[i] = 0;
-                }
-            }
         }
     }
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index 1127c3a7dd7..f9b42e7c320 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -167,6 +167,13 @@ void MergeTreeDataPartWriterWide::write(const Block & block, const IColumn::Perm
     if (compute_granularity)
     {
         size_t index_granularity_for_block = computeIndexGranularity(block);
+        if (rows_written_in_last_mark > 0)
+        {
+            size_t rows_left_in_last_mark = index_granularity.getMarkRows(getCurrentMark()) - rows_written_in_last_mark;
+            if (rows_left_in_last_mark > index_granularity_for_block)
+                adjustLastMarkAndFlushToDisk();
+        }
+
         fillIndexGranularity(index_granularity_for_block, block.rows());
     }
 
@@ -281,9 +288,6 @@ void MergeTreeDataPartWriterWide::writeSingleGranule(
     IDataType::SerializeBinaryBulkSettings & serialize_settings,
     const Granule & granule)
 {
-    if (granule.mark_on_start)
-        writeSingleMark(name, type, offset_columns, granule.granularity_rows, serialize_settings.path);
-
     type.serializeBinaryBulkWithMultipleStreams(column, granule.start_row, granule.granularity_rows, serialize_settings, serialization_state);
 
     /// So that instead of the marks pointing to the end of the compressed block, there were marks pointing to the beginning of the next one.
@@ -309,6 +313,9 @@ void MergeTreeDataPartWriterWide::writeColumn(
     WrittenOffsetColumns & offset_columns,
     const Granules & granules)
 {
+    if (granules.empty())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Empty granules for column {}, current mark {}", backQuoteIfNeed(name), getCurrentMark());
+
     auto [it, inserted] = serialization_states.emplace(name, nullptr);
 
     if (inserted)
@@ -329,6 +336,13 @@ void MergeTreeDataPartWriterWide::writeColumn(
         if (granule.granularity_rows > 0)
             data_written = true;
 
+        if (granule.mark_on_start)
+        {
+            if (last_non_written_marks.count(name))
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "We have to add new mark for column, but already have non written mark. Current mark {}, total marks {}, offset {}", getCurrentMark(), index_granularity.getMarksCount(), rows_written_in_last_mark);
+            last_non_written_marks[name] = getCurrentMarksForColumn(name, type, offset_columns, serialize_settings.path);
+        }
+
         writeSingleGranule(
            name,
            type,
@@ -338,6 +352,17 @@ void MergeTreeDataPartWriterWide::writeColumn(
            serialize_settings,
            granule
         );
+
+        if (granule.isCompleted())
+        {
+            auto it = last_non_written_marks.find(name);
+            if (it == last_non_written_marks.end())
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "No mark was saved for incomplete granule for column {}", backQuoteIfNeed(name));
+
+            for (const auto & mark : it->second)
+                flushMarkToFile(mark, index_granularity.getMarkRows(granule.mark_number));
+            last_non_written_marks.erase(it);
+        }
     }
 
     type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
@@ -404,8 +429,8 @@ void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name,
 
         if (index_granularity_rows != index_granularity.getMarkRows(mark_num))
             throw Exception(
-                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), in-memory {}, on disk {}",
-                mark_num, offset_in_compressed_file, offset_in_decompressed_block, index_granularity.getMarkRows(mark_num), index_granularity_rows);
+                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), in-memory {}, on disk {}, total marks {}",
+                mark_num, offset_in_compressed_file, offset_in_decompressed_block, index_granularity.getMarkRows(mark_num), index_granularity_rows, index_granularity.getMarksCount());
 
         auto column = type.createColumn();
 
@@ -415,7 +440,9 @@ void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name,
         {
             must_be_last = true;
         }
-        else if (column->size() != index_granularity_rows)
+
+        /// Now they must be equal
+        if (column->size() != index_granularity_rows)
         {
             throw Exception(
                 ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), actually in bin file {}, in mrk file {}",
@@ -445,6 +472,8 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
     serialize_settings.low_cardinality_max_dictionary_size = global_settings.low_cardinality_max_dictionary_size;
     serialize_settings.low_cardinality_use_single_dictionary_for_part = global_settings.low_cardinality_use_single_dictionary_for_part != 0;
     WrittenOffsetColumns offset_columns;
+    if (rows_written_in_last_mark > 0)
+        adjustLastMarkAndFlushToDisk();
 
     bool write_final_mark = (with_final_mark && data_written);
 
@@ -537,4 +566,28 @@ void MergeTreeDataPartWriterWide::fillIndexGranularity(size_t index_granularity_
         rows_in_block);
 }
 
+
+void MergeTreeDataPartWriterWide::adjustLastMarkAndFlushToDisk()
+{
+    if (getCurrentMark() != index_granularity.getMarksCount() - 1)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Non last mark {} having rows offset {}, total marks {}", getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
+
+    if (last_non_written_marks.empty())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "No saved marks for last mark {} having rows offset {}, total marks {}", getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
+
+    index_granularity.popMark();
+    index_granularity.appendMark(rows_written_in_last_mark);
+    for (const auto & [name, marks] : last_non_written_marks)
+        for (const auto & mark : marks)
+            flushMarkToFile(mark, index_granularity.getMarkRows(getCurrentMark()));
+
+    last_non_written_marks.clear();
+
+    for (size_t i = 0; i < skip_indices.size(); ++i)
+        ++skip_index_accumulated_marks[i];
+
+    setCurrentMark(getCurrentMark() + 1);
+    rows_written_in_last_mark = 0;
+}
+
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
index cffedd66a32..afaf5958338 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
@@ -99,6 +99,8 @@ private:
     /// in our index_granularity array.
     void shiftCurrentMark(const Granules & granules_written);
 
+    void adjustLastMarkAndFlushToDisk();
+
     IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns) const;
 
     using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
@@ -108,6 +110,8 @@ private:
 
     using ColumnStreams = std::map<String, StreamPtr>;
     ColumnStreams column_streams;
+    using MarksForColumns = std::unordered_map<String, StreamsWithMarks>;
+    MarksForColumns last_non_written_marks;
 
     /// How many rows we have already written in the current mark.
     /// More than zero when incoming blocks are smaller then their granularity.

From 1f105145994d55e903a2e0343ddaa2e706a92ff6 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:02:02 +0300
Subject: [PATCH 170/504] another attempt

---
 src/Functions/FunctionBinaryArithmetic.h      | 59 +++++++++++--------
 .../01603_decimal_mult_float.reference        |  4 ++
 2 files changed, 40 insertions(+), 23 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 4b1018b4ab6..1ed97535232 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1015,36 +1015,45 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                FieldType left_scale; // not really a Decimal scale, depends on the latter branches.
-                FieldType right_scale;
+                FieldType left_scale_factor; //Used by OpImpl if the left column is not const
+                FieldType right_scale_factor;
+
+                FieldType left_scale {0}; // Used by conversion functions if the left column is const
+                FieldType right_scale {0};
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                    left_scale = right.getScaleMultiplier();
+                    left_scale_factor = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
-                    left_scale = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
+                    left_scale_factor = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
                 else if constexpr (left_is_decimal)
+                {
+                    left_scale_factor = left.getScaleMultiplier();
                     left_scale = left.getScale();
+                }
                 else
-                    left_scale = 0.0; //won't be used, just to silence the warning
+                    left_scale_factor = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
-                    right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+                    right_scale_factor = type.scaleFactorFor(right, is_multiply || is_division);
                 else if constexpr (right_is_decimal)
+                {
+                    right_scale_factor = right.getScaleMultiplier();
                     right_scale = right.getScale();
+                }
                 else
-                    right_scale = 0.0; //same
+                    right_scale_factor = 0.0; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
-                        col_left_const, left_scale);
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
-                        col_right_const, right_scale);
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left_scale);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right_scale);
 
                     auto res = check_decimal_overflow ?
-                        OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(const_a, const_b, left_scale, right_scale) :
-                        OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(const_a, const_b, left_scale, right_scale);
+                        OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
+                            const_a, const_b, left_scale_factor, right_scale_factor) :
+                        OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
+                            const_a, const_b, left_scale_factor, right_scale_factor);
 
                     if constexpr (result_is_decimal)
                         return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
@@ -1064,29 +1073,33 @@ public:
                 if (col_left && col_right)
                 {
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
+                        OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
                     else
-                        OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
+                        OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
                 }
                 else if (col_left_const && col_right)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
-                        col_left_const, left_scale);
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left_scale);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(const_a, col_right->getData(), vec_res, left_scale, right_scale);
+                        OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
+                            const_a, col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
                     else
-                        OpImpl::template constantVector<left_is_decimal, right_is_decimal>(const_a, col_right->getData(), vec_res, left_scale, right_scale);
+                        OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
+                            const_a, col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
                 }
                 else if (col_left && col_right_const)
                 {
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
-                        col_right_const, right_scale);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right_scale);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(col_left->getData(), const_b, vec_res, left_scale, right_scale);
+                        OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), const_b, vec_res, left_scale_factor, right_scale_factor);
                     else
-                        OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(col_left->getData(), const_b, vec_res, left_scale, right_scale);
+                        OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), const_b, vec_res, left_scale_factor, right_scale_factor);
                 }
                 else
                     return nullptr;
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
index 318564b2d1a..fd2bc721de3 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -6,3 +6,7 @@
 56622.689999999995
 598.8376688440277
 2.9941884322906844
+1641.386318314034
+1641.386318314034
+1641.3863211679272
+1641.3863211679272

From 6c91b3dba0945a0aa40288291c2c93727ff69c59 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:10:37 +0300
Subject: [PATCH 171/504] added explicit scale calc

---
 src/Functions/FunctionBinaryArithmetic.h | 55 ++++++++++--------------
 1 file changed, 23 insertions(+), 32 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 1ed97535232..2581e256727 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1015,45 +1015,34 @@ public:
                         return decimalResultType<is_multiply, is_division>(left, right);
                 }();
 
-                FieldType left_scale_factor; //Used by OpImpl if the left column is not const
-                FieldType right_scale_factor;
-
-                FieldType left_scale {0}; // Used by conversion functions if the left column is const
-                FieldType right_scale {0};
+                FieldType left_scale;
+                FieldType right_scale;
 
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                    left_scale_factor = right.getScaleMultiplier();
+                    left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
-                    left_scale_factor = type.scaleFactorFor(left, is_multiply); //both are decimal, so multiply on 1 / scale
-                else if constexpr (left_is_decimal)
-                {
-                    left_scale_factor = left.getScaleMultiplier();
-                    left_scale = left.getScale();
-                }
+                    left_scale = type.scaleFactorFor(left, is_multiply);
                 else
-                    left_scale_factor = 0.0; //won't be used, just to silence the warning
+                    left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
-                    right_scale_factor = type.scaleFactorFor(right, is_multiply || is_division);
-                else if constexpr (right_is_decimal)
-                {
-                    right_scale_factor = right.getScaleMultiplier();
-                    right_scale = right.getScale();
-                }
+                    right_scale = type.scaleFactorFor(right, is_multiply || is_division);
                 else
-                    right_scale_factor = 0.0; //same
+                    right_scale = 0.0; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left_scale);
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right_scale);
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
+                        col_left_const, left.getScale());
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
+                        col_right_const, right.getScale());
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
-                            const_a, const_b, left_scale_factor, right_scale_factor) :
+                            const_a, const_b, left_scale, right_scale) :
                         OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
-                            const_a, const_b, left_scale_factor, right_scale_factor);
+                            const_a, const_b, left_scale, right_scale);
 
                     if constexpr (result_is_decimal)
                         return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
@@ -1074,32 +1063,34 @@ public:
                 {
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
+                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                     else
                         OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
+                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left_const && col_right)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left_scale);
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
+                        col_left_const, left.getScale());
 
                     if (check_decimal_overflow)
                         OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
-                            const_a, col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
+                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
                     else
                         OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
-                            const_a, col_right->getData(), vec_res, left_scale_factor, right_scale_factor);
+                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left && col_right_const)
                 {
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right_scale);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
+                        col_right_const, right.getScale());
 
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), const_b, vec_res, left_scale_factor, right_scale_factor);
+                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
                     else
                         OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), const_b, vec_res, left_scale_factor, right_scale_factor);
+                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
                 }
                 else
                     return nullptr;

From 611d400106040e5099dc47a4b7ec71f807a9f324 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:13:15 +0300
Subject: [PATCH 172/504] fixed compile error

---
 src/Functions/FunctionBinaryArithmetic.h | 18 +++++++-----------
 1 file changed, 7 insertions(+), 11 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 2581e256727..6639e0c60ef 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -957,15 +957,15 @@ public:
     }
 
     template <class T, class ResultDataType>
-    static auto helperGetOrConvert(const auto & col, const auto & scale)
+    static auto helperGetOrConvert(const auto & col_const, const auto & col)
     {
         using ResultType = typename ResultDataType::FieldType;
         using NativeResultType = typename NativeType<ResultType>::Type;
 
         if constexpr (IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
-            return DecimalUtils::convertTo<NativeResultType>(col->template getValue<T>(), scale);
+            return DecimalUtils::convertTo<NativeResultType>(col_const->template getValue<T>(), col.getScale());
         else
-            return col->template getValue<T>();
+            return col_const->template getValue<T>();
     }
 
     template <typename A, typename B>
@@ -1033,10 +1033,8 @@ public:
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
-                        col_left_const, left.getScale());
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
-                        col_right_const, right.getScale());
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
                     auto res = check_decimal_overflow ?
                         OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
@@ -1070,8 +1068,7 @@ public:
                 }
                 else if (col_left_const && col_right)
                 {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(
-                        col_left_const, left.getScale());
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
@@ -1082,8 +1079,7 @@ public:
                 }
                 else if (col_left && col_right_const)
                 {
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(
-                        col_right_const, right.getScale());
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
                     if (check_decimal_overflow)
                         OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(

From 764d4622886384b5a60229bb2a1c88d5eeac5764 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:16:54 +0300
Subject: [PATCH 173/504] updated test with possible right values

---
 .../0_stateless/01603_decimal_mult_float.reference   | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
index fd2bc721de3..08e0fc89922 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -2,11 +2,11 @@
 10.165
 0.00012000000000000002
 150.16500000000002
-7775.900000000001
-56622.689999999995
+7.775900000000001
+56.622689999999995
 598.8376688440277
 2.9941884322906844
-1641.386318314034
-1641.386318314034
-1641.3863211679272
-1641.3863211679272
+1.641386318314034
+1.641386318314034
+1.6413863211679272
+1.6413863211679272

From bc651b304750d28ca722430ad3b4c433b8ab82eb Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:21:32 +0300
Subject: [PATCH 174/504] added the suboptimal solution

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 6639e0c60ef..1510bc73c57 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1022,11 +1022,15 @@ public:
                     left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
                     left_scale = type.scaleFactorFor(left, is_multiply);
+                else if constexpr(left_is_decimal) // BUG precision loss
+                    left_scale = DecimalUtils::convertTo<FieldType>(left.getScaleMultiplier());
                 else
                     left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+                else if constexpr(right_is_decimal) //BUG precision loss
+                    right_scale = DecimalUtils::convertTo<FieldType>(right.getScaleMultiplier());
                 else
                     right_scale = 0.0; //same
 

From 6488846c9200c92629953bde0e6412a83fbdae14 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:23:51 +0300
Subject: [PATCH 175/504] signature fix

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 1510bc73c57..07152a16eb8 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1023,14 +1023,14 @@ public:
                 else if constexpr (result_is_decimal)
                     left_scale = type.scaleFactorFor(left, is_multiply);
                 else if constexpr(left_is_decimal) // BUG precision loss
-                    left_scale = DecimalUtils::convertTo<FieldType>(left.getScaleMultiplier());
+                    left_scale = DecimalUtils::convertTo<FieldType>(left.getScaleMultiplier(), left.getScale());
                 else
                     left_scale = 0.0; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
                 else if constexpr(right_is_decimal) //BUG precision loss
-                    right_scale = DecimalUtils::convertTo<FieldType>(right.getScaleMultiplier());
+                    right_scale = DecimalUtils::convertTo<FieldType>(right.getScaleMultiplier(), right.getScale());
                 else
                     right_scale = 0.0; //same
 

From 01c8d5dccc6522f11a0fbfa5456bb898c1105d19 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 18 Dec 2020 17:24:44 +0300
Subject: [PATCH 176/504] Use replace fiber to callback in
 ReadBufferFromPocoSocket

---
 src/CMakeLists.txt                            |  1 -
 src/Client/Connection.cpp                     |  6 +-
 src/Client/Connection.h                       |  5 +-
 src/Client/MultiplexedConnections.cpp         |  4 +-
 src/Client/MultiplexedConnections.h           |  4 +-
 src/Common/Config/CMakeLists.txt              |  1 -
 src/Common/ZooKeeper/CMakeLists.txt           |  2 +-
 .../RemoteQueryExecutorReadContext.h          | 57 +++++++++----------
 src/IO/ReadBufferFromPocoSocket.cpp           | 10 ++--
 src/IO/ReadBufferFromPocoSocket.h             |  6 +-
 10 files changed, 43 insertions(+), 53 deletions(-)

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index f3afbeb61dc..4771afd4c24 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -428,7 +428,6 @@ if (USE_ROCKSDB)
     dbms_target_include_directories(SYSTEM BEFORE PUBLIC ${ROCKSDB_INCLUDE_DIR})
 endif()
 
-target_link_libraries (clickhouse_common_io PRIVATE _boost_context)
 dbms_target_link_libraries(PRIVATE _boost_context)
 
 if (ENABLE_TESTS AND USE_GTEST)
diff --git a/src/Client/Connection.cpp b/src/Client/Connection.cpp
index 16f0786c70b..ef114490c51 100644
--- a/src/Client/Connection.cpp
+++ b/src/Client/Connection.cpp
@@ -742,10 +742,10 @@ std::optional<UInt64> Connection::checkPacket(size_t timeout_microseconds)
 }
 
 
-Packet Connection::receivePacket(Fiber * fiber)
+Packet Connection::receivePacket(std::function<void(Poco::Net::Socket &)> async_callback)
 {
-    in->setFiber(fiber);
-    SCOPE_EXIT(in->setFiber(nullptr));
+    in->setAsyncCallback(std::move(async_callback));
+    SCOPE_EXIT(in->setAsyncCallback({}));
 
     try
     {
diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index d5d202d9ac2..c2a5d60b9b6 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -171,7 +171,8 @@ public:
     std::optional<UInt64> checkPacket(size_t timeout_microseconds = 0);
 
     /// Receive packet from server.
-    Packet receivePacket(Fiber * fiber = nullptr);
+    /// Each time read blocks and async_callback is set, it will be called. You can poll socket inside it.
+    Packet receivePacket(std::function<void(Poco::Net::Socket &)> async_callback = {});
 
     /// If not connected yet, or if connection is broken - then connect. If cannot connect - throw an exception.
     void forceConnected(const ConnectionTimeouts & timeouts);
@@ -190,8 +191,6 @@ public:
     size_t outBytesCount() const { return out ? out->count() : 0; }
     size_t inBytesCount() const { return in ? in->count() : 0; }
 
-    Poco::Net::Socket & getSocket() { return in->getSocket(); }
-
 private:
     String host;
     UInt16 port;
diff --git a/src/Client/MultiplexedConnections.cpp b/src/Client/MultiplexedConnections.cpp
index 87f8ff34c2f..ed7aad0a515 100644
--- a/src/Client/MultiplexedConnections.cpp
+++ b/src/Client/MultiplexedConnections.cpp
@@ -237,7 +237,7 @@ std::string MultiplexedConnections::dumpAddressesUnlocked() const
     return buf.str();
 }
 
-Packet MultiplexedConnections::receivePacketUnlocked(Fiber * fiber)
+Packet MultiplexedConnections::receivePacketUnlocked(std::function<void(Poco::Net::Socket &)> async_callback)
 {
     if (!sent_query)
         throw Exception("Cannot receive packets: no query sent.", ErrorCodes::LOGICAL_ERROR);
@@ -249,7 +249,7 @@ Packet MultiplexedConnections::receivePacketUnlocked(Fiber * fiber)
     if (current_connection == nullptr)
         throw Exception("Logical error: no available replica", ErrorCodes::NO_AVAILABLE_REPLICA);
 
-    Packet packet = current_connection->receivePacket(fiber);
+    Packet packet = current_connection->receivePacket(std::move(async_callback));
 
     switch (packet.type)
     {
diff --git a/src/Client/MultiplexedConnections.h b/src/Client/MultiplexedConnections.h
index 42d86086ddf..46312ae339d 100644
--- a/src/Client/MultiplexedConnections.h
+++ b/src/Client/MultiplexedConnections.h
@@ -67,11 +67,9 @@ public:
     /// Without locking, because sendCancel() does not change the state of the replicas.
     bool hasActiveConnections() const { return active_connection_count > 0; }
 
-    Poco::Net::Socket & getSocket() { return current_connection->getSocket(); }
-
 private:
     /// Internal version of `receivePacket` function without locking.
-    Packet receivePacketUnlocked(Fiber * fiber = nullptr);
+    Packet receivePacketUnlocked(std::function<void(Poco::Net::Socket &)> async_callback = {});
 
     /// Internal version of `dumpAddresses` function without locking.
     std::string dumpAddressesUnlocked() const;
diff --git a/src/Common/Config/CMakeLists.txt b/src/Common/Config/CMakeLists.txt
index b7a7f8ecda4..a7914fb17ec 100644
--- a/src/Common/Config/CMakeLists.txt
+++ b/src/Common/Config/CMakeLists.txt
@@ -14,5 +14,4 @@ target_link_libraries(clickhouse_common_config
         Poco::XML
     PRIVATE
         string_utils
-        _boost_context
 )
diff --git a/src/Common/ZooKeeper/CMakeLists.txt b/src/Common/ZooKeeper/CMakeLists.txt
index 7012f263d00..ef32d9266c0 100644
--- a/src/Common/ZooKeeper/CMakeLists.txt
+++ b/src/Common/ZooKeeper/CMakeLists.txt
@@ -4,7 +4,7 @@ add_headers_and_sources(clickhouse_common_zookeeper .)
 
 add_library(clickhouse_common_zookeeper ${clickhouse_common_zookeeper_headers} ${clickhouse_common_zookeeper_sources})
 
-target_link_libraries (clickhouse_common_zookeeper PUBLIC clickhouse_common_io common PRIVATE string_utils _boost_context)
+target_link_libraries (clickhouse_common_zookeeper PUBLIC clickhouse_common_io common PRIVATE string_utils)
 
 if (ENABLE_TESTS)
     add_subdirectory (tests)
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index 9765367b647..de215d46eb3 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -29,10 +29,10 @@ public:
     FiberStack stack;
     boost::context::fiber fiber;
     std::mutex fiber_lock;
-    // std::unique_lock<std::mutex> * connection_lock;
 
     Poco::Timespan receive_timeout;
     MultiplexedConnections & connections;
+    Poco::Net::Socket * last_used_socket = nullptr;
 
     TimerDescriptor timer{CLOCK_MONOTONIC, 0};
     int socket_fd = -1;
@@ -66,8 +66,7 @@ public:
                 throwFromErrno("Cannot add timer descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
         }
 
-        auto lock = std::make_unique<std::unique_lock<std::mutex>>(connections.cancel_mutex, std::defer_lock);
-        auto routine = Routine{connections, *this, std::move(lock)};
+        auto routine = Routine{connections, *this};
         // connection_lock = routine.connection_lock.get();
         fiber = boost::context::fiber(std::allocator_arg_t(), stack, std::move(routine));
     }
@@ -104,7 +103,8 @@ public:
         }
         catch (DB::Exception & e)
         {
-            e.addMessage(" while reading from socket ({})", connections.getSocket().peerAddress().toString());
+            if (last_used_socket)
+                e.addMessage(" while reading from socket ({})", last_used_socket->peerAddress().toString());
             throw;
         }
     }
@@ -165,32 +165,12 @@ public:
             if (!fiber)
                 return false;
 
-//            if (!connection_lock->owns_lock())
-//                connection_lock->lock();
-
             fiber = std::move(fiber).resume();
-
-//            if (!is_read_in_progress)
-//                connection_lock->unlock();
         }
 
         if (exception)
             std::rethrow_exception(std::move(exception));
 
-        if (is_read_in_progress)
-        {
-            auto & socket = connections.getSocket();
-            try
-            {
-                setSocket(socket);
-            }
-            catch (DB::Exception & e)
-            {
-                e.addMessage(" while reading from socket ({})", socket.peerAddress().toString());
-                throw;
-            }
-        }
-
         return true;
     }
 
@@ -221,7 +201,29 @@ public:
     {
         MultiplexedConnections & connections;
         Self & read_context;
-        std::unique_ptr<std::unique_lock<std::mutex>> connection_lock;
+
+        struct ReadCallback
+        {
+            Self & read_context;
+            Fiber & fiber;
+
+            void operator()(Poco::Net::Socket & socket)
+            {
+                try
+                {
+                    read_context.setSocket(socket);
+                }
+                catch (DB::Exception & e)
+                {
+                    e.addMessage(" while reading from socket ({})", socket.peerAddress().toString());
+                    throw;
+                }
+
+                read_context.is_read_in_progress = true;
+                fiber = std::move(fiber).resume();
+                read_context.is_read_in_progress = false;
+            }
+        };
 
         Fiber operator()(Fiber && sink) const
         {
@@ -229,10 +231,7 @@ public:
             {
                 while (true)
                 {
-                    read_context.is_read_in_progress = true;
-                    read_context.packet = connections.receivePacketUnlocked(&sink);
-                    read_context.is_read_in_progress = false;
-
+                    read_context.packet = connections.receivePacketUnlocked(ReadCallback{read_context, sink});
                     sink = std::move(sink).resume();
                 }
             }
diff --git a/src/IO/ReadBufferFromPocoSocket.cpp b/src/IO/ReadBufferFromPocoSocket.cpp
index 7498fb869e0..2c13446e693 100644
--- a/src/IO/ReadBufferFromPocoSocket.cpp
+++ b/src/IO/ReadBufferFromPocoSocket.cpp
@@ -29,7 +29,7 @@ bool ReadBufferFromPocoSocket::nextImpl()
     Stopwatch watch;
 
     int flags = 0;
-    if (fiber)
+    if (async_callback)
         flags |= MSG_DONTWAIT;
 
     /// Add more details to exceptions.
@@ -37,14 +37,12 @@ bool ReadBufferFromPocoSocket::nextImpl()
     {
         bytes_read = socket.impl()->receiveBytes(internal_buffer.begin(), internal_buffer.size(), flags);
 
-        /// If fiber is specified, and read is blocking, run fiber and try again later.
+        /// If async_callback is specified, and read is blocking, run async_callback and try again later.
         /// It is expected that file descriptor may be polled externally.
         /// Note that receive timeout is not checked here. External code should check it while polling.
-        while (bytes_read < 0 && fiber && errno == EAGAIN)
+        while (bytes_read < 0 && async_callback && errno == EAGAIN)
         {
-            //fiber->fd = socket.impl()->sockfd();
-            //fiber->timeout = socket.impl()->getReceiveTimeout();
-            *fiber = std::move(*fiber).resume();
+            async_callback(socket);
             bytes_read = socket.impl()->receiveBytes(internal_buffer.begin(), internal_buffer.size(), flags);
         }
     }
diff --git a/src/IO/ReadBufferFromPocoSocket.h b/src/IO/ReadBufferFromPocoSocket.h
index 2311d20cced..8064cd39246 100644
--- a/src/IO/ReadBufferFromPocoSocket.h
+++ b/src/IO/ReadBufferFromPocoSocket.h
@@ -4,7 +4,6 @@
 
 #include <IO/ReadBuffer.h>
 #include <IO/BufferWithOwnMemory.h>
-#include <Common/Fiber.h>
 
 namespace DB
 {
@@ -29,11 +28,10 @@ public:
 
     bool poll(size_t timeout_microseconds);
 
-    void setFiber(Fiber * fiber_) { fiber = fiber_; }
-    Poco::Net::Socket & getSocket() { return socket; }
+    void setAsyncCallback(std::function<void(Poco::Net::Socket &)> async_callback_) { async_callback = std::move(async_callback_); }
 
 private:
-    Fiber * fiber = nullptr;
+    std::function<void(Poco::Net::Socket &)> async_callback;
 };
 
 }

From fa1f7f40f5875b6e538c9779ae5f4b0258c44c46 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:32:44 +0300
Subject: [PATCH 177/504] another try

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 07152a16eb8..2864c2f9db5 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1025,14 +1025,14 @@ public:
                 else if constexpr(left_is_decimal) // BUG precision loss
                     left_scale = DecimalUtils::convertTo<FieldType>(left.getScaleMultiplier(), left.getScale());
                 else
-                    left_scale = 0.0; //won't be used, just to silence the warning
+                    left_scale = 1;
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
                 else if constexpr(right_is_decimal) //BUG precision loss
                     right_scale = DecimalUtils::convertTo<FieldType>(right.getScaleMultiplier(), right.getScale());
                 else
-                    right_scale = 0.0; //same
+                    right_scale = 1; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)

From 440212d01796fef91ff20ed0d518d55c758c7c7c Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 18 Dec 2020 17:34:22 +0300
Subject: [PATCH 178/504] fix the reference

---
 tests/queries/0_stateless/01591_window_functions.reference | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 32574813f1e..67d37653965 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -1,3 +1,7 @@
+set allow_experimental_window_functions = 1;
+
+-- just something basic
+
 select number, count() over (partition by intDiv(number, 3) order by number) from numbers(10);
 
 -- proper calculation across blocks

From 3c01cfe630bb24a24d91583d7137d57be49cb5b4 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:34:51 +0300
Subject: [PATCH 179/504] no scale added

---
 src/Functions/FunctionBinaryArithmetic.h | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 2864c2f9db5..4222244c08b 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1022,17 +1022,13 @@ public:
                     left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
                     left_scale = type.scaleFactorFor(left, is_multiply);
-                else if constexpr(left_is_decimal) // BUG precision loss
-                    left_scale = DecimalUtils::convertTo<FieldType>(left.getScaleMultiplier(), left.getScale());
                 else
-                    left_scale = 1;
+                    left_scale = 1; //won't be used
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
-                else if constexpr(right_is_decimal) //BUG precision loss
-                    right_scale = DecimalUtils::convertTo<FieldType>(right.getScaleMultiplier(), right.getScale());
                 else
-                    right_scale = 1; //same
+                    right_scale = 1; //w
 
                 /// non-vector result
                 if (col_left_const && col_right_const)

From 58ef4c34c070c08a7f9b6cc5fd337b91781f6141 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:44:24 +0300
Subject: [PATCH 180/504] prep for debug version

---
 src/Functions/FunctionBinaryArithmetic.h | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 4222244c08b..b46440a027f 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1022,13 +1022,17 @@ public:
                     left_scale = right.getScaleMultiplier();
                 else if constexpr (result_is_decimal)
                     left_scale = type.scaleFactorFor(left, is_multiply);
+                else if constexpr(left_is_decimal) // BUG precision loss
+                    left_scale = left.getScale();
                 else
-                    left_scale = 1; //won't be used
+                    left_scale = 1; //won't be used, just to silence the warning
 
                 if constexpr (result_is_decimal)
                     right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+                else if constexpr(right_is_decimal)
+                    right_scale = right.getScale();
                 else
-                    right_scale = 1; //w
+                    right_scale = 1; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)

From cd598034cd7d83466e1b336d548db187b5be99f8 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 18 Dec 2020 17:44:31 +0300
Subject: [PATCH 181/504] Fix add tests

---
 .../MergeTreeDataPartWriterCompact.cpp        | 19 ++++++-------
 .../MergeTreeDataPartWriterOnDisk.cpp         |  4 +--
 .../MergeTree/MergeTreeDataPartWriterOnDisk.h | 17 ++++-------
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 26 ++++++++---------
 .../MergeTree/MergeTreeDataPartWriterWide.h   |  7 +++++
 ...aptive_granularity_block_borders.reference |  2 ++
 ...605_adaptive_granularity_block_borders.sql | 28 +++++++++++++++++++
 7 files changed, 66 insertions(+), 37 deletions(-)
 create mode 100644 tests/queries/0_stateless/01605_adaptive_granularity_block_borders.reference
 create mode 100644 tests/queries/0_stateless/01605_adaptive_granularity_block_borders.sql

diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
index 389d2950f65..d5e7009efd6 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
@@ -93,12 +93,12 @@ Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity,
 
         result.emplace_back(Granule{
             .start_row = current_row,
-            .granularity_rows = expected_rows_in_mark,
-            .block_rows = std::min(rows_left_in_block, expected_rows_in_mark),
+            .rows_to_write = std::min(rows_left_in_block, expected_rows_in_mark),
             .mark_number = current_mark,
-            .mark_on_start = true
+            .mark_on_start = true,
+            .is_complete = (rows_left_in_block >= expected_rows_in_mark)
         });
-        current_row += expected_rows_in_mark;
+        current_row += result.back().rows_to_write;
         current_mark++;
     }
 
@@ -173,8 +173,7 @@ void MergeTreeDataPartWriterCompact::writeDataBlock(const Block & block, const G
 {
     for (const auto & granule : granules)
     {
-        if (granule.granularity_rows)
-            data_written = true;
+        data_written = true;
 
         auto name_and_type = columns_list.begin();
         for (size_t i = 0; i < columns_list.size(); ++i, ++name_and_type)
@@ -206,13 +205,13 @@ void MergeTreeDataPartWriterCompact::writeDataBlock(const Block & block, const G
             writeIntBinary(plain_hashing.count(), marks);
             writeIntBinary(UInt64(0), marks);
 
-            writeColumnSingleGranule(block.getByName(name_and_type->name), stream_getter, granule.start_row, granule.granularity_rows);
+            writeColumnSingleGranule(block.getByName(name_and_type->name), stream_getter, granule.start_row, granule.rows_to_write);
 
             /// Each type always have at least one substream
             prev_stream->hashing_buf.next(); //-V522
         }
 
-        writeIntBinary(granule.block_rows, marks);
+        writeIntBinary(granule.rows_to_write, marks);
     }
 }
 
@@ -222,11 +221,11 @@ void MergeTreeDataPartWriterCompact::finishDataSerialization(IMergeTreeDataPart:
     {
         auto block = header.cloneWithColumns(columns_buffer.releaseColumns());
         auto granules_to_write = getGranulesToWrite(index_granularity, block.rows(), getCurrentMark(), /* last_block = */ true);
-        if (!granules_to_write.back().isCompleted())
+        if (!granules_to_write.back().is_complete)
         {
             /// Correct last mark as it should contain exact amount of rows.
             index_granularity.popMark();
-            index_granularity.appendMark(granules_to_write.back().block_rows);
+            index_granularity.appendMark(granules_to_write.back().rows_to_write);
         }
         writeDataBlockPrimaryIndexAndSkipIndices(block, granules_to_write);
     }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
index 150bdad55f2..8e6ffe9ee68 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
@@ -240,8 +240,8 @@ void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block
             }
 
             size_t pos = granule.start_row;
-            skip_indices_aggregators[i]->update(skip_indexes_block, &pos, granule.granularity_rows);
-            if (granule.isCompleted())
+            skip_indices_aggregators[i]->update(skip_indexes_block, &pos, granule.rows_to_write);
+            if (granule.is_complete)
                 ++skip_index_accumulated_marks[i];
         }
     }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
index bc09e0e61e1..a7b84c95e0a 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
@@ -20,25 +20,18 @@ struct Granule
 {
     /// Start row in block for granule
     size_t start_row;
-    /// Amount of rows which granule have to contain according to index
-    /// granularity.
-    /// NOTE: Sometimes it's not equal to actually written rows, for example
-    /// for the last granule if it's smaller than computed granularity.
-    size_t granularity_rows;
     /// Amount of rows from block which have to be written to disk from start_row
-    size_t block_rows;
+    size_t rows_to_write;
     /// Global mark number in the list of all marks (index_granularity) for this part
     size_t mark_number;
     /// Should writer write mark for the first of this granule to disk.
     /// NOTE: Sometimes we don't write mark for the start row, because
     /// this granule can be continuation of the previous one.
     bool mark_on_start;
-
-    /// Is this granule contain amout of rows equal to the value in index granularity
-    bool isCompleted() const
-    {
-        return granularity_rows == block_rows;
-    }
+    /// if true: When this granule will be written to disk all rows for corresponding mark will
+    /// be wrtten. It doesn't mean that rows_to_write == index_granularity.getMarkRows(mark_number),
+    /// We may have a lot of small blocks between two marks and this may be the last one.
+    bool is_complete;
 };
 
 /// Multiple granules to write for concrete block.
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index f9b42e7c320..ca0e4e014a5 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -33,12 +33,12 @@ Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity,
         size_t rows_left_in_block = block_rows - current_row;
         result.emplace_back(Granule{
             .start_row = current_row,
-            .granularity_rows = rows_left_in_last_mark,
-            .block_rows = std::min(rows_left_in_block, rows_left_in_last_mark),
+            .rows_to_write = std::min(rows_left_in_block, rows_left_in_last_mark),
             .mark_number = current_mark,
             .mark_on_start = false, /// Don't mark this granule because we have already marked it
+            .is_complete = (rows_left_in_block >= rows_left_in_last_mark),
         });
-        current_row += rows_left_in_last_mark;
+        current_row += result.back().rows_to_write;
         current_mark++;
     }
 
@@ -51,12 +51,12 @@ Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity,
         /// save incomplete granule
         result.emplace_back(Granule{
             .start_row = current_row,
-            .granularity_rows = expected_rows_in_mark,
-            .block_rows = std::min(rows_left_in_block, expected_rows_in_mark),
+            .rows_to_write = std::min(rows_left_in_block, expected_rows_in_mark),
             .mark_number = current_mark,
             .mark_on_start = true,
+            .is_complete = (rows_left_in_block >= expected_rows_in_mark),
         });
-        current_row += expected_rows_in_mark;
+        current_row += result.back().rows_to_write;
         current_mark++;
     }
 
@@ -136,11 +136,12 @@ IDataType::OutputStreamGetter MergeTreeDataPartWriterWide::createStreamGetter(
     };
 }
 
+
 void MergeTreeDataPartWriterWide::shiftCurrentMark(const Granules & granules_written)
 {
     auto last_granule = granules_written.back();
     /// If we didn't finished last granule than we will continue to write it from new block
-    if (!last_granule.isCompleted())
+    if (!last_granule.is_complete)
     {
         /// Shift forward except last granule
         setCurrentMark(getCurrentMark() + granules_written.size() - 1);
@@ -148,9 +149,9 @@ void MergeTreeDataPartWriterWide::shiftCurrentMark(const Granules & granules_wri
         /// We wrote whole block in the same granule, but didn't finished it.
         /// So add written rows to rows written in last_mark
         if (still_in_the_same_granule)
-            rows_written_in_last_mark += last_granule.block_rows;
+            rows_written_in_last_mark += last_granule.rows_to_write;
         else
-            rows_written_in_last_mark = last_granule.block_rows;
+            rows_written_in_last_mark = last_granule.rows_to_write;
     }
     else
     {
@@ -288,7 +289,7 @@ void MergeTreeDataPartWriterWide::writeSingleGranule(
     IDataType::SerializeBinaryBulkSettings & serialize_settings,
     const Granule & granule)
 {
-    type.serializeBinaryBulkWithMultipleStreams(column, granule.start_row, granule.granularity_rows, serialize_settings, serialization_state);
+    type.serializeBinaryBulkWithMultipleStreams(column, granule.start_row, granule.rows_to_write, serialize_settings, serialization_state);
 
     /// So that instead of the marks pointing to the end of the compressed block, there were marks pointing to the beginning of the next one.
     type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
@@ -333,8 +334,7 @@ void MergeTreeDataPartWriterWide::writeColumn(
 
     for (const auto & granule : granules)
     {
-        if (granule.granularity_rows > 0)
-            data_written = true;
+        data_written = true;
 
         if (granule.mark_on_start)
         {
@@ -353,7 +353,7 @@ void MergeTreeDataPartWriterWide::writeColumn(
            granule
         );
 
-        if (granule.isCompleted())
+        if (granule.is_complete)
         {
             auto it = last_non_written_marks.find(name);
             if (it == last_non_written_marks.end())
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
index afaf5958338..3fc3e3bfa3d 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
@@ -99,6 +99,11 @@ private:
     /// in our index_granularity array.
     void shiftCurrentMark(const Granules & granules_written);
 
+    /// Change rows in the last mark in index_granularity to rows_written_in_last_mark.
+    /// Flush all marks from last_non_written_marks to disk and increment current mark.
+    ///
+    /// This function used when blocks change granularity drastically and we have unfinished mark.
+    /// Also useful to have exact amount of rows in last (non-final) mark.
     void adjustLastMarkAndFlushToDisk();
 
     IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns) const;
@@ -110,6 +115,8 @@ private:
 
     using ColumnStreams = std::map<String, StreamPtr>;
     ColumnStreams column_streams;
+    /// Non written marks to disk (for each column). Waiting until all rows for
+    /// this marks will be written to disk.
     using MarksForColumns = std::unordered_map<String, StreamsWithMarks>;
     MarksForColumns last_non_written_marks;
 
diff --git a/tests/queries/0_stateless/01605_adaptive_granularity_block_borders.reference b/tests/queries/0_stateless/01605_adaptive_granularity_block_borders.reference
new file mode 100644
index 00000000000..81c7e6e4df0
--- /dev/null
+++ b/tests/queries/0_stateless/01605_adaptive_granularity_block_borders.reference
@@ -0,0 +1,2 @@
+849
+102400
diff --git a/tests/queries/0_stateless/01605_adaptive_granularity_block_borders.sql b/tests/queries/0_stateless/01605_adaptive_granularity_block_borders.sql
new file mode 100644
index 00000000000..a73045f5a6f
--- /dev/null
+++ b/tests/queries/0_stateless/01605_adaptive_granularity_block_borders.sql
@@ -0,0 +1,28 @@
+DROP TABLE IF EXISTS adaptive_table;
+
+--- If granularity of consequent blocks differs a lot, then adaptive
+--- granularity will adjust amout of marks correctly. Data for test empirically
+--- derived, it's quite hard to get good parameters.
+
+CREATE TABLE adaptive_table(
+    key UInt64,
+    value String
+) ENGINE MergeTree()
+ORDER BY key
+SETTINGS index_granularity_bytes=1048576, min_bytes_for_wide_part = 0, enable_vertical_merge_algorithm = 0;
+
+SET max_block_size=900;
+
+-- There are about 900 marks for our settings.
+INSERT INTO adaptive_table SELECT number, if(number > 700, randomPrintableASCII(102400), randomPrintableASCII(1)) FROM numbers(10000);
+
+OPTIMIZE TABLE adaptive_table FINAL;
+
+SELECT marks FROM system.parts WHERE table = 'adaptive_table' and database=currentDatabase() and active;
+
+-- If we have computed granularity incorrectly than we will exceed this limit.
+SET max_memory_usage='30M';
+
+SELECT max(length(value)) FROM adaptive_table;
+
+DROP TABLE IF EXISTS adaptive_table;

From 6b956a2a0dbcf0b80ef976116006d14db60abebe Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:50:16 +0300
Subject: [PATCH 182/504] possible fix for non-const mult

---
 src/Functions/FunctionBinaryArithmetic.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index b46440a027f..30799085acc 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -219,7 +219,8 @@ struct DecimalBinaryOperation
                                        NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
     {
         size_t size = a.size();
-        if constexpr (is_plus_minus_compare)
+
+        if constexpr (is_plus_minus_compare || is_multiply)
         {
             if (scale_a != 1)
             {

From 7611f8d15b13130b2374a5f0025ed5907e77c29e Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 17:54:48 +0300
Subject: [PATCH 183/504] compile fix

---
 src/Functions/FunctionBinaryArithmetic.h | 57 ++++++++++++------------
 1 file changed, 29 insertions(+), 28 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 30799085acc..efee38e5926 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -204,6 +204,8 @@ struct DecimalBinaryOperation
     static constexpr bool is_plus_minus_compare = is_plus_minus || is_compare;
     static constexpr bool can_overflow = is_plus_minus || is_multiply;
 
+    static constexpr bool is_plus_minus_compare_mult = is_plus_minus || is_compare || is_multiply;
+
     using ResultType = ResultType_;
     using NativeResultType = typename NativeType<ResultType>::Type;
     using Op = std::conditional_t<is_float_division,
@@ -220,7 +222,7 @@ struct DecimalBinaryOperation
     {
         size_t size = a.size();
 
-        if constexpr (is_plus_minus_compare || is_multiply)
+        if constexpr (is_plus_minus_compare_mult)
         {
             if (scale_a != 1)
             {
@@ -351,37 +353,36 @@ private:
     template <bool scale_left>
     static NO_SANITIZE_UNDEFINED NativeResultType applyScaled(NativeResultType a, NativeResultType b, NativeResultType scale)
     {
-        if constexpr (is_plus_minus_compare)
+        static_assert(is_plus_minus_compare_mult);
+
+        NativeResultType res;
+
+        if constexpr (check_overflow)
         {
-            NativeResultType res;
-
-            if constexpr (check_overflow)
-            {
-                bool overflow = false;
-                if constexpr (scale_left)
-                    overflow |= common::mulOverflow(a, scale, a);
-                else
-                    overflow |= common::mulOverflow(b, scale, b);
-
-                if constexpr (can_overflow)
-                    overflow |= Op::template apply<NativeResultType>(a, b, res);
-                else
-                    res = Op::template apply<NativeResultType>(a, b);
-
-                if (overflow)
-                    throw Exception("Decimal math overflow", ErrorCodes::DECIMAL_OVERFLOW);
-            }
+            bool overflow = false;
+            if constexpr (scale_left)
+                overflow |= common::mulOverflow(a, scale, a);
             else
-            {
-                if constexpr (scale_left)
-                    a *= scale;
-                else
-                    b *= scale;
-                res = Op::template apply<NativeResultType>(a, b);
-            }
+                overflow |= common::mulOverflow(b, scale, b);
 
-            return res;
+            if constexpr (can_overflow)
+                overflow |= Op::template apply<NativeResultType>(a, b, res);
+            else
+                res = Op::template apply<NativeResultType>(a, b);
+
+            if (overflow)
+                throw Exception("Decimal math overflow", ErrorCodes::DECIMAL_OVERFLOW);
         }
+        else
+        {
+            if constexpr (scale_left)
+                a *= scale;
+            else
+                b *= scale;
+            res = Op::template apply<NativeResultType>(a, b);
+        }
+
+        return res;
     }
 
     template <bool is_decimal_a>

From 9d24cafccdaadf9803d02183a5af4a0856b407aa Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:06:24 +0300
Subject: [PATCH 184/504] added vector mult branch

---
 src/Functions/FunctionBinaryArithmetic.h | 24 +++++++++++++++++++-----
 1 file changed, 19 insertions(+), 5 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index efee38e5926..c79c6256172 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -186,8 +186,7 @@ inline constexpr const auto & undec(const T & x)
 }
 
 /// Binary operations for Decimals need scale args
-/// +|- scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
-/// *   no agrs scale. ScaleR = Scale1 + Scale2;
+/// +|-* scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
 /// /   first arg scale. ScaleR = Scale1 (scale_a = DecimalType<B>::getScale()).
 template <template <typename, typename> typename Operation, typename ResultType_, bool check_overflow = true>
 struct DecimalBinaryOperation
@@ -222,7 +221,7 @@ struct DecimalBinaryOperation
     {
         size_t size = a.size();
 
-        if constexpr (is_plus_minus_compare_mult)
+        if constexpr (is_plus_minus_compare)
         {
             if (scale_a != 1)
             {
@@ -237,6 +236,21 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b))
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    c[i] = apply(scale_a * undec(a[i]), undec(b[i]));
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    c[i] = apply(undec(a[i]), undec(b[i]) * scale_b);
+                return;
+            }
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -353,13 +367,13 @@ private:
     template <bool scale_left>
     static NO_SANITIZE_UNDEFINED NativeResultType applyScaled(NativeResultType a, NativeResultType b, NativeResultType scale)
     {
-        static_assert(is_plus_minus_compare_mult);
-
+        static_assert(is_plus_minus_compare);
         NativeResultType res;
 
         if constexpr (check_overflow)
         {
             bool overflow = false;
+
             if constexpr (scale_left)
                 overflow |= common::mulOverflow(a, scale, a);
             else

From 855c4bc7d72212f07dc54715037cf9a5d70a03a4 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 18 Dec 2020 18:12:31 +0300
Subject: [PATCH 185/504] Added comments.

---
 src/Common/TimerDescriptor.h                         |  1 +
 src/DataStreams/RemoteQueryExecutorReadContext.h     |  3 +++
 src/Processors/Executors/PipelineExecutor.h          |  4 ++--
 .../{AsyncTaskQueue.cpp => PollingQueue.cpp}         | 12 ++++++------
 .../Executors/{AsyncTaskQueue.h => PollingQueue.h}   | 11 ++++++-----
 src/Processors/ya.make                               |  2 +-
 6 files changed, 19 insertions(+), 14 deletions(-)
 rename src/Processors/Executors/{AsyncTaskQueue.cpp => PollingQueue.cpp} (89%)
 rename src/Processors/Executors/{AsyncTaskQueue.h => PollingQueue.h} (85%)

diff --git a/src/Common/TimerDescriptor.h b/src/Common/TimerDescriptor.h
index 0e24c78696c..ddb8f2a1367 100644
--- a/src/Common/TimerDescriptor.h
+++ b/src/Common/TimerDescriptor.h
@@ -5,6 +5,7 @@
 namespace DB
 {
 
+/// Wrapper over timerfd.
 class TimerDescriptor
 {
 private:
diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index de215d46eb3..fc81132ea18 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -237,6 +237,9 @@ public:
             }
             catch (const boost::context::detail::forced_unwind &)
             {
+                /// This exception is thrown by fiber implementation in case if fiber is being deleted but hasn't exited
+                /// It should not be caught or it will segfault.
+                /// Other exceptions must be caught
                 throw;
             }
             catch (...)
diff --git a/src/Processors/Executors/PipelineExecutor.h b/src/Processors/Executors/PipelineExecutor.h
index 16ace7b40bb..213446ad43f 100644
--- a/src/Processors/Executors/PipelineExecutor.h
+++ b/src/Processors/Executors/PipelineExecutor.h
@@ -1,7 +1,7 @@
 #pragma once
 
 #include <Processors/IProcessor.h>
-#include <Processors/Executors/AsyncTaskQueue.h>
+#include <Processors/Executors/PollingQueue.h>
 #include <Processors/Executors/ThreadsQueue.h>
 #include <Processors/Executors/TasksQueue.h>
 #include <Processors/Executors/ExecutingGraph.h>
@@ -61,7 +61,7 @@ private:
     /// Queue which stores tasks where processors returned Async status after prepare.
     /// If multiple threads are using, main thread will wait for async tasks.
     /// For single thread, will wait for async tasks only when task_queue is empty.
-    AsyncTaskQueue async_task_queue;
+    PollingQueue async_task_queue;
     size_t num_waiting_async_tasks = 0;
 
     ThreadsQueue threads_queue;
diff --git a/src/Processors/Executors/AsyncTaskQueue.cpp b/src/Processors/Executors/PollingQueue.cpp
similarity index 89%
rename from src/Processors/Executors/AsyncTaskQueue.cpp
rename to src/Processors/Executors/PollingQueue.cpp
index 621f526d80b..7383824a592 100644
--- a/src/Processors/Executors/AsyncTaskQueue.cpp
+++ b/src/Processors/Executors/PollingQueue.cpp
@@ -1,4 +1,4 @@
-#include <Processors/Executors/AsyncTaskQueue.h>
+#include <Processors/Executors/PollingQueue.h>
 
 #if defined(OS_LINUX)
 
@@ -18,7 +18,7 @@ namespace ErrorCodes
 }
 
 
-AsyncTaskQueue::AsyncTaskQueue()
+PollingQueue::PollingQueue()
 {
     epoll_fd = epoll_create(1);
     if (-1 == epoll_fd)
@@ -35,14 +35,14 @@ AsyncTaskQueue::AsyncTaskQueue()
         throwFromErrno("Cannot add pipe descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
 }
 
-AsyncTaskQueue::~AsyncTaskQueue()
+PollingQueue::~PollingQueue()
 {
     close(epoll_fd);
     close(pipe_fd[0]);
     close(pipe_fd[1]);
 }
 
-void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
+void PollingQueue::addTask(size_t thread_number, void * data, int fd)
 {
     std::uintptr_t key = reinterpret_cast<uintptr_t>(data);
     if (tasks.count(key))
@@ -58,7 +58,7 @@ void AsyncTaskQueue::addTask(size_t thread_number, void * data, int fd)
         throwFromErrno("Cannot add socket descriptor to epoll", ErrorCodes::CANNOT_OPEN_FILE);
 }
 
-AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & lock)
+PollingQueue::TaskData PollingQueue::wait(std::unique_lock<std::mutex> & lock)
 {
     if (is_finished)
         return {};
@@ -95,7 +95,7 @@ AsyncTaskQueue::TaskData AsyncTaskQueue::wait(std::unique_lock<std::mutex> & loc
     return res;
 }
 
-void AsyncTaskQueue::finish()
+void PollingQueue::finish()
 {
     is_finished = true;
     tasks.clear();
diff --git a/src/Processors/Executors/AsyncTaskQueue.h b/src/Processors/Executors/PollingQueue.h
similarity index 85%
rename from src/Processors/Executors/AsyncTaskQueue.h
rename to src/Processors/Executors/PollingQueue.h
index 1a4bdec5dd3..5588bb87436 100644
--- a/src/Processors/Executors/AsyncTaskQueue.h
+++ b/src/Processors/Executors/PollingQueue.h
@@ -9,7 +9,9 @@ namespace DB
 {
 
 #if defined(OS_LINUX)
-class AsyncTaskQueue
+
+/// This queue is used to poll descriptors. Generally, just a wrapper over epoll.
+class PollingQueue
 {
 public:
     struct TaskData
@@ -19,7 +21,6 @@ public:
         void * data = nullptr;
         int fd = -1;
 
-
         explicit operator bool() const { return data; }
     };
 
@@ -30,8 +31,8 @@ private:
     std::unordered_map<std::uintptr_t, TaskData> tasks;
 
 public:
-    AsyncTaskQueue();
-    ~AsyncTaskQueue();
+    PollingQueue();
+    ~PollingQueue();
 
     size_t size() const { return tasks.size(); }
     bool empty() const { return tasks.empty(); }
@@ -41,7 +42,7 @@ public:
 
     /// Wait for any descriptor. If no descriptors in queue, blocks.
     /// Returns ptr which was inserted into queue or nullptr if finished was called.
-    /// Lock is used to wait on condvar.
+    /// Lock is unlocked during waiting.
     TaskData wait(std::unique_lock<std::mutex> & lock);
 
     /// Interrupt waiting.
diff --git a/src/Processors/ya.make b/src/Processors/ya.make
index bb99cabb1dd..0d573328bfa 100644
--- a/src/Processors/ya.make
+++ b/src/Processors/ya.make
@@ -14,7 +14,7 @@ SRCS(
     Chunk.cpp
     ConcatProcessor.cpp
     DelayedPortsProcessor.cpp
-    Executors/AsyncTaskQueue.cpp
+    Executors/PollingQueue.cpp
     Executors/ExecutingGraph.cpp
     Executors/PipelineExecutingBlockInputStream.cpp
     Executors/PipelineExecutor.cpp

From d22d045a59a8e2db0b7bff0f4bd9f11ac48aa48f Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:12:31 +0300
Subject: [PATCH 186/504] fixed the impl

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index c79c6256172..e8391463e05 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -241,13 +241,13 @@ struct DecimalBinaryOperation
             if (scale_a != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = apply(scale_a * undec(a[i]), undec(b[i]));
+                    c[i] = scale_a * apply(undec(a[i]), undec(b[i]));
                 return;
             }
             else if (scale_b != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = apply(undec(a[i]), undec(b[i]) * scale_b);
+                    c[i] = scale_b * apply(undec(a[i]), undec(b[i]));
                 return;
             }
         }

From 1c885274ff13afd13a90f3f2349fe7359a330309 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 18 Dec 2020 18:13:30 +0300
Subject: [PATCH 187/504] Fix yamake.

---
 src/Processors/ya.make | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Processors/ya.make b/src/Processors/ya.make
index 0d573328bfa..2d437cb09bd 100644
--- a/src/Processors/ya.make
+++ b/src/Processors/ya.make
@@ -14,10 +14,10 @@ SRCS(
     Chunk.cpp
     ConcatProcessor.cpp
     DelayedPortsProcessor.cpp
-    Executors/PollingQueue.cpp
     Executors/ExecutingGraph.cpp
     Executors/PipelineExecutingBlockInputStream.cpp
     Executors/PipelineExecutor.cpp
+    Executors/PollingQueue.cpp
     Executors/PullingAsyncPipelineExecutor.cpp
     Executors/PullingPipelineExecutor.cpp
     ForkProcessor.cpp

From 0dca8562373c495ba9170ecac3bf298df022927e Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:16:38 +0300
Subject: [PATCH 188/504] solution fix

---
 src/Functions/FunctionBinaryArithmetic.h | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index e8391463e05..76f9b497367 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -241,13 +241,15 @@ struct DecimalBinaryOperation
             if (scale_a != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = scale_a * apply(undec(a[i]), undec(b[i]));
+                    // BUG very bad solution
+                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<UInt64>(scale_a);
                 return;
             }
             else if (scale_b != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = scale_b * apply(undec(a[i]), undec(b[i]));
+                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<UInt64>(scale_a);
+
                 return;
             }
         }

From 6f1c0b118aea3ba96af3ba1a882c0fa4e6417316 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:17:03 +0300
Subject: [PATCH 189/504] typo

---
 src/Functions/FunctionBinaryArithmetic.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 76f9b497367..39543b83ef1 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -248,7 +248,7 @@ struct DecimalBinaryOperation
             else if (scale_b != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<UInt64>(scale_a);
+                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<UInt64>(scale_b);
 
                 return;
             }

From ed6d74b21837532773a3eaf870e3b6abb5547725 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:21:10 +0300
Subject: [PATCH 190/504] another bad solution

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 39543b83ef1..b2ea79cea61 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -242,13 +242,13 @@ struct DecimalBinaryOperation
             {
                 for (size_t i = 0; i < size; ++i)
                     // BUG very bad solution
-                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<UInt64>(scale_a);
+                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
                 return;
             }
             else if (scale_b != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<UInt64>(scale_b);
+                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
 
                 return;
             }

From fbaa8d3951f4f6056e92ab68adf6647dc2cc5caf Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:24:31 +0300
Subject: [PATCH 191/504] fix

---
 src/Functions/FunctionBinaryArithmetic.h | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index b2ea79cea61..f3904ae764e 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -236,19 +236,24 @@ struct DecimalBinaryOperation
                 return;
             }
         }
-        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b))
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
         {
             if (scale_a != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    // BUG very bad solution
-                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
                 return;
             }
             else if (scale_b != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
 
                 return;
             }

From a19934886028e6a8bf5320f15eea056c5fad3eb5 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:26:58 +0300
Subject: [PATCH 192/504] fix2

---
 src/Functions/FunctionBinaryArithmetic.h | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index f3904ae764e..e0745e65c55 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -292,6 +292,28 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)

From e2556659e04c705e370e5ac5c79e06b84d1a27a8 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:28:21 +0300
Subject: [PATCH 193/504] fix3

---
 src/Functions/FunctionBinaryArithmetic.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index e0745e65c55..276b2e11fb1 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -298,18 +298,18 @@ struct DecimalBinaryOperation
             {
                 for (size_t i = 0; i < size; ++i)
                     if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
                     else
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
                 return;
             }
             else if (scale_b != 1)
             {
                 for (size_t i = 0; i < size; ++i)
                     if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
                     else
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
 
                 return;
             }

From 985876b737d8564a2850e54cefa88063c6da2f2d Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:37:00 +0300
Subject: [PATCH 194/504] another fix

---
 src/Functions/FunctionBinaryArithmetic.h      | 23 +++++++++++++++++++
 .../01603_decimal_mult_float.reference        |  8 +++----
 .../0_stateless/01603_decimal_mult_float.sql  |  4 +++-
 3 files changed, 30 insertions(+), 5 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 276b2e11fb1..756a6c2e4d3 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -348,6 +348,29 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(a, undec(a[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(a, undec(a[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
+
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
index 08e0fc89922..c25beb42a27 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -3,10 +3,10 @@
 0.00012000000000000002
 150.16500000000002
 7.775900000000001
-56.622689999999995
-598.8376688440277
+56622.689999999995
+5.988376688440276
 2.9941884322906844
 1.641386318314034
 1.641386318314034
-1.6413863211679272
-1.6413863211679272
+1.6413863192433382
+1.6413863192433382
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
index 8de35f0b3f5..407fe3649ec 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.sql
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -14,7 +14,9 @@ INSERT INTO test01603(f64) SELECT 1 / (number + 1) FROM system.numbers LIMIT 100
 SELECT sum(d * 1.1) FROM test01603;
 SELECT sum(8.01 * d) FROM test01603;
 SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
-SELECT sum(toDecimal64(0.4, 2) * f32) FROM test01603;
+SELECT sum(toDecimal64(40, 2) * f32) FROM test01603;
+SELECT sum(f64 * toDecimal64(0.1, 2)) FROM test01603;
+SELECT sum(toDecimal64(0.3, 2) * f32) FROM test01603;
 
 SELECT sum(f64 * d) FROM test01603;
 SELECT sum(d * f64) FROM test01603;

From 0e235d35f80af6298c5373602c065969474dccfc Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:48:37 +0300
Subject: [PATCH 195/504] aa

---
 .gitmodules                                   |   2 +-
 CHANGELOG.md                                  | 123 ++++
 base/common/CMakeLists.txt                    |   1 +
 base/common/ReadlineLineReader.cpp            |   5 +-
 base/common/ReplxxLineReader.cpp              |  10 +-
 base/common/getMemoryAmount.cpp               |   3 +-
 base/common/getPageSize.cpp                   |   8 +
 base/common/getPageSize.h                     |   6 +
 base/common/setTerminalEcho.cpp               |   5 +-
 base/common/ya.make                           |   1 +
 cmake/sanitize.cmake                          |  21 +-
 contrib/boost                                 |   2 +-
 contrib/krb5                                  |   2 +-
 contrib/libc-headers                          |   2 +-
 contrib/libgsasl                              |   2 +-
 contrib/libhdfs3                              |   2 +-
 contrib/libhdfs3-cmake/CMakeLists.txt         |  17 +-
 contrib/librdkafka                            |   2 +-
 contrib/libunwind                             |   2 +-
 debian/control                                |   2 +-
 docker/images.json                            |  11 +-
 docker/packager/binary/build.sh               |   1 +
 docker/packager/packager                      |   4 +
 docker/server/README.md                       |  66 ++
 docker/test/base/Dockerfile                   |   2 +-
 .../integration/kerberized_hadoop/Dockerfile  |  18 +
 docker/test/integration/runner/Dockerfile     |   5 +-
 .../runner/compose/docker_compose_hdfs.yml    |   1 +
 .../docker_compose_kerberized_hdfs.yml        |  29 +
 .../zzz-perf-comparison-tweaks-config.xml     |   1 +
 docker/test/stateful_with_coverage/Dockerfile |   2 -
 .../table-engines/integrations/hdfs.md        |  89 +++
 docs/en/interfaces/cli.md                     |   2 +-
 docs/en/introduction/performance.md           |   2 +-
 .../system-tables/distribution_queue.md       |   2 +-
 .../system-tables/replication_queue.md        |  81 ++
 docs/en/operations/tips.md                    |  17 +
 .../aggregate-functions/combinators.md        |   4 +
 .../functions/array-functions.md              |  18 +
 .../functions/type-conversion-functions.md    |  57 ++
 docs/en/sql-reference/statements/system.md    |   2 +-
 docs/es/interfaces/cli.md                     |   2 +-
 docs/ja/interfaces/cli.md                     |   2 +-
 docs/ru/faq/general/index.md                  |  25 +
 .../faq/{general.md => general/mapreduce.md}  |  52 +-
 docs/ru/faq/index.md                          |  11 +
 docs/ru/faq/integration/file-export.md        |  37 +
 docs/ru/faq/integration/index.md              |  19 +
 docs/ru/faq/integration/oracle-odbc.md        |  15 +
 docs/ru/faq/operations/index.md               |  18 +
 docs/ru/faq/use-cases/index.md                |  14 +
 docs/ru/interfaces/cli.md                     |   2 +-
 .../system-tables/replication_queue.md        |  81 ++
 .../external-dicts-dict-sources.md            |   2 +-
 .../functions/date-time-functions.md          |  12 +
 docs/tr/interfaces/cli.md                     |   2 +-
 docs/zh/sql-reference/statements/misc.md      |  48 +-
 .../sql-reference/statements/select/limit.md  |  10 +-
 programs/client/Client.cpp                    |   2 +-
 programs/odbc-bridge/ColumnInfoHandler.cpp    |   2 +-
 programs/server/Server.cpp                    |  68 +-
 programs/server/config.xml                    | 189 +++--
 src/Access/ContextAccess.cpp                  |   9 +-
 src/Access/SettingsConstraints.cpp            |  19 +
 src/AggregateFunctions/AggregateFunctionAvg.h |   2 +-
 .../AggregateFunctionFactory.cpp              |  12 +-
 .../AggregateFunctionSimpleState.cpp          |  32 +
 .../AggregateFunctionSimpleState.h            |  77 ++
 .../registerAggregateFunctions.cpp            |   2 +
 src/AggregateFunctions/ya.make                |   1 +
 src/CMakeLists.txt                            |   8 +-
 src/Client/Connection.h                       |   5 +-
 src/Columns/ColumnMap.cpp                     | 241 ++++++
 src/Columns/ColumnMap.h                       |  92 +++
 src/Columns/ya.make                           |   1 +
 src/Common/Allocator.h                        |   5 +-
 src/Common/Arena.h                            |  10 +-
 src/Common/ArrayCache.h                       |   5 +-
 src/Common/Config/ConfigProcessor.cpp         |   4 +-
 src/Common/ErrorCodes.cpp                     |   1 +
 src/Common/FieldVisitors.cpp                  | 102 +++
 src/Common/FieldVisitors.h                    |  33 +
 src/Common/IFactoryWithAliases.h              |   5 +
 src/Common/MemoryStatisticsOS.cpp             |  15 +-
 src/Common/StringSearcher.h                   |   3 +-
 src/Common/ThreadProfileEvents.cpp            |  15 +-
 src/Common/ThreadStatus.cpp                   |  48 ++
 src/Common/UInt128.h                          | 127 ++--
 src/Common/ZooKeeper/TestKeeperStorage.cpp    |  15 +
 src/Common/config.h.in                        |   1 +
 src/Common/ya.make                            |   4 +-
 src/Common/ya.make.in                         |   4 +-
 src/Core/AccurateComparison.h                 |  29 +-
 src/Core/DecimalFunctions.h                   |  49 +-
 src/Core/Field.cpp                            | 378 +++-------
 src/Core/Field.h                              |  24 +-
 src/Core/Settings.cpp                         |   2 -
 src/Core/Settings.h                           |   7 +-
 src/Core/SortCursor.h                         |  39 +-
 src/Core/Types.h                              |   2 +
 .../MergingSortedBlockInputStream.cpp         |   2 +-
 src/DataStreams/RemoteQueryExecutor.h         |   3 +-
 .../DataTypeCustomSimpleAggregateFunction.cpp |  22 +-
 .../DataTypeCustomSimpleAggregateFunction.h   |   1 +
 src/DataTypes/DataTypeFactory.cpp             |   1 +
 src/DataTypes/DataTypeFactory.h               |   1 +
 src/DataTypes/DataTypeMap.cpp                 | 375 ++++++++++
 src/DataTypes/DataTypeMap.h                   | 103 +++
 src/DataTypes/DataTypesDecimal.h              | 134 +++-
 src/DataTypes/FieldToDataType.cpp             |  19 +
 src/DataTypes/FieldToDataType.h               |   1 +
 src/DataTypes/IDataType.h                     |   4 +
 src/DataTypes/ya.make                         |   1 +
 src/Databases/DatabaseAtomic.cpp              |  12 +-
 src/Databases/DatabaseAtomic.h                |   8 +-
 src/Databases/DatabaseFactory.cpp             |  33 +-
 src/Databases/DatabaseLazy.cpp                |   8 +
 src/Databases/DatabaseOnDisk.cpp              |  97 ++-
 src/Databases/DatabaseOnDisk.h                |  10 +
 src/Databases/DatabaseOrdinary.cpp            |  13 +
 src/Databases/DatabasesCommon.cpp             |  10 +-
 src/Databases/IDatabase.h                     |  13 +
 .../MySQL/DatabaseConnectionMySQL.cpp         |   7 +-
 src/Databases/MySQL/DatabaseConnectionMySQL.h |   2 +
 .../MySQL/DatabaseMaterializeMySQL.cpp        | 266 ++++---
 .../MySQL/DatabaseMaterializeMySQL.h          |  49 +-
 .../MySQL/DatabaseMaterializeTablesIterator.h |   6 +-
 .../MySQL/MaterializeMySQLSyncThread.cpp      |  14 +-
 src/Disks/S3/DiskS3.cpp                       |  17 +-
 src/Disks/S3/DiskS3.h                         |   4 +-
 src/Disks/S3/registerDiskS3.cpp               |   2 +-
 src/Functions/FunctionBase64Conversion.h      |   4 +-
 src/Functions/FunctionBinaryArithmetic.h      | 268 +++++--
 src/Functions/FunctionFactory.cpp             |  11 +-
 src/Functions/FunctionsConversion.cpp         |  11 +-
 src/Functions/FunctionsConversion.h           | 697 +++++++++++++-----
 src/Functions/array/arrayAggregation.cpp      | 318 ++++++++
 src/Functions/array/arrayElement.cpp          | 279 ++++++-
 src/Functions/array/arrayEnumerateRanked.h    |   4 +-
 src/Functions/array/arrayIntersect.cpp        |   8 +-
 src/Functions/array/arraySum.cpp              | 146 ----
 src/Functions/map.cpp                         | 140 ++++
 src/Functions/multiply.cpp                    |  11 +-
 src/Functions/registerFunctions.cpp           |   2 +
 .../registerFunctionsHigherOrder.cpp          |   4 +-
 src/Functions/toFixedString.h                 |  61 +-
 src/Functions/toModifiedJulianDay.cpp         |  11 +-
 src/Functions/trap.cpp                        |  11 +-
 src/Functions/ya.make                         |   3 +-
 src/IO/HDFSCommon.cpp                         |  62 --
 src/IO/HDFSCommon.h                           |  58 --
 src/IO/MMapReadBufferFromFileDescriptor.cpp   |   5 +-
 src/IO/S3Common.cpp                           |   8 +-
 src/IO/WriteBufferFromS3.cpp                  | 183 ++---
 src/IO/WriteBufferFromS3.h                    |  32 +-
 src/IO/WriteHelpers.h                         |  21 +-
 src/IO/readFloatText.h                        |  38 +-
 src/IO/tests/o_direct_and_dirty_pages.cpp     |  10 +-
 src/IO/ya.make                                |   2 +
 src/IO/ya.make.in                             |   2 +
 src/Interpreters/ActionsDAG.cpp               |   4 +-
 src/Interpreters/Context.cpp                  |   2 +
 src/Interpreters/DatabaseCatalog.cpp          |  37 +-
 .../IInterpreterUnionOrSelectQuery.h          |   2 +
 src/Interpreters/InterpreterCreateQuery.cpp   |  50 +-
 src/Interpreters/InterpreterDropQuery.cpp     |  48 +-
 src/Interpreters/InterpreterSelectQuery.cpp   |   7 +-
 .../InterpreterSelectWithUnionQuery.cpp       | 142 +++-
 .../InterpreterShowTablesQuery.cpp            |  20 +
 src/Interpreters/MergeJoin.cpp                |  22 +-
 src/Interpreters/Set.cpp                      |  21 +-
 src/Interpreters/castColumn.cpp               |  36 +-
 src/Interpreters/castColumn.h                 |   5 +-
 src/Interpreters/convertFieldToType.cpp       |   1 +
 src/Interpreters/loadMetadata.cpp             |   1 +
 src/Interpreters/tests/internal_iotop.cpp     |   4 +-
 src/Parsers/ASTDropQuery.cpp                  |   3 +
 src/Parsers/ASTDropQuery.h                    |   3 +
 src/Parsers/ASTFunction.cpp                   |  14 +
 src/Parsers/ASTShowTablesQuery.cpp            |   6 +
 src/Parsers/ASTShowTablesQuery.h              |   2 +
 src/Parsers/ExpressionElementParsers.cpp      |   6 +
 src/Parsers/ExpressionElementParsers.h        |  12 +
 src/Parsers/ExpressionListParsers.cpp         |   1 +
 src/Parsers/ParserCreateQuery.cpp             |  12 +-
 src/Parsers/ParserDropQuery.cpp               |  70 +-
 src/Parsers/ParserDropQuery.h                 |   2 +-
 src/Parsers/ParserExplainQuery.cpp            |   9 +-
 src/Parsers/ParserExplainQuery.h              |   4 +
 src/Parsers/ParserQuery.cpp                   |   2 +-
 src/Parsers/ParserQueryWithOutput.cpp         |   2 +-
 src/Parsers/ParserQueryWithOutput.h           |   3 +
 src/Parsers/ParserShowTablesQuery.cpp         |  25 +
 src/Parsers/tests/select_parser.cpp           |   2 +-
 src/Parsers/tests/select_parser_fuzzer.cpp    |   2 +-
 .../Impl/ConstantExpressionTemplate.cpp       |  33 +-
 .../Algorithms/AggregatingSortedAlgorithm.cpp |   6 +-
 .../Algorithms/CollapsingSortedAlgorithm.cpp  |   6 +-
 .../Algorithms/CollapsingSortedAlgorithm.h    |   6 +-
 .../GraphiteRollupSortedAlgorithm.cpp         |   8 +-
 .../Merges/Algorithms/IMergingAlgorithm.h     |   2 +
 .../IMergingAlgorithmWithDelayedChunk.cpp     |   4 +-
 .../IMergingAlgorithmWithSharedChunks.cpp     |   4 +-
 .../Algorithms/MergingSortedAlgorithm.cpp     |   2 +-
 .../Algorithms/MergingSortedAlgorithm.h       |   6 +-
 .../Algorithms/ReplacingSortedAlgorithm.cpp   |   2 +-
 src/Processors/Merges/Algorithms/RowRef.h     |   4 +-
 .../Algorithms/SummingSortedAlgorithm.cpp     |   4 +-
 .../VersionedCollapsingAlgorithm.cpp          |   2 +-
 .../Merges/CollapsingSortedTransform.h        |   4 +-
 .../Transforms/SortingTransform.cpp           |   2 +-
 src/Server/TestKeeperTCPHandler.cpp           |  42 +-
 src/Storages/AlterCommands.cpp                |   4 +-
 src/Storages/HDFS/HDFSCommon.cpp              | 194 +++++
 src/Storages/HDFS/HDFSCommon.h                | 114 +++
 .../HDFS}/ReadBufferFromHDFS.cpp              |  20 +-
 .../HDFS}/ReadBufferFromHDFS.h                |   9 +-
 src/Storages/{ => HDFS}/StorageHDFS.cpp       |  14 +-
 src/Storages/{ => HDFS}/StorageHDFS.h         |   0
 .../HDFS}/WriteBufferFromHDFS.cpp             |  17 +-
 .../HDFS}/WriteBufferFromHDFS.h               |   2 +-
 src/Storages/Kafka/KafkaSettings.h            |   2 +-
 .../MergeTree/IMergeTreeDataPartWriter.cpp    |  60 +-
 .../MergeTree/IMergeTreeDataPartWriter.h      |  60 +-
 .../MergeTree/IMergedBlockOutputStream.cpp    |  19 -
 .../MergeTree/IMergedBlockOutputStream.h      |   2 -
 src/Storages/MergeTree/KeyCondition.cpp       |   2 +-
 .../MergeTree/MergeTreeBlockOutputStream.cpp  |  12 +-
 .../MergeTree/MergeTreeBlockOutputStream.h    |   5 +-
 src/Storages/MergeTree/MergeTreeData.cpp      |  24 +-
 src/Storages/MergeTree/MergeTreeData.h        |   1 -
 .../MergeTree/MergeTreeDataMergerMutator.cpp  |  10 +
 .../MergeTreeDataPartWriterCompact.cpp        | 190 +++--
 .../MergeTreeDataPartWriterCompact.h          |  27 +-
 .../MergeTreeDataPartWriterInMemory.cpp       |  12 +-
 .../MergeTreeDataPartWriterInMemory.h         |   9 +-
 .../MergeTreeDataPartWriterOnDisk.cpp         | 134 ++--
 .../MergeTree/MergeTreeDataPartWriterOnDisk.h |  92 ++-
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 372 +++++++---
 .../MergeTree/MergeTreeDataPartWriterWide.h   |  68 +-
 .../MergeTree/MergeTreeDataWriter.cpp         | 137 +++-
 src/Storages/MergeTree/MergeTreeDataWriter.h  |   4 +-
 src/Storages/MergeTree/MergeTreeIOSettings.h  |   3 +
 .../MergeTree/MergeTreeIndexGranularity.cpp   |  12 +
 .../MergeTree/MergeTreeIndexGranularity.h     |   8 +-
 .../MergeTree/MergedBlockOutputStream.cpp     |  21 +-
 .../MergedColumnOnlyOutputStream.cpp          |  21 +-
 .../MergeTree/MergedColumnOnlyOutputStream.h  |   2 +-
 .../ReplicatedMergeTreeBlockOutputStream.cpp  |  16 +-
 .../ReplicatedMergeTreeBlockOutputStream.h    |   5 +-
 src/Storages/MergeTree/SimpleMergeSelector.h  |  77 ++
 src/Storages/SetSettings.h                    |   3 +-
 src/Storages/StorageDictionary.cpp            |   4 +-
 src/Storages/StorageJoin.cpp                  |  22 +-
 src/Storages/StorageJoin.h                    |   2 +-
 src/Storages/StorageMaterializeMySQL.cpp      |   7 +-
 src/Storages/StorageMaterializeMySQL.h        |  10 +-
 src/Storages/StorageMerge.cpp                 |  31 +-
 src/Storages/StorageMerge.h                   |  12 +-
 src/Storages/StorageMergeTree.cpp             |   2 +-
 src/Storages/StorageReplicatedMergeTree.cpp   |  11 +-
 src/Storages/StorageReplicatedMergeTree.h     |   3 +-
 src/Storages/StorageS3.cpp                    |  23 +-
 src/Storages/StorageS3.h                      |   4 +-
 src/Storages/StorageSet.cpp                   |  57 +-
 src/Storages/StorageSet.h                     |   7 +-
 src/Storages/StorageStripeLog.cpp             |   4 +-
 src/Storages/System/StorageSystemColumns.cpp  |  54 +-
 .../System/StorageSystemDatabases.cpp         |  19 +-
 src/Storages/System/StorageSystemTables.cpp   |  17 +-
 ...est_aux_funcs_for_adaptive_granularity.cpp |   1 +
 src/TableFunctions/TableFunctionFile.h        |   2 +-
 src/TableFunctions/TableFunctionHDFS.cpp      |   2 +-
 src/TableFunctions/TableFunctionMerge.cpp     |  77 +-
 src/TableFunctions/TableFunctionMerge.h       |   4 +-
 src/TableFunctions/TableFunctionS3.cpp        |   2 +
 tests/CMakeLists.txt                          |  10 -
 tests/clickhouse-test                         |  78 +-
 .../config.d/max_concurrent_queries.xml       |   4 +
 tests/config/install.sh                       |   1 +
 tests/config/top_level_domains                |   1 -
 .../top_level_domains/public_suffix_list.dat  |   0
 tests/instructions/easy_tasks_sorted_ru.md    |  28 +-
 tests/integration/README.md                   |  13 +-
 tests/integration/helpers/cluster.py          |  61 +-
 tests/integration/helpers/hdfs_api.py         | 185 ++++-
 .../test_allowed_url_from_config/test.py      |   6 +-
 .../configs/users.xml                         |   8 +
 .../test_graphite_merge_tree/test.py          |   3 +-
 .../test_join_set_family_s3/__init__.py       |   0
 .../configs/config.d/log_conf.xml             |  12 +
 .../configs/config.xml                        |  28 +
 .../test_join_set_family_s3/configs/minio.xml |  14 +
 .../test_join_set_family_s3/configs/ssl.xml   |  12 +
 .../test_join_set_family_s3/configs/users.xml |  23 +
 .../test_join_set_family_s3/test.py           |  97 +++
 .../configs/users.xml                         |   2 +
 .../configs/users_db_atomic.xml               |  19 +
 .../materialize_with_ddl.py                   |  52 +-
 .../test_materialize_mysql_database/test.py   | 103 +--
 .../test_postgresql_protocol/test.py          |   2 +-
 .../configs/users.d/settings.xml              |   1 +
 .../test_redirect_url_storage/test.py         |  15 +-
 .../configs/users.d/another_user.xml          |   7 +-
 tests/integration/test_storage_hdfs/test.py   |  57 +-
 tests/integration/test_storage_kafka/test.py  |  61 ++
 .../test_storage_kerberized_hdfs/__init__.py  |   0
 .../configs/hdfs.xml                          |  13 +
 .../configs/log_conf.xml                      |  11 +
 .../hdfs_configs/bootstrap.sh                 | 262 +++++++
 .../kerberos_image_config.sh                  | 139 ++++
 .../secrets/krb.conf                          |  24 +
 .../secrets/krb_long.conf                     |  24 +
 .../test_storage_kerberized_hdfs/test.py      | 108 +++
 tests/integration/test_storage_s3/test.py     |   3 +-
 .../__init__.py                               |   0
 .../test.py                                   |  55 ++
 .../configs/logs_conf.xml                     |  12 +
 .../test_testkeeper_back_to_back/test.py      | 125 +++-
 .../00083_create_merge_tree_zookeeper.sql     |   2 +
 .../00327_summing_composite_nested.sql        |   2 +
 .../00443_optimize_final_vertical_merge.sh    |   6 +-
 ...ed_storage_definition_syntax_zookeeper.sql |   2 +
 .../00564_versioned_collapsing_merge_tree.sql |   2 +
 ...77_replacing_merge_tree_vertical_merge.sql |   2 +
 .../0_stateless/00616_final_single_part.sql   |   2 +
 ...ptimize_on_nonleader_replica_zookeeper.sql |   4 +-
 ...00660_optimize_final_without_partition.sql |   2 +
 ...replicated_without_partition_zookeeper.sql |   2 +
 .../0_stateless/00682_empty_parts_merge.sh    |  14 +-
 ...block_size_system_tables_columns.reference |   2 +-
 ...sert_have_data_before_quorum_zookeeper.sql |   2 +-
 ...ert_lost_part_and_alive_part_zookeeper.sql |   4 +-
 ...0732_quorum_insert_lost_part_zookeeper.sql |   4 +-
 ..._old_data_and_without_quorum_zookeeper.sql |   2 +-
 ...m_insert_simple_test_1_parts_zookeeper.sql |   2 +-
 ...m_insert_simple_test_2_parts_zookeeper.sql |   2 +-
 .../00754_alter_modify_order_by.sql           |   1 +
 ...r_modify_order_by_replicated_zookeeper.sql |   2 +
 ...tem_parts_race_condition_drop_zookeeper.sh |   2 +-
 ...030_incorrect_count_summing_merge_tree.sql |   2 +
 ...invalidate_query_switchover_long.reference |   2 +-
 ...ionary_invalidate_query_switchover_long.sh |   2 +-
 .../01056_create_table_as.reference           |   1 +
 .../0_stateless/01056_create_table_as.sql     |   4 +
 ...81_PartialSortingTransform_full_column.sql |   2 -
 ..._zookeeper_mutations_and_insert_quorum.sql |   2 +-
 ...98_temporary_and_external_tables.reference |   7 +
 .../01098_temporary_and_external_tables.sh    |  27 +-
 .../01187_set_profile_as_setting.reference    |   8 +
 .../01187_set_profile_as_setting.sh           |  15 +
 ...01285_data_skip_index_over_aggregation.sql |   2 +
 .../0_stateless/01293_show_settings.reference |   5 +
 .../0_stateless/01293_show_settings.sql       |   3 +
 .../01322_cast_keep_nullable.reference        |   1 +
 .../0_stateless/01322_cast_keep_nullable.sql  |   2 +-
 .../0_stateless/01323_add_scalars_in_time.sql |   2 +
 ...mming_merge_tree_exclude_partition_key.sql |   2 +
 ...1451_replicated_detach_drop_and_quorum.sql |   2 +-
 .../01459_manual_write_to_replicas_quorum.sh  |   2 +-
 ..._select_sequence_consistency_zookeeper.sql |   2 +-
 .../01550_create_map_type.reference           |  29 +
 .../0_stateless/01550_create_map_type.sql     |  65 ++
 .../01550_type_map_formats.reference          |  46 ++
 .../0_stateless/01550_type_map_formats.sql    |  19 +
 .../01550_type_map_formats_input.reference    |   5 +
 .../01550_type_map_formats_input.sh           |  21 +
 .../01556_accurate_cast_or_null.reference     |  30 +
 .../01556_accurate_cast_or_null.sql           |  35 +
 .../01560_optimize_on_insert.reference        |  13 +
 .../0_stateless/01560_optimize_on_insert.sql  |  35 +
 ...gregator_combinator_simple_state.reference |  14 +
 ...570_aggregator_combinator_simple_state.sql |  17 +
 .../01581_to_int256_uint256_inf_nan.reference |   6 -
 .../01581_to_int256_uint256_inf_nan.sql       |  10 -
 .../01581_to_int_inf_nan.reference            |   0
 .../0_stateless/01581_to_int_inf_nan.sql      |  10 +
 .../01596_setting_limit_offset.reference      |  63 ++
 .../01596_setting_limit_offset.sql            |  30 +
 .../01600_detach_permanently.reference        | 102 +++
 .../0_stateless/01600_detach_permanently.sh   | 125 ++++
 ...emerge_sort_lowered_memory_bytes_ratio.sql |  28 +-
 .../01600_select_in_different_types.reference |  11 +
 .../01600_select_in_different_types.sql       |  35 +
 .../0_stateless/01601_accurate_cast.reference |   8 +
 .../0_stateless/01601_accurate_cast.sql       |  24 +
 .../01601_detach_permanently.reference        |  79 ++
 .../0_stateless/01601_detach_permanently.sql  | 214 ++++++
 ...01_temporary_table_session_scope.reference |   5 +
 .../01601_temporary_table_session_scope.sh    |  19 +
 .../01602_array_aggregation.reference         |  60 ++
 .../0_stateless/01602_array_aggregation.sql   |  60 ++
 ...1602_max_distributed_connections.reference |   0
 .../01602_max_distributed_connections.sh      |  15 +
 .../01602_modified_julian_day_msan.reference  |   1 +
 .../01602_modified_julian_day_msan.sql        |   4 +
 ...temporary_table_in_system_tables.reference |  10 +
 ...01602_temporary_table_in_system_tables.sql |  18 +
 .../01603_decimal_mult_float.reference        |  12 +
 .../0_stateless/01603_decimal_mult_float.sql  |  26 +
 ...603_insert_select_too_many_parts.reference |   1 +
 .../01603_insert_select_too_many_parts.sql    |  14 +
 .../01603_remove_column_ttl.reference         |   6 +
 .../0_stateless/01603_remove_column_ttl.sql   |  30 +
 .../01603_rename_overwrite_bug.reference      |   4 +
 .../01603_rename_overwrite_bug.sql            |  22 +
 ...4_explain_ast_of_nonselect_query.reference |   8 +
 .../01604_explain_ast_of_nonselect_query.sql  |   1 +
 .../queries/0_stateless/arcadia_skip_list.txt |   2 +
 tests/queries/skip_list.json                  | 172 +++++
 .../tests/privileges/create/create_table.py   |   3 +
 .../convert-month-partitioned-parts/main.cpp  |   1 +
 utils/db-generator/query_db_generator.cpp     |   2 +-
 utils/iotest/iotest.cpp                       |   3 +-
 utils/iotest/iotest_aio.cpp                   |   3 +-
 utils/zookeeper-test/main.cpp                 |  97 ++-
 416 files changed, 10085 insertions(+), 2626 deletions(-)
 create mode 100644 base/common/getPageSize.cpp
 create mode 100644 base/common/getPageSize.h
 create mode 100644 docker/test/integration/kerberized_hadoop/Dockerfile
 create mode 100644 docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml
 create mode 100644 docs/en/operations/system-tables/replication_queue.md
 create mode 100644 docs/ru/faq/general/index.md
 rename docs/ru/faq/{general.md => general/mapreduce.md} (60%)
 create mode 100644 docs/ru/faq/integration/file-export.md
 create mode 100644 docs/ru/faq/integration/index.md
 create mode 100644 docs/ru/faq/integration/oracle-odbc.md
 create mode 100644 docs/ru/faq/operations/index.md
 create mode 100644 docs/ru/faq/use-cases/index.md
 create mode 100644 docs/ru/operations/system-tables/replication_queue.md
 create mode 100644 src/AggregateFunctions/AggregateFunctionSimpleState.cpp
 create mode 100644 src/AggregateFunctions/AggregateFunctionSimpleState.h
 create mode 100644 src/Columns/ColumnMap.cpp
 create mode 100644 src/Columns/ColumnMap.h
 create mode 100644 src/DataTypes/DataTypeMap.cpp
 create mode 100644 src/DataTypes/DataTypeMap.h
 create mode 100644 src/Functions/array/arrayAggregation.cpp
 delete mode 100644 src/Functions/array/arraySum.cpp
 create mode 100644 src/Functions/map.cpp
 delete mode 100644 src/IO/HDFSCommon.cpp
 delete mode 100644 src/IO/HDFSCommon.h
 create mode 100644 src/Storages/HDFS/HDFSCommon.cpp
 create mode 100644 src/Storages/HDFS/HDFSCommon.h
 rename src/{IO => Storages/HDFS}/ReadBufferFromHDFS.cpp (81%)
 rename src/{IO => Storages/HDFS}/ReadBufferFromHDFS.h (69%)
 rename src/Storages/{ => HDFS}/StorageHDFS.cpp (95%)
 rename src/Storages/{ => HDFS}/StorageHDFS.h (100%)
 rename src/{IO => Storages/HDFS}/WriteBufferFromHDFS.cpp (84%)
 rename src/{IO => Storages/HDFS}/WriteBufferFromHDFS.h (80%)
 create mode 100644 tests/config/config.d/max_concurrent_queries.xml
 delete mode 120000 tests/config/top_level_domains
 rename {docker/test/performance-comparison => tests}/config/top_level_domains/public_suffix_list.dat (100%)
 create mode 100644 tests/integration/test_graphite_merge_tree/configs/users.xml
 create mode 100644 tests/integration/test_join_set_family_s3/__init__.py
 create mode 100644 tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/config.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/minio.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/ssl.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/users.xml
 create mode 100644 tests/integration/test_join_set_family_s3/test.py
 create mode 100644 tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/__init__.py
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml
 create mode 100755 tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/test.py
 create mode 100644 tests/integration/test_table_functions_access_rights/__init__.py
 create mode 100644 tests/integration/test_table_functions_access_rights/test.py
 create mode 100644 tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml
 create mode 100644 tests/queries/0_stateless/01187_set_profile_as_setting.reference
 create mode 100755 tests/queries/0_stateless/01187_set_profile_as_setting.sh
 create mode 100644 tests/queries/0_stateless/01293_show_settings.reference
 create mode 100644 tests/queries/0_stateless/01293_show_settings.sql
 create mode 100644 tests/queries/0_stateless/01550_create_map_type.reference
 create mode 100644 tests/queries/0_stateless/01550_create_map_type.sql
 create mode 100644 tests/queries/0_stateless/01550_type_map_formats.reference
 create mode 100644 tests/queries/0_stateless/01550_type_map_formats.sql
 create mode 100644 tests/queries/0_stateless/01550_type_map_formats_input.reference
 create mode 100755 tests/queries/0_stateless/01550_type_map_formats_input.sh
 create mode 100644 tests/queries/0_stateless/01556_accurate_cast_or_null.reference
 create mode 100644 tests/queries/0_stateless/01556_accurate_cast_or_null.sql
 create mode 100644 tests/queries/0_stateless/01560_optimize_on_insert.reference
 create mode 100644 tests/queries/0_stateless/01560_optimize_on_insert.sql
 create mode 100644 tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference
 create mode 100644 tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql
 delete mode 100644 tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference
 delete mode 100644 tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql
 create mode 100644 tests/queries/0_stateless/01581_to_int_inf_nan.reference
 create mode 100644 tests/queries/0_stateless/01581_to_int_inf_nan.sql
 create mode 100644 tests/queries/0_stateless/01596_setting_limit_offset.reference
 create mode 100644 tests/queries/0_stateless/01596_setting_limit_offset.sql
 create mode 100644 tests/queries/0_stateless/01600_detach_permanently.reference
 create mode 100755 tests/queries/0_stateless/01600_detach_permanently.sh
 create mode 100644 tests/queries/0_stateless/01600_select_in_different_types.reference
 create mode 100644 tests/queries/0_stateless/01600_select_in_different_types.sql
 create mode 100644 tests/queries/0_stateless/01601_accurate_cast.reference
 create mode 100644 tests/queries/0_stateless/01601_accurate_cast.sql
 create mode 100644 tests/queries/0_stateless/01601_detach_permanently.reference
 create mode 100644 tests/queries/0_stateless/01601_detach_permanently.sql
 create mode 100644 tests/queries/0_stateless/01601_temporary_table_session_scope.reference
 create mode 100755 tests/queries/0_stateless/01601_temporary_table_session_scope.sh
 create mode 100644 tests/queries/0_stateless/01602_array_aggregation.reference
 create mode 100644 tests/queries/0_stateless/01602_array_aggregation.sql
 create mode 100644 tests/queries/0_stateless/01602_max_distributed_connections.reference
 create mode 100755 tests/queries/0_stateless/01602_max_distributed_connections.sh
 create mode 100644 tests/queries/0_stateless/01602_modified_julian_day_msan.reference
 create mode 100644 tests/queries/0_stateless/01602_modified_julian_day_msan.sql
 create mode 100644 tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference
 create mode 100644 tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql
 create mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.reference
 create mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.sql
 create mode 100644 tests/queries/0_stateless/01603_insert_select_too_many_parts.reference
 create mode 100644 tests/queries/0_stateless/01603_insert_select_too_many_parts.sql
 create mode 100644 tests/queries/0_stateless/01603_remove_column_ttl.reference
 create mode 100644 tests/queries/0_stateless/01603_remove_column_ttl.sql
 create mode 100644 tests/queries/0_stateless/01603_rename_overwrite_bug.reference
 create mode 100644 tests/queries/0_stateless/01603_rename_overwrite_bug.sql
 create mode 100644 tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference
 create mode 100644 tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql

diff --git a/.gitmodules b/.gitmodules
index 1d996fb84cc..23deab0eed0 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -183,7 +183,7 @@
 	url = https://github.com/kthohr/stats.git
 [submodule "contrib/krb5"]
 	path = contrib/krb5
-	url = https://github.com/krb5/krb5
+	url = https://github.com/ClickHouse-Extras/krb5
 [submodule "contrib/cyrus-sasl"]
 	path = contrib/cyrus-sasl
 	url = https://github.com/cyrusimap/cyrus-sasl
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4d525798505..6d9174c7d07 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,126 @@
+### ClickHouse release 20.12
+
+### ClickHouse release v20.12.3.3-stable, 2020-12-13
+
+#### Backward Incompatible Change
+
+* Enable `use_compact_format_in_distributed_parts_names` by default (see the documentation for the reference). [#16728](https://github.com/ClickHouse/ClickHouse/pull/16728) ([Azat Khuzhin](https://github.com/azat)).
+* Accept user settings related to file formats (e.g. `format_csv_delimiter`) in the `SETTINGS` clause when creating a table that uses `File` engine, and use these settings in all `INSERT`s and `SELECT`s. The file format settings changed in the current user session, or in the `SETTINGS` clause of a DML query itself, no longer affect the query. [#16591](https://github.com/ClickHouse/ClickHouse/pull/16591) ([Alexander Kuzmenkov](https://github.com/akuzm)).
+
+#### New Feature
+
+* add `*.xz` compression/decompression support.It enables using `*.xz` in `file()` function. This closes [#8828](https://github.com/ClickHouse/ClickHouse/issues/8828). [#16578](https://github.com/ClickHouse/ClickHouse/pull/16578) ([Abi Palagashvili](https://github.com/fibersel)).
+* Introduce the query `ALTER TABLE ... DROP|DETACH PART 'part_name'`. [#15511](https://github.com/ClickHouse/ClickHouse/pull/15511) ([nvartolomei](https://github.com/nvartolomei)).
+* Added new ALTER UPDATE/DELETE IN PARTITION syntax. [#13403](https://github.com/ClickHouse/ClickHouse/pull/13403) ([Vladimir Chebotarev](https://github.com/excitoon)).
+* Allow formatting named tuples as JSON objects when using JSON input/output formats, controlled by the `output_format_json_named_tuples_as_objects` setting, disabled by default. [#17175](https://github.com/ClickHouse/ClickHouse/pull/17175) ([Alexander Kuzmenkov](https://github.com/akuzm)).
+* Add a possibility to input enum value as it's id in TSV and CSV formats by default. [#16834](https://github.com/ClickHouse/ClickHouse/pull/16834) ([Kruglov Pavel](https://github.com/Avogar)).
+* Add COLLATE support for Nullable, LowCardinality, Array and Tuple, where nested type is String. Also refactor the code associated with collations in ColumnString.cpp. [#16273](https://github.com/ClickHouse/ClickHouse/pull/16273) ([Kruglov Pavel](https://github.com/Avogar)).
+* New `tcpPort` function returns TCP port listened by this server. [#17134](https://github.com/ClickHouse/ClickHouse/pull/17134) ([Ivan](https://github.com/abyss7)).
+* Add new math functions: `acosh`, `asinh`, `atan2`, `atanh`, `cosh`, `hypot`, `log1p`, `sinh`. [#16636](https://github.com/ClickHouse/ClickHouse/pull/16636) ([Konstantin Malanchev](https://github.com/hombit)).
+* Possibility to distribute the merges between different replicas. Introduces the `execute_merges_on_single_replica_time_threshold` mergetree setting. [#16424](https://github.com/ClickHouse/ClickHouse/pull/16424) ([filimonov](https://github.com/filimonov)).
+* Add setting `aggregate_functions_null_for_empty` for SQL standard compatibility. This option will rewrite all aggregate functions in a query, adding -OrNull suffix to them. Implements [10273](https://github.com/ClickHouse/ClickHouse/issues/10273). [#16123](https://github.com/ClickHouse/ClickHouse/pull/16123) ([flynn](https://github.com/ucasFL)).
+* Updated DateTime, DateTime64 parsing to accept string Date literal format. [#16040](https://github.com/ClickHouse/ClickHouse/pull/16040) ([Maksim Kita](https://github.com/kitaisreal)).
+* Make it possible to change the path to history file in `clickhouse-client` using the `--history_file` parameter. [#15960](https://github.com/ClickHouse/ClickHouse/pull/15960) ([Maksim Kita](https://github.com/kitaisreal)).
+
+#### Bug Fix
+
+* Fix the issue when server can stop accepting connections in very rare cases. [#17542](https://github.com/ClickHouse/ClickHouse/pull/17542) ([Amos Bird](https://github.com/amosbird)).
+* Fixed `Function not implemented` error when executing `RENAME` query in `Atomic` database with ClickHouse running on Windows Subsystem for Linux. Fixes [#17661](https://github.com/ClickHouse/ClickHouse/issues/17661). [#17664](https://github.com/ClickHouse/ClickHouse/pull/17664) ([tavplubix](https://github.com/tavplubix)).
+* Do not restore parts from WAL if `in_memory_parts_enable_wal` is disabled. [#17802](https://github.com/ClickHouse/ClickHouse/pull/17802) ([detailyang](https://github.com/detailyang)).
+* fix incorrect initialization of `max_compress_block_size` of MergeTreeWriterSettings with `min_compress_block_size`. [#17833](https://github.com/ClickHouse/ClickHouse/pull/17833) ([flynn](https://github.com/ucasFL)).
+* Exception message about max table size to drop was displayed incorrectly. [#17764](https://github.com/ClickHouse/ClickHouse/pull/17764) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed possible segfault when there is not enough space when inserting into `Distributed` table. [#17737](https://github.com/ClickHouse/ClickHouse/pull/17737) ([tavplubix](https://github.com/tavplubix)).
+* Fixed problem when ClickHouse fails to resume connection to MySQL servers. [#17681](https://github.com/ClickHouse/ClickHouse/pull/17681) ([Alexander Kazakov](https://github.com/Akazz)).
+* In might be determined incorrectly if cluster is circular- (cross-) replicated or not when executing `ON CLUSTER` query due to race condition when `pool_size` > 1. It's fixed. [#17640](https://github.com/ClickHouse/ClickHouse/pull/17640) ([tavplubix](https://github.com/tavplubix)).
+* Exception `fmt::v7::format_error` can be logged in background for MergeTree tables. This fixes [#17613](https://github.com/ClickHouse/ClickHouse/issues/17613). [#17615](https://github.com/ClickHouse/ClickHouse/pull/17615) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* When clickhouse-client is used in interactive mode with multiline queries, single line comment was erronously extended till the end of query. This fixes [#13654](https://github.com/ClickHouse/ClickHouse/issues/13654). [#17565](https://github.com/ClickHouse/ClickHouse/pull/17565) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix alter query hang when the corresponding mutation was killed on the different replica. Fixes [#16953](https://github.com/ClickHouse/ClickHouse/issues/16953). [#17499](https://github.com/ClickHouse/ClickHouse/pull/17499) ([alesapin](https://github.com/alesapin)).
+* Fix issue when mark cache size was underestimated by clickhouse. It may happen when there are a lot of tiny files with marks. [#17496](https://github.com/ClickHouse/ClickHouse/pull/17496) ([alesapin](https://github.com/alesapin)).
+* Fix `ORDER BY` with enabled setting `optimize_redundant_functions_in_order_by`. [#17471](https://github.com/ClickHouse/ClickHouse/pull/17471) ([Anton Popov](https://github.com/CurtizJ)).
+* Fix duplicates after `DISTINCT` which were possible because of incorrect optimization. Fixes [#17294](https://github.com/ClickHouse/ClickHouse/issues/17294). [#17296](https://github.com/ClickHouse/ClickHouse/pull/17296) ([li chengxiang](https://github.com/chengxianglibra)). [#17439](https://github.com/ClickHouse/ClickHouse/pull/17439) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fix crash while reading from `JOIN` table with `LowCardinality` types. Fixes [#17228](https://github.com/ClickHouse/ClickHouse/issues/17228). [#17397](https://github.com/ClickHouse/ClickHouse/pull/17397) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* fix `toInt256(inf)` stack overflow. Int256 is an experimental feature. Closed [#17235](https://github.com/ClickHouse/ClickHouse/issues/17235). [#17257](https://github.com/ClickHouse/ClickHouse/pull/17257) ([flynn](https://github.com/ucasFL)).
+* Fix possible `Unexpected packet Data received from client` error logged for Distributed queries with `LIMIT`. [#17254](https://github.com/ClickHouse/ClickHouse/pull/17254) ([Azat Khuzhin](https://github.com/azat)).
+* Fix set index invalidation when there are const columns in the subquery. This fixes [#17246](https://github.com/ClickHouse/ClickHouse/issues/17246). [#17249](https://github.com/ClickHouse/ClickHouse/pull/17249) ([Amos Bird](https://github.com/amosbird)).
+* Fix possible wrong index analysis when the types of the index comparison are different. This fixes [#17122](https://github.com/ClickHouse/ClickHouse/issues/17122). [#17145](https://github.com/ClickHouse/ClickHouse/pull/17145) ([Amos Bird](https://github.com/amosbird)).
+* Fix ColumnConst comparison which leads to crash. This fixed [#17088](https://github.com/ClickHouse/ClickHouse/issues/17088) . [#17135](https://github.com/ClickHouse/ClickHouse/pull/17135) ([Amos Bird](https://github.com/amosbird)).
+* Multiple fixed for MaterializeMySQL (experimental feature). Fixes [#16923](https://github.com/ClickHouse/ClickHouse/issues/16923) Fixes [#15883](https://github.com/ClickHouse/ClickHouse/issues/15883) Fix MaterializeMySQL SYNC failure when the modify MySQL binlog_checksum. [#17091](https://github.com/ClickHouse/ClickHouse/pull/17091) ([Winter Zhang](https://github.com/zhang2014)).
+* Fix bug when `ON CLUSTER` queries may hang forever for non-leader ReplicatedMergeTreeTables. [#17089](https://github.com/ClickHouse/ClickHouse/pull/17089) ([alesapin](https://github.com/alesapin)).
+* Fixed crash on `CREATE TABLE ... AS some_table` query when `some_table` was created `AS table_function()` Fixes [#16944](https://github.com/ClickHouse/ClickHouse/issues/16944). [#17072](https://github.com/ClickHouse/ClickHouse/pull/17072) ([tavplubix](https://github.com/tavplubix)).
+* Bug unfinished implementation for funciton fuzzBits, related issue: [#16980](https://github.com/ClickHouse/ClickHouse/issues/16980). [#17051](https://github.com/ClickHouse/ClickHouse/pull/17051) ([hexiaoting](https://github.com/hexiaoting)).
+* Fix LLVM's libunwind in the case when CFA register is RAX. This is the [bug](https://bugs.llvm.org/show_bug.cgi?id=48186) in [LLVM's libunwind](https://github.com/llvm/llvm-project/tree/master/libunwind). We already have workarounds for this bug. [#17046](https://github.com/ClickHouse/ClickHouse/pull/17046) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Avoid unnecessary network errors for remote queries which may be cancelled while execution, like queries with `LIMIT`. [#17006](https://github.com/ClickHouse/ClickHouse/pull/17006) ([Azat Khuzhin](https://github.com/azat)).
+* Fix `optimize_distributed_group_by_sharding_key` setting (that is disabled by default) for query with OFFSET only. [#16996](https://github.com/ClickHouse/ClickHouse/pull/16996) ([Azat Khuzhin](https://github.com/azat)).
+* Fix for Merge tables over Distributed tables with JOIN. [#16993](https://github.com/ClickHouse/ClickHouse/pull/16993) ([Azat Khuzhin](https://github.com/azat)).
+* Fixed wrong result in big integers (128, 256 bit) when casting from double. Big integers support is experimental. [#16986](https://github.com/ClickHouse/ClickHouse/pull/16986) ([Mike](https://github.com/myrrc)).
+* Fix possible server crash after `ALTER TABLE ... MODIFY COLUMN ... NewType` when `SELECT` have `WHERE` expression on altering column and alter doesn't finished yet. [#16968](https://github.com/ClickHouse/ClickHouse/pull/16968) ([Amos Bird](https://github.com/amosbird)).
+* Blame info was not calculated correctly in `clickhouse-git-import`. [#16959](https://github.com/ClickHouse/ClickHouse/pull/16959) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix order by optimization with monotonous functions. Fixes [#16107](https://github.com/ClickHouse/ClickHouse/issues/16107). [#16956](https://github.com/ClickHouse/ClickHouse/pull/16956) ([Anton Popov](https://github.com/CurtizJ)).
+* Fix optimization of group by with enabled setting `optimize_aggregators_of_group_by_keys` and joins. Fixes [#12604](https://github.com/ClickHouse/ClickHouse/issues/12604). [#16951](https://github.com/ClickHouse/ClickHouse/pull/16951) ([Anton Popov](https://github.com/CurtizJ)).
+* Fix possible error `Illegal type of argument` for queries with `ORDER BY`. Fixes [#16580](https://github.com/ClickHouse/ClickHouse/issues/16580). [#16928](https://github.com/ClickHouse/ClickHouse/pull/16928) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fix strange code in InterpreterShowAccessQuery. [#16866](https://github.com/ClickHouse/ClickHouse/pull/16866) ([tavplubix](https://github.com/tavplubix)).
+* Prevent clickhouse server crashes when using the function `timeSeriesGroupSum`. The function is removed from newer ClickHouse releases. [#16865](https://github.com/ClickHouse/ClickHouse/pull/16865) ([filimonov](https://github.com/filimonov)).
+* Fix rare silent crashes when query profiler is on and ClickHouse is installed on OS with glibc version that has (supposedly) broken asynchronous unwind tables for some functions. This fixes [#15301](https://github.com/ClickHouse/ClickHouse/issues/15301). This fixes [#13098](https://github.com/ClickHouse/ClickHouse/issues/13098). [#16846](https://github.com/ClickHouse/ClickHouse/pull/16846) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix crash when using `any` without any arguments. This is for [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803) . cc @azat. [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
+* If no memory can be allocated while writing table metadata on disk, broken metadata file can be written. [#16772](https://github.com/ClickHouse/ClickHouse/pull/16772) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix trivial query optimization with partition predicate. [#16767](https://github.com/ClickHouse/ClickHouse/pull/16767) ([Azat Khuzhin](https://github.com/azat)).
+* Fix `IN` operator over several columns and tuples with enabled `transform_null_in` setting. Fixes [#15310](https://github.com/ClickHouse/ClickHouse/issues/15310). [#16722](https://github.com/ClickHouse/ClickHouse/pull/16722) ([Anton Popov](https://github.com/CurtizJ)).
+* Return number of affected rows for INSERT queries via MySQL protocol. Previously ClickHouse used to always return 0, it's fixed. Fixes [#16605](https://github.com/ClickHouse/ClickHouse/issues/16605). [#16715](https://github.com/ClickHouse/ClickHouse/pull/16715) ([Winter Zhang](https://github.com/zhang2014)).
+* Fix remote query failure when using 'if' suffix aggregate function. Fixes [#16574](https://github.com/ClickHouse/ClickHouse/issues/16574) Fixes [#16231](https://github.com/ClickHouse/ClickHouse/issues/16231) [#16610](https://github.com/ClickHouse/ClickHouse/pull/16610) ([Winter Zhang](https://github.com/zhang2014)).
+* Fix inconsistent behavior caused by `select_sequential_consistency` for optimized trivial count query and system.tables. [#16309](https://github.com/ClickHouse/ClickHouse/pull/16309) ([Hao Chen](https://github.com/haoch)).
+
+#### Improvement
+
+* Remove empty parts after they were pruned by TTL, mutation, or collapsing merge algorithm. [#16895](https://github.com/ClickHouse/ClickHouse/pull/16895) ([Anton Popov](https://github.com/CurtizJ)).
+* Enable compact format of directories for asynchronous sends in Distributed tables: `use_compact_format_in_distributed_parts_names` is set to 1 by default. [#16788](https://github.com/ClickHouse/ClickHouse/pull/16788) ([Azat Khuzhin](https://github.com/azat)).
+* Abort multipart upload if no data was written to S3. [#16840](https://github.com/ClickHouse/ClickHouse/pull/16840) ([Pavel Kovalenko](https://github.com/Jokser)).
+* Reresolve the IP of the `format_avro_schema_registry_url` in case of errors. [#16985](https://github.com/ClickHouse/ClickHouse/pull/16985) ([filimonov](https://github.com/filimonov)).
+* Mask password in data_path in the system.distribution_queue. [#16727](https://github.com/ClickHouse/ClickHouse/pull/16727) ([Azat Khuzhin](https://github.com/azat)).
+* Throw error when use column transformer replaces non existing column. [#16183](https://github.com/ClickHouse/ClickHouse/pull/16183) ([hexiaoting](https://github.com/hexiaoting)).
+* Turn off parallel parsing when there is no enough memory for all threads to work simultaneously. Also there could be exceptions like "Memory limit exceeded" when somebody will try to insert extremely huge rows (> min_chunk_bytes_for_parallel_parsing), because each piece to parse has to be independent set of strings (one or more). [#16721](https://github.com/ClickHouse/ClickHouse/pull/16721) ([Nikita Mikhaylov](https://github.com/nikitamikhaylov)).
+* Install script should always create subdirs in config folders. This is only relevant for Docker build with custom config. [#16936](https://github.com/ClickHouse/ClickHouse/pull/16936) ([filimonov](https://github.com/filimonov)).
+* Correct grammar in error message in JSONEachRow, JSONCompactEachRow, and RegexpRow input formats. [#17205](https://github.com/ClickHouse/ClickHouse/pull/17205) ([nico piderman](https://github.com/sneako)).
+* Set default `host` and `port` parameters for `SOURCE(CLICKHOUSE(...))` to current instance and set default `user` value to `'default'`. [#16997](https://github.com/ClickHouse/ClickHouse/pull/16997) ([vdimir](https://github.com/vdimir)).
+* Throw an informative error message when doing `ATTACH/DETACH TABLE <DICTIONARY>`. Before this PR, `detach table <dict>` works but leads to an ill-formed in-memory metadata. [#16885](https://github.com/ClickHouse/ClickHouse/pull/16885) ([Amos Bird](https://github.com/amosbird)).
+* Add cutToFirstSignificantSubdomainWithWWW(). [#16845](https://github.com/ClickHouse/ClickHouse/pull/16845) ([Azat Khuzhin](https://github.com/azat)).
+* Server refused to startup with exception message if wrong config is given (`metric_log`.`collect_interval_milliseconds` is missing). [#16815](https://github.com/ClickHouse/ClickHouse/pull/16815) ([Ivan](https://github.com/abyss7)).
+* Better exception message when configuration for distributed DDL is absent. This fixes [#5075](https://github.com/ClickHouse/ClickHouse/issues/5075). [#16769](https://github.com/ClickHouse/ClickHouse/pull/16769) ([Nikita Mikhaylov](https://github.com/nikitamikhaylov)).
+* Usability improvement: better suggestions in syntax error message when `CODEC` expression is misplaced in `CREATE TABLE` query. This fixes [#12493](https://github.com/ClickHouse/ClickHouse/issues/12493). [#16768](https://github.com/ClickHouse/ClickHouse/pull/16768) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Remove empty directories for async INSERT at start of Distributed engine. [#16729](https://github.com/ClickHouse/ClickHouse/pull/16729) ([Azat Khuzhin](https://github.com/azat)).
+* Workaround for use S3 with nginx server as proxy. Nginx currenty does not accept urls with empty path like `http://domain.com?delete`, but vanilla aws-sdk-cpp produces this kind of urls. This commit uses patched aws-sdk-cpp version, which makes urls with "/" as path in this cases, like `http://domain.com/?delete`. [#16709](https://github.com/ClickHouse/ClickHouse/pull/16709) ([ianton-ru](https://github.com/ianton-ru)).
+* Allow `reinterpretAs*` functions to work for integers and floats of the same size. Implements [16640](https://github.com/ClickHouse/ClickHouse/issues/16640). [#16657](https://github.com/ClickHouse/ClickHouse/pull/16657) ([flynn](https://github.com/ucasFL)).
+* Now, `<auxiliary_zookeepers>` configuration can be changed in `config.xml` and reloaded without server startup. [#16627](https://github.com/ClickHouse/ClickHouse/pull/16627) ([Amos Bird](https://github.com/amosbird)).
+* Support SNI in https connections to remote resources. This will allow to connect to Cloudflare servers that require SNI. This fixes [#10055](https://github.com/ClickHouse/ClickHouse/issues/10055). [#16252](https://github.com/ClickHouse/ClickHouse/pull/16252) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Make it possible to connect to `clickhouse-server` secure endpoint which requires SNI. This is possible when `clickhouse-server` is hosted behind TLS proxy. [#16938](https://github.com/ClickHouse/ClickHouse/pull/16938) ([filimonov](https://github.com/filimonov)).
+* Fix possible stack overflow if a loop of materialized views is created. This closes [#15732](https://github.com/ClickHouse/ClickHouse/issues/15732). [#16048](https://github.com/ClickHouse/ClickHouse/pull/16048) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Simplify the implementation of background tasks processing for the MergeTree table engines family. There should be no visible changes for user. [#15983](https://github.com/ClickHouse/ClickHouse/pull/15983) ([alesapin](https://github.com/alesapin)).
+* Improvement for MaterializeMySQL (experimental feature). Throw exception about right sync privileges when MySQL sync user has error privileges. [#15977](https://github.com/ClickHouse/ClickHouse/pull/15977) ([TCeason](https://github.com/TCeason)).
+* Made `indexOf()` use BloomFilter. [#14977](https://github.com/ClickHouse/ClickHouse/pull/14977) ([achimbab](https://github.com/achimbab)).
+
+#### Performance Improvement
+
+* Use Floyd-Rivest algorithm, it is the best for the ClickHouse use case of partial sorting. Bechmarks are in https://github.com/danlark1/miniselect and [here](https://drive.google.com/drive/folders/1DHEaeXgZuX6AJ9eByeZ8iQVQv0ueP8XM). [#16825](https://github.com/ClickHouse/ClickHouse/pull/16825) ([Danila Kutenin](https://github.com/danlark1)).
+* Now `ReplicatedMergeTree` tree engines family uses a separate thread pool for replicated fetches. Size of the pool limited by setting `background_fetches_pool_size` which can be tuned with a server restart. The default value of the setting is 3 and it means that the maximum amount of parallel fetches is equal to 3 (and it allows to utilize 10G network). Fixes #520. [#16390](https://github.com/ClickHouse/ClickHouse/pull/16390) ([alesapin](https://github.com/alesapin)).
+* Fixed uncontrolled growth of the state of `quantileTDigest`. [#16680](https://github.com/ClickHouse/ClickHouse/pull/16680) ([hrissan](https://github.com/hrissan)).
+* Add `VIEW` subquery description to `EXPLAIN`. Limit push down optimisation for `VIEW`. Add local replicas of `Distributed` to query plan. [#14936](https://github.com/ClickHouse/ClickHouse/pull/14936) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fix optimize_read_in_order/optimize_aggregation_in_order with max_threads > 0 and expression in ORDER BY. [#16637](https://github.com/ClickHouse/ClickHouse/pull/16637) ([Azat Khuzhin](https://github.com/azat)).
+* Fix performance of reading from `Merge` tables over huge number of `MergeTree` tables. Fixes [#7748](https://github.com/ClickHouse/ClickHouse/issues/7748). [#16988](https://github.com/ClickHouse/ClickHouse/pull/16988) ([Anton Popov](https://github.com/CurtizJ)).
+* Now we can safely prune partitions with exact match. Useful case: Suppose table is partitioned by `intHash64(x) % 100` and the query has condition on `intHash64(x) % 100` verbatim, not on x. [#16253](https://github.com/ClickHouse/ClickHouse/pull/16253) ([Amos Bird](https://github.com/amosbird)).
+
+#### Experimental Feature
+
+* Add `EmbeddedRocksDB` table engine (can be used for dictionaries). [#15073](https://github.com/ClickHouse/ClickHouse/pull/15073) ([sundyli](https://github.com/sundy-li)).
+
+#### Build/Testing/Packaging Improvement
+
+* Improvements in test coverage building images. [#17233](https://github.com/ClickHouse/ClickHouse/pull/17233) ([alesapin](https://github.com/alesapin)).
+* Update embedded timezone data to version 2020d (also update cctz to the latest master). [#17204](https://github.com/ClickHouse/ClickHouse/pull/17204) ([filimonov](https://github.com/filimonov)).
+* Fix UBSan report in Poco. This closes [#12719](https://github.com/ClickHouse/ClickHouse/issues/12719). [#16765](https://github.com/ClickHouse/ClickHouse/pull/16765) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Do not instrument 3rd-party libraries with UBSan. [#16764](https://github.com/ClickHouse/ClickHouse/pull/16764) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix UBSan report in cache dictionaries. This closes [#12641](https://github.com/ClickHouse/ClickHouse/issues/12641). [#16763](https://github.com/ClickHouse/ClickHouse/pull/16763) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix UBSan report when trying to convert infinite floating point number to integer. This closes [#14190](https://github.com/ClickHouse/ClickHouse/issues/14190). [#16677](https://github.com/ClickHouse/ClickHouse/pull/16677) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+
+
 ## ClickHouse release 20.11
 
 ### ClickHouse release v20.11.3.3-stable, 2020-11-13
diff --git a/base/common/CMakeLists.txt b/base/common/CMakeLists.txt
index 9e4462c821a..cea52b443dd 100644
--- a/base/common/CMakeLists.txt
+++ b/base/common/CMakeLists.txt
@@ -6,6 +6,7 @@ set (SRCS
     demangle.cpp
     getFQDNOrHostName.cpp
     getMemoryAmount.cpp
+    getPageSize.cpp
     getThreadId.cpp
     JSON.cpp
     LineReader.cpp
diff --git a/base/common/ReadlineLineReader.cpp b/base/common/ReadlineLineReader.cpp
index 4e67ed5e4a3..397a7dd7543 100644
--- a/base/common/ReadlineLineReader.cpp
+++ b/base/common/ReadlineLineReader.cpp
@@ -1,4 +1,5 @@
 #include <common/ReadlineLineReader.h>
+#include <common/errnoToString.h>
 #include <ext/scope_guard.h>
 
 #include <errno.h>
@@ -69,7 +70,7 @@ ReadlineLineReader::ReadlineLineReader(
     {
         int res = read_history(history_file_path.c_str());
         if (res)
-            std::cerr << "Cannot read history from file " + history_file_path + ": "+ strerror(errno) << std::endl;
+            std::cerr << "Cannot read history from file " + history_file_path + ": "+ errnoToString(errno) << std::endl;
     }
 
     /// Added '.' to the default list. Because it is used to separate database and table.
@@ -107,7 +108,7 @@ ReadlineLineReader::ReadlineLineReader(
     };
 
     if (signal(SIGINT, clear_prompt_or_exit) == SIG_ERR)
-        throw std::runtime_error(std::string("Cannot set signal handler for readline: ") + strerror(errno));
+        throw std::runtime_error(std::string("Cannot set signal handler for readline: ") + errnoToString(errno));
 
     rl_variable_bind("completion-ignore-case", "on");
     // TODO: it doesn't work
diff --git a/base/common/ReplxxLineReader.cpp b/base/common/ReplxxLineReader.cpp
index 4eb7b065fe3..111428447af 100644
--- a/base/common/ReplxxLineReader.cpp
+++ b/base/common/ReplxxLineReader.cpp
@@ -47,7 +47,7 @@ ReplxxLineReader::ReplxxLineReader(
             {
                 if (!rx.history_load(history_file_path))
                 {
-                    rx.print("Loading history failed: %s\n", strerror(errno));
+                    rx.print("Loading history failed: %s\n", errnoToString(errno).c_str());
                 }
 
                 if (flock(history_file_fd, LOCK_UN))
@@ -88,7 +88,7 @@ ReplxxLineReader::ReplxxLineReader(
 ReplxxLineReader::~ReplxxLineReader()
 {
     if (close(history_file_fd))
-        rx.print("Close of history file failed: %s\n", strerror(errno));
+        rx.print("Close of history file failed: %s\n", errnoToString(errno).c_str());
 }
 
 LineReader::InputStatus ReplxxLineReader::readOneLine(const String & prompt)
@@ -113,7 +113,7 @@ void ReplxxLineReader::addToHistory(const String & line)
     // and that is why flock() is added here.
     bool locked = false;
     if (flock(history_file_fd, LOCK_EX))
-        rx.print("Lock of history file failed: %s\n", strerror(errno));
+        rx.print("Lock of history file failed: %s\n", errnoToString(errno).c_str());
     else
         locked = true;
 
@@ -121,10 +121,10 @@ void ReplxxLineReader::addToHistory(const String & line)
 
     // flush changes to the disk
     if (!rx.history_save(history_file_path))
-        rx.print("Saving history failed: %s\n", strerror(errno));
+        rx.print("Saving history failed: %s\n", errnoToString(errno).c_str());
 
     if (locked && 0 != flock(history_file_fd, LOCK_UN))
-        rx.print("Unlock of history file failed: %s\n", strerror(errno));
+        rx.print("Unlock of history file failed: %s\n", errnoToString(errno).c_str());
 }
 
 void ReplxxLineReader::enableBracketedPaste()
diff --git a/base/common/getMemoryAmount.cpp b/base/common/getMemoryAmount.cpp
index e7d284354f9..1f1b2b8c5dd 100644
--- a/base/common/getMemoryAmount.cpp
+++ b/base/common/getMemoryAmount.cpp
@@ -1,5 +1,6 @@
 #include <stdexcept>
 #include "common/getMemoryAmount.h"
+#include "common/getPageSize.h"
 
 #include <unistd.h>
 #include <sys/types.h>
@@ -18,7 +19,7 @@ uint64_t getMemoryAmountOrZero()
     if (num_pages <= 0)
         return 0;
 
-    int64_t page_size = sysconf(_SC_PAGESIZE);
+    int64_t page_size = getPageSize();
     if (page_size <= 0)
         return 0;
 
diff --git a/base/common/getPageSize.cpp b/base/common/getPageSize.cpp
new file mode 100644
index 00000000000..e96149cc1fb
--- /dev/null
+++ b/base/common/getPageSize.cpp
@@ -0,0 +1,8 @@
+#include "common/getPageSize.h"
+
+#include <unistd.h>
+
+Int64 getPageSize()
+{
+    return sysconf(_SC_PAGESIZE);
+}
diff --git a/base/common/getPageSize.h b/base/common/getPageSize.h
new file mode 100644
index 00000000000..8a35acbfe9b
--- /dev/null
+++ b/base/common/getPageSize.h
@@ -0,0 +1,6 @@
+#pragma once
+
+#include "common/types.h"
+
+/// Get memory page size
+Int64 getPageSize();
diff --git a/base/common/setTerminalEcho.cpp b/base/common/setTerminalEcho.cpp
index 11f6c1db3f0..658f27705ba 100644
--- a/base/common/setTerminalEcho.cpp
+++ b/base/common/setTerminalEcho.cpp
@@ -1,6 +1,7 @@
 // https://stackoverflow.com/questions/1413445/reading-a-password-from-stdcin
 
 #include <common/setTerminalEcho.h>
+#include <common/errnoToString.h>
 #include <stdexcept>
 #include <cstring>
 #include <string>
@@ -31,7 +32,7 @@ void setTerminalEcho(bool enable)
 #else
     struct termios tty;
     if (tcgetattr(STDIN_FILENO, &tty))
-        throw std::runtime_error(std::string("setTerminalEcho failed get: ") + strerror(errno));
+        throw std::runtime_error(std::string("setTerminalEcho failed get: ") + errnoToString(errno));
     if (!enable)
         tty.c_lflag &= ~ECHO;
     else
@@ -39,6 +40,6 @@ void setTerminalEcho(bool enable)
 
     auto ret = tcsetattr(STDIN_FILENO, TCSANOW, &tty);
     if (ret)
-        throw std::runtime_error(std::string("setTerminalEcho failed set: ") + strerror(errno));
+        throw std::runtime_error(std::string("setTerminalEcho failed set: ") + errnoToString(errno));
 #endif
 }
diff --git a/base/common/ya.make b/base/common/ya.make
index 9b38e3919be..fd88760c933 100644
--- a/base/common/ya.make
+++ b/base/common/ya.make
@@ -47,6 +47,7 @@ SRCS(
     errnoToString.cpp
     getFQDNOrHostName.cpp
     getMemoryAmount.cpp
+    getPageSize.cpp
     getResource.cpp
     getThreadId.cpp
     mremap.cpp
diff --git a/cmake/sanitize.cmake b/cmake/sanitize.cmake
index 7a9e263778e..72bdf27004e 100644
--- a/cmake/sanitize.cmake
+++ b/cmake/sanitize.cmake
@@ -41,9 +41,10 @@ if (SANITIZE)
         if (COMPILER_CLANG)
             set (TSAN_FLAGS "${TSAN_FLAGS} -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/tsan_suppressions.txt")
         else()
-            message (WARNING "TSAN suppressions was not passed to the compiler (since the compiler is not clang)")
-            message (WARNING "Use the following command to pass them manually:")
-            message (WARNING "    export TSAN_OPTIONS=\"$TSAN_OPTIONS suppressions=${CMAKE_SOURCE_DIR}/tests/tsan_suppressions.txt\"")
+            set (MESSAGE "TSAN suppressions was not passed to the compiler (since the compiler is not clang)\n")
+            set (MESSAGE "${MESSAGE}Use the following command to pass them manually:\n")
+            set (MESSAGE "${MESSAGE}    export TSAN_OPTIONS=\"$TSAN_OPTIONS suppressions=${CMAKE_SOURCE_DIR}/tests/tsan_suppressions.txt\"")
+            message (WARNING "${MESSAGE}")
         endif()
 
 
@@ -57,8 +58,18 @@ if (SANITIZE)
         endif ()
 
     elseif (SANITIZE STREQUAL "undefined")
-        set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SAN_FLAGS} -fsanitize=undefined -fno-sanitize-recover=all -fno-sanitize=float-divide-by-zero -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt")
-        set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SAN_FLAGS} -fsanitize=undefined -fno-sanitize-recover=all -fno-sanitize=float-divide-by-zero -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt")
+        set (UBSAN_FLAGS "-fsanitize=undefined -fno-sanitize-recover=all -fno-sanitize=float-divide-by-zero")
+        if (COMPILER_CLANG)
+            set (UBSAN_FLAGS "${UBSAN_FLAGS} -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt")
+        else()
+            set (MESSAGE "UBSAN suppressions was not passed to the compiler (since the compiler is not clang)\n")
+            set (MESSAGE "${MESSAGE}Use the following command to pass them manually:\n")
+            set (MESSAGE "${MESSAGE}        export UBSAN_OPTIONS=\"$UBSAN_OPTIONS suppressions=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt\"")
+            message (WARNING "${MESSAGE}")
+        endif()
+
+        set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SAN_FLAGS} ${UBSAN_FLAGS}")
+        set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SAN_FLAGS} ${UBSAN_FLAGS}")
         if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
             set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fsanitize=undefined")
         endif()
diff --git a/contrib/boost b/contrib/boost
index a7ceabe4747..0b98b443aa7 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit a7ceabe4747ecc3309dd3dcd9de4b29660dfd298
+Subproject commit 0b98b443aa7bb77d65efd7b23b3b8c8a0ab5f1f3
diff --git a/contrib/krb5 b/contrib/krb5
index 99f7ad2831a..90ff6f4f8c6 160000
--- a/contrib/krb5
+++ b/contrib/krb5
@@ -1 +1 @@
-Subproject commit 99f7ad2831a01f264c07eed42a0a3a9336b86184
+Subproject commit 90ff6f4f8c695d6bf1aaba78a9b8942be92141c2
diff --git a/contrib/libc-headers b/contrib/libc-headers
index 92c74f938cf..a720b7105a6 160000
--- a/contrib/libc-headers
+++ b/contrib/libc-headers
@@ -1 +1 @@
-Subproject commit 92c74f938cf2c4dd529cae4f3d2923d153b029a7
+Subproject commit a720b7105a610acbd7427eea475a5b6810c151eb
diff --git a/contrib/libgsasl b/contrib/libgsasl
index 140fb582505..383ee28e82f 160000
--- a/contrib/libgsasl
+++ b/contrib/libgsasl
@@ -1 +1 @@
-Subproject commit 140fb58250588c8323285b75fcf127c4adc33dfa
+Subproject commit 383ee28e82f69fa16ed43b48bd9c8ee5b313ab84
diff --git a/contrib/libhdfs3 b/contrib/libhdfs3
index 30552ac527f..095b9d48b40 160000
--- a/contrib/libhdfs3
+++ b/contrib/libhdfs3
@@ -1 +1 @@
-Subproject commit 30552ac527f2c14070d834e171493b2e7f662375
+Subproject commit 095b9d48b400abb72d967cb0539af13b1e3d90cf
diff --git a/contrib/libhdfs3-cmake/CMakeLists.txt b/contrib/libhdfs3-cmake/CMakeLists.txt
index 3aa5cb970db..60f4376bdea 100644
--- a/contrib/libhdfs3-cmake/CMakeLists.txt
+++ b/contrib/libhdfs3-cmake/CMakeLists.txt
@@ -17,7 +17,12 @@ if (NOT USE_INTERNAL_PROTOBUF_LIBRARY AND PROTOBUF_OLD_ABI_COMPAT)
     endif ()
 endif()
 
-set(WITH_KERBEROS false)
+if (${ENABLE_LIBRARIES} AND ${ENABLE_KRB5})
+    SET(WITH_KERBEROS 1)
+else()
+    SET(WITH_KERBEROS 0)
+endif()
+
 # project and source dir
 set(HDFS3_ROOT_DIR ${ClickHouse_SOURCE_DIR}/contrib/libhdfs3)
 set(HDFS3_SOURCE_DIR ${HDFS3_ROOT_DIR}/src)
@@ -28,11 +33,6 @@ set(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMake" ${CMAKE_MODULE_PATH})
 include(Platform)
 include(Options)
 
-# prefer shared libraries
-if (WITH_KERBEROS)
-    find_package(KERBEROS REQUIRED)
-endif()
-
 # source
 set(PROTO_FILES
     #${HDFS3_SOURCE_DIR}/proto/encryption.proto
@@ -207,14 +207,11 @@ target_include_directories(hdfs3 PRIVATE ${HDFS3_COMMON_DIR})
 target_include_directories(hdfs3 PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 
 target_include_directories(hdfs3 PRIVATE ${LIBGSASL_INCLUDE_DIR})
-if (WITH_KERBEROS)
-    target_include_directories(hdfs3 PRIVATE ${KERBEROS_INCLUDE_DIRS})
-endif()
 target_include_directories(hdfs3 PRIVATE ${LIBXML2_INCLUDE_DIR})
 
 target_link_libraries(hdfs3 PRIVATE ${LIBGSASL_LIBRARY})
 if (WITH_KERBEROS)
-    target_link_libraries(hdfs3 PRIVATE ${KERBEROS_LIBRARIES})
+    target_link_libraries(hdfs3 PRIVATE ${KRB5_LIBRARY})
 endif()
 target_link_libraries(hdfs3 PRIVATE ${LIBXML2_LIBRARIES})
 
diff --git a/contrib/librdkafka b/contrib/librdkafka
index 9902bc4fb18..f2f6616419d 160000
--- a/contrib/librdkafka
+++ b/contrib/librdkafka
@@ -1 +1 @@
-Subproject commit 9902bc4fb18bb441fa55ca154b341cdda191e5d3
+Subproject commit f2f6616419d567c9198aef0d1133a2e9b4f02276
diff --git a/contrib/libunwind b/contrib/libunwind
index 51b84d9b6d2..8fe25d7dc70 160000
--- a/contrib/libunwind
+++ b/contrib/libunwind
@@ -1 +1 @@
-Subproject commit 51b84d9b6d2548f1cbdcafe622d5a753853b6149
+Subproject commit 8fe25d7dc70f2a4ea38c3e5a33fa9d4199b67a5a
diff --git a/debian/control b/debian/control
index 9b34e982698..809b5bb5d2a 100644
--- a/debian/control
+++ b/debian/control
@@ -40,7 +40,7 @@ Description: Common files for ClickHouse
 Package: clickhouse-server
 Architecture: all
 Depends: ${shlibs:Depends}, ${misc:Depends}, clickhouse-common-static (= ${binary:Version}), adduser
-Recommends: libcap2-bin
+Recommends: libcap2-bin, krb5-user
 Replaces: clickhouse-server-common, clickhouse-server-base
 Provides: clickhouse-server-common
 Description: Server binary for ClickHouse
diff --git a/docker/images.json b/docker/images.json
index f5b10a14313..5657a748579 100644
--- a/docker/images.json
+++ b/docker/images.json
@@ -58,8 +58,7 @@
     "docker/test/stateless": {
         "name": "yandex/clickhouse-stateless-test",
         "dependent": [
-            "docker/test/stateful",
-            "docker/test/stateful_with_coverage"
+            "docker/test/stateful"
         ]
     },
     "docker/test/stateless_pytest": {
@@ -68,7 +67,9 @@
     },
     "docker/test/stateless_with_coverage": {
         "name": "yandex/clickhouse-stateless-test-with-coverage",
-        "dependent": []
+        "dependent": [
+            "docker/test/stateful_with_coverage"
+        ]
     },
     "docker/test/unit": {
         "name": "yandex/clickhouse-unit-test",
@@ -157,5 +158,9 @@
          "name": "yandex/clickhouse-stateless-unbundled-test",
          "dependent": [
          ]
+    },
+    "docker/test/integration/kerberized_hadoop": {
+        "name": "yandex/clickhouse-kerberized-hadoop",
+        "dependent": []
     }
 }
diff --git a/docker/packager/binary/build.sh b/docker/packager/binary/build.sh
index 4ba2f5f87c7..c46e3b23994 100755
--- a/docker/packager/binary/build.sh
+++ b/docker/packager/binary/build.sh
@@ -31,6 +31,7 @@ find . -name '*.so.*' -print -exec mv '{}' /output \;
 if [ "performance" == "$COMBINED_OUTPUT" ]
 then
     cp -r ../tests/performance /output
+    cp -r ../tests/config/top_level_domains  /
     cp -r ../docker/test/performance-comparison/config /output ||:
     rm /output/unit_tests_dbms ||:
     rm /output/clickhouse-odbc-bridge ||:
diff --git a/docker/packager/packager b/docker/packager/packager
index 6d075195003..65c03cc10e3 100755
--- a/docker/packager/packager
+++ b/docker/packager/packager
@@ -148,6 +148,10 @@ def parse_env_variables(build_type, compiler, sanitizer, package_type, image_typ
 
     if split_binary:
         cmake_flags.append('-DUSE_STATIC_LIBRARIES=0 -DSPLIT_SHARED_LIBRARIES=1 -DCLICKHOUSE_SPLIT_BINARY=1')
+        # We can't always build utils because it requires too much space, but
+        # we have to build them at least in some way in CI. The split build is
+        # probably the least heavy disk-wise.
+        cmake_flags.append('-DENABLE_UTILS=1')
 
     if clang_tidy:
         cmake_flags.append('-DENABLE_CLANG_TIDY=1')
diff --git a/docker/server/README.md b/docker/server/README.md
index e8e8d326de7..d8e9204dffa 100644
--- a/docker/server/README.md
+++ b/docker/server/README.md
@@ -15,6 +15,8 @@ For more information and documentation see https://clickhouse.yandex/.
 $ docker run -d --name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
 ```
 
+By default ClickHouse will be accessible only via docker network. See the [networking section below](#networking).
+
 ### connect to it from a native client
 ```bash
 $ docker run -it --rm --link some-clickhouse-server:clickhouse-server yandex/clickhouse-client --host clickhouse-server
@@ -22,6 +24,70 @@ $ docker run -it --rm --link some-clickhouse-server:clickhouse-server yandex/cli
 
 More information about [ClickHouse client](https://clickhouse.yandex/docs/en/interfaces/cli/).
 
+### connect to it using curl
+
+```bash
+echo "SELECT 'Hello, ClickHouse!'" | docker run -i --rm --link some-clickhouse-server:clickhouse-server curlimages/curl 'http://clickhouse-server:8123/?query=' -s --data-binary @-
+```
+More information about [ClickHouse HTTP Interface](https://clickhouse.tech/docs/en/interfaces/http/).
+
+### stopping / removing the containter
+
+```bash
+$ docker stop some-clickhouse-server
+$ docker rm some-clickhouse-server
+```
+
+### networking
+
+You can expose you ClickHouse running in docker by [mapping particular port](https://docs.docker.com/config/containers/container-networking/) from inside container to a host ports:
+
+```bash
+$ docker run -d -p 18123:8123 -p19000:9000 --name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+$ echo 'SELECT version()' | curl 'http://localhost:18123/' --data-binary @-
+20.12.3.3
+```
+
+or by allowing container to use [host ports directly](https://docs.docker.com/network/host/) using `--network=host` (also allows archiving better network performance):
+
+```bash
+$ docker run -d --network=host --name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+$ echo 'SELECT version()' | curl 'http://localhost:8123/' --data-binary @-
+20.12.3.3
+```
+
+### Volumes 
+
+Typically you may want to mount the following folders inside your container to archieve persistency:
+
+* `/var/lib/clickhouse/` - main folder where ClickHouse stores the data
+* `/val/log/clickhouse-server/` - logs
+
+```bash
+$ docker run -d \
+	-v $(realpath ./ch_data):/var/lib/clickhouse/ \
+	-v $(realpath ./ch_logs):/var/log/clickhouse-server/ \
+	--name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+```
+
+You may also want to mount:
+
+* `/etc/clickhouse-server/config.d/*.xml` - files with server configuration adjustmenets
+* `/etc/clickhouse-server/usert.d/*.xml` - files with use settings adjustmenets
+* `/docker-entrypoint-initdb.d/` - folder with database initialization scripts (see below).
+
+### Linux capabilities 
+
+ClickHouse has some advanced functionality which requite enabling several [linux capabilities](https://man7.org/linux/man-pages/man7/capabilities.7.html).
+
+It is optional and can be enabled using the following [docker command line agruments](https://docs.docker.com/engine/reference/run/#runtime-privilege-and-linux-capabilities):
+
+```bash
+$ docker run -d \
+	--cap-add=SYS_NICE --cap-add=NET_ADMIN --cap-add=IPC_LOCK \
+	--name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+```
+
 ## Configuration
 
 Container exposes 8123 port for [HTTP interface](https://clickhouse.yandex/docs/en/interfaces/http_interface/) and 9000 port for [native client](https://clickhouse.yandex/docs/en/interfaces/tcp/).
diff --git a/docker/test/base/Dockerfile b/docker/test/base/Dockerfile
index 61a40673a96..e8653c2122e 100644
--- a/docker/test/base/Dockerfile
+++ b/docker/test/base/Dockerfile
@@ -1,5 +1,5 @@
 # docker build -t yandex/clickhouse-test-base .
-FROM ubuntu:19.10
+FROM ubuntu:20.04
 
 ENV DEBIAN_FRONTEND=noninteractive LLVM_VERSION=11
 
diff --git a/docker/test/integration/kerberized_hadoop/Dockerfile b/docker/test/integration/kerberized_hadoop/Dockerfile
new file mode 100644
index 00000000000..6a2fd96e7a7
--- /dev/null
+++ b/docker/test/integration/kerberized_hadoop/Dockerfile
@@ -0,0 +1,18 @@
+# docker build -t yandex/clickhouse-kerberized-hadoop .
+
+FROM sequenceiq/hadoop-docker:2.7.0
+RUN sed -i -e 's/^\#baseurl/baseurl/' /etc/yum.repos.d/CentOS-Base.repo
+RUN sed -i -e 's/^mirrorlist/#mirrorlist/' /etc/yum.repos.d/CentOS-Base.repo
+RUN sed -i -e 's#http://mirror.centos.org/#http://vault.centos.org/#' /etc/yum.repos.d/CentOS-Base.repo
+RUN yum clean all && \
+	rpm --rebuilddb &&	\
+	yum -y update && \
+	yum -y install yum-plugin-ovl && \
+	yum --quiet -y install krb5-workstation.x86_64
+RUN cd /tmp && \
+	curl http://archive.apache.org/dist/commons/daemon/source/commons-daemon-1.0.15-src.tar.gz   -o  commons-daemon-1.0.15-src.tar.gz && \
+	tar xzf commons-daemon-1.0.15-src.tar.gz && \
+	cd commons-daemon-1.0.15-src/src/native/unix && \
+	./configure && \
+	make && \
+	cp ./jsvc /usr/local/hadoop/sbin
diff --git a/docker/test/integration/runner/Dockerfile b/docker/test/integration/runner/Dockerfile
index 36188fc4a63..9b51891ccf5 100644
--- a/docker/test/integration/runner/Dockerfile
+++ b/docker/test/integration/runner/Dockerfile
@@ -29,6 +29,8 @@ RUN apt-get update \
     libcurl4-openssl-dev \
     gdb \
     software-properties-common \
+    libkrb5-dev \
+    krb5-user \
     && rm -rf \
         /var/lib/apt/lists/* \
         /var/cache/debconf \
@@ -75,7 +77,8 @@ RUN python3 -m pip install \
     pytest-timeout \
     redis \
     tzlocal \
-    urllib3
+    urllib3 \
+    requests-kerberos
 
 COPY modprobe.sh /usr/local/bin/modprobe
 COPY dockerd-entrypoint.sh /usr/local/bin/
diff --git a/docker/test/integration/runner/compose/docker_compose_hdfs.yml b/docker/test/integration/runner/compose/docker_compose_hdfs.yml
index d19aacfc8d4..b8cd7f64273 100644
--- a/docker/test/integration/runner/compose/docker_compose_hdfs.yml
+++ b/docker/test/integration/runner/compose/docker_compose_hdfs.yml
@@ -2,6 +2,7 @@ version: '2.3'
 services:
     hdfs1:
         image: sequenceiq/hadoop-docker:2.7.0
+        hostname: hdfs1
         restart: always
         ports:
           - 50075:50075
diff --git a/docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml b/docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml
new file mode 100644
index 00000000000..a74476613f3
--- /dev/null
+++ b/docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml
@@ -0,0 +1,29 @@
+version: '2.3'
+
+services:
+  kerberizedhdfs1:
+    cap_add:
+      - DAC_READ_SEARCH
+    image: yandex/clickhouse-kerberized-hadoop:16621
+    hostname: kerberizedhdfs1
+    restart: always
+    volumes:
+        - ${KERBERIZED_HDFS_DIR}/../../hdfs_configs/bootstrap.sh:/etc/bootstrap.sh:ro
+        - ${KERBERIZED_HDFS_DIR}/secrets:/usr/local/hadoop/etc/hadoop/conf
+        - ${KERBERIZED_HDFS_DIR}/secrets/krb_long.conf:/etc/krb5.conf:ro
+    ports:
+      - 1006:1006
+      - 50070:50070
+      - 9000:9000
+    depends_on:
+        - hdfskerberos
+    entrypoint: /etc/bootstrap.sh -d
+
+  hdfskerberos:
+    image: yandex/clickhouse-kerberos-kdc:${DOCKER_KERBEROS_KDC_TAG}
+    hostname: hdfskerberos
+    volumes:
+        - ${KERBERIZED_HDFS_DIR}/secrets:/tmp/keytab
+        - ${KERBERIZED_HDFS_DIR}/../../kerberos_image_config.sh:/config.sh
+        - /dev/urandom:/dev/random
+    ports: [88, 749]
diff --git a/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml b/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml
index 81dab1a48b0..ee2006201b0 100644
--- a/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml
+++ b/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml
@@ -3,6 +3,7 @@
     <mysql_port remove="remove"/>
     <interserver_http_port remove="remove"/>
     <tcp_with_proxy_port remove="remove"/>
+    <test_keeper_server remove="remove"/>
     <listen_host>::</listen_host>
 
     <logger>
diff --git a/docker/test/stateful_with_coverage/Dockerfile b/docker/test/stateful_with_coverage/Dockerfile
index ac6645b9463..07acf4ed4e6 100644
--- a/docker/test/stateful_with_coverage/Dockerfile
+++ b/docker/test/stateful_with_coverage/Dockerfile
@@ -1,8 +1,6 @@
 # docker build -t yandex/clickhouse-stateful-test-with-coverage .
 FROM yandex/clickhouse-stateless-test-with-coverage
 
-RUN echo "deb [trusted=yes] http://apt.llvm.org/bionic/ llvm-toolchain-bionic-9 main" >> /etc/apt/sources.list
-
 RUN apt-get update -y \
     && env DEBIAN_FRONTEND=noninteractive \
         apt-get install --yes --no-install-recommends \
diff --git a/docs/en/engines/table-engines/integrations/hdfs.md b/docs/en/engines/table-engines/integrations/hdfs.md
index f36a4771222..5c36e3f1c21 100644
--- a/docs/en/engines/table-engines/integrations/hdfs.md
+++ b/docs/en/engines/table-engines/integrations/hdfs.md
@@ -108,6 +108,95 @@ Create table with files named `file000`, `file001`, … , `file999`:
 ``` sql
 CREATE TABLE big_table (name String, value UInt32) ENGINE = HDFS('hdfs://hdfs1:9000/big_dir/file{0..9}{0..9}{0..9}', 'CSV')
 ```
+## Configuration {#configuration}
+
+Similar to GraphiteMergeTree, the HDFS engine supports extended configuration using the ClickHouse config file. There are two configuration keys that you can use: global (`hdfs`) and user-level (`hdfs_*`). The global configuration is applied first, and then the user-level configuration is applied (if it exists).
+
+``` xml
+  <!-- Global configuration options for HDFS engine type -->
+  <hdfs>
+	<hadoop_kerberos_keytab>/tmp/keytab/clickhouse.keytab</hadoop_kerberos_keytab>
+	<hadoop_kerberos_principal>clickuser@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+	<hadoop_security_authentication>kerberos</hadoop_security_authentication>
+  </hdfs>
+
+  <!-- Configuration specific for user "root" -->
+  <hdfs_root>
+	<hadoop_kerberos_principal>root@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+  </hdfs_root>
+```
+
+### List of possible configuration options with default values
+#### Supported by libhdfs3
+
+
+| **parameter**                                         | **default value**       |
+| rpc\_client\_connect\_tcpnodelay                      | true                    |
+| dfs\_client\_read\_shortcircuit                       | true                    |
+| output\_replace-datanode-on-failure                   | true                    |
+| input\_notretry-another-node                          | false                   |
+| input\_localread\_mappedfile                          | true                    |
+| dfs\_client\_use\_legacy\_blockreader\_local          | false                   |
+| rpc\_client\_ping\_interval                           | 10  * 1000              |
+| rpc\_client\_connect\_timeout                         | 600 * 1000              |
+| rpc\_client\_read\_timeout                            | 3600 * 1000             |
+| rpc\_client\_write\_timeout                           | 3600 * 1000             |
+| rpc\_client\_socekt\_linger\_timeout                  | -1                      |
+| rpc\_client\_connect\_retry                           | 10                      |
+| rpc\_client\_timeout                                  | 3600 * 1000             |
+| dfs\_default\_replica                                 | 3                       |
+| input\_connect\_timeout                               | 600 * 1000              |
+| input\_read\_timeout                                  | 3600 * 1000             |
+| input\_write\_timeout                                 | 3600 * 1000             |
+| input\_localread\_default\_buffersize                 | 1 * 1024 * 1024         |
+| dfs\_prefetchsize                                     | 10                      |
+| input\_read\_getblockinfo\_retry                      | 3                       |
+| input\_localread\_blockinfo\_cachesize                | 1000                    |
+| input\_read\_max\_retry                               | 60                      |
+| output\_default\_chunksize                            | 512                     |
+| output\_default\_packetsize                           | 64 * 1024               |
+| output\_default\_write\_retry                         | 10                      |
+| output\_connect\_timeout                              | 600 * 1000              |
+| output\_read\_timeout                                 | 3600 * 1000             |
+| output\_write\_timeout                                | 3600 * 1000             |
+| output\_close\_timeout                                | 3600 * 1000             |
+| output\_packetpool\_size                              | 1024                    |
+| output\_heeartbeat\_interval                          | 10 * 1000               |
+| dfs\_client\_failover\_max\_attempts                  | 15                      |
+| dfs\_client\_read\_shortcircuit\_streams\_cache\_size | 256                     |
+| dfs\_client\_socketcache\_expiryMsec                  | 3000                    |
+| dfs\_client\_socketcache\_capacity                    | 16                      |
+| dfs\_default\_blocksize                               | 64 * 1024 * 1024        |
+| dfs\_default\_uri                                     | "hdfs://localhost:9000" |
+| hadoop\_security\_authentication                      | "simple"                |
+| hadoop\_security\_kerberos\_ticket\_cache\_path       | ""                      |
+| dfs\_client\_log\_severity                            | "INFO"                  |
+| dfs\_domain\_socket\_path                             | ""                      |
+
+
+[HDFS Configuration Reference ](https://hawq.apache.org/docs/userguide/2.3.0.0-incubating/reference/HDFSConfigurationParameterReference.html) might explain some parameters.
+
+
+#### ClickHouse extras {#clickhouse-extras}
+
+| **parameter**                                         | **default value**       |
+|hadoop\_kerberos\_keytab                               | ""                      |
+|hadoop\_kerberos\_principal                            | ""                      |
+|hadoop\_kerberos\_kinit\_command                       | kinit                   |
+
+#### Limitations {#limitations}
+
+  * hadoop\_security\_kerberos\_ticket\_cache\_path can be global only, not user specific
+
+## Kerberos support {#kerberos-support}
+
+If hadoop\_security\_authentication parameter has value 'kerberos', ClickHouse authentifies via Kerberos facility.
+Parameters [here](#clickhouse-extras) and hadoop\_security\_kerberos\_ticket\_cache\_path may be of help.
+Note that due to libhdfs3 limitations only old-fashioned approach is supported,
+datanode communications are not secured by SASL (HADOOP\_SECURE\_DN\_USER is a reliable indicator of such
+security approach). Use tests/integration/test\_storage\_kerberized\_hdfs/hdfs_configs/bootstrap.sh for reference.
+
+If hadoop\_kerberos\_keytab, hadoop\_kerberos\_principal or hadoop\_kerberos\_kinit\_command is specified, kinit will be invoked. hadoop\_kerberos\_keytab and hadoop\_kerberos\_principal are mandatory in this case. kinit tool and krb5 configuration files are required.
 
 ## Virtual Columns {#virtual-columns}
 
diff --git a/docs/en/interfaces/cli.md b/docs/en/interfaces/cli.md
index a2ea6edf24a..7700ed36ca5 100644
--- a/docs/en/interfaces/cli.md
+++ b/docs/en/interfaces/cli.md
@@ -57,7 +57,7 @@ The command line is based on ‘replxx’ (similar to ‘readline’). In other
 
 By default, the format used is PrettyCompact. You can change the format in the FORMAT clause of the query, or by specifying `\G` at the end of the query, using the `--format` or `--vertical` argument in the command line, or using the client configuration file.
 
-To exit the client, press Ctrl+D (or Ctrl+C), or enter one of the following instead of a query: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+To exit the client, press Ctrl+D, or enter one of the following instead of a query: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 When processing a query, the client shows:
 
diff --git a/docs/en/introduction/performance.md b/docs/en/introduction/performance.md
index d2780aedccb..6e5710347a1 100644
--- a/docs/en/introduction/performance.md
+++ b/docs/en/introduction/performance.md
@@ -25,6 +25,6 @@ Under the same conditions, ClickHouse can handle several hundred queries per sec
 
 ## Performance When Inserting Data {#performance-when-inserting-data}
 
-We recommend inserting data in packets of at least 1000 rows, or no more than a single request per second. When inserting to a MergeTree table from a tab-separated dump, the insertion speed can be from 50 to 200 MB/s. If the inserted rows are around 1 Kb in size, the speed will be from 50,000 to 200,000 rows per second. If the rows are small, the performance can be higher in rows per second (on Banner System data -`>` 500,000 rows per second; on Graphite data -`>` 1,000,000 rows per second). To improve performance, you can make multiple INSERT queries in parallel, which scales linearly.
+We recommend inserting data in packets of at least 1000 rows, or no more than a single request per second. When inserting to a MergeTree table from a tab-separated dump, the insertion speed can be from 50 to 200 MB/s. If the inserted rows are around 1 KB in size, the speed will be from 50,000 to 200,000 rows per second. If the rows are small, the performance can be higher in rows per second (on Banner System data -`>` 500,000 rows per second; on Graphite data -`>` 1,000,000 rows per second). To improve performance, you can make multiple INSERT queries in parallel, which scales linearly.
 
 {## [Original article](https://clickhouse.tech/docs/en/introduction/performance/) ##}
diff --git a/docs/en/operations/system-tables/distribution_queue.md b/docs/en/operations/system-tables/distribution_queue.md
index 39d06c49e53..9068e7e860c 100644
--- a/docs/en/operations/system-tables/distribution_queue.md
+++ b/docs/en/operations/system-tables/distribution_queue.md
@@ -39,7 +39,7 @@ data_compressed_bytes: 499
 last_exception:        
 ```
 
-**See also**
+**See Also**
 
 -   [Distributed table engine](../../engines/table-engines/special/distributed.md)
 
diff --git a/docs/en/operations/system-tables/replication_queue.md b/docs/en/operations/system-tables/replication_queue.md
new file mode 100644
index 00000000000..aa379caa46c
--- /dev/null
+++ b/docs/en/operations/system-tables/replication_queue.md
@@ -0,0 +1,81 @@
+# system.replication_queue {#system_tables-replication_queue}
+
+Contains information about tasks from replication queues stored in ZooKeeper for tables in the `ReplicatedMergeTree` family.
+
+Columns:
+
+-   `database` ([String](../../sql-reference/data-types/string.md)) — Name of the database.
+
+-   `table` ([String](../../sql-reference/data-types/string.md)) — Name of the table.
+
+-   `replica_name` ([String](../../sql-reference/data-types/string.md)) — Replica name in ZooKeeper. Different replicas of the same table have different names.
+
+-   `position` ([UInt32](../../sql-reference/data-types/int-uint.md)) — Position of the task in the queue.
+
+-   `node_name` ([String](../../sql-reference/data-types/string.md)) — Node name in ZooKeeper.
+
+-   `type` ([String](../../sql-reference/data-types/string.md)) — Type of the task in the queue: `GET_PARTS`, `MERGE_PARTS`, `DETACH_PARTS`, `DROP_PARTS`, or `MUTATE_PARTS`.
+
+-   `create_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — Date and time when the task was submitted for execution.
+
+-   `required_quorum` ([UInt32](../../sql-reference/data-types/int-uint.md)) — The number of replicas waiting for the task to complete with confirmation of completion. This column is only relevant for the `GET_PARTS` task.
+
+-   `source_replica` ([String](../../sql-reference/data-types/string.md)) — Name of the source replica.
+
+-   `new_part_name` ([String](../../sql-reference/data-types/string.md)) — Name of the new part.
+
+-   `parts_to_merge` ([Array](../../sql-reference/data-types/array.md) ([String](../../sql-reference/data-types/string.md))) — Names of parts to merge or update.
+
+-   `is_detach` ([UInt8](../../sql-reference/data-types/int-uint.md)) — The flag indicates whether the `DETACH_PARTS` task is in the queue.
+
+-   `is_currently_executing` ([UInt8](../../sql-reference/data-types/int-uint.md)) — The flag indicates whether a specific task is being performed right now.
+
+-   `num_tries` ([UInt32](../../sql-reference/data-types/int-uint.md)) — The number of failed attempts to complete the task.
+
+-   `last_exception` ([String](../../sql-reference/data-types/string.md)) — Text message about the last error that occurred (if any).
+
+-   `last_attempt_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — Date and time when the task was last attempted.
+
+-   `num_postponed` ([UInt32](../../sql-reference/data-types/int-uint.md)) — The number of postponed tasks.
+
+-   `postpone_reason` ([String](../../sql-reference/data-types/string.md)) — The reason why the task was postponed.
+
+-   `last_postpone_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — Date and time when the task was last postponed.
+
+-   `merge_type` ([String](../../sql-reference/data-types/string.md)) — Type of the current merge. Empty if it's a mutation.
+
+**Example**
+
+``` sql
+SELECT * FROM system.replication_queue LIMIT 1 FORMAT Vertical;
+```
+
+``` text
+Row 1:
+──────
+database:               merge
+table:                  visits_v2
+replica_name:           mtgiga001-1t.metrika.yandex.net
+position:               15
+node_name:              queue-0009325559
+type:                   MERGE_PARTS
+create_time:            2020-12-07 14:04:21
+required_quorum:        0
+source_replica:         mtgiga001-1t.metrika.yandex.net
+new_part_name:          20201130_121373_121384_2
+parts_to_merge:         ['20201130_121373_121378_1','20201130_121379_121379_0','20201130_121380_121380_0','20201130_121381_121381_0','20201130_121382_121382_0','20201130_121383_121383_0','20201130_121384_121384_0']
+is_detach:              0
+is_currently_executing: 0
+num_tries:              36
+last_exception:         Code: 226, e.displayText() = DB::Exception: Marks file '/opt/clickhouse/data/merge/visits_v2/tmp_fetch_20201130_121373_121384_2/CounterID.mrk' doesn't exist (version 20.8.7.15 (official build))
+last_attempt_time:      2020-12-08 17:35:54
+num_postponed:          0
+postpone_reason:        
+last_postpone_time:     1970-01-01 03:00:00
+```
+
+**See Also**
+
+-   [Managing ReplicatedMergeTree Tables](../../sql-reference/statements/system.md/#query-language-system-replicated)
+
+[Original article](https://clickhouse.tech/docs/en/operations/system_tables/replication_queue) <!--hide-->
diff --git a/docs/en/operations/tips.md b/docs/en/operations/tips.md
index 024eba1c899..e62dea0b04e 100644
--- a/docs/en/operations/tips.md
+++ b/docs/en/operations/tips.md
@@ -91,6 +91,23 @@ The Linux kernel prior to 3.2 had a multitude of problems with IPv6 implementati
 
 Use at least a 10 GB network, if possible. 1 Gb will also work, but it will be much worse for patching replicas with tens of terabytes of data, or for processing distributed queries with a large amount of intermediate data.
 
+## Hypervisor configuration
+
+If you are using OpenStack, set
+```
+cpu_mode=host-passthrough
+```
+in nova.conf.
+
+If you are using libvirt, set
+```
+<cpu mode='host-passthrough'/>
+```
+in XML configuration.
+
+This is important for ClickHouse to be able to get correct information with `cpuid` instruction.
+Otherwise you may get `Illegal instruction` crashes when hypervisor is run on old CPU models.
+
 ## ZooKeeper {#zookeeper}
 
 You are probably already using ZooKeeper for other purposes. You can use the same installation of ZooKeeper, if it isn’t already overloaded.
diff --git a/docs/en/sql-reference/aggregate-functions/combinators.md b/docs/en/sql-reference/aggregate-functions/combinators.md
index 554969b80a3..e4f7ba0db81 100644
--- a/docs/en/sql-reference/aggregate-functions/combinators.md
+++ b/docs/en/sql-reference/aggregate-functions/combinators.md
@@ -25,6 +25,10 @@ Example 2: `uniqArray(arr)` – Counts the number of unique elements in all ‘a
 
 -If and -Array can be combined. However, ‘Array’ must come first, then ‘If’. Examples: `uniqArrayIf(arr, cond)`, `quantilesTimingArrayIf(level1, level2)(arr, cond)`. Due to this order, the ‘cond’ argument won’t be an array.
 
+## -SimpleState {#agg-functions-combinator-simplestate}
+
+If you apply this combinator, the aggregate function returns the same value but with a different type. This is an `SimpleAggregateFunction(...)` that can be stored in a table to work with [AggregatingMergeTree](../../engines/table-engines/mergetree-family/aggregatingmergetree.md) table engines.
+
 ## -State {#agg-functions-combinator-state}
 
 If you apply this combinator, the aggregate function doesn’t return the resulting value (such as the number of unique values for the [uniq](../../sql-reference/aggregate-functions/reference/uniq.md#agg_function-uniq) function), but an intermediate state of the aggregation (for `uniq`, this is the hash table for calculating the number of unique values). This is an `AggregateFunction(...)` that can be used for further processing or stored in a table to finish aggregating later.
diff --git a/docs/en/sql-reference/functions/array-functions.md b/docs/en/sql-reference/functions/array-functions.md
index abdf778c089..c4b7156ad95 100644
--- a/docs/en/sql-reference/functions/array-functions.md
+++ b/docs/en/sql-reference/functions/array-functions.md
@@ -1288,12 +1288,30 @@ Returns the index of the first element in the `arr1` array for which `func` retu
 
 Note that the `arrayFirstIndex` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You must pass a lambda function to it as the first argument, and it can’t be omitted.
 
+## arrayMin(\[func,\] arr1, …) {#array-min}
+
+Returns the sum of the `func` values. If the function is omitted, it just returns the min of the array elements.
+
+Note that the `arrayMin` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
+
+## arrayMax(\[func,\] arr1, …) {#array-max}
+
+Returns the sum of the `func` values. If the function is omitted, it just returns the min of the array elements.
+
+Note that the `arrayMax` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
+
 ## arraySum(\[func,\] arr1, …) {#array-sum}
 
 Returns the sum of the `func` values. If the function is omitted, it just returns the sum of the array elements.
 
 Note that the `arraySum` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
 
+## arrayAvg(\[func,\] arr1, …) {#array-avg}
+
+Returns the sum of the `func` values. If the function is omitted, it just returns the average of the array elements.
+
+Note that the `arrayAvg` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
+
 ## arrayCumSum(\[func,\] arr1, …) {#arraycumsumfunc-arr1}
 
 Returns an array of partial sums of elements in the source array (a running sum). If the `func` function is specified, then the values of the array elements are converted by this function before summing.
diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index f8458b27a22..6237cd6a976 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -430,6 +430,63 @@ SELECT toTypeName(CAST(x, 'Nullable(UInt16)')) FROM t_null
 
 -   [cast_keep_nullable](../../operations/settings/settings.md#cast_keep_nullable) setting
 
+## accurateCast(x, T) {#type_conversion_function-accurate-cast}
+
+Converts ‘x’ to the ‘t’ data type. The differente from cast(x, T) is that accurateCast
+does not allow overflow of numeric types during cast if type value x does not fit
+bounds of type T.
+
+Example
+``` sql
+SELECT cast(-1, 'UInt8') as uint8; 
+```
+
+
+``` text
+┌─uint8─┐
+│   255 │
+└───────┘
+```
+
+```sql
+SELECT accurateCast(-1, 'UInt8') as uint8;
+```
+
+``` text
+Code: 70. DB::Exception: Received from localhost:9000. DB::Exception: Value in column Int8 cannot be safely converted into type UInt8: While processing accurateCast(-1, 'UInt8') AS uint8.
+
+```
+
+## accurateCastOrNull(x, T) {#type_conversion_function-accurate-cast_or_null}
+
+Converts ‘x’ to the ‘t’ data type. Always returns nullable type and returns NULL 
+if the casted value is not representable in the target type.
+
+Example:
+
+``` sql
+SELECT
+    accurateCastOrNull(-1, 'UInt8') as uint8,
+    accurateCastOrNull(128, 'Int8') as int8,
+    accurateCastOrNull('Test', 'FixedString(2)') as fixed_string
+```
+
+``` text
+┌─uint8─┬─int8─┬─fixed_string─┐
+│  ᴺᵁᴸᴸ │ ᴺᵁᴸᴸ │ ᴺᵁᴸᴸ         │
+└───────┴──────┴──────────────┘┘
+```
+
+``` sql
+SELECT toTypeName(accurateCastOrNull(5, 'UInt8'))
+```
+
+``` text
+┌─toTypeName(accurateCastOrNull(5, 'UInt8'))─┐
+│ Nullable(UInt8)                            │
+└────────────────────────────────────────────┘
+```
+
 ## toInterval(Year\|Quarter\|Month\|Week\|Day\|Hour\|Minute\|Second) {#function-tointerval}
 
 Converts a Number type argument to an [Interval](../../sql-reference/data-types/special-data-types/interval.md) data type.
diff --git a/docs/en/sql-reference/statements/system.md b/docs/en/sql-reference/statements/system.md
index ddba1443d04..bb279703cc2 100644
--- a/docs/en/sql-reference/statements/system.md
+++ b/docs/en/sql-reference/statements/system.md
@@ -204,7 +204,7 @@ SYSTEM STOP MOVES [[db.]merge_tree_family_table_name]
 
 ## Managing ReplicatedMergeTree Tables {#query-language-system-replicated}
 
-ClickHouse can manage background replication related processes in [ReplicatedMergeTree](../../engines/table-engines/mergetree-family/replication/#table_engines-replication) tables.
+ClickHouse can manage background replication related processes in [ReplicatedMergeTree](../../engines/table-engines/mergetree-family/replication.md#table_engines-replication) tables.
 
 ### STOP FETCHES {#query_language-system-stop-fetches}
 
diff --git a/docs/es/interfaces/cli.md b/docs/es/interfaces/cli.md
index 1e4852be53f..395f9831a4e 100644
--- a/docs/es/interfaces/cli.md
+++ b/docs/es/interfaces/cli.md
@@ -59,7 +59,7 @@ La línea de comandos se basa en ‘replxx’ (similar a ‘readline’). En otr
 
 De forma predeterminada, el formato utilizado es PrettyCompact. Puede cambiar el formato en la cláusula FORMAT de la consulta o especificando `\G` al final de la consulta, utilizando el `--format` o `--vertical` en la línea de comandos, o utilizando el archivo de configuración del cliente.
 
-Para salir del cliente, presione Ctrl+D (o Ctrl+C) o introduzca una de las siguientes opciones en lugar de una consulta: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+Para salir del cliente, presione Ctrl+D o introduzca una de las siguientes opciones en lugar de una consulta: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 Al procesar una consulta, el cliente muestra:
 
diff --git a/docs/ja/interfaces/cli.md b/docs/ja/interfaces/cli.md
index 0c10ee902f5..7037fcf5512 100644
--- a/docs/ja/interfaces/cli.md
+++ b/docs/ja/interfaces/cli.md
@@ -59,7 +59,7 @@ $ cat file.csv | clickhouse-client --database=test --query="INSERT INTO test FOR
 
 既定では、使用される形式はPrettyCompactです。 クエリのFORMAT句で書式を変更するか、次のように指定することができます `\G` クエリの最後に、 `--format` または `--vertical` コマンドラインでの引数、またはクライアント構成ファイルの使用。
 
-クライアントを終了するには、Ctrl+D(またはCtrl+C)を押すか、クエリの代わりに次のいずれかを入力します: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+クライアントを終了するには、Ctrl+Dを押すか、クエリの代わりに次のいずれかを入力します: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 が処理クエリー、クライアントを示し:
 
diff --git a/docs/ru/faq/general/index.md b/docs/ru/faq/general/index.md
new file mode 100644
index 00000000000..cf0f8451450
--- /dev/null
+++ b/docs/ru/faq/general/index.md
@@ -0,0 +1,25 @@
+---
+title: General questions about ClickHouse
+toc_hidden_folder: true
+toc_priority: 1
+toc_title: Общие вопросы
+---
+
+# Общие вопросы о ClickHouse {#obshchie-voprosy}
+
+Вопросы:
+
+-   Что такое ClickHouse?
+-   Почему ClickHouse такой быстрый?
+-   Кто пользуется ClickHouse?
+-   Что обозначает название “ClickHouse”?
+-   Что значит “Не тормозит”?
+-   Что такое OLAP?
+-   Что такое колоночная база данных?
+-   [Почему бы не использовать системы типа MapReduce?](mapreduce.md)
+
+!!! info "Если вы не нашли то, что искали:"
+    Загляните в другие категории F.A.Q. или поищите в других разделах документации, ориентируйтесь по оглавлению слева.
+
+{## [Original article](https://clickhouse.tech/docs/ru/faq/general/) ##}
+
diff --git a/docs/ru/faq/general.md b/docs/ru/faq/general/mapreduce.md
similarity index 60%
rename from docs/ru/faq/general.md
rename to docs/ru/faq/general/mapreduce.md
index 470b8dabbf1..645391493c3 100644
--- a/docs/ru/faq/general.md
+++ b/docs/ru/faq/general/mapreduce.md
@@ -1,8 +1,12 @@
-# Общие вопросы {#obshchie-voprosy}
+---
+title: Why not use something like MapReduce?
+toc_hidden: true
+toc_priority: 110
+---
 
 ## Почему бы не использовать системы типа MapReduce? {#pochemu-by-ne-ispolzovat-sistemy-tipa-mapreduce}
 
-Системами типа MapReduce будем называть системы распределённых вычислений, в которых операция reduce сделана на основе распределённой сортировки. Наиболее распространённым opensource решением данного класса является [Apache Hadoop](http://hadoop.apache.org). Яндекс использует собственное решение — YT.
+Системами типа MapReduce будем называть системы распределённых вычислений, в которых операция reduce сделана на основе распределённой сортировки. Наиболее распространённым opensource решением данного класса является [Apache Hadoop](http://hadoop.apache.org). Яндекс использует собственное решение — YT. 
 
 Такие системы не подходят для онлайн запросов в силу слишком большой latency. То есть, не могут быть использованы в качестве бэкенда для веб-интерфейса.
 Такие системы не подходят для обновления данных в реальном времени.
@@ -10,47 +14,3 @@
 Распределённая сортировка является основной причиной тормозов при выполнении несложных map-reduce задач.
 
 Большинство реализаций MapReduce позволяют выполнять произвольный код на кластере. Но для OLAP задач лучше подходит декларативный язык запросов, который позволяет быстро проводить исследования. Для примера, для Hadoop существует Hive и Pig. Также смотрите Cloudera Impala, Shark (устаревший) для Spark, а также Spark SQL, Presto, Apache Drill. Впрочем, производительность при выполнении таких задач является сильно неоптимальной по сравнению со специализированными системами, а сравнительно высокая latency не позволяет использовать эти системы в качестве бэкенда для веб-интерфейса.
-
-## Что делать, если у меня проблема с кодировками при использовании Oracle через ODBC? {#oracle-odbc-encodings}
-
-Если вы используете Oracle через драйвер ODBC в качестве источника внешних словарей, необходимо задать правильное значение для переменной окружения `NLS_LANG` в `/etc/default/clickhouse`. Подробнее читайте в [Oracle NLS_LANG FAQ](https://www.oracle.com/technetwork/products/globalization/nls-lang-099431.html).
-
-**Пример**
-
-``` sql
-NLS_LANG=RUSSIAN_RUSSIA.UTF8
-```
-
-## Как экспортировать данные из ClickHouse в файл? {#how-to-export-to-file}
-
-### Секция INTO OUTFILE {#sektsiia-into-outfile}
-
-Добавьте секцию [INTO OUTFILE](../sql-reference/statements/select/into-outfile.md#into-outfile-clause) к своему запросу.
-
-Например:
-
-``` sql
-SELECT * FROM table INTO OUTFILE 'file'
-```
-
-По умолчанию, для выдачи данных ClickHouse использует формат [TabSeparated](../interfaces/formats.md#tabseparated). Чтобы выбрать [формат данных](../interfaces/formats.md), используйте [секцию FORMAT](../sql-reference/statements/select/format.md#format-clause).
-
-Например:
-
-``` sql
-SELECT * FROM table INTO OUTFILE 'file' FORMAT CSV
-```
-
-### Таблица с движком File {#tablitsa-s-dvizhkom-file}
-
-Смотрите [File](../engines/table-engines/special/file.md).
-
-### Перенаправление в командой строке {#perenapravlenie-v-komandoi-stroke}
-
-``` sql
-$ clickhouse-client --query "SELECT * from table" --format FormatName > result.txt
-```
-
-Смотрите [clickhouse-client](../interfaces/cli.md).
-
-[Оригинальная статья](https://clickhouse.tech/docs/en/faq/general/) <!--hide-->
diff --git a/docs/ru/faq/index.md b/docs/ru/faq/index.md
index d9eccf59c5c..611e604bc8e 100644
--- a/docs/ru/faq/index.md
+++ b/docs/ru/faq/index.md
@@ -4,3 +4,14 @@ toc_hidden: true
 toc_priority: 76
 ---
 
+# Содержание F.A.Q. {#soderzhanie}
+
+В этом разделе документации собрали вопросы о ClickHouse, которые задают чаще всего.
+
+Категории:
+
+-   **[Общие вопросы](../faq/general/index.md)** 
+-   **[Применение](../faq/use-cases/index.md)**
+-   **[Операции](../faq/operations/index.md)**  
+-   **[Интеграция](../faq/integration/index.md)**
+
diff --git a/docs/ru/faq/integration/file-export.md b/docs/ru/faq/integration/file-export.md
new file mode 100644
index 00000000000..d21ffbd3349
--- /dev/null
+++ b/docs/ru/faq/integration/file-export.md
@@ -0,0 +1,37 @@
+---
+title: How do I export data from ClickHouse to a file?
+toc_hidden: true
+toc_priority: 10
+---
+
+## Как экспортировать данные из ClickHouse в файл? {#how-to-export-to-file-rus}
+
+### Секция INTO OUTFILE {#sektsiia-into-outfile-rus}
+
+Добавьте секцию [INTO OUTFILE](../../sql-reference/statements/select/into-outfile.md#into-outfile-clause) к своему запросу.
+
+Например:
+
+``` sql
+SELECT * FROM table INTO OUTFILE 'file'
+```
+
+По умолчанию, для выдачи данных ClickHouse использует формат [TabSeparated](../../interfaces/formats.md#tabseparated). Чтобы выбрать [формат данных](../../interfaces/formats.md), используйте секцию [FORMAT](../../sql-reference/statements/select/format.md#format-clause).
+
+Например:
+
+``` sql
+SELECT * FROM table INTO OUTFILE 'file' FORMAT CSV
+```
+
+## Таблица с движком File {#using-a-file-engine-table}
+
+Смотрите [File](../../engines/table-engines/special/file.md).
+
+## Перенаправление в командой строке {#using-command-line-redirection}
+
+``` bash
+$ clickhouse-client --query "SELECT * from table" --format FormatName > result.txt
+```
+
+Смотрите [clickhouse-client](../../interfaces/cli.md).
diff --git a/docs/ru/faq/integration/index.md b/docs/ru/faq/integration/index.md
new file mode 100644
index 00000000000..59fd84a8139
--- /dev/null
+++ b/docs/ru/faq/integration/index.md
@@ -0,0 +1,19 @@
+---
+title: Questions about integrating ClickHouse and other systems
+toc_hidden_folder: true
+toc_priority: 4
+toc_title: Интеграция
+---
+
+# Вопросы об интеграции ClickHouse с другими системами {#question-about-integrating-clickhouse-and-other-systems-rus}
+
+Вопросы:
+
+-   [Как экспортировать данные из ClickHouse в файл?](file-export.md)
+-   Как импортировать JSON в ClickHouse?
+-   [Что делать, если у меня проблема с кодировками при использовании Oracle через ODBC?](oracle-odbc.md)
+
+!!! info "Если вы не нашли то, что искали"
+    Загляните в другие подразделы F.A.Q. или поищите в остальных разделах документации, ориентируйтесь по оглавлению слева.
+
+{## [Original article](https://clickhouse.tech/docs/ru/faq/integration/) ##}
diff --git a/docs/ru/faq/integration/oracle-odbc.md b/docs/ru/faq/integration/oracle-odbc.md
new file mode 100644
index 00000000000..f41f1ae7fc1
--- /dev/null
+++ b/docs/ru/faq/integration/oracle-odbc.md
@@ -0,0 +1,15 @@
+---
+title: What if I have a problem with encodings when using Oracle via ODBC?
+toc_hidden: true
+toc_priority: 20
+---
+
+## Что делать, если у меня проблема с кодировками при использовании Oracle через ODBC? {#oracle-odbc-encodings-rus}
+
+Если вы используете Oracle через драйвер ODBC в качестве источника внешних словарей, необходимо задать правильное значение для переменной окружения `NLS_LANG` в `/etc/default/clickhouse`. Подробнее читайте в [Oracle NLS_LANG FAQ](https://www.oracle.com/technetwork/products/globalization/nls-lang-099431.html).
+
+**Пример**
+
+``` sql
+NLS_LANG=RUSSIAN_RUSSIA.UTF8
+```
\ No newline at end of file
diff --git a/docs/ru/faq/operations/index.md b/docs/ru/faq/operations/index.md
new file mode 100644
index 00000000000..98a04213fb7
--- /dev/null
+++ b/docs/ru/faq/operations/index.md
@@ -0,0 +1,18 @@
+---
+title: Question about operating ClickHouse servers and clusters
+toc_hidden_folder: true
+toc_priority: 3
+toc_title: Операции
+---
+
+# Вопросы о производительности серверов и кластеров ClickHouse {#voprosy-ob-operating-clickhouse-servers-and-clusters}
+
+Вопросы:
+
+-   Which ClickHouse version to use in production?
+-   Is it possible to delete old records from a ClickHouse table?
+
+    !!! info "Don’t see what you were looking for?"
+        Check out [other F.A.Q. categories](../../faq/index.md) or browse around main documentation articles found in the left sidebar.
+
+    {## [Original article](https://clickhouse.tech/docs/en/faq/production/) ##}
diff --git a/docs/ru/faq/use-cases/index.md b/docs/ru/faq/use-cases/index.md
new file mode 100644
index 00000000000..6e7f2b36fe0
--- /dev/null
+++ b/docs/ru/faq/use-cases/index.md
@@ -0,0 +1,14 @@
+---
+title: Questions about ClickHouse use cases
+toc_hidden_folder: true
+toc_priority: 2
+toc_title: Применение
+---
+
+# Вопросы о применении ClickHouse {#voprosy-o-primenenii}
+
+Вопросы:
+
+-   Can I use ClickHouse as a time-series database?
+-   Can I use ClickHouse as a key-value storage?
+
diff --git a/docs/ru/interfaces/cli.md b/docs/ru/interfaces/cli.md
index 222af33f952..e37018acdfb 100644
--- a/docs/ru/interfaces/cli.md
+++ b/docs/ru/interfaces/cli.md
@@ -63,7 +63,7 @@ $ cat file.csv | clickhouse-client --database=test --query="INSERT INTO test FOR
 
 По умолчанию, в качестве формата, используется формат PrettyCompact (красивые таблички). Вы можете изменить формат с помощью секции FORMAT запроса, или с помощью указания `\G` на конце запроса, с помощью аргумента командной строки `--format` или `--vertical`, или с помощью конфигурационного файла клиента.
 
-Чтобы выйти из клиента, нажмите Ctrl+D (или Ctrl+C), или наберите вместо запроса одно из: «exit», «quit», «logout», «учше», «йгше», «дщпщге», «exit;», «quit;», «logout;», «учшеж», «йгшеж», «дщпщгеж», «q», «й», «q», «Q», «:q», «й», «Й», «Жй»
+Чтобы выйти из клиента, нажмите Ctrl+D, или наберите вместо запроса одно из: «exit», «quit», «logout», «учше», «йгше», «дщпщге», «exit;», «quit;», «logout;», «учшеж», «йгшеж», «дщпщгеж», «q», «й», «q», «Q», «:q», «й», «Й», «Жй»
 
 При выполнении запроса, клиент показывает:
 
diff --git a/docs/ru/operations/system-tables/replication_queue.md b/docs/ru/operations/system-tables/replication_queue.md
new file mode 100644
index 00000000000..47f64aea55d
--- /dev/null
+++ b/docs/ru/operations/system-tables/replication_queue.md
@@ -0,0 +1,81 @@
+# system.replication_queue {#system_tables-replication_queue}
+
+Содержит информацию о задачах из очередей репликации, хранящихся в ZooKeeper, для таблиц семейства `ReplicatedMergeTree`.
+
+Столбцы:
+
+-   `database` ([String](../../sql-reference/data-types/string.md)) — имя базы данных.
+
+-   `table` ([String](../../sql-reference/data-types/string.md)) — имя таблицы.
+
+-   `replica_name` ([String](../../sql-reference/data-types/string.md)) — имя реплики в ZooKeeper. Разные реплики одной и той же таблицы имеют различные имена.
+
+-   `position` ([UInt32](../../sql-reference/data-types/int-uint.md)) — позиция задачи в очереди.
+
+-   `node_name` ([String](../../sql-reference/data-types/string.md)) — имя узла в ZooKeeper.
+
+-   `type` ([String](../../sql-reference/data-types/string.md)) — тип задачи в очереди: `GET_PARTS`, `MERGE_PARTS`, `DETACH_PARTS`, `DROP_PARTS` или `MUTATE_PARTS`.
+
+-   `create_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — дата и время отправки задачи на выполнение.
+
+-   `required_quorum` ([UInt32](../../sql-reference/data-types/int-uint.md)) — количество реплик, ожидающих завершения задачи, с подтверждением о завершении. Этот столбец актуален только для задачи `GET_PARTS`.
+
+-   `source_replica` ([String](../../sql-reference/data-types/string.md)) — имя исходной реплики.
+
+-   `new_part_name` ([String](../../sql-reference/data-types/string.md)) — имя нового куска.
+
+-   `parts_to_merge` ([Array](../../sql-reference/data-types/array.md) ([String](../../sql-reference/data-types/string.md))) — имена кусков, которые требуется смержить или обновить.
+
+-   `is_detach` ([UInt8](../../sql-reference/data-types/int-uint.md)) — флаг, указывающий на присутствие в очереди задачи `DETACH_PARTS`.
+
+-   `is_currently_executing` ([UInt8](../../sql-reference/data-types/int-uint.md)) — флаг, указывающий на выполнение конкретной задачи на данный момент.
+
+-   `num_tries` ([UInt32](../../sql-reference/data-types/int-uint.md)) — количество неудачных попыток выполнить задачу.
+
+-   `last_exception` ([String](../../sql-reference/data-types/string.md)) — текст сообщения о последней возникшей ошибке, если таковые имеются.
+
+-   `last_attempt_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — дата и время последней попытки выполнить задачу.
+
+-   `num_postponed` ([UInt32](../../sql-reference/data-types/int-uint.md)) — количество отложенных задач.
+
+-   `postpone_reason` ([String](../../sql-reference/data-types/string.md)) — причина, по которой была отложена задача.
+
+-   `last_postpone_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — дата и время, когда была отложена задача в последний раз.
+
+-   `merge_type` ([String](../../sql-reference/data-types/string.md)) — тип текущего слияния. Пусто, если это мутация.
+
+**Пример**
+
+``` sql
+SELECT * FROM system.replication_queue LIMIT 1 FORMAT Vertical;
+```
+
+``` text
+Row 1:
+──────
+database:               merge
+table:                  visits_v2
+replica_name:           mtgiga001-1t.metrika.yandex.net
+position:               15
+node_name:              queue-0009325559
+type:                   MERGE_PARTS
+create_time:            2020-12-07 14:04:21
+required_quorum:        0
+source_replica:         mtgiga001-1t.metrika.yandex.net
+new_part_name:          20201130_121373_121384_2
+parts_to_merge:         ['20201130_121373_121378_1','20201130_121379_121379_0','20201130_121380_121380_0','20201130_121381_121381_0','20201130_121382_121382_0','20201130_121383_121383_0','20201130_121384_121384_0']
+is_detach:              0
+is_currently_executing: 0
+num_tries:              36
+last_exception:         Code: 226, e.displayText() = DB::Exception: Marks file '/opt/clickhouse/data/merge/visits_v2/tmp_fetch_20201130_121373_121384_2/CounterID.mrk' doesn't exist (version 20.8.7.15 (official build))
+last_attempt_time:      2020-12-08 17:35:54
+num_postponed:          0
+postpone_reason:        
+last_postpone_time:     1970-01-01 03:00:00
+```
+
+**Смотрите также**
+
+-   [Управление таблицами ReplicatedMergeTree](../../sql-reference/statements/system.md/#query-language-system-replicated)
+
+[Оригинальная статья](https://clickhouse.tech/docs/ru/operations/system_tables/replication_queue) <!--hide-->
diff --git a/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md b/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md
index 0015edfdf72..1367216e401 100644
--- a/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md
+++ b/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md
@@ -199,7 +199,7 @@ SOURCE(ODBC(
 
 ClickHouse получает от ODBC-драйвера информацию о квотировании и квотирует настройки в запросах к драйверу, поэтому имя таблицы нужно указывать в соответствии с регистром имени таблицы в базе данных.
 
-Если у вас есть проблемы с кодировками при использовании Oracle, ознакомьтесь с соответствующим разделом [FAQ](../../../faq/general.md#oracle-odbc-encodings).
+Если у вас есть проблемы с кодировками при использовании Oracle, ознакомьтесь с соответствующим разделом [FAQ](../../../faq/integration/oracle-odbc.md).
 
 ### Выявленная уязвимость в функционировании ODBC словарей {#vyiavlennaia-uiazvimost-v-funktsionirovanii-odbc-slovarei}
 
diff --git a/docs/ru/sql-reference/functions/date-time-functions.md b/docs/ru/sql-reference/functions/date-time-functions.md
index b7a077b3bd6..31482cde77f 100644
--- a/docs/ru/sql-reference/functions/date-time-functions.md
+++ b/docs/ru/sql-reference/functions/date-time-functions.md
@@ -593,6 +593,18 @@ SELECT dateDiff('hour', toDateTime('2018-01-01 22:00:00'), toDateTime('2018-01-0
 Например, `timeSlots(toDateTime('2012-01-01 12:20:00'), toUInt32(600)) = [toDateTime('2012-01-01 12:00:00'), toDateTime('2012-01-01 12:30:00')]`.
 Это нужно для поиска хитов, входящих в соответствующий визит.
 
+## toYYYYMM
+
+Переводит дату или дату со временем в число типа UInt32, содержащее номер года и месяца (YYYY * 100 + MM).
+
+## toYYYYMMDD
+
+Переводит дату или дату со временем в число типа UInt32, содержащее номер года, месяца и дня (YYYY * 10000 + MM * 100 + DD).
+
+## toYYYYMMDDhhmmss
+
+Переводит дату или дату со временем в число типа UInt64 содержащее номер года, месяца, дня и время (YYYY * 10000000000 + MM * 100000000 + DD * 1000000 + hh * 10000 + mm * 100 + ss).
+
 ## formatDateTime {#formatdatetime}
 
 Функция преобразует дату-и-время в строку по заданному шаблону. Важно: шаблон — константное выражение, поэтому использовать разные шаблоны в одной колонке не получится.
diff --git a/docs/tr/interfaces/cli.md b/docs/tr/interfaces/cli.md
index de62b3003af..dc0ca9e7591 100644
--- a/docs/tr/interfaces/cli.md
+++ b/docs/tr/interfaces/cli.md
@@ -59,7 +59,7 @@ Komut satırı dayanmaktadır ‘replxx’ (benzer ‘readline’). Başka bir d
 
 Varsayılan olarak, kullanılan biçim PrettyCompact. Sorgunun biçim yan tümcesinde veya belirterek biçimi değiştirebilirsiniz `\G` sorgunun sonunda, `--format` veya `--vertical` komut satırında veya istemci yapılandırma dosyasını kullanarak bağımsız değişken.
 
-İstemciden çıkmak için Ctrl+D (veya Ctrl+C) tuşlarına basın veya bir sorgu yerine aşağıdakilerden birini girin: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+İstemciden çıkmak için Ctrl+D tuşlarına basın veya bir sorgu yerine aşağıdakilerden birini girin: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 Bir sorguyu işlerken, istemci şunları gösterir:
 
diff --git a/docs/zh/sql-reference/statements/misc.md b/docs/zh/sql-reference/statements/misc.md
index b4297d1ed4f..26b3997a4d6 100644
--- a/docs/zh/sql-reference/statements/misc.md
+++ b/docs/zh/sql-reference/statements/misc.md
@@ -9,19 +9,21 @@ toc_title: "\u5176\u4ED6"
 
 ## ATTACH {#attach}
 
-这个查询是完全一样的 `CREATE`，但是
+与`CREATE`类似，但有所区别
 
--   而不是这个词 `CREATE` 它使用这个词 `ATTACH`.
--   查询不会在磁盘上创建数据，但假定数据已经在适当的位置，只是将有关表的信息添加到服务器。
-    执行附加查询后，服务器将知道表的存在。
+-   使用关键词 `ATTACH`
+-   查询不会在磁盘上创建数据。但会假定数据已经在对应位置存放，同时将与表相关的信息添加到服务器。
+    执行 `ATTACH` 查询后，服务器将知道表已经被创建。
 
-如果表之前已分离 (`DETACH`），意味着其结构是已知的，可以使用速记而不限定该结构。
+如果表之前已分离 (`DETACH`），意味着其结构是已知的，可以使用简要的写法来建立表，即不需要定义表结构的Schema细节。
 
 ``` sql
 ATTACH TABLE [IF NOT EXISTS] [db.]name [ON CLUSTER cluster]
 ```
 
-启动服务器时使用此查询。 服务器将表元数据作为文件存储 `ATTACH` 查询，它只是在启动时运行（除了在服务器上显式创建的系统表）。
+启动服务器时会自动触发此查询。 
+
+服务器将表的元数据作为文件存储 `ATTACH` 查询，它只是在启动时运行。有些表例外，如系统表，它们是在服务器上显式指定的。
 
 ## CHECK TABLE {#check-table}
 
@@ -31,13 +33,12 @@ ATTACH TABLE [IF NOT EXISTS] [db.]name [ON CLUSTER cluster]
 CHECK TABLE [db.]name
 ```
 
-该 `CHECK TABLE` 查询将实际文件大小与存储在服务器上的预期值进行比较。 如果文件大小与存储的值不匹配，则表示数据已损坏。 例如，这可能是由查询执行期间的系统崩溃引起的。
+`CHECK TABLE` 查询会比较存储在服务器上的实际文件大小与预期值。 如果文件大小与存储的值不匹配，则表示数据已损坏。 例如，这可能是由查询执行期间的系统崩溃引起的。
 
-查询响应包含 `result` 具有单行的列。 该行的值为
-[布尔值](../../sql-reference/data-types/boolean.md) 类型:
+查询返回一行结果，列名为 `result`, 该行的值为 [布尔值](../../sql-reference/data-types/boolean.md) 类型:
 
--   0-表中的数据已损坏。
--   1-数据保持完整性。
+-   0-表中的数据已损坏；
+-   1-数据保持完整性；
 
 该 `CHECK TABLE` 查询支持下表引擎:
 
@@ -56,13 +57,14 @@ CHECK TABLE [db.]name
 
 如果表已损坏，则可以将未损坏的数据复制到另一个表。 要做到这一点:
 
-1.  创建一个与损坏的表结构相同的新表。 要做到这一点，请执行查询 `CREATE TABLE <new_table_name> AS <damaged_table_name>`.
+1.  创建一个与损坏的表结构相同的新表。 请执行查询 `CREATE TABLE <new_table_name> AS <damaged_table_name>`.
 2.  将 [max_threads](../../operations/settings/settings.md#settings-max_threads) 值设置为1，以在单个线程中处理下一个查询。 要这样做，请运行查询 `SET max_threads = 1`.
 3.  执行查询 `INSERT INTO <new_table_name> SELECT * FROM <damaged_table_name>`. 此请求将未损坏的数据从损坏的表复制到另一个表。 只有损坏部分之前的数据才会被复制。
 4.  重新启动 `clickhouse-client` 以重置 `max_threads` 值。
 
 ## DESCRIBE TABLE {#misc-describe-table}
 
+查看表的描述信息，返回各列的Schema，语法如下：
 ``` sql
 DESC|DESCRIBE TABLE [db.]table [INTO OUTFILE filename] [FORMAT format]
 ```
@@ -73,24 +75,25 @@ DESC|DESCRIBE TABLE [db.]table [INTO OUTFILE filename] [FORMAT format]
 -   `type`— 列的类型。
 -   `default_type` — [默认表达式](create.md#create-default-values) (`DEFAULT`, `MATERIALIZED` 或 `ALIAS`)中使用的子句。 如果没有指定默认表达式，则列包含一个空字符串。
 -   `default_expression` — `DEFAULT` 子句中指定的值。
--   `comment_expression` — 注释。
+-   `comment_expression` — 注释信息。
 
 嵌套数据结构以 “expanded” 格式输出。 每列分别显示，列名后加点号。
 
 ## DETACH {#detach}
 
-从服务器中删除有关 ‘name’ 表的信息。 服务器停止了解该表的存在。
+从服务器中删除目标表信息（删除对象是表), 执行查询后,服务器视作该表已经不存在。
 
 ``` sql
 DETACH TABLE [IF EXISTS] [db.]name [ON CLUSTER cluster]
 ```
 
 这不会删除表的数据或元数据。 在下一次服务器启动时，服务器将读取元数据并再次查找该表。
-同样，可以使用 `ATTACH` 查询重新连接一个 “detached” 的表（系统表除外，没有为它们存储元数据）。
+也可以不停止服务器的情况下，使用前面介绍的 `ATTACH` 查询来重新关联该表（系统表除外，没有为它们存储元数据）。
 
 ## DROP {#drop}
 
 删除已经存在的实体。如果指定 `IF EXISTS`， 则如果实体不存在，则不返回错误。
+建议使用时添加 `IF EXISTS` 修饰符。
 
 ## DROP DATABASE {#drop-database}
 
@@ -135,7 +138,7 @@ DROP USER [IF EXISTS] name [,...] [ON CLUSTER cluster_name]
 
 删除角色。
 
-已删除的角色将从授予该角色的所有实体撤销。
+同时该角色所拥有的权限也会被收回。
 
 语法:
 
@@ -199,6 +202,8 @@ EXISTS [TEMPORARY] [TABLE|DICTIONARY] [db.]name [INTO OUTFILE filename] [FORMAT
 
 ## KILL QUERY {#kill-query-statement}
 
+
+
 ``` sql
 KILL QUERY [ON CLUSTER cluster]
   WHERE <where expression to SELECT FROM system.processes query>
@@ -219,16 +224,17 @@ KILL QUERY WHERE query_id='2-857d-4a57-9ee0-327da5d60a90'
 KILL QUERY WHERE user='username' SYNC
 ```
 
-只读用户只能停止自己的查询。
+只读用户只能停止自己提交的查询。
 
-默认情况下，使用异步版本的查询 (`ASYNC`），不等待确认查询已停止。
+默认情况下，使用异步版本的查询 (`ASYNC`），不需要等待确认查询已停止。
 
-同步版本 (`SYNC`）等待所有查询停止，并在停止时显示有关每个进程的信息。
-响应包含 `kill_status` 列，该列可以采用以下值:
+而相对的，终止同步版本 (`SYNC`）的查询会显示每步停止时间。
+
+返回信息包含 `kill_status` 列，该列可以采用以下值:
 
 1.  ‘finished’ – 查询已成功终止。
 2.  ‘waiting’ – 发送查询信号终止后，等待查询结束。
-3.  其他值解释为什么查询不能停止。
+3.  其他值，会解释为什么查询不能停止。
 
 测试查询 (`TEST`）仅检查用户的权限，并显示要停止的查询列表。
 
diff --git a/docs/zh/sql-reference/statements/select/limit.md b/docs/zh/sql-reference/statements/select/limit.md
index b079248deca..d6d827552b0 100644
--- a/docs/zh/sql-reference/statements/select/limit.md
+++ b/docs/zh/sql-reference/statements/select/limit.md
@@ -14,7 +14,7 @@ toc_title: LIMIT
 
 ## LIMIT … WITH TIES 修饰符 {#limit-with-ties}
 
-如果为 `LIMIT n[,m]` 设置了 `WITH TIES` ，并且声明了 `ORDER BY expr_list`, you will get in result first `n` or `n,m` rows and all rows with same `ORDER BY` fields values equal to row at position `n` for `LIMIT n` and `m` for `LIMIT n,m`.
+如果为 `LIMIT n[,m]` 设置了 `WITH TIES` ，并且声明了 `ORDER BY expr_list`, 除了得到无修饰符的结果（正常情况下的 `limit n`, 前n行数据), 还会返回与第`n`行具有相同排序字段的行(即如果第n+1行的字段与第n行 拥有相同的排序字段，同样返回该结果.
 
 此修饰符可以与： [ORDER BY … WITH FILL modifier](../../../sql-reference/statements/select/order-by.md#orderby-with-fill) 组合使用.
 
@@ -38,7 +38,7 @@ SELECT * FROM (
 └───┘
 ```
 
-单子执行了 `WITH TIES` 修饰符后
+添加 `WITH TIES` 修饰符后
 
 ``` sql
 SELECT * FROM (
@@ -59,4 +59,8 @@ SELECT * FROM (
 └───┘
 ```
 
-cause row number 6 have same value “2” for field `n` as row number 5
+虽然指定了`LIMIT 5`, 但第6行的`n`字段值为2，与第5行相同，因此也作为满足条件的记录返回。
+简而言之，该修饰符可理解为是否增加“并列行”的数据。
+
+``` sql，
+``` sql
diff --git a/programs/client/Client.cpp b/programs/client/Client.cpp
index 92ae37123fa..9d0227a51e3 100644
--- a/programs/client/Client.cpp
+++ b/programs/client/Client.cpp
@@ -2335,7 +2335,7 @@ public:
                 "Suggestion limit for how many databases, tables and columns to fetch.")
             ("multiline,m", "multiline")
             ("multiquery,n", "multiquery")
-            ("queries-file,qf", po::value<std::string>(), "file path with queries to execute")
+            ("queries-file", po::value<std::string>(), "file path with queries to execute")
             ("format,f", po::value<std::string>(), "default output format")
             ("testmode,T", "enable test hints in comments")
             ("ignore-error", "do not stop processing in multiquery mode")
diff --git a/programs/odbc-bridge/ColumnInfoHandler.cpp b/programs/odbc-bridge/ColumnInfoHandler.cpp
index 9e2307715c5..ee4daa3e16d 100644
--- a/programs/odbc-bridge/ColumnInfoHandler.cpp
+++ b/programs/odbc-bridge/ColumnInfoHandler.cpp
@@ -115,7 +115,7 @@ void ODBCColumnsInfoHandler::handleRequest(Poco::Net::HTTPServerRequest & reques
         std::string name = schema_name.empty() ? backQuoteIfNeed(table_name) : backQuoteIfNeed(schema_name) + "." + backQuoteIfNeed(table_name);
         WriteBufferFromOwnString buf;
         std::string input = "SELECT * FROM " + name + " WHERE 1 = 0";
-        ParserQueryWithOutput parser;
+        ParserQueryWithOutput parser(input.data() + input.size());
         ASTPtr select = parseQuery(parser, input.data(), input.data() + input.size(), "", context_settings.max_query_size, context_settings.max_parser_depth);
 
         IAST::FormatSettings settings(buf, true);
diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index 28d8301b920..1eeec398f44 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -139,6 +139,28 @@ void setupTmpPath(Poco::Logger * log, const std::string & path)
     }
 }
 
+int waitServersToFinish(std::vector<DB::ProtocolServerAdapter> & servers, size_t seconds_to_wait)
+{
+    const int sleep_max_ms = 1000 * seconds_to_wait;
+    const int sleep_one_ms = 100;
+    int sleep_current_ms = 0;
+    int current_connections = 0;
+    while (sleep_current_ms < sleep_max_ms)
+    {
+        current_connections = 0;
+        for (auto & server : servers)
+        {
+            server.stop();
+            current_connections += server.currentConnections();
+        }
+        if (!current_connections)
+            break;
+        sleep_current_ms += sleep_one_ms;
+        std::this_thread::sleep_for(std::chrono::milliseconds(sleep_one_ms));
+    }
+    return current_connections;
+}
+
 }
 
 namespace DB
@@ -366,7 +388,7 @@ int Server::main(const std::vector<std::string> & /*args*/)
     LOG_WARNING(log, "Server was built in debug mode. It will work slowly.");
 #endif
 
-#if defined(ADDRESS_SANITIZER) || defined(THREAD_SANITIZER) || defined(MEMORY_SANITIZER)
+#if defined(SANITIZER)
     LOG_WARNING(log, "Server was built with sanitizer. It will work slowly.");
 #endif
 
@@ -794,8 +816,29 @@ int Server::main(const std::vector<std::string> & /*args*/)
 
         LOG_DEBUG(log, "Shut down storages.");
 
-        for (auto & server : servers_to_start_before_tables)
-            server.stop();
+        if (!servers_to_start_before_tables.empty())
+        {
+            LOG_DEBUG(log, "Waiting for current connections to servers for tables to finish.");
+            int current_connections = 0;
+            for (auto & server : servers_to_start_before_tables)
+            {
+                server.stop();
+                current_connections += server.currentConnections();
+            }
+
+            if (current_connections)
+                LOG_INFO(log, "Closed all listening sockets. Waiting for {} outstanding connections.", current_connections);
+            else
+                LOG_INFO(log, "Closed all listening sockets.");
+
+            if (current_connections > 0)
+                current_connections = waitServersToFinish(servers_to_start_before_tables, config().getInt("shutdown_wait_unfinished", 5));
+
+            if (current_connections)
+                LOG_INFO(log, "Closed connections to servers for tables. But {} remain. Probably some tables of other users cannot finish their connections after context shutdown.", current_connections);
+            else
+                LOG_INFO(log, "Closed connections to servers for tables.");
+        }
 
         /** Explicitly destroy Context. It is more convenient than in destructor of Server, because logger is still available.
           * At this moment, no one could own shared part of Context.
@@ -1167,24 +1210,7 @@ int Server::main(const std::vector<std::string> & /*args*/)
             global_context->getProcessList().killAllQueries();
 
             if (current_connections)
-            {
-                const int sleep_max_ms = 1000 * config().getInt("shutdown_wait_unfinished", 5);
-                const int sleep_one_ms = 100;
-                int sleep_current_ms = 0;
-                while (sleep_current_ms < sleep_max_ms)
-                {
-                    current_connections = 0;
-                    for (auto & server : servers)
-                    {
-                        server.stop();
-                        current_connections += server.currentConnections();
-                    }
-                    if (!current_connections)
-                        break;
-                    sleep_current_ms += sleep_one_ms;
-                    std::this_thread::sleep_for(std::chrono::milliseconds(sleep_one_ms));
-                }
-            }
+                current_connections = waitServersToFinish(servers, config().getInt("shutdown_wait_unfinished", 5));
 
             if (current_connections)
                 LOG_INFO(log, "Closed connections. But {} remain."
diff --git a/programs/server/config.xml b/programs/server/config.xml
index f41c346bbed..e05ef61eed7 100644
--- a/programs/server/config.xml
+++ b/programs/server/config.xml
@@ -48,97 +48,121 @@
         -->
     </logger>
 
-    <send_crash_reports>
-        <!-- Changing <enabled> to true allows sending crash reports to -->
-        <!-- the ClickHouse core developers team via Sentry https://sentry.io -->
-        <!-- Doing so at least in pre-production environments is highly appreciated -->
-        <enabled>false</enabled>
-        <!-- Change <anonymize> to true if you don't feel comfortable attaching the server hostname to the crash report -->
-        <anonymize>false</anonymize>
-        <!-- Default endpoint should be changed to different Sentry DSN only if you have -->
-        <!-- some in-house engineers or hired consultants who're going to debug ClickHouse issues for you -->
-        <endpoint>https://6f33034cfe684dd7a3ab9875e57b1c8d@o388870.ingest.sentry.io/5226277</endpoint>
-    </send_crash_reports>
+    <!-- It is the name that will be shown in the clickhouse-client.
+         By default, anything with "production" will be highlighted in red in query prompt.
+    -->
+    <!--display_name>production</display_name-->
 
-    <!--display_name>production</display_name--> <!-- It is the name that will be shown in the client -->
+    <!-- Port for HTTP API. See also 'https_port' for secure connections.
+         This interface is also used by ODBC and JDBC drivers (DataGrip, Dbeaver, ...)
+         and by most of web interfaces (embedded UI, Grafana, Redash, ...).
+      -->
     <http_port>8123</http_port>
+
+    <!-- Port for interaction by native protocol with:
+         - clickhouse-client and other native ClickHouse tools (clickhouse-benchmark, clickhouse-copier);
+         - clickhouse-server with other clickhouse-servers for distributed query processing;
+         - ClickHouse drivers and applications supporting native protocol
+         (this protocol is also informally called as "the TCP protocol");
+         See also 'tcp_port_secure' for secure connections.
+    -->
     <tcp_port>9000</tcp_port>
+
+    <!-- Compatibility with MySQL protocol.
+         ClickHouse will pretend to be MySQL for applications connecting to this port.
+    -->
     <mysql_port>9004</mysql_port>
 
-    <!-- For HTTPS and SSL over native protocol. -->
-    <!--
-    <https_port>8443</https_port>
-    <tcp_port_secure>9440</tcp_port_secure>
+    <!-- Compatibility with PostgreSQL protocol.
+         ClickHouse will pretend to be PostgreSQL for applications connecting to this port.
     -->
+    <!-- <postgresql_port>9005</postgresql_port> -->
 
-    <!-- TCP with PROXY protocol (PROXY header sent for every connection) -->
-    <!--
-    <tcp_with_proxy_port>9010</tcp_with_proxy_port>
+    <!-- HTTP API with TLS (HTTPS).
+         You have to configure certificate to enable this interface.
+         See the openSSL section below.
     -->
+    <!-- <https_port>8443</https_port> -->
 
-    <!-- Used with https_port and tcp_port_secure. Full ssl options list: https://github.com/ClickHouse-Extras/poco/blob/master/NetSSL_OpenSSL/include/Poco/Net/SSLManager.h#L71 -->
-    <openSSL>
-        <server> <!-- Used for https server AND secure tcp port -->
-            <!-- openssl req -subj "/CN=localhost" -new -newkey rsa:2048 -days 365 -nodes -x509 -keyout /etc/clickhouse-server/server.key -out /etc/clickhouse-server/server.crt -->
-            <certificateFile>/etc/clickhouse-server/server.crt</certificateFile>
-            <privateKeyFile>/etc/clickhouse-server/server.key</privateKeyFile>
-            <!-- openssl dhparam -out /etc/clickhouse-server/dhparam.pem 4096 -->
-            <dhParamsFile>/etc/clickhouse-server/dhparam.pem</dhParamsFile>
-            <verificationMode>none</verificationMode>
-            <loadDefaultCAFile>true</loadDefaultCAFile>
-            <cacheSessions>true</cacheSessions>
-            <disableProtocols>sslv2,sslv3</disableProtocols>
-            <preferServerCiphers>true</preferServerCiphers>
-        </server>
-
-        <client> <!-- Used for connecting to https dictionary source and secured Zookeeper communication -->
-            <loadDefaultCAFile>true</loadDefaultCAFile>
-            <cacheSessions>true</cacheSessions>
-            <disableProtocols>sslv2,sslv3</disableProtocols>
-            <preferServerCiphers>true</preferServerCiphers>
-            <!-- Use for self-signed: <verificationMode>none</verificationMode> -->
-            <invalidCertificateHandler>
-                <!-- Use for self-signed: <name>AcceptCertificateHandler</name> -->
-                <name>RejectCertificateHandler</name>
-            </invalidCertificateHandler>
-        </client>
-    </openSSL>
-
-    <!-- Default root page on http[s] server. For example load UI from https://tabix.io/ when opening http://localhost:8123 -->
-    <!--
-    <http_server_default_response><![CDATA[<html ng-app="SMI2"><head><base href="http://ui.tabix.io/"></head><body><div ui-view="" class="content-ui"></div><script src="http://loader.tabix.io/master.js"></script></body></html>]]></http_server_default_response>
+    <!-- Native interface with TLS.
+         You have to configure certificate to enable this interface.
+         See the openSSL section below.
     -->
+    <!-- <tcp_port_secure>9440</tcp_port_secure> -->
 
-    <!-- Port for communication between replicas. Used for data exchange. -->
+    <!-- Native interface wrapped with PROXYv1 protocol
+         PROXYv1 header sent for every connection.
+         ClickHouse will extract information about proxy-forwarded client address from the header.
+    -->
+    <!-- <tcp_with_proxy_port>9011</tcp_with_proxy_port> -->
+
+    <!-- Port for communication between replicas. Used for data exchange.
+         It provides low-level data access between servers.
+         This port should not be accessible from untrusted networks.
+         See also 'interserver_http_credentials'.
+         Data transferred over connections to this port should not go through untrusted networks.
+         See also 'interserver_https_port'.
+      -->
     <interserver_http_port>9009</interserver_http_port>
 
+    <!-- Port for communication between replicas with TLS.
+         You have to configure certificate to enable this interface.
+         See the openSSL section below.
+         See also 'interserver_http_credentials'.
+      -->
+    <!-- <interserver_https_port>9010</interserver_https_port> -->
+
     <!-- Hostname that is used by other replicas to request this server.
          If not specified, than it is determined analogous to 'hostname -f' command.
-         This setting could be used to switch replication to another network interface.
+         This setting could be used to switch replication to another network interface
+         (the server may be connected to multiple networks via multiple addresses)
       -->
     <!--
     <interserver_http_host>example.yandex.ru</interserver_http_host>
     -->
 
-    <!-- Listen specified host. use :: (wildcard IPv6 address), if you want to accept connections both with IPv4 and IPv6 from everywhere. -->
+    <!-- You can specify credentials for authenthication between replicas.
+         This is required when interserver_https_port is accessible from untrusted networks,
+         and also recommended to avoid SSRF attacks from possibly compromised services in your network.
+      -->
+    <!--<interserver_http_credentials>
+        <user>interserver</user>
+        <password></password>
+    </interserver_http_credentials>-->
+
+    <!-- Listen specified address.
+         Use :: (wildcard IPv6 address), if you want to accept connections both with IPv4 and IPv6 from everywhere.
+         Notes:
+         If you open connections from wildcard address, make sure that at least one of the following measures applied:
+         - server is protected by firewall and not accessible from untrusted networks;
+         - all users are restricted to subset of network addresses (see users.xml);
+         - all users have strong passwords, only secure (TLS) interfaces are accessible, or connections are only made via TLS interfaces.
+         - users without password have readonly access.
+         See also: https://www.shodan.io/search?query=clickhouse
+      -->
     <!-- <listen_host>::</listen_host> -->
-    <!-- Same for hosts with disabled ipv6: -->
+
+    <!-- Same for hosts without support for IPv6: -->
     <!-- <listen_host>0.0.0.0</listen_host> -->
 
-    <!-- Default values - try listen localhost on ipv4 and ipv6: -->
+    <!-- Default values - try listen localhost on IPv4 and IPv6. -->
     <!--
     <listen_host>::1</listen_host>
     <listen_host>127.0.0.1</listen_host>
     -->
-    <!-- Don't exit if ipv6 or ipv4 unavailable, but listen_host with this protocol specified -->
+
+    <!-- Don't exit if IPv6 or IPv4 networks are unavailable while trying to listen. -->
     <!-- <listen_try>0</listen_try> -->
 
-    <!-- Allow listen on same address:port -->
+    <!-- Allow multiple servers to listen on the same address:port. This is not recommended.
+      -->
     <!-- <listen_reuse_port>0</listen_reuse_port> -->
 
     <!-- <listen_backlog>64</listen_backlog> -->
 
     <max_connections>4096</max_connections>
+
+    <!-- For 'Connection: keep-alive' in HTTP 1.1 -->
     <keep_alive_timeout>3</keep_alive_timeout>
 
     <!-- gRPC protocol (see src/Server/grpc_protos/clickhouse_grpc.proto for the API) -->
@@ -172,6 +196,43 @@
         <verbose_logs>false</verbose_logs>
     </grpc>
 
+    <!-- Used with https_port and tcp_port_secure. Full ssl options list: https://github.com/ClickHouse-Extras/poco/blob/master/NetSSL_OpenSSL/include/Poco/Net/SSLManager.h#L71 -->
+    <openSSL>
+        <server> <!-- Used for https server AND secure tcp port -->
+            <!-- openssl req -subj "/CN=localhost" -new -newkey rsa:2048 -days 365 -nodes -x509 -keyout /etc/clickhouse-server/server.key -out /etc/clickhouse-server/server.crt -->
+            <certificateFile>/etc/clickhouse-server/server.crt</certificateFile>
+            <privateKeyFile>/etc/clickhouse-server/server.key</privateKeyFile>
+            <!-- dhparams are optional. You can delete the <dhParamsFile> element.
+                 To generate dhparams, use the following command:
+                  openssl dhparam -out /etc/clickhouse-server/dhparam.pem 4096 
+                 Only file format with BEGIN DH PARAMETERS is supported.
+              -->
+            <dhParamsFile>/etc/clickhouse-server/dhparam.pem</dhParamsFile>
+            <verificationMode>none</verificationMode>
+            <loadDefaultCAFile>true</loadDefaultCAFile>
+            <cacheSessions>true</cacheSessions>
+            <disableProtocols>sslv2,sslv3</disableProtocols>
+            <preferServerCiphers>true</preferServerCiphers>
+        </server>
+
+        <client> <!-- Used for connecting to https dictionary source and secured Zookeeper communication -->
+            <loadDefaultCAFile>true</loadDefaultCAFile>
+            <cacheSessions>true</cacheSessions>
+            <disableProtocols>sslv2,sslv3</disableProtocols>
+            <preferServerCiphers>true</preferServerCiphers>
+            <!-- Use for self-signed: <verificationMode>none</verificationMode> -->
+            <invalidCertificateHandler>
+                <!-- Use for self-signed: <name>AcceptCertificateHandler</name> -->
+                <name>RejectCertificateHandler</name>
+            </invalidCertificateHandler>
+        </client>
+    </openSSL>
+
+    <!-- Default root page on http[s] server. For example load UI from https://tabix.io/ when opening http://localhost:8123 -->
+    <!--
+    <http_server_default_response><![CDATA[<html ng-app="SMI2"><head><base href="http://ui.tabix.io/"></head><body><div ui-view="" class="content-ui"></div><script src="http://loader.tabix.io/master.js"></script></body></html>]]></http_server_default_response>
+    -->
+
     <!-- Maximum number of concurrent queries. -->
     <max_concurrent_queries>100</max_concurrent_queries>
 
@@ -605,7 +666,7 @@
                 event_date + INTERVAL 1 WEEK
                 event_date + INTERVAL 7 DAY DELETE
                 event_date + INTERVAL 2 WEEK TO DISK 'bbb'
-        
+
         <ttl>event_date + INTERVAL 30 DAY DELETE</ttl>
         -->
 
@@ -887,6 +948,18 @@
     </http_handlers>
     -->
 
+    <send_crash_reports>
+        <!-- Changing <enabled> to true allows sending crash reports to -->
+        <!-- the ClickHouse core developers team via Sentry https://sentry.io -->
+        <!-- Doing so at least in pre-production environments is highly appreciated -->
+        <enabled>false</enabled>
+        <!-- Change <anonymize> to true if you don't feel comfortable attaching the server hostname to the crash report -->
+        <anonymize>false</anonymize>
+        <!-- Default endpoint should be changed to different Sentry DSN only if you have -->
+        <!-- some in-house engineers or hired consultants who're going to debug ClickHouse issues for you -->
+        <endpoint>https://6f33034cfe684dd7a3ab9875e57b1c8d@o388870.ingest.sentry.io/5226277</endpoint>
+    </send_crash_reports>
+
     <!-- Uncomment to disable ClickHouse internal DNS caching. -->
     <!-- <disable_internal_dns_cache>1</disable_internal_dns_cache> -->
 </yandex>
diff --git a/src/Access/ContextAccess.cpp b/src/Access/ContextAccess.cpp
index a74f6e8e7a0..a55d0768a7e 100644
--- a/src/Access/ContextAccess.cpp
+++ b/src/Access/ContextAccess.cpp
@@ -392,9 +392,12 @@ bool ContextAccess::checkAccessImpl2(const AccessFlags & flags, const Args &...
     if (!getUser())
         return access_denied("User has been dropped", ErrorCodes::UNKNOWN_USER);
 
-    /// If the current user was allowed to create a temporary table
-    /// then he is allowed to do with it whatever he wants.
-    if ((sizeof...(args) >= 2) && (getDatabase(args...) == DatabaseCatalog::TEMPORARY_DATABASE))
+    /// Access to temporary tables is controlled in an unusual way, not like normal tables.
+    /// Creating of temporary tables is controlled by AccessType::CREATE_TEMPORARY_TABLES grant,
+    /// and other grants are considered as always given.
+    /// The DatabaseCatalog class won't resolve StorageID for temporary tables
+    /// which shouldn't be accessed.
+    if (getDatabase(args...) == DatabaseCatalog::TEMPORARY_DATABASE)
         return access_granted();
 
     auto acs = getAccessRightsWithImplicit();
diff --git a/src/Access/SettingsConstraints.cpp b/src/Access/SettingsConstraints.cpp
index c9a6b6f6b7a..7dce4d96b33 100644
--- a/src/Access/SettingsConstraints.cpp
+++ b/src/Access/SettingsConstraints.cpp
@@ -156,6 +156,25 @@ bool SettingsConstraints::checkImpl(const Settings & current_settings, SettingCh
 {
     const String & setting_name = change.name;
 
+    if (setting_name == "profile")
+    {
+        /// TODO Check profile settings in Context::setProfile(...), not here. It will be backward incompatible.
+        const String & profile_name = change.value.safeGet<String>();
+        const auto & profile_settings_changes = manager->getProfileSettings(profile_name);
+        try
+        {
+            /// NOTE We cannot use CLAMP_ON_VIOLATION here, because we cannot modify elements of profile_settings_changes
+            for (auto change_copy : *profile_settings_changes)
+                checkImpl(current_settings, change_copy, THROW_ON_VIOLATION);
+        }
+        catch (Exception & e)
+        {
+            e.addMessage(", while trying to set settings profile {}", profile_name);
+            throw;
+        }
+        return true;
+    }
+
     bool cannot_cast;
     auto cast_value = [&](const Field & x) -> Field
     {
diff --git a/src/AggregateFunctions/AggregateFunctionAvg.h b/src/AggregateFunctions/AggregateFunctionAvg.h
index fca9df9dd98..d07ff5db2f2 100644
--- a/src/AggregateFunctions/AggregateFunctionAvg.h
+++ b/src/AggregateFunctions/AggregateFunctionAvg.h
@@ -33,7 +33,7 @@ struct AvgFraction
 
     /// Allow division by zero as sometimes we need to return NaN.
     /// Invoked only is either Numerator or Denominator are Decimal.
-    Float64 NO_SANITIZE_UNDEFINED divideIfAnyDecimal(UInt32 num_scale, UInt32 denom_scale) const
+    Float64 NO_SANITIZE_UNDEFINED divideIfAnyDecimal(UInt32 num_scale, UInt32 denom_scale [[maybe_unused]]) const
     {
         if constexpr (IsDecimalNumber<Numerator> && IsDecimalNumber<Denominator>)
         {
diff --git a/src/AggregateFunctions/AggregateFunctionFactory.cpp b/src/AggregateFunctions/AggregateFunctionFactory.cpp
index 1c66650be45..d1dcb8cf455 100644
--- a/src/AggregateFunctions/AggregateFunctionFactory.cpp
+++ b/src/AggregateFunctions/AggregateFunctionFactory.cpp
@@ -18,6 +18,7 @@
 #include <Poco/String.h>
 #include "registerAggregateFunctions.h"
 
+#include <Functions/FunctionFactory.h>
 
 namespace DB
 {
@@ -135,12 +136,17 @@ AggregateFunctionPtr AggregateFunctionFactory::getImpl(
         return combinator->transformAggregateFunction(nested_function, out_properties, argument_types, parameters);
     }
 
+
+    String extra_info;
+    if (FunctionFactory::instance().hasNameOrAlias(name))
+        extra_info = ". There is an ordinary function with the same name, but aggregate function is expected here";
+
     auto hints = this->getHints(name);
     if (!hints.empty())
-        throw Exception(fmt::format("Unknown aggregate function {}. Maybe you meant: {}", name, toString(hints)),
-            ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION);
+        throw Exception(ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION,
+                        "Unknown aggregate function {}{}. Maybe you meant: {}", name, extra_info, toString(hints));
     else
-        throw Exception(fmt::format("Unknown aggregate function {}", name), ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION);
+        throw Exception(ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION, "Unknown aggregate function {}{}", name, extra_info);
 }
 
 
diff --git a/src/AggregateFunctions/AggregateFunctionSimpleState.cpp b/src/AggregateFunctions/AggregateFunctionSimpleState.cpp
new file mode 100644
index 00000000000..8a5a71d6806
--- /dev/null
+++ b/src/AggregateFunctions/AggregateFunctionSimpleState.cpp
@@ -0,0 +1,32 @@
+#include <AggregateFunctions/AggregateFunctionCombinatorFactory.h>
+#include <AggregateFunctions/AggregateFunctionSimpleState.h>
+
+namespace DB
+{
+namespace
+{
+    class AggregateFunctionCombinatorSimpleState final : public IAggregateFunctionCombinator
+    {
+    public:
+        String getName() const override { return "SimpleState"; }
+
+        DataTypes transformArguments(const DataTypes & arguments) const override { return arguments; }
+
+        AggregateFunctionPtr transformAggregateFunction(
+            const AggregateFunctionPtr & nested_function,
+            const AggregateFunctionProperties &,
+            const DataTypes & arguments,
+            const Array & params) const override
+        {
+            return std::make_shared<AggregateFunctionSimpleState>(nested_function, arguments, params);
+        }
+    };
+
+}
+
+void registerAggregateFunctionCombinatorSimpleState(AggregateFunctionCombinatorFactory & factory)
+{
+    factory.registerCombinator(std::make_shared<AggregateFunctionCombinatorSimpleState>());
+}
+
+}
diff --git a/src/AggregateFunctions/AggregateFunctionSimpleState.h b/src/AggregateFunctions/AggregateFunctionSimpleState.h
new file mode 100644
index 00000000000..eefe00e77c0
--- /dev/null
+++ b/src/AggregateFunctions/AggregateFunctionSimpleState.h
@@ -0,0 +1,77 @@
+#pragma once
+
+#include <AggregateFunctions/IAggregateFunction.h>
+#include <DataTypes/DataTypeCustomSimpleAggregateFunction.h>
+#include <DataTypes/DataTypeFactory.h>
+
+namespace DB
+{
+
+/** Not an aggregate function, but an adapter of aggregate functions.
+  * Aggregate functions with the `SimpleState` suffix is almost identical to the corresponding ones,
+  * except the return type becomes DataTypeCustomSimpleAggregateFunction.
+  */
+class AggregateFunctionSimpleState final : public IAggregateFunctionHelper<AggregateFunctionSimpleState>
+{
+private:
+    AggregateFunctionPtr nested_func;
+    DataTypes arguments;
+    Array params;
+
+public:
+    AggregateFunctionSimpleState(AggregateFunctionPtr nested_, const DataTypes & arguments_, const Array & params_)
+        : IAggregateFunctionHelper<AggregateFunctionSimpleState>(arguments_, params_)
+        , nested_func(nested_)
+        , arguments(arguments_)
+        , params(params_)
+    {
+    }
+
+    String getName() const override { return nested_func->getName() + "SimpleState"; }
+
+    DataTypePtr getReturnType() const override
+    {
+        DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(nested_func);
+        // Need to make a clone because it'll be customized.
+        auto storage_type = DataTypeFactory::instance().get(nested_func->getReturnType()->getName());
+        DataTypeCustomNamePtr custom_name
+            = std::make_unique<DataTypeCustomSimpleAggregateFunction>(nested_func, DataTypes{nested_func->getReturnType()}, params);
+        storage_type->setCustomization(std::make_unique<DataTypeCustomDesc>(std::move(custom_name), nullptr));
+        return storage_type;
+    }
+
+    void create(AggregateDataPtr place) const override { nested_func->create(place); }
+
+    void destroy(AggregateDataPtr place) const noexcept override { nested_func->destroy(place); }
+
+    bool hasTrivialDestructor() const override { return nested_func->hasTrivialDestructor(); }
+
+    size_t sizeOfData() const override { return nested_func->sizeOfData(); }
+
+    size_t alignOfData() const override { return nested_func->alignOfData(); }
+
+    void add(AggregateDataPtr place, const IColumn ** columns, size_t row_num, Arena * arena) const override
+    {
+        nested_func->add(place, columns, row_num, arena);
+    }
+
+    void merge(AggregateDataPtr place, ConstAggregateDataPtr rhs, Arena * arena) const override { nested_func->merge(place, rhs, arena); }
+
+    void serialize(ConstAggregateDataPtr place, WriteBuffer & buf) const override { nested_func->serialize(place, buf); }
+
+    void deserialize(AggregateDataPtr place, ReadBuffer & buf, Arena * arena) const override
+    {
+        nested_func->deserialize(place, buf, arena);
+    }
+
+    void insertResultInto(AggregateDataPtr place, IColumn & to, Arena * arena) const override
+    {
+        nested_func->insertResultInto(place, to, arena);
+    }
+
+    bool allocatesMemoryInArena() const override { return nested_func->allocatesMemoryInArena(); }
+
+    AggregateFunctionPtr getNestedFunction() const { return nested_func; }
+};
+
+}
diff --git a/src/AggregateFunctions/registerAggregateFunctions.cpp b/src/AggregateFunctions/registerAggregateFunctions.cpp
index 927c7e6d6b1..228ca2f30ab 100644
--- a/src/AggregateFunctions/registerAggregateFunctions.cpp
+++ b/src/AggregateFunctions/registerAggregateFunctions.cpp
@@ -47,6 +47,7 @@ class AggregateFunctionCombinatorFactory;
 void registerAggregateFunctionCombinatorIf(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorArray(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorForEach(AggregateFunctionCombinatorFactory &);
+void registerAggregateFunctionCombinatorSimpleState(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorState(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorMerge(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorNull(AggregateFunctionCombinatorFactory &);
@@ -104,6 +105,7 @@ void registerAggregateFunctions()
         registerAggregateFunctionCombinatorIf(factory);
         registerAggregateFunctionCombinatorArray(factory);
         registerAggregateFunctionCombinatorForEach(factory);
+        registerAggregateFunctionCombinatorSimpleState(factory);
         registerAggregateFunctionCombinatorState(factory);
         registerAggregateFunctionCombinatorMerge(factory);
         registerAggregateFunctionCombinatorNull(factory);
diff --git a/src/AggregateFunctions/ya.make b/src/AggregateFunctions/ya.make
index f05b8ef55d2..8faf738fd98 100644
--- a/src/AggregateFunctions/ya.make
+++ b/src/AggregateFunctions/ya.make
@@ -41,6 +41,7 @@ SRCS(
     AggregateFunctionRetention.cpp
     AggregateFunctionSequenceMatch.cpp
     AggregateFunctionSimpleLinearRegression.cpp
+    AggregateFunctionSimpleState.cpp
     AggregateFunctionState.cpp
     AggregateFunctionStatistics.cpp
     AggregateFunctionStatisticsSimple.cpp
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 6021065f937..46956181974 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -88,6 +88,10 @@ if (USE_AWS_S3)
     add_headers_and_sources(dbms Disks/S3)
 endif()
 
+if (USE_HDFS)
+    add_headers_and_sources(dbms Storages/HDFS)
+endif()
+
 list (APPEND clickhouse_common_io_sources ${CONFIG_BUILD})
 list (APPEND clickhouse_common_io_headers ${CONFIG_VERSION} ${CONFIG_COMMON})
 
@@ -389,8 +393,8 @@ if (USE_GRPC)
 endif()
 
 if (USE_HDFS)
-    target_link_libraries (clickhouse_common_io PUBLIC ${HDFS3_LIBRARY})
-    target_include_directories (clickhouse_common_io SYSTEM BEFORE PUBLIC ${HDFS3_INCLUDE_DIR})
+    dbms_target_link_libraries(PRIVATE ${HDFS3_LIBRARY})
+    dbms_target_include_directories (SYSTEM BEFORE PUBLIC ${HDFS3_INCLUDE_DIR})
 endif()
 
 if (USE_AWS_S3)
diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index 30a74ec73aa..265a9913b8e 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -5,8 +5,9 @@
 #include <Poco/Net/StreamSocket.h>
 
 #include <Common/Throttler.h>
-#include <Common/config.h>
-
+#if !defined(ARCADIA_BUILD)
+#   include <Common/config.h>
+#endif
 #include <Core/Block.h>
 #include <Core/Defines.h>
 #include <IO/Progress.h>
diff --git a/src/Columns/ColumnMap.cpp b/src/Columns/ColumnMap.cpp
new file mode 100644
index 00000000000..c210da2aa11
--- /dev/null
+++ b/src/Columns/ColumnMap.cpp
@@ -0,0 +1,241 @@
+#include <Columns/ColumnMap.h>
+#include <Columns/IColumnImpl.h>
+#include <DataStreams/ColumnGathererStream.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/Operators.h>
+#include <ext/map.h>
+#include <ext/range.h>
+#include <Common/typeid_cast.h>
+#include <Common/assert_cast.h>
+#include <Common/WeakHash.h>
+#include <Core/Field.h>
+
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_COLUMN;
+    extern const int NOT_IMPLEMENTED;
+    extern const int LOGICAL_ERROR;
+}
+
+
+std::string ColumnMap::getName() const
+{
+    WriteBufferFromOwnString res;
+    const auto & nested_tuple = getNestedData();
+    res << "Map(" << nested_tuple.getColumn(0).getName()
+        << ", " << nested_tuple.getColumn(1).getName() << ")";
+
+    return res.str();
+}
+
+ColumnMap::ColumnMap(MutableColumnPtr && nested_)
+    : nested(std::move(nested_))
+{
+    const auto * column_array = typeid_cast<const ColumnArray *>(nested.get());
+    if (!column_array)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ColumnMap can be created only from array of tuples");
+
+    const auto * column_tuple = typeid_cast<const ColumnTuple *>(column_array->getDataPtr().get());
+    if (!column_tuple)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ColumnMap can be created only from array of tuples");
+
+    if (column_tuple->getColumns().size() != 2)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ColumnMap should contain only 2 subcolumns: keys and values");
+
+    for (const auto & column : column_tuple->getColumns())
+        if (isColumnConst(*column))
+            throw Exception(ErrorCodes::ILLEGAL_COLUMN, "ColumnMap cannot have ColumnConst as its element");
+}
+
+MutableColumnPtr ColumnMap::cloneEmpty() const
+{
+    return ColumnMap::create(nested->cloneEmpty());
+}
+
+MutableColumnPtr ColumnMap::cloneResized(size_t new_size) const
+{
+    return ColumnMap::create(nested->cloneResized(new_size));
+}
+
+Field ColumnMap::operator[](size_t n) const
+{
+    auto array = DB::get<Array>((*nested)[n]);
+    return Map(std::make_move_iterator(array.begin()), std::make_move_iterator(array.end()));
+}
+
+void ColumnMap::get(size_t n, Field & res) const
+{
+    const auto & offsets = getNestedColumn().getOffsets();
+    size_t offset = offsets[n - 1];
+    size_t size = offsets[n] - offsets[n - 1];
+
+    res = Map(size);
+    auto & map = DB::get<Map &>(res);
+
+    for (size_t i = 0; i < size; ++i)
+        getNestedData().get(offset + i, map[i]);
+}
+
+StringRef ColumnMap::getDataAt(size_t) const
+{
+    throw Exception("Method getDataAt is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
+}
+
+void ColumnMap::insertData(const char *, size_t)
+{
+    throw Exception("Method insertData is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
+}
+
+void ColumnMap::insert(const Field & x)
+{
+    const auto & map = DB::get<const Map &>(x);
+    nested->insert(Array(map.begin(), map.end()));
+}
+
+void ColumnMap::insertDefault()
+{
+    nested->insertDefault();
+}
+void ColumnMap::popBack(size_t n)
+{
+    nested->popBack(n);
+}
+
+StringRef ColumnMap::serializeValueIntoArena(size_t n, Arena & arena, char const *& begin) const
+{
+    return nested->serializeValueIntoArena(n, arena, begin);
+}
+
+const char * ColumnMap::deserializeAndInsertFromArena(const char * pos)
+{
+    return nested->deserializeAndInsertFromArena(pos);
+}
+
+void ColumnMap::updateHashWithValue(size_t n, SipHash & hash) const
+{
+    nested->updateHashWithValue(n, hash);
+}
+
+void ColumnMap::updateWeakHash32(WeakHash32 & hash) const
+{
+    nested->updateWeakHash32(hash);
+}
+
+void ColumnMap::updateHashFast(SipHash & hash) const
+{
+    nested->updateHashFast(hash);
+}
+
+void ColumnMap::insertRangeFrom(const IColumn & src, size_t start, size_t length)
+{
+    nested->insertRangeFrom(
+        assert_cast<const ColumnMap &>(src).getNestedColumn(),
+        start, length);
+}
+
+ColumnPtr ColumnMap::filter(const Filter & filt, ssize_t result_size_hint) const
+{
+    auto filtered = nested->filter(filt, result_size_hint);
+    return ColumnMap::create(filtered);
+}
+
+ColumnPtr ColumnMap::permute(const Permutation & perm, size_t limit) const
+{
+    auto permuted = nested->permute(perm, limit);
+    return ColumnMap::create(std::move(permuted));
+}
+
+ColumnPtr ColumnMap::index(const IColumn & indexes, size_t limit) const
+{
+    auto res = nested->index(indexes, limit);
+    return ColumnMap::create(std::move(res));
+}
+
+ColumnPtr ColumnMap::replicate(const Offsets & offsets) const
+{
+    auto replicated = nested->replicate(offsets);
+    return ColumnMap::create(std::move(replicated));
+}
+
+MutableColumns ColumnMap::scatter(ColumnIndex num_columns, const Selector & selector) const
+{
+    auto scattered_columns = nested->scatter(num_columns, selector);
+    MutableColumns res;
+    res.reserve(num_columns);
+    for (auto && scattered : scattered_columns)
+        res.push_back(ColumnMap::create(std::move(scattered)));
+
+    return res;
+}
+
+int ColumnMap::compareAt(size_t n, size_t m, const IColumn & rhs, int nan_direction_hint) const
+{
+    const auto & rhs_map = assert_cast<const ColumnMap &>(rhs);
+    return nested->compareAt(n, m, rhs_map.getNestedColumn(), nan_direction_hint);
+}
+
+void ColumnMap::compareColumn(const IColumn & rhs, size_t rhs_row_num,
+                                PaddedPODArray<UInt64> * row_indexes, PaddedPODArray<Int8> & compare_results,
+                                int direction, int nan_direction_hint) const
+{
+    return doCompareColumn<ColumnMap>(assert_cast<const ColumnMap &>(rhs), rhs_row_num, row_indexes,
+                                        compare_results, direction, nan_direction_hint);
+}
+
+void ColumnMap::getPermutation(bool reverse, size_t limit, int nan_direction_hint, Permutation & res) const
+{
+    nested->getPermutation(reverse, limit, nan_direction_hint, res);
+}
+
+void ColumnMap::updatePermutation(bool reverse, size_t limit, int nan_direction_hint, IColumn::Permutation & res, EqualRanges & equal_range) const
+{
+    nested->updatePermutation(reverse, limit, nan_direction_hint, res, equal_range);
+}
+
+void ColumnMap::gather(ColumnGathererStream & gatherer)
+{
+    gatherer.gather(*this);
+}
+
+void ColumnMap::reserve(size_t n)
+{
+    nested->reserve(n);
+}
+
+size_t ColumnMap::byteSize() const
+{
+    return nested->byteSize();
+}
+
+size_t ColumnMap::allocatedBytes() const
+{
+    return nested->allocatedBytes();
+}
+
+void ColumnMap::protect()
+{
+    nested->protect();
+}
+
+void ColumnMap::getExtremes(Field & min, Field & max) const
+{
+    nested->getExtremes(min, max);
+}
+
+void ColumnMap::forEachSubcolumn(ColumnCallback callback)
+{
+    nested->forEachSubcolumn(callback);
+}
+
+bool ColumnMap::structureEquals(const IColumn & rhs) const
+{
+    if (const auto * rhs_map = typeid_cast<const ColumnMap *>(&rhs))
+        return nested->structureEquals(*rhs_map->nested);
+    return false;
+}
+
+}
diff --git a/src/Columns/ColumnMap.h b/src/Columns/ColumnMap.h
new file mode 100644
index 00000000000..08197c3e1e9
--- /dev/null
+++ b/src/Columns/ColumnMap.h
@@ -0,0 +1,92 @@
+#pragma once
+
+#include <Core/Block.h>
+#include <Columns/ColumnArray.h>
+#include <Columns/ColumnVector.h>
+#include <Columns/ColumnTuple.h>
+
+namespace DB
+{
+
+/** Column, that stores a nested Array(Tuple(key, value)) column.
+ */
+class ColumnMap final : public COWHelper<IColumn, ColumnMap>
+{
+private:
+    friend class COWHelper<IColumn, ColumnMap>;
+
+    WrappedPtr nested;
+
+    explicit ColumnMap(MutableColumnPtr && nested_);
+
+    ColumnMap(const ColumnMap &) = default;
+
+public:
+    /** Create immutable column using immutable arguments. This arguments may be shared with other columns.
+      * Use IColumn::mutate in order to make mutable column and mutate shared nested columns.
+      */
+    using Base = COWHelper<IColumn, ColumnMap>;
+
+    static Ptr create(const ColumnPtr & keys, const ColumnPtr & values, const ColumnPtr & offsets)
+    {
+        auto nested_column = ColumnArray::create(ColumnTuple::create(Columns{keys, values}), offsets);
+        return ColumnMap::create(nested_column);
+    }
+
+    static Ptr create(const ColumnPtr & column) { return ColumnMap::create(column->assumeMutable()); }
+    static Ptr create(ColumnPtr && arg) { return create(arg); }
+
+    template <typename Arg, typename = typename std::enable_if<std::is_rvalue_reference<Arg &&>::value>::type>
+    static MutablePtr create(Arg && arg) { return Base::create(std::forward<Arg>(arg)); }
+
+    std::string getName() const override;
+    const char * getFamilyName() const override { return "Map"; }
+    TypeIndex getDataType() const override { return TypeIndex::Map; }
+
+    MutableColumnPtr cloneEmpty() const override;
+    MutableColumnPtr cloneResized(size_t size) const override;
+
+    size_t size() const override { return nested->size(); }
+
+    Field operator[](size_t n) const override;
+    void get(size_t n, Field & res) const override;
+
+    StringRef getDataAt(size_t n) const override;
+    void insertData(const char * pos, size_t length) override;
+    void insert(const Field & x) override;
+    void insertDefault() override;
+    void popBack(size_t n) override;
+    StringRef serializeValueIntoArena(size_t n, Arena & arena, char const *& begin) const override;
+    const char * deserializeAndInsertFromArena(const char * pos) override;
+    void updateHashWithValue(size_t n, SipHash & hash) const override;
+    void updateWeakHash32(WeakHash32 & hash) const override;
+    void updateHashFast(SipHash & hash) const override;
+    void insertRangeFrom(const IColumn & src, size_t start, size_t length) override;
+    ColumnPtr filter(const Filter & filt, ssize_t result_size_hint) const override;
+    ColumnPtr permute(const Permutation & perm, size_t limit) const override;
+    ColumnPtr index(const IColumn & indexes, size_t limit) const override;
+    ColumnPtr replicate(const Offsets & offsets) const override;
+    MutableColumns scatter(ColumnIndex num_columns, const Selector & selector) const override;
+    void gather(ColumnGathererStream & gatherer_stream) override;
+    int compareAt(size_t n, size_t m, const IColumn & rhs, int nan_direction_hint) const override;
+    void compareColumn(const IColumn & rhs, size_t rhs_row_num,
+                       PaddedPODArray<UInt64> * row_indexes, PaddedPODArray<Int8> & compare_results,
+                       int direction, int nan_direction_hint) const override;
+    void getExtremes(Field & min, Field & max) const override;
+    void getPermutation(bool reverse, size_t limit, int nan_direction_hint, Permutation & res) const override;
+    void updatePermutation(bool reverse, size_t limit, int nan_direction_hint, IColumn::Permutation & res, EqualRanges & equal_range) const override;
+    void reserve(size_t n) override;
+    size_t byteSize() const override;
+    size_t allocatedBytes() const override;
+    void protect() override;
+    void forEachSubcolumn(ColumnCallback callback) override;
+    bool structureEquals(const IColumn & rhs) const override;
+
+    const ColumnArray & getNestedColumn() const { return assert_cast<const ColumnArray &>(*nested); }
+    ColumnArray & getNestedColumn() { return assert_cast<ColumnArray &>(*nested); }
+
+    const ColumnTuple & getNestedData() const { return assert_cast<const ColumnTuple &>(getNestedColumn().getData()); }
+    ColumnTuple & getNestedData() { return assert_cast<ColumnTuple &>(getNestedColumn().getData()); }
+};
+
+}
diff --git a/src/Columns/ya.make b/src/Columns/ya.make
index f1a8cb9f32e..2affaeb0fc6 100644
--- a/src/Columns/ya.make
+++ b/src/Columns/ya.make
@@ -24,6 +24,7 @@ SRCS(
     ColumnFixedString.cpp
     ColumnFunction.cpp
     ColumnLowCardinality.cpp
+    ColumnMap.cpp
     ColumnNullable.cpp
     ColumnString.cpp
     ColumnTuple.cpp
diff --git a/src/Common/Allocator.h b/src/Common/Allocator.h
index 6ac60020379..0964bb9f522 100644
--- a/src/Common/Allocator.h
+++ b/src/Common/Allocator.h
@@ -26,6 +26,7 @@
     #define DISABLE_MREMAP 1
 #endif
 #include <common/mremap.h>
+#include <common/getPageSize.h>
 
 #include <Common/MemoryTracker.h>
 #include <Common/Exception.h>
@@ -59,7 +60,6 @@
   */
 extern const size_t MMAP_THRESHOLD;
 
-static constexpr size_t MMAP_MIN_ALIGNMENT = 4096;
 static constexpr size_t MALLOC_MIN_ALIGNMENT = 8;
 
 namespace DB
@@ -194,10 +194,11 @@ private:
     void * allocNoTrack(size_t size, size_t alignment)
     {
         void * buf;
+        size_t mmap_min_alignment = ::getPageSize();
 
         if (size >= MMAP_THRESHOLD)
         {
-            if (alignment > MMAP_MIN_ALIGNMENT)
+            if (alignment > mmap_min_alignment)
                 throw DB::Exception(fmt::format("Too large alignment {}: more than page size when allocating {}.",
                     ReadableSize(alignment), ReadableSize(size)), DB::ErrorCodes::BAD_ARGUMENTS);
 
diff --git a/src/Common/Arena.h b/src/Common/Arena.h
index 44a9b444ff2..d41dd9484d4 100644
--- a/src/Common/Arena.h
+++ b/src/Common/Arena.h
@@ -83,10 +83,11 @@ private:
     /// Last contiguous chunk of memory.
     Chunk * head;
     size_t size_in_bytes;
+    size_t page_size;
 
-    static size_t roundUpToPageSize(size_t s)
+    static size_t roundUpToPageSize(size_t s, size_t page_size)
     {
-        return (s + 4096 - 1) / 4096 * 4096;
+        return (s + page_size - 1) / page_size * page_size;
     }
 
     /// If chunks size is less than 'linear_growth_threshold', then use exponential growth, otherwise - linear growth
@@ -113,7 +114,7 @@ private:
         }
 
         assert(size_after_grow >= min_next_size);
-        return roundUpToPageSize(size_after_grow);
+        return roundUpToPageSize(size_after_grow, page_size);
     }
 
     /// Add next contiguous chunk of memory with size not less than specified.
@@ -129,7 +130,8 @@ private:
 public:
     Arena(size_t initial_size_ = 4096, size_t growth_factor_ = 2, size_t linear_growth_threshold_ = 128 * 1024 * 1024)
         : growth_factor(growth_factor_), linear_growth_threshold(linear_growth_threshold_),
-        head(new Chunk(initial_size_, nullptr)), size_in_bytes(head->size())
+        head(new Chunk(initial_size_, nullptr)), size_in_bytes(head->size()),
+        page_size(static_cast<size_t>(::getPageSize()))
     {
     }
 
diff --git a/src/Common/ArrayCache.h b/src/Common/ArrayCache.h
index caa26d237bf..6883d5f8041 100644
--- a/src/Common/ArrayCache.h
+++ b/src/Common/ArrayCache.h
@@ -13,6 +13,8 @@
 #include <boost/noncopyable.hpp>
 #include <ext/scope_guard.h>
 
+#include <common/getPageSize.h>
+
 #include <Common/Exception.h>
 #include <Common/randomSeed.h>
 #include <Common/formatReadable.h>
@@ -326,8 +328,6 @@ private:
         return (x + (rounding - 1)) / rounding * rounding;
     }
 
-    static constexpr size_t page_size = 4096;
-
     /// Sizes and addresses of allocated memory will be aligned to specified boundary.
     static constexpr size_t alignment = 16;
 
@@ -505,6 +505,7 @@ private:
 
         /// If nothing was found and total size of allocated chunks plus required size is lower than maximum,
         ///  allocate a new chunk.
+        size_t page_size = static_cast<size_t>(::getPageSize());
         size_t required_chunk_size = std::max(min_chunk_size, roundUp(size, page_size));
         if (total_chunks_size + required_chunk_size <= max_total_size)
         {
diff --git a/src/Common/Config/ConfigProcessor.cpp b/src/Common/Config/ConfigProcessor.cpp
index 3e06be94b11..5399826fe20 100644
--- a/src/Common/Config/ConfigProcessor.cpp
+++ b/src/Common/Config/ConfigProcessor.cpp
@@ -16,9 +16,11 @@
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/Exception.h>
 #include <common/getResource.h>
+#include <common/errnoToString.h>
 #include <IO/WriteBufferFromString.h>
 #include <IO/Operators.h>
 
+
 #define PREPROCESSED_SUFFIX "-preprocessed"
 
 
@@ -234,7 +236,7 @@ static std::string layerFromHost()
 {
     utsname buf;
     if (uname(&buf))
-        throw Poco::Exception(std::string("uname failed: ") + std::strerror(errno));
+        throw Poco::Exception(std::string("uname failed: ") + errnoToString(errno));
 
     std::string layer = numberFromHost(buf.nodename);
     if (layer.empty())
diff --git a/src/Common/ErrorCodes.cpp b/src/Common/ErrorCodes.cpp
index 1e381808d16..f9908e4e612 100644
--- a/src/Common/ErrorCodes.cpp
+++ b/src/Common/ErrorCodes.cpp
@@ -529,6 +529,7 @@
     M(560, ZSTD_ENCODER_FAILED) \
     M(561, ZSTD_DECODER_FAILED) \
     M(562, TLD_LIST_NOT_FOUND) \
+    M(563, CANNOT_READ_MAP_FROM_TEXT) \
     \
     M(999, KEEPER_EXCEPTION) \
     M(1000, POCO_EXCEPTION) \
diff --git a/src/Common/FieldVisitors.cpp b/src/Common/FieldVisitors.cpp
index ba3187ed039..dae153bd8d2 100644
--- a/src/Common/FieldVisitors.cpp
+++ b/src/Common/FieldVisitors.cpp
@@ -93,6 +93,22 @@ String FieldVisitorDump::operator() (const Tuple & x) const
     return wb.str();
 }
 
+String FieldVisitorDump::operator() (const Map & x) const
+{
+    WriteBufferFromOwnString wb;
+
+    wb << "Map_(";
+    for (auto it = x.begin(); it != x.end(); ++it)
+    {
+        if (it != x.begin())
+            wb << ", ";
+        wb << applyVisitor(*this, *it);
+    }
+    wb << ')';
+
+    return wb.str();
+}
+
 String FieldVisitorDump::operator() (const AggregateFunctionStateData & x) const
 {
     WriteBufferFromOwnString wb;
@@ -176,6 +192,82 @@ String FieldVisitorToString::operator() (const Tuple & x) const
     return wb.str();
 }
 
+String FieldVisitorToString::operator() (const Map & x) const
+{
+    WriteBufferFromOwnString wb;
+
+    wb << '(';
+    for (auto it = x.begin(); it != x.end(); ++it)
+    {
+        if (it != x.begin())
+            wb << ", ";
+        wb << applyVisitor(*this, *it);
+    }
+    wb << ')';
+
+    return wb.str();
+}
+
+
+void FieldVisitorWriteBinary::operator() (const Null &, WriteBuffer &) const { }
+void FieldVisitorWriteBinary::operator() (const UInt64 & x, WriteBuffer & buf) const { DB::writeVarUInt(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Int64 & x, WriteBuffer & buf) const { DB::writeVarInt(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Float64 & x, WriteBuffer & buf) const { DB::writeFloatBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const String & x, WriteBuffer & buf) const { DB::writeStringBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const UInt128 & x, WriteBuffer & buf) const { DB::writeBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Int128 & x, WriteBuffer & buf) const { DB::writeVarInt(x, buf); }
+void FieldVisitorWriteBinary::operator() (const UInt256 & x, WriteBuffer & buf) const { DB::writeBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Int256 & x, WriteBuffer & buf) const { DB::writeBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal32> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal64> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal128> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal256> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const AggregateFunctionStateData & x, WriteBuffer & buf) const
+{
+    DB::writeStringBinary(x.name, buf);
+    DB::writeStringBinary(x.data, buf);
+}
+
+void FieldVisitorWriteBinary::operator() (const Array & x, WriteBuffer & buf) const
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (size_t i = 0; i < size; ++i)
+    {
+        const UInt8 type = x[i].getType();
+        DB::writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, x[i]);
+    }
+}
+
+void FieldVisitorWriteBinary::operator() (const Tuple & x, WriteBuffer & buf) const
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (size_t i = 0; i < size; ++i)
+    {
+        const UInt8 type = x[i].getType();
+        DB::writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, x[i]);
+    }
+}
+
+
+void FieldVisitorWriteBinary::operator() (const Map & x, WriteBuffer & buf) const
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (size_t i = 0; i < size; ++i)
+    {
+        const UInt8 type = x[i].getType();
+        writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, x[i]);
+    }
+}
+
 
 FieldVisitorHash::FieldVisitorHash(SipHash & hash_) : hash(hash_) {}
 
@@ -238,6 +330,16 @@ void FieldVisitorHash::operator() (const Tuple & x) const
         applyVisitor(*this, elem);
 }
 
+void FieldVisitorHash::operator() (const Map & x) const
+{
+    UInt8 type = Field::Types::Map;
+    hash.update(type);
+    hash.update(x.size());
+
+    for (const auto & elem : x)
+        applyVisitor(*this, elem);
+}
+
 void FieldVisitorHash::operator() (const Array & x) const
 {
     UInt8 type = Field::Types::Array;
diff --git a/src/Common/FieldVisitors.h b/src/Common/FieldVisitors.h
index 79d5af88904..314609ee1a0 100644
--- a/src/Common/FieldVisitors.h
+++ b/src/Common/FieldVisitors.h
@@ -77,6 +77,7 @@ public:
     String operator() (const String & x) const;
     String operator() (const Array & x) const;
     String operator() (const Tuple & x) const;
+    String operator() (const Map & x) const;
     String operator() (const DecimalField<Decimal32> & x) const;
     String operator() (const DecimalField<Decimal64> & x) const;
     String operator() (const DecimalField<Decimal128> & x) const;
@@ -88,6 +89,30 @@ public:
 };
 
 
+class FieldVisitorWriteBinary
+{
+public:
+    void operator() (const Null & x, WriteBuffer & buf) const;
+    void operator() (const UInt64 & x, WriteBuffer & buf) const;
+    void operator() (const UInt128 & x, WriteBuffer & buf) const;
+    void operator() (const Int64 & x, WriteBuffer & buf) const;
+    void operator() (const Int128 & x, WriteBuffer & buf) const;
+    void operator() (const Float64 & x, WriteBuffer & buf) const;
+    void operator() (const String & x, WriteBuffer & buf) const;
+    void operator() (const Array & x, WriteBuffer & buf) const;
+    void operator() (const Tuple & x, WriteBuffer & buf) const;
+    void operator() (const Map & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal32> & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal64> & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal128> & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal256> & x, WriteBuffer & buf) const;
+    void operator() (const AggregateFunctionStateData & x, WriteBuffer & buf) const;
+
+    void operator() (const UInt256 & x, WriteBuffer & buf) const;
+    void operator() (const Int256 & x, WriteBuffer & buf) const;
+};
+
+
 /** Print readable and unique text dump of field type and value. */
 class FieldVisitorDump : public StaticVisitor<String>
 {
@@ -101,6 +126,7 @@ public:
     String operator() (const String & x) const;
     String operator() (const Array & x) const;
     String operator() (const Tuple & x) const;
+    String operator() (const Map & x) const;
     String operator() (const DecimalField<Decimal32> & x) const;
     String operator() (const DecimalField<Decimal64> & x) const;
     String operator() (const DecimalField<Decimal128> & x) const;
@@ -137,6 +163,11 @@ public:
         throw Exception("Cannot convert Tuple to " + demangle(typeid(T).name()), ErrorCodes::CANNOT_CONVERT_TYPE);
     }
 
+    T operator() (const Map &) const
+    {
+        throw Exception("Cannot convert Map to " + demangle(typeid(T).name()), ErrorCodes::CANNOT_CONVERT_TYPE);
+    }
+
     T operator() (const UInt64 & x) const { return T(x); }
     T operator() (const Int64 & x) const { return T(x); }
     T operator() (const Int128 & x) const { return T(x); }
@@ -226,6 +257,7 @@ public:
     void operator() (const String & x) const;
     void operator() (const Array & x) const;
     void operator() (const Tuple & x) const;
+    void operator() (const Map & x) const;
     void operator() (const DecimalField<Decimal32> & x) const;
     void operator() (const DecimalField<Decimal64> & x) const;
     void operator() (const DecimalField<Decimal128> & x) const;
@@ -268,6 +300,7 @@ public:
     bool operator() (String &) const { throw Exception("Cannot sum Strings", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (Array &) const { throw Exception("Cannot sum Arrays", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (Tuple &) const { throw Exception("Cannot sum Tuples", ErrorCodes::LOGICAL_ERROR); }
+    bool operator() (Map &) const { throw Exception("Cannot sum Maps", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (UInt128 &) const { throw Exception("Cannot sum UUIDs", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (AggregateFunctionStateData &) const { throw Exception("Cannot sum AggregateFunctionStates", ErrorCodes::LOGICAL_ERROR); }
 
diff --git a/src/Common/IFactoryWithAliases.h b/src/Common/IFactoryWithAliases.h
index 11ebf31db33..49c03049b92 100644
--- a/src/Common/IFactoryWithAliases.h
+++ b/src/Common/IFactoryWithAliases.h
@@ -106,6 +106,11 @@ public:
         return aliases.count(name) || case_insensitive_aliases.count(name);
     }
 
+    bool hasNameOrAlias(const String & name) const
+    {
+        return getMap().count(name) || getCaseInsensitiveMap().count(name) || isAlias(name);
+    }
+
     virtual ~IFactoryWithAliases() override {}
 
 private:
diff --git a/src/Common/MemoryStatisticsOS.cpp b/src/Common/MemoryStatisticsOS.cpp
index 7c65461aeb6..f1022d887c1 100644
--- a/src/Common/MemoryStatisticsOS.cpp
+++ b/src/Common/MemoryStatisticsOS.cpp
@@ -8,10 +8,11 @@
 
 #include "MemoryStatisticsOS.h"
 
+#include <common/logger_useful.h>
+#include <common/getPageSize.h>
 #include <Common/Exception.h>
 #include <IO/ReadBufferFromMemory.h>
 #include <IO/ReadHelpers.h>
-#include <common/logger_useful.h>
 
 
 namespace DB
@@ -26,7 +27,6 @@ namespace ErrorCodes
 }
 
 static constexpr auto filename = "/proc/self/statm";
-static constexpr size_t PAGE_SIZE = 4096;
 
 MemoryStatisticsOS::MemoryStatisticsOS()
 {
@@ -93,11 +93,12 @@ MemoryStatisticsOS::Data MemoryStatisticsOS::get() const
     skipWhitespaceIfAny(in);
     readIntText(data.data_and_stack, in);
 
-    data.virt *= PAGE_SIZE;
-    data.resident *= PAGE_SIZE;
-    data.shared *= PAGE_SIZE;
-    data.code *= PAGE_SIZE;
-    data.data_and_stack *= PAGE_SIZE;
+    size_t page_size = static_cast<size_t>(::getPageSize());
+    data.virt *= page_size;
+    data.resident *= page_size;
+    data.shared *= page_size;
+    data.code *= page_size;
+    data.data_and_stack *= page_size;
 
     return data;
 }
diff --git a/src/Common/StringSearcher.h b/src/Common/StringSearcher.h
index a4da499837b..e39253c439f 100644
--- a/src/Common/StringSearcher.h
+++ b/src/Common/StringSearcher.h
@@ -1,5 +1,6 @@
 #pragma once
 
+#include <common/getPageSize.h>
 #include <Common/Exception.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/UTF8Helpers.h>
@@ -37,7 +38,7 @@ struct StringSearcherBase
 {
 #ifdef __SSE2__
     static constexpr auto n = sizeof(__m128i);
-    const int page_size = getpagesize();
+    const int page_size = ::getPageSize();
 
     bool pageSafe(const void * const ptr) const
     {
diff --git a/src/Common/ThreadProfileEvents.cpp b/src/Common/ThreadProfileEvents.cpp
index c8a4555dabf..e6336baecda 100644
--- a/src/Common/ThreadProfileEvents.cpp
+++ b/src/Common/ThreadProfileEvents.cpp
@@ -21,6 +21,9 @@
 #include <sys/types.h>
 #include <dirent.h>
 
+#include <common/errnoToString.h>
+
+
 namespace DB
 {
 
@@ -246,7 +249,7 @@ static void enablePerfEvent(int event_fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Can't enable perf event with file descriptor {}: '{}' ({})",
-            event_fd, strerror(errno), errno);
+            event_fd, errnoToString(errno), errno);
     }
 }
 
@@ -256,7 +259,7 @@ static void disablePerfEvent(int event_fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Can't disable perf event with file descriptor {}: '{}' ({})",
-            event_fd, strerror(errno), errno);
+            event_fd, errnoToString(errno), errno);
     }
 }
 
@@ -266,7 +269,7 @@ static void releasePerfEvent(int event_fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Can't close perf event file descriptor {}: {} ({})",
-            event_fd, strerror(errno), errno);
+            event_fd, errnoToString(errno), errno);
     }
 }
 
@@ -284,7 +287,7 @@ static bool validatePerfEventDescriptor(int & fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Error while checking availability of event descriptor {}: {} ({})",
-            fd, strerror(errno), errno);
+            fd, errnoToString(errno), errno);
 
         disablePerfEvent(fd);
         releasePerfEvent(fd);
@@ -391,7 +394,7 @@ bool PerfEventsCounters::processThreadLocalChanges(const std::string & needed_ev
             LOG_WARNING(&Poco::Logger::get("PerfEvents"),
                 "Failed to open perf event {} (event_type={}, event_config={}): "
                 "'{}' ({})", event_info.settings_name, event_info.event_type,
-                event_info.event_config, strerror(errno), errno);
+                event_info.event_config, errnoToString(errno), errno);
         }
     }
 
@@ -477,7 +480,7 @@ void PerfEventsCounters::finalizeProfileEvents(ProfileEvents::Counters & profile
         {
             LOG_WARNING(&Poco::Logger::get("PerfEvents"),
                 "Can't read event value from file descriptor {}: '{}' ({})",
-                fd, strerror(errno), errno);
+                fd, errnoToString(errno), errno);
             current_values[i] = {};
         }
     }
diff --git a/src/Common/ThreadStatus.cpp b/src/Common/ThreadStatus.cpp
index 3215bb55537..4bba3d8f4eb 100644
--- a/src/Common/ThreadStatus.cpp
+++ b/src/Common/ThreadStatus.cpp
@@ -2,11 +2,14 @@
 #include <Common/ThreadProfileEvents.h>
 #include <Common/QueryProfiler.h>
 #include <Common/ThreadStatus.h>
+#include <common/errnoToString.h>
 #include <Interpreters/OpenTelemetrySpanLog.h>
 
 #include <Poco/Logger.h>
 #include <common/getThreadId.h>
 
+#include <signal.h>
+
 
 namespace DB
 {
@@ -21,6 +24,11 @@ namespace ErrorCodes
 thread_local ThreadStatus * current_thread = nullptr;
 thread_local ThreadStatus * main_thread = nullptr;
 
+#if !defined(SANITIZER) && !defined(ARCADIA_BUILD)
+    alignas(4096) static thread_local char alt_stack[4096];
+    static thread_local bool has_alt_stack = false;
+#endif
+
 
 ThreadStatus::ThreadStatus()
     : thread_id{getThreadId()}
@@ -34,6 +42,46 @@ ThreadStatus::ThreadStatus()
 
     /// NOTE: It is important not to do any non-trivial actions (like updating ProfileEvents or logging) before ThreadStatus is created
     /// Otherwise it could lead to SIGSEGV due to current_thread dereferencing
+
+    /// Will set alternative signal stack to provide diagnostics for stack overflow errors.
+    /// If not already installed for current thread.
+    /// Sanitizer makes larger stack usage and also it's incompatible with alternative stack by default (it sets up and relies on its own).
+#if !defined(SANITIZER) && !defined(ARCADIA_BUILD)
+    if (!has_alt_stack)
+    {
+        /// Don't repeat tries even if not installed successfully.
+        has_alt_stack = true;
+
+        /// We have to call 'sigaltstack' before first 'sigaction'. (It does not work other way, for unknown reason).
+        stack_t altstack_description{};
+        altstack_description.ss_sp = alt_stack;
+        altstack_description.ss_flags = 0;
+        altstack_description.ss_size = sizeof(alt_stack);
+
+        if (0 != sigaltstack(&altstack_description, nullptr))
+        {
+            LOG_WARNING(log, "Cannot set alternative signal stack for thread, {}", errnoToString(errno));
+        }
+        else
+        {
+            /// Obtain existing sigaction and modify it by adding a flag.
+            struct sigaction action{};
+            if (0 != sigaction(SIGSEGV, nullptr, &action))
+            {
+                LOG_WARNING(log, "Cannot obtain previous signal action to set alternative signal stack for thread, {}", errnoToString(errno));
+            }
+            else if (!(action.sa_flags & SA_ONSTACK))
+            {
+                action.sa_flags |= SA_ONSTACK;
+
+                if (0 != sigaction(SIGSEGV, &action, nullptr))
+                {
+                    LOG_WARNING(log, "Cannot set action with alternative signal stack for thread, {}", errnoToString(errno));
+                }
+            }
+        }
+    }
+#endif
 }
 
 ThreadStatus::~ThreadStatus()
diff --git a/src/Common/UInt128.h b/src/Common/UInt128.h
index baa33578089..06fddee8dc9 100644
--- a/src/Common/UInt128.h
+++ b/src/Common/UInt128.h
@@ -28,23 +28,28 @@ struct UInt128
     UInt64 low;
     UInt64 high;
 
+    /// TODO: Make this constexpr. Currently it is used in unions
+    /// and union cannot contain member with non trivial constructor
+    /// constructor must be non user provided but compiler cannot constexpr constructor
+    /// if members low and high are not initialized, if we default member initialize them
+    /// constructor becomes non trivial.
     UInt128() = default;
-    explicit UInt128(const UInt64 low_, const UInt64 high_) : low(low_), high(high_) {}
+    explicit constexpr UInt128(const UInt64 low_, const UInt64 high_) : low(low_), high(high_) { }
 
     /// We need Int128 to UInt128 conversion or AccurateComparison will call greaterOp<Int128, UInt64> instead of greaterOp<Int128, UInt128>
-    explicit UInt128(const Int128 rhs) : low(rhs), high(rhs >> 64) {}
-    explicit UInt128(const Int64 rhs) : low(rhs), high() {}
-    explicit UInt128(const Int32 rhs) : low(rhs), high() {}
-    explicit UInt128(const Int16 rhs) : low(rhs), high() {}
-    explicit UInt128(const Int8 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt8 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt16 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt32 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt64 rhs) : low(rhs), high() {}
-    explicit UInt128(const Float32 rhs) : low(rhs), high() {}
-    explicit UInt128(const Float64 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int128 rhs) : low(rhs), high(rhs >> 64) {}
+    explicit constexpr UInt128(const Int64 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int32 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int16 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int8 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt8 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt16 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt32 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt64 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Float32 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Float64 rhs) : low(rhs), high() {}
 
-    auto tuple() const { return std::tie(high, low); }
+    constexpr auto tuple() const { return std::tie(high, low); }
 
     String toHexString() const
     {
@@ -53,31 +58,31 @@ struct UInt128
         return res;
     }
 
-    bool inline operator== (const UInt128 rhs) const { return tuple() == rhs.tuple(); }
-    bool inline operator!= (const UInt128 rhs) const { return tuple() != rhs.tuple(); }
-    bool inline operator<  (const UInt128 rhs) const { return tuple() < rhs.tuple(); }
-    bool inline operator<= (const UInt128 rhs) const { return tuple() <= rhs.tuple(); }
-    bool inline operator>  (const UInt128 rhs) const { return tuple() > rhs.tuple(); }
-    bool inline operator>= (const UInt128 rhs) const { return tuple() >= rhs.tuple(); }
+    constexpr bool operator== (const UInt128 rhs) const { return tuple() == rhs.tuple(); }
+    constexpr bool operator!= (const UInt128 rhs) const { return tuple() != rhs.tuple(); }
+    constexpr bool operator<  (const UInt128 rhs) const { return tuple() < rhs.tuple(); }
+    constexpr bool operator<= (const UInt128 rhs) const { return tuple() <= rhs.tuple(); }
+    constexpr bool operator>  (const UInt128 rhs) const { return tuple() > rhs.tuple(); }
+    constexpr bool operator>= (const UInt128 rhs) const { return tuple() >= rhs.tuple(); }
 
-    bool inline operator == (const Int128 rhs) const { return *this == UInt128(rhs, rhs >> 64) && rhs >= 0; }
-    bool inline operator != (const Int128 rhs) const { return *this != UInt128(rhs, rhs >> 64) || rhs < 0; }
-    bool inline operator >= (const Int128 rhs) const { return *this >= UInt128(rhs, rhs >> 64) || rhs < 0; }
-    bool inline operator >  (const Int128 rhs) const { return *this >  UInt128(rhs, rhs >> 64) || rhs < 0; }
-    bool inline operator <= (const Int128 rhs) const { return *this <= UInt128(rhs, rhs >> 64) && rhs >= 0; }
-    bool inline operator <  (const Int128 rhs) const { return *this <  UInt128(rhs, rhs >> 64) && rhs >= 0; }
+    constexpr bool operator == (const Int128 rhs) const { return *this == UInt128(rhs, rhs >> 64) && rhs >= 0; }
+    constexpr bool operator != (const Int128 rhs) const { return *this != UInt128(rhs, rhs >> 64) || rhs < 0; }
+    constexpr bool operator >= (const Int128 rhs) const { return *this >= UInt128(rhs, rhs >> 64) || rhs < 0; }
+    constexpr bool operator >  (const Int128 rhs) const { return *this >  UInt128(rhs, rhs >> 64) || rhs < 0; }
+    constexpr bool operator <= (const Int128 rhs) const { return *this <= UInt128(rhs, rhs >> 64) && rhs >= 0; }
+    constexpr bool operator <  (const Int128 rhs) const { return *this <  UInt128(rhs, rhs >> 64) && rhs >= 0; }
 
-    bool inline operator >  (const Int256 rhs) const { return (rhs < 0) || ((Int256(high) << 64) + low) > rhs; }
-    bool inline operator >  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) > rhs; }
-    bool inline operator <  (const Int256 rhs) const { return (rhs >= 0) && ((Int256(high) << 64) + low) < rhs; }
-    bool inline operator <  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) < rhs; }
+    constexpr bool operator >  (const Int256 rhs) const { return (rhs < 0) || ((Int256(high) << 64) + low) > rhs; }
+    constexpr bool operator >  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) > rhs; }
+    constexpr bool operator <  (const Int256 rhs) const { return (rhs >= 0) && ((Int256(high) << 64) + low) < rhs; }
+    constexpr bool operator <  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) < rhs; }
 
-    template <typename T> bool inline operator== (const T rhs) const { return *this == UInt128(rhs); }
-    template <typename T> bool inline operator!= (const T rhs) const { return *this != UInt128(rhs); }
-    template <typename T> bool inline operator>= (const T rhs) const { return *this >= UInt128(rhs); }
-    template <typename T> bool inline operator>  (const T rhs) const { return *this >  UInt128(rhs); }
-    template <typename T> bool inline operator<= (const T rhs) const { return *this <= UInt128(rhs); }
-    template <typename T> bool inline operator<  (const T rhs) const { return *this <  UInt128(rhs); }
+    template <typename T> constexpr bool operator== (const T rhs) const { return *this == UInt128(rhs); }
+    template <typename T> constexpr bool operator!= (const T rhs) const { return *this != UInt128(rhs); }
+    template <typename T> constexpr bool operator>= (const T rhs) const { return *this >= UInt128(rhs); }
+    template <typename T> constexpr bool operator>  (const T rhs) const { return *this >  UInt128(rhs); }
+    template <typename T> constexpr bool operator<= (const T rhs) const { return *this <= UInt128(rhs); }
+    template <typename T> constexpr bool operator<  (const T rhs) const { return *this <  UInt128(rhs); }
 
     template <typename T> explicit operator T() const
     {
@@ -91,15 +96,15 @@ struct UInt128
 #pragma GCC diagnostic pop
 #endif
 
-    UInt128 & operator= (const UInt64 rhs) { low = rhs; high = 0; return *this; }
+    constexpr UInt128 & operator= (const UInt64 rhs) { low = rhs; high = 0; return *this; }
 };
 
-template <typename T> bool inline operator == (T a, const UInt128 b) { return b.operator==(a); }
-template <typename T> bool inline operator != (T a, const UInt128 b) { return b.operator!=(a); }
-template <typename T> bool inline operator >= (T a, const UInt128 b) { return b <= a; }
-template <typename T> bool inline operator >  (T a, const UInt128 b) { return b < a; }
-template <typename T> bool inline operator <= (T a, const UInt128 b) { return b >= a; }
-template <typename T> bool inline operator <  (T a, const UInt128 b) { return b > a; }
+template <typename T> constexpr bool operator == (T a, const UInt128 b) { return b.operator==(a); }
+template <typename T> constexpr bool operator != (T a, const UInt128 b) { return b.operator!=(a); }
+template <typename T> constexpr bool operator >= (T a, const UInt128 b) { return b <= a; }
+template <typename T> constexpr bool operator >  (T a, const UInt128 b) { return b < a; }
+template <typename T> constexpr bool operator <= (T a, const UInt128 b) { return b >= a; }
+template <typename T> constexpr bool operator <  (T a, const UInt128 b) { return b > a; }
 
 template <> inline constexpr bool IsNumber<UInt128> = true;
 template <> struct TypeName<UInt128> { static constexpr const char * get() { return "UInt128"; } };
@@ -246,4 +251,42 @@ template <> struct hash<DB::UInt128>
     }
 };
 
+template<>
+class numeric_limits<DB::UInt128>
+{
+public:
+    static constexpr bool is_specialized = true;
+    static constexpr bool is_signed = ::is_signed<DB::UInt128>::value;
+    static constexpr bool is_integer = ::is_integer<DB::UInt128>::value;
+    static constexpr bool is_exact = true;
+    static constexpr bool has_infinity = false;
+    static constexpr bool has_quiet_NaN = false;
+    static constexpr bool has_signaling_NaN = false;
+    static constexpr std::float_denorm_style has_denorm = std::denorm_absent;
+    static constexpr bool has_denorm_loss = false;
+    static constexpr std::float_round_style round_style = std::round_toward_zero;
+    static constexpr bool is_iec559 = false;
+    static constexpr bool is_bounded = true;
+    static constexpr bool is_modulo = true;
+    static constexpr int digits = std::numeric_limits<UInt64>::digits * 2;
+    static constexpr int digits10 = digits * 0.30103 /*std::log10(2)*/;
+    static constexpr int max_digits10 = 0;
+    static constexpr int radix = 2;
+    static constexpr int min_exponent = 0;
+    static constexpr int min_exponent10 = 0;
+    static constexpr int max_exponent = 0;
+    static constexpr int max_exponent10 = 0;
+    static constexpr bool traps = true;
+    static constexpr bool tinyness_before = false;
+
+    static constexpr DB::UInt128 min() noexcept { return DB::UInt128(0, 0); }
+
+    static constexpr DB::UInt128 max() noexcept
+    {
+        return DB::UInt128(std::numeric_limits<UInt64>::max(), std::numeric_limits<UInt64>::max());
+    }
+
+    static constexpr DB::UInt128 lowest() noexcept { return min(); }
+};
+
 }
diff --git a/src/Common/ZooKeeper/TestKeeperStorage.cpp b/src/Common/ZooKeeper/TestKeeperStorage.cpp
index e3e4edc23a7..6b6ae93bbb9 100644
--- a/src/Common/ZooKeeper/TestKeeperStorage.cpp
+++ b/src/Common/ZooKeeper/TestKeeperStorage.cpp
@@ -798,6 +798,21 @@ void TestKeeperStorage::clearDeadWatches(int64_t session_id)
                 if (watches_for_path.empty())
                     watches.erase(watch);
             }
+
+            auto list_watch = list_watches.find(watch_path);
+            if (list_watch != list_watches.end())
+            {
+                auto & list_watches_for_path = list_watch->second;
+                for (auto w_it = list_watches_for_path.begin(); w_it != list_watches_for_path.end();)
+                {
+                    if (w_it->session_id == session_id)
+                        w_it = list_watches_for_path.erase(w_it);
+                    else
+                        ++w_it;
+                }
+                if (list_watches_for_path.empty())
+                    list_watches.erase(list_watch);
+            }
         }
         sessions_and_watchers.erase(watches_it);
     }
diff --git a/src/Common/config.h.in b/src/Common/config.h.in
index 3a67b2a8f2c..ee2bfe3df53 100644
--- a/src/Common/config.h.in
+++ b/src/Common/config.h.in
@@ -5,6 +5,7 @@
 #cmakedefine01 USE_RE2_ST
 #cmakedefine01 USE_SSL
 #cmakedefine01 USE_HDFS
+#cmakedefine01 USE_INTERNAL_HDFS3_LIBRARY
 #cmakedefine01 USE_AWS_S3
 #cmakedefine01 USE_BROTLI
 #cmakedefine01 USE_UNWIND
diff --git a/src/Common/ya.make b/src/Common/ya.make
index e515741a272..c8be43f98a2 100644
--- a/src/Common/ya.make
+++ b/src/Common/ya.make
@@ -7,7 +7,7 @@ ADDINCL (
     GLOBAL clickhouse/src
     contrib/libs/libcpuid
     contrib/libs/libunwind/include
-    GLOBAL contrib/restricted/ryu
+    GLOBAL contrib/restricted/dragonbox
 )
 
 PEERDIR(
@@ -18,7 +18,7 @@ PEERDIR(
     contrib/libs/openssl
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/re2
-    contrib/restricted/ryu
+    contrib/restricted/dragonbox
 )
 
 INCLUDE(${ARCADIA_ROOT}/clickhouse/cmake/yandex/ya.make.versions.inc)
diff --git a/src/Common/ya.make.in b/src/Common/ya.make.in
index d913832b91b..210ecab6ef6 100644
--- a/src/Common/ya.make.in
+++ b/src/Common/ya.make.in
@@ -6,7 +6,7 @@ ADDINCL (
     GLOBAL clickhouse/src
     contrib/libs/libcpuid
     contrib/libs/libunwind/include
-    GLOBAL contrib/restricted/ryu
+    GLOBAL contrib/restricted/dragonbox
 )
 
 PEERDIR(
@@ -17,7 +17,7 @@ PEERDIR(
     contrib/libs/openssl
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/re2
-    contrib/restricted/ryu
+    contrib/restricted/dragonbox
 )
 
 INCLUDE(${ARCADIA_ROOT}/clickhouse/cmake/yandex/ya.make.versions.inc)
diff --git a/src/Core/AccurateComparison.h b/src/Core/AccurateComparison.h
index 500346872db..c533cf0c7fb 100644
--- a/src/Core/AccurateComparison.h
+++ b/src/Core/AccurateComparison.h
@@ -515,11 +515,32 @@ inline bool NO_SANITIZE_UNDEFINED convertNumeric(From value, To & result)
         return true;
     }
 
-    /// Note that NaNs doesn't compare equal to anything, but they are still in range of any Float type.
-    if (isNaN(value) && std::is_floating_point_v<To>)
+    if constexpr (std::is_floating_point_v<From> && std::is_floating_point_v<To>)
     {
-        result = value;
-        return true;
+        /// Note that NaNs doesn't compare equal to anything, but they are still in range of any Float type.
+        if (isNaN(value))
+        {
+            result = value;
+            return true;
+        }
+
+        if (value == std::numeric_limits<From>::infinity())
+        {
+            result = std::numeric_limits<To>::infinity();
+            return true;
+        }
+
+        if (value == -std::numeric_limits<From>::infinity())
+        {
+            result = -std::numeric_limits<To>::infinity();
+            return true;
+        }
+    }
+
+    if (accurate::greaterOp(value, std::numeric_limits<To>::max())
+        || accurate::greaterOp(std::numeric_limits<To>::lowest(), value))
+    {
+        return false;
     }
 
     result = static_cast<To>(value);
diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h
index b82cfd88e98..2b916cbf538 100644
--- a/src/Core/DecimalFunctions.h
+++ b/src/Core/DecimalFunctions.h
@@ -6,6 +6,7 @@
 #include <common/arithmeticOverflow.h>
 
 #include <limits>
+#include <type_traits>
 
 
 namespace DB
@@ -206,23 +207,32 @@ inline typename DecimalType::NativeType getFractionalPart(const DecimalType & de
 }
 
 /// Decimal to integer/float conversion
-template <typename To, typename DecimalType>
-To convertTo(const DecimalType & decimal, size_t scale)
+template <typename To, typename DecimalType, typename ReturnType>
+ReturnType convertToImpl(const DecimalType & decimal, size_t scale, To & result)
 {
     using NativeT = typename DecimalType::NativeType;
+    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
 
     if constexpr (std::is_floating_point_v<To>)
     {
-        return static_cast<To>(decimal.value) / static_cast<To>(scaleMultiplier<NativeT>(scale));
+        result = static_cast<To>(decimal.value) / static_cast<To>(scaleMultiplier<NativeT>(scale));
     }
     else if constexpr (is_integer_v<To> && (sizeof(To) >= sizeof(NativeT)))
     {
         NativeT whole = getWholePart(decimal, scale);
 
         if constexpr (is_unsigned_v<To>)
+        {
             if (whole < 0)
-                throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
-        return static_cast<To>(whole);
+            {
+                if constexpr (throw_exception)
+                    throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
+                else
+                    return ReturnType(true);
+            }
+        }
+
+        result = static_cast<To>(whole);
     }
     else if constexpr (is_integer_v<To>)
     {
@@ -235,9 +245,34 @@ To convertTo(const DecimalType & decimal, size_t scale)
         static const constexpr CastTo max_to = std::numeric_limits<ToNativeT>::max();
 
         if (whole < min_to || whole > max_to)
-            throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
-        return static_cast<CastTo>(whole);
+        {
+            if constexpr (throw_exception)
+                throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return ReturnType(true);
+        }
+
+        result = static_cast<CastTo>(whole);
     }
+
+    return ReturnType(true);
+}
+
+
+template <typename To, typename DecimalType>
+To convertTo(const DecimalType & decimal, size_t scale)
+{
+    To result;
+
+    convertToImpl<To, DecimalType, void>(decimal, scale, result);
+
+    return result;
+}
+
+template <typename To, typename DecimalType>
+bool tryConvertTo(const DecimalType & decimal, size_t scale, To & result)
+{
+    return convertToImpl<To, DecimalType, bool>(decimal, scale, result);
 }
 
 template <bool is_multiply, bool is_division, typename T, typename U, template <typename> typename DecimalType>
diff --git a/src/Core/Field.cpp b/src/Core/Field.cpp
index 6fda3b39b3b..93107d7bb2c 100644
--- a/src/Core/Field.cpp
+++ b/src/Core/Field.cpp
@@ -17,6 +17,63 @@ namespace ErrorCodes
     extern const int DECIMAL_OVERFLOW;
 }
 
+inline Field getBinaryValue(UInt8 type, ReadBuffer & buf)
+{
+    switch (type)
+    {
+        case Field::Types::Null: {
+            return DB::Field();
+        }
+        case Field::Types::UInt64: {
+            UInt64 value;
+            DB::readVarUInt(value, buf);
+            return value;
+        }
+        case Field::Types::UInt128: {
+            UInt128 value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Int64: {
+            Int64 value;
+            DB::readVarInt(value, buf);
+            return value;
+        }
+        case Field::Types::Float64: {
+            Float64 value;
+            DB::readFloatBinary(value, buf);
+            return value;
+        }
+        case Field::Types::String: {
+            std::string value;
+            DB::readStringBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Array: {
+            Array value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Tuple: {
+            Tuple value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Map: {
+            Map value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::AggregateFunctionState: {
+            AggregateFunctionStateData value;
+            DB::readStringBinary(value.name, buf);
+            DB::readStringBinary(value.data, buf);
+            return value;
+        }
+    }
+    return DB::Field();
+}
+
 void readBinary(Array & x, ReadBuffer & buf)
 {
     size_t size;
@@ -25,73 +82,7 @@ void readBinary(Array & x, ReadBuffer & buf)
     DB::readBinary(size, buf);
 
     for (size_t index = 0; index < size; ++index)
-    {
-        switch (type)
-        {
-            case Field::Types::Null:
-            {
-                x.push_back(DB::Field());
-                break;
-            }
-            case Field::Types::UInt64:
-            {
-                UInt64 value;
-                DB::readVarUInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                UInt128 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                Int64 value;
-                DB::readVarInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                Float64 value;
-                DB::readFloatBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::String:
-            {
-                std::string value;
-                DB::readStringBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                Array value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                Tuple value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                AggregateFunctionStateData value;
-                DB::readStringBinary(value.name, buf);
-                DB::readStringBinary(value.data, buf);
-                x.push_back(value);
-                break;
-            }
-        }
-    }
+        x.push_back(getBinaryValue(type, buf));
 }
 
 void writeBinary(const Array & x, WriteBuffer & buf)
@@ -104,53 +95,7 @@ void writeBinary(const Array & x, WriteBuffer & buf)
     DB::writeBinary(size, buf);
 
     for (const auto & elem : x)
-    {
-        switch (type)
-        {
-            case Field::Types::Null: break;
-            case Field::Types::UInt64:
-            {
-                DB::writeVarUInt(get<UInt64>(elem), buf);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                DB::writeBinary(get<UInt128>(elem), buf);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                DB::writeVarInt(get<Int64>(elem), buf);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                DB::writeFloatBinary(get<Float64>(elem), buf);
-                break;
-            }
-            case Field::Types::String:
-            {
-                DB::writeStringBinary(get<std::string>(elem), buf);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                DB::writeBinary(get<Array>(elem), buf);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                DB::writeBinary(get<Tuple>(elem), buf);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().name, buf);
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().data, buf);
-                break;
-            }
-        }
-    }
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, elem);
 }
 
 void writeText(const Array & x, WriteBuffer & buf)
@@ -168,93 +113,7 @@ void readBinary(Tuple & x, ReadBuffer & buf)
     {
         UInt8 type;
         DB::readBinary(type, buf);
-
-        switch (type)
-        {
-            case Field::Types::Null:
-            {
-                x.push_back(DB::Field());
-                break;
-            }
-            case Field::Types::UInt64:
-            {
-                UInt64 value;
-                DB::readVarUInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                UInt128 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                Int64 value;
-                DB::readVarInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int128:
-            {
-                Int64 value;
-                DB::readVarInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                Float64 value;
-                DB::readFloatBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::String:
-            {
-                std::string value;
-                DB::readStringBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::UInt256:
-            {
-                UInt256 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int256:
-            {
-                Int256 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                Array value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                Tuple value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                AggregateFunctionStateData value;
-                DB::readStringBinary(value.name, buf);
-                DB::readStringBinary(value.data, buf);
-                x.push_back(value);
-                break;
-            }
-        }
+        x.push_back(getBinaryValue(type, buf));
     }
 }
 
@@ -267,67 +126,7 @@ void writeBinary(const Tuple & x, WriteBuffer & buf)
     {
         const UInt8 type = elem.getType();
         DB::writeBinary(type, buf);
-
-        switch (type)
-        {
-            case Field::Types::Null: break;
-            case Field::Types::UInt64:
-            {
-                DB::writeVarUInt(get<UInt64>(elem), buf);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                DB::writeBinary(get<UInt128>(elem), buf);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                DB::writeVarInt(get<Int64>(elem), buf);
-                break;
-            }
-            case Field::Types::Int128:
-            {
-                DB::writeVarInt(get<Int64>(elem), buf);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                DB::writeFloatBinary(get<Float64>(elem), buf);
-                break;
-            }
-            case Field::Types::String:
-            {
-                DB::writeStringBinary(get<std::string>(elem), buf);
-                break;
-            }
-            case Field::Types::UInt256:
-            {
-                DB::writeBinary(get<UInt256>(elem), buf);
-                break;
-            }
-            case Field::Types::Int256:
-            {
-                DB::writeBinary(get<Int256>(elem), buf);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                DB::writeBinary(get<Array>(elem), buf);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                DB::writeBinary(get<Tuple>(elem), buf);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().name, buf);
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().data, buf);
-                break;
-            }
-        }
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, elem);
     }
 }
 
@@ -336,6 +135,37 @@ void writeText(const Tuple & x, WriteBuffer & buf)
     writeFieldText(DB::Field(x), buf);
 }
 
+void readBinary(Map & x, ReadBuffer & buf)
+{
+    size_t size;
+    DB::readBinary(size, buf);
+
+    for (size_t index = 0; index < size; ++index)
+    {
+        UInt8 type;
+        DB::readBinary(type, buf);
+        x.push_back(getBinaryValue(type, buf));
+    }
+}
+
+void writeBinary(const Map & x, WriteBuffer & buf)
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (const auto & elem : x)
+    {
+        const UInt8 type = elem.getType();
+        DB::writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, elem);
+    }
+}
+
+void writeText(const Map & x, WriteBuffer & buf)
+{
+    writeFieldText(DB::Field(x), buf);
+}
+
 template <typename T>
 void readQuoted(DecimalField<T> & x, ReadBuffer & buf)
 {
@@ -530,6 +360,30 @@ Field Field::restoreFromDump(const std::string_view & dump_)
         return tuple;
     }
 
+    prefix = std::string_view{"Map_("};
+    if (dump.starts_with(prefix))
+    {
+        std::string_view tail = dump.substr(prefix.length());
+        trimLeft(tail);
+        Map map;
+        while (tail != ")")
+        {
+            size_t separator = tail.find_first_of(",)");
+            if (separator == std::string_view::npos)
+                show_error();
+            bool comma = (tail[separator] == ',');
+            std::string_view element = tail.substr(0, separator);
+            tail.remove_prefix(separator);
+            if (comma)
+                tail.remove_prefix(1);
+            trimLeft(tail);
+            if (!comma && tail != ")")
+                show_error();
+            map.push_back(Field::restoreFromDump(element));
+        }
+        return map;
+    }
+
     prefix = std::string_view{"AggregateFunctionState_("};
     if (dump.starts_with(prefix))
     {
diff --git a/src/Core/Field.h b/src/Core/Field.h
index 33f3a984c5c..3a52186167f 100644
--- a/src/Core/Field.h
+++ b/src/Core/Field.h
@@ -51,6 +51,9 @@ struct X : public FieldVector \
 DEFINE_FIELD_VECTOR(Array);
 DEFINE_FIELD_VECTOR(Tuple);
 
+/// An array with the following structure: [(key1, value1), (key2, value2), ...]
+DEFINE_FIELD_VECTOR(Map);
+
 #undef DEFINE_FIELD_VECTOR
 
 struct AggregateFunctionStateData
@@ -206,6 +209,7 @@ template <> struct NearestFieldTypeImpl<std::string_view> { using Type = String;
 template <> struct NearestFieldTypeImpl<String> { using Type = String; };
 template <> struct NearestFieldTypeImpl<Array> { using Type = Array; };
 template <> struct NearestFieldTypeImpl<Tuple> { using Type = Tuple; };
+template <> struct NearestFieldTypeImpl<Map> { using Type = Map; };
 template <> struct NearestFieldTypeImpl<bool> { using Type = UInt64; };
 template <> struct NearestFieldTypeImpl<Null> { using Type = Null; };
 
@@ -259,6 +263,7 @@ public:
             Decimal256 = 23,
             UInt256 = 24,
             Int256  = 25,
+            Map = 26,
         };
 
         static const int MIN_NON_POD = 16;
@@ -276,6 +281,7 @@ public:
                 case String:  return "String";
                 case Array:   return "Array";
                 case Tuple:   return "Tuple";
+                case Map:     return "Map";
                 case Decimal32:  return "Decimal32";
                 case Decimal64:  return "Decimal64";
                 case Decimal128: return "Decimal128";
@@ -464,6 +470,7 @@ public:
             case Types::String:  return get<String>()  < rhs.get<String>();
             case Types::Array:   return get<Array>()   < rhs.get<Array>();
             case Types::Tuple:   return get<Tuple>()   < rhs.get<Tuple>();
+            case Types::Map:     return get<Map>()     < rhs.get<Map>();
             case Types::Decimal32:  return get<DecimalField<Decimal32>>()  < rhs.get<DecimalField<Decimal32>>();
             case Types::Decimal64:  return get<DecimalField<Decimal64>>()  < rhs.get<DecimalField<Decimal64>>();
             case Types::Decimal128: return get<DecimalField<Decimal128>>() < rhs.get<DecimalField<Decimal128>>();
@@ -499,6 +506,7 @@ public:
             case Types::String:  return get<String>()  <= rhs.get<String>();
             case Types::Array:   return get<Array>()   <= rhs.get<Array>();
             case Types::Tuple:   return get<Tuple>()   <= rhs.get<Tuple>();
+            case Types::Map:     return get<Map>()     <= rhs.get<Map>();
             case Types::Decimal32:  return get<DecimalField<Decimal32>>()  <= rhs.get<DecimalField<Decimal32>>();
             case Types::Decimal64:  return get<DecimalField<Decimal64>>()  <= rhs.get<DecimalField<Decimal64>>();
             case Types::Decimal128: return get<DecimalField<Decimal128>>() <= rhs.get<DecimalField<Decimal128>>();
@@ -536,6 +544,7 @@ public:
             case Types::String:  return get<String>()  == rhs.get<String>();
             case Types::Array:   return get<Array>()   == rhs.get<Array>();
             case Types::Tuple:   return get<Tuple>()   == rhs.get<Tuple>();
+            case Types::Map:     return get<Map>()     == rhs.get<Map>();
             case Types::UInt128: return get<UInt128>() == rhs.get<UInt128>();
             case Types::Int128:  return get<Int128>()  == rhs.get<Int128>();
             case Types::Decimal32:  return get<DecimalField<Decimal32>>()  == rhs.get<DecimalField<Decimal32>>();
@@ -575,6 +584,7 @@ public:
             case Types::String:  return f(field.template get<String>());
             case Types::Array:   return f(field.template get<Array>());
             case Types::Tuple:   return f(field.template get<Tuple>());
+            case Types::Map:     return f(field.template get<Map>());
             case Types::Decimal32:  return f(field.template get<DecimalField<Decimal32>>());
             case Types::Decimal64:  return f(field.template get<DecimalField<Decimal64>>());
             case Types::Decimal128: return f(field.template get<DecimalField<Decimal128>>());
@@ -600,7 +610,7 @@ public:
 
 private:
     std::aligned_union_t<DBMS_MIN_FIELD_SIZE - sizeof(Types::Which),
-        Null, UInt64, UInt128, Int64, Int128, Float64, String, Array, Tuple,
+        Null, UInt64, UInt128, Int64, Int128, Float64, String, Array, Tuple, Map,
         DecimalField<Decimal32>, DecimalField<Decimal64>, DecimalField<Decimal128>, DecimalField<Decimal256>,
         AggregateFunctionStateData,
         UInt256, Int256
@@ -699,6 +709,9 @@ private:
             case Types::Tuple:
                 destroy<Tuple>();
                 break;
+            case Types::Map:
+                destroy<Map>();
+                break;
             case Types::AggregateFunctionState:
                 destroy<AggregateFunctionStateData>();
                 break;
@@ -729,6 +742,7 @@ template <> struct Field::TypeToEnum<Float64> { static const Types::Which value
 template <> struct Field::TypeToEnum<String>  { static const Types::Which value = Types::String; };
 template <> struct Field::TypeToEnum<Array>   { static const Types::Which value = Types::Array; };
 template <> struct Field::TypeToEnum<Tuple>   { static const Types::Which value = Types::Tuple; };
+template <> struct Field::TypeToEnum<Map>     { static const Types::Which value = Types::Map; };
 template <> struct Field::TypeToEnum<DecimalField<Decimal32>>{ static const Types::Which value = Types::Decimal32; };
 template <> struct Field::TypeToEnum<DecimalField<Decimal64>>{ static const Types::Which value = Types::Decimal64; };
 template <> struct Field::TypeToEnum<DecimalField<Decimal128>>{ static const Types::Which value = Types::Decimal128; };
@@ -747,6 +761,7 @@ template <> struct Field::EnumToType<Field::Types::Float64> { using Type = Float
 template <> struct Field::EnumToType<Field::Types::String>  { using Type = String; };
 template <> struct Field::EnumToType<Field::Types::Array>   { using Type = Array; };
 template <> struct Field::EnumToType<Field::Types::Tuple>   { using Type = Tuple; };
+template <> struct Field::EnumToType<Field::Types::Map>     { using Type = Map; };
 template <> struct Field::EnumToType<Field::Types::Decimal32> { using Type = DecimalField<Decimal32>; };
 template <> struct Field::EnumToType<Field::Types::Decimal64> { using Type = DecimalField<Decimal64>; };
 template <> struct Field::EnumToType<Field::Types::Decimal128> { using Type = DecimalField<Decimal128>; };
@@ -814,6 +829,7 @@ T safeGet(Field & field)
 
 template <> struct TypeName<Array> { static std::string get() { return "Array"; } };
 template <> struct TypeName<Tuple> { static std::string get() { return "Tuple"; } };
+template <> struct TypeName<Map>   { static std::string get() { return "Map"; } };
 template <> struct TypeName<AggregateFunctionStateData> { static std::string get() { return "AggregateFunctionState"; } };
 
 template <typename T>
@@ -900,6 +916,12 @@ void writeBinary(const Tuple & x, WriteBuffer & buf);
 
 void writeText(const Tuple & x, WriteBuffer & buf);
 
+void readBinary(Map & x, ReadBuffer & buf);
+[[noreturn]] inline void readText(Map &, ReadBuffer &) { throw Exception("Cannot read Map.", ErrorCodes::NOT_IMPLEMENTED); }
+[[noreturn]] inline void readQuoted(Map &, ReadBuffer &) { throw Exception("Cannot read Map.", ErrorCodes::NOT_IMPLEMENTED); }
+void writeBinary(const Map & x, WriteBuffer & buf);
+void writeText(const Map & x, WriteBuffer & buf);
+[[noreturn]] inline void writeQuoted(const Map &, WriteBuffer &) { throw Exception("Cannot write Map quoted.", ErrorCodes::NOT_IMPLEMENTED); }
 
 __attribute__ ((noreturn)) inline void writeText(const AggregateFunctionStateData &, WriteBuffer &)
 {
diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp
index 4e59293c88d..2a8dd2acdc0 100644
--- a/src/Core/Settings.cpp
+++ b/src/Core/Settings.cpp
@@ -16,10 +16,8 @@ namespace ErrorCodes
     extern const int UNKNOWN_ELEMENT_IN_CONFIG;
 }
 
-
 IMPLEMENT_SETTINGS_TRAITS(SettingsTraits, LIST_OF_SETTINGS)
 
-
 /** Set the settings from the profile (in the server configuration, many settings can be listed in one profile).
     * The profile can also be set using the `set` functions, like the `profile` setting.
     */
diff --git a/src/Core/Settings.h b/src/Core/Settings.h
index 86b9d1a197e..0555326022f 100644
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@@ -65,6 +65,7 @@ class IColumn;
     M(UInt64, distributed_connections_pool_size, DBMS_DEFAULT_DISTRIBUTED_CONNECTIONS_POOL_SIZE, "Maximum number of connections with one remote server in the pool.", 0) \
     M(UInt64, connections_with_failover_max_tries, DBMS_CONNECTION_POOL_WITH_FAILOVER_DEFAULT_MAX_TRIES, "The maximum number of attempts to connect to replicas.", 0) \
     M(UInt64, s3_min_upload_part_size, 512*1024*1024, "The minimum size of part to upload during multipart upload to S3.", 0) \
+    M(UInt64, s3_max_single_part_upload_size, 64*1024*1024, "The maximum size of object to upload using singlepart upload to S3.", 0) \
     M(UInt64, s3_max_redirects, 10, "Max number of S3 redirects hops allowed.", 0) \
     M(Bool, extremes, false, "Calculate minimums and maximums of the result columns. They can be output in JSON-formats.", IMPORTANT) \
     M(Bool, use_uncompressed_cache, true, "Whether to use the cache of uncompressed blocks.", 0) \
@@ -165,7 +166,7 @@ class IColumn;
     \
     M(UInt64, insert_quorum, 0, "For INSERT queries in the replicated table, wait writing for the specified number of replicas and linearize the addition of the data. 0 - disabled.", 0) \
     M(Milliseconds, insert_quorum_timeout, 600000, "", 0) \
-    M(Bool, insert_quorum_parallel, false, "For quorum INSERT queries - enable to make parallel inserts without linearizability", 0) \
+    M(Bool, insert_quorum_parallel, true, "For quorum INSERT queries - enable to make parallel inserts without linearizability", 0) \
     M(UInt64, select_sequential_consistency, 0, "For SELECT queries from the replicated table, throw an exception if the replica does not have a chunk written with the quorum; do not read the parts that have not yet been written with the quorum.", 0) \
     M(UInt64, table_function_remote_max_addresses, 1000, "The maximum number of different shards and the maximum number of replicas of one shard in the `remote` function.", 0) \
     M(Milliseconds, read_backoff_min_latency_ms, 1000, "Setting to reduce the number of threads in case of slow reads. Pay attention only to reads that took at least that much time.", 0) \
@@ -238,6 +239,8 @@ class IColumn;
       * Almost all limits apply to each stream individually. \
       */ \
     \
+    M(UInt64, limit, 0, "Limit on read rows from the most 'end' result for select query, default 0 means no limit length", 0) \
+    M(UInt64, offset, 0, "Offset on read rows from the most 'end' result for select query", 0) \
     M(UInt64, max_rows_to_read, 0, "Limit on read rows from the most 'deep' sources. That is, only in the deepest subquery. When reading from a remote server, it is only checked on a remote server.", 0) \
     M(UInt64, max_bytes_to_read, 0, "Limit on read bytes (after decompression) from the most 'deep' sources. That is, only in the deepest subquery. When reading from a remote server, it is only checked on a remote server.", 0) \
     M(OverflowMode, read_overflow_mode, OverflowMode::THROW, "What to do when the limit is exceeded.", 0) \
@@ -400,6 +403,8 @@ class IColumn;
     M(Bool, enable_global_with_statement, false, "Propagate WITH statements to UNION queries and all subqueries", 0) \
     M(Bool, aggregate_functions_null_for_empty, false, "Rewrite all aggregate functions in a query, adding -OrNull suffix to them", 0) \
     M(Bool, optimize_skip_merged_partitions, false, "Skip partitions with one part with level > 0 in optimize final", 0) \
+    M(Bool, optimize_on_insert, true, "Do the same transformation for inserted block of data as if merge was done on this block.", 0) \
+    M(Bool, allow_experimental_map_type, false, "Allow data type Map", 0) \
     \
     M(Bool, use_antlr_parser, false, "Parse incoming queries using ANTLR-generated parser", 0) \
     \
diff --git a/src/Core/SortCursor.h b/src/Core/SortCursor.h
index 7a222f70199..f383c3ded8e 100644
--- a/src/Core/SortCursor.h
+++ b/src/Core/SortCursor.h
@@ -30,7 +30,6 @@ struct SortCursorImpl
     ColumnRawPtrs all_columns;
     SortDescription desc;
     size_t sort_columns_size = 0;
-    size_t pos = 0;
     size_t rows = 0;
 
     /** Determines order if comparing columns are equal.
@@ -49,15 +48,20 @@ struct SortCursorImpl
     /** Is there at least one column with Collator. */
     bool has_collation = false;
 
+    /** We could use SortCursorImpl in case when columns aren't sorted
+      *  but we have their sorted permutation
+      */
+    IColumn::Permutation * permutation = nullptr;
+
     SortCursorImpl() {}
 
-    SortCursorImpl(const Block & block, const SortDescription & desc_, size_t order_ = 0)
+    SortCursorImpl(const Block & block, const SortDescription & desc_, size_t order_ = 0, IColumn::Permutation * perm = nullptr)
         : desc(desc_), sort_columns_size(desc.size()), order(order_), need_collation(desc.size())
     {
-        reset(block);
+        reset(block, perm);
     }
 
-    SortCursorImpl(const Columns & columns, const SortDescription & desc_, size_t order_ = 0)
+    SortCursorImpl(const Columns & columns, const SortDescription & desc_, size_t order_ = 0, IColumn::Permutation * perm = nullptr)
         : desc(desc_), sort_columns_size(desc.size()), order(order_), need_collation(desc.size())
     {
         for (auto & column_desc : desc)
@@ -66,19 +70,19 @@ struct SortCursorImpl
                 throw Exception("SortDescription should contain column position if SortCursor was used without header.",
                         ErrorCodes::LOGICAL_ERROR);
         }
-        reset(columns, {});
+        reset(columns, {}, perm);
     }
 
     bool empty() const { return rows == 0; }
 
     /// Set the cursor to the beginning of the new block.
-    void reset(const Block & block)
+    void reset(const Block & block, IColumn::Permutation * perm = nullptr)
     {
-        reset(block.getColumns(), block);
+        reset(block.getColumns(), block, perm);
     }
 
     /// Set the cursor to the beginning of the new block.
-    void reset(const Columns & columns, const Block & block)
+    void reset(const Columns & columns, const Block & block, IColumn::Permutation * perm = nullptr)
     {
         all_columns.clear();
         sort_columns.clear();
@@ -96,18 +100,33 @@ struct SortCursorImpl
                                    : column_desc.column_number;
             sort_columns.push_back(columns[column_number].get());
 
-            need_collation[j] = desc[j].collator != nullptr && sort_columns.back()->isCollationSupported();    /// TODO Nullable(String)
+            need_collation[j] = desc[j].collator != nullptr && sort_columns.back()->isCollationSupported();
             has_collation |= need_collation[j];
         }
 
         pos = 0;
         rows = all_columns[0]->size();
+        permutation = perm;
     }
 
+    size_t getRow() const
+    {
+        if (permutation)
+            return (*permutation)[pos];
+        return pos;
+    }
+
+    /// We need a possibility to change pos (see MergeJoin).
+    size_t & getPosRef() { return pos; }
+
     bool isFirst() const { return pos == 0; }
     bool isLast() const { return pos + 1 >= rows; }
     bool isValid() const { return pos < rows; }
     void next() { ++pos; }
+
+/// Prevent using pos instead of getRow()
+private:
+    size_t pos;
 };
 
 using SortCursorImpls = std::vector<SortCursorImpl>;
@@ -127,7 +146,7 @@ struct SortCursorHelper
 
     bool ALWAYS_INLINE greater(const SortCursorHelper & rhs) const
     {
-        return derived().greaterAt(rhs.derived(), impl->pos, rhs.impl->pos);
+        return derived().greaterAt(rhs.derived(), impl->getRow(), rhs.impl->getRow());
     }
 
     /// Inverted so that the priority queue elements are removed in ascending order.
diff --git a/src/Core/Types.h b/src/Core/Types.h
index 84c676b2204..0d578e17517 100644
--- a/src/Core/Types.h
+++ b/src/Core/Types.h
@@ -56,6 +56,7 @@ enum class TypeIndex
     Function,
     AggregateFunction,
     LowCardinality,
+    Map,
 };
 #if !__clang__
 #pragma GCC diagnostic pop
@@ -267,6 +268,7 @@ inline constexpr const char * getTypeName(TypeIndex idx)
         case TypeIndex::Function:   return "Function";
         case TypeIndex::AggregateFunction: return "AggregateFunction";
         case TypeIndex::LowCardinality: return "LowCardinality";
+        case TypeIndex::Map:        return "Map";
     }
 
     __builtin_unreachable();
diff --git a/src/DataStreams/MergingSortedBlockInputStream.cpp b/src/DataStreams/MergingSortedBlockInputStream.cpp
index ee6b93dc8e4..b7396a23d6a 100644
--- a/src/DataStreams/MergingSortedBlockInputStream.cpp
+++ b/src/DataStreams/MergingSortedBlockInputStream.cpp
@@ -222,7 +222,7 @@ void MergingSortedBlockInputStream::merge(MutableColumns & merged_columns, TSort
 //        std::cerr << "total_merged_rows: " << total_merged_rows << ", merged_rows: " << merged_rows << "\n";
 //        std::cerr << "Inserting row\n";
         for (size_t i = 0; i < num_columns; ++i)
-            merged_columns[i]->insertFrom(*current->all_columns[i], current->pos);
+            merged_columns[i]->insertFrom(*current->all_columns[i], current->getRow());
 
         if (out_row_sources_buf)
         {
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index eb03472504d..cec92a5f0e9 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -3,6 +3,7 @@
 #include <Client/ConnectionPool.h>
 #include <Client/MultiplexedConnections.h>
 #include <Storages/IStorage_fwd.h>
+#include <Interpreters/Context.h>
 #include <Interpreters/StorageID.h>
 
 namespace DB
@@ -93,7 +94,7 @@ private:
 
     const String query;
     String query_id = "";
-    const Context & context;
+    Context context;
 
     ProgressCallback progress_callback;
     ProfileInfoCallback profile_info_callback;
diff --git a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp
index d82adf33108..6b2f94aa7b5 100644
--- a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp
+++ b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp
@@ -25,10 +25,19 @@ namespace ErrorCodes
     extern const int LOGICAL_ERROR;
 }
 
-static const std::vector<String> supported_functions{"any", "anyLast", "min",
-    "max", "sum", "sumWithOverflow", "groupBitAnd", "groupBitOr", "groupBitXor",
-    "sumMap", "minMap", "maxMap", "groupArrayArray", "groupUniqArrayArray"};
+void DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(const AggregateFunctionPtr & function)
+{
+    static const std::vector<String> supported_functions{"any", "anyLast", "min",
+        "max", "sum", "sumWithOverflow", "groupBitAnd", "groupBitOr", "groupBitXor",
+        "sumMap", "minMap", "maxMap", "groupArrayArray", "groupUniqArrayArray"};
 
+    // check function
+    if (std::find(std::begin(supported_functions), std::end(supported_functions), function->getName()) == std::end(supported_functions))
+    {
+        throw Exception("Unsupported aggregate function " + function->getName() + ", supported functions are " + boost::algorithm::join(supported_functions, ","),
+                ErrorCodes::BAD_ARGUMENTS);
+    }
+}
 
 String DataTypeCustomSimpleAggregateFunction::getName() const
 {
@@ -114,12 +123,7 @@ static std::pair<DataTypePtr, DataTypeCustomDescPtr> create(const ASTPtr & argum
     AggregateFunctionProperties properties;
     function = AggregateFunctionFactory::instance().get(function_name, argument_types, params_row, properties);
 
-    // check function
-    if (std::find(std::begin(supported_functions), std::end(supported_functions), function->getName()) == std::end(supported_functions))
-    {
-        throw Exception("Unsupported aggregate function " + function->getName() + ", supported functions are " + boost::algorithm::join(supported_functions, ","),
-                        ErrorCodes::BAD_ARGUMENTS);
-    }
+    DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(function);
 
     DataTypePtr storage_type = DataTypeFactory::instance().get(argument_types[0]->getName());
 
diff --git a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h
index 0b75cd0ef88..44b007ab07f 100644
--- a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h
+++ b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h
@@ -37,6 +37,7 @@ public:
 
     const AggregateFunctionPtr getFunction() const { return function; }
     String getName() const override;
+    static void checkSupportedFunctions(const AggregateFunctionPtr & function);
 };
 
 }
diff --git a/src/DataTypes/DataTypeFactory.cpp b/src/DataTypes/DataTypeFactory.cpp
index 5052a065163..1ff00d97e84 100644
--- a/src/DataTypes/DataTypeFactory.cpp
+++ b/src/DataTypes/DataTypeFactory.cpp
@@ -180,6 +180,7 @@ DataTypeFactory::DataTypeFactory()
     registerDataTypeDomainIPv4AndIPv6(*this);
     registerDataTypeDomainSimpleAggregateFunction(*this);
     registerDataTypeDomainGeo(*this);
+    registerDataTypeMap(*this);
 }
 
 DataTypeFactory & DataTypeFactory::instance()
diff --git a/src/DataTypes/DataTypeFactory.h b/src/DataTypes/DataTypeFactory.h
index ea77c50170c..192b1beae5d 100644
--- a/src/DataTypes/DataTypeFactory.h
+++ b/src/DataTypes/DataTypeFactory.h
@@ -73,6 +73,7 @@ void registerDataTypeFixedString(DataTypeFactory & factory);
 void registerDataTypeEnum(DataTypeFactory & factory);
 void registerDataTypeArray(DataTypeFactory & factory);
 void registerDataTypeTuple(DataTypeFactory & factory);
+void registerDataTypeMap(DataTypeFactory & factory);
 void registerDataTypeNullable(DataTypeFactory & factory);
 void registerDataTypeNothing(DataTypeFactory & factory);
 void registerDataTypeUUID(DataTypeFactory & factory);
diff --git a/src/DataTypes/DataTypeMap.cpp b/src/DataTypes/DataTypeMap.cpp
new file mode 100644
index 00000000000..2047928721a
--- /dev/null
+++ b/src/DataTypes/DataTypeMap.cpp
@@ -0,0 +1,375 @@
+#include <Common/StringUtils/StringUtils.h>
+#include <Columns/ColumnMap.h>
+#include <Columns/ColumnArray.h>
+#include <Core/Field.h>
+#include <Formats/FormatSettings.h>
+#include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypeArray.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeFactory.h>
+#include <Parsers/IAST.h>
+#include <Parsers/ASTNameTypePair.h>
+#include <Common/typeid_cast.h>
+#include <Common/assert_cast.h>
+#include <Common/quoteString.h>
+#include <IO/WriteHelpers.h>
+#include <IO/ReadHelpers.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/ReadBufferFromString.h>
+#include <IO/Operators.h>
+
+#include <ext/map.h>
+#include <ext/enumerate.h>
+
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+    extern const int CANNOT_READ_MAP_FROM_TEXT;
+}
+
+
+DataTypeMap::DataTypeMap(const DataTypes & elems_)
+{
+    assert(elems_.size() == 2);
+    key_type = elems_[0];
+    value_type = elems_[1];
+
+    nested = std::make_shared<DataTypeArray>(
+        std::make_shared<DataTypeTuple>(DataTypes{key_type, value_type}, Names{"keys", "values"}));
+}
+
+DataTypeMap::DataTypeMap(const DataTypePtr & key_type_, const DataTypePtr & value_type_)
+    : key_type(key_type_), value_type(value_type_)
+    , nested(std::make_shared<DataTypeArray>(
+        std::make_shared<DataTypeTuple>(DataTypes{key_type_, value_type_}, Names{"keys", "values"}))) {}
+
+std::string DataTypeMap::doGetName() const
+{
+    WriteBufferFromOwnString s;
+    s << "Map(" << key_type->getName() << "," << value_type->getName() << ")";
+
+    return s.str();
+}
+
+static const IColumn & extractNestedColumn(const IColumn & column)
+{
+    return assert_cast<const ColumnMap &>(column).getNestedColumn();
+}
+
+static IColumn & extractNestedColumn(IColumn & column)
+{
+    return assert_cast<ColumnMap &>(column).getNestedColumn();
+}
+
+void DataTypeMap::serializeBinary(const Field & field, WriteBuffer & ostr) const
+{
+    const auto & map = get<const Map &>(field);
+    writeVarUInt(map.size(), ostr);
+    for (const auto & elem : map)
+    {
+        const auto & tuple = elem.safeGet<const Tuple>();
+        assert(tuple.size() == 2);
+        key_type->serializeBinary(tuple[0], ostr);
+        value_type->serializeBinary(tuple[1], ostr);
+    }
+}
+
+void DataTypeMap::deserializeBinary(Field & field, ReadBuffer & istr) const
+{
+    size_t size;
+    readVarUInt(size, istr);
+    field = Map(size);
+    for (auto & elem : field.get<Map &>())
+    {
+        Tuple tuple(2);
+        key_type->deserializeBinary(tuple[0], istr);
+        value_type->deserializeBinary(tuple[1], istr);
+        elem = std::move(tuple);
+    }
+}
+
+void DataTypeMap::serializeBinary(const IColumn & column, size_t row_num, WriteBuffer & ostr) const
+{
+    nested->serializeBinary(extractNestedColumn(column), row_num, ostr);
+}
+
+void DataTypeMap::deserializeBinary(IColumn & column, ReadBuffer & istr) const
+{
+    nested->deserializeBinary(extractNestedColumn(column), istr);
+}
+
+
+template <typename Writer>
+void DataTypeMap::serializeTextImpl(const IColumn & column, size_t row_num, WriteBuffer & ostr, Writer && writer) const
+{
+    const auto & column_map = assert_cast<const ColumnMap &>(column);
+
+    const auto & nested_array = column_map.getNestedColumn();
+    const auto & nested_tuple = column_map.getNestedData();
+    const auto & offsets = nested_array.getOffsets();
+
+    size_t offset = offsets[row_num - 1];
+    size_t next_offset = offsets[row_num];
+
+    writeChar('{', ostr);
+    for (size_t i = offset; i < next_offset; ++i)
+    {
+        if (i != offset)
+            writeChar(',', ostr);
+        writer(key_type, nested_tuple.getColumn(0), i);
+        writeChar(':', ostr);
+        writer(value_type, nested_tuple.getColumn(1), i);
+    }
+    writeChar('}', ostr);
+}
+
+template <typename Reader>
+void DataTypeMap::deserializeTextImpl(IColumn & column, ReadBuffer & istr, bool need_safe_get_int_key, Reader && reader) const
+{
+    auto & column_map = assert_cast<ColumnMap &>(column);
+
+    auto & nested_array = column_map.getNestedColumn();
+    auto & nested_tuple = column_map.getNestedData();
+    auto & offsets = nested_array.getOffsets();
+
+    auto & key_column = nested_tuple.getColumn(0);
+    auto & value_column = nested_tuple.getColumn(1);
+
+    size_t size = 0;
+    assertChar('{', istr);
+
+    try
+    {
+        bool first = true;
+        while (!istr.eof() && *istr.position() != '}')
+        {
+            if (!first)
+            {
+                if (*istr.position() == ',')
+                    ++istr.position();
+                else
+                    throw Exception("Cannot read Map from text", ErrorCodes::CANNOT_READ_MAP_FROM_TEXT);
+            }
+
+            first = false;
+
+            skipWhitespaceIfAny(istr);
+
+            if (*istr.position() == '}')
+                break;
+
+            if (need_safe_get_int_key)
+            {
+                ReadBuffer::Position tmp = istr.position();
+                while (*tmp != ':' && *tmp != '}')
+                    ++tmp;
+                *tmp = ' ';
+                reader(key_type, key_column);
+            }
+            else
+            {
+                reader(key_type, key_column);
+                skipWhitespaceIfAny(istr);
+                assertChar(':', istr);
+            }
+
+            ++size;
+            skipWhitespaceIfAny(istr);
+            reader(value_type, value_column);
+
+            skipWhitespaceIfAny(istr);
+        }
+
+        offsets.push_back(offsets.back() + size);
+        assertChar('}', istr);
+    }
+    catch (...)
+    {
+        throw;
+    }
+}
+
+void DataTypeMap::serializeText(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    serializeTextImpl(column, row_num, ostr,
+        [&](const DataTypePtr & subcolumn_type, const IColumn & subcolumn, size_t pos)
+        {
+            subcolumn_type->serializeAsTextQuoted(subcolumn, pos, ostr, settings);
+        });
+}
+
+void DataTypeMap::deserializeText(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    // need_safe_get_int_key is set for Interger to prevent to readIntTextUnsafe
+    bool need_safe_get_int_key = isInteger(key_type);
+
+    deserializeTextImpl(column, istr, need_safe_get_int_key,
+        [&](const DataTypePtr & subcolumn_type, IColumn & subcolumn)
+        {
+            subcolumn_type->deserializeAsTextQuoted(subcolumn, istr, settings);
+        });
+}
+
+
+void DataTypeMap::serializeTextJSON(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    serializeTextImpl(column, row_num, ostr,
+        [&](const DataTypePtr & subcolumn_type, const IColumn & subcolumn, size_t pos)
+        {
+            subcolumn_type->serializeAsTextJSON(subcolumn, pos, ostr, settings);
+        });
+}
+
+void DataTypeMap::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    // need_safe_get_int_key is set for Interger to prevent to readIntTextUnsafe
+    bool need_safe_get_int_key = isInteger(key_type);
+
+    deserializeTextImpl(column, istr, need_safe_get_int_key,
+        [&](const DataTypePtr & subcolumn_type, IColumn & subcolumn)
+        {
+            subcolumn_type->deserializeAsTextJSON(subcolumn, istr, settings);
+        });
+}
+
+void DataTypeMap::serializeTextXML(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    const auto & column_map = assert_cast<const ColumnMap &>(column);
+    const auto & offsets = column_map.getNestedColumn().getOffsets();
+
+    size_t offset = offsets[row_num - 1];
+    size_t next_offset = offsets[row_num];
+
+    const auto & nested_data = column_map.getNestedData();
+
+    writeCString("<map>", ostr);
+    for (size_t i = offset; i < next_offset; ++i)
+    {
+        writeCString("<elem>", ostr);
+        writeCString("<key>", ostr);
+        key_type->serializeAsTextXML(nested_data.getColumn(0), i, ostr, settings);
+        writeCString("</key>", ostr);
+
+        writeCString("<value>", ostr);
+        value_type->serializeAsTextXML(nested_data.getColumn(1), i, ostr, settings);
+        writeCString("</value>", ostr);
+        writeCString("</elem>", ostr);
+    }
+    writeCString("</map>", ostr);
+}
+
+void DataTypeMap::serializeTextCSV(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    WriteBufferFromOwnString wb;
+    serializeText(column, row_num, wb, settings);
+    writeCSV(wb.str(), ostr);
+}
+
+void DataTypeMap::deserializeTextCSV(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    String s;
+    readCSV(s, istr, settings.csv);
+    ReadBufferFromString rb(s);
+    deserializeText(column, rb, settings);
+}
+
+
+void DataTypeMap::enumerateStreams(const StreamCallback & callback, SubstreamPath & path) const
+{
+    nested->enumerateStreams(callback, path);
+}
+
+void DataTypeMap::serializeBinaryBulkStatePrefix(
+    SerializeBinaryBulkSettings & settings,
+    SerializeBinaryBulkStatePtr & state) const
+{
+    nested->serializeBinaryBulkStatePrefix(settings, state);
+}
+
+void DataTypeMap::serializeBinaryBulkStateSuffix(
+    SerializeBinaryBulkSettings & settings,
+    SerializeBinaryBulkStatePtr & state) const
+{
+    nested->serializeBinaryBulkStateSuffix(settings, state);
+}
+
+void DataTypeMap::deserializeBinaryBulkStatePrefix(
+        DeserializeBinaryBulkSettings & settings,
+        DeserializeBinaryBulkStatePtr & state) const
+{
+    nested->deserializeBinaryBulkStatePrefix(settings, state);
+}
+
+
+void DataTypeMap::serializeBinaryBulkWithMultipleStreams(
+    const IColumn & column,
+    size_t offset,
+    size_t limit,
+    SerializeBinaryBulkSettings & settings,
+    SerializeBinaryBulkStatePtr & state) const
+{
+    nested->serializeBinaryBulkWithMultipleStreams(extractNestedColumn(column), offset, limit, settings, state);
+}
+
+void DataTypeMap::deserializeBinaryBulkWithMultipleStreams(
+    IColumn & column,
+    size_t limit,
+    DeserializeBinaryBulkSettings & settings,
+    DeserializeBinaryBulkStatePtr & state) const
+{
+    nested->deserializeBinaryBulkWithMultipleStreams(extractNestedColumn(column), limit, settings, state);
+}
+
+void DataTypeMap::serializeProtobuf(const IColumn & column, size_t row_num, ProtobufWriter & protobuf, size_t & value_index) const
+{
+    nested->serializeProtobuf(extractNestedColumn(column), row_num, protobuf, value_index);
+}
+
+void DataTypeMap::deserializeProtobuf(IColumn & column, ProtobufReader & protobuf, bool allow_add_row, bool & row_added) const
+{
+    nested->deserializeProtobuf(extractNestedColumn(column), protobuf, allow_add_row, row_added);
+}
+
+MutableColumnPtr DataTypeMap::createColumn() const
+{
+    return ColumnMap::create(nested->createColumn());
+}
+
+Field DataTypeMap::getDefault() const
+{
+    return Map();
+}
+
+bool DataTypeMap::equals(const IDataType & rhs) const
+{
+    if (typeid(rhs) != typeid(*this))
+        return false;
+
+    const DataTypeMap & rhs_map = static_cast<const DataTypeMap &>(rhs);
+    return nested->equals(*rhs_map.nested);
+}
+
+static DataTypePtr create(const ASTPtr & arguments)
+{
+    if (!arguments || arguments->children.size() != 2)
+        throw Exception("Map data type family must have two arguments: key and value types", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+    DataTypes nested_types;
+    nested_types.reserve(arguments->children.size());
+
+    for (const ASTPtr & child : arguments->children)
+        nested_types.emplace_back(DataTypeFactory::instance().get(child));
+
+    return std::make_shared<DataTypeMap>(nested_types);
+}
+
+
+void registerDataTypeMap(DataTypeFactory & factory)
+{
+    factory.registerDataType("Map", create);
+}
+}
diff --git a/src/DataTypes/DataTypeMap.h b/src/DataTypes/DataTypeMap.h
new file mode 100644
index 00000000000..a52969428f4
--- /dev/null
+++ b/src/DataTypes/DataTypeMap.h
@@ -0,0 +1,103 @@
+#pragma once
+
+#include <DataTypes/DataTypeWithSimpleSerialization.h>
+
+
+namespace DB
+{
+
+/** Map data type.
+  * Map is implemented as two arrays of keys and values.
+  * Serialization of type 'Map(K, V)' is similar to serialization.
+  * of 'Array(Tuple(keys K, values V))' or in other words of 'Nested(keys K, valuev V)'.
+  */
+class DataTypeMap final : public DataTypeWithSimpleSerialization
+{
+private:
+    DataTypePtr key_type;
+    DataTypePtr value_type;
+
+    /// 'nested' is an Array(Tuple(key_type, value_type))
+    DataTypePtr nested;
+
+public:
+    static constexpr bool is_parametric = true;
+
+    DataTypeMap(const DataTypes & elems);
+    DataTypeMap(const DataTypePtr & key_type_, const DataTypePtr & value_type_);
+
+    TypeIndex getTypeId() const override { return TypeIndex::Map; }
+    std::string doGetName() const override;
+    const char * getFamilyName() const override { return "Map"; }
+
+    bool canBeInsideNullable() const override { return false; }
+
+    void serializeBinary(const Field & field, WriteBuffer & ostr) const override;
+    void deserializeBinary(Field & field, ReadBuffer & istr) const override;
+    void serializeBinary(const IColumn & column, size_t row_num, WriteBuffer & ostr) const override;
+    void deserializeBinary(IColumn & column, ReadBuffer & istr) const override;
+    void serializeText(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+    void deserializeText(IColumn & column, ReadBuffer & istr, const FormatSettings &) const override;
+    void serializeTextJSON(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+    void deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings &) const override;
+    void serializeTextXML(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+
+    void serializeTextCSV(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+    void deserializeTextCSV(IColumn & column, ReadBuffer & istr, const FormatSettings &) const override;
+
+
+    /** Each sub-column in a map is serialized in separate stream.
+      */
+    void enumerateStreams(const StreamCallback & callback, SubstreamPath & path) const override;
+
+    void serializeBinaryBulkStatePrefix(
+           SerializeBinaryBulkSettings & settings,
+           SerializeBinaryBulkStatePtr & state) const override;
+
+    void serializeBinaryBulkStateSuffix(
+           SerializeBinaryBulkSettings & settings,
+           SerializeBinaryBulkStatePtr & state) const override;
+
+    void deserializeBinaryBulkStatePrefix(
+           DeserializeBinaryBulkSettings & settings,
+           DeserializeBinaryBulkStatePtr & state) const override;
+
+    void serializeBinaryBulkWithMultipleStreams(
+           const IColumn & column,
+           size_t offset,
+           size_t limit,
+           SerializeBinaryBulkSettings & settings,
+           SerializeBinaryBulkStatePtr & state) const override;
+
+    void deserializeBinaryBulkWithMultipleStreams(
+           IColumn & column,
+           size_t limit,
+           DeserializeBinaryBulkSettings & settings,
+           DeserializeBinaryBulkStatePtr & state) const override;
+
+    void serializeProtobuf(const IColumn & column, size_t row_num, ProtobufWriter & protobuf, size_t & value_index) const override;
+    void deserializeProtobuf(IColumn & column, ProtobufReader & protobuf, bool allow_add_row, bool & row_added) const override;
+
+    MutableColumnPtr createColumn() const override;
+
+    Field getDefault() const override;
+
+    bool equals(const IDataType & rhs) const override;
+    bool isComparable() const override { return key_type->isComparable() && value_type->isComparable(); }
+    bool isParametric() const override { return true; }
+    bool haveSubtypes() const override { return true; }
+
+    const DataTypePtr & getKeyType() const { return key_type; }
+    const DataTypePtr & getValueType() const { return value_type; }
+    DataTypes getKeyValueTypes() const { return {key_type, value_type}; }
+
+private:
+    template <typename Writer>
+    void serializeTextImpl(const IColumn & column, size_t row_num, WriteBuffer & ostr, Writer && writer) const;
+
+    template <typename Reader>
+    void deserializeTextImpl(IColumn & column, ReadBuffer & istr, bool need_safe_get_int_key, Reader && reader) const;
+};
+
+}
+
diff --git a/src/DataTypes/DataTypesDecimal.h b/src/DataTypes/DataTypesDecimal.h
index 079812c2e74..3f7b4e2ac63 100644
--- a/src/DataTypes/DataTypesDecimal.h
+++ b/src/DataTypes/DataTypesDecimal.h
@@ -96,22 +96,29 @@ inline UInt32 getDecimalScale(const DataTypeDecimal<T> & data_type)
     return data_type.getScale();
 }
 
-template <typename FromDataType, typename ToDataType>
-inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
-convertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to)
+template <typename FromDataType, typename ToDataType, typename ReturnType = void>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, ReturnType>
+convertDecimalsImpl(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to, typename ToDataType::FieldType& result)
 {
     using FromFieldType = typename FromDataType::FieldType;
     using ToFieldType = typename ToDataType::FieldType;
     using MaxFieldType = std::conditional_t<(sizeof(FromFieldType) > sizeof(ToFieldType)), FromFieldType, ToFieldType>;
     using MaxNativeType = typename MaxFieldType::NativeType;
 
+    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
+
     MaxNativeType converted_value;
     if (scale_to > scale_from)
     {
         converted_value = DecimalUtils::scaleMultiplier<MaxNativeType>(scale_to - scale_from);
         if (common::mulOverflow(static_cast<MaxNativeType>(value.value), converted_value, converted_value))
-            throw Exception(std::string(ToDataType::family_name) + " convert overflow",
-                            ErrorCodes::DECIMAL_OVERFLOW);
+        {
+            if constexpr (throw_exception)
+                throw Exception(std::string(ToDataType::family_name) + " convert overflow",
+                                ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return ReturnType(false);
+        }
     }
     else
         converted_value = value.value / DecimalUtils::scaleMultiplier<MaxNativeType>(scale_from - scale_to);
@@ -120,35 +127,87 @@ convertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_fro
     {
         if (converted_value < std::numeric_limits<typename ToFieldType::NativeType>::min() ||
             converted_value > std::numeric_limits<typename ToFieldType::NativeType>::max())
-            throw Exception(std::string(ToDataType::family_name) + " convert overflow",
-                            ErrorCodes::DECIMAL_OVERFLOW);
+        {
+            if constexpr (throw_exception)
+                throw Exception(std::string(ToDataType::family_name) + " convert overflow",
+                                ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return ReturnType(false);
+        }
     }
 
-    return static_cast<typename ToFieldType::NativeType>(converted_value);
+    result = static_cast<typename ToFieldType::NativeType>(converted_value);
+
+    return ReturnType(true);
+}
+
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
+convertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to)
+{
+    using ToFieldType = typename ToDataType::FieldType;
+    ToFieldType result;
+
+    convertDecimalsImpl<FromDataType, ToDataType, void>(value, scale_from, scale_to, result);
+
+    return result;
+}
+
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, bool>
+tryConvertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to, typename ToDataType::FieldType& result)
+{
+    return convertDecimalsImpl<FromDataType, ToDataType, bool>(value, scale_from, scale_to, result);
+}
+
+template <typename FromDataType, typename ToDataType, typename ReturnType>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsNumber<typename ToDataType::FieldType>, ReturnType>
+convertFromDecimalImpl(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
+{
+    using FromFieldType = typename FromDataType::FieldType;
+    using ToFieldType = typename ToDataType::FieldType;
+
+    return DecimalUtils::convertToImpl<ToFieldType, FromFieldType, ReturnType>(value, scale, result);
 }
 
 template <typename FromDataType, typename ToDataType>
 inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsNumber<typename ToDataType::FieldType>, typename ToDataType::FieldType>
 convertFromDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
 {
-    using ToFieldType = typename ToDataType::FieldType;
+    typename ToDataType::FieldType result;
 
-    return DecimalUtils::convertTo<ToFieldType>(value, scale);
+    convertFromDecimalImpl<FromDataType, ToDataType, void>(value, scale, result);
+
+    return result;
 }
 
 template <typename FromDataType, typename ToDataType>
-inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
-convertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsNumber<typename ToDataType::FieldType>, bool>
+tryConvertFromDecimal(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
+{
+    return convertFromDecimalImpl<FromDataType, ToDataType, bool>(value, scale, result);
+}
+
+template <typename FromDataType, typename ToDataType, typename ReturnType>
+inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, ReturnType>
+convertToDecimalImpl(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
 {
     using FromFieldType = typename FromDataType::FieldType;
     using ToFieldType = typename ToDataType::FieldType;
     using ToNativeType = typename ToFieldType::NativeType;
 
+    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
+
     if constexpr (std::is_floating_point_v<FromFieldType>)
     {
         if (!std::isfinite(value))
-            throw Exception(std::string(ToDataType::family_name) + " convert overflow. Cannot convert infinity or NaN to decimal",
-                            ErrorCodes::DECIMAL_OVERFLOW);
+        {
+            if constexpr (throw_exception)
+                throw Exception(std::string(ToDataType::family_name) + " convert overflow. Cannot convert infinity or NaN to decimal",
+                                ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return false;
+        }
 
         auto out = value * static_cast<FromFieldType>(DecimalUtils::scaleMultiplier<ToNativeType>(scale));
         if constexpr (std::is_same_v<ToNativeType, Int128>)
@@ -157,29 +216,60 @@ convertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
             static constexpr Int128 max_int128 = maxInt128();
 
             if (out <= static_cast<ToNativeType>(min_int128) || out >= static_cast<ToNativeType>(max_int128))
-                throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
-                                ErrorCodes::DECIMAL_OVERFLOW);
+            {
+                if constexpr (throw_exception)
+                    throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
+                                    ErrorCodes::DECIMAL_OVERFLOW);
+                else
+                    return ReturnType(false);
+            }
         }
         else
         {
             if (out <= static_cast<FromFieldType>(std::numeric_limits<ToNativeType>::min()) ||
                 out >= static_cast<FromFieldType>(std::numeric_limits<ToNativeType>::max()))
-                throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
-                                ErrorCodes::DECIMAL_OVERFLOW);
+            {
+                if constexpr (throw_exception)
+                    throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
+                                    ErrorCodes::DECIMAL_OVERFLOW);
+                else
+                    return ReturnType(false);
+            }
         }
-        return static_cast<ToNativeType>(out);
+
+        result = static_cast<ToNativeType>(out);
+
+        return ReturnType(true);
     }
     else
     {
         if constexpr (is_big_int_v<FromFieldType>)
-            return convertDecimals<DataTypeDecimal<Decimal256>, ToDataType>(static_cast<Int256>(value), 0, scale);
+            return ReturnType(convertDecimalsImpl<DataTypeDecimal<Decimal256>, ToDataType, ReturnType>(static_cast<Int256>(value), 0, scale, result));
         else if constexpr (std::is_same_v<FromFieldType, UInt64>)
-            return convertDecimals<DataTypeDecimal<Decimal128>, ToDataType>(value, 0, scale);
+            return ReturnType(convertDecimalsImpl<DataTypeDecimal<Decimal128>, ToDataType, ReturnType>(value, 0, scale, result));
         else
-            return convertDecimals<DataTypeDecimal<Decimal64>, ToDataType>(value, 0, scale);
+            return ReturnType(convertDecimalsImpl<DataTypeDecimal<Decimal64>, ToDataType, ReturnType>(value, 0, scale, result));
     }
 }
 
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
+convertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
+{
+    typename ToDataType::FieldType result;
+
+    convertToDecimalImpl<FromDataType, ToDataType, void>(value, scale, result);
+
+    return result;
+}
+
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, bool>
+tryConvertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
+{
+    return convertToDecimalImpl<FromDataType, ToDataType, bool>(value, scale, result);
+}
+
 template <typename T>
 inline DataTypePtr createDecimalMaxPrecision(UInt64 scale)
 {
diff --git a/src/DataTypes/FieldToDataType.cpp b/src/DataTypes/FieldToDataType.cpp
index e01b264df81..95f03385121 100644
--- a/src/DataTypes/FieldToDataType.cpp
+++ b/src/DataTypes/FieldToDataType.cpp
@@ -1,6 +1,7 @@
 #include <Common/FieldVisitors.h>
 #include <DataTypes/FieldToDataType.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypesDecimal.h>
 #include <DataTypes/DataTypeString.h>
@@ -118,6 +119,24 @@ DataTypePtr FieldToDataType::operator() (const Tuple & tuple) const
     return std::make_shared<DataTypeTuple>(element_types);
 }
 
+DataTypePtr FieldToDataType::operator() (const Map & map) const
+{
+    DataTypes key_types;
+    DataTypes value_types;
+    key_types.reserve(map.size());
+    value_types.reserve(map.size());
+
+    for (const auto & elem : map)
+    {
+        const auto & tuple = elem.safeGet<const Tuple &>();
+        assert(tuple.size() == 2);
+        key_types.push_back(applyVisitor(FieldToDataType(), tuple[0]));
+        value_types.push_back(applyVisitor(FieldToDataType(), tuple[1]));
+    }
+
+    return std::make_shared<DataTypeMap>(getLeastSupertype(key_types), getLeastSupertype(value_types));
+}
+
 DataTypePtr FieldToDataType::operator() (const AggregateFunctionStateData & x) const
 {
     const auto & name = static_cast<const AggregateFunctionStateData &>(x).name;
diff --git a/src/DataTypes/FieldToDataType.h b/src/DataTypes/FieldToDataType.h
index 0f8cbfd266c..39457564a2f 100644
--- a/src/DataTypes/FieldToDataType.h
+++ b/src/DataTypes/FieldToDataType.h
@@ -26,6 +26,7 @@ public:
     DataTypePtr operator() (const String & x) const;
     DataTypePtr operator() (const Array & x) const;
     DataTypePtr operator() (const Tuple & tuple) const;
+    DataTypePtr operator() (const Map & map) const;
     DataTypePtr operator() (const DecimalField<Decimal32> & x) const;
     DataTypePtr operator() (const DecimalField<Decimal64> & x) const;
     DataTypePtr operator() (const DecimalField<Decimal128> & x) const;
diff --git a/src/DataTypes/IDataType.h b/src/DataTypes/IDataType.h
index 4601378e8c2..060572ee997 100644
--- a/src/DataTypes/IDataType.h
+++ b/src/DataTypes/IDataType.h
@@ -91,6 +91,8 @@ public:
 
             TupleElement,
 
+            MapElement,
+
             DictionaryKeys,
             DictionaryIndexes,
         };
@@ -449,6 +451,7 @@ public:
     static bool isSpecialCompressionAllowed(const SubstreamPath & path);
 private:
     friend class DataTypeFactory;
+    friend class AggregateFunctionSimpleState;
     /// Customize this DataType
     void setCustomization(DataTypeCustomDescPtr custom_desc_) const;
 
@@ -517,6 +520,7 @@ struct WhichDataType
     constexpr bool isUUID() const { return idx == TypeIndex::UUID; }
     constexpr bool isArray() const { return idx == TypeIndex::Array; }
     constexpr bool isTuple() const { return idx == TypeIndex::Tuple; }
+    constexpr bool isMap() const {return idx == TypeIndex::Map; }
     constexpr bool isSet() const { return idx == TypeIndex::Set; }
     constexpr bool isInterval() const { return idx == TypeIndex::Interval; }
 
diff --git a/src/DataTypes/ya.make b/src/DataTypes/ya.make
index b570adbdc7a..db6793db6ca 100644
--- a/src/DataTypes/ya.make
+++ b/src/DataTypes/ya.make
@@ -27,6 +27,7 @@ SRCS(
     DataTypeInterval.cpp
     DataTypeLowCardinality.cpp
     DataTypeLowCardinalityHelpers.cpp
+    DataTypeMap.cpp
     DataTypeNothing.cpp
     DataTypeNullable.cpp
     DataTypeNumberBase.cpp
diff --git a/src/Databases/DatabaseAtomic.cpp b/src/Databases/DatabaseAtomic.cpp
index af8b751e787..17a91a1fff9 100644
--- a/src/Databases/DatabaseAtomic.cpp
+++ b/src/Databases/DatabaseAtomic.cpp
@@ -35,8 +35,8 @@ public:
 };
 
 
-DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, Context & context_)
-    : DatabaseOrdinary(name_, std::move(metadata_path_), "store/", "DatabaseAtomic (" + name_ + ")", context_)
+DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const String & logger_name, const Context & context_)
+    : DatabaseOrdinary(name_, std::move(metadata_path_), "store/", logger_name, context_)
     , path_to_table_symlinks(global_context.getPath() + "data/" + escapeForFileName(name_) + "/")
     , path_to_metadata_symlink(global_context.getPath() + "metadata/" + escapeForFileName(name_))
     , db_uuid(uuid)
@@ -46,6 +46,11 @@ DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, C
     tryCreateMetadataSymlink();
 }
 
+DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const Context & context_)
+    : DatabaseAtomic(name_, std::move(metadata_path_), uuid, "DatabaseAtomic (" + name_ + ")", context_)
+{
+}
+
 String DatabaseAtomic::getTableDataPath(const String & table_name) const
 {
     std::lock_guard lock(mutex);
@@ -212,6 +217,9 @@ void DatabaseAtomic::renameTable(const Context & context, const String & table_n
     if (is_dictionary && !inside_database)
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Cannot move dictionary to other database");
 
+    if (!exchange)
+        other_db.checkMetadataFilenameAvailabilityUnlocked(to_table_name, inside_database ? db_lock : other_db_lock);
+
     StoragePtr table = getTableUnlocked(table_name, db_lock);
     table->checkTableCanBeRenamed();
     assert_can_move_mat_view(table);
diff --git a/src/Databases/DatabaseAtomic.h b/src/Databases/DatabaseAtomic.h
index 82408ff3ab3..1b1c0cd4353 100644
--- a/src/Databases/DatabaseAtomic.h
+++ b/src/Databases/DatabaseAtomic.h
@@ -20,8 +20,8 @@ namespace DB
 class DatabaseAtomic : public DatabaseOrdinary
 {
 public:
-
-    DatabaseAtomic(String name_, String metadata_path_, UUID uuid, Context & context_);
+    DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const String & logger_name, const Context & context_);
+    DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const Context & context_);
 
     String getEngineName() const override { return "Atomic"; }
     UUID getUUID() const override { return db_uuid; }
@@ -51,14 +51,14 @@ public:
     void loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach) override;
 
     /// Atomic database cannot be detached if there is detached table which still in use
-    void assertCanBeDetached(bool cleanup);
+    void assertCanBeDetached(bool cleanup) override;
 
     UUID tryGetTableUUID(const String & table_name) const override;
 
     void tryCreateSymlink(const String & table_name, const String & actual_data_path, bool if_data_path_exist = false);
     void tryRemoveSymlink(const String & table_name);
 
-    void waitDetachedTableNotInUse(const UUID & uuid);
+    void waitDetachedTableNotInUse(const UUID & uuid) override;
 
 private:
     void commitAlterTable(const StorageID & table_id, const String & table_metadata_tmp_path, const String & table_metadata_path) override;
diff --git a/src/Databases/DatabaseFactory.cpp b/src/Databases/DatabaseFactory.cpp
index 2bb4c595c05..0ac80706bd8 100644
--- a/src/Databases/DatabaseFactory.cpp
+++ b/src/Databases/DatabaseFactory.cpp
@@ -120,27 +120,32 @@ DatabasePtr DatabaseFactory::getImpl(const ASTCreateQuery & create, const String
             const auto & [remote_host_name, remote_port] = parseAddress(host_name_and_port, 3306);
             auto mysql_pool = mysqlxx::Pool(mysql_database_name, remote_host_name, mysql_user_name, mysql_user_password, remote_port);
 
-            if (engine_name == "MaterializeMySQL")
+            if (engine_name == "MySQL")
             {
-                MySQLClient client(remote_host_name, remote_port, mysql_user_name, mysql_user_password);
+                auto mysql_database_settings = std::make_unique<ConnectionMySQLSettings>();
 
-                auto materialize_mode_settings = std::make_unique<MaterializeMySQLSettings>();
+                mysql_database_settings->loadFromQueryContext(context);
+                mysql_database_settings->loadFromQuery(*engine_define); /// higher priority
 
-                if (engine_define->settings)
-                    materialize_mode_settings->loadFromQuery(*engine_define);
-
-                return std::make_shared<DatabaseMaterializeMySQL>(
-                    context, database_name, metadata_path, engine_define, mysql_database_name, std::move(mysql_pool), std::move(client)
-                    , std::move(materialize_mode_settings));
+                return std::make_shared<DatabaseConnectionMySQL>(
+                    context, database_name, metadata_path, engine_define, mysql_database_name, std::move(mysql_database_settings), std::move(mysql_pool));
             }
 
-            auto mysql_database_settings = std::make_unique<ConnectionMySQLSettings>();
+            MySQLClient client(remote_host_name, remote_port, mysql_user_name, mysql_user_password);
 
-            mysql_database_settings->loadFromQueryContext(context);
-            mysql_database_settings->loadFromQuery(*engine_define); /// higher priority
+            auto materialize_mode_settings = std::make_unique<MaterializeMySQLSettings>();
 
-            return std::make_shared<DatabaseConnectionMySQL>(
-                context, database_name, metadata_path, engine_define, mysql_database_name, std::move(mysql_database_settings), std::move(mysql_pool));
+            if (engine_define->settings)
+                materialize_mode_settings->loadFromQuery(*engine_define);
+
+            if (create.uuid == UUIDHelpers::Nil)
+                return std::make_shared<DatabaseMaterializeMySQL<DatabaseOrdinary>>(
+                    context, database_name, metadata_path, uuid, mysql_database_name, std::move(mysql_pool), std::move(client)
+                    , std::move(materialize_mode_settings));
+            else
+                return std::make_shared<DatabaseMaterializeMySQL<DatabaseAtomic>>(
+                    context, database_name, metadata_path, uuid, mysql_database_name, std::move(mysql_pool), std::move(client)
+                    , std::move(materialize_mode_settings));
         }
         catch (...)
         {
diff --git a/src/Databases/DatabaseLazy.cpp b/src/Databases/DatabaseLazy.cpp
index 0119f17f843..f297bf2c82f 100644
--- a/src/Databases/DatabaseLazy.cpp
+++ b/src/Databases/DatabaseLazy.cpp
@@ -42,6 +42,14 @@ void DatabaseLazy::loadStoredObjects(
     iterateMetadataFiles(context, [this](const String & file_name)
     {
         const std::string table_name = file_name.substr(0, file_name.size() - 4);
+
+        auto detached_permanently_flag = Poco::File(getMetadataPath() + "/" + file_name + detached_suffix);
+        if (detached_permanently_flag.exists())
+        {
+            LOG_DEBUG(log, "Skipping permanently detached table {}.", backQuote(table_name));
+            return;
+        }
+
         attachTable(table_name, nullptr, {});
     });
 }
diff --git a/src/Databases/DatabaseOnDisk.cpp b/src/Databases/DatabaseOnDisk.cpp
index 1e6b4019c4b..2a8276ec6db 100644
--- a/src/Databases/DatabaseOnDisk.cpp
+++ b/src/Databases/DatabaseOnDisk.cpp
@@ -164,20 +164,38 @@ void DatabaseOnDisk::createTable(
     /// But there is protection from it - see using DDLGuard in InterpreterCreateQuery.
 
     if (isDictionaryExist(table_name))
-        throw Exception("Dictionary " + backQuote(getDatabaseName()) + "." + backQuote(table_name) + " already exists.",
-            ErrorCodes::DICTIONARY_ALREADY_EXISTS);
+        throw Exception(ErrorCodes::DICTIONARY_ALREADY_EXISTS, "Dictionary {}.{} already exists", backQuote(getDatabaseName()), backQuote(table_name));
 
     if (isTableExist(table_name, global_context))
-        throw Exception("Table " + backQuote(getDatabaseName()) + "." + backQuote(table_name) + " already exists.", ErrorCodes::TABLE_ALREADY_EXISTS);
+        throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists", backQuote(getDatabaseName()), backQuote(table_name));
+
+    String table_metadata_path = getObjectMetadataPath(table_name);
 
     if (create.attach_short_syntax)
     {
         /// Metadata already exists, table was detached
         attachTable(table_name, table, getTableDataPath(create));
+        removeDetachedPermanentlyFlag(table_name, table_metadata_path);
         return;
     }
 
-    String table_metadata_path = getObjectMetadataPath(table_name);
+    if (!create.attach)
+        checkMetadataFilenameAvailability(table_name);
+
+    if (create.attach && Poco::File(table_metadata_path).exists())
+    {
+        ASTPtr ast_detached = parseQueryFromMetadata(log, context, table_metadata_path);
+        auto & create_detached = ast_detached->as<ASTCreateQuery &>();
+
+        // either both should be Nil, either values should be equal
+        if (create.uuid != create_detached.uuid)
+            throw Exception(
+                    ErrorCodes::TABLE_ALREADY_EXISTS,
+                    "Table {}.{} already exist (detached permanently). To attach it back "
+                    "you need to use short ATTACH syntax or a full statement with the same UUID",
+                    backQuote(getDatabaseName()), backQuote(table_name));
+    }
+
     String table_metadata_tmp_path = table_metadata_path + create_suffix;
     String statement;
 
@@ -194,6 +212,26 @@ void DatabaseOnDisk::createTable(
     }
 
     commitCreateTable(create, table, table_metadata_tmp_path, table_metadata_path);
+
+    removeDetachedPermanentlyFlag(table_name, table_metadata_path);
+}
+
+/// If the table was detached permanently we will have a flag file with
+/// .sql.detached extension, is not needed anymore since we attached the table back
+void DatabaseOnDisk::removeDetachedPermanentlyFlag(const String & table_name, const String & table_metadata_path) const
+{
+    try
+    {
+        auto detached_permanently_flag = Poco::File(table_metadata_path + detached_suffix);
+
+        if (detached_permanently_flag.exists())
+            detached_permanently_flag.remove();
+    }
+    catch (Exception & e)
+    {
+        e.addMessage("while trying to remove permanenty detached flag. Table {}.{} may still be marked as permanently detached, and will not be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
+        throw;
+    }
 }
 
 void DatabaseOnDisk::commitCreateTable(const ASTCreateQuery & query, const StoragePtr & table,
@@ -215,6 +253,22 @@ void DatabaseOnDisk::commitCreateTable(const ASTCreateQuery & query, const Stora
     }
 }
 
+void DatabaseOnDisk::detachTablePermanently(const String & table_name)
+{
+    auto table = detachTable(table_name);
+
+    Poco::File detached_permanently_flag(getObjectMetadataPath(table_name) + detached_suffix);
+    try
+    {
+        detached_permanently_flag.createFile();
+    }
+    catch (Exception & e)
+    {
+        e.addMessage("while trying to set permanenty detached flag. Table {}.{} may be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
+        throw;
+    }
+}
+
 void DatabaseOnDisk::dropTable(const Context & context, const String & table_name, bool /*no_delay*/)
 {
     String table_metadata_path = getObjectMetadataPath(table_name);
@@ -253,6 +307,27 @@ void DatabaseOnDisk::dropTable(const Context & context, const String & table_nam
     Poco::File(table_metadata_path_drop).remove();
 }
 
+void DatabaseOnDisk::checkMetadataFilenameAvailability(const String & to_table_name) const
+{
+    std::unique_lock lock(mutex);
+    checkMetadataFilenameAvailabilityUnlocked(to_table_name, lock);
+}
+
+void DatabaseOnDisk::checkMetadataFilenameAvailabilityUnlocked(const String & to_table_name, std::unique_lock<std::mutex> &) const
+{
+    String table_metadata_path = getObjectMetadataPath(to_table_name);
+
+    if (Poco::File(table_metadata_path).exists())
+    {
+        auto detached_permanently_flag = Poco::File(table_metadata_path + detached_suffix);
+
+        if (detached_permanently_flag.exists())
+            throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists (detached permanently)", backQuote(database_name), backQuote(to_table_name));
+        else
+            throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists (detached)", backQuote(database_name), backQuote(to_table_name));
+    }
+}
+
 void DatabaseOnDisk::renameTable(
         const Context & context,
         const String & table_name,
@@ -299,6 +374,9 @@ void DatabaseOnDisk::renameTable(
         if (from_atomic_to_ordinary)
             create.uuid = UUIDHelpers::Nil;
 
+        if (auto * target_db = dynamic_cast<DatabaseOnDisk *>(&to_database))
+            target_db->checkMetadataFilenameAvailability(to_table_name);
+
         /// Notify the table that it is renamed. It will move data to new path (if it stores data on disk) and update StorageID
         table->rename(to_database.getTableDataPath(create), StorageID(create));
     }
@@ -328,6 +406,8 @@ void DatabaseOnDisk::renameTable(
     }
 }
 
+
+/// It returns create table statement (even if table is detached)
 ASTPtr DatabaseOnDisk::getCreateTableQueryImpl(const String & table_name, const Context &, bool throw_on_error) const
 {
     ASTPtr ast;
@@ -400,7 +480,7 @@ void DatabaseOnDisk::iterateMetadataFiles(const Context & context, const Iterati
 {
     auto process_tmp_drop_metadata_file = [&](const String & file_name)
     {
-        assert(getEngineName() != "Atomic");
+        assert(getUUID() == UUIDHelpers::Nil);
         static const char * tmp_drop_ext = ".sql.tmp_drop";
         const std::string object_name = file_name.substr(0, file_name.size() - strlen(tmp_drop_ext));
         if (Poco::File(context.getPath() + getDataPath() + '/' + object_name).exists())
@@ -430,8 +510,11 @@ void DatabaseOnDisk::iterateMetadataFiles(const Context & context, const Iterati
         if (endsWith(dir_it.name(), ".sql.bak"))
             continue;
 
-        static const char * tmp_drop_ext = ".sql.tmp_drop";
-        if (endsWith(dir_it.name(), tmp_drop_ext))
+        /// Permanently detached table flag
+        if (endsWith(dir_it.name(), ".sql.detached"))
+            continue;
+
+        if (endsWith(dir_it.name(), ".sql.tmp_drop"))
         {
             /// There are files that we tried to delete previously
             metadata_files.emplace(dir_it.name(), false);
diff --git a/src/Databases/DatabaseOnDisk.h b/src/Databases/DatabaseOnDisk.h
index 586491d4d29..fff2a259911 100644
--- a/src/Databases/DatabaseOnDisk.h
+++ b/src/Databases/DatabaseOnDisk.h
@@ -39,6 +39,8 @@ public:
         const StoragePtr & table,
         const ASTPtr & query) override;
 
+    void detachTablePermanently(const String & table_name) override;
+
     void dropTable(
         const Context & context,
         const String & table_name,
@@ -67,9 +69,14 @@ public:
 
     static ASTPtr parseQueryFromMetadata(Poco::Logger * log, const Context & context, const String & metadata_file_path, bool throw_on_error = true, bool remove_empty = false);
 
+    /// will throw when the table we want to attach already exists (in active / detached / detached permanently form)
+    void checkMetadataFilenameAvailability(const String & to_table_name) const;
+    void checkMetadataFilenameAvailabilityUnlocked(const String & to_table_name, std::unique_lock<std::mutex> &) const;
+
 protected:
     static constexpr const char * create_suffix = ".tmp";
     static constexpr const char * drop_suffix = ".tmp_drop";
+    static constexpr const char * detached_suffix = ".detached";
 
     using IteratingFunction = std::function<void(const String &)>;
 
@@ -87,6 +94,9 @@ protected:
 
     const String metadata_path;
     const String data_path;
+
+private:
+    void removeDetachedPermanentlyFlag(const String & table_name, const String & table_metadata_path) const;
 };
 
 }
diff --git a/src/Databases/DatabaseOrdinary.cpp b/src/Databases/DatabaseOrdinary.cpp
index 24e2bdcd6b2..58b4ad72a1b 100644
--- a/src/Databases/DatabaseOrdinary.cpp
+++ b/src/Databases/DatabaseOrdinary.cpp
@@ -135,6 +135,19 @@ void DatabaseOrdinary::loadStoredObjects(Context & context, bool has_force_resto
             {
                 auto * create_query = ast->as<ASTCreateQuery>();
                 create_query->database = database_name;
+
+                auto detached_permanently_flag = Poco::File(full_path.string() + detached_suffix);
+                if (detached_permanently_flag.exists())
+                {
+                    /// FIXME: even if we don't load the table we can still mark the uuid of it as taken.
+                    /// if (create_query->uuid != UUIDHelpers::Nil)
+                    ///     DatabaseCatalog::instance().addUUIDMapping(create_query->uuid);
+
+                    const std::string table_name = file_name.substr(0, file_name.size() - 4);
+                    LOG_DEBUG(log, "Skipping permanently detached table {}.", backQuote(table_name));
+                    return;
+                }
+
                 std::lock_guard lock{file_names_mutex};
                 file_names[file_name] = ast;
                 total_dictionaries += create_query->is_dictionary;
diff --git a/src/Databases/DatabasesCommon.cpp b/src/Databases/DatabasesCommon.cpp
index c5df954c2da..2d3d206162b 100644
--- a/src/Databases/DatabasesCommon.cpp
+++ b/src/Databases/DatabasesCommon.cpp
@@ -72,7 +72,7 @@ StoragePtr DatabaseWithOwnTablesBase::detachTableUnlocked(const String & table_n
 
     auto it = tables.find(table_name);
     if (it == tables.end())
-        throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist.",
+        throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist",
                         backQuote(database_name), backQuote(table_name));
     res = it->second;
     tables.erase(it);
@@ -80,7 +80,7 @@ StoragePtr DatabaseWithOwnTablesBase::detachTableUnlocked(const String & table_n
     auto table_id = res->getStorageID();
     if (table_id.hasUUID())
     {
-        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getEngineName() == "Atomic");
+        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getUUID() != UUIDHelpers::Nil);
         DatabaseCatalog::instance().removeUUIDMapping(table_id.uuid);
     }
 
@@ -102,7 +102,7 @@ void DatabaseWithOwnTablesBase::attachTableUnlocked(const String & table_name, c
 
     if (table_id.hasUUID())
     {
-        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getEngineName() == "Atomic");
+        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getUUID() != UUIDHelpers::Nil);
         DatabaseCatalog::instance().addUUIDMapping(table_id.uuid, shared_from_this(), table);
     }
 
@@ -131,7 +131,7 @@ void DatabaseWithOwnTablesBase::shutdown()
         kv.second->shutdown();
         if (table_id.hasUUID())
         {
-            assert(getDatabaseName() == DatabaseCatalog::TEMPORARY_DATABASE || getEngineName() == "Atomic");
+            assert(getDatabaseName() == DatabaseCatalog::TEMPORARY_DATABASE || getUUID() != UUIDHelpers::Nil);
             DatabaseCatalog::instance().removeUUIDMapping(table_id.uuid);
         }
     }
@@ -157,7 +157,7 @@ StoragePtr DatabaseWithOwnTablesBase::getTableUnlocked(const String & table_name
     auto it = tables.find(table_name);
     if (it != tables.end())
         return it->second;
-    throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist.",
+    throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist",
                     backQuote(database_name), backQuote(table_name));
 }
 
diff --git a/src/Databases/IDatabase.h b/src/Databases/IDatabase.h
index fadec5fe7a9..fc821fcab30 100644
--- a/src/Databases/IDatabase.h
+++ b/src/Databases/IDatabase.h
@@ -221,6 +221,8 @@ public:
     }
 
     /// Add a table to the database, but do not add it to the metadata. The database may not support this method.
+    ///
+    /// Note: ATTACH TABLE statement actually uses createTable method.
     virtual void attachTable(const String & /*name*/, const StoragePtr & /*table*/, [[maybe_unused]] const String & relative_table_path = {})
     {
         throw Exception("There is no ATTACH TABLE query for Database" + getEngineName(), ErrorCodes::NOT_IMPLEMENTED);
@@ -245,6 +247,13 @@ public:
         throw Exception("There is no DETACH DICTIONARY query for Database" + getEngineName(), ErrorCodes::NOT_IMPLEMENTED);
     }
 
+    /// Forget about the table without deleting it's data, but rename metadata file to prevent reloading it
+    /// with next restart. The database may not support this method.
+    virtual void detachTablePermanently(const String & /*name*/)
+    {
+        throw Exception("There is no DETACH TABLE PERMANENTLY query for Database" + getEngineName(), ErrorCodes::NOT_IMPLEMENTED);
+    }
+
     /// Rename the table and possibly move the table to another database.
     virtual void renameTable(
         const Context & /*context*/,
@@ -334,6 +343,10 @@ public:
     /// All tables and dictionaries should be detached before detaching the database.
     virtual bool shouldBeEmptyOnDetach() const { return true; }
 
+    virtual void assertCanBeDetached(bool /*cleanup*/) {}
+
+    virtual void waitDetachedTableNotInUse(const UUID & /*uuid*/) { assert(false); }
+
     /// Ask all tables to complete the background threads they are using and delete all table objects.
     virtual void shutdown() = 0;
 
diff --git a/src/Databases/MySQL/DatabaseConnectionMySQL.cpp b/src/Databases/MySQL/DatabaseConnectionMySQL.cpp
index 45483055739..35b016f255b 100644
--- a/src/Databases/MySQL/DatabaseConnectionMySQL.cpp
+++ b/src/Databases/MySQL/DatabaseConnectionMySQL.cpp
@@ -395,7 +395,7 @@ void DatabaseConnectionMySQL::loadStoredObjects(Context &, bool, bool /*force_at
     }
 }
 
-void DatabaseConnectionMySQL::dropTable(const Context &, const String & table_name, bool /*no_delay*/)
+void DatabaseConnectionMySQL::detachTablePermanently(const String & table_name)
 {
     std::lock_guard<std::mutex> lock{mutex};
 
@@ -429,6 +429,11 @@ void DatabaseConnectionMySQL::dropTable(const Context &, const String & table_na
     table_iter->second.second->is_dropped = true;
 }
 
+void DatabaseConnectionMySQL::dropTable(const Context &, const String & table_name, bool /*no_delay*/)
+{
+    detachTablePermanently(table_name);
+}
+
 DatabaseConnectionMySQL::~DatabaseConnectionMySQL()
 {
     try
diff --git a/src/Databases/MySQL/DatabaseConnectionMySQL.h b/src/Databases/MySQL/DatabaseConnectionMySQL.h
index d8694e71db2..3e305fcb20d 100644
--- a/src/Databases/MySQL/DatabaseConnectionMySQL.h
+++ b/src/Databases/MySQL/DatabaseConnectionMySQL.h
@@ -72,6 +72,8 @@ public:
 
     StoragePtr detachTable(const String & table_name) override;
 
+    void detachTablePermanently(const String & table_name) override;
+
     void dropTable(const Context &, const String & table_name, bool no_delay) override;
 
     void attachTable(const String & table_name, const StoragePtr & storage, const String & relative_table_path) override;
diff --git a/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp b/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp
index b5231a23d7e..6a9f1e37f8e 100644
--- a/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp
+++ b/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp
@@ -8,6 +8,7 @@
 
 #    include <Interpreters/Context.h>
 #    include <Databases/DatabaseOrdinary.h>
+#    include <Databases/DatabaseAtomic.h>
 #    include <Databases/MySQL/DatabaseMaterializeTablesIterator.h>
 #    include <Databases/MySQL/MaterializeMySQLSyncThread.h>
 #    include <Parsers/ASTCreateQuery.h>
@@ -22,21 +23,37 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int NOT_IMPLEMENTED;
+    extern const int LOGICAL_ERROR;
 }
 
-DatabaseMaterializeMySQL::DatabaseMaterializeMySQL(
-    const Context & context, const String & database_name_, const String & metadata_path_, const IAST * database_engine_define_
-    , const String & mysql_database_name_, mysqlxx::Pool && pool_, MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_)
-    : IDatabase(database_name_), global_context(context.getGlobalContext()), engine_define(database_engine_define_->clone())
-    , nested_database(std::make_shared<DatabaseOrdinary>(database_name_, metadata_path_, context))
-    , settings(std::move(settings_)), log(&Poco::Logger::get("DatabaseMaterializeMySQL"))
+template<>
+DatabaseMaterializeMySQL<DatabaseOrdinary>::DatabaseMaterializeMySQL(
+    const Context & context, const String & database_name_, const String & metadata_path_, UUID /*uuid*/,
+    const String & mysql_database_name_, mysqlxx::Pool && pool_, MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_)
+    : DatabaseOrdinary(database_name_
+                      , metadata_path_
+                      , "data/" + escapeForFileName(database_name_) + "/"
+                      , "DatabaseMaterializeMySQL<Ordinary> (" + database_name_ + ")", context
+                      )
+    , settings(std::move(settings_))
     , materialize_thread(context, database_name_, mysql_database_name_, std::move(pool_), std::move(client_), settings.get())
 {
 }
 
-void DatabaseMaterializeMySQL::rethrowExceptionIfNeed() const
+template<>
+DatabaseMaterializeMySQL<DatabaseAtomic>::DatabaseMaterializeMySQL(
+    const Context & context, const String & database_name_, const String & metadata_path_, UUID uuid,
+    const String & mysql_database_name_, mysqlxx::Pool && pool_, MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_)
+    : DatabaseAtomic(database_name_, metadata_path_, uuid, "DatabaseMaterializeMySQL<Atomic> (" + database_name_ + ")", context)
+    , settings(std::move(settings_))
+    , materialize_thread(context, database_name_, mysql_database_name_, std::move(pool_), std::move(client_), settings.get())
 {
-    std::unique_lock<std::mutex> lock(mutex);
+}
+
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::rethrowExceptionIfNeed() const
+{
+    std::unique_lock<std::mutex> lock(Base::mutex);
 
     if (!settings->allows_query_when_mysql_lost && exception)
     {
@@ -46,129 +63,71 @@ void DatabaseMaterializeMySQL::rethrowExceptionIfNeed() const
         }
         catch (Exception & ex)
         {
+            /// This method can be called from multiple threads
+            /// and Exception can be modified concurrently by calling addMessage(...),
+            /// so we rethrow a copy.
             throw Exception(ex);
         }
     }
 }
 
-void DatabaseMaterializeMySQL::setException(const std::exception_ptr & exception_)
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::setException(const std::exception_ptr & exception_)
 {
-    std::unique_lock<std::mutex> lock(mutex);
+    std::unique_lock<std::mutex> lock(Base::mutex);
     exception = exception_;
 }
 
-ASTPtr DatabaseMaterializeMySQL::getCreateDatabaseQuery() const
-{
-    const auto & create_query = std::make_shared<ASTCreateQuery>();
-    create_query->database = database_name;
-    create_query->set(create_query->storage, engine_define);
-    return create_query;
-}
-
-void DatabaseMaterializeMySQL::loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach)
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach)
 {
+    Base::loadStoredObjects(context, has_force_restore_data_flag, force_attach);
     try
     {
-        std::unique_lock<std::mutex> lock(mutex);
-        nested_database->loadStoredObjects(context, has_force_restore_data_flag, force_attach);
         materialize_thread.startSynchronization();
+        started_up = true;
     }
     catch (...)
     {
-        tryLogCurrentException(log, "Cannot load MySQL nested database stored objects.");
+        tryLogCurrentException(Base::log, "Cannot load MySQL nested database stored objects.");
 
         if (!force_attach)
             throw;
     }
 }
 
-void DatabaseMaterializeMySQL::shutdown()
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::createTable(const Context & context, const String & name, const StoragePtr & table, const ASTPtr & query)
 {
-    materialize_thread.stopSynchronization();
-
-    auto iterator = nested_database->getTablesIterator(global_context, {});
-
-    /// We only shutdown the table, The tables is cleaned up when destructed database
-    for (; iterator->isValid(); iterator->next())
-        iterator->table()->shutdown();
+    assertCalledFromSyncThreadOrDrop("create table");
+    Base::createTable(context, name, table, query);
 }
 
-bool DatabaseMaterializeMySQL::empty() const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::dropTable(const Context & context, const String & name, bool no_delay)
 {
-    return nested_database->empty();
+    assertCalledFromSyncThreadOrDrop("drop table");
+    Base::dropTable(context, name, no_delay);
 }
 
-String DatabaseMaterializeMySQL::getDataPath() const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::attachTable(const String & name, const StoragePtr & table, const String & relative_table_path)
 {
-    return nested_database->getDataPath();
+    assertCalledFromSyncThreadOrDrop("attach table");
+    Base::attachTable(name, table, relative_table_path);
 }
 
-String DatabaseMaterializeMySQL::getMetadataPath() const
+template<typename Base>
+StoragePtr DatabaseMaterializeMySQL<Base>::detachTable(const String & name)
 {
-    return nested_database->getMetadataPath();
+    assertCalledFromSyncThreadOrDrop("detach table");
+    return Base::detachTable(name);
 }
 
-String DatabaseMaterializeMySQL::getTableDataPath(const String & table_name) const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::renameTable(const Context & context, const String & name, IDatabase & to_database, const String & to_name, bool exchange, bool dictionary)
 {
-    return nested_database->getTableDataPath(table_name);
-}
-
-String DatabaseMaterializeMySQL::getTableDataPath(const ASTCreateQuery & query) const
-{
-    return nested_database->getTableDataPath(query);
-}
-
-String DatabaseMaterializeMySQL::getObjectMetadataPath(const String & table_name) const
-{
-    return nested_database->getObjectMetadataPath(table_name);
-}
-
-UUID DatabaseMaterializeMySQL::tryGetTableUUID(const String & table_name) const
-{
-    return nested_database->tryGetTableUUID(table_name);
-}
-
-time_t DatabaseMaterializeMySQL::getObjectMetadataModificationTime(const String & name) const
-{
-    return nested_database->getObjectMetadataModificationTime(name);
-}
-
-void DatabaseMaterializeMySQL::createTable(const Context & context, const String & name, const StoragePtr & table, const ASTPtr & query)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support create table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->createTable(context, name, table, query);
-}
-
-void DatabaseMaterializeMySQL::dropTable(const Context & context, const String & name, bool no_delay)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support drop table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->dropTable(context, name, no_delay);
-}
-
-void DatabaseMaterializeMySQL::attachTable(const String & name, const StoragePtr & table, const String & relative_table_path)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support attach table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->attachTable(name, table, relative_table_path);
-}
-
-StoragePtr DatabaseMaterializeMySQL::detachTable(const String & name)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support detach table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    return nested_database->detachTable(name);
-}
-
-void DatabaseMaterializeMySQL::renameTable(const Context & context, const String & name, IDatabase & to_database, const String & to_name, bool exchange, bool dictionary)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support rename table.", ErrorCodes::NOT_IMPLEMENTED);
+    assertCalledFromSyncThreadOrDrop("rename table");
 
     if (exchange)
         throw Exception("MaterializeMySQL database not support exchange table.", ErrorCodes::NOT_IMPLEMENTED);
@@ -176,57 +135,37 @@ void DatabaseMaterializeMySQL::renameTable(const Context & context, const String
     if (dictionary)
         throw Exception("MaterializeMySQL database not support rename dictionary.", ErrorCodes::NOT_IMPLEMENTED);
 
-    if (to_database.getDatabaseName() != getDatabaseName())
+    if (to_database.getDatabaseName() != Base::getDatabaseName())
         throw Exception("Cannot rename with other database for MaterializeMySQL database.", ErrorCodes::NOT_IMPLEMENTED);
 
-    nested_database->renameTable(context, name, *nested_database, to_name, exchange, dictionary);
+    Base::renameTable(context, name, *this, to_name, exchange, dictionary);
 }
 
-void DatabaseMaterializeMySQL::alterTable(const Context & context, const StorageID & table_id, const StorageInMemoryMetadata & metadata)
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::alterTable(const Context & context, const StorageID & table_id, const StorageInMemoryMetadata & metadata)
 {
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support alter table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->alterTable(context, table_id, metadata);
+    assertCalledFromSyncThreadOrDrop("alter table");
+    Base::alterTable(context, table_id, metadata);
 }
 
-bool DatabaseMaterializeMySQL::shouldBeEmptyOnDetach() const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::drop(const Context & context)
 {
-    return false;
+    /// Remove metadata info
+    Poco::File metadata(Base::getMetadataPath() + "/.metadata");
+
+    if (metadata.exists())
+        metadata.remove(false);
+
+    Base::drop(context);
 }
 
-void DatabaseMaterializeMySQL::drop(const Context & context)
-{
-    if (nested_database->shouldBeEmptyOnDetach())
-    {
-        for (auto iterator = nested_database->getTablesIterator(context, {}); iterator->isValid(); iterator->next())
-        {
-            TableExclusiveLockHolder table_lock = iterator->table()->lockExclusively(
-                context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
-
-            nested_database->dropTable(context, iterator->name(), true);
-        }
-
-        /// Remove metadata info
-        Poco::File metadata(getMetadataPath() + "/.metadata");
-
-        if (metadata.exists())
-            metadata.remove(false);
-    }
-
-    nested_database->drop(context);
-}
-
-bool DatabaseMaterializeMySQL::isTableExist(const String & name, const Context & context) const
-{
-    return nested_database->isTableExist(name, context);
-}
-
-StoragePtr DatabaseMaterializeMySQL::tryGetTable(const String & name, const Context & context) const
+template<typename Base>
+StoragePtr DatabaseMaterializeMySQL<Base>::tryGetTable(const String & name, const Context & context) const
 {
     if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
     {
-        StoragePtr nested_storage = nested_database->tryGetTable(name, context);
+        StoragePtr nested_storage = Base::tryGetTable(name, context);
 
         if (!nested_storage)
             return {};
@@ -234,20 +173,71 @@ StoragePtr DatabaseMaterializeMySQL::tryGetTable(const String & name, const Cont
         return std::make_shared<StorageMaterializeMySQL>(std::move(nested_storage), this);
     }
 
-    return nested_database->tryGetTable(name, context);
+    return Base::tryGetTable(name, context);
 }
 
-DatabaseTablesIteratorPtr DatabaseMaterializeMySQL::getTablesIterator(const Context & context, const FilterByNameFunction & filter_by_table_name)
+template<typename Base>
+DatabaseTablesIteratorPtr DatabaseMaterializeMySQL<Base>::getTablesIterator(const Context & context, const DatabaseOnDisk::FilterByNameFunction & filter_by_table_name)
 {
     if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
     {
-        DatabaseTablesIteratorPtr iterator = nested_database->getTablesIterator(context, filter_by_table_name);
+        DatabaseTablesIteratorPtr iterator = Base::getTablesIterator(context, filter_by_table_name);
         return std::make_unique<DatabaseMaterializeTablesIterator>(std::move(iterator), this);
     }
 
-    return nested_database->getTablesIterator(context, filter_by_table_name);
+    return Base::getTablesIterator(context, filter_by_table_name);
 }
 
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::assertCalledFromSyncThreadOrDrop(const char * method) const
+{
+    if (!MaterializeMySQLSyncThread::isMySQLSyncThread() && started_up)
+        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "MaterializeMySQL database not support {}", method);
+}
+
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::shutdownSynchronizationThread()
+{
+    materialize_thread.stopSynchronization();
+    started_up = false;
+}
+
+template<typename Database, template<class> class Helper, typename... Args>
+auto castToMaterializeMySQLAndCallHelper(Database * database, Args && ... args)
+{
+    using Ordinary = DatabaseMaterializeMySQL<DatabaseOrdinary>;
+    using Atomic = DatabaseMaterializeMySQL<DatabaseAtomic>;
+    using ToOrdinary = typename std::conditional_t<std::is_const_v<Database>, const Ordinary *, Ordinary *>;
+    using ToAtomic = typename std::conditional_t<std::is_const_v<Database>, const Atomic *, Atomic *>;
+    if (auto * database_materialize = typeid_cast<ToOrdinary>(database))
+        return (database_materialize->*Helper<Ordinary>::v)(std::forward<Args>(args)...);
+    if (auto * database_materialize = typeid_cast<ToAtomic>(database))
+        return (database_materialize->*Helper<Atomic>::v)(std::forward<Args>(args)...);
+
+    throw Exception("LOGICAL_ERROR: cannot cast to DatabaseMaterializeMySQL, it is a bug.", ErrorCodes::LOGICAL_ERROR);
+}
+
+template<typename T> struct HelperSetException { static constexpr auto v = &T::setException; };
+void setSynchronizationThreadException(const DatabasePtr & materialize_mysql_db, const std::exception_ptr & exception)
+{
+    castToMaterializeMySQLAndCallHelper<IDatabase, HelperSetException>(materialize_mysql_db.get(), exception);
+}
+
+template<typename T> struct HelperStopSync { static constexpr auto v = &T::shutdownSynchronizationThread; };
+void stopDatabaseSynchronization(const DatabasePtr & materialize_mysql_db)
+{
+    castToMaterializeMySQLAndCallHelper<IDatabase, HelperStopSync>(materialize_mysql_db.get());
+}
+
+template<typename T> struct HelperRethrow { static constexpr auto v = &T::rethrowExceptionIfNeed; };
+void rethrowSyncExceptionIfNeed(const IDatabase * materialize_mysql_db)
+{
+    castToMaterializeMySQLAndCallHelper<const IDatabase, HelperRethrow>(materialize_mysql_db);
+}
+
+template class DatabaseMaterializeMySQL<DatabaseOrdinary>;
+template class DatabaseMaterializeMySQL<DatabaseAtomic>;
+
 }
 
 #endif
diff --git a/src/Databases/MySQL/DatabaseMaterializeMySQL.h b/src/Databases/MySQL/DatabaseMaterializeMySQL.h
index 799db65b481..e1229269a33 100644
--- a/src/Databases/MySQL/DatabaseMaterializeMySQL.h
+++ b/src/Databases/MySQL/DatabaseMaterializeMySQL.h
@@ -17,48 +17,34 @@ namespace DB
  *
  *  All table structure and data will be written to the local file system
  */
-class DatabaseMaterializeMySQL : public IDatabase
+template<typename Base>
+class DatabaseMaterializeMySQL : public Base
 {
 public:
+
     DatabaseMaterializeMySQL(
-        const Context & context, const String & database_name_, const String & metadata_path_,
-        const IAST * database_engine_define_, const String & mysql_database_name_, mysqlxx::Pool && pool_,
+        const Context & context, const String & database_name_, const String & metadata_path_, UUID uuid,
+        const String & mysql_database_name_, mysqlxx::Pool && pool_,
         MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_);
 
     void rethrowExceptionIfNeed() const;
 
     void setException(const std::exception_ptr & exception);
 protected:
-    const Context & global_context;
 
-    ASTPtr engine_define;
-    DatabasePtr nested_database;
     std::unique_ptr<MaterializeMySQLSettings> settings;
 
-    Poco::Logger * log;
     MaterializeMySQLSyncThread materialize_thread;
 
     std::exception_ptr exception;
 
+    std::atomic_bool started_up{false};
+
 public:
     String getEngineName() const override { return "MaterializeMySQL"; }
 
-    ASTPtr getCreateDatabaseQuery() const override;
-
     void loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach) override;
 
-    void shutdown() override;
-
-    bool empty() const override;
-
-    String getDataPath() const override;
-
-    String getTableDataPath(const String & table_name) const override;
-
-    String getTableDataPath(const ASTCreateQuery & query) const override;
-
-    UUID tryGetTableUUID(const String & table_name) const override;
-
     void createTable(const Context & context, const String & name, const StoragePtr & table, const ASTPtr & query) override;
 
     void dropTable(const Context & context, const String & name, bool no_delay) override;
@@ -71,23 +57,22 @@ public:
 
     void alterTable(const Context & context, const StorageID & table_id, const StorageInMemoryMetadata & metadata) override;
 
-    time_t getObjectMetadataModificationTime(const String & name) const override;
-
-    String getMetadataPath() const override;
-
-    String getObjectMetadataPath(const String & table_name) const override;
-
-    bool shouldBeEmptyOnDetach() const override;
-
     void drop(const Context & context) override;
 
-    bool isTableExist(const String & name, const Context & context) const override;
-
     StoragePtr tryGetTable(const String & name, const Context & context) const override;
 
-    DatabaseTablesIteratorPtr getTablesIterator(const Context & context, const FilterByNameFunction & filter_by_table_name) override;
+    DatabaseTablesIteratorPtr getTablesIterator(const Context & context, const DatabaseOnDisk::FilterByNameFunction & filter_by_table_name) override;
+
+    void assertCalledFromSyncThreadOrDrop(const char * method) const;
+
+    void shutdownSynchronizationThread();
 };
 
+
+void setSynchronizationThreadException(const DatabasePtr & materialize_mysql_db, const std::exception_ptr & exception);
+void stopDatabaseSynchronization(const DatabasePtr & materialize_mysql_db);
+void rethrowSyncExceptionIfNeed(const IDatabase * materialize_mysql_db);
+
 }
 
 #endif
diff --git a/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h b/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h
index 86a5cbf8206..54031de40a2 100644
--- a/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h
+++ b/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h
@@ -2,7 +2,6 @@
 
 #include <Databases/IDatabase.h>
 #include <Storages/StorageMaterializeMySQL.h>
-#include <Databases/MySQL/DatabaseMaterializeMySQL.h>
 
 namespace DB
 {
@@ -30,7 +29,7 @@ public:
 
     UUID uuid() const override { return nested_iterator->uuid(); }
 
-    DatabaseMaterializeTablesIterator(DatabaseTablesIteratorPtr nested_iterator_, DatabaseMaterializeMySQL * database_)
+    DatabaseMaterializeTablesIterator(DatabaseTablesIteratorPtr nested_iterator_, const IDatabase * database_)
         : nested_iterator(std::move(nested_iterator_)), database(database_)
     {
     }
@@ -38,8 +37,7 @@ public:
 private:
     mutable std::vector<StoragePtr> tables;
     DatabaseTablesIteratorPtr nested_iterator;
-    DatabaseMaterializeMySQL * database;
-
+    const IDatabase * database;
 };
 
 }
diff --git a/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp b/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp
index bc7c9e4c554..1da033fa4b3 100644
--- a/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp
+++ b/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp
@@ -71,15 +71,6 @@ static BlockIO tryToExecuteQuery(const String & query_to_execute, Context & quer
     }
 }
 
-static inline DatabaseMaterializeMySQL & getDatabase(const String & database_name)
-{
-    DatabasePtr database = DatabaseCatalog::instance().getDatabase(database_name);
-
-    if (DatabaseMaterializeMySQL * database_materialize = typeid_cast<DatabaseMaterializeMySQL *>(database.get()))
-        return *database_materialize;
-
-    throw Exception("LOGICAL_ERROR: cannot cast to DatabaseMaterializeMySQL, it is a bug.", ErrorCodes::LOGICAL_ERROR);
-}
 
 MaterializeMySQLSyncThread::~MaterializeMySQLSyncThread()
 {
@@ -190,7 +181,8 @@ void MaterializeMySQLSyncThread::synchronization()
     {
         client.disconnect();
         tryLogCurrentException(log);
-        getDatabase(database_name).setException(std::current_exception());
+        auto db = DatabaseCatalog::instance().getDatabase(database_name);
+        setSynchronizationThreadException(db, std::current_exception());
     }
 }
 
@@ -343,7 +335,7 @@ std::optional<MaterializeMetadata> MaterializeMySQLSyncThread::prepareSynchroniz
             opened_transaction = false;
 
             MaterializeMetadata metadata(
-                connection, getDatabase(database_name).getMetadataPath() + "/.metadata", mysql_database_name, opened_transaction);
+                connection, DatabaseCatalog::instance().getDatabase(database_name)->getMetadataPath() + "/.metadata", mysql_database_name, opened_transaction);
 
             if (!metadata.need_dumping_tables.empty())
             {
diff --git a/src/Disks/S3/DiskS3.cpp b/src/Disks/S3/DiskS3.cpp
index aa5088373b3..4786c05f8b0 100644
--- a/src/Disks/S3/DiskS3.cpp
+++ b/src/Disks/S3/DiskS3.cpp
@@ -348,11 +348,11 @@ public:
         DiskS3::Metadata metadata_,
         const String & s3_path_,
         std::optional<DiskS3::ObjectMetadata> object_metadata_,
-        bool is_multipart,
         size_t min_upload_part_size,
+        size_t max_single_part_upload_size,
         size_t buf_size_)
         : WriteBufferFromFileBase(buf_size_, nullptr, 0)
-        , impl(WriteBufferFromS3(client_ptr_, bucket_, metadata_.s3_root_path + s3_path_, min_upload_part_size, is_multipart,std::move(object_metadata_), buf_size_))
+        , impl(WriteBufferFromS3(client_ptr_, bucket_, metadata_.s3_root_path + s3_path_, min_upload_part_size, max_single_part_upload_size,std::move(object_metadata_), buf_size_))
         , metadata(std::move(metadata_))
         , s3_path(s3_path_)
     {
@@ -542,7 +542,7 @@ DiskS3::DiskS3(
     String s3_root_path_,
     String metadata_path_,
     size_t min_upload_part_size_,
-    size_t min_multi_part_upload_size_,
+    size_t max_single_part_upload_size_,
     size_t min_bytes_for_seek_,
     bool send_metadata_)
     : IDisk(std::make_unique<AsyncExecutor>())
@@ -553,7 +553,7 @@ DiskS3::DiskS3(
     , s3_root_path(std::move(s3_root_path_))
     , metadata_path(std::move(metadata_path_))
     , min_upload_part_size(min_upload_part_size_)
-    , min_multi_part_upload_size(min_multi_part_upload_size_)
+    , max_single_part_upload_size(max_single_part_upload_size_)
     , min_bytes_for_seek(min_bytes_for_seek_)
     , send_metadata(send_metadata_)
 {
@@ -665,7 +665,7 @@ std::unique_ptr<ReadBufferFromFileBase> DiskS3::readFile(const String & path, si
     return std::make_unique<SeekAvoidingReadBuffer>(std::move(reader), min_bytes_for_seek);
 }
 
-std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path, size_t buf_size, WriteMode mode, size_t estimated_size, size_t)
+std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path, size_t buf_size, WriteMode mode, size_t, size_t)
 {
     bool exist = exists(path);
     if (exist && readMeta(path).read_only)
@@ -674,7 +674,6 @@ std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path,
     /// Path to store new S3 object.
     auto s3_path = getRandomName();
     auto object_metadata = createObjectMetadata(path);
-    bool is_multipart = estimated_size >= min_multi_part_upload_size;
     if (!exist || mode == WriteMode::Rewrite)
     {
         /// If metadata file exists - remove and create new.
@@ -687,7 +686,8 @@ std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path,
 
         LOG_DEBUG(&Poco::Logger::get("DiskS3"), "Write to file by path: {}. New S3 path: {}", backQuote(metadata_path + path), s3_root_path + s3_path);
 
-        return std::make_unique<WriteIndirectBufferFromS3>(client, bucket, metadata, s3_path, object_metadata, is_multipart, min_upload_part_size, buf_size);
+        return std::make_unique<WriteIndirectBufferFromS3>(
+            client, bucket, metadata, s3_path, object_metadata, min_upload_part_size, max_single_part_upload_size, buf_size);
     }
     else
     {
@@ -696,7 +696,8 @@ std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path,
         LOG_DEBUG(&Poco::Logger::get("DiskS3"), "Append to file by path: {}. New S3 path: {}. Existing S3 objects: {}.",
             backQuote(metadata_path + path), s3_root_path + s3_path, metadata.s3_objects.size());
 
-        return std::make_unique<WriteIndirectBufferFromS3>(client, bucket, metadata, s3_path, object_metadata, is_multipart, min_upload_part_size, buf_size);
+        return std::make_unique<WriteIndirectBufferFromS3>(
+            client, bucket, metadata, s3_path, object_metadata, min_upload_part_size, max_single_part_upload_size, buf_size);
     }
 }
 
diff --git a/src/Disks/S3/DiskS3.h b/src/Disks/S3/DiskS3.h
index 5bc8d51abb2..f62c603adda 100644
--- a/src/Disks/S3/DiskS3.h
+++ b/src/Disks/S3/DiskS3.h
@@ -34,7 +34,7 @@ public:
         String s3_root_path_,
         String metadata_path_,
         size_t min_upload_part_size_,
-        size_t min_multi_part_upload_size_,
+        size_t max_single_part_upload_size_,
         size_t min_bytes_for_seek_,
         bool send_metadata_);
 
@@ -133,7 +133,7 @@ private:
     const String s3_root_path;
     const String metadata_path;
     size_t min_upload_part_size;
-    size_t min_multi_part_upload_size;
+    size_t max_single_part_upload_size;
     size_t min_bytes_for_seek;
     bool send_metadata;
 
diff --git a/src/Disks/S3/registerDiskS3.cpp b/src/Disks/S3/registerDiskS3.cpp
index ff92c295fe4..fd658d95327 100644
--- a/src/Disks/S3/registerDiskS3.cpp
+++ b/src/Disks/S3/registerDiskS3.cpp
@@ -148,7 +148,7 @@ void registerDiskS3(DiskFactory & factory)
             uri.key,
             metadata_path,
             context.getSettingsRef().s3_min_upload_part_size,
-            config.getUInt64(config_prefix + ".min_multi_part_upload_size", 10 * 1024 * 1024),
+            context.getSettingsRef().s3_max_single_part_upload_size,
             config.getUInt64(config_prefix + ".min_bytes_for_seek", 1024 * 1024),
             config.getBool(config_prefix + ".send_object_metadata", false));
 
diff --git a/src/Functions/FunctionBase64Conversion.h b/src/Functions/FunctionBase64Conversion.h
index 4bc2a779cf4..4887e9fe8cc 100644
--- a/src/Functions/FunctionBase64Conversion.h
+++ b/src/Functions/FunctionBase64Conversion.h
@@ -158,7 +158,9 @@ public:
             __msan_unpoison(dst_pos, outlen);
 
             source += srclen + 1;
-            dst_pos += outlen + 1;
+            dst_pos += outlen;
+            *dst_pos = '\0';
+            dst_pos += 1;
 
             dst_offsets[row] = dst_pos - dst;
             src_offset_prev = src_offsets[row];
diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 8c684d3578f..4a54e1ffb2b 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -3,6 +3,8 @@
 // Include this first, because `#define _asan_poison_address` from
 // llvm/Support/Compiler.h conflicts with its forward declaration in
 // sanitizer/asan_interface.h
+#include <memory>
+#include <type_traits>
 #include <Common/Arena.h>
 
 #include <DataTypes/DataTypesNumber.h>
@@ -20,6 +22,7 @@
 #include <Columns/ColumnFixedString.h>
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnAggregateFunction.h>
+#include "Core/DecimalFunctions.h"
 #include "IFunctionImpl.h"
 #include "FunctionHelpers.h"
 #include "IsOperation.h"
@@ -43,6 +46,8 @@
 #    pragma GCC diagnostic pop
 #endif
 
+#include <cassert>
+
 
 namespace DB
 {
@@ -181,8 +186,7 @@ inline constexpr const auto & undec(const T & x)
 }
 
 /// Binary operations for Decimals need scale args
-/// +|- scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
-/// *   no agrs scale. ScaleR = Scale1 + Scale2;
+/// +|-* scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
 /// /   first arg scale. ScaleR = Scale1 (scale_a = DecimalType<B>::getScale()).
 template <template <typename, typename> typename Operation, typename ResultType_, bool check_overflow = true>
 struct DecimalBinaryOperation
@@ -199,19 +203,24 @@ struct DecimalBinaryOperation
     static constexpr bool is_plus_minus_compare = is_plus_minus || is_compare;
     static constexpr bool can_overflow = is_plus_minus || is_multiply;
 
+    static constexpr bool is_plus_minus_compare_mult = is_plus_minus || is_compare || is_multiply;
+
     using ResultType = ResultType_;
     using NativeResultType = typename NativeType<ResultType>::Type;
     using Op = std::conditional_t<is_float_division,
         DivideIntegralImpl<NativeResultType, NativeResultType>, /// substitute divide by intDiv (throw on division by zero)
         Operation<NativeResultType, NativeResultType>>;
 
-    using ArrayC = typename ColumnDecimal<ResultType>::Container;
+    using ArrayC = typename std::conditional_t<IsDecimalNumber<ResultType>,
+        ColumnDecimal<ResultType>,
+        ColumnVector<ResultType>>::Container;
 
     template <bool is_decimal_a, bool is_decimal_b, typename ArrayA, typename ArrayB>
     static void NO_INLINE vectorVector(const ArrayA & a, const ArrayB & b, ArrayC & c,
                                        NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
     {
         size_t size = a.size();
+
         if constexpr (is_plus_minus_compare)
         {
             if (scale_a != 1)
@@ -227,6 +236,28 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -261,6 +292,28 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -295,6 +348,29 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
+
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -343,37 +419,36 @@ private:
     template <bool scale_left>
     static NO_SANITIZE_UNDEFINED NativeResultType applyScaled(NativeResultType a, NativeResultType b, NativeResultType scale)
     {
-        if constexpr (is_plus_minus_compare)
+        static_assert(is_plus_minus_compare);
+        NativeResultType res;
+
+        if constexpr (check_overflow)
         {
-            NativeResultType res;
+            bool overflow = false;
 
-            if constexpr (check_overflow)
-            {
-                bool overflow = false;
-                if constexpr (scale_left)
-                    overflow |= common::mulOverflow(a, scale, a);
-                else
-                    overflow |= common::mulOverflow(b, scale, b);
-
-                if constexpr (can_overflow)
-                    overflow |= Op::template apply<NativeResultType>(a, b, res);
-                else
-                    res = Op::template apply<NativeResultType>(a, b);
-
-                if (overflow)
-                    throw Exception("Decimal math overflow", ErrorCodes::DECIMAL_OVERFLOW);
-            }
+            if constexpr (scale_left)
+                overflow |= common::mulOverflow(a, scale, a);
             else
-            {
-                if constexpr (scale_left)
-                    a *= scale;
-                else
-                    b *= scale;
-                res = Op::template apply<NativeResultType>(a, b);
-            }
+                overflow |= common::mulOverflow(b, scale, b);
 
-            return res;
+            if constexpr (can_overflow)
+                overflow |= Op::template apply<NativeResultType>(a, b, res);
+            else
+                res = Op::template apply<NativeResultType>(a, b);
+
+            if (overflow)
+                throw Exception("Decimal math overflow", ErrorCodes::DECIMAL_OVERFLOW);
         }
+        else
+        {
+            if constexpr (scale_left)
+                a *= scale;
+            else
+                b *= scale;
+            res = Op::template apply<NativeResultType>(a, b);
+        }
+
+        return res;
     }
 
     template <bool is_decimal_a>
@@ -467,26 +542,36 @@ public:
         Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>, RightDataType>,
         Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
+
+        /// e.g Decimal * Float64 = Float64
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>,
+
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
         Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
+
         /// number <op> number -> see corresponding impl
         Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
             DataTypeFromFieldType<typename Op::ResultType>>,
+
         /// Date + Integral -> Date
         /// Integral + Date -> Date
         Case<IsOperation<Operation>::plus, Switch<
             Case<IsIntegral<RightDataType>, LeftDataType>,
             Case<IsIntegral<LeftDataType>, RightDataType>>>,
+
         /// Date - Date     -> Int32
         /// Date - Integral -> Date
         Case<IsOperation<Operation>::minus, Switch<
             Case<std::is_same_v<LeftDataType, RightDataType>, DataTypeInt32>,
             Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, LeftDataType>>>,
+
         /// least(Date, Date) -> Date
         /// greatest(Date, Date) -> Date
         Case<std::is_same_v<LeftDataType, RightDataType> && (IsOperation<Operation>::least || IsOperation<Operation>::greatest),
             LeftDataType>,
+
         /// Date % Int32 -> Int32
         /// Date % Float -> Float64
         Case<IsOperation<Operation>::modulo, Switch<
@@ -494,7 +579,6 @@ public:
             Case<IsDateOrDateTime<LeftDataType> && IsFloatingPoint<RightDataType>, DataTypeFloat64>>>>;
 };
 
-
 template <template <typename, typename> class Op, typename Name, bool valid_on_default_arguments = true>
 class FunctionBinaryArithmetic : public IFunction
 {
@@ -536,7 +620,13 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+        return castType(left, [&](const auto & left_)
+        {
+            return castType(right, [&](const auto & right_)
+            {
+                return f(left_, right_);
+            });
+        });
     }
 
     static FunctionOverloadResolverPtr
@@ -790,10 +880,12 @@ public:
         }
 
         DataTypePtr type_res;
-        bool valid = castBothTypes(arguments[0].get(), arguments[1].get(), [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(arguments[0].get(), arguments[1].get(), [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
@@ -810,6 +902,7 @@ public:
             else
             {
                 using ResultDataType = typename BinaryOperationTraits<Op, LeftDataType, RightDataType>::ResultDataType;
+
                 if constexpr (!std::is_same_v<ResultDataType, InvalidType>)
                 {
                     if constexpr (IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>)
@@ -817,6 +910,10 @@ public:
                         ResultDataType result_type = decimalResultType<is_multiply, is_division>(left, right);
                         type_res = std::make_shared<ResultDataType>(result_type.getPrecision(), result_type.getScale());
                     }
+                    else if constexpr ((IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>) ||
+                        (IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>))
+                        type_res = std::make_shared<std::conditional_t<IsFloatingPoint<LeftDataType>,
+                            LeftDataType, RightDataType>>();
                     else if constexpr (IsDataTypeDecimal<LeftDataType>)
                         type_res = std::make_shared<LeftDataType>(left.getPrecision(), left.getScale());
                     else if constexpr (IsDataTypeDecimal<RightDataType>)
@@ -841,10 +938,15 @@ public:
             }
             return false;
         });
-        if (!valid)
-            throw Exception("Illegal types " + arguments[0]->getName() + " and " + arguments[1]->getName() + " of arguments of function " + String(name),
-                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-        return type_res;
+
+        if (valid)
+            return type_res;
+
+        throw Exception(
+            "Illegal types " + arguments[0]->getName() +
+            " and " + arguments[1]->getName() +
+            " of arguments of function " + String(name),
+            ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
     }
 
     ColumnPtr executeFixedString(const ColumnsWithTypeAndName & arguments) const
@@ -922,6 +1024,18 @@ public:
         return nullptr;
     }
 
+    template <class T, class ResultDataType>
+    static auto helperGetOrConvert(const auto & col_const, const auto & col)
+    {
+        using ResultType = typename ResultDataType::FieldType;
+        using NativeResultType = typename NativeType<ResultType>::Type;
+
+        if constexpr (IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
+            return DecimalUtils::convertTo<NativeResultType>(col_const->template getValue<T>(), col.getScale());
+        else
+            return col_const->template getValue<T>();
+    }
+
     template <typename A, typename B>
     ColumnPtr executeNumeric(const ColumnsWithTypeAndName & arguments, const A & left, const B & right) const
     {
@@ -954,59 +1068,97 @@ public:
                 using NativeResultType = typename NativeType<ResultType>::Type;
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
+                using FieldType = typename ResultDataType::FieldType;
 
-                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
+                static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
+                static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
+                static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
 
-                static constexpr const bool dec_a = IsDecimalNumber<T0>;
-                static constexpr const bool dec_b = IsDecimalNumber<T1>;
+                const ResultDataType type = [&] {
+                    if constexpr (left_is_decimal && IsFloatingPoint<RightDataType>)
+                        return RightDataType();
+                    else if constexpr (right_is_decimal && IsFloatingPoint<LeftDataType>)
+                        return LeftDataType();
+                    else
+                        return decimalResultType<is_multiply, is_division>(left, right);
+                }();
+
+                FieldType left_scale;
+                FieldType right_scale;
 
-                typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
-                typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                    scale_a = right.getScaleMultiplier();
+                    left_scale = right.getScaleMultiplier();
+                else if constexpr (result_is_decimal)
+                    left_scale = type.scaleFactorFor(left, is_multiply);
+                else if constexpr(left_is_decimal) // BUG precision loss
+                    left_scale = left.getScale();
+                else
+                    left_scale = 1; //won't be used, just to silence the warning
+
+                if constexpr (result_is_decimal)
+                    right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+                else if constexpr(right_is_decimal)
+                    right_scale = right.getScale();
+                else
+                    right_scale = 1; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
                     auto res = check_decimal_overflow ?
-                        OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
-                        OpImpl::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b);
+                        OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
+                            const_a, const_b, left_scale, right_scale) :
+                        OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
+                            const_a, const_b, left_scale, right_scale);
 
-                    return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
+                    if constexpr (result_is_decimal)
+                        return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
                             col_left_const->size(), toField(res, type.getScale()));
+                    else
+                         return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
-                col_res = ColVecResult::create(0, type.getScale());
+                if constexpr (result_is_decimal)
+                    col_res = ColVecResult::create(0, type.getScale());
+                else
+                    col_res = ColVecResult::create(0);
+
                 auto & vec_res = col_res->getData();
                 vec_res.resize(col_left_raw->size());
 
                 if (col_left && col_right)
                 {
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorVector<dec_a, dec_b>(col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template vectorVector<dec_a, dec_b>(col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left_const && col_right)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
+                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
+                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left && col_right_const)
                 {
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                        OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                        OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
                 }
                 else
                     return nullptr;
@@ -1073,10 +1225,12 @@ public:
         const auto * left_generic = left_argument.type.get();
         const auto * right_generic = right_argument.type.get();
         ColumnPtr res;
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
diff --git a/src/Functions/FunctionFactory.cpp b/src/Functions/FunctionFactory.cpp
index fbc8e11a9c9..945a78c2734 100644
--- a/src/Functions/FunctionFactory.cpp
+++ b/src/Functions/FunctionFactory.cpp
@@ -8,6 +8,8 @@
 
 #include <IO/WriteHelpers.h>
 
+#include <AggregateFunctions/AggregateFunctionFactory.h>
+
 namespace DB
 {
 
@@ -46,12 +48,15 @@ FunctionOverloadResolverImplPtr FunctionFactory::getImpl(
     auto res = tryGetImpl(name, context);
     if (!res)
     {
+        String extra_info;
+        if (AggregateFunctionFactory::instance().hasNameOrAlias(name))
+            extra_info = ". There is an aggregate function with the same name, but ordinary function is expected here";
+
         auto hints = this->getHints(name);
         if (!hints.empty())
-            throw Exception("Unknown function " + name + ". Maybe you meant: " + toString(hints),
-                            ErrorCodes::UNKNOWN_FUNCTION);
+            throw Exception(ErrorCodes::UNKNOWN_FUNCTION, "Unknown function {}{}. Maybe you meant: {}", name, extra_info, toString(hints));
         else
-            throw Exception("Unknown function " + name, ErrorCodes::UNKNOWN_FUNCTION);
+            throw Exception(ErrorCodes::UNKNOWN_FUNCTION, "Unknown function {}{}", name, extra_info);
     }
     return res;
 }
diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp
index df962800385..257b852ecd8 100644
--- a/src/Functions/FunctionsConversion.cpp
+++ b/src/Functions/FunctionsConversion.cpp
@@ -1,16 +1,10 @@
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionsConversion.h>
-#include <Interpreters/Context.h>
 
 
 namespace DB
 {
 
-FunctionOverloadResolverImplPtr CastOverloadResolver::create(const Context & context)
-{
-    return createImpl(context.getSettingsRef().cast_keep_nullable);
-}
-
 void registerFunctionFixedString(FunctionFactory & factory);
 
 void registerFunctionsConversion(FunctionFactory & factory)
@@ -44,7 +38,10 @@ void registerFunctionsConversion(FunctionFactory & factory)
     registerFunctionFixedString(factory);
 
     factory.registerFunction<FunctionToUnixTimestamp>();
-    factory.registerFunction<CastOverloadResolver>(FunctionFactory::CaseInsensitive);
+
+    factory.registerFunction<CastOverloadResolver<CastType::nonAccurate>>(FunctionFactory::CaseInsensitive);
+    factory.registerFunction<CastOverloadResolver<CastType::accurate>>();
+    factory.registerFunction<CastOverloadResolver<CastType::accurateOrNull>>();
 
     factory.registerFunction<FunctionToUInt8OrZero>();
     factory.registerFunction<FunctionToUInt16OrZero>();
diff --git a/src/Functions/FunctionsConversion.h b/src/Functions/FunctionsConversion.h
index 1e4bd5c26d6..92607f718d5 100644
--- a/src/Functions/FunctionsConversion.h
+++ b/src/Functions/FunctionsConversion.h
@@ -20,6 +20,7 @@
 #include <DataTypes/DataTypeEnum.h>
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeNothing.h>
 #include <DataTypes/DataTypeUUID.h>
@@ -32,17 +33,20 @@
 #include <Columns/ColumnArray.h>
 #include <Columns/ColumnNullable.h>
 #include <Columns/ColumnTuple.h>
+#include <Columns/ColumnMap.h>
 #include <Columns/ColumnsCommon.h>
 #include <Common/FieldVisitors.h>
 #include <Common/assert_cast.h>
 #include <Common/quoteString.h>
+#include <Core/AccurateComparison.h>
 #include <Functions/IFunctionAdaptors.h>
 #include <Functions/FunctionsMiscellaneous.h>
 #include <Functions/FunctionHelpers.h>
 #include <Functions/DateTimeTransforms.h>
+#include <Functions/toFixedString.h>
 #include <DataTypes/DataTypeLowCardinality.h>
 #include <Columns/ColumnLowCardinality.h>
-#include <Functions/toFixedString.h>
+#include <Interpreters/Context.h>
 
 
 namespace DB
@@ -94,6 +98,15 @@ inline UInt32 extractToDecimalScale(const ColumnWithTypeAndName & named_column)
 /// Function toUnixTimestamp has exactly the same implementation as toDateTime of String type.
 struct NameToUnixTimestamp { static constexpr auto name = "toUnixTimestamp"; };
 
+struct AccurateConvertStrategyAdditions
+{
+    UInt32 scale { 0 };
+};
+
+struct AccurateOrNullConvertStrategyAdditions
+{
+    UInt32 scale { 0 };
+};
 
 /** Conversion of number types to each other, enums to numbers, dates and datetimes to numbers and back: done by straight assignment.
   *  (Date is represented internally as number of days from some day; DateTime - as unix timestamp)
@@ -106,7 +119,7 @@ struct ConvertImpl
 
     template <typename Additions = void *>
     static ColumnPtr NO_SANITIZE_UNDEFINED execute(
-        const ColumnsWithTypeAndName & arguments, const DataTypePtr & /*result_type*/, size_t /*input_rows_count*/,
+        const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type [[maybe_unused]], size_t /*input_rows_count*/,
         Additions additions [[maybe_unused]] = Additions())
     {
         const ColumnWithTypeAndName & named_from = arguments[0];
@@ -136,7 +149,17 @@ struct ConvertImpl
             typename ColVecTo::MutablePtr col_to = nullptr;
             if constexpr (IsDataTypeDecimal<ToDataType>)
             {
-                UInt32 scale = additions;
+                UInt32 scale;
+                if constexpr (std::is_same_v<Additions, AccurateConvertStrategyAdditions>
+                    || std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                {
+                    scale = additions.scale;
+                }
+                else
+                {
+                    scale = additions;
+                }
+
                 col_to = ColVecTo::create(0, scale);
             }
             else
@@ -147,36 +170,106 @@ struct ConvertImpl
             size_t size = vec_from.size();
             vec_to.resize(size);
 
-            for (size_t i = 0; i < size; ++i)
+            ColumnUInt8::MutablePtr col_null_map_to;
+            ColumnUInt8::Container * vec_null_map_to [[maybe_unused]] = nullptr;
+            if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
             {
-                if constexpr (IsDataTypeDecimal<FromDataType> || IsDataTypeDecimal<ToDataType>)
-                {
-                    if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>)
-                        vec_to[i] = convertDecimals<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), vec_to.getScale());
-                    else if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeNumber<ToDataType>)
-                        vec_to[i] = convertFromDecimal<FromDataType, ToDataType>(vec_from[i], vec_from.getScale());
-                    else if constexpr (IsDataTypeNumber<FromDataType> && IsDataTypeDecimal<ToDataType>)
-                        vec_to[i] = convertToDecimal<FromDataType, ToDataType>(vec_from[i], vec_to.getScale());
-                    else
-                        throw Exception("Unsupported data type in conversion function", ErrorCodes::CANNOT_CONVERT_TYPE);
-                }
-                else if constexpr (is_big_int_v<FromFieldType> || is_big_int_v<ToFieldType>)
-                {
-                    if constexpr (std::is_same_v<FromFieldType, UInt128> || std::is_same_v<ToFieldType, UInt128>)
-                        throw Exception("Unexpected UInt128 to big int conversion", ErrorCodes::NOT_IMPLEMENTED);
-                    /// If From Data is Nan or Inf, throw exception
-                    else if (!isFinite(vec_from[i]))
-                        throw Exception("Unexpected inf or nan to big int conversion", ErrorCodes::NOT_IMPLEMENTED);
-                    else
-                        vec_to[i] = bigint_cast<ToFieldType>(vec_from[i]);
-                }
-                else if constexpr (std::is_same_v<ToFieldType, UInt128> && sizeof(FromFieldType) <= sizeof(UInt64))
-                    vec_to[i] = static_cast<ToFieldType>(static_cast<UInt64>(vec_from[i]));
-                else
-                    vec_to[i] = static_cast<ToFieldType>(vec_from[i]);
+                col_null_map_to = ColumnUInt8::create(size, false);
+                vec_null_map_to = &col_null_map_to->getData();
             }
 
-            return col_to;
+            for (size_t i = 0; i < size; ++i)
+            {
+                if constexpr ((is_big_int_v<FromFieldType> || is_big_int_v<ToFieldType>) &&
+                    (std::is_same_v<FromFieldType, UInt128> || std::is_same_v<ToFieldType, UInt128>))
+                {
+                    if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                        (*vec_null_map_to)[i] = true;
+                    else
+                        throw Exception("Unexpected UInt128 to big int conversion", ErrorCodes::NOT_IMPLEMENTED);
+                }
+                else
+                {
+                    if constexpr (IsDataTypeDecimal<FromDataType> || IsDataTypeDecimal<ToDataType>)
+                    {
+                        if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                        {
+                            ToFieldType result;
+                            bool convert_result = false;
+
+                            if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                convert_result = tryConvertDecimals<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), vec_to.getScale(), result);
+                            else if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeNumber<ToDataType>)
+                                convert_result = tryConvertFromDecimal<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), result);
+                            else if constexpr (IsDataTypeNumber<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                convert_result = tryConvertToDecimal<FromDataType, ToDataType>(vec_from[i], vec_to.getScale(), result);
+
+                            if (convert_result)
+                                vec_to[i] = result;
+                            else
+                                (*vec_null_map_to)[i] = true;
+                        }
+                        else
+                        {
+                            if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                vec_to[i] = convertDecimals<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), vec_to.getScale());
+                            else if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeNumber<ToDataType>)
+                                vec_to[i] = convertFromDecimal<FromDataType, ToDataType>(vec_from[i], vec_from.getScale());
+                            else if constexpr (IsDataTypeNumber<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                vec_to[i] = convertToDecimal<FromDataType, ToDataType>(vec_from[i], vec_to.getScale());
+                            else
+                                throw Exception("Unsupported data type in conversion function", ErrorCodes::CANNOT_CONVERT_TYPE);
+                        }
+                    }
+                    else
+                    {
+                        /// If From Data is Nan or Inf and we convert to integer type, throw exception
+                        if constexpr (std::is_floating_point_v<FromFieldType> && !std::is_floating_point_v<ToFieldType>)
+                        {
+                            if (!isFinite(vec_from[i]))
+                            {
+                                if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                                {
+                                    (*vec_null_map_to)[i] = true;
+                                    continue;
+                                }
+                                else
+                                    throw Exception("Unexpected inf or nan to integer conversion", ErrorCodes::CANNOT_CONVERT_TYPE);
+                            }
+                        }
+
+                        if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>
+                                || std::is_same_v<Additions, AccurateConvertStrategyAdditions>)
+                        {
+                            bool convert_result = accurate::convertNumeric(vec_from[i], vec_to[i]);
+
+                            if (!convert_result)
+                            {
+                                if (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                                {
+                                    (*vec_null_map_to)[i] = true;
+                                }
+                                else
+                                {
+                                    throw Exception(
+                                        "Value in column " + named_from.column->getName() + " cannot be safely converted into type "
+                                            + result_type->getName(),
+                                        ErrorCodes::CANNOT_CONVERT_TYPE);
+                                }
+                            }
+                        }
+                        else
+                        {
+                            vec_to[i] = static_cast<ToFieldType>(vec_from[i]);
+                        }
+                    }
+                }
+            }
+
+            if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                return ColumnNullable::create(std::move(col_to), std::move(col_null_map_to));
+            else
+                return col_to;
         }
         else
             throw Exception("Illegal column " + named_from.column->getName() + " of first argument of function " + Name::name,
@@ -943,7 +1036,9 @@ struct ConvertImpl<DataTypeString, DataTypeUInt32, NameToUnixTimestamp>
 template <typename T, typename Name>
 struct ConvertImpl<std::enable_if_t<!T::is_parametric, T>, T, Name>
 {
-    static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/)
+    template <typename Additions = void *>
+    static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/,
+        Additions additions [[maybe_unused]] = Additions())
     {
         return arguments[0].column;
     }
@@ -1929,9 +2024,15 @@ private:
     std::optional<Diagnostic> diagnostic;
 };
 
-
 struct NameCast { static constexpr auto name = "CAST"; };
 
+enum class CastType
+{
+    nonAccurate,
+    accurate,
+    accurateOrNull
+};
+
 class FunctionCast final : public IFunctionBaseImpl
 {
 public:
@@ -1940,9 +2041,11 @@ public:
     using Diagnostic = ExecutableFunctionCast::Diagnostic;
 
     FunctionCast(const char * name_, MonotonicityForRange && monotonicity_for_range_
-        , const DataTypes & argument_types_, const DataTypePtr & return_type_, std::optional<Diagnostic> diagnostic_)
-        : name(name_), monotonicity_for_range(monotonicity_for_range_)
+        , const DataTypes & argument_types_, const DataTypePtr & return_type_
+        , std::optional<Diagnostic> diagnostic_, CastType cast_type_)
+        : name(name_), monotonicity_for_range(std::move(monotonicity_for_range_))
         , argument_types(argument_types_), return_type(return_type_), diagnostic(std::move(diagnostic_))
+        , cast_type(cast_type_)
     {
     }
 
@@ -1989,70 +2092,118 @@ private:
     DataTypePtr return_type;
 
     std::optional<Diagnostic> diagnostic;
+    CastType cast_type;
 
-    template <typename DataType>
-    WrapperType createWrapper(const DataTypePtr & from_type, const DataType * const, bool requested_result_is_nullable) const
+    WrapperType createFunctionAdaptor(FunctionPtr function, const DataTypePtr & from_type) const
     {
-        FunctionPtr function;
+        auto function_adaptor = FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
+                                    .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
+
+        return [function_adaptor]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        {
+            return function_adaptor->execute(arguments, result_type, input_rows_count);
+        };
+    }
+
+    static WrapperType createToNullableColumnWrapper()
+    {
+        return [] (ColumnsWithTypeAndName &, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        {
+            ColumnPtr res = result_type->createColumn();
+            ColumnUInt8::Ptr col_null_map_to = ColumnUInt8::create(input_rows_count, true);
+            return ColumnNullable::create(res->cloneResized(input_rows_count), std::move(col_null_map_to));
+        };
+    }
+
+    template <typename ToDataType>
+    WrapperType createWrapper(const DataTypePtr & from_type, const ToDataType * const to_type, bool requested_result_is_nullable) const
+    {
+        TypeIndex from_type_index = from_type->getTypeId();
+        WhichDataType which(from_type_index);
+        bool can_apply_accurate_cast = (cast_type == CastType::accurate || cast_type == CastType::accurateOrNull)
+            && (which.isInt() || which.isUInt() || which.isFloat());
 
         if (requested_result_is_nullable && checkAndGetDataType<DataTypeString>(from_type.get()))
         {
             /// In case when converting to Nullable type, we apply different parsing rule,
             /// that will not throw an exception but return NULL in case of malformed input.
-            function = FunctionConvertFromString<DataType, NameCast, ConvertFromStringExceptionMode::Null>::create();
+            FunctionPtr function = FunctionConvertFromString<ToDataType, NameCast, ConvertFromStringExceptionMode::Null>::create();
+            return createFunctionAdaptor(function, from_type);
         }
-        else
-            function = FunctionTo<DataType>::Type::create();
-
-        auto function_adaptor =
-                FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
-
-        return [function_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        else if (!can_apply_accurate_cast)
         {
-            return function_adaptor->execute(arguments, result_type, input_rows_count);
+            FunctionPtr function = FunctionTo<ToDataType>::Type::create();
+            return createFunctionAdaptor(function, from_type);
+        }
+
+        auto wrapper_cast_type = cast_type;
+
+        return [wrapper_cast_type, from_type_index, to_type]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *column_nullable, size_t input_rows_count)
+        {
+            ColumnPtr result_column;
+            auto res = callOnIndexAndDataType<ToDataType>(from_type_index, [&](const auto & types) -> bool {
+                using Types = std::decay_t<decltype(types)>;
+                using LeftDataType = typename Types::LeftType;
+                using RightDataType = typename Types::RightType;
+
+                if constexpr (IsDataTypeNumber<LeftDataType> && IsDataTypeNumber<RightDataType>)
+                {
+                    if (wrapper_cast_type == CastType::accurate)
+                    {
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, AccurateConvertStrategyAdditions());
+                    }
+                    else
+                    {
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, AccurateOrNullConvertStrategyAdditions());
+                    }
+
+                    return true;
+                }
+
+                return false;
+            });
+
+            /// Additionally check if callOnIndexAndDataType wasn't called at all.
+            if (!res)
+            {
+                if (wrapper_cast_type == CastType::accurateOrNull)
+                {
+                    auto nullable_column_wrapper = FunctionCast::createToNullableColumnWrapper();
+                    return nullable_column_wrapper(arguments, result_type, column_nullable, input_rows_count);
+                }
+                else
+                {
+                    throw Exception{"Conversion from " + std::string(getTypeName(from_type_index)) + " to " + to_type->getName() + " is not supported",
+                        ErrorCodes::CANNOT_CONVERT_TYPE};
+                }
+            }
+
+            return result_column;
         };
     }
 
-    static WrapperType createStringWrapper(const DataTypePtr & from_type)
+    WrapperType createStringWrapper(const DataTypePtr & from_type) const
     {
         FunctionPtr function = FunctionToString::create();
-
-        auto function_adaptor =
-                FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
-
-        return [function_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
-        {
-            return function_adaptor->execute(arguments, result_type, input_rows_count);
-        };
+        return createFunctionAdaptor(function, from_type);
     }
 
-    static WrapperType createFixedStringWrapper(const DataTypePtr & from_type, const size_t N)
+    WrapperType createFixedStringWrapper(const DataTypePtr & from_type, const size_t N) const
     {
         if (!isStringOrFixedString(from_type))
             throw Exception{"CAST AS FixedString is only implemented for types String and FixedString", ErrorCodes::NOT_IMPLEMENTED};
 
-        return [N] (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable *, size_t /*input_rows_count*/)
+        bool exception_mode_null = cast_type == CastType::accurateOrNull;
+        return [exception_mode_null, N] (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable *, size_t /*input_rows_count*/)
         {
-            return FunctionToFixedString::executeForN(arguments, N);
-        };
-    }
-
-    static WrapperType createUUIDWrapper(const DataTypePtr & from_type, const DataTypeUUID * const, bool requested_result_is_nullable)
-    {
-        if (requested_result_is_nullable)
-            throw Exception{"CAST AS Nullable(UUID) is not implemented", ErrorCodes::NOT_IMPLEMENTED};
-
-        FunctionPtr function = FunctionTo<DataTypeUUID>::Type::create();
-
-        auto function_adaptor =
-                FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
-
-        return [function_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
-        {
-            return function_adaptor->execute(arguments, result_type, input_rows_count);
+            if (exception_mode_null)
+                return FunctionToFixedString::executeForN<ConvertToFixedStringExceptionMode::Null>(arguments, N);
+            else
+                return FunctionToFixedString::executeForN<ConvertToFixedStringExceptionMode::Throw>(arguments, N);
         };
     }
 
@@ -2064,41 +2215,73 @@ private:
         UInt32 scale = to_type->getScale();
 
         WhichDataType which(type_index);
-        bool ok = which.isNativeInt() ||
-            which.isNativeUInt() ||
-            which.isDecimal() ||
-            which.isFloat() ||
-            which.isDateOrDateTime() ||
-            which.isStringOrFixedString();
+        bool ok = which.isNativeInt() || which.isNativeUInt() || which.isDecimal() || which.isFloat() || which.isDateOrDateTime()
+            || which.isStringOrFixedString();
         if (!ok)
-            throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
-                ErrorCodes::CANNOT_CONVERT_TYPE};
+        {
+            if (cast_type == CastType::accurateOrNull)
+                return createToNullableColumnWrapper();
+            else
+                throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
+                    ErrorCodes::CANNOT_CONVERT_TYPE};
+        }
 
-        return [type_index, scale, to_type] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        auto wrapper_cast_type = cast_type;
+
+        return [wrapper_cast_type, type_index, scale, to_type]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *column_nullable, size_t input_rows_count)
         {
             ColumnPtr result_column;
-            auto res = callOnIndexAndDataType<ToDataType>(type_index, [&](const auto & types) -> bool
-            {
+            auto res = callOnIndexAndDataType<ToDataType>(type_index, [&](const auto & types) -> bool {
                 using Types = std::decay_t<decltype(types)>;
                 using LeftDataType = typename Types::LeftType;
                 using RightDataType = typename Types::RightType;
 
+                if constexpr (IsDataTypeDecimalOrNumber<LeftDataType> && IsDataTypeDecimalOrNumber<RightDataType>)
+                {
+                    if (wrapper_cast_type == CastType::accurate)
+                    {
+                        AccurateConvertStrategyAdditions additions;
+                        additions.scale = scale;
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, additions);
+
+                        return true;
+                    }
+                    else if (wrapper_cast_type == CastType::accurateOrNull)
+                    {
+                        AccurateOrNullConvertStrategyAdditions additions;
+                        additions.scale = scale;
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, additions);
+
+                        return true;
+                    }
+                }
+
                 result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(arguments, result_type, input_rows_count, scale);
+
                 return true;
             });
 
             /// Additionally check if callOnIndexAndDataType wasn't called at all.
             if (!res)
             {
-                throw Exception{"Conversion from " + std::string(getTypeName(type_index)) + " to " + to_type->getName() +
-                                " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+                if (wrapper_cast_type == CastType::accurateOrNull)
+                {
+                    auto nullable_column_wrapper = FunctionCast::createToNullableColumnWrapper();
+                    return nullable_column_wrapper(arguments, result_type, column_nullable, input_rows_count);
+                }
+                else
+                    throw Exception{"Conversion from " + std::string(getTypeName(type_index)) + " to " + to_type->getName() + " is not supported",
+                        ErrorCodes::CANNOT_CONVERT_TYPE};
             }
 
             return result_column;
         };
     }
 
-    static WrapperType createAggregateFunctionWrapper(const DataTypePtr & from_type_untyped, const DataTypeAggregateFunction * to_type)
+    WrapperType createAggregateFunctionWrapper(const DataTypePtr & from_type_untyped, const DataTypeAggregateFunction * to_type) const
     {
         /// Conversion from String through parsing.
         if (checkAndGetDataType<DataTypeString>(from_type_untyped.get()))
@@ -2109,8 +2292,13 @@ private:
             };
         }
         else
-            throw Exception{"Conversion from " + from_type_untyped->getName() + " to " + to_type->getName() +
-                " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+        {
+            if (cast_type == CastType::accurateOrNull)
+                return createToNullableColumnWrapper();
+            else
+                throw Exception{"Conversion from " + from_type_untyped->getName() + " to " + to_type->getName() +
+                    " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+        }
     }
 
     WrapperType createArrayWrapper(const DataTypePtr & from_type_untyped, const DataTypeArray * to_type) const
@@ -2167,6 +2355,20 @@ private:
         };
     }
 
+    using ElementWrappers = std::vector<WrapperType>;
+
+    ElementWrappers getElementWrappers(const DataTypes & from_element_types, const DataTypes & to_element_types) const
+    {
+        ElementWrappers element_wrappers;
+        element_wrappers.reserve(from_element_types.size());
+
+        /// Create conversion wrapper for each element in tuple
+        for (const auto idx_type : ext::enumerate(from_element_types))
+            element_wrappers.push_back(prepareUnpackDictionaries(idx_type.second, to_element_types[idx_type.first]));
+
+        return element_wrappers;
+    }
+
     WrapperType createTupleWrapper(const DataTypePtr & from_type_untyped, const DataTypeTuple * to_type) const
     {
         /// Conversion from String through parsing.
@@ -2189,12 +2391,7 @@ private:
 
         const auto & from_element_types = from_type->getElements();
         const auto & to_element_types = to_type->getElements();
-        std::vector<WrapperType> element_wrappers;
-        element_wrappers.reserve(from_element_types.size());
-
-        /// Create conversion wrapper for each element in tuple
-        for (const auto idx_type : ext::enumerate(from_type->getElements()))
-            element_wrappers.push_back(prepareUnpackDictionaries(idx_type.second, to_element_types[idx_type.first]));
+        auto element_wrappers = getElementWrappers(from_element_types, to_element_types);
 
         return [element_wrappers, from_element_types, to_element_types]
             (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
@@ -2217,6 +2414,121 @@ private:
         };
     }
 
+    /// The case of: tuple([key1, key2, ..., key_n], [value1, value2, ..., value_n])
+    WrapperType createTupleToMapWrapper(const DataTypes & from_kv_types, const DataTypes & to_kv_types) const
+    {
+        return [element_wrappers = getElementWrappers(from_kv_types, to_kv_types), from_kv_types, to_kv_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
+        {
+            const auto * col = arguments.front().column.get();
+            const auto & column_tuple = assert_cast<const ColumnTuple &>(*col);
+
+            if (column_tuple.getColumn(0).size() != column_tuple.getColumn(1).size())
+                throw Exception(ErrorCodes::TYPE_MISMATCH,
+                    "CAST AS Map can only be performed from tuple of arrays with equal sizes."
+                    " Size of keys: {}. Size of values: {}", column_tuple.getColumn(0).size(), column_tuple.getColumn(1).size());
+
+            ColumnPtr offsets;
+            Columns converted_columns(2);
+            for (size_t i = 0; i < 2; ++i)
+            {
+                const auto & column_array = assert_cast<const ColumnArray &>(column_tuple.getColumn(i));
+                ColumnsWithTypeAndName element = {{column_array.getDataPtr(), from_kv_types[i], ""}};
+                converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
+
+                if (!offsets)
+                    offsets = column_array.getOffsetsPtr();
+            }
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], offsets);
+        };
+    }
+
+    WrapperType createMapToMapWrrapper(const DataTypes & from_kv_types, const DataTypes & to_kv_types) const
+    {
+        return [element_wrappers = getElementWrappers(from_kv_types, to_kv_types), from_kv_types, to_kv_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
+        {
+            const auto * col = arguments.front().column.get();
+            const auto & column_map = typeid_cast<const ColumnMap &>(*col);
+            const auto & nested_data = column_map.getNestedData();
+
+            Columns converted_columns(2);
+            for (size_t i = 0; i < 2; ++i)
+            {
+                ColumnsWithTypeAndName element = {{nested_data.getColumnPtr(i), from_kv_types[i], ""}};
+                converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
+            }
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], column_map.getNestedColumn().getOffsetsPtr());
+        };
+    }
+
+    /// The case of: [(key1, value1), (key2, value2), ...]
+    WrapperType createArrayToMapWrrapper(const DataTypes & from_kv_types, const DataTypes & to_kv_types) const
+    {
+        return [element_wrappers = getElementWrappers(from_kv_types, to_kv_types), from_kv_types, to_kv_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
+        {
+            const auto * col = arguments.front().column.get();
+            const auto & column_array = typeid_cast<const ColumnArray &>(*col);
+            const auto & nested_data = typeid_cast<const ColumnTuple &>(column_array.getData());
+
+            Columns converted_columns(2);
+            for (size_t i = 0; i < 2; ++i)
+            {
+                ColumnsWithTypeAndName element = {{nested_data.getColumnPtr(i), from_kv_types[i], ""}};
+                converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
+            }
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], column_array.getOffsetsPtr());
+        };
+    }
+
+
+    WrapperType createMapWrapper(const DataTypePtr & from_type_untyped, const DataTypeMap * to_type) const
+    {
+        if (const auto * from_tuple = checkAndGetDataType<DataTypeTuple>(from_type_untyped.get()))
+        {
+            if (from_tuple->getElements().size() != 2)
+                throw Exception{"CAST AS Map from tuple requeires 2 elements.\n"
+                    "Left type: " + from_tuple->getName() + ", right type: " + to_type->getName(), ErrorCodes::TYPE_MISMATCH};
+
+            DataTypes from_kv_types;
+            const auto & to_kv_types = to_type->getKeyValueTypes();
+
+            for (const auto & elem : from_tuple->getElements())
+            {
+                const auto * type_array = checkAndGetDataType<DataTypeArray>(elem.get());
+                if (!type_array)
+                    throw Exception(ErrorCodes::TYPE_MISMATCH,
+                        "CAST AS Map can only be performed from tuples of array. Got: {}", from_tuple->getName());
+
+                from_kv_types.push_back(type_array->getNestedType());
+            }
+
+            return createTupleToMapWrapper(from_kv_types, to_kv_types);
+        }
+        else if (const auto * from_array = typeid_cast<const DataTypeArray *>(from_type_untyped.get()))
+        {
+            const auto * nested_tuple = typeid_cast<const DataTypeTuple *>(from_array->getNestedType().get());
+            if (!nested_tuple || nested_tuple->getElements().size() != 2)
+                throw Exception{"CAST AS Map from array requeires nested tuple of 2 elements.\n"
+                    "Left type: " + from_array->getName() + ", right type: " + to_type->getName(), ErrorCodes::TYPE_MISMATCH};
+
+            return createArrayToMapWrrapper(nested_tuple->getElements(), to_type->getKeyValueTypes());
+        }
+        else if (const auto * from_type = checkAndGetDataType<DataTypeMap>(from_type_untyped.get()))
+        {
+            return createMapToMapWrrapper(from_type->getKeyValueTypes(), to_type->getKeyValueTypes());
+        }
+        else
+        {
+            throw Exception{"Unsupported types to CAST AS Map\n"
+                "Left type: " + from_type_untyped->getName() + ", right type: " + to_type->getName(), ErrorCodes::TYPE_MISMATCH};
+        }
+    }
+
     template <typename FieldType>
     WrapperType createEnumWrapper(const DataTypePtr & from_type, const DataTypeEnum<FieldType> * to_type) const
     {
@@ -2235,17 +2547,16 @@ private:
         else if (isNativeNumber(from_type) || isEnum(from_type))
         {
             auto function = Function::create();
-            auto func_or_adaptor = FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                    .build(ColumnsWithTypeAndName{{nullptr, from_type, "" }});
-
-            return [func_or_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
-            {
-                return func_or_adaptor->execute(arguments, result_type, input_rows_count);
-            };
+            return createFunctionAdaptor(function, from_type);
         }
         else
-            throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() +
-                " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+        {
+            if (cast_type == CastType::accurateOrNull)
+                return createToNullableColumnWrapper();
+            else
+                throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
+                    ErrorCodes::CANNOT_CONVERT_TYPE};
+        }
     }
 
     template <typename EnumTypeFrom, typename EnumTypeTo>
@@ -2346,7 +2657,16 @@ private:
         if (from_type->onlyNull())
         {
             if (!to_nested->isNullable())
-                throw Exception{"Cannot convert NULL to a non-nullable type", ErrorCodes::CANNOT_CONVERT_TYPE};
+            {
+                if (cast_type == CastType::accurateOrNull)
+                {
+                    return createToNullableColumnWrapper();
+                }
+                else
+                {
+                    throw Exception{"Cannot convert NULL to a non-nullable type", ErrorCodes::CANNOT_CONVERT_TYPE};
+                }
+            }
 
             return [](ColumnsWithTypeAndName &, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
             {
@@ -2536,7 +2856,8 @@ private:
                 std::is_same_v<ToDataType, DataTypeFloat32> ||
                 std::is_same_v<ToDataType, DataTypeFloat64> ||
                 std::is_same_v<ToDataType, DataTypeDate> ||
-                std::is_same_v<ToDataType, DataTypeDateTime>)
+                std::is_same_v<ToDataType, DataTypeDateTime> ||
+                std::is_same_v<ToDataType, DataTypeUUID>)
             {
                 ret = createWrapper(from_type, checkAndGetDataType<ToDataType>(to_type.get()), requested_result_is_nullable);
                 return true;
@@ -2558,14 +2879,6 @@ private:
                 ret = createDecimalWrapper(from_type, checkAndGetDataType<ToDataType>(to_type.get()));
                 return true;
             }
-            if constexpr (std::is_same_v<ToDataType, DataTypeUUID>)
-            {
-                if (isStringOrFixedString(from_type))
-                {
-                    ret = createUUIDWrapper(from_type, checkAndGetDataType<ToDataType>(to_type.get()), requested_result_is_nullable);
-                    return true;
-                }
-            }
 
             return false;
         };
@@ -2584,6 +2897,8 @@ private:
                 return createArrayWrapper(from_type, checkAndGetDataType<DataTypeArray>(to_type.get()));
             case TypeIndex::Tuple:
                 return createTupleWrapper(from_type, checkAndGetDataType<DataTypeTuple>(to_type.get()));
+            case TypeIndex::Map:
+                return createMapWrapper(from_type, checkAndGetDataType<DataTypeMap>(to_type.get()));
 
             case TypeIndex::AggregateFunction:
                 return createAggregateFunctionWrapper(from_type, checkAndGetDataType<DataTypeAggregateFunction>(to_type.get()));
@@ -2591,20 +2906,91 @@ private:
                 break;
         }
 
-        throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
-            ErrorCodes::CANNOT_CONVERT_TYPE};
+        if (cast_type == CastType::accurateOrNull)
+            return createToNullableColumnWrapper();
+        else
+            throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
+                ErrorCodes::CANNOT_CONVERT_TYPE};
     }
 };
 
+class MonotonicityHelper
+{
+public:
+    using MonotonicityForRange = FunctionCast::MonotonicityForRange;
+
+    template <typename DataType>
+    static auto monotonicityForType(const DataType * const)
+    {
+        return FunctionTo<DataType>::Type::Monotonic::get;
+    }
+
+    static MonotonicityForRange getMonotonicityInformation(const DataTypePtr & from_type, const IDataType * to_type)
+    {
+        if (const auto type = checkAndGetDataType<DataTypeUInt8>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt16>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt32>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt64>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt256>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt8>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt16>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt32>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt64>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt128>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt256>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeFloat32>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeFloat64>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeDate>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeDateTime>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeString>(to_type))
+            return monotonicityForType(type);
+        if (isEnum(from_type))
+        {
+            if (const auto type = checkAndGetDataType<DataTypeEnum8>(to_type))
+                return monotonicityForType(type);
+            if (const auto type = checkAndGetDataType<DataTypeEnum16>(to_type))
+                return monotonicityForType(type);
+        }
+        /// other types like Null, FixedString, Array and Tuple have no monotonicity defined
+        return {};
+    }
+};
+
+template<CastType cast_type>
 class CastOverloadResolver : public IFunctionOverloadResolverImpl
 {
 public:
     using MonotonicityForRange = FunctionCast::MonotonicityForRange;
     using Diagnostic = FunctionCast::Diagnostic;
 
-    static constexpr auto name = "CAST";
+    static constexpr auto accurate_cast_name = "accurateCast";
+    static constexpr auto accurate_cast_or_null_name = "accurateCastOrNull";
+    static constexpr auto cast_name = "CAST";
+
+    static constexpr auto name = cast_type == CastType::accurate
+        ? accurate_cast_name
+        : (cast_type == CastType::accurateOrNull ? accurate_cast_or_null_name : cast_name);
+
+    static FunctionOverloadResolverImplPtr create(const Context & context)
+    {
+        return createImpl(context.getSettingsRef().cast_keep_nullable);
+    }
 
-    static FunctionOverloadResolverImplPtr create(const Context & context);
     static FunctionOverloadResolverImplPtr createImpl(bool keep_nullable, std::optional<Diagnostic> diagnostic = {})
     {
         return std::make_unique<CastOverloadResolver>(keep_nullable, std::move(diagnostic));
@@ -2630,8 +3016,8 @@ protected:
         for (size_t i = 0; i < arguments.size(); ++i)
             data_types[i] = arguments[i].type;
 
-        auto monotonicity = getMonotonicityInformation(arguments.front().type, return_type.get());
-        return std::make_unique<FunctionCast>(name, std::move(monotonicity), data_types, return_type, diagnostic);
+        auto monotonicity = MonotonicityHelper::getMonotonicityInformation(arguments.front().type, return_type.get());
+        return std::make_unique<FunctionCast>(name, std::move(monotonicity), data_types, return_type, diagnostic, cast_type);
     }
 
     DataTypePtr getReturnType(const ColumnsWithTypeAndName & arguments) const override
@@ -2649,9 +3035,17 @@ protected:
                 ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
         DataTypePtr type = DataTypeFactory::instance().get(type_col->getValue<String>());
-        if (keep_nullable && arguments.front().type->isNullable())
+
+        if constexpr (cast_type == CastType::accurateOrNull)
+        {
             return makeNullable(type);
-        return type;
+        }
+        else
+        {
+            if (keep_nullable && arguments.front().type->isNullable())
+                return makeNullable(type);
+            return type;
+        }
     }
 
     bool useDefaultImplementationForNulls() const override { return false; }
@@ -2660,57 +3054,6 @@ protected:
 private:
     bool keep_nullable;
     std::optional<Diagnostic> diagnostic;
-
-    template <typename DataType>
-    static auto monotonicityForType(const DataType * const)
-    {
-        return FunctionTo<DataType>::Type::Monotonic::get;
-    }
-
-    static MonotonicityForRange getMonotonicityInformation(const DataTypePtr & from_type, const IDataType * to_type)
-    {
-        if (const auto * type = checkAndGetDataType<DataTypeUInt8>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt16>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt32>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt64>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt256>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt8>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt16>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt32>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt64>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt128>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt256>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeFloat32>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeFloat64>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeDate>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeDateTime>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeString>(to_type))
-            return monotonicityForType(type);
-        if (isEnum(from_type))
-        {
-            if (const auto * type = checkAndGetDataType<DataTypeEnum8>(to_type))
-                return monotonicityForType(type);
-            if (const auto * type = checkAndGetDataType<DataTypeEnum16>(to_type))
-                return monotonicityForType(type);
-        }
-        /// other types like Null, FixedString, Array and Tuple have no monotonicity defined
-        return {};
-    }
 };
 
 }
diff --git a/src/Functions/array/arrayAggregation.cpp b/src/Functions/array/arrayAggregation.cpp
new file mode 100644
index 00000000000..992a331d05b
--- /dev/null
+++ b/src/Functions/array/arrayAggregation.cpp
@@ -0,0 +1,318 @@
+#include <DataTypes/DataTypesNumber.h>
+#include <DataTypes/DataTypesDecimal.h>
+#include <DataTypes/DataTypeDateTime64.h>
+#include <Columns/ColumnsNumber.h>
+#include <Columns/ColumnDecimal.h>
+#include "FunctionArrayMapped.h"
+#include <Functions/FunctionFactory.h>
+
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+    extern const int ILLEGAL_COLUMN;
+}
+
+enum class AggregateOperation
+{
+    min,
+    max,
+    sum,
+    average
+};
+
+/**
+ * During array aggregation we derive result type from operation.
+ * For array min or array max we use array element as result type.
+ * For array average we use Float64.
+ * For array sum for for big integers, we use same type representation, decimal numbers we use Decimal128,
+ * for floating point numbers Float64, for numeric unsigned Int64, and for numeric signed UInt64.
+ */
+
+template <typename ArrayElement, AggregateOperation operation>
+struct ArrayAggregateResultImpl;
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::min>
+{
+    using Result = ArrayElement;
+};
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::max>
+{
+    using Result = ArrayElement;
+};
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::average>
+{
+    using Result = Float64;
+};
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::sum>
+{
+    using Result =
+        std::conditional_t<std::is_same_v<ArrayElement, Int128>, Int128,
+            std::conditional_t<std::is_same_v<ArrayElement, Int256>, Int256,
+                std::conditional_t<std::is_same_v<ArrayElement, UInt256>, UInt256,
+                    std::conditional_t<IsDecimalNumber<ArrayElement>, Decimal128,
+                        std::conditional_t<std::is_floating_point_v<ArrayElement>, Float64,
+                            std::conditional_t<std::is_signed_v<ArrayElement>, Int64,
+                                UInt64>>>>>>;
+};
+
+template <typename ArrayElement, AggregateOperation operation>
+using ArrayAggregateResult = typename ArrayAggregateResultImpl<ArrayElement, operation>::Result;
+
+template<AggregateOperation aggregate_operation>
+struct ArrayAggregateImpl
+{
+    static bool needBoolean() { return false; }
+    static bool needExpression() { return false; }
+    static bool needOneArray() { return false; }
+
+    static DataTypePtr getReturnType(const DataTypePtr & expression_return, const DataTypePtr & /*array_element*/)
+    {
+        DataTypePtr result;
+
+        auto call = [&](const auto & types)
+        {
+            using Types = std::decay_t<decltype(types)>;
+            using DataType = typename Types::LeftType;
+
+            if constexpr (aggregate_operation == AggregateOperation::average)
+            {
+                result = std::make_shared<DataTypeFloat64>();
+
+                return true;
+            }
+            else if constexpr (IsDataTypeNumber<DataType>)
+            {
+                using NumberReturnType = ArrayAggregateResult<typename DataType::FieldType, aggregate_operation>;
+                result = std::make_shared<DataTypeNumber<NumberReturnType>>();
+
+                return true;
+            }
+            else if constexpr (IsDataTypeDecimal<DataType> && !IsDataTypeDateOrDateTime<DataType>)
+            {
+                using DecimalReturnType = ArrayAggregateResult<typename DataType::FieldType, aggregate_operation>;
+                UInt32 scale = getDecimalScale(*expression_return);
+                result = std::make_shared<DataTypeDecimal<DecimalReturnType>>(DecimalUtils::maxPrecision<DecimalReturnType>(), scale);
+
+                return true;
+            }
+
+            return false;
+        };
+
+        if (!callOnIndexAndDataType<void>(expression_return->getTypeId(), call))
+        {
+            throw Exception(
+                "array aggregation function cannot be performed on type " + expression_return->getName(),
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        }
+
+        return result;
+    }
+
+    template <typename Element>
+    static bool executeType(const ColumnPtr & mapped, const ColumnArray::Offsets & offsets, ColumnPtr & res_ptr)
+    {
+        using Result = ArrayAggregateResult<Element, aggregate_operation>;
+        using ColVecType = std::conditional_t<IsDecimalNumber<Element>, ColumnDecimal<Element>, ColumnVector<Element>>;
+        using ColVecResult = std::conditional_t<IsDecimalNumber<Result>, ColumnDecimal<Result>, ColumnVector<Result>>;
+
+        /// For average of array we return Float64 as result, but we want to keep precision
+        /// so we convert to Float64 as last step, but intermediate sum is represented as result of sum operation
+        static constexpr bool is_average_operation = aggregate_operation == AggregateOperation::average;
+        using SummAggregationType = ArrayAggregateResult<Element, AggregateOperation::sum>;
+
+        using AggregationType = std::conditional_t<is_average_operation, SummAggregationType, Result>;
+
+
+        const ColVecType * column = checkAndGetColumn<ColVecType>(&*mapped);
+
+        /// Constant case.
+        if (!column)
+        {
+            const ColumnConst * column_const = checkAndGetColumnConst<ColVecType>(&*mapped);
+
+            if (!column_const)
+                return false;
+
+            const AggregationType x = column_const->template getValue<Element>(); // NOLINT
+            const typename ColVecType::Container & data
+                = checkAndGetColumn<ColVecType>(&column_const->getDataColumn())->getData();
+
+            typename ColVecResult::MutablePtr res_column;
+            if constexpr (IsDecimalNumber<Element>)
+            {
+                res_column = ColVecResult::create(offsets.size(), data.getScale());
+            }
+            else
+                res_column = ColVecResult::create(offsets.size());
+
+            typename ColVecResult::Container & res = res_column->getData();
+
+            size_t pos = 0;
+            for (size_t i = 0; i < offsets.size(); ++i)
+            {
+                if constexpr (aggregate_operation == AggregateOperation::sum)
+                {
+                    size_t array_size = offsets[i] - pos;
+                    /// Just multiply the value by array size.
+                    res[i] = x * array_size;
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::min ||
+                                aggregate_operation == AggregateOperation::max)
+                {
+                    res[i] = x;
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::average)
+                {
+                    if constexpr (IsDecimalNumber<Element>)
+                    {
+                        res[i] = DecimalUtils::convertTo<Result>(x, data.getScale());
+                    }
+                    else
+                    {
+                        res[i] = x;
+                    }
+                }
+
+                pos = offsets[i];
+            }
+
+            res_ptr = std::move(res_column);
+            return true;
+        }
+
+        const typename ColVecType::Container & data = column->getData();
+
+        typename ColVecResult::MutablePtr res_column;
+        if constexpr (IsDecimalNumber<Element>)
+            res_column = ColVecResult::create(offsets.size(), data.getScale());
+        else
+            res_column = ColVecResult::create(offsets.size());
+
+        typename ColVecResult::Container & res = res_column->getData();
+
+        size_t pos = 0;
+        for (size_t i = 0; i < offsets.size(); ++i)
+        {
+            AggregationType s = 0;
+
+            /// Array is empty
+            if (offsets[i] == pos)
+            {
+                res[i] = s;
+                continue;
+            }
+
+            size_t count = 1;
+            s = data[pos]; // NOLINT
+            ++pos;
+
+            for (; pos < offsets[i]; ++pos)
+            {
+                auto element = data[pos];
+
+                if constexpr (aggregate_operation == AggregateOperation::sum ||
+                            aggregate_operation == AggregateOperation::average)
+                {
+                    s += element;
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::min)
+                {
+                    if (element < s)
+                    {
+                        s = element;
+                    }
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::max)
+                {
+                    if (element > s)
+                    {
+                        s = element;
+                    }
+                }
+
+                ++count;
+            }
+
+            if constexpr (aggregate_operation == AggregateOperation::average)
+            {
+                if constexpr (IsDecimalNumber<Element>)
+                {
+                    s = s / count;
+                    res[i] = DecimalUtils::convertTo<Result>(s, data.getScale());
+                }
+                else
+                {
+                    res[i] = static_cast<Result>(s) / count;
+                }
+            }
+            else
+            {
+                res[i] = s;
+            }
+        }
+
+        res_ptr = std::move(res_column);
+        return true;
+    }
+
+    static ColumnPtr execute(const ColumnArray & array, ColumnPtr mapped)
+    {
+        const IColumn::Offsets & offsets = array.getOffsets();
+        ColumnPtr res;
+
+        if (executeType<UInt8>(mapped, offsets, res) ||
+            executeType<UInt16>(mapped, offsets, res) ||
+            executeType<UInt32>(mapped, offsets, res) ||
+            executeType<UInt64>(mapped, offsets, res) ||
+            executeType<UInt256>(mapped, offsets, res) ||
+            executeType<Int8>(mapped, offsets, res) ||
+            executeType<Int16>(mapped, offsets, res) ||
+            executeType<Int32>(mapped, offsets, res) ||
+            executeType<Int64>(mapped, offsets, res) ||
+            executeType<Int128>(mapped, offsets, res) ||
+            executeType<Int256>(mapped, offsets, res) ||
+            executeType<Float32>(mapped, offsets, res) ||
+            executeType<Float64>(mapped, offsets, res) ||
+            executeType<Decimal32>(mapped, offsets, res) ||
+            executeType<Decimal64>(mapped, offsets, res) ||
+            executeType<Decimal128>(mapped, offsets, res))
+            return res;
+        else
+            throw Exception("Unexpected column for arraySum: " + mapped->getName(), ErrorCodes::ILLEGAL_COLUMN);
+    }
+};
+
+struct NameArrayMin { static constexpr auto name = "arrayMin"; };
+using FunctionArrayMin = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::min>, NameArrayMin>;
+
+struct NameArrayMax { static constexpr auto name = "arrayMax"; };
+using FunctionArrayMax = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::max>, NameArrayMax>;
+
+struct NameArraySum { static constexpr auto name = "arraySum"; };
+using FunctionArraySum = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::sum>, NameArraySum>;
+
+struct NameArrayAverage { static constexpr auto name = "arrayAvg"; };
+using FunctionArrayAverage = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::average>, NameArrayAverage>;
+
+void registerFunctionArrayAggregation(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionArrayMin>();
+    factory.registerFunction<FunctionArrayMax>();
+    factory.registerFunction<FunctionArraySum>();
+    factory.registerFunction<FunctionArrayAverage>();
+}
+
+}
+
diff --git a/src/Functions/array/arrayElement.cpp b/src/Functions/array/arrayElement.cpp
index 46a384db449..310b6b5b450 100644
--- a/src/Functions/array/arrayElement.cpp
+++ b/src/Functions/array/arrayElement.cpp
@@ -4,12 +4,16 @@
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypesNumber.h>
 #include <Core/ColumnNumbers.h>
 #include <Columns/ColumnArray.h>
+#include <Core/Field.h>
 #include <Columns/ColumnNullable.h>
 #include <Columns/ColumnsNumber.h>
 #include <Columns/ColumnString.h>
 #include <Columns/ColumnTuple.h>
+#include <Columns/ColumnMap.h>
 #include <Common/typeid_cast.h>
 #include <Common/assert_cast.h>
 
@@ -80,6 +84,38 @@ private:
     /** For a tuple array, the function is evaluated component-wise for each element of the tuple.
       */
     ColumnPtr executeTuple(const ColumnsWithTypeAndName & arguments, size_t input_rows_count) const;
+
+    /** For a map the function finds the matched value for a key.
+     *  Currently implemented just as linear seach in array.
+     *  However, optimizations are possible.
+     */
+    ColumnPtr executeMap(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const;
+
+    using Offsets = ColumnArray::Offsets;
+
+    static bool matchKeyToIndex(const IColumn & data, const Offsets & offsets,
+        const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs);
+
+    static bool matchKeyToIndexConst(const IColumn & data, const Offsets & offsets,
+        const Field & index, PaddedPODArray<UInt64> & matched_idxs);
+
+    template <typename DataType>
+    static bool matchKeyToIndexNumber(const IColumn & data, const Offsets & offsets,
+        const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs);
+
+    template <typename DataType>
+    static bool matchKeyToIndexNumberConst(const IColumn & data, const Offsets & offsets,
+        const Field & index, PaddedPODArray<UInt64> & matched_idxs);
+
+    static bool matchKeyToIndexString(const IColumn & data, const Offsets & offsets,
+        const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs);
+
+    static bool matchKeyToIndexStringConst(const IColumn & data, const Offsets & offsets,
+        const Field & index, PaddedPODArray<UInt64> & matched_idxs);
+
+    template <typename Matcher>
+    static void executeMatchKeyToIndex(const Offsets & offsets,
+        PaddedPODArray<UInt64> & matched_idxs, const Matcher & matcher);
 };
 
 
@@ -626,10 +662,8 @@ ColumnPtr FunctionArrayElement::executeArgument(
     const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, ArrayImpl::NullMapBuilder & builder, size_t input_rows_count) const
 {
     auto index = checkAndGetColumn<ColumnVector<IndexType>>(arguments[1].column.get());
-
     if (!index)
         return nullptr;
-
     const auto & index_data = index->getData();
 
     if (builder)
@@ -704,6 +738,236 @@ ColumnPtr FunctionArrayElement::executeTuple(const ColumnsWithTypeAndName & argu
     return ColumnTuple::create(result_tuple_columns);
 }
 
+namespace
+{
+
+struct MatcherString
+{
+    const ColumnString & data;
+    const ColumnString & index;
+
+    bool match(size_t row_data, size_t row_index) const
+    {
+        auto data_ref = data.getDataAt(row_data);
+        auto index_ref = index.getDataAt(row_index);
+        return memequalSmallAllowOverflow15(index_ref.data, index_ref.size, data_ref.data, data_ref.size);
+    }
+};
+
+struct MatcherStringConst
+{
+    const ColumnString & data;
+    const String & index;
+
+    bool match(size_t row_data, size_t /* row_index */) const
+    {
+        auto data_ref = data.getDataAt(row_data);
+        return index.size() == data_ref.size && memcmp(index.data(), data_ref.data, data_ref.size) == 0;
+    }
+};
+
+template <typename T>
+struct MatcherNumber
+{
+    const PaddedPODArray<T> & data;
+    const PaddedPODArray<T> & index;
+
+    bool match(size_t row_data, size_t row_index) const
+    {
+        return data[row_data] == index[row_index];
+    }
+};
+
+template <typename T>
+struct MatcherNumberConst
+{
+    const PaddedPODArray<T> & data;
+    T index;
+
+    bool match(size_t row_data, size_t /* row_index */) const
+    {
+        return data[row_data] == index;
+    }
+};
+
+}
+
+template <typename Matcher>
+void FunctionArrayElement::executeMatchKeyToIndex(
+    const Offsets & offsets, PaddedPODArray<UInt64> & matched_idxs, const Matcher & matcher)
+{
+    size_t rows = offsets.size();
+    for (size_t i = 0; i < rows; ++i)
+    {
+        bool matched = false;
+        size_t begin = offsets[i - 1];
+        size_t end = offsets[i];
+        for (size_t j = begin; j < end; ++j)
+        {
+            if (matcher.match(j, i))
+            {
+                matched_idxs.push_back(j - begin + 1);
+                matched = true;
+                break;
+            }
+        }
+
+        if (!matched)
+            matched_idxs.push_back(0);
+    }
+}
+
+bool FunctionArrayElement::matchKeyToIndexStringConst(
+    const IColumn & data, const Offsets & offsets,
+    const Field & index, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * data_string = checkAndGetColumn<ColumnString>(&data);
+    if (!data_string)
+        return false;
+
+    if (index.getType() != Field::Types::String)
+        return false;
+
+    MatcherStringConst matcher{*data_string, get<const String &>(index)};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+bool FunctionArrayElement::matchKeyToIndexString(
+    const IColumn & data, const Offsets & offsets,
+    const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * index_string = checkAndGetColumn<ColumnString>(arguments[1].column.get());
+    if (!index_string)
+        return false;
+
+    const auto * data_string = checkAndGetColumn<ColumnString>(&data);
+    if (!data_string)
+        return false;
+
+    MatcherString matcher{*data_string, *index_string};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+template <typename DataType>
+bool FunctionArrayElement::matchKeyToIndexNumberConst(
+    const IColumn & data, const Offsets & offsets,
+    const Field & index, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * data_numeric = checkAndGetColumn<ColumnVector<DataType>>(&data);
+    if (!data_numeric)
+        return false;
+
+    if (index.getType() != Field::Types::UInt64 && index.getType() != Field::Types::Int64)
+        return false;
+
+    MatcherNumberConst<DataType> matcher{data_numeric->getData(), get<DataType>(index)};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+template <typename DataType>
+bool FunctionArrayElement::matchKeyToIndexNumber(
+    const IColumn & data, const Offsets & offsets,
+    const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * index_numeric = checkAndGetColumn<ColumnVector<DataType>>(arguments[1].column.get());
+    if (!index_numeric)
+        return false;
+
+    const auto * data_numeric = checkAndGetColumn<ColumnVector<DataType>>(&data);
+    if (!data_numeric)
+        return false;
+
+    MatcherNumber<DataType> matcher{data_numeric->getData(), index_numeric->getData()};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+bool FunctionArrayElement::matchKeyToIndex(
+    const IColumn & data, const Offsets & offsets,
+    const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs)
+{
+    return matchKeyToIndexNumber<UInt8>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<UInt16>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<UInt32>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<UInt64>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int8>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int16>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int32>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int64>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexString(data, offsets, arguments, matched_idxs);
+}
+
+bool FunctionArrayElement::matchKeyToIndexConst(
+    const IColumn & data, const Offsets & offsets,
+    const Field & index, PaddedPODArray<UInt64> & matched_idxs)
+{
+    return matchKeyToIndexNumberConst<UInt8>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<UInt16>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<UInt32>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<UInt64>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int8>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int16>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int32>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int64>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexStringConst(data, offsets, index, matched_idxs);
+}
+
+ColumnPtr FunctionArrayElement::executeMap(
+    const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const
+{
+    const ColumnMap * col_map = typeid_cast<const ColumnMap *>(arguments[0].column.get());
+    if (!col_map)
+        return nullptr;
+
+    const auto & nested_column = col_map->getNestedColumn();
+    const auto & keys_data = col_map->getNestedData().getColumn(0);
+    const auto & values_data = col_map->getNestedData().getColumn(1);
+    const auto & offsets = nested_column.getOffsets();
+
+    /// At first step calculate indices in array of values for requested keys.
+    auto indices_column = DataTypeNumber<UInt64>().createColumn();
+    indices_column->reserve(input_rows_count);
+    auto & indices_data = assert_cast<ColumnVector<UInt64> &>(*indices_column).getData();
+
+    if (!isColumnConst(*arguments[1].column))
+    {
+        if (input_rows_count > 0 && !matchKeyToIndex(keys_data, offsets, arguments, indices_data))
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                "Illegal types of arguments: {}, {} for function ",
+                arguments[0].type->getName(), arguments[1].type->getName(), getName());
+    }
+    else
+    {
+        Field index = (*arguments[1].column)[0];
+
+        // Get Matched key's value
+        if (input_rows_count > 0 && !matchKeyToIndexConst(keys_data, offsets, index, indices_data))
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                "Illegal types of arguments: {}, {} for function {}",
+                arguments[0].type->getName(), arguments[1].type->getName(), getName());
+    }
+
+    /// Prepare arguments to call arrayElement for array with values and calculated indices at previous step.
+    ColumnsWithTypeAndName new_arguments =
+    {
+        {
+            ColumnArray::create(values_data.getPtr(), nested_column.getOffsetsPtr()),
+            std::make_shared<DataTypeArray>(result_type),
+            ""
+        },
+        {
+            std::move(indices_column),
+            std::make_shared<DataTypeNumber<UInt64>>(),
+            ""
+        }
+    };
+
+    return executeImpl(new_arguments, result_type, input_rows_count);
+}
+
 String FunctionArrayElement::getName() const
 {
     return name;
@@ -711,7 +975,10 @@ String FunctionArrayElement::getName() const
 
 DataTypePtr FunctionArrayElement::getReturnTypeImpl(const DataTypes & arguments) const
 {
-    const DataTypeArray * array_type = checkAndGetDataType<DataTypeArray>(arguments[0].get());
+    if (const auto * map_type = checkAndGetDataType<DataTypeMap>(arguments[0].get()))
+        return map_type->getValueType();
+
+    const auto * array_type = checkAndGetDataType<DataTypeArray>(arguments[0].get());
     if (!array_type)
     {
         throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
@@ -734,6 +1001,10 @@ ColumnPtr FunctionArrayElement::executeImpl(const ColumnsWithTypeAndName & argum
     /// Check nullability.
     bool is_array_of_nullable = false;
 
+    const ColumnMap * col_map = checkAndGetColumn<ColumnMap>(arguments[0].column.get());
+    if (col_map)
+        return executeMap(arguments, result_type, input_rows_count);
+
     const ColumnArray * col_array = nullptr;
     const ColumnArray * col_const_array = nullptr;
 
@@ -826,7 +1097,7 @@ ColumnPtr FunctionArrayElement::perform(const ColumnsWithTypeAndName & arguments
             || (res = executeArgument<Int16>(arguments, result_type, builder, input_rows_count))
             || (res = executeArgument<Int32>(arguments, result_type, builder, input_rows_count))
             || (res = executeArgument<Int64>(arguments, result_type, builder, input_rows_count))))
-        throw Exception("Second argument for function " + getName() + " must must have UInt or Int type.",
+        throw Exception("Second argument for function " + getName() + " must have UInt or Int type.",
                         ErrorCodes::ILLEGAL_COLUMN);
     }
     else
diff --git a/src/Functions/array/arrayEnumerateRanked.h b/src/Functions/array/arrayEnumerateRanked.h
index a0d865f08fe..ce82b8bcd86 100644
--- a/src/Functions/array/arrayEnumerateRanked.h
+++ b/src/Functions/array/arrayEnumerateRanked.h
@@ -304,10 +304,10 @@ void FunctionArrayEnumerateRankedExtended<Derived>::executeMethodImpl(
     const size_t depth_to_look = arrays_depths.max_array_depth;
     const auto & offsets = *offsets_by_depth[depth_to_look - 1];
 
-    using Map = ClearableHashMapWithStackMemory<UInt128, UInt32,
+    using Container = ClearableHashMapWithStackMemory<UInt128, UInt32,
         UInt128TrivialHash, INITIAL_SIZE_DEGREE>;
 
-    Map indices;
+    Container indices;
 
     std::vector<size_t> indices_by_depth(depth_to_look);
     std::vector<size_t> current_offset_n_by_depth(depth_to_look);
diff --git a/src/Functions/array/arrayIntersect.cpp b/src/Functions/array/arrayIntersect.cpp
index 1eb3e7ac6c3..736069d6060 100644
--- a/src/Functions/array/arrayIntersect.cpp
+++ b/src/Functions/array/arrayIntersect.cpp
@@ -438,22 +438,22 @@ ColumnPtr FunctionArrayIntersect::executeImpl(const ColumnsWithTypeAndName & arg
 template <typename T, size_t>
 void FunctionArrayIntersect::NumberExecutor::operator()()
 {
-    using Map = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
+    using Container = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
         INITIAL_SIZE_DEGREE>;
 
     if (!result && typeid_cast<const DataTypeNumber<T> *>(data_type.get()))
-        result = execute<Map, ColumnVector<T>, true>(arrays, ColumnVector<T>::create());
+        result = execute<Container, ColumnVector<T>, true>(arrays, ColumnVector<T>::create());
 }
 
 template <typename T, size_t>
 void FunctionArrayIntersect::DecimalExecutor::operator()()
 {
-    using Map = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
+    using Container = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
         INITIAL_SIZE_DEGREE>;
 
     if (!result)
         if (auto * decimal = typeid_cast<const DataTypeDecimal<T> *>(data_type.get()))
-            result = execute<Map, ColumnDecimal<T>, true>(arrays, ColumnDecimal<T>::create(0, decimal->getScale()));
+            result = execute<Container, ColumnDecimal<T>, true>(arrays, ColumnDecimal<T>::create(0, decimal->getScale()));
 }
 
 template <typename Map, typename ColumnType, bool is_numeric_column>
diff --git a/src/Functions/array/arraySum.cpp b/src/Functions/array/arraySum.cpp
deleted file mode 100644
index 1aedcb6ef92..00000000000
--- a/src/Functions/array/arraySum.cpp
+++ /dev/null
@@ -1,146 +0,0 @@
-#include <DataTypes/DataTypesNumber.h>
-#include <DataTypes/DataTypesDecimal.h>
-#include <Columns/ColumnsNumber.h>
-#include <Columns/ColumnDecimal.h>
-#include "FunctionArrayMapped.h"
-#include <Functions/FunctionFactory.h>
-
-
-namespace DB
-{
-
-namespace ErrorCodes
-{
-    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
-    extern const int ILLEGAL_COLUMN;
-}
-
-struct ArraySumImpl
-{
-    static bool needBoolean() { return false; }
-    static bool needExpression() { return false; }
-    static bool needOneArray() { return false; }
-
-    static DataTypePtr getReturnType(const DataTypePtr & expression_return, const DataTypePtr & /*array_element*/)
-    {
-        WhichDataType which(expression_return);
-
-        if (which.isNativeUInt())
-            return std::make_shared<DataTypeUInt64>();
-
-        if (which.isNativeInt())
-            return std::make_shared<DataTypeInt64>();
-
-        if (which.isFloat())
-            return std::make_shared<DataTypeFloat64>();
-
-        if (which.isDecimal())
-        {
-            UInt32 scale = getDecimalScale(*expression_return);
-            return std::make_shared<DataTypeDecimal<Decimal128>>(DecimalUtils::maxPrecision<Decimal128>(), scale);
-        }
-
-        throw Exception("arraySum cannot add values of type " + expression_return->getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-    }
-
-    template <typename Element, typename Result>
-    static bool executeType(const ColumnPtr & mapped, const ColumnArray::Offsets & offsets, ColumnPtr & res_ptr)
-    {
-        using ColVecType = std::conditional_t<IsDecimalNumber<Element>, ColumnDecimal<Element>, ColumnVector<Element>>;
-        using ColVecResult = std::conditional_t<IsDecimalNumber<Result>, ColumnDecimal<Result>, ColumnVector<Result>>;
-
-        const ColVecType * column = checkAndGetColumn<ColVecType>(&*mapped);
-
-        /// Constant case.
-        if (!column)
-        {
-            const ColumnConst * column_const = checkAndGetColumnConst<ColVecType>(&*mapped);
-
-            if (!column_const)
-                return false;
-
-            const Result x = column_const->template getValue<Element>(); // NOLINT
-
-            typename ColVecResult::MutablePtr res_column;
-            if constexpr (IsDecimalNumber<Element>)
-            {
-                const typename ColVecType::Container & data =
-                    checkAndGetColumn<ColVecType>(&column_const->getDataColumn())->getData();
-                res_column = ColVecResult::create(offsets.size(), data.getScale());
-            }
-            else
-                res_column = ColVecResult::create(offsets.size());
-
-            typename ColVecResult::Container & res = res_column->getData();
-
-            size_t pos = 0;
-            for (size_t i = 0; i < offsets.size(); ++i)
-            {
-                /// Just multiply the value by array size.
-                res[i] = x * (offsets[i] - pos);
-                pos = offsets[i];
-            }
-
-            res_ptr = std::move(res_column);
-            return true;
-        }
-
-        const typename ColVecType::Container & data = column->getData();
-
-        typename ColVecResult::MutablePtr res_column;
-        if constexpr (IsDecimalNumber<Element>)
-            res_column = ColVecResult::create(offsets.size(), data.getScale());
-        else
-            res_column = ColVecResult::create(offsets.size());
-
-        typename ColVecResult::Container & res = res_column->getData();
-
-        size_t pos = 0;
-        for (size_t i = 0; i < offsets.size(); ++i)
-        {
-            Result s = 0;
-            for (; pos < offsets[i]; ++pos)
-            {
-                s += data[pos];
-            }
-            res[i] = s;
-        }
-
-        res_ptr = std::move(res_column);
-        return true;
-    }
-
-    static ColumnPtr execute(const ColumnArray & array, ColumnPtr mapped)
-    {
-        const IColumn::Offsets & offsets = array.getOffsets();
-        ColumnPtr res;
-
-        if (executeType< UInt8 , UInt64>(mapped, offsets, res) ||
-            executeType< UInt16, UInt64>(mapped, offsets, res) ||
-            executeType< UInt32, UInt64>(mapped, offsets, res) ||
-            executeType< UInt64, UInt64>(mapped, offsets, res) ||
-            executeType<  Int8 ,  Int64>(mapped, offsets, res) ||
-            executeType<  Int16,  Int64>(mapped, offsets, res) ||
-            executeType<  Int32,  Int64>(mapped, offsets, res) ||
-            executeType<  Int64,  Int64>(mapped, offsets, res) ||
-            executeType<Float32,Float64>(mapped, offsets, res) ||
-            executeType<Float64,Float64>(mapped, offsets, res) ||
-            executeType<Decimal32, Decimal128>(mapped, offsets, res) ||
-            executeType<Decimal64, Decimal128>(mapped, offsets, res) ||
-            executeType<Decimal128, Decimal128>(mapped, offsets, res))
-            return res;
-        else
-            throw Exception("Unexpected column for arraySum: " + mapped->getName(), ErrorCodes::ILLEGAL_COLUMN);
-    }
-};
-
-struct NameArraySum { static constexpr auto name = "arraySum"; };
-using FunctionArraySum = FunctionArrayMapped<ArraySumImpl, NameArraySum>;
-
-void registerFunctionArraySum(FunctionFactory & factory)
-{
-    factory.registerFunction<FunctionArraySum>();
-}
-
-}
-
diff --git a/src/Functions/map.cpp b/src/Functions/map.cpp
new file mode 100644
index 00000000000..5993ab3706e
--- /dev/null
+++ b/src/Functions/map.cpp
@@ -0,0 +1,140 @@
+#include <Functions/IFunctionImpl.h>
+#include <Functions/FunctionFactory.h>
+#include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Columns/ColumnMap.h>
+#include <Columns/ColumnArray.h>
+#include <DataTypes/getLeastSupertype.h>
+#include <Interpreters/castColumn.h>
+#include <memory>
+
+
+namespace DB
+{
+namespace ErrorCodes
+{
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+}
+
+namespace
+{
+
+// map(x, y, ...) is a function that allows you to make key-value pair
+class FunctionMap : public IFunction
+{
+public:
+    static constexpr auto name = "map";
+
+    static FunctionPtr create(const Context &)
+    {
+        return std::make_shared<FunctionMap>();
+    }
+
+    String getName() const override
+    {
+        return name;
+    }
+
+    bool isVariadic() const override
+    {
+        return true;
+    }
+
+    size_t getNumberOfArguments() const override
+    {
+        return 0;
+    }
+
+    bool isInjective(const ColumnsWithTypeAndName &) const override
+    {
+        return true;
+    }
+
+    bool useDefaultImplementationForNulls() const override { return false; }
+    bool useDefaultImplementationForConstants() const override { return true; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (arguments.size() % 2 != 0)
+            throw Exception("Function " + getName() + " even number of arguments.", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+        DataTypes keys, values;
+        for (size_t i = 0; i < arguments.size(); i += 2)
+        {
+            keys.emplace_back(arguments[i]);
+            values.emplace_back(arguments[i + 1]);
+        }
+
+        DataTypes tmp;
+        tmp.emplace_back(getLeastSupertype(keys));
+        tmp.emplace_back(getLeastSupertype(values));
+        return std::make_shared<DataTypeMap>(tmp);
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
+    {
+        size_t num_elements = arguments.size();
+
+        if (num_elements == 0)
+            return result_type->createColumnConstWithDefaultValue(input_rows_count);
+
+        const auto & result_type_map = static_cast<const DataTypeMap &>(*result_type);
+        const DataTypePtr & key_type = result_type_map.getKeyType();
+        const DataTypePtr & value_type = result_type_map.getValueType();
+
+        Columns columns_holder(num_elements);
+        ColumnRawPtrs column_ptrs(num_elements);
+
+        for (size_t i = 0; i < num_elements; ++i)
+        {
+            const auto & arg = arguments[i];
+            const auto to_type = i % 2 == 0 ? key_type : value_type;
+
+            ColumnPtr preprocessed_column = castColumn(arg, to_type);
+            preprocessed_column = preprocessed_column->convertToFullColumnIfConst();
+
+            columns_holder[i] = std::move(preprocessed_column);
+            column_ptrs[i] = columns_holder[i].get();
+        }
+
+        /// Create and fill the result map.
+
+        MutableColumnPtr keys_data = key_type->createColumn();
+        MutableColumnPtr values_data = value_type->createColumn();
+        MutableColumnPtr offsets = DataTypeNumber<IColumn::Offset>().createColumn();
+
+        size_t total_elements = input_rows_count * num_elements / 2;
+        keys_data->reserve(total_elements);
+        values_data->reserve(total_elements);
+        offsets->reserve(input_rows_count);
+
+        IColumn::Offset current_offset = 0;
+        for (size_t i = 0; i < input_rows_count; ++i)
+        {
+            for (size_t j = 0; j < num_elements; j += 2)
+            {
+                keys_data->insertFrom(*column_ptrs[j], i);
+                values_data->insertFrom(*column_ptrs[j + 1], i);
+            }
+
+            current_offset += num_elements / 2;
+            offsets->insert(current_offset);
+        }
+
+        auto nested_column = ColumnArray::create(
+            ColumnTuple::create(Columns{std::move(keys_data), std::move(values_data)}),
+            std::move(offsets));
+
+        return ColumnMap::create(nested_column);
+    }
+};
+
+}
+
+void registerFunctionsMap(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionMap>();
+}
+
+}
diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index 7552af7dbf1..54695a1087d 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -1,3 +1,4 @@
+#include <type_traits>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionBinaryArithmetic.h>
 #include <common/arithmeticOverflow.h>
@@ -25,11 +26,17 @@ struct MultiplyImpl
             return static_cast<Result>(a) * b;
     }
 
-    /// Apply operation and check overflow. It's used for Deciamal operations. @returns true if overflowed, false otherwise.
+    /// Apply operation and check overflow. It's used for Decimal operations. @returns true if overflowed, false otherwise.
     template <typename Result = ResultType>
     static inline bool apply(A a, B b, Result & c)
     {
-        return common::mulOverflow(static_cast<Result>(a), b, c);
+        if constexpr (std::is_same_v<Result, float> || std::is_same_v<Result, double>)
+        {
+            c = static_cast<Result>(a) * b;
+            return false;
+        }
+        else
+            return common::mulOverflow(static_cast<Result>(a), b, c);
     }
 
 #if USE_EMBEDDED_COMPILER
diff --git a/src/Functions/registerFunctions.cpp b/src/Functions/registerFunctions.cpp
index 6dfebc13665..3f75746f861 100644
--- a/src/Functions/registerFunctions.cpp
+++ b/src/Functions/registerFunctions.cpp
@@ -10,6 +10,7 @@ namespace DB
 void registerFunctionsArithmetic(FunctionFactory &);
 void registerFunctionsArray(FunctionFactory &);
 void registerFunctionsTuple(FunctionFactory &);
+void registerFunctionsMap(FunctionFactory &);
 void registerFunctionsBitmap(FunctionFactory &);
 void registerFunctionsCoding(FunctionFactory &);
 void registerFunctionsComparison(FunctionFactory &);
@@ -64,6 +65,7 @@ void registerFunctions()
     registerFunctionsArithmetic(factory);
     registerFunctionsArray(factory);
     registerFunctionsTuple(factory);
+    registerFunctionsMap(factory);
 #if !defined(ARCADIA_BUILD)
     registerFunctionsBitmap(factory);
 #endif
diff --git a/src/Functions/registerFunctionsHigherOrder.cpp b/src/Functions/registerFunctionsHigherOrder.cpp
index 08938ef6534..d3621a03ecd 100644
--- a/src/Functions/registerFunctionsHigherOrder.cpp
+++ b/src/Functions/registerFunctionsHigherOrder.cpp
@@ -9,7 +9,7 @@ void registerFunctionArrayCount(FunctionFactory & factory);
 void registerFunctionArrayExists(FunctionFactory & factory);
 void registerFunctionArrayAll(FunctionFactory & factory);
 void registerFunctionArrayCompact(FunctionFactory & factory);
-void registerFunctionArraySum(FunctionFactory & factory);
+void registerFunctionArrayAggregation(FunctionFactory & factory);
 void registerFunctionArrayFirst(FunctionFactory & factory);
 void registerFunctionArrayFirstIndex(FunctionFactory & factory);
 void registerFunctionsArrayFill(FunctionFactory & factory);
@@ -27,7 +27,7 @@ void registerFunctionsHigherOrder(FunctionFactory & factory)
     registerFunctionArrayExists(factory);
     registerFunctionArrayAll(factory);
     registerFunctionArrayCompact(factory);
-    registerFunctionArraySum(factory);
+    registerFunctionArrayAggregation(factory);
     registerFunctionArrayFirst(factory);
     registerFunctionArrayFirstIndex(factory);
     registerFunctionsArrayFill(factory);
diff --git a/src/Functions/toFixedString.h b/src/Functions/toFixedString.h
index de1ff7e8084..23c9e92faaf 100644
--- a/src/Functions/toFixedString.h
+++ b/src/Functions/toFixedString.h
@@ -5,6 +5,8 @@
 #include <DataTypes/DataTypesNumber.h>
 #include <Columns/ColumnString.h>
 #include <Columns/ColumnFixedString.h>
+#include <Columns/ColumnsNumber.h>
+#include <Columns/ColumnNullable.h>
 #include <IO/WriteHelpers.h>
 
 
@@ -18,6 +20,11 @@ namespace ErrorCodes
     extern const int NOT_IMPLEMENTED;
 }
 
+enum class ConvertToFixedStringExceptionMode
+{
+    Throw,
+    Null
+};
 
 /** Conversion to fixed string is implemented only for strings.
   */
@@ -55,13 +62,22 @@ public:
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
     {
         const auto n = arguments[1].column->getUInt(0);
-        return executeForN(arguments, n);
+        return executeForN<ConvertToFixedStringExceptionMode::Throw>(arguments, n);
     }
 
+    template<ConvertToFixedStringExceptionMode exception_mode>
     static ColumnPtr executeForN(const ColumnsWithTypeAndName & arguments, const size_t n)
     {
         const auto & column = arguments[0].column;
 
+        ColumnUInt8::MutablePtr col_null_map_to;
+        ColumnUInt8::Container * vec_null_map_to [[maybe_unused]] = nullptr;
+        if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Null)
+        {
+            col_null_map_to = ColumnUInt8::create(column->size(), false);
+            vec_null_map_to = &col_null_map_to->getData();
+        }
+
         if (const auto * column_string = checkAndGetColumn<ColumnString>(column.get()))
         {
             auto column_fixed = ColumnFixedString::create(n);
@@ -77,18 +93,42 @@ public:
                 const size_t off = i ? in_offsets[i - 1] : 0;
                 const size_t len = in_offsets[i] - off - 1;
                 if (len > n)
-                    throw Exception("String too long for type FixedString(" + toString(n) + ")",
-                        ErrorCodes::TOO_LARGE_STRING_SIZE);
+                {
+                    if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Throw)
+                    {
+                        throw Exception("String too long for type FixedString(" + toString(n) + ")",
+                            ErrorCodes::TOO_LARGE_STRING_SIZE);
+                    }
+                    else
+                    {
+                        (*vec_null_map_to)[i] = true;
+                        continue;
+                    }
+                }
                 memcpy(&out_chars[i * n], &in_chars[off], len);
             }
 
-            return column_fixed;
+            if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Null)
+                return ColumnNullable::create(std::move(column_fixed), std::move(col_null_map_to));
+            else
+                return column_fixed;
         }
         else if (const auto * column_fixed_string = checkAndGetColumn<ColumnFixedString>(column.get()))
         {
             const auto src_n = column_fixed_string->getN();
             if (src_n > n)
-                throw Exception{"String too long for type FixedString(" + toString(n) + ")", ErrorCodes::TOO_LARGE_STRING_SIZE};
+            {
+                if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Throw)
+                {
+                    throw Exception{"String too long for type FixedString(" + toString(n) + ")", ErrorCodes::TOO_LARGE_STRING_SIZE};
+                }
+                else
+                {
+                    auto column_fixed = ColumnFixedString::create(n);
+                    std::fill(vec_null_map_to->begin(), vec_null_map_to->end(), true);
+                    return ColumnNullable::create(column_fixed->cloneResized(column->size()), std::move(col_null_map_to));
+                }
+            }
 
             auto column_fixed = ColumnFixedString::create(n);
 
@@ -103,7 +143,16 @@ public:
             return column_fixed;
         }
         else
-            throw Exception("Unexpected column: " + column->getName(), ErrorCodes::ILLEGAL_COLUMN);
+        {
+            if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Throw)
+                throw Exception("Unexpected column: " + column->getName(), ErrorCodes::ILLEGAL_COLUMN);
+            else
+            {
+                auto column_fixed = ColumnFixedString::create(n);
+                std::fill(vec_null_map_to->begin(), vec_null_map_to->end(), true);
+                return ColumnNullable::create(column_fixed->cloneResized(column->size()), std::move(col_null_map_to));
+            }
+        }
     }
 };
 
diff --git a/src/Functions/toModifiedJulianDay.cpp b/src/Functions/toModifiedJulianDay.cpp
index df3066a5005..32e9eb23987 100644
--- a/src/Functions/toModifiedJulianDay.cpp
+++ b/src/Functions/toModifiedJulianDay.cpp
@@ -62,11 +62,11 @@ namespace DB
             typename ColVecTo::Container & vec_to = col_to->getData();
 
             ColumnUInt8::MutablePtr col_null_map_to;
-            ColumnUInt8::Container * vec_null_map_to [[maybe_unused]] = nullptr;
+            UInt8 * vec_null_map_to [[maybe_unused]] = nullptr;
             if constexpr (nullOnErrors)
             {
                 col_null_map_to = ColumnUInt8::create(input_rows_count);
-                vec_null_map_to = &col_null_map_to->getData();
+                vec_null_map_to = col_null_map_to->getData().data();
             }
 
             size_t current_offset = 0;
@@ -83,12 +83,15 @@ namespace DB
                     {
                         const GregorianDate<> date(read_buffer);
                         vec_to[i] = date.toModifiedJulianDay<typename ToDataType::FieldType>();
-                        (*vec_null_map_to)[i] = false;
+                        vec_null_map_to[i] = false;
                     }
                     catch (const Exception & e)
                     {
                         if (e.code() == ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED || e.code() == ErrorCodes::CANNOT_PARSE_DATE)
-                            (*vec_null_map_to)[i] = true;
+                        {
+                            vec_to[i] = static_cast<Int32>(0);
+                            vec_null_map_to[i] = true;
+                        }
                         else
                             throw;
                     }
diff --git a/src/Functions/trap.cpp b/src/Functions/trap.cpp
index 34553f4b266..38cf53948e6 100644
--- a/src/Functions/trap.cpp
+++ b/src/Functions/trap.cpp
@@ -61,9 +61,10 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
-    [[clang::optnone]] void executeImpl(Block & columns, const ColumnNumbers & arguments, size_t result, size_t input_rows_count) const override
+    [[clang::optnone]]
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & block, const DataTypePtr & result_type, size_t input_rows_count) const override
     {
-        if (const ColumnConst * column = checkAndGetColumnConst<ColumnString>(columns[arguments[0]].column.get()))
+        if (const ColumnConst * column = checkAndGetColumnConst<ColumnString>(block[0].column.get()))
         {
             String mode = column->getValue<String>();
 
@@ -135,6 +136,10 @@ public:
             {
                 (void)context.getCurrentQueryId();
             }
+            else if (mode == "stack overflow")
+            {
+                executeImpl(block, result_type, input_rows_count);
+            }
             else if (mode == "mmap many")
             {
                 std::vector<void *> maps;
@@ -160,7 +165,7 @@ public:
         else
             throw Exception("The only argument for function " + getName() + " must be constant String", ErrorCodes::ILLEGAL_COLUMN);
 
-        columns[result].column = columns[result].type->createColumnConst(input_rows_count, 0ULL);
+        return result_type->createColumnConst(input_rows_count, 0ULL);
     }
 };
 
diff --git a/src/Functions/ya.make b/src/Functions/ya.make
index 12013dd22b6..6e5d832db77 100644
--- a/src/Functions/ya.make
+++ b/src/Functions/ya.make
@@ -120,6 +120,7 @@ SRCS(
     appendTrailingCharIfAbsent.cpp
     array/array.cpp
     array/arrayAUC.cpp
+    array/arrayAggregation.cpp
     array/arrayAll.cpp
     array/arrayCompact.cpp
     array/arrayConcat.cpp
@@ -155,7 +156,6 @@ SRCS(
     array/arraySlice.cpp
     array/arraySort.cpp
     array/arraySplit.cpp
-    array/arraySum.cpp
     array/arrayUniq.cpp
     array/arrayWithConstant.cpp
     array/arrayZip.cpp
@@ -325,6 +325,7 @@ SRCS(
     lowCardinalityKeys.cpp
     lower.cpp
     lowerUTF8.cpp
+    map.cpp
     match.cpp
     materialize.cpp
     minus.cpp
diff --git a/src/IO/HDFSCommon.cpp b/src/IO/HDFSCommon.cpp
deleted file mode 100644
index 4789dd2b0f4..00000000000
--- a/src/IO/HDFSCommon.cpp
+++ /dev/null
@@ -1,62 +0,0 @@
-#include <IO/HDFSCommon.h>
-#include <Poco/URI.h>
-
-#if USE_HDFS
-#include <Common/Exception.h>
-
-namespace DB
-{
-namespace ErrorCodes
-{
-extern const int BAD_ARGUMENTS;
-extern const int NETWORK_ERROR;
-}
-
-HDFSBuilderPtr createHDFSBuilder(const std::string & uri_str)
-{
-    const Poco::URI uri(uri_str);
-    const auto & host = uri.getHost();
-    auto port = uri.getPort();
-    const std::string path = "//";
-    if (host.empty())
-        throw Exception("Illegal HDFS URI: " + uri.toString(), ErrorCodes::BAD_ARGUMENTS);
-
-    HDFSBuilderPtr builder(hdfsNewBuilder());
-    if (builder == nullptr)
-        throw Exception("Unable to create builder to connect to HDFS: " + uri.toString() + " " + std::string(hdfsGetLastError()),
-            ErrorCodes::NETWORK_ERROR);
-    hdfsBuilderConfSetStr(builder.get(), "input.read.timeout", "60000"); // 1 min
-    hdfsBuilderConfSetStr(builder.get(), "input.write.timeout", "60000"); // 1 min
-    hdfsBuilderConfSetStr(builder.get(), "input.connect.timeout", "60000"); // 1 min
-
-    std::string user_info = uri.getUserInfo();
-    if (!user_info.empty() && user_info.front() != ':')
-    {
-        std::string user;
-        size_t delim_pos = user_info.find(':');
-        if (delim_pos != std::string::npos)
-            user = user_info.substr(0, delim_pos);
-        else
-            user = user_info;
-
-        hdfsBuilderSetUserName(builder.get(), user.c_str());
-    }
-    hdfsBuilderSetNameNode(builder.get(), host.c_str());
-    if (port != 0)
-    {
-        hdfsBuilderSetNameNodePort(builder.get(), port);
-    }
-    return builder;
-}
-
-HDFSFSPtr createHDFSFS(hdfsBuilder * builder)
-{
-    HDFSFSPtr fs(hdfsBuilderConnect(builder));
-    if (fs == nullptr)
-        throw Exception("Unable to connect to HDFS: " + std::string(hdfsGetLastError()),
-            ErrorCodes::NETWORK_ERROR);
-
-    return fs;
-}
-}
-#endif
diff --git a/src/IO/HDFSCommon.h b/src/IO/HDFSCommon.h
deleted file mode 100644
index 3d212ddc347..00000000000
--- a/src/IO/HDFSCommon.h
+++ /dev/null
@@ -1,58 +0,0 @@
-#pragma once
-#include <Common/config.h>
-#include <memory>
-#include <type_traits>
-
-#if USE_HDFS
-#include <hdfs/hdfs.h>
-
-namespace DB
-{
-namespace detail
-{
-struct HDFSBuilderDeleter
-{
-    void operator()(hdfsBuilder * builder_ptr)
-    {
-        hdfsFreeBuilder(builder_ptr);
-    }
-};
-struct HDFSFsDeleter
-{
-    void operator()(hdfsFS fs_ptr)
-    {
-        hdfsDisconnect(fs_ptr);
-    }
-};
-
-}
-
-struct HDFSFileInfo
-{
-    hdfsFileInfo * file_info;
-    int length;
-
-    HDFSFileInfo()
-        : file_info(nullptr)
-        , length(0)
-    {
-    }
-    HDFSFileInfo(const HDFSFileInfo & other) = delete;
-    HDFSFileInfo(HDFSFileInfo && other) = default;
-    HDFSFileInfo & operator=(const HDFSFileInfo & other) = delete;
-    HDFSFileInfo & operator=(HDFSFileInfo && other) = default;
-
-    ~HDFSFileInfo()
-    {
-        hdfsFreeFileInfo(file_info, length);
-    }
-};
-using HDFSBuilderPtr = std::unique_ptr<hdfsBuilder, detail::HDFSBuilderDeleter>;
-using HDFSFSPtr = std::unique_ptr<std::remove_pointer_t<hdfsFS>, detail::HDFSFsDeleter>;
-
-// set read/connect timeout, default value in libhdfs3 is about 1 hour, and too large
-/// TODO Allow to tune from query Settings.
-HDFSBuilderPtr createHDFSBuilder(const std::string & uri_str);
-HDFSFSPtr createHDFSFS(hdfsBuilder * builder);
-}
-#endif
diff --git a/src/IO/MMapReadBufferFromFileDescriptor.cpp b/src/IO/MMapReadBufferFromFileDescriptor.cpp
index bbeec8959b4..2034a116e0c 100644
--- a/src/IO/MMapReadBufferFromFileDescriptor.cpp
+++ b/src/IO/MMapReadBufferFromFileDescriptor.cpp
@@ -6,6 +6,7 @@
 #include <Common/ProfileEvents.h>
 #include <Common/formatReadable.h>
 #include <Common/Exception.h>
+#include <common/getPageSize.h>
 #include <IO/WriteHelpers.h>
 #include <IO/MMapReadBufferFromFileDescriptor.h>
 
@@ -38,7 +39,9 @@ void MMapReadBufferFromFileDescriptor::init(int fd_, size_t offset, size_t lengt
                 ErrorCodes::CANNOT_ALLOCATE_MEMORY);
 
         BufferBase::set(static_cast<char *>(buf), length, 0);
-        ReadBuffer::padded = (length % 4096) > 0 && (length % 4096) <= (4096 - 15); /// TODO determine page size
+
+        size_t page_size = static_cast<size_t>(::getPageSize());
+        ReadBuffer::padded = (length % page_size) > 0 && (length % page_size) <= (page_size - 15);
     }
 }
 
diff --git a/src/IO/S3Common.cpp b/src/IO/S3Common.cpp
index aad6e984568..06c51e058a0 100644
--- a/src/IO/S3Common.cpp
+++ b/src/IO/S3Common.cpp
@@ -368,8 +368,12 @@ namespace S3
                 throw Exception(
                     "Bucket name length is out of bounds in virtual hosted style S3 URI: " + bucket + " (" + uri.toString() + ")", ErrorCodes::BAD_ARGUMENTS);
 
-            /// Remove leading '/' from path to extract key.
-            key = uri.getPath().substr(1);
+            if (!uri.getPath().empty())
+            {
+                /// Remove leading '/' from path to extract key.
+                key = uri.getPath().substr(1);
+            }
+
             if (key.empty() || key == "/")
                 throw Exception("Key name is empty in virtual hosted style S3 URI: " + key + " (" + uri.toString() + ")", ErrorCodes::BAD_ARGUMENTS);
             boost::to_upper(name);
diff --git a/src/IO/WriteBufferFromS3.cpp b/src/IO/WriteBufferFromS3.cpp
index e4fb0634bd0..09aabb1b21d 100644
--- a/src/IO/WriteBufferFromS3.cpp
+++ b/src/IO/WriteBufferFromS3.cpp
@@ -5,12 +5,9 @@
 #    include <IO/WriteBufferFromS3.h>
 #    include <IO/WriteHelpers.h>
 
-#    include <aws/core/utils/memory/stl/AWSStreamFwd.h>
-#    include <aws/core/utils/memory/stl/AWSStringStream.h>
 #    include <aws/s3/S3Client.h>
-#    include <aws/s3/model/CompleteMultipartUploadRequest.h>
-#    include <aws/s3/model/AbortMultipartUploadRequest.h>
 #    include <aws/s3/model/CreateMultipartUploadRequest.h>
+#    include <aws/s3/model/CompleteMultipartUploadRequest.h>
 #    include <aws/s3/model/PutObjectRequest.h>
 #    include <aws/s3/model/UploadPartRequest.h>
 #    include <common/logger_useful.h>
@@ -42,22 +39,19 @@ WriteBufferFromS3::WriteBufferFromS3(
     const String & bucket_,
     const String & key_,
     size_t minimum_upload_part_size_,
-    bool is_multipart_,
+    size_t max_single_part_upload_size_,
     std::optional<std::map<String, String>> object_metadata_,
     size_t buffer_size_)
     : BufferWithOwnMemory<WriteBuffer>(buffer_size_, nullptr, 0)
-    , is_multipart(is_multipart_)
     , bucket(bucket_)
     , key(key_)
     , object_metadata(std::move(object_metadata_))
     , client_ptr(std::move(client_ptr_))
-    , minimum_upload_part_size{minimum_upload_part_size_}
-    , temporary_buffer{std::make_unique<WriteBufferFromOwnString>()}
-    , last_part_size{0}
-{
-    if (is_multipart)
-        initiate();
-}
+    , minimum_upload_part_size(minimum_upload_part_size_)
+    , max_single_part_upload_size(max_single_part_upload_size_)
+    , temporary_buffer(Aws::MakeShared<Aws::StringStream>("temporary buffer"))
+    , last_part_size(0)
+{ }
 
 void WriteBufferFromS3::nextImpl()
 {
@@ -68,16 +62,17 @@ void WriteBufferFromS3::nextImpl()
 
     ProfileEvents::increment(ProfileEvents::S3WriteBytes, offset());
 
-    if (is_multipart)
-    {
-        last_part_size += offset();
+    last_part_size += offset();
 
-        if (last_part_size > minimum_upload_part_size)
-        {
-            writePart(temporary_buffer->str());
-            last_part_size = 0;
-            temporary_buffer->restart();
-        }
+    /// Data size exceeds singlepart upload threshold, need to use multipart upload.
+    if (multipart_upload_id.empty() && last_part_size > max_single_part_upload_size)
+        createMultipartUpload();
+
+    if (!multipart_upload_id.empty() && last_part_size > minimum_upload_part_size)
+    {
+        writePart();
+        last_part_size = 0;
+        temporary_buffer = Aws::MakeShared<Aws::StringStream>("temporary buffer");
     }
 }
 
@@ -88,17 +83,23 @@ void WriteBufferFromS3::finalize()
 
 void WriteBufferFromS3::finalizeImpl()
 {
-    if (!finalized)
+    if (finalized)
+        return;
+
+    next();
+
+    if (multipart_upload_id.empty())
     {
-        next();
-
-        if (is_multipart)
-            writePart(temporary_buffer->str());
-
-        complete();
-
-        finalized = true;
+        makeSinglepartUpload();
     }
+    else
+    {
+        /// Write rest of the data as last part.
+        writePart();
+        completeMultipartUpload();
+    }
+
+    finalized = true;
 }
 
 WriteBufferFromS3::~WriteBufferFromS3()
@@ -113,7 +114,7 @@ WriteBufferFromS3::~WriteBufferFromS3()
     }
 }
 
-void WriteBufferFromS3::initiate()
+void WriteBufferFromS3::createMultipartUpload()
 {
     Aws::S3::Model::CreateMultipartUploadRequest req;
     req.SetBucket(bucket);
@@ -125,17 +126,17 @@ void WriteBufferFromS3::initiate()
 
     if (outcome.IsSuccess())
     {
-        upload_id = outcome.GetResult().GetUploadId();
-        LOG_DEBUG(log, "Multipart upload initiated. Upload id: {}", upload_id);
+        multipart_upload_id = outcome.GetResult().GetUploadId();
+        LOG_DEBUG(log, "Multipart upload has created. Upload id: {}", multipart_upload_id);
     }
     else
         throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
 
-void WriteBufferFromS3::writePart(const String & data)
+void WriteBufferFromS3::writePart()
 {
-    if (data.empty())
+    if (temporary_buffer->tellp() <= 0)
         return;
 
     if (part_tags.size() == S3_WARN_MAX_PARTS)
@@ -149,93 +150,71 @@ void WriteBufferFromS3::writePart(const String & data)
     req.SetBucket(bucket);
     req.SetKey(key);
     req.SetPartNumber(part_tags.size() + 1);
-    req.SetUploadId(upload_id);
-    req.SetContentLength(data.size());
-    req.SetBody(std::make_shared<Aws::StringStream>(data));
+    req.SetUploadId(multipart_upload_id);
+    req.SetContentLength(temporary_buffer->tellp());
+    req.SetBody(temporary_buffer);
 
     auto outcome = client_ptr->UploadPart(req);
 
-    LOG_TRACE(log, "Writing part. Bucket: {}, Key: {}, Upload_id: {}, Data size: {}", bucket, key, upload_id, data.size());
+    LOG_TRACE(log, "Writing part. Bucket: {}, Key: {}, Upload_id: {}, Data size: {}", bucket, key, multipart_upload_id, temporary_buffer->tellp());
 
     if (outcome.IsSuccess())
     {
         auto etag = outcome.GetResult().GetETag();
         part_tags.push_back(etag);
-        LOG_DEBUG(log, "Writing part finished. Total parts: {}, Upload_id: {}, Etag: {}", part_tags.size(), upload_id, etag);
+        LOG_DEBUG(log, "Writing part finished. Total parts: {}, Upload_id: {}, Etag: {}", part_tags.size(), multipart_upload_id, etag);
     }
     else
         throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
-
-void WriteBufferFromS3::complete()
+void WriteBufferFromS3::completeMultipartUpload()
 {
-    if (is_multipart)
+    LOG_DEBUG(log, "Completing multipart upload. Bucket: {}, Key: {}, Upload_id: {}", bucket, key, multipart_upload_id);
+
+    Aws::S3::Model::CompleteMultipartUploadRequest req;
+    req.SetBucket(bucket);
+    req.SetKey(key);
+    req.SetUploadId(multipart_upload_id);
+
+    Aws::S3::Model::CompletedMultipartUpload multipart_upload;
+    for (size_t i = 0; i < part_tags.size(); ++i)
     {
-        if (part_tags.empty())
-        {
-            LOG_DEBUG(log, "Multipart upload has no data. Aborting it. Bucket: {}, Key: {}, Upload_id: {}", bucket, key, upload_id);
-
-            Aws::S3::Model::AbortMultipartUploadRequest req;
-            req.SetBucket(bucket);
-            req.SetKey(key);
-            req.SetUploadId(upload_id);
-
-            auto outcome = client_ptr->AbortMultipartUpload(req);
-
-            if (outcome.IsSuccess())
-                LOG_DEBUG(log, "Aborting multipart upload completed. Upload_id: {}", upload_id);
-            else
-                throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
-
-            return;
-        }
-
-        LOG_DEBUG(log, "Completing multipart upload. Bucket: {}, Key: {}, Upload_id: {}", bucket, key, upload_id);
-
-        Aws::S3::Model::CompleteMultipartUploadRequest req;
-        req.SetBucket(bucket);
-        req.SetKey(key);
-        req.SetUploadId(upload_id);
-
-        Aws::S3::Model::CompletedMultipartUpload multipart_upload;
-        for (size_t i = 0; i < part_tags.size(); ++i)
-        {
-            Aws::S3::Model::CompletedPart part;
-            multipart_upload.AddParts(part.WithETag(part_tags[i]).WithPartNumber(i + 1));
-        }
-
-        req.SetMultipartUpload(multipart_upload);
-
-        auto outcome = client_ptr->CompleteMultipartUpload(req);
-
-        if (outcome.IsSuccess())
-            LOG_DEBUG(log, "Multipart upload completed. Upload_id: {}", upload_id);
-        else
-            throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
+        Aws::S3::Model::CompletedPart part;
+        multipart_upload.AddParts(part.WithETag(part_tags[i]).WithPartNumber(i + 1));
     }
+
+    req.SetMultipartUpload(multipart_upload);
+
+    auto outcome = client_ptr->CompleteMultipartUpload(req);
+
+    if (outcome.IsSuccess())
+        LOG_DEBUG(log, "Multipart upload has completed. Upload_id: {}", multipart_upload_id);
     else
-    {
-        LOG_DEBUG(log, "Making single part upload. Bucket: {}, Key: {}", bucket, key);
+        throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
+}
 
-        Aws::S3::Model::PutObjectRequest req;
-        req.SetBucket(bucket);
-        req.SetKey(key);
-        if (object_metadata.has_value())
-            req.SetMetadata(object_metadata.value());
+void WriteBufferFromS3::makeSinglepartUpload()
+{
+    if (temporary_buffer->tellp() <= 0)
+        return;
 
-        /// This could be improved using an adapter to WriteBuffer.
-        const std::shared_ptr<Aws::IOStream> input_data = Aws::MakeShared<Aws::StringStream>("temporary buffer", temporary_buffer->str());
-        temporary_buffer = std::make_unique<WriteBufferFromOwnString>();
-        req.SetBody(input_data);
+    LOG_DEBUG(log, "Making single part upload. Bucket: {}, Key: {}", bucket, key);
 
-        auto outcome = client_ptr->PutObject(req);
+    Aws::S3::Model::PutObjectRequest req;
+    req.SetBucket(bucket);
+    req.SetKey(key);
+    req.SetContentLength(temporary_buffer->tellp());
+    req.SetBody(temporary_buffer);
+    if (object_metadata.has_value())
+        req.SetMetadata(object_metadata.value());
 
-        if (outcome.IsSuccess())
-            LOG_DEBUG(log, "Single part upload has completed. Bucket: {}, Key: {}", bucket, key);
-        else
-            throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
-    }
+    auto outcome = client_ptr->PutObject(req);
+
+    if (outcome.IsSuccess())
+        LOG_DEBUG(log, "Single part upload has completed. Bucket: {}, Key: {}", bucket, key);
+    else
+        throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
 }
diff --git a/src/IO/WriteBufferFromS3.h b/src/IO/WriteBufferFromS3.h
index 9bd7b896d7f..9e4b056603a 100644
--- a/src/IO/WriteBufferFromS3.h
+++ b/src/IO/WriteBufferFromS3.h
@@ -6,11 +6,13 @@
 
 #    include <memory>
 #    include <vector>
+#    include <common/logger_useful.h>
 #    include <common/types.h>
+
 #    include <IO/BufferWithOwnMemory.h>
-#    include <IO/HTTPCommon.h>
 #    include <IO/WriteBuffer.h>
-#    include <IO/WriteBufferFromString.h>
+
+#    include <aws/core/utils/memory/stl/AWSStringStream.h>
 
 namespace Aws::S3
 {
@@ -19,24 +21,30 @@ class S3Client;
 
 namespace DB
 {
-/* Perform S3 HTTP PUT request.
+
+/**
+ * Buffer to write a data to a S3 object with specified bucket and key.
+ * If data size written to the buffer is less than 'max_single_part_upload_size' write is performed using singlepart upload.
+ * In another case multipart upload is used:
+ * Data is divided on chunks with size greater than 'minimum_upload_part_size'. Last chunk can be less than this threshold.
+ * Each chunk is written as a part to S3.
  */
 class WriteBufferFromS3 : public BufferWithOwnMemory<WriteBuffer>
 {
 private:
-    bool is_multipart;
-
     String bucket;
     String key;
     std::optional<std::map<String, String>> object_metadata;
     std::shared_ptr<Aws::S3::S3Client> client_ptr;
     size_t minimum_upload_part_size;
-    std::unique_ptr<WriteBufferFromOwnString> temporary_buffer;
+    size_t max_single_part_upload_size;
+    /// Buffer to accumulate data.
+    std::shared_ptr<Aws::StringStream> temporary_buffer;
     size_t last_part_size;
 
     /// Upload in S3 is made in parts.
     /// We initiate upload, then upload each part and get ETag as a response, and then finish upload with listing all our parts.
-    String upload_id;
+    String multipart_upload_id;
     std::vector<String> part_tags;
 
     Poco::Logger * log = &Poco::Logger::get("WriteBufferFromS3");
@@ -47,7 +55,7 @@ public:
         const String & bucket_,
         const String & key_,
         size_t minimum_upload_part_size_,
-        bool is_multipart,
+        size_t max_single_part_upload_size_,
         std::optional<std::map<String, String>> object_metadata_ = std::nullopt,
         size_t buffer_size_ = DBMS_DEFAULT_BUFFER_SIZE);
 
@@ -61,9 +69,11 @@ public:
 private:
     bool finalized = false;
 
-    void initiate();
-    void writePart(const String & data);
-    void complete();
+    void createMultipartUpload();
+    void writePart();
+    void completeMultipartUpload();
+
+    void makeSinglepartUpload();
 
     void finalizeImpl();
 };
diff --git a/src/IO/WriteHelpers.h b/src/IO/WriteHelpers.h
index e5c6f5b4ab8..f89422f9715 100644
--- a/src/IO/WriteHelpers.h
+++ b/src/IO/WriteHelpers.h
@@ -29,11 +29,14 @@
 #include <IO/DoubleConverter.h>
 #include <IO/WriteBufferFromString.h>
 
-/// There is no dragonbox in Arcadia
-#if !defined(ARCADIA_BUILD)
-#   include <dragonbox/dragonbox_to_chars.h>
-#else
-#   include <ryu/ryu.h>
+#ifdef __clang__
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunused-parameter"
+#pragma clang diagnostic ignored "-Wsign-compare"
+#endif
+#include <dragonbox/dragonbox_to_chars.h>
+#ifdef __clang__
+#pragma clang diagnostic pop
 #endif
 
 #include <Formats/FormatSettings.h>
@@ -233,22 +236,14 @@ inline size_t writeFloatTextFastPath(T x, char * buffer)
         if (DecomposedFloat64(x).is_inside_int64())
             result = itoa(Int64(x), buffer) - buffer;
         else
-#if !defined(ARCADIA_BUILD)
             result = jkj::dragonbox::to_chars_n(x, buffer) - buffer;
-#else
-            result = d2s_buffered_n(x, buffer);
-#endif
     }
     else
     {
         if (DecomposedFloat32(x).is_inside_int32())
             result = itoa(Int32(x), buffer) - buffer;
         else
-#if !defined(ARCADIA_BUILD)
             result = jkj::dragonbox::to_chars_n(x, buffer) - buffer;
-#else
-            result = f2s_buffered_n(x, buffer);
-#endif
     }
 
     if (result <= 0)
diff --git a/src/IO/readFloatText.h b/src/IO/readFloatText.h
index a721642a185..5647d574c62 100644
--- a/src/IO/readFloatText.h
+++ b/src/IO/readFloatText.h
@@ -5,8 +5,15 @@
 #include <common/shift10.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <double-conversion/double-conversion.h>
-#include <fast_float/fast_float.h>
 
+#ifdef __clang__
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunneeded-internal-declaration"
+#endif
+#include <fast_float/fast_float.h>
+#ifdef __clang__
+#pragma clang diagnostic pop
+#endif
 
 /** Methods for reading floating point numbers from text with decimal representation.
   * There are "precise", "fast" and "simple" implementations.
@@ -249,6 +256,19 @@ ReturnType readFloatTextPreciseImpl(T & x, ReadBuffer & buf)
 }
 
 
+// credit: https://johnnylee-sde.github.io/Fast-numeric-string-to-int/
+static inline bool is_made_of_eight_digits_fast(uint64_t val) noexcept
+{
+    return (((val & 0xF0F0F0F0F0F0F0F0) | (((val + 0x0606060606060606) & 0xF0F0F0F0F0F0F0F0) >> 4)) == 0x3333333333333333);
+}
+
+static inline bool is_made_of_eight_digits_fast(const char * chars) noexcept
+{
+    uint64_t val;
+    ::memcpy(&val, chars, 8);
+    return is_made_of_eight_digits_fast(val);
+}
+
 template <size_t N, typename T>
 static inline void readUIntTextUpToNSignificantDigits(T & x, ReadBuffer & buf)
 {
@@ -266,9 +286,6 @@ static inline void readUIntTextUpToNSignificantDigits(T & x, ReadBuffer & buf)
             else
                 return;
         }
-
-        while (!buf.eof() && isNumericASCII(*buf.position()))
-            ++buf.position();
     }
     else
     {
@@ -283,10 +300,16 @@ static inline void readUIntTextUpToNSignificantDigits(T & x, ReadBuffer & buf)
             else
                 return;
         }
-
-        while (!buf.eof() && isNumericASCII(*buf.position()))
-            ++buf.position();
     }
+
+    while (!buf.eof() && (buf.position() + 8 <= buf.buffer().end()) &&
+         is_made_of_eight_digits_fast(buf.position()))
+    {
+        buf.position() += 8;
+    }
+
+    while (!buf.eof() && isNumericASCII(*buf.position()))
+        ++buf.position();
 }
 
 
@@ -319,7 +342,6 @@ ReturnType readFloatTextFastImpl(T & x, ReadBuffer & in)
         ++in.position();
     }
 
-
     auto count_after_sign = in.count();
 
     constexpr int significant_digits = std::numeric_limits<UInt64>::digits10;
diff --git a/src/IO/tests/o_direct_and_dirty_pages.cpp b/src/IO/tests/o_direct_and_dirty_pages.cpp
index 327d6357e9b..c60141d286f 100644
--- a/src/IO/tests/o_direct_and_dirty_pages.cpp
+++ b/src/IO/tests/o_direct_and_dirty_pages.cpp
@@ -3,6 +3,7 @@
 #include <iostream>
 
 #include <common/types.h>
+#include <common/getPageSize.h>
 #include <IO/WriteHelpers.h>
 #include <IO/ReadHelpers.h>
 #include <IO/WriteBufferFromFile.h>
@@ -16,6 +17,7 @@ int main(int, char **)
     {
         static const size_t N = 100000;
         static const size_t BUF_SIZE = 1048576;
+        size_t page_size = static_cast<size_t>(::getPageSize());
 
         ReadBufferFromFile rand_in("/dev/urandom");
         unsigned rand = 0;
@@ -33,7 +35,7 @@ int main(int, char **)
         }
 
         {
-            ReadBufferFromFile rb("test1", BUF_SIZE, O_RDONLY | O_DIRECT, nullptr, 4096);
+            ReadBufferFromFile rb("test1", BUF_SIZE, O_RDONLY | O_DIRECT, nullptr, page_size);
             String res;
             for (size_t i = 0; i < N; ++i)
                 readStringBinary(res, rb);
@@ -44,14 +46,14 @@ int main(int, char **)
         /// Write to file with O_DIRECT, read as usual.
 
         {
-            WriteBufferFromFile wb("test2", BUF_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, 4096);
+            WriteBufferFromFile wb("test2", BUF_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, page_size);
 
             for (size_t i = 0; i < N; ++i)
                 writeStringBinary(test, wb);
 
-            if (wb.offset() % 4096 != 0)
+            if (wb.offset() % page_size != 0)
             {
-                size_t pad = 4096 - wb.offset() % 4096;
+                size_t pad = page_size - wb.offset() % page_size;
                 memset(wb.position(), 0, pad);
                 wb.position() += pad;
             }
diff --git a/src/IO/ya.make b/src/IO/ya.make
index c2913587bb6..4dc3afb2f11 100644
--- a/src/IO/ya.make
+++ b/src/IO/ya.make
@@ -5,6 +5,7 @@ LIBRARY()
 
 ADDINCL(
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 PEERDIR(
@@ -13,6 +14,7 @@ PEERDIR(
     contrib/libs/brotli/enc
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 
diff --git a/src/IO/ya.make.in b/src/IO/ya.make.in
index 9b824228b92..252e102a819 100644
--- a/src/IO/ya.make.in
+++ b/src/IO/ya.make.in
@@ -4,6 +4,7 @@ LIBRARY()
 
 ADDINCL(
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 PEERDIR(
@@ -12,6 +13,7 @@ PEERDIR(
     contrib/libs/brotli/enc
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 
diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index 75915bdad1a..62cb6047704 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -662,10 +662,10 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
             auto * right_arg = const_cast<Node *>(&actions_dag->addColumn(std::move(column), true));
             auto * left_arg = src_node;
 
-            CastOverloadResolver::Diagnostic diagnostic = {src_node->result_name, res_elem.name};
+            FunctionCast::Diagnostic diagnostic = {src_node->result_name, res_elem.name};
             FunctionOverloadResolverPtr func_builder_cast =
                     std::make_shared<FunctionOverloadResolverAdaptor>(
-                            CastOverloadResolver::createImpl(false, std::move(diagnostic)));
+                            CastOverloadResolver<CastType::nonAccurate>::createImpl(false, std::move(diagnostic)));
 
             Inputs children = { left_arg, right_arg };
             src_node = &actions_dag->addFunction(func_builder_cast, std::move(children), {}, true);
diff --git a/src/Interpreters/Context.cpp b/src/Interpreters/Context.cpp
index 9955c1ca11f..337753ba0b8 100644
--- a/src/Interpreters/Context.cpp
+++ b/src/Interpreters/Context.cpp
@@ -445,6 +445,8 @@ struct ContextShared
 
         /// Stop trace collector if any
         trace_collector.reset();
+        /// Stop zookeeper connection
+        zookeeper.reset();
         /// Stop test_keeper storage
         test_keeper_storage.reset();
     }
diff --git a/src/Interpreters/DatabaseCatalog.cpp b/src/Interpreters/DatabaseCatalog.cpp
index ea1984c96fb..e9caaab4ef9 100644
--- a/src/Interpreters/DatabaseCatalog.cpp
+++ b/src/Interpreters/DatabaseCatalog.cpp
@@ -4,7 +4,7 @@
 #include <Storages/IStorage.h>
 #include <Databases/IDatabase.h>
 #include <Databases/DatabaseMemory.h>
-#include <Databases/DatabaseAtomic.h>
+#include <Databases/DatabaseOnDisk.h>
 #include <Poco/File.h>
 #include <Common/quoteString.h>
 #include <Storages/StorageMemory.h>
@@ -16,6 +16,15 @@
 #include <Common/CurrentMetrics.h>
 #include <common/logger_useful.h>
 
+#if !defined(ARCADIA_BUILD)
+#    include "config_core.h"
+#endif
+
+#if USE_MYSQL
+#    include <Databases/MySQL/MaterializeMySQLSyncThread.h>
+#    include <Storages/StorageMaterializeMySQL.h>
+#endif
+
 #include <filesystem>
 
 
@@ -202,7 +211,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
     if (!table_id)
     {
         if (exception)
-            exception->emplace("Cannot find table: StorageID is empty", ErrorCodes::UNKNOWN_TABLE);
+            exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Cannot find table: StorageID is empty");
         return {};
     }
 
@@ -214,9 +223,18 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         {
             assert(!db_and_table.first && !db_and_table.second);
             if (exception)
-                exception->emplace("Table " + table_id.getNameForLogs() + " doesn't exist.", ErrorCodes::UNKNOWN_TABLE);
+                exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Table {} doesn't exist", table_id.getNameForLogs());
             return {};
         }
+
+#if USE_MYSQL
+        /// It's definitely not the best place for this logic, but behaviour must be consistent with DatabaseMaterializeMySQL::tryGetTable(...)
+        if (db_and_table.first->getEngineName() == "MaterializeMySQL")
+        {
+            if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
+                db_and_table.second = std::make_shared<StorageMaterializeMySQL>(std::move(db_and_table.second), db_and_table.first.get());
+        }
+#endif
         return db_and_table;
     }
 
@@ -226,7 +244,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         /// If table_id has no UUID, then the name of database was specified by user and table_id was not resolved through context.
         /// Do not allow access to TEMPORARY_DATABASE because it contains all temporary tables of all contexts and users.
         if (exception)
-            exception->emplace("Direct access to `" + String(TEMPORARY_DATABASE) + "` database is not allowed.", ErrorCodes::DATABASE_ACCESS_DENIED);
+            exception->emplace(ErrorCodes::DATABASE_ACCESS_DENIED, "Direct access to `{}` database is not allowed", String(TEMPORARY_DATABASE));
         return {};
     }
 
@@ -237,8 +255,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         if (databases.end() == it)
         {
             if (exception)
-                exception->emplace("Database " + backQuoteIfNeed(table_id.getDatabaseName()) + " doesn't exist",
-                                   ErrorCodes::UNKNOWN_DATABASE);
+                exception->emplace(ErrorCodes::UNKNOWN_DATABASE, "Database {} doesn't exist", backQuoteIfNeed(table_id.getDatabaseName()));
             return {};
         }
         database = it->second;
@@ -246,7 +263,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
 
     auto table = database->tryGetTable(table_id.table_name, context);
     if (!table && exception)
-            exception->emplace("Table " + table_id.getNameForLogs() + " doesn't exist.", ErrorCodes::UNKNOWN_TABLE);
+            exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Table {} doesn't exist", table_id.getNameForLogs());
     if (!table)
         database = nullptr;
 
@@ -286,7 +303,6 @@ void DatabaseCatalog::attachDatabase(const String & database_name, const Databas
     assertDatabaseDoesntExistUnlocked(database_name);
     databases.emplace(database_name, database);
     UUID db_uuid = database->getUUID();
-    assert((db_uuid != UUIDHelpers::Nil) ^ (dynamic_cast<DatabaseAtomic *>(database.get()) == nullptr));
     if (db_uuid != UUIDHelpers::Nil)
         db_uuid_map.emplace(db_uuid, database);
 }
@@ -313,9 +329,8 @@ DatabasePtr DatabaseCatalog::detachDatabase(const String & database_name, bool d
             if (!db->empty())
                 throw Exception("New table appeared in database being dropped or detached. Try again.",
                                 ErrorCodes::DATABASE_NOT_EMPTY);
-            auto * database_atomic = typeid_cast<DatabaseAtomic *>(db.get());
-            if (!drop && database_atomic)
-                database_atomic->assertCanBeDetached(false);
+            if (!drop)
+                db->assertCanBeDetached(false);
         }
         catch (...)
         {
diff --git a/src/Interpreters/IInterpreterUnionOrSelectQuery.h b/src/Interpreters/IInterpreterUnionOrSelectQuery.h
index 67e3cf25e5a..aefa78260c7 100644
--- a/src/Interpreters/IInterpreterUnionOrSelectQuery.h
+++ b/src/Interpreters/IInterpreterUnionOrSelectQuery.h
@@ -34,6 +34,8 @@ protected:
     Block result_header;
     SelectQueryOptions options;
     size_t max_streams = 1;
+    bool settings_limit_offset_needed = false;
+    bool settings_limit_offset_done = false;
 };
 }
 
diff --git a/src/Interpreters/InterpreterCreateQuery.cpp b/src/Interpreters/InterpreterCreateQuery.cpp
index d57a14a61dc..2446075d38a 100644
--- a/src/Interpreters/InterpreterCreateQuery.cpp
+++ b/src/Interpreters/InterpreterCreateQuery.cpp
@@ -157,6 +157,35 @@ BlockIO InterpreterCreateQuery::createDatabase(ASTCreateQuery & create)
         if (!create.attach && fs::exists(metadata_path))
             throw Exception(ErrorCodes::DATABASE_ALREADY_EXISTS, "Metadata directory {} already exists", metadata_path.string());
     }
+    else if (create.storage->engine->name == "MaterializeMySQL")
+    {
+        /// It creates nested database with Ordinary or Atomic engine depending on UUID in query and default engine setting.
+        /// Do nothing if it's an internal ATTACH on server startup or short-syntax ATTACH query from user,
+        /// because we got correct query from the metadata file in this case.
+        /// If we got query from user, then normalize it first.
+        bool attach_from_user = create.attach && !internal && !create.attach_short_syntax;
+        bool create_from_user = !create.attach;
+
+        if (create_from_user)
+        {
+            const auto & default_engine = context.getSettingsRef().default_database_engine.value;
+            if (create.uuid == UUIDHelpers::Nil && default_engine == DefaultDatabaseEngine::Atomic)
+                create.uuid = UUIDHelpers::generateV4();    /// Will enable Atomic engine for nested database
+        }
+        else if (attach_from_user && create.uuid == UUIDHelpers::Nil)
+        {
+            /// Ambiguity is possible: should we attach nested database as Ordinary
+            /// or throw "UUID must be specified" for Atomic? So we suggest short syntax for Ordinary.
+            throw Exception("Use short attach syntax ('ATTACH DATABASE name;' without engine) to attach existing database "
+                            "or specify UUID to attach new database with Atomic engine", ErrorCodes::INCORRECT_QUERY);
+        }
+
+        /// Set metadata path according to nested engine
+        if (create.uuid == UUIDHelpers::Nil)
+            metadata_path = metadata_path / "metadata" / database_name_escaped;
+        else
+            metadata_path = metadata_path / "store" / DatabaseCatalog::getPathForUUID(create.uuid);
+    }
     else
     {
         bool is_on_cluster = context.getClientInfo().query_kind == ClientInfo::QueryKind::SECONDARY_QUERY;
@@ -586,6 +615,23 @@ void InterpreterCreateQuery::validateTableStructure(const ASTCreateQuery & creat
             }
         }
     }
+
+    if (!create.attach && !settings.allow_experimental_map_type)
+    {
+        for (const auto & name_and_type_pair : properties.columns.getAllPhysical())
+        {
+            WhichDataType which(*name_and_type_pair.type);
+            if (which.isMap())
+            {
+                const auto & type_name = name_and_type_pair.type->getName();
+                String message = "Cannot create table with column '" + name_and_type_pair.name + "' which type is '"
+                                 + type_name + "' because experimental Map type is not allowed. "
+                                 + "Set 'allow_experimental_map_type = 1' setting to enable";
+                throw Exception(message, ErrorCodes::ILLEGAL_COLUMN);
+            }
+        }
+
+    }
 }
 
 void InterpreterCreateQuery::setEngine(ASTCreateQuery & create) const
@@ -655,7 +701,7 @@ void InterpreterCreateQuery::assertOrSetUUID(ASTCreateQuery & create, const Data
 
     bool from_path = create.attach_from_path.has_value();
 
-    if (database->getEngineName() == "Atomic")
+    if (database->getUUID() != UUIDHelpers::Nil)
     {
         if (create.attach && !from_path && create.uuid == UUIDHelpers::Nil)
         {
@@ -705,7 +751,7 @@ BlockIO InterpreterCreateQuery::createTable(ASTCreateQuery & create)
         auto database = DatabaseCatalog::instance().getDatabase(database_name);
         bool if_not_exists = create.if_not_exists;
 
-        // Table SQL definition is available even if the table is detached
+        // Table SQL definition is available even if the table is detached (even permanently)
         auto query = database->getCreateTableQuery(create.table, context);
         create = query->as<ASTCreateQuery &>(); // Copy the saved create query, but use ATTACH instead of CREATE
         if (create.is_dictionary)
diff --git a/src/Interpreters/InterpreterDropQuery.cpp b/src/Interpreters/InterpreterDropQuery.cpp
index 5f7f70fdda4..47bafdbdcb4 100644
--- a/src/Interpreters/InterpreterDropQuery.cpp
+++ b/src/Interpreters/InterpreterDropQuery.cpp
@@ -11,7 +11,14 @@
 #include <Common/escapeForFileName.h>
 #include <Common/quoteString.h>
 #include <Common/typeid_cast.h>
-#include <Databases/DatabaseAtomic.h>
+
+#if !defined(ARCADIA_BUILD)
+#    include "config_core.h"
+#endif
+
+#if USE_MYSQL
+#   include <Databases/MySQL/DatabaseMaterializeMySQL.h>
+#endif
 
 
 namespace DB
@@ -23,6 +30,7 @@ namespace ErrorCodes
     extern const int SYNTAX_ERROR;
     extern const int UNKNOWN_TABLE;
     extern const int UNKNOWN_DICTIONARY;
+    extern const int NOT_IMPLEMENTED;
 }
 
 
@@ -48,6 +56,8 @@ BlockIO InterpreterDropQuery::execute()
     {
         if (!drop.is_dictionary)
             return executeToTable(drop);
+        else if (drop.permanently && drop.kind == ASTDropQuery::Kind::Detach)
+            throw Exception("DETACH PERMANENTLY is not implemented for dictionaries", ErrorCodes::NOT_IMPLEMENTED);
         else
             return executeToDictionary(drop.database, drop.table, drop.kind, drop.if_exists, drop.temporary, drop.no_ddl_lock);
     }
@@ -66,10 +76,7 @@ void InterpreterDropQuery::waitForTableToBeActuallyDroppedOrDetached(const ASTDr
     if (query.kind == ASTDropQuery::Kind::Drop)
         DatabaseCatalog::instance().waitTableFinallyDropped(uuid_to_wait);
     else if (query.kind == ASTDropQuery::Kind::Detach)
-    {
-        if (auto * atomic = typeid_cast<DatabaseAtomic *>(db.get()))
-            atomic->waitDetachedTableNotInUse(uuid_to_wait);
-    }
+        db->waitDetachedTableNotInUse(uuid_to_wait);
 }
 
 BlockIO InterpreterDropQuery::executeToTable(const ASTDropQuery & query)
@@ -122,10 +129,20 @@ BlockIO InterpreterDropQuery::executeToTableImpl(const ASTDropQuery & query, Dat
             table->checkTableCanBeDetached();
             table->shutdown();
             TableExclusiveLockHolder table_lock;
-            if (database->getEngineName() != "Atomic")
+            if (database->getUUID() == UUIDHelpers::Nil)
                 table_lock = table->lockExclusively(context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
-            /// Drop table from memory, don't touch data and metadata
-            database->detachTable(table_id.table_name);
+
+            if (query.permanently)
+            {
+                /// Drop table from memory, don't touch data, metadata file renamed and will be skipped during server restart
+                database->detachTablePermanently(table_id.table_name);
+            }
+            else
+            {
+                /// Drop table from memory, don't touch data and metadata
+                database->detachTable(table_id.table_name);
+            }
+
         }
         else if (query.kind == ASTDropQuery::Kind::Truncate)
         {
@@ -145,7 +162,7 @@ BlockIO InterpreterDropQuery::executeToTableImpl(const ASTDropQuery & query, Dat
             table->shutdown();
 
             TableExclusiveLockHolder table_lock;
-            if (database->getEngineName() != "Atomic")
+            if (database->getUUID() == UUIDHelpers::Nil)
                 table_lock = table->lockExclusively(context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
 
             database->dropTable(context, table_id.table_name, query.no_delay);
@@ -282,6 +299,14 @@ BlockIO InterpreterDropQuery::executeToDatabaseImpl(const ASTDropQuery & query,
             bool drop = query.kind == ASTDropQuery::Kind::Drop;
             context.checkAccess(AccessType::DROP_DATABASE, database_name);
 
+            if (query.kind == ASTDropQuery::Kind::Detach && query.permanently)
+                throw Exception("DETACH PERMANENTLY is not implemented for databases", ErrorCodes::NOT_IMPLEMENTED);
+
+#if USE_MYSQL
+            if (database->getEngineName() == "MaterializeMySQL")
+                stopDatabaseSynchronization(database);
+#endif
+
             if (database->shouldBeEmptyOnDetach())
             {
                 /// DETACH or DROP all tables and dictionaries inside database.
@@ -312,9 +337,8 @@ BlockIO InterpreterDropQuery::executeToDatabaseImpl(const ASTDropQuery & query,
             /// Protects from concurrent CREATE TABLE queries
             auto db_guard = DatabaseCatalog::instance().getExclusiveDDLGuardForDatabase(database_name);
 
-            auto * database_atomic = typeid_cast<DatabaseAtomic *>(database.get());
-            if (!drop && database_atomic)
-                database_atomic->assertCanBeDetached(true);
+            if (!drop)
+                database->assertCanBeDetached(true);
 
             /// DETACH or DROP database itself
             DatabaseCatalog::instance().detachDatabase(database_name, drop, database->shouldBeEmptyOnDetach());
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index ee07a2cde87..aee4eae8782 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1552,7 +1552,12 @@ void InterpreterSelectQuery::executeFetchColumns(
         throw Exception("Logical error in InterpreterSelectQuery: nowhere to read", ErrorCodes::LOGICAL_ERROR);
 
     /// Specify the number of threads only if it wasn't specified in storage.
-    if (!query_plan.getMaxThreads())
+    ///
+    /// But in case of remote query and prefer_localhost_replica=1 (default)
+    /// The inner local query (that is done in the same process, without
+    /// network interaction), it will setMaxThreads earlier and distributed
+    /// query will not update it.
+    if (!query_plan.getMaxThreads() || is_remote)
         query_plan.setMaxThreads(max_threads_execute_query);
 
     /// Aliases in table declaration.
diff --git a/src/Interpreters/InterpreterSelectWithUnionQuery.cpp b/src/Interpreters/InterpreterSelectWithUnionQuery.cpp
index a836dc8b271..e6610df43ff 100644
--- a/src/Interpreters/InterpreterSelectWithUnionQuery.cpp
+++ b/src/Interpreters/InterpreterSelectWithUnionQuery.cpp
@@ -9,10 +9,14 @@
 #include <Processors/QueryPlan/IQueryPlanStep.h>
 #include <Processors/QueryPlan/QueryPlan.h>
 #include <Processors/QueryPlan/UnionStep.h>
+#include <Processors/QueryPlan/LimitStep.h>
+#include <Processors/QueryPlan/OffsetStep.h>
 #include <Common/typeid_cast.h>
 
 #include <Interpreters/InDepthNodeVisitor.h>
 
+#include <algorithm>
+
 namespace DB
 {
 
@@ -128,24 +132,28 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
     const ASTPtr & query_ptr_, const Context & context_, const SelectQueryOptions & options_, const Names & required_result_column_names)
     : IInterpreterUnionOrSelectQuery(query_ptr_, context_, options_)
 {
-    auto & ast = query_ptr->as<ASTSelectWithUnionQuery &>();
+    ASTSelectWithUnionQuery * ast = query_ptr->as<ASTSelectWithUnionQuery>();
+
+    const Settings & settings = context->getSettingsRef();
+    if (options.subquery_depth == 0 && (settings.limit > 0 || settings.offset > 0))
+        settings_limit_offset_needed = true;
 
     /// Normalize AST Tree
-    if (!ast.is_normalized)
+    if (!ast->is_normalized)
     {
-        CustomizeASTSelectWithUnionQueryNormalizeVisitor::Data union_default_mode{context->getSettingsRef().union_default_mode};
+        CustomizeASTSelectWithUnionQueryNormalizeVisitor::Data union_default_mode{settings.union_default_mode};
         CustomizeASTSelectWithUnionQueryNormalizeVisitor(union_default_mode).visit(query_ptr);
 
         /// After normalization, if it only has one ASTSelectWithUnionQuery child,
         /// we can lift it up, this can reduce one unnecessary recursion later.
-        if (ast.list_of_selects->children.size() == 1 && ast.list_of_selects->children.at(0)->as<ASTSelectWithUnionQuery>())
+        if (ast->list_of_selects->children.size() == 1 && ast->list_of_selects->children.at(0)->as<ASTSelectWithUnionQuery>())
         {
-            query_ptr = std::move(ast.list_of_selects->children.at(0));
-            ast = query_ptr->as<ASTSelectWithUnionQuery &>();
+            query_ptr = std::move(ast->list_of_selects->children.at(0));
+            ast = query_ptr->as<ASTSelectWithUnionQuery>();
         }
     }
 
-    size_t num_children = ast.list_of_selects->children.size();
+    size_t num_children = ast->list_of_selects->children.size();
     if (!num_children)
         throw Exception("Logical error: no children in ASTSelectWithUnionQuery", ErrorCodes::LOGICAL_ERROR);
 
@@ -161,7 +169,7 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
         /// Result header if there are no filtering by 'required_result_column_names'.
         /// We use it to determine positions of 'required_result_column_names' in SELECT clause.
 
-        Block full_result_header = getCurrentChildResultHeader(ast.list_of_selects->children.at(0), required_result_column_names);
+        Block full_result_header = getCurrentChildResultHeader(ast->list_of_selects->children.at(0), required_result_column_names);
 
         std::vector<size_t> positions_of_required_result_columns(required_result_column_names.size());
 
@@ -171,7 +179,7 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
         for (size_t query_num = 1; query_num < num_children; ++query_num)
         {
             Block full_result_header_for_current_select
-                = getCurrentChildResultHeader(ast.list_of_selects->children.at(query_num), required_result_column_names);
+                = getCurrentChildResultHeader(ast->list_of_selects->children.at(query_num), required_result_column_names);
 
             if (full_result_header_for_current_select.columns() != full_result_header.columns())
                 throw Exception("Different number of columns in UNION ALL elements:\n"
@@ -186,13 +194,59 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
         }
     }
 
+    if (num_children == 1 && settings_limit_offset_needed)
+    {
+        const ASTPtr first_select_ast = ast->list_of_selects->children.at(0);
+        ASTSelectQuery * select_query = first_select_ast->as<ASTSelectQuery>();
+
+        if (!select_query->withFill() && !select_query->limit_with_ties)
+        {
+            UInt64 limit_length = 0;
+            UInt64 limit_offset = 0;
+
+            const ASTPtr limit_offset_ast = select_query->limitOffset();
+            if (limit_offset_ast)
+            {
+                limit_offset = limit_offset_ast->as<ASTLiteral &>().value.safeGet<UInt64>();
+                UInt64 new_limit_offset = settings.offset + limit_offset;
+                limit_offset_ast->as<ASTLiteral &>().value = Field(new_limit_offset);
+            }
+            else if (settings.offset)
+            {
+                ASTPtr new_limit_offset_ast = std::make_shared<ASTLiteral>(Field(UInt64(settings.offset)));
+                select_query->setExpression(ASTSelectQuery::Expression::LIMIT_OFFSET, std::move(new_limit_offset_ast));
+            }
+
+            const ASTPtr limit_length_ast = select_query->limitLength();
+            if (limit_length_ast)
+            {
+                limit_length = limit_length_ast->as<ASTLiteral &>().value.safeGet<UInt64>();
+
+                UInt64 new_limit_length = 0;
+                if (settings.offset == 0)
+                    new_limit_length = std::min(limit_length, UInt64(settings.limit));
+                else if (settings.offset < limit_length)
+                    new_limit_length =  settings.limit ? std::min(UInt64(settings.limit), limit_length - settings.offset) : (limit_length - settings.offset);
+
+                limit_length_ast->as<ASTLiteral &>().value = Field(new_limit_length);
+            }
+            else if (settings.limit)
+            {
+                ASTPtr new_limit_length_ast = std::make_shared<ASTLiteral>(Field(UInt64(settings.limit)));
+                select_query->setExpression(ASTSelectQuery::Expression::LIMIT_LENGTH, std::move(new_limit_length_ast));
+            }
+
+            settings_limit_offset_done = true;
+        }
+    }
+
     for (size_t query_num = 0; query_num < num_children; ++query_num)
     {
         const Names & current_required_result_column_names
             = query_num == 0 ? required_result_column_names : required_result_column_names_for_other_selects[query_num];
 
         nested_interpreters.emplace_back(
-            buildCurrentChildInterpreter(ast.list_of_selects->children.at(query_num), current_required_result_column_names));
+            buildCurrentChildInterpreter(ast->list_of_selects->children.at(query_num), current_required_result_column_names));
     }
 
     /// Determine structure of the result.
@@ -293,39 +347,57 @@ void InterpreterSelectWithUnionQuery::buildQueryPlan(QueryPlan & query_plan)
 {
     // auto num_distinct_union = optimizeUnionList();
     size_t num_plans = nested_interpreters.size();
+    const Settings & settings = context->getSettingsRef();
 
     /// Skip union for single interpreter.
     if (num_plans == 1)
     {
         nested_interpreters.front()->buildQueryPlan(query_plan);
-        return;
+    }
+    else
+    {
+        std::vector<std::unique_ptr<QueryPlan>> plans(num_plans);
+        DataStreams data_streams(num_plans);
+
+        for (size_t i = 0; i < num_plans; ++i)
+        {
+            plans[i] = std::make_unique<QueryPlan>();
+            nested_interpreters[i]->buildQueryPlan(*plans[i]);
+            data_streams[i] = plans[i]->getCurrentDataStream();
+        }
+
+        auto max_threads = context->getSettingsRef().max_threads;
+        auto union_step = std::make_unique<UnionStep>(std::move(data_streams), result_header, max_threads);
+
+        query_plan.unitePlans(std::move(union_step), std::move(plans));
+
+        const auto & query = query_ptr->as<ASTSelectWithUnionQuery &>();
+        if (query.union_mode == ASTSelectWithUnionQuery::Mode::DISTINCT)
+        {
+            /// Add distinct transform
+            SizeLimits limits(settings.max_rows_in_distinct, settings.max_bytes_in_distinct, settings.distinct_overflow_mode);
+
+            auto distinct_step
+                = std::make_unique<DistinctStep>(query_plan.getCurrentDataStream(), limits, 0, result_header.getNames(), false);
+
+            query_plan.addStep(std::move(distinct_step));
+        }
     }
 
-    std::vector<std::unique_ptr<QueryPlan>> plans(num_plans);
-    DataStreams data_streams(num_plans);
-
-    for (size_t i = 0; i < num_plans; ++i)
+    if (settings_limit_offset_needed && !settings_limit_offset_done)
     {
-        plans[i] = std::make_unique<QueryPlan>();
-        nested_interpreters[i]->buildQueryPlan(*plans[i]);
-        data_streams[i] = plans[i]->getCurrentDataStream();
-    }
-
-    auto max_threads = context->getSettingsRef().max_threads;
-    auto union_step = std::make_unique<UnionStep>(std::move(data_streams), result_header, max_threads);
-
-    query_plan.unitePlans(std::move(union_step), std::move(plans));
-
-    const auto & query = query_ptr->as<ASTSelectWithUnionQuery &>();
-    if (query.union_mode == ASTSelectWithUnionQuery::Mode::DISTINCT)
-    {
-        /// Add distinct transform
-        const Settings & settings = context->getSettingsRef();
-        SizeLimits limits(settings.max_rows_in_distinct, settings.max_bytes_in_distinct, settings.distinct_overflow_mode);
-
-        auto distinct_step = std::make_unique<DistinctStep>(query_plan.getCurrentDataStream(), limits, 0, result_header.getNames(), false);
-
-        query_plan.addStep(std::move(distinct_step));
+        if (settings.limit > 0)
+        {
+            auto limit = std::make_unique<LimitStep>(query_plan.getCurrentDataStream(), settings.limit, settings.offset);
+            limit->setStepDescription("LIMIT OFFSET for SETTINGS");
+            query_plan.addStep(std::move(limit));
+        }
+        else
+        {
+            auto offset = std::make_unique<OffsetStep>(query_plan.getCurrentDataStream(), settings.offset);
+            offset->setStepDescription("OFFSET for SETTINGS");
+            query_plan.addStep(std::move(offset));
+        }
     }
 
 }
diff --git a/src/Interpreters/InterpreterShowTablesQuery.cpp b/src/Interpreters/InterpreterShowTablesQuery.cpp
index 861fa8c3505..49c55fa3a93 100644
--- a/src/Interpreters/InterpreterShowTablesQuery.cpp
+++ b/src/Interpreters/InterpreterShowTablesQuery.cpp
@@ -79,6 +79,26 @@ String InterpreterShowTablesQuery::getRewrittenQuery()
         return rewritten_query.str();
     }
 
+    /// SHOW SETTINGS
+    if (query.m_settings)
+    {
+        WriteBufferFromOwnString rewritten_query;
+        rewritten_query << "SELECT name, type, value FROM system.settings";
+
+        if (query.changed)
+            rewritten_query << " WHERE changed = 1";
+
+        if (!query.like.empty())
+        {
+            rewritten_query
+                << (query.changed ? " AND name " : " WHERE name ")
+                << (query.case_insensitive_like ? "ILIKE " : "LIKE ")
+                << DB::quote << query.like;
+        }
+
+        return rewritten_query.str();
+    }
+
     if (query.temporary && !query.from.empty())
         throw Exception("The `FROM` and `TEMPORARY` cannot be used together in `SHOW TABLES`", ErrorCodes::SYNTAX_ERROR);
 
diff --git a/src/Interpreters/MergeJoin.cpp b/src/Interpreters/MergeJoin.cpp
index 60357187270..4839980eac7 100644
--- a/src/Interpreters/MergeJoin.cpp
+++ b/src/Interpreters/MergeJoin.cpp
@@ -180,12 +180,16 @@ class MergeJoinCursor
 public:
     MergeJoinCursor(const Block & block, const SortDescription & desc_)
         : impl(SortCursorImpl(block, desc_))
-    {}
+    {
+        /// SortCursorImpl can work with permutation, but MergeJoinCursor can't.
+        if (impl.permutation)
+            throw Exception("Logical error: MergeJoinCursor doesn't support permutation", ErrorCodes::LOGICAL_ERROR);
+    }
 
-    size_t position() const { return impl.pos; }
+    size_t position() const { return impl.getRow(); }
     size_t end() const { return impl.rows; }
-    bool atEnd() const { return impl.pos >= impl.rows; }
-    void nextN(size_t num) { impl.pos += num; }
+    bool atEnd() const { return impl.getRow() >= impl.rows; }
+    void nextN(size_t num) { impl.getPosRef() += num; }
 
     void setCompareNullability(const MergeJoinCursor & rhs)
     {
@@ -254,10 +258,10 @@ private:
             else if (cmp > 0)
                 rhs.impl.next();
             else if (!cmp)
-                return Range{impl.pos, rhs.impl.pos, getEqualLength(), rhs.getEqualLength()};
+                return Range{impl.getRow(), rhs.impl.getRow(), getEqualLength(), rhs.getEqualLength()};
         }
 
-        return Range{impl.pos, rhs.impl.pos, 0, 0};
+        return Range{impl.getRow(), rhs.impl.getRow(), 0, 0};
     }
 
     template <bool left_nulls, bool right_nulls>
@@ -268,7 +272,7 @@ private:
             const auto * left_column = impl.sort_columns[i];
             const auto * right_column = rhs.impl.sort_columns[i];
 
-            int res = nullableCompareAt<left_nulls, right_nulls>(*left_column, *right_column, impl.pos, rhs.impl.pos);
+            int res = nullableCompareAt<left_nulls, right_nulls>(*left_column, *right_column, impl.getRow(), rhs.impl.getRow());
             if (res)
                 return res;
         }
@@ -278,11 +282,11 @@ private:
     /// Expects !atEnd()
     size_t getEqualLength()
     {
-        size_t pos = impl.pos + 1;
+        size_t pos = impl.getRow() + 1;
         for (; pos < impl.rows; ++pos)
             if (!samePrev(pos))
                 break;
-        return pos - impl.pos;
+        return pos - impl.getRow();
     }
 
     /// Expects lhs_pos > 0
diff --git a/src/Interpreters/Set.cpp b/src/Interpreters/Set.cpp
index 08e6107590e..dcf7b3b3c24 100644
--- a/src/Interpreters/Set.cpp
+++ b/src/Interpreters/Set.cpp
@@ -23,6 +23,7 @@
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Interpreters/NullableUtils.h>
 #include <Interpreters/sortBlock.h>
+#include <Interpreters/castColumn.h>
 #include <Interpreters/Context.h>
 
 #include <Storages/MergeTree/KeyCondition.h>
@@ -30,7 +31,6 @@
 #include <ext/range.h>
 #include <DataTypes/DataTypeLowCardinality.h>
 
-
 namespace DB
 {
 
@@ -251,11 +251,26 @@ ColumnPtr Set::execute(const Block & block, bool negative) const
 
     /// The constant columns to the left of IN are not supported directly. For this, they first materialize.
     Columns materialized_columns;
+    materialized_columns.reserve(num_key_columns);
 
     for (size_t i = 0; i < num_key_columns; ++i)
     {
-        checkTypesEqual(i, block.safeGetByPosition(i).type);
-        materialized_columns.emplace_back(block.safeGetByPosition(i).column->convertToFullColumnIfConst());
+        ColumnPtr result;
+
+        const auto & column_before_cast = block.safeGetByPosition(i);
+        ColumnWithTypeAndName column_to_cast
+            = {column_before_cast.column->convertToFullColumnIfConst(), column_before_cast.type, column_before_cast.name};
+
+        if (!transform_null_in && data_types[i]->canBeInsideNullable())
+        {
+            result = castColumnAccurateOrNull(column_to_cast, data_types[i]);
+        }
+        else
+        {
+            result = castColumnAccurate(column_to_cast, data_types[i]);
+        }
+
+        materialized_columns.emplace_back() = result;
         key_columns.emplace_back() = materialized_columns.back().get();
     }
 
diff --git a/src/Interpreters/castColumn.cpp b/src/Interpreters/castColumn.cpp
index 249538ec2e0..e8f8bd3bed1 100644
--- a/src/Interpreters/castColumn.cpp
+++ b/src/Interpreters/castColumn.cpp
@@ -1,15 +1,12 @@
-#include <Core/Field.h>
 #include <Interpreters/castColumn.h>
-#include <Interpreters/ExpressionActions.h>
-#include <DataTypes/DataTypeString.h>
-#include <Functions/IFunctionAdaptors.h>
-#include <Functions/FunctionsConversion.h>
 
+#include <Functions/FunctionsConversion.h>
 
 namespace DB
 {
 
-ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+template <CastType cast_type = CastType::nonAccurate>
+static ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
 {
     if (arg.type->equals(*type))
         return arg.column;
@@ -25,10 +22,33 @@ ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type
     };
 
     FunctionOverloadResolverPtr func_builder_cast =
-        std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver::createImpl(false));
+        std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver<cast_type>::createImpl(false));
 
     auto func_cast = func_builder_cast->build(arguments);
-    return func_cast->execute(arguments, type, arg.column->size());
+
+    if constexpr (cast_type == CastType::accurateOrNull)
+    {
+        return func_cast->execute(arguments, makeNullable(type), arg.column->size());
+    }
+    else
+    {
+        return func_cast->execute(arguments, type, arg.column->size());
+    }
+}
+
+ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+{
+    return castColumn<CastType::nonAccurate>(arg, type);
+}
+
+ColumnPtr castColumnAccurate(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+{
+    return castColumn<CastType::accurate>(arg, type);
+}
+
+ColumnPtr castColumnAccurateOrNull(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+{
+    return castColumn<CastType::accurateOrNull>(arg, type);
 }
 
 }
diff --git a/src/Interpreters/castColumn.h b/src/Interpreters/castColumn.h
index 44c7f071c26..fcbea0f4646 100644
--- a/src/Interpreters/castColumn.h
+++ b/src/Interpreters/castColumn.h
@@ -2,8 +2,11 @@
 
 #include <Core/ColumnWithTypeAndName.h>
 
-
 namespace DB
 {
+
 ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type);
+ColumnPtr castColumnAccurate(const ColumnWithTypeAndName & arg, const DataTypePtr & type);
+ColumnPtr castColumnAccurateOrNull(const ColumnWithTypeAndName & arg, const DataTypePtr & type);
+
 }
diff --git a/src/Interpreters/convertFieldToType.cpp b/src/Interpreters/convertFieldToType.cpp
index 18a3eeecb46..73bf493fa65 100644
--- a/src/Interpreters/convertFieldToType.cpp
+++ b/src/Interpreters/convertFieldToType.cpp
@@ -5,6 +5,7 @@
 
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypesDecimal.h>
 #include <DataTypes/DataTypeString.h>
diff --git a/src/Interpreters/loadMetadata.cpp b/src/Interpreters/loadMetadata.cpp
index bb476999f8c..71d3c7e6e5b 100644
--- a/src/Interpreters/loadMetadata.cpp
+++ b/src/Interpreters/loadMetadata.cpp
@@ -108,6 +108,7 @@ void loadMetadata(Context & context, const String & default_database_name)
 
         if (!it->isDirectory())
         {
+            /// TODO: DETACH DATABASE PERMANENTLY ?
             if (endsWith(it.name(), ".sql"))
             {
                 String db_name = it.name().substr(0, it.name().size() - 4);
diff --git a/src/Interpreters/tests/internal_iotop.cpp b/src/Interpreters/tests/internal_iotop.cpp
index 3d17ca32f3f..6025a46e9b7 100644
--- a/src/Interpreters/tests/internal_iotop.cpp
+++ b/src/Interpreters/tests/internal_iotop.cpp
@@ -5,6 +5,7 @@
 #include <Common/TaskStatsInfoGetter.h>
 #include <Poco/File.h>
 #include <Common/Stopwatch.h>
+#include <common/getPageSize.h>
 #include <common/getThreadId.h>
 #include <IO/WriteBufferFromString.h>
 #include <linux/taskstats.h>
@@ -61,8 +62,9 @@ static void do_io(size_t id)
     std::string path_dst = "test_out_" + std::to_string(id);
 
     {
+        size_t page_size = static_cast<size_t>(::getPageSize());
         ReadBufferFromFile rb("/dev/urandom");
-        WriteBufferFromFile wb(path_dst, DBMS_DEFAULT_BUFFER_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, 4096);
+        WriteBufferFromFile wb(path_dst, DBMS_DEFAULT_BUFFER_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, page_size);
         copyData(rb, wb, copy_size);
         wb.close();
     }
diff --git a/src/Parsers/ASTDropQuery.cpp b/src/Parsers/ASTDropQuery.cpp
index b09b588ca6e..6ea6c81c3be 100644
--- a/src/Parsers/ASTDropQuery.cpp
+++ b/src/Parsers/ASTDropQuery.cpp
@@ -67,6 +67,9 @@ void ASTDropQuery::formatQueryImpl(const FormatSettings & settings, FormatState
 
     formatOnCluster(settings);
 
+    if (permanently)
+        settings.ostr << " PERMANENTLY";
+
     if (no_delay)
         settings.ostr << (settings.hilite ? hilite_keyword : "") << " NO DELAY" << (settings.hilite ? hilite_none : "");
 }
diff --git a/src/Parsers/ASTDropQuery.h b/src/Parsers/ASTDropQuery.h
index 5096975c219..b612618aaec 100644
--- a/src/Parsers/ASTDropQuery.h
+++ b/src/Parsers/ASTDropQuery.h
@@ -33,6 +33,9 @@ public:
 
     bool no_delay{false};
 
+    // We detach the object permanently, so it will not be reattached back during server restart.
+    bool permanently{false};
+
     /** Get the text that identifies this element. */
     String getID(char) const override;
     ASTPtr clone() const override;
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index 4cca8ad0807..c86a8e5c139 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -3,6 +3,7 @@
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ASTWithAlias.h>
 #include <Parsers/ASTSubquery.h>
+#include <Parsers/ASTExpressionList.h>
 #include <IO/WriteHelpers.h>
 #include <IO/WriteBufferFromString.h>
 #include <Common/SipHash.h>
@@ -395,6 +396,19 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
             settings.ostr << (settings.hilite ? hilite_operator : "") << ')' << (settings.hilite ? hilite_none : "");
             written = true;
         }
+
+        if (!written && 0 == strcmp(name.c_str(), "map"))
+        {
+            settings.ostr << (settings.hilite ? hilite_operator : "") << '{' << (settings.hilite ? hilite_none : "");
+            for (size_t i = 0; i < arguments->children.size(); ++i)
+            {
+                if (i != 0)
+                    settings.ostr << ", ";
+                arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
+            }
+            settings.ostr << (settings.hilite ? hilite_operator : "") << '}' << (settings.hilite ? hilite_none : "");
+            written = true;
+        }
     }
 
     if (!written)
diff --git a/src/Parsers/ASTShowTablesQuery.cpp b/src/Parsers/ASTShowTablesQuery.cpp
index cd83bae06d9..61d68c4a273 100644
--- a/src/Parsers/ASTShowTablesQuery.cpp
+++ b/src/Parsers/ASTShowTablesQuery.cpp
@@ -55,6 +55,12 @@ void ASTShowTablesQuery::formatQueryImpl(const FormatSettings & settings, Format
         settings.ostr << (settings.hilite ? hilite_keyword : "") << "SHOW CLUSTER" << (settings.hilite ? hilite_none : "");
         settings.ostr << " " << backQuoteIfNeed(cluster_str);
     }
+    else if (m_settings)
+    {
+        settings.ostr << (settings.hilite ? hilite_keyword : "") << "SHOW " << (changed ? "CHANGED " : "") << "SETTINGS" <<
+            (settings.hilite ? hilite_none : "");
+        formatLike(settings);
+    }
     else
     {
         settings.ostr << (settings.hilite ? hilite_keyword : "") << "SHOW " << (temporary ? "TEMPORARY " : "") <<
diff --git a/src/Parsers/ASTShowTablesQuery.h b/src/Parsers/ASTShowTablesQuery.h
index 43976e8a958..57383dff66f 100644
--- a/src/Parsers/ASTShowTablesQuery.h
+++ b/src/Parsers/ASTShowTablesQuery.h
@@ -18,6 +18,8 @@ public:
     bool clusters{false};
     bool cluster{false};
     bool dictionaries{false};
+    bool m_settings{false};
+    bool changed{false};
     bool temporary{false};
 
     String cluster_str;
diff --git a/src/Parsers/ExpressionElementParsers.cpp b/src/Parsers/ExpressionElementParsers.cpp
index 6fbe9abc96e..8706aecb7ac 100644
--- a/src/Parsers/ExpressionElementParsers.cpp
+++ b/src/Parsers/ExpressionElementParsers.cpp
@@ -1110,6 +1110,10 @@ bool ParserCollectionOfLiterals<Collection>::parseImpl(Pos & pos, ASTPtr & node,
             {
                 ++pos;
             }
+            else if (pos->type == TokenType::Colon && std::is_same_v<Collection, Map> && arr.size() % 2 == 1)
+            {
+                ++pos;
+            }
             else
             {
                 expected.add(pos, "comma or closing bracket");
@@ -1130,6 +1134,7 @@ bool ParserCollectionOfLiterals<Collection>::parseImpl(Pos & pos, ASTPtr & node,
 
 template bool ParserCollectionOfLiterals<Array>::parseImpl(Pos & pos, ASTPtr & node, Expected & expected);
 template bool ParserCollectionOfLiterals<Tuple>::parseImpl(Pos & pos, ASTPtr & node, Expected & expected);
+template bool ParserCollectionOfLiterals<Map>::parseImpl(Pos & pos, ASTPtr & node, Expected & expected);
 
 bool ParserLiteral::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
 {
@@ -1564,6 +1569,7 @@ bool ParserExpressionElement::parseImpl(Pos & pos, ASTPtr & node, Expected & exp
 {
     return ParserSubquery().parse(pos, node, expected)
         || ParserTupleOfLiterals().parse(pos, node, expected)
+        || ParserMapOfLiterals().parse(pos, node, expected)
         || ParserParenthesisExpression().parse(pos, node, expected)
         || ParserArrayOfLiterals().parse(pos, node, expected)
         || ParserArray().parse(pos, node, expected)
diff --git a/src/Parsers/ExpressionElementParsers.h b/src/Parsers/ExpressionElementParsers.h
index 4713a7ea8a6..4033a6e4f61 100644
--- a/src/Parsers/ExpressionElementParsers.h
+++ b/src/Parsers/ExpressionElementParsers.h
@@ -270,6 +270,18 @@ protected:
     }
 };
 
+class ParserMapOfLiterals : public IParserBase
+{
+public:
+    ParserCollectionOfLiterals<Map> map_parser{TokenType::OpeningCurlyBrace, TokenType::ClosingCurlyBrace};
+protected:
+    const char * getName() const override { return "map"; }
+    bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override
+    {
+        return map_parser.parse(pos, node, expected);
+    }
+};
+
 class ParserArrayOfLiterals : public IParserBase
 {
 public:
diff --git a/src/Parsers/ExpressionListParsers.cpp b/src/Parsers/ExpressionListParsers.cpp
index dda1417ca97..f50cf71f54d 100644
--- a/src/Parsers/ExpressionListParsers.cpp
+++ b/src/Parsers/ExpressionListParsers.cpp
@@ -100,6 +100,7 @@ bool ParserList::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
     auto list = std::make_shared<ASTExpressionList>(result_separator);
     list->children = std::move(elements);
     node = list;
+
     return true;
 }
 
diff --git a/src/Parsers/ParserCreateQuery.cpp b/src/Parsers/ParserCreateQuery.cpp
index b08646bcf9d..5fe6a4eee80 100644
--- a/src/Parsers/ParserCreateQuery.cpp
+++ b/src/Parsers/ParserCreateQuery.cpp
@@ -449,12 +449,22 @@ bool ParserCreateTableQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
         if (!s_rparen.ignore(pos, expected))
             return false;
 
-        if (!storage_p.parse(pos, storage, expected) && !is_temporary)
+        auto storage_parse_result = storage_p.parse(pos, storage, expected);
+
+        if (storage_parse_result && s_as.ignore(pos, expected))
+        {
+            if (!select_p.parse(pos, select, expected))
+                return false;
+        }
+
+        if (!storage_parse_result && !is_temporary)
         {
             if (!s_as.ignore(pos, expected))
                 return false;
             if (!table_function_p.parse(pos, as_table_function, expected))
+            {
                 return false;
+            }
         }
     }
     else
diff --git a/src/Parsers/ParserDropQuery.cpp b/src/Parsers/ParserDropQuery.cpp
index e09642e7b84..5400f33fbd9 100644
--- a/src/Parsers/ParserDropQuery.cpp
+++ b/src/Parsers/ParserDropQuery.cpp
@@ -11,7 +11,7 @@ namespace DB
 namespace
 {
 
-bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool optional_table_keyword = false)
+bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, const ASTDropQuery::Kind kind)
 {
     ParserKeyword s_temporary("TEMPORARY");
     ParserKeyword s_table("TABLE");
@@ -21,6 +21,7 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
     ParserToken s_dot(TokenType::Dot);
     ParserKeyword s_if_exists("IF EXISTS");
     ParserIdentifier name_p;
+    ParserKeyword s_permanently("PERMANENTLY");
     ParserKeyword s_no_delay("NO DELAY");
     ParserKeyword s_sync("SYNC");
 
@@ -32,6 +33,7 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
     bool is_dictionary = false;
     bool is_view = false;
     bool no_delay = false;
+    bool permanently = false;
 
     if (s_database.ignore(pos, expected))
     {
@@ -40,15 +42,6 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
 
         if (!name_p.parse(pos, database, expected))
             return false;
-
-        if (ParserKeyword{"ON"}.ignore(pos, expected))
-        {
-            if (!ASTQueryWithOnCluster::parse(pos, cluster_str, expected))
-                return false;
-        }
-
-        if (s_no_delay.ignore(pos, expected) || s_sync.ignore(pos, expected))
-            no_delay = true;
     }
     else
     {
@@ -59,7 +52,8 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
         else if (s_temporary.ignore(pos, expected))
             temporary = true;
 
-        if (!is_view && !is_dictionary && (!s_table.ignore(pos, expected) && !optional_table_keyword))
+        /// for TRUNCATE queries TABLE keyword is assumed as default and can be skipped
+        if (!is_view && !is_dictionary && (!s_table.ignore(pos, expected) && kind != ASTDropQuery::Kind::Truncate))
         {
             return false;
         }
@@ -76,26 +70,32 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
             if (!name_p.parse(pos, table, expected))
                 return false;
         }
-
-        if (ParserKeyword{"ON"}.ignore(pos, expected))
-        {
-            if (!ASTQueryWithOnCluster::parse(pos, cluster_str, expected))
-                return false;
-        }
-
-        if (s_no_delay.ignore(pos, expected) || s_sync.ignore(pos, expected))
-            no_delay = true;
     }
 
+    /// common for tables / dictionaries / databases
+    if (ParserKeyword{"ON"}.ignore(pos, expected))
+    {
+        if (!ASTQueryWithOnCluster::parse(pos, cluster_str, expected))
+            return false;
+    }
+
+    if (kind == ASTDropQuery::Kind::Detach && s_permanently.ignore(pos, expected))
+        permanently = true;
+
+    /// actually for TRUNCATE NO DELAY / SYNC means nothing
+    if (s_no_delay.ignore(pos, expected) || s_sync.ignore(pos, expected))
+        no_delay = true;
+
     auto query = std::make_shared<ASTDropQuery>();
     node = query;
 
-    query->kind = ASTDropQuery::Kind::Drop;
+    query->kind = kind;
     query->if_exists = if_exists;
     query->temporary = temporary;
     query->is_dictionary = is_dictionary;
     query->is_view = is_view;
     query->no_delay = no_delay;
+    query->permanently = permanently;
 
     tryGetIdentifierNameInto(database, query->database);
     tryGetIdentifierNameInto(table, query->table);
@@ -105,28 +105,6 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
     return true;
 }
 
-bool parseDetachQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected)
-{
-    if (parseDropQuery(pos, node, expected))
-    {
-        auto * drop_query = node->as<ASTDropQuery>();
-        drop_query->kind = ASTDropQuery::Kind::Detach;
-        return true;
-    }
-    return false;
-}
-
-bool parseTruncateQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected)
-{
-    if (parseDropQuery(pos, node, expected, true))
-    {
-        auto * drop_query = node->as<ASTDropQuery>();
-        drop_query->kind = ASTDropQuery::Kind::Truncate;
-        return true;
-    }
-    return false;
-}
-
 }
 
 bool ParserDropQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
@@ -136,11 +114,11 @@ bool ParserDropQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
     ParserKeyword s_truncate("TRUNCATE");
 
     if (s_drop.ignore(pos, expected))
-        return parseDropQuery(pos, node, expected);
+        return parseDropQuery(pos, node, expected, ASTDropQuery::Kind::Drop);
     else if (s_detach.ignore(pos, expected))
-        return parseDetachQuery(pos, node, expected);
+        return parseDropQuery(pos, node, expected, ASTDropQuery::Kind::Detach);
     else if (s_truncate.ignore(pos, expected))
-        return parseTruncateQuery(pos, node, expected);
+        return parseDropQuery(pos, node, expected, ASTDropQuery::Kind::Truncate);
     else
         return false;
 }
diff --git a/src/Parsers/ParserDropQuery.h b/src/Parsers/ParserDropQuery.h
index 069b9c34ddd..39ff5b7c4fd 100644
--- a/src/Parsers/ParserDropQuery.h
+++ b/src/Parsers/ParserDropQuery.h
@@ -8,7 +8,7 @@ namespace DB
 {
 
 /** Query like this:
-  * DROP|DETACH|TRUNCATE TABLE [IF EXISTS] [db.]name
+  * DROP|DETACH|TRUNCATE TABLE [IF EXISTS] [db.]name [PERMANENTLY]
   *
   * Or:
   * DROP DATABASE [IF EXISTS] db
diff --git a/src/Parsers/ParserExplainQuery.cpp b/src/Parsers/ParserExplainQuery.cpp
index fe48addd0cd..7a7abba94a9 100644
--- a/src/Parsers/ParserExplainQuery.cpp
+++ b/src/Parsers/ParserExplainQuery.cpp
@@ -5,6 +5,7 @@
 #include <Parsers/ParserCreateQuery.h>
 #include <Parsers/ParserSelectWithUnionQuery.h>
 #include <Parsers/ParserSetQuery.h>
+#include <Parsers/ParserQuery.h>
 
 namespace DB
 {
@@ -51,7 +52,13 @@ bool ParserExplainQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected
     ParserCreateTableQuery create_p;
     ParserSelectWithUnionQuery select_p;
     ASTPtr query;
-    if (select_p.parse(pos, query, expected) ||
+    if (kind == ASTExplainQuery::ExplainKind::ParsedAST)
+    {
+        ParserQuery p(end);
+        if (p.parse(pos, query, expected))
+            explain_query->setExplainedQuery(std::move(query));
+    }
+    else if (select_p.parse(pos, query, expected) ||
         create_p.parse(pos, query, expected))
         explain_query->setExplainedQuery(std::move(query));
     else
diff --git a/src/Parsers/ParserExplainQuery.h b/src/Parsers/ParserExplainQuery.h
index b58597b9248..a1865e30239 100644
--- a/src/Parsers/ParserExplainQuery.h
+++ b/src/Parsers/ParserExplainQuery.h
@@ -9,8 +9,12 @@ namespace DB
 class ParserExplainQuery : public IParserBase
 {
 protected:
+    const char * end;
+
     const char * getName() const override { return "EXPLAIN"; }
     bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override;
+public:
+    ParserExplainQuery(const char* end_) : end(end_) {}
 };
 
 }
diff --git a/src/Parsers/ParserQuery.cpp b/src/Parsers/ParserQuery.cpp
index 89ddd1797c0..4550bdc8a75 100644
--- a/src/Parsers/ParserQuery.cpp
+++ b/src/Parsers/ParserQuery.cpp
@@ -26,7 +26,7 @@ namespace DB
 
 bool ParserQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
 {
-    ParserQueryWithOutput query_with_output_p;
+    ParserQueryWithOutput query_with_output_p(end);
     ParserInsertQuery insert_p(end);
     ParserUseQuery use_p;
     ParserSetQuery set_p;
diff --git a/src/Parsers/ParserQueryWithOutput.cpp b/src/Parsers/ParserQueryWithOutput.cpp
index 76a3e8b95f8..d5aa1e47533 100644
--- a/src/Parsers/ParserQueryWithOutput.cpp
+++ b/src/Parsers/ParserQueryWithOutput.cpp
@@ -48,7 +48,7 @@ bool ParserQueryWithOutput::parseImpl(Pos & pos, ASTPtr & node, Expected & expec
     ParserShowCreateAccessEntityQuery show_create_access_entity_p;
     ParserShowGrantsQuery show_grants_p;
     ParserShowPrivilegesQuery show_privileges_p;
-    ParserExplainQuery explain_p;
+    ParserExplainQuery explain_p(end);
 
     ASTPtr query;
 
diff --git a/src/Parsers/ParserQueryWithOutput.h b/src/Parsers/ParserQueryWithOutput.h
index 110cf923147..854d5a74ffd 100644
--- a/src/Parsers/ParserQueryWithOutput.h
+++ b/src/Parsers/ParserQueryWithOutput.h
@@ -11,8 +11,11 @@ namespace DB
 class ParserQueryWithOutput : public IParserBase
 {
 protected:
+    const char * end;
     const char * getName() const override { return "Query with output"; }
     bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override;
+public:
+    ParserQueryWithOutput(const char * end_) : end(end_) {}
 };
 
 }
diff --git a/src/Parsers/ParserShowTablesQuery.cpp b/src/Parsers/ParserShowTablesQuery.cpp
index 4586e10a8a3..6022a31034a 100644
--- a/src/Parsers/ParserShowTablesQuery.cpp
+++ b/src/Parsers/ParserShowTablesQuery.cpp
@@ -24,6 +24,8 @@ bool ParserShowTablesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expec
     ParserKeyword s_clusters("CLUSTERS");
     ParserKeyword s_cluster("CLUSTER");
     ParserKeyword s_dictionaries("DICTIONARIES");
+    ParserKeyword s_settings("SETTINGS");
+    ParserKeyword s_changed("CHANGED");
     ParserKeyword s_from("FROM");
     ParserKeyword s_in("IN");
     ParserKeyword s_not("NOT");
@@ -99,6 +101,29 @@ bool ParserShowTablesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expec
 
         query->cluster_str = std::move(cluster_str);
     }
+    else if (bool changed = s_changed.ignore(pos); changed || s_settings.ignore(pos))
+    {
+        query->m_settings = true;
+
+        if (changed)
+        {
+            query->changed = true;
+            if (!s_settings.ignore(pos, expected))
+                return false;
+        }
+
+        /// Not expected due to "SHOW SETTINGS PROFILES"
+        if (bool insensitive = s_ilike.ignore(pos); insensitive || s_like.ignore(pos))
+        {
+            if (insensitive)
+                query->case_insensitive_like = true;
+
+            if (!like_p.parse(pos, like, expected))
+                return false;
+        }
+        else
+            return false;
+    }
     else
     {
         if (s_temporary.ignore(pos))
diff --git a/src/Parsers/tests/select_parser.cpp b/src/Parsers/tests/select_parser.cpp
index 7c18563659d..15295170c6b 100644
--- a/src/Parsers/tests/select_parser.cpp
+++ b/src/Parsers/tests/select_parser.cpp
@@ -22,7 +22,7 @@ try
         " INTO OUTFILE 'test.out'"
         " FORMAT TabSeparated";
 
-    ParserQueryWithOutput parser;
+    ParserQueryWithOutput parser(input.data() + input.size());
     ASTPtr ast = parseQuery(parser, input.data(), input.data() + input.size(), "", 0, 0);
 
     std::cout << "Success." << std::endl;
diff --git a/src/Parsers/tests/select_parser_fuzzer.cpp b/src/Parsers/tests/select_parser_fuzzer.cpp
index fc4670333fe..4848b285c07 100644
--- a/src/Parsers/tests/select_parser_fuzzer.cpp
+++ b/src/Parsers/tests/select_parser_fuzzer.cpp
@@ -11,7 +11,7 @@ try
 {
     std::string input = std::string(reinterpret_cast<const char*>(data), size);
 
-    DB::ParserQueryWithOutput parser;
+    DB::ParserQueryWithOutput parser(input.data() + input.size());
     DB::ASTPtr ast = parseQuery(parser, input.data(), input.data() + input.size(), "", 0, 0);
 
     DB::formatAST(*ast, std::cerr);
diff --git a/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp b/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp
index 68854e53a02..d7a65c2f15d 100644
--- a/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp
+++ b/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp
@@ -1,11 +1,13 @@
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnTuple.h>
+#include <Columns/ColumnMap.h>
 #include <Columns/ColumnsNumber.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeString.h>
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/FieldToDataType.h>
 #include <Processors/Formats/IRowInputFormat.h>
 #include <Functions/FunctionFactory.h>
@@ -46,6 +48,7 @@ struct SpecialParserType
     bool is_nullable = false;
     bool is_array = false;
     bool is_tuple = false;
+    bool is_map = false;
     /// Type and nullability
     std::vector<std::pair<Field::Types::Which, bool>> nested_types;
 
@@ -119,6 +122,10 @@ static void fillLiteralInfo(DataTypes & nested_types, LiteralInfo & info)
         {
             field_type = Field::Types::Tuple;
         }
+        else if (type_info.isMap())
+        {
+            field_type = Field::Types::Map;
+        }
         else
             throw Exception("Unexpected literal type inside Array: " + nested_type->getName() + ". It's a bug",
                             ErrorCodes::LOGICAL_ERROR);
@@ -196,6 +203,13 @@ private:
                 if (not_null == array.end())
                     return true;
             }
+            else if (literal->value.getType() == Field::Types::Map)
+            {
+                const Map & map = literal->value.get<Map>();
+                if (map.size() % 2)
+                    return false;
+            }
+
             String column_name = "_dummy_" + std::to_string(replaced_literals.size());
             replaced_literals.emplace_back(literal, column_name, force_nullable);
             setDataType(replaced_literals.back());
@@ -243,6 +257,15 @@ private:
             fillLiteralInfo(nested_types, info);
             info.type = std::make_shared<DataTypeTuple>(nested_types);
         }
+        else if (field_type == Field::Types::Map)
+        {
+            info.special_parser.is_map = true;
+
+            info.type = applyVisitor(FieldToDataType(), info.literal->value);
+            auto nested_types = assert_cast<const DataTypeMap &>(*info.type).getKeyValueTypes();
+            fillLiteralInfo(nested_types, info);
+            info.type = std::make_shared<DataTypeMap>(nested_types);
+        }
         else
             throw Exception(String("Unexpected literal type ") + info.literal->value.getTypeName() + ". It's a bug",
                             ErrorCodes::LOGICAL_ERROR);
@@ -453,17 +476,19 @@ bool ConstantExpressionTemplate::parseLiteralAndAssertType(ReadBuffer & istr, co
     /// If literal does not fit entirely in the buffer, parsing error will happen.
     /// However, it's possible to deduce new template (or use template from cache) after error like it was template mismatch.
 
-    if (type_info.is_array || type_info.is_tuple)
+    if (type_info.is_array || type_info.is_tuple || type_info.is_map)
     {
         /// TODO faster way to check types without using Parsers
         ParserArrayOfLiterals parser_array;
         ParserTupleOfLiterals parser_tuple;
+        ParserMapOfLiterals parser_map;
 
         Tokens tokens_number(istr.position(), istr.buffer().end());
         IParser::Pos iterator(tokens_number, settings.max_parser_depth);
         Expected expected;
         ASTPtr ast;
-        if (!parser_array.parse(iterator, ast, expected) && !parser_tuple.parse(iterator, ast, expected))
+        if (!parser_array.parse(iterator, ast, expected) && !parser_tuple.parse(iterator, ast, expected)
+            && !parser_map.parse(iterator, ast, expected))
             return false;
 
         istr.position() = const_cast<char *>(iterator->begin);
@@ -474,8 +499,10 @@ bool ConstantExpressionTemplate::parseLiteralAndAssertType(ReadBuffer & istr, co
         DataTypes nested_types;
         if (type_info.is_array)
             nested_types = { assert_cast<const DataTypeArray &>(*collection_type).getNestedType() };
-        else
+        else if (type_info.is_tuple)
             nested_types = assert_cast<const DataTypeTuple &>(*collection_type).getElements();
+        else
+            nested_types = assert_cast<const DataTypeMap &>(*collection_type).getKeyValueTypes();
 
         for (size_t i = 0; i < nested_types.size(); ++i)
         {
diff --git a/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp
index 11d68372b10..6b2ee1c8039 100644
--- a/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp
@@ -257,12 +257,12 @@ void AggregatingSortedAlgorithm::AggregatingMergedData::addRow(SortCursor & curs
         throw Exception("Can't add a row to the group because it was not started.", ErrorCodes::LOGICAL_ERROR);
 
     for (auto & desc : def.columns_to_aggregate)
-        desc.column->insertMergeFrom(*cursor->all_columns[desc.column_number], cursor->pos);
+        desc.column->insertMergeFrom(*cursor->all_columns[desc.column_number], cursor->getRow());
 
     for (auto & desc : def.columns_to_simple_aggregate)
     {
         auto & col = cursor->all_columns[desc.column_number];
-        desc.add_function(desc.function.get(), desc.state.data(), &col, cursor->pos, arena.get());
+        desc.add_function(desc.function.get(), desc.state.data(), &col, cursor->getRow(), arena.get());
     }
 }
 
@@ -352,7 +352,7 @@ IMergingAlgorithm::Status AggregatingSortedAlgorithm::merge()
                 return Status(merged_data.pull());
             }
 
-            merged_data.startGroup(current->all_columns, current->pos);
+            merged_data.startGroup(current->all_columns, current->getRow());
         }
 
         merged_data.addRow(current);
diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp
index 8f6888b9a7a..ccb66259e2e 100644
--- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp
@@ -26,9 +26,9 @@ CollapsingSortedAlgorithm::CollapsingSortedAlgorithm(
     const String & sign_column,
     bool only_positive_sign_,
     size_t max_block_size,
+    Poco::Logger * log_,
     WriteBuffer * out_row_sources_buf_,
-    bool use_average_block_sizes,
-    Poco::Logger * log_)
+    bool use_average_block_sizes)
     : IMergingAlgorithmWithSharedChunks(num_inputs, std::move(description_), out_row_sources_buf_, max_row_refs)
     , merged_data(header.cloneEmptyColumns(), use_average_block_sizes, max_block_size)
     , sign_column_number(header.getPositionByName(sign_column))
@@ -123,7 +123,7 @@ IMergingAlgorithm::Status CollapsingSortedAlgorithm::merge()
             return Status(current.impl->order);
         }
 
-        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->pos];
+        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->getRow()];
 
         RowRef current_row;
         setRowRef(current_row, current);
diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h
index d95fac2f02b..028715f715b 100644
--- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h
+++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h
@@ -33,9 +33,9 @@ public:
         const String & sign_column,
         bool only_positive_sign_, /// For select final. Skip rows with sum(sign) < 0.
         size_t max_block_size,
-        WriteBuffer * out_row_sources_buf_,
-        bool use_average_block_sizes,
-        Poco::Logger * log_);
+        Poco::Logger * log_,
+        WriteBuffer * out_row_sources_buf_ = nullptr,
+        bool use_average_block_sizes = false);
 
     Status merge() override;
 
diff --git a/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp
index 0e704e5a05b..df10fb26d40 100644
--- a/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp
@@ -164,12 +164,12 @@ IMergingAlgorithm::Status GraphiteRollupSortedAlgorithm::merge()
             return Status(current.impl->order);
         }
 
-        StringRef next_path = current->all_columns[columns_definition.path_column_num]->getDataAt(current->pos);
+        StringRef next_path = current->all_columns[columns_definition.path_column_num]->getDataAt(current->getRow());
         bool new_path = is_first || next_path != current_group_path;
 
         is_first = false;
 
-        time_t next_row_time = current->all_columns[columns_definition.time_column_num]->getUInt(current->pos);
+        time_t next_row_time = current->all_columns[columns_definition.time_column_num]->getUInt(current->getRow());
         /// Is new key before rounding.
         bool is_new_key = new_path || next_row_time != current_time;
 
@@ -227,7 +227,7 @@ IMergingAlgorithm::Status GraphiteRollupSortedAlgorithm::merge()
         /// and for rows with same maximum version - only last row.
         if (is_new_key
             || current->all_columns[columns_definition.version_column_num]->compareAt(
-                current->pos, current_subgroup_newest_row.row_num,
+                current->getRow(), current_subgroup_newest_row.row_num,
                 *(*current_subgroup_newest_row.all_columns)[columns_definition.version_column_num],
                 /* nan_direction_hint = */ 1) >= 0)
         {
@@ -263,7 +263,7 @@ IMergingAlgorithm::Status GraphiteRollupSortedAlgorithm::merge()
 
 void GraphiteRollupSortedAlgorithm::startNextGroup(SortCursor & cursor, Graphite::RollupRule next_rule)
 {
-    merged_data.startNextGroup(cursor->all_columns, cursor->pos, next_rule, columns_definition);
+    merged_data.startNextGroup(cursor->all_columns, cursor->getRow(), next_rule, columns_definition);
 }
 
 void GraphiteRollupSortedAlgorithm::finishCurrentGroup()
diff --git a/src/Processors/Merges/Algorithms/IMergingAlgorithm.h b/src/Processors/Merges/Algorithms/IMergingAlgorithm.h
index f86be2a7d1b..5c8d18875e7 100644
--- a/src/Processors/Merges/Algorithms/IMergingAlgorithm.h
+++ b/src/Processors/Merges/Algorithms/IMergingAlgorithm.h
@@ -29,6 +29,8 @@ public:
         /// between different algorithm objects in parallel FINAL.
         bool skip_last_row = false;
 
+        IColumn::Permutation * permutation = nullptr;
+
         void swap(Input & other)
         {
             chunk.swap(other.chunk);
diff --git a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp
index 0b13d689636..e4c60d7609c 100644
--- a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp
+++ b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp
@@ -22,7 +22,7 @@ void IMergingAlgorithmWithDelayedChunk::initializeQueue(Inputs inputs)
         if (!current_inputs[source_num].chunk)
             continue;
 
-        cursors[source_num] = SortCursorImpl(current_inputs[source_num].chunk.getColumns(), description, source_num);
+        cursors[source_num] = SortCursorImpl(current_inputs[source_num].chunk.getColumns(), description, source_num, current_inputs[source_num].permutation);
     }
 
     queue = SortingHeap<SortCursor>(cursors);
@@ -37,7 +37,7 @@ void IMergingAlgorithmWithDelayedChunk::updateCursor(Input & input, size_t sourc
     last_chunk_sort_columns = std::move(cursors[source_num].sort_columns);
 
     current_input.swap(input);
-    cursors[source_num].reset(current_input.chunk.getColumns(), {});
+    cursors[source_num].reset(current_input.chunk.getColumns(), {}, current_input.permutation);
 
     queue.push(cursors[source_num]);
 }
diff --git a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp
index 39abe5c0ec7..97abffdc167 100644
--- a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp
+++ b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp
@@ -39,7 +39,7 @@ void IMergingAlgorithmWithSharedChunks::initialize(Inputs inputs)
 
         source.skip_last_row = inputs[source_num].skip_last_row;
         source.chunk = chunk_allocator.alloc(inputs[source_num].chunk);
-        cursors[source_num] = SortCursorImpl(source.chunk->getColumns(), description, source_num);
+        cursors[source_num] = SortCursorImpl(source.chunk->getColumns(), description, source_num, inputs[source_num].permutation);
 
         source.chunk->all_columns = cursors[source_num].all_columns;
         source.chunk->sort_columns = cursors[source_num].sort_columns;
@@ -55,7 +55,7 @@ void IMergingAlgorithmWithSharedChunks::consume(Input & input, size_t source_num
     auto & source = sources[source_num];
     source.skip_last_row = input.skip_last_row;
     source.chunk = chunk_allocator.alloc(input.chunk);
-    cursors[source_num].reset(source.chunk->getColumns(), {});
+    cursors[source_num].reset(source.chunk->getColumns(), {}, input.permutation);
 
     source.chunk->all_columns = cursors[source_num].all_columns;
     source.chunk->sort_columns = cursors[source_num].sort_columns;
diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp
index ee13ef70203..511bd9dd74f 100644
--- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp
@@ -139,7 +139,7 @@ IMergingAlgorithm::Status MergingSortedAlgorithm::mergeImpl(TSortingHeap & queue
 
         //std::cerr << "total_merged_rows: " << total_merged_rows << ", merged_rows: " << merged_rows << "\n";
         //std::cerr << "Inserting row\n";
-        merged_data.insertRow(current->all_columns, current->pos, current->rows);
+        merged_data.insertRow(current->all_columns, current->getRow(), current->rows);
 
         if (out_row_sources_buf)
         {
diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h
index 531b2636747..63dced26dd4 100644
--- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h
+++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h
@@ -18,9 +18,9 @@ public:
         size_t num_inputs,
         SortDescription description_,
         size_t max_block_size,
-        UInt64 limit_,
-        WriteBuffer * out_row_sources_buf_,
-        bool use_average_block_sizes);
+        UInt64 limit_ = 0,
+        WriteBuffer * out_row_sources_buf_ = nullptr,
+        bool use_average_block_sizes = false);
 
     void addInput();
 
diff --git a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp
index 3ee0df0efd8..132241844d7 100644
--- a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp
@@ -73,7 +73,7 @@ IMergingAlgorithm::Status ReplacingSortedAlgorithm::merge()
         if (version_column_number == -1
             || selected_row.empty()
             || current->all_columns[version_column_number]->compareAt(
-                current->pos, selected_row.row_num,
+                current->getRow(), selected_row.row_num,
                 *(*selected_row.all_columns)[version_column_number],
                 /* nan_direction_hint = */ 1) >= 0)
         {
diff --git a/src/Processors/Merges/Algorithms/RowRef.h b/src/Processors/Merges/Algorithms/RowRef.h
index 1b4da9781f8..e4610c88581 100644
--- a/src/Processors/Merges/Algorithms/RowRef.h
+++ b/src/Processors/Merges/Algorithms/RowRef.h
@@ -136,7 +136,7 @@ struct RowRef
     {
         sort_columns = cursor.impl->sort_columns.data();
         num_columns = cursor.impl->sort_columns.size();
-        row_num = cursor.impl->pos;
+        row_num = cursor.impl->getRow();
     }
 
     static bool checkEquals(size_t size, const IColumn ** lhs, size_t lhs_row, const IColumn ** rhs, size_t rhs_row)
@@ -192,7 +192,7 @@ struct RowRefWithOwnedChunk
     void set(SortCursor & cursor, SharedChunkPtr chunk)
     {
         owned_chunk = std::move(chunk);
-        row_num = cursor.impl->pos;
+        row_num = cursor.impl->getRow();
         all_columns = &owned_chunk->all_columns;
         sort_columns = &owned_chunk->sort_columns;
     }
diff --git a/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp
index 17e5e4364ff..f558a6f0677 100644
--- a/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp
@@ -688,10 +688,10 @@ IMergingAlgorithm::Status SummingSortedAlgorithm::merge()
                 return Status(merged_data.pull());
             }
 
-            merged_data.startGroup(current->all_columns, current->pos);
+            merged_data.startGroup(current->all_columns, current->getRow());
         }
         else
-            merged_data.addRow(current->all_columns, current->pos);
+            merged_data.addRow(current->all_columns, current->getRow());
 
         if (!current->isLast())
         {
diff --git a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp
index 6273bd28371..672242b253b 100644
--- a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp
@@ -73,7 +73,7 @@ IMergingAlgorithm::Status VersionedCollapsingAlgorithm::merge()
 
         RowRef current_row;
 
-        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->pos];
+        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->getRow()];
 
         setRowRef(current_row, current);
 
diff --git a/src/Processors/Merges/CollapsingSortedTransform.h b/src/Processors/Merges/CollapsingSortedTransform.h
index 4e65504a101..9e6bd306eee 100644
--- a/src/Processors/Merges/CollapsingSortedTransform.h
+++ b/src/Processors/Merges/CollapsingSortedTransform.h
@@ -27,9 +27,9 @@ public:
             sign_column,
             only_positive_sign,
             max_block_size,
+            &Poco::Logger::get("CollapsingSortedTransform"),
             out_row_sources_buf_,
-            use_average_block_sizes,
-            &Poco::Logger::get("CollapsingSortedTransform"))
+            use_average_block_sizes)
     {
     }
 
diff --git a/src/Processors/Transforms/SortingTransform.cpp b/src/Processors/Transforms/SortingTransform.cpp
index 03c8e87ce7a..11f23530c9e 100644
--- a/src/Processors/Transforms/SortingTransform.cpp
+++ b/src/Processors/Transforms/SortingTransform.cpp
@@ -100,7 +100,7 @@ Chunk MergeSorter::mergeImpl(TSortingHeap & queue)
 
         /// Append a row from queue.
         for (size_t i = 0; i < num_columns; ++i)
-            merged_columns[i]->insertFrom(*current->all_columns[i], current->pos);
+            merged_columns[i]->insertFrom(*current->all_columns[i], current->getRow());
 
         ++total_merged_rows;
         ++merged_rows;
diff --git a/src/Server/TestKeeperTCPHandler.cpp b/src/Server/TestKeeperTCPHandler.cpp
index ffb73985824..ad5c3898bb4 100644
--- a/src/Server/TestKeeperTCPHandler.cpp
+++ b/src/Server/TestKeeperTCPHandler.cpp
@@ -419,28 +419,40 @@ Coordination::OpNum TestKeeperTCPHandler::receiveRequest()
     Coordination::ZooKeeperRequestPtr request = Coordination::ZooKeeperRequestFactory::instance().get(opnum);
     request->xid = xid;
     request->readImpl(*in);
-    int response_fd = poll_wrapper->getResponseFD();
     auto promise = std::make_shared<std::promise<Coordination::ZooKeeperResponsePtr>>();
-    zkutil::ResponseCallback callback = [response_fd, promise] (const Coordination::ZooKeeperResponsePtr & response)
+    if (opnum != Coordination::OpNum::Close)
     {
-        promise->set_value(response);
-        [[maybe_unused]] int result = write(response_fd, &RESPONSE_BYTE, sizeof(RESPONSE_BYTE));
-    };
-
-    if (request->has_watch)
-    {
-        auto watch_promise = std::make_shared<std::promise<Coordination::ZooKeeperResponsePtr>>();
-        zkutil::ResponseCallback watch_callback = [response_fd, watch_promise] (const Coordination::ZooKeeperResponsePtr & response)
+        int response_fd = poll_wrapper->getResponseFD();
+        zkutil::ResponseCallback callback = [response_fd, promise] (const Coordination::ZooKeeperResponsePtr & response)
         {
-            watch_promise->set_value(response);
-            [[maybe_unused]] int result = write(response_fd, &WATCH_RESPONSE_BYTE, sizeof(WATCH_RESPONSE_BYTE));
+            promise->set_value(response);
+            [[maybe_unused]] int result = write(response_fd, &RESPONSE_BYTE, sizeof(RESPONSE_BYTE));
         };
-        test_keeper_storage->putRequest(request, session_id, callback, watch_callback);
-        responses.push(promise->get_future());
-        watch_responses.emplace_back(watch_promise->get_future());
+
+        if (request->has_watch)
+        {
+            auto watch_promise = std::make_shared<std::promise<Coordination::ZooKeeperResponsePtr>>();
+            zkutil::ResponseCallback watch_callback = [response_fd, watch_promise] (const Coordination::ZooKeeperResponsePtr & response)
+            {
+                watch_promise->set_value(response);
+                [[maybe_unused]] int result = write(response_fd, &WATCH_RESPONSE_BYTE, sizeof(WATCH_RESPONSE_BYTE));
+            };
+            test_keeper_storage->putRequest(request, session_id, callback, watch_callback);
+            responses.push(promise->get_future());
+            watch_responses.emplace_back(watch_promise->get_future());
+        }
+        else
+        {
+            test_keeper_storage->putRequest(request, session_id, callback);
+            responses.push(promise->get_future());
+        }
     }
     else
     {
+        zkutil::ResponseCallback callback = [promise] (const Coordination::ZooKeeperResponsePtr & response)
+        {
+            promise->set_value(response);
+        };
         test_keeper_storage->putRequest(request, session_id, callback);
         responses.push(promise->get_future());
     }
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp
index 559121d6911..59b40ffdfbb 100644
--- a/src/Storages/AlterCommands.cpp
+++ b/src/Storages/AlterCommands.cpp
@@ -793,6 +793,7 @@ void AlterCommands::apply(StorageInMemoryMetadata & metadata, const Context & co
         throw DB::Exception("Alter commands is not prepared. Cannot apply. It's a bug", ErrorCodes::LOGICAL_ERROR);
 
     auto metadata_copy = metadata;
+
     for (const AlterCommand & command : *this)
         if (!command.ignore)
             command.apply(metadata_copy, context);
@@ -823,6 +824,7 @@ void AlterCommands::apply(StorageInMemoryMetadata & metadata, const Context & co
 
     /// Changes in columns may lead to changes in TTL expressions.
     auto column_ttl_asts = metadata_copy.columns.getColumnTTLs();
+    metadata_copy.column_ttls_by_name.clear();
     for (const auto & [name, ast] : column_ttl_asts)
     {
         auto new_ttl_entry = TTLDescription::getTTLFromAST(ast, metadata_copy.columns, context, metadata_copy.primary_key);
@@ -830,7 +832,7 @@ void AlterCommands::apply(StorageInMemoryMetadata & metadata, const Context & co
     }
 
     if (metadata_copy.table_ttl.definition_ast != nullptr)
-        metadata.table_ttl = TTLTableDescription::getTTLForTableFromAST(
+        metadata_copy.table_ttl = TTLTableDescription::getTTLForTableFromAST(
             metadata_copy.table_ttl.definition_ast, metadata_copy.columns, context, metadata_copy.primary_key);
 
     metadata = std::move(metadata_copy);
diff --git a/src/Storages/HDFS/HDFSCommon.cpp b/src/Storages/HDFS/HDFSCommon.cpp
new file mode 100644
index 00000000000..e5ec8a06139
--- /dev/null
+++ b/src/Storages/HDFS/HDFSCommon.cpp
@@ -0,0 +1,194 @@
+#include <Storages/HDFS/HDFSCommon.h>
+#include <Poco/URI.h>
+#include <boost/algorithm/string/replace.hpp>
+
+#if USE_HDFS
+#include <Common/ShellCommand.h>
+#include <Common/Exception.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/Operators.h>
+#include <common/logger_useful.h>
+
+namespace DB
+{
+namespace ErrorCodes
+{
+extern const int BAD_ARGUMENTS;
+extern const int NETWORK_ERROR;
+extern const int EXCESSIVE_ELEMENT_IN_CONFIG;
+extern const int NO_ELEMENTS_IN_CONFIG;
+}
+
+const String HDFSBuilderWrapper::CONFIG_PREFIX = "hdfs";
+
+void HDFSBuilderWrapper::loadFromConfig(const Poco::Util::AbstractConfiguration & config,
+    const String & config_path, bool isUser)
+{
+    Poco::Util::AbstractConfiguration::Keys keys;
+
+    config.keys(config_path, keys);
+    for (const auto & key : keys)
+    {
+        const String key_path = config_path + "." + key;
+
+        String key_name;
+        if (key == "hadoop_kerberos_keytab")
+        {
+            need_kinit = true;
+            hadoop_kerberos_keytab = config.getString(key_path);
+            continue;
+        }
+        else if (key == "hadoop_kerberos_principal")
+        {
+            need_kinit = true;
+            hadoop_kerberos_principal = config.getString(key_path);
+
+#if USE_INTERNAL_HDFS3_LIBRARY
+            hdfsBuilderSetPrincipal(hdfs_builder, hadoop_kerberos_principal.c_str());
+#endif
+
+            continue;
+        }
+        else if (key == "hadoop_kerberos_kinit_command")
+        {
+            need_kinit = true;
+            hadoop_kerberos_kinit_command = config.getString(key_path);
+            continue;
+        }
+        else if (key == "hadoop_security_kerberos_ticket_cache_path")
+        {
+            if (isUser)
+            {
+                throw Exception("hadoop.security.kerberos.ticket.cache.path cannot be set per user",
+                    ErrorCodes::EXCESSIVE_ELEMENT_IN_CONFIG);
+            }
+
+            hadoop_security_kerberos_ticket_cache_path = config.getString(key_path);
+            // standard param - pass further
+        }
+
+        key_name = boost::replace_all_copy(key, "_", ".");
+
+        const auto & [k,v] = keep(key_name, config.getString(key_path));
+        hdfsBuilderConfSetStr(hdfs_builder, k.c_str(), v.c_str());
+    }
+}
+
+String HDFSBuilderWrapper::getKinitCmd()
+{
+
+    if (hadoop_kerberos_keytab.empty() || hadoop_kerberos_principal.empty())
+    {
+        throw Exception("Not enough parameters to run kinit",
+            ErrorCodes::NO_ELEMENTS_IN_CONFIG);
+    }
+
+    WriteBufferFromOwnString ss;
+
+    String cache_name =  hadoop_security_kerberos_ticket_cache_path.empty() ?
+        String() :
+        (String(" -c \"") + hadoop_security_kerberos_ticket_cache_path + "\"");
+
+    // command to run looks like
+    // kinit -R -t /keytab_dir/clickhouse.keytab -k somebody@TEST.CLICKHOUSE.TECH || ..
+    ss << hadoop_kerberos_kinit_command << cache_name <<
+        " -R -t \"" << hadoop_kerberos_keytab << "\" -k " << hadoop_kerberos_principal <<
+        "|| " << hadoop_kerberos_kinit_command << cache_name << " -t \"" <<
+        hadoop_kerberos_keytab << "\" -k " << hadoop_kerberos_principal;
+    return ss.str();
+}
+
+void HDFSBuilderWrapper::runKinit()
+{
+    String cmd = getKinitCmd();
+    LOG_DEBUG(&Poco::Logger::get("HDFSClient"), "running kinit: {}", cmd);
+
+    std::unique_lock<std::mutex> lck(kinit_mtx);
+
+    auto command = ShellCommand::execute(cmd);
+    auto status = command->tryWait();
+    if (status)
+    {
+        throw Exception("kinit failure: " + cmd, ErrorCodes::BAD_ARGUMENTS);
+    }
+}
+
+HDFSBuilderWrapper createHDFSBuilder(const String & uri_str, const Poco::Util::AbstractConfiguration & config)
+{
+    const Poco::URI uri(uri_str);
+    const auto & host = uri.getHost();
+    auto port = uri.getPort();
+    const String path = "//";
+    if (host.empty())
+        throw Exception("Illegal HDFS URI: " + uri.toString(), ErrorCodes::BAD_ARGUMENTS);
+
+    HDFSBuilderWrapper builder;
+    if (builder.get() == nullptr)
+        throw Exception("Unable to create builder to connect to HDFS: " +
+            uri.toString() + " " + String(hdfsGetLastError()),
+            ErrorCodes::NETWORK_ERROR);
+
+    hdfsBuilderConfSetStr(builder.get(), "input.read.timeout", "60000"); // 1 min
+    hdfsBuilderConfSetStr(builder.get(), "input.write.timeout", "60000"); // 1 min
+    hdfsBuilderConfSetStr(builder.get(), "input.connect.timeout", "60000"); // 1 min
+
+    String user_info = uri.getUserInfo();
+    String user;
+    if (!user_info.empty() && user_info.front() != ':')
+    {
+        size_t delim_pos = user_info.find(':');
+        if (delim_pos != String::npos)
+            user = user_info.substr(0, delim_pos);
+        else
+            user = user_info;
+
+        hdfsBuilderSetUserName(builder.get(), user.c_str());
+    }
+    hdfsBuilderSetNameNode(builder.get(), host.c_str());
+    if (port != 0)
+    {
+        hdfsBuilderSetNameNodePort(builder.get(), port);
+    }
+
+    if (config.has(HDFSBuilderWrapper::CONFIG_PREFIX))
+    {
+        builder.loadFromConfig(config, HDFSBuilderWrapper::CONFIG_PREFIX);
+    }
+
+    if (!user.empty())
+    {
+        String user_config_prefix = HDFSBuilderWrapper::CONFIG_PREFIX + "_" + user;
+        if (config.has(user_config_prefix))
+        {
+#if USE_INTERNAL_HDFS3_LIBRARY
+            builder.loadFromConfig(config, user_config_prefix, true);
+#else
+            throw Exception("Multi user HDFS configuration required internal libhdfs3",
+                ErrorCodes::EXCESSIVE_ELEMENT_IN_CONFIG);
+#endif
+        }
+    }
+
+    if (builder.need_kinit)
+    {
+        builder.runKinit();
+    }
+
+    return builder;
+}
+
+std::mutex HDFSBuilderWrapper::kinit_mtx;
+
+HDFSFSPtr createHDFSFS(hdfsBuilder * builder)
+{
+    HDFSFSPtr fs(hdfsBuilderConnect(builder));
+    if (fs == nullptr)
+        throw Exception("Unable to connect to HDFS: " + String(hdfsGetLastError()),
+            ErrorCodes::NETWORK_ERROR);
+
+    return fs;
+}
+
+}
+
+#endif
diff --git a/src/Storages/HDFS/HDFSCommon.h b/src/Storages/HDFS/HDFSCommon.h
new file mode 100644
index 00000000000..fa1ca88464e
--- /dev/null
+++ b/src/Storages/HDFS/HDFSCommon.h
@@ -0,0 +1,114 @@
+#pragma once
+
+#include <Common/config.h>
+
+#if USE_HDFS
+#include <memory>
+#include <type_traits>
+#include <vector>
+
+#include <hdfs/hdfs.h>
+#include <common/types.h>
+#include <mutex>
+
+#include <Interpreters/Context.h>
+#include <Poco/Util/AbstractConfiguration.h>
+
+
+namespace DB
+{
+namespace detail
+{
+    struct HDFSFsDeleter
+    {
+        void operator()(hdfsFS fs_ptr)
+        {
+            hdfsDisconnect(fs_ptr);
+        }
+    };
+}
+
+struct HDFSFileInfo
+{
+    hdfsFileInfo * file_info;
+    int length;
+
+    HDFSFileInfo()
+        : file_info(nullptr)
+        , length(0)
+    {
+    }
+    HDFSFileInfo(const HDFSFileInfo & other) = delete;
+    HDFSFileInfo(HDFSFileInfo && other) = default;
+    HDFSFileInfo & operator=(const HDFSFileInfo & other) = delete;
+    HDFSFileInfo & operator=(HDFSFileInfo && other) = default;
+
+    ~HDFSFileInfo()
+    {
+        hdfsFreeFileInfo(file_info, length);
+    }
+};
+
+class HDFSBuilderWrapper
+{
+    hdfsBuilder * hdfs_builder;
+    String hadoop_kerberos_keytab;
+    String hadoop_kerberos_principal;
+    String hadoop_kerberos_kinit_command = "kinit";
+    String hadoop_security_kerberos_ticket_cache_path;
+
+    static std::mutex kinit_mtx;
+
+    std::vector<std::pair<String, String>> config_stor;
+
+    // hdfs builder relies on an external config data storage
+    std::pair<String, String>& keep(const String & k, const String & v)
+    {
+        return config_stor.emplace_back(std::make_pair(k, v));
+    }
+
+    bool need_kinit{false};
+
+    static const String CONFIG_PREFIX;
+
+private:
+
+    void loadFromConfig(const Poco::Util::AbstractConfiguration & config, const String & config_path, bool isUser = false);
+
+    String getKinitCmd();
+
+    void runKinit();
+
+public:
+
+    hdfsBuilder *
+    get()
+    {
+        return hdfs_builder;
+    }
+
+    HDFSBuilderWrapper()
+        : hdfs_builder(hdfsNewBuilder())
+    {
+    }
+
+    ~HDFSBuilderWrapper()
+    {
+        hdfsFreeBuilder(hdfs_builder);
+
+    }
+
+    HDFSBuilderWrapper(const HDFSBuilderWrapper &) = delete;
+    HDFSBuilderWrapper(HDFSBuilderWrapper &&) = default;
+
+    friend HDFSBuilderWrapper createHDFSBuilder(const String & uri_str, const Poco::Util::AbstractConfiguration &);
+};
+
+using HDFSFSPtr = std::unique_ptr<std::remove_pointer_t<hdfsFS>, detail::HDFSFsDeleter>;
+
+// set read/connect timeout, default value in libhdfs3 is about 1 hour, and too large
+/// TODO Allow to tune from query Settings.
+HDFSBuilderWrapper createHDFSBuilder(const String & uri_str, const Poco::Util::AbstractConfiguration &);
+HDFSFSPtr createHDFSFS(hdfsBuilder * builder);
+}
+#endif
diff --git a/src/IO/ReadBufferFromHDFS.cpp b/src/Storages/HDFS/ReadBufferFromHDFS.cpp
similarity index 81%
rename from src/IO/ReadBufferFromHDFS.cpp
rename to src/Storages/HDFS/ReadBufferFromHDFS.cpp
index 0b9f75d8c46..f3b0e3022f1 100644
--- a/src/IO/ReadBufferFromHDFS.cpp
+++ b/src/Storages/HDFS/ReadBufferFromHDFS.cpp
@@ -1,7 +1,7 @@
 #include "ReadBufferFromHDFS.h"
 
 #if USE_HDFS
-#include <IO/HDFSCommon.h>
+#include <Storages/HDFS/HDFSCommon.h>
 #include <hdfs/hdfs.h>
 #include <mutex>
 
@@ -23,15 +23,16 @@ struct ReadBufferFromHDFS::ReadBufferFromHDFSImpl
 
     std::string hdfs_uri;
     hdfsFile fin;
-    HDFSBuilderPtr builder;
+    HDFSBuilderWrapper builder;
     HDFSFSPtr fs;
 
-    explicit ReadBufferFromHDFSImpl(const std::string & hdfs_name_)
-        : hdfs_uri(hdfs_name_)
+    explicit ReadBufferFromHDFSImpl(const std::string & hdfs_name_,
+        const Poco::Util::AbstractConfiguration & config_)
+        : hdfs_uri(hdfs_name_),
+          builder(createHDFSBuilder(hdfs_uri, config_))
     {
         std::lock_guard lock(hdfs_init_mutex);
 
-        builder = createHDFSBuilder(hdfs_uri);
         fs = createHDFSFS(builder.get());
         const size_t begin_of_path = hdfs_uri.find('/', hdfs_uri.find("//") + 2);
         const std::string path = hdfs_uri.substr(begin_of_path);
@@ -58,11 +59,14 @@ struct ReadBufferFromHDFS::ReadBufferFromHDFSImpl
     }
 };
 
+
 std::mutex ReadBufferFromHDFS::ReadBufferFromHDFSImpl::hdfs_init_mutex;
 
-ReadBufferFromHDFS::ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size)
-    : BufferWithOwnMemory<ReadBuffer>(buf_size)
-    , impl(std::make_unique<ReadBufferFromHDFSImpl>(hdfs_name_))
+ReadBufferFromHDFS::ReadBufferFromHDFS(const std::string & hdfs_name_,
+    const Poco::Util::AbstractConfiguration & config_,
+    size_t buf_size_)
+    : BufferWithOwnMemory<ReadBuffer>(buf_size_)
+    , impl(std::make_unique<ReadBufferFromHDFSImpl>(hdfs_name_, config_))
 {
 }
 
diff --git a/src/IO/ReadBufferFromHDFS.h b/src/Storages/HDFS/ReadBufferFromHDFS.h
similarity index 69%
rename from src/IO/ReadBufferFromHDFS.h
rename to src/Storages/HDFS/ReadBufferFromHDFS.h
index 58479d45660..8d26c001b2e 100644
--- a/src/IO/ReadBufferFromHDFS.h
+++ b/src/Storages/HDFS/ReadBufferFromHDFS.h
@@ -8,6 +8,13 @@
 #include <string>
 #include <memory>
 
+#include <hdfs/hdfs.h>
+
+#include <common/types.h>
+
+#include <Interpreters/Context.h>
+
+
 namespace DB
 {
 /** Accepts HDFS path to file and opens it.
@@ -18,7 +25,7 @@ class ReadBufferFromHDFS : public BufferWithOwnMemory<ReadBuffer>
     struct ReadBufferFromHDFSImpl;
     std::unique_ptr<ReadBufferFromHDFSImpl> impl;
 public:
-    ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE);
+    ReadBufferFromHDFS(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration &, size_t buf_size_ = DBMS_DEFAULT_BUFFER_SIZE);
     ~ReadBufferFromHDFS() override;
 
     bool nextImpl() override;
diff --git a/src/Storages/StorageHDFS.cpp b/src/Storages/HDFS/StorageHDFS.cpp
similarity index 95%
rename from src/Storages/StorageHDFS.cpp
rename to src/Storages/HDFS/StorageHDFS.cpp
index 0a4729b1c06..f50b4722ce1 100644
--- a/src/Storages/StorageHDFS.cpp
+++ b/src/Storages/HDFS/StorageHDFS.cpp
@@ -3,15 +3,15 @@
 #if USE_HDFS
 
 #include <Storages/StorageFactory.h>
-#include <Storages/StorageHDFS.h>
+#include <Storages/HDFS/StorageHDFS.h>
 #include <Interpreters/Context.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Parsers/ASTLiteral.h>
 #include <IO/ReadHelpers.h>
-#include <IO/ReadBufferFromHDFS.h>
-#include <IO/WriteBufferFromHDFS.h>
+#include <Storages/HDFS/ReadBufferFromHDFS.h>
+#include <Storages/HDFS/WriteBufferFromHDFS.h>
 #include <IO/WriteHelpers.h>
-#include <IO/HDFSCommon.h>
+#include <Storages/HDFS/HDFSCommon.h>
 #include <Formats/FormatFactory.h>
 #include <DataTypes/DataTypeString.h>
 #include <DataStreams/IBlockOutputStream.h>
@@ -121,7 +121,7 @@ public:
                 current_path = uri + path;
 
                 auto compression = chooseCompressionMethod(path, compression_method);
-                auto read_buf = wrapReadBufferWithCompressionMethod(std::make_unique<ReadBufferFromHDFS>(current_path), compression);
+                auto read_buf = wrapReadBufferWithCompressionMethod(std::make_unique<ReadBufferFromHDFS>(current_path, context.getGlobalContext().getConfigRef()), compression);
                 auto input_stream = FormatFactory::instance().getInput(format, *read_buf, sample_block, context, max_block_size);
 
                 reader = std::make_shared<OwningBlockInputStream<ReadBuffer>>(input_stream, std::move(read_buf));
@@ -180,7 +180,7 @@ public:
         const CompressionMethod compression_method)
         : sample_block(sample_block_)
     {
-        write_buf = wrapWriteBufferWithCompressionMethod(std::make_unique<WriteBufferFromHDFS>(uri), compression_method, 3);
+        write_buf = wrapWriteBufferWithCompressionMethod(std::make_unique<WriteBufferFromHDFS>(uri, context.getGlobalContext().getConfigRef()), compression_method, 3);
         writer = FormatFactory::instance().getOutput(format, *write_buf, sample_block, context);
     }
 
@@ -274,7 +274,7 @@ Pipe StorageHDFS::read(
     const String path_from_uri = uri.substr(begin_of_path);
     const String uri_without_path = uri.substr(0, begin_of_path);
 
-    HDFSBuilderPtr builder = createHDFSBuilder(uri_without_path + "/");
+    HDFSBuilderWrapper builder = createHDFSBuilder(uri_without_path + "/", context_.getGlobalContext().getConfigRef());
     HDFSFSPtr fs = createHDFSFS(builder.get());
 
     auto sources_info = std::make_shared<HDFSSource::SourcesInfo>();
diff --git a/src/Storages/StorageHDFS.h b/src/Storages/HDFS/StorageHDFS.h
similarity index 100%
rename from src/Storages/StorageHDFS.h
rename to src/Storages/HDFS/StorageHDFS.h
diff --git a/src/IO/WriteBufferFromHDFS.cpp b/src/Storages/HDFS/WriteBufferFromHDFS.cpp
similarity index 84%
rename from src/IO/WriteBufferFromHDFS.cpp
rename to src/Storages/HDFS/WriteBufferFromHDFS.cpp
index 2176c77b914..3873eace10d 100644
--- a/src/IO/WriteBufferFromHDFS.cpp
+++ b/src/Storages/HDFS/WriteBufferFromHDFS.cpp
@@ -2,8 +2,9 @@
 
 #if USE_HDFS
 
-#include <IO/WriteBufferFromHDFS.h>
-#include <IO/HDFSCommon.h>
+#include <Interpreters/Context.h>
+#include <Storages/HDFS/WriteBufferFromHDFS.h>
+#include <Storages/HDFS/HDFSCommon.h>
 #include <hdfs/hdfs.h>
 
 
@@ -23,12 +24,12 @@ struct WriteBufferFromHDFS::WriteBufferFromHDFSImpl
 {
     std::string hdfs_uri;
     hdfsFile fout;
-    HDFSBuilderPtr builder;
+    HDFSBuilderWrapper builder;
     HDFSFSPtr fs;
 
-    explicit WriteBufferFromHDFSImpl(const std::string & hdfs_name_)
+    explicit WriteBufferFromHDFSImpl(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration & config_)
         : hdfs_uri(hdfs_name_)
-        , builder(createHDFSBuilder(hdfs_uri))
+        , builder(createHDFSBuilder(hdfs_uri,config_))
         , fs(createHDFSFS(builder.get()))
     {
         const size_t begin_of_path = hdfs_uri.find('/', hdfs_uri.find("//") + 2);
@@ -72,9 +73,9 @@ struct WriteBufferFromHDFS::WriteBufferFromHDFSImpl
     }
 };
 
-WriteBufferFromHDFS::WriteBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size)
-    : BufferWithOwnMemory<WriteBuffer>(buf_size)
-    , impl(std::make_unique<WriteBufferFromHDFSImpl>(hdfs_name_))
+WriteBufferFromHDFS::WriteBufferFromHDFS(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration & config_, size_t buf_size_)
+    : BufferWithOwnMemory<WriteBuffer>(buf_size_)
+    , impl(std::make_unique<WriteBufferFromHDFSImpl>(hdfs_name_, config_))
 {
 }
 
diff --git a/src/IO/WriteBufferFromHDFS.h b/src/Storages/HDFS/WriteBufferFromHDFS.h
similarity index 80%
rename from src/IO/WriteBufferFromHDFS.h
rename to src/Storages/HDFS/WriteBufferFromHDFS.h
index 22eab5c4d5d..339ec0c2a7c 100644
--- a/src/IO/WriteBufferFromHDFS.h
+++ b/src/Storages/HDFS/WriteBufferFromHDFS.h
@@ -18,7 +18,7 @@ class WriteBufferFromHDFS : public BufferWithOwnMemory<WriteBuffer>
     struct WriteBufferFromHDFSImpl;
     std::unique_ptr<WriteBufferFromHDFSImpl> impl;
 public:
-    WriteBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE);
+    WriteBufferFromHDFS(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration &, size_t buf_size_ = DBMS_DEFAULT_BUFFER_SIZE);
     WriteBufferFromHDFS(WriteBufferFromHDFS &&) = default;
 
     void nextImpl() override;
diff --git a/src/Storages/Kafka/KafkaSettings.h b/src/Storages/Kafka/KafkaSettings.h
index 53fbe8adc6b..1df10d16339 100644
--- a/src/Storages/Kafka/KafkaSettings.h
+++ b/src/Storages/Kafka/KafkaSettings.h
@@ -20,7 +20,7 @@ class ASTStorage;
     M(Milliseconds, kafka_poll_timeout_ms, 0, "Timeout for single poll from Kafka.", 0) \
     /* default is min(max_block_size, kafka_max_block_size)*/ \
     M(UInt64, kafka_poll_max_batch_size, 0, "Maximum amount of messages to be polled in a single Kafka poll.", 0) \
-    /* default is = min_insert_block_size / kafka_num_consumers  */ \
+    /* default is = max_insert_block_size / kafka_num_consumers  */ \
     M(UInt64, kafka_max_block_size, 0, "Number of row collected by poll(s) for flushing data from Kafka.", 0) \
     /* default is stream_flush_interval_ms */ \
     M(Milliseconds, kafka_flush_interval_ms, 0, "Timeout for flushing data from Kafka.", 0) \
diff --git a/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp b/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
index 143a9920b93..df905215df1 100644
--- a/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
+++ b/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
@@ -3,33 +3,53 @@
 namespace DB
 {
 
-IMergeTreeDataPartWriter::IMergeTreeDataPartWriter(
-    const MergeTreeData::DataPartPtr & data_part_,
-    const NamesAndTypesList & columns_list_,
-    const StorageMetadataPtr & metadata_snapshot_,
-    const MergeTreeWriterSettings & settings_)
-    : data_part(data_part_)
-    , storage(data_part_->storage)
-    , metadata_snapshot(metadata_snapshot_)
-    , columns_list(columns_list_)
-    , settings(settings_)
-    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity){}
+Block getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation)
+{
+    Block result;
+    for (size_t i = 0, size = names.size(); i < size; ++i)
+    {
+        const auto & name = names[i];
+        result.insert(i, block.getByName(name));
+
+        /// Reorder primary key columns in advance and add them to `primary_key_columns`.
+        if (permutation)
+        {
+            auto & column = result.getByPosition(i);
+            column.column = column.column->permute(*permutation, 0);
+        }
+    }
+
+    return result;
+}
+
+Block permuteBlockIfNeeded(const Block & block, const IColumn::Permutation * permutation)
+{
+    Block result;
+    for (size_t i = 0; i < block.columns(); ++i)
+    {
+        result.insert(i, block.getByPosition(i));
+        if (permutation)
+        {
+            auto & column = result.getByPosition(i);
+            column.column = column.column->permute(*permutation, 0);
+        }
+    }
+    return result;
+}
 
 IMergeTreeDataPartWriter::IMergeTreeDataPartWriter(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
     const StorageMetadataPtr & metadata_snapshot_,
-    const MergeTreeIndices & skip_indices_,
-    const MergeTreeIndexGranularity & index_granularity_,
-    const MergeTreeWriterSettings & settings_)
+    const MergeTreeWriterSettings & settings_,
+    const MergeTreeIndexGranularity & index_granularity_)
     : data_part(data_part_)
     , storage(data_part_->storage)
     , metadata_snapshot(metadata_snapshot_)
     , columns_list(columns_list_)
-    , skip_indices(skip_indices_)
-    , index_granularity(index_granularity_)
     , settings(settings_)
-    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity) {}
+    , index_granularity(index_granularity_)
+    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity){}
 
 Columns IMergeTreeDataPartWriter::releaseIndexColumns()
 {
@@ -38,12 +58,6 @@ Columns IMergeTreeDataPartWriter::releaseIndexColumns()
         std::make_move_iterator(index_columns.end()));
 }
 
-void IMergeTreeDataPartWriter::next()
-{
-    current_mark = next_mark;
-    index_offset = next_index_offset;
-}
-
 IMergeTreeDataPartWriter::~IMergeTreeDataPartWriter() = default;
 
 }
diff --git a/src/Storages/MergeTree/IMergeTreeDataPartWriter.h b/src/Storages/MergeTree/IMergeTreeDataPartWriter.h
index 4a42a58a65b..5e12d5da678 100644
--- a/src/Storages/MergeTree/IMergeTreeDataPartWriter.h
+++ b/src/Storages/MergeTree/IMergeTreeDataPartWriter.h
@@ -13,6 +13,9 @@
 namespace DB
 {
 
+Block getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation);
+
+Block permuteBlockIfNeeded(const Block & block, const IColumn::Permutation * permutation);
 
 /// Writes data part to disk in different formats.
 /// Calculates and serializes primary and skip indices if needed.
@@ -23,68 +26,29 @@ public:
         const MergeTreeData::DataPartPtr & data_part_,
         const NamesAndTypesList & columns_list_,
         const StorageMetadataPtr & metadata_snapshot_,
-        const MergeTreeWriterSettings & settings_);
-
-    IMergeTreeDataPartWriter(
-        const MergeTreeData::DataPartPtr & data_part_,
-        const NamesAndTypesList & columns_list_,
-        const StorageMetadataPtr & metadata_snapshot_,
-        const MergeTreeIndices & skip_indices_,
-        const MergeTreeIndexGranularity & index_granularity_,
-        const MergeTreeWriterSettings & settings_);
+        const MergeTreeWriterSettings & settings_,
+        const MergeTreeIndexGranularity & index_granularity_ = {});
 
     virtual ~IMergeTreeDataPartWriter();
 
-    virtual void write(
-        const Block & block, const IColumn::Permutation * permutation = nullptr,
-        /* Blocks with already sorted index columns */
-        const Block & primary_key_block = {}, const Block & skip_indexes_block = {}) = 0;
+    virtual void write(const Block & block, const IColumn::Permutation * permutation) = 0;
 
-    virtual void calculateAndSerializePrimaryIndex(const Block & /* primary_index_block */) {}
-    virtual void calculateAndSerializeSkipIndices(const Block & /* skip_indexes_block */) {}
-
-    /// Shift mark and offset to prepare read next mark.
-    /// You must call it after calling write method and optionally
-    ///  calling calculations of primary and skip indices.
-    void next();
-
-    virtual void initSkipIndices() {}
-    virtual void initPrimaryIndex() {}
-
-    virtual void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) = 0;
-    virtual void finishPrimaryIndexSerialization(MergeTreeData::DataPart::Checksums & /* checksums */, bool /* sync */) {}
-    virtual void finishSkipIndicesSerialization(MergeTreeData::DataPart::Checksums & /* checksums */, bool /* sync */) {}
+    virtual void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) = 0;
 
     Columns releaseIndexColumns();
     const MergeTreeIndexGranularity & getIndexGranularity() const { return index_granularity; }
-    const MergeTreeIndices & getSkipIndices() { return skip_indices; }
 
 protected:
-    size_t getCurrentMark() const { return current_mark; }
-    size_t getIndexOffset() const { return index_offset; }
 
-    using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
-    using SerializationStates = std::unordered_map<String, SerializationState>;
-
-    MergeTreeData::DataPartPtr data_part;
+    const MergeTreeData::DataPartPtr data_part;
     const MergeTreeData & storage;
-    StorageMetadataPtr metadata_snapshot;
-    NamesAndTypesList columns_list;
-    MergeTreeIndices skip_indices;
+    const StorageMetadataPtr metadata_snapshot;
+    const NamesAndTypesList columns_list;
+    const MergeTreeWriterSettings settings;
     MergeTreeIndexGranularity index_granularity;
-    MergeTreeWriterSettings settings;
-    bool with_final_mark;
-
-    size_t next_mark = 0;
-    size_t next_index_offset = 0;
+    const bool with_final_mark;
 
     MutableColumns index_columns;
-
-private:
-    /// Data is already written up to this mark.
-    size_t current_mark = 0;
-    /// The offset to the first row of the block for which you want to write the index.
-    size_t index_offset = 0;
 };
 
 }
diff --git a/src/Storages/MergeTree/IMergedBlockOutputStream.cpp b/src/Storages/MergeTree/IMergedBlockOutputStream.cpp
index e5cc6d6ecbf..2a52d7d53f0 100644
--- a/src/Storages/MergeTree/IMergedBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/IMergedBlockOutputStream.cpp
@@ -15,25 +15,6 @@ IMergedBlockOutputStream::IMergedBlockOutputStream(
 {
 }
 
-Block IMergedBlockOutputStream::getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation)
-{
-    Block result;
-    for (size_t i = 0, size = names.size(); i < size; ++i)
-    {
-        const auto & name = names[i];
-        result.insert(i, block.getByName(name));
-
-        /// Reorder primary key columns in advance and add them to `primary_key_columns`.
-        if (permutation)
-        {
-            auto & column = result.getByPosition(i);
-            column.column = column.column->permute(*permutation, 0);
-        }
-    }
-
-    return result;
-}
-
 NameSet IMergedBlockOutputStream::removeEmptyColumnsFromPart(
     const MergeTreeDataPartPtr & data_part,
     NamesAndTypesList & columns,
diff --git a/src/Storages/MergeTree/IMergedBlockOutputStream.h b/src/Storages/MergeTree/IMergedBlockOutputStream.h
index 7ec9f85ff28..ed8da4d334b 100644
--- a/src/Storages/MergeTree/IMergedBlockOutputStream.h
+++ b/src/Storages/MergeTree/IMergedBlockOutputStream.h
@@ -42,8 +42,6 @@ protected:
     VolumePtr volume;
     String part_path;
 
-    static Block getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation);
-
     IMergeTreeDataPart::MergeTreeWriterPtr writer;
 };
 
diff --git a/src/Storages/MergeTree/KeyCondition.cpp b/src/Storages/MergeTree/KeyCondition.cpp
index 5903dc89e9b..ead93a6e1d9 100644
--- a/src/Storages/MergeTree/KeyCondition.cpp
+++ b/src/Storages/MergeTree/KeyCondition.cpp
@@ -1135,7 +1135,7 @@ bool KeyCondition::tryParseAtomFromAST(const ASTPtr & node, const Context & cont
                         ColumnsWithTypeAndName arguments{
                             {nullptr, key_expr_type, ""}, {DataTypeString().createColumnConst(1, common_type->getName()), common_type, ""}};
                         FunctionOverloadResolverPtr func_builder_cast
-                                = std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver::createImpl(false));
+                                = std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver<CastType::nonAccurate>::createImpl(false));
                         auto func_cast = func_builder_cast->build(arguments);
 
                         /// If we know the given range only contains one value, then we treat all functions as positive monotonic.
diff --git a/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp b/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp
index 40714e5af31..904081cc1df 100644
--- a/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp
@@ -13,16 +13,22 @@ Block MergeTreeBlockOutputStream::getHeader() const
 }
 
 
+void MergeTreeBlockOutputStream::writePrefix()
+{
+    /// Only check "too many parts" before write,
+    /// because interrupting long-running INSERT query in the middle is not convenient for users.
+    storage.delayInsertOrThrowIfNeeded();
+}
+
+
 void MergeTreeBlockOutputStream::write(const Block & block)
 {
-    storage.delayInsertOrThrowIfNeeded();
-
     auto part_blocks = storage.writer.splitBlockIntoParts(block, max_parts_per_block, metadata_snapshot);
     for (auto & current_block : part_blocks)
     {
         Stopwatch watch;
 
-        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot);
+        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot, optimize_on_insert);
         storage.renameTempPartAndAdd(part, &storage.increment);
 
         PartLog::addNewPart(storage.global_context, part, watch.elapsed());
diff --git a/src/Storages/MergeTree/MergeTreeBlockOutputStream.h b/src/Storages/MergeTree/MergeTreeBlockOutputStream.h
index 8aae7f3e625..8caa53382dc 100644
--- a/src/Storages/MergeTree/MergeTreeBlockOutputStream.h
+++ b/src/Storages/MergeTree/MergeTreeBlockOutputStream.h
@@ -14,20 +14,23 @@ class StorageMergeTree;
 class MergeTreeBlockOutputStream : public IBlockOutputStream
 {
 public:
-    MergeTreeBlockOutputStream(StorageMergeTree & storage_, const StorageMetadataPtr metadata_snapshot_, size_t max_parts_per_block_)
+    MergeTreeBlockOutputStream(StorageMergeTree & storage_, const StorageMetadataPtr metadata_snapshot_, size_t max_parts_per_block_, bool optimize_on_insert_)
         : storage(storage_)
         , metadata_snapshot(metadata_snapshot_)
         , max_parts_per_block(max_parts_per_block_)
+        , optimize_on_insert(optimize_on_insert_)
     {
     }
 
     Block getHeader() const override;
     void write(const Block & block) override;
+    void writePrefix() override;
 
 private:
     StorageMergeTree & storage;
     StorageMetadataPtr metadata_snapshot;
     size_t max_parts_per_block;
+    bool optimize_on_insert;
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeData.cpp b/src/Storages/MergeTree/MergeTreeData.cpp
index c3a599665bb..9fa19859c7f 100644
--- a/src/Storages/MergeTree/MergeTreeData.cpp
+++ b/src/Storages/MergeTree/MergeTreeData.cpp
@@ -2380,8 +2380,6 @@ void MergeTreeData::delayInsertOrThrowIfNeeded(Poco::Event * until) const
     }
 
     const size_t parts_count_in_partition = getMaxPartsCountForPartition();
-    if (parts_count_in_partition < settings->parts_to_delay_insert)
-        return;
 
     if (parts_count_in_partition >= settings->parts_to_throw_insert)
     {
@@ -2389,6 +2387,9 @@ void MergeTreeData::delayInsertOrThrowIfNeeded(Poco::Event * until) const
         throw Exception("Too many parts (" + toString(parts_count_in_partition) + "). Merges are processing significantly slower than inserts.", ErrorCodes::TOO_MANY_PARTS);
     }
 
+    if (parts_count_in_partition < settings->parts_to_delay_insert)
+        return;
+
     const size_t max_k = settings->parts_to_throw_insert - settings->parts_to_delay_insert; /// always > 0
     const size_t k = 1 + parts_count_in_partition - settings->parts_to_delay_insert; /// from 1 to max_k
     const double delay_milliseconds = ::pow(settings->max_delay_to_insert * 1000, static_cast<double>(k) / max_k);
@@ -2406,24 +2407,6 @@ void MergeTreeData::delayInsertOrThrowIfNeeded(Poco::Event * until) const
         std::this_thread::sleep_for(std::chrono::milliseconds(static_cast<size_t>(delay_milliseconds)));
 }
 
-void MergeTreeData::throwInsertIfNeeded() const
-{
-    const auto settings = getSettings();
-    const size_t parts_count_in_total = getPartsCount();
-    if (parts_count_in_total >= settings->max_parts_in_total)
-    {
-        ProfileEvents::increment(ProfileEvents::RejectedInserts);
-        throw Exception("Too many parts (" + toString(parts_count_in_total) + ") in all partitions in total. This indicates wrong choice of partition key. The threshold can be modified with 'max_parts_in_total' setting in <merge_tree> element in config.xml or with per-table setting.", ErrorCodes::TOO_MANY_PARTS);
-    }
-
-    const size_t parts_count_in_partition = getMaxPartsCountForPartition();
-
-    if (parts_count_in_partition >= settings->parts_to_throw_insert)
-    {
-        ProfileEvents::increment(ProfileEvents::RejectedInserts);
-        throw Exception("Too many parts (" + toString(parts_count_in_partition) + "). Merges are processing significantly slower than inserts.", ErrorCodes::TOO_MANY_PARTS);
-    }
-}
 
 MergeTreeData::DataPartPtr MergeTreeData::getActiveContainingPart(
     const MergeTreePartInfo & part_info, MergeTreeData::DataPartState state, DataPartsLock & /*lock*/) const
@@ -2451,6 +2434,7 @@ MergeTreeData::DataPartPtr MergeTreeData::getActiveContainingPart(
     return nullptr;
 }
 
+
 void MergeTreeData::swapActivePart(MergeTreeData::DataPartPtr part_copy)
 {
     auto lock = lockParts();
diff --git a/src/Storages/MergeTree/MergeTreeData.h b/src/Storages/MergeTree/MergeTreeData.h
index 38e6c34e232..382f53f40a4 100644
--- a/src/Storages/MergeTree/MergeTreeData.h
+++ b/src/Storages/MergeTree/MergeTreeData.h
@@ -423,7 +423,6 @@ public:
     /// If the table contains too many active parts, sleep for a while to give them time to merge.
     /// If until is non-null, wake up from the sleep earlier if the event happened.
     void delayInsertOrThrowIfNeeded(Poco::Event * until = nullptr) const;
-    void throwInsertIfNeeded() const;
 
     /// Renames temporary part to a permanent part and adds it to the parts set.
     /// It is assumed that the part does not intersect with existing parts.
diff --git a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
index 1065b992396..e375e653f8f 100644
--- a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
@@ -235,6 +235,7 @@ SelectPartsDecision MergeTreeDataMergerMutator::selectPartsToMerge(
     /// Previous part only in boundaries of partition frame
     const MergeTreeData::DataPartPtr * prev_part = nullptr;
 
+    size_t parts_selected_precondition = 0;
     for (const MergeTreeData::DataPartPtr & part : data_parts)
     {
         const String & partition_id = part->info.partition_id;
@@ -282,6 +283,8 @@ SelectPartsDecision MergeTreeDataMergerMutator::selectPartsToMerge(
         part_info.compression_codec_desc = part->default_codec->getFullCodecDesc();
         part_info.shall_participate_in_merges = has_volumes_with_disabled_merges ? part->shallParticipateInMerges(storage_policy) : true;
 
+        ++parts_selected_precondition;
+
         parts_ranges.back().emplace_back(part_info);
 
         /// Check for consistency of data parts. If assertion is failed, it requires immediate investigation.
@@ -294,6 +297,13 @@ SelectPartsDecision MergeTreeDataMergerMutator::selectPartsToMerge(
         prev_part = &part;
     }
 
+    if (parts_selected_precondition == 0)
+    {
+        if (out_disable_reason)
+            *out_disable_reason = "No parts satisfy preconditions for merge";
+        return SelectPartsDecision::CANNOT_SELECT;
+    }
+
     IMergeSelector::PartsRange parts_to_merge;
 
     if (metadata_snapshot->hasAnyTTL() && merge_with_ttl_allowed && !ttl_merges_blocker.isCancelled())
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
index 0b22f1271e3..389d2950f65 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
@@ -4,6 +4,11 @@
 namespace DB
 {
 
+namespace ErrorCodes
+{
+    extern const int LOGICAL_ERROR;
+}
+
 MergeTreeDataPartWriterCompact::MergeTreeDataPartWriterCompact(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
@@ -64,9 +69,64 @@ void MergeTreeDataPartWriterCompact::addStreams(const String & name, const IData
     type.enumerateStreams(callback, stream_path);
 }
 
-void MergeTreeDataPartWriterCompact::write(
-    const Block & block, const IColumn::Permutation * permutation,
-    const Block & primary_key_block, const Block & skip_indexes_block)
+namespace
+{
+
+/// Get granules for block using index_granularity
+Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity, size_t block_rows, size_t current_mark, bool last_block)
+{
+    if (current_mark >= index_granularity.getMarksCount())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Request to get granules from mark {} but index granularity size is {}", current_mark, index_granularity.getMarksCount());
+
+    Granules result;
+    size_t current_row = 0;
+    while (current_row < block_rows)
+    {
+        size_t expected_rows_in_mark = index_granularity.getMarkRows(current_mark);
+        size_t rows_left_in_block = block_rows - current_row;
+        if (rows_left_in_block < expected_rows_in_mark && !last_block)
+        {
+            /// Invariant: we always have equal amount of rows for block in compact parts because we accumulate them in buffer.
+            /// The only exclusion is the last block, when we cannot accumulate more rows.
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Required to write {} rows, but only {} rows was written for the non last granule", expected_rows_in_mark, rows_left_in_block);
+        }
+
+        result.emplace_back(Granule{
+            .start_row = current_row,
+            .granularity_rows = expected_rows_in_mark,
+            .block_rows = std::min(rows_left_in_block, expected_rows_in_mark),
+            .mark_number = current_mark,
+            .mark_on_start = true
+        });
+        current_row += expected_rows_in_mark;
+        current_mark++;
+    }
+
+    return result;
+}
+
+/// Write single granule of one column (rows between 2 marks)
+void writeColumnSingleGranule(
+    const ColumnWithTypeAndName & column,
+    IDataType::OutputStreamGetter stream_getter,
+    size_t from_row,
+    size_t number_of_rows)
+{
+    IDataType::SerializeBinaryBulkStatePtr state;
+    IDataType::SerializeBinaryBulkSettings serialize_settings;
+
+    serialize_settings.getter = stream_getter;
+    serialize_settings.position_independent_encoding = true;
+    serialize_settings.low_cardinality_max_dictionary_size = 0;
+
+    column.type->serializeBinaryBulkStatePrefix(serialize_settings, state);
+    column.type->serializeBinaryBulkWithMultipleStreams(*column.column, from_row, number_of_rows, serialize_settings, state);
+    column.type->serializeBinaryBulkStateSuffix(serialize_settings, state);
+}
+
+}
+
+void MergeTreeDataPartWriterCompact::write(const Block & block, const IColumn::Permutation * permutation)
 {
     /// Fill index granularity for this block
     /// if it's unknown (in case of insert data or horizontal merge,
@@ -77,58 +137,43 @@ void MergeTreeDataPartWriterCompact::write(
         fillIndexGranularity(index_granularity_for_block, block.rows());
     }
 
-    Block result_block;
-
-    if (permutation)
-    {
-        for (const auto & it : columns_list)
-        {
-            if (primary_key_block.has(it.name))
-                result_block.insert(primary_key_block.getByName(it.name));
-            else if (skip_indexes_block.has(it.name))
-                result_block.insert(skip_indexes_block.getByName(it.name));
-            else
-            {
-                auto column = block.getByName(it.name);
-                column.column = column.column->permute(*permutation, 0);
-                result_block.insert(column);
-            }
-        }
-    }
-    else
-    {
-        result_block = block;
-    }
+    Block result_block = permuteBlockIfNeeded(block, permutation);
 
     if (!header)
         header = result_block.cloneEmpty();
 
     columns_buffer.add(result_block.mutateColumns());
-    size_t last_mark_rows = index_granularity.getLastMarkRows();
+    size_t current_mark_rows = index_granularity.getMarkRows(getCurrentMark());
     size_t rows_in_buffer = columns_buffer.size();
 
-    if (rows_in_buffer < last_mark_rows)
+    if (rows_in_buffer >= current_mark_rows)
     {
-        /// If it's not enough rows for granule, accumulate blocks
-        ///  and save how much rows we already have.
-        next_index_offset = last_mark_rows - rows_in_buffer;
-        return;
+        Block flushed_block = header.cloneWithColumns(columns_buffer.releaseColumns());
+        auto granules_to_write = getGranulesToWrite(index_granularity, flushed_block.rows(), getCurrentMark(), /* last_block = */ false);
+        writeDataBlockPrimaryIndexAndSkipIndices(flushed_block, granules_to_write);
+        setCurrentMark(getCurrentMark() + granules_to_write.size());
     }
-
-    writeBlock(header.cloneWithColumns(columns_buffer.releaseColumns()));
 }
 
-void MergeTreeDataPartWriterCompact::writeBlock(const Block & block)
+void MergeTreeDataPartWriterCompact::writeDataBlockPrimaryIndexAndSkipIndices(const Block & block, const Granules & granules_to_write)
 {
-    size_t total_rows = block.rows();
-    size_t from_mark = getCurrentMark();
-    size_t current_row = 0;
+    writeDataBlock(block, granules_to_write);
 
-    while (current_row < total_rows)
+    if (settings.rewrite_primary_key)
     {
-        size_t rows_to_write = index_granularity.getMarkRows(from_mark);
+        Block primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), nullptr);
+        calculateAndSerializePrimaryIndex(primary_key_block, granules_to_write);
+    }
 
-        if (rows_to_write)
+    Block skip_indices_block = getBlockAndPermute(block, getSkipIndicesColumns(), nullptr);
+    calculateAndSerializeSkipIndices(skip_indices_block, granules_to_write);
+}
+
+void MergeTreeDataPartWriterCompact::writeDataBlock(const Block & block, const Granules & granules)
+{
+    for (const auto & granule : granules)
+    {
+        if (granule.granularity_rows)
             data_written = true;
 
         auto name_and_type = columns_list.begin();
@@ -161,53 +206,30 @@ void MergeTreeDataPartWriterCompact::writeBlock(const Block & block)
             writeIntBinary(plain_hashing.count(), marks);
             writeIntBinary(UInt64(0), marks);
 
-            writeColumnSingleGranule(block.getByName(name_and_type->name), stream_getter, current_row, rows_to_write);
+            writeColumnSingleGranule(block.getByName(name_and_type->name), stream_getter, granule.start_row, granule.granularity_rows);
 
             /// Each type always have at least one substream
             prev_stream->hashing_buf.next(); //-V522
         }
 
-        ++from_mark;
-        size_t rows_written = total_rows - current_row;
-        current_row += rows_to_write;
-
-        /// Correct last mark as it should contain exact amount of rows.
-        if (current_row >= total_rows && rows_written != rows_to_write)
-        {
-            rows_to_write = rows_written;
-            index_granularity.popMark();
-            index_granularity.appendMark(rows_written);
-        }
-
-        writeIntBinary(rows_to_write, marks);
+        writeIntBinary(granule.block_rows, marks);
     }
-
-    next_index_offset = 0;
-    next_mark = from_mark;
-}
-
-void MergeTreeDataPartWriterCompact::writeColumnSingleGranule(
-    const ColumnWithTypeAndName & column,
-    IDataType::OutputStreamGetter stream_getter,
-    size_t from_row,
-    size_t number_of_rows)
-{
-    IDataType::SerializeBinaryBulkStatePtr state;
-    IDataType::SerializeBinaryBulkSettings serialize_settings;
-
-    serialize_settings.getter = stream_getter;
-    serialize_settings.position_independent_encoding = true;
-    serialize_settings.low_cardinality_max_dictionary_size = 0;
-
-    column.type->serializeBinaryBulkStatePrefix(serialize_settings, state);
-    column.type->serializeBinaryBulkWithMultipleStreams(*column.column, from_row, number_of_rows, serialize_settings, state);
-    column.type->serializeBinaryBulkStateSuffix(serialize_settings, state);
 }
 
 void MergeTreeDataPartWriterCompact::finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync)
 {
     if (columns_buffer.size() != 0)
-        writeBlock(header.cloneWithColumns(columns_buffer.releaseColumns()));
+    {
+        auto block = header.cloneWithColumns(columns_buffer.releaseColumns());
+        auto granules_to_write = getGranulesToWrite(index_granularity, block.rows(), getCurrentMark(), /* last_block = */ true);
+        if (!granules_to_write.back().isCompleted())
+        {
+            /// Correct last mark as it should contain exact amount of rows.
+            index_granularity.popMark();
+            index_granularity.appendMark(granules_to_write.back().block_rows);
+        }
+        writeDataBlockPrimaryIndexAndSkipIndices(block, granules_to_write);
+    }
 
 #ifndef NDEBUG
     /// Offsets should be 0, because compressed block is written for every granule.
@@ -269,9 +291,13 @@ static void fillIndexGranularityImpl(
 
 void MergeTreeDataPartWriterCompact::fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block)
 {
+    size_t index_offset = 0;
+    if (index_granularity.getMarksCount() > getCurrentMark())
+        index_offset = index_granularity.getMarkRows(getCurrentMark()) - columns_buffer.size();
+
     fillIndexGranularityImpl(
         index_granularity,
-        getIndexOffset(),
+        index_offset,
         index_granularity_for_block,
         rows_in_block);
 }
@@ -328,4 +354,14 @@ size_t MergeTreeDataPartWriterCompact::ColumnsBuffer::size() const
     return accumulated_columns.at(0)->size();
 }
 
+void MergeTreeDataPartWriterCompact::finish(IMergeTreeDataPart::Checksums & checksums, bool sync)
+{
+    finishDataSerialization(checksums, sync);
+
+    if (settings.rewrite_primary_key)
+        finishPrimaryIndexSerialization(checksums, sync);
+
+    finishSkipIndicesSerialization(checksums, sync);
+}
+
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h
index a6187676dde..2913960fd35 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h
@@ -18,16 +18,22 @@ public:
         const MergeTreeWriterSettings & settings,
         const MergeTreeIndexGranularity & index_granularity);
 
-    void write(const Block & block, const IColumn::Permutation * permutation,
-        const Block & primary_key_block, const Block & skip_indexes_block) override;
+    void write(const Block & block, const IColumn::Permutation * permutation) override;
 
-    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
-
-protected:
-    void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) override;
+    void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
 
 private:
-    void writeBlock(const Block & block);
+    /// Finish serialization of the data. Flush rows in buffer to disk, compute checksums.
+    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync);
+
+    void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) override;
+
+    /// Write block of rows into .bin file and marks in .mrk files
+    void writeDataBlock(const Block & block, const Granules & granules);
+
+    /// Write block of rows into .bin file and marks in .mrk files, primary index in .idx file
+    /// and skip indices in their corresponding files.
+    void writeDataBlockPrimaryIndexAndSkipIndices(const Block & block, const Granules & granules);
 
     void addToChecksums(MergeTreeDataPartChecksums & checksums);
 
@@ -78,13 +84,6 @@ private:
     /// marks -> marks_file
     std::unique_ptr<WriteBufferFromFileBase> marks_file;
     HashingWriteBuffer marks;
-
-    /// Write single granule of one column (rows between 2 marks)
-    static void writeColumnSingleGranule(
-        const ColumnWithTypeAndName & column,
-        IDataType::OutputStreamGetter stream_getter,
-        size_t from_row,
-        size_t number_of_rows);
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp
index f0738a1130a..5efcc63b2fc 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp
@@ -19,12 +19,15 @@ MergeTreeDataPartWriterInMemory::MergeTreeDataPartWriterInMemory(
     , part_in_memory(part_) {}
 
 void MergeTreeDataPartWriterInMemory::write(
-    const Block & block, const IColumn::Permutation * permutation,
-    const Block & primary_key_block, const Block & /* skip_indexes_block */)
+    const Block & block, const IColumn::Permutation * permutation)
 {
     if (part_in_memory->block)
         throw Exception("DataPartWriterInMemory supports only one write", ErrorCodes::LOGICAL_ERROR);
 
+    Block primary_key_block;
+    if (settings.rewrite_primary_key)
+        primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), permutation);
+
     Block result_block;
     if (permutation)
     {
@@ -50,6 +53,9 @@ void MergeTreeDataPartWriterInMemory::write(
     if (with_final_mark)
         index_granularity.appendMark(0);
     part_in_memory->block = std::move(result_block);
+
+    if (settings.rewrite_primary_key)
+        calculateAndSerializePrimaryIndex(primary_key_block);
 }
 
 void MergeTreeDataPartWriterInMemory::calculateAndSerializePrimaryIndex(const Block & primary_index_block)
@@ -70,7 +76,7 @@ void MergeTreeDataPartWriterInMemory::calculateAndSerializePrimaryIndex(const Bl
     }
 }
 
-void MergeTreeDataPartWriterInMemory::finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool)
+void MergeTreeDataPartWriterInMemory::finish(IMergeTreeDataPart::Checksums & checksums, bool /* sync */)
 {
     /// If part is empty we still need to initialize block by empty columns.
     if (!part_in_memory->block)
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h
index 990661aad1b..495585ad4d2 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h
@@ -16,14 +16,13 @@ public:
         const MergeTreeWriterSettings & settings_);
 
     /// You can write only one block. In-memory part can be written only at INSERT.
-    void write(const Block & block, const IColumn::Permutation * permutation,
-        const Block & primary_key_block, const Block & skip_indexes_block) override;
+    void write(const Block & block, const IColumn::Permutation * permutation) override;
 
-    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
-
-    void calculateAndSerializePrimaryIndex(const Block & primary_index_block) override;
+    void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
 
 private:
+    void calculateAndSerializePrimaryIndex(const Block & primary_index_block);
+
     DataPartInMemoryPtr part_in_memory;
 };
 
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
index 0ec10f9db76..5190b439809 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
@@ -70,14 +70,14 @@ MergeTreeDataPartWriterOnDisk::MergeTreeDataPartWriterOnDisk(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
     const StorageMetadataPtr & metadata_snapshot_,
-    const std::vector<MergeTreeIndexPtr> & indices_to_recalc_,
+    const MergeTreeIndices & indices_to_recalc_,
     const String & marks_file_extension_,
     const CompressionCodecPtr & default_codec_,
     const MergeTreeWriterSettings & settings_,
     const MergeTreeIndexGranularity & index_granularity_)
     : IMergeTreeDataPartWriter(data_part_,
-        columns_list_, metadata_snapshot_, indices_to_recalc_,
-        index_granularity_, settings_)
+        columns_list_, metadata_snapshot_, settings_, index_granularity_)
+    , skip_indices(indices_to_recalc_)
     , part_path(data_part_->getFullRelativePath())
     , marks_file_extension(marks_file_extension_)
     , default_codec(default_codec_)
@@ -89,6 +89,10 @@ MergeTreeDataPartWriterOnDisk::MergeTreeDataPartWriterOnDisk(
     auto disk = data_part->volume->getDisk();
     if (!disk->exists(part_path))
         disk->createDirectories(part_path);
+
+    if (settings.rewrite_primary_key)
+        initPrimaryIndex();
+    initSkipIndices();
 }
 
 // Implementation is split into static functions for ability
@@ -129,17 +133,7 @@ static size_t computeIndexGranularityImpl(
     return index_granularity_for_block;
 }
 
-static void fillIndexGranularityImpl(
-    MergeTreeIndexGranularity & index_granularity,
-    size_t index_offset,
-    size_t index_granularity_for_block,
-    size_t rows_in_block)
-{
-    for (size_t current_row = index_offset; current_row < rows_in_block; current_row += index_granularity_for_block)
-        index_granularity.appendMark(index_granularity_for_block);
-}
-
-size_t MergeTreeDataPartWriterOnDisk::computeIndexGranularity(const Block & block)
+size_t MergeTreeDataPartWriterOnDisk::computeIndexGranularity(const Block & block) const
 {
     const auto storage_settings = storage.getSettings();
     return computeIndexGranularityImpl(
@@ -150,15 +144,6 @@ size_t MergeTreeDataPartWriterOnDisk::computeIndexGranularity(const Block & bloc
             settings.can_use_adaptive_granularity);
 }
 
-void MergeTreeDataPartWriterOnDisk::fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block)
-{
-    fillIndexGranularityImpl(
-        index_granularity,
-        getIndexOffset(),
-        index_granularity_for_block,
-        rows_in_block);
-}
-
 void MergeTreeDataPartWriterOnDisk::initPrimaryIndex()
 {
     if (metadata_snapshot->hasPrimaryKey())
@@ -166,8 +151,6 @@ void MergeTreeDataPartWriterOnDisk::initPrimaryIndex()
         index_file_stream = data_part->volume->getDisk()->writeFile(part_path + "primary.idx", DBMS_DEFAULT_BUFFER_SIZE, WriteMode::Rewrite);
         index_stream = std::make_unique<HashingWriteBuffer>(*index_file_stream);
     }
-
-    primary_index_initialized = true;
 }
 
 void MergeTreeDataPartWriterOnDisk::initSkipIndices()
@@ -184,18 +167,12 @@ void MergeTreeDataPartWriterOnDisk::initSkipIndices()
                         default_codec, settings.max_compress_block_size,
                         0, settings.aio_threshold));
         skip_indices_aggregators.push_back(index_helper->createIndexAggregator());
-        skip_index_filling.push_back(0);
+        skip_index_accumulated_marks.push_back(0);
     }
-
-    skip_indices_initialized = true;
 }
 
-void MergeTreeDataPartWriterOnDisk::calculateAndSerializePrimaryIndex(const Block & primary_index_block)
+void MergeTreeDataPartWriterOnDisk::calculateAndSerializePrimaryIndex(const Block & primary_index_block, const Granules & granules_to_write)
 {
-    if (!primary_index_initialized)
-        throw Exception("Primary index is not initialized", ErrorCodes::LOGICAL_ERROR);
-
-    size_t rows = primary_index_block.rows();
     size_t primary_columns_num = primary_index_block.columns();
     if (index_columns.empty())
     {
@@ -215,98 +192,62 @@ void MergeTreeDataPartWriterOnDisk::calculateAndSerializePrimaryIndex(const Bloc
     MemoryTracker::BlockerInThread temporarily_disable_memory_tracker;
 
     /// Write index. The index contains Primary Key value for each `index_granularity` row.
-
-    size_t current_row = getIndexOffset();
-    size_t total_marks = index_granularity.getMarksCount();
-
-    while (index_mark < total_marks && current_row < rows)
+    for (const auto & granule : granules_to_write)
     {
-        if (metadata_snapshot->hasPrimaryKey())
+        if (metadata_snapshot->hasPrimaryKey() && granule.mark_on_start)
         {
             for (size_t j = 0; j < primary_columns_num; ++j)
             {
                 const auto & primary_column = primary_index_block.getByPosition(j);
-                index_columns[j]->insertFrom(*primary_column.column, current_row);
-                primary_column.type->serializeBinary(*primary_column.column, current_row, *index_stream);
+                index_columns[j]->insertFrom(*primary_column.column, granule.start_row);
+                primary_column.type->serializeBinary(*primary_column.column, granule.start_row, *index_stream);
             }
         }
-
-        current_row += index_granularity.getMarkRows(index_mark++);
     }
-
     /// store last index row to write final mark at the end of column
     for (size_t j = 0; j < primary_columns_num; ++j)
         last_block_index_columns[j] = primary_index_block.getByPosition(j).column;
 }
 
-void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block & skip_indexes_block)
+void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block & skip_indexes_block, const Granules & granules_to_write)
 {
-    if (!skip_indices_initialized)
-        throw Exception("Skip indices are not initialized", ErrorCodes::LOGICAL_ERROR);
-
-    size_t rows = skip_indexes_block.rows();
-    size_t skip_index_current_data_mark = 0;
-
     /// Filling and writing skip indices like in MergeTreeDataPartWriterWide::writeColumn
     for (size_t i = 0; i < skip_indices.size(); ++i)
     {
         const auto index_helper = skip_indices[i];
         auto & stream = *skip_indices_streams[i];
-        size_t prev_pos = 0;
-        skip_index_current_data_mark = skip_index_data_mark;
-        while (prev_pos < rows)
+        for (const auto & granule : granules_to_write)
         {
-            UInt64 limit = 0;
-            size_t current_index_offset = getIndexOffset();
-            if (prev_pos == 0 && current_index_offset != 0)
+            if (skip_indices_aggregators[i]->empty() && granule.mark_on_start)
             {
-                limit = current_index_offset;
-            }
-            else if (skip_index_current_data_mark == index_granularity.getMarksCount())
-            {
-                /// Case, when last granule was exceeded and no new granule was created.
-                limit = rows - prev_pos;
-            }
-            else
-            {
-                limit = index_granularity.getMarkRows(skip_index_current_data_mark);
-                if (skip_indices_aggregators[i]->empty())
-                {
-                    skip_indices_aggregators[i] = index_helper->createIndexAggregator();
-                    skip_index_filling[i] = 0;
+                skip_indices_aggregators[i] = index_helper->createIndexAggregator();
 
-                    if (stream.compressed.offset() >= settings.min_compress_block_size)
-                        stream.compressed.next();
+                if (stream.compressed.offset() >= settings.min_compress_block_size)
+                    stream.compressed.next();
 
-                    writeIntBinary(stream.plain_hashing.count(), stream.marks);
-                    writeIntBinary(stream.compressed.offset(), stream.marks);
-                    /// Actually this numbers is redundant, but we have to store them
-                    /// to be compatible with normal .mrk2 file format
-                    if (settings.can_use_adaptive_granularity)
-                        writeIntBinary(1UL, stream.marks);
-                }
-                /// this mark is aggregated, go to the next one
-                skip_index_current_data_mark++;
+                writeIntBinary(stream.plain_hashing.count(), stream.marks);
+                writeIntBinary(stream.compressed.offset(), stream.marks);
+                /// Actually this numbers is redundant, but we have to store them
+                /// to be compatible with normal .mrk2 file format
+                if (settings.can_use_adaptive_granularity)
+                    writeIntBinary(1UL, stream.marks);
             }
 
-            size_t pos = prev_pos;
-            skip_indices_aggregators[i]->update(skip_indexes_block, &pos, limit);
-
-            if (pos == prev_pos + limit)
+            size_t pos = granule.start_row;
+            skip_indices_aggregators[i]->update(skip_indexes_block, &pos, granule.granularity_rows);
+            if (granule.isCompleted())
             {
-                ++skip_index_filling[i];
+                ++skip_index_accumulated_marks[i];
 
                 /// write index if it is filled
-                if (skip_index_filling[i] == index_helper->index.granularity)
+                if (skip_index_accumulated_marks[i] == index_helper->index.granularity)
                 {
                     skip_indices_aggregators[i]->getGranuleAndReset()->serializeBinary(stream.compressed);
-                    skip_index_filling[i] = 0;
+                    skip_index_accumulated_marks[i] = 0;
                 }
             }
-            prev_pos = pos;
         }
     }
-    skip_index_data_mark = skip_index_current_data_mark;
 }
 
 void MergeTreeDataPartWriterOnDisk::finishPrimaryIndexSerialization(
@@ -360,7 +301,16 @@ void MergeTreeDataPartWriterOnDisk::finishSkipIndicesSerialization(
 
     skip_indices_streams.clear();
     skip_indices_aggregators.clear();
-    skip_index_filling.clear();
+    skip_index_accumulated_marks.clear();
+}
+
+Names MergeTreeDataPartWriterOnDisk::getSkipIndicesColumns() const
+{
+    std::unordered_set<String> skip_indexes_column_names_set;
+    for (const auto & index : skip_indices)
+        std::copy(index->index.column_names.cbegin(), index->index.column_names.cend(),
+                  std::inserter(skip_indexes_column_names_set, skip_indexes_column_names_set.end()));
+    return Names(skip_indexes_column_names_set.begin(), skip_indexes_column_names_set.end());
 }
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
index 55da413c9fb..bc09e0e61e1 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
@@ -14,6 +14,35 @@
 namespace DB
 {
 
+/// Single unit for writing data to disk. Contains information about
+/// amount of rows to write and marks.
+struct Granule
+{
+    /// Start row in block for granule
+    size_t start_row;
+    /// Amount of rows which granule have to contain according to index
+    /// granularity.
+    /// NOTE: Sometimes it's not equal to actually written rows, for example
+    /// for the last granule if it's smaller than computed granularity.
+    size_t granularity_rows;
+    /// Amount of rows from block which have to be written to disk from start_row
+    size_t block_rows;
+    /// Global mark number in the list of all marks (index_granularity) for this part
+    size_t mark_number;
+    /// Should writer write mark for the first of this granule to disk.
+    /// NOTE: Sometimes we don't write mark for the start row, because
+    /// this granule can be continuation of the previous one.
+    bool mark_on_start;
+
+    /// Is this granule contain amout of rows equal to the value in index granularity
+    bool isCompleted() const
+    {
+        return granularity_rows == block_rows;
+    }
+};
+
+/// Multiple granules to write for concrete block.
+using Granules = std::vector<Granule>;
 
 /// Writes data part to disk in different formats.
 /// Calculates and serializes primary and skip indices if needed.
@@ -71,42 +100,46 @@ public:
         const MergeTreeWriterSettings & settings,
         const MergeTreeIndexGranularity & index_granularity);
 
-    void calculateAndSerializePrimaryIndex(const Block & primary_index_block) final;
-    void calculateAndSerializeSkipIndices(const Block & skip_indexes_block) final;
-
-     /// Count index_granularity for block and store in `index_granularity`
-    size_t computeIndexGranularity(const Block & block);
-    virtual void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block);
-
-    void initSkipIndices() final;
-    void initPrimaryIndex() final;
-
-    void finishPrimaryIndexSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync) final;
-    void finishSkipIndicesSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync) final;
-
     void setWrittenOffsetColumns(WrittenOffsetColumns * written_offset_columns_)
     {
         written_offset_columns = written_offset_columns_;
     }
 
 protected:
-    using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
-    using SerializationStates = std::unordered_map<String, SerializationState>;
+     /// Count index_granularity for block and store in `index_granularity`
+    size_t computeIndexGranularity(const Block & block) const;
 
-    String part_path;
+    /// Write primary index according to granules_to_write
+    void calculateAndSerializePrimaryIndex(const Block & primary_index_block, const Granules & granules_to_write);
+    /// Write skip indices according to granules_to_write. Skip indices also have their own marks
+    /// and one skip index granule can contain multiple "normal" marks. So skip indices serialization
+    /// require additional state: skip_indices_aggregators and skip_index_accumulated_marks
+    void calculateAndSerializeSkipIndices(const Block & skip_indexes_block, const Granules & granules_to_write);
+
+    /// Finishes primary index serialization: write final primary index row (if required) and compute checksums
+    void finishPrimaryIndexSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync);
+    /// Finishes skip indices serialization: write all accumulated data to disk and compute checksums
+    void finishSkipIndicesSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync);
+
+    /// Get global number of the current which we are writing (or going to start to write)
+    size_t getCurrentMark() const { return current_mark; }
+
+    void setCurrentMark(size_t mark) { current_mark = mark; }
+
+    /// Get unique non ordered skip indices column.
+    Names getSkipIndicesColumns() const;
+
+    const MergeTreeIndices skip_indices;
+
+    const String part_path;
     const String marks_file_extension;
-    CompressionCodecPtr default_codec;
+    const CompressionCodecPtr default_codec;
 
-    bool compute_granularity;
-    bool need_finish_last_granule;
-
-    /// Number of marsk in data from which skip indices have to start
-    /// aggregation. I.e. it's data mark number, not skip indices mark.
-    size_t skip_index_data_mark = 0;
+    const bool compute_granularity;
 
     std::vector<StreamPtr> skip_indices_streams;
     MergeTreeIndexAggregators skip_indices_aggregators;
-    std::vector<size_t> skip_index_filling;
+    std::vector<size_t> skip_index_accumulated_marks;
 
     std::unique_ptr<WriteBufferFromFileBase> index_file_stream;
     std::unique_ptr<HashingWriteBuffer> index_stream;
@@ -116,15 +149,18 @@ protected:
     Columns last_block_index_columns;
 
     bool data_written = false;
-    bool primary_index_initialized = false;
-    bool skip_indices_initialized = false;
 
     /// To correctly write Nested elements column-by-column.
     WrittenOffsetColumns * written_offset_columns = nullptr;
 
+    /// Data is already written up to this mark.
+    size_t current_mark = 0;
+
 private:
-    /// Index is already serialized up to this mark.
-    size_t index_mark = 0;
+    void initSkipIndices();
+    void initPrimaryIndex();
+
+    virtual void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) = 0;
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index c15c39e7b7f..1127c3a7dd7 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -1,6 +1,7 @@
 #include <Storages/MergeTree/MergeTreeDataPartWriterWide.h>
 #include <Interpreters/Context.h>
 #include <Compression/CompressionFactory.h>
+#include <Compression/CompressedReadBufferFromFile.h>
 
 namespace DB
 {
@@ -14,6 +15,56 @@ namespace
     constexpr auto DATA_FILE_EXTENSION = ".bin";
 }
 
+namespace
+{
+
+/// Get granules for block using index_granularity
+Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity, size_t block_rows, size_t current_mark, size_t rows_written_in_last_mark)
+{
+    if (current_mark >= index_granularity.getMarksCount())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Request to get granules from mark {} but index granularity size is {}", current_mark, index_granularity.getMarksCount());
+
+    Granules result;
+    size_t current_row = 0;
+    /// When our last mark is not finished yet and we have to write in rows into it
+    if (rows_written_in_last_mark > 0)
+    {
+        size_t rows_left_in_last_mark = index_granularity.getMarkRows(current_mark) - rows_written_in_last_mark;
+        size_t rows_left_in_block = block_rows - current_row;
+        result.emplace_back(Granule{
+            .start_row = current_row,
+            .granularity_rows = rows_left_in_last_mark,
+            .block_rows = std::min(rows_left_in_block, rows_left_in_last_mark),
+            .mark_number = current_mark,
+            .mark_on_start = false, /// Don't mark this granule because we have already marked it
+        });
+        current_row += rows_left_in_last_mark;
+        current_mark++;
+    }
+
+    /// Calculating normal granules for block
+    while (current_row < block_rows)
+    {
+        size_t expected_rows_in_mark = index_granularity.getMarkRows(current_mark);
+        size_t rows_left_in_block  = block_rows - current_row;
+        /// If we have less rows in block than expected in granularity
+        /// save incomplete granule
+        result.emplace_back(Granule{
+            .start_row = current_row,
+            .granularity_rows = expected_rows_in_mark,
+            .block_rows = std::min(rows_left_in_block, expected_rows_in_mark),
+            .mark_number = current_mark,
+            .mark_on_start = true,
+        });
+        current_row += expected_rows_in_mark;
+        current_mark++;
+    }
+
+    return result;
+}
+
+}
+
 MergeTreeDataPartWriterWide::MergeTreeDataPartWriterWide(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
@@ -69,7 +120,7 @@ void MergeTreeDataPartWriterWide::addStreams(
 
 
 IDataType::OutputStreamGetter MergeTreeDataPartWriterWide::createStreamGetter(
-        const String & name, WrittenOffsetColumns & offset_columns)
+        const String & name, WrittenOffsetColumns & offset_columns) const
 {
     return [&, this] (const IDataType::SubstreamPath & substream_path) -> WriteBuffer *
     {
@@ -81,13 +132,34 @@ IDataType::OutputStreamGetter MergeTreeDataPartWriterWide::createStreamGetter(
         if (is_offsets && offset_columns.count(stream_name))
             return nullptr;
 
-        return &column_streams[stream_name]->compressed;
+        return &column_streams.at(stream_name)->compressed;
     };
 }
 
-void MergeTreeDataPartWriterWide::write(const Block & block,
-    const IColumn::Permutation * permutation,
-    const Block & primary_key_block, const Block & skip_indexes_block)
+void MergeTreeDataPartWriterWide::shiftCurrentMark(const Granules & granules_written)
+{
+    auto last_granule = granules_written.back();
+    /// If we didn't finished last granule than we will continue to write it from new block
+    if (!last_granule.isCompleted())
+    {
+        /// Shift forward except last granule
+        setCurrentMark(getCurrentMark() + granules_written.size() - 1);
+        bool still_in_the_same_granule = granules_written.size() == 1;
+        /// We wrote whole block in the same granule, but didn't finished it.
+        /// So add written rows to rows written in last_mark
+        if (still_in_the_same_granule)
+            rows_written_in_last_mark += last_granule.block_rows;
+        else
+            rows_written_in_last_mark = last_granule.block_rows;
+    }
+    else
+    {
+        setCurrentMark(getCurrentMark() + granules_written.size());
+        rows_written_in_last_mark = 0;
+    }
+}
+
+void MergeTreeDataPartWriterWide::write(const Block & block, const IColumn::Permutation * permutation)
 {
     /// Fill index granularity for this block
     /// if it's unknown (in case of insert data or horizontal merge,
@@ -98,7 +170,14 @@ void MergeTreeDataPartWriterWide::write(const Block & block,
         fillIndexGranularity(index_granularity_for_block, block.rows());
     }
 
+    auto granules_to_write = getGranulesToWrite(index_granularity, block.rows(), getCurrentMark(), rows_written_in_last_mark);
+
     auto offset_columns = written_offset_columns ? *written_offset_columns : WrittenOffsetColumns{};
+    Block primary_key_block;
+    if (settings.rewrite_primary_key)
+        primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), permutation);
+
+    Block skip_indexes_block = getBlockAndPermute(block, getSkipIndicesColumns(), permutation);
 
     auto it = columns_list.begin();
     for (size_t i = 0; i < columns_list.size(); ++i, ++it)
@@ -110,25 +189,32 @@ void MergeTreeDataPartWriterWide::write(const Block & block,
             if (primary_key_block.has(it->name))
             {
                 const auto & primary_column = *primary_key_block.getByName(it->name).column;
-                writeColumn(column.name, *column.type, primary_column, offset_columns);
+                writeColumn(column.name, *column.type, primary_column, offset_columns, granules_to_write);
             }
             else if (skip_indexes_block.has(it->name))
             {
                 const auto & index_column = *skip_indexes_block.getByName(it->name).column;
-                writeColumn(column.name, *column.type, index_column, offset_columns);
+                writeColumn(column.name, *column.type, index_column, offset_columns, granules_to_write);
             }
             else
             {
                 /// We rearrange the columns that are not included in the primary key here; Then the result is released - to save RAM.
                 ColumnPtr permuted_column = column.column->permute(*permutation, 0);
-                writeColumn(column.name, *column.type, *permuted_column, offset_columns);
+                writeColumn(column.name, *column.type, *permuted_column, offset_columns, granules_to_write);
             }
         }
         else
         {
-            writeColumn(column.name, *column.type, *column.column, offset_columns);
+            writeColumn(column.name, *column.type, *column.column, offset_columns, granules_to_write);
         }
     }
+
+    if (settings.rewrite_primary_key)
+        calculateAndSerializePrimaryIndex(primary_key_block, granules_to_write);
+
+    calculateAndSerializeSkipIndices(skip_indexes_block, granules_to_write);
+
+    shiftCurrentMark(granules_to_write);
 }
 
 void MergeTreeDataPartWriterWide::writeSingleMark(
@@ -138,44 +224,67 @@ void MergeTreeDataPartWriterWide::writeSingleMark(
     size_t number_of_rows,
     DB::IDataType::SubstreamPath & path)
 {
-    type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
-     {
-         bool is_offsets = !substream_path.empty() && substream_path.back().type == IDataType::Substream::ArraySizes;
-
-         String stream_name = IDataType::getFileNameForStream(name, substream_path);
-
-         /// Don't write offsets more than one time for Nested type.
-         if (is_offsets && offset_columns.count(stream_name))
-             return;
-
-         Stream & stream = *column_streams[stream_name];
-
-         /// There could already be enough data to compress into the new block.
-         if (stream.compressed.offset() >= settings.min_compress_block_size)
-             stream.compressed.next();
-
-         writeIntBinary(stream.plain_hashing.count(), stream.marks);
-         writeIntBinary(stream.compressed.offset(), stream.marks);
-         if (settings.can_use_adaptive_granularity)
-             writeIntBinary(number_of_rows, stream.marks);
-     }, path);
+    StreamsWithMarks marks = getCurrentMarksForColumn(name, type, offset_columns, path);
+    for (const auto & mark : marks)
+        flushMarkToFile(mark, number_of_rows);
 }
 
-size_t MergeTreeDataPartWriterWide::writeSingleGranule(
+void MergeTreeDataPartWriterWide::flushMarkToFile(const StreamNameAndMark & stream_with_mark, size_t rows_in_mark)
+{
+    Stream & stream = *column_streams[stream_with_mark.stream_name];
+    writeIntBinary(stream_with_mark.mark.offset_in_compressed_file, stream.marks);
+    writeIntBinary(stream_with_mark.mark.offset_in_decompressed_block, stream.marks);
+    if (settings.can_use_adaptive_granularity)
+        writeIntBinary(rows_in_mark, stream.marks);
+}
+
+StreamsWithMarks MergeTreeDataPartWriterWide::getCurrentMarksForColumn(
+    const String & name,
+    const IDataType & type,
+    WrittenOffsetColumns & offset_columns,
+    DB::IDataType::SubstreamPath & path)
+{
+    StreamsWithMarks result;
+    type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
+    {
+        bool is_offsets = !substream_path.empty() && substream_path.back().type == IDataType::Substream::ArraySizes;
+
+        String stream_name = IDataType::getFileNameForStream(name, substream_path);
+
+        /// Don't write offsets more than one time for Nested type.
+        if (is_offsets && offset_columns.count(stream_name))
+            return;
+
+        Stream & stream = *column_streams[stream_name];
+
+        /// There could already be enough data to compress into the new block.
+        if (stream.compressed.offset() >= settings.min_compress_block_size)
+            stream.compressed.next();
+
+        StreamNameAndMark stream_with_mark;
+        stream_with_mark.stream_name = stream_name;
+        stream_with_mark.mark.offset_in_compressed_file = stream.plain_hashing.count();
+        stream_with_mark.mark.offset_in_decompressed_block = stream.compressed.offset();
+
+        result.push_back(stream_with_mark);
+    }, path);
+
+    return result;
+}
+
+void MergeTreeDataPartWriterWide::writeSingleGranule(
     const String & name,
     const IDataType & type,
     const IColumn & column,
     WrittenOffsetColumns & offset_columns,
     IDataType::SerializeBinaryBulkStatePtr & serialization_state,
     IDataType::SerializeBinaryBulkSettings & serialize_settings,
-    size_t from_row,
-    size_t number_of_rows,
-    bool write_marks)
+    const Granule & granule)
 {
-    if (write_marks)
-        writeSingleMark(name, type, offset_columns, number_of_rows, serialize_settings.path);
+    if (granule.mark_on_start)
+        writeSingleMark(name, type, offset_columns, granule.granularity_rows, serialize_settings.path);
 
-    type.serializeBinaryBulkWithMultipleStreams(column, from_row, number_of_rows, serialize_settings, serialization_state);
+    type.serializeBinaryBulkWithMultipleStreams(column, granule.start_row, granule.granularity_rows, serialize_settings, serialization_state);
 
     /// So that instead of the marks pointing to the end of the compressed block, there were marks pointing to the beginning of the next one.
     type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
@@ -190,8 +299,6 @@ size_t MergeTreeDataPartWriterWide::writeSingleGranule(
 
         column_streams[stream_name]->compressed.nextIfAtEnd();
     }, serialize_settings.path);
-
-    return from_row + number_of_rows;
 }
 
 /// Column must not be empty. (column.size() !== 0)
@@ -199,9 +306,11 @@ void MergeTreeDataPartWriterWide::writeColumn(
     const String & name,
     const IDataType & type,
     const IColumn & column,
-    WrittenOffsetColumns & offset_columns)
+    WrittenOffsetColumns & offset_columns,
+    const Granules & granules)
 {
     auto [it, inserted] = serialization_states.emplace(name, nullptr);
+
     if (inserted)
     {
         IDataType::SerializeBinaryBulkSettings serialize_settings;
@@ -215,48 +324,20 @@ void MergeTreeDataPartWriterWide::writeColumn(
     serialize_settings.low_cardinality_max_dictionary_size = global_settings.low_cardinality_max_dictionary_size;
     serialize_settings.low_cardinality_use_single_dictionary_for_part = global_settings.low_cardinality_use_single_dictionary_for_part != 0;
 
-    size_t total_rows = column.size();
-    size_t current_row = 0;
-    size_t current_column_mark = getCurrentMark();
-    size_t current_index_offset = getIndexOffset();
-    while (current_row < total_rows)
+    for (const auto & granule : granules)
     {
-        size_t rows_to_write;
-        bool write_marks = true;
-
-        /// If there is `index_offset`, then the first mark goes not immediately, but after this number of rows.
-        if (current_row == 0 && current_index_offset != 0)
-        {
-            write_marks = false;
-            rows_to_write = current_index_offset;
-        }
-        else
-        {
-            if (index_granularity.getMarksCount() <= current_column_mark)
-                throw Exception(
-                    "Incorrect size of index granularity expect mark " + toString(current_column_mark) + " totally have marks " + toString(index_granularity.getMarksCount()),
-                    ErrorCodes::LOGICAL_ERROR);
-
-            rows_to_write = index_granularity.getMarkRows(current_column_mark);
-        }
-
-        if (rows_to_write != 0)
+        if (granule.granularity_rows > 0)
             data_written = true;
 
-        current_row = writeSingleGranule(
-            name,
-            type,
-            column,
-            offset_columns,
-            it->second,
-            serialize_settings,
-            current_row,
-            rows_to_write,
-            write_marks
+        writeSingleGranule(
+           name,
+           type,
+           column,
+           offset_columns,
+           it->second,
+           serialize_settings,
+           granule
         );
-
-        if (write_marks)
-            current_column_mark++;
     }
 
     type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
@@ -268,9 +349,93 @@ void MergeTreeDataPartWriterWide::writeColumn(
             offset_columns.insert(stream_name);
         }
     }, serialize_settings.path);
+}
+
+
+void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name, const IDataType & type)
+{
+    if (!type.isValueRepresentedByNumber() || type.haveSubtypes())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot validate column of non fixed type {}", type.getName());
+
+    auto disk = data_part->volume->getDisk();
+    String mrk_path = fullPath(disk, part_path + name + marks_file_extension);
+    String bin_path = fullPath(disk, part_path + name + DATA_FILE_EXTENSION);
+    DB::ReadBufferFromFile mrk_in(mrk_path);
+    DB::CompressedReadBufferFromFile bin_in(bin_path, 0, 0, 0);
+    bool must_be_last = false;
+    UInt64 offset_in_compressed_file = 0;
+    UInt64 offset_in_decompressed_block = 0;
+    UInt64 index_granularity_rows = 0;
+
+    size_t mark_num;
+
+    for (mark_num = 0; !mrk_in.eof(); ++mark_num)
+    {
+        if (mark_num > index_granularity.getMarksCount())
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Incorrect number of marks in memory {}, on disk (at least) {}", index_granularity.getMarksCount(), mark_num + 1);
+
+        DB::readBinary(offset_in_compressed_file, mrk_in);
+        DB::readBinary(offset_in_decompressed_block, mrk_in);
+        if (settings.can_use_adaptive_granularity)
+            DB::readBinary(index_granularity_rows, mrk_in);
+        else
+            index_granularity_rows = storage.getSettings()->index_granularity;
+
+        if (must_be_last)
+        {
+            if (index_granularity_rows != 0)
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "We ran out of binary data but still have non empty mark #{} with rows number {}", mark_num, index_granularity_rows);
+
+            if (!mrk_in.eof())
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "Mark #{} must be last, but we still have some to read", mark_num);
+
+            break;
+        }
+
+        if (index_granularity_rows == 0)
+        {
+            auto column = type.createColumn();
+
+            type.deserializeBinaryBulk(*column, bin_in, 1000000000, 0.0);
+
+            throw Exception(ErrorCodes::LOGICAL_ERROR,
+                        "Still have {} rows in bin stream, last mark #{} index granularity size {}, last rows {}", column->size(), mark_num, index_granularity.getMarksCount(), index_granularity_rows);
+        }
+
+        if (index_granularity_rows != index_granularity.getMarkRows(mark_num))
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), in-memory {}, on disk {}",
+                mark_num, offset_in_compressed_file, offset_in_decompressed_block, index_granularity.getMarkRows(mark_num), index_granularity_rows);
+
+        auto column = type.createColumn();
+
+        type.deserializeBinaryBulk(*column, bin_in, index_granularity_rows, 0.0);
+
+        if (bin_in.eof())
+        {
+            must_be_last = true;
+        }
+        else if (column->size() != index_granularity_rows)
+        {
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), actually in bin file {}, in mrk file {}",
+                mark_num, offset_in_compressed_file, offset_in_decompressed_block, column->size(), index_granularity.getMarkRows(mark_num));
+        }
+    }
+
+    if (!mrk_in.eof())
+        throw Exception(ErrorCodes::LOGICAL_ERROR,
+                        "Still have something in marks stream, last mark #{} index granularity size {}, last rows {}", mark_num, index_granularity.getMarksCount(), index_granularity_rows);
+    if (!bin_in.eof())
+    {
+        auto column = type.createColumn();
+
+        type.deserializeBinaryBulk(*column, bin_in, 1000000000, 0.0);
+
+        throw Exception(ErrorCodes::LOGICAL_ERROR,
+                        "Still have {} rows in bin stream, last mark #{} index granularity size {}, last rows {}", column->size(), mark_num, index_granularity.getMarksCount(), index_granularity_rows);
+    }
 
-    next_mark = current_column_mark;
-    next_index_offset = current_row - total_rows;
 }
 
 void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync)
@@ -294,12 +459,9 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
             }
 
             if (write_final_mark)
-            {
                 writeFinalMark(it->name, it->type, offset_columns, serialize_settings.path);
-            }
         }
     }
-
     for (auto & stream : column_streams)
     {
         stream.second->finalize();
@@ -310,6 +472,24 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
 
     column_streams.clear();
     serialization_states.clear();
+
+#ifndef NDEBUG
+    for (const auto & column : columns_list)
+    {
+        if (column.type->isValueRepresentedByNumber() && !column.type->haveSubtypes())
+            validateColumnOfFixedSize(column.name, *column.type);
+    }
+#endif
+
+}
+
+void MergeTreeDataPartWriterWide::finish(IMergeTreeDataPart::Checksums & checksums, bool sync)
+{
+    finishDataSerialization(checksums, sync);
+    if (settings.rewrite_primary_key)
+        finishPrimaryIndexSerialization(checksums, sync);
+
+    finishSkipIndicesSerialization(checksums, sync);
 }
 
 void MergeTreeDataPartWriterWide::writeFinalMark(
@@ -331,4 +511,30 @@ void MergeTreeDataPartWriterWide::writeFinalMark(
     }, path);
 }
 
+static void fillIndexGranularityImpl(
+    MergeTreeIndexGranularity & index_granularity,
+    size_t index_offset,
+    size_t index_granularity_for_block,
+    size_t rows_in_block)
+{
+    for (size_t current_row = index_offset; current_row < rows_in_block; current_row += index_granularity_for_block)
+        index_granularity.appendMark(index_granularity_for_block);
+}
+
+void MergeTreeDataPartWriterWide::fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block)
+{
+    if (getCurrentMark() < index_granularity.getMarksCount() && getCurrentMark() != index_granularity.getMarksCount() - 1)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Trying to add marks, while current mark {}, but total marks {}", getCurrentMark(), index_granularity.getMarksCount());
+
+    size_t index_offset = 0;
+    if (rows_written_in_last_mark != 0)
+        index_offset = index_granularity.getLastMarkRows() - rows_written_in_last_mark;
+
+    fillIndexGranularityImpl(
+        index_granularity,
+        index_offset,
+        index_granularity_for_block,
+        rows_in_block);
+}
+
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
index aeb686a70a1..cffedd66a32 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
@@ -4,13 +4,19 @@
 namespace DB
 {
 
+struct StreamNameAndMark
+{
+    String stream_name;
+    MarkInCompressedFile mark;
+};
+
+using StreamsWithMarks = std::vector<StreamNameAndMark>;
+using ColumnNameToMark = std::unordered_map<String, StreamsWithMarks>;
+
 /// Writes data part in wide format.
 class MergeTreeDataPartWriterWide : public MergeTreeDataPartWriterOnDisk
 {
 public:
-
-    using ColumnToSize = std::map<std::string, UInt64>;
-
     MergeTreeDataPartWriterWide(
         const MergeTreeData::DataPartPtr & data_part,
         const NamesAndTypesList & columns_list,
@@ -21,14 +27,15 @@ public:
         const MergeTreeWriterSettings & settings,
         const MergeTreeIndexGranularity & index_granularity);
 
-    void write(const Block & block, const IColumn::Permutation * permutation,
-        const Block & primary_key_block, const Block & skip_indexes_block) override;
+    void write(const Block & block, const IColumn::Permutation * permutation) override;
 
-    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
-
-    IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns);
+    void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) final;
 
 private:
+    /// Finish serialization of data: write final mark if required and compute checksums
+    /// Also validate written data in debug mode
+    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync);
+
     /// Write data of one column.
     /// Return how many marks were written and
     /// how many rows were written for last mark
@@ -36,21 +43,32 @@ private:
         const String & name,
         const IDataType & type,
         const IColumn & column,
-        WrittenOffsetColumns & offset_columns);
+        WrittenOffsetColumns & offset_columns,
+        const Granules & granules);
 
-    /// Write single granule of one column (rows between 2 marks)
-    size_t writeSingleGranule(
+    /// Write single granule of one column.
+    void writeSingleGranule(
         const String & name,
         const IDataType & type,
         const IColumn & column,
         WrittenOffsetColumns & offset_columns,
         IDataType::SerializeBinaryBulkStatePtr & serialization_state,
         IDataType::SerializeBinaryBulkSettings & serialize_settings,
-        size_t from_row,
-        size_t number_of_rows,
-        bool write_marks);
+        const Granule & granule);
 
-    /// Write mark for column
+    /// Take offsets from column and return as MarkInCompressed file with stream name
+    StreamsWithMarks getCurrentMarksForColumn(
+        const String & name,
+        const IDataType & type,
+        WrittenOffsetColumns & offset_columns,
+        DB::IDataType::SubstreamPath & path);
+
+    /// Write mark to disk using stream and rows count
+    void flushMarkToFile(
+        const StreamNameAndMark & stream_with_mark,
+        size_t rows_in_mark);
+
+    /// Write mark for column taking offsets from column stream
     void writeSingleMark(
         const String & name,
         const IDataType & type,
@@ -70,10 +88,30 @@ private:
         const ASTPtr & effective_codec_desc,
         size_t estimated_size);
 
+    /// Method for self check (used in debug-build only). Checks that written
+    /// data and corresponding marks are consistent. Otherwise throws logical
+    /// errors.
+    void validateColumnOfFixedSize(const String & name, const IDataType & type);
+
+    void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) override;
+
+    /// Use information from just written granules to shift current mark
+    /// in our index_granularity array.
+    void shiftCurrentMark(const Granules & granules_written);
+
+    IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns) const;
+
+    using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
+    using SerializationStates = std::unordered_map<String, SerializationState>;
+
     SerializationStates serialization_states;
 
     using ColumnStreams = std::map<String, StreamPtr>;
     ColumnStreams column_streams;
+
+    /// How many rows we have already written in the current mark.
+    /// More than zero when incoming blocks are smaller then their granularity.
+    size_t rows_written_in_last_mark = 0;
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataWriter.cpp b/src/Storages/MergeTree/MergeTreeDataWriter.cpp
index e65a8e22f32..c93d4bceba0 100644
--- a/src/Storages/MergeTree/MergeTreeDataWriter.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataWriter.cpp
@@ -16,6 +16,14 @@
 
 #include <Parsers/queryToString.h>
 
+#include <Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/MergingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/SummingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/AggregatingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h>
+#include <Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.h>
+
 namespace ProfileEvents
 {
     extern const Event MergeTreeDataWriterBlocks;
@@ -194,7 +202,74 @@ BlocksWithPartition MergeTreeDataWriter::splitBlockIntoParts(const Block & block
     return result;
 }
 
-MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPartition & block_with_partition, const StorageMetadataPtr & metadata_snapshot)
+Block MergeTreeDataWriter::mergeBlock(const Block & block, SortDescription sort_description, Names & partition_key_columns, IColumn::Permutation *& permutation)
+{
+    size_t block_size = block.rows();
+
+    auto get_merging_algorithm = [&]() -> std::shared_ptr<IMergingAlgorithm>
+    {
+        switch (data.merging_params.mode)
+        {
+            /// There is nothing to merge in single block in ordinary MergeTree
+            case MergeTreeData::MergingParams::Ordinary:
+                return nullptr;
+            case MergeTreeData::MergingParams::Replacing:
+                return std::make_shared<ReplacingSortedAlgorithm>(
+                    block, 1, sort_description, data.merging_params.version_column, block_size + 1);
+            case MergeTreeData::MergingParams::Collapsing:
+                return std::make_shared<CollapsingSortedAlgorithm>(
+                    block, 1, sort_description, data.merging_params.sign_column,
+                    false, block_size + 1, &Poco::Logger::get("MergeTreeBlockOutputStream"));
+            case MergeTreeData::MergingParams::Summing:
+                return std::make_shared<SummingSortedAlgorithm>(
+                    block, 1, sort_description, data.merging_params.columns_to_sum,
+                    partition_key_columns, block_size + 1);
+            case MergeTreeData::MergingParams::Aggregating:
+                return std::make_shared<AggregatingSortedAlgorithm>(block, 1, sort_description, block_size + 1);
+            case MergeTreeData::MergingParams::VersionedCollapsing:
+                return std::make_shared<VersionedCollapsingAlgorithm>(
+                    block, 1, sort_description, data.merging_params.sign_column, block_size + 1);
+            case MergeTreeData::MergingParams::Graphite:
+                return std::make_shared<GraphiteRollupSortedAlgorithm>(
+                    block, 1, sort_description, block_size + 1, data.merging_params.graphite_params, time(nullptr));
+        }
+
+        __builtin_unreachable();
+    };
+
+    auto merging_algorithm = get_merging_algorithm();
+    if (!merging_algorithm)
+        return block;
+
+    Chunk chunk(block.getColumns(), block_size);
+
+    IMergingAlgorithm::Input input;
+    input.set(std::move(chunk));
+    input.permutation = permutation;
+
+    IMergingAlgorithm::Inputs inputs;
+    inputs.push_back(std::move(input));
+    merging_algorithm->initialize(std::move(inputs));
+
+    IMergingAlgorithm::Status status = merging_algorithm->merge();
+
+    /// Check that after first merge merging_algorithm is waiting for data from input 0.
+    if (status.required_source != 0)
+        throw Exception("Logical error: required source after the first merge is not 0.", ErrorCodes::LOGICAL_ERROR);
+
+    status = merging_algorithm->merge();
+
+    /// Check that merge is finished.
+    if (!status.is_finished)
+        throw Exception("Logical error: merge is not finished after the second merge.", ErrorCodes::LOGICAL_ERROR);
+
+    /// Merged Block is sorted and we don't need to use permutation anymore
+    permutation = nullptr;
+
+    return block.cloneWithColumns(status.chunk.getColumns());
+}
+
+MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPartition & block_with_partition, const StorageMetadataPtr & metadata_snapshot, bool optimize_on_insert)
 {
     Block & block = block_with_partition.block;
 
@@ -228,6 +303,38 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPa
     else
         part_name = new_part_info.getPartName();
 
+    /// If we need to calculate some columns to sort.
+    if (metadata_snapshot->hasSortingKey() || metadata_snapshot->hasSecondaryIndices())
+        data.getSortingKeyAndSkipIndicesExpression(metadata_snapshot)->execute(block);
+
+    Names sort_columns = metadata_snapshot->getSortingKeyColumns();
+    SortDescription sort_description;
+    size_t sort_columns_size = sort_columns.size();
+    sort_description.reserve(sort_columns_size);
+
+    for (size_t i = 0; i < sort_columns_size; ++i)
+        sort_description.emplace_back(block.getPositionByName(sort_columns[i]), 1, 1);
+
+    ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocks);
+
+    /// Sort
+    IColumn::Permutation * perm_ptr = nullptr;
+    IColumn::Permutation perm;
+    if (!sort_description.empty())
+    {
+        if (!isAlreadySorted(block, sort_description))
+        {
+            stableGetPermutation(block, sort_description, perm);
+            perm_ptr = &perm;
+        }
+        else
+            ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocksAlreadySorted);
+    }
+
+    Names partition_key_columns = metadata_snapshot->getPartitionKey().column_names;
+    if (optimize_on_insert)
+        block = mergeBlock(block, sort_description, partition_key_columns, perm_ptr);
+
     /// Size of part would not be greater than block.bytes() + epsilon
     size_t expected_size = block.bytes();
 
@@ -274,34 +381,6 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPa
             sync_guard.emplace(disk, full_path);
     }
 
-    /// If we need to calculate some columns to sort.
-    if (metadata_snapshot->hasSortingKey() || metadata_snapshot->hasSecondaryIndices())
-        data.getSortingKeyAndSkipIndicesExpression(metadata_snapshot)->execute(block);
-
-    Names sort_columns = metadata_snapshot->getSortingKeyColumns();
-    SortDescription sort_description;
-    size_t sort_columns_size = sort_columns.size();
-    sort_description.reserve(sort_columns_size);
-
-    for (size_t i = 0; i < sort_columns_size; ++i)
-        sort_description.emplace_back(block.getPositionByName(sort_columns[i]), 1, 1);
-
-    ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocks);
-
-    /// Sort
-    IColumn::Permutation * perm_ptr = nullptr;
-    IColumn::Permutation perm;
-    if (!sort_description.empty())
-    {
-        if (!isAlreadySorted(block, sort_description))
-        {
-            stableGetPermutation(block, sort_description, perm);
-            perm_ptr = &perm;
-        }
-        else
-            ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocksAlreadySorted);
-    }
-
     if (metadata_snapshot->hasRowsTTL())
         updateTTL(metadata_snapshot->getRowsTTL(), new_data_part->ttl_infos, new_data_part->ttl_infos.table_ttl, block, true);
 
diff --git a/src/Storages/MergeTree/MergeTreeDataWriter.h b/src/Storages/MergeTree/MergeTreeDataWriter.h
index b4ad936672c..685d1adf947 100644
--- a/src/Storages/MergeTree/MergeTreeDataWriter.h
+++ b/src/Storages/MergeTree/MergeTreeDataWriter.h
@@ -45,7 +45,9 @@ public:
     /** All rows must correspond to same partition.
       * Returns part with unique name starting with 'tmp_', yet not added to MergeTreeData.
       */
-    MergeTreeData::MutableDataPartPtr writeTempPart(BlockWithPartition & block, const StorageMetadataPtr & metadata_snapshot);
+    MergeTreeData::MutableDataPartPtr writeTempPart(BlockWithPartition & block, const StorageMetadataPtr & metadata_snapshot, bool optimize_on_insert);
+
+    Block mergeBlock(const Block & block, SortDescription sort_description, Names & partition_key_columns, IColumn::Permutation *& permutation);
 
 private:
     MergeTreeData & data;
diff --git a/src/Storages/MergeTree/MergeTreeIOSettings.h b/src/Storages/MergeTree/MergeTreeIOSettings.h
index 41d9be07c70..7d8dda0c3ca 100644
--- a/src/Storages/MergeTree/MergeTreeIOSettings.h
+++ b/src/Storages/MergeTree/MergeTreeIOSettings.h
@@ -25,6 +25,7 @@ struct MergeTreeWriterSettings
         const MergeTreeSettingsPtr & storage_settings,
         bool can_use_adaptive_granularity_,
         size_t aio_threshold_,
+        bool rewrite_primary_key_,
         bool blocks_are_granules_size_ = false)
         : min_compress_block_size(
             storage_settings->min_compress_block_size ? storage_settings->min_compress_block_size : global_settings.min_compress_block_size)
@@ -33,6 +34,7 @@ struct MergeTreeWriterSettings
                                                         : global_settings.max_compress_block_size)
         , aio_threshold(aio_threshold_)
         , can_use_adaptive_granularity(can_use_adaptive_granularity_)
+        , rewrite_primary_key(rewrite_primary_key_)
         , blocks_are_granules_size(blocks_are_granules_size_)
     {
     }
@@ -41,6 +43,7 @@ struct MergeTreeWriterSettings
     size_t max_compress_block_size;
     size_t aio_threshold;
     bool can_use_adaptive_granularity;
+    bool rewrite_primary_key;
     bool blocks_are_granules_size;
 
     /// true if we write temporary files during alter.
diff --git a/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp b/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp
index bca0d0cb883..8dc353714cb 100644
--- a/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp
+++ b/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp
@@ -7,6 +7,7 @@ namespace DB
 {
 namespace ErrorCodes
 {
+    extern const int LOGICAL_ERROR;
 }
 
 MergeTreeIndexGranularity::MergeTreeIndexGranularity(const std::vector<size_t> & marks_rows_partial_sums_)
@@ -20,6 +21,17 @@ MergeTreeIndexGranularity::MergeTreeIndexGranularity(size_t marks_count, size_t
 {
 }
 
+/// Rows after mark to next mark
+size_t MergeTreeIndexGranularity::getMarkRows(size_t mark_index) const
+{
+    if (mark_index >= getMarksCount())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Trying to get non existing mark {}, while size is {}", mark_index, getMarksCount());
+    if (mark_index == 0)
+        return marks_rows_partial_sums[0];
+    else
+        return marks_rows_partial_sums[mark_index] - marks_rows_partial_sums[mark_index - 1];
+}
+
 size_t MergeTreeIndexGranularity::getMarkStartingRow(size_t mark_index) const
 {
     if (mark_index == 0)
diff --git a/src/Storages/MergeTree/MergeTreeIndexGranularity.h b/src/Storages/MergeTree/MergeTreeIndexGranularity.h
index 5aefd0f102b..0923a496174 100644
--- a/src/Storages/MergeTree/MergeTreeIndexGranularity.h
+++ b/src/Storages/MergeTree/MergeTreeIndexGranularity.h
@@ -47,13 +47,7 @@ public:
     size_t getMarksCountWithoutFinal() const { return getMarksCount() - hasFinalMark(); }
 
     /// Rows after mark to next mark
-    inline size_t getMarkRows(size_t mark_index) const
-    {
-        if (mark_index == 0)
-            return marks_rows_partial_sums[0];
-        else
-            return marks_rows_partial_sums[mark_index] - marks_rows_partial_sums[mark_index - 1];
-    }
+    size_t getMarkRows(size_t mark_index) const;
 
     /// Return amount of rows before mark
     size_t getMarkStartingRow(size_t mark_index) const;
diff --git a/src/Storages/MergeTree/MergedBlockOutputStream.cpp b/src/Storages/MergeTree/MergedBlockOutputStream.cpp
index 00a4c37c60d..e45aaac6169 100644
--- a/src/Storages/MergeTree/MergedBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/MergedBlockOutputStream.cpp
@@ -51,6 +51,7 @@ MergedBlockOutputStream::MergedBlockOutputStream(
         storage.getSettings(),
         data_part->index_granularity_info.is_adaptive,
         aio_threshold,
+        /* rewrite_primary_key = */ true,
         blocks_are_granules_size);
 
     if (aio_threshold > 0 && !merged_column_to_size.empty())
@@ -67,8 +68,6 @@ MergedBlockOutputStream::MergedBlockOutputStream(
         volume->getDisk()->createDirectories(part_path);
 
     writer = data_part->getWriter(columns_list, metadata_snapshot, skip_indices, default_codec, writer_settings);
-    writer->initPrimaryIndex();
-    writer->initSkipIndices();
 }
 
 /// If data is pre-sorted.
@@ -103,9 +102,7 @@ void MergedBlockOutputStream::writeSuffixAndFinalizePart(
         checksums = std::move(*additional_column_checksums);
 
     /// Finish columns serialization.
-    writer->finishDataSerialization(checksums, sync);
-    writer->finishPrimaryIndexSerialization(checksums, sync);
-    writer->finishSkipIndicesSerialization(checksums, sync);
+    writer->finish(checksums, sync);
 
     NamesAndTypesList part_columns;
     if (!total_columns_list)
@@ -219,19 +216,7 @@ void MergedBlockOutputStream::writeImpl(const Block & block, const IColumn::Perm
     if (!rows)
         return;
 
-    std::unordered_set<String> skip_indexes_column_names_set;
-    for (const auto & index : metadata_snapshot->getSecondaryIndices())
-        std::copy(index.column_names.cbegin(), index.column_names.cend(),
-                std::inserter(skip_indexes_column_names_set, skip_indexes_column_names_set.end()));
-    Names skip_indexes_column_names(skip_indexes_column_names_set.begin(), skip_indexes_column_names_set.end());
-
-    Block primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), permutation);
-    Block skip_indexes_block = getBlockAndPermute(block, skip_indexes_column_names, permutation);
-
-    writer->write(block, permutation, primary_key_block, skip_indexes_block);
-    writer->calculateAndSerializeSkipIndices(skip_indexes_block);
-    writer->calculateAndSerializePrimaryIndex(primary_key_block);
-    writer->next();
+    writer->write(block, permutation);
 
     rows_count += rows;
 }
diff --git a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp
index 12b14d13656..87f67ee92a5 100644
--- a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp
+++ b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp
@@ -14,7 +14,7 @@ MergedColumnOnlyOutputStream::MergedColumnOnlyOutputStream(
     const StorageMetadataPtr & metadata_snapshot_,
     const Block & header_,
     CompressionCodecPtr default_codec,
-    const std::vector<MergeTreeIndexPtr> & indices_to_recalc,
+    const MergeTreeIndices & indices_to_recalc,
     WrittenOffsetColumns * offset_columns_,
     const MergeTreeIndexGranularity & index_granularity,
     const MergeTreeIndexGranularityInfo * index_granularity_info)
@@ -28,7 +28,8 @@ MergedColumnOnlyOutputStream::MergedColumnOnlyOutputStream(
         global_settings,
         storage_settings,
         index_granularity_info ? index_granularity_info->is_adaptive : data_part->storage.canUseAdaptiveGranularity(),
-        global_settings.min_bytes_to_use_direct_io);
+        global_settings.min_bytes_to_use_direct_io,
+        /* rewrite_primary_key = */false);
 
     writer = data_part->getWriter(
         header.getNamesAndTypesList(),
@@ -43,25 +44,14 @@ MergedColumnOnlyOutputStream::MergedColumnOnlyOutputStream(
         throw Exception("MergedColumnOnlyOutputStream supports only parts stored on disk", ErrorCodes::NOT_IMPLEMENTED);
 
     writer_on_disk->setWrittenOffsetColumns(offset_columns_);
-    writer_on_disk->initSkipIndices();
 }
 
 void MergedColumnOnlyOutputStream::write(const Block & block)
 {
-    std::unordered_set<String> skip_indexes_column_names_set;
-    for (const auto & index : writer->getSkipIndices())
-        std::copy(index->index.column_names.cbegin(), index->index.column_names.cend(),
-                  std::inserter(skip_indexes_column_names_set, skip_indexes_column_names_set.end()));
-    Names skip_indexes_column_names(skip_indexes_column_names_set.begin(), skip_indexes_column_names_set.end());
-
-    Block skip_indexes_block = getBlockAndPermute(block, skip_indexes_column_names, nullptr);
-
     if (!block.rows())
         return;
 
-    writer->write(block);
-    writer->calculateAndSerializeSkipIndices(skip_indexes_block);
-    writer->next();
+    writer->write(block, nullptr);
 }
 
 void MergedColumnOnlyOutputStream::writeSuffix()
@@ -77,8 +67,7 @@ MergedColumnOnlyOutputStream::writeSuffixAndGetChecksums(
 {
     /// Finish columns serialization.
     MergeTreeData::DataPart::Checksums checksums;
-    writer->finishDataSerialization(checksums, sync);
-    writer->finishSkipIndicesSerialization(checksums, sync);
+    writer->finish(checksums, sync);
 
     auto columns = new_part->getColumns();
 
diff --git a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h
index 507a964ede0..393058dc82e 100644
--- a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h
+++ b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h
@@ -18,7 +18,7 @@ public:
         const StorageMetadataPtr & metadata_snapshot_,
         const Block & header_,
         CompressionCodecPtr default_codec_,
-        const std::vector<MergeTreeIndexPtr> & indices_to_recalc_,
+        const MergeTreeIndices & indices_to_recalc_,
         WrittenOffsetColumns * offset_columns_ = nullptr,
         const MergeTreeIndexGranularity & index_granularity = {},
         const MergeTreeIndexGranularityInfo * index_granularity_info_ = nullptr);
diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp
index 9bf1de5d03c..54248e6032c 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp
@@ -40,7 +40,8 @@ ReplicatedMergeTreeBlockOutputStream::ReplicatedMergeTreeBlockOutputStream(
     size_t quorum_timeout_ms_,
     size_t max_parts_per_block_,
     bool quorum_parallel_,
-    bool deduplicate_)
+    bool deduplicate_,
+    bool optimize_on_insert_)
     : storage(storage_)
     , metadata_snapshot(metadata_snapshot_)
     , quorum(quorum_)
@@ -49,6 +50,7 @@ ReplicatedMergeTreeBlockOutputStream::ReplicatedMergeTreeBlockOutputStream(
     , quorum_parallel(quorum_parallel_)
     , deduplicate(deduplicate_)
     , log(&Poco::Logger::get(storage.getLogName() + " (Replicated OutputStream)"))
+    , optimize_on_insert(optimize_on_insert_)
 {
     /// The quorum value `1` has the same meaning as if it is disabled.
     if (quorum == 1)
@@ -121,8 +123,6 @@ void ReplicatedMergeTreeBlockOutputStream::write(const Block & block)
 {
     last_block_is_duplicate = false;
 
-    storage.delayInsertOrThrowIfNeeded(&storage.partial_shutdown_event);
-
     auto zookeeper = storage.getZooKeeper();
     assertSessionIsNotExpired(zookeeper);
 
@@ -142,7 +142,7 @@ void ReplicatedMergeTreeBlockOutputStream::write(const Block & block)
 
         /// Write part to the filesystem under temporary name. Calculate a checksum.
 
-        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot);
+        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot, optimize_on_insert);
 
         String block_id;
 
@@ -491,6 +491,8 @@ void ReplicatedMergeTreeBlockOutputStream::commitPart(
                 if (!zookeeper->tryGet(quorum_info.status_path, value, nullptr, event))
                     break;
 
+                LOG_TRACE(log, "Quorum node {} still exists, will wait for updates", quorum_info.status_path);
+
                 ReplicatedMergeTreeQuorumEntry quorum_entry(value);
 
                 /// If the node has time to disappear, and then appear again for the next insert.
@@ -499,6 +501,8 @@ void ReplicatedMergeTreeBlockOutputStream::commitPart(
 
                 if (!event->tryWait(quorum_timeout_ms))
                     throw Exception("Timeout while waiting for quorum", ErrorCodes::TIMEOUT_EXCEEDED);
+
+                LOG_TRACE(log, "Quorum {} updated, will check quorum node still exists", quorum_info.status_path);
             }
 
             /// And what if it is possible that the current replica at this time has ceased to be active
@@ -522,7 +526,9 @@ void ReplicatedMergeTreeBlockOutputStream::commitPart(
 
 void ReplicatedMergeTreeBlockOutputStream::writePrefix()
 {
-    storage.throwInsertIfNeeded();
+    /// Only check "too many parts" before write,
+    /// because interrupting long-running INSERT query in the middle is not convenient for users.
+    storage.delayInsertOrThrowIfNeeded(&storage.partial_shutdown_event);
 }
 
 
diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h
index 97c094c1128..3ac2c4bcfcb 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h
@@ -29,7 +29,8 @@ public:
         size_t quorum_timeout_ms_,
         size_t max_parts_per_block_,
         bool quorum_parallel_,
-        bool deduplicate_);
+        bool deduplicate_,
+        bool optimize_on_insert);
 
     Block getHeader() const override;
     void writePrefix() override;
@@ -71,6 +72,8 @@ private:
 
     using Logger = Poco::Logger;
     Poco::Logger * log;
+
+    bool optimize_on_insert;
 };
 
 }
diff --git a/src/Storages/MergeTree/SimpleMergeSelector.h b/src/Storages/MergeTree/SimpleMergeSelector.h
index fe57c40320a..4f277ad74cd 100644
--- a/src/Storages/MergeTree/SimpleMergeSelector.h
+++ b/src/Storages/MergeTree/SimpleMergeSelector.h
@@ -3,6 +3,83 @@
 #include <Storages/MergeTree/MergeSelector.h>
 
 
+/**
+We have a set of data parts that is dynamically changing - new data parts are added and there is background merging process.
+Background merging process periodically selects continuous range of data parts to merge.
+
+It tries to optimize the following metrics:
+1. Write amplification: total amount of data written on disk (source data + merges) relative to the amount of source data.
+It can be also considered as the total amount of work for merges.
+2. The number of data parts in the set at the random moment of time (average + quantiles).
+
+Also taking the following considerations:
+1. Older data parts should be merged less frequently than newer data parts.
+2. Larger data parts should be merged less frequently than smaller data parts.
+3. If no new parts arrive, we should continue to merge existing data parts to eventually optimize the table.
+4. Never allow too many parts, because it will slow down SELECT queries significantly.
+5. Multiple background merges can run concurrently but not too many.
+
+It is not possible to optimize both metrics, because they contradict to each other.
+To lower the number of parts we can merge eagerly but write amplification will increase.
+Then we need some balance between optimization of these two metrics.
+
+But some optimizations may improve both metrics.
+
+For example, we can look at the "merge tree" - the tree of data parts that were merged.
+If the tree is perfectly balanced then its depth is proportonal to the log(data size),
+the total amount of work is proportional to data_size * log(data_size)
+and the write amplification is proportional to log(data_size).
+If it's not balanced (e.g. every new data part is always merged with existing data parts),
+its depth is proportional to the data size, total amount of work is proportional to data_size^2.
+
+We can also control the "base of the logarithm" - you can look it as the number of data parts
+that are merged at once (the tree "arity"). But as the data parts are of different size, we should generalize it:
+calculate the ratio between total size of merged parts to the size of the largest part participated in merge.
+For example, if we merge 4 parts of size 5, 3, 1, 1 - then "base" will be 2 (10 / 5).
+
+Base of the logarithm (simply called `base` in `SimpleMergeSelector`) is the main knob to control the write amplification.
+The more it is, the less is write amplification but we will have more data parts on average.
+
+To fit all the considerations, we also adjust `base` depending on total parts count,
+parts size and parts age, with linear interpolation (then `base` is not a constant but a function of multiple variables,
+looking like a section of hyperplane).
+
+
+Then we apply the algorithm to select the optimal range of data parts to merge.
+There is a proof that this algorithm is optimal if we look in the future only by single step.
+
+The best range of data parts is selected.
+
+We also apply some tunes:
+- there is a fixed const of merging small parts (that is added to the size of data part before all estimations);
+- there are some heuristics to "stick" ranges to large data parts.
+
+It's still unclear if this algorithm is good or optimal at all. It's unclear if this algorithm is using the optimal coefficients.
+
+To test and optimize SimpleMergeSelector, we apply the following methods:
+- insert/merge simulator: a model simulating parts insertion and merging;
+merge selecting algorithm is applied and the relevant metrics are calculated to allow to tune the algorithm;
+- insert/merge simulator on real `system.part_log` from production - it gives realistic information about inserted data parts:
+their sizes, at what time intervals they are inserted;
+
+There is a research thesis dedicated to optimization of merge algorithm:
+https://presentations.clickhouse.tech/hse_2019/merge_algorithm.pptx
+
+This work made attempt to variate the coefficients in SimpleMergeSelector and to solve the optimization task:
+maybe some change in coefficients will give a clear win on all metrics. Surprisingly enough, it has found
+that our selection of coefficients is near optimal. It has found slightly more optimal coefficients,
+but I decided not to use them, because the representativeness of the test data is in question.
+
+This work did not make any attempt to propose any other algorithm.
+This work did not make any attempt to analyze the task with analytical methods.
+That's why I still believe that there are many opportunities to optimize the merge selection algorithm.
+
+Please do not mix the task with a similar task in other LSM-based systems (like RocksDB).
+Their problem statement is subtly different. Our set of data parts is consisted of data parts
+that are completely independent in stored data. Ranges of primary keys in data parts can intersect.
+When doing SELECT we read from all data parts. INSERTed data parts comes with unknown size...
+*/
+
 namespace DB
 {
 
diff --git a/src/Storages/SetSettings.h b/src/Storages/SetSettings.h
index c5926e409c2..8e4706f2b9e 100644
--- a/src/Storages/SetSettings.h
+++ b/src/Storages/SetSettings.h
@@ -10,7 +10,8 @@ class ASTStorage;
 
 
 #define SET_RELATED_SETTINGS(M) \
-    M(Bool, persistent, true, "Disable setting to avoid the overhead of writing to disk for StorageSet", 0)
+    M(Bool, persistent, true, "Disable setting to avoid the overhead of writing to disk for StorageSet", 0) \
+    M(String, disk, "default", "Name of the disk used to persist set data", 0)
 
 #define LIST_OF_SET_SETTINGS(M) \
     SET_RELATED_SETTINGS(M) \
diff --git a/src/Storages/StorageDictionary.cpp b/src/Storages/StorageDictionary.cpp
index 4896ace4e7a..32fe7b4c026 100644
--- a/src/Storages/StorageDictionary.cpp
+++ b/src/Storages/StorageDictionary.cpp
@@ -121,9 +121,9 @@ StorageDictionary::StorageDictionary(
 void StorageDictionary::checkTableCanBeDropped() const
 {
     if (location == Location::SameDatabaseAndNameAsDictionary)
-        throw Exception("Cannot detach dictionary " + backQuote(dictionary_name) + " as table, use DETACH DICTIONARY query", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
+        throw Exception("Cannot drop/detach dictionary " + backQuote(dictionary_name) + " as table, use DROP DICTIONARY or DETACH DICTIONARY query instead", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
     if (location == Location::DictionaryDatabase)
-        throw Exception("Cannot detach table " + getStorageID().getFullTableName() + " from a database with DICTIONARY engine", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
+        throw Exception("Cannot drop/detach table " + getStorageID().getFullTableName() + " from a database with DICTIONARY engine", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
 }
 
 void StorageDictionary::checkTableCanBeDetached() const
diff --git a/src/Storages/StorageJoin.cpp b/src/Storages/StorageJoin.cpp
index 728f16ec6d8..8d4f0b3b3be 100644
--- a/src/Storages/StorageJoin.cpp
+++ b/src/Storages/StorageJoin.cpp
@@ -8,6 +8,7 @@
 #include <Core/ColumnNumbers.h>
 #include <DataStreams/IBlockInputStream.h>
 #include <DataTypes/NestedUtils.h>
+#include <Disks/IDisk.h>
 #include <Interpreters/joinDispatch.h>
 #include <Interpreters/TableJoin.h>
 #include <Interpreters/castColumn.h>
@@ -35,6 +36,7 @@ namespace ErrorCodes
 }
 
 StorageJoin::StorageJoin(
+    DiskPtr disk_,
     const String & relative_path_,
     const StorageID & table_id_,
     const Names & key_names_,
@@ -45,9 +47,8 @@ StorageJoin::StorageJoin(
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
     bool overwrite_,
-    const Context & context_,
     bool persistent_)
-    : StorageSetOrJoinBase{relative_path_, table_id_, columns_, constraints_, context_, persistent_}
+    : StorageSetOrJoinBase{disk_, relative_path_, table_id_, columns_, constraints_, persistent_}
     , key_names(key_names_)
     , use_nulls(use_nulls_)
     , limits(limits_)
@@ -69,9 +70,9 @@ StorageJoin::StorageJoin(
 void StorageJoin::truncate(
     const ASTPtr &, const StorageMetadataPtr & metadata_snapshot, const Context &, TableExclusiveLockHolder&)
 {
-    Poco::File(path).remove(true);
-    Poco::File(path).createDirectories();
-    Poco::File(path + "tmp/").createDirectories();
+    disk->removeRecursive(path);
+    disk->createDirectories(path);
+    disk->createDirectories(path + "tmp/");
 
     increment = 0;
     join = std::make_shared<HashJoin>(table_join, metadata_snapshot->getSampleBlock().sortColumns(), overwrite);
@@ -124,6 +125,7 @@ void registerStorageJoin(StorageFactory & factory)
         auto join_any_take_last_row = settings.join_any_take_last_row;
         auto old_any_join = settings.any_join_distinct_right_table_keys;
         bool persistent = true;
+        String disk_name = "default";
 
         if (args.storage_def && args.storage_def->settings)
         {
@@ -141,6 +143,8 @@ void registerStorageJoin(StorageFactory & factory)
                     join_any_take_last_row = setting.value;
                 else if (setting.name == "any_join_distinct_right_table_keys")
                     old_any_join = setting.value;
+                else if (setting.name == "disk")
+                    disk_name = setting.value.get<String>();
                 else if (setting.name == "persistent")
                 {
                     auto join_settings = std::make_unique<JoinSettings>();
@@ -148,12 +152,12 @@ void registerStorageJoin(StorageFactory & factory)
                     persistent = join_settings->persistent;
                 }
                 else
-                    throw Exception(
-                        "Unknown setting " + setting.name + " for storage " + args.engine_name,
-                        ErrorCodes::BAD_ARGUMENTS);
+                    throw Exception("Unknown setting " + setting.name + " for storage " + args.engine_name, ErrorCodes::BAD_ARGUMENTS);
             }
         }
 
+        DiskPtr disk = args.context.getDisk(disk_name);
+
         if (engine_args.size() < 3)
             throw Exception(
                 "Storage Join requires at least 3 parameters: Join(ANY|ALL|SEMI|ANTI, LEFT|INNER|RIGHT, keys...).",
@@ -219,6 +223,7 @@ void registerStorageJoin(StorageFactory & factory)
         }
 
         return StorageJoin::create(
+            disk,
             args.relative_data_path,
             args.table_id,
             key_names,
@@ -229,7 +234,6 @@ void registerStorageJoin(StorageFactory & factory)
             args.columns,
             args.constraints,
             join_any_take_last_row,
-            args.context,
             persistent);
     };
 
diff --git a/src/Storages/StorageJoin.h b/src/Storages/StorageJoin.h
index df250b89cb1..c453c036b65 100644
--- a/src/Storages/StorageJoin.h
+++ b/src/Storages/StorageJoin.h
@@ -67,6 +67,7 @@ private:
 
 protected:
     StorageJoin(
+        DiskPtr disk_,
         const String & relative_path_,
         const StorageID & table_id_,
         const Names & key_names_,
@@ -76,7 +77,6 @@ protected:
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
         bool overwrite,
-        const Context & context_,
         bool persistent_);
 };
 
diff --git a/src/Storages/StorageMaterializeMySQL.cpp b/src/Storages/StorageMaterializeMySQL.cpp
index 9be285adc69..721221e3fdc 100644
--- a/src/Storages/StorageMaterializeMySQL.cpp
+++ b/src/Storages/StorageMaterializeMySQL.cpp
@@ -21,12 +21,13 @@
 #include <Processors/Pipe.h>
 #include <Processors/Transforms/FilterTransform.h>
 
+#include <Databases/MySQL/DatabaseMaterializeMySQL.h>
 #include <Storages/SelectQueryInfo.h>
 
 namespace DB
 {
 
-StorageMaterializeMySQL::StorageMaterializeMySQL(const StoragePtr & nested_storage_, const DatabaseMaterializeMySQL * database_)
+StorageMaterializeMySQL::StorageMaterializeMySQL(const StoragePtr & nested_storage_, const IDatabase * database_)
     : StorageProxy(nested_storage_->getStorageID()), nested_storage(nested_storage_), database(database_)
 {
     auto nested_memory_metadata = nested_storage->getInMemoryMetadata();
@@ -45,7 +46,7 @@ Pipe StorageMaterializeMySQL::read(
     unsigned int num_streams)
 {
     /// If the background synchronization thread has exception.
-    database->rethrowExceptionIfNeed();
+    rethrowSyncExceptionIfNeed(database);
 
     NameSet column_names_set = NameSet(column_names.begin(), column_names.end());
     auto lock = nested_storage->lockForShare(context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
@@ -106,7 +107,7 @@ Pipe StorageMaterializeMySQL::read(
 NamesAndTypesList StorageMaterializeMySQL::getVirtuals() const
 {
     /// If the background synchronization thread has exception.
-    database->rethrowExceptionIfNeed();
+    rethrowSyncExceptionIfNeed(database);
     return nested_storage->getVirtuals();
 }
 
diff --git a/src/Storages/StorageMaterializeMySQL.h b/src/Storages/StorageMaterializeMySQL.h
index ff86c7abdc2..ea90c1ffc9e 100644
--- a/src/Storages/StorageMaterializeMySQL.h
+++ b/src/Storages/StorageMaterializeMySQL.h
@@ -5,7 +5,6 @@
 #if USE_MYSQL
 
 #include <Storages/StorageProxy.h>
-#include <Databases/MySQL/DatabaseMaterializeMySQL.h>
 
 namespace DB
 {
@@ -21,7 +20,7 @@ class StorageMaterializeMySQL final : public ext::shared_ptr_helper<StorageMater
 public:
     String getName() const override { return "MaterializeMySQL"; }
 
-    StorageMaterializeMySQL(const StoragePtr & nested_storage_, const DatabaseMaterializeMySQL * database_);
+    StorageMaterializeMySQL(const StoragePtr & nested_storage_, const IDatabase * database_);
 
     Pipe read(
         const Names & column_names, const StorageMetadataPtr & metadata_snapshot, SelectQueryInfo & query_info,
@@ -32,15 +31,18 @@ public:
     NamesAndTypesList getVirtuals() const override;
     ColumnSizeByName getColumnSizes() const override;
 
-private:
     StoragePtr getNested() const override { return nested_storage; }
+
+    void drop() override { nested_storage->drop(); }
+
+private:
     [[noreturn]] void throwNotAllowed() const
     {
         throw Exception("This method is not allowed for MaterializeMySQL", ErrorCodes::NOT_IMPLEMENTED);
     }
 
     StoragePtr nested_storage;
-    const DatabaseMaterializeMySQL * database;
+    const IDatabase * database;
 };
 
 }
diff --git a/src/Storages/StorageMerge.cpp b/src/Storages/StorageMerge.cpp
index 97f4ccd0bba..63d28eaff77 100644
--- a/src/Storages/StorageMerge.cpp
+++ b/src/Storages/StorageMerge.cpp
@@ -72,11 +72,27 @@ StorageMerge::StorageMerge(
     const StorageID & table_id_,
     const ColumnsDescription & columns_,
     const String & source_database_,
-    const String & table_name_regexp_,
+    const Strings & source_tables_,
     const Context & context_)
     : IStorage(table_id_)
     , source_database(source_database_)
-    , table_name_regexp(table_name_regexp_)
+    , source_tables(std::in_place, source_tables_.begin(), source_tables_.end())
+    , global_context(context_.getGlobalContext())
+{
+    StorageInMemoryMetadata storage_metadata;
+    storage_metadata.setColumns(columns_);
+    setInMemoryMetadata(storage_metadata);
+}
+
+StorageMerge::StorageMerge(
+    const StorageID & table_id_,
+    const ColumnsDescription & columns_,
+    const String & source_database_,
+    const String & source_table_regexp_,
+    const Context & context_)
+    : IStorage(table_id_)
+    , source_database(source_database_)
+    , source_table_regexp(source_table_regexp_)
     , global_context(context_.getGlobalContext())
 {
     StorageInMemoryMetadata storage_metadata;
@@ -439,8 +455,17 @@ DatabaseTablesIteratorPtr StorageMerge::getDatabaseIterator(const Context & cont
         e.addMessage("while getting table iterator of Merge table. Maybe caused by two Merge tables that will endlessly try to read each other's data");
         throw;
     }
+
     auto database = DatabaseCatalog::instance().getDatabase(source_database);
-    auto table_name_match = [this](const String & table_name_) { return table_name_regexp.match(table_name_); };
+
+    auto table_name_match = [this](const String & table_name_) -> bool
+    {
+        if (source_tables)
+            return source_tables->count(table_name_);
+        else
+            return source_table_regexp->match(table_name_);
+    };
+
     return database->getTablesIterator(context, table_name_match);
 }
 
diff --git a/src/Storages/StorageMerge.h b/src/Storages/StorageMerge.h
index 681ea7015e7..223383a6975 100644
--- a/src/Storages/StorageMerge.h
+++ b/src/Storages/StorageMerge.h
@@ -48,7 +48,8 @@ public:
 
 private:
     String source_database;
-    OptimizedRegularExpression table_name_regexp;
+    std::optional<std::unordered_set<String>> source_tables;
+    std::optional<OptimizedRegularExpression> source_table_regexp;
     const Context & global_context;
 
     using StorageWithLockAndName = std::tuple<StoragePtr, TableLockHolder, String>;
@@ -72,7 +73,14 @@ protected:
         const StorageID & table_id_,
         const ColumnsDescription & columns_,
         const String & source_database_,
-        const String & table_name_regexp_,
+        const Strings & source_tables_,
+        const Context & context_);
+
+    StorageMerge(
+        const StorageID & table_id_,
+        const ColumnsDescription & columns_,
+        const String & source_database_,
+        const String & source_table_regexp_,
         const Context & context_);
 
     Pipe createSources(
diff --git a/src/Storages/StorageMergeTree.cpp b/src/Storages/StorageMergeTree.cpp
index eb28ccfa3d5..591f9840896 100644
--- a/src/Storages/StorageMergeTree.cpp
+++ b/src/Storages/StorageMergeTree.cpp
@@ -233,7 +233,7 @@ BlockOutputStreamPtr StorageMergeTree::write(const ASTPtr & /*query*/, const Sto
 
     const auto & settings = context.getSettingsRef();
     return std::make_shared<MergeTreeBlockOutputStream>(
-        *this, metadata_snapshot, settings.max_partitions_per_insert_block);
+        *this, metadata_snapshot, settings.max_partitions_per_insert_block, context.getSettingsRef().optimize_on_insert);
 }
 
 void StorageMergeTree::checkTableCanBeDropped() const
diff --git a/src/Storages/StorageReplicatedMergeTree.cpp b/src/Storages/StorageReplicatedMergeTree.cpp
index 31b04664b17..f91d83b6811 100644
--- a/src/Storages/StorageReplicatedMergeTree.cpp
+++ b/src/Storages/StorageReplicatedMergeTree.cpp
@@ -3235,6 +3235,8 @@ void StorageReplicatedMergeTree::updateQuorum(const String & part_name, bool is_
         ReplicatedMergeTreeQuorumEntry quorum_entry(value);
         if (quorum_entry.part_name != part_name)
         {
+            LOG_TRACE(log, "Quorum {}, already achieved for part {} current part {}",
+                      quorum_status_path, part_name, quorum_entry.part_name);
             /// The quorum has already been achieved. Moreover, another INSERT with a quorum has already started.
             break;
         }
@@ -3244,6 +3246,8 @@ void StorageReplicatedMergeTree::updateQuorum(const String & part_name, bool is_
         if (quorum_entry.replicas.size() >= quorum_entry.required_number_of_replicas)
         {
             /// The quorum is reached. Delete the node, and update information about the last part that was successfully written with quorum.
+            LOG_TRACE(log, "Got {} replicas confirmed quorum {}, going to remove node",
+                      quorum_entry.replicas.size(), quorum_status_path);
 
             Coordination::Requests ops;
             Coordination::Responses responses;
@@ -3291,6 +3295,8 @@ void StorageReplicatedMergeTree::updateQuorum(const String & part_name, bool is_
         }
         else
         {
+            LOG_TRACE(log, "Quorum {} still not satisfied (have only {} replicas), updating node",
+                      quorum_status_path, quorum_entry.replicas.size());
             /// We update the node, registering there one more replica.
             auto code = zookeeper->trySet(quorum_status_path, quorum_entry.toString(), stat.version);
 
@@ -3861,7 +3867,8 @@ BlockOutputStreamPtr StorageReplicatedMergeTree::write(const ASTPtr & /*query*/,
         query_settings.insert_quorum_timeout.totalMilliseconds(),
         query_settings.max_partitions_per_insert_block,
         query_settings.insert_quorum_parallel,
-        deduplicate);
+        deduplicate,
+        context.getSettingsRef().optimize_on_insert);
 }
 
 
@@ -4444,7 +4451,7 @@ PartitionCommandsResultInfo StorageReplicatedMergeTree::attachPartition(
     PartsTemporaryRename renamed_parts(*this, "detached/");
     MutableDataPartsVector loaded_parts = tryLoadPartsToAttach(partition, attach_part, query_context, renamed_parts);
 
-    ReplicatedMergeTreeBlockOutputStream output(*this, metadata_snapshot, 0, 0, 0, false, false);   /// TODO Allow to use quorum here.
+    ReplicatedMergeTreeBlockOutputStream output(*this, metadata_snapshot, 0, 0, 0, false, false, false);   /// TODO Allow to use quorum here.
     for (size_t i = 0; i < loaded_parts.size(); ++i)
     {
         String old_name = loaded_parts[i]->name;
diff --git a/src/Storages/StorageReplicatedMergeTree.h b/src/Storages/StorageReplicatedMergeTree.h
index f1094793944..d396f32dcca 100644
--- a/src/Storages/StorageReplicatedMergeTree.h
+++ b/src/Storages/StorageReplicatedMergeTree.h
@@ -39,7 +39,8 @@ namespace DB
   * - the structure of the table (/metadata, /columns)
   * - action log with data (/log/log-...,/replicas/replica_name/queue/queue-...);
   * - a replica list (/replicas), and replica activity tag (/replicas/replica_name/is_active), replica addresses (/replicas/replica_name/host);
-  * - select the leader replica (/leader_election) - this is the replica that assigns the merge;
+  * - select the leader replica (/leader_election) - these are the replicas that assigning merges, mutations and partition manipulations
+  *   (after ClickHouse version 20.5 we allow multiple leaders to act concurrently);
   * - a set of parts of data on each replica (/replicas/replica_name/parts);
   * - list of the last N blocks of data with checksum, for deduplication (/blocks);
   * - the list of incremental block numbers (/block_numbers) that we are about to insert,
diff --git a/src/Storages/StorageS3.cpp b/src/Storages/StorageS3.cpp
index fff44bb1f4c..0465cfdffba 100644
--- a/src/Storages/StorageS3.cpp
+++ b/src/Storages/StorageS3.cpp
@@ -141,17 +141,18 @@ namespace
     public:
         StorageS3BlockOutputStream(
             const String & format,
-            UInt64 min_upload_part_size,
             const Block & sample_block_,
             const Context & context,
             const CompressionMethod compression_method,
             const std::shared_ptr<Aws::S3::S3Client> & client,
             const String & bucket,
-            const String & key)
+            const String & key,
+            size_t min_upload_part_size,
+            size_t max_single_part_upload_size)
             : sample_block(sample_block_)
         {
             write_buf = wrapWriteBufferWithCompressionMethod(
-                std::make_unique<WriteBufferFromS3>(client, bucket, key, min_upload_part_size, true), compression_method, 3);
+                std::make_unique<WriteBufferFromS3>(client, bucket, key, min_upload_part_size, max_single_part_upload_size), compression_method, 3);
             writer = FormatFactory::instance().getOutput(format, *write_buf, sample_block, context);
         }
 
@@ -192,6 +193,7 @@ StorageS3::StorageS3(
     const StorageID & table_id_,
     const String & format_name_,
     UInt64 min_upload_part_size_,
+    UInt64 max_single_part_upload_size_,
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
     const Context & context_,
@@ -201,6 +203,7 @@ StorageS3::StorageS3(
     , global_context(context_.getGlobalContext())
     , format_name(format_name_)
     , min_upload_part_size(min_upload_part_size_)
+    , max_single_part_upload_size(max_single_part_upload_size_)
     , compression_method(compression_method_)
     , name(uri_.storage_name)
 {
@@ -331,9 +334,15 @@ Pipe StorageS3::read(
 BlockOutputStreamPtr StorageS3::write(const ASTPtr & /*query*/, const StorageMetadataPtr & metadata_snapshot, const Context & /*context*/)
 {
     return std::make_shared<StorageS3BlockOutputStream>(
-        format_name, min_upload_part_size, metadata_snapshot->getSampleBlock(),
-        global_context, chooseCompressionMethod(uri.endpoint, compression_method),
-        client, uri.bucket, uri.key);
+        format_name,
+        metadata_snapshot->getSampleBlock(),
+        global_context,
+        chooseCompressionMethod(uri.endpoint, compression_method),
+        client,
+        uri.bucket,
+        uri.key,
+        min_upload_part_size,
+        max_single_part_upload_size);
 }
 
 void registerStorageS3Impl(const String & name, StorageFactory & factory)
@@ -362,6 +371,7 @@ void registerStorageS3Impl(const String & name, StorageFactory & factory)
         }
 
         UInt64 min_upload_part_size = args.local_context.getSettingsRef().s3_min_upload_part_size;
+        UInt64 max_single_part_upload_size = args.local_context.getSettingsRef().s3_max_single_part_upload_size;
 
         String compression_method;
         String format_name;
@@ -383,6 +393,7 @@ void registerStorageS3Impl(const String & name, StorageFactory & factory)
             args.table_id,
             format_name,
             min_upload_part_size,
+            max_single_part_upload_size,
             args.columns,
             args.constraints,
             args.context,
diff --git a/src/Storages/StorageS3.h b/src/Storages/StorageS3.h
index 96f0cf02e88..f436fb85c90 100644
--- a/src/Storages/StorageS3.h
+++ b/src/Storages/StorageS3.h
@@ -31,6 +31,7 @@ public:
         const StorageID & table_id_,
         const String & format_name_,
         UInt64 min_upload_part_size_,
+        UInt64 max_single_part_upload_size_,
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
         const Context & context_,
@@ -59,7 +60,8 @@ private:
     const Context & global_context;
 
     String format_name;
-    UInt64 min_upload_part_size;
+    size_t min_upload_part_size;
+    size_t max_single_part_upload_size;
     String compression_method;
     std::shared_ptr<Aws::S3::S3Client> client;
     String name;
diff --git a/src/Storages/StorageSet.cpp b/src/Storages/StorageSet.cpp
index d691a185e1e..e518c7da0e4 100644
--- a/src/Storages/StorageSet.cpp
+++ b/src/Storages/StorageSet.cpp
@@ -6,12 +6,11 @@
 #include <Compression/CompressedWriteBuffer.h>
 #include <DataStreams/NativeBlockOutputStream.h>
 #include <DataStreams/NativeBlockInputStream.h>
+#include <Disks/IDisk.h>
 #include <Common/formatReadable.h>
-#include <Common/escapeForFileName.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <Interpreters/Set.h>
 #include <Interpreters/Context.h>
-#include <Poco/DirectoryIterator.h>
 #include <Parsers/ASTCreateQuery.h>
 #include <Parsers/ASTLiteral.h>
 
@@ -49,7 +48,7 @@ private:
     String backup_path;
     String backup_tmp_path;
     String backup_file_name;
-    WriteBufferFromFile backup_buf;
+    std::unique_ptr<WriteBufferFromFileBase> backup_buf;
     CompressedWriteBuffer compressed_backup_buf;
     NativeBlockOutputStream backup_stream;
     bool persistent;
@@ -68,8 +67,8 @@ SetOrJoinBlockOutputStream::SetOrJoinBlockOutputStream(
     , backup_path(backup_path_)
     , backup_tmp_path(backup_tmp_path_)
     , backup_file_name(backup_file_name_)
-    , backup_buf(backup_tmp_path + backup_file_name)
-    , compressed_backup_buf(backup_buf)
+    , backup_buf(table_.disk->writeFile(backup_tmp_path + backup_file_name))
+    , compressed_backup_buf(*backup_buf)
     , backup_stream(compressed_backup_buf, 0, metadata_snapshot->getSampleBlock())
     , persistent(persistent_)
 {
@@ -92,9 +91,10 @@ void SetOrJoinBlockOutputStream::writeSuffix()
     {
         backup_stream.flush();
         compressed_backup_buf.next();
-        backup_buf.next();
+        backup_buf->next();
+        backup_buf->finalize();
 
-        Poco::File(backup_tmp_path + backup_file_name).renameTo(backup_path + backup_file_name);
+        table.disk->replaceFile(backup_tmp_path + backup_file_name, backup_path + backup_file_name);
     }
 }
 
@@ -107,13 +107,14 @@ BlockOutputStreamPtr StorageSetOrJoinBase::write(const ASTPtr & /*query*/, const
 
 
 StorageSetOrJoinBase::StorageSetOrJoinBase(
+    DiskPtr disk_,
     const String & relative_path_,
     const StorageID & table_id_,
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
-    const Context & context_,
     bool persistent_)
     : IStorage(table_id_),
+    disk(disk_),
     persistent(persistent_)
 {
     StorageInMemoryMetadata storage_metadata;
@@ -125,19 +126,18 @@ StorageSetOrJoinBase::StorageSetOrJoinBase(
     if (relative_path_.empty())
         throw Exception("Join and Set storages require data path", ErrorCodes::INCORRECT_FILE_NAME);
 
-    base_path = context_.getPath();
-    path = base_path + relative_path_;
+    path = relative_path_;
 }
 
 
 StorageSet::StorageSet(
+    DiskPtr disk_,
     const String & relative_path_,
     const StorageID & table_id_,
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
-    const Context & context_,
     bool persistent_)
-    : StorageSetOrJoinBase{relative_path_, table_id_, columns_, constraints_, context_, persistent_},
+    : StorageSetOrJoinBase{disk_, relative_path_, table_id_, columns_, constraints_, persistent_},
     set(std::make_shared<Set>(SizeLimits(), false, true))
 {
 
@@ -158,9 +158,9 @@ std::optional<UInt64> StorageSet::totalBytes(const Settings &) const { return se
 
 void StorageSet::truncate(const ASTPtr &, const StorageMetadataPtr & metadata_snapshot, const Context &, TableExclusiveLockHolder &)
 {
-    Poco::File(path).remove(true);
-    Poco::File(path).createDirectories();
-    Poco::File(path + "tmp/").createDirectories();
+    disk->removeRecursive(path);
+    disk->createDirectories(path);
+    disk->createDirectories(path + "tmp/");
 
     Block header = metadata_snapshot->getSampleBlock();
     header = header.sortColumns();
@@ -173,24 +173,23 @@ void StorageSet::truncate(const ASTPtr &, const StorageMetadataPtr & metadata_sn
 
 void StorageSetOrJoinBase::restore()
 {
-    Poco::File tmp_dir(path + "tmp/");
-    if (!tmp_dir.exists())
+    if (!disk->exists(path + "tmp/"))
     {
-        tmp_dir.createDirectories();
+        disk->createDirectories(path + "tmp/");
         return;
     }
 
     static const char * file_suffix = ".bin";
     static const auto file_suffix_size = strlen(".bin");
 
-    Poco::DirectoryIterator dir_end;
-    for (Poco::DirectoryIterator dir_it(path); dir_end != dir_it; ++dir_it)
+    for (auto dir_it{disk->iterateDirectory(path)}; dir_it->isValid(); dir_it->next())
     {
-        const auto & name = dir_it.name();
+        const auto & name = dir_it->name();
+        const auto & file_path = dir_it->path();
 
-        if (dir_it->isFile()
+        if (disk->isFile(file_path)
             && endsWith(name, file_suffix)
-            && dir_it->getSize() > 0)
+            && disk->getFileSize(file_path) > 0)
         {
             /// Calculate the maximum number of available files with a backup to add the following files with large numbers.
             UInt64 file_num = parse<UInt64>(name.substr(0, name.size() - file_suffix_size));
@@ -205,8 +204,8 @@ void StorageSetOrJoinBase::restore()
 
 void StorageSetOrJoinBase::restoreFromFile(const String & file_path)
 {
-    ReadBufferFromFile backup_buf(file_path);
-    CompressedReadBuffer compressed_backup_buf(backup_buf);
+    auto backup_buf = disk->readFile(file_path);
+    CompressedReadBuffer compressed_backup_buf(*backup_buf);
     NativeBlockInputStream backup_stream(compressed_backup_buf, 0);
 
     backup_stream.readPrefix();
@@ -226,10 +225,9 @@ void StorageSetOrJoinBase::restoreFromFile(const String & file_path)
 void StorageSetOrJoinBase::rename(const String & new_path_to_table_data, const StorageID & new_table_id)
 {
     /// Rename directory with data.
-    String new_path = base_path + new_path_to_table_data;
-    Poco::File(path).renameTo(new_path);
+    disk->replaceFile(path, new_path_to_table_data);
 
-    path = new_path;
+    path = new_path_to_table_data;
     renameInMemory(new_table_id);
 }
 
@@ -251,7 +249,8 @@ void registerStorageSet(StorageFactory & factory)
             set_settings->loadFromQuery(*args.storage_def);
         }
 
-        return StorageSet::create(args.relative_data_path, args.table_id, args.columns, args.constraints, args.context, set_settings->persistent);
+        DiskPtr disk = args.context.getDisk(set_settings->disk);
+        return StorageSet::create(disk, args.relative_data_path, args.table_id, args.columns, args.constraints, set_settings->persistent);
     }, StorageFactory::StorageFeatures{ .supports_settings = true, });
 }
 
diff --git a/src/Storages/StorageSet.h b/src/Storages/StorageSet.h
index 0f976fce17d..9b9078f7dd5 100644
--- a/src/Storages/StorageSet.h
+++ b/src/Storages/StorageSet.h
@@ -2,6 +2,7 @@
 
 #include <ext/shared_ptr_helper.h>
 
+#include <Interpreters/Context.h>
 #include <Storages/IStorage.h>
 #include <Storages/SetSettings.h>
 
@@ -29,14 +30,14 @@ public:
 
 protected:
     StorageSetOrJoinBase(
+        DiskPtr disk_,
         const String & relative_path_,
         const StorageID & table_id_,
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
-        const Context & context_,
         bool persistent_);
 
-    String base_path;
+    DiskPtr disk;
     String path;
     bool persistent;
 
@@ -85,11 +86,11 @@ private:
 
 protected:
     StorageSet(
+        DiskPtr disk_,
         const String & relative_path_,
         const StorageID & table_id_,
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
-        const Context & context_,
         bool persistent_);
 };
 
diff --git a/src/Storages/StorageStripeLog.cpp b/src/Storages/StorageStripeLog.cpp
index a265df856c1..bc6afffddeb 100644
--- a/src/Storages/StorageStripeLog.cpp
+++ b/src/Storages/StorageStripeLog.cpp
@@ -40,8 +40,6 @@
 namespace DB
 {
 
-#define INDEX_BUFFER_SIZE 4096
-
 namespace ErrorCodes
 {
     extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
@@ -319,7 +317,7 @@ Pipe StorageStripeLog::read(
         return Pipe(std::make_shared<NullSource>(metadata_snapshot->getSampleBlockForColumns(column_names, getVirtuals(), getStorageID())));
     }
 
-    CompressedReadBufferFromFile index_in(disk->readFile(index_file, INDEX_BUFFER_SIZE));
+    CompressedReadBufferFromFile index_in(disk->readFile(index_file, 4096));
     std::shared_ptr<const IndexForNativeFormat> index{std::make_shared<IndexForNativeFormat>(index_in, column_names_set)};
 
     size_t size = index->blocks.size();
diff --git a/src/Storages/System/StorageSystemColumns.cpp b/src/Storages/System/StorageSystemColumns.cpp
index b0371fa12de..6726d502071 100644
--- a/src/Storages/System/StorageSystemColumns.cpp
+++ b/src/Storages/System/StorageSystemColumns.cpp
@@ -272,17 +272,28 @@ Pipe StorageSystemColumns::read(
     Pipes pipes;
 
     {
-        Databases databases = DatabaseCatalog::instance().getDatabases();
-
         /// Add `database` column.
         MutableColumnPtr database_column_mut = ColumnString::create();
-        for (const auto & database : databases)
+
+        const auto databases = DatabaseCatalog::instance().getDatabases();
+        for (const auto & [database_name, database] : databases)
         {
+            if (database_name == DatabaseCatalog::TEMPORARY_DATABASE)
+                continue; /// We don't want to show the internal database for temporary tables in system.columns
+
             /// We are skipping "Lazy" database because we cannot afford initialization of all its tables.
             /// This should be documented.
 
-            if (database.second->getEngineName() != "Lazy")
-                database_column_mut->insert(database.first);
+            if (database->getEngineName() != "Lazy")
+                database_column_mut->insert(database_name);
+        }
+
+        Tables external_tables;
+        if (context.hasSessionContext())
+        {
+            external_tables = context.getSessionContext().getExternalTables();
+            if (!external_tables.empty())
+                database_column_mut->insertDefault(); /// Empty database for external tables.
         }
 
         block_to_filter.insert(ColumnWithTypeAndName(std::move(database_column_mut), std::make_shared<DataTypeString>(), "database"));
@@ -297,29 +308,36 @@ Pipe StorageSystemColumns::read(
         }
 
         ColumnPtr & database_column = block_to_filter.getByName("database").column;
-        size_t rows = database_column->size();
 
         /// Add `table` column.
         MutableColumnPtr table_column_mut = ColumnString::create();
-        IColumn::Offsets offsets(rows);
-        for (size_t i = 0; i < rows; ++i)
+        IColumn::Offsets offsets(database_column->size());
+
+        for (size_t i = 0; i < database_column->size(); ++i)
         {
             const std::string database_name = (*database_column)[i].get<std::string>();
-            const DatabasePtr database = databases.at(database_name);
-            offsets[i] = i ? offsets[i - 1] : 0;
-
-            for (auto iterator = database->getTablesIterator(context); iterator->isValid(); iterator->next())
+            if (database_name.empty())
             {
-                if (const auto & table = iterator->table())
+                for (auto & [table_name, table] : external_tables)
                 {
-                    const String & table_name = iterator->name();
-                    storages.emplace(std::piecewise_construct,
-                        std::forward_as_tuple(database_name, table_name),
-                        std::forward_as_tuple(table));
+                    storages[{"", table_name}] = table;
                     table_column_mut->insert(table_name);
-                    ++offsets[i];
                 }
             }
+            else
+            {
+                const DatabasePtr & database = databases.at(database_name);
+                for (auto iterator = database->getTablesIterator(context); iterator->isValid(); iterator->next())
+                {
+                    if (const auto & table = iterator->table())
+                    {
+                        const String & table_name = iterator->name();
+                        storages[{database_name, table_name}] = table;
+                        table_column_mut->insert(table_name);
+                    }
+                }
+            }
+            offsets[i] = table_column_mut->size();
         }
 
         database_column = database_column->replicate(offsets);
diff --git a/src/Storages/System/StorageSystemDatabases.cpp b/src/Storages/System/StorageSystemDatabases.cpp
index 2a89bf34186..88ac987014d 100644
--- a/src/Storages/System/StorageSystemDatabases.cpp
+++ b/src/Storages/System/StorageSystemDatabases.cpp
@@ -25,17 +25,20 @@ void StorageSystemDatabases::fillData(MutableColumns & res_columns, const Contex
     const auto access = context.getAccess();
     const bool check_access_for_databases = !access->isGranted(AccessType::SHOW_DATABASES);
 
-    auto databases = DatabaseCatalog::instance().getDatabases();
-    for (const auto & database : databases)
+    const auto databases = DatabaseCatalog::instance().getDatabases();
+    for (const auto & [database_name, database] : databases)
     {
-        if (check_access_for_databases && !access->isGranted(AccessType::SHOW_DATABASES, database.first))
+        if (check_access_for_databases && !access->isGranted(AccessType::SHOW_DATABASES, database_name))
             continue;
 
-        res_columns[0]->insert(database.first);
-        res_columns[1]->insert(database.second->getEngineName());
-        res_columns[2]->insert(context.getPath() + database.second->getDataPath());
-        res_columns[3]->insert(database.second->getMetadataPath());
-        res_columns[4]->insert(database.second->getUUID());
+        if (database_name == DatabaseCatalog::TEMPORARY_DATABASE)
+            continue; /// We don't want to show the internal database for temporary tables in system.databases
+
+        res_columns[0]->insert(database_name);
+        res_columns[1]->insert(database->getEngineName());
+        res_columns[2]->insert(context.getPath() + database->getDataPath());
+        res_columns[3]->insert(database->getMetadataPath());
+        res_columns[4]->insert(database->getUUID());
    }
 }
 
diff --git a/src/Storages/System/StorageSystemTables.cpp b/src/Storages/System/StorageSystemTables.cpp
index c9b3223ee38..363a2a20828 100644
--- a/src/Storages/System/StorageSystemTables.cpp
+++ b/src/Storages/System/StorageSystemTables.cpp
@@ -65,8 +65,15 @@ StorageSystemTables::StorageSystemTables(const StorageID & table_id_)
 static ColumnPtr getFilteredDatabases(const ASTPtr & query, const Context & context)
 {
     MutableColumnPtr column = ColumnString::create();
-    for (const auto & db : DatabaseCatalog::instance().getDatabases())
-        column->insert(db.first);
+
+    const auto databases = DatabaseCatalog::instance().getDatabases();
+    for (const auto & database_name : databases | boost::adaptors::map_keys)
+    {
+        if (database_name == DatabaseCatalog::TEMPORARY_DATABASE)
+            continue; /// We don't want to show the internal database for temporary tables in system.tables
+
+        column->insert(database_name);
+    }
 
     Block block { ColumnWithTypeAndName(std::move(column), std::make_shared<DataTypeString>(), "database") };
     VirtualColumnUtils::filterBlockWithQuery(query, block, context);
@@ -192,7 +199,11 @@ protected:
 
                         // create_table_query
                         if (columns_mask[src_index++])
-                            res_columns[res_index++]->insertDefault();
+                        {
+                            auto temp_db = DatabaseCatalog::instance().getDatabaseForTemporaryTables();
+                            ASTPtr ast = temp_db ? temp_db->tryGetCreateTableQuery(table.second->getStorageID().getTableName(), context) : nullptr;
+                            res_columns[res_index++]->insert(ast ? queryToString(ast) : "");
+                        }
 
                         // engine_full
                         if (columns_mask[src_index++])
diff --git a/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp b/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp
index 4e7bff0ef41..0a3b3bd328f 100644
--- a/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp
+++ b/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp
@@ -4,6 +4,7 @@
 
 // I know that inclusion of .cpp is not good at all
 #include <Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp> // NOLINT
+#include <Storages/MergeTree/MergeTreeDataPartWriterWide.cpp> // NOLINT
 
 using namespace DB;
 static Block getBlockWithSize(size_t required_size_in_bytes, size_t size_of_row_in_bytes)
diff --git a/src/TableFunctions/TableFunctionFile.h b/src/TableFunctions/TableFunctionFile.h
index 91e2225a6d0..02704e4bf7f 100644
--- a/src/TableFunctions/TableFunctionFile.h
+++ b/src/TableFunctions/TableFunctionFile.h
@@ -6,8 +6,8 @@
 namespace DB
 {
 class Context;
+
 /* file(path, format, structure) - creates a temporary storage from file
- *
  *
  * The file must be in the clickhouse data directory.
  * The relative path begins with the clickhouse data directory.
diff --git a/src/TableFunctions/TableFunctionHDFS.cpp b/src/TableFunctions/TableFunctionHDFS.cpp
index e2f227ef7b5..700cb93ca06 100644
--- a/src/TableFunctions/TableFunctionHDFS.cpp
+++ b/src/TableFunctions/TableFunctionHDFS.cpp
@@ -2,7 +2,7 @@
 #include "registerTableFunctions.h"
 
 #if USE_HDFS
-#include <Storages/StorageHDFS.h>
+#include <Storages/HDFS/StorageHDFS.h>
 #include <Storages/ColumnsDescription.h>
 #include <TableFunctions/TableFunctionFactory.h>
 #include <TableFunctions/TableFunctionHDFS.h>
diff --git a/src/TableFunctions/TableFunctionMerge.cpp b/src/TableFunctions/TableFunctionMerge.cpp
index a878909e29d..c5fb9a7686d 100644
--- a/src/TableFunctions/TableFunctionMerge.cpp
+++ b/src/TableFunctions/TableFunctionMerge.cpp
@@ -6,6 +6,7 @@
 #include <TableFunctions/ITableFunction.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Interpreters/Context.h>
+#include <Access/ContextAccess.h>
 #include <TableFunctions/TableFunctionMerge.h>
 #include <TableFunctions/TableFunctionFactory.h>
 #include <TableFunctions/registerTableFunctions.h>
@@ -14,37 +15,24 @@
 namespace DB
 {
 
-
 namespace ErrorCodes
 {
     extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
     extern const int UNKNOWN_TABLE;
 }
 
-
-static NamesAndTypesList chooseColumns(const String & source_database, const String & table_name_regexp_, const Context & context)
+namespace
 {
-    OptimizedRegularExpression table_name_regexp(table_name_regexp_);
-    auto table_name_match = [&](const String & table_name) { return table_name_regexp.match(table_name); };
-
-    StoragePtr any_table;
-
+    [[noreturn]] void throwNoTablesMatchRegexp(const String & source_database, const String & source_table_regexp)
     {
-        auto database = DatabaseCatalog::instance().getDatabase(source_database);
-        auto iterator = database->getTablesIterator(context, table_name_match);
-
-        if (iterator->isValid())
-            if (const auto & table = iterator->table())
-                any_table = table;
+        throw Exception(
+            "Error while executing table function merge. In database " + source_database
+                + " no one matches regular expression: " + source_table_regexp,
+            ErrorCodes::UNKNOWN_TABLE);
     }
-
-    if (!any_table)
-        throw Exception("Error while executing table function merge. In database " + source_database + " no one matches regular expression: "
-            + table_name_regexp_, ErrorCodes::UNKNOWN_TABLE);
-
-    return any_table->getInMemoryMetadataPtr()->getColumns().getAllPhysical();
 }
 
+
 void TableFunctionMerge::parseArguments(const ASTPtr & ast_function, const Context & context)
 {
     ASTs & args_func = ast_function->children;
@@ -65,21 +53,64 @@ void TableFunctionMerge::parseArguments(const ASTPtr & ast_function, const Conte
     args[1] = evaluateConstantExpressionAsLiteral(args[1], context);
 
     source_database = args[0]->as<ASTLiteral &>().value.safeGet<String>();
-    table_name_regexp = args[1]->as<ASTLiteral &>().value.safeGet<String>();
+    source_table_regexp = args[1]->as<ASTLiteral &>().value.safeGet<String>();
 }
 
+
+const Strings & TableFunctionMerge::getSourceTables(const Context & context) const
+{
+    if (source_tables)
+        return *source_tables;
+
+    auto database = DatabaseCatalog::instance().getDatabase(source_database);
+
+    OptimizedRegularExpression re(source_table_regexp);
+    auto table_name_match = [&](const String & table_name_) { return re.match(table_name_); };
+
+    auto access = context.getAccess();
+    bool granted_show_on_all_tables = access->isGranted(AccessType::SHOW_TABLES, source_database);
+    bool granted_select_on_all_tables = access->isGranted(AccessType::SELECT, source_database);
+
+    source_tables.emplace();
+    for (auto it = database->getTablesIterator(context, table_name_match); it->isValid(); it->next())
+    {
+        if (!it->table())
+            continue;
+        bool granted_show = granted_show_on_all_tables || access->isGranted(AccessType::SHOW_TABLES, source_database, it->name());
+        if (!granted_show)
+            continue;
+        if (!granted_select_on_all_tables)
+            access->checkAccess(AccessType::SELECT, source_database, it->name());
+        source_tables->emplace_back(it->name());
+    }
+
+    if (source_tables->empty())
+        throwNoTablesMatchRegexp(source_database, source_table_regexp);
+
+    return *source_tables;
+}
+
+
 ColumnsDescription TableFunctionMerge::getActualTableStructure(const Context & context) const
 {
-    return ColumnsDescription{chooseColumns(source_database, table_name_regexp, context)};
+    for (const auto & table_name : getSourceTables(context))
+    {
+        auto storage = DatabaseCatalog::instance().tryGetTable(StorageID{source_database, table_name}, context);
+        if (storage)
+            return ColumnsDescription{storage->getInMemoryMetadataPtr()->getColumns().getAllPhysical()};
+    }
+
+    throwNoTablesMatchRegexp(source_database, source_table_regexp);
 }
 
+
 StoragePtr TableFunctionMerge::executeImpl(const ASTPtr & /*ast_function*/, const Context & context, const std::string & table_name, ColumnsDescription /*cached_columns*/) const
 {
     auto res = StorageMerge::create(
         StorageID(getDatabaseName(), table_name),
         getActualTableStructure(context),
         source_database,
-        table_name_regexp,
+        getSourceTables(context),
         context);
 
     res->startup();
diff --git a/src/TableFunctions/TableFunctionMerge.h b/src/TableFunctions/TableFunctionMerge.h
index a9c4dd6b038..8f9f4522d17 100644
--- a/src/TableFunctions/TableFunctionMerge.h
+++ b/src/TableFunctions/TableFunctionMerge.h
@@ -19,11 +19,13 @@ private:
     StoragePtr executeImpl(const ASTPtr & ast_function, const Context & context, const std::string & table_name, ColumnsDescription cached_columns) const override;
     const char * getStorageTypeName() const override { return "Merge"; }
 
+    const Strings & getSourceTables(const Context & context) const;
     ColumnsDescription getActualTableStructure(const Context & context) const override;
     void parseArguments(const ASTPtr & ast_function, const Context & context) override;
 
     String source_database;
-    String table_name_regexp;
+    String source_table_regexp;
+    mutable std::optional<Strings> source_tables;
 };
 
 
diff --git a/src/TableFunctions/TableFunctionS3.cpp b/src/TableFunctions/TableFunctionS3.cpp
index dfe1cf6e792..cc7877b204e 100644
--- a/src/TableFunctions/TableFunctionS3.cpp
+++ b/src/TableFunctions/TableFunctionS3.cpp
@@ -67,6 +67,7 @@ StoragePtr TableFunctionS3::executeImpl(const ASTPtr & /*ast_function*/, const C
     Poco::URI uri (filename);
     S3::URI s3_uri (uri);
     UInt64 min_upload_part_size = context.getSettingsRef().s3_min_upload_part_size;
+    UInt64 max_single_part_upload_size = context.getSettingsRef().s3_max_single_part_upload_size;
 
     StoragePtr storage = StorageS3::create(
             s3_uri,
@@ -75,6 +76,7 @@ StoragePtr TableFunctionS3::executeImpl(const ASTPtr & /*ast_function*/, const C
             StorageID(getDatabaseName(), table_name),
             format,
             min_upload_part_size,
+            max_single_part_upload_size,
             getActualTableStructure(context),
             ConstraintsDescription{},
             const_cast<Context &>(context),
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index 9e5a2e29dc9..3ef09e5658f 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -15,16 +15,6 @@ install (
     COMPONENT clickhouse
     PATTERN "CMakeLists.txt" EXCLUDE
     PATTERN ".gitignore" EXCLUDE
-    PATTERN "top_level_domains" EXCLUDE
-)
-
-# Dereference symlink
-get_filename_component(TOP_LEVEL_DOMAINS_ABS_DIR config/top_level_domains REALPATH)
-install (
-    DIRECTORY "${TOP_LEVEL_DOMAINS_ABS_DIR}"
-    DESTINATION ${CMAKE_INSTALL_DATAROOTDIR}/clickhouse-test/config
-    USE_SOURCE_PERMISSIONS
-    COMPONENT clickhouse
 )
 
 install (FILES server-test.xml DESTINATION  ${CLICKHOUSE_ETC_DIR}/clickhouse-server COMPONENT clickhouse)
diff --git a/tests/clickhouse-test b/tests/clickhouse-test
index 3702268819b..45cde00037f 100755
--- a/tests/clickhouse-test
+++ b/tests/clickhouse-test
@@ -15,6 +15,7 @@ from subprocess import check_call
 from subprocess import Popen
 from subprocess import PIPE
 from subprocess import CalledProcessError
+from subprocess import TimeoutExpired
 from datetime import datetime
 from time import time, sleep
 from errno import ESRCH
@@ -114,6 +115,7 @@ def get_db_engine(args):
 def run_single_test(args, ext, server_logs_level, client_options, case_file, stdout_file, stderr_file):
     # print(client_options)
 
+    start_time = datetime.now()
     if args.database:
         database = args.database
         os.environ.setdefault("CLICKHOUSE_DATABASE", database)
@@ -129,7 +131,11 @@ def run_single_test(args, ext, server_logs_level, client_options, case_file, std
         database = 'test_{suffix}'.format(suffix=random_str())
 
         clickhouse_proc_create = Popen(shlex.split(args.client), stdin=PIPE, stdout=PIPE, stderr=PIPE, universal_newlines=True)
-        clickhouse_proc_create.communicate(("CREATE DATABASE " + database + get_db_engine(args)))
+        try:
+            clickhouse_proc_create.communicate(("CREATE DATABASE " + database + get_db_engine(args)), timeout=args.timeout)
+        except TimeoutExpired:
+            total_time = (datetime.now() - start_time).total_seconds()
+            return clickhouse_proc_create, "", "Timeout creating database {} before test".format(database), total_time
 
         os.environ["CLICKHOUSE_DATABASE"] = database
 
@@ -152,14 +158,26 @@ def run_single_test(args, ext, server_logs_level, client_options, case_file, std
     # print(command)
 
     proc = Popen(command, shell=True, env=os.environ)
-    start_time = datetime.now()
 
     while (datetime.now() - start_time).total_seconds() < args.timeout and proc.poll() is None:
         sleep(0.01)
 
     if not args.database:
         clickhouse_proc_create = Popen(shlex.split(args.client), stdin=PIPE, stdout=PIPE, stderr=PIPE, universal_newlines=True)
-        clickhouse_proc_create.communicate(("DROP DATABASE " + database))
+        seconds_left = max(args.timeout - (datetime.now() - start_time).total_seconds(), 10)
+        try:
+            clickhouse_proc_create.communicate(("DROP DATABASE " + database), timeout=seconds_left)
+        except TimeoutExpired:
+            # kill test process because it can also hung
+            if proc.returncode is None:
+                try:
+                    proc.kill()
+                except OSError as e:
+                    if e.errno != ESRCH:
+                        raise
+
+            total_time = (datetime.now() - start_time).total_seconds()
+            return clickhouse_proc_create, "", "Timeout dropping database {} after test".format(database), total_time
 
     total_time = (datetime.now() - start_time).total_seconds()
 
@@ -305,7 +323,7 @@ def run_tests_array(all_tests_with_params):
 
                     if args.testname:
                         clickhouse_proc = Popen(shlex.split(args.client), stdin=PIPE, stdout=PIPE, stderr=PIPE, universal_newlines=True)
-                        clickhouse_proc.communicate(("SELECT 'Running test {suite}/{case} from pid={pid}';".format(pid = os.getpid(), case = case, suite = suite)))
+                        clickhouse_proc.communicate(("SELECT 'Running test {suite}/{case} from pid={pid}';".format(pid = os.getpid(), case = case, suite = suite)), timeout=10)
 
                         if clickhouse_proc.returncode != 0:
                             failures += 1
@@ -330,6 +348,8 @@ def run_tests_array(all_tests_with_params):
                         print(MSG_FAIL, end='')
                         print_test_time(total_time)
                         print(" - Timeout!")
+                        if stderr:
+                            print(stderr)
                     else:
                         counter = 1
                         while proc.returncode != 0 and need_retry(stderr):
@@ -562,6 +582,9 @@ def main(args):
     else:
         args.skip = tests_to_skip_from_list
 
+    if args.use_skip_list and not args.sequential:
+        args.sequential = collect_sequential_list(args.skip_list_path)
+
     base_dir = os.path.abspath(args.queries)
     tmp_dir = os.path.abspath(args.tmp)
 
@@ -675,10 +698,19 @@ def main(args):
                 all_tests = [t for t in all_tests if any([re.search(r, t) for r in args.test])]
             all_tests.sort(key=key_func)
 
+            parallel_tests = []
+            sequential_tests = []
+            for test in all_tests:
+                if any(s in test for s in args.sequential):
+                    sequential_tests.append(test)
+                else:
+                    parallel_tests.append(test)
+
+            print("Found", len(parallel_tests), "parallel tests and", len(sequential_tests), "sequential tests")
             run_n, run_total = args.parallel.split('/')
             run_n = float(run_n)
             run_total = float(run_total)
-            tests_n = len(all_tests)
+            tests_n = len(parallel_tests)
             if run_total > tests_n:
                 run_total = tests_n
             if run_n > run_total:
@@ -690,18 +722,20 @@ def main(args):
             if jobs > run_total:
                 run_total = jobs
 
-            batch_size = len(all_tests) // jobs
-            all_tests_array = []
-            for i in range(0, len(all_tests), batch_size):
-                all_tests_array.append((all_tests[i:i+batch_size], suite, suite_dir, suite_tmp_dir, run_total))
+            batch_size = len(parallel_tests) // jobs
+            parallel_tests_array = []
+            for i in range(0, len(parallel_tests), batch_size):
+                parallel_tests_array.append((parallel_tests[i:i+batch_size], suite, suite_dir, suite_tmp_dir, run_total))
 
             if jobs > 1:
                 with closing(multiprocessing.Pool(processes=jobs)) as pool:
-                    pool.map(run_tests_array, all_tests_array)
-            else:
-                run_tests_array(all_tests_array[int(run_n)-1])
+                    pool.map(run_tests_array, parallel_tests_array)
 
-            total_tests_run += tests_n
+                run_tests_array((sequential_tests, suite, suite_dir, suite_tmp_dir, run_total))
+                total_tests_run += len(sequential_tests) + len(parallel_tests)
+            else:
+                run_tests_array((all_tests, suite, suite_dir, suite_tmp_dir, run_total))
+                total_tests_run += len(all_tests)
 
     if args.hung_check:
 
@@ -792,6 +826,20 @@ def collect_tests_to_skip(skip_list_path, build_flags):
 
     return result
 
+
+def collect_sequential_list(skip_list_path):
+    if not os.path.exists(skip_list_path):
+        return set([])
+
+    with open(skip_list_path, 'r') as skip_list_file:
+        content = skip_list_file.read()
+        # allows to have comments in skip_list.json
+        skip_dict = json.loads(json_minify(content))
+        if 'parallel' in skip_dict:
+            return skip_dict['parallel']
+        return set([])
+
+
 if __name__ == '__main__':
     parser=ArgumentParser(description='ClickHouse functional tests')
     parser.add_argument('-q', '--queries', help='Path to queries dir')
@@ -828,6 +876,7 @@ if __name__ == '__main__':
     parser.add_argument('--no-stateless', action='store_true', help='Disable all stateless tests')
     parser.add_argument('--no-stateful', action='store_true', help='Disable all stateful tests')
     parser.add_argument('--skip', nargs='+', help="Skip these tests")
+    parser.add_argument('--sequential', nargs='+', help="Run these tests sequentially even if --parallel specified")
     parser.add_argument('--no-long', action='store_false', dest='no_long', help='Do not run long tests')
     parser.add_argument('--client-option', nargs='+', help='Specify additional client argument')
     parser.add_argument('--print-time', action='store_true', dest='print_time', help='Print test time')
@@ -871,6 +920,9 @@ if __name__ == '__main__':
     if args.skip_list_path is None:
         args.skip_list_path = os.path.join(args.queries, 'skip_list.json')
 
+    if args.sequential is None:
+        args.sequential = set([])
+
     if args.tmp is None:
         args.tmp = args.queries
     if args.client is None:
diff --git a/tests/config/config.d/max_concurrent_queries.xml b/tests/config/config.d/max_concurrent_queries.xml
new file mode 100644
index 00000000000..f326c293244
--- /dev/null
+++ b/tests/config/config.d/max_concurrent_queries.xml
@@ -0,0 +1,4 @@
+<!-- Ignore limit of concurrent queries in tests -->
+<yandex>
+    <max_concurrent_queries>0</max_concurrent_queries>
+</yandex>
diff --git a/tests/config/install.sh b/tests/config/install.sh
index 416cc21893b..9965e1fb1ad 100755
--- a/tests/config/install.sh
+++ b/tests/config/install.sh
@@ -27,6 +27,7 @@ ln -sf $SRC_PATH/config.d/secure_ports.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/clusters.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/graphite.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/database_atomic.xml $DEST_SERVER_PATH/config.d/
+ln -sf $SRC_PATH/config.d/max_concurrent_queries.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/test_cluster_with_incorrect_pw.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/test_keeper_port.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/logging_no_rotate.xml $DEST_SERVER_PATH/config.d/
diff --git a/tests/config/top_level_domains b/tests/config/top_level_domains
deleted file mode 120000
index 7e12ab4ba2c..00000000000
--- a/tests/config/top_level_domains
+++ /dev/null
@@ -1 +0,0 @@
-../../docker/test/performance-comparison/config/top_level_domains
\ No newline at end of file
diff --git a/docker/test/performance-comparison/config/top_level_domains/public_suffix_list.dat b/tests/config/top_level_domains/public_suffix_list.dat
similarity index 100%
rename from docker/test/performance-comparison/config/top_level_domains/public_suffix_list.dat
rename to tests/config/top_level_domains/public_suffix_list.dat
diff --git a/tests/instructions/easy_tasks_sorted_ru.md b/tests/instructions/easy_tasks_sorted_ru.md
index 2dd60f97db3..2c38dcc2c8f 100644
--- a/tests/instructions/easy_tasks_sorted_ru.md
+++ b/tests/instructions/easy_tasks_sorted_ru.md
@@ -16,7 +16,7 @@
 
 `context.setSetting` - для выставления `max_memory_usage` и других.
 
-## Битовые операции для FixedString.
+## + Битовые операции для FixedString.
 
 bitAnd, bitOr, bitNot, bitXor для значения типа FixedString, интерпретируемого как набор бит.
 
@@ -26,7 +26,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 ```
 Потом используйте их в вашей функции.
 
-## Добавить generic вариант функций least, greatest.
+## + Добавить generic вариант функций least, greatest.
 
 `SELECT least(123, 456)` - работает.
 
@@ -38,7 +38,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 Смотрим, что все файлы в прикрепляемых кусках от правильного пользователя.
 
-## COLLATE должно работать для Nullable(String).
+## + COLLATE должно работать для Nullable(String).
 
 В ClickHouse есть возможность указать collation для сортировки строк. Это не работает для `Nullable(String)`.
 
@@ -46,13 +46,13 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 Состояния агрегатных функций могут быть записаны в дамп и считаны из него. Но десериализация состояний агрегатных функций небезопасна. Аккуратно выбранные пользовательские данные могут привести к segfault или порче памяти. Поэтому нужно просто сделать настройку, которая запрещает читать AggregateFunction из пользовательских данных.
 
-## В статистику jemalloc добавить информацию по arenas.
+## + В статистику jemalloc добавить информацию по arenas.
 
 В `system.asynchronous_metrics` - суммарный размер арен.
 
 # Более сложные задачи
 
-## Layout внешних словарей "direct".
+## + Layout внешних словарей "direct".
 
 Как cache, но без кэша — всегда прямой запрос в источник.
 
@@ -60,7 +60,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 ## Агрегатные функции для статистических тестов (e.g. тест нормальности распределения) и статистик.
 
-## Функции создания и обновления состояния агрегатной функции по одному кортежу аргументов.
+## + Функции создания и обновления состояния агрегатной функции по одному кортежу аргументов.
 
 В ClickHouse есть понятие - состояние вычисления агрегатной функции. Состояния агрегатных функций можно записывать в таблицы, складывать, финализировать и т. п. https://clickhouse.yandex/docs/ru/data_types/nested_data_structures/aggregatefunction/
 
@@ -85,7 +85,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 `changeYear(datetime, 2019)`
 
-## Исправить мерцание прогресс-бара в clickhouse-client.
+## + Исправить мерцание прогресс-бара в clickhouse-client.
 
 Это заметно при работе с серверами с большим пингом.
 Прогресс бар не должен мерцать.
@@ -107,15 +107,15 @@ https://dev.mysql.com/doc/refman/8.0/en/date-and-time-functions.html#function_da
 
 position с конца строки.
 
-## Функция indexOf должна поддерживать Enum-ы без cast-а.
+## + Функция indexOf должна поддерживать Enum-ы без cast-а.
 
 `indexOf(arr, 'hello')`, `indexOf(arr, 1)` должны работать, если arr имеет тип `Array(Enum8('hello' = 1, 'world' = 2))`
 
-## Комбинатор агрегатных функций Distinct.
+## + Комбинатор агрегатных функций Distinct.
 
 Пример: `avgDistinct(x)` - вычислить среднее по всем различным переданным значениям.
 
-## Метрики количества ошибок.
+## + Метрики количества ошибок.
 
 Добавляем счётчики всех ошибок (ErrorCodes) по аналогии с ProfileEvents. Кроме количества запоминаем также время последней ошибки, стек трейс, сообщение. Добавляем системную таблицу system.errors. Отправка в Graphite.
 
@@ -125,9 +125,9 @@ position с конца строки.
 
 ## Запрос CREATE OR REPLACE TABLE
 
-Атомарно (под блокировкой) удаляет таблицу перед созданием новой, если такая была.
+Атомарно удаляет таблицу перед созданием новой, если такая была.
 
-## Приведение типов для IN (subquery).
+## * Приведение типов для IN (subquery).
 
 `SELECT 1 IN (SELECT -1 UNION ALL SELECT 1)`
 
@@ -155,7 +155,7 @@ https://clickhouse.yandex/docs/en/query_language/create/#create-table
 
 Подключить одну из key-value БД как источник.
 
-## Движок таблиц Mongo, табличная функция mongo.
+## + Движок таблиц Mongo, табличная функция mongo.
 
 Возможность легко импортировать данные из MongoDB.
 
@@ -201,7 +201,7 @@ https://clickhouse.yandex/docs/en/operations/table_engines/external_data/
 
 ## Настройка rollup_use_nulls.
 
-## Настройка cast_keep_nullable.
+## + Настройка cast_keep_nullable.
 
 ## Функция bitEquals для сравнения произвольных типов данных побитово.
 
diff --git a/tests/integration/README.md b/tests/integration/README.md
index 0886dc2cfac..cdfb6b1a70a 100644
--- a/tests/integration/README.md
+++ b/tests/integration/README.md
@@ -38,7 +38,8 @@ sudo -H pip install \
     pytest-timeout \
     redis \
     tzlocal \
-    urllib3
+    urllib3 \
+    requests-kerberos
 ```
 
 (highly not recommended) If you really want to use OS packages on modern debian/ubuntu instead of "pip": `sudo apt install -y docker docker-compose python3-pytest python3-dicttoxml python3-docker python3-pymysql python3-pymongo python3-tzlocal python3-kazoo python3-psycopg2 kafka-python python3-pytest-timeout python3-minio`
@@ -135,3 +136,13 @@ named `test.py` containing tests in it. All functions with names starting with `
 To assert that two TSV files must be equal, wrap them in the `TSV` class and use the regular `assert`
 statement. Example: `assert TSV(result) == TSV(reference)`. In case the assertion fails, `pytest`
 will automagically detect the types of variables and only the small diff of two files is printed.
+
+### Troubleshooting
+
+If tests failing for misterious reasons, this may help:
+
+```
+sudo service docker stop
+sudo bash -c 'rm -rf /var/lib/docker/*'
+sudo service docker start
+```
diff --git a/tests/integration/helpers/cluster.py b/tests/integration/helpers/cluster.py
index 3967d016591..a65a420cd5b 100644
--- a/tests/integration/helpers/cluster.py
+++ b/tests/integration/helpers/cluster.py
@@ -137,6 +137,7 @@ class ClickHouseCluster:
         self.with_rabbitmq = False
         self.with_odbc_drivers = False
         self.with_hdfs = False
+        self.with_kerberized_hdfs = False
         self.with_mongo = False
         self.with_net_trics = False
         self.with_redis = False
@@ -172,7 +173,7 @@ class ClickHouseCluster:
                      macros=None,
                      with_zookeeper=False, with_mysql=False, with_kafka=False, with_kerberized_kafka=False, with_rabbitmq=False,
                      clickhouse_path_dir=None,
-                     with_odbc_drivers=False, with_postgres=False, with_hdfs=False, with_mongo=False,
+                     with_odbc_drivers=False, with_postgres=False, with_hdfs=False, with_kerberized_hdfs=False, with_mongo=False,
                      with_redis=False, with_minio=False, with_cassandra=False,
                      hostname=None, env_variables=None, image="yandex/clickhouse-integration-test", tag=None,
                      stay_alive=False, ipv4_address=None, ipv6_address=None, with_installed_binary=False, tmpfs=None,
@@ -210,6 +211,7 @@ class ClickHouseCluster:
             with_kafka=with_kafka,
             with_kerberized_kafka=with_kerberized_kafka,
             with_rabbitmq=with_rabbitmq,
+            with_kerberized_hdfs=with_kerberized_hdfs,
             with_mongo=with_mongo,
             with_redis=with_redis,
             with_minio=with_minio,
@@ -315,6 +317,14 @@ class ClickHouseCluster:
                                   p.join(docker_compose_yml_dir, 'docker_compose_hdfs.yml')]
             cmds.append(self.base_hdfs_cmd)
 
+        if with_kerberized_hdfs and not self.with_kerberized_hdfs:
+            self.with_kerberized_hdfs = True
+            self.base_cmd.extend(['--file', p.join(docker_compose_yml_dir, 'docker_compose_kerberized_hdfs.yml')])
+            self.base_kerberized_hdfs_cmd = ['docker-compose', '--project-directory', self.base_dir, '--project-name',
+                                             self.project_name, '--file',
+                                             p.join(docker_compose_yml_dir, 'docker_compose_kerberized_hdfs.yml')]
+            cmds.append(self.base_kerberized_hdfs_cmd)
+
         if with_mongo and not self.with_mongo:
             self.with_mongo = True
             self.base_cmd.extend(['--file', p.join(docker_compose_yml_dir, 'docker_compose_mongo.yml')])
@@ -476,12 +486,35 @@ class ClickHouseCluster:
 
         raise Exception("Cannot wait ZooKeeper container")
 
+    def make_hdfs_api(self, timeout=60, kerberized=False):
+        if kerberized:
+            keytab = p.abspath(p.join(self.instances['node1'].path, "secrets/clickhouse.keytab"))
+            krb_conf = p.abspath(p.join(self.instances['node1'].path, "secrets/krb_long.conf"))
+            hdfs_ip = self.get_instance_ip('kerberizedhdfs1')
+            # print("kerberizedhdfs1 ip ", hdfs_ip)
+            kdc_ip = self.get_instance_ip('hdfskerberos')
+            # print("kdc_ip ", kdc_ip)
+            self.hdfs_api = HDFSApi(user="root",
+                               timeout=timeout,
+                               kerberized=True,
+                               principal="root@TEST.CLICKHOUSE.TECH",
+                               keytab=keytab,
+                               krb_conf=krb_conf,
+                               host="kerberizedhdfs1",
+                               protocol="http",
+                               proxy_port=50070,
+                               data_port=1006,
+                               hdfs_ip=hdfs_ip,
+                               kdc_ip=kdc_ip)
+        else:
+            self.hdfs_api = HDFSApi(user="root", host="hdfs1")
+
+
     def wait_hdfs_to_start(self, timeout=60):
-        hdfs_api = HDFSApi("root")
         start = time.time()
         while time.time() - start < timeout:
             try:
-                hdfs_api.write_data("/somefilewithrandomname222", "1")
+                self.hdfs_api.write_data("/somefilewithrandomname222", "1")
                 print("Connected to HDFS and SafeMode disabled! ")
                 return
             except Exception as ex:
@@ -620,6 +653,7 @@ class ClickHouseCluster:
                 self.wait_schema_registry_to_start(120)
 
             if self.with_kerberized_kafka and self.base_kerberized_kafka_cmd:
+                print('Setup kerberized kafka')
                 env = os.environ.copy()
                 env['KERBERIZED_KAFKA_DIR'] = instance.path + '/'
                 subprocess.check_call(self.base_kerberized_kafka_cmd + common_opts + ['--renew-anon-volumes'], env=env)
@@ -631,8 +665,17 @@ class ClickHouseCluster:
             if self.with_hdfs and self.base_hdfs_cmd:
                 print('Setup HDFS')
                 subprocess_check_call(self.base_hdfs_cmd + common_opts)
+                self.make_hdfs_api()
                 self.wait_hdfs_to_start(120)
 
+            if self.with_kerberized_hdfs and self.base_kerberized_hdfs_cmd:
+                print('Setup kerberized HDFS')
+                env = os.environ.copy()
+                env['KERBERIZED_HDFS_DIR'] = instance.path + '/'
+                subprocess.check_call(self.base_kerberized_hdfs_cmd + common_opts, env=env)
+                self.make_hdfs_api(kerberized=True)
+                self.wait_hdfs_to_start(timeout=300)
+
             if self.with_mongo and self.base_mongo_cmd:
                 print('Setup Mongo')
                 subprocess_check_call(self.base_mongo_cmd + common_opts)
@@ -840,8 +883,8 @@ class ClickHouseInstance:
     def __init__(
             self, cluster, base_path, name, base_config_dir, custom_main_configs, custom_user_configs,
             custom_dictionaries,
-            macros, with_zookeeper, zookeeper_config_path, with_mysql, with_kafka, with_kerberized_kafka, with_rabbitmq, with_mongo,
-            with_redis, with_minio,
+            macros, with_zookeeper, zookeeper_config_path, with_mysql, with_kafka, with_kerberized_kafka, with_rabbitmq, with_kerberized_hdfs,
+            with_mongo, with_redis, with_minio,
             with_cassandra, server_bin_path, odbc_bridge_bin_path, clickhouse_path_dir, with_odbc_drivers,
             hostname=None, env_variables=None,
             image="yandex/clickhouse-integration-test", tag="latest",
@@ -871,6 +914,7 @@ class ClickHouseInstance:
         self.with_kafka = with_kafka
         self.with_kerberized_kafka = with_kerberized_kafka
         self.with_rabbitmq = with_rabbitmq
+        self.with_kerberized_hdfs = with_kerberized_hdfs
         self.with_mongo = with_mongo
         self.with_redis = with_redis
         self.with_minio = with_minio
@@ -885,7 +929,7 @@ class ClickHouseInstance:
         else:
             self.odbc_ini_path = ""
 
-        if with_kerberized_kafka:
+        if with_kerberized_kafka or with_kerberized_hdfs:
             self.keytab_path = '- ' + os.path.dirname(self.docker_compose_path) + "/secrets:/tmp/keytab"
             self.krb5_conf = '- ' + os.path.dirname(self.docker_compose_path) + "/secrets/krb.conf:/etc/krb5.conf:ro"
         else:
@@ -1231,7 +1275,7 @@ class ClickHouseInstance:
         if self.with_zookeeper:
             shutil.copy(self.zookeeper_config_path, conf_d_dir)
 
-        if self.with_kerberized_kafka:
+        if self.with_kerberized_kafka or self.with_kerberized_hdfs:
             shutil.copytree(self.kerberos_secrets_dir, p.abspath(p.join(self.path, 'secrets')))
 
         # Copy config.d configs
@@ -1272,6 +1316,9 @@ class ClickHouseInstance:
         if self.with_kerberized_kafka:
             depends_on.append("kerberized_kafka1")
 
+        if self.with_kerberized_hdfs:
+            depends_on.append("kerberizedhdfs1")
+
         if self.with_rabbitmq:
             depends_on.append("rabbitmq1")
 
diff --git a/tests/integration/helpers/hdfs_api.py b/tests/integration/helpers/hdfs_api.py
index d3afa9be837..cb742662855 100644
--- a/tests/integration/helpers/hdfs_api.py
+++ b/tests/integration/helpers/hdfs_api.py
@@ -2,45 +2,145 @@
 import io
 import gzip
 import subprocess
+import time
 from tempfile import NamedTemporaryFile
-
 import requests
+import requests_kerberos as reqkerb
+import socket
+import tempfile
+import logging
+import os
 
+g_dns_hook = None
+
+def custom_getaddrinfo(*args):
+    # print("from custom_getaddrinfo g_dns_hook is None ", g_dns_hook is None)
+    ret = g_dns_hook.custom_getaddrinfo(*args)
+    # print("g_dns_hook.custom_getaddrinfo result", ret)
+    return ret
+
+
+class mk_krb_conf(object):
+    def __init__(self, krb_conf, kdc_ip):
+        self.krb_conf = krb_conf
+        self.kdc_ip = kdc_ip
+        self.amended_krb_conf = None
+    def __enter__(self):
+        with open(self.krb_conf) as f:
+            content = f.read()
+        amended_content = content.replace('hdfskerberos', self.kdc_ip)
+        self.amended_krb_conf = tempfile.NamedTemporaryFile(delete=False, mode="w+")
+        self.amended_krb_conf.write(amended_content)
+        self.amended_krb_conf.close()
+        return self.amended_krb_conf.name
+    def __exit__(self, type, value, traceback):
+        if self.amended_krb_conf is not None:
+            self.amended_krb_conf.close()
+
+# tweak dns resolution to connect to localhost where api_host is in URL
+class dns_hook(object):
+    def __init__(self, hdfs_api):
+        # print("dns_hook.init ", hdfs_api.kerberized, hdfs_api.host, hdfs_api.data_port, hdfs_api.proxy_port)
+        self.hdfs_api = hdfs_api
+    def __enter__(self):
+        global g_dns_hook
+        g_dns_hook = self
+        # print("g_dns_hook is None ", g_dns_hook is None)
+        self.original_getaddrinfo = socket.getaddrinfo
+        socket.getaddrinfo = custom_getaddrinfo
+        return self
+    def __exit__(self, type, value, traceback):
+        global g_dns_hook
+        g_dns_hook = None
+        socket.getaddrinfo = self.original_getaddrinfo
+    def custom_getaddrinfo(self, *args):
+        (hostname, port) = args[:2]
+        # print("top of custom_getaddrinfo", hostname, port)
+
+        if hostname == self.hdfs_api.host and (port == self.hdfs_api.data_port or port == self.hdfs_api.proxy_port):
+            # print("dns_hook substitute")
+            return [(socket.AF_INET, 1, 6, '', ("127.0.0.1", port))]
+        else:
+            return self.original_getaddrinfo(*args)
 
 class HDFSApi(object):
-    def __init__(self, user):
-        self.host = "localhost"
-        self.http_proxy_port = "50070"
-        self.http_data_port = "50075"
+    def __init__(self, user, timeout=100, kerberized=False, principal=None,
+                 keytab=None, krb_conf=None,
+                 host = "localhost", protocol = "http",
+                 proxy_port = 50070, data_port = 50075, hdfs_ip = None, kdc_ip = None):
+        self.host = host
+        self.protocol = protocol
+        self.proxy_port = proxy_port
+        self.data_port = data_port
         self.user = user
+        self.kerberized = kerberized
+        self.principal = principal
+        self.keytab = keytab
+        self.timeout = timeout
+        self.hdfs_ip = hdfs_ip
+        self.kdc_ip = kdc_ip
+        self.krb_conf = krb_conf
+
+        # logging.basicConfig(level=logging.DEBUG)
+        # logging.getLogger().setLevel(logging.DEBUG)
+        # requests_log = logging.getLogger("requests.packages.urllib3")
+        # requests_log.setLevel(logging.DEBUG)
+        # requests_log.propagate = True
+
+        if kerberized:
+            self._run_kinit()
+            self.kerberos_auth = reqkerb.HTTPKerberosAuth(mutual_authentication=reqkerb.DISABLED, hostname_override=self.host, principal=self.principal)
+                #principal=self.principal,
+                #hostname_override=self.host, principal=self.principal)
+                # , mutual_authentication=reqkerb.REQUIRED, force_preemptive=True)
+        else:
+            self.kerberos_auth = None
+
+    def _run_kinit(self):
+        if self.principal is None or self.keytab is None:
+            raise Exception("kerberos principal and keytab are required")
+
+        with mk_krb_conf(self.krb_conf, self.kdc_ip) as instantiated_krb_conf:
+            # print("instantiated_krb_conf ", instantiated_krb_conf)
+
+            os.environ["KRB5_CONFIG"] = instantiated_krb_conf
+
+            cmd = "(kinit -R -t {keytab} -k {principal} || (sleep 5 && kinit -R -t {keytab} -k {principal})) ; klist".format(instantiated_krb_conf=instantiated_krb_conf, keytab=self.keytab, principal=self.principal)
+
+            # print(cmd)
+
+            start = time.time()
+
+            while time.time() - start < self.timeout:
+                try:
+                    subprocess.call(cmd, shell=True)
+                    print("KDC started, kinit successfully run")
+                    return
+                except Exception as ex:
+                    print("Can't run kinit ... waiting {}".format(str(ex)))
+                    time.sleep(1)
+
+        raise Exception("Kinit running failure")
 
     def read_data(self, path, universal_newlines=True):
-        response = requests.get(
-            "http://{host}:{port}/webhdfs/v1{path}?op=OPEN".format(host=self.host, port=self.http_proxy_port,
-                                                                   path=path), allow_redirects=False)
+        with dns_hook(self):
+            response = requests.get("{protocol}://{host}:{port}/webhdfs/v1{path}?op=OPEN".format(protocol=self.protocol, host=self.host, port=self.proxy_port, path=path), headers={'host': 'localhost'}, allow_redirects=False, verify=False, auth=self.kerberos_auth)
         if response.status_code != 307:
             response.raise_for_status()
-        additional_params = '&'.join(response.headers['Location'].split('&')[1:2])
-        response_data = requests.get(
-            "http://{host}:{port}/webhdfs/v1{path}?op=OPEN&{params}".format(host=self.host, port=self.http_data_port,
-                                                                            path=path, params=additional_params))
+        # additional_params = '&'.join(response.headers['Location'].split('&')[1:2])
+        url = "{location}".format(location=response.headers['Location'])
+        # print("redirected to ", url)
+        with dns_hook(self):
+            response_data = requests.get(url,
+                                         headers={'host': 'localhost'},
+                                         verify=False, auth=self.kerberos_auth)
         if response_data.status_code != 200:
             response_data.raise_for_status()
-
         if universal_newlines:
             return response_data.text
         else:
             return response_data.content
 
-    # Requests can't put file
-    def _curl_to_put(self, filename, path, params):
-        url = "http://{host}:{port}/webhdfs/v1{path}?op=CREATE&{params}".format(host=self.host,
-                                                                                port=self.http_data_port, path=path,
-                                                                                params=params)
-        cmd = "curl -s -i -X PUT -T {fname} '{url}'".format(fname=filename, url=url)
-        output = subprocess.check_output(cmd, shell=True, universal_newlines=True)
-        return output
-
     def write_data(self, path, content):
         named_file = NamedTemporaryFile(mode='wb+')
         fpath = named_file.name
@@ -48,19 +148,44 @@ class HDFSApi(object):
             content = content.encode()
         named_file.write(content)
         named_file.flush()
-        response = requests.put(
-            "http://{host}:{port}/webhdfs/v1{path}?op=CREATE".format(host=self.host, port=self.http_proxy_port,
-                                                                     path=path, user=self.user),
-            allow_redirects=False
-        )
+
+
+        if self.kerberized:
+            self._run_kinit()
+            self.kerberos_auth = reqkerb.HTTPKerberosAuth(mutual_authentication=reqkerb.DISABLED, hostname_override=self.host, principal=self.principal)
+            # print(self.kerberos_auth)
+
+        with dns_hook(self):
+            response = requests.put(
+                "{protocol}://{host}:{port}/webhdfs/v1{path}?op=CREATE".format(protocol=self.protocol, host=self.host,
+                                                                               port=self.proxy_port,
+                                                                               path=path, user=self.user),
+                allow_redirects=False,
+                headers={'host': 'localhost'},
+                params={'overwrite' : 'true'},
+                verify=False, auth=self.kerberos_auth
+            )
         if response.status_code != 307:
+            # print(response.headers)
             response.raise_for_status()
 
         additional_params = '&'.join(
             response.headers['Location'].split('&')[1:2] + ["user.name={}".format(self.user), "overwrite=true"])
-        output = self._curl_to_put(fpath, path, additional_params)
-        if "201 Created" not in output:
-            raise Exception("Can't create file on hdfs:\n {}".format(output))
+
+        with dns_hook(self), open(fpath, mode="rb") as fh:
+            file_data = fh.read()
+            protocol = "http" # self.protocol
+            response = requests.put(
+                "{location}".format(location=response.headers['Location']),
+                data=file_data,
+                headers={'content-type':'text/plain', 'host': 'localhost'},
+                params={'file': path, 'user.name' : self.user},
+                allow_redirects=False, verify=False, auth=self.kerberos_auth
+            )
+            # print(response)
+            if response.status_code != 201:
+                response.raise_for_status()
+
 
     def write_gzip_data(self, path, content):
         if isinstance(content, str):
diff --git a/tests/integration/test_allowed_url_from_config/test.py b/tests/integration/test_allowed_url_from_config/test.py
index 44715d92121..6442937c8f4 100644
--- a/tests/integration/test_allowed_url_from_config/test.py
+++ b/tests/integration/test_allowed_url_from_config/test.py
@@ -1,6 +1,5 @@
 import pytest
 from helpers.cluster import ClickHouseCluster
-from helpers.hdfs_api import HDFSApi
 
 cluster = ClickHouseCluster(__file__)
 node1 = cluster.add_instance('node1', main_configs=['configs/config_with_hosts.xml'])
@@ -101,9 +100,8 @@ def test_table_function_remote(start_cluster):
 
 
 def test_redirect(start_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\t\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\t\n"
+    start_cluster.hdfs_api.write_data("/simple_storage", "1\t\n")
+    assert start_cluster.hdfs_api.read_data("/simple_storage") == "1\t\n"
     node7.query(
         "CREATE TABLE table_test_7_1 (word String) ENGINE=URL('http://hdfs1:50070/webhdfs/v1/simple_storage?op=OPEN&namenoderpcaddress=hdfs1:9000&offset=0', CSV)")
     assert "not allowed" in node7.query_and_get_error("SET max_http_get_redirects=1; SELECT * from table_test_7_1")
diff --git a/tests/integration/test_graphite_merge_tree/configs/users.xml b/tests/integration/test_graphite_merge_tree/configs/users.xml
new file mode 100644
index 00000000000..cdd437797ce
--- /dev/null
+++ b/tests/integration/test_graphite_merge_tree/configs/users.xml
@@ -0,0 +1,8 @@
+<?xml version="1.0"?>
+<yandex>
+    <profiles>
+        <default>
+            <optimize_on_insert>0</optimize_on_insert>
+        </default>
+    </profiles>
+</yandex>
diff --git a/tests/integration/test_graphite_merge_tree/test.py b/tests/integration/test_graphite_merge_tree/test.py
index 502004d2dfe..ee8a18f1ca5 100644
--- a/tests/integration/test_graphite_merge_tree/test.py
+++ b/tests/integration/test_graphite_merge_tree/test.py
@@ -8,7 +8,8 @@ from helpers.test_tools import TSV
 
 cluster = ClickHouseCluster(__file__)
 instance = cluster.add_instance('instance',
-                                main_configs=['configs/graphite_rollup.xml'])
+                                main_configs=['configs/graphite_rollup.xml'],
+                                user_configs=["configs/users.xml"])
 q = instance.query
 
 
diff --git a/tests/integration/test_join_set_family_s3/__init__.py b/tests/integration/test_join_set_family_s3/__init__.py
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml b/tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml
new file mode 100644
index 00000000000..318a6bca95d
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml
@@ -0,0 +1,12 @@
+<yandex>
+    <shutdown_wait_unfinished>3</shutdown_wait_unfinished>
+    <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/log.log</log>
+        <errorlog>/var/log/clickhouse-server/log.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+        <stderr>/var/log/clickhouse-server/stderr.log</stderr>
+        <stdout>/var/log/clickhouse-server/stdout.log</stdout>
+    </logger>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/config.xml b/tests/integration/test_join_set_family_s3/configs/config.xml
new file mode 100644
index 00000000000..5b9b5b5843a
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/config.xml
@@ -0,0 +1,28 @@
+<?xml version="1.0"?>
+<yandex>
+    <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/clickhouse-server.log</log>
+        <errorlog>/var/log/clickhouse-server/clickhouse-server.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+    </logger>
+
+    <tcp_port>9000</tcp_port>
+    <listen_host>127.0.0.1</listen_host>
+
+    <openSSL>
+        <client>
+            <cacheSessions>true</cacheSessions>
+            <verificationMode>none</verificationMode>
+            <invalidCertificateHandler>
+                <name>AcceptCertificateHandler</name>
+            </invalidCertificateHandler>
+        </client>
+    </openSSL>
+
+    <max_concurrent_queries>500</max_concurrent_queries>
+    <mark_cache_size>5368709120</mark_cache_size>
+    <path>./clickhouse/</path>
+    <users_config>users.xml</users_config>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/minio.xml b/tests/integration/test_join_set_family_s3/configs/minio.xml
new file mode 100644
index 00000000000..7337be1ad94
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/minio.xml
@@ -0,0 +1,14 @@
+<?xml version="1.0"?>
+<yandex>
+    <storage_configuration>
+        <disks>
+            <s3>
+                <type>s3</type>
+                <endpoint>http://minio1:9001/root/data/</endpoint>
+                <access_key_id>minio</access_key_id>
+                <secret_access_key>minio123</secret_access_key>
+                <send_object_metadata>true</send_object_metadata>
+            </s3>
+        </disks>
+    </storage_configuration>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/ssl.xml b/tests/integration/test_join_set_family_s3/configs/ssl.xml
new file mode 100644
index 00000000000..95cdc918bd0
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/ssl.xml
@@ -0,0 +1,12 @@
+<?xml version="1.0"?>
+<yandex>
+    <openSSL>
+        <client>
+            <cacheSessions>true</cacheSessions>
+            <verificationMode>none</verificationMode>
+            <invalidCertificateHandler>
+                <name>AcceptCertificateHandler</name>
+            </invalidCertificateHandler>
+        </client>
+    </openSSL>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/users.xml b/tests/integration/test_join_set_family_s3/configs/users.xml
new file mode 100644
index 00000000000..6061af8e33d
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/users.xml
@@ -0,0 +1,23 @@
+<?xml version="1.0"?>
+<yandex>
+    <profiles>
+        <default>
+        </default>
+    </profiles>
+
+    <users>
+        <default>
+            <password></password>
+            <networks incl="networks" replace="replace">
+                <ip>::/0</ip>
+            </networks>
+            <profile>default</profile>
+            <quota>default</quota>
+        </default>
+    </users>
+
+    <quotas>
+        <default>
+        </default>
+    </quotas>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/test.py b/tests/integration/test_join_set_family_s3/test.py
new file mode 100644
index 00000000000..2a20c8eade9
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/test.py
@@ -0,0 +1,97 @@
+import logging
+import sys
+
+import pytest
+from helpers.cluster import ClickHouseCluster
+
+logging.getLogger().setLevel(logging.INFO)
+logging.getLogger().addHandler(logging.StreamHandler())
+
+
+@pytest.fixture(scope="module")
+def cluster():
+    try:
+        cluster = ClickHouseCluster(__file__)
+        cluster.add_instance("node",
+                             main_configs=["configs/minio.xml", "configs/ssl.xml", "configs/config.d/log_conf.xml"],
+                             with_minio=True, stay_alive=True)
+        logging.info("Starting cluster...")
+        cluster.start()
+
+        logging.info("Cluster started")
+
+        yield cluster
+    finally:
+        cluster.shutdown()
+
+
+def assert_objects_count(cluster, objects_count, path='data/'):
+    minio = cluster.minio_client
+    s3_objects = list(minio.list_objects(cluster.minio_bucket, path))
+    if objects_count != len(s3_objects):
+        for s3_object in s3_objects:
+            object_meta = minio.stat_object(cluster.minio_bucket, s3_object.object_name)
+            logging.info("Existing S3 object: %s", str(object_meta))
+        assert objects_count == len(s3_objects)
+
+
+def test_set_s3(cluster):
+    node = cluster.instances["node"]
+
+    node.query("CREATE TABLE testLocalSet (n UInt64) Engine = Set")
+    node.query("CREATE TABLE testS3Set (n UInt64) Engine = Set SETTINGS disk='s3'")
+
+    node.query("INSERT INTO TABLE testLocalSet VALUES (1)")
+    node.query("INSERT INTO TABLE testS3Set VALUES (1)")
+
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n1\t1\n0\t0\n"
+    assert_objects_count(cluster, 1)
+
+    node.query("INSERT INTO TABLE testLocalSet VALUES (2)")
+    node.query("INSERT INTO TABLE testS3Set VALUES (2)")
+
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n1\t1\n1\t1\n"
+    assert_objects_count(cluster, 2)
+
+    node.restart_clickhouse()
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n1\t1\n1\t1\n"
+
+    node.query("TRUNCATE TABLE testLocalSet")
+    node.query("TRUNCATE TABLE testS3Set")
+
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n0\t0\n0\t0\n"
+    assert_objects_count(cluster, 0)
+
+    node.query("DROP TABLE testLocalSet")
+    node.query("DROP TABLE testS3Set")
+
+
+def test_join_s3(cluster):
+    node = cluster.instances["node"]
+
+    node.query("CREATE TABLE testLocalJoin(`id` UInt64, `val` String) ENGINE = Join(ANY, LEFT, id)")
+    node.query("CREATE TABLE testS3Join(`id` UInt64, `val` String) ENGINE = Join(ANY, LEFT, id) SETTINGS disk='s3'")
+
+    node.query("INSERT INTO testLocalJoin VALUES (1, 'a')")
+    node.query("INSERT INTO testS3Join VALUES (1, 'a')")
+
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\na\ta\n\t\n"
+    assert_objects_count(cluster, 1)
+
+    node.query("INSERT INTO testLocalJoin VALUES (2, 'b')")
+    node.query("INSERT INTO testS3Join VALUES (2, 'b')")
+
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\na\ta\nb\tb\n"
+    assert_objects_count(cluster, 2)
+
+    node.restart_clickhouse()
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\na\ta\nb\tb\n"
+
+    node.query("TRUNCATE TABLE testLocalJoin")
+    node.query("TRUNCATE TABLE testS3Join")
+
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\n\t\n\t\n"
+    assert_objects_count(cluster, 0)
+
+    node.query("DROP TABLE testLocalJoin")
+    node.query("DROP TABLE testS3Join")
diff --git a/tests/integration/test_materialize_mysql_database/configs/users.xml b/tests/integration/test_materialize_mysql_database/configs/users.xml
index f6df1c30fc4..196ea5cfb99 100644
--- a/tests/integration/test_materialize_mysql_database/configs/users.xml
+++ b/tests/integration/test_materialize_mysql_database/configs/users.xml
@@ -4,6 +4,8 @@
         <default>
             <allow_experimental_database_materialize_mysql>1</allow_experimental_database_materialize_mysql>
             <allow_introspection_functions>1</allow_introspection_functions>
+            <optimize_on_insert>0</optimize_on_insert>
+            <default_database_engine>Ordinary</default_database_engine>
         </default>
     </profiles>
 
diff --git a/tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml b/tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml
new file mode 100644
index 00000000000..3add72ec554
--- /dev/null
+++ b/tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml
@@ -0,0 +1,19 @@
+<?xml version="1.0"?>
+<yandex>
+    <profiles>
+        <default>
+            <allow_experimental_database_materialize_mysql>1</allow_experimental_database_materialize_mysql>
+            <default_database_engine>Atomic</default_database_engine>
+        </default>
+    </profiles>
+
+    <users>
+        <default>
+            <password></password>
+            <networks incl="networks" replace="replace">
+                <ip>::/0</ip>
+            </networks>
+            <profile>default</profile>
+        </default>
+    </users>
+</yandex>
diff --git a/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py b/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
index 56b6d8b3a15..47e325d0e96 100644
--- a/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
+++ b/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
@@ -15,6 +15,7 @@ from multiprocessing.dummy import Pool
 
 def check_query(clickhouse_node, query, result_set, retry_count=60, interval_seconds=3):
     lastest_result = ''
+
     for i in range(retry_count):
         try:
             lastest_result = clickhouse_node.query(query)
@@ -35,6 +36,7 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
     clickhouse_node.query("DROP DATABASE IF EXISTS test_database")
     mysql_node.query("CREATE DATABASE test_database DEFAULT CHARACTER SET 'utf8'")
     # existed before the mapping was created
+
     mysql_node.query("CREATE TABLE test_database.test_table_1 ("
                      "`key` INT NOT NULL PRIMARY KEY, "
                      "unsigned_tiny_int TINYINT UNSIGNED, tiny_int TINYINT, "
@@ -51,9 +53,10 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
                      "_date Date, _datetime DateTime, _timestamp TIMESTAMP, _bool BOOLEAN) ENGINE = InnoDB;")
 
     # it already has some data
-    mysql_node.query(
-        "INSERT INTO test_database.test_table_1 VALUES(1, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', "
-        "'2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', true);")
+    mysql_node.query("""
+        INSERT INTO test_database.test_table_1 VALUES(1, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 
+        '2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', true);
+        """)
 
     clickhouse_node.query(
         "CREATE DATABASE test_database ENGINE = MaterializeMySQL('{}:3306', 'test_database', 'root', 'clickhouse')".format(
@@ -65,9 +68,10 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
                 "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
                 "2020-01-01 00:00:00\t2020-01-01 00:00:00\t1\n")
 
-    mysql_node.query(
-        "INSERT INTO test_database.test_table_1 VALUES(2, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', "
-        "'2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', false);")
+    mysql_node.query("""
+        INSERT INTO test_database.test_table_1 VALUES(2, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 
+        '2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', false);
+        """)
 
     check_query(clickhouse_node, "SELECT * FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
                 "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
@@ -76,14 +80,16 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
 
     mysql_node.query("UPDATE test_database.test_table_1 SET unsigned_tiny_int = 2 WHERE `key` = 1")
 
-    check_query(clickhouse_node, "SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,"
-                                 " small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, "
-                                 " unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, "
-                                 " _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ "
-                                 " _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
-                "1\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
-                "2020-01-01 00:00:00\t1\n2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
-                "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t0\n")
+    check_query(clickhouse_node, """
+        SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,
+         small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, 
+         unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, 
+         _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ 
+         _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV
+        """,
+        "1\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
+        "2020-01-01 00:00:00\t1\n2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
+        "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t0\n")
 
     # update primary key
     mysql_node.query("UPDATE test_database.test_table_1 SET `key` = 3 WHERE `unsigned_tiny_int` = 2")
@@ -472,10 +478,16 @@ def select_without_columns(clickhouse_node, mysql_node, service_name):
     clickhouse_node.query("CREATE VIEW v AS SELECT * FROM db.t")
     mysql_node.query("INSERT INTO db.t VALUES (1, 1), (2, 2)")
     mysql_node.query("DELETE FROM db.t WHERE a=2;")
-    check_query(clickhouse_node, "SELECT count((_sign, _version)) FROM db.t FORMAT TSV", "3\n")
 
-    assert clickhouse_node.query("SELECT count(_sign) FROM db.t FORMAT TSV") == "2\n"
-    assert clickhouse_node.query("SELECT count(_version) FROM db.t FORMAT TSV") == "2\n"
+    optimize_on_insert = clickhouse_node.query("SELECT value FROM system.settings WHERE name='optimize_on_insert'").strip()
+    if optimize_on_insert == "0":
+        res = ["3\n", "2\n", "2\n"]
+    else:
+        res = ["2\n", "2\n", "1\n"]
+    check_query(clickhouse_node, "SELECT count((_sign, _version)) FROM db.t FORMAT TSV", res[0])
+
+    assert clickhouse_node.query("SELECT count(_sign) FROM db.t FORMAT TSV") == res[1]
+    assert clickhouse_node.query("SELECT count(_version) FROM db.t FORMAT TSV") == res[2]
 
     assert clickhouse_node.query("SELECT count() FROM db.t FORMAT TSV") == "1\n"
     assert clickhouse_node.query("SELECT count(*) FROM db.t FORMAT TSV") == "1\n"
@@ -556,6 +568,12 @@ def err_sync_user_privs_with_materialize_mysql_database(clickhouse_node, mysql_n
     assert 'MySQL SYNC USER ACCESS ERR:' in str(exception.value)
     assert "priv_err_db" not in clickhouse_node.query("SHOW DATABASES")
 
+    mysql_node.query("GRANT SELECT ON priv_err_db.* TO 'test'@'%'")
+    time.sleep(3)
+    clickhouse_node.query("ATTACH DATABASE priv_err_db")
+    clickhouse_node.query("DROP DATABASE priv_err_db")
+    mysql_node.query("REVOKE SELECT ON priv_err_db.* FROM 'test'@'%'")
+
     mysql_node.query("DROP DATABASE priv_err_db;")
     mysql_node.query("DROP USER 'test'@'%'")
 
diff --git a/tests/integration/test_materialize_mysql_database/test.py b/tests/integration/test_materialize_mysql_database/test.py
index ed37e3d7502..1b7aa041540 100644
--- a/tests/integration/test_materialize_mysql_database/test.py
+++ b/tests/integration/test_materialize_mysql_database/test.py
@@ -14,7 +14,10 @@ from . import materialize_with_ddl
 DOCKER_COMPOSE_PATH = get_docker_compose_path()
 
 cluster = ClickHouseCluster(__file__)
-clickhouse_node = cluster.add_instance('node1', user_configs=["configs/users.xml"], with_mysql=False, stay_alive=True)
+
+node_db_ordinary = cluster.add_instance('node1', user_configs=["configs/users.xml"], with_mysql=False, stay_alive=True)
+node_db_atomic = cluster.add_instance('node2', user_configs=["configs/users_db_atomic.xml"], with_mysql=False, stay_alive=True)
+
 
 @pytest.fixture(scope="module")
 def started_cluster():
@@ -119,39 +122,30 @@ def started_mysql_8_0():
                                '--remove-orphans'])
 
 
-def test_materialize_database_dml_with_mysql_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_dml_with_mysql_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.dml_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
     materialize_with_ddl.materialize_mysql_database_with_datetime_and_decimal(clickhouse_node, started_mysql_5_7, "mysql1")
 
-
-def test_materialize_database_dml_with_mysql_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_dml_with_mysql_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.dml_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
     materialize_with_ddl.materialize_mysql_database_with_datetime_and_decimal(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_mysql_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
+    materialize_with_ddl.drop_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.create_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_add_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_drop_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    # mysql 5.7 cannot support alter rename column
+    # materialize_with_ddl.alter_rename_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_modify_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
 
-
-def test_materialize_database_ddl_with_mysql_5_7(started_cluster, started_mysql_5_7):
-    try:
-        materialize_with_ddl.drop_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.create_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.alter_add_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                              "mysql1")
-        materialize_with_ddl.alter_drop_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                               "mysql1")
-        # mysql 5.7 cannot support alter rename column
-        # materialize_with_ddl.alter_rename_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.alter_rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                                "mysql1")
-        materialize_with_ddl.alter_modify_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                                 "mysql1")
-    except:
-        print((clickhouse_node.query(
-            "select '\n', thread_id, query_id, arrayStringConcat(arrayMap(x -> concat(demangle(addressToSymbol(x)), '\n    ', addressToLine(x)), trace), '\n') AS sym from system.stack_trace format TSVRaw")))
-        raise
-
-
-def test_materialize_database_ddl_with_mysql_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_mysql_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.drop_table_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
     materialize_with_ddl.create_table_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
     materialize_with_ddl.rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
@@ -166,61 +160,72 @@ def test_materialize_database_ddl_with_mysql_8_0(started_cluster, started_mysql_
     materialize_with_ddl.alter_modify_column_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0,
                                                                              "mysql8_0")
 
-
-def test_materialize_database_ddl_with_empty_transaction_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_empty_transaction_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.query_event_with_empty_transaction(clickhouse_node, started_mysql_5_7, "mysql1")
 
-
-def test_materialize_database_ddl_with_empty_transaction_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_empty_transaction_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.query_event_with_empty_transaction(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
 
-def test_select_without_columns_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_select_without_columns_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.select_without_columns(clickhouse_node, started_mysql_5_7, "mysql1")
 
 
-def test_select_without_columns_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_select_without_columns_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.select_without_columns(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
 
-def test_insert_with_modify_binlog_checksum_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_insert_with_modify_binlog_checksum_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.insert_with_modify_binlog_checksum(clickhouse_node, started_mysql_5_7, "mysql1")
 
 
-def test_insert_with_modify_binlog_checksum_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_insert_with_modify_binlog_checksum_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.insert_with_modify_binlog_checksum(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
 
-def test_materialize_database_err_sync_user_privs_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_err_sync_user_privs_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.err_sync_user_privs_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_materialize_database_err_sync_user_privs_8_0(started_cluster, started_mysql_8_0):
+
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_err_sync_user_privs_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.err_sync_user_privs_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_network_partition_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_network_partition_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.network_partition_test(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_network_partition_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_network_partition_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.network_partition_test(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_mysql_kill_sync_thread_restore_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_kill_sync_thread_restore_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.mysql_kill_sync_thread_restore_test(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_mysql_kill_sync_thread_restore_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_kill_sync_thread_restore_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.mysql_kill_sync_thread_restore_test(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_mysql_killed_while_insert_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_killed_while_insert_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.mysql_killed_while_insert(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_mysql_killed_while_insert_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_killed_while_insert_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.mysql_killed_while_insert(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_clickhouse_killed_while_insert_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_clickhouse_killed_while_insert_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.clickhouse_killed_while_insert(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_clickhouse_killed_while_insert_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_clickhouse_killed_while_insert_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.clickhouse_killed_while_insert(clickhouse_node, started_mysql_8_0, "mysql8_0")
diff --git a/tests/integration/test_postgresql_protocol/test.py b/tests/integration/test_postgresql_protocol/test.py
index 52c911cb939..3436fc5bace 100644
--- a/tests/integration/test_postgresql_protocol/test.py
+++ b/tests/integration/test_postgresql_protocol/test.py
@@ -120,7 +120,7 @@ def test_python_client(server_address):
         cur.execute('select name from tables;')
 
     assert exc_info.value.args == (
-        "Query execution failed.\nDB::Exception: Table default.tables doesn't exist.\nSSL connection has been closed unexpectedly\n",)
+        "Query execution failed.\nDB::Exception: Table default.tables doesn't exist\nSSL connection has been closed unexpectedly\n",)
 
     ch = py_psql.connect(host=server_address, port=server_port, user='default', password='123', database='')
     cur = ch.cursor()
diff --git a/tests/integration/test_quorum_inserts/configs/users.d/settings.xml b/tests/integration/test_quorum_inserts/configs/users.d/settings.xml
index 29b762656aa..62d0d1ec39d 100644
--- a/tests/integration/test_quorum_inserts/configs/users.d/settings.xml
+++ b/tests/integration/test_quorum_inserts/configs/users.d/settings.xml
@@ -5,6 +5,7 @@
             <max_memory_usage>10000000000</max_memory_usage>
             <use_uncompressed_cache>0</use_uncompressed_cache>
             <insert_quorum>2</insert_quorum>
+            <insert_quorum_parallel>0</insert_quorum_parallel>
             <select_sequential_consistency>1</select_sequential_consistency>
         </default>
     </profiles>
diff --git a/tests/integration/test_redirect_url_storage/test.py b/tests/integration/test_redirect_url_storage/test.py
index f93548af0db..f2731794d43 100644
--- a/tests/integration/test_redirect_url_storage/test.py
+++ b/tests/integration/test_redirect_url_storage/test.py
@@ -17,9 +17,8 @@ def started_cluster():
 
 
 def test_url_without_redirect(started_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    started_cluster.hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
 
     # access datanode port directly
     node1.query(
@@ -28,9 +27,8 @@ def test_url_without_redirect(started_cluster):
 
 
 def test_url_with_redirect_not_allowed(started_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    started_cluster.hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
 
     # access proxy port without allowing redirects
     node1.query(
@@ -40,9 +38,8 @@ def test_url_with_redirect_not_allowed(started_cluster):
 
 
 def test_url_with_redirect_allowed(started_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    started_cluster.hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
 
     # access proxy port with allowing redirects
     # http://localhost:50070/webhdfs/v1/b?op=OPEN&namenoderpcaddress=hdfs1:9000&offset=0
diff --git a/tests/integration/test_row_policy/configs/users.d/another_user.xml b/tests/integration/test_row_policy/configs/users.d/another_user.xml
index fb9608e5313..89e16e94c83 100644
--- a/tests/integration/test_row_policy/configs/users.d/another_user.xml
+++ b/tests/integration/test_row_policy/configs/users.d/another_user.xml
@@ -1,5 +1,10 @@
 <?xml version="1.0"?>
 <yandex>
+    <profiles>
+        <default>
+            <optimize_on_insert>0</optimize_on_insert>
+        </default>
+    </profiles>
     <users>
         <another>
             <password/>
@@ -10,4 +15,4 @@
             <quota>default</quota>
         </another>
     </users>
-</yandex>
\ No newline at end of file
+</yandex>
diff --git a/tests/integration/test_storage_hdfs/test.py b/tests/integration/test_storage_hdfs/test.py
index 59944a3725e..a6c8b7e1ee9 100644
--- a/tests/integration/test_storage_hdfs/test.py
+++ b/tests/integration/test_storage_hdfs/test.py
@@ -24,18 +24,14 @@ def started_cluster():
 
 
 def test_read_write_storage(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table SimpleHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/simple_storage', 'TSV')")
     node1.query("insert into SimpleHDFSStorage values (1, 'Mark', 72.53)")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
     assert node1.query("select * from SimpleHDFSStorage") == "1\tMark\t72.53\n"
 
 
 def test_read_write_storage_with_globs(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table HDFSStorageWithRange (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/storage{1..5}', 'TSV')")
     node1.query(
@@ -46,8 +42,8 @@ def test_read_write_storage_with_globs(started_cluster):
         "create table HDFSStorageWithAsterisk (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/storage*', 'TSV')")
 
     for i in ["1", "2", "3"]:
-        hdfs_api.write_data("/storage" + i, i + "\tMark\t72.53\n")
-        assert hdfs_api.read_data("/storage" + i) == i + "\tMark\t72.53\n"
+        started_cluster.hdfs_api.write_data("/storage" + i, i + "\tMark\t72.53\n")
+        assert started_cluster.hdfs_api.read_data("/storage" + i) == i + "\tMark\t72.53\n"
 
     assert node1.query("select count(*) from HDFSStorageWithRange") == "3\n"
     assert node1.query("select count(*) from HDFSStorageWithEnum") == "3\n"
@@ -77,25 +73,23 @@ def test_read_write_storage_with_globs(started_cluster):
 
 
 def test_read_write_table(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
-    hdfs_api.write_data("/simple_table_function", data)
+    started_cluster.hdfs_api.write_data("/simple_table_function", data)
 
-    assert hdfs_api.read_data("/simple_table_function") == data
+    assert started_cluster.hdfs_api.read_data("/simple_table_function") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64')") == data
 
 
 def test_write_table(started_cluster):
-    hdfs_api = HDFSApi("root")
 
     node1.query(
         "create table OtherHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/other_storage', 'TSV')")
     node1.query("insert into OtherHDFSStorage values (10, 'tomas', 55.55), (11, 'jack', 32.54)")
 
     result = "10\ttomas\t55.55\n11\tjack\t32.54\n"
-    assert hdfs_api.read_data("/other_storage") == result
+    assert started_cluster.hdfs_api.read_data("/other_storage") == result
     assert node1.query("select * from OtherHDFSStorage order by id") == result
 
 
@@ -120,15 +114,14 @@ def test_bad_hdfs_uri(started_cluster):
         print(ex)
         assert "Unable to open HDFS file" in str(ex)
 
-
+@pytest.mark.timeout(800)
 def test_globs_in_read_table(started_cluster):
-    hdfs_api = HDFSApi("root")
     some_data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
     globs_dir = "/dir_for_test_with_globs/"
     files = ["dir1/dir_dir/file1", "dir2/file2", "simple_table_function", "dir/file", "some_dir/dir1/file",
              "some_dir/dir2/file", "some_dir/file", "table1_function", "table2_function", "table3_function"]
     for filename in files:
-        hdfs_api.write_data(globs_dir + filename, some_data)
+        started_cluster.hdfs_api.write_data(globs_dir + filename, some_data)
 
     test_requests = [("dir{1..5}/dir_dir/file1", 1, 1),
                      ("*_table_functio?", 1, 1),
@@ -145,6 +138,7 @@ def test_globs_in_read_table(started_cluster):
 
     for pattern, paths_amount, files_amount in test_requests:
         inside_table_func = "'hdfs://hdfs1:9000" + globs_dir + pattern + "', 'TSV', 'id UInt64, text String, number Float64'"
+        print("inside_table_func ", inside_table_func)
         assert node1.query("select * from hdfs(" + inside_table_func + ")") == paths_amount * some_data
         assert node1.query("select count(distinct _path) from hdfs(" + inside_table_func + ")").rstrip() == str(
             paths_amount)
@@ -153,64 +147,61 @@ def test_globs_in_read_table(started_cluster):
 
 
 def test_read_write_gzip_table(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_gzip_data("/simple_table_function.gz", data)
+    started_cluster.hdfs_api.write_gzip_data("/simple_table_function.gz", data)
 
-    assert hdfs_api.read_gzip_data("/simple_table_function.gz") == data
+    assert started_cluster.hdfs_api.read_gzip_data("/simple_table_function.gz") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function.gz', 'TSV', 'id UInt64, text String, number Float64')") == data
 
 
 def test_read_write_gzip_table_with_parameter_gzip(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_gzip_data("/simple_table_function", data)
+    started_cluster.hdfs_api.write_gzip_data("/simple_table_function", data)
 
-    assert hdfs_api.read_gzip_data("/simple_table_function") == data
+    assert started_cluster.hdfs_api.read_gzip_data("/simple_table_function") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64', 'gzip')") == data
 
 
 def test_read_write_table_with_parameter_none(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_data("/simple_table_function.gz", data)
+    started_cluster.hdfs_api.write_data("/simple_table_function.gz", data)
 
-    assert hdfs_api.read_data("/simple_table_function.gz") == data
+    assert started_cluster.hdfs_api.read_data("/simple_table_function.gz") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function.gz', 'TSV', 'id UInt64, text String, number Float64', 'none')") == data
 
 
 def test_read_write_gzip_table_with_parameter_auto_gz(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_gzip_data("/simple_table_function.gz", data)
+    started_cluster.hdfs_api.write_gzip_data("/simple_table_function.gz", data)
 
-    assert hdfs_api.read_gzip_data("/simple_table_function.gz") == data
+    assert started_cluster.hdfs_api.read_gzip_data("/simple_table_function.gz") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function.gz', 'TSV', 'id UInt64, text String, number Float64', 'auto')") == data
 
 
 def test_write_gz_storage(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table GZHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/storage.gz', 'TSV')")
     node1.query("insert into GZHDFSStorage values (1, 'Mark', 72.53)")
-    assert hdfs_api.read_gzip_data("/storage.gz") == "1\tMark\t72.53\n"
+    assert started_cluster.hdfs_api.read_gzip_data("/storage.gz") == "1\tMark\t72.53\n"
     assert node1.query("select * from GZHDFSStorage") == "1\tMark\t72.53\n"
 
 
 def test_write_gzip_storage(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table GZIPHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/gzip_storage', 'TSV', 'gzip')")
     node1.query("insert into GZIPHDFSStorage values (1, 'Mark', 72.53)")
-    assert hdfs_api.read_gzip_data("/gzip_storage") == "1\tMark\t72.53\n"
+    assert started_cluster.hdfs_api.read_gzip_data("/gzip_storage") == "1\tMark\t72.53\n"
     assert node1.query("select * from GZIPHDFSStorage") == "1\tMark\t72.53\n"
+
+if __name__ == '__main__':
+    cluster.start()
+    input("Cluster created, press any key to destroy...")
+    cluster.shutdown()
diff --git a/tests/integration/test_storage_kafka/test.py b/tests/integration/test_storage_kafka/test.py
index 06290053efa..07d2bcb60c0 100644
--- a/tests/integration/test_storage_kafka/test.py
+++ b/tests/integration/test_storage_kafka/test.py
@@ -9,6 +9,7 @@ import time
 import avro.schema
 from confluent_kafka.avro.cached_schema_registry_client import CachedSchemaRegistryClient
 from confluent_kafka.avro.serializer.message_serializer import MessageSerializer
+from confluent_kafka import admin
 
 import kafka.errors
 import pytest
@@ -1161,6 +1162,66 @@ def test_kafka_materialized_view(kafka_cluster):
 
     kafka_check_result(result, True)
 
+@pytest.mark.timeout(180)
+def test_librdkafka_snappy_regression(kafka_cluster):
+    """
+    Regression for UB in snappy-c (that is used in librdkafka),
+    backport pr is [1].
+
+      [1]: https://github.com/ClickHouse-Extras/librdkafka/pull/3
+
+    Example of corruption:
+
+        2020.12.10 09:59:56.831507 [ 20 ] {} <Error> void DB::StorageKafka::threadFunc(size_t): Code: 27, e.displayText() = DB::Exception: Cannot parse input: expected '"' before: 'foo"}': (while reading the value of key value): (at row 1)
+, Stack trace (when copying this message, always include the lines below):
+    """
+
+    # create topic with snappy compression
+    admin_client = admin.AdminClient({'bootstrap.servers': 'localhost:9092'})
+    topic_snappy = admin.NewTopic(topic='snappy_regression', num_partitions=1, replication_factor=1, config={
+        'compression.type': 'snappy',
+    })
+    admin_client.create_topics(new_topics=[topic_snappy], validate_only=False)
+
+    instance.query('''
+        CREATE TABLE test.kafka (key UInt64, value String)
+            ENGINE = Kafka
+            SETTINGS kafka_broker_list = 'kafka1:19092',
+                     kafka_topic_list = 'snappy_regression',
+                     kafka_group_name = 'ch_snappy_regression',
+                     kafka_format = 'JSONEachRow';
+    ''')
+
+    messages = []
+    expected = []
+    # To trigger this regression there should duplicated messages
+    # Orignal reproducer is:
+    #
+    #     $ gcc --version |& fgrep gcc
+    #     gcc (GCC) 10.2.0
+    #     $ yes foobarbaz | fold -w 80 | head -n10 >| in-…
+    #     $ make clean && make CFLAGS='-Wall -g -O2 -ftree-loop-vectorize -DNDEBUG=1 -DSG=1 -fPIC'
+    #     $ ./verify in
+    #     final comparision of in failed at 20 of 100
+    value = 'foobarbaz'*10
+    number_of_messages = 50
+    for i in range(number_of_messages):
+        messages.append(json.dumps({'key': i, 'value': value}))
+        expected.append(f'{i}\t{value}')
+    kafka_produce('snappy_regression', messages)
+
+    expected = '\n'.join(expected)
+
+    while True:
+        result = instance.query('SELECT * FROM test.kafka')
+        rows = len(result.strip('\n').split('\n'))
+        print(rows)
+        if rows == number_of_messages:
+            break
+
+    assert TSV(result) == TSV(expected)
+
+    instance.query('DROP TABLE test.kafka')
 
 @pytest.mark.timeout(180)
 def test_kafka_materialized_view_with_subquery(kafka_cluster):
diff --git a/tests/integration/test_storage_kerberized_hdfs/__init__.py b/tests/integration/test_storage_kerberized_hdfs/__init__.py
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml b/tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml
new file mode 100644
index 00000000000..3a93fa603c7
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml
@@ -0,0 +1,13 @@
+<yandex>
+  <hdfs>
+		<hadoop_kerberos_keytab>/tmp/keytab/clickhouse.keytab</hadoop_kerberos_keytab>
+		<hadoop_kerberos_principal>root@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+		<hadoop_security_authentication>kerberos</hadoop_security_authentication>
+  </hdfs>
+  <hdfs_suser>
+		<hadoop_kerberos_principal>specuser@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+  </hdfs_suser>
+  <hdfs_dedicatedcachepath>
+		<hadoop_security_kerberos_ticket_cache_path>/tmp/kerb_cache</hadoop_security_kerberos_ticket_cache_path>
+  </hdfs_dedicatedcachepath>
+</yandex>
diff --git a/tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml b/tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml
new file mode 100644
index 00000000000..0de2745ca4c
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml
@@ -0,0 +1,11 @@
+<yandex>
+  <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/log.log</log>
+        <errorlog>/var/log/clickhouse-server/log.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+        <stderr>/var/log/clickhouse-server/stderr.log</stderr>
+        <stdout>/var/log/clickhouse-server/stdout.log</stdout>
+    </logger>
+</yandex>
diff --git a/tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh b/tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh
new file mode 100755
index 00000000000..38f098ae1e1
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh
@@ -0,0 +1,262 @@
+#!/bin/bash
+
+: "${HADOOP_PREFIX:=/usr/local/hadoop}"
+
+cat >> $HADOOP_PREFIX/etc/hadoop/hadoop-env.sh <<EOF
+export HADOOP_SECURE_DN_USER=hdfs
+export HADOOP_SECURE_DN_PID_DIR=$HADOOP_PREFIX/pid
+export HADOOP_SECURE_DN_LOG_DIR=$HADOOP_PREFIX/logs/hdfs
+export JSVC_HOME=$HADOOP_PREFIX/sbin
+EOF
+$HADOOP_PREFIX/etc/hadoop/hadoop-env.sh
+
+mkdir -p "${HADOOP_SECURE_DN_PID_DIR}"
+mkdir -p "${HADOOP_SECURE_DN_LOG_DIR}"
+
+rm /tmp/*.pid
+
+# installing libraries if any - (resource urls added comma separated to the ACP system variable)
+cd "${HADOOP_PREFIX}/share/hadoop/common" || exit
+for cp in ${ACP//,/ }; do  echo "== ${cp}"; curl -LO "${cp}" ; done;
+cd - || exit
+
+# altering the core-site configuration
+sed "s/HOSTNAME/${HOSTNAME}/" /usr/local/hadoop/etc/hadoop/core-site.xml.template | grep -v '/configuration' > /usr/local/hadoop/etc/hadoop/core-site.xml
+
+cat >> /usr/local/hadoop/etc/hadoop/core-site.xml << EOF
+  <property>
+    <name>hadoop.security.authentication</name>
+    <value>kerberos</value> <!-- A value of "simple" would disable security. -->
+  </property>
+   <property>
+     <name>hadoop.security.authorization</name>
+     <value>true</value>
+   </property>
+  <property>
+    <name>fs.defaultFS</name>
+    <value>hdfs://kerberizedhdfs1:9000</value>
+  </property>
+  <property>
+    <name>fs.default.name</name>
+    <value>hdfs://kerberizedhdfs1:9000</value>
+  </property>
+  <!--
+  <property>
+      <name>hadoop.rpc.protection</name>
+      <value>privacy</value>
+  </property>
+  -->
+</configuration>
+EOF
+
+
+cat > /usr/local/hadoop/etc/hadoop/hdfs-site.xml << EOF
+<configuration>
+<!--
+<property>
+	<name>dfs.permissions.enabled</name>
+	<value>false</value>
+</property>
+-->
+<property>
+  <name>dfs.replication</name>
+  <value>1</value>
+</property>
+<!-- General HDFS security config -->
+<property>
+  <name>dfs.block.access.token.enable</name>
+  <value>true</value>
+</property>
+<!-- NameNode security config -->
+<property>
+  <name>dfs.namenode.keytab.file</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+<property>
+  <name>dfs.namenode.kerberos.principal</name>
+  <value>hdfs/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<property>
+  <name>dfs.namenode.kerberos.internal.spnego.principal</name>
+  <value>HTTP/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<!-- Secondary NameNode security config -->
+<property>
+  <name>dfs.secondary.namenode.keytab.file</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+<property>
+  <name>dfs.secondary.namenode.kerberos.principal</name>
+  <value>hdfs/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<property>
+  <name>dfs.secondary.namenode.kerberos.internal.spnego.principal</name>
+  <value>HTTP/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<!-- DataNode security config
+<property>
+   <name>dfs.data.transfer.protectionл</name>
+   <value>integrity</value>
+</property>
+-->
+<property>
+  <name>dfs.datanode.data.dir.perm</name>
+  <value>700</value>
+</property>
+<property>
+  <name>dfs.datanode.address</name>
+  <value>0.0.0.0:1004</value>
+</property>
+<property>
+  <name>dfs.datanode.http.address</name>
+  <value>0.0.0.0:1006</value>
+</property>
+<!--
+<property>
+  <name>dfs.http.policy</name>
+  <value>HTTPS_ONLY</value>
+</property>
+-->
+
+
+<property>
+  <name>dfs.datanode.keytab.file</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+<property>
+  <name>dfs.datanode.kerberos.principal</name>
+  <value>hdfs/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+
+<!-- Web Authentication config -->
+<property>
+    <name>dfs.webhdfs.enabled</name>
+    <value>true</value>
+</property>
+<property>
+<name>dfs.encrypt.data.transfer</name>
+<value>false</value>
+</property>
+<property>
+  <name>dfs.web.authentication.kerberos.principal</name>
+  <value>HTTP/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<property>
+    <name>dfs.web.authentication.kerberos.keytab</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+</configuration>
+EOF
+
+
+
+# cat > /usr/local/hadoop/etc/hadoop/ssl-server.xml << EOF
+# <configuration>
+# <property>
+#   <name>ssl.server.truststore.location</name>
+#   <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.jks</value>
+# </property>
+# <property>
+#   <name>ssl.server.truststore.password</name>
+#   <value>masterkey</value>
+# </property>
+# <property>
+#   <name>ssl.server.keystore.location</name>
+#   <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.jks</value>
+# </property>
+# <property>
+#   <name>ssl.server.keystore.password</name>
+#   <value>masterkey</value>
+# </property>
+# <property>
+#   <name>ssl.server.keystore.keypassword</name>
+#   <value>masterkey</value>
+# </property>
+# </configuration>
+# EOF
+
+cat > /usr/local/hadoop/etc/hadoop/log4j.properties << EOF
+# Set everything to be logged to the console
+log4j.rootCategory=DEBUG, console
+log4j.appender.console=org.apache.log4j.ConsoleAppender
+log4j.appender.console.target=System.err
+log4j.appender.console.layout=org.apache.log4j.PatternLayout
+log4j.appender.console.layout.ConversionPattern=%d{yy/MM/dd HH:mm:ss} %p %c{1}: %m%n
+
+# Set the default spark-shell log level to WARN. When running the spark-shell, the
+# log level for this class is used to overwrite the root logger's log level, so that
+# the user can have different defaults for the shell and regular Spark apps.
+log4j.logger.org.apache.spark.repl.Main=INFO
+
+# Settings to quiet third party logs that are too verbose
+log4j.logger.org.spark_project.jetty=DEBUG
+log4j.logger.org.spark_project.jetty.util.component.AbstractLifeCycle=ERROR
+log4j.logger.org.apache.spark.repl.SparkIMain$exprTyper=INFO
+log4j.logger.org.apache.spark.repl.SparkILoop$SparkILoopInterpreter=INFO
+log4j.logger.org.apache.parquet=ERROR
+log4j.logger.parquet=ERROR
+
+# SPARK-9183: Settings to avoid annoying messages when looking up nonexistent UDFs in SparkSQL with Hive support
+log4j.logger.org.apache.hadoop.hive.metastore.RetryingHMSHandler=FATAL
+log4j.logger.org.apache.hadoop.hive.ql.exec.FunctionRegistry=ERROR
+
+log4j.logger.org.apache.spark.deploy=DEBUG
+log4j.logger.org.apache.spark.executor=DEBUG
+log4j.logger.org.apache.spark.scheduler=DEBUG
+EOF
+
+useradd -u 1098 hdfs
+
+# keytool -genkey -alias kerberized_hdfs1.test.clickhouse.tech -keyalg rsa -keysize 1024 -dname "CN=kerberized_hdfs1.test.clickhouse.tech" -keypass masterkey -keystore /usr/local/hadoop/etc/hadoop/conf/hdfs.jks -storepass masterkey
+keytool -genkey -alias kerberizedhdfs1 -keyalg rsa -keysize 1024 -dname "CN=kerberizedhdfs1" -keypass masterkey -keystore /usr/local/hadoop/etc/hadoop/conf/hdfs.jks -storepass masterkey
+
+chmod g+r /usr/local/hadoop/etc/hadoop/conf/hdfs.jks
+
+
+service sshd start
+
+# yum --quiet --assumeyes install krb5-workstation.x86_64
+# yum --quiet --assumeyes install tcpdump
+
+# cd /tmp
+# curl http://archive.apache.org/dist/commons/daemon/source/commons-daemon-1.0.15-src.tar.gz   -o  commons-daemon-1.0.15-src.tar.gz
+# tar xzf commons-daemon-1.0.15-src.tar.gz
+# cd commons-daemon-1.0.15-src/src/native/unix
+# ./configure && make
+# cp ./jsvc /usr/local/hadoop/sbin
+
+
+until kinit -kt /usr/local/hadoop/etc/hadoop/conf/hdfs.keytab hdfs/kerberizedhdfs1@TEST.CLICKHOUSE.TECH; do sleep 2; done
+echo "KDC is up and ready to go... starting up"
+
+$HADOOP_PREFIX/sbin/start-dfs.sh
+$HADOOP_PREFIX/sbin/start-yarn.sh
+$HADOOP_PREFIX/sbin/mr-jobhistory-daemon.sh start historyserver
+
+chmod a+r /usr/local/hadoop/etc/hadoop/conf/hdfs.keytab   # create dedicated keytab for hdfsuser
+
+$HADOOP_PREFIX/sbin/start-secure-dns.sh
+sleep 3
+
+/usr/local/hadoop/bin/hdfs dfsadmin -safemode leave
+
+/usr/local/hadoop/bin/hdfs dfs -mkdir /user/specuser
+/usr/local/hadoop/bin/hdfs dfs -chown specuser /user/specuser
+
+kdestroy
+
+
+
+# adduser --groups hdfs hdfsuser
+
+# /usr/local/hadoop/sbin/hadoop-daemon.sh --config /usr/local/hadoop/etc/hadoop/ --script /usr/local/hadoop/sbin/hdfs start namenode
+# /usr/local/hadoop/sbin/hadoop-daemon.sh --config /usr/local/hadoop/etc/hadoop/ --script /usr/local/hadoop/sbin/hdfs start datanode
+
+
+if [[ $1 == "-d" ]]; then
+  while true; do sleep 1000; done
+fi
+
+if [[ $1 == "-bash" ]]; then
+  /bin/bash
+fi
diff --git a/tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh b/tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh
new file mode 100644
index 00000000000..daa060cbe29
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh
@@ -0,0 +1,139 @@
+#!/bin/bash
+
+
+set -x  #  trace
+
+: "${REALM:=TEST.CLICKHOUSE.TECH}"
+: "${DOMAIN_REALM:=test.clickhouse.tech}"
+: "${KERB_MASTER_KEY:=masterkey}"
+: "${KERB_ADMIN_USER:=admin}"
+: "${KERB_ADMIN_PASS:=admin}"
+
+create_config() {
+  : "${KDC_ADDRESS:=$(hostname -f)}"
+
+  cat>/etc/krb5.conf<<EOF
+[logging]
+ default = FILE:/var/log/kerberos/krb5libs.log
+ kdc = FILE:/var/log/kerberos/krb5kdc.log
+ admin_server = FILE:/var/log/kerberos/kadmind.log
+
+[libdefaults]
+ default_realm = $REALM
+ dns_lookup_realm = false
+ dns_lookup_kdc = false
+ ticket_lifetime = 15d
+ renew_lifetime = 15d
+ forwardable = true
+ # WARNING: We use weaker key types to simplify testing as stronger key types
+ # require the enhanced security JCE policy file to be installed. You should
+ # NOT run with this configuration in production or any real environment. You
+ # have been warned.
+ default_tkt_enctypes = des-cbc-md5 des-cbc-crc des3-cbc-sha1
+ default_tgs_enctypes = des-cbc-md5 des-cbc-crc des3-cbc-sha1
+ permitted_enctypes = des-cbc-md5 des-cbc-crc des3-cbc-sha1
+
+[realms]
+ $REALM = {
+  kdc = $KDC_ADDRESS
+  admin_server = $KDC_ADDRESS
+ }
+
+[domain_realm]
+ .$DOMAIN_REALM = $REALM
+ $DOMAIN_REALM = $REALM
+EOF
+
+cat>/var/kerberos/krb5kdc/kdc.conf<<EOF
+[kdcdefaults]
+ kdc_ports = 88
+ kdc_tcp_ports = 88
+
+[realms]
+ $REALM = {
+  acl_file = /var/kerberos/krb5kdc/kadm5.acl
+  dict_file = /usr/share/dict/words
+  admin_keytab = /var/kerberos/krb5kdc/kadm5.keytab
+  # WARNING: We use weaker key types to simplify testing as stronger key types
+  # require the enhanced security JCE policy file to be installed. You should
+  # NOT run with this configuration in production or any real environment. You
+  # have been warned.
+  master_key_type = des3-hmac-sha1
+  supported_enctypes = arcfour-hmac:normal des3-hmac-sha1:normal des-cbc-crc:normal des:normal des:v4 des:norealm des:onlyrealm des:afs3
+  default_principal_flags = +preauth
+ }
+EOF
+}
+
+create_db() {
+  /usr/sbin/kdb5_util -P $KERB_MASTER_KEY -r $REALM create -s
+}
+
+start_kdc() {
+  mkdir -p /var/log/kerberos
+
+  /etc/rc.d/init.d/krb5kdc start
+  /etc/rc.d/init.d/kadmin start
+
+  chkconfig krb5kdc on
+  chkconfig kadmin on
+}
+
+restart_kdc() {
+  /etc/rc.d/init.d/krb5kdc restart
+  /etc/rc.d/init.d/kadmin restart
+}
+
+create_admin_user() {
+  kadmin.local -q "addprinc -pw $KERB_ADMIN_PASS $KERB_ADMIN_USER/admin"
+  echo "*/admin@$REALM *" > /var/kerberos/krb5kdc/kadm5.acl
+}
+
+create_keytabs() {
+
+
+  # kadmin.local -q "addprinc -randkey hdfs/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+  # kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab hdfs/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+
+  # kadmin.local -q "addprinc -randkey HTTP/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+  # kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab HTTP/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+  kadmin.local -q "addprinc -randkey hdfs/kerberizedhdfs1@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab hdfs/kerberizedhdfs1@${REALM}"
+
+  kadmin.local -q "addprinc -randkey HTTP/kerberizedhdfs1@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab HTTP/kerberizedhdfs1@${REALM}"
+
+  kadmin.local -q "addprinc -randkey hdfsuser/node1@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab hdfsuser/node1@${REALM}"
+  kadmin.local -q "addprinc -randkey hdfsuser@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab hdfsuser@${REALM}"
+  kadmin.local -q "addprinc -randkey root@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab root@${REALM}"
+  kadmin.local -q "addprinc -randkey specuser@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab specuser@${REALM}"
+
+  chmod g+r /tmp/keytab/clickhouse.keytab
+}
+
+main() {
+
+  if [ ! -f /kerberos_initialized ]; then
+    create_config
+    create_db
+    create_admin_user
+    start_kdc
+
+    touch /kerberos_initialized
+  fi
+
+  if [ ! -f /var/kerberos/krb5kdc/principal ]; then
+    while true; do sleep 1000; done
+  else
+    start_kdc
+		create_keytabs
+    tail -F /var/log/kerberos/krb5kdc.log
+  fi
+
+}
+
+[[ "$0" == "${BASH_SOURCE[0]}" ]] && main "$@"
diff --git a/tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf b/tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf
new file mode 100644
index 00000000000..2c1e6f15f77
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf
@@ -0,0 +1,24 @@
+[logging]
+ default = FILE:/var/log/krb5libs.log
+ kdc = FILE:/var/log/krb5kdc.log
+ admin_server = FILE:/var/log/kadmind.log
+
+[libdefaults]
+ default_realm = TEST.CLICKHOUSE.TECH
+ dns_lookup_realm = false
+ dns_lookup_kdc = false
+ ticket_lifetime = 15s
+ forwardable = true
+ default_tgs_enctypes = des3-hmac-sha1
+ default_tkt_enctypes = des3-hmac-sha1
+ permitted_enctypes = des3-hmac-sha1
+
+[realms]
+ TEST.CLICKHOUSE.TECH = {
+  kdc = hdfskerberos
+  admin_server = hdfskerberos
+ }
+
+[domain_realm]
+ .test.clickhouse.tech = TEST.CLICKHOUSE.TECH
+ test.clickhouse.tech = TEST.CLICKHOUSE.TECH
diff --git a/tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf b/tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf
new file mode 100644
index 00000000000..ec1c54240e8
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf
@@ -0,0 +1,24 @@
+[logging]
+ default = FILE:/var/log/krb5libs.log
+ kdc = FILE:/var/log/krb5kdc.log
+ admin_server = FILE:/var/log/kadmind.log
+
+[libdefaults]
+ default_realm = TEST.CLICKHOUSE.TECH
+ dns_lookup_realm = false
+ dns_lookup_kdc = false
+ ticket_lifetime = 15d
+ forwardable = true
+ default_tgs_enctypes = des3-hmac-sha1
+ default_tkt_enctypes = des3-hmac-sha1
+ permitted_enctypes = des3-hmac-sha1
+
+[realms]
+ TEST.CLICKHOUSE.TECH = {
+  kdc = hdfskerberos
+  admin_server = hdfskerberos
+ }
+
+[domain_realm]
+ .test.clickhouse.tech = TEST.CLICKHOUSE.TECH
+ test.clickhouse.tech = TEST.CLICKHOUSE.TECH
diff --git a/tests/integration/test_storage_kerberized_hdfs/test.py b/tests/integration/test_storage_kerberized_hdfs/test.py
new file mode 100644
index 00000000000..a2a2a4ef88e
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/test.py
@@ -0,0 +1,108 @@
+import time
+import pytest
+
+import os
+
+from helpers.cluster import ClickHouseCluster
+import subprocess
+
+cluster = ClickHouseCluster(__file__)
+node1 = cluster.add_instance('node1', with_kerberized_hdfs=True, user_configs=[], main_configs=['configs/log_conf.xml', 'configs/hdfs.xml'])
+
+@pytest.fixture(scope="module")
+def started_cluster():
+    try:
+        cluster.start()
+
+        yield cluster
+
+    except Exception as ex:
+        print(ex)
+        raise ex
+    finally:
+        cluster.shutdown()
+
+def test_read_table(started_cluster):
+    data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
+    started_cluster.hdfs_api.write_data("/simple_table_function", data)
+
+    api_read = started_cluster.hdfs_api.read_data("/simple_table_function")
+    assert api_read == data
+
+    select_read = node1.query("select * from hdfs('hdfs://kerberizedhdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64')")
+    assert select_read == data
+
+
+def test_read_write_storage(started_cluster):
+    node1.query("create table SimpleHDFSStorage2 (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://kerberizedhdfs1:9000/simple_storage1', 'TSV')")
+    node1.query("insert into SimpleHDFSStorage2 values (1, 'Mark', 72.53)")
+
+    api_read = started_cluster.hdfs_api.read_data("/simple_storage1")
+    assert api_read == "1\tMark\t72.53\n"
+
+    select_read = node1.query("select * from SimpleHDFSStorage2")
+    assert select_read == "1\tMark\t72.53\n"
+
+
+def test_write_storage_not_expired(started_cluster):
+    node1.query("create table SimpleHDFSStorageNotExpired (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://kerberizedhdfs1:9000/simple_storage_not_expired', 'TSV')")
+
+    time.sleep(45)   # wait for ticket expiration
+    node1.query("insert into SimpleHDFSStorageNotExpired values (1, 'Mark', 72.53)")
+
+    api_read = started_cluster.hdfs_api.read_data("/simple_storage_not_expired")
+    assert api_read == "1\tMark\t72.53\n"
+
+    select_read = node1.query("select * from SimpleHDFSStorageNotExpired")
+    assert select_read == "1\tMark\t72.53\n"
+
+
+def test_two_users(started_cluster):
+    node1.query("create table HDFSStorOne (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://kerberizedhdfs1:9000/storage_user_one', 'TSV')")
+    node1.query("insert into HDFSStorOne values (1, 'Real', 86.00)")
+
+    node1.query("create table HDFSStorTwo (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://suser@kerberizedhdfs1:9000/user/specuser/storage_user_two', 'TSV')")
+    node1.query("insert into HDFSStorTwo values (1, 'Ideal', 74.00)")
+
+    select_read_1 = node1.query("select * from hdfs('hdfs://kerberizedhdfs1:9000/user/specuser/storage_user_two', 'TSV', 'id UInt64, text String, number Float64')")
+
+    select_read_2 = node1.query("select * from hdfs('hdfs://suser@kerberizedhdfs1:9000/storage_user_one', 'TSV', 'id UInt64, text String, number Float64')")
+
+def test_read_table_expired(started_cluster):
+    data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
+    started_cluster.hdfs_api.write_data("/simple_table_function_relogin", data)
+
+    started_cluster.pause_container('hdfskerberos')
+    time.sleep(45)
+
+    try:
+        select_read = node1.query("select * from hdfs('hdfs://reloginuser&kerberizedhdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64')")
+        assert False, "Exception have to be thrown"
+    except Exception as ex:
+        assert "DB::Exception: kinit failure:" in str(ex)
+
+    started_cluster.unpause_container('hdfskerberos')
+
+
+def test_prohibited(started_cluster):
+    node1.query("create table HDFSStorTwoProhibited (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://suser@kerberizedhdfs1:9000/storage_user_two_prohibited', 'TSV')")
+    try:
+        node1.query("insert into HDFSStorTwoProhibited values (1, 'SomeOne', 74.00)")
+        assert False, "Exception have to be thrown"
+    except Exception as ex:
+        assert "Unable to open HDFS file: /storage_user_two_prohibited error: Permission denied: user=specuser, access=WRITE" in str(ex)
+
+
+def test_cache_path(started_cluster):
+    node1.query("create table HDFSStorCachePath (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://dedicatedcachepath@kerberizedhdfs1:9000/storage_dedicated_cache_path', 'TSV')")
+    try:
+        node1.query("insert into HDFSStorCachePath values (1, 'FatMark', 92.53)")
+        assert False, "Exception have to be thrown"
+    except Exception as ex:
+        assert "DB::Exception: hadoop.security.kerberos.ticket.cache.path cannot be set per user" in str(ex)
+
+
+if __name__ == '__main__':
+    cluster.start()
+    input("Cluster created, press any key to destroy...")
+    cluster.shutdown()
diff --git a/tests/integration/test_storage_s3/test.py b/tests/integration/test_storage_s3/test.py
index ca7206ca3b5..9e49c5ebf8f 100644
--- a/tests/integration/test_storage_s3/test.py
+++ b/tests/integration/test_storage_s3/test.py
@@ -306,7 +306,8 @@ def test_multipart_put(cluster, maybe_auth, positive):
         cluster.minio_redirect_host, cluster.minio_redirect_port, bucket, filename, maybe_auth, table_format)
 
     try:
-        run_query(instance, put_query, stdin=csv_data, settings={'s3_min_upload_part_size': min_part_size_bytes})
+        run_query(instance, put_query, stdin=csv_data, settings={'s3_min_upload_part_size': min_part_size_bytes,
+                                                                 's3_max_single_part_upload_size': 0})
     except helpers.client.QueryRuntimeException:
         if positive:
             raise
diff --git a/tests/integration/test_table_functions_access_rights/__init__.py b/tests/integration/test_table_functions_access_rights/__init__.py
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/integration/test_table_functions_access_rights/test.py b/tests/integration/test_table_functions_access_rights/test.py
new file mode 100644
index 00000000000..bd2f767413a
--- /dev/null
+++ b/tests/integration/test_table_functions_access_rights/test.py
@@ -0,0 +1,55 @@
+import pytest
+from helpers.cluster import ClickHouseCluster
+from helpers.test_tools import TSV
+
+cluster = ClickHouseCluster(__file__)
+instance = cluster.add_instance('instance')
+
+
+@pytest.fixture(scope="module", autouse=True)
+def started_cluster():
+    try:
+        cluster.start()
+
+        instance.query("CREATE TABLE table1(x UInt32) ENGINE = MergeTree ORDER BY tuple()")
+        instance.query("CREATE TABLE table2(x UInt32) ENGINE = MergeTree ORDER BY tuple()")
+        instance.query("INSERT INTO table1 VALUES (1)")
+        instance.query("INSERT INTO table2 VALUES (2)")
+
+        yield cluster
+
+    finally:
+        cluster.shutdown()
+
+
+@pytest.fixture(autouse=True)
+def cleanup_after_test():
+    try:
+        yield
+    finally:
+        instance.query("DROP USER IF EXISTS A")
+
+
+def test_merge():
+    select_query = "SELECT * FROM merge('default', 'table[0-9]+') ORDER BY x"
+    assert instance.query(select_query) == "1\n2\n"
+
+    instance.query("CREATE USER A")
+    assert "it's necessary to have grant CREATE TEMPORARY TABLE ON *.*" in instance.query_and_get_error(select_query, user = 'A')
+
+    instance.query("GRANT CREATE TEMPORARY TABLE ON *.* TO A")
+    assert "no one matches regular expression" in instance.query_and_get_error(select_query, user = 'A')
+    
+    instance.query("GRANT SELECT ON default.table1 TO A")
+    assert instance.query(select_query, user = 'A') == "1\n"
+
+    instance.query("GRANT SELECT ON default.* TO A")
+    assert instance.query(select_query, user = 'A') == "1\n2\n"
+
+    instance.query("REVOKE SELECT ON default.table1 FROM A")
+    assert instance.query(select_query, user = 'A') == "2\n"
+
+    instance.query("REVOKE ALL ON default.* FROM A")
+    instance.query("GRANT SELECT ON default.table1 TO A")
+    instance.query("GRANT INSERT ON default.table2 TO A")
+    assert "it's necessary to have grant SELECT ON default.table2" in instance.query_and_get_error(select_query, user = 'A')
diff --git a/tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml b/tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml
new file mode 100644
index 00000000000..318a6bca95d
--- /dev/null
+++ b/tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml
@@ -0,0 +1,12 @@
+<yandex>
+    <shutdown_wait_unfinished>3</shutdown_wait_unfinished>
+    <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/log.log</log>
+        <errorlog>/var/log/clickhouse-server/log.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+        <stderr>/var/log/clickhouse-server/stderr.log</stderr>
+        <stdout>/var/log/clickhouse-server/stdout.log</stdout>
+    </logger>
+</yandex>
diff --git a/tests/integration/test_testkeeper_back_to_back/test.py b/tests/integration/test_testkeeper_back_to_back/test.py
index 04a2150e7f8..f74d6a4c646 100644
--- a/tests/integration/test_testkeeper_back_to_back/test.py
+++ b/tests/integration/test_testkeeper_back_to_back/test.py
@@ -4,9 +4,10 @@ import random
 import string
 import os
 import time
+from multiprocessing.dummy import Pool
 
 cluster = ClickHouseCluster(__file__)
-node = cluster.add_instance('node', main_configs=['configs/enable_test_keeper.xml'], with_zookeeper=True)
+node = cluster.add_instance('node', main_configs=['configs/enable_test_keeper.xml', 'configs/logs_conf.xml'], with_zookeeper=True)
 from kazoo.client import KazooClient
 
 _genuine_zk_instance = None
@@ -400,3 +401,125 @@ def test_end_of_session(started_cluster):
                     zk.close()
         except:
             pass
+
+def test_end_of_watches_session(started_cluster):
+    fake_zk1 = None
+    fake_zk2 = None
+    try:
+        fake_zk1 = KazooClient(hosts=cluster.get_instance_ip("node") + ":9181")
+        fake_zk1.start()
+
+        fake_zk2 = KazooClient(hosts=cluster.get_instance_ip("node") + ":9181")
+        fake_zk2.start()
+
+        fake_zk1.create("/test_end_of_watches_session")
+
+        dummy_set = 0
+        def dummy_callback(event):
+            nonlocal dummy_set
+            dummy_set += 1
+            print(event)
+
+        for child_node in range(100):
+            fake_zk1.create("/test_end_of_watches_session/" + str(child_node))
+            fake_zk1.get_children("/test_end_of_watches_session/" + str(child_node), watch=dummy_callback)
+
+        fake_zk2.get_children("/test_end_of_watches_session/" + str(0), watch=dummy_callback)
+        fake_zk2.get_children("/test_end_of_watches_session/" + str(1), watch=dummy_callback)
+
+        fake_zk1.stop()
+        fake_zk1.close()
+
+        for child_node in range(100):
+            fake_zk2.create("/test_end_of_watches_session/" + str(child_node) + "/" + str(child_node), b"somebytes")
+
+        assert dummy_set == 2
+    finally:
+        try:
+            for zk in [fake_zk1, fake_zk2]:
+                if zk:
+                    zk.stop()
+                    zk.close()
+        except:
+            pass
+
+def test_concurrent_watches(started_cluster):
+    fake_zk = get_fake_zk()
+    fake_zk.restart()
+    global_path = "/test_concurrent_watches_0"
+    fake_zk.create(global_path)
+
+    dumb_watch_triggered_counter = 0
+    all_paths_triggered = []
+
+    existing_path = []
+    all_paths_created = []
+    watches_created = 0
+    def create_path_and_watch(i):
+        nonlocal watches_created
+        nonlocal all_paths_created
+        fake_zk.ensure_path(global_path + "/" + str(i))
+        # new function each time
+        def dumb_watch(event):
+            nonlocal dumb_watch_triggered_counter
+            dumb_watch_triggered_counter += 1
+            nonlocal all_paths_triggered
+            all_paths_triggered.append(event.path)
+
+        fake_zk.get(global_path + "/" + str(i), watch=dumb_watch)
+        all_paths_created.append(global_path + "/" + str(i))
+        watches_created += 1
+        existing_path.append(i)
+
+    trigger_called = 0
+    def trigger_watch(i):
+        nonlocal trigger_called
+        trigger_called += 1
+        fake_zk.set(global_path + "/" + str(i), b"somevalue")
+        try:
+            existing_path.remove(i)
+        except:
+            pass
+
+    def call(total):
+        for i in range(total):
+            create_path_and_watch(random.randint(0, 1000))
+            time.sleep(random.random() % 0.5)
+            try:
+                rand_num = random.choice(existing_path)
+                trigger_watch(rand_num)
+            except:
+                pass
+        while existing_path:
+            try:
+                rand_num = random.choice(existing_path)
+                trigger_watch(rand_num)
+            except:
+                pass
+
+    p = Pool(10)
+    arguments = [100] * 10
+    watches_must_be_created = sum(arguments)
+    watches_trigger_must_be_called = sum(arguments)
+    watches_must_be_triggered = sum(arguments)
+    p.map(call, arguments)
+    p.close()
+
+    # waiting for late watches
+    for i in range(50):
+        if dumb_watch_triggered_counter == watches_must_be_triggered:
+            break
+
+        time.sleep(0.1)
+
+    assert watches_created == watches_must_be_created
+    assert trigger_called >= watches_trigger_must_be_called
+    assert len(existing_path) == 0
+    if dumb_watch_triggered_counter != watches_must_be_triggered:
+        print("All created paths", all_paths_created)
+        print("All triggerred paths", all_paths_triggered)
+        print("All paths len", len(all_paths_created))
+        print("All triggered len", len(all_paths_triggered))
+        print("Diff", list(set(all_paths_created) - set(all_paths_triggered)))
+
+    assert dumb_watch_triggered_counter == watches_must_be_triggered
diff --git a/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql b/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql
index 120d599bd35..998a4517163 100644
--- a/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql
+++ b/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS merge_tree;
 DROP TABLE IF EXISTS collapsing_merge_tree;
 DROP TABLE IF EXISTS versioned_collapsing_merge_tree;
diff --git a/tests/queries/0_stateless/00327_summing_composite_nested.sql b/tests/queries/0_stateless/00327_summing_composite_nested.sql
index b61bc71b892..9be21e87abf 100644
--- a/tests/queries/0_stateless/00327_summing_composite_nested.sql
+++ b/tests/queries/0_stateless/00327_summing_composite_nested.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS summing_composite_key;
 CREATE TABLE summing_composite_key (d Date, k UInt64, FirstMap Nested(k1 UInt32, k2ID Int8, s Float64), SecondMap Nested(k1ID UInt64, k2Key String, k3Type Int32, s Int64)) ENGINE = SummingMergeTree(d, k, 1);
 
diff --git a/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh b/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh
index 3cea3a65cdf..cc19a608c20 100755
--- a/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh
+++ b/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh
@@ -42,13 +42,13 @@ $CLICKHOUSE_CLIENT -q "INSERT INTO $name (date, Sign, ki) SELECT
 toDate(0) AS date,
 toInt8(1) AS Sign,
 toUInt64(0) AS ki
-FROM system.numbers LIMIT 9000"
+FROM system.numbers LIMIT 9000" --server_logs_file=/dev/null
 
 $CLICKHOUSE_CLIENT -q "INSERT INTO $name (date, Sign, ki) SELECT
 toDate(0) AS date,
 toInt8(1) AS Sign,
 number AS ki
-FROM system.numbers LIMIT 9000, 9000"
+FROM system.numbers LIMIT 9000, 9000" --server_logs_file=/dev/null
 
 $CLICKHOUSE_CLIENT -q "INSERT INTO $name SELECT
 toDate(0) AS date,
@@ -67,7 +67,7 @@ number AS di09,
 number AS di10,
 [number, number+1] AS \`n.i\`,
 [hex(number), hex(number+1)] AS \`n.s\`
-FROM system.numbers LIMIT $res_rows"
+FROM system.numbers LIMIT $res_rows" --server_logs_file=/dev/null
 
 while [[ $(get_num_parts) -ne 1 ]] ; do $CLICKHOUSE_CLIENT -q "OPTIMIZE TABLE $name PARTITION 197001" --server_logs_file=/dev/null; done
 
diff --git a/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql b/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql
index 4ed053f5953..ef8655a1861 100644
--- a/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql
+++ b/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 SELECT '*** Replicated with sampling ***';
 
 DROP TABLE IF EXISTS replicated_with_sampling;
diff --git a/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql b/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql
index b7824e7efdc..634b9781c7a 100644
--- a/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql
+++ b/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql
@@ -1,3 +1,5 @@
+set optimize_on_insert = 0;
+
 drop table if exists mult_tab;
 create table mult_tab (date Date, value String, version UInt64, sign Int8) engine = VersionedCollapsingMergeTree(date, (date), 8192, sign, version);
 insert into mult_tab select '2018-01-31', 'str_' || toString(number), 0, if(number % 2, 1, -1) from system.numbers limit 10;
diff --git a/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql b/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql
index 1bfbdaf75c5..8c51a6f34da 100644
--- a/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql
+++ b/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql
@@ -1,3 +1,5 @@
+set optimize_on_insert = 0;
+
 drop table if exists tab_00577;
 create table tab_00577 (date Date, version UInt64, val UInt64) engine = ReplacingMergeTree(version) partition by date order by date settings enable_vertical_merge_algorithm = 1, vertical_merge_algorithm_min_rows_to_activate = 1, vertical_merge_algorithm_min_columns_to_activate = 0;
 insert into tab_00577 values ('2018-01-01', 2, 2), ('2018-01-01', 1, 1);
diff --git a/tests/queries/0_stateless/00616_final_single_part.sql b/tests/queries/0_stateless/00616_final_single_part.sql
index df65123e29b..6618d0b1252 100644
--- a/tests/queries/0_stateless/00616_final_single_part.sql
+++ b/tests/queries/0_stateless/00616_final_single_part.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS test_00616;
 DROP TABLE IF EXISTS replacing_00616;
 
diff --git a/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql b/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql
index f488502b13b..715129cc699 100644
--- a/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql
+++ b/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql
@@ -9,8 +9,8 @@ CREATE TABLE rename2 (p Int64, i Int64, v UInt64) ENGINE = ReplicatedReplacingMe
 INSERT INTO rename1 VALUES (0, 1, 0);
 INSERT INTO rename1 VALUES (0, 1, 1);
 
-OPTIMIZE TABLE rename1;
-OPTIMIZE TABLE rename2;
+OPTIMIZE TABLE rename1 FINAL;
+OPTIMIZE TABLE rename2 FINAL;
 SELECT * FROM rename1;
 
 RENAME TABLE rename2 TO rename3;
diff --git a/tests/queries/0_stateless/00660_optimize_final_without_partition.sql b/tests/queries/0_stateless/00660_optimize_final_without_partition.sql
index 8c1f2ebd361..6545ad6e85b 100644
--- a/tests/queries/0_stateless/00660_optimize_final_without_partition.sql
+++ b/tests/queries/0_stateless/00660_optimize_final_without_partition.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS partitioned_by_tuple;
 
 CREATE TABLE partitioned_by_tuple (d Date, x UInt8, w String, y UInt8) ENGINE SummingMergeTree (y) PARTITION BY (d, x) ORDER BY (d, x, w);
diff --git a/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql b/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql
index 77747bc0383..033202e04aa 100644
--- a/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql
+++ b/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS partitioned_by_tuple_replica1_00661;
 DROP TABLE IF EXISTS partitioned_by_tuple_replica2_00661;
 CREATE TABLE partitioned_by_tuple_replica1_00661(d Date, x UInt8, w String, y UInt8) ENGINE = ReplicatedSummingMergeTree('/clickhouse/tables/test/partitioned_by_tuple_00661', '1') PARTITION BY (d, x) ORDER BY (d, x, w);
diff --git a/tests/queries/0_stateless/00682_empty_parts_merge.sh b/tests/queries/0_stateless/00682_empty_parts_merge.sh
index 03915e3f5e4..1a336580b3a 100755
--- a/tests/queries/0_stateless/00682_empty_parts_merge.sh
+++ b/tests/queries/0_stateless/00682_empty_parts_merge.sh
@@ -4,7 +4,7 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 . "$CURDIR"/../shell_config.sh
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS ordinary_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE ordinary_00682(k UInt32) ENGINE MergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE ordinary_00682(k UInt32) ENGINE MergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO ordinary_00682(k) VALUES (1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO ordinary_00682(k) VALUES (1)"
@@ -21,7 +21,7 @@ ${CLICKHOUSE_CLIENT} --query="SELECT '*** Vertical merge ***'"
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS vertical_00682"
 ${CLICKHOUSE_CLIENT} --query="CREATE TABLE vertical_00682(k UInt32, v UInt32) ENGINE MergeTree ORDER BY k \
-    SETTINGS enable_vertical_merge_algorithm=1, \
+    SETTINGS enable_vertical_merge_algorithm=1, remove_empty_parts=0, \
              vertical_merge_algorithm_min_rows_to_activate=0, \
              vertical_merge_algorithm_min_columns_to_activate=0"
 
@@ -37,7 +37,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE vertical_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS summing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE summing_00682(k UInt32, v UInt32) ENGINE SummingMergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE summing_00682(k UInt32, v UInt32) ENGINE SummingMergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO summing_00682(k, v) VALUES (1, 1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO summing_00682(k, v) VALUES (1, 2)"
@@ -51,7 +51,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE summing_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS aggregating_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE aggregating_00682(k UInt32, v AggregateFunction(count)) ENGINE AggregatingMergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE aggregating_00682(k UInt32, v AggregateFunction(count)) ENGINE AggregatingMergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO aggregating_00682(k) VALUES (1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO aggregating_00682(k) VALUES (1)"
@@ -65,7 +65,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE aggregating_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS replacing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE replacing_00682(k UInt32, v String) ENGINE ReplacingMergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE replacing_00682(k UInt32, v String) ENGINE ReplacingMergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO replacing_00682(k, v) VALUES (1, 'a')"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO replacing_00682(k, v) VALUES (1, 'b')"
@@ -79,7 +79,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE replacing_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS collapsing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE collapsing_00682(k UInt32, v String, s Int8) ENGINE CollapsingMergeTree(s) ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE collapsing_00682(k UInt32, v String, s Int8) ENGINE CollapsingMergeTree(s) ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO collapsing_00682(k, v, s) VALUES (1, 'a', 1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO collapsing_00682(k, v, s) VALUES (2, 'b', 1)"
@@ -93,7 +93,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE collapsing_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS versioned_collapsing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE versioned_collapsing_00682(k UInt32, val String, ver UInt32, s Int8) ENGINE VersionedCollapsingMergeTree(s, ver) ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE versioned_collapsing_00682(k UInt32, val String, ver UInt32, s Int8) ENGINE VersionedCollapsingMergeTree(s, ver) ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO versioned_collapsing_00682(k, val, ver, s) VALUES (1, 'a', 0, 1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO versioned_collapsing_00682(k, val, ver, s) VALUES (2, 'b', 0, 1)"
diff --git a/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference b/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference
index f8fe62bc722..724d74d19c6 100644
--- a/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference
+++ b/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference
@@ -2,7 +2,7 @@
 1
 1
 1
-	t_00693	Memory	1	[]		1970-01-01 00:00:00	[]	[]		Memory						\N	\N
+	t_00693	Memory	1	[]		1970-01-01 00:00:00	[]	[]	CREATE TEMPORARY TABLE t_00693 (`x` UInt8) ENGINE = Memory	Memory						\N	\N
 1
 1
 1
diff --git a/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql
index 3483e575fbb..a1fc25fbf0b 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql
@@ -17,7 +17,7 @@ SET select_sequential_consistency=1;
 SELECT x FROM quorum1 ORDER BY x;
 SELECT x FROM quorum2 ORDER BY x;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 INSERT INTO quorum1 VALUES (4, '1990-11-15');
 INSERT INTO quorum1 VALUES (5, '1990-11-15');
diff --git a/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql
index 42d1c9060fe..914d98ff3f5 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost_alive', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost_alive', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 INSERT INTO quorum1 VALUES (1, '2018-11-15');
@@ -30,7 +30,7 @@ SET select_sequential_consistency=1;
 
 SYSTEM START FETCHES quorum1;
 SYSTEM SYNC REPLICA quorum1;
- 
+
 SELECT x FROM quorum1 ORDER BY x;
 SELECT x FROM quorum2 ORDER BY x;
 
diff --git a/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql
index 9de73ac98e1..18cad0dda41 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 SET insert_quorum_timeout=0;
@@ -27,7 +27,7 @@ SET insert_quorum_timeout=100;
 
 SYSTEM START FETCHES quorum1;
 SYSTEM SYNC REPLICA quorum1;
- 
+
 SELECT x FROM quorum1 ORDER BY x;
 SELECT x FROM quorum2 ORDER BY x;
 
diff --git a/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql
index 74f05d26b88..11c8324ac25 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql
@@ -13,7 +13,7 @@ INSERT INTO quorum1 VALUES (3, '2020-12-16');
 SYSTEM SYNC REPLICA quorum2;
 
 SET select_sequential_consistency=1;
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 SET insert_quorum_timeout=0;
 
diff --git a/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql
index fe64b9bc150..a4e262ce3c4 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum1', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum1', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 INSERT INTO quorum1 VALUES (1, '2018-11-15');
diff --git a/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql
index 07738ed0abd..40c08fd926f 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum2', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum2', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 INSERT INTO quorum1 VALUES (1, '2018-11-15');
diff --git a/tests/queries/0_stateless/00754_alter_modify_order_by.sql b/tests/queries/0_stateless/00754_alter_modify_order_by.sql
index f8c584ed052..a09d824c928 100644
--- a/tests/queries/0_stateless/00754_alter_modify_order_by.sql
+++ b/tests/queries/0_stateless/00754_alter_modify_order_by.sql
@@ -1,4 +1,5 @@
 SET send_logs_level = 'fatal';
+SET optimize_on_insert = 0;
 
 DROP TABLE IF EXISTS old_style;
 CREATE TABLE old_style(d Date, x UInt32) ENGINE MergeTree(d, x, 8192);
diff --git a/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql b/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql
index 4b150fb3826..809adfaa498 100644
--- a/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql
+++ b/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 SET send_logs_level = 'fatal';
 
 DROP TABLE IF EXISTS old_style;
diff --git a/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh b/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh
index d4344e6e8bd..0c13ad2d573 100755
--- a/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh
+++ b/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh
@@ -98,6 +98,6 @@ timeout $TIMEOUT bash -c thread6 2>&1 | grep "was not completely removed from Zo
 wait
 
 for i in {0..9}; do
-    $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS alter_table_$i" &
+    $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS alter_table_$i" 2>&1 | grep "was not completely removed from ZooKeeper" &
 done
 wait
diff --git a/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql b/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql
index 0b5845d3b04..90b1660e546 100644
--- a/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql
+++ b/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 select '-- SummingMergeTree with Nullable column without duplicates.';  
 
 drop table if exists tst;
diff --git a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference
index 1fca8dab675..e362d692f42 100644
--- a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference
+++ b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference
@@ -1,5 +1,5 @@
 122
 
-Table dictdb.dict_invalidate doesn\'t exist.
+Table dictdb.dict_invalidate doesn\'t exist
 
 133
diff --git a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh
index 3dd3c0ee363..18f04032933 100755
--- a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh
+++ b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh
@@ -51,7 +51,7 @@ function check_exception_detected()
 export -f check_exception_detected;
 timeout 30 bash -c check_exception_detected 2> /dev/null
 
-$CLICKHOUSE_CLIENT --query "SELECT last_exception FROM system.dictionaries WHERE database = 'dictdb' AND name = 'invalidate'" 2>&1 | grep -Eo "Table dictdb.dict_invalidate .* exist."
+$CLICKHOUSE_CLIENT --query "SELECT last_exception FROM system.dictionaries WHERE database = 'dictdb' AND name = 'invalidate'" 2>&1 | grep -Eo "Table dictdb.dict_invalidate .* exist"
 
 $CLICKHOUSE_CLIENT --query "
 CREATE TABLE dictdb.dict_invalidate
diff --git a/tests/queries/0_stateless/01056_create_table_as.reference b/tests/queries/0_stateless/01056_create_table_as.reference
index e69de29bb2d..c6e14cb675e 100644
--- a/tests/queries/0_stateless/01056_create_table_as.reference
+++ b/tests/queries/0_stateless/01056_create_table_as.reference
@@ -0,0 +1 @@
+1	String
diff --git a/tests/queries/0_stateless/01056_create_table_as.sql b/tests/queries/0_stateless/01056_create_table_as.sql
index bf2a143fa5a..c27f30b61d5 100644
--- a/tests/queries/0_stateless/01056_create_table_as.sql
+++ b/tests/queries/0_stateless/01056_create_table_as.sql
@@ -49,3 +49,7 @@ DROP DICTIONARY dict;
 DROP TABLE test_01056_dict_data.dict_data;
 
 DROP DATABASE test_01056_dict_data;
+
+CREATE TABLE t1 (x String) ENGINE = Memory AS SELECT 1;
+SELECT x, toTypeName(x) FROM t1;
+DROP TABLE t1;
diff --git a/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql b/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql
index 146107d6f3d..768a20c8ca4 100644
--- a/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql
+++ b/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql
@@ -9,8 +9,6 @@ select 1 from remote('127.{1,2}', currentDatabase(), test_01081) lhs join system
 --     Code: 171. DB::Exception: Received from localhost:9000. DB::Exception: Received from 127.2:9000. DB::Exception: Block structure mismatch in  function connect between PartialSortingTransform and LazyOutputFormat stream: different columns:
 --     _dummy Int Int32(size = 0), 1 UInt8 UInt8(size = 0)
 --     _dummy Int Int32(size = 0), 1 UInt8 Const(size = 0, UInt8(size = 1)).
---
--- With experimental_use_processors=1 (default at the time of writing).
 
 insert into test_01081 select * from system.numbers limit 10;
 select 1 from remote('127.{1,2}', currentDatabase(), test_01081) lhs join system.one as rhs on rhs.dummy = 1 order by 1;
diff --git a/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql b/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql
index 8c4ba1fd318..5597bf523e4 100644
--- a/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql
+++ b/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql
@@ -4,7 +4,7 @@ DROP TABLE IF EXISTS mutations_and_quorum2;
 CREATE TABLE mutations_and_quorum1 (`server_date` Date, `something` String) ENGINE = ReplicatedMergeTree('/clickhouse/tables/test_01090/mutations_and_quorum', '1') PARTITION BY toYYYYMM(server_date) ORDER BY (server_date, something);
 CREATE TABLE mutations_and_quorum2 (`server_date` Date, `something` String) ENGINE = ReplicatedMergeTree('/clickhouse/tables/test_01090/mutations_and_quorum', '2') PARTITION BY toYYYYMM(server_date) ORDER BY (server_date, something);
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 INSERT INTO mutations_and_quorum1 VALUES ('2019-01-01', 'test1'), ('2019-02-01', 'test2'), ('2019-03-01', 'test3'), ('2019-04-01', 'test4'), ('2019-05-01', 'test1'), ('2019-06-01', 'test2'), ('2019-07-01', 'test3'), ('2019-08-01', 'test4'), ('2019-09-01', 'test1'), ('2019-10-01', 'test2'), ('2019-11-01', 'test3'), ('2019-12-01', 'test4');
 
diff --git a/tests/queries/0_stateless/01098_temporary_and_external_tables.reference b/tests/queries/0_stateless/01098_temporary_and_external_tables.reference
index c3e51f64b2d..1ddb10581a8 100644
--- a/tests/queries/0_stateless/01098_temporary_and_external_tables.reference
+++ b/tests/queries/0_stateless/01098_temporary_and_external_tables.reference
@@ -1,2 +1,9 @@
+CREATE TEMPORARY TABLE tmp_table\n(\n    `n` UInt64\n)\nENGINE = Memory AS\nSELECT number AS n\nFROM numbers(42)
+CREATE TEMPORARY TABLE tmp_table\n(\n    `n` UInt64\n)\nENGINE = Memory AS\nSELECT number AS n\nFROM numbers(42)
+CREATE TEMPORARY TABLE tmp_table\n(\n    `n` UInt64\n)\nENGINE = Memory AS\nSELECT number AS n\nFROM numbers(42)
+42
+OK
 OK
 3	9
+0
+1
diff --git a/tests/queries/0_stateless/01098_temporary_and_external_tables.sh b/tests/queries/0_stateless/01098_temporary_and_external_tables.sh
index 67f57567298..54834107728 100755
--- a/tests/queries/0_stateless/01098_temporary_and_external_tables.sh
+++ b/tests/queries/0_stateless/01098_temporary_and_external_tables.sh
@@ -6,20 +6,23 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 url_without_session="https://${CLICKHOUSE_HOST}:${CLICKHOUSE_PORT_HTTPS}/?"
 url="${url_without_session}session_id=test_01098"
 
-${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "CREATE TEMPORARY TABLE tmp_table AS SELECT number AS n FROM numbers(42)" > /dev/null;
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "DROP TEMPORARY TABLE IF EXISTS tmp_table"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "CREATE TEMPORARY TABLE tmp_table AS SELECT number AS n FROM numbers(42)"
 
-name_expr="'\`' || database || '\`.\`' || name || '\`'"
-full_tmp_name=$(echo "SELECT $name_expr FROM system.tables WHERE database='_temporary_and_external_tables' AND create_table_query LIKE '%tmp_table%'" | ${CLICKHOUSE_CURL} -m 30 -sSgk "$url" -d @-)
+id=$(echo "SELECT uuid FROM system.tables WHERE name='tmp_table' AND is_temporary" | ${CLICKHOUSE_CURL} -m 31 -sSgk "$url" -d @-)
+internal_table_name="_temporary_and_external_tables.\`_tmp_$id\`"
 
-echo "SELECT * FROM $full_tmp_name" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url" -d @- | grep -F "Code: 291" > /dev/null && echo "OK"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "SHOW CREATE TEMPORARY TABLE tmp_table"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "SHOW CREATE TABLE $internal_table_name"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url_without_session" --data "SHOW CREATE TABLE $internal_table_name"
 
-echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+sum((number+GLOBAL+IN+(SELECT+number+AS+n+FROM+remote('127.0.0.2',+numbers(5))+WHERE+n+GLOBAL+IN+(SELECT+*+FROM+tmp_table)+AND+n+GLOBAL+NOT+IN+(SELECT+*+FROM+file)+))+AS+res),+sum(number*res)+FROM+remote('127.0.0.2',+numbers(10))";
+echo "SELECT COUNT() FROM tmp_table" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url" -d @-
+echo "SELECT COUNT() FROM $internal_table_name" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url" -d @- | grep -F "Code: 291" > /dev/null && echo "OK"
+echo "SELECT COUNT() FROM $internal_table_name" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url_without_session" -d @- | grep -F "Code: 291" > /dev/null && echo "OK"
+
+echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+sum((number+GLOBAL+IN+(SELECT+number+AS+n+FROM+remote('127.0.0.2',+numbers(5))+WHERE+n+GLOBAL+IN+(SELECT+*+FROM+tmp_table)+AND+n+GLOBAL+NOT+IN+(SELECT+*+FROM+file)+))+AS+res),+sum(number*res)+FROM+remote('127.0.0.2',+numbers(10))"
+
+echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+_1%2BsleepEachRow(3)+FROM+file" &
 
-echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+sleepEachRow(3)+FROM+file" > /dev/null &
-sleep 1
-full_tmp_names=$(echo "SELECT $name_expr FROM system.tables WHERE database='_temporary_and_external_tables' FORMAT TSV" | ${CLICKHOUSE_CURL} -m 30 -sSgk "$url_without_session" -d @-)
-for name in $full_tmp_names
-do
-  ${CLICKHOUSE_CURL} -m 30 -sSk "${url_without_session}query=SHOW+CREATE+TABLE+$name" 1>/dev/null 2>/dev/null
-done;
 wait
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "DROP TEMPORARY TABLE tmp_table"
diff --git a/tests/queries/0_stateless/01187_set_profile_as_setting.reference b/tests/queries/0_stateless/01187_set_profile_as_setting.reference
new file mode 100644
index 00000000000..ae1cc97819b
--- /dev/null
+++ b/tests/queries/0_stateless/01187_set_profile_as_setting.reference
@@ -0,0 +1,8 @@
+0	0
+0	0
+OK
+1	1
+2	1
+2	1
+OK
+1	1
diff --git a/tests/queries/0_stateless/01187_set_profile_as_setting.sh b/tests/queries/0_stateless/01187_set_profile_as_setting.sh
new file mode 100755
index 00000000000..8247ab8870a
--- /dev/null
+++ b/tests/queries/0_stateless/01187_set_profile_as_setting.sh
@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+$CLICKHOUSE_CLIENT -n -m -q "select value, changed from system.settings where name='readonly';"
+$CLICKHOUSE_CLIENT -n -m -q "set profile='default'; select value, changed from system.settings where name='readonly';"
+$CLICKHOUSE_CLIENT -n -m -q "set profile='readonly'; select value, changed from system.settings where name='readonly';" 2>&1| grep -Fa "Cannot modify 'send_logs_level' setting in readonly mode" > /dev/null && echo "OK"
+CLICKHOUSE_CLIENT=$(echo ${CLICKHOUSE_CLIENT} | sed 's/'"--send_logs_level=${CLICKHOUSE_CLIENT_SERVER_LOGS_LEVEL}"'/--send_logs_level=fatal/g')
+$CLICKHOUSE_CLIENT -n -m -q "set profile='readonly'; select value, changed from system.settings where name='readonly';"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=select+value,changed+from+system.settings+where+name='readonly'"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&profile=default&query=select+value,changed+from+system.settings+where+name='readonly'"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&profile=readonly&query=select+value,changed+from+system.settings+where+name='readonly'" 2>&1| grep -Fa "Cannot modify 'readonly' setting in readonly mode" > /dev/null && echo "OK"
+echo "select value, changed from system.settings where name='readonly';" | ${CLICKHOUSE_CURL} -sSg "${CLICKHOUSE_URL}&profile=readonly" -d @-
diff --git a/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql b/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql
index 575b4dfa4a5..dba734454c0 100644
--- a/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql
+++ b/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS data_01285;
 
 SET max_threads=1;
diff --git a/tests/queries/0_stateless/01293_show_settings.reference b/tests/queries/0_stateless/01293_show_settings.reference
new file mode 100644
index 00000000000..feebc9e473d
--- /dev/null
+++ b/tests/queries/0_stateless/01293_show_settings.reference
@@ -0,0 +1,5 @@
+send_timeout	Seconds	300
+connect_timeout	Seconds	10
+connect_timeout_with_failover_ms	Milliseconds	50
+connect_timeout_with_failover_secure_ms	Milliseconds	100
+max_memory_usage	UInt64	10000000000
diff --git a/tests/queries/0_stateless/01293_show_settings.sql b/tests/queries/0_stateless/01293_show_settings.sql
new file mode 100644
index 00000000000..08f00ed201c
--- /dev/null
+++ b/tests/queries/0_stateless/01293_show_settings.sql
@@ -0,0 +1,3 @@
+show settings like 'send_timeout';
+SHOW SETTINGS ILIKE '%CONNECT_timeout%';
+SHOW CHANGED SETTINGS ILIKE '%MEMORY%';
diff --git a/tests/queries/0_stateless/01322_cast_keep_nullable.reference b/tests/queries/0_stateless/01322_cast_keep_nullable.reference
index 8ad99a10170..cfe80a7c1b1 100644
--- a/tests/queries/0_stateless/01322_cast_keep_nullable.reference
+++ b/tests/queries/0_stateless/01322_cast_keep_nullable.reference
@@ -4,6 +4,7 @@
 1	Nullable(Int32)
 2	Nullable(Float32)
 2	Nullable(UInt8)
+00000000-0000-0002-0000-000000000000	Nullable(UUID)
 3	Nullable(Int32)
 \N	Nullable(Int32)
 42	Nullable(Int32)
diff --git a/tests/queries/0_stateless/01322_cast_keep_nullable.sql b/tests/queries/0_stateless/01322_cast_keep_nullable.sql
index 10918717469..ef5f5747a7d 100644
--- a/tests/queries/0_stateless/01322_cast_keep_nullable.sql
+++ b/tests/queries/0_stateless/01322_cast_keep_nullable.sql
@@ -10,7 +10,7 @@ SELECT CAST(toNullable(toInt8(1)) AS Int32) as x, toTypeName(x);
 
 SELECT CAST(toNullable(toFloat32(2)), 'Float32') as x, toTypeName(x);
 SELECT CAST(toNullable(toFloat32(2)), 'UInt8') as x, toTypeName(x);
-SELECT CAST(toNullable(toFloat32(2)), 'UUID') as x, toTypeName(x); -- { serverError 70 }
+SELECT CAST(toNullable(toFloat32(2)), 'UUID') as x, toTypeName(x);
 
 SELECT CAST(if(1 = 1, toNullable(toInt8(3)), NULL) AS Int32) as x, toTypeName(x);
 SELECT CAST(if(1 = 0, toNullable(toInt8(3)), NULL) AS Int32) as x, toTypeName(x);
diff --git a/tests/queries/0_stateless/01323_add_scalars_in_time.sql b/tests/queries/0_stateless/01323_add_scalars_in_time.sql
index 2d7cf270017..2ee5603f760 100644
--- a/tests/queries/0_stateless/01323_add_scalars_in_time.sql
+++ b/tests/queries/0_stateless/01323_add_scalars_in_time.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS tags;
 
 CREATE TABLE tags (
diff --git a/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql b/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql
index 790fbca6b73..c5a874efe09 100644
--- a/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql
+++ b/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS tt_01373;
 
 CREATE TABLE tt_01373
diff --git a/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql b/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql
index fca14b81e27..72e588c5c02 100644
--- a/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql
+++ b/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql
@@ -22,7 +22,7 @@ SELECT * FROM replica2;
 -- drop of empty partition works
 ALTER TABLE replica2 DROP PARTITION ID 'all';
 
-SET insert_quorum=2;
+SET insert_quorum = 2, insert_quorum_parallel = 0;
 
 INSERT INTO replica2 VALUES (1);
 
diff --git a/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh b/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh
index 48cf034a4c3..e033e0d5b72 100755
--- a/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh
+++ b/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh
@@ -17,7 +17,7 @@ done
 function thread {
     for x in {0..99}; do
         while true; do
-            $CLICKHOUSE_CLIENT --insert_quorum 5 --query "INSERT INTO r$1 SELECT $x" 2>&1 | grep -qF 'Quorum for previous write has not been satisfied yet' || break
+            $CLICKHOUSE_CLIENT --insert_quorum 5 --insert_quorum_parallel 0 --query "INSERT INTO r$1 SELECT $x" 2>&1 | grep -qF 'Quorum for previous write has not been satisfied yet' || break
         done
     done
 }
diff --git a/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql b/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql
index 6a6be32278d..0eb9f20ad91 100644
--- a/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql
+++ b/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql
@@ -17,7 +17,7 @@ SYSTEM SYNC REPLICA quorum3;
 
 SET select_sequential_consistency=0;
 SET optimize_trivial_count_query=1;
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 SYSTEM STOP FETCHES quorum1;
 
diff --git a/tests/queries/0_stateless/01550_create_map_type.reference b/tests/queries/0_stateless/01550_create_map_type.reference
new file mode 100644
index 00000000000..877199e92d5
--- /dev/null
+++ b/tests/queries/0_stateless/01550_create_map_type.reference
@@ -0,0 +1,29 @@
+zhangsan
+lisi
+1111	2222
+1112	2224
+1113	2226
+female
+zhangsan
+gender
+1116
+1117
+1118
+1119
+[]
+[]
+[]
+[]
+[]
+[]
+[0,2,0]
+[1,2,3]
+[1,3,2]
+[2,4,4]
+[3,5,6]
+[4,6,8]
+[5,7,10]
+[100,20,90]
+{1:'1',2:'2',3:'foo'}	1
+200000	560000	0
+200000	560000	0
diff --git a/tests/queries/0_stateless/01550_create_map_type.sql b/tests/queries/0_stateless/01550_create_map_type.sql
new file mode 100644
index 00000000000..7a08886a5d8
--- /dev/null
+++ b/tests/queries/0_stateless/01550_create_map_type.sql
@@ -0,0 +1,65 @@
+set allow_experimental_map_type = 1;
+
+-- String type
+drop table if exists table_map;
+create table table_map (a Map(String, String)) engine = Memory;
+insert into table_map values ({'name':'zhangsan', 'gender':'male'}), ({'name':'lisi', 'gender':'female'});
+select a['name'] from table_map;
+drop table if exists table_map;
+
+
+drop table if exists table_map;
+create table table_map (a Map(String, UInt64)) engine = MergeTree() order by a;
+insert into table_map select map('key1', number, 'key2', number * 2) from numbers(1111, 3);
+select a['key1'], a['key2'] from table_map;
+drop table if exists table_map;
+
+-- MergeTree Engine
+drop table if exists table_map;
+create table table_map (a Map(String, String), b String) engine = MergeTree() order by a;
+insert into table_map values ({'name':'zhangsan', 'gender':'male'}, 'name'), ({'name':'lisi', 'gender':'female'}, 'gender');
+select a[b] from table_map;
+select b from table_map where a = map('name','lisi', 'gender', 'female');
+drop table if exists table_map;
+
+-- Int type
+drop table if exists table_map;
+create table table_map(a Map(UInt8, UInt64), b UInt8) Engine = MergeTree() order by b;
+insert into table_map select map(number, number+5), number from numbers(1111,4);
+select a[b] from table_map;
+drop table if exists table_map;
+
+
+-- Array Type
+drop table if exists table_map;
+create table table_map(a Map(String, Array(UInt8))) Engine = MergeTree() order by a;
+insert into table_map values(map('k1', [1,2,3], 'k2', [4,5,6])), (map('k0', [], 'k1', [100,20,90]));
+insert into table_map select map('k1', [number, number + 2, number * 2]) from numbers(6);
+insert into table_map select map('k2', [number, number + 2, number * 2]) from numbers(6);
+select a['k1'] as col1 from table_map order by col1;
+drop table if exists table_map;
+
+SELECT CAST(([1, 2, 3], ['1', '2', 'foo']), 'Map(UInt8, String)') AS map, map[1];
+
+CREATE TABLE table_map (n UInt32, m Map(String, Int))
+ENGINE = MergeTree ORDER BY n SETTINGS min_bytes_for_wide_part = 0;
+
+-- coversion from Tuple(Array(K), Array(V))
+INSERT INTO table_map SELECT number, (arrayMap(x -> toString(x), range(number % 10 + 2)), range(number % 10 + 2)) FROM numbers(100000);
+-- coversion from Array(Tuple(K, V))
+INSERT INTO table_map SELECT number, arrayMap(x -> (toString(x), x), range(number % 10 + 2)) FROM numbers(100000);
+SELECT sum(m['1']), sum(m['7']), sum(m['100']) FROM table_map;
+
+DROP TABLE IF EXISTS table_map;
+
+CREATE TABLE table_map (n UInt32, m Map(String, Int))
+ENGINE = MergeTree ORDER BY n;
+
+-- coversion from Tuple(Array(K), Array(V))
+INSERT INTO table_map SELECT number, (arrayMap(x -> toString(x), range(number % 10 + 2)), range(number % 10 + 2)) FROM numbers(100000);
+-- coversion from Array(Tuple(K, V))
+INSERT INTO table_map SELECT number, arrayMap(x -> (toString(x), x), range(number % 10 + 2)) FROM numbers(100000);
+SELECT sum(m['1']), sum(m['7']), sum(m['100']) FROM table_map;
+
+DROP TABLE IF EXISTS table_map;
+
diff --git a/tests/queries/0_stateless/01550_type_map_formats.reference b/tests/queries/0_stateless/01550_type_map_formats.reference
new file mode 100644
index 00000000000..ca081db75a2
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats.reference
@@ -0,0 +1,46 @@
+JSON
+{
+	"meta":
+	[
+		{
+			"name": "m",
+			"type": "Map(String,UInt32)"
+		},
+		{
+			"name": "m1",
+			"type": "Map(String,Date)"
+		},
+		{
+			"name": "m2",
+			"type": "Map(String,Array(UInt32))"
+		}
+	],
+
+	"data":
+	[
+		{
+			"m": {"k1":1,"k2":2,"k3":3},
+			"m1": {"k1":"2020-05-05"},
+			"m2": {"k1":[],"k2":[7,8]}
+		},
+		{
+			"m": {"k1":10,"k3":30},
+			"m1": {"k2":"2020-06-06"},
+			"m2": {}
+		}
+	],
+
+	"rows": 2
+}
+JSONEachRow
+{"m":{"k1":1,"k2":2,"k3":3},"m1":{"k1":"2020-05-05"},"m2":{"k1":[],"k2":[7,8]}}
+{"m":{"k1":10,"k3":30},"m1":{"k2":"2020-06-06"},"m2":{}}
+CSV
+"{'k1':1,'k2':2,'k3':3}","{'k1':'2020-05-05'}","{'k1':[],'k2':[7,8]}"
+"{'k1':10,'k3':30}","{'k2':'2020-06-06'}","{}"
+TSV
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':10,'k3':30}	{'k2':'2020-06-06'}	{}
+TSKV
+m={'k1':1,'k2':2,'k3':3}	m1={'k1':'2020-05-05'}	m2={'k1':[],'k2':[7,8]}
+m={'k1':10,'k3':30}	m1={'k2':'2020-06-06'}	m2={}
diff --git a/tests/queries/0_stateless/01550_type_map_formats.sql b/tests/queries/0_stateless/01550_type_map_formats.sql
new file mode 100644
index 00000000000..6a09fca9398
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats.sql
@@ -0,0 +1,19 @@
+SET allow_experimental_map_type = 1;
+SET output_format_write_statistics = 0;
+
+DROP TABLE IF EXISTS map_formats;
+CREATE TABLE map_formats (m Map(String, UInt32), m1 Map(String, Date), m2 Map(String, Array(UInt32))) ENGINE = Log;
+
+INSERT INTO map_formats VALUES(map('k1', 1, 'k2', 2, 'k3', 3), map('k1', toDate('2020-05-05')), map('k1', [], 'k2', [7, 8]));
+INSERT INTO map_formats VALUES(map('k1', 10, 'k3', 30), map('k2', toDate('2020-06-06')), map());
+
+SELECT 'JSON';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT JSON;
+SELECT 'JSONEachRow';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT JSONEachRow;
+SELECT 'CSV';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT CSV;
+SELECT 'TSV';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT TSV;
+SELECT 'TSKV';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT TSKV;
diff --git a/tests/queries/0_stateless/01550_type_map_formats_input.reference b/tests/queries/0_stateless/01550_type_map_formats_input.reference
new file mode 100644
index 00000000000..a360e6971d9
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats_input.reference
@@ -0,0 +1,5 @@
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
diff --git a/tests/queries/0_stateless/01550_type_map_formats_input.sh b/tests/queries/0_stateless/01550_type_map_formats_input.sh
new file mode 100755
index 00000000000..e62002dc6ce
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats_input.sh
@@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+$CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS map_formats_input"
+$CLICKHOUSE_CLIENT -q "CREATE TABLE map_formats_input (m Map(String, UInt32), m1 Map(String, Date), m2 Map(String, Array(UInt32))) ENGINE = Log;" --allow_experimental_map_type 1
+
+$CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT JSONEachRow" <<< '{"m":{"k1":1,"k2":2,"k3":3},"m1":{"k1":"2020-05-05"},"m2":{"k1":[],"k2":[7,8]}}'
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
+$CLICKHOUSE_CLIENT -q "TRUNCATE TABLE map_formats_input"
+
+$CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT CSV" <<< "\"{'k1':1,'k2':2,'k3':3}\",\"{'k1':'2020-05-05'}\",\"{'k1':[],'k2':[7,8]}\""
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
+$CLICKHOUSE_CLIENT -q "TRUNCATE TABLE map_formats_input"
+
+$CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT TSV" <<< "{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}"
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
+
+$CLICKHOUSE_CLIENT -q 'SELECT * FROM map_formats_input FORMAT Native' | $CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT Native"
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
diff --git a/tests/queries/0_stateless/01556_accurate_cast_or_null.reference b/tests/queries/0_stateless/01556_accurate_cast_or_null.reference
new file mode 100644
index 00000000000..4287c755c3f
--- /dev/null
+++ b/tests/queries/0_stateless/01556_accurate_cast_or_null.reference
@@ -0,0 +1,30 @@
+\N
+5
+\N
+\N
+5
+\N
+\N
+5
+\N
+\N
+5
+\N
+5
+\N
+5
+\N
+\N
+1.000000000
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
diff --git a/tests/queries/0_stateless/01556_accurate_cast_or_null.sql b/tests/queries/0_stateless/01556_accurate_cast_or_null.sql
new file mode 100644
index 00000000000..b11ed5939fa
--- /dev/null
+++ b/tests/queries/0_stateless/01556_accurate_cast_or_null.sql
@@ -0,0 +1,35 @@
+SELECT accurateCastOrNull(-1, 'UInt8');
+SELECT accurateCastOrNull(5, 'UInt8');
+SELECT accurateCastOrNull(257, 'UInt8');
+SELECT accurateCastOrNull(-1, 'UInt16');
+SELECT accurateCastOrNull(5, 'UInt16');
+SELECT accurateCastOrNull(65536, 'UInt16');
+SELECT accurateCastOrNull(-1, 'UInt32');
+SELECT accurateCastOrNull(5, 'UInt32');
+SELECT accurateCastOrNull(4294967296, 'UInt32');
+SELECT accurateCastOrNull(-1, 'UInt64');
+SELECT accurateCastOrNull(5, 'UInt64');
+SELECT accurateCastOrNull(-1, 'UInt256');
+SELECT accurateCastOrNull(5, 'UInt256');
+
+SELECT accurateCastOrNull(-129, 'Int8');
+SELECT accurateCastOrNull(5, 'Int8');
+SELECT accurateCastOrNull(128, 'Int8');
+
+SELECT accurateCastOrNull(10, 'Decimal32(9)');
+SELECT accurateCastOrNull(1, 'Decimal32(9)');
+SELECT accurateCastOrNull(-10, 'Decimal32(9)');
+
+SELECT accurateCastOrNull('123', 'FixedString(2)');
+
+SELECT accurateCastOrNull(inf, 'Int64');
+SELECT accurateCastOrNull(inf, 'Int128');
+SELECT accurateCastOrNull(inf, 'Int256');
+SELECT accurateCastOrNull(nan, 'Int64');
+SELECT accurateCastOrNull(nan, 'Int128');
+SELECT accurateCastOrNull(nan, 'Int256');
+
+SELECT accurateCastOrNull(inf, 'UInt64');
+SELECT accurateCastOrNull(inf, 'UInt256');
+SELECT accurateCastOrNull(nan, 'UInt64');
+SELECT accurateCastOrNull(nan, 'UInt256');
diff --git a/tests/queries/0_stateless/01560_optimize_on_insert.reference b/tests/queries/0_stateless/01560_optimize_on_insert.reference
new file mode 100644
index 00000000000..7ace2043be0
--- /dev/null
+++ b/tests/queries/0_stateless/01560_optimize_on_insert.reference
@@ -0,0 +1,13 @@
+Replacing Merge Tree
+1	2020-01-01 00:00:00
+2	2020-01-02 00:00:00
+Collapsing Merge Tree
+1	1	2020-01-01 00:00:00
+Versioned Collapsing Merge Tree
+1	1	2	2020-01-01 00:00:00
+Summing Merge Tree
+1	6	2020-01-01 00:00:00
+2	6	2020-01-02 00:00:00
+Aggregating Merge Tree
+1	5	2020-01-01 00:00:00
+2	5	2020-01-02 00:00:00
diff --git a/tests/queries/0_stateless/01560_optimize_on_insert.sql b/tests/queries/0_stateless/01560_optimize_on_insert.sql
new file mode 100644
index 00000000000..9f6dac686bb
--- /dev/null
+++ b/tests/queries/0_stateless/01560_optimize_on_insert.sql
@@ -0,0 +1,35 @@
+SELECT 'Replacing Merge Tree';
+DROP TABLE IF EXISTS replacing_merge_tree;
+CREATE TABLE replacing_merge_tree (key UInt32, date Datetime) ENGINE=ReplacingMergeTree() PARTITION BY date ORDER BY key;
+INSERT INTO replacing_merge_tree VALUES (1, '2020-01-01'), (2, '2020-01-02'), (1, '2020-01-01'), (2, '2020-01-02');
+SELECT * FROM replacing_merge_tree ORDER BY key;
+DROP TABLE replacing_merge_tree;
+
+SELECT 'Collapsing Merge Tree';
+DROP TABLE IF EXISTS collapsing_merge_tree;
+CREATE TABLE collapsing_merge_tree (key UInt32, sign Int8, date Datetime) ENGINE=CollapsingMergeTree(sign) PARTITION BY date ORDER BY key;
+INSERT INTO collapsing_merge_tree VALUES (1, 1, '2020-01-01'), (2, 1, '2020-01-02'), (1, -1, '2020-01-01'), (2, -1, '2020-01-02'), (1, 1, '2020-01-01');
+SELECT * FROM collapsing_merge_tree ORDER BY key;
+DROP TABLE collapsing_merge_tree;
+
+SELECT 'Versioned Collapsing Merge Tree';
+DROP TABLE IF EXISTS versioned_collapsing_merge_tree;
+CREATE TABLE versioned_collapsing_merge_tree (key UInt32, sign Int8, version Int32, date Datetime) ENGINE=VersionedCollapsingMergeTree(sign, version) PARTITION BY date ORDER BY (key, version);
+INSERT INTO versioned_collapsing_merge_tree VALUES (1, 1, 1, '2020-01-01'), (1, -1, 1, '2020-01-01'), (1, 1, 2, '2020-01-01');
+SELECT * FROM versioned_collapsing_merge_tree ORDER BY key;
+DROP TABLE versioned_collapsing_merge_tree;
+
+SELECT 'Summing Merge Tree';
+DROP TABLE IF EXISTS summing_merge_tree;
+CREATE TABLE summing_merge_tree (key UInt32, val UInt32, date Datetime) ENGINE=SummingMergeTree(val) PARTITION BY date ORDER BY key;
+INSERT INTO summing_merge_tree VALUES (1, 1, '2020-01-01'), (2, 1, '2020-01-02'), (1, 5, '2020-01-01'), (2, 5, '2020-01-02');
+SELECT * FROM summing_merge_tree ORDER BY key;
+DROP TABLE summing_merge_tree;
+
+SELECT 'Aggregating Merge Tree';
+DROP TABLE IF EXISTS aggregating_merge_tree;
+CREATE TABLE aggregating_merge_tree (key UInt32, val SimpleAggregateFunction(max, UInt32), date Datetime) ENGINE=AggregatingMergeTree() PARTITION BY date ORDER BY key;
+INSERT INTO aggregating_merge_tree VALUES (1, 1, '2020-01-01'), (2, 1, '2020-01-02'), (1, 5, '2020-01-01'), (2, 5, '2020-01-02');
+SELECT * FROM aggregating_merge_tree ORDER BY key;
+DROP TABLE aggregating_merge_tree;
+
diff --git a/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference
new file mode 100644
index 00000000000..1c7908bf830
--- /dev/null
+++ b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference
@@ -0,0 +1,14 @@
+SimpleAggregateFunction(any, UInt64)	0
+SimpleAggregateFunction(anyLast, UInt64)	0
+SimpleAggregateFunction(min, UInt64)	0
+SimpleAggregateFunction(max, UInt64)	0
+SimpleAggregateFunction(sum, UInt64)	0
+SimpleAggregateFunction(sumWithOverflow, UInt64)	0
+SimpleAggregateFunction(groupBitAnd, UInt64)	0
+SimpleAggregateFunction(groupBitOr, UInt64)	0
+SimpleAggregateFunction(groupBitXor, UInt64)	0
+SimpleAggregateFunction(sumMap, Tuple(Array(UInt64), Array(UInt64)))	([],[])
+SimpleAggregateFunction(minMap, Tuple(Array(UInt64), Array(UInt64)))	([0],[0])
+SimpleAggregateFunction(maxMap, Tuple(Array(UInt64), Array(UInt64)))	([0],[0])
+SimpleAggregateFunction(groupArrayArray, Array(UInt64))	[0]
+SimpleAggregateFunction(groupUniqArrayArray, Array(UInt64))	[0]
diff --git a/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql
new file mode 100644
index 00000000000..00a12a69d16
--- /dev/null
+++ b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql
@@ -0,0 +1,17 @@
+with anySimpleState(number) as c select toTypeName(c), c from numbers(1);
+with anyLastSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with minSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with maxSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with sumSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with sumWithOverflowSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with groupBitAndSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with groupBitOrSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with groupBitXorSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with sumMapSimpleState(([number], [number])) as c select toTypeName(c), c from numbers(1);
+with minMapSimpleState(([number], [number])) as c select toTypeName(c), c from numbers(1);
+with maxMapSimpleState(([number], [number])) as c select toTypeName(c), c from numbers(1);
+with groupArrayArraySimpleState([number]) as c select toTypeName(c), c from numbers(1);
+with groupUniqArrayArraySimpleState([number]) as c select toTypeName(c), c from numbers(1);
+
+-- non-SimpleAggregateFunction
+with countSimpleState(number) as c select toTypeName(c), c from numbers(1); -- { serverError 36 }
diff --git a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference b/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference
deleted file mode 100644
index 37883cc6c47..00000000000
--- a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference
+++ /dev/null
@@ -1,6 +0,0 @@
--9223372036854775808
-170141183460469231731687303715884105727
--9223372036854775808
-170141183460469231731687303715884105727
-0
-0
diff --git a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql b/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql
deleted file mode 100644
index b28fc80774f..00000000000
--- a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql
+++ /dev/null
@@ -1,10 +0,0 @@
-SELECT toInt64(inf);
-SELECT toInt128(inf);
-SELECT toInt256(inf); -- { serverError 48 }
-SELECT toInt64(nan);
-SELECT toInt128(nan);
-SELECT toInt256(nan); -- { serverError 48 }
-SELECT toUInt64(inf);
-SELECT toUInt256(inf); -- { serverError 48 }
-SELECT toUInt64(nan);
-SELECT toUInt256(nan); -- { serverError 48 }
diff --git a/tests/queries/0_stateless/01581_to_int_inf_nan.reference b/tests/queries/0_stateless/01581_to_int_inf_nan.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01581_to_int_inf_nan.sql b/tests/queries/0_stateless/01581_to_int_inf_nan.sql
new file mode 100644
index 00000000000..4959b4d61e9
--- /dev/null
+++ b/tests/queries/0_stateless/01581_to_int_inf_nan.sql
@@ -0,0 +1,10 @@
+SELECT toInt64(inf); -- { serverError 70 }
+SELECT toInt128(inf); -- { serverError 70 }
+SELECT toInt256(inf); -- { serverError 70 }
+SELECT toInt64(nan); -- { serverError 70 }
+SELECT toInt128(nan); -- { serverError 70 }
+SELECT toInt256(nan); -- { serverError 70 }
+SELECT toUInt64(inf); -- { serverError 70 }
+SELECT toUInt256(inf); -- { serverError 70 }
+SELECT toUInt64(nan); -- { serverError 70 }
+SELECT toUInt256(nan); -- { serverError 70 }
diff --git a/tests/queries/0_stateless/01596_setting_limit_offset.reference b/tests/queries/0_stateless/01596_setting_limit_offset.reference
new file mode 100644
index 00000000000..96483268d43
--- /dev/null
+++ b/tests/queries/0_stateless/01596_setting_limit_offset.reference
@@ -0,0 +1,63 @@
+0
+1
+2
+3
+4
+15
+15
+16
+16
+17
+30
+30
+31
+31
+32
+102
+103
+104
+105
+105
+106
+107
+108
+109
+105
+106
+107
+108
+109
+60
+60
+61
+61
+62
+62
+63
+63
+64
+64
+60
+35
+35
+36
+36
+37
+37
+38
+38
+39
+39
+105
+106
+107
+108
+109
+12
+13
+13
+14
+14
+15
+15
+16
diff --git a/tests/queries/0_stateless/01596_setting_limit_offset.sql b/tests/queries/0_stateless/01596_setting_limit_offset.sql
new file mode 100644
index 00000000000..7c009a95837
--- /dev/null
+++ b/tests/queries/0_stateless/01596_setting_limit_offset.sql
@@ -0,0 +1,30 @@
+DROP TABLE IF EXISTS test;
+CREATE TABLE test (i UInt64) Engine = MergeTree() order by i;
+INSERT INTO test SELECT number FROM numbers(100);
+INSERT INTO test SELECT number FROM numbers(10,100);
+OPTIMIZE TABLE test FINAL;
+
+-- Only set limit
+SET limit = 5;
+SELECT * FROM test; -- 5 rows 
+SELECT * FROM test OFFSET 20; -- 5 rows 
+SELECT * FROM (SELECT i FROM test LIMIT 10 OFFSET 50) TMP; -- 5 rows 
+SELECT * FROM test LIMIT 4 OFFSET 192; -- 4 rows 
+SELECT * FROM test LIMIT 10 OFFSET 195; -- 5 rows
+
+-- Only set offset
+SET limit = 0;
+SET offset = 195;
+SELECT * FROM test; -- 5 rows
+SELECT * FROM test OFFSET 20; -- no result
+SELECT * FROM test LIMIT 100; -- no result
+SET offset = 10;
+SELECT * FROM test LIMIT 20 OFFSET 100; -- 10 rows
+SELECT * FROM test LIMIT 11 OFFSET 100; -- 1 rows
+
+-- offset and limit together
+SET limit = 10;
+SELECT * FROM test LIMIT 50 OFFSET 50; -- 10 rows 
+SELECT * FROM test LIMIT 50 OFFSET 190; -- 0 rows
+SELECT * FROM test LIMIT 50 OFFSET 185; -- 5 rows 
+SELECT * FROM test LIMIT 18 OFFSET 5; -- 8 rows 
diff --git a/tests/queries/0_stateless/01600_detach_permanently.reference b/tests/queries/0_stateless/01600_detach_permanently.reference
new file mode 100644
index 00000000000..98ed3b6762d
--- /dev/null
+++ b/tests/queries/0_stateless/01600_detach_permanently.reference
@@ -0,0 +1,102 @@
+##################
+setup env
+##################
+db_ordinary.log_table 1
+>table detached!
+db_ordinary.log_table 2
+>Table is back after restart
+10
+db_ordinary.log_table 3
+>table detached (permanently)!
+db_ordinary.log_table 4
+>table is still detached (after restart)!
+db_ordinary.log_table 5
+>Table is back after attach
+db_ordinary.log_table 6
+10
+##################
+db_ordinary.mt_table 1
+>table detached!
+db_ordinary.mt_table 2
+>Table is back after restart
+10
+db_ordinary.mt_table 3
+>table detached (permanently)!
+db_ordinary.mt_table 4
+>table is still detached (after restart)!
+db_ordinary.mt_table 5
+>Table is back after attach
+db_ordinary.mt_table 6
+10
+##################
+db_ordinary.null_table 1
+>table detached!
+db_ordinary.null_table 2
+>Table is back after restart
+0
+db_ordinary.null_table 3
+>table detached (permanently)!
+db_ordinary.null_table 4
+>table is still detached (after restart)!
+db_ordinary.null_table 5
+>Table is back after attach
+db_ordinary.null_table 6
+0
+##################
+db_atomic.log_table 1
+>table detached!
+db_atomic.log_table 2
+>Table is back after restart
+10
+db_atomic.log_table 3
+>table detached (permanently)!
+db_atomic.log_table 4
+>table is still detached (after restart)!
+db_atomic.log_table 5
+>Table is back after attach
+db_atomic.log_table 6
+10
+##################
+db_atomic.mt_table 1
+>table detached!
+db_atomic.mt_table 2
+>Table is back after restart
+10
+db_atomic.mt_table 3
+>table detached (permanently)!
+db_atomic.mt_table 4
+>table is still detached (after restart)!
+db_atomic.mt_table 5
+>Table is back after attach
+db_atomic.mt_table 6
+10
+##################
+db_atomic.null_table 1
+>table detached!
+db_atomic.null_table 2
+>Table is back after restart
+0
+db_atomic.null_table 3
+>table detached (permanently)!
+db_atomic.null_table 4
+>table is still detached (after restart)!
+db_atomic.null_table 5
+>Table is back after attach
+db_atomic.null_table 6
+0
+##################
+test for MV
+MV is working
+Usual detach works immediately till restart
+Usual detach activates after restart
+Permanent detach works immediately
+Permanent detach still works after restart
+View can be reattached
+##################
+test for MV with inner table
+MV is working
+1
+View can be reattached
+##################
+DETACH DATABASE is not implemented (proper error)
+1
diff --git a/tests/queries/0_stateless/01600_detach_permanently.sh b/tests/queries/0_stateless/01600_detach_permanently.sh
new file mode 100755
index 00000000000..e897f80a33f
--- /dev/null
+++ b/tests/queries/0_stateless/01600_detach_permanently.sh
@@ -0,0 +1,125 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+## tests with real clickhouse restart would be a bit to heavy,
+## to ensure the table will not reappear back clickhose-local is enough.
+
+WORKING_FOLDER_01600="${CLICKHOUSE_TMP}/01600_detach_permanently"
+rm -rf "${WORKING_FOLDER_01600}"
+mkdir -p "${WORKING_FOLDER_01600}"
+
+clickhouse_local() {
+    local query="$1"
+    shift
+    ${CLICKHOUSE_LOCAL} --query "$query" $@ -- --path="${WORKING_FOLDER_01600}"
+}
+
+test_detach_attach_sequence() {
+    local db="$1"
+    local table="$2"
+    echo "##################"
+
+    echo "${db}.${table} 1"
+    # normal DETACH - while process is running (clickhouse-local here, same for server) table is detached.
+    clickhouse_local "DETACH TABLE ${db}.${table}; SELECT if( count() = 0, '>table detached!', '>Fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    # but once we restart the precess (either clickhouse-local either clickhouse server) the table is back.
+    echo "${db}.${table} 2"
+    clickhouse_local "SELECT if(name='${table}', '>Table is back after restart', '>fail') FROM system.tables WHERE database='${db}' AND name='${table}'; SELECT count() FROM ${db}.${table};"
+
+    # permanent DETACH - table is detached, and metadata file renamed, prevening further attach
+    echo "${db}.${table} 3"
+    clickhouse_local "DETACH TABLE ${db}.${table} PERMANENTLY; SELECT if( count() = 0, '>table detached (permanently)!', '>Fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    # still detached after restart
+    echo "${db}.${table} 4"
+    clickhouse_local "SELECT if( count() = 0, '>table is still detached (after restart)!', '>Fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    # but can be reattached
+    echo "${db}.${table} 5"
+    clickhouse_local "ATTACH TABLE ${db}.${table}; SELECT if(name='${table}', '>Table is back after attach', '>fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    echo "${db}.${table} 6"
+    clickhouse_local "SELECT count() FROM ${db}.${table};"
+}
+
+echo "##################"
+echo "setup env"
+
+clickhouse_local "DROP DATABASE IF EXISTS db_ordinary SYNC;"
+clickhouse_local "DROP DATABASE IF EXISTS db_atomic SYNC;"
+
+clickhouse_local "CREATE DATABASE db_ordinary Engine=Ordinary"
+clickhouse_local "CREATE DATABASE db_atomic Engine=Atomic"
+
+clickhouse_local "CREATE TABLE db_ordinary.log_table Engine=Log AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_ordinary.mt_table Engine=MergeTree ORDER BY tuple() AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_ordinary.null_table Engine=Null AS SELECT * FROM numbers(10)"
+
+clickhouse_local "CREATE TABLE db_atomic.log_table Engine=Log AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_atomic.mt_table Engine=MergeTree ORDER BY tuple() AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_atomic.null_table Engine=Null AS SELECT * FROM numbers(10)"
+
+test_detach_attach_sequence "db_ordinary" "log_table"
+test_detach_attach_sequence "db_ordinary" "mt_table"
+test_detach_attach_sequence "db_ordinary" "null_table"
+
+test_detach_attach_sequence "db_atomic" "log_table"
+test_detach_attach_sequence "db_atomic" "mt_table"
+test_detach_attach_sequence "db_atomic" "null_table"
+
+echo "##################"
+echo "test for MV"
+clickhouse_local "CREATE TABLE db_ordinary.src Engine=Null AS system.numbers"
+clickhouse_local "CREATE TABLE db_ordinary.dst Engine=Log AS system.numbers"
+clickhouse_local "CREATE MATERIALIZED VIEW db_ordinary.src2dst_mv_to TO db_ordinary.dst AS SELECT * FROM db_ordinary.src"
+
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 10, 'MV is working', 'MV failed') from db_ordinary.dst;"
+
+clickhouse_local "DETACH VIEW db_ordinary.src2dst_mv_to; INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 10, 'Usual detach works immediately till restart', 'Usual detach failed') from db_ordinary.dst;"
+
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'Usual detach activates after restart', 'Usual detach reactivation failed') from db_ordinary.dst;"
+
+clickhouse_local "DETACH VIEW db_ordinary.src2dst_mv_to PERMANENTLY; INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'Permanent detach works immediately', 'Permanent detach failed') from db_ordinary.dst;"
+
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'Permanent detach still works after restart', 'Permanent detach reactivated!') from db_ordinary.dst;"
+
+## Quite silly: ATTACH MATERIALIZED VIEW don't work with short syntax (w/o select), but i can attach it using ATTACH TABLE ...
+clickhouse_local "ATTACH TABLE db_ordinary.src2dst_mv_to"
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 30, 'View can be reattached', 'can not reattach permanently detached view') from db_ordinary.dst;"
+
+clickhouse_local "DROP VIEW db_ordinary.src2dst_mv_to SYNC"
+
+echo "##################"
+echo "test for MV with inner table"
+clickhouse_local "CREATE MATERIALIZED VIEW db_ordinary.src_mv_with_inner Engine=Log AS SELECT * FROM db_ordinary.src"
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+
+clickhouse_local "SELECT if(count() = 10, 'MV is working', 'MV failed') FROM db_ordinary.src_mv_with_inner"
+
+clickhouse_local "DETACH VIEW db_ordinary.src_mv_with_inner PERMANENTLY; INSERT INTO db_ordinary.src SELECT * FROM numbers(10)" --stacktrace
+clickhouse_local "SELECT if(count() = 10, 'MV can be detached permanently', 'MV detach failed') FROM db_ordinary.src_mv_with_inner" 2>&1 | grep -c "db_ordinary.src_mv_with_inner doesn't exist"
+## Quite silly: ATTACH MATERIALIZED VIEW don't work with short syntax (w/o select), but i can attach it using ATTACH TABLE ...
+clickhouse_local "ATTACH TABLE db_ordinary.src_mv_with_inner"
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'View can be reattached', 'can not reattach permanently detached view') from db_ordinary.src_mv_with_inner;"
+
+## clickhouse_local can't work with dicts...
+# mkdir -p "${WORKING_FOLDER_01600}/user_files"
+# echo "1" > "${WORKING_FOLDER_01600}/user_files/dummy_dict.tsv"
+# clickhouse_local "DROP DICTIONARY db_ordinary.dummy; CREATE DICTIONARY db_ordinary.dummy (id UInt64) PRIMARY KEY id LAYOUT(FLAT()) SOURCE(FILE(path 'dummy_dict.tsv' format 'TabSeparated')) LIFETIME(MIN 1 MAX 10); DETACH DICTIONARY db_ordinary.dummy PERMANENTLY; SELECT dictGet('db_ordinary.dummy','val',toUInt64(1));"
+
+echo "##################"
+echo "DETACH DATABASE is not implemented (proper error)"
+clickhouse_local "DETACH DATABASE db_ordinary PERMANENTLY;" 2>&1 | grep -c 'DETACH PERMANENTLY is not implemented'
+
+# clean up
+rm -rf "${WORKING_FOLDER_01600}"
diff --git a/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql b/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql
index 9569324ba4d..b33b74c918d 100644
--- a/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql
+++ b/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql
@@ -1,17 +1,29 @@
 -- Check remerge_sort_lowered_memory_bytes_ratio setting
 
-set max_memory_usage='4Gi';
+set max_memory_usage='300Mi';
 -- enter remerge once limit*2 is reached
 set max_bytes_before_remerge_sort='10Mi';
 -- more blocks
 set max_block_size=40960;
 
--- default 2, slightly not enough:
---
---     MergeSortingTransform: Memory usage is lowered from 1.91 GiB to 980.00 MiB
+-- remerge_sort_lowered_memory_bytes_ratio default 2, slightly not enough
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 819200 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 186.25 MiB to 95.00 MiB
 --     MergeSortingTransform: Re-merging is not useful (memory usage was not lowered by remerge_sort_lowered_memory_bytes_ratio=2.0)
-select number k, repeat(toString(number), 101) v1, repeat(toString(number), 102) v2, repeat(toString(number), 103) v3 from numbers(toUInt64(10e6)) order by k limit 400e3 format Null; -- { serverError 241 }
-select number k, repeat(toString(number), 101) v1, repeat(toString(number), 102) v2, repeat(toString(number), 103) v3 from numbers(toUInt64(10e6)) order by k limit 400e3 settings remerge_sort_lowered_memory_bytes_ratio=2. format Null; -- { serverError 241 }
+select number k, repeat(toString(number), 11) v1, repeat(toString(number), 12) v2 from numbers(toUInt64(3e6)) order by k limit 400e3 format Null; -- { serverError 241 }
+select number k, repeat(toString(number), 11) v1, repeat(toString(number), 12) v2 from numbers(toUInt64(3e6)) order by k limit 400e3 settings remerge_sort_lowered_memory_bytes_ratio=2. format Null; -- { serverError 241 }
 
--- 1.91/0.98=1.94 is good
-select number k, repeat(toString(number), 101) v1, repeat(toString(number), 102) v2, repeat(toString(number), 103) v3 from numbers(toUInt64(10e6)) order by k limit 400e3 settings remerge_sort_lowered_memory_bytes_ratio=1.9 format Null;
+-- remerge_sort_lowered_memory_bytes_ratio 1.9 is good (need at least 1.91/0.98=1.94)
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 819200 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 186.25 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+select number k, repeat(toString(number), 11) v1, repeat(toString(number), 12) v2 from numbers(toUInt64(3e6)) order by k limit 400e3 settings remerge_sort_lowered_memory_bytes_ratio=1.9 format Null;
diff --git a/tests/queries/0_stateless/01600_select_in_different_types.reference b/tests/queries/0_stateless/01600_select_in_different_types.reference
new file mode 100644
index 00000000000..626fb152744
--- /dev/null
+++ b/tests/queries/0_stateless/01600_select_in_different_types.reference
@@ -0,0 +1,11 @@
+1
+0
+1
+1
+2
+2
+1
+1
+1
+0
+0
diff --git a/tests/queries/0_stateless/01600_select_in_different_types.sql b/tests/queries/0_stateless/01600_select_in_different_types.sql
new file mode 100644
index 00000000000..25d37c122e0
--- /dev/null
+++ b/tests/queries/0_stateless/01600_select_in_different_types.sql
@@ -0,0 +1,35 @@
+SELECT 1 IN (SELECT 1);
+SELECT -1 IN (SELECT 1);
+
+DROP TABLE IF EXISTS select_in_test;
+
+CREATE TABLE select_in_test(value UInt8) ENGINE=TinyLog;
+INSERT INTO select_in_test VALUES (1), (2), (3);
+
+SELECT value FROM select_in_test WHERE value IN (-1);
+SELECT value FROM select_in_test WHERE value IN (SELECT -1);
+
+SELECT value FROM select_in_test WHERE value IN (1);
+SELECT value FROM select_in_test WHERE value IN (SELECT 1);
+
+DROP TABLE select_in_test;
+
+CREATE TABLE select_in_test(value Int8) ENGINE=TinyLog;
+INSERT INTO select_in_test VALUES (-1), (2), (3);
+
+SELECT value FROM select_in_test WHERE value IN (1);
+SELECT value FROM select_in_test WHERE value IN (SELECT 1);
+
+SELECT value FROM select_in_test WHERE value IN (2);
+SELECT value FROM select_in_test WHERE value IN (SELECT 2);
+
+DROP TABLE select_in_test;
+
+SELECT 1 IN (1);
+SELECT '1' IN (SELECT 1);
+
+SELECT 1 IN (SELECT 1) SETTINGS transform_null_in = 1;
+SELECT 1 IN (SELECT 'a') SETTINGS transform_null_in = 1;
+SELECT 'a' IN (SELECT 1) SETTINGS transform_null_in = 1; -- { serverError 6 }
+SELECT 1 IN (SELECT -1) SETTINGS transform_null_in = 1;
+SELECT -1 IN (SELECT 1) SETTINGS transform_null_in = 1; -- { serverError 70 }
diff --git a/tests/queries/0_stateless/01601_accurate_cast.reference b/tests/queries/0_stateless/01601_accurate_cast.reference
new file mode 100644
index 00000000000..6ccedbdcfea
--- /dev/null
+++ b/tests/queries/0_stateless/01601_accurate_cast.reference
@@ -0,0 +1,8 @@
+5
+5
+5
+5
+5
+5
+1.000000000
+12
diff --git a/tests/queries/0_stateless/01601_accurate_cast.sql b/tests/queries/0_stateless/01601_accurate_cast.sql
new file mode 100644
index 00000000000..b5fd4fb04a4
--- /dev/null
+++ b/tests/queries/0_stateless/01601_accurate_cast.sql
@@ -0,0 +1,24 @@
+SELECT accurateCast(-1, 'UInt8'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt8');
+SELECT accurateCast(257, 'UInt8'); -- { serverError 70 }
+SELECT accurateCast(-1, 'UInt16'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt16');
+SELECT accurateCast(65536, 'UInt16'); -- { serverError 70 }
+SELECT accurateCast(-1, 'UInt32'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt32');
+SELECT accurateCast(4294967296, 'UInt32'); -- { serverError 70 }
+SELECT accurateCast(-1, 'UInt64'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt64');
+SELECT accurateCast(-1, 'UInt256'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt256');
+
+SELECT accurateCast(-129, 'Int8'); -- { serverError 70 }
+SELECT accurateCast(5, 'Int8');
+SELECT accurateCast(128, 'Int8'); -- { serverError 70 }
+
+SELECT accurateCast(10, 'Decimal32(9)'); -- { serverError 407 }
+SELECT accurateCast(1, 'Decimal32(9)');
+SELECT accurateCast(-10, 'Decimal32(9)'); -- { serverError 407 }
+
+SELECT accurateCast('123', 'FixedString(2)'); -- { serverError 131 }
+SELECT accurateCast('12', 'FixedString(2)');
diff --git a/tests/queries/0_stateless/01601_detach_permanently.reference b/tests/queries/0_stateless/01601_detach_permanently.reference
new file mode 100644
index 00000000000..6683603c972
--- /dev/null
+++ b/tests/queries/0_stateless/01601_detach_permanently.reference
@@ -0,0 +1,79 @@
+database atomic tests
+can not create table with same name as detached permanently
+can not detach twice
+can not drop detached
+can not replace with the other table
+can still show the create statement
+Row 1:
+──────
+statement: CREATE TABLE test1601_detach_permanently_atomic.test_name_reuse
+(
+    `number` UInt64
+)
+ENGINE = MergeTree
+ORDER BY tuple()
+SETTINGS index_granularity = 8192
+can not attach with bad uuid
+can attach with short syntax
+100
+can not detach permanently the table which is already detached (temporary)
+100
+After database reattachement the table is back (it was detached temporary)
+And we can detach it permanently
+After database reattachement the table is still absent (it was detached permamently)
+And we can not detach it permanently
+But we can attach it back
+And detach permanently again to check how database drop will behave
+DROP database
+-----------------------
+database ordinary tests
+can not create table with same name as detached permanently
+can not detach twice
+can not drop detached
+can not replace with the other table
+can still show the create statement
+Row 1:
+──────
+statement: CREATE TABLE test1601_detach_permanently_ordinary.test_name_reuse
+(
+    `number` UInt64
+)
+ENGINE = MergeTree
+ORDER BY tuple()
+SETTINGS index_granularity = 8192
+can attach with full syntax
+can attach with short syntax
+can not detach permanently the table which is already detached (temporary)
+After database reattachement the table is back (it was detached temporary)
+And we can detach it permanently
+After database reattachement the table is still absent (it was detached permamently)
+And we can not detach it permanently
+But we can attach it back
+And detach permanently again to check how database drop will behave
+DROP database - Directory not empty error, but database deteched
+DROP database - now success
+-----------------------
+database lazy tests
+can not create table with same name as detached permanently
+can not detach twice
+can not drop detached
+can not replace with the other table
+can still show the create statement
+Row 1:
+──────
+statement: CREATE TABLE test1601_detach_permanently_lazy.test_name_reuse
+(
+    `number` UInt64
+)
+ENGINE = Log
+can attach with full syntax
+can attach with short syntax
+can not detach permanently the table which is already detached (temporary)
+After database reattachement the table is back (it was detached temporary)
+And we can detach it permanently
+After database reattachement the table is still absent (it was detached permamently)
+And we can not detach it permanently
+But we can attach it back
+And detach permanently again to check how database drop will behave
+DROP database - Directory not empty error, but database deteched
+DROP database - now success
diff --git a/tests/queries/0_stateless/01601_detach_permanently.sql b/tests/queries/0_stateless/01601_detach_permanently.sql
new file mode 100644
index 00000000000..0e21fd8f463
--- /dev/null
+++ b/tests/queries/0_stateless/01601_detach_permanently.sql
@@ -0,0 +1,214 @@
+SELECT 'database atomic tests';
+
+DROP DATABASE IF EXISTS test1601_detach_permanently_atomic;
+CREATE DATABASE test1601_detach_permanently_atomic Engine=Atomic;
+
+create table test1601_detach_permanently_atomic.test_name_reuse (number UInt64) engine=MergeTree order by tuple();
+
+INSERT INTO test1601_detach_permanently_atomic.test_name_reuse SELECT * FROM numbers(100);
+
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY;
+
+SELECT 'can not create table with same name as detached permanently';
+create table test1601_detach_permanently_atomic.test_name_reuse (number UInt64) engine=MergeTree order by tuple(); -- { serverError 57 }
+
+SELECT 'can not detach twice';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY; -- { serverError 60 }
+DETACH table test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can not drop detached';
+drop table test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 60 }
+
+create table test1601_detach_permanently_atomic.test_name_rename_attempt (number UInt64) engine=MergeTree order by tuple();
+
+SELECT 'can not replace with the other table';
+RENAME TABLE test1601_detach_permanently_atomic.test_name_rename_attempt TO test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 57 }
+EXCHANGE TABLES test1601_detach_permanently_atomic.test_name_rename_attempt AND test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can still show the create statement';
+SHOW CREATE TABLE test1601_detach_permanently_atomic.test_name_reuse FORMAT Vertical;
+
+SELECT 'can not attach with bad uuid';
+ATTACH TABLE test1601_detach_permanently_atomic.test_name_reuse UUID '00000000-0000-0000-0000-000000000001'　(`number` UInt64　)　ENGINE = MergeTree　ORDER BY tuple()　SETTINGS index_granularity = 8192 ;  -- { serverError 57 }
+
+SELECT 'can attach with short syntax';
+ATTACH TABLE test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT count() FROM test1601_detach_permanently_atomic.test_name_reuse;
+
+DETACH table test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT 'can not detach permanently the table which is already detached (temporary)';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+DETACH DATABASE test1601_detach_permanently_atomic;
+ATTACH DATABASE test1601_detach_permanently_atomic;
+
+SELECT count() FROM test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT 'After database reattachement the table is back (it was detached temporary)';
+SELECT 'And we can detach it permanently';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY;
+
+DETACH DATABASE test1601_detach_permanently_atomic;
+ATTACH DATABASE test1601_detach_permanently_atomic;
+
+SELECT 'After database reattachement the table is still absent (it was detached permamently)';
+SELECT 'And we can not detach it permanently';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+SELECT 'But we can attach it back';
+ATTACH TABLE test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT 'And detach permanently again to check how database drop will behave';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY;
+
+SELECT 'DROP database';
+DROP DATABASE test1601_detach_permanently_atomic SYNC;
+
+SELECT '-----------------------';
+SELECT 'database ordinary tests';
+
+DROP DATABASE IF EXISTS test1601_detach_permanently_ordinary;
+CREATE DATABASE test1601_detach_permanently_ordinary Engine=Ordinary;
+
+create table test1601_detach_permanently_ordinary.test_name_reuse (number UInt64) engine=MergeTree order by tuple();
+
+INSERT INTO test1601_detach_permanently_ordinary.test_name_reuse SELECT * FROM numbers(100);
+
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+SELECT 'can not create table with same name as detached permanently';
+create table test1601_detach_permanently_ordinary.test_name_reuse (number UInt64) engine=MergeTree order by tuple(); -- { serverError 57 }
+
+SELECT 'can not detach twice';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY; -- { serverError 60 }
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can not drop detached';
+drop table test1601_detach_permanently_ordinary.test_name_reuse; -- { serverError 60 }
+
+create table test1601_detach_permanently_ordinary.test_name_rename_attempt (number UInt64) engine=MergeTree order by tuple();
+
+SELECT 'can not replace with the other table';
+RENAME TABLE test1601_detach_permanently_ordinary.test_name_rename_attempt TO test1601_detach_permanently_ordinary.test_name_reuse; -- { serverError 57 }
+
+SELECT 'can still show the create statement';
+SHOW CREATE TABLE test1601_detach_permanently_ordinary.test_name_reuse FORMAT Vertical;
+
+SELECT 'can attach with full syntax';
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse (`number` UInt64　)　ENGINE = MergeTree　ORDER BY tuple()　SETTINGS index_granularity = 8192;
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+SELECT 'can attach with short syntax';
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse;
+
+SELECT 'can not detach permanently the table which is already detached (temporary)';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+DETACH DATABASE test1601_detach_permanently_ordinary;
+ATTACH DATABASE test1601_detach_permanently_ordinary;
+
+SELECT 'After database reattachement the table is back (it was detached temporary)';
+SELECT 'And we can detach it permanently';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+DETACH DATABASE test1601_detach_permanently_ordinary;
+ATTACH DATABASE test1601_detach_permanently_ordinary;
+
+SELECT 'After database reattachement the table is still absent (it was detached permamently)';
+SELECT 'And we can not detach it permanently';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+SELECT 'But we can attach it back';
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+
+SELECT 'And detach permanently again to check how database drop will behave';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+SELECT 'DROP database - Directory not empty error, but database deteched';
+DROP DATABASE test1601_detach_permanently_ordinary; -- { serverError 1000 }
+
+ATTACH DATABASE test1601_detach_permanently_ordinary;
+
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+DROP TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+
+SELECT 'DROP database - now success';
+DROP DATABASE test1601_detach_permanently_ordinary;
+
+
+SELECT '-----------------------';
+SELECT 'database lazy tests';
+
+DROP DATABASE IF EXISTS test1601_detach_permanently_lazy;
+CREATE DATABASE test1601_detach_permanently_lazy Engine=Lazy(10);
+
+create table test1601_detach_permanently_lazy.test_name_reuse (number UInt64) engine=Log;
+
+INSERT INTO test1601_detach_permanently_lazy.test_name_reuse SELECT * FROM numbers(100);
+
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+SELECT 'can not create table with same name as detached permanently';
+create table test1601_detach_permanently_lazy.test_name_reuse (number UInt64) engine=Log; -- { serverError 57 }
+
+SELECT 'can not detach twice';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY; -- { serverError 60 }
+DETACH table test1601_detach_permanently_lazy.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can not drop detached';
+drop table test1601_detach_permanently_lazy.test_name_reuse; -- { serverError 60 }
+
+create table test1601_detach_permanently_lazy.test_name_rename_attempt (number UInt64) engine=Log;
+
+SELECT 'can not replace with the other table';
+RENAME TABLE test1601_detach_permanently_lazy.test_name_rename_attempt TO test1601_detach_permanently_lazy.test_name_reuse; -- { serverError 57 }
+
+SELECT 'can still show the create statement';
+SHOW CREATE TABLE test1601_detach_permanently_lazy.test_name_reuse FORMAT Vertical;
+
+SELECT 'can attach with full syntax';
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse (`number` UInt64　)　ENGINE = Log;
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+SELECT 'can attach with short syntax';
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse;
+
+DETACH table test1601_detach_permanently_lazy.test_name_reuse;
+
+SELECT 'can not detach permanently the table which is already detached (temporary)';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+DETACH DATABASE test1601_detach_permanently_lazy;
+ATTACH DATABASE test1601_detach_permanently_lazy;
+
+SELECT 'After database reattachement the table is back (it was detached temporary)';
+SELECT 'And we can detach it permanently';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+DETACH DATABASE test1601_detach_permanently_lazy;
+ATTACH DATABASE test1601_detach_permanently_lazy;
+
+SELECT 'After database reattachement the table is still absent (it was detached permamently)';
+SELECT 'And we can not detach it permanently';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+SELECT 'But we can attach it back';
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse;
+
+SELECT 'And detach permanently again to check how database drop will behave';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+SELECT 'DROP database - Directory not empty error, but database deteched';
+DROP DATABASE test1601_detach_permanently_lazy; -- { serverError 1000 }
+
+ATTACH DATABASE test1601_detach_permanently_lazy;
+
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse;
+DROP TABLE test1601_detach_permanently_lazy.test_name_reuse;
+
+SELECT 'DROP database - now success';
+DROP DATABASE test1601_detach_permanently_lazy;
diff --git a/tests/queries/0_stateless/01601_temporary_table_session_scope.reference b/tests/queries/0_stateless/01601_temporary_table_session_scope.reference
new file mode 100644
index 00000000000..84b736f86c9
--- /dev/null
+++ b/tests/queries/0_stateless/01601_temporary_table_session_scope.reference
@@ -0,0 +1,5 @@
+CREATE TEMPORARY TABLE tmptable (`x` UInt32) ENGINE = Memory
+CREATE TEMPORARY TABLE tmptable (`y` Float64, `z` String) ENGINE = Memory
+x	UInt32	1
+y	Float64	1
+z	String	2
diff --git a/tests/queries/0_stateless/01601_temporary_table_session_scope.sh b/tests/queries/0_stateless/01601_temporary_table_session_scope.sh
new file mode 100755
index 00000000000..fbc45aec216
--- /dev/null
+++ b/tests/queries/0_stateless/01601_temporary_table_session_scope.sh
@@ -0,0 +1,19 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+IF+EXISTS+tmptable&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+IF+EXISTS+tmptable&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=CREATE+TEMPORARY+TABLE+tmptable(x+UInt32)&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=CREATE+TEMPORARY+TABLE+tmptable(y+Float64,+z+String)&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+create_table_query+FROM+system.tables+WHERE+database=''+AND+name='tmptable'&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+create_table_query+FROM+system.tables+WHERE+database=''+AND+name='tmptable'&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+name,type,position+FROM+system.columns+WHERE+database=''+AND+table='tmptable'&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+name,type,position+FROM+system.columns+WHERE+database=''+AND+table='tmptable'&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+tmptable&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+tmptable&session_id=session_1601b"
diff --git a/tests/queries/0_stateless/01602_array_aggregation.reference b/tests/queries/0_stateless/01602_array_aggregation.reference
new file mode 100644
index 00000000000..4baef982f29
--- /dev/null
+++ b/tests/queries/0_stateless/01602_array_aggregation.reference
@@ -0,0 +1,60 @@
+Array min 	1
+Array max 	6
+Array sum 	21
+Array avg 	3.5
+Table array int min
+1
+0
+1
+Table array int max
+6
+0
+3
+Table array int sum
+21
+0
+6
+Table array int avg
+3.5
+0
+2
+Table array decimal min
+1.00000000
+0.00000000
+1.00000000
+Table array decimal max
+6.00000000
+0.00000000
+3.00000000
+Table array decimal sum
+21.00000000
+0.00000000
+6.00000000
+Table array decimal avg
+3.5
+0
+2
+Types of aggregation result array min
+Int8	Int16	Int32	Int64
+UInt8	UInt16	UInt32	UInt64
+Int128	Int256	UInt256
+Float32	Float64
+Decimal(9, 8)	Decimal(18, 8)	Decimal(38, 8)
+Types of aggregation result array max
+Int8	Int16	Int32	Int64
+UInt8	UInt16	UInt32	UInt64
+Int128	Int256	UInt256
+Float32	Float64
+Decimal(9, 8)	Decimal(18, 8)	Decimal(38, 8)
+Types of aggregation result array summ
+Int64	Int64	Int64	Int64
+UInt64	UInt64	UInt64	UInt64
+Int128	Int256	UInt256
+Float64	Float64
+Decimal(38, 8)	Decimal(38, 8)	Decimal(38, 8)
+Types of aggregation result array avg
+Float64	Float64	Float64	Float64
+Float64	Float64	Float64	Float64
+Float64	Float64	Float64
+Float64	Float64
+Float64	Float64	Float64
diff --git a/tests/queries/0_stateless/01602_array_aggregation.sql b/tests/queries/0_stateless/01602_array_aggregation.sql
new file mode 100644
index 00000000000..7159eb1442b
--- /dev/null
+++ b/tests/queries/0_stateless/01602_array_aggregation.sql
@@ -0,0 +1,60 @@
+SELECT 'Array min ', (arrayMin(array(1,2,3,4,5,6)));
+SELECT 'Array max ', (arrayMax(array(1,2,3,4,5,6)));
+SELECT 'Array sum ', (arraySum(array(1,2,3,4,5,6)));
+SELECT 'Array avg ', (arrayAvg(array(1,2,3,4,5,6)));
+
+DROP TABLE IF EXISTS test_aggregation;
+CREATE TABLE test_aggregation (x Array(Int)) ENGINE=TinyLog;
+
+INSERT INTO test_aggregation VALUES ([1,2,3,4,5,6]), ([]), ([1,2,3]);
+
+SELECT 'Table array int min';
+SELECT arrayMin(x) FROM test_aggregation;
+SELECT 'Table array int max';
+SELECT arrayMax(x) FROM test_aggregation;
+SELECT 'Table array int sum';
+SELECT arraySum(x) FROM test_aggregation;
+SELECT 'Table array int avg';
+SELECT arrayAvg(x) FROM test_aggregation;
+
+DROP TABLE test_aggregation;
+
+CREATE TABLE test_aggregation (x Array(Decimal64(8))) ENGINE=TinyLog;
+
+INSERT INTO test_aggregation VALUES ([1,2,3,4,5,6]), ([]), ([1,2,3]);
+
+SELECT 'Table array decimal min';
+SELECT arrayMin(x) FROM test_aggregation;
+SELECT 'Table array decimal max';
+SELECT arrayMax(x) FROM test_aggregation;
+SELECT 'Table array decimal sum';
+SELECT arraySum(x) FROM test_aggregation;
+SELECT 'Table array decimal avg';
+SELECT arrayAvg(x) FROM test_aggregation;
+
+DROP TABLE test_aggregation;
+
+SELECT 'Types of aggregation result array min';
+SELECT toTypeName(arrayMin([toInt8(0)])), toTypeName(arrayMin([toInt16(0)])), toTypeName(arrayMin([toInt32(0)])), toTypeName(arrayMin([toInt64(0)]));
+SELECT toTypeName(arrayMin([toUInt8(0)])), toTypeName(arrayMin([toUInt16(0)])), toTypeName(arrayMin([toUInt32(0)])), toTypeName(arrayMin([toUInt64(0)]));
+SELECT toTypeName(arrayMin([toInt128(0)])), toTypeName(arrayMin([toInt256(0)])), toTypeName(arrayMin([toUInt256(0)]));
+SELECT toTypeName(arrayMin([toFloat32(0)])), toTypeName(arrayMin([toFloat64(0)]));
+SELECT toTypeName(arrayMin([toDecimal32(0, 8)])), toTypeName(arrayMin([toDecimal64(0, 8)])), toTypeName(arrayMin([toDecimal128(0, 8)]));
+SELECT 'Types of aggregation result array max';
+SELECT toTypeName(arrayMax([toInt8(0)])), toTypeName(arrayMax([toInt16(0)])), toTypeName(arrayMax([toInt32(0)])), toTypeName(arrayMax([toInt64(0)]));
+SELECT toTypeName(arrayMax([toUInt8(0)])), toTypeName(arrayMax([toUInt16(0)])), toTypeName(arrayMax([toUInt32(0)])), toTypeName(arrayMax([toUInt64(0)]));
+SELECT toTypeName(arrayMax([toInt128(0)])), toTypeName(arrayMax([toInt256(0)])), toTypeName(arrayMax([toUInt256(0)]));
+SELECT toTypeName(arrayMax([toFloat32(0)])), toTypeName(arrayMax([toFloat64(0)]));
+SELECT toTypeName(arrayMax([toDecimal32(0, 8)])), toTypeName(arrayMax([toDecimal64(0, 8)])), toTypeName(arrayMax([toDecimal128(0, 8)]));
+SELECT 'Types of aggregation result array summ';
+SELECT toTypeName(arraySum([toInt8(0)])), toTypeName(arraySum([toInt16(0)])), toTypeName(arraySum([toInt32(0)])), toTypeName(arraySum([toInt64(0)]));
+SELECT toTypeName(arraySum([toUInt8(0)])), toTypeName(arraySum([toUInt16(0)])), toTypeName(arraySum([toUInt32(0)])), toTypeName(arraySum([toUInt64(0)]));
+SELECT toTypeName(arraySum([toInt128(0)])), toTypeName(arraySum([toInt256(0)])), toTypeName(arraySum([toUInt256(0)]));
+SELECT toTypeName(arraySum([toFloat32(0)])), toTypeName(arraySum([toFloat64(0)]));
+SELECT toTypeName(arraySum([toDecimal32(0, 8)])), toTypeName(arraySum([toDecimal64(0, 8)])), toTypeName(arraySum([toDecimal128(0, 8)]));
+SELECT 'Types of aggregation result array avg';
+SELECT toTypeName(arrayAvg([toInt8(0)])), toTypeName(arrayAvg([toInt16(0)])), toTypeName(arrayAvg([toInt32(0)])), toTypeName(arrayAvg([toInt64(0)]));
+SELECT toTypeName(arrayAvg([toUInt8(0)])), toTypeName(arrayAvg([toUInt16(0)])), toTypeName(arrayAvg([toUInt32(0)])), toTypeName(arrayAvg([toUInt64(0)]));
+SELECT toTypeName(arrayAvg([toInt128(0)])), toTypeName(arrayAvg([toInt256(0)])), toTypeName(arrayAvg([toUInt256(0)]));
+SELECT toTypeName(arrayAvg([toFloat32(0)])), toTypeName(arrayAvg([toFloat64(0)]));
+SELECT toTypeName(arrayAvg([toDecimal32(0, 8)])), toTypeName(arrayAvg([toDecimal64(0, 8)])), toTypeName(arrayAvg([toDecimal128(0, 8)]));
diff --git a/tests/queries/0_stateless/01602_max_distributed_connections.reference b/tests/queries/0_stateless/01602_max_distributed_connections.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01602_max_distributed_connections.sh b/tests/queries/0_stateless/01602_max_distributed_connections.sh
new file mode 100755
index 00000000000..8c19b6f5bb7
--- /dev/null
+++ b/tests/queries/0_stateless/01602_max_distributed_connections.sh
@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+common_opts=(
+    "--format=Null"
+
+    "--max_threads=1"
+    "--max_distributed_connections=3"
+)
+
+# NOTE: the test use higher timeout to avoid flakiness.
+timeout 9s ${CLICKHOUSE_CLIENT} "$@" "${common_opts[@]}" -q "select sleep(3) from remote('127.{1,2,3,4,5}', system.one)" --prefer_localhost_replica=0
+timeout 9s ${CLICKHOUSE_CLIENT} "$@" "${common_opts[@]}" -q "select sleep(3) from remote('127.{1,2,3,4,5}', system.one)" --prefer_localhost_replica=1
diff --git a/tests/queries/0_stateless/01602_modified_julian_day_msan.reference b/tests/queries/0_stateless/01602_modified_julian_day_msan.reference
new file mode 100644
index 00000000000..587579af915
--- /dev/null
+++ b/tests/queries/0_stateless/01602_modified_julian_day_msan.reference
@@ -0,0 +1 @@
+Ok.
diff --git a/tests/queries/0_stateless/01602_modified_julian_day_msan.sql b/tests/queries/0_stateless/01602_modified_julian_day_msan.sql
new file mode 100644
index 00000000000..d18665f0bcf
--- /dev/null
+++ b/tests/queries/0_stateless/01602_modified_julian_day_msan.sql
@@ -0,0 +1,4 @@
+SELECT tryBase64Decode(( SELECT countSubstrings(toModifiedJulianDayOrNull('\0'), '') ) AS n, ( SELECT regionIn('l. ') ) AS srocpnuv); -- { serverError 43 }
+SELECT countSubstrings(toModifiedJulianDayOrNull('\0'), ''); -- { serverError 43 }
+SELECT countSubstrings(toInt32OrNull('123qwe123'), ''); -- { serverError 43 }
+SELECT 'Ok.';
diff --git a/tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference
new file mode 100644
index 00000000000..5b321d6c87a
--- /dev/null
+++ b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference
@@ -0,0 +1,10 @@
+	test_01602a	CREATE TEMPORARY TABLE test_01602a (`x` UInt32) ENGINE = Memory	Memory	Memory	1
+	test_01602b	CREATE TEMPORARY TABLE test_01602b (`y` Float64, `z` String) ENGINE = Memory	Memory	Memory	1
+	test_01602a	x	UInt32	1			0	0	0		0	0	0	0	
+	test_01602b	y	Float64	1			0	0	0		0	0	0	0	
+	test_01602b	z	String	2			0	0	0		0	0	0	0	
+CREATE TEMPORARY TABLE test_01602a\n(\n    `x` UInt32\n)\nENGINE = Memory
+CREATE TEMPORARY TABLE test_01602b\n(\n    `y` Float64,\n    `z` String\n)\nENGINE = Memory
+0
+0
+0
diff --git a/tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql
new file mode 100644
index 00000000000..4eb679d7307
--- /dev/null
+++ b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql
@@ -0,0 +1,18 @@
+DROP TEMPORARY TABLE IF EXISTS test_01602a;
+DROP TEMPORARY TABLE IF EXISTS test_01602b;
+
+CREATE TEMPORARY TABLE test_01602a(x UInt32);
+CREATE TEMPORARY TABLE test_01602b(y Float64, z String);
+
+SELECT database, name, create_table_query, engine, engine_full, is_temporary FROM system.tables WHERE name LIKE 'test_01602%' ORDER BY name;
+SELECT * FROM system.columns WHERE table LIKE 'test_01602%' ORDER BY table, name;
+
+SHOW CREATE TEMPORARY TABLE test_01602a;
+SHOW CREATE TEMPORARY TABLE test_01602b;
+
+SELECT COUNT() FROM system.databases WHERE name='_temporary_and_external_tables';
+SELECT COUNT() FROM system.tables WHERE database='_temporary_and_external_tables';
+SELECT COUNT() FROM system.columns WHERE database='_temporary_and_external_tables';
+
+DROP TEMPORARY TABLE test_01602a;
+DROP TEMPORARY TABLE test_01602b;
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
new file mode 100644
index 00000000000..c25beb42a27
--- /dev/null
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -0,0 +1,12 @@
+2.4
+10.165
+0.00012000000000000002
+150.16500000000002
+7.775900000000001
+56622.689999999995
+5.988376688440276
+2.9941884322906844
+1.641386318314034
+1.641386318314034
+1.6413863192433382
+1.6413863192433382
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
new file mode 100644
index 00000000000..407fe3649ec
--- /dev/null
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -0,0 +1,26 @@
+SELECT toDecimal32(2, 2) * 1.2;
+SELECT toDecimal64(0.5, 2) * 20.33;
+SELECT 0.00001 * toDecimal32(12, 2);
+SELECT 30.033 * toDecimal32(5, 1);
+
+CREATE TABLE IF NOT EXISTS test01603 (
+    f64 Float64,
+    d Decimal64(3) DEFAULT toDecimal32(f64, 3),
+    f32 Float32 DEFAULT f64
+) ENGINE=MergeTree() ORDER BY f32;
+
+INSERT INTO test01603(f64) SELECT 1 / (number + 1) FROM system.numbers LIMIT 1000;
+
+SELECT sum(d * 1.1) FROM test01603;
+SELECT sum(8.01 * d) FROM test01603;
+SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
+SELECT sum(toDecimal64(40, 2) * f32) FROM test01603;
+SELECT sum(f64 * toDecimal64(0.1, 2)) FROM test01603;
+SELECT sum(toDecimal64(0.3, 2) * f32) FROM test01603;
+
+SELECT sum(f64 * d) FROM test01603;
+SELECT sum(d * f64) FROM test01603;
+SELECT sum(f32 * d) FROM test01603;
+SELECT sum(d * f32) FROM test01603;
+
+DROP TABLE IF EXISTS test01603;
diff --git a/tests/queries/0_stateless/01603_insert_select_too_many_parts.reference b/tests/queries/0_stateless/01603_insert_select_too_many_parts.reference
new file mode 100644
index 00000000000..f599e28b8ab
--- /dev/null
+++ b/tests/queries/0_stateless/01603_insert_select_too_many_parts.reference
@@ -0,0 +1 @@
+10
diff --git a/tests/queries/0_stateless/01603_insert_select_too_many_parts.sql b/tests/queries/0_stateless/01603_insert_select_too_many_parts.sql
new file mode 100644
index 00000000000..a56b680e212
--- /dev/null
+++ b/tests/queries/0_stateless/01603_insert_select_too_many_parts.sql
@@ -0,0 +1,14 @@
+DROP TABLE IF EXISTS too_many_parts;
+CREATE TABLE too_many_parts (x UInt64) ENGINE = MergeTree ORDER BY tuple() SETTINGS parts_to_delay_insert = 5, parts_to_throw_insert = 5;
+
+SYSTEM STOP MERGES too_many_parts;
+SET max_block_size = 1, min_insert_block_size_rows = 0, min_insert_block_size_bytes = 0;
+
+-- exception is not thrown if threshold is exceeded when multi-block INSERT is already started.
+INSERT INTO too_many_parts SELECT * FROM numbers(10);
+SELECT count() FROM too_many_parts;
+
+-- exception is thrown if threshold is exceeded on new INSERT.
+INSERT INTO too_many_parts SELECT * FROM numbers(10); -- { serverError 252 }
+
+DROP TABLE too_many_parts;
diff --git a/tests/queries/0_stateless/01603_remove_column_ttl.reference b/tests/queries/0_stateless/01603_remove_column_ttl.reference
new file mode 100644
index 00000000000..bb0d0b8aafe
--- /dev/null
+++ b/tests/queries/0_stateless/01603_remove_column_ttl.reference
@@ -0,0 +1,6 @@
+1	32
+2	0
+CREATE TABLE default.table_with_column_ttl\n(\n    `EventTime` DateTime,\n    `UserID` UInt64,\n    `Age` UInt8\n)\nENGINE = MergeTree\nORDER BY tuple()\nSETTINGS min_bytes_for_wide_part = 0, index_granularity = 8192
+1	32
+2	0
+3	27
diff --git a/tests/queries/0_stateless/01603_remove_column_ttl.sql b/tests/queries/0_stateless/01603_remove_column_ttl.sql
new file mode 100644
index 00000000000..556d968d270
--- /dev/null
+++ b/tests/queries/0_stateless/01603_remove_column_ttl.sql
@@ -0,0 +1,30 @@
+DROP TABLE IF EXISTS table_with_column_ttl;
+CREATE TABLE table_with_column_ttl
+(
+    EventTime DateTime,
+    UserID UInt64,
+    Age UInt8 TTL EventTime + INTERVAL 3 MONTH
+)
+ENGINE MergeTree()
+ORDER BY tuple()
+SETTINGS min_bytes_for_wide_part = 0; -- column TTL doesn't work for compact parts
+
+INSERT INTO table_with_column_ttl VALUES (now(), 1, 32);
+
+INSERT INTO table_with_column_ttl VALUES (now() - INTERVAL 4 MONTH, 2, 45);
+
+OPTIMIZE TABLE table_with_column_ttl FINAL;
+
+SELECT UserID, Age FROM table_with_column_ttl ORDER BY UserID;
+
+ALTER TABLE table_with_column_ttl MODIFY COLUMN Age REMOVE TTL;
+
+SHOW CREATE TABLE table_with_column_ttl;
+
+INSERT INTO table_with_column_ttl VALUES (now() - INTERVAL 10 MONTH, 3, 27);
+
+OPTIMIZE TABLE table_with_column_ttl FINAL;
+
+SELECT UserID, Age FROM table_with_column_ttl ORDER BY UserID;
+
+DROP TABLE table_with_column_ttl;
diff --git a/tests/queries/0_stateless/01603_rename_overwrite_bug.reference b/tests/queries/0_stateless/01603_rename_overwrite_bug.reference
new file mode 100644
index 00000000000..2c98ac3d71a
--- /dev/null
+++ b/tests/queries/0_stateless/01603_rename_overwrite_bug.reference
@@ -0,0 +1,4 @@
+0
+200
+0
+200
diff --git a/tests/queries/0_stateless/01603_rename_overwrite_bug.sql b/tests/queries/0_stateless/01603_rename_overwrite_bug.sql
new file mode 100644
index 00000000000..c857e08b243
--- /dev/null
+++ b/tests/queries/0_stateless/01603_rename_overwrite_bug.sql
@@ -0,0 +1,22 @@
+DROP database IF EXISTS test_1603_rename_bug_ordinary;
+create database test_1603_rename_bug_ordinary engine=Ordinary;
+create table test_1603_rename_bug_ordinary.foo engine=Memory as select * from numbers(100);
+create table test_1603_rename_bug_ordinary.bar engine=Log as select * from numbers(200);
+detach table test_1603_rename_bug_ordinary.foo;
+rename table test_1603_rename_bug_ordinary.bar to test_1603_rename_bug_ordinary.foo; -- { serverError 57 }
+attach table test_1603_rename_bug_ordinary.foo;
+SELECT count() from test_1603_rename_bug_ordinary.foo;
+SELECT count() from test_1603_rename_bug_ordinary.bar;
+DROP DATABASE test_1603_rename_bug_ordinary;
+
+-- was not broken, adding just in case.
+DROP database IF EXISTS test_1603_rename_bug_atomic;
+create database test_1603_rename_bug_atomic engine=Atomic;
+create table test_1603_rename_bug_atomic.foo engine=Memory as select * from numbers(100);
+create table test_1603_rename_bug_atomic.bar engine=Log as select * from numbers(200);
+detach table test_1603_rename_bug_atomic.foo;
+rename table test_1603_rename_bug_atomic.bar to test_1603_rename_bug_atomic.foo; -- { serverError 57 }
+attach table test_1603_rename_bug_atomic.foo;
+SELECT count() from test_1603_rename_bug_atomic.foo;
+SELECT count() from test_1603_rename_bug_atomic.bar;
+DROP DATABASE test_1603_rename_bug_atomic;
diff --git a/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference
new file mode 100644
index 00000000000..e674ac5fabb
--- /dev/null
+++ b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference
@@ -0,0 +1,8 @@
+AlterQuery  t1 (children 1)
+ ExpressionList (children 1)
+  AlterCommand 25 (children 1)
+   Function equals (children 1)
+    ExpressionList (children 2)
+     Identifier date
+     Function today (children 1)
+      ExpressionList
diff --git a/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql
new file mode 100644
index 00000000000..413acb789e0
--- /dev/null
+++ b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql
@@ -0,0 +1 @@
+explain ast alter table t1 delete where date = today()
diff --git a/tests/queries/0_stateless/arcadia_skip_list.txt b/tests/queries/0_stateless/arcadia_skip_list.txt
index 172dd579481..c297698641b 100644
--- a/tests/queries/0_stateless/arcadia_skip_list.txt
+++ b/tests/queries/0_stateless/arcadia_skip_list.txt
@@ -174,5 +174,7 @@
 01584_distributed_buffer_cannot_find_column
 01018_ip_dictionary
 01582_distinct_subquery_groupby
+01558_ttest
 01558_ttest_scipy
 01561_mann_whitney_scipy
+01601_custom_tld
diff --git a/tests/queries/skip_list.json b/tests/queries/skip_list.json
index 62921b17597..f3cc522be97 100644
--- a/tests/queries/skip_list.json
+++ b/tests/queries/skip_list.json
@@ -106,5 +106,177 @@
     "polymorphic-parts": [
         "01508_partition_pruning", /// bug, shoud be fixed
         "01482_move_to_prewhere_and_cast" /// bug, shoud be fixed
+    ],
+    "parallel":
+    [
+        /// Pessimistic list of tests which work badly in parallel.
+        /// Probably they need better investigation.
+        "00062_replicated_merge_tree_alter_zookeeper",
+        "00109_shard_totals_after_having",
+        "00110_external_sort",
+        "00116_storage_set",
+        "00121_drop_column_zookeeper",
+        "00133_long_shard_memory_tracker_and_exception_safety",
+        "00180_attach_materialized_view",
+        "00226_zookeeper_deduplication_and_unexpected_parts",
+        "00236_replicated_drop_on_non_leader_zookeeper",
+        "00305_http_and_readonly",
+        "00311_array_primary_key",
+        "00417_kill_query",
+        "00423_storage_log_single_thread",
+        "00429_long_http_bufferization",
+        "00446_clear_column_in_partition_concurrent_zookeeper",
+        "00446_clear_column_in_partition_zookeeper",
+        "00463_long_sessions_in_http_interface",
+        "00505_shard_secure",
+        "00508_materialized_view_to",
+        "00516_deduplication_after_drop_partition_zookeeper",
+        "00534_functions_bad_arguments10",
+        "00552_or_nullable",
+        "00564_versioned_collapsing_merge_tree",
+        "00571_non_exist_database_when_create_materializ_view",
+        "00575_illegal_column_exception_when_drop_depen_column",
+        "00599_create_view_with_subquery",
+        "00612_http_max_query_size",
+        "00619_union_highlite",
+        "00620_optimize_on_nonleader_replica_zookeeper",
+        "00625_arrays_in_nested",
+        "00626_replace_partition_from_table",
+        "00626_replace_partition_from_table_zookeeper",
+        "00633_materialized_view_and_too_many_parts_zookeeper",
+        "00652_mergetree_mutations",
+        "00652_replicated_mutations_zookeeper",
+        "00682_empty_parts_merge",
+        "00688_low_cardinality_serialization",
+        "00693_max_block_size_system_tables_columns",
+        "00699_materialized_view_mutations",
+        "00701_rollup",
+        "00715_fetch_merged_or_mutated_part_zookeeper",
+        "00751_default_databasename_for_view",
+        "00753_alter_attach",
+        "00754_alter_modify_column_partitions",
+        "00754_alter_modify_order_by_replicated_zookeeper",
+        "00763_long_lock_buffer_alter_destination_table",
+        "00804_test_alter_compression_codecs",
+        "00804_test_custom_compression_codecs",
+        "00804_test_custom_compression_codes_log_storages",
+        "00804_test_delta_codec_compression",
+        "00834_cancel_http_readonly_queries_on_client_close",
+        "00834_kill_mutation",
+        "00834_kill_mutation_replicated_zookeeper",
+        "00840_long_concurrent_select_and_drop_deadlock",
+        "00899_long_attach_memory_limit",
+        "00910_zookeeper_custom_compression_codecs_replicated",
+        "00926_adaptive_index_granularity_merge_tree",
+        "00926_adaptive_index_granularity_pk",
+        "00926_adaptive_index_granularity_replacing_merge_tree",
+        "00926_zookeeper_adaptive_index_granularity_replicated_merge_tree",
+        "00933_alter_ttl",
+        "00933_reserved_word",
+        "00933_test_fix_extra_seek_on_compressed_cache",
+        "00933_ttl_replicated_zookeeper",
+        "00933_ttl_with_default",
+        "00955_test_final_mark",
+        "00976_ttl_with_old_parts",
+        "00980_merge_alter_settings",
+        "00980_zookeeper_merge_tree_alter_settings",
+        "00988_constraints_replication_zookeeper",
+        "00989_parallel_parts_loading",
+        "00993_system_parts_race_condition_drop_zookeeper",
+        "01013_sync_replica_timeout_zookeeper",
+        "01014_lazy_database_concurrent_recreate_reattach_and_show_tables",
+        "01015_attach_part",
+        "01018_ddl_dictionaries_concurrent_requrests",
+        "01018_ddl_dictionaries_create",
+        "01018_ddl_dictionaries_select",
+        "01021_only_tuple_columns",
+        "01031_mutations_interpreter_and_context",
+        "01033_dictionaries_lifetime",
+        "01035_concurrent_move_partition_from_table_zookeeper",
+        "01045_zookeeper_system_mutations_with_parts_names",
+        "01053_ssd_dictionary",
+        "01055_compact_parts_1",
+        "01060_avro",
+        "01060_shutdown_table_after_detach",
+        "01070_materialize_ttl",
+        "01070_modify_ttl",
+        "01070_mutations_with_dependencies",
+        "01071_live_view_detach_dependency",
+        "01071_prohibition_secondary_index_with_old_format_merge_tree",
+        "01073_attach_if_not_exists",
+        "01076_parallel_alter_replicated_zookeeper",
+        "01079_parallel_alter_add_drop_column_zookeeper",
+        "01079_parallel_alter_detach_table_zookeeper",
+        "01083_expressions_in_engine_arguments",
+        "01085_max_distributed_connections_http",
+        "01092_memory_profiler",
+        "01098_temporary_and_external_tables",
+        "01107_atomic_db_detach_attach",
+        "01108_restart_replicas_rename_deadlock_zookeeper",
+        "01110_dictionary_layout_without_arguments",
+        "01114_database_atomic",
+        "01127_month_partitioning_consistency_select",
+        "01130_in_memory_parts_partitons",
+        "01135_default_and_alter_zookeeper",
+        "01148_zookeeper_path_macros_unfolding",
+        "01190_full_attach_syntax",
+        "01193_metadata_loading",
+        "01200_mutations_memory_consumption",
+        "01238_http_memory_tracking",
+        "01249_bad_arguments_for_bloom_filter",
+        "01251_dict_is_in_infinite_loop",
+        "01254_dict_load_after_detach_attach",
+        "01259_dictionary_custom_settings_ddl",
+        "01267_alter_default_key_columns_zookeeper",
+        "01268_dictionary_direct_layout",
+        "01269_alias_type_differs",
+        "01272_suspicious_codecs",
+        "01277_alter_rename_column_constraint_zookeeper",
+        "01280_ssd_complex_key_dictionary",
+        "01280_ttl_where_group_by",
+        "01281_group_by_limit_memory_tracking",
+        "01281_unsucceeded_insert_select_queries_counter",
+        "01293_system_distribution_queue",
+        "01294_lazy_database_concurrent",
+        "01294_lazy_database_concurrent_recreate_reattach_and_show_tables",
+        "01305_replica_create_drop_zookeeper",
+        "01307_multiple_leaders_zookeeper",
+        "01318_long_unsuccessful_mutation_zookeeper",
+        "01319_manual_write_to_replicas",
+        "01338_long_select_and_alter",
+        "01338_long_select_and_alter_zookeeper",
+        "01355_alter_column_with_order",
+        "01355_ilike",
+        "01357_version_collapsing_attach_detach_zookeeper",
+        "01375_compact_parts_codecs",
+        "01378_alter_rename_with_ttl_zookeeper",
+        "01388_clear_all_columns",
+        "01396_inactive_replica_cleanup_nodes_zookeeper",
+        "01412_cache_dictionary_race",
+        "01414_mutations_and_errors_zookeeper",
+        "01415_inconsistent_merge_tree_settings",
+        "01415_sticking_mutations",
+        "01417_freeze_partition_verbose",
+        "01417_freeze_partition_verbose_zookeeper",
+        "01430_modify_sample_by_zookeeper",
+        "01454_storagememory_data_race_challenge",
+        "01456_modify_column_type_via_add_drop_update",
+        "01457_create_as_table_function_structure",
+        "01459_manual_write_to_replicas",
+        "01460_DistributedFilesToInsert",
+        "01465_ttl_recompression",
+        "01471_calculate_ttl_during_merge",
+        "01493_alter_remove_properties_zookeeper",
+        "01493_storage_set_persistency",
+        "01494_storage_join_persistency",
+        "01516_drop_table_stress",
+        "01541_max_memory_usage_for_user",
+        "attach",
+        "ddl_dictionaries",
+        "dictionary",
+        "limit_memory",
+        "live_view",
+        "memory_leak",
+        "memory_limit"
     ]
 }
diff --git a/tests/testflows/rbac/tests/privileges/create/create_table.py b/tests/testflows/rbac/tests/privileges/create/create_table.py
index 3985679a08a..c39b0c9a5e4 100644
--- a/tests/testflows/rbac/tests/privileges/create/create_table.py
+++ b/tests/testflows/rbac/tests/privileges/create/create_table.py
@@ -706,6 +706,9 @@ def create_as_merge(self, node=None):
                 with When("I grant CREATE TABLE privilege to a user"):
                     node.query(f"GRANT CREATE TABLE ON {table_name} TO {user_name}")
 
+                with And("I grant SELECT privilege to a user to allow executing the table function merge()"):
+                    node.query(f"GRANT SELECT ON {source_table_name} TO {user_name}")
+
                 with Then("I try to create a table as another table"):
                     node.query(f"CREATE TABLE {table_name} AS merge(default,'{source_table_name}')", settings = [("user", f"{user_name}")])
 
diff --git a/utils/convert-month-partitioned-parts/main.cpp b/utils/convert-month-partitioned-parts/main.cpp
index af8e221a10b..bce1e08077c 100644
--- a/utils/convert-month-partitioned-parts/main.cpp
+++ b/utils/convert-month-partitioned-parts/main.cpp
@@ -1,4 +1,5 @@
 #include <DataTypes/DataTypeDate.h>
+#include <DataTypes/DataTypesNumber.h>
 #include <IO/HashingWriteBuffer.h>
 #include <IO/ReadBufferFromFile.h>
 #include <IO/WriteBufferFromFile.h>
diff --git a/utils/db-generator/query_db_generator.cpp b/utils/db-generator/query_db_generator.cpp
index fd310fe1897..b6e287278ff 100644
--- a/utils/db-generator/query_db_generator.cpp
+++ b/utils/db-generator/query_db_generator.cpp
@@ -1255,10 +1255,10 @@ void parseSelectQuery(DB::ASTPtr ast, TableList & all_tables)
 
 TableList getTablesFromSelect(std::vector<std::string> queries)
 {
-    DB::ParserQueryWithOutput parser;
     TableList result;
     for (std::string & query : queries)
     {
+        DB::ParserQueryWithOutput parser(query.data() + query.size());
         DB::ASTPtr ast = parseQuery(parser, query.data(), query.data() + query.size(), "", 0, 0);
         for (auto & select : getSelect(ast))
         {
diff --git a/utils/iotest/iotest.cpp b/utils/iotest/iotest.cpp
index e578a539bcd..5b28dbc95f9 100644
--- a/utils/iotest/iotest.cpp
+++ b/utils/iotest/iotest.cpp
@@ -6,6 +6,7 @@
 #include <Common/Stopwatch.h>
 #include <Common/ThreadPool.h>
 #include <Common/randomSeed.h>
+#include <common/getPageSize.h>
 
 #include <cstdlib>
 #include <iomanip>
@@ -46,7 +47,7 @@ void thread(int fd, int mode, size_t min_offset, size_t max_offset, size_t block
 {
     using namespace DB;
 
-    Memory<> direct_buf(block_size, sysconf(_SC_PAGESIZE));
+    Memory<> direct_buf(block_size, ::getPageSize());
     std::vector<char> simple_buf(block_size);
 
     char * buf;
diff --git a/utils/iotest/iotest_aio.cpp b/utils/iotest/iotest_aio.cpp
index 24508c1dd9f..66de4713893 100644
--- a/utils/iotest/iotest_aio.cpp
+++ b/utils/iotest/iotest_aio.cpp
@@ -14,6 +14,7 @@ int main(int, char **) { return 0; }
 #include <Common/ThreadPool.h>
 #include <Common/Stopwatch.h>
 #include <Common/randomSeed.h>
+#include <common/getPageSize.h>
 #include <pcg_random.hpp>
 #include <IO/BufferWithOwnMemory.h>
 #include <IO/ReadHelpers.h>
@@ -52,7 +53,7 @@ void thread(int fd, int mode, size_t min_offset, size_t max_offset, size_t block
 
     std::vector<Memory<>> buffers(buffers_count);
     for (size_t i = 0; i < buffers_count; ++i)
-        buffers[i] = Memory<>(block_size, sysconf(_SC_PAGESIZE));
+        buffers[i] = Memory<>(block_size, ::getPageSize());
 
     pcg64_fast rng(randomSeed());
 
diff --git a/utils/zookeeper-test/main.cpp b/utils/zookeeper-test/main.cpp
index fe7bf93fd15..8f8aac00866 100644
--- a/utils/zookeeper-test/main.cpp
+++ b/utils/zookeeper-test/main.cpp
@@ -6,10 +6,15 @@
 #include <common/LineReader.h>
 #include <common/logger_useful.h>
 #include <fmt/format.h>
+#include <random>
+#include <iterator>
+#include <algorithm>
+#include <chrono>
 
 #include <iostream>
 #include <sstream>
 #include <exception>
+#include <future>
 
 using namespace std;
 
@@ -144,9 +149,98 @@ void testMultiRequest(zkutil::ZooKeeper & zk)
     checkEq(zk, "/data/multirequest", "bbb");
 }
 
+std::mutex elements_mutex;
+std::vector<int> current_elements;
+std::atomic<int> watches_triggered = 0;
+
+void triggerWatch(const Coordination::WatchResponse &)
+{
+    watches_triggered++;
+}
+
+template<typename Iter, typename RandomGenerator>
+Iter select_randomly(Iter start, Iter end, RandomGenerator& g)
+{
+    std::uniform_int_distribution<> dis(0, std::distance(start, end) - 1);
+    std::advance(start, dis(g));
+    return start;
+}
+
+template<typename Iter>
+Iter select_randomly(Iter start, Iter end)
+{
+    static std::random_device rd;
+    static std::mt19937 gen(rd());
+    return select_randomly(start, end, gen);
+}
+
+std::atomic<int> element_counter = 0;
+std::atomic<int> failed_setup_counter = 0;
+
+void createPathAndSetWatch(zkutil::ZooKeeper & zk, const String & path_prefix, size_t total)
+{
+    for (size_t i = 0; i < total; ++i)
+    {
+        int element = element_counter++;
+        zk.createIfNotExists(path_prefix + "/" + std::to_string(element), "");
+
+        std::string result;
+        if (!zk.tryGetWatch(path_prefix + "/" + std::to_string(element), result, nullptr, triggerWatch))
+            failed_setup_counter++;
+
+        {
+            std::lock_guard lock(elements_mutex);
+            current_elements.push_back(element);
+        }
+
+        std::this_thread::sleep_for(std::chrono::milliseconds(200));
+
+        {
+            std::lock_guard lock(elements_mutex);
+            if (current_elements.empty())
+                continue;
+            element = *select_randomly(current_elements.begin(), current_elements.end());
+            current_elements.erase(std::remove(current_elements.begin(), current_elements.end(), element), current_elements.end());
+        }
+        zk.tryRemove(path_prefix + "/" + std::to_string(element));
+    }
+
+}
+
+void tryConcurrentWatches(zkutil::ZooKeeper & zk)
+{
+    std::string path_prefix = "/concurrent_watches";
+    std::vector<std::future<void>> asyncs;
+    zk.createIfNotExists(path_prefix, "");
+    for (size_t i = 0; i < 100; ++i)
+    {
+        auto callback = [&zk, path_prefix] ()
+        {
+            createPathAndSetWatch(zk, path_prefix, 100);
+        };
+        asyncs.push_back(std::async(std::launch::async, callback));
+    }
+
+    for (auto & async : asyncs)
+    {
+        async.wait();
+    }
+
+    size_t counter = 0;
+    while (watches_triggered != 100 * 100)
+    {
+        std::this_thread::sleep_for(std::chrono::milliseconds(100));
+        if (counter++ > 20)
+            break;
+    }
+
+    std::cerr << "Failed setup counter:" << failed_setup_counter << std::endl;
+    std::cerr << "Current elements size:" << current_elements.size() << std::endl;
+    std::cerr << "WatchesTriggered:" << watches_triggered << std::endl;
+}
+
 int main(int argc, char *argv[])
 {
-
     if (argc != 2)
     {
         std::cerr << "usage: " << argv[0] << " hosts" << std::endl;
@@ -168,6 +262,7 @@ int main(int argc, char *argv[])
         testMultiRequest(zk);
         testCreateSetWatchEvent(zk);
         testCreateListWatchEvent(zk);
+        tryConcurrentWatches(zk);
     }
     catch (...)
     {

From 61e189d84bff6a32acbe15b3ea9ed92a9cc98890 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:50:05 +0300
Subject: [PATCH 196/504] another fix

---
 src/Functions/FunctionBinaryArithmetic.h | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 4a54e1ffb2b..490fd631a33 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -203,8 +203,6 @@ struct DecimalBinaryOperation
     static constexpr bool is_plus_minus_compare = is_plus_minus || is_compare;
     static constexpr bool can_overflow = is_plus_minus || is_multiply;
 
-    static constexpr bool is_plus_minus_compare_mult = is_plus_minus || is_compare || is_multiply;
-
     using ResultType = ResultType_;
     using NativeResultType = typename NativeType<ResultType>::Type;
     using Op = std::conditional_t<is_float_division,
@@ -354,9 +352,9 @@ struct DecimalBinaryOperation
             {
                 for (size_t i = 0; i < size; ++i)
                     if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                        c[i] = apply(a, undec(a[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
                     else
-                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                        c[i] = apply(a, undec(a[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
                 return;
             }
             else if (scale_b != 1)

From 17331d2732d3de57870db222a2dc80441aa6febd Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:51:05 +0300
Subject: [PATCH 197/504] another fix

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 490fd631a33..1d1711b79d2 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -352,9 +352,9 @@ struct DecimalBinaryOperation
             {
                 for (size_t i = 0; i < size; ++i)
                     if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(a, undec(a[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
                     else
-                        c[i] = apply(a, undec(a[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
                 return;
             }
             else if (scale_b != 1)

From c8c99f47326eeafbb3fdb7527b7bdafa7d0cb1da Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 18 Dec 2020 18:51:37 +0300
Subject: [PATCH 198/504] Fix tests

---
 .../00933_test_fix_extra_seek_on_compressed_cache.reference     | 2 +-
 .../00961_checksums_in_system_parts_columns_table.reference     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/queries/0_stateless/00933_test_fix_extra_seek_on_compressed_cache.reference b/tests/queries/0_stateless/00933_test_fix_extra_seek_on_compressed_cache.reference
index 797f208c02b..7a08495654c 100644
--- a/tests/queries/0_stateless/00933_test_fix_extra_seek_on_compressed_cache.reference
+++ b/tests/queries/0_stateless/00933_test_fix_extra_seek_on_compressed_cache.reference
@@ -1 +1 @@
-0	36	14
+0	36	13
diff --git a/tests/queries/0_stateless/00961_checksums_in_system_parts_columns_table.reference b/tests/queries/0_stateless/00961_checksums_in_system_parts_columns_table.reference
index 099fe566817..186f2feab79 100644
--- a/tests/queries/0_stateless/00961_checksums_in_system_parts_columns_table.reference
+++ b/tests/queries/0_stateless/00961_checksums_in_system_parts_columns_table.reference
@@ -1 +1 @@
-20000101_1_1_0	test_00961	5f2e2d4bbc14336f44037e3ac667f247	ed226557cd4e18ecf3ae06c6d5e6725c	da96ff1e527a8a1f908ddf2b1d0af239
+20000101_1_1_0	test_00961	b5fce9c4ef1ca42ce4ed027389c208d2	fc3b062b646cd23d4c23d7f5920f89ae	da96ff1e527a8a1f908ddf2b1d0af239

From 2c8582a2a8679d65e40797bdb04323f39417050c Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 18:52:33 +0300
Subject: [PATCH 199/504] Squashed commit of the following:
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

commit 17331d2732d3de57870db222a2dc80441aa6febd
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:51:05 2020 +0300

    another fix

commit 61e189d84bff6a32acbe15b3ea9ed92a9cc98890
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:50:05 2020 +0300

    another fix

commit cb2ddc7d01c810939cc24781e9511da7cf5d5294
Merge: 0e235d35f8 985876b737
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:49:18 2020 +0300

    aMerge branch 'improvement/multiply-decimals-and-floats' of github:myrrc/clickhouse into improvement/multiply-decimals-and-floats

commit 0e235d35f80af6298c5373602c065969474dccfc
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:48:37 2020 +0300

    aa

commit 985876b737d8564a2850e54cefa88063c6da2f2d
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:37:00 2020 +0300

    another fix

commit e2556659e04c705e370e5ac5c79e06b84d1a27a8
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:28:21 2020 +0300

    fix3

commit a19934886028e6a8bf5320f15eea056c5fad3eb5
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:26:58 2020 +0300

    fix2

commit fbaa8d3951f4f6056e92ab68adf6647dc2cc5caf
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:24:31 2020 +0300

    fix

commit ed6d74b21837532773a3eaf870e3b6abb5547725
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:21:10 2020 +0300

    another bad solution

commit 6f1c0b118aea3ba96af3ba1a882c0fa4e6417316
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:17:03 2020 +0300

    typo

commit 0dca8562373c495ba9170ecac3bf298df022927e
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:16:38 2020 +0300

    solution fix

commit d22d045a59a8e2db0b7bff0f4bd9f11ac48aa48f
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:12:31 2020 +0300

    fixed the impl

commit 9d24cafccdaadf9803d02183a5af4a0856b407aa
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 18:06:24 2020 +0300

    added vector mult branch

commit 7611f8d15b13130b2374a5f0025ed5907e77c29e
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:54:48 2020 +0300

    compile fix

commit 6b956a2a0dbcf0b80ef976116006d14db60abebe
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:50:16 2020 +0300

    possible fix for non-const mult

commit 58ef4c34c070c08a7f9b6cc5fd337b91781f6141
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:44:24 2020 +0300

    prep for debug version

commit 3c01cfe630bb24a24d91583d7137d57be49cb5b4
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:34:51 2020 +0300

    no scale added

commit fa1f7f40f5875b6e538c9779ae5f4b0258c44c46
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:32:44 2020 +0300

    another try

commit 6488846c9200c92629953bde0e6412a83fbdae14
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:23:51 2020 +0300

    signature fix

commit bc651b304750d28ca722430ad3b4c433b8ab82eb
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:21:32 2020 +0300

    added the suboptimal solution

commit 764d4622886384b5a60229bb2a1c88d5eeac5764
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:16:54 2020 +0300

    updated test with possible right values

commit 611d400106040e5099dc47a4b7ec71f807a9f324
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:13:15 2020 +0300

    fixed compile error

commit 6c91b3dba0945a0aa40288291c2c93727ff69c59
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:10:37 2020 +0300

    added explicit scale calc

commit 1f105145994d55e903a2e0343ddaa2e706a92ff6
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 17:02:02 2020 +0300

    another attempt

commit 07bc3a2a4d26b6be743e511aac4dfddfb343ee97
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 16:49:00 2020 +0300

    reverted the overengineering

commit c4ea410a4da0872035f3ec0a0f57b87aa66999a3
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 16:45:44 2020 +0300

    updated the comments and branches

commit 15f0915ad32913a0032b497acad3367937cf2416
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 16:35:37 2020 +0300

    replaced ternary if

commit 84013d253dff75b1f522ac1756fb8c966a51f88c
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 16:33:00 2020 +0300

    fix: ternary if

commit f34c0500dbd9b3d9af75be1151238cd3668e9b1e
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 16:29:53 2020 +0300

    enlarged the test, updated scales calc

commit d4f8c46210fce2491cd4ad49e41f9e1bf82d0170
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 16:04:19 2020 +0300

    fixed the test

commit 283268c50feefcadfe914a286361d96d4c248f4b
Merge: 5b6143bf2f 6a9397a64b
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 15:42:39 2020 +0300

    Merge branch 'improvement/multiply-decimals-and-floats' of github:myrrc/clickhouse into improvement/multiply-decimals-and-floats

commit 5b6143bf2f3281a9e57f9316cf7c8c313f1b8ee6
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 15:42:12 2020 +0300

    fix style error

commit 05abfb04bde8685588886c75aff415c4d7aad001
Merge: a50e9548fd 26637bd7ee
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 15:42:01 2020 +0300

    Merge remote-tracking branch 'upstream/master' into improvement/multiply-decimals-and-floats

commit 26637bd7ee690fdbe902b4cfd076edb51dadc076
Merge: 0d9519a0a2 911a1cd48f
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 18 12:41:16 2020 +0300

    Merge pull request #17943 from ClickHouse/try_rewrite_writer

    Small simplification of MergeTreeDataWriter

commit 0d9519a0a2ca464e752b2593f7b061cb63eca4a4
Merge: 6ee2ada592 e4433157e7
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 08:08:26 2020 +0300

    Merge pull request #17567 from ClickHouse/insert_quorum_parallel-default

    Insert quorum parallel by default

commit 6ee2ada592a7015bebeee21320dfb90a58d8a934
Merge: 55604bde47 fb3d235e84
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 08:07:46 2020 +0300

    Merge pull request #18204 from ClickHouse/remove-strerror

    Do not use non thread-safe function `strerror`

commit 55604bde47f52e72998a1c3adaf50d4d1e1dfb68
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 08:07:09 2020 +0300

    Whitespaces

commit c343fa1ae4322f408f9ce42c58539e3a5269df2a
Merge: 6a5ce2eea7 4a1618d0b8
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 08:03:11 2020 +0300

    Merge pull request #18202 from ClickHouse/update-libc-headers-2

    Update libc headers

commit 6a5ce2eea7378e2378b54812aa176680b663455a
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 07:58:56 2020 +0300

    Update config.xml

commit 44734a829ea4add9ba88be3844408c1a1484f8e5
Merge: aa7793e38f b4a59aecfe
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 06:39:35 2020 +0300

    Merge pull request #18207 from ClickHouse/better-config

    Better config: example ports, security advices

commit b4a59aecfe7a74e74cc8921be5e2d2454fdfc6d7
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 06:39:21 2020 +0300

    Update config.xml

commit aa7793e38f7bdc2de2878787c7f5b0f58675494a
Merge: 8bc9d8ed7e d198f03301
Author: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date:   Fri Dec 18 01:54:42 2020 +0300

    Merge pull request #18084 from nikitamikhaylov/fix-arcadia-third

    Try to fix Arcadia

commit 8bc9d8ed7e7140c66d5da9c56c79aaeb357d89f1
Merge: 6fc3ca8b7b 52ef4289fe
Author: tavplubix <avtokmakov@yandex-team.ru>
Date:   Fri Dec 18 01:50:32 2020 +0300

    Merge pull request #18198 from ClickHouse/materialize_mysql_optimize_on_insert

    Enable optimize_on_insert for MaterializeMySQL

commit 6a9397a64bc7c3d71520d515d917a13ceab476f3
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 01:35:37 2020 +0300

    Update FunctionBinaryArithmetic.h

commit a50e9548fd25eb9783844e181eff44cbc215cd17
Merge: 43348bd708 0462340d8a
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 01:26:23 2020 +0300

    Merge branch 'improvement/multiply-decimals-and-floats' of github:myrrc/clickhouse into improvement/multiply-decimals-and-floats

commit 43348bd7087b7f46b36ec60d4890287d4de78144
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 01:25:30 2020 +0300

    fixed the possible values case

commit 0462340d8a6c02a455b254cd12a6beb4c759c3d0
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 01:03:25 2020 +0300

    Update FunctionBinaryArithmetic.h

commit 5a78eafb0547b53df329c9a2b4fb7d3d1d8d7b6b
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 00:59:39 2020 +0300

    Update FunctionBinaryArithmetic.h

commit 40c702c54a4d676fef6ce22b583a4523c287af2f
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 00:56:42 2020 +0300

    Update multiply.cpp

commit 6fc3ca8b7b42879d36a9dfd1756b9eb772128814
Merge: 8140fbd0fb 0c04a263e5
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 00:45:27 2020 +0300

    Merge pull request #16724 from kitaisreal/added-function-accurate-cast-or-null

    Added function accurateCastOrNull, allow different types inside IN subquery

commit 92d8840ae5ae49e95341b9eb39061a455fa6e6d5
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Fri Dec 18 00:31:09 2020 +0300

    Better config: example ports, security advices

commit 8140fbd0fb5b177f8d862306fb9dacd2c6254cb2
Merge: 799997db35 aecdd02be4
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 18 00:22:19 2020 +0300

    Merge pull request #18200 from ClickHouse/more_logs_during_quorum

    More logs during quorum insert

commit fe1b39625664c7397d7c0f18f650a1a605be2f35
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 00:14:06 2020 +0300

    added test reference

commit bf5e836414f68f49d363708c8d32b0a9f0d576cc
Author: myrrc <me@myrrec.space>
Date:   Fri Dec 18 00:03:45 2020 +0300

    updated the test

commit 7db7a9ef5b2727196f037af3e5d4f0834fbdc3cf
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:55:59 2020 +0300

    added another constexpr branch

commit 205db4fa92a3f9086fc154997a9dc283719aefbb
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:50:59 2020 +0300

    fixed the scale bug, added the test (wo reference)

commit b5e6b233eea10c5ba62ea615492f50b7363bf114
Merge: 4f7d019a9e 61d9d213e9
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:20:14 2020 +0300

    Merge branch 'improvement/multiply-decimals-and-floats' of github:myrrc/clickhouse into improvement/multiply-decimals-and-floats

commit 4f7d019a9ebd22b06b3607895a2160fb78c41f4a
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:19:57 2020 +0300

    Squashed commit of the following:

    commit 61d9d213e91f7fbc0381a2ef81b1e941d7fd2506
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 23:14:02 2020 +0300

        fixed another variable init

    commit 73f3195fe17610f814453b27c8f4bbd15a54820d
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 23:12:41 2020 +0300

        fixed variable init

    commit 488f4ef42e90988667f77ec5b6d7e25fef9a3a03
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 23:09:33 2020 +0300

        fixed compile condition

    commit e50af17545827eb9ef657ff7d17c319afb17428e
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 23:05:10 2020 +0300

        fixed template args1

    commit f93385dcf01697a3cd5e77d8ef7f5b69aa8c90ac
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 23:01:29 2020 +0300

        replaced lambda with helper function

    commit 5974e784c29b54c59b1d84a1cecd1049cbc5e5a5
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:52:52 2020 +0300

        added the conversion func

    commit 6eecef35a6b09f1d83f5e1c75db8200d4a438fe5
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:35:46 2020 +0300

        fixed result type trait

    commit 31e3ca16703e06b5ac5cd3bc39631d0dfdaa12cc
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:26:15 2020 +0300

        restored type signature

    commit 613f4e0f04405a9ce5901f36f82935e32c1ca03b
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:24:09 2020 +0300

        removed explicit types

    commit 1c52842e835a661a9fb36693a45ff70a1be96633
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:21:23 2020 +0300

        simplified the condition

    commit c61af32360f95b5cffa8827b77e603fcb7c01740
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:12:12 2020 +0300

        fixed const column branch

    commit fd1d24c928acdcf8d0009ba975209fc2876ebfd0
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:10:33 2020 +0300

        fixed overflow check

    commit b428f8fcd88c8bcc29fe3aac86361df672d5d0e4
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 22:01:04 2020 +0300

        fixed scales

    commit 39c82fe50cb3652ae1f426650634bb89468023f5
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:52:19 2020 +0300

        fixed result type

    commit 54887dacbf2a1243e1dc70dee259787141592ab2
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:38:36 2020 +0300

        fix scale init

    commit 3e9428515c28aa6b9f73fa1a9ff02f7115f913a8
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:36:15 2020 +0300

        fix deleted constuctor

    commit ea2da764f6bd4e1500adaebfc9794879bf414526
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:33:51 2020 +0300

        fixed result data type init

    commit 1f67b0ce8d68693262a6e714d5c9f2f40b74c9a1
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:28:51 2020 +0300

        reverting if constexpr

    commit 494e9066c33c9949d2063f8a63ce1f31120c8805
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:17:38 2020 +0300

        fixing first compile bug

    commit d24fc58c94e270f48834451f4807aba10fe330fc
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:12:24 2020 +0300

        case fix 2

    commit 9f217c5df57e6fe8d8ca5d823d4889e39830ffc1
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 21:11:39 2020 +0300

        case fix

    commit fa6a1d3b4a70f3217df426934b14de4c6a166e6d
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 20:55:35 2020 +0300

        fix

    commit af8a72a45f2162f69c3cc0034af607cc17d288f9
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 20:16:37 2020 +0300

        fix

    commit 737d4fc22fd2565e5551b15f6cd96aa419a271cd
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 19:43:29 2020 +0300

        debug fix

    commit fcacf5c510c1ea34f3980863c92b34a191523f7d
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 19:31:54 2020 +0300

        another try

    commit 49f3b71b5f1c731c98eb0d90421bef968a675e64
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 18:48:22 2020 +0300

        another fix

    commit 0d319c8c612285fc249a6aea9382e9aad3b8e7c4
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 18:47:14 2020 +0300

        fix

    commit 9d5998fd10812ae4cf6b786dbe11fb947b6ec35b
    Author: myrrc <me@myrrec.space>
    Date:   Thu Dec 17 18:46:01 2020 +0300

        added debug info

    commit 17d996054ecd77d59246a26b48ff3f2436c17d3a
    Author: myrrc <me@myrrec.space>
    Date:   Wed Dec 16 18:47:24 2020 +0300

        style fix

    commit 02baaddc13b5cb40f77b44bcc104421030963012
    Author: myrrc <me@myrrec.space>
    Date:   Wed Dec 16 18:20:41 2020 +0300

        removed some casts

commit 799997db35856f1a71329bc6a14cc53348b34d1f
Merge: 30c98adde4 832d99515b
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 23:16:58 2020 +0300

    Merge pull request #16621 from arenadata/ADQM-148

    HDFS configuration and Kerberos Authenticaiton

commit 61d9d213e91f7fbc0381a2ef81b1e941d7fd2506
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:14:02 2020 +0300

    fixed another variable init

commit 30c98adde46e6d242d8bc27e075d68c4177c06f1
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 23:13:49 2020 +0300

    Fix local variable in clickhouse-test

commit 73f3195fe17610f814453b27c8f4bbd15a54820d
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:12:41 2020 +0300

    fixed variable init

commit 488f4ef42e90988667f77ec5b6d7e25fef9a3a03
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:09:33 2020 +0300

    fixed compile condition

commit e50af17545827eb9ef657ff7d17c319afb17428e
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:05:10 2020 +0300

    fixed template args1

commit f93385dcf01697a3cd5e77d8ef7f5b69aa8c90ac
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 23:01:29 2020 +0300

    replaced lambda with helper function

commit b9c537ac37facaf84ef2beafd933c48027c61cac
Merge: d9125a4822 da18af96a3
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Thu Dec 17 22:59:33 2020 +0300

    Merge pull request #18104 from vitlibar/fix-access-rights-for-merge-table-function-2

    Fix access rights required for the merge() table function #2.

commit d9125a4822109d2f60f35ee96222c77c73e00ee8
Merge: dbfe50c0c1 72df75fc50
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Thu Dec 17 22:59:09 2020 +0300

    Merge pull request #18014 from vitlibar/fix-showing-temporary-tables-in-system-tables

    Fix showing temporary tables in system tables

commit 5974e784c29b54c59b1d84a1cecd1049cbc5e5a5
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:52:52 2020 +0300

    added the conversion func

commit dbfe50c0c1593f0bf3820cf7e36107c31c12a167
Merge: 82ed528d98 a207e3db75
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 22:52:15 2020 +0300

    Merge pull request #18199 from ClickHouse/add_one_more_test_in_util

    Add one more case in zk test util

commit 82ed528d980b9b021eafe7ffb94cb7e89f6c0b6f
Merge: f82274a722 fa266790b4
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 22:51:52 2020 +0300

    Merge pull request #18191 from ClickHouse/hard_watches_test

    Test watches for in-memory zookeeper

commit e4433157e7daf43fb57123818f2f88fa3d72a7bc
Merge: c2b0f9d530 f82274a722
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 22:42:32 2020 +0300

    Merge branch 'master' into insert_quorum_parallel-default

commit c2b0f9d5306831fca199dcc9275925dcd496461e
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 22:42:15 2020 +0300

    Update integration test

commit 6eecef35a6b09f1d83f5e1c75db8200d4a438fe5
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:35:46 2020 +0300

    fixed result type trait

commit f82274a72297e863e38dc94d636fe2d7bd5f5d57
Merge: e9d97160f4 fd7fe0f84c
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 22:34:05 2020 +0300

    Merge pull request #16527 from kitaisreal/mmap-read-buffer-from-file-descriptor-use-getpagesize

    MMapReadBufferFromFileDescriptor use getpagesize

commit e9d97160f4431e5b460ca77c1851146d24981791
Merge: dd78c9dc75 f7f8ff8758
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 22:28:22 2020 +0300

    Merge pull request #16346 from ClickHouse/sigaltstack

    Provide diagnostics on stack overflow

commit 31e3ca16703e06b5ac5cd3bc39631d0dfdaa12cc
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:26:15 2020 +0300

    restored type signature

commit 613f4e0f04405a9ce5901f36f82935e32c1ca03b
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:24:09 2020 +0300

    removed explicit types

commit 1c52842e835a661a9fb36693a45ff70a1be96633
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:21:23 2020 +0300

    simplified the condition

commit f7f8ff8758615dcdc923ce45ec327e7f2b1cd29c
Merge: 01d3aef890 dd78c9dc75
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 22:16:30 2020 +0300

    Merge branch 'master' into sigaltstack

commit c61af32360f95b5cffa8827b77e603fcb7c01740
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:12:12 2020 +0300

    fixed const column branch

commit fd1d24c928acdcf8d0009ba975209fc2876ebfd0
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:10:33 2020 +0300

    fixed overflow check

commit 0c04a263e539ce530bc7a4ea98f57b2c59f35919
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Dec 17 22:04:26 2020 +0300

    Fixed ya.make

commit b428f8fcd88c8bcc29fe3aac86361df672d5d0e4
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 22:01:04 2020 +0300

    fixed scales

commit 39c82fe50cb3652ae1f426650634bb89468023f5
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:52:19 2020 +0300

    fixed result type

commit 54887dacbf2a1243e1dc70dee259787141592ab2
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:38:36 2020 +0300

    fix scale init

commit 3e9428515c28aa6b9f73fa1a9ff02f7115f913a8
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:36:15 2020 +0300

    fix deleted constuctor

commit ea2da764f6bd4e1500adaebfc9794879bf414526
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:33:51 2020 +0300

    fixed result data type init

commit 9aaa2f934580122bb3dc84950f6ca84ffa65327b
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Dec 17 21:32:25 2020 +0300

    Fixed compilation timeout

commit 1f67b0ce8d68693262a6e714d5c9f2f40b74c9a1
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:28:51 2020 +0300

    reverting if constexpr

commit fb3d235e8473d70500bf0e11788a9ec782466f37
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 21:08:42 2020 +0300

    Do not use strerror

commit 494e9066c33c9949d2063f8a63ce1f31120c8805
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:17:38 2020 +0300

    fixing first compile bug

commit d24fc58c94e270f48834451f4807aba10fe330fc
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:12:24 2020 +0300

    case fix 2

commit 9f217c5df57e6fe8d8ca5d823d4889e39830ffc1
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 21:11:39 2020 +0300

    case fix

commit fa6a1d3b4a70f3217df426934b14de4c6a166e6d
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 20:55:35 2020 +0300

    fix

commit dd78c9dc75ad684b4cad458a5e87f3e28d3ba775
Merge: a77acd5aa7 b25f126569
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 20:44:44 2020 +0300

    Merge pull request #17566 from ClickHouse/relax-too-many-parts-1

    Relax "Too many parts" threshold, part 1

commit 4a1618d0b80b6ec1d692a42a0a96ee92bbf26dbd
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 20:25:51 2020 +0300

    Update libc headers

commit af8a72a45f2162f69c3cc0034af607cc17d288f9
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 20:16:37 2020 +0300

    fix

commit 737d4fc22fd2565e5551b15f6cd96aa419a271cd
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 19:43:29 2020 +0300

    debug fix

commit fcacf5c510c1ea34f3980863c92b34a191523f7d
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 19:31:54 2020 +0300

    another try

commit a207e3db7555b2984275a55a2116af48ece3af48
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 19:21:46 2020 +0300

    Fix bad sleep

commit aecdd02be4880b262fa7d4e1cf2e6b28027c4e59
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 19:14:42 2020 +0300

    Better message

commit f45993cd1e4bd651b25ce73cfab1d75b775690ed
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 19:13:01 2020 +0300

    More logs during quorum insert

commit 49f3b71b5f1c731c98eb0d90421bef968a675e64
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 18:48:22 2020 +0300

    another fix

commit 0d319c8c612285fc249a6aea9382e9aad3b8e7c4
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 18:47:14 2020 +0300

    fix

commit d198f033017dc5da327855eba62acb2657bd69cb
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 18:46:26 2020 +0300

    suppress warnings

commit 9d5998fd10812ae4cf6b786dbe11fb947b6ec35b
Author: myrrc <me@myrrec.space>
Date:   Thu Dec 17 18:46:01 2020 +0300

    added debug info

commit a77acd5aa753b154e475b7c6706274fbae9db8df
Author: Anton Popov <pad11rus@gmail.com>
Date:   Thu Dec 17 18:12:24 2020 +0300

    Fix flaky test

commit e80b93939f451202e579624602ef3ec8ac596a07
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 18:09:09 2020 +0300

    Add one more case in zk test uti

commit cb91d64fe7a9ba589faf52a384da9f0c10392953
Author: hexiaoting <hewenting_ict@163.com>
Date:   Thu Dec 17 22:16:14 2020 +0800

    Add new feature: limit length and offset setting (#17633)

    Co-authored-by: Alexander Kazakov <Akazz@users.noreply.github.com>

commit 52ef4289feca44bdd75a188c6fa558842ce6203a
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Thu Dec 17 16:05:38 2020 +0300

    enable optimize_on_insert for MaterializeMySQL

commit 0f3582883c3b1b75ef96dda9ceed5c4b2e7897d1
Merge: d302685e72 8f49946a72
Author: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 15:44:55 2020 +0300

    Merge pull request #18147 from kitaisreal/array-aggregation-big-integer-types-support

    Added array aggregation big integer types support

commit d302685e726b6960b31f1df29d4f01cd5656b907
Merge: 7e73b6acd8 3e07094364
Author: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 15:12:05 2020 +0300

    Merge pull request #18172 from nikitamikhaylov/julian-day-msan

    toModifiedJulianDayOrNull msan fix

commit 3e070943644630c3297d67c878b39b5e000c60a9
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 15:11:32 2020 +0300

    style

commit fa266790b42671f9b9f55354cf72c7a76498c534
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 14:29:29 2020 +0300

    Better test

commit 7e73b6acd874e47e3724716a8707a2b0386c70bf
Merge: d82c23d5cf 294e8bbb4e
Author: tavplubix <avtokmakov@yandex-team.ru>
Date:   Thu Dec 17 14:26:51 2020 +0300

    Merge pull request #17642 from filimonov/detach-permanently

    DETACH TABLE PERMANENTLY

commit b6bd84df844e3c0760847fcbe040a72eb284d61c
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 14:25:59 2020 +0300

    One more test for watches

commit d82c23d5cf8ab49905212b0699bd43a2668fbd81
Merge: 10449695bc c6ac831d4e
Author: Kruglov Pavel <48961922+Avogar@users.noreply.github.com>
Date:   Thu Dec 17 14:00:05 2020 +0300

    Merge pull request #16853 from amosbird/ss

    Add -SimpleState combinator

commit 911a1cd48f66ba270d91b1fa8b1e3abb5b435b03
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 13:25:31 2020 +0300

    Suppress gdb warning

commit da18af96a33765a74cf5f762ff5d2a3623e0510b
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Thu Dec 10 23:16:53 2020 +0300

    Fix access rights required for the merge() table function.

commit fd7fe0f84cc53de082c530c9e133e06dd6dccff7
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Dec 17 12:10:46 2020 +0300

    Updated ya.make

commit 832d99515b4f4d68a3121579dbdeec153900af51
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Thu Dec 17 12:05:49 2020 +0300

    tag for yandex/clickhouse-kerberized-hadoop

commit 72df75fc50877eee367ece609c602a9af518374f
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Fri Dec 11 23:48:27 2020 +0300

    Add tests.

commit af3ccb49f1e917d0f703dc9ba7102c2a6a18e29e
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Fri Dec 11 22:44:18 2020 +0300

    Show create query for temporary tables in system.tables

commit ede755d3bced701606b048ae7d4983c1451a3f69
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Fri Dec 11 23:32:57 2020 +0300

    Show temporary tables in system.columns as tables with empty database.

commit d7a3cc8c909a6d35737968ea16b3fad2ee92190b
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Fri Dec 11 22:42:14 2020 +0300

    Don't show the "_temporary_and_external_tables" database in system tables
    system.databases, system.tables, system.columns.

commit b74e7accf15c7fe3880827ce9103a650da3c942c
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 11:17:39 2020 +0300

    Review fixes

commit f2957d6f0545ce4aeb023bb7b71a0a49d2e888ad
Merge: 1bd2f49fb5 10449695bc
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 10:40:25 2020 +0300

    Merge branch 'master' into try_rewrite_writer

commit 10449695bc0f9dbed35b1e00303f86867d4861ce
Author: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date:   Thu Dec 17 10:32:11 2020 +0300

    Update boost.

commit c91c5d0a686c5bfbb883318428b9c0d2c3091985
Author: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date:   Thu Dec 17 10:29:47 2020 +0300

    Update boost.

commit 66228f405456b74488fdb54d2fb79af9e26af567
Merge: bc0186eede 1f3702ff9e
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 17 10:15:30 2020 +0300

    Merge pull request #18065 from ClickHouse/fix_watch_race_testkeeper

    Remove dead list watches from TestKeeperStorage

commit bc0186eede039b4c6ebe407fd427806f5bd6b28d
Merge: 81c00a4e94 6c8f844442
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 07:19:04 2020 +0300

    Merge pull request #18168 from ClickHouse/unknown_function_better_exception

    Better exception message for unknown function

commit 81c00a4e94c2fc3017263aa0f6e0d4cddf219b2c
Merge: b159e15b1b 76e73e4e8b
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 07:18:53 2020 +0300

    Merge pull request #18167 from ClickHouse/fix_checking_profile_settings_constraints

    Fix `Unknown setting profile` error on attempt to set settings profile

commit 01d3aef89096564bf399efc8b8616fe7b3a9a48d
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 05:53:43 2020 +0300

    Disable trap back

commit b159e15b1b544de418344d78fe15cc856f902b4a
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 05:50:04 2020 +0300

    Update easy_tasks_sorted_ru.md

commit 22363f6086a31f2cf7ea397b81f424a395ce23ce
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 05:29:27 2020 +0300

    better

commit db362e4585ae9de6c189fc8a3ea2bca7dcb5c192
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 05:20:56 2020 +0300

    done

commit 4e8978e02222a2810a3e698f4203f2d12931c08e
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 05:21:27 2020 +0300

    Merge with master

commit 7f280bef37226f9d8dbd2873bde368206aecfb7d
Merge: 2e232bad74 16e524e0de
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 05:09:10 2020 +0300

    Merge branch 'master' into sigaltstack

commit 7abc58526ab64ff98f13d1906c36547bdaf04da9
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 04:55:07 2020 +0300

    Update test

commit 6a7c73a7bdc931d1a7f98fe543efab0970ca4dae
Merge: f46eb6f2be 16e524e0de
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 04:54:17 2020 +0300

    Merge branch 'master' into insert_quorum_parallel-default

commit b25f126569e4870cdfcc6cba4858122ba5edd6b7
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 04:52:27 2020 +0300

    Added a test

commit 9a017b77df614a2618f9ffcd706b77f970ce582d
Merge: 2e6bedce91 16e524e0de
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 04:40:39 2020 +0300

    Merge branch 'master' into relax-too-many-parts-1

commit d2653f91e1bbaec641c832a9764948abf3d24cf7
Merge: 792d8981ab 4edbb42ba1
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 03:40:13 2020 +0300

    Merge pull request #18112 from GrigoryPervakov/join_set_s3

    Use IDisk in Set and Join storages

commit 792d8981ab74a58f2a909a143276ed6ef2ed8dac
Merge: 37da513c36 60b916fe0b
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 03:36:07 2020 +0300

    Merge pull request #18115 from AnaUvarova/ana-uvarova-DOCSUP-2030-rus-faq-order

    DOCSUP-2030 Rus FAQ order

commit 37da513c3619d14ea87c307beae3179ce0f2bc38
Merge: 484f20f77d d96ca3fb8d
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 03:35:19 2020 +0300

    Merge pull request #18136 from taiyang-li/explain_ast_nonselect_query

    Explain ast support nonselect querys

commit 484f20f77d0eb6d96e312f434b4a63f8ef16d909
Merge: 908784a0d9 849db47f8a
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 03:34:39 2020 +0300

    Merge pull request #18140 from ClickHouse/fix_flaky_test

    Trying to fix 00620_optimize_on_nonleader_test

commit 908784a0d982e3e09bab3876498d3047222f2f76
Merge: 2e2cc6d344 8139bab240
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 03:32:10 2020 +0300

    Merge pull request #18148 from ClickHouse/suppress_error_in_test

    Suppress error in 00993_system_parts_race_condition_drop_zookeeper

commit 71bd7e55d30d6aabf9cdf7607b828d60b8d82ae9
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 03:15:13 2020 +0300

    Bump

commit 2e2cc6d344d4615b691c4c330bb5b7ff909e1ab9
Merge: a999850378 2413e82e96
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Thu Dec 17 01:19:08 2020 +0300

    Merge pull request #18141 from staticor/patch-5

    Update misc.md

commit d70f58a98f8cf1b325823dd6c366120ce77090d8
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Thu Dec 17 00:52:04 2020 +0300

    remove ryu

commit 6c8f844442102da23fb7e4cd5b7fdddb92f8cef9
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Thu Dec 17 00:42:14 2020 +0300

    better exception message for unknown function

commit 685099af7f530b75772cee40d9c3c25579a30eec
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Dec 17 00:23:41 2020 +0300

    Move getPageSize in common

commit 8f49946a72b2bc9766896c732f34dce5b10db5aa
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Dec 17 00:21:44 2020 +0300

    Reused Result type

commit 4b32bb323ac71948728892e275a83d0f7e55b5cc
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Dec 17 00:04:03 2020 +0300

    Fixed style issues

commit 76e73e4e8b466706b0ddc3ff337b4427e0c569ff
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Wed Dec 16 23:49:56 2020 +0300

    fix 'Unknown setting profile' error

commit 5a096a75b8f4e903970d5078447680e0c2b43aef
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 16 23:10:40 2020 +0300

    Fixed issues with average

commit 1f3702ff9eac498dec5472680e1d75ab423cf502
Merge: 422467628a a999850378
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 23:03:07 2020 +0300

    Merge branch 'master' into fix_watch_race_testkeeper

commit a999850378b01ffe6bc6378df4002ace98bae89a
Merge: 521e3fc4ae dbbe761fa1
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 22:05:16 2020 +0300

    Merge pull request #18130 from ClickHouse/fix_remove_column_ttl

    Fix remove TTL for column

commit 521e3fc4ae26dd56cda52bd52bef7fed9a9c9df4
Merge: 170cd132da 153610b2c3
Author: Ilya Yatsishin <2159081+qoega@users.noreply.github.com>
Date:   Wed Dec 16 21:43:32 2020 +0300

    Merge pull request #17910 from sevirov/sevirov-DOCSUP-4919-document_replication_queue_system_table

    DOCSUP-4919: Document the system.replication_queue system table

commit 60b916fe0b72509a0151eb5fd2c0556db34a0257
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Wed Dec 16 21:24:06 2020 +0300

    polishing links

commit 23270915d37a05491cfae7ed5b170501349524fb
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 21:03:39 2020 +0300

    better

commit 170cd132da9350d30cac224a937098b5cbf0e2b1
Merge: b8250882ac 95c35b318e
Author: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
Date:   Wed Dec 16 20:49:43 2020 +0300

    Merge pull request #18113 from azat/tld-list-symlink-fix

    Avoid using symlinks for top_level_domains

commit b8250882ac371cff376cf0b3d46126510ad1ee3c
Merge: 16e524e0de a7933ddf31
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 20:35:18 2020 +0300

    Merge pull request #18056 from zhangjmruc/master

    Fix issue #17160 by adding show settings like '...'

commit 294e8bbb4e74532af994d0118217ee5c4e4e18f0
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Wed Dec 16 18:16:47 2020 +0100

    Fix2

commit d96ca3fb8da73ad973c7d701a0e0250b89991d35
Author: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date:   Wed Dec 16 20:11:32 2020 +0300

    Fixing build.

commit 16e524e0de74c0be2a6ebfe7b94e629f0e3502d0
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 19:58:00 2020 +0300

    Remove outdated info from the docs

commit 422467628a1d3972d7b43117746c26e645cbf528
Merge: b347fa322c d873439846
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 19:45:57 2020 +0300

    Merge branch 'master' into fix_watch_race_testkeeper

commit d873439846b576cf884df01a9b23e513a918568e
Merge: d30099fd78 9ad2a76d51
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 19:29:32 2020 +0300

    Merge pull request #15236 from ClickHouse/run_func_tests_in_parallel

    Collect list of non-parallel test

commit 8139bab240d2a94095707254d6a62e18910597ab
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 19:12:57 2020 +0300

    More relaxed check

commit 9cc6fec36cf9b704858bfdf6dc06794a6813bda7
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 19:10:46 2020 +0300

    Supress error in 00993_system_parts_race_condition_drop_zookeeper

commit 80dbbc7e58b372f7455419b726e558a63177e936
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 16 18:59:53 2020 +0300

    Style check fix

commit b82770c2ad8ef58b89aeea2b4ac4e22e2305e07c
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 16 18:47:30 2020 +0300

    Added array aggregation big integer types support

commit 17d996054ecd77d59246a26b48ff3f2436c17d3a
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 18:47:24 2020 +0300

    style fix

commit 50c29d9ed007e744718725f8a1bfe94037723725
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 18:34:29 2020 +0300

    better

commit d30099fd78164498432cfcf3c2917d506d4a4f2c
Merge: d046164049 44d2194326
Author: Nikolai Kochetov <KochetovNicolai@users.noreply.github.com>
Date:   Wed Dec 16 18:23:45 2020 +0300

    Merge pull request #18132 from ClickHouse/try-fix-mysql-integration-test

    Try fix integration tests.

commit d04616404937d55915faf7e38599fa3f31400776
Merge: b1dc8074ef 69ed3d1cb5
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 18:22:38 2020 +0300

    Merge pull request #18128 from ClickHouse/boringssl-krb5-2

    Port Kerberos to BoringSSL

commit 02baaddc13b5cb40f77b44bcc104421030963012
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 18:20:41 2020 +0300

    removed some casts

commit f143a37693027395c873eab0e53085569508c5e1
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 18:11:26 2020 +0300

    simplified case to multiply only

commit f14cafee51ad974d7de3f3ab775345310976c99d
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Wed Dec 16 16:06:43 2020 +0100

    Fix tests

commit d7145a5ee15fa81b64c3204a75e66d40ed70267b
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 18:01:18 2020 +0300

    working

commit 484e6e7de4d1d1d4f95002bd4193ae4925a4fdcf
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 16 17:38:17 2020 +0300

    Removed getPageSize from small files

commit 849db47f8adf9423e14a79e414f8a9b05c24088b
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 17:31:17 2020 +0300

    Better exception messages

commit b1dc8074efbc000c1509d9f2210a5343fabb77ce
Merge: d99792e614 45361f752c
Author: Anton Popov <pad11rus@gmail.com>
Date:   Wed Dec 16 16:58:02 2020 +0300

    Merge pull request #17829 from CurtizJ/merging-map

    Merging data type Map #15806

commit 74ee36bdb8f40bc306323c139b6d0119839fe61b
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Wed Dec 16 16:39:14 2020 +0300

    done

commit 2413e82e9639faf90e40ad7e9f5e757db20326ae
Author: Steve-金勇 <staticor@me.com>
Date:   Wed Dec 16 21:35:55 2020 +0800

    Update misc.md

    Update translation by machine, and improve document more human-readable.

commit c15f5fba3b2547b9c0a8d6b6c8eff6692e53738c
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 16:16:24 2020 +0300

    Fix 00620_optimize_on_nonleader_test

commit b347fa322c1c19ddc3db53be4976ce13d4ceea2f
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 15:58:00 2020 +0300

    Trying to fix races one more time

commit 5521abba1e18497c47e4d1626ce4c7a4c66b9898
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 15:57:42 2020 +0300

    fix boost

commit c906b1ca769679af8bf8f0ed1db8fd80877fa416
Author: myrrc <me@myrrec.space>
Date:   Wed Dec 16 15:21:23 2020 +0300

    added the initial solution

commit 4edbb42ba14ad7fd9bf757c1bec946f8b4f72e08
Author: Pervakov Grigorii <pervakovg@yandex-team.ru>
Date:   Tue Dec 15 19:45:13 2020 +0300

    Use IDisk in Set and Join storages

commit 0a2bda002d7d94ed2931aafb1d17a6a72a5f3d44
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 14:46:58 2020 +0300

    remove last commit

commit 61c1db38234f5b84bed5815d8cf6b7890863b6dc
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 14:45:40 2020 +0300

    skip test

commit bcd6fc1c88f334674c97a4f1fbaa7c8877893041
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 02:53:25 2020 +0300

    better

commit 18d52dbc63b40784e70e57ce980b9252c37b932d
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Wed Dec 16 02:30:35 2020 +0300

    better

commit 20aaf48e38db2ba399ebb61aa004fcad4ffdb7bd
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Tue Dec 15 22:56:47 2020 +0300

    pragmas

commit 5c4e1d1db54d81acdbfa4843f81c3ffbf8ecae52
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Tue Dec 15 20:53:10 2020 +0300

    better

commit 3288f1fd0cb3bbf7b316deb929b0e797d0173b4b
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Tue Dec 15 16:00:40 2020 +0300

    ya make

commit c3a6499f6df3e50178c301e2627147f2ec4cc723
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Tue Dec 15 14:48:24 2020 +0300

    fix ya.make

commit 4ccdb3ca208a90963464b911b371aab0ed47c53b
Author: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date:   Tue Dec 15 06:07:43 2020 +0300

    done

commit a5a33f1963aa0ab18e7784f2891d692171f3e023
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Tue Dec 15 14:52:10 2020 +0300

    rpm rebuilddb and yum update in kerberized_hadoop/Dockerfile

commit 39877b1a6fda55fede6aa69c59cecb4496cdd539
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Tue Dec 15 12:46:02 2020 +0300

    yum clean all in kerberized_hadoop/Dockerfile

commit a4dad0acb7f326e526a91a81bad4ef6072fc10ff
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Mon Dec 14 21:35:13 2020 +0300

    CentOS 6 became too old

commit 09d1e896741dfd57d1ad5d56805742b8b98d922a
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Mon Dec 14 15:07:05 2020 +0300

    yandex/clickhouse-kerberized-hadoop + missed header

commit 407565b47f77bc8265bf3fa43ddfe365605d1c2d
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Fri Dec 11 12:16:58 2020 +0300

    fix - wrong error code

commit 9701d6a2d5fd1db05449f95fc5d2a12573ca488f
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Fri Dec 11 00:52:05 2020 +0300

    style fix per code review, doc improvement, params consistency check

commit 815856bc3d24bf553cadfce5333710d808066fc5
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Sat Nov 28 01:17:05 2020 +0300

    pass config, not context to createHDFSBuilder

commit fc0a1af2141199af5315cf7453bda16dc089d45a
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Thu Nov 19 00:08:17 2020 +0300

    doc minor changes, cleanup, krb5-user as a recommended package

commit 7e9781485900af0240af7eabf69d7f2b537d4a78
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Fri Oct 30 22:40:16 2020 +0300

    cleanup, fixes, new submodules, ShellCommand, WriteBufferFromString

commit d1d657335b62bedaf4dba7075a35131aafd1df82
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Thu Oct 29 23:40:47 2020 +0300

    USE_INTERNAL_HDFS3_LIBRARY + bash lint

commit d9166a0645981bbf6c79f109b2916ae86eea0e8a
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Mon Sep 28 20:20:04 2020 +0300

    HADOOP_SECURE_DN_USER way, kinit thread, docker capabilities

commit eb10948ca2c928f8b10a3676162210722962db2d
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Thu Sep 10 13:02:46 2020 +0300

    kerberized_hdfs cluster started

commit 185ac744b1f67cfde0d7151556cc9309c1b98b46
Author: Ilya Golshtein <gie@arenadata.io>
Date:   Wed Sep 9 15:13:20 2020 +0300

    kerberized hdfs compiled

commit 7d23451e8126bb9cf19f05b56b2325a76fc19b14
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Oct 31 00:24:16 2020 +0300

    Replaced usages of 4096 constant with getPageSize

commit dbb2fbcdd54197c7689f4a5670b2834d8c24bb17
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Oct 29 22:52:12 2020 +0300

    Unified usages of getPageSize

    1. Introduced getPageSize function.
    2. Replaced usages of getpagesize with getPageSize function.
    3. Replaced usages of sysconf(_SG_PAGESIZE) with getPageSize function.

commit 058fbd52223af7c78dac4468d5d964cc865bef50
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Oct 29 21:49:44 2020 +0300

    MMapReadBufferFromFileDescriptor use getpagesize

commit dbbe761fa10d088eb73c962e727b88f1006ace3a
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 13:23:11 2020 +0300

    Fix test

commit cd7827f70c65b7c9042b9a722aed77b4ee3dd50b
Author: taiyang-li <taiyang_li@163.com>
Date:   Wed Dec 16 18:00:21 2020 +0800

    explain ast support nonselect queries

commit ea4d11cb73bef08653858207e0be5e84cde69ddf
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 13:04:46 2020 +0300

    Wait for connections to special servers

commit 44d2194326804a36163a25c6e967b27a96cf279c
Author: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date:   Wed Dec 16 11:52:04 2020 +0300

    Try fix integration tests.

commit 54aeb8fcff0fb50365748f33a2129aee75ee6789
Merge: 2a4a6b22b1 d99792e614
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 11:25:19 2020 +0300

    Merge branch 'master' into fix_remove_column_ttl

commit d99792e61404ff0ee805f7a999e3c433482476f3
Merge: 37363ae89d 86fc9ea4df
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 10:54:58 2020 +0300

    Merge pull request #18105 from ClickHouse/fix_docker_images

    Fix dependencies for docker stateful_with_coverage

commit 2a4a6b22b19638d18bf51e7b83c8f0a47fd4c2fb
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 16 10:26:18 2020 +0300

    Fix remove TTL for column

commit 37363ae89dd70abb8e23b6bf52bdc248c9f8bbba
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 09:43:05 2020 +0300

    Comment update

commit 69ed3d1cb5ba5abe7dd23f1901aa0b905f4547e5
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 09:40:30 2020 +0300

    Update submodule URL

commit e3b490e9b305c2e6b16bf37271d2f13c3ca8d1d7
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 09:36:26 2020 +0300

    Port Kerberos to BoringSSL

commit 42b401c89390e21537db28398d99c319c1fe0e80
Merge: 4bd5d3b662 cf35e144ff
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 09:00:38 2020 +0300

    Merge pull request #18127 from chienlungcheung/patch-1

    Correct the unit from Kb to KB in the last section.

commit cf35e144ff65c8e7ea16579dcb6451f46224214a
Author: Chienlung Cheung <chienlungcheung@gmail.com>
Date:   Wed Dec 16 13:55:17 2020 +0800

    correct the unit from Kb to KB in the last section.

commit 4bd5d3b662f642d38158482a285366f318711625
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 06:36:43 2020 +0300

    Comment update

commit 62ce33153a45bc3355555cc6c3a75920425fa4b2
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 06:34:25 2020 +0300

    Added valuable comment

commit 88c5031dfa14f7805f0aa3a8e3570efc3cac4da7
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 06:27:28 2020 +0300

    Add a commit #16595

commit ca9ccf4850208d617085fa150b69b5ef35633925
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 06:03:43 2020 +0300

    Comment update

commit a7933ddf31f26719475cb3bb10d7c8ab162cde1d
Merge: b56523fd94 eed767bbab
Author: jianmei zhang <zhangjmruc@sina.com>
Date:   Wed Dec 16 09:38:38 2020 +0800

    Merge and rebase with master

commit 45361f752ca591c079888483341c1d1798b3a85f
Author: Anton Popov <pad11rus@gmail.com>
Date:   Wed Dec 16 02:36:03 2020 +0300

    fix test

commit b485b5bd1cc7edf8a4b0a76599f040e42dc6d954
Author: Anton Popov <pad11rus@gmail.com>
Date:   Wed Dec 16 01:22:21 2020 +0300

    minor fixes and comments added

commit aec820362a535f30b9ab04f84117ab27e10cea91
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 00:28:53 2020 +0300

    Update DecimalFunctions.h

commit eed767bbab573f6f72d69f93ddc2c577539bfab2
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Wed Dec 16 00:02:44 2020 +0300

    Update tips.md

commit b842cb04551560ba3bb5bb0c31bc4cb2f2f271c0
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 23:38:36 2020 +0300

    one more try

commit 4e3cc606e64abba6700dbaea7d5078ab73f0d3c0
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Tue Dec 15 23:36:45 2020 +0300

    Fixed tests

commit 97c93fa984a4cab53620c77119c18d2e7fbe8128
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 23:23:58 2020 +0300

    one more try

commit 86fc9ea4df3efaaef44b48d1504ef6a9599bec34
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 23:18:53 2020 +0300

    Update base image to 20.04

commit 0887cf476d4e0534c648479ff1f2423aec975326
Merge: b0d27d6747 2e4aa6b9e1
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 23:09:37 2020 +0300

    Merge pull request #18098 from ClickHouse/fix_clickhouse_test

    Fix database drop timeout in clickhouse-test

commit 93b58321d0b4212cd5a21d7466901458e45d1956
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 23:00:30 2020 +0300

    fixes

commit acfc3d4281be996520acaf0900ab9688bc502e47
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 22:47:24 2020 +0300

    host fixes

commit 8aaef7f634ec57e744f2d8a37458169052287b7d
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 22:34:11 2020 +0300

    host fixes

commit 99ecb8ae6831a95dfe69e7598f5a6e1c71b5f9f5
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 22:02:37 2020 +0300

    host fixes

commit b0d27d6747b32b8edb0c869e14a34615661761c6
Merge: f1e1672b54 3ec48cec53
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Tue Dec 15 22:01:57 2020 +0300

    Merge pull request #18082 from den-crane/patch-3

    Doc. toYYYYMM  translation to Russian

commit f1e1672b54045893ecc544d827e708b108d91740
Merge: 01286a6a09 96f22a5bcc
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Tue Dec 15 22:01:08 2020 +0300

    Merge pull request #18094 from staticor/patch-4

    Update limit.md

commit 95c35b318e55362c9a18d2ea723092ef9fc1f02f
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Tue Dec 15 21:48:24 2020 +0300

    Avoid using symlinks for top_level_domains

    Move the main copy into the tests/config and ship it to the perf tests
    via the perf package.

commit 51f53ce8e9cc4c916c4266d8a82857fb319663c3
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Tue Dec 15 21:54:16 2020 +0300

    Added tryParseDecimal functions

commit e51cf03f21c7abd1f1c352cc23414f14d296dc79
Merge: ba63635251 01286a6a09
Author: Anton Popov <pad11rus@gmail.com>
Date:   Tue Dec 15 21:08:01 2020 +0300

    Merge remote-tracking branch 'upstream/master' into HEAD

commit ba6363525190c432262239f68fb5bb532ca3e4fb
Author: Anton Popov <pad11rus@gmail.com>
Date:   Tue Dec 15 21:07:13 2020 +0300

    add tests for formats with type map

commit 9df7ecb8e640b22977f177a4a6af16936304e746
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Tue Dec 15 18:06:23 2020 +0100

    fixes

commit 2e4aa6b9e113afdac58d4b5b41069b16ec6fb33e
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 19:20:09 2020 +0300

    Kill not only database proc

commit 01286a6a095d9abe6b77c0d61514d1a979be746a
Author: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date:   Tue Dec 15 19:18:28 2020 +0300

    Update libunwind.

commit e10ee944bed26fc67a14e57fc6d025dd96794f32
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 19:17:42 2020 +0300

    links fixes

commit a325c0db744374610543ae7f8b79f2a6d14f796c
Author: ana-uvarova <ana-uvarova@yandex-team.ru>
Date:   Tue Dec 15 19:07:25 2020 +0300

    base

commit 1bd2f49fb5f699e62f1aab311725163c16e5e5dc
Merge: 8670836573 063488d943
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 19:05:04 2020 +0300

    Merge branch 'master' into try_rewrite_writer

commit dff71850a8d39ac42af7d877f75fcdae1da08402
Merge: 9d17f01dc9 063488d943
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 19:04:24 2020 +0300

    Merge branch 'master' into fix_watch_race_testkeeper

commit 9ad2a76d5151c613d281deaf7d66ebc895506c06
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 18:27:47 2020 +0300

    Update run.sh

commit 554fd738e9a783f1b7f69cae7b2e2a7d3f796a05
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 18:22:21 2020 +0300

    Remove strange line

commit ccabb4680ddad1bfbbdd2550353d890136ade9da
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 18:19:56 2020 +0300

    Fix dependencies

commit 063488d94350bd7f8c0c41dc0b8009f48d7e9be8
Merge: 31bef163ba 22aba554c0
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Tue Dec 15 18:07:42 2020 +0300

    Merge pull request #18103 from ClickHouse/revert-17983-fix-access-rights-for-merge-table-function

    Revert "Fix access rights required for the merge() table function."

commit 22aba554c0fa42a8664a944e266fd69d7189e55f
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Tue Dec 15 18:07:20 2020 +0300

    Revert "Fix access rights required for the merge() table function."

commit 27ff3a5214ab639712a48279c2afa6dd594dc837
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Tue Dec 15 15:46:13 2020 +0100

    Rework using a separate flag file instead of renames

commit 31bef163ba4c37b48f59d56423aea2ef80d46da5
Merge: b252321ef7 c3a4e5e7b7
Author: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date:   Tue Dec 15 17:34:38 2020 +0300

    Merge pull request #17848 from azat/max_distributed_connections-fix

    Fix max_distributed_connections

commit dd09e1d78396775f4f04750f4404a1ee53680a7d
Author: Anton Popov <pad11rus@gmail.com>
Date:   Tue Dec 15 17:34:37 2020 +0300

    fix type map ser/de to json

commit b252321ef7ea531351a29ca93c5851ce718ea260
Merge: def080c8ba 6defcbb662
Author: Nikolai Kochetov <KochetovNicolai@users.noreply.github.com>
Date:   Tue Dec 15 16:51:54 2020 +0300

    Merge pull request #16954 from Avogar/optimize-data-on-insert

    Optimize data on INSERT

commit 6defcbb66215d775e4b736c646dcacbf693c0d83
Merge: c8c543ca94 def080c8ba
Author: Nikolai Kochetov <KochetovNicolai@users.noreply.github.com>
Date:   Tue Dec 15 16:50:42 2020 +0300

    Merge branch 'master' into optimize-data-on-insert

commit 873d9d1e1e22dd6a6f09c6919bfa1cf68cc4a55b
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 16:34:53 2020 +0300

    Better

commit 9bb0326fc6bac0922b5a94f03fe4102745bb1e41
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 16:33:14 2020 +0300

    Fix database drop timeout in clickhouse-test

commit 96f22a5bcc8aef0159805337319d767c54208243
Author: Steve-金勇 <staticor@me.com>
Date:   Tue Dec 15 18:55:52 2020 +0800

    Update limit.md

    Update translation of LIMIT , and providing more details for with ties modifier.

    Make document more human-readable for Chinese.

commit 9d17f01dc9b5c7703cea8bbf0b2cd2a4679e143a
Merge: 73780838a1 def080c8ba
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 13:41:04 2020 +0300

    Merge branch 'master' into fix_watch_race_testkeeper

commit 8670836573d09282d29fe75ae9e9c43eb234bdff
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 13:34:28 2020 +0300

    Add some comments

commit b56523fd94db0cdcde4809f247424d53fcfd2f1f
Author: jianmei zhang <zhangjmruc@sina.com>
Date:   Tue Dec 15 17:56:54 2020 +0800

    Update test case to avoid the latest added parameter

commit 74c2211510564bcc19667d31c9f0e873b5d8a6c5
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 12:54:48 2020 +0300

    More clear

commit def080c8bae1fb60a105e38527e1452e83f207f2
Author: Alexander Kazakov <Akazz@users.noreply.github.com>
Date:   Tue Dec 15 11:16:25 2020 +0300

    Fixed unsafe ast rewrite (#18064)

    Fixed unsafe assignment to a reference in InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery()

commit a50187c9954c05fc7b70aa0a7b8d323f1fa4057c
Author: alesapin <alesapin@gmail.com>
Date:   Tue Dec 15 11:04:37 2020 +0300

    Slightly better

commit 508b50d121fc03c26fa9898fadc16ebb9b76f10d
Author: jianmei zhang <zhangjmruc@sina.com>
Date:   Tue Dec 15 14:44:39 2020 +0800

    support SHOW CHANGED SETTINGS query and add test cases, also do code changes to compile SHOW SETTINGS PROFILES.

commit 5a8c750b0bda0e7255bf698d2989a1b1750955cd
Merge: 45b9d79f81 9310437920
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Tue Dec 15 09:17:52 2020 +0300

    Merge pull request #17983 from vitlibar/fix-access-rights-for-merge-table-function

    Fix access rights required for the merge() table function.

commit 45b9d79f819820e30c46f6f5ed306f58cfe09afe
Merge: a0405182ed 8de0c0e8fc
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Tue Dec 15 07:51:14 2020 +0300

    Merge pull request #18081 from azat/gcc10-sanitizers

    Do not pass -fsanitize-blacklist for gcc (it does not support it) for UBSAN

commit 3ec48cec53f32374df82825b5534f2131e2018ec
Author: Denny Crane <deniszhuravlov@gmail.com>
Date:   Mon Dec 14 22:19:06 2020 -0400

    Update date-time-functions.md

    toYYYYMM  translation to Russian

commit a0405182ed16815c7b01dd4f9f3ef7e0cb559a9b
Merge: 49f65ecf9d fe6f477881
Author: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
Date:   Tue Dec 15 03:12:23 2020 +0300

    Merge pull request #18066 from ClickHouse/aku/utils-split-build

    Build utils in CI, at least in split build

commit 8de0c0e8fcf975ce074f0f19032f9ed37c8021bf
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Tue Dec 15 02:24:50 2020 +0300

    Fix warning message for unusupported -fsanitize-blacklist

commit 519d89dc9fde458b2c61636ef84c0ee5b0ecf5b1
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Tue Dec 15 02:19:37 2020 +0300

    Do not pass -fsanitize-blacklist for gcc (it does not support it) for UBSAN

commit c78861e86e1f7c6fce2dd3cc3e6b1d9561a229c6
Merge: afc2d7099b 49f65ecf9d
Author: Anton Popov <pad11rus@gmail.com>
Date:   Tue Dec 15 01:45:01 2020 +0300

    Merge remote-tracking branch 'upstream/master' into HEAD

commit afc2d7099b71c786620c2728f944cc992fa98b22
Author: Anton Popov <pad11rus@gmail.com>
Date:   Tue Dec 15 01:16:04 2020 +0300

    add conversion from array to map

commit 49f65ecf9de008a6a51b6d169e24646ed7af91ba
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 23:07:28 2020 +0300

    Update CHANGELOG.md

commit 91be7a6f27a0901384ed95653a76f98aac03a094
Merge: 363ff79cca 5da9315ceb
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 23:03:21 2020 +0300

    Merge pull request #16809 from kitaisreal/read-float-8digits-unroll-performance-improvement

    ReadFloat 8digits unroll performance improvement

commit 363ff79cca91696e391ebd55e08aa87d3d4b0ab4
Merge: 491e086435 346c241c20
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 23:00:43 2020 +0300

    Merge pull request #17934 from Jokser/s3-adaptive-single-part-upload

    Adaptive choose of single/multi part upload in WriteBufferFromS3

commit 491e086435ce06821e9f0d781a963b66bfafeb9b
Merge: 906415540a bf23bc6f0c
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 22:57:27 2020 +0300

    Merge pull request #18057 from azat/fix-01600_remerge_sort_lowered_memory_bytes_ratio-flap

    Fix 01600_remerge_sort_lowered_memory_bytes_ratio flap

commit 906415540aab3a9789650c87ca337efb0d92b356
Merge: 7184746f45 14dedae9f1
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 22:57:02 2020 +0300

    Merge pull request #18060 from kitaisreal/create-query-create-as-select-with-columns-specification

    CreateQuery create as select with columns specification

commit 7184746f45793b8ecabf0a7d8616551730bfea2f
Merge: 5f424f7c5c 60efab623d
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 22:56:39 2020 +0300

    Merge pull request #18059 from excitoon/patch-2

    Fixed `std::out_of_range: basic_string` in S3 URL parsing

commit 17b43ca721c48190085b3b930e87960f7a61474f
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Tue Dec 8 13:54:33 2020 +0300

    Fixed tests

commit 16a30eb4aabc4072ad51da17868e2e31c6cec727
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Tue Dec 8 10:43:18 2020 +0300

    Updated non-finite floating conversion into int

commit e945821e1f7d5c3c0119c193c99409212fca04a7
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 7 20:34:16 2020 +0300

    Update castColumn.h

commit 26e582272f321937749b3477ca4aada3293cf706
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 7 20:34:01 2020 +0300

    Update castColumn.h

commit 6873b1358753436879665a191f605499dc54b497
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 6 14:58:54 2020 +0300

    Fixed clang build issue

commit 0b565b679e40015c7a8b0e2f958f07ed8ab9d540
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Dec 5 17:55:37 2020 +0300

    Fixed Set execute for columns that cannot be inside Nullable

commit 18dc118298ffb0b49792012cf772058cd088b226
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Dec 5 16:46:14 2020 +0300

    Fixed compile issues

commit 0464859cfeb9d25af5c614173558adb62f7b9920
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Dec 5 16:18:56 2020 +0300

    Updated usage of different types during IN query

    1. Added accurateCast function.
    2. Use accurateCast in Set during execute.
    3. Added accurateCast tests.
    4. Updated select_in_different_types tests.

commit 6f4bc77dbcf35e18f85c41c61739da2355ca6834
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 2 15:28:29 2020 +0300

    Added equality comparsion

commit f4b8e8ef9931a2b1d2fb4710286cbc9d066ae956
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 2 13:09:13 2020 +0300

    Allow different types inside IN subquery

commit dff31e8de7f158e0dcfa4e43b6663edd130d83e8
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Nov 30 11:26:00 2020 +0300

    Fixed compile issues

commit 0dac432942c85c8803bdf8c38e54e66a3ec85867
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Nov 28 15:55:29 2020 +0300

    Added const to arguments

commit dd072a20c9ab58e0f06d1438dc75252ff301bc3b
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Nov 15 22:24:15 2020 +0300

    Fixed compile issues

commit 7b66dc374c660102dbd34ee532e1bbe8a46a32e8
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Nov 15 13:03:53 2020 +0300

    Fixed clang compile issues

commit 4435642e39ae5c6f763d2356c10e8ac32020f7c6
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Nov 12 18:56:17 2020 +0300

    Fixed style issues

commit db5d0944ed47f7cd9a789ebc9e691fa2dfb3271c
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Nov 12 14:27:02 2020 +0300

    Added Decimal support

commit 476fb3b677edbcaf1528011e373deee88198c0d3
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Nov 12 13:32:37 2020 +0300

    Updated documentation

commit baad78d2f2d13c5873591999765eb6b838558919
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Nov 12 12:28:49 2020 +0300

    Fixed setting null value in vector. Updated tests.

commit 4cd5025e83849790bb897d66b20fb0ee9bda85de
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Tue Nov 10 16:18:58 2020 +0300

    Added numeric limits for UInt128

commit d636e1618ba8c055d8f09da004e8bcb2b7afca86
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Thu Nov 5 22:09:17 2020 +0300

    Added function accurate cast or null

commit fe6f4778814ded00009fe94a82344ac5529ebba8
Author: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date:   Mon Dec 14 20:39:52 2020 +0300

    Build utils in CI, at least in split build

    Currently we don't build them at all, so they were broken by https://github.com/ClickHouse/ClickHouse/pull/17988

commit 5f424f7c5cc8384cbbf87aabcc54018513d73b31
Author: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
Date:   Mon Dec 14 20:08:32 2020 +0300

    Update zzz-perf-comparison-tweaks-config.xml

commit 19af691aa6dd64c2ccb32e34e78b659d15352edb
Merge: e7f93b5563 b08842d44e
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 19:44:13 2020 +0300

    Merge pull request #17738 from vklimontovich/docker-cli-example

    Docker cli example

commit 73780838a14beabd4f2bcde74067117cbf7141c5
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 19:01:29 2020 +0300

    Remove dead list watches

commit b08842d44e9c50f4cba6b97f1da95981dea7de23
Author: filimonov <1549571+filimonov@users.noreply.github.com>
Date:   Mon Dec 14 16:54:37 2020 +0100

    Update README.md

commit 1436986a86528072e128e4f81c771c3e5e4860b4
Author: filimonov <1549571+filimonov@users.noreply.github.com>
Date:   Mon Dec 14 16:53:58 2020 +0100

    Update README.md

commit bca41bb7fd1275ed989113ef2f729e4f4fa89eac
Author: filimonov <1549571+filimonov@users.noreply.github.com>
Date:   Mon Dec 14 16:50:48 2020 +0100

    Update README.md

commit e7f93b5563fd6f25925788171a7d5b4b764c8918
Merge: 2edabd5519 653ddab63e
Author: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
Date:   Mon Dec 14 17:54:47 2020 +0300

    Merge pull request #18058 from azat/perf-test-zk-port-fix

    Remove test_keeper_server in perf tests

commit 2edabd5519d1eecd256f0a03311bb91dd2cc7344
Author: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
Date:   Mon Dec 14 17:45:58 2020 +0300

    Update zzz-perf-comparison-tweaks-config.xml

commit 24abcc970bc2cd1e4c6c15eb9b045c4956b9e57a
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Mon Dec 14 15:21:31 2020 +0100

    Tests and fixes

commit 87ea1528cb50bd0b86ec828eb38a3fb7773d6fb9
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Tue Dec 8 16:54:49 2020 +0100

    support for Lazy database

commit 4ba80996b0e5f8dbf051f59e7dcf493420f19903
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Tue Dec 8 16:48:30 2020 +0100

    Detach permanently for DatabaseConnectionMySQL

commit 799141362d5aed0697d576153ae04208afa0ca7c
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Tue Dec 8 11:00:32 2020 +0100

    Prevent attempts to recreate the table which was detached permanently

commit ad9cb580a9a864409e2ce90a99ff90e09858888a
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Mon Dec 7 16:19:05 2020 +0100

    Fix

commit 2aeb72c09ad189bcce4e9c75e0d2ccd599903ac0
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Mon Dec 7 15:01:09 2020 +0100

    Fix for tiny issue #13850

commit 3f5e5eb15ad8684e492d7b8fabee4f73c5141810
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Mon Dec 7 14:55:24 2020 +0100

    Handle not implemented cases properly

commit f6097cf818302d2873e03fa94eec323e92609d90
Author: Mikhail Filimonov <mfilimonov@altinity.com>
Date:   Mon Nov 30 18:52:32 2020 +0100

    Adding DETACH TABLE PERMANENTLY

commit 34d1ad1acae26e0e2387229df64aa422e09ee4e2
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 17:27:39 2020 +0300

    Fix writer compact

commit dd2ae6926dd3ef82176146c75d5f437c1c0a8a67
Merge: 7a8510d1c4 7f7eed4031
Author: tavplubix <avtokmakov@yandex-team.ru>
Date:   Mon Dec 14 16:27:13 2020 +0300

    Merge pull request #14849 from ClickHouse/allow_atomic_database_inside_materialize_mysql

    Allow Atomic database inside MaterializeMySQL

commit 35d2a689dd39bf3c632ccf93377cc53304a7c1fd
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 16:06:11 2020 +0300

    One more time

commit 2a2b02d395cea92db344a669939e9c3d4ced68a3
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 16:05:48 2020 +0300

    One more time

commit 9e98cdad5a2356150cd336ad74ba191e7bce71b6
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 16:05:07 2020 +0300

    Accident changes

commit 47e3f619e5f75d3444dc9b7aae25973f104c29a7
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 16:01:01 2020 +0300

    Remove redundant code

commit a87fb3fae0c3e9894f60118e95e320ffce9b1a5d
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 15:52:15 2020 +0300

    Fix style

commit 303c12d2e4f1019b91843d222a0e4c63e7060dba
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 15:51:14 2020 +0300

    Slightly better

commit 14dedae9f1b665569cc438ea0f43904b61480190
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Dec 14 15:50:11 2020 +0300

    Updated sql test file

commit 7a8510d1c454f335dc53dc86cd0f9542ca6eb169
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 15:45:13 2020 +0300

    Update CHANGELOG.md

commit ad70b0bd9776856ff9a20ebb8fee5a3402088d63
Merge: 0bc2603b1d 9f4929bdce
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 15:43:51 2020 +0300

    Merge pull request #18062 from ClickHouse/changelog-20.12

    Add changelog for 20.12

commit 9f4929bdce52f09b980f0d2ed79a5ed4d1fe19f9
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 15:30:37 2020 +0300

    Add changelog for 20.12

commit 55f991adde8d67d379754a4fdddd1347d7af4d9a
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Dec 14 15:23:53 2020 +0300

    Updated test

commit 6fb23da851b845149abef338823b4f4dfcd55e04
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 15:03:49 2020 +0300

    Refactoring

commit 70be03aa7d879485660f9d36549d04e560783303
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 14:36:15 2020 +0300

    Working compact parts

commit 0bc2603b1de56dc67b2c6c74e97a69efe9ecf5e3
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 14:14:59 2020 +0300

    Update 01081_PartialSortingTransform_full_column.sql

commit ab17927c853739537682b9c7050aec4cd03c3bbe
Merge: d55c5bec94 8a20d449d9
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 14:11:09 2020 +0300

    Merge pull request #18032 from kitaisreal/array-aggregation-functions

    Array aggregation functions

commit fbcf85cb4261bd0ff2f0b93bc65dd2f67b5407a9
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 14:06:02 2020 +0300

    Something working for wide parts

commit 8a20d449d999bb53d5941b8f7847a2aff9e9c4b3
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 14:01:31 2020 +0300

    Update arrayAggregation.cpp

commit d55c5bec9481833d4cd2ea86c0cf29269ff28c25
Merge: 493ee6732b 2bf226decc
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 13:59:11 2020 +0300

    Merge pull request #18053 from azat/librdkafka-snappy-UB-fix

    Fix possible corruption in librdkafka snappy decompression

commit 493ee6732b4eff62da489657d8f02f5a4640f0ad
Merge: aac8b85b34 1bccd6dff6
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 13:58:45 2020 +0300

    Merge pull request #18052 from kitaisreal/memory-storage-read-optimization

    MemoryStorage read optimization

commit aac8b85b34d788359cebb589416489a281d732a9
Merge: 86cef85186 6341521b10
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 13:58:15 2020 +0300

    Merge pull request #18054 from ClickHouse/remove-obsolete-settings

    Remove obsolete settings

commit 60efab623d5295b64d17ca7331b31d00f8c24762
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 13:53:40 2020 +0300

    Update S3Common.cpp

commit 86cef85186af389e9ef60f8ca2ec2ccfc7d49182
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 13:46:39 2020 +0300

    Remove old changelog generator to avoid confusion

commit 5f7c867da33c9bcede56c36096b77e1f0b8925f6
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Dec 14 12:33:16 2020 +0300

    Added parse error if no select after as keyword

commit d8534ae1afe618c8e390788b8debbe6bee50143a
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Dec 14 12:30:08 2020 +0300

    CreateQuery create as select with columns specification

commit a98e2311f8f5bac70fc7a77229e12dd13b7484d5
Merge: e7ffae89e9 1b3893bcab
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Mon Dec 14 12:10:17 2020 +0300

    Merge pull request #17908 from vitlibar/fix-checking-introspection-grants

    Fix checking introspection grants

commit 418c974eafcfa9b191fc2a914b0e918a480b0833
Author: Vladimir Chebotarev <vladimir.chebotarev@gmail.com>
Date:   Mon Dec 14 11:17:29 2020 +0300

    Fixed `std::out_of_range: basic_string` in S3 URL parsing.

commit 346c241c2066d4c5413a2380cd7f0b7f5e98f43f
Merge: 59a5e63417 e7ffae89e9
Author: Pavel Kovalenko <jokserfn@yandex-team.ru>
Date:   Mon Dec 14 11:00:30 2020 +0300

    Merge remote-tracking branch 'origin/master' into s3-adaptive-single-part-upload

    # Conflicts:
    #	src/Disks/S3/DiskS3.cpp

commit 653ddab63ebedd43ecdb50a7f2230fa08d458ced
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Mon Dec 14 10:59:01 2020 +0300

    Remove test_keeper_server in perf tests

    To avoid port overlaps for right and left server (anyway it is not
    used).

commit bf23bc6f0c9f8e392845b6d612ef1272dc8e59a5
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Mon Dec 14 10:46:30 2020 +0300

    Fix 01600_remerge_sort_lowered_memory_bytes_ratio flap

commit 12b6a54a2473ba9decd774b146b8616da6847b7d
Merge: fd695c9ff4 e7ffae89e9
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 10:30:01 2020 +0300

    Merge branch 'master' into try_rewrite_writer

commit fd695c9ff40e4afa266f68523a7be107569ef2f5
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 14 10:28:42 2020 +0300

    Dive into debug

commit f829fbb9284650b3eaee1dcc371d421bf9378688
Merge: 2169e8f7f4 e7ffae89e9
Author: jianmei zhang <zhangjmruc@sina.com>
Date:   Mon Dec 14 11:21:42 2020 +0800

    Merge remote-tracking branch 'upstream/master'

commit 2169e8f7f441d11be2a1eb2290f79429f59d13ae
Author: jianmei zhang <zhangjmruc@sina.com>
Date:   Mon Dec 14 10:33:51 2020 +0800

    Fix issue #17160 by adding show settings like '...'

commit 5da9315ceb726e91548a58a735b6e509363a8062
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 05:22:17 2020 +0300

    Update readFloatText.h

commit 66b2424e3362299b8a4f655464ae23bdede687c9
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 05:20:00 2020 +0300

    Update readFloatText.h

commit 6341521b104c3170ec5e28e2c152d40c43309e75
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 02:40:11 2020 +0300

    Remove obsolete settings

commit e7ffae89e946e88cebde0f3de464b5d545f3c3dd
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 02:31:35 2020 +0300

    Update build.sh

commit 52f03d942e1942415e94d3f40ce532ca267d1de8
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Dec 14 02:31:18 2020 +0300

    Update Dockerfile

commit 1bccd6dff6528c30cd6b053337b4f7012f416340
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Dec 14 00:50:55 2020 +0300

    Fixed style issues

commit 2bf226deccbe86fbd52de7628d09426d53ae2158
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Mon Dec 14 00:38:10 2020 +0300

    Fix typo in KafkaSettings comments

commit 3b431b054ec732b0cfb7538fd6a1ba01bd22f1be
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Mon Dec 14 00:34:38 2020 +0300

    Bump librdkafka to fix UB in snappy decompression.

    In incremental_copy_fast_path there is undefined behavior (and in some
    other places too).

    And under this circumstances gcc10 with -O1 -ftree-loop-vectorize (or
    simply -O3), due to loop unroll, generates code that do copy by 16 bytes
    at a time for the second loop (MOVDQU+MOVUPS), while this is not correct
    since the memory may be overlapped and may be changed in the previous
    iteration.

    Updated version includes [1], that fixes those UB, by using memcpy over
    direct store/load since these days direct store/loads looks redundant.
    Even on ARM.

      [1]: https://github.com/ClickHouse-Extras/librdkafka/pull/3

    NOTE: clang is fine, and other older versions of gcc too

commit 2e569899d1cdf61e27b15335f2535f82a108fad4
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Thu Dec 10 10:18:49 2020 +0300

    Add a test for kafka with snappy compression method (regression for UB in snappy)

    I've checked it manually and it works with clang build and does not with
    gcc (there are NULL bytes in the middle of value).

commit 69127ca0ed33d75af28021554d4b408f6efca687
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Dec 14 00:21:25 2020 +0300

    Changed BlocksList to Blocks

commit 153610b2c395b430f3d0b3196948554d3a60ec9d
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Sun Dec 13 23:44:09 2020 +0300

    Update distribution_queue.md

    Поменял also на Also.

commit 698737681abd500f871eb30d707254175831be36
Merge: 629f5f0797 8702809c9f
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Sun Dec 13 23:41:22 2020 +0300

    Merge remote-tracking branch 'remotes/clickhouse/master' into sevirov-DOCSUP-4919-document_replication_queue_system_table

commit 9310437920f111249dea294a75c6d2aeed5e0973
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Thu Dec 10 23:16:53 2020 +0300

    Fix access rights required for the merge() table function.

commit 3c6df61e5c863e868541e63fe6c99038aeaec04b
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 22:24:15 2020 +0300

    MemoryStorage read optimization

commit 8702809c9fa9ba4ec6f99a4765be646045aef5d9
Merge: 246e1fbb2f 0954cc2b5c
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 21:28:03 2020 +0300

    Merge pull request #17464 from amosbird/ddlworkermetric

    ddlworker max entry metric

commit 0be51cbabd94adb256514549577929d2f5d13a29
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 21:27:48 2020 +0300

    Updated tests

commit 246e1fbb2f108ef382fa7e67a144ee3e022530ca
Merge: 4a2aaa68a8 149431ffc3
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 21:09:09 2020 +0300

    Merge pull request #18048 from ClickHouse/play-auto-dark-theme

    Automatic support for dark theme in Play interface

commit 4a2aaa68a87150f98ef7daa047ee2bcdce89310c
Merge: 9116eb7aca 6d87743b0e
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 21:08:50 2020 +0300

    Merge pull request #18049 from ClickHouse/datasets-cdn

    CDN for datasets

commit 6d87743b0e53af533caf42417b0d43b3392ba897
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 21:06:27 2020 +0300

    CDN for datasets

commit 149431ffc32c1fd0e712aa5cf7c8fd84a433d921
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 20:54:57 2020 +0300

    Automatic support for dark theme in Play interface

commit 9116eb7aca749240bd23ae56da7e6d2e3fea6795
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 19:35:56 2020 +0300

    Add checksums

commit d75c7ec30c35dc0892bdb49cf3b0f17bee665145
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 19:28:45 2020 +0300

    Redundancy.

commit 629f5f07972d7af7cd2139074e8d0b93a786f4ca
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Sun Dec 13 18:42:34 2020 +0300

    Update replication_queue.md

    Выполнил перевод на русский язык, внес изменения в английскую версию.

commit d8484d9c31c18d56d60ded9729b8514ccc322911
Merge: 73890fa3a7 ef064696e7
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 18:28:52 2020 +0300

    Merge pull request #18033 from ClickHouse/encode-xml-rename-files

    encodeXMLComponent: rename files after #17659

commit 73890fa3a76583df3afbe3c3c0609c78b0db0cd1
Merge: cd5f827265 b394487d4f
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 18:27:52 2020 +0300

    Merge pull request #18046 from baibaichen/feature/update_doc

    update Mac-OS build documentation

commit cd5f8272657d6ea94c99813d3a1f919aaae07517
Merge: 443d1724b3 06b047235a
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 18:27:01 2020 +0300

    Merge pull request #18047 from kitaisreal/space-saving-remove-last-element-from-map-fix

    HashTable erase added tests

commit 7f7eed4031b403c8edc01cddbd14b31cd644b152
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Sun Dec 13 17:57:15 2020 +0300

    fix test

commit 0c80fce7f0326dde97e37a711335b7ea5d8554cf
Merge: ed355f8663 443d1724b3
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Sun Dec 13 17:56:50 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit 5d6036e9c8c285d2f00ce01c9cd0afc4c89d9488
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 14:38:56 2020 +0300

    Fixed style issues

commit 06b047235aa4c0ccdc16b3847455622a41f8f515
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 14:35:52 2020 +0300

    Fixed style issues

commit 547f452ba98baade61171448d37efe9d5deada39
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 13:30:25 2020 +0300

    HashTable erase added tests

commit b394487d4f45821a9cebe019c7bf71c48f690113
Author: chang.chen <baibaichen@gmail.com>
Date:   Sun Dec 13 16:10:07 2020 +0800

    update Mac-OS build documentation

commit 443d1724b3f65989e184138a6332f0cfabacac20
Merge: a714b82f9b 065a64cc37
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:54:14 2020 +0300

    Merge pull request #18031 from sevirov/sevirov-DOCSUP-4709-document_distribution_queue_system_table

    DOCSUP-4709: Document system.distribution_queue table

commit a714b82f9b875582cec6af0b3436a835de645ce5
Merge: 785317e542 784a484d80
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:52:39 2020 +0300

    Merge pull request #18036 from ClickHouse/benchmark-macbook2020

    Added results from MacBook Pro

commit 784a484d80d486dff837f4e4a18d54ad1de1ce87
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:51:56 2020 +0300

    Added results from MacBook Pro

commit 785317e5427f39ba13d9a64fdfab2c414cef4375
Merge: bce02de691 10beda9d39
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:44:47 2020 +0300

    Merge pull request #18035 from ClickHouse/benchmark-results-seodo

    Add benchmark results Yiğit Konur and Metehan Çetinkaya of seo.do

commit 10beda9d39f86b0c858bba9a36e2e319d140ff0d
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:44:07 2020 +0300

    Add benchmark results Yiğit Konur and Metehan Çetinkaya of seo.do

commit bce02de691a64f07cc9a4e6392705b8e5935f0ff
Merge: 436adf34df ccd794a676
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:38:50 2020 +0300

    Merge pull request #17539 from azat/remerge_sort_lowered_memory_bytes_ratio

    Add remerge_sort_lowered_memory_bytes_ratio setting

commit 436adf34df642b3f49ea430f843f9d35f054f818
Merge: 4ce148d2ea 8d79ae56f0
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:37:29 2020 +0300

    Merge pull request #17881 from gyuton/gyuton-DOCSUP-4263-Update_the_sql_statement_ORDER_BY

    DOCSUP-4263: Updated the sql statement order by

commit 4ce148d2ea9c6aa45186399004751f29e47058b1
Merge: 8df4789113 17a8a4e966
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:34:46 2020 +0300

    Merge pull request #18034 from edani/master

    Expand max_parallel_replicas documentation

commit 8df478911323771ff94aada88cf79a82b6a7822e
Merge: 04e222f6f3 6f50a86025
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 04:09:51 2020 +0300

    Merge pull request #17845 from kitaisreal/space-saving-remove-last-element-from-map-fix

    SpaceSaving remove last element from map fix

commit 04e222f6f38acb0d9395751aeb8c5096fb484f0a
Merge: 2ad5c8eed3 41bc43325b
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 01:12:05 2020 +0300

    Merge pull request #17986 from azat/fwd-decl

    More forward declaration for generic headers

commit 17a8a4e9664fabed5b370b37e148139ba698acf5
Author: Eric Daniel <6210036+edani@users.noreply.github.com>
Date:   Sat Dec 12 13:57:01 2020 -0800

    Expand max_parallel_replicas documentation

    Based on discussion in #17807

commit 5ae1652d51f3ce24f4b76d71922de783d180664e
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 00:47:17 2020 +0300

    Fixed aggregate function avg

commit 2ad5c8eed34ce0de0b2d7a5bdaac2dafc4cec2eb
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 00:45:23 2020 +0300

    Allow server to start with TestKeeper port in development environment

commit b810b28b085756f22439cba83c7b15680f40c005
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 00:40:34 2020 +0300

    Added comment

commit 4887d59b064ba4e88fcd629a712a864f4d0cfa70
Merge: 2eabd41749 c2c497e290
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 00:39:41 2020 +0300

    Merge pull request #18027 from Jokser/fix-flappy-log-family-s3

    Fix flappy test_log_family_s3

commit 0dcbcec19ae89305afa12df5c28c87b40de88787
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 13 00:29:09 2020 +0300

    Added tests

commit ef064696e77aa4d6e9fc72bf3b412145ddaefaf2
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Dec 13 00:17:37 2020 +0300

    Add perf test

commit 532557ecfd463d1e1de6cca6263b9cf526934c34
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 23:47:37 2020 +0300

    encodeXMLComponent: rename files after #17659

commit 18bde6ef020e6e3f1c6e92c7e765eb7810b90dce
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Dec 12 23:47:23 2020 +0300

    Fixed compile issues

commit 8d79ae56f0180dd4e9838bd79b3d9243af0eb7ce
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 23:07:24 2020 +0300

    created lowcardinality files for ja and tr

commit 3d261c82e9a7d391c4aef23bd688bf36ce48e1b5
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Dec 12 22:50:28 2020 +0300

    Array aggregation functions

commit 065a64cc3701a122f21cd6b64c8285edb20f801e
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Sat Dec 12 22:43:52 2020 +0300

    Update distribution_queue.md

    Добавил ссылку на движок таблиц Distributed.

commit a244a4f0e34bba5d0f18d395582b7eb60b454a9a
Merge: c03e91eec0 6e417088dc
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Sat Dec 12 22:28:09 2020 +0300

    Merge branch 'sevirov-DOCSUP-4709-document_distribution_queue_system_table' of https://github.com/sevirov/ClickHouse into sevirov-DOCSUP-4709-document_distribution_queue_system_table

commit c03e91eec0d1029def0f44096c3200eba5b6cd0e
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Sat Dec 12 22:27:01 2020 +0300

    Create distribution_queue.md

    Выполнил перевод на русский язык.

commit 6f50a860250a2e52cd9947b0a82ce6548fdb1d35
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 18:57:07 2020 +0300

    Update gtest_hash_table.cpp

commit 41bc43325bb7c63369ac206bd2db102b4882f640
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Fri Dec 11 21:10:37 2020 +0300

    Make 01290_max_execution_speed_distributed less flaky

commit 77a1d00dee7157ff89a73a6a10a2cb3a0aa27de1
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Fri Dec 11 21:10:36 2020 +0300

    Add new remote context as interpreter context to the Pipe

commit 5b3ab488610b19f829507f04c7b74b0fec6d52bc
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Fri Dec 11 01:05:02 2020 +0300

    More forward declaration for generic headers

    The following headers are pretty generic, so use forward declaration as
    much as possible:
    - Context.h
    - Settings.h
    - ConnectionTimeouts.h
    (Also this shows that some missing some includes -- this has been fixed)

    And split ConnectionTimeouts.h into ConnectionTimeoutsContext.h (since
    module part cannot be added for it, due to recursive build dependencies
    that will be introduced)

    Also remove Settings from the RemoteBlockInputStream/RemoteQueryExecutor
    and just pass the context, since settings was passed only in speicifc
    places, that can allow making a copy of Context (i.e. Copier).

    Approx results (How much units will be recompiled after changing file X?):

    - ConnectionTimeouts.h
      - mainline: 100

    - Context.h:
      - mainline: ~800
      - patched:  415

    - Settings.h:
      - mainline: 900-1K
      - patched:  440 (most of them because of the Context.h)

commit c2c497e290af35cd4945f23136a8f07de737156d
Author: Pavel Kovalenko <jokserfn@yandex-team.ru>
Date:   Sat Dec 12 16:44:19 2020 +0300

    Use s3 disk explicitly for log tables in test_log_family_s3.

commit 07c829304326eb52d6bc2e2ae8750d91fa0fb06a
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sat Dec 12 15:08:46 2020 +0300

    Added tests

commit c7d48c48ce247bca0e4a89cdf8cb6eb122dff3de
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 11:58:07 2020 +0300

    more troubleshooting

commit fa9257cd82997c187d5f74771e35d6ee2533c4ed
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 11:14:46 2020 +0300

    fix

commit 1567444628f82bdf79e7012bd3c01174ea4c3cd5
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 10:36:39 2020 +0300

    troubleshooting - fixed links to original articles

commit 0954cc2b5c193eb721cb2a0ef244bdfe9276c322
Author: Amos Bird <amosbird@gmail.com>
Date:   Sun Nov 29 21:28:11 2020 +0800

    test is not working

commit 8d0e949462ac55ae531e30209eb5a47da2ac7a33
Author: Amos Bird <amosbird@gmail.com>
Date:   Fri Nov 27 12:27:43 2020 +0800

    ddlworker max entry metric

commit 9b1d91911009c672e2f6689804193de83b5ab0e4
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 06:39:22 2020 +0300

    more troubleshooting

commit 3794cb2d3f299c9a0953d4d4c8c6021be1f3501e
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 05:48:23 2020 +0300

    troubleshooting the error

commit c74011851acb5dfa14cfe00885370781d53be90f
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 04:58:07 2020 +0300

    Update HashTable.h

commit d94d9aa8d72ca3fee79ecd1a0d0e465ae6cde6ff
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 04:17:32 2020 +0300

    Update HashTable.h

commit d5f87efba5de521627f7786b3459f9c392ce15e4
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 04:16:58 2020 +0300

    fixed a typo in translation

commit 99212dcb7316cde90490968a9fb4eb4f3b39210b
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 04:06:37 2020 +0300

    added translation

commit 9e48b9b656ab904456deb035fd41191b8d72d583
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 03:47:43 2020 +0300

    Update HashTable.h

commit cabe1463d10874c011350b636e7d8a104b7b8e01
Author: George <gyuton@yandex-team.ru>
Date:   Sat Dec 12 03:39:45 2020 +0300

    fixed an error

commit eba1299845e0b490a1d813e10ab25802177baa00
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 03:38:55 2020 +0300

    Update HashTable.h

commit 245dea6f2ea380732a1f64e07fd630e599567933
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 03:33:57 2020 +0300

    Update HashTable.h

commit 7e2bc1e7a650d3f032e6eccea0628f7ad2b48f11
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 03:30:12 2020 +0300

    Update HashTable.h

commit 6e7366f921ec86fa1a31bbe85db4a9c26f9a6e45
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Sat Dec 12 03:28:41 2020 +0300

    Update HashTable.h

commit 59a5e63417148956cf79c65d0979a6cbb456caf7
Merge: 5802633f95 f1a145a2b2
Author: Pavel Kovalenko <jokserfn@yandex-team.ru>
Date:   Fri Dec 11 18:28:41 2020 +0300

    Merge remote-tracking branch 'origin/master' into s3-adaptive-single-part-upload

    # Conflicts:
    #	src/Disks/S3/DiskS3.cpp
    #	src/Disks/S3/DiskS3.h
    #	src/Disks/S3/registerDiskS3.cpp
    #	src/IO/WriteBufferFromS3.cpp
    #	src/IO/WriteBufferFromS3.h

commit a24a76828ede28d530d589a074ecbed6ab19a9f2
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 11 18:18:51 2020 +0300

    Fix

commit 4b83707c6326b4069585038f0fd6dd8b13a18f21
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 11 18:00:58 2020 +0300

    Compileable code for some reason

commit c6ac831d4eddf3eeeefa6f4c770c1e6bf6172ba9
Author: Amos Bird <amosbird@gmail.com>
Date:   Fri Dec 11 22:44:22 2020 +0800

    Fix type

commit b70b98c4154d15206d99afda2a9af6362d2842e4
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 9 23:56:15 2020 +0300

    Fix style issues

commit 4ae56197149238bb31c18d11a957cf23434fc246
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 9 22:44:43 2020 +0300

    Optimized erase in HashTable

commit bf5d75853cf3840de8ee11f1bd976080965f922f
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Wed Dec 9 15:47:22 2020 +0300

    Added erase into HashTable

commit bfceb066594fd2d9093710345be5e58cf532c88c
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 6 22:16:21 2020 +0300

    Fixed test

commit d8b78682e1ee1387f34da67ab1422263ecfd0285
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Sun Dec 6 22:14:40 2020 +0300

    SpaceSaving remove last element from map fix

commit ed355f866372504763d36983254dcca446f18281
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Fri Dec 11 16:50:45 2020 +0300

    fix

commit 1b3893bcab7b7f95d30ce5bb23764de27181a690
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Tue Dec 8 21:09:13 2020 +0300

    User with allow_ddl=0 cannot do DDL but can grant DDL.

commit 648be453a41415dc24c3b425a790411fff5c3122
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Tue Dec 8 21:07:50 2020 +0300

    User with allow_introspection_functions=0 cannot call introspection functions but can grant INTROSPECTION.

commit 710ba6f6173e11f3613f76a707c10b436f255813
Author: Vitaly Baranov <vitbar@yandex-team.ru>
Date:   Tue Dec 8 17:24:32 2020 +0300

    Simplify class ContextAccess.

commit 57c30c4164d19a1060f463b68813cc731f4fe22b
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 11 16:20:19 2020 +0300

    Garbage

commit 58a31e4882c859c3e7ce8717359049182a910f92
Merge: 0ba712a2d3 fb1221148d
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Fri Dec 11 16:14:03 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit 7ffd76a6051da99c12c84a8a950d32e1ba98ff81
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 11 11:41:02 2020 +0300

    Fix bug

commit cbba73672e89fd503201157a659b522a0439e2fe
Merge: 1cb06bd975 e11790f61a
Author: alesapin <alesapin@gmail.com>
Date:   Fri Dec 11 11:09:47 2020 +0300

    Merge branch 'master' into try_rewrite_writer

commit 5b1e5679b4a292e33ee5e60c0ba9cefa1e8388bd
Author: Amos Bird <amosbird@gmail.com>
Date:   Wed Nov 11 12:05:54 2020 +0800

    Add -SimpleState combinator

commit 1cb06bd97570403084c5d6958872a05abff8345c
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 10 19:29:10 2020 +0300

    Split one method

commit 90fa9a20733072861f752797b73c909e6617ac09
Merge: f3d1dc360c b1c00d4045
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 10 18:36:10 2020 +0300

    Merge branch 'master' into try_rewrite_writer

commit f3d1dc360c89d68b7693690264546fc5f97160cb
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 10 12:24:52 2020 +0300

    Fix style

commit 0704b478ee6c741cb86016ec8d3c1ab0b8d65338
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 10 12:22:43 2020 +0300

    Less constuctors

commit 2787b615b0d5b43ca56770f01f688a87bbbb06d0
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 10 11:57:52 2020 +0300

    Make interface as simple as possible

commit 7979b3655c46fbcb12fce8c69d04f9ba25417777
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 9 21:22:07 2020 +0300

    Better formating

commit 3b88c077edd8481a3b3bb2ac698cde87628abaa0
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 9 21:19:49 2020 +0300

    Mechanical refactoring

commit af73db93e6b05073997508956165c974b3f91a8d
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 9 21:10:09 2020 +0300

    Mechanical refactoring

commit 7b88f9a115dd11dce65ec34079a2a9def35cde1b
Merge: fa64d314df 89a3a86b71
Author: alesapin <alesapin@gmail.com>
Date:   Wed Dec 9 19:23:53 2020 +0300

    Merge branch 'master' into try_rewrite_writer

commit 5802633f957bd5ee29611a16116219d4b32cbc79
Author: Pavel Kovalenko <jokserfn@yandex-team.ru>
Date:   Wed Dec 9 18:38:02 2020 +0300

    Fix parameter name in constructor of DiskS3

commit c8c543ca9416fb193a13545d3f23d6b3a60c17c3
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Wed Dec 9 18:07:58 2020 +0300

    Add throwing exceptions

commit db7a87089b69744f60edd506f0287e3ae8363b35
Author: Pavel Kovalenko <jokserfn@yandex-team.ru>
Date:   Wed Dec 9 17:09:04 2020 +0300

    Adaptive choose of single/multi part upload in WriteBufferFromS3.

commit 215c5675dd470292a34a02fcc2b0d2c2b78be122
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Tue Dec 8 23:25:45 2020 +0300

    Update replication_queue.md

    Изменил описание столбца node_name.

commit ce721f95ee9a1578167651c7f101f978814d79fc
Author: Dmitriy <sevirov@yandex-team.ru>
Date:   Tue Dec 8 21:46:25 2020 +0300

    Create replication_queue.md

    Задокументировал системную таблицу replication_queue.

commit 291959c1b35c7fbf62cd45dafbb13741cbfb10e7
Author: Kruglov Pavel <48961922+Avogar@users.noreply.github.com>
Date:   Tue Dec 8 19:19:51 2020 +0300

    Update comment

commit 0ba712a2d36a394f9af876f4f3204eb21856e349
Merge: a0ef3a113a d9b52f94f7
Author: Ivan <5627721+abyss7@users.noreply.github.com>
Date:   Tue Dec 8 16:45:51 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit e19eb6f17ad8d330f2b3206fedacd9a8cdbd8b8a
Merge: 5ae6c6dab9 d9b52f94f7
Author: Kruglov Pavel <48961922+Avogar@users.noreply.github.com>
Date:   Tue Dec 8 15:57:46 2020 +0300

    Merge branch 'master' into optimize-data-on-insert

commit 4e49c59ed0beb0ccf54b0704588c5648fc9d9e51
Author: George <gyuton@yandex-team.ru>
Date:   Tue Dec 8 04:55:57 2020 +0300

    minor fix

commit 6be3933ceae3399135ea7dfd5f2cfe47da485c6a
Author: George <gyuton@yandex-team.ru>
Date:   Tue Dec 8 04:27:26 2020 +0300

    fixed links v2

commit 8760fc12272a8d30a253b7bf2f54f00cd9a89e8a
Author: George <gyuton@yandex-team.ru>
Date:   Tue Dec 8 04:07:34 2020 +0300

    Fixed links

commit 1f0cfbaca89701946793f2d11c3a9e8c84c03326
Author: George <gyuton@yandex-team.ru>
Date:   Tue Dec 8 03:51:41 2020 +0300

    Added Collation description

commit 371cb23946d1f809aabda0c3b8726d95ffbc742f
Merge: b68aa1c76d dd8362a5eb
Author: alesapin <alesapin@gmail.com>
Date:   Mon Dec 7 13:30:41 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit c3a4e5e7b78ddbbca38efe46bc53e00829fa3a49
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Mon Dec 7 01:12:54 2020 +0300

    Fix max_distributed_connections (one more time)

    With prefer_localhost_replica=1 max_distributed_connections uses
    max_threads before this patch.

    Fixes: #14936
    Previous fix: #9673

commit ce03a896a17f8fadb47df55799fbfe3b0ac336ee
Author: Anton Popov <pad11rus@gmail.com>
Date:   Sat Dec 5 22:51:16 2020 +0300

    remove unused files

commit 7aaa2c052ff1035ef7b0211145f75bcddf708d74
Merge: 6b09669d66 67c82eb61b
Author: Anton Popov <pad11rus@gmail.com>
Date:   Sat Dec 5 22:41:46 2020 +0300

    Merge branch 'master' into merging-map

commit 6b09669d6600c6551a4fd6ff5be3892c267f85eb
Author: Anton Popov <pad11rus@gmail.com>
Date:   Thu Dec 3 06:52:41 2020 +0300

    data type map: several fixes

commit 5ae6c6dab9b06acc1707db44077346e7f0208bdd
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Dec 4 20:40:28 2020 +0300

    Fix build error

commit 4b58528b9ed886eea538d07249d51515a1c214a0
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Dec 4 19:25:30 2020 +0300

    Rename getPos to getRow, change mergeBlock, pass setting instead of context

commit 905ba78adce1fcf9bc2f8fa7090472bc34ff4f8f
Merge: e5588b94aa 7c2a04d77d
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Dec 4 18:56:46 2020 +0300

    Merge branch 'master' of github.com:ClickHouse/ClickHouse into optimize-data-on-insert

commit b68aa1c76d0f439d3538a0ca5e816170422b944d
Merge: 9df414fa0c 2b3281888a
Author: alesapin <alesapin@gmail.com>
Date:   Thu Dec 3 17:22:09 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 9bf5ab9f4c91015ed8fe0fc99cba64acf7309f45
Author: Vladimir Klimontovich <v@klimontovich.com>
Date:   Wed Dec 2 19:01:31 2020 +0300

    Removed unnecessary mapping

commit 274c87d4c78bae11dd5b804156f5c64329e891ce
Author: Vladimir Klimontovich <v@klimontovich.com>
Date:   Wed Dec 2 19:00:23 2020 +0300

    Remove extra space

commit a0ef3a113a3796149071adfa535982ef76e34610
Merge: e82c63e05c 9b72323594
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Wed Dec 2 18:29:02 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit d402d68071e2ed4e9e9f8a0c81d5d4c9c9566b41
Author: Vladimir Klimontovich <v@klimontovich.com>
Date:   Wed Dec 2 18:26:16 2020 +0300

    Added HTTP CLI example and expose 8123 port

    Exposing the 8123 (HTTP) interface in basic example and giving CLI connection instruction makes it easier to start. For a lot of developers, curl is a native and very familiar tool

commit 12604ce0e64d2930021e783900d7c39289f3e7b0
Merge: 21bbc7bc19 5b3a7f5c98
Author: Anton Popov <pad11rus@gmail.com>
Date:   Tue Dec 1 14:52:30 2020 +0300

    Merge branch 'dev_map2' of git://github.com/hexiaoting/ClickHouse into merging-map

commit ccd794a676d2e9bc19367f6043ab81414011c9b6
Author: Azat Khuzhin <a3at.mail@gmail.com>
Date:   Sat Nov 28 19:32:28 2020 +0300

    Add remerge_sort_lowered_memory_bytes_ratio setting

    Sometimes default ratio is not good enough (2), since it depends on lots
    of factors:
    - size of LIMIT
    - max_memory_usage
    - ...

    So add a separate setting for it.

    But note that it make sense to set
    remerge_sort_lowered_memory_bytes_ratio only in range (1, 2]

    Since <= 1 will just use more CPU, and > 2 will use more RAM and will
    unlikely be faster.

commit f46eb6f2bed21808a23da2686d61a8b0db6da864
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Nov 29 18:16:47 2020 +0300

    Update tests

commit 7d79a3cc35859ae9adb92254f23f8138c72ba4a1
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Nov 29 18:13:57 2020 +0300

    Make insert_quorum_parallel by default

commit 2e6bedce91cd40a274faa04f42da4e7104c6e0da
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Nov 29 18:08:02 2020 +0300

    Relax "Too many parts" threshold, part 1

commit 5b3a7f5c98332506428ae75fab413396bd85da80
Merge: 74d875c837 747453b008
Author: hexiaoting <hewenting_ict@163.com>
Date:   Thu Nov 26 16:43:27 2020 +0800

    Merge remote-tracking branch 'origin/master' into dev_map2

commit 74d875c8374a631ef94ea97f802711694478ae01
Merge: c7c14cc6ad b494b18748
Author: hexiaoting <hewenting_ict@163.com>
Date:   Thu Nov 26 15:13:52 2020 +0800

    Merge remote-tracking branch 'origin/master' into dev_map2

commit e82c63e05cac931dd282ce0597c45385e01094c8
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Nov 24 15:28:54 2020 +0300

    fixes

commit dce0bf8efc025332b923a11d70cd392fc8937b0b
Merge: 5f6bbe8967 45f09fe293
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Nov 24 15:28:31 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit e5588b94aa960d0b7a900269d355d3ae5b17691f
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 20 13:38:53 2020 +0300

    Fix style and tests

commit 625962fb5950f222ebbf5dfb6a4cd430c94cb245
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 20 03:19:36 2020 +0300

    Remove debug logging

commit c3c32ff9f8c93a4a5b0f779b8e9161625a8f8ecb
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 20 03:16:57 2020 +0300

    Minor change

commit 608722b6ab143747d5ee5b8cdbc38a4eecf864a6
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 20 02:59:58 2020 +0300

    Fix integration tests

commit 3ee497258601649923e5b0a48b0ed3524e260dda
Merge: cc1b16ea81 4b7a7ca12a
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 20 01:23:51 2020 +0300

    Merge branch 'master' of github.com:ClickHouse/ClickHouse into optimize-data-on-insert

commit cc1b16ea815059f2f72a1290085cc8ddbe3543f0
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Wed Nov 18 13:40:58 2020 +0300

    Fix test

commit 547ec19fb35de4e5b599e07857b0a7d91a9c8c57
Merge: 4ed7657008 0bc60e2d53
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Wed Nov 18 12:01:59 2020 +0300

    Merge branch 'master' of github.com:ClickHouse/ClickHouse into optimize-data-on-insert

commit 4ed765700818c8735560f358fe545fe20e37656f
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Tue Nov 17 22:07:19 2020 +0300

    fix typo

commit 53b8ff42c2daac587b7bdef50d376dd216243690
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Tue Nov 17 19:56:36 2020 +0300

    Add test and set setting to 0 in some other tests

commit 437bbef5b9c52f51db02e7a28ce39dd90d593c2a
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Nov 9 21:56:49 2020 +0300

    Fix marker check

commit 1b9c20a5df3eb5e214e20f315e24047faa6b195e
Author: Maksim Kita <kitaetoya@gmail.com>
Date:   Mon Nov 9 14:36:44 2020 +0300

    ReadFloat 8digits unroll performance improvement

commit 264e9daf6f9b22a52f50a9696915e2c4f42c349b
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 13 11:16:30 2020 +0300

    Fix style

commit 6a57c0a8cf5217e90ef8c5daa6f006436856e033
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Fri Nov 13 10:54:05 2020 +0300

    Move merge in MergeTreeDataWriter

commit 8d5e0784d357f756c5f515e682a9ca317b16095d
Author: Pavel Kruglov <avogar@sandbox-633380738>
Date:   Thu Nov 12 23:29:36 2020 +0300

    Add setting optimize_on_insert

commit 9df414fa0c1feaf81c8f266b655b7063906b3e70
Author: alesapin <alesapin@gmail.com>
Date:   Tue Nov 10 13:54:56 2020 +0300

    One more test to skip

commit fe4bd6a7d23650c45be371c6a563b07ff45ce609
Merge: e2ece9ef0a 059357d51e
Author: alesapin <alesapin@gmail.com>
Date:   Tue Nov 10 13:54:43 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit c7c14cc6ad9723b04dbc95bd1c096f91a711c967
Author: hexiaoting <hewenting_ict@163.com>
Date:   Tue Nov 10 11:24:39 2020 +0800

    fix bug

commit e2ece9ef0a07ef59461d434e65c08bb691955459
Merge: 8fa6831185 24e57c659a
Author: alesapin <alesapin@gmail.com>
Date:   Mon Nov 9 13:24:47 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit ef0b78cf126f61a2f31d88119929e8544ef17b8a
Author: hexiaoting <hewenting_ict@163.com>
Date:   Fri Nov 6 17:21:26 2020 +0800

    style error fix

commit 28e87da9768af95ca8c50cb77880037dd0bc97f3
Merge: 66bca89001 ccf57300eb
Author: hexiaoting <hewenting_ict@163.com>
Date:   Fri Nov 6 15:07:41 2020 +0800

    remove map literal and Merge remote-tracking branch 'origin/master' into dev_map2

commit 8fa68311858d32b7fee312b0a4fbe84ff31b7677
Merge: 5622a3f57f 01898be459
Author: alesapin <alesapin@gmail.com>
Date:   Tue Nov 3 10:16:28 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 66bca89001a034b6cf1f02f99560e827ed486323
Author: hexiaoting <hewenting_ict@163.com>
Date:   Mon Nov 2 17:23:02 2020 +0800

    Style error fix

commit 5622a3f57fb753a259f8d187ef5661171b86e096
Author: alesapin <alesapin@gmail.com>
Date:   Mon Nov 2 10:46:43 2020 +0300

    Update skip list json

commit 3fe6f1e3fdfbf3fb57ab6330be857b05b4a3d8fa
Author: hexiaoting <hewenting_ict@163.com>
Date:   Mon Nov 2 15:10:29 2020 +0800

    Add test case for tuple convert to map

commit 3a70cf3773270fcb6e79001a5a011dedcc562ac8
Merge: e1ff6c3870 dc5fc9ddcb
Author: hexiaoting <hewenting_ict@163.com>
Date:   Mon Nov 2 14:46:56 2020 +0800

    Merge remote-tracking branch 'origin/master' into dev_map2

commit e1ff6c3870615360c452d71a4caf683cf149a49b
Author: hexiaoting <hewenting_ict@163.com>
Date:   Mon Nov 2 14:46:32 2020 +0800

    Update codes

commit 619fe6b10bb2879c009fc677d9e0e306c2c0d6ba
Merge: 2ba68d7494 ceab1add26
Author: hexiaoting <hewenting_ict@163.com>
Date:   Mon Nov 2 14:07:00 2020 +0800

    Merge remote-tracking branch 'hexiaoting/dev_map2' into dev_map2

commit 2ba68d74947a84d2f6008fa6a3bceab3bad0ed5b
Author: hexiaoting <hewenting_ict@163.com>
Date:   Mon Nov 2 14:05:53 2020 +0800

    Add map function

commit 4e45253c02ee5fbc4ccbd36714caf5fe7f5765d8
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 29 19:05:56 2020 +0300

    Update skip_list.json

commit b5553c0c2359e4db595228f8aca1eef7bc83ed99
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 29 15:48:02 2020 +0300

    Update skip_list.json

commit 8a6574767ae3b3e9d3f6da417c7b998389035ca1
Merge: 6345769b01 2e8594159c
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 29 12:57:56 2020 +0300

    Merge branch 'run_func_tests_in_parallel' of github.com:ClickHouse/ClickHouse into run_func_tests_in_parallel

commit 6345769b01a7c8fe9e617b13f01593af8d659427
Merge: 362b2c1cb2 745cb4ab2f
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 29 12:56:54 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 2e8594159cc927628e64c7c5d3b896e8f682148f
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 28 22:25:57 2020 +0300

    Update skip_list.json

commit 964149586eb9a42f4ebd8763637efc3054fd4c44
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 28 18:20:21 2020 +0300

    Update tests/queries/skip_list.json

    Co-authored-by: Azat Khuzhin <a3at.mail@gmail.com>

commit ceab1add26f4c12a40ab7c3b82f840d330293ade
Merge: 483be134b2 e0d85ffd1d
Author: hexiaoting <hewenting_ict@163.com>
Date:   Wed Oct 28 19:13:40 2020 +0800

    Merge pull request #1 from Enmk/ColumnMap_fixes

    Allow casting Tuple as Map.

commit 362b2c1cb225b59eb664aafef73dec63b8434e06
Merge: b9f63645b6 4a63f1271b
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 28 14:12:45 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit e0d85ffd1d36cd0fb87a5d450493bf155a2e3797
Author: Vasily Nemkov <V.Nemkov@gmail.com>
Date:   Mon Oct 26 21:20:45 2020 +0300

    Allow casting Tuple as Map.

    SELECT CAST(([1, 2, 3], ['1', '2', 'foo']), 'Map(UInt8, String)') AS map

commit 2e232bad74a82b54611d4e0da5d41733c4fab6d0
Merge: aa3a43c964 cc5f15da29
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 22:51:58 2020 +0300

    Merge branch 'master' into sigaltstack

commit aa3a43c9649c1458282601a1bd6ca0741fc29fe0
Merge: 0ba2fa8105 a591d1f32a
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 22:51:55 2020 +0300

    Merge branch 'sigaltstack' of github.com:yandex/ClickHouse into sigaltstack

commit a591d1f32afcb8e7eb0b69583523cc726ad1a719
Author: alexey-milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 22:46:25 2020 +0300

    Update ThreadStatus.cpp

commit b9f63645b647ac7778717d54a8cbd91b86020910
Author: tavplubix <avtokmakov@yandex-team.ru>
Date:   Mon Oct 26 12:45:30 2020 +0300

    Update 01107_atomic_db_detach_attach.sh

commit eab3371d5f5d43230d136305692d3f7f69fc4306
Author: tavplubix <avtokmakov@yandex-team.ru>
Date:   Mon Oct 26 12:42:54 2020 +0300

    Update 01114_database_atomic.sh

commit 0ba2fa81058b19148dcc002fb7445d11b517a3ed
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 11:56:30 2020 +0300

    Fix "Arcadia"

commit c1cce28470c4740bd5055d4321d2a14d4044769a
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 26 09:25:20 2020 +0300

    Update skip_list.json

commit 46fc6733e20bbab97da195d42a0c4455c952effb
Merge: b472bd35b7 18458f36e7
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 08:23:47 2020 +0300

    Merge branch 'master' into sigaltstack

commit b472bd35b70be33d48613efc6ec3c9976acc89e7
Merge: 88e846b36d bfe17d7dca
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 07:42:33 2020 +0300

    Merge branch 'master' into sigaltstack

commit 88e846b36d01a121ff50f11d96d42dbf25f1c706
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 07:42:22 2020 +0300

    Fix build on Mac OS

commit 068235c2d7a92143117f30607eb38d16b5ceb2b1
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 05:18:36 2020 +0300

    Disable alt stack for sanitizers

commit 1228396effa5d57da4e73820a6f7c0a969f1920c
Merge: 4e2c0ba226 772faeb74f
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Mon Oct 26 05:09:33 2020 +0300

    Merge branch 'master' into sigaltstack

commit 4e2c0ba226f29f1aecab837aa2996e71222086f5
Author: Alexey Milovidov <milovidov@yandex-team.ru>
Date:   Sun Oct 25 03:10:05 2020 +0300

    Provide diagnostics on stack overflow

commit 96a10f3664257100621edb0cb789fe6b04123d66
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 23 15:19:03 2020 +0300

    Update skip_list.json

commit bbc421acb749606fa51efb1629ea2c37bcdea473
Merge: b158a22a4f 6285fd1378
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 23 13:22:34 2020 +0300

    Merge branch 'run_func_tests_in_parallel' of github.com:ClickHouse/ClickHouse into run_func_tests_in_parallel

commit b158a22a4f5743545c0bcabc3536a7d440e54174
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 23 13:21:40 2020 +0300

    Bump CI

commit 483be134b2718211d155f1ef72ce4eeda942c2f0
Merge: 521b07d37a 0082c56b67
Author: hexiaoting <hewenting_ict@163.com>
Date:   Fri Oct 23 16:36:17 2020 +0800

    Support MergeTree Engine

commit 6285fd13783b1c587da7d7002cd772089d511991
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 22 17:41:34 2020 +0300

    Update skip_list.json

commit d04edef048cb079387301be672b8ba1c35c91e3b
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 22 10:20:45 2020 +0300

    More tests to skip

commit fedf169bb6259dc0b0d94d082a963947376ae3fe
Merge: 96daa31436 0082c56b67
Author: alesapin <alesapin@gmail.com>
Date:   Thu Oct 22 10:19:48 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 96daa31436649261089fdb432389160681a8d03a
Merge: 620c8e3d38 4261455621
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 21 22:53:18 2020 +0300

    Merge branch 'run_func_tests_in_parallel' of github.com:ClickHouse/ClickHouse into run_func_tests_in_parallel

commit 620c8e3d3813a93cc433575a271488c6ae278268
Merge: cc43b002cc 6a57bc8c0b
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 21 22:52:30 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 4261455621270c2039c8318492f3ef40e6defb95
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 21 17:18:48 2020 +0300

    Update skip_list.json

commit cc43b002cc4f3c34d70243bb39b846b93700b345
Merge: df4620b7bd c53f59dece
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 21 12:51:43 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit df4620b7bd38ac87134358696e718c8dc9197457
Merge: 85ca6f7e6a b4f0e08369
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 20 17:15:56 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 521b07d37a8393346b914d384c9d02e5ed25b2ae
Author: hexiaoting <hewenting_ict@163.com>
Date:   Sun Oct 18 20:40:51 2020 +0800

    Fix build error

commit fcd35b06193a251eca24cf0ebf7d16840d9a37df
Merge: db8ba39c9c 6dc5cb166f
Author: hexiaoting <hewenting_ict@163.com>
Date:   Sun Oct 18 16:18:13 2020 +0800

    Merge branch 'master' into dev_map2

commit db8ba39c9c622d89bf29739757c507ab66376534
Author: hexiaoting <hewenting_ict@163.com>
Date:   Sat Oct 17 23:45:35 2020 +0800

    Implement Integer And Array for value type

commit 85ca6f7e6a08ee6d6008589059a64e25b8027bf7
Merge: 8f3d6dae2d e89a3b5d09
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 16 13:43:30 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 8f3d6dae2d88113e0e5070a5a7589ddd56751c53
Merge: 09bb6e9944 0fa35cdc80
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 16 10:26:57 2020 +0300

    Merge branch 'run_func_tests_in_parallel' of github.com:ClickHouse/ClickHouse into run_func_tests_in_parallel

commit 09bb6e99445883cd4ac41dadd36f47fb21b22358
Merge: 9a235cfb8b 0bb4480fee
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 16 10:26:33 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit cb9e88efa327da17d9d7a27138c622a689bbb646
Author: hexiaoting <hewenting_ict@163.com>
Date:   Wed Oct 14 15:35:22 2020 +0800

    Update with fresh master to build sucessfully.

commit 4f3f22198723733feb426f4111ef4a00b6ef8a33
Author: hexiaoting <hewenting_ict@163.com>
Date:   Tue Oct 13 15:02:03 2020 +0800

    Fix build error

commit 03102fead3fba3d6d26a2c5bc102028082668a1e
Author: hexiaoting <hewenting_ict@163.com>
Date:   Sun Oct 11 22:22:54 2020 +0800

    Fix style check error.

commit 974352a8e1d6b2f90c9a97b2598fa91daff9b49d
Author: hexiaoting <hewenting_ict@163.com>
Date:   Sat Oct 10 18:16:05 2020 +0800

    Fix style check errors.

commit cdb4dd57a63c5dc7855ba3eb6d4b471da329f993
Author: hexiaoting <hewenting_ict@163.com>
Date:   Sat Oct 10 14:49:03 2020 +0800

    Add new DataType Map(key,value)

commit 0fa35cdc8004a4e813405010ae4234d2edcfc8d1
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Oct 13 19:03:25 2020 +0300

    use async drop in some tests

commit 9a235cfb8bdff69169fcf7b9ddce669173248f57
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 13 17:29:24 2020 +0300

    Database atomic sync drop detach

commit 7d735b4725f430b19ca57591ff0557207c92771d
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 13 13:40:00 2020 +0300

    Update max_concurrent_queries.xml

commit 2909378bbc8c0de0be303abe7d473869499c77d1
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 13 09:52:37 2020 +0300

    Bump CI

commit 97e89afea7b934d108c90adf10dd0273b7762562
Merge: d75d486029 38c7132c0f
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 12 22:40:57 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit d75d486029beed8795f46cfe31539ed99afab76c
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 12 13:20:52 2020 +0300

    Update skip_list.json

commit 2a63a619cc5d4be8fd179170def377718d589204
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 9 22:42:51 2020 +0300

    Update skip_list.json

commit e112fbe2a972cfc2b8563f331a7d2287528b5d96
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 9 14:51:34 2020 +0300

    Update skip_list.json

commit cd28dab66624b64466b1c66e6addfcc377e909b7
Merge: 7b3d430149 e6c6930397
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 9 10:26:40 2020 +0300

    Merge branch 'run_func_tests_in_parallel' of github.com:ClickHouse/ClickHouse into run_func_tests_in_parallel

commit 7b3d4301494bd5fd65346c12cfc9dcda1d6e6aa8
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 9 10:25:50 2020 +0300

    Add more tests

commit 2b66180c704100ad9ef551358719b276fbe357c7
Merge: 9982c17814 66b7748b63
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 9 10:25:35 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit e6c6930397f2987d9c23cea4b1fcb8cf417b538b
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 7 23:01:03 2020 +0300

    Update skip_list.json

commit e7d1d34a605b49d46aee37b0676ad4952399862d
Author: alesapin <alesapin@gmail.com>
Date:   Wed Oct 7 12:43:49 2020 +0300

    Update skip_list.json

commit d2762624f5abd1ac37e479475f5357a486eb2809
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 6 22:45:17 2020 +0300

    Update skip_list.json

commit 9982c17814c2e705769efc80989b54bf785a4a6e
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 6 19:02:33 2020 +0300

    Skip more

commit 4f6d794816a50feb8cfc897c6c52e7145970e269
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 6 14:41:35 2020 +0300

    Trying 8

commit a5f843976136935b110bde7bb1ebb8ec3a582f68
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 6 12:18:49 2020 +0300

    Trying 12 threads

commit 4330fc0f39232559a69d4cdb91a1c06f36c8cb6d
Merge: a257751924 a49591b250
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 6 10:26:55 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit a257751924ea4fcbbb9b6eebcb35a39ae8a92f5a
Author: alesapin <alesapin@gmail.com>
Date:   Tue Oct 6 10:13:55 2020 +0300

    Ignore max concurrent queries in tests

commit 99ea5cca1014aee22bcc31cac14bf3ab6a634eea
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 23:30:14 2020 +0300

    Bump

commit 766802ae4d20254c2e2f607620fbb3df0e52d7c3
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 22:34:50 2020 +0300

    Bump

commit 2f9a6fe4791fb6efd384aa0ecb902aeb0787748c
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 18:49:40 2020 +0300

    More tests to skip

commit b2f2a274c03c5d422b949bd239771ab2cc1bf8c4
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 16:08:39 2020 +0300

    One more time

commit 8561a116200704c5b2489c30208209df7834021b
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 12:56:48 2020 +0300

    Revert "Better fix"

    This reverts commit 0496987c8fe113b3716551cefb37a654597887d2.

commit 0496987c8fe113b3716551cefb37a654597887d2
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 12:56:20 2020 +0300

    Better fix

commit 916cef04872af698d0ec7400d10aec9af13e547b
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 12:45:11 2020 +0300

    Bump CI

commit edd5a4b9362950cd21b87f65dfc7f0813d0cc29d
Author: alesapin <alesapin@gmail.com>
Date:   Mon Oct 5 10:15:08 2020 +0300

    Less tests

commit 933999b4aecf8af38dfa06aa12336444dc7c3cb0
Author: alesapin <alesapin@gmail.com>
Date:   Sun Oct 4 11:37:11 2020 +0300

    less tests

commit 7ccf1cc3162abe1c58e72399a0f61d8e09afd950
Author: alesapin <alesapin@gmail.com>
Date:   Sun Oct 4 11:36:34 2020 +0300

    Less tests

commit 744eb8a2e4bfb01b954441c2add1ffb5d19e681a
Author: alesapin <alesapin@gmail.com>
Date:   Sat Oct 3 16:10:10 2020 +0300

    Trying 6

commit 7cef000b7251bf0040500649f69f232bade1d218
Author: alesapin <alesapin@gmail.com>
Date:   Sat Oct 3 10:53:02 2020 +0300

    More tests to skip

commit 182f5105657956262d6c7ac7fd290698eb91a1fb
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 2 21:58:11 2020 +0300

    Try 8 threads

commit d8f7dce28698e8d869fe8d81083ef1fa2a3a8539
Merge: 626e9a82cf 77c9f7af79
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 2 21:45:43 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 5f6bbe896703fa792638d40bfada8b665ebc2c8a
Merge: 32776da69a 9cb3c743bd
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Fri Oct 2 20:14:08 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit 32776da69a67fa29bad7529d520c7f252ed693c0
Merge: aa5e5013f2 404c199448
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Fri Oct 2 20:09:37 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit 626e9a82cf88f910e12d9e87eb414456e2f37ecf
Merge: 91ba9c2e5f 7d75866722
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 2 17:34:20 2020 +0300

    Merge branch 'fix_ilike' into run_func_tests_in_parallel

commit 91ba9c2e5fd95338bcd332d5937afe37c9e87838
Merge: 76e37a6406 c0d1416bbd
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 2 10:44:34 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 76e37a64064b6a2e13d93057bb11d2e6a7ae89f4
Author: alesapin <alesapin@gmail.com>
Date:   Fri Oct 2 10:43:21 2020 +0300

    More tests to skip

commit aa5e5013f2be29ec2121740df3120e88a66f45f2
Merge: 1c37ac7c9d bbbe51033d
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Sun Sep 27 14:11:53 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit 2452b400938868fd1daefb7cf2fe7d5b1061c330
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 22:44:13 2020 +0300

    Skip more tests

commit 7c1c548b43f2f1c58c702822da6e903c78df9f9c
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 22:43:35 2020 +0300

    Disable more tests

commit 4ee1621ebf0b1169121092802bd84fb0df722a47
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 19:55:05 2020 +0300

    Disable all tests with attach

commit 6669b7da2976de14d10579b52f944894732f6b80
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 19:41:09 2020 +0300

    More parallel tests

commit bedab199af266e8652b7f202961f149ac2bd98c2
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 16:20:16 2020 +0300

    Fix category name

commit 66be1580fdbef72cc63b63a5a5193e1853aa361a
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 16:18:08 2020 +0300

    Correctly merge with master

commit 987a46faee2789c62c76b42263214ee4a2137872
Merge: 8e9013561d 8f21c73b76
Author: alesapin <alesapin@gmail.com>
Date:   Fri Sep 25 16:16:42 2020 +0300

    Merge branch 'master' into run_func_tests_in_parallel

commit 8e9013561d1cd1f98bf042625fbaeba14e62725b
Author: alesapin <alesapin@gmail.com>
Date:   Thu Sep 24 17:54:10 2020 +0300

    Trying to split tests in sequential and parallel

commit 1c37ac7c9d7bce591b150725d9d3114180376f78
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Thu Sep 17 21:19:02 2020 +0300

    fix

commit 41e99cf261719df7977ae8fd62d94abc4aa046f3
Merge: 18e48838bb 9b6c62e82b
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Thu Sep 17 20:28:50 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit 18e48838bb136c2297fc30d256203e0832a4fa5d
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Sep 15 17:22:20 2020 +0300

    fix build

commit 5455068226399a8c1065022ac3ee32608f0a254a
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Sep 15 17:17:30 2020 +0300

    try fix flacky test

commit 0ecf8e595f40205259524482aebdcfc0170d8735
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Sep 15 16:30:30 2020 +0300

    cleanup code, add test

commit 3671c91c3768dc0f514a514c1b9e8f20493a5fa9
Merge: a812f7eb75 1d8ee50769
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Tue Sep 15 12:43:01 2020 +0300

    Merge branch 'master' into allow_atomic_database_inside_materialize_mysql

commit a812f7eb756fe768932c57b55c6ec0a2e0f7c0c6
Author: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date:   Mon Sep 14 22:25:02 2020 +0300

    allow using Atomic database inside MaterializeMySQL
---
 .gitmodules                                   |   2 +-
 CHANGELOG.md                                  | 123 ++++
 base/common/CMakeLists.txt                    |   1 +
 base/common/ReadlineLineReader.cpp            |   5 +-
 base/common/ReplxxLineReader.cpp              |  10 +-
 base/common/getMemoryAmount.cpp               |   3 +-
 base/common/getPageSize.cpp                   |   8 +
 base/common/getPageSize.h                     |   6 +
 base/common/setTerminalEcho.cpp               |   5 +-
 base/common/ya.make                           |   1 +
 cmake/sanitize.cmake                          |  21 +-
 contrib/boost                                 |   2 +-
 contrib/krb5                                  |   2 +-
 contrib/libc-headers                          |   2 +-
 contrib/libgsasl                              |   2 +-
 contrib/libhdfs3                              |   2 +-
 contrib/libhdfs3-cmake/CMakeLists.txt         |  17 +-
 contrib/librdkafka                            |   2 +-
 contrib/libunwind                             |   2 +-
 debian/control                                |   2 +-
 docker/builder/Dockerfile                     |   4 +-
 docker/builder/build.sh                       |   2 +-
 docker/images.json                            |  11 +-
 docker/packager/binary/build.sh               |   1 +
 docker/packager/packager                      |   4 +
 docker/server/README.md                       |  66 ++
 docker/test/base/Dockerfile                   |   2 +-
 .../integration/kerberized_hadoop/Dockerfile  |  18 +
 docker/test/integration/runner/Dockerfile     |   5 +-
 .../runner/compose/docker_compose_hdfs.yml    |   1 +
 .../docker_compose_kerberized_hdfs.yml        |  29 +
 .../zzz-perf-comparison-tweaks-config.xml     |   1 +
 docker/test/stateful_with_coverage/Dockerfile |   2 -
 docs/en/development/build-osx.md              |   4 +-
 docs/en/development/developer-instruction.md  |   4 +-
 .../table-engines/integrations/hdfs.md        |  89 +++
 .../getting-started/example-datasets/index.md |   1 +
 .../example-datasets/metrica.md               |  16 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/en/getting-started/tutorial.md           |   4 +-
 docs/en/interfaces/cli.md                     |   2 +-
 docs/en/introduction/performance.md           |   2 +-
 docs/en/operations/performance-test.md        |   2 +-
 docs/en/operations/settings/settings.md       |  11 +-
 .../system-tables/distribution_queue.md       |   4 +
 .../system-tables/replication_queue.md        |  81 ++
 docs/en/operations/tips.md                    |  17 +
 .../aggregate-functions/combinators.md        |   4 +
 .../data-types/lowcardinality.md              |   2 +
 .../functions/array-functions.md              |  18 +
 .../functions/type-conversion-functions.md    |  57 ++
 docs/en/sql-reference/operators/in.md         |  22 +
 .../statements/select/order-by.md             | 180 ++++-
 docs/en/sql-reference/statements/system.md    |   2 +-
 docs/es/development/developer-instruction.md  |   4 +-
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/es/getting-started/tutorial.md           |   4 +-
 docs/es/interfaces/cli.md                     |   2 +-
 docs/es/operations/performance-test.md        |   2 +-
 docs/fa/development/developer-instruction.md  |   4 +-
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/fa/getting-started/tutorial.md           |   4 +-
 docs/fa/operations/performance-test.md        |   2 +-
 docs/fr/development/developer-instruction.md  |   4 +-
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/fr/getting-started/tutorial.md           |   4 +-
 docs/fr/operations/performance-test.md        |   2 +-
 docs/ja/development/developer-instruction.md  |   4 +-
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/ja/getting-started/tutorial.md           |   4 +-
 docs/ja/interfaces/cli.md                     |   2 +-
 docs/ja/operations/performance-test.md        |   2 +-
 .../data-types/lowcardinality.md              |   1 +
 docs/ru/development/developer-instruction.md  |   4 +-
 docs/ru/faq/general/index.md                  |  25 +
 .../faq/{general.md => general/mapreduce.md}  |  52 +-
 docs/ru/faq/index.md                          |  11 +
 docs/ru/faq/integration/file-export.md        |  37 +
 docs/ru/faq/integration/index.md              |  19 +
 docs/ru/faq/integration/oracle-odbc.md        |  15 +
 docs/ru/faq/operations/index.md               |  18 +
 docs/ru/faq/use-cases/index.md                |  14 +
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/ru/getting-started/tutorial.md           |   4 +-
 docs/ru/interfaces/cli.md                     |   2 +-
 .../system-tables/distribution_queue.md       |  46 ++
 .../system-tables/replication_queue.md        |  81 ++
 .../data-types/lowcardinality.md              |   2 +
 .../external-dicts-dict-sources.md            |   2 +-
 .../functions/date-time-functions.md          |  12 +
 .../statements/select/order-by.md             | 180 ++++-
 docs/tr/development/developer-instruction.md  |   4 +-
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/tr/getting-started/tutorial.md           |   4 +-
 docs/tr/interfaces/cli.md                     |   2 +-
 docs/tr/operations/performance-test.md        |   2 +-
 .../data-types/lowcardinality.md              |   1 +
 docs/zh/development/developer-instruction.md  |   4 +-
 .../example-datasets/metrica.md               |  10 +-
 .../example-datasets/nyc-taxi.md              |   2 +-
 .../example-datasets/ontime.md                |   2 +-
 docs/zh/getting-started/tutorial.md           |   4 +-
 docs/zh/operations/performance-test.md        |   2 +-
 docs/zh/sql-reference/statements/misc.md      |  48 +-
 .../sql-reference/statements/select/limit.md  |  10 +-
 programs/benchmark/Benchmark.cpp              |   4 +-
 programs/client/Client.cpp                    |   2 +-
 programs/copier/ClusterCopier.cpp             |  15 +-
 programs/copier/ClusterCopier.h               |   2 +-
 programs/copier/ClusterCopierApp.cpp          |   1 +
 programs/copier/TaskCluster.h                 |   5 +-
 programs/odbc-bridge/ColumnInfoHandler.cpp    |   2 +-
 programs/server/Server.cpp                    |  68 +-
 programs/server/config.d/test_keeper_port.xml |   1 +
 programs/server/config.xml                    | 189 +++--
 programs/server/play.html                     |  17 +
 src/Access/ContextAccess.cpp                  | 528 ++++++-------
 src/Access/ContextAccess.h                    |  56 +-
 src/Access/SettingsConstraints.cpp            |  19 +
 src/AggregateFunctions/AggregateFunctionAvg.h |   2 +-
 .../AggregateFunctionFactory.cpp              |  12 +-
 .../AggregateFunctionSimpleState.cpp          |  32 +
 .../AggregateFunctionSimpleState.h            |  77 ++
 .../registerAggregateFunctions.cpp            |   2 +
 src/AggregateFunctions/ya.make                |   1 +
 src/CMakeLists.txt                            |   8 +-
 src/Client/Connection.cpp                     |   1 +
 src/Client/Connection.h                       |   6 +-
 src/Client/ConnectionPool.h                   |   2 +-
 src/Client/ConnectionPoolWithFailover.cpp     |   1 +
 src/Columns/ColumnMap.cpp                     | 241 ++++++
 src/Columns/ColumnMap.h                       |  92 +++
 src/Columns/ya.make                           |   1 +
 src/Common/Allocator.h                        |   5 +-
 src/Common/Arena.h                            |  10 +-
 src/Common/ArrayCache.h                       |   5 +-
 src/Common/Config/ConfigProcessor.cpp         |   4 +-
 src/Common/CurrentMetrics.cpp                 |   1 +
 src/Common/ErrorCodes.cpp                     |   1 +
 src/Common/FieldVisitors.cpp                  | 102 +++
 src/Common/FieldVisitors.h                    |  33 +
 src/Common/HashTable/HashTable.h              | 120 ++-
 src/Common/IFactoryWithAliases.h              |   5 +
 src/Common/MemoryStatisticsOS.cpp             |  15 +-
 src/Common/SpaceSaving.h                      |   1 +
 src/Common/StringSearcher.h                   |   3 +-
 src/Common/ThreadProfileEvents.cpp            |  15 +-
 src/Common/ThreadStatus.cpp                   |  48 ++
 src/Common/UInt128.h                          | 127 ++--
 src/Common/XDBCBridgeHelper.h                 |   1 +
 src/Common/ZooKeeper/TestKeeperStorage.cpp    |  15 +
 src/Common/config.h.in                        |   1 +
 src/Common/tests/CMakeLists.txt               |   3 -
 src/Common/tests/gtest_hash_table.cpp         | 319 ++++++++
 src/Common/tests/hash_table.cpp               |  50 --
 src/Common/ya.make                            |   4 +-
 src/Common/ya.make.in                         |   4 +-
 src/Core/AccurateComparison.h                 |  29 +-
 src/Core/DecimalFunctions.h                   |  49 +-
 src/Core/Field.cpp                            | 378 +++-------
 src/Core/Field.h                              |  24 +-
 src/Core/Settings.cpp                         |   2 -
 src/Core/Settings.h                           |  10 +-
 src/Core/SortCursor.h                         |  39 +-
 src/Core/Types.h                              |   2 +
 .../MergingSortedBlockInputStream.cpp         |   2 +-
 src/DataStreams/RemoteBlockInputStream.cpp    |  12 +-
 src/DataStreams/RemoteBlockInputStream.h      |  12 +-
 src/DataStreams/RemoteQueryExecutor.cpp       |  19 +-
 src/DataStreams/RemoteQueryExecutor.h         |  15 +-
 .../DataTypeCustomSimpleAggregateFunction.cpp |  22 +-
 .../DataTypeCustomSimpleAggregateFunction.h   |   1 +
 src/DataTypes/DataTypeFactory.cpp             |   1 +
 src/DataTypes/DataTypeFactory.h               |   1 +
 src/DataTypes/DataTypeMap.cpp                 | 375 ++++++++++
 src/DataTypes/DataTypeMap.h                   | 103 +++
 src/DataTypes/DataTypesDecimal.h              | 134 +++-
 src/DataTypes/FieldToDataType.cpp             |  19 +
 src/DataTypes/FieldToDataType.h               |   1 +
 src/DataTypes/IDataType.h                     |   4 +
 src/DataTypes/ya.make                         |   1 +
 src/Databases/DatabaseAtomic.cpp              |  12 +-
 src/Databases/DatabaseAtomic.h                |   8 +-
 src/Databases/DatabaseFactory.cpp             |  33 +-
 src/Databases/DatabaseLazy.cpp                |   8 +
 src/Databases/DatabaseOnDisk.cpp              |  97 ++-
 src/Databases/DatabaseOnDisk.h                |  10 +
 src/Databases/DatabaseOrdinary.cpp            |  13 +
 src/Databases/DatabasesCommon.cpp             |  10 +-
 src/Databases/IDatabase.h                     |  13 +
 .../MySQL/DatabaseConnectionMySQL.cpp         |   7 +-
 src/Databases/MySQL/DatabaseConnectionMySQL.h |   2 +
 .../MySQL/DatabaseMaterializeMySQL.cpp        | 266 ++++---
 .../MySQL/DatabaseMaterializeMySQL.h          |  49 +-
 .../MySQL/DatabaseMaterializeTablesIterator.h |   6 +-
 .../MySQL/MaterializeMySQLSyncThread.cpp      |  14 +-
 src/Dictionaries/HTTPDictionarySource.cpp     |   1 +
 src/Dictionaries/XDBCDictionarySource.cpp     |   1 +
 src/Disks/S3/DiskS3.cpp                       |  19 +-
 src/Disks/S3/DiskS3.h                         |   4 +-
 src/Disks/S3/registerDiskS3.cpp               |   2 +-
 src/Formats/FormatFactory.h                   |   1 -
 src/Functions/FunctionBase64Conversion.h      |   4 +-
 src/Functions/FunctionBinaryArithmetic.h      | 266 +++++--
 src/Functions/FunctionFactory.cpp             |  11 +-
 src/Functions/FunctionFactory.h               |   1 -
 src/Functions/FunctionsConversion.cpp         |  11 +-
 src/Functions/FunctionsConversion.h           | 697 +++++++++++++-----
 src/Functions/array/arrayAggregation.cpp      | 318 ++++++++
 src/Functions/array/arrayElement.cpp          | 279 ++++++-
 src/Functions/array/arrayEnumerateRanked.h    |   4 +-
 src/Functions/array/arrayIntersect.cpp        |   8 +-
 src/Functions/array/arraySum.cpp              | 146 ----
 .../{encodeXML.cpp => encodeXMLComponent.cpp} |  11 +-
 src/Functions/map.cpp                         | 140 ++++
 src/Functions/multiply.cpp                    |  11 +-
 src/Functions/registerFunctions.cpp           |   2 +
 .../registerFunctionsHigherOrder.cpp          |   4 +-
 src/Functions/registerFunctionsString.cpp     |   4 +-
 src/Functions/tcpPort.cpp                     |   1 +
 src/Functions/toFixedString.h                 |  61 +-
 src/Functions/toModifiedJulianDay.cpp         |  11 +-
 src/Functions/trap.cpp                        |  11 +-
 src/Functions/ya.make                         |   5 +-
 src/IO/ConnectionTimeouts.h                   |  27 +-
 src/IO/ConnectionTimeoutsContext.h            |  30 +
 src/IO/HDFSCommon.cpp                         |  62 --
 src/IO/HDFSCommon.h                           |  58 --
 src/IO/MMapReadBufferFromFileDescriptor.cpp   |   5 +-
 src/IO/S3Common.cpp                           |   9 +-
 src/IO/WriteBufferFromS3.cpp                  | 183 ++---
 src/IO/WriteBufferFromS3.h                    |  32 +-
 src/IO/WriteHelpers.h                         |  21 +-
 src/IO/readFloatText.h                        |  38 +-
 src/IO/tests/o_direct_and_dirty_pages.cpp     |  10 +-
 src/IO/ya.make                                |   2 +
 src/IO/ya.make.in                             |   2 +
 src/Interpreters/ActionsDAG.cpp               |   4 +-
 src/Interpreters/Cluster.cpp                  |   1 +
 src/Interpreters/Cluster.h                    |  12 +-
 .../ClusterProxy/IStreamFactory.h             |   2 +-
 .../ClusterProxy/SelectStreamFactory.cpp      |  14 +-
 .../ClusterProxy/SelectStreamFactory.h        |   2 +-
 .../ClusterProxy/executeQuery.cpp             |   9 +-
 src/Interpreters/ClusterProxy/executeQuery.h  |   2 +-
 src/Interpreters/Context.cpp                  |   2 +
 src/Interpreters/DDLWorker.cpp                |  22 +
 src/Interpreters/DDLWorker.h                  |   3 +
 src/Interpreters/DatabaseCatalog.cpp          |  37 +-
 src/Interpreters/ExpressionActions.h          |   1 -
 src/Interpreters/ExpressionAnalyzer.cpp       |   5 +
 src/Interpreters/ExpressionAnalyzer.h         |   7 +-
 .../IInterpreterUnionOrSelectQuery.h          |   2 +
 src/Interpreters/InterpreterCreateQuery.cpp   |  50 +-
 src/Interpreters/InterpreterDropQuery.cpp     |  48 +-
 src/Interpreters/InterpreterInsertQuery.cpp   |   1 +
 src/Interpreters/InterpreterSelectQuery.cpp   |   9 +-
 .../InterpreterSelectWithUnionQuery.cpp       | 142 +++-
 .../InterpreterShowTablesQuery.cpp            |  20 +
 src/Interpreters/MergeJoin.cpp                |  24 +-
 src/Interpreters/Set.cpp                      |  21 +-
 src/Interpreters/castColumn.cpp               |  36 +-
 src/Interpreters/castColumn.h                 |   5 +-
 src/Interpreters/convertFieldToType.cpp       |   1 +
 src/Interpreters/loadMetadata.cpp             |   1 +
 .../tests/gtest_cycle_aliases.cpp             |   1 +
 src/Interpreters/tests/internal_iotop.cpp     |   4 +-
 src/Parsers/ASTDropQuery.cpp                  |   3 +
 src/Parsers/ASTDropQuery.h                    |   3 +
 src/Parsers/ASTFunction.cpp                   |  14 +
 src/Parsers/ASTShowTablesQuery.cpp            |   6 +
 src/Parsers/ASTShowTablesQuery.h              |   2 +
 src/Parsers/ExpressionElementParsers.cpp      |   6 +
 src/Parsers/ExpressionElementParsers.h        |  12 +
 src/Parsers/ExpressionListParsers.cpp         |   1 +
 src/Parsers/IParser.h                         |   1 -
 src/Parsers/ParserCreateQuery.cpp             |  12 +-
 src/Parsers/ParserDropQuery.cpp               |  70 +-
 src/Parsers/ParserDropQuery.h                 |   2 +-
 src/Parsers/ParserExplainQuery.cpp            |   9 +-
 src/Parsers/ParserExplainQuery.h              |   4 +
 src/Parsers/ParserQuery.cpp                   |   2 +-
 src/Parsers/ParserQueryWithOutput.cpp         |   2 +-
 src/Parsers/ParserQueryWithOutput.h           |   3 +
 src/Parsers/ParserSetQuery.cpp                |   1 +
 src/Parsers/ParserSetQuery.h                  |   2 +
 src/Parsers/ParserShowTablesQuery.cpp         |  25 +
 src/Parsers/tests/select_parser.cpp           |   2 +-
 src/Parsers/tests/select_parser_fuzzer.cpp    |   2 +-
 .../Impl/ConstantExpressionTemplate.cpp       |  33 +-
 .../Algorithms/AggregatingSortedAlgorithm.cpp |   6 +-
 .../Algorithms/CollapsingSortedAlgorithm.cpp  |   6 +-
 .../Algorithms/CollapsingSortedAlgorithm.h    |   6 +-
 .../GraphiteRollupSortedAlgorithm.cpp         |   8 +-
 .../Merges/Algorithms/IMergingAlgorithm.h     |   2 +
 .../IMergingAlgorithmWithDelayedChunk.cpp     |   4 +-
 .../IMergingAlgorithmWithSharedChunks.cpp     |   4 +-
 .../Algorithms/MergingSortedAlgorithm.cpp     |   2 +-
 .../Algorithms/MergingSortedAlgorithm.h       |   6 +-
 .../Algorithms/ReplacingSortedAlgorithm.cpp   |   2 +-
 src/Processors/Merges/Algorithms/RowRef.h     |   4 +-
 .../Algorithms/SummingSortedAlgorithm.cpp     |   4 +-
 .../VersionedCollapsingAlgorithm.cpp          |   2 +-
 .../Merges/CollapsingSortedTransform.h        |   4 +-
 src/Processors/QueryPlan/MergeSortingStep.cpp |   3 +
 src/Processors/QueryPlan/MergeSortingStep.h   |   2 +
 .../Transforms/MergeSortingTransform.cpp      |  10 +-
 .../Transforms/MergeSortingTransform.h        |   2 +
 .../Transforms/SortingTransform.cpp           |   2 +-
 src/Server/HTTPHandler.cpp                    |   1 +
 src/Server/InterserverIOHTTPHandler.cpp       |   1 +
 src/Server/StaticRequestHandler.cpp           |   1 +
 src/Server/TestKeeperTCPHandler.cpp           |  42 +-
 src/Storages/AlterCommands.cpp                |   4 +-
 src/Storages/Distributed/DirectoryMonitor.cpp |   1 +
 .../DistributedBlockOutputStream.cpp          |   1 +
 src/Storages/HDFS/HDFSCommon.cpp              | 194 +++++
 src/Storages/HDFS/HDFSCommon.h                | 114 +++
 .../HDFS}/ReadBufferFromHDFS.cpp              |  20 +-
 .../HDFS}/ReadBufferFromHDFS.h                |   9 +-
 src/Storages/{ => HDFS}/StorageHDFS.cpp       |  14 +-
 src/Storages/{ => HDFS}/StorageHDFS.h         |   0
 .../HDFS}/WriteBufferFromHDFS.cpp             |  17 +-
 .../HDFS}/WriteBufferFromHDFS.h               |   2 +-
 src/Storages/Kafka/KafkaSettings.h            |   2 +-
 .../MergeTree/IMergeTreeDataPartWriter.cpp    |  60 +-
 .../MergeTree/IMergeTreeDataPartWriter.h      |  60 +-
 .../MergeTree/IMergedBlockOutputStream.cpp    |  19 -
 .../MergeTree/IMergedBlockOutputStream.h      |   2 -
 src/Storages/MergeTree/KeyCondition.cpp       |   2 +-
 .../MergeTree/MergeTreeBlockOutputStream.cpp  |  12 +-
 .../MergeTree/MergeTreeBlockOutputStream.h    |   5 +-
 src/Storages/MergeTree/MergeTreeData.cpp      |  24 +-
 src/Storages/MergeTree/MergeTreeData.h        |   1 -
 .../MergeTree/MergeTreeDataMergerMutator.cpp  |  10 +
 .../MergeTreeDataPartWriterCompact.cpp        | 190 +++--
 .../MergeTreeDataPartWriterCompact.h          |  27 +-
 .../MergeTreeDataPartWriterInMemory.cpp       |  12 +-
 .../MergeTreeDataPartWriterInMemory.h         |   9 +-
 .../MergeTreeDataPartWriterOnDisk.cpp         | 134 ++--
 .../MergeTree/MergeTreeDataPartWriterOnDisk.h |  92 ++-
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 372 +++++++---
 .../MergeTree/MergeTreeDataPartWriterWide.h   |  68 +-
 .../MergeTree/MergeTreeDataWriter.cpp         | 137 +++-
 src/Storages/MergeTree/MergeTreeDataWriter.h  |   4 +-
 src/Storages/MergeTree/MergeTreeIOSettings.h  |   3 +
 .../MergeTree/MergeTreeIndexGranularity.cpp   |  12 +
 .../MergeTree/MergeTreeIndexGranularity.h     |   8 +-
 src/Storages/MergeTree/MergeTreeSettings.cpp  |   1 +
 .../MergeTree/MergedBlockOutputStream.cpp     |  21 +-
 .../MergedColumnOnlyOutputStream.cpp          |  21 +-
 .../MergeTree/MergedColumnOnlyOutputStream.h  |   2 +-
 .../ReplicatedMergeTreeBlockOutputStream.cpp  |  16 +-
 .../ReplicatedMergeTreeBlockOutputStream.h    |   5 +-
 src/Storages/MergeTree/SimpleMergeSelector.h  |  77 ++
 src/Storages/RabbitMQ/StorageRabbitMQ.cpp     |   1 +
 src/Storages/RabbitMQ/StorageRabbitMQ.h       |   3 +-
 src/Storages/SetSettings.h                    |   3 +-
 src/Storages/StorageDictionary.cpp            |   4 +-
 src/Storages/StorageJoin.cpp                  |  22 +-
 src/Storages/StorageJoin.h                    |   2 +-
 src/Storages/StorageMaterializeMySQL.cpp      |   7 +-
 src/Storages/StorageMaterializeMySQL.h        |  10 +-
 src/Storages/StorageMemory.cpp                |  96 ++-
 src/Storages/StorageMemory.h                  |   2 +-
 src/Storages/StorageMerge.cpp                 |  31 +-
 src/Storages/StorageMerge.h                   |  12 +-
 src/Storages/StorageMergeTree.cpp             |   2 +-
 src/Storages/StorageReplicatedMergeTree.cpp   |  12 +-
 src/Storages/StorageReplicatedMergeTree.h     |   3 +-
 src/Storages/StorageS3.cpp                    |  23 +-
 src/Storages/StorageS3.h                      |   4 +-
 src/Storages/StorageSet.cpp                   |  57 +-
 src/Storages/StorageSet.h                     |   7 +-
 src/Storages/StorageStripeLog.cpp             |   4 +-
 src/Storages/StorageURL.cpp                   |   2 +
 src/Storages/StorageURL.h                     |   5 +-
 src/Storages/StorageXDBC.cpp                  |   1 +
 src/Storages/System/StorageSystemColumns.cpp  |  54 +-
 .../System/StorageSystemDatabases.cpp         |  19 +-
 .../System/StorageSystemDistributionQueue.cpp |   1 +
 src/Storages/System/StorageSystemTables.cpp   |  17 +-
 src/Storages/getStructureOfRemoteTable.cpp    |   2 +-
 ...est_aux_funcs_for_adaptive_granularity.cpp |   1 +
 src/TableFunctions/ITableFunctionXDBC.cpp     |   1 +
 src/TableFunctions/TableFunctionFile.h        |   2 +-
 src/TableFunctions/TableFunctionHDFS.cpp      |   2 +-
 src/TableFunctions/TableFunctionMerge.cpp     |  77 +-
 src/TableFunctions/TableFunctionMerge.h       |   4 +-
 src/TableFunctions/TableFunctionS3.cpp        |   2 +
 tests/CMakeLists.txt                          |  10 -
 tests/clickhouse-test                         |  78 +-
 .../config.d/max_concurrent_queries.xml       |   4 +
 tests/config/install.sh                       |   1 +
 tests/config/top_level_domains                |   1 -
 .../top_level_domains/public_suffix_list.dat  |   0
 tests/instructions/easy_tasks_sorted_ru.md    |  28 +-
 tests/integration/README.md                   |  13 +-
 tests/integration/helpers/cluster.py          |  61 +-
 tests/integration/helpers/hdfs_api.py         | 185 ++++-
 .../test_allowed_url_from_config/test.py      |   6 +-
 .../configs/users.xml                         |   8 +
 .../test_graphite_merge_tree/test.py          |   3 +-
 .../test_join_set_family_s3/__init__.py       |   0
 .../configs/config.d/log_conf.xml             |  12 +
 .../configs/config.xml                        |  28 +
 .../test_join_set_family_s3/configs/minio.xml |  14 +
 .../test_join_set_family_s3/configs/ssl.xml   |  12 +
 .../test_join_set_family_s3/configs/users.xml |  23 +
 .../test_join_set_family_s3/test.py           |  97 +++
 .../test_log_family_s3/configs/minio.xml      |   4 +-
 tests/integration/test_log_family_s3/test.py  |   3 +-
 .../configs/users.xml                         |   2 +
 .../configs/users_db_atomic.xml               |  19 +
 .../materialize_with_ddl.py                   |  52 +-
 .../test_materialize_mysql_database/test.py   | 103 +--
 .../test_postgresql_protocol/test.py          |   2 +-
 .../configs/users.d/settings.xml              |   1 +
 .../test_redirect_url_storage/test.py         |  15 +-
 .../configs/users.d/another_user.xml          |   7 +-
 .../integration/test_settings_profile/test.py |  32 +-
 tests/integration/test_storage_hdfs/test.py   |  57 +-
 tests/integration/test_storage_kafka/test.py  |  61 ++
 .../test_storage_kerberized_hdfs/__init__.py  |   0
 .../configs/hdfs.xml                          |  13 +
 .../configs/log_conf.xml                      |  11 +
 .../hdfs_configs/bootstrap.sh                 | 262 +++++++
 .../kerberos_image_config.sh                  | 139 ++++
 .../secrets/krb.conf                          |  24 +
 .../secrets/krb_long.conf                     |  24 +
 .../test_storage_kerberized_hdfs/test.py      | 108 +++
 tests/integration/test_storage_s3/test.py     |   3 +-
 .../__init__.py                               |   0
 .../test.py                                   |  55 ++
 .../configs/logs_conf.xml                     |  12 +
 .../test_testkeeper_back_to_back/test.py      | 125 +++-
 tests/performance/encodeXMLComponent.xml      |   7 +
 .../00083_create_merge_tree_zookeeper.sql     |   2 +
 .../00327_summing_composite_nested.sql        |   2 +
 .../00443_optimize_final_vertical_merge.sh    |   6 +-
 .../queries/0_stateless/00453_top_k.reference |   7 +
 tests/queries/0_stateless/00453_top_k.sql     |  16 +-
 ...ed_storage_definition_syntax_zookeeper.sql |   2 +
 .../00564_versioned_collapsing_merge_tree.sql |   2 +
 ...77_replacing_merge_tree_vertical_merge.sql |   2 +
 .../0_stateless/00616_final_single_part.sql   |   2 +
 ...ptimize_on_nonleader_replica_zookeeper.sql |   4 +-
 ...00660_optimize_final_without_partition.sql |   2 +
 ...replicated_without_partition_zookeeper.sql |   2 +
 .../0_stateless/00682_empty_parts_merge.sh    |  14 +-
 ...block_size_system_tables_columns.reference |   2 +-
 ...sert_have_data_before_quorum_zookeeper.sql |   2 +-
 ...ert_lost_part_and_alive_part_zookeeper.sql |   4 +-
 ...0732_quorum_insert_lost_part_zookeeper.sql |   4 +-
 ..._old_data_and_without_quorum_zookeeper.sql |   2 +-
 ...m_insert_simple_test_1_parts_zookeeper.sql |   2 +-
 ...m_insert_simple_test_2_parts_zookeeper.sql |   2 +-
 .../00754_alter_modify_order_by.sql           |   1 +
 ...r_modify_order_by_replicated_zookeeper.sql |   2 +
 ...tem_parts_race_condition_drop_zookeeper.sh |   2 +-
 ...030_incorrect_count_summing_merge_tree.sql |   2 +
 ...invalidate_query_switchover_long.reference |   2 +-
 ...ionary_invalidate_query_switchover_long.sh |   2 +-
 .../01056_create_table_as.reference           |   1 +
 .../0_stateless/01056_create_table_as.sql     |   4 +
 ...81_PartialSortingTransform_full_column.sql |   2 -
 ..._zookeeper_mutations_and_insert_quorum.sql |   2 +-
 ...98_temporary_and_external_tables.reference |   7 +
 .../01098_temporary_and_external_tables.sh    |  27 +-
 .../01187_set_profile_as_setting.reference    |   8 +
 .../01187_set_profile_as_setting.sh           |  15 +
 ...01285_data_skip_index_over_aggregation.sql |   2 +
 .../01290_max_execution_speed_distributed.sql |  14 +-
 .../0_stateless/01293_show_settings.reference |   5 +
 .../0_stateless/01293_show_settings.sql       |   3 +
 .../01322_cast_keep_nullable.reference        |   1 +
 .../0_stateless/01322_cast_keep_nullable.sql  |   2 +-
 .../0_stateless/01323_add_scalars_in_time.sql |   2 +
 ...mming_merge_tree_exclude_partition_key.sql |   2 +
 ...1451_replicated_detach_drop_and_quorum.sql |   2 +-
 .../01459_manual_write_to_replicas_quorum.sh  |   2 +-
 ..._select_sequence_consistency_zookeeper.sql |   2 +-
 .../01550_create_map_type.reference           |  29 +
 .../0_stateless/01550_create_map_type.sql     |  65 ++
 .../01550_type_map_formats.reference          |  46 ++
 .../0_stateless/01550_type_map_formats.sql    |  19 +
 .../01550_type_map_formats_input.reference    |   5 +
 .../01550_type_map_formats_input.sh           |  21 +
 .../01556_accurate_cast_or_null.reference     |  30 +
 .../01556_accurate_cast_or_null.sql           |  35 +
 .../01560_optimize_on_insert.reference        |  13 +
 .../0_stateless/01560_optimize_on_insert.sql  |  35 +
 ...gregator_combinator_simple_state.reference |  14 +
 ...570_aggregator_combinator_simple_state.sql |  17 +
 .../01581_to_int256_uint256_inf_nan.reference |   6 -
 .../01581_to_int256_uint256_inf_nan.sql       |  10 -
 .../01581_to_int_inf_nan.reference            |   0
 .../0_stateless/01581_to_int_inf_nan.sql      |  10 +
 .../01596_setting_limit_offset.reference      |  63 ++
 .../01596_setting_limit_offset.sql            |  30 +
 .../01600_detach_permanently.reference        | 102 +++
 .../0_stateless/01600_detach_permanently.sh   | 125 ++++
 ..._sort_lowered_memory_bytes_ratio.reference |   0
 ...emerge_sort_lowered_memory_bytes_ratio.sql |  29 +
 .../01600_select_in_different_types.reference |  11 +
 .../01600_select_in_different_types.sql       |  35 +
 .../0_stateless/01601_accurate_cast.reference |   8 +
 .../0_stateless/01601_accurate_cast.sql       |  24 +
 .../01601_detach_permanently.reference        |  79 ++
 .../0_stateless/01601_detach_permanently.sql  | 214 ++++++
 ...01_temporary_table_session_scope.reference |   5 +
 .../01601_temporary_table_session_scope.sh    |  19 +
 .../01602_array_aggregation.reference         |  60 ++
 .../0_stateless/01602_array_aggregation.sql   |  60 ++
 ...1602_max_distributed_connections.reference |   0
 .../01602_max_distributed_connections.sh      |  15 +
 .../01602_modified_julian_day_msan.reference  |   1 +
 .../01602_modified_julian_day_msan.sql        |   4 +
 ...temporary_table_in_system_tables.reference |  10 +
 ...01602_temporary_table_in_system_tables.sql |  18 +
 .../01603_decimal_mult_float.reference        |  12 +
 .../0_stateless/01603_decimal_mult_float.sql  |  26 +
 ...603_insert_select_too_many_parts.reference |   1 +
 .../01603_insert_select_too_many_parts.sql    |  14 +
 .../01603_remove_column_ttl.reference         |   6 +
 .../0_stateless/01603_remove_column_ttl.sql   |  30 +
 .../01603_rename_overwrite_bug.reference      |   4 +
 .../01603_rename_overwrite_bug.sql            |  22 +
 ...4_explain_ast_of_nonselect_query.reference |   8 +
 .../01604_explain_ast_of_nonselect_query.sql  |   1 +
 .../queries/0_stateless/arcadia_skip_list.txt |   2 +
 tests/queries/skip_list.json                  | 172 +++++
 .../tests/privileges/create/create_table.py   |   3 +
 .../convert-month-partitioned-parts/main.cpp  |   1 +
 utils/db-generator/query_db_generator.cpp     |   2 +-
 utils/iotest/iotest.cpp                       |   3 +-
 utils/iotest/iotest_aio.cpp                   |   3 +-
 utils/make_changelog.py                       | 507 -------------
 utils/zookeeper-test/main.cpp                 |  97 ++-
 website/benchmark/hardware/index.html         |   3 +-
 .../results/do_storage_optimized.json         | 380 ++++++++++
 ...018.json => macbook_pro_core_i7_2018.json} |   0
 .../results/macbook_pro_core_i7_2020.json     |  54 ++
 556 files changed, 12092 insertions(+), 3787 deletions(-)
 create mode 100644 base/common/getPageSize.cpp
 create mode 100644 base/common/getPageSize.h
 create mode 100644 docker/test/integration/kerberized_hadoop/Dockerfile
 create mode 100644 docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml
 create mode 100644 docs/en/operations/system-tables/replication_queue.md
 create mode 100644 docs/ja/sql-reference/data-types/lowcardinality.md
 create mode 100644 docs/ru/faq/general/index.md
 rename docs/ru/faq/{general.md => general/mapreduce.md} (60%)
 create mode 100644 docs/ru/faq/integration/file-export.md
 create mode 100644 docs/ru/faq/integration/index.md
 create mode 100644 docs/ru/faq/integration/oracle-odbc.md
 create mode 100644 docs/ru/faq/operations/index.md
 create mode 100644 docs/ru/faq/use-cases/index.md
 create mode 100644 docs/ru/operations/system-tables/distribution_queue.md
 create mode 100644 docs/ru/operations/system-tables/replication_queue.md
 create mode 100644 docs/tr/sql-reference/data-types/lowcardinality.md
 create mode 120000 programs/server/config.d/test_keeper_port.xml
 create mode 100644 src/AggregateFunctions/AggregateFunctionSimpleState.cpp
 create mode 100644 src/AggregateFunctions/AggregateFunctionSimpleState.h
 create mode 100644 src/Columns/ColumnMap.cpp
 create mode 100644 src/Columns/ColumnMap.h
 create mode 100644 src/Common/tests/gtest_hash_table.cpp
 delete mode 100644 src/Common/tests/hash_table.cpp
 create mode 100644 src/DataTypes/DataTypeMap.cpp
 create mode 100644 src/DataTypes/DataTypeMap.h
 create mode 100644 src/Functions/array/arrayAggregation.cpp
 delete mode 100644 src/Functions/array/arraySum.cpp
 rename src/Functions/{encodeXML.cpp => encodeXMLComponent.cpp} (93%)
 create mode 100644 src/Functions/map.cpp
 create mode 100644 src/IO/ConnectionTimeoutsContext.h
 delete mode 100644 src/IO/HDFSCommon.cpp
 delete mode 100644 src/IO/HDFSCommon.h
 create mode 100644 src/Storages/HDFS/HDFSCommon.cpp
 create mode 100644 src/Storages/HDFS/HDFSCommon.h
 rename src/{IO => Storages/HDFS}/ReadBufferFromHDFS.cpp (81%)
 rename src/{IO => Storages/HDFS}/ReadBufferFromHDFS.h (69%)
 rename src/Storages/{ => HDFS}/StorageHDFS.cpp (95%)
 rename src/Storages/{ => HDFS}/StorageHDFS.h (100%)
 rename src/{IO => Storages/HDFS}/WriteBufferFromHDFS.cpp (84%)
 rename src/{IO => Storages/HDFS}/WriteBufferFromHDFS.h (80%)
 create mode 100644 tests/config/config.d/max_concurrent_queries.xml
 delete mode 120000 tests/config/top_level_domains
 rename {docker/test/performance-comparison => tests}/config/top_level_domains/public_suffix_list.dat (100%)
 create mode 100644 tests/integration/test_graphite_merge_tree/configs/users.xml
 create mode 100644 tests/integration/test_join_set_family_s3/__init__.py
 create mode 100644 tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/config.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/minio.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/ssl.xml
 create mode 100644 tests/integration/test_join_set_family_s3/configs/users.xml
 create mode 100644 tests/integration/test_join_set_family_s3/test.py
 create mode 100644 tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/__init__.py
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml
 create mode 100755 tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf
 create mode 100644 tests/integration/test_storage_kerberized_hdfs/test.py
 create mode 100644 tests/integration/test_table_functions_access_rights/__init__.py
 create mode 100644 tests/integration/test_table_functions_access_rights/test.py
 create mode 100644 tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml
 create mode 100644 tests/performance/encodeXMLComponent.xml
 create mode 100644 tests/queries/0_stateless/01187_set_profile_as_setting.reference
 create mode 100755 tests/queries/0_stateless/01187_set_profile_as_setting.sh
 create mode 100644 tests/queries/0_stateless/01293_show_settings.reference
 create mode 100644 tests/queries/0_stateless/01293_show_settings.sql
 create mode 100644 tests/queries/0_stateless/01550_create_map_type.reference
 create mode 100644 tests/queries/0_stateless/01550_create_map_type.sql
 create mode 100644 tests/queries/0_stateless/01550_type_map_formats.reference
 create mode 100644 tests/queries/0_stateless/01550_type_map_formats.sql
 create mode 100644 tests/queries/0_stateless/01550_type_map_formats_input.reference
 create mode 100755 tests/queries/0_stateless/01550_type_map_formats_input.sh
 create mode 100644 tests/queries/0_stateless/01556_accurate_cast_or_null.reference
 create mode 100644 tests/queries/0_stateless/01556_accurate_cast_or_null.sql
 create mode 100644 tests/queries/0_stateless/01560_optimize_on_insert.reference
 create mode 100644 tests/queries/0_stateless/01560_optimize_on_insert.sql
 create mode 100644 tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference
 create mode 100644 tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql
 delete mode 100644 tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference
 delete mode 100644 tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql
 create mode 100644 tests/queries/0_stateless/01581_to_int_inf_nan.reference
 create mode 100644 tests/queries/0_stateless/01581_to_int_inf_nan.sql
 create mode 100644 tests/queries/0_stateless/01596_setting_limit_offset.reference
 create mode 100644 tests/queries/0_stateless/01596_setting_limit_offset.sql
 create mode 100644 tests/queries/0_stateless/01600_detach_permanently.reference
 create mode 100755 tests/queries/0_stateless/01600_detach_permanently.sh
 create mode 100644 tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.reference
 create mode 100644 tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql
 create mode 100644 tests/queries/0_stateless/01600_select_in_different_types.reference
 create mode 100644 tests/queries/0_stateless/01600_select_in_different_types.sql
 create mode 100644 tests/queries/0_stateless/01601_accurate_cast.reference
 create mode 100644 tests/queries/0_stateless/01601_accurate_cast.sql
 create mode 100644 tests/queries/0_stateless/01601_detach_permanently.reference
 create mode 100644 tests/queries/0_stateless/01601_detach_permanently.sql
 create mode 100644 tests/queries/0_stateless/01601_temporary_table_session_scope.reference
 create mode 100755 tests/queries/0_stateless/01601_temporary_table_session_scope.sh
 create mode 100644 tests/queries/0_stateless/01602_array_aggregation.reference
 create mode 100644 tests/queries/0_stateless/01602_array_aggregation.sql
 create mode 100644 tests/queries/0_stateless/01602_max_distributed_connections.reference
 create mode 100755 tests/queries/0_stateless/01602_max_distributed_connections.sh
 create mode 100644 tests/queries/0_stateless/01602_modified_julian_day_msan.reference
 create mode 100644 tests/queries/0_stateless/01602_modified_julian_day_msan.sql
 create mode 100644 tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference
 create mode 100644 tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql
 create mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.reference
 create mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.sql
 create mode 100644 tests/queries/0_stateless/01603_insert_select_too_many_parts.reference
 create mode 100644 tests/queries/0_stateless/01603_insert_select_too_many_parts.sql
 create mode 100644 tests/queries/0_stateless/01603_remove_column_ttl.reference
 create mode 100644 tests/queries/0_stateless/01603_remove_column_ttl.sql
 create mode 100644 tests/queries/0_stateless/01603_rename_overwrite_bug.reference
 create mode 100644 tests/queries/0_stateless/01603_rename_overwrite_bug.sql
 create mode 100644 tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference
 create mode 100644 tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql
 delete mode 100755 utils/make_changelog.py
 create mode 100644 website/benchmark/hardware/results/do_storage_optimized.json
 rename website/benchmark/hardware/results/{core_i7_macbook_pro_2018.json => macbook_pro_core_i7_2018.json} (100%)
 create mode 100644 website/benchmark/hardware/results/macbook_pro_core_i7_2020.json

diff --git a/.gitmodules b/.gitmodules
index 1d996fb84cc..23deab0eed0 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -183,7 +183,7 @@
 	url = https://github.com/kthohr/stats.git
 [submodule "contrib/krb5"]
 	path = contrib/krb5
-	url = https://github.com/krb5/krb5
+	url = https://github.com/ClickHouse-Extras/krb5
 [submodule "contrib/cyrus-sasl"]
 	path = contrib/cyrus-sasl
 	url = https://github.com/cyrusimap/cyrus-sasl
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4d525798505..6d9174c7d07 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,126 @@
+### ClickHouse release 20.12
+
+### ClickHouse release v20.12.3.3-stable, 2020-12-13
+
+#### Backward Incompatible Change
+
+* Enable `use_compact_format_in_distributed_parts_names` by default (see the documentation for the reference). [#16728](https://github.com/ClickHouse/ClickHouse/pull/16728) ([Azat Khuzhin](https://github.com/azat)).
+* Accept user settings related to file formats (e.g. `format_csv_delimiter`) in the `SETTINGS` clause when creating a table that uses `File` engine, and use these settings in all `INSERT`s and `SELECT`s. The file format settings changed in the current user session, or in the `SETTINGS` clause of a DML query itself, no longer affect the query. [#16591](https://github.com/ClickHouse/ClickHouse/pull/16591) ([Alexander Kuzmenkov](https://github.com/akuzm)).
+
+#### New Feature
+
+* add `*.xz` compression/decompression support.It enables using `*.xz` in `file()` function. This closes [#8828](https://github.com/ClickHouse/ClickHouse/issues/8828). [#16578](https://github.com/ClickHouse/ClickHouse/pull/16578) ([Abi Palagashvili](https://github.com/fibersel)).
+* Introduce the query `ALTER TABLE ... DROP|DETACH PART 'part_name'`. [#15511](https://github.com/ClickHouse/ClickHouse/pull/15511) ([nvartolomei](https://github.com/nvartolomei)).
+* Added new ALTER UPDATE/DELETE IN PARTITION syntax. [#13403](https://github.com/ClickHouse/ClickHouse/pull/13403) ([Vladimir Chebotarev](https://github.com/excitoon)).
+* Allow formatting named tuples as JSON objects when using JSON input/output formats, controlled by the `output_format_json_named_tuples_as_objects` setting, disabled by default. [#17175](https://github.com/ClickHouse/ClickHouse/pull/17175) ([Alexander Kuzmenkov](https://github.com/akuzm)).
+* Add a possibility to input enum value as it's id in TSV and CSV formats by default. [#16834](https://github.com/ClickHouse/ClickHouse/pull/16834) ([Kruglov Pavel](https://github.com/Avogar)).
+* Add COLLATE support for Nullable, LowCardinality, Array and Tuple, where nested type is String. Also refactor the code associated with collations in ColumnString.cpp. [#16273](https://github.com/ClickHouse/ClickHouse/pull/16273) ([Kruglov Pavel](https://github.com/Avogar)).
+* New `tcpPort` function returns TCP port listened by this server. [#17134](https://github.com/ClickHouse/ClickHouse/pull/17134) ([Ivan](https://github.com/abyss7)).
+* Add new math functions: `acosh`, `asinh`, `atan2`, `atanh`, `cosh`, `hypot`, `log1p`, `sinh`. [#16636](https://github.com/ClickHouse/ClickHouse/pull/16636) ([Konstantin Malanchev](https://github.com/hombit)).
+* Possibility to distribute the merges between different replicas. Introduces the `execute_merges_on_single_replica_time_threshold` mergetree setting. [#16424](https://github.com/ClickHouse/ClickHouse/pull/16424) ([filimonov](https://github.com/filimonov)).
+* Add setting `aggregate_functions_null_for_empty` for SQL standard compatibility. This option will rewrite all aggregate functions in a query, adding -OrNull suffix to them. Implements [10273](https://github.com/ClickHouse/ClickHouse/issues/10273). [#16123](https://github.com/ClickHouse/ClickHouse/pull/16123) ([flynn](https://github.com/ucasFL)).
+* Updated DateTime, DateTime64 parsing to accept string Date literal format. [#16040](https://github.com/ClickHouse/ClickHouse/pull/16040) ([Maksim Kita](https://github.com/kitaisreal)).
+* Make it possible to change the path to history file in `clickhouse-client` using the `--history_file` parameter. [#15960](https://github.com/ClickHouse/ClickHouse/pull/15960) ([Maksim Kita](https://github.com/kitaisreal)).
+
+#### Bug Fix
+
+* Fix the issue when server can stop accepting connections in very rare cases. [#17542](https://github.com/ClickHouse/ClickHouse/pull/17542) ([Amos Bird](https://github.com/amosbird)).
+* Fixed `Function not implemented` error when executing `RENAME` query in `Atomic` database with ClickHouse running on Windows Subsystem for Linux. Fixes [#17661](https://github.com/ClickHouse/ClickHouse/issues/17661). [#17664](https://github.com/ClickHouse/ClickHouse/pull/17664) ([tavplubix](https://github.com/tavplubix)).
+* Do not restore parts from WAL if `in_memory_parts_enable_wal` is disabled. [#17802](https://github.com/ClickHouse/ClickHouse/pull/17802) ([detailyang](https://github.com/detailyang)).
+* fix incorrect initialization of `max_compress_block_size` of MergeTreeWriterSettings with `min_compress_block_size`. [#17833](https://github.com/ClickHouse/ClickHouse/pull/17833) ([flynn](https://github.com/ucasFL)).
+* Exception message about max table size to drop was displayed incorrectly. [#17764](https://github.com/ClickHouse/ClickHouse/pull/17764) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed possible segfault when there is not enough space when inserting into `Distributed` table. [#17737](https://github.com/ClickHouse/ClickHouse/pull/17737) ([tavplubix](https://github.com/tavplubix)).
+* Fixed problem when ClickHouse fails to resume connection to MySQL servers. [#17681](https://github.com/ClickHouse/ClickHouse/pull/17681) ([Alexander Kazakov](https://github.com/Akazz)).
+* In might be determined incorrectly if cluster is circular- (cross-) replicated or not when executing `ON CLUSTER` query due to race condition when `pool_size` > 1. It's fixed. [#17640](https://github.com/ClickHouse/ClickHouse/pull/17640) ([tavplubix](https://github.com/tavplubix)).
+* Exception `fmt::v7::format_error` can be logged in background for MergeTree tables. This fixes [#17613](https://github.com/ClickHouse/ClickHouse/issues/17613). [#17615](https://github.com/ClickHouse/ClickHouse/pull/17615) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* When clickhouse-client is used in interactive mode with multiline queries, single line comment was erronously extended till the end of query. This fixes [#13654](https://github.com/ClickHouse/ClickHouse/issues/13654). [#17565](https://github.com/ClickHouse/ClickHouse/pull/17565) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix alter query hang when the corresponding mutation was killed on the different replica. Fixes [#16953](https://github.com/ClickHouse/ClickHouse/issues/16953). [#17499](https://github.com/ClickHouse/ClickHouse/pull/17499) ([alesapin](https://github.com/alesapin)).
+* Fix issue when mark cache size was underestimated by clickhouse. It may happen when there are a lot of tiny files with marks. [#17496](https://github.com/ClickHouse/ClickHouse/pull/17496) ([alesapin](https://github.com/alesapin)).
+* Fix `ORDER BY` with enabled setting `optimize_redundant_functions_in_order_by`. [#17471](https://github.com/ClickHouse/ClickHouse/pull/17471) ([Anton Popov](https://github.com/CurtizJ)).
+* Fix duplicates after `DISTINCT` which were possible because of incorrect optimization. Fixes [#17294](https://github.com/ClickHouse/ClickHouse/issues/17294). [#17296](https://github.com/ClickHouse/ClickHouse/pull/17296) ([li chengxiang](https://github.com/chengxianglibra)). [#17439](https://github.com/ClickHouse/ClickHouse/pull/17439) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fix crash while reading from `JOIN` table with `LowCardinality` types. Fixes [#17228](https://github.com/ClickHouse/ClickHouse/issues/17228). [#17397](https://github.com/ClickHouse/ClickHouse/pull/17397) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* fix `toInt256(inf)` stack overflow. Int256 is an experimental feature. Closed [#17235](https://github.com/ClickHouse/ClickHouse/issues/17235). [#17257](https://github.com/ClickHouse/ClickHouse/pull/17257) ([flynn](https://github.com/ucasFL)).
+* Fix possible `Unexpected packet Data received from client` error logged for Distributed queries with `LIMIT`. [#17254](https://github.com/ClickHouse/ClickHouse/pull/17254) ([Azat Khuzhin](https://github.com/azat)).
+* Fix set index invalidation when there are const columns in the subquery. This fixes [#17246](https://github.com/ClickHouse/ClickHouse/issues/17246). [#17249](https://github.com/ClickHouse/ClickHouse/pull/17249) ([Amos Bird](https://github.com/amosbird)).
+* Fix possible wrong index analysis when the types of the index comparison are different. This fixes [#17122](https://github.com/ClickHouse/ClickHouse/issues/17122). [#17145](https://github.com/ClickHouse/ClickHouse/pull/17145) ([Amos Bird](https://github.com/amosbird)).
+* Fix ColumnConst comparison which leads to crash. This fixed [#17088](https://github.com/ClickHouse/ClickHouse/issues/17088) . [#17135](https://github.com/ClickHouse/ClickHouse/pull/17135) ([Amos Bird](https://github.com/amosbird)).
+* Multiple fixed for MaterializeMySQL (experimental feature). Fixes [#16923](https://github.com/ClickHouse/ClickHouse/issues/16923) Fixes [#15883](https://github.com/ClickHouse/ClickHouse/issues/15883) Fix MaterializeMySQL SYNC failure when the modify MySQL binlog_checksum. [#17091](https://github.com/ClickHouse/ClickHouse/pull/17091) ([Winter Zhang](https://github.com/zhang2014)).
+* Fix bug when `ON CLUSTER` queries may hang forever for non-leader ReplicatedMergeTreeTables. [#17089](https://github.com/ClickHouse/ClickHouse/pull/17089) ([alesapin](https://github.com/alesapin)).
+* Fixed crash on `CREATE TABLE ... AS some_table` query when `some_table` was created `AS table_function()` Fixes [#16944](https://github.com/ClickHouse/ClickHouse/issues/16944). [#17072](https://github.com/ClickHouse/ClickHouse/pull/17072) ([tavplubix](https://github.com/tavplubix)).
+* Bug unfinished implementation for funciton fuzzBits, related issue: [#16980](https://github.com/ClickHouse/ClickHouse/issues/16980). [#17051](https://github.com/ClickHouse/ClickHouse/pull/17051) ([hexiaoting](https://github.com/hexiaoting)).
+* Fix LLVM's libunwind in the case when CFA register is RAX. This is the [bug](https://bugs.llvm.org/show_bug.cgi?id=48186) in [LLVM's libunwind](https://github.com/llvm/llvm-project/tree/master/libunwind). We already have workarounds for this bug. [#17046](https://github.com/ClickHouse/ClickHouse/pull/17046) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Avoid unnecessary network errors for remote queries which may be cancelled while execution, like queries with `LIMIT`. [#17006](https://github.com/ClickHouse/ClickHouse/pull/17006) ([Azat Khuzhin](https://github.com/azat)).
+* Fix `optimize_distributed_group_by_sharding_key` setting (that is disabled by default) for query with OFFSET only. [#16996](https://github.com/ClickHouse/ClickHouse/pull/16996) ([Azat Khuzhin](https://github.com/azat)).
+* Fix for Merge tables over Distributed tables with JOIN. [#16993](https://github.com/ClickHouse/ClickHouse/pull/16993) ([Azat Khuzhin](https://github.com/azat)).
+* Fixed wrong result in big integers (128, 256 bit) when casting from double. Big integers support is experimental. [#16986](https://github.com/ClickHouse/ClickHouse/pull/16986) ([Mike](https://github.com/myrrc)).
+* Fix possible server crash after `ALTER TABLE ... MODIFY COLUMN ... NewType` when `SELECT` have `WHERE` expression on altering column and alter doesn't finished yet. [#16968](https://github.com/ClickHouse/ClickHouse/pull/16968) ([Amos Bird](https://github.com/amosbird)).
+* Blame info was not calculated correctly in `clickhouse-git-import`. [#16959](https://github.com/ClickHouse/ClickHouse/pull/16959) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix order by optimization with monotonous functions. Fixes [#16107](https://github.com/ClickHouse/ClickHouse/issues/16107). [#16956](https://github.com/ClickHouse/ClickHouse/pull/16956) ([Anton Popov](https://github.com/CurtizJ)).
+* Fix optimization of group by with enabled setting `optimize_aggregators_of_group_by_keys` and joins. Fixes [#12604](https://github.com/ClickHouse/ClickHouse/issues/12604). [#16951](https://github.com/ClickHouse/ClickHouse/pull/16951) ([Anton Popov](https://github.com/CurtizJ)).
+* Fix possible error `Illegal type of argument` for queries with `ORDER BY`. Fixes [#16580](https://github.com/ClickHouse/ClickHouse/issues/16580). [#16928](https://github.com/ClickHouse/ClickHouse/pull/16928) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fix strange code in InterpreterShowAccessQuery. [#16866](https://github.com/ClickHouse/ClickHouse/pull/16866) ([tavplubix](https://github.com/tavplubix)).
+* Prevent clickhouse server crashes when using the function `timeSeriesGroupSum`. The function is removed from newer ClickHouse releases. [#16865](https://github.com/ClickHouse/ClickHouse/pull/16865) ([filimonov](https://github.com/filimonov)).
+* Fix rare silent crashes when query profiler is on and ClickHouse is installed on OS with glibc version that has (supposedly) broken asynchronous unwind tables for some functions. This fixes [#15301](https://github.com/ClickHouse/ClickHouse/issues/15301). This fixes [#13098](https://github.com/ClickHouse/ClickHouse/issues/13098). [#16846](https://github.com/ClickHouse/ClickHouse/pull/16846) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix crash when using `any` without any arguments. This is for [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803) . cc @azat. [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
+* If no memory can be allocated while writing table metadata on disk, broken metadata file can be written. [#16772](https://github.com/ClickHouse/ClickHouse/pull/16772) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix trivial query optimization with partition predicate. [#16767](https://github.com/ClickHouse/ClickHouse/pull/16767) ([Azat Khuzhin](https://github.com/azat)).
+* Fix `IN` operator over several columns and tuples with enabled `transform_null_in` setting. Fixes [#15310](https://github.com/ClickHouse/ClickHouse/issues/15310). [#16722](https://github.com/ClickHouse/ClickHouse/pull/16722) ([Anton Popov](https://github.com/CurtizJ)).
+* Return number of affected rows for INSERT queries via MySQL protocol. Previously ClickHouse used to always return 0, it's fixed. Fixes [#16605](https://github.com/ClickHouse/ClickHouse/issues/16605). [#16715](https://github.com/ClickHouse/ClickHouse/pull/16715) ([Winter Zhang](https://github.com/zhang2014)).
+* Fix remote query failure when using 'if' suffix aggregate function. Fixes [#16574](https://github.com/ClickHouse/ClickHouse/issues/16574) Fixes [#16231](https://github.com/ClickHouse/ClickHouse/issues/16231) [#16610](https://github.com/ClickHouse/ClickHouse/pull/16610) ([Winter Zhang](https://github.com/zhang2014)).
+* Fix inconsistent behavior caused by `select_sequential_consistency` for optimized trivial count query and system.tables. [#16309](https://github.com/ClickHouse/ClickHouse/pull/16309) ([Hao Chen](https://github.com/haoch)).
+
+#### Improvement
+
+* Remove empty parts after they were pruned by TTL, mutation, or collapsing merge algorithm. [#16895](https://github.com/ClickHouse/ClickHouse/pull/16895) ([Anton Popov](https://github.com/CurtizJ)).
+* Enable compact format of directories for asynchronous sends in Distributed tables: `use_compact_format_in_distributed_parts_names` is set to 1 by default. [#16788](https://github.com/ClickHouse/ClickHouse/pull/16788) ([Azat Khuzhin](https://github.com/azat)).
+* Abort multipart upload if no data was written to S3. [#16840](https://github.com/ClickHouse/ClickHouse/pull/16840) ([Pavel Kovalenko](https://github.com/Jokser)).
+* Reresolve the IP of the `format_avro_schema_registry_url` in case of errors. [#16985](https://github.com/ClickHouse/ClickHouse/pull/16985) ([filimonov](https://github.com/filimonov)).
+* Mask password in data_path in the system.distribution_queue. [#16727](https://github.com/ClickHouse/ClickHouse/pull/16727) ([Azat Khuzhin](https://github.com/azat)).
+* Throw error when use column transformer replaces non existing column. [#16183](https://github.com/ClickHouse/ClickHouse/pull/16183) ([hexiaoting](https://github.com/hexiaoting)).
+* Turn off parallel parsing when there is no enough memory for all threads to work simultaneously. Also there could be exceptions like "Memory limit exceeded" when somebody will try to insert extremely huge rows (> min_chunk_bytes_for_parallel_parsing), because each piece to parse has to be independent set of strings (one or more). [#16721](https://github.com/ClickHouse/ClickHouse/pull/16721) ([Nikita Mikhaylov](https://github.com/nikitamikhaylov)).
+* Install script should always create subdirs in config folders. This is only relevant for Docker build with custom config. [#16936](https://github.com/ClickHouse/ClickHouse/pull/16936) ([filimonov](https://github.com/filimonov)).
+* Correct grammar in error message in JSONEachRow, JSONCompactEachRow, and RegexpRow input formats. [#17205](https://github.com/ClickHouse/ClickHouse/pull/17205) ([nico piderman](https://github.com/sneako)).
+* Set default `host` and `port` parameters for `SOURCE(CLICKHOUSE(...))` to current instance and set default `user` value to `'default'`. [#16997](https://github.com/ClickHouse/ClickHouse/pull/16997) ([vdimir](https://github.com/vdimir)).
+* Throw an informative error message when doing `ATTACH/DETACH TABLE <DICTIONARY>`. Before this PR, `detach table <dict>` works but leads to an ill-formed in-memory metadata. [#16885](https://github.com/ClickHouse/ClickHouse/pull/16885) ([Amos Bird](https://github.com/amosbird)).
+* Add cutToFirstSignificantSubdomainWithWWW(). [#16845](https://github.com/ClickHouse/ClickHouse/pull/16845) ([Azat Khuzhin](https://github.com/azat)).
+* Server refused to startup with exception message if wrong config is given (`metric_log`.`collect_interval_milliseconds` is missing). [#16815](https://github.com/ClickHouse/ClickHouse/pull/16815) ([Ivan](https://github.com/abyss7)).
+* Better exception message when configuration for distributed DDL is absent. This fixes [#5075](https://github.com/ClickHouse/ClickHouse/issues/5075). [#16769](https://github.com/ClickHouse/ClickHouse/pull/16769) ([Nikita Mikhaylov](https://github.com/nikitamikhaylov)).
+* Usability improvement: better suggestions in syntax error message when `CODEC` expression is misplaced in `CREATE TABLE` query. This fixes [#12493](https://github.com/ClickHouse/ClickHouse/issues/12493). [#16768](https://github.com/ClickHouse/ClickHouse/pull/16768) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Remove empty directories for async INSERT at start of Distributed engine. [#16729](https://github.com/ClickHouse/ClickHouse/pull/16729) ([Azat Khuzhin](https://github.com/azat)).
+* Workaround for use S3 with nginx server as proxy. Nginx currenty does not accept urls with empty path like `http://domain.com?delete`, but vanilla aws-sdk-cpp produces this kind of urls. This commit uses patched aws-sdk-cpp version, which makes urls with "/" as path in this cases, like `http://domain.com/?delete`. [#16709](https://github.com/ClickHouse/ClickHouse/pull/16709) ([ianton-ru](https://github.com/ianton-ru)).
+* Allow `reinterpretAs*` functions to work for integers and floats of the same size. Implements [16640](https://github.com/ClickHouse/ClickHouse/issues/16640). [#16657](https://github.com/ClickHouse/ClickHouse/pull/16657) ([flynn](https://github.com/ucasFL)).
+* Now, `<auxiliary_zookeepers>` configuration can be changed in `config.xml` and reloaded without server startup. [#16627](https://github.com/ClickHouse/ClickHouse/pull/16627) ([Amos Bird](https://github.com/amosbird)).
+* Support SNI in https connections to remote resources. This will allow to connect to Cloudflare servers that require SNI. This fixes [#10055](https://github.com/ClickHouse/ClickHouse/issues/10055). [#16252](https://github.com/ClickHouse/ClickHouse/pull/16252) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Make it possible to connect to `clickhouse-server` secure endpoint which requires SNI. This is possible when `clickhouse-server` is hosted behind TLS proxy. [#16938](https://github.com/ClickHouse/ClickHouse/pull/16938) ([filimonov](https://github.com/filimonov)).
+* Fix possible stack overflow if a loop of materialized views is created. This closes [#15732](https://github.com/ClickHouse/ClickHouse/issues/15732). [#16048](https://github.com/ClickHouse/ClickHouse/pull/16048) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Simplify the implementation of background tasks processing for the MergeTree table engines family. There should be no visible changes for user. [#15983](https://github.com/ClickHouse/ClickHouse/pull/15983) ([alesapin](https://github.com/alesapin)).
+* Improvement for MaterializeMySQL (experimental feature). Throw exception about right sync privileges when MySQL sync user has error privileges. [#15977](https://github.com/ClickHouse/ClickHouse/pull/15977) ([TCeason](https://github.com/TCeason)).
+* Made `indexOf()` use BloomFilter. [#14977](https://github.com/ClickHouse/ClickHouse/pull/14977) ([achimbab](https://github.com/achimbab)).
+
+#### Performance Improvement
+
+* Use Floyd-Rivest algorithm, it is the best for the ClickHouse use case of partial sorting. Bechmarks are in https://github.com/danlark1/miniselect and [here](https://drive.google.com/drive/folders/1DHEaeXgZuX6AJ9eByeZ8iQVQv0ueP8XM). [#16825](https://github.com/ClickHouse/ClickHouse/pull/16825) ([Danila Kutenin](https://github.com/danlark1)).
+* Now `ReplicatedMergeTree` tree engines family uses a separate thread pool for replicated fetches. Size of the pool limited by setting `background_fetches_pool_size` which can be tuned with a server restart. The default value of the setting is 3 and it means that the maximum amount of parallel fetches is equal to 3 (and it allows to utilize 10G network). Fixes #520. [#16390](https://github.com/ClickHouse/ClickHouse/pull/16390) ([alesapin](https://github.com/alesapin)).
+* Fixed uncontrolled growth of the state of `quantileTDigest`. [#16680](https://github.com/ClickHouse/ClickHouse/pull/16680) ([hrissan](https://github.com/hrissan)).
+* Add `VIEW` subquery description to `EXPLAIN`. Limit push down optimisation for `VIEW`. Add local replicas of `Distributed` to query plan. [#14936](https://github.com/ClickHouse/ClickHouse/pull/14936) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fix optimize_read_in_order/optimize_aggregation_in_order with max_threads > 0 and expression in ORDER BY. [#16637](https://github.com/ClickHouse/ClickHouse/pull/16637) ([Azat Khuzhin](https://github.com/azat)).
+* Fix performance of reading from `Merge` tables over huge number of `MergeTree` tables. Fixes [#7748](https://github.com/ClickHouse/ClickHouse/issues/7748). [#16988](https://github.com/ClickHouse/ClickHouse/pull/16988) ([Anton Popov](https://github.com/CurtizJ)).
+* Now we can safely prune partitions with exact match. Useful case: Suppose table is partitioned by `intHash64(x) % 100` and the query has condition on `intHash64(x) % 100` verbatim, not on x. [#16253](https://github.com/ClickHouse/ClickHouse/pull/16253) ([Amos Bird](https://github.com/amosbird)).
+
+#### Experimental Feature
+
+* Add `EmbeddedRocksDB` table engine (can be used for dictionaries). [#15073](https://github.com/ClickHouse/ClickHouse/pull/15073) ([sundyli](https://github.com/sundy-li)).
+
+#### Build/Testing/Packaging Improvement
+
+* Improvements in test coverage building images. [#17233](https://github.com/ClickHouse/ClickHouse/pull/17233) ([alesapin](https://github.com/alesapin)).
+* Update embedded timezone data to version 2020d (also update cctz to the latest master). [#17204](https://github.com/ClickHouse/ClickHouse/pull/17204) ([filimonov](https://github.com/filimonov)).
+* Fix UBSan report in Poco. This closes [#12719](https://github.com/ClickHouse/ClickHouse/issues/12719). [#16765](https://github.com/ClickHouse/ClickHouse/pull/16765) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Do not instrument 3rd-party libraries with UBSan. [#16764](https://github.com/ClickHouse/ClickHouse/pull/16764) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix UBSan report in cache dictionaries. This closes [#12641](https://github.com/ClickHouse/ClickHouse/issues/12641). [#16763](https://github.com/ClickHouse/ClickHouse/pull/16763) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fix UBSan report when trying to convert infinite floating point number to integer. This closes [#14190](https://github.com/ClickHouse/ClickHouse/issues/14190). [#16677](https://github.com/ClickHouse/ClickHouse/pull/16677) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+
+
 ## ClickHouse release 20.11
 
 ### ClickHouse release v20.11.3.3-stable, 2020-11-13
diff --git a/base/common/CMakeLists.txt b/base/common/CMakeLists.txt
index 9e4462c821a..cea52b443dd 100644
--- a/base/common/CMakeLists.txt
+++ b/base/common/CMakeLists.txt
@@ -6,6 +6,7 @@ set (SRCS
     demangle.cpp
     getFQDNOrHostName.cpp
     getMemoryAmount.cpp
+    getPageSize.cpp
     getThreadId.cpp
     JSON.cpp
     LineReader.cpp
diff --git a/base/common/ReadlineLineReader.cpp b/base/common/ReadlineLineReader.cpp
index 4e67ed5e4a3..397a7dd7543 100644
--- a/base/common/ReadlineLineReader.cpp
+++ b/base/common/ReadlineLineReader.cpp
@@ -1,4 +1,5 @@
 #include <common/ReadlineLineReader.h>
+#include <common/errnoToString.h>
 #include <ext/scope_guard.h>
 
 #include <errno.h>
@@ -69,7 +70,7 @@ ReadlineLineReader::ReadlineLineReader(
     {
         int res = read_history(history_file_path.c_str());
         if (res)
-            std::cerr << "Cannot read history from file " + history_file_path + ": "+ strerror(errno) << std::endl;
+            std::cerr << "Cannot read history from file " + history_file_path + ": "+ errnoToString(errno) << std::endl;
     }
 
     /// Added '.' to the default list. Because it is used to separate database and table.
@@ -107,7 +108,7 @@ ReadlineLineReader::ReadlineLineReader(
     };
 
     if (signal(SIGINT, clear_prompt_or_exit) == SIG_ERR)
-        throw std::runtime_error(std::string("Cannot set signal handler for readline: ") + strerror(errno));
+        throw std::runtime_error(std::string("Cannot set signal handler for readline: ") + errnoToString(errno));
 
     rl_variable_bind("completion-ignore-case", "on");
     // TODO: it doesn't work
diff --git a/base/common/ReplxxLineReader.cpp b/base/common/ReplxxLineReader.cpp
index 4eb7b065fe3..111428447af 100644
--- a/base/common/ReplxxLineReader.cpp
+++ b/base/common/ReplxxLineReader.cpp
@@ -47,7 +47,7 @@ ReplxxLineReader::ReplxxLineReader(
             {
                 if (!rx.history_load(history_file_path))
                 {
-                    rx.print("Loading history failed: %s\n", strerror(errno));
+                    rx.print("Loading history failed: %s\n", errnoToString(errno).c_str());
                 }
 
                 if (flock(history_file_fd, LOCK_UN))
@@ -88,7 +88,7 @@ ReplxxLineReader::ReplxxLineReader(
 ReplxxLineReader::~ReplxxLineReader()
 {
     if (close(history_file_fd))
-        rx.print("Close of history file failed: %s\n", strerror(errno));
+        rx.print("Close of history file failed: %s\n", errnoToString(errno).c_str());
 }
 
 LineReader::InputStatus ReplxxLineReader::readOneLine(const String & prompt)
@@ -113,7 +113,7 @@ void ReplxxLineReader::addToHistory(const String & line)
     // and that is why flock() is added here.
     bool locked = false;
     if (flock(history_file_fd, LOCK_EX))
-        rx.print("Lock of history file failed: %s\n", strerror(errno));
+        rx.print("Lock of history file failed: %s\n", errnoToString(errno).c_str());
     else
         locked = true;
 
@@ -121,10 +121,10 @@ void ReplxxLineReader::addToHistory(const String & line)
 
     // flush changes to the disk
     if (!rx.history_save(history_file_path))
-        rx.print("Saving history failed: %s\n", strerror(errno));
+        rx.print("Saving history failed: %s\n", errnoToString(errno).c_str());
 
     if (locked && 0 != flock(history_file_fd, LOCK_UN))
-        rx.print("Unlock of history file failed: %s\n", strerror(errno));
+        rx.print("Unlock of history file failed: %s\n", errnoToString(errno).c_str());
 }
 
 void ReplxxLineReader::enableBracketedPaste()
diff --git a/base/common/getMemoryAmount.cpp b/base/common/getMemoryAmount.cpp
index e7d284354f9..1f1b2b8c5dd 100644
--- a/base/common/getMemoryAmount.cpp
+++ b/base/common/getMemoryAmount.cpp
@@ -1,5 +1,6 @@
 #include <stdexcept>
 #include "common/getMemoryAmount.h"
+#include "common/getPageSize.h"
 
 #include <unistd.h>
 #include <sys/types.h>
@@ -18,7 +19,7 @@ uint64_t getMemoryAmountOrZero()
     if (num_pages <= 0)
         return 0;
 
-    int64_t page_size = sysconf(_SC_PAGESIZE);
+    int64_t page_size = getPageSize();
     if (page_size <= 0)
         return 0;
 
diff --git a/base/common/getPageSize.cpp b/base/common/getPageSize.cpp
new file mode 100644
index 00000000000..e96149cc1fb
--- /dev/null
+++ b/base/common/getPageSize.cpp
@@ -0,0 +1,8 @@
+#include "common/getPageSize.h"
+
+#include <unistd.h>
+
+Int64 getPageSize()
+{
+    return sysconf(_SC_PAGESIZE);
+}
diff --git a/base/common/getPageSize.h b/base/common/getPageSize.h
new file mode 100644
index 00000000000..8a35acbfe9b
--- /dev/null
+++ b/base/common/getPageSize.h
@@ -0,0 +1,6 @@
+#pragma once
+
+#include "common/types.h"
+
+/// Get memory page size
+Int64 getPageSize();
diff --git a/base/common/setTerminalEcho.cpp b/base/common/setTerminalEcho.cpp
index 11f6c1db3f0..658f27705ba 100644
--- a/base/common/setTerminalEcho.cpp
+++ b/base/common/setTerminalEcho.cpp
@@ -1,6 +1,7 @@
 // https://stackoverflow.com/questions/1413445/reading-a-password-from-stdcin
 
 #include <common/setTerminalEcho.h>
+#include <common/errnoToString.h>
 #include <stdexcept>
 #include <cstring>
 #include <string>
@@ -31,7 +32,7 @@ void setTerminalEcho(bool enable)
 #else
     struct termios tty;
     if (tcgetattr(STDIN_FILENO, &tty))
-        throw std::runtime_error(std::string("setTerminalEcho failed get: ") + strerror(errno));
+        throw std::runtime_error(std::string("setTerminalEcho failed get: ") + errnoToString(errno));
     if (!enable)
         tty.c_lflag &= ~ECHO;
     else
@@ -39,6 +40,6 @@ void setTerminalEcho(bool enable)
 
     auto ret = tcsetattr(STDIN_FILENO, TCSANOW, &tty);
     if (ret)
-        throw std::runtime_error(std::string("setTerminalEcho failed set: ") + strerror(errno));
+        throw std::runtime_error(std::string("setTerminalEcho failed set: ") + errnoToString(errno));
 #endif
 }
diff --git a/base/common/ya.make b/base/common/ya.make
index 9b38e3919be..fd88760c933 100644
--- a/base/common/ya.make
+++ b/base/common/ya.make
@@ -47,6 +47,7 @@ SRCS(
     errnoToString.cpp
     getFQDNOrHostName.cpp
     getMemoryAmount.cpp
+    getPageSize.cpp
     getResource.cpp
     getThreadId.cpp
     mremap.cpp
diff --git a/cmake/sanitize.cmake b/cmake/sanitize.cmake
index 7a9e263778e..72bdf27004e 100644
--- a/cmake/sanitize.cmake
+++ b/cmake/sanitize.cmake
@@ -41,9 +41,10 @@ if (SANITIZE)
         if (COMPILER_CLANG)
             set (TSAN_FLAGS "${TSAN_FLAGS} -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/tsan_suppressions.txt")
         else()
-            message (WARNING "TSAN suppressions was not passed to the compiler (since the compiler is not clang)")
-            message (WARNING "Use the following command to pass them manually:")
-            message (WARNING "    export TSAN_OPTIONS=\"$TSAN_OPTIONS suppressions=${CMAKE_SOURCE_DIR}/tests/tsan_suppressions.txt\"")
+            set (MESSAGE "TSAN suppressions was not passed to the compiler (since the compiler is not clang)\n")
+            set (MESSAGE "${MESSAGE}Use the following command to pass them manually:\n")
+            set (MESSAGE "${MESSAGE}    export TSAN_OPTIONS=\"$TSAN_OPTIONS suppressions=${CMAKE_SOURCE_DIR}/tests/tsan_suppressions.txt\"")
+            message (WARNING "${MESSAGE}")
         endif()
 
 
@@ -57,8 +58,18 @@ if (SANITIZE)
         endif ()
 
     elseif (SANITIZE STREQUAL "undefined")
-        set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SAN_FLAGS} -fsanitize=undefined -fno-sanitize-recover=all -fno-sanitize=float-divide-by-zero -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt")
-        set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SAN_FLAGS} -fsanitize=undefined -fno-sanitize-recover=all -fno-sanitize=float-divide-by-zero -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt")
+        set (UBSAN_FLAGS "-fsanitize=undefined -fno-sanitize-recover=all -fno-sanitize=float-divide-by-zero")
+        if (COMPILER_CLANG)
+            set (UBSAN_FLAGS "${UBSAN_FLAGS} -fsanitize-blacklist=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt")
+        else()
+            set (MESSAGE "UBSAN suppressions was not passed to the compiler (since the compiler is not clang)\n")
+            set (MESSAGE "${MESSAGE}Use the following command to pass them manually:\n")
+            set (MESSAGE "${MESSAGE}        export UBSAN_OPTIONS=\"$UBSAN_OPTIONS suppressions=${CMAKE_SOURCE_DIR}/tests/ubsan_suppressions.txt\"")
+            message (WARNING "${MESSAGE}")
+        endif()
+
+        set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SAN_FLAGS} ${UBSAN_FLAGS}")
+        set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SAN_FLAGS} ${UBSAN_FLAGS}")
         if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
             set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fsanitize=undefined")
         endif()
diff --git a/contrib/boost b/contrib/boost
index a7ceabe4747..0b98b443aa7 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit a7ceabe4747ecc3309dd3dcd9de4b29660dfd298
+Subproject commit 0b98b443aa7bb77d65efd7b23b3b8c8a0ab5f1f3
diff --git a/contrib/krb5 b/contrib/krb5
index 99f7ad2831a..90ff6f4f8c6 160000
--- a/contrib/krb5
+++ b/contrib/krb5
@@ -1 +1 @@
-Subproject commit 99f7ad2831a01f264c07eed42a0a3a9336b86184
+Subproject commit 90ff6f4f8c695d6bf1aaba78a9b8942be92141c2
diff --git a/contrib/libc-headers b/contrib/libc-headers
index 92c74f938cf..a720b7105a6 160000
--- a/contrib/libc-headers
+++ b/contrib/libc-headers
@@ -1 +1 @@
-Subproject commit 92c74f938cf2c4dd529cae4f3d2923d153b029a7
+Subproject commit a720b7105a610acbd7427eea475a5b6810c151eb
diff --git a/contrib/libgsasl b/contrib/libgsasl
index 140fb582505..383ee28e82f 160000
--- a/contrib/libgsasl
+++ b/contrib/libgsasl
@@ -1 +1 @@
-Subproject commit 140fb58250588c8323285b75fcf127c4adc33dfa
+Subproject commit 383ee28e82f69fa16ed43b48bd9c8ee5b313ab84
diff --git a/contrib/libhdfs3 b/contrib/libhdfs3
index 30552ac527f..095b9d48b40 160000
--- a/contrib/libhdfs3
+++ b/contrib/libhdfs3
@@ -1 +1 @@
-Subproject commit 30552ac527f2c14070d834e171493b2e7f662375
+Subproject commit 095b9d48b400abb72d967cb0539af13b1e3d90cf
diff --git a/contrib/libhdfs3-cmake/CMakeLists.txt b/contrib/libhdfs3-cmake/CMakeLists.txt
index 3aa5cb970db..60f4376bdea 100644
--- a/contrib/libhdfs3-cmake/CMakeLists.txt
+++ b/contrib/libhdfs3-cmake/CMakeLists.txt
@@ -17,7 +17,12 @@ if (NOT USE_INTERNAL_PROTOBUF_LIBRARY AND PROTOBUF_OLD_ABI_COMPAT)
     endif ()
 endif()
 
-set(WITH_KERBEROS false)
+if (${ENABLE_LIBRARIES} AND ${ENABLE_KRB5})
+    SET(WITH_KERBEROS 1)
+else()
+    SET(WITH_KERBEROS 0)
+endif()
+
 # project and source dir
 set(HDFS3_ROOT_DIR ${ClickHouse_SOURCE_DIR}/contrib/libhdfs3)
 set(HDFS3_SOURCE_DIR ${HDFS3_ROOT_DIR}/src)
@@ -28,11 +33,6 @@ set(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMake" ${CMAKE_MODULE_PATH})
 include(Platform)
 include(Options)
 
-# prefer shared libraries
-if (WITH_KERBEROS)
-    find_package(KERBEROS REQUIRED)
-endif()
-
 # source
 set(PROTO_FILES
     #${HDFS3_SOURCE_DIR}/proto/encryption.proto
@@ -207,14 +207,11 @@ target_include_directories(hdfs3 PRIVATE ${HDFS3_COMMON_DIR})
 target_include_directories(hdfs3 PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 
 target_include_directories(hdfs3 PRIVATE ${LIBGSASL_INCLUDE_DIR})
-if (WITH_KERBEROS)
-    target_include_directories(hdfs3 PRIVATE ${KERBEROS_INCLUDE_DIRS})
-endif()
 target_include_directories(hdfs3 PRIVATE ${LIBXML2_INCLUDE_DIR})
 
 target_link_libraries(hdfs3 PRIVATE ${LIBGSASL_LIBRARY})
 if (WITH_KERBEROS)
-    target_link_libraries(hdfs3 PRIVATE ${KERBEROS_LIBRARIES})
+    target_link_libraries(hdfs3 PRIVATE ${KRB5_LIBRARY})
 endif()
 target_link_libraries(hdfs3 PRIVATE ${LIBXML2_LIBRARIES})
 
diff --git a/contrib/librdkafka b/contrib/librdkafka
index 9902bc4fb18..f2f6616419d 160000
--- a/contrib/librdkafka
+++ b/contrib/librdkafka
@@ -1 +1 @@
-Subproject commit 9902bc4fb18bb441fa55ca154b341cdda191e5d3
+Subproject commit f2f6616419d567c9198aef0d1133a2e9b4f02276
diff --git a/contrib/libunwind b/contrib/libunwind
index 51b84d9b6d2..8fe25d7dc70 160000
--- a/contrib/libunwind
+++ b/contrib/libunwind
@@ -1 +1 @@
-Subproject commit 51b84d9b6d2548f1cbdcafe622d5a753853b6149
+Subproject commit 8fe25d7dc70f2a4ea38c3e5a33fa9d4199b67a5a
diff --git a/debian/control b/debian/control
index 9b34e982698..809b5bb5d2a 100644
--- a/debian/control
+++ b/debian/control
@@ -40,7 +40,7 @@ Description: Common files for ClickHouse
 Package: clickhouse-server
 Architecture: all
 Depends: ${shlibs:Depends}, ${misc:Depends}, clickhouse-common-static (= ${binary:Version}), adduser
-Recommends: libcap2-bin
+Recommends: libcap2-bin, krb5-user
 Replaces: clickhouse-server-common, clickhouse-server-base
 Provides: clickhouse-server-common
 Description: Server binary for ClickHouse
diff --git a/docker/builder/Dockerfile b/docker/builder/Dockerfile
index 68245a92c58..199b5217d79 100644
--- a/docker/builder/Dockerfile
+++ b/docker/builder/Dockerfile
@@ -1,6 +1,6 @@
-FROM ubuntu:19.10
+FROM ubuntu:20.04
 
-ENV DEBIAN_FRONTEND=noninteractive LLVM_VERSION=10
+ENV DEBIAN_FRONTEND=noninteractive LLVM_VERSION=11
 
 RUN apt-get update \
     && apt-get install ca-certificates lsb-release wget gnupg apt-transport-https \
diff --git a/docker/builder/build.sh b/docker/builder/build.sh
index d814bcdf2b4..d4cf662e91b 100755
--- a/docker/builder/build.sh
+++ b/docker/builder/build.sh
@@ -4,7 +4,7 @@ set -e
 #ccache -s # uncomment to display CCache statistics
 mkdir -p /server/build_docker
 cd /server/build_docker
-cmake -G Ninja /server "-DCMAKE_C_COMPILER=$(command -v gcc-9)" "-DCMAKE_CXX_COMPILER=$(command -v g++-9)"
+cmake -G Ninja /server "-DCMAKE_C_COMPILER=$(command -v clang-11)" "-DCMAKE_CXX_COMPILER=$(command -v clang++-11)"
 
 # Set the number of build jobs to the half of number of virtual CPU cores (rounded up).
 # By default, ninja use all virtual CPU cores, that leads to very high memory consumption without much improvement in build time.
diff --git a/docker/images.json b/docker/images.json
index f5b10a14313..5657a748579 100644
--- a/docker/images.json
+++ b/docker/images.json
@@ -58,8 +58,7 @@
     "docker/test/stateless": {
         "name": "yandex/clickhouse-stateless-test",
         "dependent": [
-            "docker/test/stateful",
-            "docker/test/stateful_with_coverage"
+            "docker/test/stateful"
         ]
     },
     "docker/test/stateless_pytest": {
@@ -68,7 +67,9 @@
     },
     "docker/test/stateless_with_coverage": {
         "name": "yandex/clickhouse-stateless-test-with-coverage",
-        "dependent": []
+        "dependent": [
+            "docker/test/stateful_with_coverage"
+        ]
     },
     "docker/test/unit": {
         "name": "yandex/clickhouse-unit-test",
@@ -157,5 +158,9 @@
          "name": "yandex/clickhouse-stateless-unbundled-test",
          "dependent": [
          ]
+    },
+    "docker/test/integration/kerberized_hadoop": {
+        "name": "yandex/clickhouse-kerberized-hadoop",
+        "dependent": []
     }
 }
diff --git a/docker/packager/binary/build.sh b/docker/packager/binary/build.sh
index 4ba2f5f87c7..c46e3b23994 100755
--- a/docker/packager/binary/build.sh
+++ b/docker/packager/binary/build.sh
@@ -31,6 +31,7 @@ find . -name '*.so.*' -print -exec mv '{}' /output \;
 if [ "performance" == "$COMBINED_OUTPUT" ]
 then
     cp -r ../tests/performance /output
+    cp -r ../tests/config/top_level_domains  /
     cp -r ../docker/test/performance-comparison/config /output ||:
     rm /output/unit_tests_dbms ||:
     rm /output/clickhouse-odbc-bridge ||:
diff --git a/docker/packager/packager b/docker/packager/packager
index 6d075195003..65c03cc10e3 100755
--- a/docker/packager/packager
+++ b/docker/packager/packager
@@ -148,6 +148,10 @@ def parse_env_variables(build_type, compiler, sanitizer, package_type, image_typ
 
     if split_binary:
         cmake_flags.append('-DUSE_STATIC_LIBRARIES=0 -DSPLIT_SHARED_LIBRARIES=1 -DCLICKHOUSE_SPLIT_BINARY=1')
+        # We can't always build utils because it requires too much space, but
+        # we have to build them at least in some way in CI. The split build is
+        # probably the least heavy disk-wise.
+        cmake_flags.append('-DENABLE_UTILS=1')
 
     if clang_tidy:
         cmake_flags.append('-DENABLE_CLANG_TIDY=1')
diff --git a/docker/server/README.md b/docker/server/README.md
index e8e8d326de7..d8e9204dffa 100644
--- a/docker/server/README.md
+++ b/docker/server/README.md
@@ -15,6 +15,8 @@ For more information and documentation see https://clickhouse.yandex/.
 $ docker run -d --name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
 ```
 
+By default ClickHouse will be accessible only via docker network. See the [networking section below](#networking).
+
 ### connect to it from a native client
 ```bash
 $ docker run -it --rm --link some-clickhouse-server:clickhouse-server yandex/clickhouse-client --host clickhouse-server
@@ -22,6 +24,70 @@ $ docker run -it --rm --link some-clickhouse-server:clickhouse-server yandex/cli
 
 More information about [ClickHouse client](https://clickhouse.yandex/docs/en/interfaces/cli/).
 
+### connect to it using curl
+
+```bash
+echo "SELECT 'Hello, ClickHouse!'" | docker run -i --rm --link some-clickhouse-server:clickhouse-server curlimages/curl 'http://clickhouse-server:8123/?query=' -s --data-binary @-
+```
+More information about [ClickHouse HTTP Interface](https://clickhouse.tech/docs/en/interfaces/http/).
+
+### stopping / removing the containter
+
+```bash
+$ docker stop some-clickhouse-server
+$ docker rm some-clickhouse-server
+```
+
+### networking
+
+You can expose you ClickHouse running in docker by [mapping particular port](https://docs.docker.com/config/containers/container-networking/) from inside container to a host ports:
+
+```bash
+$ docker run -d -p 18123:8123 -p19000:9000 --name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+$ echo 'SELECT version()' | curl 'http://localhost:18123/' --data-binary @-
+20.12.3.3
+```
+
+or by allowing container to use [host ports directly](https://docs.docker.com/network/host/) using `--network=host` (also allows archiving better network performance):
+
+```bash
+$ docker run -d --network=host --name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+$ echo 'SELECT version()' | curl 'http://localhost:8123/' --data-binary @-
+20.12.3.3
+```
+
+### Volumes 
+
+Typically you may want to mount the following folders inside your container to archieve persistency:
+
+* `/var/lib/clickhouse/` - main folder where ClickHouse stores the data
+* `/val/log/clickhouse-server/` - logs
+
+```bash
+$ docker run -d \
+	-v $(realpath ./ch_data):/var/lib/clickhouse/ \
+	-v $(realpath ./ch_logs):/var/log/clickhouse-server/ \
+	--name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+```
+
+You may also want to mount:
+
+* `/etc/clickhouse-server/config.d/*.xml` - files with server configuration adjustmenets
+* `/etc/clickhouse-server/usert.d/*.xml` - files with use settings adjustmenets
+* `/docker-entrypoint-initdb.d/` - folder with database initialization scripts (see below).
+
+### Linux capabilities 
+
+ClickHouse has some advanced functionality which requite enabling several [linux capabilities](https://man7.org/linux/man-pages/man7/capabilities.7.html).
+
+It is optional and can be enabled using the following [docker command line agruments](https://docs.docker.com/engine/reference/run/#runtime-privilege-and-linux-capabilities):
+
+```bash
+$ docker run -d \
+	--cap-add=SYS_NICE --cap-add=NET_ADMIN --cap-add=IPC_LOCK \
+	--name some-clickhouse-server --ulimit nofile=262144:262144 yandex/clickhouse-server
+```
+
 ## Configuration
 
 Container exposes 8123 port for [HTTP interface](https://clickhouse.yandex/docs/en/interfaces/http_interface/) and 9000 port for [native client](https://clickhouse.yandex/docs/en/interfaces/tcp/).
diff --git a/docker/test/base/Dockerfile b/docker/test/base/Dockerfile
index 61a40673a96..e8653c2122e 100644
--- a/docker/test/base/Dockerfile
+++ b/docker/test/base/Dockerfile
@@ -1,5 +1,5 @@
 # docker build -t yandex/clickhouse-test-base .
-FROM ubuntu:19.10
+FROM ubuntu:20.04
 
 ENV DEBIAN_FRONTEND=noninteractive LLVM_VERSION=11
 
diff --git a/docker/test/integration/kerberized_hadoop/Dockerfile b/docker/test/integration/kerberized_hadoop/Dockerfile
new file mode 100644
index 00000000000..6a2fd96e7a7
--- /dev/null
+++ b/docker/test/integration/kerberized_hadoop/Dockerfile
@@ -0,0 +1,18 @@
+# docker build -t yandex/clickhouse-kerberized-hadoop .
+
+FROM sequenceiq/hadoop-docker:2.7.0
+RUN sed -i -e 's/^\#baseurl/baseurl/' /etc/yum.repos.d/CentOS-Base.repo
+RUN sed -i -e 's/^mirrorlist/#mirrorlist/' /etc/yum.repos.d/CentOS-Base.repo
+RUN sed -i -e 's#http://mirror.centos.org/#http://vault.centos.org/#' /etc/yum.repos.d/CentOS-Base.repo
+RUN yum clean all && \
+	rpm --rebuilddb &&	\
+	yum -y update && \
+	yum -y install yum-plugin-ovl && \
+	yum --quiet -y install krb5-workstation.x86_64
+RUN cd /tmp && \
+	curl http://archive.apache.org/dist/commons/daemon/source/commons-daemon-1.0.15-src.tar.gz   -o  commons-daemon-1.0.15-src.tar.gz && \
+	tar xzf commons-daemon-1.0.15-src.tar.gz && \
+	cd commons-daemon-1.0.15-src/src/native/unix && \
+	./configure && \
+	make && \
+	cp ./jsvc /usr/local/hadoop/sbin
diff --git a/docker/test/integration/runner/Dockerfile b/docker/test/integration/runner/Dockerfile
index 36188fc4a63..9b51891ccf5 100644
--- a/docker/test/integration/runner/Dockerfile
+++ b/docker/test/integration/runner/Dockerfile
@@ -29,6 +29,8 @@ RUN apt-get update \
     libcurl4-openssl-dev \
     gdb \
     software-properties-common \
+    libkrb5-dev \
+    krb5-user \
     && rm -rf \
         /var/lib/apt/lists/* \
         /var/cache/debconf \
@@ -75,7 +77,8 @@ RUN python3 -m pip install \
     pytest-timeout \
     redis \
     tzlocal \
-    urllib3
+    urllib3 \
+    requests-kerberos
 
 COPY modprobe.sh /usr/local/bin/modprobe
 COPY dockerd-entrypoint.sh /usr/local/bin/
diff --git a/docker/test/integration/runner/compose/docker_compose_hdfs.yml b/docker/test/integration/runner/compose/docker_compose_hdfs.yml
index d19aacfc8d4..b8cd7f64273 100644
--- a/docker/test/integration/runner/compose/docker_compose_hdfs.yml
+++ b/docker/test/integration/runner/compose/docker_compose_hdfs.yml
@@ -2,6 +2,7 @@ version: '2.3'
 services:
     hdfs1:
         image: sequenceiq/hadoop-docker:2.7.0
+        hostname: hdfs1
         restart: always
         ports:
           - 50075:50075
diff --git a/docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml b/docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml
new file mode 100644
index 00000000000..a74476613f3
--- /dev/null
+++ b/docker/test/integration/runner/compose/docker_compose_kerberized_hdfs.yml
@@ -0,0 +1,29 @@
+version: '2.3'
+
+services:
+  kerberizedhdfs1:
+    cap_add:
+      - DAC_READ_SEARCH
+    image: yandex/clickhouse-kerberized-hadoop:16621
+    hostname: kerberizedhdfs1
+    restart: always
+    volumes:
+        - ${KERBERIZED_HDFS_DIR}/../../hdfs_configs/bootstrap.sh:/etc/bootstrap.sh:ro
+        - ${KERBERIZED_HDFS_DIR}/secrets:/usr/local/hadoop/etc/hadoop/conf
+        - ${KERBERIZED_HDFS_DIR}/secrets/krb_long.conf:/etc/krb5.conf:ro
+    ports:
+      - 1006:1006
+      - 50070:50070
+      - 9000:9000
+    depends_on:
+        - hdfskerberos
+    entrypoint: /etc/bootstrap.sh -d
+
+  hdfskerberos:
+    image: yandex/clickhouse-kerberos-kdc:${DOCKER_KERBEROS_KDC_TAG}
+    hostname: hdfskerberos
+    volumes:
+        - ${KERBERIZED_HDFS_DIR}/secrets:/tmp/keytab
+        - ${KERBERIZED_HDFS_DIR}/../../kerberos_image_config.sh:/config.sh
+        - /dev/urandom:/dev/random
+    ports: [88, 749]
diff --git a/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml b/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml
index 81dab1a48b0..ee2006201b0 100644
--- a/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml
+++ b/docker/test/performance-comparison/config/config.d/zzz-perf-comparison-tweaks-config.xml
@@ -3,6 +3,7 @@
     <mysql_port remove="remove"/>
     <interserver_http_port remove="remove"/>
     <tcp_with_proxy_port remove="remove"/>
+    <test_keeper_server remove="remove"/>
     <listen_host>::</listen_host>
 
     <logger>
diff --git a/docker/test/stateful_with_coverage/Dockerfile b/docker/test/stateful_with_coverage/Dockerfile
index ac6645b9463..07acf4ed4e6 100644
--- a/docker/test/stateful_with_coverage/Dockerfile
+++ b/docker/test/stateful_with_coverage/Dockerfile
@@ -1,8 +1,6 @@
 # docker build -t yandex/clickhouse-stateful-test-with-coverage .
 FROM yandex/clickhouse-stateless-test-with-coverage
 
-RUN echo "deb [trusted=yes] http://apt.llvm.org/bionic/ llvm-toolchain-bionic-9 main" >> /etc/apt/sources.list
-
 RUN apt-get update -y \
     && env DEBIAN_FRONTEND=noninteractive \
         apt-get install --yes --no-install-recommends \
diff --git a/docs/en/development/build-osx.md b/docs/en/development/build-osx.md
index 7a4387f073b..c3a0a540b6d 100644
--- a/docs/en/development/build-osx.md
+++ b/docs/en/development/build-osx.md
@@ -35,10 +35,12 @@ $ cd ClickHouse
 
 ## Build ClickHouse {#build-clickhouse}
 
+> Please note: ClickHouse doesn't support build with native Apple Clang compiler, we need use clang from LLVM.
+
 ``` bash
 $ mkdir build
 $ cd build
-$ cmake .. -DCMAKE_CXX_COMPILER=`which clang++` -DCMAKE_C_COMPILER=`which clang`
+$ cmake ..-DCMAKE_C_COMPILER=`brew --prefix llvm`/bin/clang -DCMAKE_CXX_COMPILER=`brew --prefix llvm`/bin/clang++ -DCMAKE_PREFIX_PATH=`brew --prefix llvm`
 $ ninja
 $ cd ..
 ```
diff --git a/docs/en/development/developer-instruction.md b/docs/en/development/developer-instruction.md
index dc95c3ec50b..5511e8e19c7 100644
--- a/docs/en/development/developer-instruction.md
+++ b/docs/en/development/developer-instruction.md
@@ -253,8 +253,8 @@ Developing ClickHouse often requires loading realistic datasets. It is particula
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/en/engines/table-engines/integrations/hdfs.md b/docs/en/engines/table-engines/integrations/hdfs.md
index f36a4771222..5c36e3f1c21 100644
--- a/docs/en/engines/table-engines/integrations/hdfs.md
+++ b/docs/en/engines/table-engines/integrations/hdfs.md
@@ -108,6 +108,95 @@ Create table with files named `file000`, `file001`, … , `file999`:
 ``` sql
 CREATE TABLE big_table (name String, value UInt32) ENGINE = HDFS('hdfs://hdfs1:9000/big_dir/file{0..9}{0..9}{0..9}', 'CSV')
 ```
+## Configuration {#configuration}
+
+Similar to GraphiteMergeTree, the HDFS engine supports extended configuration using the ClickHouse config file. There are two configuration keys that you can use: global (`hdfs`) and user-level (`hdfs_*`). The global configuration is applied first, and then the user-level configuration is applied (if it exists).
+
+``` xml
+  <!-- Global configuration options for HDFS engine type -->
+  <hdfs>
+	<hadoop_kerberos_keytab>/tmp/keytab/clickhouse.keytab</hadoop_kerberos_keytab>
+	<hadoop_kerberos_principal>clickuser@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+	<hadoop_security_authentication>kerberos</hadoop_security_authentication>
+  </hdfs>
+
+  <!-- Configuration specific for user "root" -->
+  <hdfs_root>
+	<hadoop_kerberos_principal>root@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+  </hdfs_root>
+```
+
+### List of possible configuration options with default values
+#### Supported by libhdfs3
+
+
+| **parameter**                                         | **default value**       |
+| rpc\_client\_connect\_tcpnodelay                      | true                    |
+| dfs\_client\_read\_shortcircuit                       | true                    |
+| output\_replace-datanode-on-failure                   | true                    |
+| input\_notretry-another-node                          | false                   |
+| input\_localread\_mappedfile                          | true                    |
+| dfs\_client\_use\_legacy\_blockreader\_local          | false                   |
+| rpc\_client\_ping\_interval                           | 10  * 1000              |
+| rpc\_client\_connect\_timeout                         | 600 * 1000              |
+| rpc\_client\_read\_timeout                            | 3600 * 1000             |
+| rpc\_client\_write\_timeout                           | 3600 * 1000             |
+| rpc\_client\_socekt\_linger\_timeout                  | -1                      |
+| rpc\_client\_connect\_retry                           | 10                      |
+| rpc\_client\_timeout                                  | 3600 * 1000             |
+| dfs\_default\_replica                                 | 3                       |
+| input\_connect\_timeout                               | 600 * 1000              |
+| input\_read\_timeout                                  | 3600 * 1000             |
+| input\_write\_timeout                                 | 3600 * 1000             |
+| input\_localread\_default\_buffersize                 | 1 * 1024 * 1024         |
+| dfs\_prefetchsize                                     | 10                      |
+| input\_read\_getblockinfo\_retry                      | 3                       |
+| input\_localread\_blockinfo\_cachesize                | 1000                    |
+| input\_read\_max\_retry                               | 60                      |
+| output\_default\_chunksize                            | 512                     |
+| output\_default\_packetsize                           | 64 * 1024               |
+| output\_default\_write\_retry                         | 10                      |
+| output\_connect\_timeout                              | 600 * 1000              |
+| output\_read\_timeout                                 | 3600 * 1000             |
+| output\_write\_timeout                                | 3600 * 1000             |
+| output\_close\_timeout                                | 3600 * 1000             |
+| output\_packetpool\_size                              | 1024                    |
+| output\_heeartbeat\_interval                          | 10 * 1000               |
+| dfs\_client\_failover\_max\_attempts                  | 15                      |
+| dfs\_client\_read\_shortcircuit\_streams\_cache\_size | 256                     |
+| dfs\_client\_socketcache\_expiryMsec                  | 3000                    |
+| dfs\_client\_socketcache\_capacity                    | 16                      |
+| dfs\_default\_blocksize                               | 64 * 1024 * 1024        |
+| dfs\_default\_uri                                     | "hdfs://localhost:9000" |
+| hadoop\_security\_authentication                      | "simple"                |
+| hadoop\_security\_kerberos\_ticket\_cache\_path       | ""                      |
+| dfs\_client\_log\_severity                            | "INFO"                  |
+| dfs\_domain\_socket\_path                             | ""                      |
+
+
+[HDFS Configuration Reference ](https://hawq.apache.org/docs/userguide/2.3.0.0-incubating/reference/HDFSConfigurationParameterReference.html) might explain some parameters.
+
+
+#### ClickHouse extras {#clickhouse-extras}
+
+| **parameter**                                         | **default value**       |
+|hadoop\_kerberos\_keytab                               | ""                      |
+|hadoop\_kerberos\_principal                            | ""                      |
+|hadoop\_kerberos\_kinit\_command                       | kinit                   |
+
+#### Limitations {#limitations}
+
+  * hadoop\_security\_kerberos\_ticket\_cache\_path can be global only, not user specific
+
+## Kerberos support {#kerberos-support}
+
+If hadoop\_security\_authentication parameter has value 'kerberos', ClickHouse authentifies via Kerberos facility.
+Parameters [here](#clickhouse-extras) and hadoop\_security\_kerberos\_ticket\_cache\_path may be of help.
+Note that due to libhdfs3 limitations only old-fashioned approach is supported,
+datanode communications are not secured by SASL (HADOOP\_SECURE\_DN\_USER is a reliable indicator of such
+security approach). Use tests/integration/test\_storage\_kerberized\_hdfs/hdfs_configs/bootstrap.sh for reference.
+
+If hadoop\_kerberos\_keytab, hadoop\_kerberos\_principal or hadoop\_kerberos\_kinit\_command is specified, kinit will be invoked. hadoop\_kerberos\_keytab and hadoop\_kerberos\_principal are mandatory in this case. kinit tool and krb5 configuration files are required.
 
 ## Virtual Columns {#virtual-columns}
 
diff --git a/docs/en/getting-started/example-datasets/index.md b/docs/en/getting-started/example-datasets/index.md
index b769e8fcb45..ed795a6c4de 100644
--- a/docs/en/getting-started/example-datasets/index.md
+++ b/docs/en/getting-started/example-datasets/index.md
@@ -10,6 +10,7 @@ This section describes how to obtain example datasets and import them into Click
 
 The list of documented datasets:
 
+-   [GitHub Events](../../getting-started/example-datasets/github-events.md)
 -   [Anonymized Yandex.Metrica Dataset](../../getting-started/example-datasets/metrica.md)
 -   [Star Schema Benchmark](../../getting-started/example-datasets/star-schema.md)
 -   [WikiStat](../../getting-started/example-datasets/wikistat.md)
diff --git a/docs/en/getting-started/example-datasets/metrica.md b/docs/en/getting-started/example-datasets/metrica.md
index b036973b255..cdbb9b56eeb 100644
--- a/docs/en/getting-started/example-datasets/metrica.md
+++ b/docs/en/getting-started/example-datasets/metrica.md
@@ -7,14 +7,14 @@ toc_title: Yandex.Metrica Data
 
 Dataset consists of two tables containing anonymized data about hits (`hits_v1`) and visits (`visits_v1`) of Yandex.Metrica. You can read more about Yandex.Metrica in [ClickHouse history](../../introduction/history.md) section.
 
-The dataset consists of two tables, either of them can be downloaded as a compressed `tsv.xz` file or as prepared partitions. In addition to that, an extended version of the `hits` table containing 100 million rows is available as TSV at https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz and as prepared partitions at https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
+The dataset consists of two tables, either of them can be downloaded as a compressed `tsv.xz` file or as prepared partitions. In addition to that, an extended version of the `hits` table containing 100 million rows is available as TSV at https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz and as prepared partitions at https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
 
 ## Obtaining Tables from Prepared Partitions {#obtaining-tables-from-prepared-partitions}
 
 Download and import hits table:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -24,7 +24,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Download and import visits:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -36,7 +36,10 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 Download and import hits from compressed TSV file:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+# Validate the checksum
+md5sum hits_v1.tsv
+# Checksum should be equal to: f3631b6295bf06989c1437491f7592cb
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -50,7 +53,10 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Download and import visits from compressed tsv-file:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+# Validate the checksum
+md5sum visits_v1.tsv
+# Checksum should be equal to: 6dafe1a0f24e59e3fc2d0fed85601de6
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/en/getting-started/example-datasets/nyc-taxi.md b/docs/en/getting-started/example-datasets/nyc-taxi.md
index 9b9a12ba724..38a87a674f9 100644
--- a/docs/en/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/en/getting-started/example-datasets/nyc-taxi.md
@@ -283,7 +283,7 @@ Among other things, you can run the OPTIMIZE query on MergeTree. But it’s not
 ## Download of Prepared Partitions {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/en/getting-started/example-datasets/ontime.md b/docs/en/getting-started/example-datasets/ontime.md
index 6f2408af3b6..5e499cafb2a 100644
--- a/docs/en/getting-started/example-datasets/ontime.md
+++ b/docs/en/getting-started/example-datasets/ontime.md
@@ -154,7 +154,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## Download of Prepared Partitions {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/en/getting-started/tutorial.md b/docs/en/getting-started/tutorial.md
index 3e051456a75..254fa37d47d 100644
--- a/docs/en/getting-started/tutorial.md
+++ b/docs/en/getting-started/tutorial.md
@@ -85,8 +85,8 @@ Now it’s time to fill our ClickHouse server with some sample data. In this tut
 ### Download and Extract Table Data {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 The extracted files are about 10GB in size.
diff --git a/docs/en/interfaces/cli.md b/docs/en/interfaces/cli.md
index a2ea6edf24a..7700ed36ca5 100644
--- a/docs/en/interfaces/cli.md
+++ b/docs/en/interfaces/cli.md
@@ -57,7 +57,7 @@ The command line is based on ‘replxx’ (similar to ‘readline’). In other
 
 By default, the format used is PrettyCompact. You can change the format in the FORMAT clause of the query, or by specifying `\G` at the end of the query, using the `--format` or `--vertical` argument in the command line, or using the client configuration file.
 
-To exit the client, press Ctrl+D (or Ctrl+C), or enter one of the following instead of a query: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+To exit the client, press Ctrl+D, or enter one of the following instead of a query: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 When processing a query, the client shows:
 
diff --git a/docs/en/introduction/performance.md b/docs/en/introduction/performance.md
index d2780aedccb..6e5710347a1 100644
--- a/docs/en/introduction/performance.md
+++ b/docs/en/introduction/performance.md
@@ -25,6 +25,6 @@ Under the same conditions, ClickHouse can handle several hundred queries per sec
 
 ## Performance When Inserting Data {#performance-when-inserting-data}
 
-We recommend inserting data in packets of at least 1000 rows, or no more than a single request per second. When inserting to a MergeTree table from a tab-separated dump, the insertion speed can be from 50 to 200 MB/s. If the inserted rows are around 1 Kb in size, the speed will be from 50,000 to 200,000 rows per second. If the rows are small, the performance can be higher in rows per second (on Banner System data -`>` 500,000 rows per second; on Graphite data -`>` 1,000,000 rows per second). To improve performance, you can make multiple INSERT queries in parallel, which scales linearly.
+We recommend inserting data in packets of at least 1000 rows, or no more than a single request per second. When inserting to a MergeTree table from a tab-separated dump, the insertion speed can be from 50 to 200 MB/s. If the inserted rows are around 1 KB in size, the speed will be from 50,000 to 200,000 rows per second. If the rows are small, the performance can be higher in rows per second (on Banner System data -`>` 500,000 rows per second; on Graphite data -`>` 1,000,000 rows per second). To improve performance, you can make multiple INSERT queries in parallel, which scales linearly.
 
 {## [Original article](https://clickhouse.tech/docs/en/introduction/performance/) ##}
diff --git a/docs/en/operations/performance-test.md b/docs/en/operations/performance-test.md
index 984bbe02174..ca805923ba9 100644
--- a/docs/en/operations/performance-test.md
+++ b/docs/en/operations/performance-test.md
@@ -27,7 +27,7 @@ wget https://raw.githubusercontent.com/ClickHouse/ClickHouse/master/benchmark/cl
 ```
 6.  Download test data according to the [Yandex.Metrica dataset](../getting-started/example-datasets/metrica.md) instruction (“hits” table containing 100 million rows).
 ```bash
-wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
 tar xvf hits_100m_obfuscated_v1.tar.xz -C .
 mv hits_100m_obfuscated_v1/* .
 ```
diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index fc921f2ef7e..093e4e50dfe 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -1093,9 +1093,14 @@ See the section “WITH TOTALS modifier”.
 
 ## max_parallel_replicas {#settings-max_parallel_replicas}
 
-The maximum number of replicas for each shard when executing a query.
-For consistency (to get different parts of the same data split), this option only works when the sampling key is set.
-Replica lag is not controlled.
+The maximum number of replicas for each shard when executing a query. In limited circumstances, this can make a query faster by executing it on more servers. This setting is only useful for replicated tables with a sampling key. There are cases where performance will not improve or even worsen:
+
+- the position of the sampling key in the partitioning key's order doesn't allow efficient range scans
+- adding a sampling key to the table makes filtering by other columns less efficient
+- the sampling key is an expression that is expensive to calculate
+- the cluster's latency distribution has a long tail, so that querying more servers increases the query's overall latency
+
+In addition, this setting will produce incorrect results when joins or subqueries are involved, and all tables don't meet certain conditions. See [Distributed Subqueries and max_parallel_replicas](../../sql-reference/operators/in.md/#max_parallel_replica-subqueries) for more details.
 
 ## compile {#compile}
 
diff --git a/docs/en/operations/system-tables/distribution_queue.md b/docs/en/operations/system-tables/distribution_queue.md
index da4effd0ff5..9068e7e860c 100644
--- a/docs/en/operations/system-tables/distribution_queue.md
+++ b/docs/en/operations/system-tables/distribution_queue.md
@@ -39,4 +39,8 @@ data_compressed_bytes: 499
 last_exception:        
 ```
 
+**See Also**
+
+-   [Distributed table engine](../../engines/table-engines/special/distributed.md)
+
 [Original article](https://clickhouse.tech/docs/en/operations/system_tables/distribution_queue) <!--hide-->
diff --git a/docs/en/operations/system-tables/replication_queue.md b/docs/en/operations/system-tables/replication_queue.md
new file mode 100644
index 00000000000..aa379caa46c
--- /dev/null
+++ b/docs/en/operations/system-tables/replication_queue.md
@@ -0,0 +1,81 @@
+# system.replication_queue {#system_tables-replication_queue}
+
+Contains information about tasks from replication queues stored in ZooKeeper for tables in the `ReplicatedMergeTree` family.
+
+Columns:
+
+-   `database` ([String](../../sql-reference/data-types/string.md)) — Name of the database.
+
+-   `table` ([String](../../sql-reference/data-types/string.md)) — Name of the table.
+
+-   `replica_name` ([String](../../sql-reference/data-types/string.md)) — Replica name in ZooKeeper. Different replicas of the same table have different names.
+
+-   `position` ([UInt32](../../sql-reference/data-types/int-uint.md)) — Position of the task in the queue.
+
+-   `node_name` ([String](../../sql-reference/data-types/string.md)) — Node name in ZooKeeper.
+
+-   `type` ([String](../../sql-reference/data-types/string.md)) — Type of the task in the queue: `GET_PARTS`, `MERGE_PARTS`, `DETACH_PARTS`, `DROP_PARTS`, or `MUTATE_PARTS`.
+
+-   `create_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — Date and time when the task was submitted for execution.
+
+-   `required_quorum` ([UInt32](../../sql-reference/data-types/int-uint.md)) — The number of replicas waiting for the task to complete with confirmation of completion. This column is only relevant for the `GET_PARTS` task.
+
+-   `source_replica` ([String](../../sql-reference/data-types/string.md)) — Name of the source replica.
+
+-   `new_part_name` ([String](../../sql-reference/data-types/string.md)) — Name of the new part.
+
+-   `parts_to_merge` ([Array](../../sql-reference/data-types/array.md) ([String](../../sql-reference/data-types/string.md))) — Names of parts to merge or update.
+
+-   `is_detach` ([UInt8](../../sql-reference/data-types/int-uint.md)) — The flag indicates whether the `DETACH_PARTS` task is in the queue.
+
+-   `is_currently_executing` ([UInt8](../../sql-reference/data-types/int-uint.md)) — The flag indicates whether a specific task is being performed right now.
+
+-   `num_tries` ([UInt32](../../sql-reference/data-types/int-uint.md)) — The number of failed attempts to complete the task.
+
+-   `last_exception` ([String](../../sql-reference/data-types/string.md)) — Text message about the last error that occurred (if any).
+
+-   `last_attempt_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — Date and time when the task was last attempted.
+
+-   `num_postponed` ([UInt32](../../sql-reference/data-types/int-uint.md)) — The number of postponed tasks.
+
+-   `postpone_reason` ([String](../../sql-reference/data-types/string.md)) — The reason why the task was postponed.
+
+-   `last_postpone_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — Date and time when the task was last postponed.
+
+-   `merge_type` ([String](../../sql-reference/data-types/string.md)) — Type of the current merge. Empty if it's a mutation.
+
+**Example**
+
+``` sql
+SELECT * FROM system.replication_queue LIMIT 1 FORMAT Vertical;
+```
+
+``` text
+Row 1:
+──────
+database:               merge
+table:                  visits_v2
+replica_name:           mtgiga001-1t.metrika.yandex.net
+position:               15
+node_name:              queue-0009325559
+type:                   MERGE_PARTS
+create_time:            2020-12-07 14:04:21
+required_quorum:        0
+source_replica:         mtgiga001-1t.metrika.yandex.net
+new_part_name:          20201130_121373_121384_2
+parts_to_merge:         ['20201130_121373_121378_1','20201130_121379_121379_0','20201130_121380_121380_0','20201130_121381_121381_0','20201130_121382_121382_0','20201130_121383_121383_0','20201130_121384_121384_0']
+is_detach:              0
+is_currently_executing: 0
+num_tries:              36
+last_exception:         Code: 226, e.displayText() = DB::Exception: Marks file '/opt/clickhouse/data/merge/visits_v2/tmp_fetch_20201130_121373_121384_2/CounterID.mrk' doesn't exist (version 20.8.7.15 (official build))
+last_attempt_time:      2020-12-08 17:35:54
+num_postponed:          0
+postpone_reason:        
+last_postpone_time:     1970-01-01 03:00:00
+```
+
+**See Also**
+
+-   [Managing ReplicatedMergeTree Tables](../../sql-reference/statements/system.md/#query-language-system-replicated)
+
+[Original article](https://clickhouse.tech/docs/en/operations/system_tables/replication_queue) <!--hide-->
diff --git a/docs/en/operations/tips.md b/docs/en/operations/tips.md
index 024eba1c899..e62dea0b04e 100644
--- a/docs/en/operations/tips.md
+++ b/docs/en/operations/tips.md
@@ -91,6 +91,23 @@ The Linux kernel prior to 3.2 had a multitude of problems with IPv6 implementati
 
 Use at least a 10 GB network, if possible. 1 Gb will also work, but it will be much worse for patching replicas with tens of terabytes of data, or for processing distributed queries with a large amount of intermediate data.
 
+## Hypervisor configuration
+
+If you are using OpenStack, set
+```
+cpu_mode=host-passthrough
+```
+in nova.conf.
+
+If you are using libvirt, set
+```
+<cpu mode='host-passthrough'/>
+```
+in XML configuration.
+
+This is important for ClickHouse to be able to get correct information with `cpuid` instruction.
+Otherwise you may get `Illegal instruction` crashes when hypervisor is run on old CPU models.
+
 ## ZooKeeper {#zookeeper}
 
 You are probably already using ZooKeeper for other purposes. You can use the same installation of ZooKeeper, if it isn’t already overloaded.
diff --git a/docs/en/sql-reference/aggregate-functions/combinators.md b/docs/en/sql-reference/aggregate-functions/combinators.md
index 554969b80a3..e4f7ba0db81 100644
--- a/docs/en/sql-reference/aggregate-functions/combinators.md
+++ b/docs/en/sql-reference/aggregate-functions/combinators.md
@@ -25,6 +25,10 @@ Example 2: `uniqArray(arr)` – Counts the number of unique elements in all ‘a
 
 -If and -Array can be combined. However, ‘Array’ must come first, then ‘If’. Examples: `uniqArrayIf(arr, cond)`, `quantilesTimingArrayIf(level1, level2)(arr, cond)`. Due to this order, the ‘cond’ argument won’t be an array.
 
+## -SimpleState {#agg-functions-combinator-simplestate}
+
+If you apply this combinator, the aggregate function returns the same value but with a different type. This is an `SimpleAggregateFunction(...)` that can be stored in a table to work with [AggregatingMergeTree](../../engines/table-engines/mergetree-family/aggregatingmergetree.md) table engines.
+
 ## -State {#agg-functions-combinator-state}
 
 If you apply this combinator, the aggregate function doesn’t return the resulting value (such as the number of unique values for the [uniq](../../sql-reference/aggregate-functions/reference/uniq.md#agg_function-uniq) function), but an intermediate state of the aggregation (for `uniq`, this is the hash table for calculating the number of unique values). This is an `AggregateFunction(...)` that can be used for further processing or stored in a table to finish aggregating later.
diff --git a/docs/en/sql-reference/data-types/lowcardinality.md b/docs/en/sql-reference/data-types/lowcardinality.md
index 36c86bf443c..e0a483973e6 100644
--- a/docs/en/sql-reference/data-types/lowcardinality.md
+++ b/docs/en/sql-reference/data-types/lowcardinality.md
@@ -57,3 +57,5 @@ Functions:
 -   [A Magical Mystery Tour of the LowCardinality Data Type](https://www.altinity.com/blog/2019/3/27/low-cardinality).
 -   [Reducing Clickhouse Storage Cost with the Low Cardinality Type – Lessons from an Instana Engineer](https://www.instana.com/blog/reducing-clickhouse-storage-cost-with-the-low-cardinality-type-lessons-from-an-instana-engineer/).
 -   [String Optimization (video presentation in Russian)](https://youtu.be/rqf-ILRgBdY?list=PL0Z2YDlm0b3iwXCpEFiOOYmwXzVmjJfEt). [Slides in English](https://github.com/yandex/clickhouse-presentations/raw/master/meetup19/string_optimization.pdf).
+
+[Original article](https://clickhouse.tech/docs/en/sql-reference/data-types/lowcardinality/) <!--hide-->
diff --git a/docs/en/sql-reference/functions/array-functions.md b/docs/en/sql-reference/functions/array-functions.md
index abdf778c089..c4b7156ad95 100644
--- a/docs/en/sql-reference/functions/array-functions.md
+++ b/docs/en/sql-reference/functions/array-functions.md
@@ -1288,12 +1288,30 @@ Returns the index of the first element in the `arr1` array for which `func` retu
 
 Note that the `arrayFirstIndex` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You must pass a lambda function to it as the first argument, and it can’t be omitted.
 
+## arrayMin(\[func,\] arr1, …) {#array-min}
+
+Returns the sum of the `func` values. If the function is omitted, it just returns the min of the array elements.
+
+Note that the `arrayMin` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
+
+## arrayMax(\[func,\] arr1, …) {#array-max}
+
+Returns the sum of the `func` values. If the function is omitted, it just returns the min of the array elements.
+
+Note that the `arrayMax` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
+
 ## arraySum(\[func,\] arr1, …) {#array-sum}
 
 Returns the sum of the `func` values. If the function is omitted, it just returns the sum of the array elements.
 
 Note that the `arraySum` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
 
+## arrayAvg(\[func,\] arr1, …) {#array-avg}
+
+Returns the sum of the `func` values. If the function is omitted, it just returns the average of the array elements.
+
+Note that the `arrayAvg` is a [higher-order function](../../sql-reference/functions/index.md#higher-order-functions). You can pass a lambda function to it as the first argument.
+
 ## arrayCumSum(\[func,\] arr1, …) {#arraycumsumfunc-arr1}
 
 Returns an array of partial sums of elements in the source array (a running sum). If the `func` function is specified, then the values of the array elements are converted by this function before summing.
diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md
index f8458b27a22..6237cd6a976 100644
--- a/docs/en/sql-reference/functions/type-conversion-functions.md
+++ b/docs/en/sql-reference/functions/type-conversion-functions.md
@@ -430,6 +430,63 @@ SELECT toTypeName(CAST(x, 'Nullable(UInt16)')) FROM t_null
 
 -   [cast_keep_nullable](../../operations/settings/settings.md#cast_keep_nullable) setting
 
+## accurateCast(x, T) {#type_conversion_function-accurate-cast}
+
+Converts ‘x’ to the ‘t’ data type. The differente from cast(x, T) is that accurateCast
+does not allow overflow of numeric types during cast if type value x does not fit
+bounds of type T.
+
+Example
+``` sql
+SELECT cast(-1, 'UInt8') as uint8; 
+```
+
+
+``` text
+┌─uint8─┐
+│   255 │
+└───────┘
+```
+
+```sql
+SELECT accurateCast(-1, 'UInt8') as uint8;
+```
+
+``` text
+Code: 70. DB::Exception: Received from localhost:9000. DB::Exception: Value in column Int8 cannot be safely converted into type UInt8: While processing accurateCast(-1, 'UInt8') AS uint8.
+
+```
+
+## accurateCastOrNull(x, T) {#type_conversion_function-accurate-cast_or_null}
+
+Converts ‘x’ to the ‘t’ data type. Always returns nullable type and returns NULL 
+if the casted value is not representable in the target type.
+
+Example:
+
+``` sql
+SELECT
+    accurateCastOrNull(-1, 'UInt8') as uint8,
+    accurateCastOrNull(128, 'Int8') as int8,
+    accurateCastOrNull('Test', 'FixedString(2)') as fixed_string
+```
+
+``` text
+┌─uint8─┬─int8─┬─fixed_string─┐
+│  ᴺᵁᴸᴸ │ ᴺᵁᴸᴸ │ ᴺᵁᴸᴸ         │
+└───────┴──────┴──────────────┘┘
+```
+
+``` sql
+SELECT toTypeName(accurateCastOrNull(5, 'UInt8'))
+```
+
+``` text
+┌─toTypeName(accurateCastOrNull(5, 'UInt8'))─┐
+│ Nullable(UInt8)                            │
+└────────────────────────────────────────────┘
+```
+
 ## toInterval(Year\|Quarter\|Month\|Week\|Day\|Hour\|Minute\|Second) {#function-tointerval}
 
 Converts a Number type argument to an [Interval](../../sql-reference/data-types/special-data-types/interval.md) data type.
diff --git a/docs/en/sql-reference/operators/in.md b/docs/en/sql-reference/operators/in.md
index eca95dbc652..bfa8b3d1003 100644
--- a/docs/en/sql-reference/operators/in.md
+++ b/docs/en/sql-reference/operators/in.md
@@ -197,3 +197,25 @@ This is more optimal than using the normal IN. However, keep the following point
 5.  If you need to use GLOBAL IN often, plan the location of the ClickHouse cluster so that a single group of replicas resides in no more than one data center with a fast network between them, so that a query can be processed entirely within a single data center.
 
 It also makes sense to specify a local table in the `GLOBAL IN` clause, in case this local table is only available on the requestor server and you want to use data from it on remote servers.
+
+### Distributed Subqueries and max_parallel_replicas {#max_parallel_replica-subqueries}
+
+When max_parallel_replicas is greater than 1, distributed queries are further transformed. For example, the following:
+
+```sql
+SEELECT CounterID, count() FROM distributed_table_1 WHERE UserID IN (SELECT UserID FROM local_table_2 WHERE CounterID < 100)
+SETTINGS max_parallel_replicas=3
+```
+
+is transformed on each server into
+
+```sql
+SELECT CounterID, count() FROM local_table_1 WHERE UserID IN (SELECT UserID FROM local_table_2 WHERE CounterID < 100)
+SETTINGS parallel_replicas_count=3, parallel_replicas_offset=M
+```
+
+where M is between 1 and 3 depending on which replica the local query is executing on. These settings affect every MergeTree-family table in the query and have the same effect as applying `SAMPLE 1/3 OFFSET (M-1)/3` on each table.
+
+Therefore adding the max_parallel_replicas setting will only produce correct results if both tables have the same replication scheme and are sampled by UserID or a subkey of it. In particular, if local_table_2 does not have a sampling key, incorrect results will be produced. The same rule applies to JOIN.
+
+One workaround if local_table_2 doesn't meet the requirements, is to use `GLOBAL IN` or `GLOBAL JOIN`.
diff --git a/docs/en/sql-reference/statements/select/order-by.md b/docs/en/sql-reference/statements/select/order-by.md
index 57e071d6734..fb1df445db1 100644
--- a/docs/en/sql-reference/statements/select/order-by.md
+++ b/docs/en/sql-reference/statements/select/order-by.md
@@ -56,10 +56,188 @@ When floating point numbers are sorted, NaNs are separate from the other values.
 
 ## Collation Support {#collation-support}
 
-For sorting by String values, you can specify collation (comparison). Example: `ORDER BY SearchPhrase COLLATE 'tr'` - for sorting by keyword in ascending order, using the Turkish alphabet, case insensitive, assuming that strings are UTF-8 encoded. `COLLATE` can be specified or not for each expression in ORDER BY independently. If `ASC` or `DESC` is specified, `COLLATE` is specified after it. When using `COLLATE`, sorting is always case-insensitive.
+For sorting by [String](../../../sql-reference/data-types/string.md) values, you can specify collation (comparison). Example: `ORDER BY SearchPhrase COLLATE 'tr'` - for sorting by keyword in ascending order, using the Turkish alphabet, case insensitive, assuming that strings are UTF-8 encoded. `COLLATE` can be specified or not for each expression in ORDER BY independently. If `ASC` or `DESC` is specified, `COLLATE` is specified after it. When using `COLLATE`, sorting is always case-insensitive.
+
+Collate is supported in [LowCardinality](../../../sql-reference/data-types/lowcardinality.md), [Nullable](../../../sql-reference/data-types/nullable.md), [Array](../../../sql-reference/data-types/array.md) and [Tuple](../../../sql-reference/data-types/tuple.md).
 
 We only recommend using `COLLATE` for final sorting of a small number of rows, since sorting with `COLLATE` is less efficient than normal sorting by bytes.
 
+## Collation Examples {#collation-examples}
+
+Example only with [String](../../../sql-reference/data-types/string.md) values:
+
+Input table:
+
+``` text
+┌─x─┬─s────┐
+│ 1 │ bca  │
+│ 2 │ ABC  │
+│ 3 │ 123a │
+│ 4 │ abc  │
+│ 5 │ BCA  │
+└───┴──────┘
+```
+
+Query:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Result:
+
+``` text
+┌─x─┬─s────┐
+│ 3 │ 123a │
+│ 4 │ abc  │
+│ 2 │ ABC  │
+│ 1 │ bca  │
+│ 5 │ BCA  │
+└───┴──────┘
+```
+
+Example with [Nullable](../../../sql-reference/data-types/nullable.md):
+
+Input table:
+
+``` text
+┌─x─┬─s────┐
+│ 1 │ bca  │
+│ 2 │ ᴺᵁᴸᴸ │
+│ 3 │ ABC  │
+│ 4 │ 123a │
+│ 5 │ abc  │
+│ 6 │ ᴺᵁᴸᴸ │
+│ 7 │ BCA  │
+└───┴──────┘
+```
+
+Query:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Result:
+
+``` text
+┌─x─┬─s────┐
+│ 4 │ 123a │
+│ 5 │ abc  │
+│ 3 │ ABC  │
+│ 1 │ bca  │
+│ 7 │ BCA  │
+│ 6 │ ᴺᵁᴸᴸ │
+│ 2 │ ᴺᵁᴸᴸ │
+└───┴──────┘
+```
+
+Example with [Array](../../../sql-reference/data-types/array.md):
+
+Input table:
+
+``` text
+┌─x─┬─s─────────────┐
+│ 1 │ ['Z']         │
+│ 2 │ ['z']         │
+│ 3 │ ['a']         │
+│ 4 │ ['A']         │
+│ 5 │ ['z','a']     │
+│ 6 │ ['z','a','a'] │
+│ 7 │ ['']          │
+└───┴───────────────┘
+```
+
+Query:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Result:
+
+``` text
+┌─x─┬─s─────────────┐
+│ 7 │ ['']          │
+│ 3 │ ['a']         │
+│ 4 │ ['A']         │
+│ 2 │ ['z']         │
+│ 5 │ ['z','a']     │
+│ 6 │ ['z','a','a'] │
+│ 1 │ ['Z']         │
+└───┴───────────────┘
+```
+
+Example with [LowCardinality](../../../sql-reference/data-types/lowcardinality.md) string:
+
+Input table:
+
+```text
+┌─x─┬─s───┐
+│ 1 │ Z   │
+│ 2 │ z   │
+│ 3 │ a   │
+│ 4 │ A   │
+│ 5 │ za  │
+│ 6 │ zaa │
+│ 7 │     │
+└───┴─────┘
+```
+
+Query:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Result:
+
+```text
+┌─x─┬─s───┐
+│ 7 │     │
+│ 3 │ a   │
+│ 4 │ A   │
+│ 2 │ z   │
+│ 1 │ Z   │
+│ 5 │ za  │
+│ 6 │ zaa │
+└───┴─────┘
+```
+
+Example with [Tuple](../../../sql-reference/data-types/tuple.md):
+
+```text
+┌─x─┬─s───────┐
+│ 1 │ (1,'Z') │
+│ 2 │ (1,'z') │
+│ 3 │ (1,'a') │
+│ 4 │ (2,'z') │
+│ 5 │ (1,'A') │
+│ 6 │ (2,'Z') │
+│ 7 │ (2,'A') │
+└───┴─────────┘
+```
+
+Query:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Result:
+
+```text
+┌─x─┬─s───────┐
+│ 3 │ (1,'a') │
+│ 5 │ (1,'A') │
+│ 2 │ (1,'z') │
+│ 1 │ (1,'Z') │
+│ 7 │ (2,'A') │
+│ 4 │ (2,'z') │
+│ 6 │ (2,'Z') │
+└───┴─────────┘
+```
+
 ## Implementation Details {#implementation-details}
 
 Less RAM is used if a small enough [LIMIT](../../../sql-reference/statements/select/limit.md) is specified in addition to `ORDER BY`. Otherwise, the amount of memory spent is proportional to the volume of data for sorting. For distributed query processing, if [GROUP BY](../../../sql-reference/statements/select/group-by.md) is omitted, sorting is partially done on remote servers, and the results are merged on the requestor server. This means that for distributed sorting, the volume of data to sort can be greater than the amount of memory on a single server.
diff --git a/docs/en/sql-reference/statements/system.md b/docs/en/sql-reference/statements/system.md
index ddba1443d04..bb279703cc2 100644
--- a/docs/en/sql-reference/statements/system.md
+++ b/docs/en/sql-reference/statements/system.md
@@ -204,7 +204,7 @@ SYSTEM STOP MOVES [[db.]merge_tree_family_table_name]
 
 ## Managing ReplicatedMergeTree Tables {#query-language-system-replicated}
 
-ClickHouse can manage background replication related processes in [ReplicatedMergeTree](../../engines/table-engines/mergetree-family/replication/#table_engines-replication) tables.
+ClickHouse can manage background replication related processes in [ReplicatedMergeTree](../../engines/table-engines/mergetree-family/replication.md#table_engines-replication) tables.
 
 ### STOP FETCHES {#query_language-system-stop-fetches}
 
diff --git a/docs/es/development/developer-instruction.md b/docs/es/development/developer-instruction.md
index 390ac55602d..0ce5d0b457a 100644
--- a/docs/es/development/developer-instruction.md
+++ b/docs/es/development/developer-instruction.md
@@ -257,8 +257,8 @@ El desarrollo de ClickHouse a menudo requiere cargar conjuntos de datos realista
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/es/getting-started/example-datasets/metrica.md b/docs/es/getting-started/example-datasets/metrica.md
index b99346eda29..0b3bc8b6833 100644
--- a/docs/es/getting-started/example-datasets/metrica.md
+++ b/docs/es/getting-started/example-datasets/metrica.md
@@ -9,14 +9,14 @@ toc_title: El Yandex.Metrica Datos
 
 El conjunto de datos consta de dos tablas que contienen datos anónimos sobre los hits (`hits_v1`) y visitas (`visits_v1`) el Yandex.Métrica. Puedes leer más sobre Yandex.Metrica en [Historial de ClickHouse](../../introduction/history.md) apartado.
 
-El conjunto de datos consta de dos tablas, cualquiera de ellas se puede descargar como `tsv.xz` o como particiones preparadas. Además, una versión extendida de la `hits` La tabla que contiene 100 millones de filas está disponible como TSV en https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz y como particiones preparadas en https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
+El conjunto de datos consta de dos tablas, cualquiera de ellas se puede descargar como `tsv.xz` o como particiones preparadas. Además, una versión extendida de la `hits` La tabla que contiene 100 millones de filas está disponible como TSV en https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz y como particiones preparadas en https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
 
 ## Obtención de tablas a partir de particiones preparadas {#obtaining-tables-from-prepared-partitions}
 
 Descargar e importar tabla de hits:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -26,7 +26,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Descargar e importar visitas:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -38,7 +38,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 Descargar e importar hits desde un archivo TSV comprimido:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -52,7 +52,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Descargue e importe visitas desde un archivo tsv comprimido:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/es/getting-started/example-datasets/nyc-taxi.md b/docs/es/getting-started/example-datasets/nyc-taxi.md
index 4a2bae83a0a..c6441311c96 100644
--- a/docs/es/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/es/getting-started/example-datasets/nyc-taxi.md
@@ -285,7 +285,7 @@ Entre otras cosas, puede ejecutar la consulta OPTIMIZE en MergeTree. Pero no es
 ## Descarga de Prepared Partitions {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/es/getting-started/example-datasets/ontime.md b/docs/es/getting-started/example-datasets/ontime.md
index b0662ef8b53..f89d74048bd 100644
--- a/docs/es/getting-started/example-datasets/ontime.md
+++ b/docs/es/getting-started/example-datasets/ontime.md
@@ -156,7 +156,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## Descarga de Prepared Partitions {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/es/getting-started/tutorial.md b/docs/es/getting-started/tutorial.md
index ccc07e50468..52699190b4d 100644
--- a/docs/es/getting-started/tutorial.md
+++ b/docs/es/getting-started/tutorial.md
@@ -87,8 +87,8 @@ Ahora es el momento de llenar nuestro servidor ClickHouse con algunos datos de m
 ### Descargar y extraer datos de tabla {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 Los archivos extraídos tienen un tamaño de aproximadamente 10 GB.
diff --git a/docs/es/interfaces/cli.md b/docs/es/interfaces/cli.md
index 1e4852be53f..395f9831a4e 100644
--- a/docs/es/interfaces/cli.md
+++ b/docs/es/interfaces/cli.md
@@ -59,7 +59,7 @@ La línea de comandos se basa en ‘replxx’ (similar a ‘readline’). En otr
 
 De forma predeterminada, el formato utilizado es PrettyCompact. Puede cambiar el formato en la cláusula FORMAT de la consulta o especificando `\G` al final de la consulta, utilizando el `--format` o `--vertical` en la línea de comandos, o utilizando el archivo de configuración del cliente.
 
-Para salir del cliente, presione Ctrl+D (o Ctrl+C) o introduzca una de las siguientes opciones en lugar de una consulta: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+Para salir del cliente, presione Ctrl+D o introduzca una de las siguientes opciones en lugar de una consulta: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 Al procesar una consulta, el cliente muestra:
 
diff --git a/docs/es/operations/performance-test.md b/docs/es/operations/performance-test.md
index d0beb7bd2b4..97444f339cd 100644
--- a/docs/es/operations/performance-test.md
+++ b/docs/es/operations/performance-test.md
@@ -48,7 +48,7 @@ Con esta instrucción, puede ejecutar una prueba de rendimiento básica de Click
 
 <!-- -->
 
-      wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+      wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
       tar xvf hits_100m_obfuscated_v1.tar.xz -C .
       mv hits_100m_obfuscated_v1/* .
 
diff --git a/docs/fa/development/developer-instruction.md b/docs/fa/development/developer-instruction.md
index ee78050da07..9284d3a511a 100644
--- a/docs/fa/development/developer-instruction.md
+++ b/docs/fa/development/developer-instruction.md
@@ -259,8 +259,8 @@ KDevelop و QTCreator دیگر از جایگزین های بسیار خوبی ا
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/fa/getting-started/example-datasets/metrica.md b/docs/fa/getting-started/example-datasets/metrica.md
index b1bdf3fd131..ac6743309ef 100644
--- a/docs/fa/getting-started/example-datasets/metrica.md
+++ b/docs/fa/getting-started/example-datasets/metrica.md
@@ -10,14 +10,14 @@ toc_title: "\u06CC\u0627\u0646\u062F\u06A9\u0633\u0627\u0637\u0644\u0627\u0639\u
 
 مجموعه داده شامل دو جدول حاوی داده های ناشناس در مورد بازدید (`hits_v1`) و بازدیدکننده داشته است (`visits_v1`) یاندکس . متریکا شما می توانید اطلاعات بیشتر در مورد یاندکس به عنوان خوانده شده.متریکا در [تاریخچه کلیک](../../introduction/history.md) بخش.
 
-مجموعه داده ها شامل دو جدول است که هر کدام می توانند به عنوان یک فشرده دانلود شوند `tsv.xz` فایل و یا به عنوان پارتیشن تهیه شده است. علاوه بر این, یک نسخه طولانی از `hits` جدول حاوی 100 میلیون ردیف به عنوان تسو در دسترس است https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz و به عنوان پارتیشن تهیه شده در https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
+مجموعه داده ها شامل دو جدول است که هر کدام می توانند به عنوان یک فشرده دانلود شوند `tsv.xz` فایل و یا به عنوان پارتیشن تهیه شده است. علاوه بر این, یک نسخه طولانی از `hits` جدول حاوی 100 میلیون ردیف به عنوان تسو در دسترس است https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz و به عنوان پارتیشن تهیه شده در https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
 
 ## اخذ جداول از پارتیشن های تهیه شده {#obtaining-tables-from-prepared-partitions}
 
 دانلود و وارد کردن جدول بازدید:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -27,7 +27,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 دانلود و وارد کردن بازدیدکننده داشته است:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -39,7 +39,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 دانلود و وارد کردن بازدید از فایل تسو فشرده:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -53,7 +53,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 دانلود و واردات بازدیدکننده داشته است از فشرده فایل:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/fa/getting-started/example-datasets/nyc-taxi.md b/docs/fa/getting-started/example-datasets/nyc-taxi.md
index 56255e1e09b..32fbd471b32 100644
--- a/docs/fa/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/fa/getting-started/example-datasets/nyc-taxi.md
@@ -286,7 +286,7 @@ SELECT formatReadableSize(sum(bytes)) FROM system.parts WHERE table = 'trips_mer
 ## دانلود پارتیشن های تهیه شده {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/fa/getting-started/example-datasets/ontime.md b/docs/fa/getting-started/example-datasets/ontime.md
index e8cc60113e7..7282c3c29bb 100644
--- a/docs/fa/getting-started/example-datasets/ontime.md
+++ b/docs/fa/getting-started/example-datasets/ontime.md
@@ -156,7 +156,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## دانلود پارتیشن های تهیه شده {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/fa/getting-started/tutorial.md b/docs/fa/getting-started/tutorial.md
index 933dae2b55f..595d17667ae 100644
--- a/docs/fa/getting-started/tutorial.md
+++ b/docs/fa/getting-started/tutorial.md
@@ -87,8 +87,8 @@ clickhouse-client --query='INSERT INTO table FORMAT TabSeparated' < data.tsv
 ### دانلود و استخراج داده های جدول {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 فایل های استخراج شده حدود 10 گیگابایت است.
diff --git a/docs/fa/operations/performance-test.md b/docs/fa/operations/performance-test.md
index 1d061769d64..4bd5cc2c15f 100644
--- a/docs/fa/operations/performance-test.md
+++ b/docs/fa/operations/performance-test.md
@@ -48,7 +48,7 @@ toc_title: "\u0633\u062E\u062A \u0627\u0641\u0632\u0627\u0631 \u062A\u0633\u062A
 
 <!-- -->
 
-      wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+      wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
       tar xvf hits_100m_obfuscated_v1.tar.xz -C .
       mv hits_100m_obfuscated_v1/* .
 
diff --git a/docs/fr/development/developer-instruction.md b/docs/fr/development/developer-instruction.md
index e78ed7ba6d9..610216925c3 100644
--- a/docs/fr/development/developer-instruction.md
+++ b/docs/fr/development/developer-instruction.md
@@ -257,8 +257,8 @@ Le développement de ClickHouse nécessite souvent le chargement d'ensembles de
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/fr/getting-started/example-datasets/metrica.md b/docs/fr/getting-started/example-datasets/metrica.md
index f9d6c7b437b..453a30d436f 100644
--- a/docs/fr/getting-started/example-datasets/metrica.md
+++ b/docs/fr/getting-started/example-datasets/metrica.md
@@ -9,14 +9,14 @@ toc_title: "Yandex.Metrica De Donn\xE9es"
 
 Dataset se compose de deux tables contenant des données anonymisées sur les hits (`hits_v1`) et les visites (`visits_v1`) de Yandex.Metrica. Vous pouvez en savoir plus sur Yandex.Metrica dans [Histoire de ClickHouse](../../introduction/history.md) section.
 
-L'ensemble de données se compose de deux tables, l'une d'elles peut être téléchargée sous forme compressée `tsv.xz` fichier ou comme partitions préparées. En outre, une version étendue de l' `hits` table contenant 100 millions de lignes est disponible comme TSV à https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz et comme partitions préparées à https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
+L'ensemble de données se compose de deux tables, l'une d'elles peut être téléchargée sous forme compressée `tsv.xz` fichier ou comme partitions préparées. En outre, une version étendue de l' `hits` table contenant 100 millions de lignes est disponible comme TSV à https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz et comme partitions préparées à https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
 
 ## Obtention de Tables à partir de Partitions préparées {#obtaining-tables-from-prepared-partitions}
 
 Télécharger et importer la table hits:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -26,7 +26,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Télécharger et importer des visites:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -38,7 +38,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 Télécharger et importer des hits à partir du fichier TSV compressé:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -52,7 +52,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Télécharger et importer des visites à partir du fichier TSV compressé:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/fr/getting-started/example-datasets/nyc-taxi.md b/docs/fr/getting-started/example-datasets/nyc-taxi.md
index e351b9ec543..46aa944e718 100644
--- a/docs/fr/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/fr/getting-started/example-datasets/nyc-taxi.md
@@ -285,7 +285,7 @@ Entre autres choses, vous pouvez exécuter la requête OPTIMIZE sur MergeTree. M
 ## Téléchargement des Partitions préparées {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/fr/getting-started/example-datasets/ontime.md b/docs/fr/getting-started/example-datasets/ontime.md
index 4040e2a1900..8d036901b99 100644
--- a/docs/fr/getting-started/example-datasets/ontime.md
+++ b/docs/fr/getting-started/example-datasets/ontime.md
@@ -156,7 +156,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## Téléchargement des Partitions préparées {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/fr/getting-started/tutorial.md b/docs/fr/getting-started/tutorial.md
index e8a0e8d81cb..b600456b484 100644
--- a/docs/fr/getting-started/tutorial.md
+++ b/docs/fr/getting-started/tutorial.md
@@ -87,8 +87,8 @@ Maintenant, il est temps de remplir notre serveur ClickHouse avec quelques exemp
 ### Télécharger et extraire les données de la Table {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 Les fichiers extraits ont une taille d'environ 10 Go.
diff --git a/docs/fr/operations/performance-test.md b/docs/fr/operations/performance-test.md
index 6093772aefe..7940b957642 100644
--- a/docs/fr/operations/performance-test.md
+++ b/docs/fr/operations/performance-test.md
@@ -48,7 +48,7 @@ Avec cette instruction, vous pouvez exécuter le test de performance clickhouse
 
 <!-- -->
 
-      wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+      wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
       tar xvf hits_100m_obfuscated_v1.tar.xz -C .
       mv hits_100m_obfuscated_v1/* .
 
diff --git a/docs/ja/development/developer-instruction.md b/docs/ja/development/developer-instruction.md
index 988becf98c3..ccc3a177d1f 100644
--- a/docs/ja/development/developer-instruction.md
+++ b/docs/ja/development/developer-instruction.md
@@ -257,8 +257,8 @@ KDevelopとQTCreatorは、ClickHouseを開発するためのIDEの他の優れ
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/ja/getting-started/example-datasets/metrica.md b/docs/ja/getting-started/example-datasets/metrica.md
index 36c6411ff2f..5caf32928c3 100644
--- a/docs/ja/getting-started/example-datasets/metrica.md
+++ b/docs/ja/getting-started/example-datasets/metrica.md
@@ -9,14 +9,14 @@ toc_title: "Yandex.Metrica データ"
 Yandex.Metricaについての詳細は [ClickHouse history](../../introduction/history.md) のセクションを参照してください。
 
 データセットは2つのテーブルから構成されており、どちらも圧縮された `tsv.xz` ファイルまたは準備されたパーティションとしてダウンロードすることができます。
-さらに、1億行を含む`hits`テーブルの拡張版が TSVとして https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz に、準備されたパーティションとして https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz にあります。
+さらに、1億行を含む`hits`テーブルの拡張版が TSVとして https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz に、準備されたパーティションとして https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz にあります。
 
 ## パーティション済みテーブルの取得 {#obtaining-tables-from-prepared-partitions}
 
 hits テーブルのダウンロードとインポート:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # ClickHouse のデータディレクトリへのパス
 # 展開されたデータのパーミッションをチェックし、必要に応じて修正します。
 sudo service clickhouse-server restart
@@ -26,7 +26,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 visits のダウンロードとインポート:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # ClickHouse のデータディレクトリへのパス
 # 展開されたデータのパーミッションをチェックし、必要に応じて修正します。
 sudo service clickhouse-server restart
@@ -38,7 +38,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
  圧縮TSVファイルのダウンロードと hits テーブルのインポート:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -52,7 +52,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 圧縮TSVファイルのダウンロードと visits テーブルのインポート:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/ja/getting-started/example-datasets/nyc-taxi.md b/docs/ja/getting-started/example-datasets/nyc-taxi.md
index 9d15ac14a36..a717b64b2b2 100644
--- a/docs/ja/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/ja/getting-started/example-datasets/nyc-taxi.md
@@ -286,7 +286,7 @@ SELECT formatReadableSize(sum(bytes)) FROM system.parts WHERE table = 'trips_mer
 ## パーティションされたデータのダウンロード {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/ja/getting-started/example-datasets/ontime.md b/docs/ja/getting-started/example-datasets/ontime.md
index 58d1999bf2b..bd049e8caad 100644
--- a/docs/ja/getting-started/example-datasets/ontime.md
+++ b/docs/ja/getting-started/example-datasets/ontime.md
@@ -154,7 +154,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## パーティション済みデータのダウンロード {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/ja/getting-started/tutorial.md b/docs/ja/getting-started/tutorial.md
index 84f834a9486..d63954684f1 100644
--- a/docs/ja/getting-started/tutorial.md
+++ b/docs/ja/getting-started/tutorial.md
@@ -92,8 +92,8 @@ ClickHouseサーバーにいくつかのサンプルデータを入れてみま
 ### テーブルデータのダウンロードと展開 {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 展開されたファイルのサイズは約10GBです。
diff --git a/docs/ja/interfaces/cli.md b/docs/ja/interfaces/cli.md
index 0c10ee902f5..7037fcf5512 100644
--- a/docs/ja/interfaces/cli.md
+++ b/docs/ja/interfaces/cli.md
@@ -59,7 +59,7 @@ $ cat file.csv | clickhouse-client --database=test --query="INSERT INTO test FOR
 
 既定では、使用される形式はPrettyCompactです。 クエリのFORMAT句で書式を変更するか、次のように指定することができます `\G` クエリの最後に、 `--format` または `--vertical` コマンドラインでの引数、またはクライアント構成ファイルの使用。
 
-クライアントを終了するには、Ctrl+D(またはCtrl+C)を押すか、クエリの代わりに次のいずれかを入力します: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+クライアントを終了するには、Ctrl+Dを押すか、クエリの代わりに次のいずれかを入力します: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 が処理クエリー、クライアントを示し:
 
diff --git a/docs/ja/operations/performance-test.md b/docs/ja/operations/performance-test.md
index ff3a0192b49..a60a8872a16 100644
--- a/docs/ja/operations/performance-test.md
+++ b/docs/ja/operations/performance-test.md
@@ -48,7 +48,7 @@ toc_title: "\u30CF\u30FC\u30C9\u30A6\u30A7\u30A2\u8A66\u9A13"
 
 <!-- -->
 
-      wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+      wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
       tar xvf hits_100m_obfuscated_v1.tar.xz -C .
       mv hits_100m_obfuscated_v1/* .
 
diff --git a/docs/ja/sql-reference/data-types/lowcardinality.md b/docs/ja/sql-reference/data-types/lowcardinality.md
new file mode 100644
index 00000000000..dd3a9aa1c0d
--- /dev/null
+++ b/docs/ja/sql-reference/data-types/lowcardinality.md
@@ -0,0 +1 @@
+../../../../en/sql-reference/data-types/lowcardinality.md
\ No newline at end of file
diff --git a/docs/ru/development/developer-instruction.md b/docs/ru/development/developer-instruction.md
index 4bdcf89004d..5d7c0caeafa 100644
--- a/docs/ru/development/developer-instruction.md
+++ b/docs/ru/development/developer-instruction.md
@@ -259,8 +259,8 @@ Mac OS X:
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/ru/faq/general/index.md b/docs/ru/faq/general/index.md
new file mode 100644
index 00000000000..cf0f8451450
--- /dev/null
+++ b/docs/ru/faq/general/index.md
@@ -0,0 +1,25 @@
+---
+title: General questions about ClickHouse
+toc_hidden_folder: true
+toc_priority: 1
+toc_title: Общие вопросы
+---
+
+# Общие вопросы о ClickHouse {#obshchie-voprosy}
+
+Вопросы:
+
+-   Что такое ClickHouse?
+-   Почему ClickHouse такой быстрый?
+-   Кто пользуется ClickHouse?
+-   Что обозначает название “ClickHouse”?
+-   Что значит “Не тормозит”?
+-   Что такое OLAP?
+-   Что такое колоночная база данных?
+-   [Почему бы не использовать системы типа MapReduce?](mapreduce.md)
+
+!!! info "Если вы не нашли то, что искали:"
+    Загляните в другие категории F.A.Q. или поищите в других разделах документации, ориентируйтесь по оглавлению слева.
+
+{## [Original article](https://clickhouse.tech/docs/ru/faq/general/) ##}
+
diff --git a/docs/ru/faq/general.md b/docs/ru/faq/general/mapreduce.md
similarity index 60%
rename from docs/ru/faq/general.md
rename to docs/ru/faq/general/mapreduce.md
index 470b8dabbf1..645391493c3 100644
--- a/docs/ru/faq/general.md
+++ b/docs/ru/faq/general/mapreduce.md
@@ -1,8 +1,12 @@
-# Общие вопросы {#obshchie-voprosy}
+---
+title: Why not use something like MapReduce?
+toc_hidden: true
+toc_priority: 110
+---
 
 ## Почему бы не использовать системы типа MapReduce? {#pochemu-by-ne-ispolzovat-sistemy-tipa-mapreduce}
 
-Системами типа MapReduce будем называть системы распределённых вычислений, в которых операция reduce сделана на основе распределённой сортировки. Наиболее распространённым opensource решением данного класса является [Apache Hadoop](http://hadoop.apache.org). Яндекс использует собственное решение — YT.
+Системами типа MapReduce будем называть системы распределённых вычислений, в которых операция reduce сделана на основе распределённой сортировки. Наиболее распространённым opensource решением данного класса является [Apache Hadoop](http://hadoop.apache.org). Яндекс использует собственное решение — YT. 
 
 Такие системы не подходят для онлайн запросов в силу слишком большой latency. То есть, не могут быть использованы в качестве бэкенда для веб-интерфейса.
 Такие системы не подходят для обновления данных в реальном времени.
@@ -10,47 +14,3 @@
 Распределённая сортировка является основной причиной тормозов при выполнении несложных map-reduce задач.
 
 Большинство реализаций MapReduce позволяют выполнять произвольный код на кластере. Но для OLAP задач лучше подходит декларативный язык запросов, который позволяет быстро проводить исследования. Для примера, для Hadoop существует Hive и Pig. Также смотрите Cloudera Impala, Shark (устаревший) для Spark, а также Spark SQL, Presto, Apache Drill. Впрочем, производительность при выполнении таких задач является сильно неоптимальной по сравнению со специализированными системами, а сравнительно высокая latency не позволяет использовать эти системы в качестве бэкенда для веб-интерфейса.
-
-## Что делать, если у меня проблема с кодировками при использовании Oracle через ODBC? {#oracle-odbc-encodings}
-
-Если вы используете Oracle через драйвер ODBC в качестве источника внешних словарей, необходимо задать правильное значение для переменной окружения `NLS_LANG` в `/etc/default/clickhouse`. Подробнее читайте в [Oracle NLS_LANG FAQ](https://www.oracle.com/technetwork/products/globalization/nls-lang-099431.html).
-
-**Пример**
-
-``` sql
-NLS_LANG=RUSSIAN_RUSSIA.UTF8
-```
-
-## Как экспортировать данные из ClickHouse в файл? {#how-to-export-to-file}
-
-### Секция INTO OUTFILE {#sektsiia-into-outfile}
-
-Добавьте секцию [INTO OUTFILE](../sql-reference/statements/select/into-outfile.md#into-outfile-clause) к своему запросу.
-
-Например:
-
-``` sql
-SELECT * FROM table INTO OUTFILE 'file'
-```
-
-По умолчанию, для выдачи данных ClickHouse использует формат [TabSeparated](../interfaces/formats.md#tabseparated). Чтобы выбрать [формат данных](../interfaces/formats.md), используйте [секцию FORMAT](../sql-reference/statements/select/format.md#format-clause).
-
-Например:
-
-``` sql
-SELECT * FROM table INTO OUTFILE 'file' FORMAT CSV
-```
-
-### Таблица с движком File {#tablitsa-s-dvizhkom-file}
-
-Смотрите [File](../engines/table-engines/special/file.md).
-
-### Перенаправление в командой строке {#perenapravlenie-v-komandoi-stroke}
-
-``` sql
-$ clickhouse-client --query "SELECT * from table" --format FormatName > result.txt
-```
-
-Смотрите [clickhouse-client](../interfaces/cli.md).
-
-[Оригинальная статья](https://clickhouse.tech/docs/en/faq/general/) <!--hide-->
diff --git a/docs/ru/faq/index.md b/docs/ru/faq/index.md
index d9eccf59c5c..611e604bc8e 100644
--- a/docs/ru/faq/index.md
+++ b/docs/ru/faq/index.md
@@ -4,3 +4,14 @@ toc_hidden: true
 toc_priority: 76
 ---
 
+# Содержание F.A.Q. {#soderzhanie}
+
+В этом разделе документации собрали вопросы о ClickHouse, которые задают чаще всего.
+
+Категории:
+
+-   **[Общие вопросы](../faq/general/index.md)** 
+-   **[Применение](../faq/use-cases/index.md)**
+-   **[Операции](../faq/operations/index.md)**  
+-   **[Интеграция](../faq/integration/index.md)**
+
diff --git a/docs/ru/faq/integration/file-export.md b/docs/ru/faq/integration/file-export.md
new file mode 100644
index 00000000000..d21ffbd3349
--- /dev/null
+++ b/docs/ru/faq/integration/file-export.md
@@ -0,0 +1,37 @@
+---
+title: How do I export data from ClickHouse to a file?
+toc_hidden: true
+toc_priority: 10
+---
+
+## Как экспортировать данные из ClickHouse в файл? {#how-to-export-to-file-rus}
+
+### Секция INTO OUTFILE {#sektsiia-into-outfile-rus}
+
+Добавьте секцию [INTO OUTFILE](../../sql-reference/statements/select/into-outfile.md#into-outfile-clause) к своему запросу.
+
+Например:
+
+``` sql
+SELECT * FROM table INTO OUTFILE 'file'
+```
+
+По умолчанию, для выдачи данных ClickHouse использует формат [TabSeparated](../../interfaces/formats.md#tabseparated). Чтобы выбрать [формат данных](../../interfaces/formats.md), используйте секцию [FORMAT](../../sql-reference/statements/select/format.md#format-clause).
+
+Например:
+
+``` sql
+SELECT * FROM table INTO OUTFILE 'file' FORMAT CSV
+```
+
+## Таблица с движком File {#using-a-file-engine-table}
+
+Смотрите [File](../../engines/table-engines/special/file.md).
+
+## Перенаправление в командой строке {#using-command-line-redirection}
+
+``` bash
+$ clickhouse-client --query "SELECT * from table" --format FormatName > result.txt
+```
+
+Смотрите [clickhouse-client](../../interfaces/cli.md).
diff --git a/docs/ru/faq/integration/index.md b/docs/ru/faq/integration/index.md
new file mode 100644
index 00000000000..59fd84a8139
--- /dev/null
+++ b/docs/ru/faq/integration/index.md
@@ -0,0 +1,19 @@
+---
+title: Questions about integrating ClickHouse and other systems
+toc_hidden_folder: true
+toc_priority: 4
+toc_title: Интеграция
+---
+
+# Вопросы об интеграции ClickHouse с другими системами {#question-about-integrating-clickhouse-and-other-systems-rus}
+
+Вопросы:
+
+-   [Как экспортировать данные из ClickHouse в файл?](file-export.md)
+-   Как импортировать JSON в ClickHouse?
+-   [Что делать, если у меня проблема с кодировками при использовании Oracle через ODBC?](oracle-odbc.md)
+
+!!! info "Если вы не нашли то, что искали"
+    Загляните в другие подразделы F.A.Q. или поищите в остальных разделах документации, ориентируйтесь по оглавлению слева.
+
+{## [Original article](https://clickhouse.tech/docs/ru/faq/integration/) ##}
diff --git a/docs/ru/faq/integration/oracle-odbc.md b/docs/ru/faq/integration/oracle-odbc.md
new file mode 100644
index 00000000000..f41f1ae7fc1
--- /dev/null
+++ b/docs/ru/faq/integration/oracle-odbc.md
@@ -0,0 +1,15 @@
+---
+title: What if I have a problem with encodings when using Oracle via ODBC?
+toc_hidden: true
+toc_priority: 20
+---
+
+## Что делать, если у меня проблема с кодировками при использовании Oracle через ODBC? {#oracle-odbc-encodings-rus}
+
+Если вы используете Oracle через драйвер ODBC в качестве источника внешних словарей, необходимо задать правильное значение для переменной окружения `NLS_LANG` в `/etc/default/clickhouse`. Подробнее читайте в [Oracle NLS_LANG FAQ](https://www.oracle.com/technetwork/products/globalization/nls-lang-099431.html).
+
+**Пример**
+
+``` sql
+NLS_LANG=RUSSIAN_RUSSIA.UTF8
+```
\ No newline at end of file
diff --git a/docs/ru/faq/operations/index.md b/docs/ru/faq/operations/index.md
new file mode 100644
index 00000000000..98a04213fb7
--- /dev/null
+++ b/docs/ru/faq/operations/index.md
@@ -0,0 +1,18 @@
+---
+title: Question about operating ClickHouse servers and clusters
+toc_hidden_folder: true
+toc_priority: 3
+toc_title: Операции
+---
+
+# Вопросы о производительности серверов и кластеров ClickHouse {#voprosy-ob-operating-clickhouse-servers-and-clusters}
+
+Вопросы:
+
+-   Which ClickHouse version to use in production?
+-   Is it possible to delete old records from a ClickHouse table?
+
+    !!! info "Don’t see what you were looking for?"
+        Check out [other F.A.Q. categories](../../faq/index.md) or browse around main documentation articles found in the left sidebar.
+
+    {## [Original article](https://clickhouse.tech/docs/en/faq/production/) ##}
diff --git a/docs/ru/faq/use-cases/index.md b/docs/ru/faq/use-cases/index.md
new file mode 100644
index 00000000000..6e7f2b36fe0
--- /dev/null
+++ b/docs/ru/faq/use-cases/index.md
@@ -0,0 +1,14 @@
+---
+title: Questions about ClickHouse use cases
+toc_hidden_folder: true
+toc_priority: 2
+toc_title: Применение
+---
+
+# Вопросы о применении ClickHouse {#voprosy-o-primenenii}
+
+Вопросы:
+
+-   Can I use ClickHouse as a time-series database?
+-   Can I use ClickHouse as a key-value storage?
+
diff --git a/docs/ru/getting-started/example-datasets/metrica.md b/docs/ru/getting-started/example-datasets/metrica.md
index e8a3163376c..3246eb5178c 100644
--- a/docs/ru/getting-started/example-datasets/metrica.md
+++ b/docs/ru/getting-started/example-datasets/metrica.md
@@ -5,14 +5,14 @@ toc_title: "\u0410\u043d\u043e\u043d\u0438\u043c\u0438\u0437\u0438\u0440\u043e\u
 
 # Анонимизированные данные Яндекс.Метрики {#anonimizirovannye-dannye-iandeks-metriki}
 
-Датасет состоит из двух таблиц, содержащих анонимизированные данные о хитах (`hits_v1`) и визитах (`visits_v1`) Яндекс.Метрики. Каждую из таблиц можно скачать в виде сжатого `.tsv.xz`-файла или в виде уже готовых партиций. Также можно скачать расширенную версию таблицы `hits`, содержащую 100 миллионов строк в виде [архива c файлами TSV](https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz) и в виде [готовых партиций](https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz).
+Датасет состоит из двух таблиц, содержащих анонимизированные данные о хитах (`hits_v1`) и визитах (`visits_v1`) Яндекс.Метрики. Каждую из таблиц можно скачать в виде сжатого `.tsv.xz`-файла или в виде уже готовых партиций. Также можно скачать расширенную версию таблицы `hits`, содержащую 100 миллионов строк в виде [архива c файлами TSV](https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz) и в виде [готовых партиций](https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz).
 
 ## Получение таблиц из партиций {#poluchenie-tablits-iz-partitsii}
 
 **Скачивание и импортирование партиций hits:**
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+$ curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 $ tar xvf hits_v1.tar -C /var/lib/clickhouse # путь к папке с данными ClickHouse
 $ # убедитесь, что установлены корректные права доступа на файлы
 $ sudo service clickhouse-server restart
@@ -22,7 +22,7 @@ $ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 **Скачивание и импортирование партиций visits:**
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+$ curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 $ tar xvf visits_v1.tar -C /var/lib/clickhouse # путь к папке с данными ClickHouse
 $ # убедитесь, что установлены корректные права доступа на файлы
 $ sudo service clickhouse-server restart
@@ -34,7 +34,7 @@ $ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 **Скачивание и импортирование hits из сжатого tsv-файла**
 
 ``` bash
-$ curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+$ curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 $ # теперь создадим таблицу
 $ clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 $ clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -48,7 +48,7 @@ $ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 **Скачивание и импортирование visits из сжатого tsv-файла**
 
 ``` bash
-$ curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+$ curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 $ # теперь создадим таблицу
 $ clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 $ clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/ru/getting-started/example-datasets/nyc-taxi.md b/docs/ru/getting-started/example-datasets/nyc-taxi.md
index 1f981324261..a4472751a99 100644
--- a/docs/ru/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/ru/getting-started/example-datasets/nyc-taxi.md
@@ -283,7 +283,7 @@ SELECT formatReadableSize(sum(bytes)) FROM system.parts WHERE table = 'trips_mer
 ## Скачивание готовых партиций {#skachivanie-gotovykh-partitsii}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # путь к папке с данными ClickHouse
 $ # убедитесь, что установлены корректные права доступа на файлы
 $ sudo service clickhouse-server restart
diff --git a/docs/ru/getting-started/example-datasets/ontime.md b/docs/ru/getting-started/example-datasets/ontime.md
index 4d3eea14da6..41a1c0d3142 100644
--- a/docs/ru/getting-started/example-datasets/ontime.md
+++ b/docs/ru/getting-started/example-datasets/ontime.md
@@ -152,7 +152,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## Скачивание готовых партиций {#skachivanie-gotovykh-partitsii}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # путь к папке с данными ClickHouse
 $ # убедитесь, что установлены корректные права доступа на файлы
 $ sudo service clickhouse-server restart
diff --git a/docs/ru/getting-started/tutorial.md b/docs/ru/getting-started/tutorial.md
index 0fc267e497c..bd987520dff 100644
--- a/docs/ru/getting-started/tutorial.md
+++ b/docs/ru/getting-started/tutorial.md
@@ -85,8 +85,8 @@ Now it’s time to fill our ClickHouse server with some sample data. In this tut
 ### Download and Extract Table Data {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 The extracted files are about 10GB in size.
diff --git a/docs/ru/interfaces/cli.md b/docs/ru/interfaces/cli.md
index 222af33f952..e37018acdfb 100644
--- a/docs/ru/interfaces/cli.md
+++ b/docs/ru/interfaces/cli.md
@@ -63,7 +63,7 @@ $ cat file.csv | clickhouse-client --database=test --query="INSERT INTO test FOR
 
 По умолчанию, в качестве формата, используется формат PrettyCompact (красивые таблички). Вы можете изменить формат с помощью секции FORMAT запроса, или с помощью указания `\G` на конце запроса, с помощью аргумента командной строки `--format` или `--vertical`, или с помощью конфигурационного файла клиента.
 
-Чтобы выйти из клиента, нажмите Ctrl+D (или Ctrl+C), или наберите вместо запроса одно из: «exit», «quit», «logout», «учше», «йгше», «дщпщге», «exit;», «quit;», «logout;», «учшеж», «йгшеж», «дщпщгеж», «q», «й», «q», «Q», «:q», «й», «Й», «Жй»
+Чтобы выйти из клиента, нажмите Ctrl+D, или наберите вместо запроса одно из: «exit», «quit», «logout», «учше», «йгше», «дщпщге», «exit;», «quit;», «logout;», «учшеж», «йгшеж», «дщпщгеж», «q», «й», «q», «Q», «:q», «й», «Й», «Жй»
 
 При выполнении запроса, клиент показывает:
 
diff --git a/docs/ru/operations/system-tables/distribution_queue.md b/docs/ru/operations/system-tables/distribution_queue.md
new file mode 100644
index 00000000000..18346b34e04
--- /dev/null
+++ b/docs/ru/operations/system-tables/distribution_queue.md
@@ -0,0 +1,46 @@
+# system.distribution_queue {#system_tables-distribution_queue}
+
+Содержит информацию о локальных файлах, которые находятся в очереди для отправки на шарды. Эти локальные файлы содержат новые куски, которые создаются путем вставки новых данных в Distributed таблицу в асинхронном режиме.
+
+Столбцы:
+
+-   `database` ([String](../../sql-reference/data-types/string.md)) — имя базы данных.
+
+-   `table` ([String](../../sql-reference/data-types/string.md)) — имя таблицы.
+
+-   `data_path` ([String](../../sql-reference/data-types/string.md)) — путь к папке с локальными файлами.
+
+-   `is_blocked` ([UInt8](../../sql-reference/data-types/int-uint.md)) — флаг, указывающий на блокировку отправки локальных файлов на шарды.
+
+-   `error_count` ([UInt64](../../sql-reference/data-types/int-uint.md)) — количество ошибок.
+
+-   `data_files` ([UInt64](../../sql-reference/data-types/int-uint.md)) — количество локальных файлов в папке.
+
+-   `data_compressed_bytes` ([UInt64](../../sql-reference/data-types/int-uint.md)) — размер сжатых данных в локальных файлах в байтах.
+
+-   `last_exception` ([String](../../sql-reference/data-types/string.md)) — текстовое сообщение о последней возникшей ошибке, если таковые имеются.
+
+**Пример**
+
+``` sql
+SELECT * FROM system.distribution_queue LIMIT 1 FORMAT Vertical;
+```
+
+``` text
+Row 1:
+──────
+database:              default
+table:                 dist
+data_path:             ./store/268/268bc070-3aad-4b1a-9cf2-4987580161af/default@127%2E0%2E0%2E2:9000/
+is_blocked:            1
+error_count:           0
+data_files:            1
+data_compressed_bytes: 499
+last_exception:        
+```
+
+**Смотрите также**
+
+-   [Движок таблиц Distributed](../../engines/table-engines/special/distributed.md)
+
+[Оригинальная статья](https://clickhouse.tech/docs/ru/operations/system_tables/distribution_queue) <!--hide-->
diff --git a/docs/ru/operations/system-tables/replication_queue.md b/docs/ru/operations/system-tables/replication_queue.md
new file mode 100644
index 00000000000..47f64aea55d
--- /dev/null
+++ b/docs/ru/operations/system-tables/replication_queue.md
@@ -0,0 +1,81 @@
+# system.replication_queue {#system_tables-replication_queue}
+
+Содержит информацию о задачах из очередей репликации, хранящихся в ZooKeeper, для таблиц семейства `ReplicatedMergeTree`.
+
+Столбцы:
+
+-   `database` ([String](../../sql-reference/data-types/string.md)) — имя базы данных.
+
+-   `table` ([String](../../sql-reference/data-types/string.md)) — имя таблицы.
+
+-   `replica_name` ([String](../../sql-reference/data-types/string.md)) — имя реплики в ZooKeeper. Разные реплики одной и той же таблицы имеют различные имена.
+
+-   `position` ([UInt32](../../sql-reference/data-types/int-uint.md)) — позиция задачи в очереди.
+
+-   `node_name` ([String](../../sql-reference/data-types/string.md)) — имя узла в ZooKeeper.
+
+-   `type` ([String](../../sql-reference/data-types/string.md)) — тип задачи в очереди: `GET_PARTS`, `MERGE_PARTS`, `DETACH_PARTS`, `DROP_PARTS` или `MUTATE_PARTS`.
+
+-   `create_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — дата и время отправки задачи на выполнение.
+
+-   `required_quorum` ([UInt32](../../sql-reference/data-types/int-uint.md)) — количество реплик, ожидающих завершения задачи, с подтверждением о завершении. Этот столбец актуален только для задачи `GET_PARTS`.
+
+-   `source_replica` ([String](../../sql-reference/data-types/string.md)) — имя исходной реплики.
+
+-   `new_part_name` ([String](../../sql-reference/data-types/string.md)) — имя нового куска.
+
+-   `parts_to_merge` ([Array](../../sql-reference/data-types/array.md) ([String](../../sql-reference/data-types/string.md))) — имена кусков, которые требуется смержить или обновить.
+
+-   `is_detach` ([UInt8](../../sql-reference/data-types/int-uint.md)) — флаг, указывающий на присутствие в очереди задачи `DETACH_PARTS`.
+
+-   `is_currently_executing` ([UInt8](../../sql-reference/data-types/int-uint.md)) — флаг, указывающий на выполнение конкретной задачи на данный момент.
+
+-   `num_tries` ([UInt32](../../sql-reference/data-types/int-uint.md)) — количество неудачных попыток выполнить задачу.
+
+-   `last_exception` ([String](../../sql-reference/data-types/string.md)) — текст сообщения о последней возникшей ошибке, если таковые имеются.
+
+-   `last_attempt_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — дата и время последней попытки выполнить задачу.
+
+-   `num_postponed` ([UInt32](../../sql-reference/data-types/int-uint.md)) — количество отложенных задач.
+
+-   `postpone_reason` ([String](../../sql-reference/data-types/string.md)) — причина, по которой была отложена задача.
+
+-   `last_postpone_time` ([Datetime](../../sql-reference/data-types/datetime.md)) — дата и время, когда была отложена задача в последний раз.
+
+-   `merge_type` ([String](../../sql-reference/data-types/string.md)) — тип текущего слияния. Пусто, если это мутация.
+
+**Пример**
+
+``` sql
+SELECT * FROM system.replication_queue LIMIT 1 FORMAT Vertical;
+```
+
+``` text
+Row 1:
+──────
+database:               merge
+table:                  visits_v2
+replica_name:           mtgiga001-1t.metrika.yandex.net
+position:               15
+node_name:              queue-0009325559
+type:                   MERGE_PARTS
+create_time:            2020-12-07 14:04:21
+required_quorum:        0
+source_replica:         mtgiga001-1t.metrika.yandex.net
+new_part_name:          20201130_121373_121384_2
+parts_to_merge:         ['20201130_121373_121378_1','20201130_121379_121379_0','20201130_121380_121380_0','20201130_121381_121381_0','20201130_121382_121382_0','20201130_121383_121383_0','20201130_121384_121384_0']
+is_detach:              0
+is_currently_executing: 0
+num_tries:              36
+last_exception:         Code: 226, e.displayText() = DB::Exception: Marks file '/opt/clickhouse/data/merge/visits_v2/tmp_fetch_20201130_121373_121384_2/CounterID.mrk' doesn't exist (version 20.8.7.15 (official build))
+last_attempt_time:      2020-12-08 17:35:54
+num_postponed:          0
+postpone_reason:        
+last_postpone_time:     1970-01-01 03:00:00
+```
+
+**Смотрите также**
+
+-   [Управление таблицами ReplicatedMergeTree](../../sql-reference/statements/system.md/#query-language-system-replicated)
+
+[Оригинальная статья](https://clickhouse.tech/docs/ru/operations/system_tables/replication_queue) <!--hide-->
diff --git a/docs/ru/sql-reference/data-types/lowcardinality.md b/docs/ru/sql-reference/data-types/lowcardinality.md
index ec9e4e7588e..d94cedd29ce 100644
--- a/docs/ru/sql-reference/data-types/lowcardinality.md
+++ b/docs/ru/sql-reference/data-types/lowcardinality.md
@@ -57,3 +57,5 @@ ORDER BY id
 - [A Magical Mystery Tour of the LowCardinality Data Type](https://www.altinity.com/blog/2019/3/27/low-cardinality).
 - [Reducing Clickhouse Storage Cost with the Low Cardinality Type – Lessons from an Instana Engineer](https://www.instana.com/blog/reducing-clickhouse-storage-cost-with-the-low-cardinality-type-lessons-from-an-instana-engineer/).
 - [String Optimization (video presentation in Russian)](https://youtu.be/rqf-ILRgBdY?list=PL0Z2YDlm0b3iwXCpEFiOOYmwXzVmjJfEt). [Slides in English](https://github.com/yandex/clickhouse-presentations/raw/master/meetup19/string_optimization.pdf).
+
+[Оригинальная статья](https://clickhouse.tech/docs/ru/sql-reference/data-types/lowcardinality/) <!--hide-->
diff --git a/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md b/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md
index 0015edfdf72..1367216e401 100644
--- a/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md
+++ b/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-sources.md
@@ -199,7 +199,7 @@ SOURCE(ODBC(
 
 ClickHouse получает от ODBC-драйвера информацию о квотировании и квотирует настройки в запросах к драйверу, поэтому имя таблицы нужно указывать в соответствии с регистром имени таблицы в базе данных.
 
-Если у вас есть проблемы с кодировками при использовании Oracle, ознакомьтесь с соответствующим разделом [FAQ](../../../faq/general.md#oracle-odbc-encodings).
+Если у вас есть проблемы с кодировками при использовании Oracle, ознакомьтесь с соответствующим разделом [FAQ](../../../faq/integration/oracle-odbc.md).
 
 ### Выявленная уязвимость в функционировании ODBC словарей {#vyiavlennaia-uiazvimost-v-funktsionirovanii-odbc-slovarei}
 
diff --git a/docs/ru/sql-reference/functions/date-time-functions.md b/docs/ru/sql-reference/functions/date-time-functions.md
index b7a077b3bd6..31482cde77f 100644
--- a/docs/ru/sql-reference/functions/date-time-functions.md
+++ b/docs/ru/sql-reference/functions/date-time-functions.md
@@ -593,6 +593,18 @@ SELECT dateDiff('hour', toDateTime('2018-01-01 22:00:00'), toDateTime('2018-01-0
 Например, `timeSlots(toDateTime('2012-01-01 12:20:00'), toUInt32(600)) = [toDateTime('2012-01-01 12:00:00'), toDateTime('2012-01-01 12:30:00')]`.
 Это нужно для поиска хитов, входящих в соответствующий визит.
 
+## toYYYYMM
+
+Переводит дату или дату со временем в число типа UInt32, содержащее номер года и месяца (YYYY * 100 + MM).
+
+## toYYYYMMDD
+
+Переводит дату или дату со временем в число типа UInt32, содержащее номер года, месяца и дня (YYYY * 10000 + MM * 100 + DD).
+
+## toYYYYMMDDhhmmss
+
+Переводит дату или дату со временем в число типа UInt64 содержащее номер года, месяца, дня и время (YYYY * 10000000000 + MM * 100000000 + DD * 1000000 + hh * 10000 + mm * 100 + ss).
+
 ## formatDateTime {#formatdatetime}
 
 Функция преобразует дату-и-время в строку по заданному шаблону. Важно: шаблон — константное выражение, поэтому использовать разные шаблоны в одной колонке не получится.
diff --git a/docs/ru/sql-reference/statements/select/order-by.md b/docs/ru/sql-reference/statements/select/order-by.md
index ea0f40b2dc0..f8b838cbd15 100644
--- a/docs/ru/sql-reference/statements/select/order-by.md
+++ b/docs/ru/sql-reference/statements/select/order-by.md
@@ -56,10 +56,188 @@ toc_title: ORDER BY
 
 ## Поддержка collation {#collation-support}
 
-Для сортировки по значениям типа String есть возможность указать collation (сравнение). Пример: `ORDER BY SearchPhrase COLLATE 'tr'` - для сортировки по поисковой фразе, по возрастанию, с учётом турецкого алфавита, регистронезависимо, при допущении, что строки в кодировке UTF-8. `COLLATE` может быть указан или не указан для каждого выражения в ORDER BY независимо. Если есть `ASC` или `DESC`, то `COLLATE` указывается после них. При использовании `COLLATE` сортировка всегда регистронезависима.
+Для сортировки по значениям типа [String](../../../sql-reference/data-types/string.md) есть возможность указать collation (сравнение). Пример: `ORDER BY SearchPhrase COLLATE 'tr'` - для сортировки по поисковой фразе, по возрастанию, с учётом турецкого алфавита, регистронезависимо, при допущении, что строки в кодировке UTF-8. `COLLATE` может быть указан или не указан для каждого выражения в ORDER BY независимо. Если есть `ASC` или `DESC`, то `COLLATE` указывается после них. При использовании `COLLATE` сортировка всегда регистронезависима.
+
+Сравнение поддерживается при использовании типов [LowCardinality](../../../sql-reference/data-types/lowcardinality.md), [Nullable](../../../sql-reference/data-types/nullable.md), [Array](../../../sql-reference/data-types/array.md) и [Tuple](../../../sql-reference/data-types/tuple.md).
 
 Рекомендуется использовать `COLLATE` только для окончательной сортировки небольшого количества строк, так как производительность сортировки с указанием `COLLATE` меньше, чем обычной сортировки по байтам.
 
+## Примеры с использованием сравнения {#collation-examples}
+
+Пример с значениями типа [String](../../../sql-reference/data-types/string.md):
+
+Входная таблица:
+
+``` text
+┌─x─┬─s────┐
+│ 1 │ bca  │
+│ 2 │ ABC  │
+│ 3 │ 123a │
+│ 4 │ abc  │
+│ 5 │ BCA  │
+└───┴──────┘
+```
+
+Запрос:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Результат:
+
+``` text
+┌─x─┬─s────┐
+│ 3 │ 123a │
+│ 4 │ abc  │
+│ 2 │ ABC  │
+│ 1 │ bca  │
+│ 5 │ BCA  │
+└───┴──────┘
+```
+
+Пример со строками типа [Nullable](../../../sql-reference/data-types/nullable.md):
+
+Входная таблица:
+
+``` text
+┌─x─┬─s────┐
+│ 1 │ bca  │
+│ 2 │ ᴺᵁᴸᴸ │
+│ 3 │ ABC  │
+│ 4 │ 123a │
+│ 5 │ abc  │
+│ 6 │ ᴺᵁᴸᴸ │
+│ 7 │ BCA  │
+└───┴──────┘
+```
+
+Запрос:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Результат:
+
+``` text
+┌─x─┬─s────┐
+│ 4 │ 123a │
+│ 5 │ abc  │
+│ 3 │ ABC  │
+│ 1 │ bca  │
+│ 7 │ BCA  │
+│ 6 │ ᴺᵁᴸᴸ │
+│ 2 │ ᴺᵁᴸᴸ │
+└───┴──────┘
+```
+
+Пример со строками в [Array](../../../sql-reference/data-types/array.md):
+
+Входная таблица:
+
+``` text
+┌─x─┬─s─────────────┐
+│ 1 │ ['Z']         │
+│ 2 │ ['z']         │
+│ 3 │ ['a']         │
+│ 4 │ ['A']         │
+│ 5 │ ['z','a']     │
+│ 6 │ ['z','a','a'] │
+│ 7 │ ['']          │
+└───┴───────────────┘
+```
+
+Запрос:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Результат:
+
+``` text
+┌─x─┬─s─────────────┐
+│ 7 │ ['']          │
+│ 3 │ ['a']         │
+│ 4 │ ['A']         │
+│ 2 │ ['z']         │
+│ 5 │ ['z','a']     │
+│ 6 │ ['z','a','a'] │
+│ 1 │ ['Z']         │
+└───┴───────────────┘
+```
+
+Пример со строками типа [LowCardinality](../../../sql-reference/data-types/lowcardinality.md):
+
+Входная таблица:
+
+```text
+┌─x─┬─s───┐
+│ 1 │ Z   │
+│ 2 │ z   │
+│ 3 │ a   │
+│ 4 │ A   │
+│ 5 │ za  │
+│ 6 │ zaa │
+│ 7 │     │
+└───┴─────┘
+```
+
+Запрос:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Результат:
+
+```text
+┌─x─┬─s───┐
+│ 7 │     │
+│ 3 │ a   │
+│ 4 │ A   │
+│ 2 │ z   │
+│ 1 │ Z   │
+│ 5 │ za  │
+│ 6 │ zaa │
+└───┴─────┘
+```
+
+Пример со строками в [Tuple](../../../sql-reference/data-types/tuple.md):
+
+```text
+┌─x─┬─s───────┐
+│ 1 │ (1,'Z') │
+│ 2 │ (1,'z') │
+│ 3 │ (1,'a') │
+│ 4 │ (2,'z') │
+│ 5 │ (1,'A') │
+│ 6 │ (2,'Z') │
+│ 7 │ (2,'A') │
+└───┴─────────┘
+```
+
+Запрос:
+
+```sql
+SELECT * FROM collate_test ORDER BY s ASC COLLATE 'en';
+```
+
+Результат:
+
+```text
+┌─x─┬─s───────┐
+│ 3 │ (1,'a') │
+│ 5 │ (1,'A') │
+│ 2 │ (1,'z') │
+│ 1 │ (1,'Z') │
+│ 7 │ (2,'A') │
+│ 4 │ (2,'z') │
+│ 6 │ (2,'Z') │
+└───┴─────────┘
+```
+
 ## Деталь реализации {#implementation-details}
 
 Если кроме `ORDER BY` указан также не слишком большой [LIMIT](limit.md), то расходуется меньше оперативки. Иначе расходуется количество памяти, пропорциональное количеству данных для сортировки. При распределённой обработке запроса, если отсутствует [GROUP BY](group-by.md), сортировка частично делается на удалённых серверах, а на сервере-инициаторе запроса производится слияние результатов. Таким образом, при распределённой сортировке, может сортироваться объём данных, превышающий размер памяти на одном сервере.
diff --git a/docs/tr/development/developer-instruction.md b/docs/tr/development/developer-instruction.md
index 51a6c4345c6..79c93104551 100644
--- a/docs/tr/development/developer-instruction.md
+++ b/docs/tr/development/developer-instruction.md
@@ -257,8 +257,8 @@ Clickhouse'un geliştirilmesi genellikle gerçekçi veri kümelerinin yüklenmes
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/tr/getting-started/example-datasets/metrica.md b/docs/tr/getting-started/example-datasets/metrica.md
index 6a727d1ab55..0ad7debf54d 100644
--- a/docs/tr/getting-started/example-datasets/metrica.md
+++ b/docs/tr/getting-started/example-datasets/metrica.md
@@ -9,14 +9,14 @@ toc_title: "\xDCye.Metrica Verileri"
 
 Veri kümesi, isabetlerle ilgili anonimleştirilmiş verileri içeren iki tablodan oluşur (`hits_v1`) ve ziyaret visitsler (`visits_v1`(kayıt olmak için).Metrica. Yandex hakkında daha fazla bilgi edinebilirsiniz.Metrica içinde [ClickHouse geçmişi](../../introduction/history.md) bölme.
 
-Veri kümesi iki tablodan oluşur, bunlardan biri sıkıştırılmış olarak indirilebilir `tsv.xz` dosya veya hazırlanmış bölümler olarak. Buna ek olarak, genişletilmiş bir sürümü `hits` 100 milyon satır içeren tablo TSV olarak mevcuttur https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz ve hazırlanan bölümler olarak https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
+Veri kümesi iki tablodan oluşur, bunlardan biri sıkıştırılmış olarak indirilebilir `tsv.xz` dosya veya hazırlanmış bölümler olarak. Buna ek olarak, genişletilmiş bir sürümü `hits` 100 milyon satır içeren tablo TSV olarak mevcuttur https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz ve hazırlanan bölümler olarak https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz.
 
 ## Hazırlanan bölümlerden tablolar elde etme {#obtaining-tables-from-prepared-partitions}
 
 İndirme ve ithalat tablo hits:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -26,7 +26,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 İndirme ve ithalat ziyaretleri:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -38,7 +38,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 Sıkıştırılmış TSV dosyasından indir ve İçe Aktar:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -52,7 +52,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 Sıkıştırılmış tsv dosyasından ziyaretleri indirin ve içe aktarın:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/tr/getting-started/example-datasets/nyc-taxi.md b/docs/tr/getting-started/example-datasets/nyc-taxi.md
index 7c2fa26eb05..ebbb595aa5e 100644
--- a/docs/tr/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/tr/getting-started/example-datasets/nyc-taxi.md
@@ -285,7 +285,7 @@ Diğer şeylerin yanı sıra, MERGETREE üzerinde en iyi duruma getirme sorgusun
 ## Hazırlanan Bölüm downloadlerin indir downloadilmesi {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/tr/getting-started/example-datasets/ontime.md b/docs/tr/getting-started/example-datasets/ontime.md
index f1d477dbc6e..5754263fc00 100644
--- a/docs/tr/getting-started/example-datasets/ontime.md
+++ b/docs/tr/getting-started/example-datasets/ontime.md
@@ -156,7 +156,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## Hazırlanan Bölüm downloadlerin indir downloadilmesi {#download-of-prepared-partitions}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/tr/getting-started/tutorial.md b/docs/tr/getting-started/tutorial.md
index 5cca914fb35..449710ba33b 100644
--- a/docs/tr/getting-started/tutorial.md
+++ b/docs/tr/getting-started/tutorial.md
@@ -87,8 +87,8 @@ clickhouse-client --query='INSERT INTO table FORMAT TabSeparated' < data.tsv
 ### Tablo verilerini indirin ve ayıklayın {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 Çıkarılan dosyalar yaklaşık 10GB boyutundadır.
diff --git a/docs/tr/interfaces/cli.md b/docs/tr/interfaces/cli.md
index de62b3003af..dc0ca9e7591 100644
--- a/docs/tr/interfaces/cli.md
+++ b/docs/tr/interfaces/cli.md
@@ -59,7 +59,7 @@ Komut satırı dayanmaktadır ‘replxx’ (benzer ‘readline’). Başka bir d
 
 Varsayılan olarak, kullanılan biçim PrettyCompact. Sorgunun biçim yan tümcesinde veya belirterek biçimi değiştirebilirsiniz `\G` sorgunun sonunda, `--format` veya `--vertical` komut satırında veya istemci yapılandırma dosyasını kullanarak bağımsız değişken.
 
-İstemciden çıkmak için Ctrl+D (veya Ctrl+C) tuşlarına basın veya bir sorgu yerine aşağıdakilerden birini girin: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
+İstemciden çıkmak için Ctrl+D tuşlarına basın veya bir sorgu yerine aşağıdakilerden birini girin: “exit”, “quit”, “logout”, “exit;”, “quit;”, “logout;”, “q”, “Q”, “:q”
 
 Bir sorguyu işlerken, istemci şunları gösterir:
 
diff --git a/docs/tr/operations/performance-test.md b/docs/tr/operations/performance-test.md
index 498f3851861..8f12d63490d 100644
--- a/docs/tr/operations/performance-test.md
+++ b/docs/tr/operations/performance-test.md
@@ -48,7 +48,7 @@ Bu talimat ile ClickHouse paketlerinin kurulumu olmadan herhangi bir sunucuda te
 
 <!-- -->
 
-      wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+      wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
       tar xvf hits_100m_obfuscated_v1.tar.xz -C .
       mv hits_100m_obfuscated_v1/* .
 
diff --git a/docs/tr/sql-reference/data-types/lowcardinality.md b/docs/tr/sql-reference/data-types/lowcardinality.md
new file mode 100644
index 00000000000..dd3a9aa1c0d
--- /dev/null
+++ b/docs/tr/sql-reference/data-types/lowcardinality.md
@@ -0,0 +1 @@
+../../../../en/sql-reference/data-types/lowcardinality.md
\ No newline at end of file
diff --git a/docs/zh/development/developer-instruction.md b/docs/zh/development/developer-instruction.md
index 3e2ccf5da35..5a5d8d82144 100644
--- a/docs/zh/development/developer-instruction.md
+++ b/docs/zh/development/developer-instruction.md
@@ -242,8 +242,8 @@ ClickHouse的架构描述可以在此处查看：https://clickhouse.tech/docs/en
 
     sudo apt install wget xz-utils
 
-    wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
-    wget https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz
+    wget https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz
 
     xz -v -d hits_v1.tsv.xz
     xz -v -d visits_v1.tsv.xz
diff --git a/docs/zh/getting-started/example-datasets/metrica.md b/docs/zh/getting-started/example-datasets/metrica.md
index 353a24ce0cb..62ba28a92ef 100644
--- a/docs/zh/getting-started/example-datasets/metrica.md
+++ b/docs/zh/getting-started/example-datasets/metrica.md
@@ -7,14 +7,14 @@ toc_title: Yandex.Metrica Data
 
 数据集由两个表组成，包含关于Yandex.Metrica的hits(`hits_v1`)和visit(`visits_v1`)的匿名数据。你可以阅读更多关于Yandex的信息。在[ClickHouse历史](../../introduction/history.md)的Metrica部分。
 
-数据集由两个表组成，他们中的任何一个都可以下载作为一个压缩`tsv.xz`的文件或准备的分区。除此之外,一个扩展版的`hits`表包含1亿行TSV在https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_100m_obfuscated_v1.tsv.xz，准备分区在https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz。
+数据集由两个表组成，他们中的任何一个都可以下载作为一个压缩`tsv.xz`的文件或准备的分区。除此之外,一个扩展版的`hits`表包含1亿行TSV在https://datasets.clickhouse.tech/hits/tsv/hits_100m_obfuscated_v1.tsv.xz，准备分区在https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz。
 
 ## 从准备好的分区获取表 {#obtaining-tables-from-prepared-partitions}
 
 下载和导入`hits`表:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar
+curl -O https://datasets.clickhouse.tech/hits/partitions/hits_v1.tar
 tar xvf hits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -24,7 +24,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 下载和导入`visits`表:
 
 ``` bash
-curl -O https://clickhouse-datasets.s3.yandex.net/visits/partitions/visits_v1.tar
+curl -O https://datasets.clickhouse.tech/visits/partitions/visits_v1.tar
 tar xvf visits_v1.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 # check permissions on unpacked data, fix if required
 sudo service clickhouse-server restart
@@ -36,7 +36,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.visits_v1"
 从TSV压缩文件下载并导入`hits`:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.hits_v1 ( WatchID UInt64,  JavaEnable UInt8,  Title String,  GoodEvent Int16,  EventTime DateTime,  EventDate Date,  CounterID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RegionID UInt32,  UserID UInt64,  CounterClass Int8,  OS UInt8,  UserAgent UInt8,  URL String,  Referer String,  URLDomain String,  RefererDomain String,  Refresh UInt8,  IsRobot UInt8,  RefererCategories Array(UInt16),  URLCategories Array(UInt16), URLRegions Array(UInt32),  RefererRegions Array(UInt32),  ResolutionWidth UInt16,  ResolutionHeight UInt16,  ResolutionDepth UInt8,  FlashMajor UInt8, FlashMinor UInt8,  FlashMinor2 String,  NetMajor UInt8,  NetMinor UInt8, UserAgentMajor UInt16,  UserAgentMinor FixedString(2),  CookieEnable UInt8, JavascriptEnable UInt8,  IsMobile UInt8,  MobilePhone UInt8,  MobilePhoneModel String,  Params String,  IPNetworkID UInt32,  TraficSourceID Int8, SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  IsArtifical UInt8,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  ClientTimeZone Int16,  ClientEventTime DateTime,  SilverlightVersion1 UInt8, SilverlightVersion2 UInt8,  SilverlightVersion3 UInt32,  SilverlightVersion4 UInt16,  PageCharset String,  CodeVersion UInt32,  IsLink UInt8,  IsDownload UInt8,  IsNotBounce UInt8,  FUniqID UInt64,  HID UInt32,  IsOldCounter UInt8, IsEvent UInt8,  IsParameter UInt8,  DontCountHits UInt8,  WithHash UInt8, HitColor FixedString(1),  UTCEventTime DateTime,  Age UInt8,  Sex UInt8,  Income UInt8,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16), RemoteIP UInt32,  RemoteIP6 FixedString(16),  WindowName Int32,  OpenerName Int32,  HistoryLength Int16,  BrowserLanguage FixedString(2),  BrowserCountry FixedString(2),  SocialNetwork String,  SocialAction String,  HTTPError UInt16, SendTiming Int32,  DNSTiming Int32,  ConnectTiming Int32,  ResponseStartTiming Int32,  ResponseEndTiming Int32,  FetchTiming Int32,  RedirectTiming Int32, DOMInteractiveTiming Int32,  DOMContentLoadedTiming Int32,  DOMCompleteTiming Int32,  LoadEventStartTiming Int32,  LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32,  FirstPaintTiming Int32,  RedirectCount Int8, SocialSourceNetworkID UInt8,  SocialSourcePage String,  ParamPrice Int64, ParamOrderID String,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16, GoalsReached Array(UInt32),  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String, UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String, FromTag String,  HasGCLID UInt8,  RefererHash UInt64,  URLHash UInt64,  CLID UInt32,  YCLID UInt64,  ShareService String,  ShareURL String,  ShareTitle String,  ParsedParams Nested(Key1 String,  Key2 String, Key3 String, Key4 String, Key5 String,  ValueDouble Float64),  IslandID FixedString(16),  RequestNum UInt32,  RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
@@ -50,7 +50,7 @@ clickhouse-client --query "SELECT COUNT(*) FROM datasets.hits_v1"
 从压缩tsv文件下载和导入`visits`:
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 # now create table
 clickhouse-client --query "CREATE DATABASE IF NOT EXISTS datasets"
 clickhouse-client --query "CREATE TABLE datasets.visits_v1 ( CounterID UInt32,  StartDate Date,  Sign Int8,  IsNew UInt8,  VisitID UInt64,  UserID UInt64,  StartTime DateTime,  Duration UInt32,  UTCStartTime DateTime,  PageViews Int32,  Hits Int32,  IsBounce UInt8,  Referer String,  StartURL String,  RefererDomain String,  StartURLDomain String,  EndURL String,  LinkURL String,  IsDownload UInt8,  TraficSourceID Int8,  SearchEngineID UInt16,  SearchPhrase String,  AdvEngineID UInt8,  PlaceID Int32,  RefererCategories Array(UInt16),  URLCategories Array(UInt16),  URLRegions Array(UInt32),  RefererRegions Array(UInt32),  IsYandex UInt8,  GoalReachesDepth Int32,  GoalReachesURL Int32,  GoalReachesAny Int32,  SocialSourceNetworkID UInt8,  SocialSourcePage String,  MobilePhoneModel String,  ClientEventTime DateTime,  RegionID UInt32,  ClientIP UInt32,  ClientIP6 FixedString(16),  RemoteIP UInt32,  RemoteIP6 FixedString(16),  IPNetworkID UInt32,  SilverlightVersion3 UInt32,  CodeVersion UInt32,  ResolutionWidth UInt16,  ResolutionHeight UInt16,  UserAgentMajor UInt16,  UserAgentMinor UInt16,  WindowClientWidth UInt16,  WindowClientHeight UInt16,  SilverlightVersion2 UInt8,  SilverlightVersion4 UInt16,  FlashVersion3 UInt16,  FlashVersion4 UInt16,  ClientTimeZone Int16,  OS UInt8,  UserAgent UInt8,  ResolutionDepth UInt8,  FlashMajor UInt8,  FlashMinor UInt8,  NetMajor UInt8,  NetMinor UInt8,  MobilePhone UInt8,  SilverlightVersion1 UInt8,  Age UInt8,  Sex UInt8,  Income UInt8,  JavaEnable UInt8,  CookieEnable UInt8,  JavascriptEnable UInt8,  IsMobile UInt8,  BrowserLanguage UInt16,  BrowserCountry UInt16,  Interests UInt16,  Robotness UInt8,  GeneralInterests Array(UInt16),  Params Array(String),  Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime,  Price Int64,  OrderID String, CurrencyID UInt32),  WatchIDs Array(UInt64),  ParamSumPrice Int64,  ParamCurrency FixedString(3),  ParamCurrencyID UInt16,  ClickLogID UInt64,  ClickEventID Int32,  ClickGoodEvent Int32,  ClickEventTime DateTime,  ClickPriorityID Int32,  ClickPhraseID Int32,  ClickPageID Int32,  ClickPlaceID Int32,  ClickTypeID Int32,  ClickResourceID Int32,  ClickCost UInt32,  ClickClientIP UInt32,  ClickDomainID UInt32,  ClickURL String,  ClickAttempt UInt8,  ClickOrderID UInt32,  ClickBannerID UInt32,  ClickMarketCategoryID UInt32,  ClickMarketPP UInt32,  ClickMarketCategoryName String,  ClickMarketPPName String,  ClickAWAPSCampaignName String,  ClickPageName String,  ClickTargetType UInt16,  ClickTargetPhraseID UInt64,  ClickContextType UInt8,  ClickSelectType Int8,  ClickOptions String,  ClickGroupBannerID Int32,  OpenstatServiceName String,  OpenstatCampaignID String,  OpenstatAdID String,  OpenstatSourceID String,  UTMSource String,  UTMMedium String,  UTMCampaign String,  UTMContent String,  UTMTerm String,  FromTag String,  HasGCLID UInt8,  FirstVisit DateTime,  PredLastVisit Date,  LastVisit Date,  TotalVisits UInt32,  TraficSource    Nested(ID Int8,  SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String),  Attendance FixedString(16),  CLID UInt32,  YCLID UInt64,  NormalizedRefererHash UInt64,  SearchPhraseHash UInt64,  RefererDomainHash UInt64,  NormalizedStartURLHash UInt64,  StartURLDomainHash UInt64,  NormalizedEndURLHash UInt64,  TopLevelDomain UInt64,  URLScheme UInt64,  OpenstatServiceNameHash UInt64,  OpenstatCampaignIDHash UInt64,  OpenstatAdIDHash UInt64,  OpenstatSourceIDHash UInt64,  UTMSourceHash UInt64,  UTMMediumHash UInt64,  UTMCampaignHash UInt64,  UTMContentHash UInt64,  UTMTermHash UInt64,  FromHash UInt64,  WebVisorEnabled UInt8,  WebVisorActivity UInt32,  ParsedParams    Nested(Key1 String,  Key2 String,  Key3 String,  Key4 String, Key5 String, ValueDouble    Float64),  Market Nested(Type UInt8, GoalID UInt32, OrderID String,  OrderPrice Int64,  PP UInt32,  DirectPlaceID UInt32,  DirectOrderID  UInt32,  DirectBannerID UInt32,  GoodID String, GoodName String, GoodQuantity Int32,  GoodPrice Int64),  IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192"
diff --git a/docs/zh/getting-started/example-datasets/nyc-taxi.md b/docs/zh/getting-started/example-datasets/nyc-taxi.md
index c6b41e9d396..0e4ff67d6e7 100644
--- a/docs/zh/getting-started/example-datasets/nyc-taxi.md
+++ b/docs/zh/getting-started/example-datasets/nyc-taxi.md
@@ -283,7 +283,7 @@ SELECT formatReadableSize(sum(bytes)) FROM system.parts WHERE table = 'trips_mer
 ## 下载预处理好的分区数据 {#xia-zai-yu-chu-li-hao-de-fen-qu-shu-ju}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/trips_mergetree/partitions/trips_mergetree.tar
+$ curl -O https://datasets.clickhouse.tech/trips_mergetree/partitions/trips_mergetree.tar
 $ tar xvf trips_mergetree.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/zh/getting-started/example-datasets/ontime.md b/docs/zh/getting-started/example-datasets/ontime.md
index 4c21eee51a2..3921f71fc7e 100644
--- a/docs/zh/getting-started/example-datasets/ontime.md
+++ b/docs/zh/getting-started/example-datasets/ontime.md
@@ -154,7 +154,7 @@ $ for i in *.zip; do echo $i; unzip -cq $i '*.csv' | sed 's/\.00//g' | clickhous
 ## 下载预处理好的分区数据 {#xia-zai-yu-chu-li-hao-de-fen-qu-shu-ju}
 
 ``` bash
-$ curl -O https://clickhouse-datasets.s3.yandex.net/ontime/partitions/ontime.tar
+$ curl -O https://datasets.clickhouse.tech/ontime/partitions/ontime.tar
 $ tar xvf ontime.tar -C /var/lib/clickhouse # path to ClickHouse data directory
 $ # check permissions of unpacked data, fix if required
 $ sudo service clickhouse-server restart
diff --git a/docs/zh/getting-started/tutorial.md b/docs/zh/getting-started/tutorial.md
index 93f368bc2dc..5e34729a9c5 100644
--- a/docs/zh/getting-started/tutorial.md
+++ b/docs/zh/getting-started/tutorial.md
@@ -85,8 +85,8 @@ clickhouse-client --query='INSERT INTO table FORMAT TabSeparated' < data.tsv
 ### 下载并提取表数据 {#download-and-extract-table-data}
 
 ``` bash
-curl https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
-curl https://clickhouse-datasets.s3.yandex.net/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
+curl https://datasets.clickhouse.tech/hits/tsv/hits_v1.tsv.xz | unxz --threads=`nproc` > hits_v1.tsv
+curl https://datasets.clickhouse.tech/visits/tsv/visits_v1.tsv.xz | unxz --threads=`nproc` > visits_v1.tsv
 ```
 
 提取的文件大小约为10GB。
diff --git a/docs/zh/operations/performance-test.md b/docs/zh/operations/performance-test.md
index cde27a8547d..7728325bb9a 100644
--- a/docs/zh/operations/performance-test.md
+++ b/docs/zh/operations/performance-test.md
@@ -48,7 +48,7 @@ toc_title: "\u6D4B\u8BD5\u786C\u4EF6"
 
 <!-- -->
 
-      wget https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_100m_obfuscated_v1.tar.xz
+      wget https://datasets.clickhouse.tech/hits/partitions/hits_100m_obfuscated_v1.tar.xz
       tar xvf hits_100m_obfuscated_v1.tar.xz -C .
       mv hits_100m_obfuscated_v1/* .
 
diff --git a/docs/zh/sql-reference/statements/misc.md b/docs/zh/sql-reference/statements/misc.md
index b4297d1ed4f..26b3997a4d6 100644
--- a/docs/zh/sql-reference/statements/misc.md
+++ b/docs/zh/sql-reference/statements/misc.md
@@ -9,19 +9,21 @@ toc_title: "\u5176\u4ED6"
 
 ## ATTACH {#attach}
 
-这个查询是完全一样的 `CREATE`，但是
+与`CREATE`类似，但有所区别
 
--   而不是这个词 `CREATE` 它使用这个词 `ATTACH`.
--   查询不会在磁盘上创建数据，但假定数据已经在适当的位置，只是将有关表的信息添加到服务器。
-    执行附加查询后，服务器将知道表的存在。
+-   使用关键词 `ATTACH`
+-   查询不会在磁盘上创建数据。但会假定数据已经在对应位置存放，同时将与表相关的信息添加到服务器。
+    执行 `ATTACH` 查询后，服务器将知道表已经被创建。
 
-如果表之前已分离 (`DETACH`），意味着其结构是已知的，可以使用速记而不限定该结构。
+如果表之前已分离 (`DETACH`），意味着其结构是已知的，可以使用简要的写法来建立表，即不需要定义表结构的Schema细节。
 
 ``` sql
 ATTACH TABLE [IF NOT EXISTS] [db.]name [ON CLUSTER cluster]
 ```
 
-启动服务器时使用此查询。 服务器将表元数据作为文件存储 `ATTACH` 查询，它只是在启动时运行（除了在服务器上显式创建的系统表）。
+启动服务器时会自动触发此查询。 
+
+服务器将表的元数据作为文件存储 `ATTACH` 查询，它只是在启动时运行。有些表例外，如系统表，它们是在服务器上显式指定的。
 
 ## CHECK TABLE {#check-table}
 
@@ -31,13 +33,12 @@ ATTACH TABLE [IF NOT EXISTS] [db.]name [ON CLUSTER cluster]
 CHECK TABLE [db.]name
 ```
 
-该 `CHECK TABLE` 查询将实际文件大小与存储在服务器上的预期值进行比较。 如果文件大小与存储的值不匹配，则表示数据已损坏。 例如，这可能是由查询执行期间的系统崩溃引起的。
+`CHECK TABLE` 查询会比较存储在服务器上的实际文件大小与预期值。 如果文件大小与存储的值不匹配，则表示数据已损坏。 例如，这可能是由查询执行期间的系统崩溃引起的。
 
-查询响应包含 `result` 具有单行的列。 该行的值为
-[布尔值](../../sql-reference/data-types/boolean.md) 类型:
+查询返回一行结果，列名为 `result`, 该行的值为 [布尔值](../../sql-reference/data-types/boolean.md) 类型:
 
--   0-表中的数据已损坏。
--   1-数据保持完整性。
+-   0-表中的数据已损坏；
+-   1-数据保持完整性；
 
 该 `CHECK TABLE` 查询支持下表引擎:
 
@@ -56,13 +57,14 @@ CHECK TABLE [db.]name
 
 如果表已损坏，则可以将未损坏的数据复制到另一个表。 要做到这一点:
 
-1.  创建一个与损坏的表结构相同的新表。 要做到这一点，请执行查询 `CREATE TABLE <new_table_name> AS <damaged_table_name>`.
+1.  创建一个与损坏的表结构相同的新表。 请执行查询 `CREATE TABLE <new_table_name> AS <damaged_table_name>`.
 2.  将 [max_threads](../../operations/settings/settings.md#settings-max_threads) 值设置为1，以在单个线程中处理下一个查询。 要这样做，请运行查询 `SET max_threads = 1`.
 3.  执行查询 `INSERT INTO <new_table_name> SELECT * FROM <damaged_table_name>`. 此请求将未损坏的数据从损坏的表复制到另一个表。 只有损坏部分之前的数据才会被复制。
 4.  重新启动 `clickhouse-client` 以重置 `max_threads` 值。
 
 ## DESCRIBE TABLE {#misc-describe-table}
 
+查看表的描述信息，返回各列的Schema，语法如下：
 ``` sql
 DESC|DESCRIBE TABLE [db.]table [INTO OUTFILE filename] [FORMAT format]
 ```
@@ -73,24 +75,25 @@ DESC|DESCRIBE TABLE [db.]table [INTO OUTFILE filename] [FORMAT format]
 -   `type`— 列的类型。
 -   `default_type` — [默认表达式](create.md#create-default-values) (`DEFAULT`, `MATERIALIZED` 或 `ALIAS`)中使用的子句。 如果没有指定默认表达式，则列包含一个空字符串。
 -   `default_expression` — `DEFAULT` 子句中指定的值。
--   `comment_expression` — 注释。
+-   `comment_expression` — 注释信息。
 
 嵌套数据结构以 “expanded” 格式输出。 每列分别显示，列名后加点号。
 
 ## DETACH {#detach}
 
-从服务器中删除有关 ‘name’ 表的信息。 服务器停止了解该表的存在。
+从服务器中删除目标表信息（删除对象是表), 执行查询后,服务器视作该表已经不存在。
 
 ``` sql
 DETACH TABLE [IF EXISTS] [db.]name [ON CLUSTER cluster]
 ```
 
 这不会删除表的数据或元数据。 在下一次服务器启动时，服务器将读取元数据并再次查找该表。
-同样，可以使用 `ATTACH` 查询重新连接一个 “detached” 的表（系统表除外，没有为它们存储元数据）。
+也可以不停止服务器的情况下，使用前面介绍的 `ATTACH` 查询来重新关联该表（系统表除外，没有为它们存储元数据）。
 
 ## DROP {#drop}
 
 删除已经存在的实体。如果指定 `IF EXISTS`， 则如果实体不存在，则不返回错误。
+建议使用时添加 `IF EXISTS` 修饰符。
 
 ## DROP DATABASE {#drop-database}
 
@@ -135,7 +138,7 @@ DROP USER [IF EXISTS] name [,...] [ON CLUSTER cluster_name]
 
 删除角色。
 
-已删除的角色将从授予该角色的所有实体撤销。
+同时该角色所拥有的权限也会被收回。
 
 语法:
 
@@ -199,6 +202,8 @@ EXISTS [TEMPORARY] [TABLE|DICTIONARY] [db.]name [INTO OUTFILE filename] [FORMAT
 
 ## KILL QUERY {#kill-query-statement}
 
+
+
 ``` sql
 KILL QUERY [ON CLUSTER cluster]
   WHERE <where expression to SELECT FROM system.processes query>
@@ -219,16 +224,17 @@ KILL QUERY WHERE query_id='2-857d-4a57-9ee0-327da5d60a90'
 KILL QUERY WHERE user='username' SYNC
 ```
 
-只读用户只能停止自己的查询。
+只读用户只能停止自己提交的查询。
 
-默认情况下，使用异步版本的查询 (`ASYNC`），不等待确认查询已停止。
+默认情况下，使用异步版本的查询 (`ASYNC`），不需要等待确认查询已停止。
 
-同步版本 (`SYNC`）等待所有查询停止，并在停止时显示有关每个进程的信息。
-响应包含 `kill_status` 列，该列可以采用以下值:
+而相对的，终止同步版本 (`SYNC`）的查询会显示每步停止时间。
+
+返回信息包含 `kill_status` 列，该列可以采用以下值:
 
 1.  ‘finished’ – 查询已成功终止。
 2.  ‘waiting’ – 发送查询信号终止后，等待查询结束。
-3.  其他值解释为什么查询不能停止。
+3.  其他值，会解释为什么查询不能停止。
 
 测试查询 (`TEST`）仅检查用户的权限，并显示要停止的查询列表。
 
diff --git a/docs/zh/sql-reference/statements/select/limit.md b/docs/zh/sql-reference/statements/select/limit.md
index b079248deca..d6d827552b0 100644
--- a/docs/zh/sql-reference/statements/select/limit.md
+++ b/docs/zh/sql-reference/statements/select/limit.md
@@ -14,7 +14,7 @@ toc_title: LIMIT
 
 ## LIMIT … WITH TIES 修饰符 {#limit-with-ties}
 
-如果为 `LIMIT n[,m]` 设置了 `WITH TIES` ，并且声明了 `ORDER BY expr_list`, you will get in result first `n` or `n,m` rows and all rows with same `ORDER BY` fields values equal to row at position `n` for `LIMIT n` and `m` for `LIMIT n,m`.
+如果为 `LIMIT n[,m]` 设置了 `WITH TIES` ，并且声明了 `ORDER BY expr_list`, 除了得到无修饰符的结果（正常情况下的 `limit n`, 前n行数据), 还会返回与第`n`行具有相同排序字段的行(即如果第n+1行的字段与第n行 拥有相同的排序字段，同样返回该结果.
 
 此修饰符可以与： [ORDER BY … WITH FILL modifier](../../../sql-reference/statements/select/order-by.md#orderby-with-fill) 组合使用.
 
@@ -38,7 +38,7 @@ SELECT * FROM (
 └───┘
 ```
 
-单子执行了 `WITH TIES` 修饰符后
+添加 `WITH TIES` 修饰符后
 
 ``` sql
 SELECT * FROM (
@@ -59,4 +59,8 @@ SELECT * FROM (
 └───┘
 ```
 
-cause row number 6 have same value “2” for field `n` as row number 5
+虽然指定了`LIMIT 5`, 但第6行的`n`字段值为2，与第5行相同，因此也作为满足条件的记录返回。
+简而言之，该修饰符可理解为是否增加“并列行”的数据。
+
+``` sql，
+``` sql
diff --git a/programs/benchmark/Benchmark.cpp b/programs/benchmark/Benchmark.cpp
index 8c69a545017..ae1d16ce402 100644
--- a/programs/benchmark/Benchmark.cpp
+++ b/programs/benchmark/Benchmark.cpp
@@ -27,6 +27,7 @@
 #include <IO/WriteHelpers.h>
 #include <IO/Operators.h>
 #include <IO/ConnectionTimeouts.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <IO/UseSSL.h>
 #include <DataStreams/RemoteBlockInputStream.h>
 #include <Interpreters/Context.h>
@@ -95,6 +96,7 @@ public:
         }
 
         global_context.makeGlobalContext();
+        global_context.setSettings(settings);
 
         std::cerr << std::fixed << std::setprecision(3);
 
@@ -404,7 +406,7 @@ private:
         Stopwatch watch;
         RemoteBlockInputStream stream(
             *(*connection_entries[connection_index]),
-            query, {}, global_context, &settings, nullptr, Scalars(), Tables(), query_processing_stage);
+            query, {}, global_context, nullptr, Scalars(), Tables(), query_processing_stage);
         if (!query_id.empty())
             stream.setQueryId(query_id);
 
diff --git a/programs/client/Client.cpp b/programs/client/Client.cpp
index 92ae37123fa..9d0227a51e3 100644
--- a/programs/client/Client.cpp
+++ b/programs/client/Client.cpp
@@ -2335,7 +2335,7 @@ public:
                 "Suggestion limit for how many databases, tables and columns to fetch.")
             ("multiline,m", "multiline")
             ("multiquery,n", "multiquery")
-            ("queries-file,qf", po::value<std::string>(), "file path with queries to execute")
+            ("queries-file", po::value<std::string>(), "file path with queries to execute")
             ("format,f", po::value<std::string>(), "default output format")
             ("testmode,T", "enable test hints in comments")
             ("ignore-error", "do not stop processing in multiquery mode")
diff --git a/programs/copier/ClusterCopier.cpp b/programs/copier/ClusterCopier.cpp
index 2f19fc47fd2..ca09e7c1889 100644
--- a/programs/copier/ClusterCopier.cpp
+++ b/programs/copier/ClusterCopier.cpp
@@ -5,6 +5,7 @@
 #include <Common/ZooKeeper/ZooKeeper.h>
 #include <Common/ZooKeeper/KeeperException.h>
 #include <Common/setThreadName.h>
+#include <IO/ConnectionTimeoutsContext.h>
 
 
 namespace DB
@@ -1588,11 +1589,14 @@ void ClusterCopier::dropParticularPartitionPieceFromAllHelpingTables(const TaskT
     LOG_DEBUG(log, "All helping tables dropped partition {}", partition_name);
 }
 
-String ClusterCopier::getRemoteCreateTable(const DatabaseAndTableName & table, Connection & connection, const Settings * settings)
+String ClusterCopier::getRemoteCreateTable(const DatabaseAndTableName & table, Connection & connection, const Settings & settings)
 {
+    Context remote_context(context);
+    remote_context.setSettings(settings);
+
     String query = "SHOW CREATE TABLE " + getQuotedTable(table);
     Block block = getBlockWithAllStreamData(std::make_shared<RemoteBlockInputStream>(
-            connection, query, InterpreterShowCreateQuery::getSampleBlock(), context, settings));
+            connection, query, InterpreterShowCreateQuery::getSampleBlock(), remote_context));
 
     return typeid_cast<const ColumnString &>(*block.safeGetByPosition(0).column).getDataAt(0).toString();
 }
@@ -1604,7 +1608,7 @@ ASTPtr ClusterCopier::getCreateTableForPullShard(const ConnectionTimeouts & time
     String create_query_pull_str = getRemoteCreateTable(
             task_shard.task_table.table_pull,
             *connection_entry,
-            &task_cluster->settings_pull);
+            task_cluster->settings_pull);
 
     ParserCreateQuery parser_create_query;
     const auto & settings = context.getSettingsRef();
@@ -1856,6 +1860,9 @@ UInt64 ClusterCopier::executeQueryOnCluster(
             auto timeouts = ConnectionTimeouts::getTCPTimeoutsWithFailover(shard_settings).getSaturated(shard_settings.max_execution_time);
             auto connections = shard.pool->getMany(timeouts, &shard_settings, pool_mode);
 
+            Context shard_context(context);
+            shard_context.setSettings(shard_settings);
+
             for (auto & connection : connections)
             {
                 if (connection.isNull())
@@ -1864,7 +1871,7 @@ UInt64 ClusterCopier::executeQueryOnCluster(
                 try
                 {
                     /// CREATE TABLE and DROP PARTITION queries return empty block
-                    RemoteBlockInputStream stream{*connection, query, Block{}, context, &shard_settings};
+                    RemoteBlockInputStream stream{*connection, query, Block{}, shard_context};
                     NullBlockOutputStream output{Block{}};
                     copyData(stream, output);
 
diff --git a/programs/copier/ClusterCopier.h b/programs/copier/ClusterCopier.h
index beaf247dfc8..9aff5493cf8 100644
--- a/programs/copier/ClusterCopier.h
+++ b/programs/copier/ClusterCopier.h
@@ -154,7 +154,7 @@ protected:
     /// table we can get rid of partition pieces (partitions in helping tables).
     void dropParticularPartitionPieceFromAllHelpingTables(const TaskTable & task_table, const String & partition_name);
 
-    String getRemoteCreateTable(const DatabaseAndTableName & table, Connection & connection, const Settings * settings = nullptr);
+    String getRemoteCreateTable(const DatabaseAndTableName & table, Connection & connection, const Settings & settings);
 
     ASTPtr getCreateTableForPullShard(const ConnectionTimeouts & timeouts, TaskShard & task_shard);
 
diff --git a/programs/copier/ClusterCopierApp.cpp b/programs/copier/ClusterCopierApp.cpp
index c2946e12c34..e3169a49ecf 100644
--- a/programs/copier/ClusterCopierApp.cpp
+++ b/programs/copier/ClusterCopierApp.cpp
@@ -1,6 +1,7 @@
 #include "ClusterCopierApp.h"
 #include <Common/StatusFile.h>
 #include <Common/TerminalSize.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <Formats/registerFormats.h>
 #include <unistd.h>
 
diff --git a/programs/copier/TaskCluster.h b/programs/copier/TaskCluster.h
index 68d98c648f5..5b28f461dd8 100644
--- a/programs/copier/TaskCluster.h
+++ b/programs/copier/TaskCluster.h
@@ -1,6 +1,7 @@
 #pragma once
 
 #include "Aliases.h"
+#include <Poco/Util/AbstractConfiguration.h>
 
 namespace DB
 {
@@ -12,7 +13,9 @@ namespace ErrorCodes
 struct TaskCluster
 {
     TaskCluster(const String & task_zookeeper_path_, const String & default_local_database_)
-            : task_zookeeper_path(task_zookeeper_path_), default_local_database(default_local_database_) {}
+            : task_zookeeper_path(task_zookeeper_path_)
+            , default_local_database(default_local_database_)
+    {}
 
     void loadTasks(const Poco::Util::AbstractConfiguration & config, const String & base_key = "");
 
diff --git a/programs/odbc-bridge/ColumnInfoHandler.cpp b/programs/odbc-bridge/ColumnInfoHandler.cpp
index 9e2307715c5..ee4daa3e16d 100644
--- a/programs/odbc-bridge/ColumnInfoHandler.cpp
+++ b/programs/odbc-bridge/ColumnInfoHandler.cpp
@@ -115,7 +115,7 @@ void ODBCColumnsInfoHandler::handleRequest(Poco::Net::HTTPServerRequest & reques
         std::string name = schema_name.empty() ? backQuoteIfNeed(table_name) : backQuoteIfNeed(schema_name) + "." + backQuoteIfNeed(table_name);
         WriteBufferFromOwnString buf;
         std::string input = "SELECT * FROM " + name + " WHERE 1 = 0";
-        ParserQueryWithOutput parser;
+        ParserQueryWithOutput parser(input.data() + input.size());
         ASTPtr select = parseQuery(parser, input.data(), input.data() + input.size(), "", context_settings.max_query_size, context_settings.max_parser_depth);
 
         IAST::FormatSettings settings(buf, true);
diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index 28d8301b920..1eeec398f44 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -139,6 +139,28 @@ void setupTmpPath(Poco::Logger * log, const std::string & path)
     }
 }
 
+int waitServersToFinish(std::vector<DB::ProtocolServerAdapter> & servers, size_t seconds_to_wait)
+{
+    const int sleep_max_ms = 1000 * seconds_to_wait;
+    const int sleep_one_ms = 100;
+    int sleep_current_ms = 0;
+    int current_connections = 0;
+    while (sleep_current_ms < sleep_max_ms)
+    {
+        current_connections = 0;
+        for (auto & server : servers)
+        {
+            server.stop();
+            current_connections += server.currentConnections();
+        }
+        if (!current_connections)
+            break;
+        sleep_current_ms += sleep_one_ms;
+        std::this_thread::sleep_for(std::chrono::milliseconds(sleep_one_ms));
+    }
+    return current_connections;
+}
+
 }
 
 namespace DB
@@ -366,7 +388,7 @@ int Server::main(const std::vector<std::string> & /*args*/)
     LOG_WARNING(log, "Server was built in debug mode. It will work slowly.");
 #endif
 
-#if defined(ADDRESS_SANITIZER) || defined(THREAD_SANITIZER) || defined(MEMORY_SANITIZER)
+#if defined(SANITIZER)
     LOG_WARNING(log, "Server was built with sanitizer. It will work slowly.");
 #endif
 
@@ -794,8 +816,29 @@ int Server::main(const std::vector<std::string> & /*args*/)
 
         LOG_DEBUG(log, "Shut down storages.");
 
-        for (auto & server : servers_to_start_before_tables)
-            server.stop();
+        if (!servers_to_start_before_tables.empty())
+        {
+            LOG_DEBUG(log, "Waiting for current connections to servers for tables to finish.");
+            int current_connections = 0;
+            for (auto & server : servers_to_start_before_tables)
+            {
+                server.stop();
+                current_connections += server.currentConnections();
+            }
+
+            if (current_connections)
+                LOG_INFO(log, "Closed all listening sockets. Waiting for {} outstanding connections.", current_connections);
+            else
+                LOG_INFO(log, "Closed all listening sockets.");
+
+            if (current_connections > 0)
+                current_connections = waitServersToFinish(servers_to_start_before_tables, config().getInt("shutdown_wait_unfinished", 5));
+
+            if (current_connections)
+                LOG_INFO(log, "Closed connections to servers for tables. But {} remain. Probably some tables of other users cannot finish their connections after context shutdown.", current_connections);
+            else
+                LOG_INFO(log, "Closed connections to servers for tables.");
+        }
 
         /** Explicitly destroy Context. It is more convenient than in destructor of Server, because logger is still available.
           * At this moment, no one could own shared part of Context.
@@ -1167,24 +1210,7 @@ int Server::main(const std::vector<std::string> & /*args*/)
             global_context->getProcessList().killAllQueries();
 
             if (current_connections)
-            {
-                const int sleep_max_ms = 1000 * config().getInt("shutdown_wait_unfinished", 5);
-                const int sleep_one_ms = 100;
-                int sleep_current_ms = 0;
-                while (sleep_current_ms < sleep_max_ms)
-                {
-                    current_connections = 0;
-                    for (auto & server : servers)
-                    {
-                        server.stop();
-                        current_connections += server.currentConnections();
-                    }
-                    if (!current_connections)
-                        break;
-                    sleep_current_ms += sleep_one_ms;
-                    std::this_thread::sleep_for(std::chrono::milliseconds(sleep_one_ms));
-                }
-            }
+                current_connections = waitServersToFinish(servers, config().getInt("shutdown_wait_unfinished", 5));
 
             if (current_connections)
                 LOG_INFO(log, "Closed connections. But {} remain."
diff --git a/programs/server/config.d/test_keeper_port.xml b/programs/server/config.d/test_keeper_port.xml
new file mode 120000
index 00000000000..f3f721caae0
--- /dev/null
+++ b/programs/server/config.d/test_keeper_port.xml
@@ -0,0 +1 @@
+../../../tests/config/config.d/test_keeper_port.xml
\ No newline at end of file
diff --git a/programs/server/config.xml b/programs/server/config.xml
index f41c346bbed..e05ef61eed7 100644
--- a/programs/server/config.xml
+++ b/programs/server/config.xml
@@ -48,97 +48,121 @@
         -->
     </logger>
 
-    <send_crash_reports>
-        <!-- Changing <enabled> to true allows sending crash reports to -->
-        <!-- the ClickHouse core developers team via Sentry https://sentry.io -->
-        <!-- Doing so at least in pre-production environments is highly appreciated -->
-        <enabled>false</enabled>
-        <!-- Change <anonymize> to true if you don't feel comfortable attaching the server hostname to the crash report -->
-        <anonymize>false</anonymize>
-        <!-- Default endpoint should be changed to different Sentry DSN only if you have -->
-        <!-- some in-house engineers or hired consultants who're going to debug ClickHouse issues for you -->
-        <endpoint>https://6f33034cfe684dd7a3ab9875e57b1c8d@o388870.ingest.sentry.io/5226277</endpoint>
-    </send_crash_reports>
+    <!-- It is the name that will be shown in the clickhouse-client.
+         By default, anything with "production" will be highlighted in red in query prompt.
+    -->
+    <!--display_name>production</display_name-->
 
-    <!--display_name>production</display_name--> <!-- It is the name that will be shown in the client -->
+    <!-- Port for HTTP API. See also 'https_port' for secure connections.
+         This interface is also used by ODBC and JDBC drivers (DataGrip, Dbeaver, ...)
+         and by most of web interfaces (embedded UI, Grafana, Redash, ...).
+      -->
     <http_port>8123</http_port>
+
+    <!-- Port for interaction by native protocol with:
+         - clickhouse-client and other native ClickHouse tools (clickhouse-benchmark, clickhouse-copier);
+         - clickhouse-server with other clickhouse-servers for distributed query processing;
+         - ClickHouse drivers and applications supporting native protocol
+         (this protocol is also informally called as "the TCP protocol");
+         See also 'tcp_port_secure' for secure connections.
+    -->
     <tcp_port>9000</tcp_port>
+
+    <!-- Compatibility with MySQL protocol.
+         ClickHouse will pretend to be MySQL for applications connecting to this port.
+    -->
     <mysql_port>9004</mysql_port>
 
-    <!-- For HTTPS and SSL over native protocol. -->
-    <!--
-    <https_port>8443</https_port>
-    <tcp_port_secure>9440</tcp_port_secure>
+    <!-- Compatibility with PostgreSQL protocol.
+         ClickHouse will pretend to be PostgreSQL for applications connecting to this port.
     -->
+    <!-- <postgresql_port>9005</postgresql_port> -->
 
-    <!-- TCP with PROXY protocol (PROXY header sent for every connection) -->
-    <!--
-    <tcp_with_proxy_port>9010</tcp_with_proxy_port>
+    <!-- HTTP API with TLS (HTTPS).
+         You have to configure certificate to enable this interface.
+         See the openSSL section below.
     -->
+    <!-- <https_port>8443</https_port> -->
 
-    <!-- Used with https_port and tcp_port_secure. Full ssl options list: https://github.com/ClickHouse-Extras/poco/blob/master/NetSSL_OpenSSL/include/Poco/Net/SSLManager.h#L71 -->
-    <openSSL>
-        <server> <!-- Used for https server AND secure tcp port -->
-            <!-- openssl req -subj "/CN=localhost" -new -newkey rsa:2048 -days 365 -nodes -x509 -keyout /etc/clickhouse-server/server.key -out /etc/clickhouse-server/server.crt -->
-            <certificateFile>/etc/clickhouse-server/server.crt</certificateFile>
-            <privateKeyFile>/etc/clickhouse-server/server.key</privateKeyFile>
-            <!-- openssl dhparam -out /etc/clickhouse-server/dhparam.pem 4096 -->
-            <dhParamsFile>/etc/clickhouse-server/dhparam.pem</dhParamsFile>
-            <verificationMode>none</verificationMode>
-            <loadDefaultCAFile>true</loadDefaultCAFile>
-            <cacheSessions>true</cacheSessions>
-            <disableProtocols>sslv2,sslv3</disableProtocols>
-            <preferServerCiphers>true</preferServerCiphers>
-        </server>
-
-        <client> <!-- Used for connecting to https dictionary source and secured Zookeeper communication -->
-            <loadDefaultCAFile>true</loadDefaultCAFile>
-            <cacheSessions>true</cacheSessions>
-            <disableProtocols>sslv2,sslv3</disableProtocols>
-            <preferServerCiphers>true</preferServerCiphers>
-            <!-- Use for self-signed: <verificationMode>none</verificationMode> -->
-            <invalidCertificateHandler>
-                <!-- Use for self-signed: <name>AcceptCertificateHandler</name> -->
-                <name>RejectCertificateHandler</name>
-            </invalidCertificateHandler>
-        </client>
-    </openSSL>
-
-    <!-- Default root page on http[s] server. For example load UI from https://tabix.io/ when opening http://localhost:8123 -->
-    <!--
-    <http_server_default_response><![CDATA[<html ng-app="SMI2"><head><base href="http://ui.tabix.io/"></head><body><div ui-view="" class="content-ui"></div><script src="http://loader.tabix.io/master.js"></script></body></html>]]></http_server_default_response>
+    <!-- Native interface with TLS.
+         You have to configure certificate to enable this interface.
+         See the openSSL section below.
     -->
+    <!-- <tcp_port_secure>9440</tcp_port_secure> -->
 
-    <!-- Port for communication between replicas. Used for data exchange. -->
+    <!-- Native interface wrapped with PROXYv1 protocol
+         PROXYv1 header sent for every connection.
+         ClickHouse will extract information about proxy-forwarded client address from the header.
+    -->
+    <!-- <tcp_with_proxy_port>9011</tcp_with_proxy_port> -->
+
+    <!-- Port for communication between replicas. Used for data exchange.
+         It provides low-level data access between servers.
+         This port should not be accessible from untrusted networks.
+         See also 'interserver_http_credentials'.
+         Data transferred over connections to this port should not go through untrusted networks.
+         See also 'interserver_https_port'.
+      -->
     <interserver_http_port>9009</interserver_http_port>
 
+    <!-- Port for communication between replicas with TLS.
+         You have to configure certificate to enable this interface.
+         See the openSSL section below.
+         See also 'interserver_http_credentials'.
+      -->
+    <!-- <interserver_https_port>9010</interserver_https_port> -->
+
     <!-- Hostname that is used by other replicas to request this server.
          If not specified, than it is determined analogous to 'hostname -f' command.
-         This setting could be used to switch replication to another network interface.
+         This setting could be used to switch replication to another network interface
+         (the server may be connected to multiple networks via multiple addresses)
       -->
     <!--
     <interserver_http_host>example.yandex.ru</interserver_http_host>
     -->
 
-    <!-- Listen specified host. use :: (wildcard IPv6 address), if you want to accept connections both with IPv4 and IPv6 from everywhere. -->
+    <!-- You can specify credentials for authenthication between replicas.
+         This is required when interserver_https_port is accessible from untrusted networks,
+         and also recommended to avoid SSRF attacks from possibly compromised services in your network.
+      -->
+    <!--<interserver_http_credentials>
+        <user>interserver</user>
+        <password></password>
+    </interserver_http_credentials>-->
+
+    <!-- Listen specified address.
+         Use :: (wildcard IPv6 address), if you want to accept connections both with IPv4 and IPv6 from everywhere.
+         Notes:
+         If you open connections from wildcard address, make sure that at least one of the following measures applied:
+         - server is protected by firewall and not accessible from untrusted networks;
+         - all users are restricted to subset of network addresses (see users.xml);
+         - all users have strong passwords, only secure (TLS) interfaces are accessible, or connections are only made via TLS interfaces.
+         - users without password have readonly access.
+         See also: https://www.shodan.io/search?query=clickhouse
+      -->
     <!-- <listen_host>::</listen_host> -->
-    <!-- Same for hosts with disabled ipv6: -->
+
+    <!-- Same for hosts without support for IPv6: -->
     <!-- <listen_host>0.0.0.0</listen_host> -->
 
-    <!-- Default values - try listen localhost on ipv4 and ipv6: -->
+    <!-- Default values - try listen localhost on IPv4 and IPv6. -->
     <!--
     <listen_host>::1</listen_host>
     <listen_host>127.0.0.1</listen_host>
     -->
-    <!-- Don't exit if ipv6 or ipv4 unavailable, but listen_host with this protocol specified -->
+
+    <!-- Don't exit if IPv6 or IPv4 networks are unavailable while trying to listen. -->
     <!-- <listen_try>0</listen_try> -->
 
-    <!-- Allow listen on same address:port -->
+    <!-- Allow multiple servers to listen on the same address:port. This is not recommended.
+      -->
     <!-- <listen_reuse_port>0</listen_reuse_port> -->
 
     <!-- <listen_backlog>64</listen_backlog> -->
 
     <max_connections>4096</max_connections>
+
+    <!-- For 'Connection: keep-alive' in HTTP 1.1 -->
     <keep_alive_timeout>3</keep_alive_timeout>
 
     <!-- gRPC protocol (see src/Server/grpc_protos/clickhouse_grpc.proto for the API) -->
@@ -172,6 +196,43 @@
         <verbose_logs>false</verbose_logs>
     </grpc>
 
+    <!-- Used with https_port and tcp_port_secure. Full ssl options list: https://github.com/ClickHouse-Extras/poco/blob/master/NetSSL_OpenSSL/include/Poco/Net/SSLManager.h#L71 -->
+    <openSSL>
+        <server> <!-- Used for https server AND secure tcp port -->
+            <!-- openssl req -subj "/CN=localhost" -new -newkey rsa:2048 -days 365 -nodes -x509 -keyout /etc/clickhouse-server/server.key -out /etc/clickhouse-server/server.crt -->
+            <certificateFile>/etc/clickhouse-server/server.crt</certificateFile>
+            <privateKeyFile>/etc/clickhouse-server/server.key</privateKeyFile>
+            <!-- dhparams are optional. You can delete the <dhParamsFile> element.
+                 To generate dhparams, use the following command:
+                  openssl dhparam -out /etc/clickhouse-server/dhparam.pem 4096 
+                 Only file format with BEGIN DH PARAMETERS is supported.
+              -->
+            <dhParamsFile>/etc/clickhouse-server/dhparam.pem</dhParamsFile>
+            <verificationMode>none</verificationMode>
+            <loadDefaultCAFile>true</loadDefaultCAFile>
+            <cacheSessions>true</cacheSessions>
+            <disableProtocols>sslv2,sslv3</disableProtocols>
+            <preferServerCiphers>true</preferServerCiphers>
+        </server>
+
+        <client> <!-- Used for connecting to https dictionary source and secured Zookeeper communication -->
+            <loadDefaultCAFile>true</loadDefaultCAFile>
+            <cacheSessions>true</cacheSessions>
+            <disableProtocols>sslv2,sslv3</disableProtocols>
+            <preferServerCiphers>true</preferServerCiphers>
+            <!-- Use for self-signed: <verificationMode>none</verificationMode> -->
+            <invalidCertificateHandler>
+                <!-- Use for self-signed: <name>AcceptCertificateHandler</name> -->
+                <name>RejectCertificateHandler</name>
+            </invalidCertificateHandler>
+        </client>
+    </openSSL>
+
+    <!-- Default root page on http[s] server. For example load UI from https://tabix.io/ when opening http://localhost:8123 -->
+    <!--
+    <http_server_default_response><![CDATA[<html ng-app="SMI2"><head><base href="http://ui.tabix.io/"></head><body><div ui-view="" class="content-ui"></div><script src="http://loader.tabix.io/master.js"></script></body></html>]]></http_server_default_response>
+    -->
+
     <!-- Maximum number of concurrent queries. -->
     <max_concurrent_queries>100</max_concurrent_queries>
 
@@ -605,7 +666,7 @@
                 event_date + INTERVAL 1 WEEK
                 event_date + INTERVAL 7 DAY DELETE
                 event_date + INTERVAL 2 WEEK TO DISK 'bbb'
-        
+
         <ttl>event_date + INTERVAL 30 DAY DELETE</ttl>
         -->
 
@@ -887,6 +948,18 @@
     </http_handlers>
     -->
 
+    <send_crash_reports>
+        <!-- Changing <enabled> to true allows sending crash reports to -->
+        <!-- the ClickHouse core developers team via Sentry https://sentry.io -->
+        <!-- Doing so at least in pre-production environments is highly appreciated -->
+        <enabled>false</enabled>
+        <!-- Change <anonymize> to true if you don't feel comfortable attaching the server hostname to the crash report -->
+        <anonymize>false</anonymize>
+        <!-- Default endpoint should be changed to different Sentry DSN only if you have -->
+        <!-- some in-house engineers or hired consultants who're going to debug ClickHouse issues for you -->
+        <endpoint>https://6f33034cfe684dd7a3ab9875e57b1c8d@o388870.ingest.sentry.io/5226277</endpoint>
+    </send_crash_reports>
+
     <!-- Uncomment to disable ClickHouse internal DNS caching. -->
     <!-- <disable_internal_dns_cache>1</disable_internal_dns_cache> -->
 </yandex>
diff --git a/programs/server/play.html b/programs/server/play.html
index 12435f55793..5c3d3566af4 100644
--- a/programs/server/play.html
+++ b/programs/server/play.html
@@ -534,6 +534,23 @@
     var theme = window.localStorage.getItem('theme');
     if (theme) {
         setColorTheme(theme);
+    } else {
+        /// Obtain system-level user preference
+        var media_query_list = window.matchMedia('prefers-color-scheme: dark')
+
+        if (media_query_list.matches) {
+            /// Set without saving to localstorage
+            document.documentElement.setAttribute('data-theme', 'dark');
+        }
+
+        /// There is a rumor that on some computers, the theme is changing automatically on day/night.
+        media_query_list.addEventListener('change', function(e) {
+            if (e.matches) {
+                document.documentElement.setAttribute('data-theme', 'dark');
+            } else {
+                document.documentElement.setAttribute('data-theme', 'light');
+            }
+        });
     }
 
     document.getElementById('toggle-light').onclick = function()
diff --git a/src/Access/ContextAccess.cpp b/src/Access/ContextAccess.cpp
index 0e4f3fe7871..a55d0768a7e 100644
--- a/src/Access/ContextAccess.cpp
+++ b/src/Access/ContextAccess.cpp
@@ -41,42 +41,7 @@ namespace
     }
 
 
-    void applyParamsToAccessRights(AccessRights & access, const ContextAccessParams & params)
-    {
-        static const AccessFlags table_ddl = AccessType::CREATE_DATABASE | AccessType::CREATE_TABLE | AccessType::CREATE_VIEW
-            | AccessType::ALTER_TABLE | AccessType::ALTER_VIEW | AccessType::DROP_DATABASE | AccessType::DROP_TABLE | AccessType::DROP_VIEW
-            | AccessType::TRUNCATE;
-
-        static const AccessFlags dictionary_ddl = AccessType::CREATE_DICTIONARY | AccessType::DROP_DICTIONARY;
-        static const AccessFlags table_and_dictionary_ddl = table_ddl | dictionary_ddl;
-        static const AccessFlags write_table_access = AccessType::INSERT | AccessType::OPTIMIZE;
-        static const AccessFlags write_dcl_access = AccessType::ACCESS_MANAGEMENT - AccessType::SHOW_ACCESS;
-
-        if (params.readonly)
-            access.revoke(write_table_access | table_and_dictionary_ddl | write_dcl_access | AccessType::SYSTEM | AccessType::KILL_QUERY);
-
-        if (params.readonly == 1)
-        {
-            /// Table functions are forbidden in readonly mode.
-            /// For readonly = 2 they're allowed.
-            access.revoke(AccessType::CREATE_TEMPORARY_TABLE);
-        }
-
-        if (!params.allow_ddl)
-            access.revoke(table_and_dictionary_ddl);
-
-        if (!params.allow_introspection)
-            access.revoke(AccessType::INTROSPECTION);
-
-        if (params.readonly)
-        {
-            /// No grant option in readonly mode.
-            access.revokeGrantOption(AccessType::ALL);
-        }
-    }
-
-
-    void addImplicitAccessRights(AccessRights & access)
+    AccessRights addImplicitAccessRights(const AccessRights & access)
     {
         auto modifier = [&](const AccessFlags & flags, const AccessFlags & min_flags_with_children, const AccessFlags & max_flags_with_children, const std::string_view & database, const std::string_view & table, const std::string_view & column) -> AccessFlags
         {
@@ -150,23 +115,12 @@ namespace
             return res;
         };
 
-        access.modifyFlags(modifier);
-
-        /// Transform access to temporary tables into access to "_temporary_and_external_tables" database.
-        if (access.isGranted(AccessType::CREATE_TEMPORARY_TABLE))
-            access.grant(AccessFlags::allTableFlags() | AccessFlags::allColumnFlags(), DatabaseCatalog::TEMPORARY_DATABASE);
+        AccessRights res = access;
+        res.modifyFlags(modifier);
 
         /// Anyone has access to the "system" database.
-        access.grant(AccessType::SELECT, DatabaseCatalog::SYSTEM_DATABASE);
-    }
-
-
-    AccessRights calculateFinalAccessRights(const AccessRights & access_from_user_and_roles, const ContextAccessParams & params)
-    {
-        AccessRights res_access = access_from_user_and_roles;
-        applyParamsToAccessRights(res_access, params);
-        addImplicitAccessRights(res_access);
-        return res_access;
+        res.grant(AccessType::SELECT, DatabaseCatalog::SYSTEM_DATABASE);
+        return res;
     }
 
 
@@ -176,6 +130,12 @@ namespace
         ids[0] = id;
         return ids;
     }
+
+    /// Helper for using in templates.
+    std::string_view getDatabase() { return {}; }
+
+    template <typename... OtherArgs>
+    std::string_view getDatabase(const std::string_view & arg1, const OtherArgs &...) { return arg1; }
 }
 
 
@@ -203,10 +163,7 @@ void ContextAccess::setUser(const UserPtr & user_) const
         /// User has been dropped.
         auto nothing_granted = std::make_shared<AccessRights>();
         access = nothing_granted;
-        access_without_readonly = nothing_granted;
-        access_with_allow_ddl = nothing_granted;
-        access_with_allow_introspection = nothing_granted;
-        access_from_user_and_roles = nothing_granted;
+        access_with_implicit = nothing_granted;
         subscription_for_user_change = {};
         subscription_for_roles_changes = {};
         enabled_roles = nullptr;
@@ -270,12 +227,8 @@ void ContextAccess::setRolesInfo(const std::shared_ptr<const EnabledRolesInfo> &
 
 void ContextAccess::calculateAccessRights() const
 {
-    access_from_user_and_roles = std::make_shared<AccessRights>(mixAccessRightsFromUserAndRoles(*user, *roles_info));
-    access = std::make_shared<AccessRights>(calculateFinalAccessRights(*access_from_user_and_roles, params));
-
-    access_without_readonly = nullptr;
-    access_with_allow_ddl = nullptr;
-    access_with_allow_introspection = nullptr;
+    access = std::make_shared<AccessRights>(mixAccessRightsFromUserAndRoles(*user, *roles_info));
+    access_with_implicit = std::make_shared<AccessRights>(addImplicitAccessRights(*access));
 
     if (trace_log)
     {
@@ -287,6 +240,7 @@ void ContextAccess::calculateAccessRights() const
         }
         LOG_TRACE(trace_log, "Settings: readonly={}, allow_ddl={}, allow_introspection_functions={}", params.readonly, params.allow_ddl, params.allow_introspection);
         LOG_TRACE(trace_log, "List of all grants: {}", access->toString());
+        LOG_TRACE(trace_log, "List of all grants including implicit: {}", access_with_implicit->toString());
     }
 }
 
@@ -340,6 +294,7 @@ std::shared_ptr<const ContextAccess> ContextAccess::getFullAccess()
     static const std::shared_ptr<const ContextAccess> res = []
     {
         auto full_access = std::shared_ptr<ContextAccess>(new ContextAccess);
+        full_access->is_full_access = true;
         full_access->access = std::make_shared<AccessRights>(AccessRights::getFullAccess());
         full_access->enabled_quota = EnabledQuota::getUnlimitedQuota();
         return full_access;
@@ -362,323 +317,306 @@ std::shared_ptr<const SettingsConstraints> ContextAccess::getSettingsConstraints
 }
 
 
-std::shared_ptr<const AccessRights> ContextAccess::getAccess() const
+std::shared_ptr<const AccessRights> ContextAccess::getAccessRights() const
 {
     std::lock_guard lock{mutex};
     return access;
 }
 
 
-template <bool grant_option, typename... Args>
-bool ContextAccess::isGrantedImpl2(const AccessFlags & flags, const Args &... args) const
+std::shared_ptr<const AccessRights> ContextAccess::getAccessRightsWithImplicit() const
 {
-    bool access_granted;
-    if constexpr (grant_option)
-        access_granted = getAccess()->hasGrantOption(flags, args...);
-    else
-        access_granted = getAccess()->isGranted(flags, args...);
-
-    if (trace_log)
-        LOG_TRACE(trace_log, "Access {}: {}{}", (access_granted ? "granted" : "denied"), (AccessRightsElement{flags, args...}.toString()),
-                  (grant_option ? " WITH GRANT OPTION" : ""));
-
-    return access_granted;
+    std::lock_guard lock{mutex};
+    return access_with_implicit;
 }
 
-template <bool grant_option>
-bool ContextAccess::isGrantedImpl(const AccessFlags & flags) const
+
+template <bool throw_if_denied, bool grant_option>
+bool ContextAccess::checkAccessImpl(const AccessFlags & flags) const
 {
-    return isGrantedImpl2<grant_option>(flags);
+    return checkAccessImpl2<throw_if_denied, grant_option>(flags);
 }
 
-template <bool grant_option, typename... Args>
-bool ContextAccess::isGrantedImpl(const AccessFlags & flags, const std::string_view & database, const Args &... args) const
+template <bool throw_if_denied, bool grant_option, typename... Args>
+bool ContextAccess::checkAccessImpl(const AccessFlags & flags, const std::string_view & database, const Args &... args) const
 {
-    return isGrantedImpl2<grant_option>(flags, database.empty() ? params.current_database : database, args...);
+    return checkAccessImpl2<throw_if_denied, grant_option>(flags, database.empty() ? params.current_database : database, args...);
 }
 
-template <bool grant_option>
-bool ContextAccess::isGrantedImpl(const AccessRightsElement & element) const
+template <bool throw_if_denied, bool grant_option>
+bool ContextAccess::checkAccessImpl(const AccessRightsElement & element) const
 {
     if (element.any_database)
-        return isGrantedImpl<grant_option>(element.access_flags);
+        return checkAccessImpl<throw_if_denied, grant_option>(element.access_flags);
     else if (element.any_table)
-        return isGrantedImpl<grant_option>(element.access_flags, element.database);
+        return checkAccessImpl<throw_if_denied, grant_option>(element.access_flags, element.database);
     else if (element.any_column)
-        return isGrantedImpl<grant_option>(element.access_flags, element.database, element.table);
+        return checkAccessImpl<throw_if_denied, grant_option>(element.access_flags, element.database, element.table);
     else
-        return isGrantedImpl<grant_option>(element.access_flags, element.database, element.table, element.columns);
+        return checkAccessImpl<throw_if_denied, grant_option>(element.access_flags, element.database, element.table, element.columns);
 }
 
-template <bool grant_option>
-bool ContextAccess::isGrantedImpl(const AccessRightsElements & elements) const
+template <bool throw_if_denied, bool grant_option>
+bool ContextAccess::checkAccessImpl(const AccessRightsElements & elements) const
 {
     for (const auto & element : elements)
-        if (!isGrantedImpl<grant_option>(element))
+        if (!checkAccessImpl<throw_if_denied, grant_option>(element))
             return false;
     return true;
 }
 
-bool ContextAccess::isGranted(const AccessFlags & flags) const { return isGrantedImpl<false>(flags); }
-bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database) const { return isGrantedImpl<false>(flags, database); }
-bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { return isGrantedImpl<false>(flags, database, table); }
-bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { return isGrantedImpl<false>(flags, database, table, column); }
-bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { return isGrantedImpl<false>(flags, database, table, columns); }
-bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { return isGrantedImpl<false>(flags, database, table, columns); }
-bool ContextAccess::isGranted(const AccessRightsElement & element) const { return isGrantedImpl<false>(element); }
-bool ContextAccess::isGranted(const AccessRightsElements & elements) const { return isGrantedImpl<false>(elements); }
-
-bool ContextAccess::hasGrantOption(const AccessFlags & flags) const { return isGrantedImpl<true>(flags); }
-bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database) const { return isGrantedImpl<true>(flags, database); }
-bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { return isGrantedImpl<true>(flags, database, table); }
-bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { return isGrantedImpl<true>(flags, database, table, column); }
-bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { return isGrantedImpl<true>(flags, database, table, columns); }
-bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { return isGrantedImpl<true>(flags, database, table, columns); }
-bool ContextAccess::hasGrantOption(const AccessRightsElement & element) const { return isGrantedImpl<true>(element); }
-bool ContextAccess::hasGrantOption(const AccessRightsElements & elements) const { return isGrantedImpl<true>(elements); }
-
-
-template <bool grant_option, typename... Args>
-void ContextAccess::checkAccessImpl2(const AccessFlags & flags, const Args &... args) const
+template <bool throw_if_denied, bool grant_option, typename... Args>
+bool ContextAccess::checkAccessImpl2(const AccessFlags & flags, const Args &... args) const
 {
-    if constexpr (grant_option)
+    auto access_granted = [&]
     {
-        if (hasGrantOption(flags, args...))
-            return;
-    }
-    else
-    {
-        if (isGranted(flags, args...))
-            return;
-    }
-
-    auto show_error = [&](const String & msg, int error_code)
-    {
-        throw Exception(user_name + ": " + msg, error_code);
+        if (trace_log)
+            LOG_TRACE(trace_log, "Access granted: {}{}", (AccessRightsElement{flags, args...}.toString()),
+                      (grant_option ? " WITH GRANT OPTION" : ""));
+        return true;
     };
 
-    std::lock_guard lock{mutex};
-
-    if (!user)
-        show_error("User has been dropped", ErrorCodes::UNKNOWN_USER);
-
-    if (grant_option && access->isGranted(flags, args...))
+    auto access_denied = [&](const String & error_msg, int error_code)
     {
-        show_error(
-            "Not enough privileges. "
-            "The required privileges have been granted, but without grant option. "
-            "To execute this query it's necessary to have the grant "
-                + AccessRightsElement{flags, args...}.toString() + " WITH GRANT OPTION",
+        if (trace_log)
+            LOG_TRACE(trace_log, "Access denied: {}{}", (AccessRightsElement{flags, args...}.toString()),
+                      (grant_option ? " WITH GRANT OPTION" : ""));
+        if constexpr (throw_if_denied)
+            throw Exception(getUserName() + ": " + error_msg, error_code);
+        return false;
+    };
+
+    if (!flags || is_full_access)
+        return access_granted();
+
+    if (!getUser())
+        return access_denied("User has been dropped", ErrorCodes::UNKNOWN_USER);
+
+    /// Access to temporary tables is controlled in an unusual way, not like normal tables.
+    /// Creating of temporary tables is controlled by AccessType::CREATE_TEMPORARY_TABLES grant,
+    /// and other grants are considered as always given.
+    /// The DatabaseCatalog class won't resolve StorageID for temporary tables
+    /// which shouldn't be accessed.
+    if (getDatabase(args...) == DatabaseCatalog::TEMPORARY_DATABASE)
+        return access_granted();
+
+    auto acs = getAccessRightsWithImplicit();
+    bool granted;
+    if constexpr (grant_option)
+        granted = acs->hasGrantOption(flags, args...);
+    else
+        granted = acs->isGranted(flags, args...);
+
+    if (!granted)
+    {
+        if (grant_option && acs->isGranted(flags, args...))
+        {
+            return access_denied(
+                "Not enough privileges. "
+                "The required privileges have been granted, but without grant option. "
+                "To execute this query it's necessary to have grant "
+                    + AccessRightsElement{flags, args...}.toString() + " WITH GRANT OPTION",
+                ErrorCodes::ACCESS_DENIED);
+        }
+
+        return access_denied(
+            "Not enough privileges. To execute this query it's necessary to have grant "
+                + AccessRightsElement{flags, args...}.toString() + (grant_option ? " WITH GRANT OPTION" : ""),
             ErrorCodes::ACCESS_DENIED);
     }
 
+    struct PrecalculatedFlags
+    {
+        const AccessFlags table_ddl = AccessType::CREATE_DATABASE | AccessType::CREATE_TABLE | AccessType::CREATE_VIEW
+            | AccessType::ALTER_TABLE | AccessType::ALTER_VIEW | AccessType::DROP_DATABASE | AccessType::DROP_TABLE | AccessType::DROP_VIEW
+            | AccessType::TRUNCATE;
+
+        const AccessFlags dictionary_ddl = AccessType::CREATE_DICTIONARY | AccessType::DROP_DICTIONARY;
+        const AccessFlags table_and_dictionary_ddl = table_ddl | dictionary_ddl;
+        const AccessFlags write_table_access = AccessType::INSERT | AccessType::OPTIMIZE;
+        const AccessFlags write_dcl_access = AccessType::ACCESS_MANAGEMENT - AccessType::SHOW_ACCESS;
+
+        const AccessFlags not_readonly_flags = write_table_access | table_and_dictionary_ddl | write_dcl_access | AccessType::SYSTEM | AccessType::KILL_QUERY;
+        const AccessFlags not_readonly_1_flags = AccessType::CREATE_TEMPORARY_TABLE;
+
+        const AccessFlags ddl_flags = table_ddl | dictionary_ddl;
+        const AccessFlags introspection_flags = AccessType::INTROSPECTION;
+    };
+    static const PrecalculatedFlags precalc;
+
     if (params.readonly)
     {
-        if (!access_without_readonly)
-        {
-            Params changed_params = params;
-            changed_params.readonly = 0;
-            access_without_readonly = std::make_shared<AccessRights>(calculateFinalAccessRights(*access_from_user_and_roles, changed_params));
-        }
-
-        if (access_without_readonly->isGranted(flags, args...))
+        if constexpr (grant_option)
+            return access_denied("Cannot change grants in readonly mode.", ErrorCodes::READONLY);
+        if ((flags & precalc.not_readonly_flags) ||
+            ((params.readonly == 1) && (flags & precalc.not_readonly_1_flags)))
         {
             if (params.interface == ClientInfo::Interface::HTTP && params.http_method == ClientInfo::HTTPMethod::GET)
-                show_error(
+            {
+                return access_denied(
                     "Cannot execute query in readonly mode. "
                     "For queries over HTTP, method GET implies readonly. You should use method POST for modifying queries",
                     ErrorCodes::READONLY);
+            }
             else
-                show_error("Cannot execute query in readonly mode", ErrorCodes::READONLY);
+                return access_denied("Cannot execute query in readonly mode", ErrorCodes::READONLY);
         }
     }
 
-    if (!params.allow_ddl)
+    if (!params.allow_ddl && !grant_option)
     {
-        if (!access_with_allow_ddl)
-        {
-            Params changed_params = params;
-            changed_params.allow_ddl = true;
-            access_with_allow_ddl = std::make_shared<AccessRights>(calculateFinalAccessRights(*access_from_user_and_roles, changed_params));
-        }
-
-        if (access_with_allow_ddl->isGranted(flags, args...))
-        {
-            show_error("Cannot execute query. DDL queries are prohibited for the user", ErrorCodes::QUERY_IS_PROHIBITED);
-        }
+        if (flags & precalc.ddl_flags)
+            return access_denied("Cannot execute query. DDL queries are prohibited for the user", ErrorCodes::QUERY_IS_PROHIBITED);
     }
 
-    if (!params.allow_introspection)
+    if (!params.allow_introspection && !grant_option)
     {
-        if (!access_with_allow_introspection)
-        {
-            Params changed_params = params;
-            changed_params.allow_introspection = true;
-            access_with_allow_introspection = std::make_shared<AccessRights>(calculateFinalAccessRights(*access_from_user_and_roles, changed_params));
-        }
-
-        if (access_with_allow_introspection->isGranted(flags, args...))
-        {
-            show_error("Introspection functions are disabled, because setting 'allow_introspection_functions' is set to 0", ErrorCodes::FUNCTION_NOT_ALLOWED);
-        }
+        if (flags & precalc.introspection_flags)
+            return access_denied("Introspection functions are disabled, because setting 'allow_introspection_functions' is set to 0", ErrorCodes::FUNCTION_NOT_ALLOWED);
     }
 
-    show_error(
-        "Not enough privileges. To execute this query it's necessary to have the grant "
-            + AccessRightsElement{flags, args...}.toString() + (grant_option ? " WITH GRANT OPTION" : ""),
-        ErrorCodes::ACCESS_DENIED);
+    return access_granted();
 }
 
-template <bool grant_option>
-void ContextAccess::checkAccessImpl(const AccessFlags & flags) const
+
+bool ContextAccess::isGranted(const AccessFlags & flags) const { return checkAccessImpl<false, false>(flags); }
+bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database) const { return checkAccessImpl<false, false>(flags, database); }
+bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { return checkAccessImpl<false, false>(flags, database, table); }
+bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { return checkAccessImpl<false, false>(flags, database, table, column); }
+bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { return checkAccessImpl<false, false>(flags, database, table, columns); }
+bool ContextAccess::isGranted(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { return checkAccessImpl<false, false>(flags, database, table, columns); }
+bool ContextAccess::isGranted(const AccessRightsElement & element) const { return checkAccessImpl<false, false>(element); }
+bool ContextAccess::isGranted(const AccessRightsElements & elements) const { return checkAccessImpl<false, false>(elements); }
+
+bool ContextAccess::hasGrantOption(const AccessFlags & flags) const { return checkAccessImpl<false, true>(flags); }
+bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database) const { return checkAccessImpl<false, true>(flags, database); }
+bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { return checkAccessImpl<false, true>(flags, database, table); }
+bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { return checkAccessImpl<false, true>(flags, database, table, column); }
+bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { return checkAccessImpl<false, true>(flags, database, table, columns); }
+bool ContextAccess::hasGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { return checkAccessImpl<false, true>(flags, database, table, columns); }
+bool ContextAccess::hasGrantOption(const AccessRightsElement & element) const { return checkAccessImpl<false, true>(element); }
+bool ContextAccess::hasGrantOption(const AccessRightsElements & elements) const { return checkAccessImpl<false, true>(elements); }
+
+void ContextAccess::checkAccess(const AccessFlags & flags) const { checkAccessImpl<true, false>(flags); }
+void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database) const { checkAccessImpl<true, false>(flags, database); }
+void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { checkAccessImpl<true, false>(flags, database, table); }
+void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { checkAccessImpl<true, false>(flags, database, table, column); }
+void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { checkAccessImpl<true, false>(flags, database, table, columns); }
+void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { checkAccessImpl<true, false>(flags, database, table, columns); }
+void ContextAccess::checkAccess(const AccessRightsElement & element) const { checkAccessImpl<true, false>(element); }
+void ContextAccess::checkAccess(const AccessRightsElements & elements) const { checkAccessImpl<true, false>(elements); }
+
+void ContextAccess::checkGrantOption(const AccessFlags & flags) const { checkAccessImpl<true, true>(flags); }
+void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database) const { checkAccessImpl<true, true>(flags, database); }
+void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { checkAccessImpl<true, true>(flags, database, table); }
+void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { checkAccessImpl<true, true>(flags, database, table, column); }
+void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { checkAccessImpl<true, true>(flags, database, table, columns); }
+void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { checkAccessImpl<true, true>(flags, database, table, columns); }
+void ContextAccess::checkGrantOption(const AccessRightsElement & element) const { checkAccessImpl<true, true>(element); }
+void ContextAccess::checkGrantOption(const AccessRightsElements & elements) const { checkAccessImpl<true, true>(elements); }
+
+
+template <bool throw_if_denied>
+bool ContextAccess::checkAdminOptionImpl(const UUID & role_id) const
 {
-    checkAccessImpl2<grant_option>(flags);
+    return checkAdminOptionImpl2<throw_if_denied>(to_array(role_id), [this](const UUID & id, size_t) { return manager->tryReadName(id); });
 }
 
-template <bool grant_option, typename... Args>
-void ContextAccess::checkAccessImpl(const AccessFlags & flags, const std::string_view & database, const Args &... args) const
+template <bool throw_if_denied>
+bool ContextAccess::checkAdminOptionImpl(const UUID & role_id, const String & role_name) const
 {
-    checkAccessImpl2<grant_option>(flags, database.empty() ? params.current_database : database, args...);
+    return checkAdminOptionImpl2<throw_if_denied>(to_array(role_id), [&role_name](const UUID &, size_t) { return std::optional<String>{role_name}; });
 }
 
-template <bool grant_option>
-void ContextAccess::checkAccessImpl(const AccessRightsElement & element) const
+template <bool throw_if_denied>
+bool ContextAccess::checkAdminOptionImpl(const UUID & role_id, const std::unordered_map<UUID, String> & names_of_roles) const
 {
-    if (element.any_database)
-        checkAccessImpl<grant_option>(element.access_flags);
-    else if (element.any_table)
-        checkAccessImpl<grant_option>(element.access_flags, element.database);
-    else if (element.any_column)
-        checkAccessImpl<grant_option>(element.access_flags, element.database, element.table);
-    else
-        checkAccessImpl<grant_option>(element.access_flags, element.database, element.table, element.columns);
+    return checkAdminOptionImpl2<throw_if_denied>(to_array(role_id), [&names_of_roles](const UUID & id, size_t) { auto it = names_of_roles.find(id); return (it != names_of_roles.end()) ? it->second : std::optional<String>{}; });
 }
 
-template <bool grant_option>
-void ContextAccess::checkAccessImpl(const AccessRightsElements & elements) const
+template <bool throw_if_denied>
+bool ContextAccess::checkAdminOptionImpl(const std::vector<UUID> & role_ids) const
 {
-    for (const auto & element : elements)
-        checkAccessImpl<grant_option>(element);
+    return checkAdminOptionImpl2<throw_if_denied>(role_ids, [this](const UUID & id, size_t) { return manager->tryReadName(id); });
 }
 
-void ContextAccess::checkAccess(const AccessFlags & flags) const { checkAccessImpl<false>(flags); }
-void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database) const { checkAccessImpl<false>(flags, database); }
-void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { checkAccessImpl<false>(flags, database, table); }
-void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { checkAccessImpl<false>(flags, database, table, column); }
-void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { checkAccessImpl<false>(flags, database, table, columns); }
-void ContextAccess::checkAccess(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { checkAccessImpl<false>(flags, database, table, columns); }
-void ContextAccess::checkAccess(const AccessRightsElement & element) const { checkAccessImpl<false>(element); }
-void ContextAccess::checkAccess(const AccessRightsElements & elements) const { checkAccessImpl<false>(elements); }
-
-void ContextAccess::checkGrantOption(const AccessFlags & flags) const { checkAccessImpl<true>(flags); }
-void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database) const { checkAccessImpl<true>(flags, database); }
-void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table) const { checkAccessImpl<true>(flags, database, table); }
-void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::string_view & column) const { checkAccessImpl<true>(flags, database, table, column); }
-void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const std::vector<std::string_view> & columns) const { checkAccessImpl<true>(flags, database, table, columns); }
-void ContextAccess::checkGrantOption(const AccessFlags & flags, const std::string_view & database, const std::string_view & table, const Strings & columns) const { checkAccessImpl<true>(flags, database, table, columns); }
-void ContextAccess::checkGrantOption(const AccessRightsElement & element) const { checkAccessImpl<true>(element); }
-void ContextAccess::checkGrantOption(const AccessRightsElements & elements) const { checkAccessImpl<true>(elements); }
-
-
-template <typename Container, typename GetNameFunction>
-bool ContextAccess::checkAdminOptionImpl(bool throw_on_error, const Container & role_ids, const GetNameFunction & get_name_function) const
+template <bool throw_if_denied>
+bool ContextAccess::checkAdminOptionImpl(const std::vector<UUID> & role_ids, const Strings & names_of_roles) const
 {
+    return checkAdminOptionImpl2<throw_if_denied>(role_ids, [&names_of_roles](const UUID &, size_t i) { return std::optional<String>{names_of_roles[i]}; });
+}
+
+template <bool throw_if_denied>
+bool ContextAccess::checkAdminOptionImpl(const std::vector<UUID> & role_ids, const std::unordered_map<UUID, String> & names_of_roles) const
+{
+    return checkAdminOptionImpl2<throw_if_denied>(role_ids, [&names_of_roles](const UUID & id, size_t) { auto it = names_of_roles.find(id); return (it != names_of_roles.end()) ? it->second : std::optional<String>{}; });
+}
+
+template <bool throw_if_denied, typename Container, typename GetNameFunction>
+bool ContextAccess::checkAdminOptionImpl2(const Container & role_ids, const GetNameFunction & get_name_function) const
+{
+    if (!std::size(role_ids) || is_full_access)
+        return true;
+
+    auto show_error = [this](const String & msg, int error_code)
+    {
+        UNUSED(this);
+        if constexpr (throw_if_denied)
+            throw Exception(getUserName() + ": " + msg, error_code);
+    };
+
+    if (!getUser())
+    {
+        show_error("User has been dropped", ErrorCodes::UNKNOWN_USER);
+        return false;
+    }
+
     if (isGranted(AccessType::ROLE_ADMIN))
         return true;
 
     auto info = getRolesInfo();
-    if (!info)
-    {
-        if (!user)
-        {
-            if (throw_on_error)
-                throw Exception(user_name + ": User has been dropped", ErrorCodes::UNKNOWN_USER);
-            else
-                return false;
-        }
-        return true;
-    }
-
     size_t i = 0;
     for (auto it = std::begin(role_ids); it != std::end(role_ids); ++it, ++i)
     {
         const UUID & role_id = *it;
-        if (info->enabled_roles_with_admin_option.count(role_id))
+        if (info && info->enabled_roles_with_admin_option.count(role_id))
             continue;
 
-        auto role_name = get_name_function(role_id, i);
-        if (!role_name)
-            role_name = "ID {" + toString(role_id) + "}";
-        String msg = "To execute this query it's necessary to have the role " + backQuoteIfNeed(*role_name) + " granted with ADMIN option";
-        if (info->enabled_roles.count(role_id))
-            msg = "Role " + backQuote(*role_name) + " is granted, but without ADMIN option. " + msg;
-        if (throw_on_error)
-            throw Exception(getUserName() + ": Not enough privileges. " + msg, ErrorCodes::ACCESS_DENIED);
-        else
-            return false;
+        if (throw_if_denied)
+        {
+            auto role_name = get_name_function(role_id, i);
+            if (!role_name)
+                role_name = "ID {" + toString(role_id) + "}";
+
+            if (info && info->enabled_roles.count(role_id))
+                show_error("Not enough privileges. "
+                           "Role " + backQuote(*role_name) + " is granted, but without ADMIN option. "
+                           "To execute this query it's necessary to have the role " + backQuoteIfNeed(*role_name) + " granted with ADMIN option.",
+                           ErrorCodes::ACCESS_DENIED);
+            else
+                show_error("Not enough privileges. "
+                           "To execute this query it's necessary to have the role " + backQuoteIfNeed(*role_name) + " granted with ADMIN option.",
+                           ErrorCodes::ACCESS_DENIED);
+        }
+
+        return false;
     }
 
     return true;
 }
 
-bool ContextAccess::hasAdminOption(const UUID & role_id) const
-{
-    return checkAdminOptionImpl(false, to_array(role_id), [this](const UUID & id, size_t) { return manager->tryReadName(id); });
-}
+bool ContextAccess::hasAdminOption(const UUID & role_id) const { return checkAdminOptionImpl<false>(role_id); }
+bool ContextAccess::hasAdminOption(const UUID & role_id, const String & role_name) const { return checkAdminOptionImpl<false>(role_id, role_name); }
+bool ContextAccess::hasAdminOption(const UUID & role_id, const std::unordered_map<UUID, String> & names_of_roles) const { return checkAdminOptionImpl<false>(role_id, names_of_roles); }
+bool ContextAccess::hasAdminOption(const std::vector<UUID> & role_ids) const { return checkAdminOptionImpl<false>(role_ids); }
+bool ContextAccess::hasAdminOption(const std::vector<UUID> & role_ids, const Strings & names_of_roles) const { return checkAdminOptionImpl<false>(role_ids, names_of_roles); }
+bool ContextAccess::hasAdminOption(const std::vector<UUID> & role_ids, const std::unordered_map<UUID, String> & names_of_roles) const { return checkAdminOptionImpl<false>(role_ids, names_of_roles); }
 
-bool ContextAccess::hasAdminOption(const UUID & role_id, const String & role_name) const
-{
-    return checkAdminOptionImpl(false, to_array(role_id), [&role_name](const UUID &, size_t) { return std::optional<String>{role_name}; });
-}
-
-bool ContextAccess::hasAdminOption(const UUID & role_id, const std::unordered_map<UUID, String> & names_of_roles) const
-{
-    return checkAdminOptionImpl(false, to_array(role_id), [&names_of_roles](const UUID & id, size_t) { auto it = names_of_roles.find(id); return (it != names_of_roles.end()) ? it->second : std::optional<String>{}; });
-}
-
-bool ContextAccess::hasAdminOption(const std::vector<UUID> & role_ids) const
-{
-    return checkAdminOptionImpl(false, role_ids, [this](const UUID & id, size_t) { return manager->tryReadName(id); });
-}
-
-bool ContextAccess::hasAdminOption(const std::vector<UUID> & role_ids, const Strings & names_of_roles) const
-{
-    return checkAdminOptionImpl(false, role_ids, [&names_of_roles](const UUID &, size_t i) { return std::optional<String>{names_of_roles[i]}; });
-}
-
-bool ContextAccess::hasAdminOption(const std::vector<UUID> & role_ids, const std::unordered_map<UUID, String> & names_of_roles) const
-{
-    return checkAdminOptionImpl(false, role_ids, [&names_of_roles](const UUID & id, size_t) { auto it = names_of_roles.find(id); return (it != names_of_roles.end()) ? it->second : std::optional<String>{}; });
-}
-
-void ContextAccess::checkAdminOption(const UUID & role_id) const
-{
-    checkAdminOptionImpl(true, to_array(role_id), [this](const UUID & id, size_t) { return manager->tryReadName(id); });
-}
-
-void ContextAccess::checkAdminOption(const UUID & role_id, const String & role_name) const
-{
-    checkAdminOptionImpl(true, to_array(role_id), [&role_name](const UUID &, size_t) { return std::optional<String>{role_name}; });
-}
-
-void ContextAccess::checkAdminOption(const UUID & role_id, const std::unordered_map<UUID, String> & names_of_roles) const
-{
-    checkAdminOptionImpl(true, to_array(role_id), [&names_of_roles](const UUID & id, size_t) { auto it = names_of_roles.find(id); return (it != names_of_roles.end()) ? it->second : std::optional<String>{}; });
-}
-
-void ContextAccess::checkAdminOption(const std::vector<UUID> & role_ids) const
-{
-    checkAdminOptionImpl(true, role_ids, [this](const UUID & id, size_t) { return manager->tryReadName(id); });
-}
-
-void ContextAccess::checkAdminOption(const std::vector<UUID> & role_ids, const Strings & names_of_roles) const
-{
-    checkAdminOptionImpl(true, role_ids, [&names_of_roles](const UUID &, size_t i) { return std::optional<String>{names_of_roles[i]}; });
-}
-
-void ContextAccess::checkAdminOption(const std::vector<UUID> & role_ids, const std::unordered_map<UUID, String> & names_of_roles) const
-{
-    checkAdminOptionImpl(true, role_ids, [&names_of_roles](const UUID & id, size_t) { auto it = names_of_roles.find(id); return (it != names_of_roles.end()) ? it->second : std::optional<String>{}; });
-}
+void ContextAccess::checkAdminOption(const UUID & role_id) const { checkAdminOptionImpl<true>(role_id); }
+void ContextAccess::checkAdminOption(const UUID & role_id, const String & role_name) const { checkAdminOptionImpl<true>(role_id, role_name); }
+void ContextAccess::checkAdminOption(const UUID & role_id, const std::unordered_map<UUID, String> & names_of_roles) const { checkAdminOptionImpl<true>(role_id, names_of_roles); }
+void ContextAccess::checkAdminOption(const std::vector<UUID> & role_ids) const { checkAdminOptionImpl<true>(role_ids); }
+void ContextAccess::checkAdminOption(const std::vector<UUID> & role_ids, const Strings & names_of_roles) const { checkAdminOptionImpl<true>(role_ids, names_of_roles); }
+void ContextAccess::checkAdminOption(const std::vector<UUID> & role_ids, const std::unordered_map<UUID, String> & names_of_roles) const { checkAdminOptionImpl<true>(role_ids, names_of_roles); }
 
 }
diff --git a/src/Access/ContextAccess.h b/src/Access/ContextAccess.h
index 319c8edb076..43e9f60a4c6 100644
--- a/src/Access/ContextAccess.h
+++ b/src/Access/ContextAccess.h
@@ -96,7 +96,8 @@ public:
     std::shared_ptr<const SettingsConstraints> getSettingsConstraints() const;
 
     /// Returns the current access rights.
-    std::shared_ptr<const AccessRights> getAccess() const;
+    std::shared_ptr<const AccessRights> getAccessRights() const;
+    std::shared_ptr<const AccessRights> getAccessRightsWithImplicit() const;
 
     /// Checks if a specified access is granted.
     bool isGranted(const AccessFlags & flags) const;
@@ -166,41 +167,45 @@ private:
     void setSettingsAndConstraints() const;
     void calculateAccessRights() const;
 
-    template <bool grant_option>
-    bool isGrantedImpl(const AccessFlags & flags) const;
+    template <bool throw_if_denied, bool grant_option>
+    bool checkAccessImpl(const AccessFlags & flags) const;
 
-    template <bool grant_option, typename... Args>
-    bool isGrantedImpl(const AccessFlags & flags, const std::string_view & database, const Args &... args) const;
+    template <bool throw_if_denied, bool grant_option, typename... Args>
+    bool checkAccessImpl(const AccessFlags & flags, const std::string_view & database, const Args &... args) const;
 
-    template <bool grant_option>
-    bool isGrantedImpl(const AccessRightsElement & element) const;
+    template <bool throw_if_denied, bool grant_option>
+    bool checkAccessImpl(const AccessRightsElement & element) const;
 
-    template <bool grant_option>
-    bool isGrantedImpl(const AccessRightsElements & elements) const;
+    template <bool throw_if_denied, bool grant_option>
+    bool checkAccessImpl(const AccessRightsElements & elements) const;
 
-    template <bool grant_option, typename... Args>
-    bool isGrantedImpl2(const AccessFlags & flags, const Args &... args) const;
+    template <bool throw_if_denied, bool grant_option, typename... Args>
+    bool checkAccessImpl2(const AccessFlags & flags, const Args &... args) const;
 
-    template <bool grant_option>
-    void checkAccessImpl(const AccessFlags & flags) const;
+    template <bool throw_if_denied>
+    bool checkAdminOptionImpl(const UUID & role_id) const;
 
-    template <bool grant_option, typename... Args>
-    void checkAccessImpl(const AccessFlags & flags, const std::string_view & database, const Args &... args) const;
+    template <bool throw_if_denied>
+    bool checkAdminOptionImpl(const UUID & role_id, const String & role_name) const;
 
-    template <bool grant_option>
-    void checkAccessImpl(const AccessRightsElement & element) const;
+    template <bool throw_if_denied>
+    bool checkAdminOptionImpl(const UUID & role_id, const std::unordered_map<UUID, String> & names_of_roles) const;
 
-    template <bool grant_option>
-    void checkAccessImpl(const AccessRightsElements & elements) const;
+    template <bool throw_if_denied>
+    bool checkAdminOptionImpl(const std::vector<UUID> & role_ids) const;
 
-    template <bool grant_option, typename... Args>
-    void checkAccessImpl2(const AccessFlags & flags, const Args &... args) const;
+    template <bool throw_if_denied>
+    bool checkAdminOptionImpl(const std::vector<UUID> & role_ids, const Strings & names_of_roles) const;
 
-    template <typename Container, typename GetNameFunction>
-    bool checkAdminOptionImpl(bool throw_on_error, const Container & role_ids, const GetNameFunction & get_name_function) const;
+    template <bool throw_if_denied>
+    bool checkAdminOptionImpl(const std::vector<UUID> & role_ids, const std::unordered_map<UUID, String> & names_of_roles) const;
+
+    template <bool throw_if_denied, typename Container, typename GetNameFunction>
+    bool checkAdminOptionImpl2(const Container & role_ids, const GetNameFunction & get_name_function) const;
 
     const AccessControlManager * manager = nullptr;
     const Params params;
+    bool is_full_access = false;
     mutable Poco::Logger * trace_log = nullptr;
     mutable UserPtr user;
     mutable String user_name;
@@ -209,13 +214,10 @@ private:
     mutable ext::scope_guard subscription_for_roles_changes;
     mutable std::shared_ptr<const EnabledRolesInfo> roles_info;
     mutable std::shared_ptr<const AccessRights> access;
+    mutable std::shared_ptr<const AccessRights> access_with_implicit;
     mutable std::shared_ptr<const EnabledRowPolicies> enabled_row_policies;
     mutable std::shared_ptr<const EnabledQuota> enabled_quota;
     mutable std::shared_ptr<const EnabledSettings> enabled_settings;
-    mutable std::shared_ptr<const AccessRights> access_without_readonly;
-    mutable std::shared_ptr<const AccessRights> access_with_allow_ddl;
-    mutable std::shared_ptr<const AccessRights> access_with_allow_introspection;
-    mutable std::shared_ptr<const AccessRights> access_from_user_and_roles;
     mutable std::mutex mutex;
 };
 
diff --git a/src/Access/SettingsConstraints.cpp b/src/Access/SettingsConstraints.cpp
index c9a6b6f6b7a..7dce4d96b33 100644
--- a/src/Access/SettingsConstraints.cpp
+++ b/src/Access/SettingsConstraints.cpp
@@ -156,6 +156,25 @@ bool SettingsConstraints::checkImpl(const Settings & current_settings, SettingCh
 {
     const String & setting_name = change.name;
 
+    if (setting_name == "profile")
+    {
+        /// TODO Check profile settings in Context::setProfile(...), not here. It will be backward incompatible.
+        const String & profile_name = change.value.safeGet<String>();
+        const auto & profile_settings_changes = manager->getProfileSettings(profile_name);
+        try
+        {
+            /// NOTE We cannot use CLAMP_ON_VIOLATION here, because we cannot modify elements of profile_settings_changes
+            for (auto change_copy : *profile_settings_changes)
+                checkImpl(current_settings, change_copy, THROW_ON_VIOLATION);
+        }
+        catch (Exception & e)
+        {
+            e.addMessage(", while trying to set settings profile {}", profile_name);
+            throw;
+        }
+        return true;
+    }
+
     bool cannot_cast;
     auto cast_value = [&](const Field & x) -> Field
     {
diff --git a/src/AggregateFunctions/AggregateFunctionAvg.h b/src/AggregateFunctions/AggregateFunctionAvg.h
index fca9df9dd98..d07ff5db2f2 100644
--- a/src/AggregateFunctions/AggregateFunctionAvg.h
+++ b/src/AggregateFunctions/AggregateFunctionAvg.h
@@ -33,7 +33,7 @@ struct AvgFraction
 
     /// Allow division by zero as sometimes we need to return NaN.
     /// Invoked only is either Numerator or Denominator are Decimal.
-    Float64 NO_SANITIZE_UNDEFINED divideIfAnyDecimal(UInt32 num_scale, UInt32 denom_scale) const
+    Float64 NO_SANITIZE_UNDEFINED divideIfAnyDecimal(UInt32 num_scale, UInt32 denom_scale [[maybe_unused]]) const
     {
         if constexpr (IsDecimalNumber<Numerator> && IsDecimalNumber<Denominator>)
         {
diff --git a/src/AggregateFunctions/AggregateFunctionFactory.cpp b/src/AggregateFunctions/AggregateFunctionFactory.cpp
index 1c66650be45..d1dcb8cf455 100644
--- a/src/AggregateFunctions/AggregateFunctionFactory.cpp
+++ b/src/AggregateFunctions/AggregateFunctionFactory.cpp
@@ -18,6 +18,7 @@
 #include <Poco/String.h>
 #include "registerAggregateFunctions.h"
 
+#include <Functions/FunctionFactory.h>
 
 namespace DB
 {
@@ -135,12 +136,17 @@ AggregateFunctionPtr AggregateFunctionFactory::getImpl(
         return combinator->transformAggregateFunction(nested_function, out_properties, argument_types, parameters);
     }
 
+
+    String extra_info;
+    if (FunctionFactory::instance().hasNameOrAlias(name))
+        extra_info = ". There is an ordinary function with the same name, but aggregate function is expected here";
+
     auto hints = this->getHints(name);
     if (!hints.empty())
-        throw Exception(fmt::format("Unknown aggregate function {}. Maybe you meant: {}", name, toString(hints)),
-            ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION);
+        throw Exception(ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION,
+                        "Unknown aggregate function {}{}. Maybe you meant: {}", name, extra_info, toString(hints));
     else
-        throw Exception(fmt::format("Unknown aggregate function {}", name), ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION);
+        throw Exception(ErrorCodes::UNKNOWN_AGGREGATE_FUNCTION, "Unknown aggregate function {}{}", name, extra_info);
 }
 
 
diff --git a/src/AggregateFunctions/AggregateFunctionSimpleState.cpp b/src/AggregateFunctions/AggregateFunctionSimpleState.cpp
new file mode 100644
index 00000000000..8a5a71d6806
--- /dev/null
+++ b/src/AggregateFunctions/AggregateFunctionSimpleState.cpp
@@ -0,0 +1,32 @@
+#include <AggregateFunctions/AggregateFunctionCombinatorFactory.h>
+#include <AggregateFunctions/AggregateFunctionSimpleState.h>
+
+namespace DB
+{
+namespace
+{
+    class AggregateFunctionCombinatorSimpleState final : public IAggregateFunctionCombinator
+    {
+    public:
+        String getName() const override { return "SimpleState"; }
+
+        DataTypes transformArguments(const DataTypes & arguments) const override { return arguments; }
+
+        AggregateFunctionPtr transformAggregateFunction(
+            const AggregateFunctionPtr & nested_function,
+            const AggregateFunctionProperties &,
+            const DataTypes & arguments,
+            const Array & params) const override
+        {
+            return std::make_shared<AggregateFunctionSimpleState>(nested_function, arguments, params);
+        }
+    };
+
+}
+
+void registerAggregateFunctionCombinatorSimpleState(AggregateFunctionCombinatorFactory & factory)
+{
+    factory.registerCombinator(std::make_shared<AggregateFunctionCombinatorSimpleState>());
+}
+
+}
diff --git a/src/AggregateFunctions/AggregateFunctionSimpleState.h b/src/AggregateFunctions/AggregateFunctionSimpleState.h
new file mode 100644
index 00000000000..eefe00e77c0
--- /dev/null
+++ b/src/AggregateFunctions/AggregateFunctionSimpleState.h
@@ -0,0 +1,77 @@
+#pragma once
+
+#include <AggregateFunctions/IAggregateFunction.h>
+#include <DataTypes/DataTypeCustomSimpleAggregateFunction.h>
+#include <DataTypes/DataTypeFactory.h>
+
+namespace DB
+{
+
+/** Not an aggregate function, but an adapter of aggregate functions.
+  * Aggregate functions with the `SimpleState` suffix is almost identical to the corresponding ones,
+  * except the return type becomes DataTypeCustomSimpleAggregateFunction.
+  */
+class AggregateFunctionSimpleState final : public IAggregateFunctionHelper<AggregateFunctionSimpleState>
+{
+private:
+    AggregateFunctionPtr nested_func;
+    DataTypes arguments;
+    Array params;
+
+public:
+    AggregateFunctionSimpleState(AggregateFunctionPtr nested_, const DataTypes & arguments_, const Array & params_)
+        : IAggregateFunctionHelper<AggregateFunctionSimpleState>(arguments_, params_)
+        , nested_func(nested_)
+        , arguments(arguments_)
+        , params(params_)
+    {
+    }
+
+    String getName() const override { return nested_func->getName() + "SimpleState"; }
+
+    DataTypePtr getReturnType() const override
+    {
+        DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(nested_func);
+        // Need to make a clone because it'll be customized.
+        auto storage_type = DataTypeFactory::instance().get(nested_func->getReturnType()->getName());
+        DataTypeCustomNamePtr custom_name
+            = std::make_unique<DataTypeCustomSimpleAggregateFunction>(nested_func, DataTypes{nested_func->getReturnType()}, params);
+        storage_type->setCustomization(std::make_unique<DataTypeCustomDesc>(std::move(custom_name), nullptr));
+        return storage_type;
+    }
+
+    void create(AggregateDataPtr place) const override { nested_func->create(place); }
+
+    void destroy(AggregateDataPtr place) const noexcept override { nested_func->destroy(place); }
+
+    bool hasTrivialDestructor() const override { return nested_func->hasTrivialDestructor(); }
+
+    size_t sizeOfData() const override { return nested_func->sizeOfData(); }
+
+    size_t alignOfData() const override { return nested_func->alignOfData(); }
+
+    void add(AggregateDataPtr place, const IColumn ** columns, size_t row_num, Arena * arena) const override
+    {
+        nested_func->add(place, columns, row_num, arena);
+    }
+
+    void merge(AggregateDataPtr place, ConstAggregateDataPtr rhs, Arena * arena) const override { nested_func->merge(place, rhs, arena); }
+
+    void serialize(ConstAggregateDataPtr place, WriteBuffer & buf) const override { nested_func->serialize(place, buf); }
+
+    void deserialize(AggregateDataPtr place, ReadBuffer & buf, Arena * arena) const override
+    {
+        nested_func->deserialize(place, buf, arena);
+    }
+
+    void insertResultInto(AggregateDataPtr place, IColumn & to, Arena * arena) const override
+    {
+        nested_func->insertResultInto(place, to, arena);
+    }
+
+    bool allocatesMemoryInArena() const override { return nested_func->allocatesMemoryInArena(); }
+
+    AggregateFunctionPtr getNestedFunction() const { return nested_func; }
+};
+
+}
diff --git a/src/AggregateFunctions/registerAggregateFunctions.cpp b/src/AggregateFunctions/registerAggregateFunctions.cpp
index 927c7e6d6b1..228ca2f30ab 100644
--- a/src/AggregateFunctions/registerAggregateFunctions.cpp
+++ b/src/AggregateFunctions/registerAggregateFunctions.cpp
@@ -47,6 +47,7 @@ class AggregateFunctionCombinatorFactory;
 void registerAggregateFunctionCombinatorIf(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorArray(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorForEach(AggregateFunctionCombinatorFactory &);
+void registerAggregateFunctionCombinatorSimpleState(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorState(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorMerge(AggregateFunctionCombinatorFactory &);
 void registerAggregateFunctionCombinatorNull(AggregateFunctionCombinatorFactory &);
@@ -104,6 +105,7 @@ void registerAggregateFunctions()
         registerAggregateFunctionCombinatorIf(factory);
         registerAggregateFunctionCombinatorArray(factory);
         registerAggregateFunctionCombinatorForEach(factory);
+        registerAggregateFunctionCombinatorSimpleState(factory);
         registerAggregateFunctionCombinatorState(factory);
         registerAggregateFunctionCombinatorMerge(factory);
         registerAggregateFunctionCombinatorNull(factory);
diff --git a/src/AggregateFunctions/ya.make b/src/AggregateFunctions/ya.make
index f05b8ef55d2..8faf738fd98 100644
--- a/src/AggregateFunctions/ya.make
+++ b/src/AggregateFunctions/ya.make
@@ -41,6 +41,7 @@ SRCS(
     AggregateFunctionRetention.cpp
     AggregateFunctionSequenceMatch.cpp
     AggregateFunctionSimpleLinearRegression.cpp
+    AggregateFunctionSimpleState.cpp
     AggregateFunctionState.cpp
     AggregateFunctionStatistics.cpp
     AggregateFunctionStatisticsSimple.cpp
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 6021065f937..46956181974 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -88,6 +88,10 @@ if (USE_AWS_S3)
     add_headers_and_sources(dbms Disks/S3)
 endif()
 
+if (USE_HDFS)
+    add_headers_and_sources(dbms Storages/HDFS)
+endif()
+
 list (APPEND clickhouse_common_io_sources ${CONFIG_BUILD})
 list (APPEND clickhouse_common_io_headers ${CONFIG_VERSION} ${CONFIG_COMMON})
 
@@ -389,8 +393,8 @@ if (USE_GRPC)
 endif()
 
 if (USE_HDFS)
-    target_link_libraries (clickhouse_common_io PUBLIC ${HDFS3_LIBRARY})
-    target_include_directories (clickhouse_common_io SYSTEM BEFORE PUBLIC ${HDFS3_INCLUDE_DIR})
+    dbms_target_link_libraries(PRIVATE ${HDFS3_LIBRARY})
+    dbms_target_include_directories (SYSTEM BEFORE PUBLIC ${HDFS3_INCLUDE_DIR})
 endif()
 
 if (USE_AWS_S3)
diff --git a/src/Client/Connection.cpp b/src/Client/Connection.cpp
index f7119195e97..8f4a64766cd 100644
--- a/src/Client/Connection.cpp
+++ b/src/Client/Connection.cpp
@@ -1,5 +1,6 @@
 #include <Poco/Net/NetException.h>
 #include <Core/Defines.h>
+#include <Core/Settings.h>
 #include <Compression/CompressedReadBuffer.h>
 #include <Compression/CompressedWriteBuffer.h>
 #include <IO/ReadBufferFromPocoSocket.h>
diff --git a/src/Client/Connection.h b/src/Client/Connection.h
index f4c25001f3e..265a9913b8e 100644
--- a/src/Client/Connection.h
+++ b/src/Client/Connection.h
@@ -5,7 +5,9 @@
 #include <Poco/Net/StreamSocket.h>
 
 #include <Common/Throttler.h>
-
+#if !defined(ARCADIA_BUILD)
+#   include <Common/config.h>
+#endif
 #include <Core/Block.h>
 #include <Core/Defines.h>
 #include <IO/Progress.h>
@@ -17,7 +19,6 @@
 
 #include <IO/ConnectionTimeouts.h>
 
-#include <Core/Settings.h>
 #include <Interpreters/TablesStatus.h>
 
 #include <Compression/ICompressionCodec.h>
@@ -31,6 +32,7 @@ namespace DB
 
 class ClientInfo;
 class Pipe;
+struct Settings;
 
 /// Struct which represents data we are going to send for external table.
 struct ExternalTableData
diff --git a/src/Client/ConnectionPool.h b/src/Client/ConnectionPool.h
index 736075a4cc1..2389cc6755d 100644
--- a/src/Client/ConnectionPool.h
+++ b/src/Client/ConnectionPool.h
@@ -1,9 +1,9 @@
 #pragma once
 
 #include <Common/PoolBase.h>
-
 #include <Client/Connection.h>
 #include <IO/ConnectionTimeouts.h>
+#include <Core/Settings.h>
 
 namespace DB
 {
diff --git a/src/Client/ConnectionPoolWithFailover.cpp b/src/Client/ConnectionPoolWithFailover.cpp
index 68f4bcd1b76..1ca61dc8059 100644
--- a/src/Client/ConnectionPoolWithFailover.cpp
+++ b/src/Client/ConnectionPoolWithFailover.cpp
@@ -4,6 +4,7 @@
 #include <Poco/Net/DNS.h>
 
 #include <Common/BitHelpers.h>
+#include <Common/quoteString.h>
 #include <common/getFQDNOrHostName.h>
 #include <Common/isLocalAddress.h>
 #include <Common/ProfileEvents.h>
diff --git a/src/Columns/ColumnMap.cpp b/src/Columns/ColumnMap.cpp
new file mode 100644
index 00000000000..c210da2aa11
--- /dev/null
+++ b/src/Columns/ColumnMap.cpp
@@ -0,0 +1,241 @@
+#include <Columns/ColumnMap.h>
+#include <Columns/IColumnImpl.h>
+#include <DataStreams/ColumnGathererStream.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/Operators.h>
+#include <ext/map.h>
+#include <ext/range.h>
+#include <Common/typeid_cast.h>
+#include <Common/assert_cast.h>
+#include <Common/WeakHash.h>
+#include <Core/Field.h>
+
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_COLUMN;
+    extern const int NOT_IMPLEMENTED;
+    extern const int LOGICAL_ERROR;
+}
+
+
+std::string ColumnMap::getName() const
+{
+    WriteBufferFromOwnString res;
+    const auto & nested_tuple = getNestedData();
+    res << "Map(" << nested_tuple.getColumn(0).getName()
+        << ", " << nested_tuple.getColumn(1).getName() << ")";
+
+    return res.str();
+}
+
+ColumnMap::ColumnMap(MutableColumnPtr && nested_)
+    : nested(std::move(nested_))
+{
+    const auto * column_array = typeid_cast<const ColumnArray *>(nested.get());
+    if (!column_array)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ColumnMap can be created only from array of tuples");
+
+    const auto * column_tuple = typeid_cast<const ColumnTuple *>(column_array->getDataPtr().get());
+    if (!column_tuple)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ColumnMap can be created only from array of tuples");
+
+    if (column_tuple->getColumns().size() != 2)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ColumnMap should contain only 2 subcolumns: keys and values");
+
+    for (const auto & column : column_tuple->getColumns())
+        if (isColumnConst(*column))
+            throw Exception(ErrorCodes::ILLEGAL_COLUMN, "ColumnMap cannot have ColumnConst as its element");
+}
+
+MutableColumnPtr ColumnMap::cloneEmpty() const
+{
+    return ColumnMap::create(nested->cloneEmpty());
+}
+
+MutableColumnPtr ColumnMap::cloneResized(size_t new_size) const
+{
+    return ColumnMap::create(nested->cloneResized(new_size));
+}
+
+Field ColumnMap::operator[](size_t n) const
+{
+    auto array = DB::get<Array>((*nested)[n]);
+    return Map(std::make_move_iterator(array.begin()), std::make_move_iterator(array.end()));
+}
+
+void ColumnMap::get(size_t n, Field & res) const
+{
+    const auto & offsets = getNestedColumn().getOffsets();
+    size_t offset = offsets[n - 1];
+    size_t size = offsets[n] - offsets[n - 1];
+
+    res = Map(size);
+    auto & map = DB::get<Map &>(res);
+
+    for (size_t i = 0; i < size; ++i)
+        getNestedData().get(offset + i, map[i]);
+}
+
+StringRef ColumnMap::getDataAt(size_t) const
+{
+    throw Exception("Method getDataAt is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
+}
+
+void ColumnMap::insertData(const char *, size_t)
+{
+    throw Exception("Method insertData is not supported for " + getName(), ErrorCodes::NOT_IMPLEMENTED);
+}
+
+void ColumnMap::insert(const Field & x)
+{
+    const auto & map = DB::get<const Map &>(x);
+    nested->insert(Array(map.begin(), map.end()));
+}
+
+void ColumnMap::insertDefault()
+{
+    nested->insertDefault();
+}
+void ColumnMap::popBack(size_t n)
+{
+    nested->popBack(n);
+}
+
+StringRef ColumnMap::serializeValueIntoArena(size_t n, Arena & arena, char const *& begin) const
+{
+    return nested->serializeValueIntoArena(n, arena, begin);
+}
+
+const char * ColumnMap::deserializeAndInsertFromArena(const char * pos)
+{
+    return nested->deserializeAndInsertFromArena(pos);
+}
+
+void ColumnMap::updateHashWithValue(size_t n, SipHash & hash) const
+{
+    nested->updateHashWithValue(n, hash);
+}
+
+void ColumnMap::updateWeakHash32(WeakHash32 & hash) const
+{
+    nested->updateWeakHash32(hash);
+}
+
+void ColumnMap::updateHashFast(SipHash & hash) const
+{
+    nested->updateHashFast(hash);
+}
+
+void ColumnMap::insertRangeFrom(const IColumn & src, size_t start, size_t length)
+{
+    nested->insertRangeFrom(
+        assert_cast<const ColumnMap &>(src).getNestedColumn(),
+        start, length);
+}
+
+ColumnPtr ColumnMap::filter(const Filter & filt, ssize_t result_size_hint) const
+{
+    auto filtered = nested->filter(filt, result_size_hint);
+    return ColumnMap::create(filtered);
+}
+
+ColumnPtr ColumnMap::permute(const Permutation & perm, size_t limit) const
+{
+    auto permuted = nested->permute(perm, limit);
+    return ColumnMap::create(std::move(permuted));
+}
+
+ColumnPtr ColumnMap::index(const IColumn & indexes, size_t limit) const
+{
+    auto res = nested->index(indexes, limit);
+    return ColumnMap::create(std::move(res));
+}
+
+ColumnPtr ColumnMap::replicate(const Offsets & offsets) const
+{
+    auto replicated = nested->replicate(offsets);
+    return ColumnMap::create(std::move(replicated));
+}
+
+MutableColumns ColumnMap::scatter(ColumnIndex num_columns, const Selector & selector) const
+{
+    auto scattered_columns = nested->scatter(num_columns, selector);
+    MutableColumns res;
+    res.reserve(num_columns);
+    for (auto && scattered : scattered_columns)
+        res.push_back(ColumnMap::create(std::move(scattered)));
+
+    return res;
+}
+
+int ColumnMap::compareAt(size_t n, size_t m, const IColumn & rhs, int nan_direction_hint) const
+{
+    const auto & rhs_map = assert_cast<const ColumnMap &>(rhs);
+    return nested->compareAt(n, m, rhs_map.getNestedColumn(), nan_direction_hint);
+}
+
+void ColumnMap::compareColumn(const IColumn & rhs, size_t rhs_row_num,
+                                PaddedPODArray<UInt64> * row_indexes, PaddedPODArray<Int8> & compare_results,
+                                int direction, int nan_direction_hint) const
+{
+    return doCompareColumn<ColumnMap>(assert_cast<const ColumnMap &>(rhs), rhs_row_num, row_indexes,
+                                        compare_results, direction, nan_direction_hint);
+}
+
+void ColumnMap::getPermutation(bool reverse, size_t limit, int nan_direction_hint, Permutation & res) const
+{
+    nested->getPermutation(reverse, limit, nan_direction_hint, res);
+}
+
+void ColumnMap::updatePermutation(bool reverse, size_t limit, int nan_direction_hint, IColumn::Permutation & res, EqualRanges & equal_range) const
+{
+    nested->updatePermutation(reverse, limit, nan_direction_hint, res, equal_range);
+}
+
+void ColumnMap::gather(ColumnGathererStream & gatherer)
+{
+    gatherer.gather(*this);
+}
+
+void ColumnMap::reserve(size_t n)
+{
+    nested->reserve(n);
+}
+
+size_t ColumnMap::byteSize() const
+{
+    return nested->byteSize();
+}
+
+size_t ColumnMap::allocatedBytes() const
+{
+    return nested->allocatedBytes();
+}
+
+void ColumnMap::protect()
+{
+    nested->protect();
+}
+
+void ColumnMap::getExtremes(Field & min, Field & max) const
+{
+    nested->getExtremes(min, max);
+}
+
+void ColumnMap::forEachSubcolumn(ColumnCallback callback)
+{
+    nested->forEachSubcolumn(callback);
+}
+
+bool ColumnMap::structureEquals(const IColumn & rhs) const
+{
+    if (const auto * rhs_map = typeid_cast<const ColumnMap *>(&rhs))
+        return nested->structureEquals(*rhs_map->nested);
+    return false;
+}
+
+}
diff --git a/src/Columns/ColumnMap.h b/src/Columns/ColumnMap.h
new file mode 100644
index 00000000000..08197c3e1e9
--- /dev/null
+++ b/src/Columns/ColumnMap.h
@@ -0,0 +1,92 @@
+#pragma once
+
+#include <Core/Block.h>
+#include <Columns/ColumnArray.h>
+#include <Columns/ColumnVector.h>
+#include <Columns/ColumnTuple.h>
+
+namespace DB
+{
+
+/** Column, that stores a nested Array(Tuple(key, value)) column.
+ */
+class ColumnMap final : public COWHelper<IColumn, ColumnMap>
+{
+private:
+    friend class COWHelper<IColumn, ColumnMap>;
+
+    WrappedPtr nested;
+
+    explicit ColumnMap(MutableColumnPtr && nested_);
+
+    ColumnMap(const ColumnMap &) = default;
+
+public:
+    /** Create immutable column using immutable arguments. This arguments may be shared with other columns.
+      * Use IColumn::mutate in order to make mutable column and mutate shared nested columns.
+      */
+    using Base = COWHelper<IColumn, ColumnMap>;
+
+    static Ptr create(const ColumnPtr & keys, const ColumnPtr & values, const ColumnPtr & offsets)
+    {
+        auto nested_column = ColumnArray::create(ColumnTuple::create(Columns{keys, values}), offsets);
+        return ColumnMap::create(nested_column);
+    }
+
+    static Ptr create(const ColumnPtr & column) { return ColumnMap::create(column->assumeMutable()); }
+    static Ptr create(ColumnPtr && arg) { return create(arg); }
+
+    template <typename Arg, typename = typename std::enable_if<std::is_rvalue_reference<Arg &&>::value>::type>
+    static MutablePtr create(Arg && arg) { return Base::create(std::forward<Arg>(arg)); }
+
+    std::string getName() const override;
+    const char * getFamilyName() const override { return "Map"; }
+    TypeIndex getDataType() const override { return TypeIndex::Map; }
+
+    MutableColumnPtr cloneEmpty() const override;
+    MutableColumnPtr cloneResized(size_t size) const override;
+
+    size_t size() const override { return nested->size(); }
+
+    Field operator[](size_t n) const override;
+    void get(size_t n, Field & res) const override;
+
+    StringRef getDataAt(size_t n) const override;
+    void insertData(const char * pos, size_t length) override;
+    void insert(const Field & x) override;
+    void insertDefault() override;
+    void popBack(size_t n) override;
+    StringRef serializeValueIntoArena(size_t n, Arena & arena, char const *& begin) const override;
+    const char * deserializeAndInsertFromArena(const char * pos) override;
+    void updateHashWithValue(size_t n, SipHash & hash) const override;
+    void updateWeakHash32(WeakHash32 & hash) const override;
+    void updateHashFast(SipHash & hash) const override;
+    void insertRangeFrom(const IColumn & src, size_t start, size_t length) override;
+    ColumnPtr filter(const Filter & filt, ssize_t result_size_hint) const override;
+    ColumnPtr permute(const Permutation & perm, size_t limit) const override;
+    ColumnPtr index(const IColumn & indexes, size_t limit) const override;
+    ColumnPtr replicate(const Offsets & offsets) const override;
+    MutableColumns scatter(ColumnIndex num_columns, const Selector & selector) const override;
+    void gather(ColumnGathererStream & gatherer_stream) override;
+    int compareAt(size_t n, size_t m, const IColumn & rhs, int nan_direction_hint) const override;
+    void compareColumn(const IColumn & rhs, size_t rhs_row_num,
+                       PaddedPODArray<UInt64> * row_indexes, PaddedPODArray<Int8> & compare_results,
+                       int direction, int nan_direction_hint) const override;
+    void getExtremes(Field & min, Field & max) const override;
+    void getPermutation(bool reverse, size_t limit, int nan_direction_hint, Permutation & res) const override;
+    void updatePermutation(bool reverse, size_t limit, int nan_direction_hint, IColumn::Permutation & res, EqualRanges & equal_range) const override;
+    void reserve(size_t n) override;
+    size_t byteSize() const override;
+    size_t allocatedBytes() const override;
+    void protect() override;
+    void forEachSubcolumn(ColumnCallback callback) override;
+    bool structureEquals(const IColumn & rhs) const override;
+
+    const ColumnArray & getNestedColumn() const { return assert_cast<const ColumnArray &>(*nested); }
+    ColumnArray & getNestedColumn() { return assert_cast<ColumnArray &>(*nested); }
+
+    const ColumnTuple & getNestedData() const { return assert_cast<const ColumnTuple &>(getNestedColumn().getData()); }
+    ColumnTuple & getNestedData() { return assert_cast<ColumnTuple &>(getNestedColumn().getData()); }
+};
+
+}
diff --git a/src/Columns/ya.make b/src/Columns/ya.make
index f1a8cb9f32e..2affaeb0fc6 100644
--- a/src/Columns/ya.make
+++ b/src/Columns/ya.make
@@ -24,6 +24,7 @@ SRCS(
     ColumnFixedString.cpp
     ColumnFunction.cpp
     ColumnLowCardinality.cpp
+    ColumnMap.cpp
     ColumnNullable.cpp
     ColumnString.cpp
     ColumnTuple.cpp
diff --git a/src/Common/Allocator.h b/src/Common/Allocator.h
index 6ac60020379..0964bb9f522 100644
--- a/src/Common/Allocator.h
+++ b/src/Common/Allocator.h
@@ -26,6 +26,7 @@
     #define DISABLE_MREMAP 1
 #endif
 #include <common/mremap.h>
+#include <common/getPageSize.h>
 
 #include <Common/MemoryTracker.h>
 #include <Common/Exception.h>
@@ -59,7 +60,6 @@
   */
 extern const size_t MMAP_THRESHOLD;
 
-static constexpr size_t MMAP_MIN_ALIGNMENT = 4096;
 static constexpr size_t MALLOC_MIN_ALIGNMENT = 8;
 
 namespace DB
@@ -194,10 +194,11 @@ private:
     void * allocNoTrack(size_t size, size_t alignment)
     {
         void * buf;
+        size_t mmap_min_alignment = ::getPageSize();
 
         if (size >= MMAP_THRESHOLD)
         {
-            if (alignment > MMAP_MIN_ALIGNMENT)
+            if (alignment > mmap_min_alignment)
                 throw DB::Exception(fmt::format("Too large alignment {}: more than page size when allocating {}.",
                     ReadableSize(alignment), ReadableSize(size)), DB::ErrorCodes::BAD_ARGUMENTS);
 
diff --git a/src/Common/Arena.h b/src/Common/Arena.h
index 44a9b444ff2..d41dd9484d4 100644
--- a/src/Common/Arena.h
+++ b/src/Common/Arena.h
@@ -83,10 +83,11 @@ private:
     /// Last contiguous chunk of memory.
     Chunk * head;
     size_t size_in_bytes;
+    size_t page_size;
 
-    static size_t roundUpToPageSize(size_t s)
+    static size_t roundUpToPageSize(size_t s, size_t page_size)
     {
-        return (s + 4096 - 1) / 4096 * 4096;
+        return (s + page_size - 1) / page_size * page_size;
     }
 
     /// If chunks size is less than 'linear_growth_threshold', then use exponential growth, otherwise - linear growth
@@ -113,7 +114,7 @@ private:
         }
 
         assert(size_after_grow >= min_next_size);
-        return roundUpToPageSize(size_after_grow);
+        return roundUpToPageSize(size_after_grow, page_size);
     }
 
     /// Add next contiguous chunk of memory with size not less than specified.
@@ -129,7 +130,8 @@ private:
 public:
     Arena(size_t initial_size_ = 4096, size_t growth_factor_ = 2, size_t linear_growth_threshold_ = 128 * 1024 * 1024)
         : growth_factor(growth_factor_), linear_growth_threshold(linear_growth_threshold_),
-        head(new Chunk(initial_size_, nullptr)), size_in_bytes(head->size())
+        head(new Chunk(initial_size_, nullptr)), size_in_bytes(head->size()),
+        page_size(static_cast<size_t>(::getPageSize()))
     {
     }
 
diff --git a/src/Common/ArrayCache.h b/src/Common/ArrayCache.h
index caa26d237bf..6883d5f8041 100644
--- a/src/Common/ArrayCache.h
+++ b/src/Common/ArrayCache.h
@@ -13,6 +13,8 @@
 #include <boost/noncopyable.hpp>
 #include <ext/scope_guard.h>
 
+#include <common/getPageSize.h>
+
 #include <Common/Exception.h>
 #include <Common/randomSeed.h>
 #include <Common/formatReadable.h>
@@ -326,8 +328,6 @@ private:
         return (x + (rounding - 1)) / rounding * rounding;
     }
 
-    static constexpr size_t page_size = 4096;
-
     /// Sizes and addresses of allocated memory will be aligned to specified boundary.
     static constexpr size_t alignment = 16;
 
@@ -505,6 +505,7 @@ private:
 
         /// If nothing was found and total size of allocated chunks plus required size is lower than maximum,
         ///  allocate a new chunk.
+        size_t page_size = static_cast<size_t>(::getPageSize());
         size_t required_chunk_size = std::max(min_chunk_size, roundUp(size, page_size));
         if (total_chunks_size + required_chunk_size <= max_total_size)
         {
diff --git a/src/Common/Config/ConfigProcessor.cpp b/src/Common/Config/ConfigProcessor.cpp
index 3e06be94b11..5399826fe20 100644
--- a/src/Common/Config/ConfigProcessor.cpp
+++ b/src/Common/Config/ConfigProcessor.cpp
@@ -16,9 +16,11 @@
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/Exception.h>
 #include <common/getResource.h>
+#include <common/errnoToString.h>
 #include <IO/WriteBufferFromString.h>
 #include <IO/Operators.h>
 
+
 #define PREPROCESSED_SUFFIX "-preprocessed"
 
 
@@ -234,7 +236,7 @@ static std::string layerFromHost()
 {
     utsname buf;
     if (uname(&buf))
-        throw Poco::Exception(std::string("uname failed: ") + std::strerror(errno));
+        throw Poco::Exception(std::string("uname failed: ") + errnoToString(errno));
 
     std::string layer = numberFromHost(buf.nodename);
     if (layer.empty())
diff --git a/src/Common/CurrentMetrics.cpp b/src/Common/CurrentMetrics.cpp
index c48e76e1d98..d3a4a41046e 100644
--- a/src/Common/CurrentMetrics.cpp
+++ b/src/Common/CurrentMetrics.cpp
@@ -56,6 +56,7 @@
     M(LocalThreadActive, "Number of threads in local thread pools running a task.") \
     M(DistributedFilesToInsert, "Number of pending files to process for asynchronous insertion into Distributed tables. Number of files for every shard is summed.") \
     M(TablesToDropQueueSize, "Number of dropped tables, that are waiting for background data removal.") \
+    M(MaxDDLEntryID, "Max processed DDL entry of DDLWorker.")           \
 
 namespace CurrentMetrics
 {
diff --git a/src/Common/ErrorCodes.cpp b/src/Common/ErrorCodes.cpp
index 1e381808d16..f9908e4e612 100644
--- a/src/Common/ErrorCodes.cpp
+++ b/src/Common/ErrorCodes.cpp
@@ -529,6 +529,7 @@
     M(560, ZSTD_ENCODER_FAILED) \
     M(561, ZSTD_DECODER_FAILED) \
     M(562, TLD_LIST_NOT_FOUND) \
+    M(563, CANNOT_READ_MAP_FROM_TEXT) \
     \
     M(999, KEEPER_EXCEPTION) \
     M(1000, POCO_EXCEPTION) \
diff --git a/src/Common/FieldVisitors.cpp b/src/Common/FieldVisitors.cpp
index ba3187ed039..dae153bd8d2 100644
--- a/src/Common/FieldVisitors.cpp
+++ b/src/Common/FieldVisitors.cpp
@@ -93,6 +93,22 @@ String FieldVisitorDump::operator() (const Tuple & x) const
     return wb.str();
 }
 
+String FieldVisitorDump::operator() (const Map & x) const
+{
+    WriteBufferFromOwnString wb;
+
+    wb << "Map_(";
+    for (auto it = x.begin(); it != x.end(); ++it)
+    {
+        if (it != x.begin())
+            wb << ", ";
+        wb << applyVisitor(*this, *it);
+    }
+    wb << ')';
+
+    return wb.str();
+}
+
 String FieldVisitorDump::operator() (const AggregateFunctionStateData & x) const
 {
     WriteBufferFromOwnString wb;
@@ -176,6 +192,82 @@ String FieldVisitorToString::operator() (const Tuple & x) const
     return wb.str();
 }
 
+String FieldVisitorToString::operator() (const Map & x) const
+{
+    WriteBufferFromOwnString wb;
+
+    wb << '(';
+    for (auto it = x.begin(); it != x.end(); ++it)
+    {
+        if (it != x.begin())
+            wb << ", ";
+        wb << applyVisitor(*this, *it);
+    }
+    wb << ')';
+
+    return wb.str();
+}
+
+
+void FieldVisitorWriteBinary::operator() (const Null &, WriteBuffer &) const { }
+void FieldVisitorWriteBinary::operator() (const UInt64 & x, WriteBuffer & buf) const { DB::writeVarUInt(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Int64 & x, WriteBuffer & buf) const { DB::writeVarInt(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Float64 & x, WriteBuffer & buf) const { DB::writeFloatBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const String & x, WriteBuffer & buf) const { DB::writeStringBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const UInt128 & x, WriteBuffer & buf) const { DB::writeBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Int128 & x, WriteBuffer & buf) const { DB::writeVarInt(x, buf); }
+void FieldVisitorWriteBinary::operator() (const UInt256 & x, WriteBuffer & buf) const { DB::writeBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const Int256 & x, WriteBuffer & buf) const { DB::writeBinary(x, buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal32> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal64> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal128> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const DecimalField<Decimal256> & x, WriteBuffer & buf) const { DB::writeBinary(x.getValue(), buf); }
+void FieldVisitorWriteBinary::operator() (const AggregateFunctionStateData & x, WriteBuffer & buf) const
+{
+    DB::writeStringBinary(x.name, buf);
+    DB::writeStringBinary(x.data, buf);
+}
+
+void FieldVisitorWriteBinary::operator() (const Array & x, WriteBuffer & buf) const
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (size_t i = 0; i < size; ++i)
+    {
+        const UInt8 type = x[i].getType();
+        DB::writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, x[i]);
+    }
+}
+
+void FieldVisitorWriteBinary::operator() (const Tuple & x, WriteBuffer & buf) const
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (size_t i = 0; i < size; ++i)
+    {
+        const UInt8 type = x[i].getType();
+        DB::writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, x[i]);
+    }
+}
+
+
+void FieldVisitorWriteBinary::operator() (const Map & x, WriteBuffer & buf) const
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (size_t i = 0; i < size; ++i)
+    {
+        const UInt8 type = x[i].getType();
+        writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, x[i]);
+    }
+}
+
 
 FieldVisitorHash::FieldVisitorHash(SipHash & hash_) : hash(hash_) {}
 
@@ -238,6 +330,16 @@ void FieldVisitorHash::operator() (const Tuple & x) const
         applyVisitor(*this, elem);
 }
 
+void FieldVisitorHash::operator() (const Map & x) const
+{
+    UInt8 type = Field::Types::Map;
+    hash.update(type);
+    hash.update(x.size());
+
+    for (const auto & elem : x)
+        applyVisitor(*this, elem);
+}
+
 void FieldVisitorHash::operator() (const Array & x) const
 {
     UInt8 type = Field::Types::Array;
diff --git a/src/Common/FieldVisitors.h b/src/Common/FieldVisitors.h
index 79d5af88904..314609ee1a0 100644
--- a/src/Common/FieldVisitors.h
+++ b/src/Common/FieldVisitors.h
@@ -77,6 +77,7 @@ public:
     String operator() (const String & x) const;
     String operator() (const Array & x) const;
     String operator() (const Tuple & x) const;
+    String operator() (const Map & x) const;
     String operator() (const DecimalField<Decimal32> & x) const;
     String operator() (const DecimalField<Decimal64> & x) const;
     String operator() (const DecimalField<Decimal128> & x) const;
@@ -88,6 +89,30 @@ public:
 };
 
 
+class FieldVisitorWriteBinary
+{
+public:
+    void operator() (const Null & x, WriteBuffer & buf) const;
+    void operator() (const UInt64 & x, WriteBuffer & buf) const;
+    void operator() (const UInt128 & x, WriteBuffer & buf) const;
+    void operator() (const Int64 & x, WriteBuffer & buf) const;
+    void operator() (const Int128 & x, WriteBuffer & buf) const;
+    void operator() (const Float64 & x, WriteBuffer & buf) const;
+    void operator() (const String & x, WriteBuffer & buf) const;
+    void operator() (const Array & x, WriteBuffer & buf) const;
+    void operator() (const Tuple & x, WriteBuffer & buf) const;
+    void operator() (const Map & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal32> & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal64> & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal128> & x, WriteBuffer & buf) const;
+    void operator() (const DecimalField<Decimal256> & x, WriteBuffer & buf) const;
+    void operator() (const AggregateFunctionStateData & x, WriteBuffer & buf) const;
+
+    void operator() (const UInt256 & x, WriteBuffer & buf) const;
+    void operator() (const Int256 & x, WriteBuffer & buf) const;
+};
+
+
 /** Print readable and unique text dump of field type and value. */
 class FieldVisitorDump : public StaticVisitor<String>
 {
@@ -101,6 +126,7 @@ public:
     String operator() (const String & x) const;
     String operator() (const Array & x) const;
     String operator() (const Tuple & x) const;
+    String operator() (const Map & x) const;
     String operator() (const DecimalField<Decimal32> & x) const;
     String operator() (const DecimalField<Decimal64> & x) const;
     String operator() (const DecimalField<Decimal128> & x) const;
@@ -137,6 +163,11 @@ public:
         throw Exception("Cannot convert Tuple to " + demangle(typeid(T).name()), ErrorCodes::CANNOT_CONVERT_TYPE);
     }
 
+    T operator() (const Map &) const
+    {
+        throw Exception("Cannot convert Map to " + demangle(typeid(T).name()), ErrorCodes::CANNOT_CONVERT_TYPE);
+    }
+
     T operator() (const UInt64 & x) const { return T(x); }
     T operator() (const Int64 & x) const { return T(x); }
     T operator() (const Int128 & x) const { return T(x); }
@@ -226,6 +257,7 @@ public:
     void operator() (const String & x) const;
     void operator() (const Array & x) const;
     void operator() (const Tuple & x) const;
+    void operator() (const Map & x) const;
     void operator() (const DecimalField<Decimal32> & x) const;
     void operator() (const DecimalField<Decimal64> & x) const;
     void operator() (const DecimalField<Decimal128> & x) const;
@@ -268,6 +300,7 @@ public:
     bool operator() (String &) const { throw Exception("Cannot sum Strings", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (Array &) const { throw Exception("Cannot sum Arrays", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (Tuple &) const { throw Exception("Cannot sum Tuples", ErrorCodes::LOGICAL_ERROR); }
+    bool operator() (Map &) const { throw Exception("Cannot sum Maps", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (UInt128 &) const { throw Exception("Cannot sum UUIDs", ErrorCodes::LOGICAL_ERROR); }
     bool operator() (AggregateFunctionStateData &) const { throw Exception("Cannot sum AggregateFunctionStates", ErrorCodes::LOGICAL_ERROR); }
 
diff --git a/src/Common/HashTable/HashTable.h b/src/Common/HashTable/HashTable.h
index a569b1c15db..2d5580bf709 100644
--- a/src/Common/HashTable/HashTable.h
+++ b/src/Common/HashTable/HashTable.h
@@ -194,9 +194,6 @@ struct HashTableCell
     /// Do the hash table need to store the zero key separately (that is, can a zero key be inserted into the hash table).
     static constexpr bool need_zero_value_storage = true;
 
-    /// Whether the cell is deleted.
-    bool isDeleted() const { return false; }
-
     /// Set the mapped value, if any (for HashMap), to the corresponding `value`.
     void setMapped(const value_type & /*value*/) {}
 
@@ -230,6 +227,9 @@ struct HashTableGrower
     UInt8 size_degree = initial_size_degree;
     static constexpr auto initial_count = 1ULL << initial_size_degree;
 
+    /// If collision resolution chains are contiguous, we can implement erase operation by moving the elements.
+    static constexpr auto performs_linear_probing_with_single_step = true;
+
     /// The size of the hash table in the cells.
     size_t bufSize() const               { return 1ULL << size_degree; }
 
@@ -277,6 +277,9 @@ template <size_t key_bits>
 struct HashTableFixedGrower
 {
     static constexpr auto initial_count = 1ULL << key_bits;
+
+    static constexpr auto performs_linear_probing_with_single_step = true;
+
     size_t bufSize() const               { return 1ULL << key_bits; }
     size_t place(size_t x) const         { return x; }
     /// You could write __builtin_unreachable(), but the compiler does not optimize everything, and it turns out less efficiently.
@@ -466,7 +469,7 @@ protected:
           */
         size_t i = 0;
         for (; i < old_size; ++i)
-            if (!buf[i].isZero(*this) && !buf[i].isDeleted())
+            if (!buf[i].isZero(*this))
                 reinsert(buf[i], buf[i].getHash(*this));
 
         /** There is also a special case:
@@ -477,7 +480,7 @@ protected:
           *    after transferring all the elements from the old halves you need to     [         o   x    ]
           *    process tail from the collision resolution chain immediately after it   [        o    x    ]
           */
-        for (; !buf[i].isZero(*this) && !buf[i].isDeleted(); ++i)
+        for (; !buf[i].isZero(*this); ++i)
             reinsert(buf[i], buf[i].getHash(*this));
 
 #ifdef DBMS_HASH_MAP_DEBUG_RESIZES
@@ -829,6 +832,7 @@ protected:
                   */
                 --m_size;
                 buf[place_value].setZero();
+                inserted = false;
                 throw;
             }
 
@@ -954,6 +958,112 @@ public:
         return const_cast<std::decay_t<decltype(*this)> *>(this)->find(x, hash_value);
     }
 
+    std::enable_if_t<Grower::performs_linear_probing_with_single_step, void>
+    ALWAYS_INLINE erase(const Key & x)
+    {
+        /** Deletion from open addressing hash table without tombstones
+          *
+          * https://en.wikipedia.org/wiki/Linear_probing
+          * https://en.wikipedia.org/wiki/Open_addressing
+          * Algorithm without recomputing hash but keep probes difference value (difference of natural cell position and inserted one)
+          * in cell https://arxiv.org/ftp/arxiv/papers/0909/0909.2547.pdf
+          *
+          * Currently we use algorithm with hash recomputing on each step from https://en.wikipedia.org/wiki/Open_addressing
+          */
+
+        if (Cell::isZero(x, *this))
+        {
+            if (this->hasZero())
+            {
+                --m_size;
+                this->clearHasZero();
+            }
+            else
+            {
+                return;
+            }
+        }
+
+        size_t hash_value = hash(x);
+        size_t erased_key_position = findCell(x, hash_value, grower.place(hash_value));
+
+        /// Key is not found
+        if (buf[erased_key_position].isZero(*this))
+        {
+            return;
+        }
+
+        /// We need to guarantee loop termination because there will be empty position
+        assert(m_size < grower.bufSize());
+
+        size_t next_position = erased_key_position;
+
+        /**
+         * During element deletion there is a possibility that the search will be broken for one
+         * of the following elements, because this place erased_key_position is empty. We will check
+         * next_element. Consider a sequence from (erased_key_position, next_element], if the
+         * optimal_position of next_element falls into it, then removing erased_key_position
+         * will not break search for next_element.
+         * If optimal_position of the element does not fall into the sequence (erased_key_position, next_element]
+         * then deleting a erased_key_position will break search for it, so we need to move next_element
+         * to erased_key_position. Now we have empty place at next_element, so we apply the identical
+         * procedure for it.
+         * If an empty element is encoutered then means that there is no more next elements for which we can
+         * break the search so we can exit.
+        */
+
+        /// Walk to the right through collision resolution chain and move elements to better positions
+        while (true)
+        {
+            next_position = grower.next(next_position);
+
+            /// If there's no more elements in the chain
+            if (buf[next_position].isZero(*this))
+                break;
+
+            /// The optimal position of the element in the cell at next_position
+            size_t optimal_position = grower.place(buf[next_position].getHash(*this));
+
+            /// If position of this element is already optimal - proceed to the next element.
+            if (optimal_position == next_position)
+                continue;
+
+            /// Cannot move this element because optimal position is after the freed place
+            /// The second condition is tricky - if the chain was overlapped before erased_key_position,
+            ///  and the optimal position is actually before in collision resolution chain:
+            ///
+            /// [*xn***----------------***]
+            ///   ^^-next elem          ^
+            ///   |                     |
+            ///   erased elem           the optimal position of the next elem
+            ///
+            /// so, the next elem should be moved to position of erased elem
+
+            /// The case of non overlapping part of chain
+            if (next_position > erased_key_position
+               && (optimal_position > erased_key_position) && (optimal_position < next_position))
+            {
+                continue;
+            }
+
+            /// The case of overlapping chain
+            if (next_position < erased_key_position
+                /// Cannot move this element because optimal position is after the freed place
+                && ((optimal_position > erased_key_position) || (optimal_position < next_position)))
+            {
+                continue;
+            }
+
+            /// Move the element to the freed place
+            memcpy(static_cast<void *>(&buf[erased_key_position]), static_cast<void *>(&buf[next_position]), sizeof(Cell));
+            /// Now we have another freed place
+            erased_key_position = next_position;
+        }
+
+        buf[erased_key_position].setZero();
+        --m_size;
+    }
+
     bool ALWAYS_INLINE has(const Key & x) const
     {
         if (Cell::isZero(x, *this))
diff --git a/src/Common/IFactoryWithAliases.h b/src/Common/IFactoryWithAliases.h
index 11ebf31db33..49c03049b92 100644
--- a/src/Common/IFactoryWithAliases.h
+++ b/src/Common/IFactoryWithAliases.h
@@ -106,6 +106,11 @@ public:
         return aliases.count(name) || case_insensitive_aliases.count(name);
     }
 
+    bool hasNameOrAlias(const String & name) const
+    {
+        return getMap().count(name) || getCaseInsensitiveMap().count(name) || isAlias(name);
+    }
+
     virtual ~IFactoryWithAliases() override {}
 
 private:
diff --git a/src/Common/MemoryStatisticsOS.cpp b/src/Common/MemoryStatisticsOS.cpp
index 7c65461aeb6..f1022d887c1 100644
--- a/src/Common/MemoryStatisticsOS.cpp
+++ b/src/Common/MemoryStatisticsOS.cpp
@@ -8,10 +8,11 @@
 
 #include "MemoryStatisticsOS.h"
 
+#include <common/logger_useful.h>
+#include <common/getPageSize.h>
 #include <Common/Exception.h>
 #include <IO/ReadBufferFromMemory.h>
 #include <IO/ReadHelpers.h>
-#include <common/logger_useful.h>
 
 
 namespace DB
@@ -26,7 +27,6 @@ namespace ErrorCodes
 }
 
 static constexpr auto filename = "/proc/self/statm";
-static constexpr size_t PAGE_SIZE = 4096;
 
 MemoryStatisticsOS::MemoryStatisticsOS()
 {
@@ -93,11 +93,12 @@ MemoryStatisticsOS::Data MemoryStatisticsOS::get() const
     skipWhitespaceIfAny(in);
     readIntText(data.data_and_stack, in);
 
-    data.virt *= PAGE_SIZE;
-    data.resident *= PAGE_SIZE;
-    data.shared *= PAGE_SIZE;
-    data.code *= PAGE_SIZE;
-    data.data_and_stack *= PAGE_SIZE;
+    size_t page_size = static_cast<size_t>(::getPageSize());
+    data.virt *= page_size;
+    data.resident *= page_size;
+    data.shared *= page_size;
+    data.code *= page_size;
+    data.data_and_stack *= page_size;
 
     return data;
 }
diff --git a/src/Common/SpaceSaving.h b/src/Common/SpaceSaving.h
index cb6fee1ad91..185b4aa90ae 100644
--- a/src/Common/SpaceSaving.h
+++ b/src/Common/SpaceSaving.h
@@ -353,6 +353,7 @@ private:
     void destroyLastElement()
     {
         auto last_element = counter_list.back();
+        counter_map.erase(last_element->key);
         arena.free(last_element->key);
         delete last_element;
         counter_list.pop_back();
diff --git a/src/Common/StringSearcher.h b/src/Common/StringSearcher.h
index a4da499837b..e39253c439f 100644
--- a/src/Common/StringSearcher.h
+++ b/src/Common/StringSearcher.h
@@ -1,5 +1,6 @@
 #pragma once
 
+#include <common/getPageSize.h>
 #include <Common/Exception.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/UTF8Helpers.h>
@@ -37,7 +38,7 @@ struct StringSearcherBase
 {
 #ifdef __SSE2__
     static constexpr auto n = sizeof(__m128i);
-    const int page_size = getpagesize();
+    const int page_size = ::getPageSize();
 
     bool pageSafe(const void * const ptr) const
     {
diff --git a/src/Common/ThreadProfileEvents.cpp b/src/Common/ThreadProfileEvents.cpp
index c8a4555dabf..e6336baecda 100644
--- a/src/Common/ThreadProfileEvents.cpp
+++ b/src/Common/ThreadProfileEvents.cpp
@@ -21,6 +21,9 @@
 #include <sys/types.h>
 #include <dirent.h>
 
+#include <common/errnoToString.h>
+
+
 namespace DB
 {
 
@@ -246,7 +249,7 @@ static void enablePerfEvent(int event_fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Can't enable perf event with file descriptor {}: '{}' ({})",
-            event_fd, strerror(errno), errno);
+            event_fd, errnoToString(errno), errno);
     }
 }
 
@@ -256,7 +259,7 @@ static void disablePerfEvent(int event_fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Can't disable perf event with file descriptor {}: '{}' ({})",
-            event_fd, strerror(errno), errno);
+            event_fd, errnoToString(errno), errno);
     }
 }
 
@@ -266,7 +269,7 @@ static void releasePerfEvent(int event_fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Can't close perf event file descriptor {}: {} ({})",
-            event_fd, strerror(errno), errno);
+            event_fd, errnoToString(errno), errno);
     }
 }
 
@@ -284,7 +287,7 @@ static bool validatePerfEventDescriptor(int & fd)
     {
         LOG_WARNING(&Poco::Logger::get("PerfEvents"),
             "Error while checking availability of event descriptor {}: {} ({})",
-            fd, strerror(errno), errno);
+            fd, errnoToString(errno), errno);
 
         disablePerfEvent(fd);
         releasePerfEvent(fd);
@@ -391,7 +394,7 @@ bool PerfEventsCounters::processThreadLocalChanges(const std::string & needed_ev
             LOG_WARNING(&Poco::Logger::get("PerfEvents"),
                 "Failed to open perf event {} (event_type={}, event_config={}): "
                 "'{}' ({})", event_info.settings_name, event_info.event_type,
-                event_info.event_config, strerror(errno), errno);
+                event_info.event_config, errnoToString(errno), errno);
         }
     }
 
@@ -477,7 +480,7 @@ void PerfEventsCounters::finalizeProfileEvents(ProfileEvents::Counters & profile
         {
             LOG_WARNING(&Poco::Logger::get("PerfEvents"),
                 "Can't read event value from file descriptor {}: '{}' ({})",
-                fd, strerror(errno), errno);
+                fd, errnoToString(errno), errno);
             current_values[i] = {};
         }
     }
diff --git a/src/Common/ThreadStatus.cpp b/src/Common/ThreadStatus.cpp
index 3215bb55537..4bba3d8f4eb 100644
--- a/src/Common/ThreadStatus.cpp
+++ b/src/Common/ThreadStatus.cpp
@@ -2,11 +2,14 @@
 #include <Common/ThreadProfileEvents.h>
 #include <Common/QueryProfiler.h>
 #include <Common/ThreadStatus.h>
+#include <common/errnoToString.h>
 #include <Interpreters/OpenTelemetrySpanLog.h>
 
 #include <Poco/Logger.h>
 #include <common/getThreadId.h>
 
+#include <signal.h>
+
 
 namespace DB
 {
@@ -21,6 +24,11 @@ namespace ErrorCodes
 thread_local ThreadStatus * current_thread = nullptr;
 thread_local ThreadStatus * main_thread = nullptr;
 
+#if !defined(SANITIZER) && !defined(ARCADIA_BUILD)
+    alignas(4096) static thread_local char alt_stack[4096];
+    static thread_local bool has_alt_stack = false;
+#endif
+
 
 ThreadStatus::ThreadStatus()
     : thread_id{getThreadId()}
@@ -34,6 +42,46 @@ ThreadStatus::ThreadStatus()
 
     /// NOTE: It is important not to do any non-trivial actions (like updating ProfileEvents or logging) before ThreadStatus is created
     /// Otherwise it could lead to SIGSEGV due to current_thread dereferencing
+
+    /// Will set alternative signal stack to provide diagnostics for stack overflow errors.
+    /// If not already installed for current thread.
+    /// Sanitizer makes larger stack usage and also it's incompatible with alternative stack by default (it sets up and relies on its own).
+#if !defined(SANITIZER) && !defined(ARCADIA_BUILD)
+    if (!has_alt_stack)
+    {
+        /// Don't repeat tries even if not installed successfully.
+        has_alt_stack = true;
+
+        /// We have to call 'sigaltstack' before first 'sigaction'. (It does not work other way, for unknown reason).
+        stack_t altstack_description{};
+        altstack_description.ss_sp = alt_stack;
+        altstack_description.ss_flags = 0;
+        altstack_description.ss_size = sizeof(alt_stack);
+
+        if (0 != sigaltstack(&altstack_description, nullptr))
+        {
+            LOG_WARNING(log, "Cannot set alternative signal stack for thread, {}", errnoToString(errno));
+        }
+        else
+        {
+            /// Obtain existing sigaction and modify it by adding a flag.
+            struct sigaction action{};
+            if (0 != sigaction(SIGSEGV, nullptr, &action))
+            {
+                LOG_WARNING(log, "Cannot obtain previous signal action to set alternative signal stack for thread, {}", errnoToString(errno));
+            }
+            else if (!(action.sa_flags & SA_ONSTACK))
+            {
+                action.sa_flags |= SA_ONSTACK;
+
+                if (0 != sigaction(SIGSEGV, &action, nullptr))
+                {
+                    LOG_WARNING(log, "Cannot set action with alternative signal stack for thread, {}", errnoToString(errno));
+                }
+            }
+        }
+    }
+#endif
 }
 
 ThreadStatus::~ThreadStatus()
diff --git a/src/Common/UInt128.h b/src/Common/UInt128.h
index baa33578089..06fddee8dc9 100644
--- a/src/Common/UInt128.h
+++ b/src/Common/UInt128.h
@@ -28,23 +28,28 @@ struct UInt128
     UInt64 low;
     UInt64 high;
 
+    /// TODO: Make this constexpr. Currently it is used in unions
+    /// and union cannot contain member with non trivial constructor
+    /// constructor must be non user provided but compiler cannot constexpr constructor
+    /// if members low and high are not initialized, if we default member initialize them
+    /// constructor becomes non trivial.
     UInt128() = default;
-    explicit UInt128(const UInt64 low_, const UInt64 high_) : low(low_), high(high_) {}
+    explicit constexpr UInt128(const UInt64 low_, const UInt64 high_) : low(low_), high(high_) { }
 
     /// We need Int128 to UInt128 conversion or AccurateComparison will call greaterOp<Int128, UInt64> instead of greaterOp<Int128, UInt128>
-    explicit UInt128(const Int128 rhs) : low(rhs), high(rhs >> 64) {}
-    explicit UInt128(const Int64 rhs) : low(rhs), high() {}
-    explicit UInt128(const Int32 rhs) : low(rhs), high() {}
-    explicit UInt128(const Int16 rhs) : low(rhs), high() {}
-    explicit UInt128(const Int8 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt8 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt16 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt32 rhs) : low(rhs), high() {}
-    explicit UInt128(const UInt64 rhs) : low(rhs), high() {}
-    explicit UInt128(const Float32 rhs) : low(rhs), high() {}
-    explicit UInt128(const Float64 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int128 rhs) : low(rhs), high(rhs >> 64) {}
+    explicit constexpr UInt128(const Int64 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int32 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int16 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Int8 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt8 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt16 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt32 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const UInt64 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Float32 rhs) : low(rhs), high() {}
+    explicit constexpr UInt128(const Float64 rhs) : low(rhs), high() {}
 
-    auto tuple() const { return std::tie(high, low); }
+    constexpr auto tuple() const { return std::tie(high, low); }
 
     String toHexString() const
     {
@@ -53,31 +58,31 @@ struct UInt128
         return res;
     }
 
-    bool inline operator== (const UInt128 rhs) const { return tuple() == rhs.tuple(); }
-    bool inline operator!= (const UInt128 rhs) const { return tuple() != rhs.tuple(); }
-    bool inline operator<  (const UInt128 rhs) const { return tuple() < rhs.tuple(); }
-    bool inline operator<= (const UInt128 rhs) const { return tuple() <= rhs.tuple(); }
-    bool inline operator>  (const UInt128 rhs) const { return tuple() > rhs.tuple(); }
-    bool inline operator>= (const UInt128 rhs) const { return tuple() >= rhs.tuple(); }
+    constexpr bool operator== (const UInt128 rhs) const { return tuple() == rhs.tuple(); }
+    constexpr bool operator!= (const UInt128 rhs) const { return tuple() != rhs.tuple(); }
+    constexpr bool operator<  (const UInt128 rhs) const { return tuple() < rhs.tuple(); }
+    constexpr bool operator<= (const UInt128 rhs) const { return tuple() <= rhs.tuple(); }
+    constexpr bool operator>  (const UInt128 rhs) const { return tuple() > rhs.tuple(); }
+    constexpr bool operator>= (const UInt128 rhs) const { return tuple() >= rhs.tuple(); }
 
-    bool inline operator == (const Int128 rhs) const { return *this == UInt128(rhs, rhs >> 64) && rhs >= 0; }
-    bool inline operator != (const Int128 rhs) const { return *this != UInt128(rhs, rhs >> 64) || rhs < 0; }
-    bool inline operator >= (const Int128 rhs) const { return *this >= UInt128(rhs, rhs >> 64) || rhs < 0; }
-    bool inline operator >  (const Int128 rhs) const { return *this >  UInt128(rhs, rhs >> 64) || rhs < 0; }
-    bool inline operator <= (const Int128 rhs) const { return *this <= UInt128(rhs, rhs >> 64) && rhs >= 0; }
-    bool inline operator <  (const Int128 rhs) const { return *this <  UInt128(rhs, rhs >> 64) && rhs >= 0; }
+    constexpr bool operator == (const Int128 rhs) const { return *this == UInt128(rhs, rhs >> 64) && rhs >= 0; }
+    constexpr bool operator != (const Int128 rhs) const { return *this != UInt128(rhs, rhs >> 64) || rhs < 0; }
+    constexpr bool operator >= (const Int128 rhs) const { return *this >= UInt128(rhs, rhs >> 64) || rhs < 0; }
+    constexpr bool operator >  (const Int128 rhs) const { return *this >  UInt128(rhs, rhs >> 64) || rhs < 0; }
+    constexpr bool operator <= (const Int128 rhs) const { return *this <= UInt128(rhs, rhs >> 64) && rhs >= 0; }
+    constexpr bool operator <  (const Int128 rhs) const { return *this <  UInt128(rhs, rhs >> 64) && rhs >= 0; }
 
-    bool inline operator >  (const Int256 rhs) const { return (rhs < 0) || ((Int256(high) << 64) + low) > rhs; }
-    bool inline operator >  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) > rhs; }
-    bool inline operator <  (const Int256 rhs) const { return (rhs >= 0) && ((Int256(high) << 64) + low) < rhs; }
-    bool inline operator <  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) < rhs; }
+    constexpr bool operator >  (const Int256 rhs) const { return (rhs < 0) || ((Int256(high) << 64) + low) > rhs; }
+    constexpr bool operator >  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) > rhs; }
+    constexpr bool operator <  (const Int256 rhs) const { return (rhs >= 0) && ((Int256(high) << 64) + low) < rhs; }
+    constexpr bool operator <  (const UInt256 rhs) const { return ((UInt256(high) << 64) + low) < rhs; }
 
-    template <typename T> bool inline operator== (const T rhs) const { return *this == UInt128(rhs); }
-    template <typename T> bool inline operator!= (const T rhs) const { return *this != UInt128(rhs); }
-    template <typename T> bool inline operator>= (const T rhs) const { return *this >= UInt128(rhs); }
-    template <typename T> bool inline operator>  (const T rhs) const { return *this >  UInt128(rhs); }
-    template <typename T> bool inline operator<= (const T rhs) const { return *this <= UInt128(rhs); }
-    template <typename T> bool inline operator<  (const T rhs) const { return *this <  UInt128(rhs); }
+    template <typename T> constexpr bool operator== (const T rhs) const { return *this == UInt128(rhs); }
+    template <typename T> constexpr bool operator!= (const T rhs) const { return *this != UInt128(rhs); }
+    template <typename T> constexpr bool operator>= (const T rhs) const { return *this >= UInt128(rhs); }
+    template <typename T> constexpr bool operator>  (const T rhs) const { return *this >  UInt128(rhs); }
+    template <typename T> constexpr bool operator<= (const T rhs) const { return *this <= UInt128(rhs); }
+    template <typename T> constexpr bool operator<  (const T rhs) const { return *this <  UInt128(rhs); }
 
     template <typename T> explicit operator T() const
     {
@@ -91,15 +96,15 @@ struct UInt128
 #pragma GCC diagnostic pop
 #endif
 
-    UInt128 & operator= (const UInt64 rhs) { low = rhs; high = 0; return *this; }
+    constexpr UInt128 & operator= (const UInt64 rhs) { low = rhs; high = 0; return *this; }
 };
 
-template <typename T> bool inline operator == (T a, const UInt128 b) { return b.operator==(a); }
-template <typename T> bool inline operator != (T a, const UInt128 b) { return b.operator!=(a); }
-template <typename T> bool inline operator >= (T a, const UInt128 b) { return b <= a; }
-template <typename T> bool inline operator >  (T a, const UInt128 b) { return b < a; }
-template <typename T> bool inline operator <= (T a, const UInt128 b) { return b >= a; }
-template <typename T> bool inline operator <  (T a, const UInt128 b) { return b > a; }
+template <typename T> constexpr bool operator == (T a, const UInt128 b) { return b.operator==(a); }
+template <typename T> constexpr bool operator != (T a, const UInt128 b) { return b.operator!=(a); }
+template <typename T> constexpr bool operator >= (T a, const UInt128 b) { return b <= a; }
+template <typename T> constexpr bool operator >  (T a, const UInt128 b) { return b < a; }
+template <typename T> constexpr bool operator <= (T a, const UInt128 b) { return b >= a; }
+template <typename T> constexpr bool operator <  (T a, const UInt128 b) { return b > a; }
 
 template <> inline constexpr bool IsNumber<UInt128> = true;
 template <> struct TypeName<UInt128> { static constexpr const char * get() { return "UInt128"; } };
@@ -246,4 +251,42 @@ template <> struct hash<DB::UInt128>
     }
 };
 
+template<>
+class numeric_limits<DB::UInt128>
+{
+public:
+    static constexpr bool is_specialized = true;
+    static constexpr bool is_signed = ::is_signed<DB::UInt128>::value;
+    static constexpr bool is_integer = ::is_integer<DB::UInt128>::value;
+    static constexpr bool is_exact = true;
+    static constexpr bool has_infinity = false;
+    static constexpr bool has_quiet_NaN = false;
+    static constexpr bool has_signaling_NaN = false;
+    static constexpr std::float_denorm_style has_denorm = std::denorm_absent;
+    static constexpr bool has_denorm_loss = false;
+    static constexpr std::float_round_style round_style = std::round_toward_zero;
+    static constexpr bool is_iec559 = false;
+    static constexpr bool is_bounded = true;
+    static constexpr bool is_modulo = true;
+    static constexpr int digits = std::numeric_limits<UInt64>::digits * 2;
+    static constexpr int digits10 = digits * 0.30103 /*std::log10(2)*/;
+    static constexpr int max_digits10 = 0;
+    static constexpr int radix = 2;
+    static constexpr int min_exponent = 0;
+    static constexpr int min_exponent10 = 0;
+    static constexpr int max_exponent = 0;
+    static constexpr int max_exponent10 = 0;
+    static constexpr bool traps = true;
+    static constexpr bool tinyness_before = false;
+
+    static constexpr DB::UInt128 min() noexcept { return DB::UInt128(0, 0); }
+
+    static constexpr DB::UInt128 max() noexcept
+    {
+        return DB::UInt128(std::numeric_limits<UInt64>::max(), std::numeric_limits<UInt64>::max());
+    }
+
+    static constexpr DB::UInt128 lowest() noexcept { return min(); }
+};
+
 }
diff --git a/src/Common/XDBCBridgeHelper.h b/src/Common/XDBCBridgeHelper.h
index ed1f63a2507..d7d3a6ba4cc 100644
--- a/src/Common/XDBCBridgeHelper.h
+++ b/src/Common/XDBCBridgeHelper.h
@@ -12,6 +12,7 @@
 #include <Poco/URI.h>
 #include <Poco/Util/AbstractConfiguration.h>
 #include <Common/ShellCommand.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <common/logger_useful.h>
 #include <ext/range.h>
 
diff --git a/src/Common/ZooKeeper/TestKeeperStorage.cpp b/src/Common/ZooKeeper/TestKeeperStorage.cpp
index e3e4edc23a7..6b6ae93bbb9 100644
--- a/src/Common/ZooKeeper/TestKeeperStorage.cpp
+++ b/src/Common/ZooKeeper/TestKeeperStorage.cpp
@@ -798,6 +798,21 @@ void TestKeeperStorage::clearDeadWatches(int64_t session_id)
                 if (watches_for_path.empty())
                     watches.erase(watch);
             }
+
+            auto list_watch = list_watches.find(watch_path);
+            if (list_watch != list_watches.end())
+            {
+                auto & list_watches_for_path = list_watch->second;
+                for (auto w_it = list_watches_for_path.begin(); w_it != list_watches_for_path.end();)
+                {
+                    if (w_it->session_id == session_id)
+                        w_it = list_watches_for_path.erase(w_it);
+                    else
+                        ++w_it;
+                }
+                if (list_watches_for_path.empty())
+                    list_watches.erase(list_watch);
+            }
         }
         sessions_and_watchers.erase(watches_it);
     }
diff --git a/src/Common/config.h.in b/src/Common/config.h.in
index 3a67b2a8f2c..ee2bfe3df53 100644
--- a/src/Common/config.h.in
+++ b/src/Common/config.h.in
@@ -5,6 +5,7 @@
 #cmakedefine01 USE_RE2_ST
 #cmakedefine01 USE_SSL
 #cmakedefine01 USE_HDFS
+#cmakedefine01 USE_INTERNAL_HDFS3_LIBRARY
 #cmakedefine01 USE_AWS_S3
 #cmakedefine01 USE_BROTLI
 #cmakedefine01 USE_UNWIND
diff --git a/src/Common/tests/CMakeLists.txt b/src/Common/tests/CMakeLists.txt
index 6a39c2f8553..cb36e2b97d2 100644
--- a/src/Common/tests/CMakeLists.txt
+++ b/src/Common/tests/CMakeLists.txt
@@ -10,9 +10,6 @@ target_link_libraries (sip_hash_perf PRIVATE clickhouse_common_io)
 add_executable (auto_array auto_array.cpp)
 target_link_libraries (auto_array PRIVATE clickhouse_common_io)
 
-add_executable (hash_table hash_table.cpp)
-target_link_libraries (hash_table PRIVATE clickhouse_common_io)
-
 add_executable (small_table small_table.cpp)
 target_link_libraries (small_table PRIVATE clickhouse_common_io)
 
diff --git a/src/Common/tests/gtest_hash_table.cpp b/src/Common/tests/gtest_hash_table.cpp
new file mode 100644
index 00000000000..41255dcbba1
--- /dev/null
+++ b/src/Common/tests/gtest_hash_table.cpp
@@ -0,0 +1,319 @@
+#include <iomanip>
+#include <iostream>
+
+#include <Interpreters/AggregationCommon.h>
+
+#include <Common/HashTable/HashMap.h>
+#include <Common/HashTable/HashSet.h>
+
+#include <IO/ReadBufferFromString.h>
+
+#include <gtest/gtest.h>
+
+/// To test dump functionality without using other hashes that can change
+template <typename T>
+struct DummyHash
+{
+    size_t operator()(T key) const { return T(key); }
+};
+
+template<typename HashTable>
+std::set<typename HashTable::value_type> convertToSet(const HashTable& table)
+{
+    std::set<typename HashTable::value_type> result;
+
+    for (auto v: table)
+        result.emplace(v.getValue());
+
+    return result;
+}
+
+
+TEST(HashTable, Insert)
+{
+    using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
+
+    Cont cont;
+
+    cont.insert(1);
+    cont.insert(2);
+
+    ASSERT_EQ(cont.size(), 2);
+}
+
+TEST(HashTable, Emplace)
+{
+    using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
+
+    Cont cont;
+
+    Cont::LookupResult it;
+    bool inserted = false;
+    cont.emplace(1, it, inserted);
+    ASSERT_EQ(it->getKey(), 1);
+    ASSERT_EQ(inserted, true);
+
+    cont.emplace(2, it, inserted);
+    ASSERT_EQ(it->getKey(), 2);
+    ASSERT_EQ(inserted, true);
+
+    cont.emplace(1, it, inserted);
+    ASSERT_EQ(it->getKey(), 1);
+    ASSERT_EQ(inserted, false);
+}
+
+TEST(HashTable, Lookup)
+{
+    using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
+
+    Cont cont;
+
+    cont.insert(1);
+    cont.insert(2);
+
+    Cont::LookupResult it = cont.find(1);
+    ASSERT_TRUE(it != nullptr);
+
+    it = cont.find(2);
+    ASSERT_TRUE(it != nullptr);
+
+    it = cont.find(3);
+    ASSERT_TRUE(it == nullptr);
+}
+
+TEST(HashTable, Iteration)
+{
+    using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
+
+    Cont cont;
+
+    cont.insert(1);
+    cont.insert(2);
+    cont.insert(3);
+
+    std::set<int> expected = {1, 2, 3};
+    std::set<int> actual = convertToSet(cont);
+
+    ASSERT_EQ(actual, expected);
+}
+
+TEST(HashTable, Erase)
+{
+    {
+        /// Check zero element deletion
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        cont.insert(0);
+
+        ASSERT_TRUE(cont.find(0) != nullptr && cont.find(0)->getKey() == 0);
+
+        cont.erase(0);
+
+        ASSERT_TRUE(cont.find(0) == nullptr);
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        /// [.(1)..............] erase of (1).
+        cont.insert(1);
+
+        ASSERT_TRUE(cont.find(1) != nullptr && cont.find(1)->getKey() == 1);
+
+        cont.erase(1);
+
+        ASSERT_TRUE(cont.find(1) == nullptr);
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        /// [.(1)(2)(3)............] erase of (1) does not break search for (2) (3).
+        cont.insert(1);
+        cont.insert(2);
+        cont.insert(3);
+        cont.erase(1);
+
+        ASSERT_TRUE(cont.find(1) == nullptr);
+        ASSERT_TRUE(cont.find(2) != nullptr && cont.find(2)->getKey() == 2);
+        ASSERT_TRUE(cont.find(3) != nullptr && cont.find(3)->getKey() == 3);
+
+        cont.erase(2);
+        cont.erase(3);
+        ASSERT_TRUE(cont.find(2) == nullptr);
+        ASSERT_TRUE(cont.find(3) == nullptr);
+        ASSERT_EQ(cont.size(), 0);
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        /// [.(1)(17).............] erase of (1) breaks search for (17) because their natural position is 1.
+        cont.insert(1);
+        cont.insert(17);
+        cont.erase(1);
+
+        ASSERT_TRUE(cont.find(1) == nullptr);
+        ASSERT_TRUE(cont.find(17) != nullptr && cont.find(17)->getKey() == 17);
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        /// [.(1)(2)(3)(17)...........] erase of (2) breaks search for (17) because their natural position is 1.
+
+        cont.insert(1);
+        cont.insert(2);
+        cont.insert(3);
+        cont.insert(17);
+        cont.erase(2);
+
+        ASSERT_TRUE(cont.find(2) == nullptr);
+        ASSERT_TRUE(cont.find(1) != nullptr && cont.find(1)->getKey() == 1);
+        ASSERT_TRUE(cont.find(3) != nullptr && cont.find(3)->getKey() == 3);
+        ASSERT_TRUE(cont.find(17) != nullptr && cont.find(17)->getKey() == 17);
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        /// [(16)(30)............(14)(15)] erase of (16) breaks search for (30) because their natural position is 14.
+        cont.insert(14);
+        cont.insert(15);
+        cont.insert(16);
+        cont.insert(30);
+        cont.erase(16);
+
+        ASSERT_TRUE(cont.find(16) == nullptr);
+        ASSERT_TRUE(cont.find(14) != nullptr && cont.find(14)->getKey() == 14);
+        ASSERT_TRUE(cont.find(15) != nullptr && cont.find(15)->getKey() == 15);
+        ASSERT_TRUE(cont.find(30) != nullptr && cont.find(30)->getKey() == 30);
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<4>>;
+        Cont cont;
+
+        /// [(16)(30)............(14)(15)] erase of (15) breaks search for (30) because their natural position is 14.
+        cont.insert(14);
+        cont.insert(15);
+        cont.insert(16);
+        cont.insert(30);
+        cont.erase(15);
+
+        ASSERT_TRUE(cont.find(15) == nullptr);
+        ASSERT_TRUE(cont.find(14) != nullptr && cont.find(14)->getKey() == 14);
+        ASSERT_TRUE(cont.find(16) != nullptr && cont.find(16)->getKey() == 16);
+        ASSERT_TRUE(cont.find(30) != nullptr && cont.find(30)->getKey() == 30);
+    }
+    {
+        using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
+        Cont cont;
+
+        for (size_t i = 0; i < 5000; ++i)
+        {
+            cont.insert(i);
+        }
+
+        for (size_t i = 0; i < 2500; ++i)
+        {
+            cont.erase(i);
+        }
+
+        for (size_t i = 5000; i < 10000; ++i)
+        {
+            cont.insert(i);
+        }
+
+        for (size_t i = 5000; i < 10000; ++i)
+        {
+            cont.erase(i);
+        }
+
+        for (size_t i = 2500; i < 5000; ++i)
+        {
+            cont.erase(i);
+        }
+
+        ASSERT_EQ(cont.size(), 0);
+    }
+}
+
+TEST(HashTable, SerializationDeserialization)
+{
+    {
+        /// Use dummy hash to make it reproducible if default hash implementation will be changed
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<1>>;
+
+        Cont cont;
+
+        cont.insert(1);
+        cont.insert(2);
+        cont.insert(3);
+
+        DB::WriteBufferFromOwnString wb;
+        cont.writeText(wb);
+
+        std::string expected = "3,1,2,3";
+
+        ASSERT_EQ(wb.str(), expected);
+
+        DB::ReadBufferFromString rb(expected);
+
+        Cont deserialized;
+        deserialized.readText(rb);
+        ASSERT_EQ(convertToSet(cont), convertToSet(deserialized));
+    }
+    {
+        using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
+
+        Cont cont;
+
+        cont.insert(1);
+        cont.insert(2);
+        cont.insert(3);
+
+        DB::WriteBufferFromOwnString wb;
+        cont.write(wb);
+
+        DB::ReadBufferFromString rb(wb.str());
+
+        Cont deserialized;
+        deserialized.read(rb);
+        ASSERT_EQ(convertToSet(cont), convertToSet(deserialized));
+    }
+    {
+        using Cont = HashSet<int, DummyHash<int>, HashTableGrower<1>>;
+        Cont cont;
+
+        DB::WriteBufferFromOwnString wb;
+        cont.writeText(wb);
+
+        std::string expected = "0";
+        ASSERT_EQ(wb.str(), expected);
+
+        DB::ReadBufferFromString rb(expected);
+
+        Cont deserialized;
+        deserialized.readText(rb);
+        ASSERT_EQ(convertToSet(cont), convertToSet(deserialized));
+    }
+    {
+        using Cont = HashSet<DB::UInt128, DB::UInt128TrivialHash>;
+        Cont cont;
+
+        DB::WriteBufferFromOwnString wb;
+        cont.write(wb);
+
+        std::string expected;
+        expected += static_cast<char>(0);
+
+        ASSERT_EQ(wb.str(), expected);
+
+        DB::ReadBufferFromString rb(expected);
+
+        Cont deserialized;
+        deserialized.read(rb);
+        ASSERT_EQ(convertToSet(cont), convertToSet(deserialized));
+    }
+}
diff --git a/src/Common/tests/hash_table.cpp b/src/Common/tests/hash_table.cpp
deleted file mode 100644
index ebc22c5b5e5..00000000000
--- a/src/Common/tests/hash_table.cpp
+++ /dev/null
@@ -1,50 +0,0 @@
-#include <iostream>
-#include <iomanip>
-
-#include <Interpreters/AggregationCommon.h>
-
-#include <Common/HashTable/HashMap.h>
-#include <Common/HashTable/HashSet.h>
-
-
-int main(int, char **)
-{
-    {
-        using Cont = HashSet<int, DefaultHash<int>, HashTableGrower<1>>;
-        Cont cont;
-
-        cont.insert(1);
-        cont.insert(2);
-
-        Cont::LookupResult it;
-        bool inserted;
-        int key = 3;
-        cont.emplace(key, it, inserted);
-        std::cerr << inserted << ", " << key << std::endl;
-
-        cont.emplace(key, it, inserted);
-        std::cerr << inserted << ", " << key << std::endl;
-
-        for (auto x : cont)
-            std::cerr << x.getValue() << std::endl;
-
-        DB::WriteBufferFromOwnString wb;
-        cont.writeText(wb);
-
-        std::cerr << "dump: " << wb.str() << std::endl;
-    }
-
-    {
-        using Cont = HashSet<
-            DB::UInt128,
-            DB::UInt128TrivialHash>;
-        Cont cont;
-
-        DB::WriteBufferFromOwnString wb;
-        cont.write(wb);
-
-        std::cerr << "dump: " << wb.str() << std::endl;
-    }
-
-    return 0;
-}
diff --git a/src/Common/ya.make b/src/Common/ya.make
index e515741a272..c8be43f98a2 100644
--- a/src/Common/ya.make
+++ b/src/Common/ya.make
@@ -7,7 +7,7 @@ ADDINCL (
     GLOBAL clickhouse/src
     contrib/libs/libcpuid
     contrib/libs/libunwind/include
-    GLOBAL contrib/restricted/ryu
+    GLOBAL contrib/restricted/dragonbox
 )
 
 PEERDIR(
@@ -18,7 +18,7 @@ PEERDIR(
     contrib/libs/openssl
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/re2
-    contrib/restricted/ryu
+    contrib/restricted/dragonbox
 )
 
 INCLUDE(${ARCADIA_ROOT}/clickhouse/cmake/yandex/ya.make.versions.inc)
diff --git a/src/Common/ya.make.in b/src/Common/ya.make.in
index d913832b91b..210ecab6ef6 100644
--- a/src/Common/ya.make.in
+++ b/src/Common/ya.make.in
@@ -6,7 +6,7 @@ ADDINCL (
     GLOBAL clickhouse/src
     contrib/libs/libcpuid
     contrib/libs/libunwind/include
-    GLOBAL contrib/restricted/ryu
+    GLOBAL contrib/restricted/dragonbox
 )
 
 PEERDIR(
@@ -17,7 +17,7 @@ PEERDIR(
     contrib/libs/openssl
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/re2
-    contrib/restricted/ryu
+    contrib/restricted/dragonbox
 )
 
 INCLUDE(${ARCADIA_ROOT}/clickhouse/cmake/yandex/ya.make.versions.inc)
diff --git a/src/Core/AccurateComparison.h b/src/Core/AccurateComparison.h
index 500346872db..c533cf0c7fb 100644
--- a/src/Core/AccurateComparison.h
+++ b/src/Core/AccurateComparison.h
@@ -515,11 +515,32 @@ inline bool NO_SANITIZE_UNDEFINED convertNumeric(From value, To & result)
         return true;
     }
 
-    /// Note that NaNs doesn't compare equal to anything, but they are still in range of any Float type.
-    if (isNaN(value) && std::is_floating_point_v<To>)
+    if constexpr (std::is_floating_point_v<From> && std::is_floating_point_v<To>)
     {
-        result = value;
-        return true;
+        /// Note that NaNs doesn't compare equal to anything, but they are still in range of any Float type.
+        if (isNaN(value))
+        {
+            result = value;
+            return true;
+        }
+
+        if (value == std::numeric_limits<From>::infinity())
+        {
+            result = std::numeric_limits<To>::infinity();
+            return true;
+        }
+
+        if (value == -std::numeric_limits<From>::infinity())
+        {
+            result = -std::numeric_limits<To>::infinity();
+            return true;
+        }
+    }
+
+    if (accurate::greaterOp(value, std::numeric_limits<To>::max())
+        || accurate::greaterOp(std::numeric_limits<To>::lowest(), value))
+    {
+        return false;
     }
 
     result = static_cast<To>(value);
diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h
index b82cfd88e98..2b916cbf538 100644
--- a/src/Core/DecimalFunctions.h
+++ b/src/Core/DecimalFunctions.h
@@ -6,6 +6,7 @@
 #include <common/arithmeticOverflow.h>
 
 #include <limits>
+#include <type_traits>
 
 
 namespace DB
@@ -206,23 +207,32 @@ inline typename DecimalType::NativeType getFractionalPart(const DecimalType & de
 }
 
 /// Decimal to integer/float conversion
-template <typename To, typename DecimalType>
-To convertTo(const DecimalType & decimal, size_t scale)
+template <typename To, typename DecimalType, typename ReturnType>
+ReturnType convertToImpl(const DecimalType & decimal, size_t scale, To & result)
 {
     using NativeT = typename DecimalType::NativeType;
+    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
 
     if constexpr (std::is_floating_point_v<To>)
     {
-        return static_cast<To>(decimal.value) / static_cast<To>(scaleMultiplier<NativeT>(scale));
+        result = static_cast<To>(decimal.value) / static_cast<To>(scaleMultiplier<NativeT>(scale));
     }
     else if constexpr (is_integer_v<To> && (sizeof(To) >= sizeof(NativeT)))
     {
         NativeT whole = getWholePart(decimal, scale);
 
         if constexpr (is_unsigned_v<To>)
+        {
             if (whole < 0)
-                throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
-        return static_cast<To>(whole);
+            {
+                if constexpr (throw_exception)
+                    throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
+                else
+                    return ReturnType(true);
+            }
+        }
+
+        result = static_cast<To>(whole);
     }
     else if constexpr (is_integer_v<To>)
     {
@@ -235,9 +245,34 @@ To convertTo(const DecimalType & decimal, size_t scale)
         static const constexpr CastTo max_to = std::numeric_limits<ToNativeT>::max();
 
         if (whole < min_to || whole > max_to)
-            throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
-        return static_cast<CastTo>(whole);
+        {
+            if constexpr (throw_exception)
+                throw Exception("Convert overflow", ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return ReturnType(true);
+        }
+
+        result = static_cast<CastTo>(whole);
     }
+
+    return ReturnType(true);
+}
+
+
+template <typename To, typename DecimalType>
+To convertTo(const DecimalType & decimal, size_t scale)
+{
+    To result;
+
+    convertToImpl<To, DecimalType, void>(decimal, scale, result);
+
+    return result;
+}
+
+template <typename To, typename DecimalType>
+bool tryConvertTo(const DecimalType & decimal, size_t scale, To & result)
+{
+    return convertToImpl<To, DecimalType, bool>(decimal, scale, result);
 }
 
 template <bool is_multiply, bool is_division, typename T, typename U, template <typename> typename DecimalType>
diff --git a/src/Core/Field.cpp b/src/Core/Field.cpp
index 6fda3b39b3b..93107d7bb2c 100644
--- a/src/Core/Field.cpp
+++ b/src/Core/Field.cpp
@@ -17,6 +17,63 @@ namespace ErrorCodes
     extern const int DECIMAL_OVERFLOW;
 }
 
+inline Field getBinaryValue(UInt8 type, ReadBuffer & buf)
+{
+    switch (type)
+    {
+        case Field::Types::Null: {
+            return DB::Field();
+        }
+        case Field::Types::UInt64: {
+            UInt64 value;
+            DB::readVarUInt(value, buf);
+            return value;
+        }
+        case Field::Types::UInt128: {
+            UInt128 value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Int64: {
+            Int64 value;
+            DB::readVarInt(value, buf);
+            return value;
+        }
+        case Field::Types::Float64: {
+            Float64 value;
+            DB::readFloatBinary(value, buf);
+            return value;
+        }
+        case Field::Types::String: {
+            std::string value;
+            DB::readStringBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Array: {
+            Array value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Tuple: {
+            Tuple value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::Map: {
+            Map value;
+            DB::readBinary(value, buf);
+            return value;
+        }
+        case Field::Types::AggregateFunctionState: {
+            AggregateFunctionStateData value;
+            DB::readStringBinary(value.name, buf);
+            DB::readStringBinary(value.data, buf);
+            return value;
+        }
+    }
+    return DB::Field();
+}
+
 void readBinary(Array & x, ReadBuffer & buf)
 {
     size_t size;
@@ -25,73 +82,7 @@ void readBinary(Array & x, ReadBuffer & buf)
     DB::readBinary(size, buf);
 
     for (size_t index = 0; index < size; ++index)
-    {
-        switch (type)
-        {
-            case Field::Types::Null:
-            {
-                x.push_back(DB::Field());
-                break;
-            }
-            case Field::Types::UInt64:
-            {
-                UInt64 value;
-                DB::readVarUInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                UInt128 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                Int64 value;
-                DB::readVarInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                Float64 value;
-                DB::readFloatBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::String:
-            {
-                std::string value;
-                DB::readStringBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                Array value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                Tuple value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                AggregateFunctionStateData value;
-                DB::readStringBinary(value.name, buf);
-                DB::readStringBinary(value.data, buf);
-                x.push_back(value);
-                break;
-            }
-        }
-    }
+        x.push_back(getBinaryValue(type, buf));
 }
 
 void writeBinary(const Array & x, WriteBuffer & buf)
@@ -104,53 +95,7 @@ void writeBinary(const Array & x, WriteBuffer & buf)
     DB::writeBinary(size, buf);
 
     for (const auto & elem : x)
-    {
-        switch (type)
-        {
-            case Field::Types::Null: break;
-            case Field::Types::UInt64:
-            {
-                DB::writeVarUInt(get<UInt64>(elem), buf);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                DB::writeBinary(get<UInt128>(elem), buf);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                DB::writeVarInt(get<Int64>(elem), buf);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                DB::writeFloatBinary(get<Float64>(elem), buf);
-                break;
-            }
-            case Field::Types::String:
-            {
-                DB::writeStringBinary(get<std::string>(elem), buf);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                DB::writeBinary(get<Array>(elem), buf);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                DB::writeBinary(get<Tuple>(elem), buf);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().name, buf);
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().data, buf);
-                break;
-            }
-        }
-    }
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, elem);
 }
 
 void writeText(const Array & x, WriteBuffer & buf)
@@ -168,93 +113,7 @@ void readBinary(Tuple & x, ReadBuffer & buf)
     {
         UInt8 type;
         DB::readBinary(type, buf);
-
-        switch (type)
-        {
-            case Field::Types::Null:
-            {
-                x.push_back(DB::Field());
-                break;
-            }
-            case Field::Types::UInt64:
-            {
-                UInt64 value;
-                DB::readVarUInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                UInt128 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                Int64 value;
-                DB::readVarInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int128:
-            {
-                Int64 value;
-                DB::readVarInt(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                Float64 value;
-                DB::readFloatBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::String:
-            {
-                std::string value;
-                DB::readStringBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::UInt256:
-            {
-                UInt256 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Int256:
-            {
-                Int256 value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                Array value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                Tuple value;
-                DB::readBinary(value, buf);
-                x.push_back(value);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                AggregateFunctionStateData value;
-                DB::readStringBinary(value.name, buf);
-                DB::readStringBinary(value.data, buf);
-                x.push_back(value);
-                break;
-            }
-        }
+        x.push_back(getBinaryValue(type, buf));
     }
 }
 
@@ -267,67 +126,7 @@ void writeBinary(const Tuple & x, WriteBuffer & buf)
     {
         const UInt8 type = elem.getType();
         DB::writeBinary(type, buf);
-
-        switch (type)
-        {
-            case Field::Types::Null: break;
-            case Field::Types::UInt64:
-            {
-                DB::writeVarUInt(get<UInt64>(elem), buf);
-                break;
-            }
-            case Field::Types::UInt128:
-            {
-                DB::writeBinary(get<UInt128>(elem), buf);
-                break;
-            }
-            case Field::Types::Int64:
-            {
-                DB::writeVarInt(get<Int64>(elem), buf);
-                break;
-            }
-            case Field::Types::Int128:
-            {
-                DB::writeVarInt(get<Int64>(elem), buf);
-                break;
-            }
-            case Field::Types::Float64:
-            {
-                DB::writeFloatBinary(get<Float64>(elem), buf);
-                break;
-            }
-            case Field::Types::String:
-            {
-                DB::writeStringBinary(get<std::string>(elem), buf);
-                break;
-            }
-            case Field::Types::UInt256:
-            {
-                DB::writeBinary(get<UInt256>(elem), buf);
-                break;
-            }
-            case Field::Types::Int256:
-            {
-                DB::writeBinary(get<Int256>(elem), buf);
-                break;
-            }
-            case Field::Types::Array:
-            {
-                DB::writeBinary(get<Array>(elem), buf);
-                break;
-            }
-            case Field::Types::Tuple:
-            {
-                DB::writeBinary(get<Tuple>(elem), buf);
-                break;
-            }
-            case Field::Types::AggregateFunctionState:
-            {
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().name, buf);
-                DB::writeStringBinary(elem.get<AggregateFunctionStateData>().data, buf);
-                break;
-            }
-        }
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, elem);
     }
 }
 
@@ -336,6 +135,37 @@ void writeText(const Tuple & x, WriteBuffer & buf)
     writeFieldText(DB::Field(x), buf);
 }
 
+void readBinary(Map & x, ReadBuffer & buf)
+{
+    size_t size;
+    DB::readBinary(size, buf);
+
+    for (size_t index = 0; index < size; ++index)
+    {
+        UInt8 type;
+        DB::readBinary(type, buf);
+        x.push_back(getBinaryValue(type, buf));
+    }
+}
+
+void writeBinary(const Map & x, WriteBuffer & buf)
+{
+    const size_t size = x.size();
+    DB::writeBinary(size, buf);
+
+    for (const auto & elem : x)
+    {
+        const UInt8 type = elem.getType();
+        DB::writeBinary(type, buf);
+        Field::dispatch([&buf] (const auto & value) { DB::FieldVisitorWriteBinary()(value, buf); }, elem);
+    }
+}
+
+void writeText(const Map & x, WriteBuffer & buf)
+{
+    writeFieldText(DB::Field(x), buf);
+}
+
 template <typename T>
 void readQuoted(DecimalField<T> & x, ReadBuffer & buf)
 {
@@ -530,6 +360,30 @@ Field Field::restoreFromDump(const std::string_view & dump_)
         return tuple;
     }
 
+    prefix = std::string_view{"Map_("};
+    if (dump.starts_with(prefix))
+    {
+        std::string_view tail = dump.substr(prefix.length());
+        trimLeft(tail);
+        Map map;
+        while (tail != ")")
+        {
+            size_t separator = tail.find_first_of(",)");
+            if (separator == std::string_view::npos)
+                show_error();
+            bool comma = (tail[separator] == ',');
+            std::string_view element = tail.substr(0, separator);
+            tail.remove_prefix(separator);
+            if (comma)
+                tail.remove_prefix(1);
+            trimLeft(tail);
+            if (!comma && tail != ")")
+                show_error();
+            map.push_back(Field::restoreFromDump(element));
+        }
+        return map;
+    }
+
     prefix = std::string_view{"AggregateFunctionState_("};
     if (dump.starts_with(prefix))
     {
diff --git a/src/Core/Field.h b/src/Core/Field.h
index 33f3a984c5c..3a52186167f 100644
--- a/src/Core/Field.h
+++ b/src/Core/Field.h
@@ -51,6 +51,9 @@ struct X : public FieldVector \
 DEFINE_FIELD_VECTOR(Array);
 DEFINE_FIELD_VECTOR(Tuple);
 
+/// An array with the following structure: [(key1, value1), (key2, value2), ...]
+DEFINE_FIELD_VECTOR(Map);
+
 #undef DEFINE_FIELD_VECTOR
 
 struct AggregateFunctionStateData
@@ -206,6 +209,7 @@ template <> struct NearestFieldTypeImpl<std::string_view> { using Type = String;
 template <> struct NearestFieldTypeImpl<String> { using Type = String; };
 template <> struct NearestFieldTypeImpl<Array> { using Type = Array; };
 template <> struct NearestFieldTypeImpl<Tuple> { using Type = Tuple; };
+template <> struct NearestFieldTypeImpl<Map> { using Type = Map; };
 template <> struct NearestFieldTypeImpl<bool> { using Type = UInt64; };
 template <> struct NearestFieldTypeImpl<Null> { using Type = Null; };
 
@@ -259,6 +263,7 @@ public:
             Decimal256 = 23,
             UInt256 = 24,
             Int256  = 25,
+            Map = 26,
         };
 
         static const int MIN_NON_POD = 16;
@@ -276,6 +281,7 @@ public:
                 case String:  return "String";
                 case Array:   return "Array";
                 case Tuple:   return "Tuple";
+                case Map:     return "Map";
                 case Decimal32:  return "Decimal32";
                 case Decimal64:  return "Decimal64";
                 case Decimal128: return "Decimal128";
@@ -464,6 +470,7 @@ public:
             case Types::String:  return get<String>()  < rhs.get<String>();
             case Types::Array:   return get<Array>()   < rhs.get<Array>();
             case Types::Tuple:   return get<Tuple>()   < rhs.get<Tuple>();
+            case Types::Map:     return get<Map>()     < rhs.get<Map>();
             case Types::Decimal32:  return get<DecimalField<Decimal32>>()  < rhs.get<DecimalField<Decimal32>>();
             case Types::Decimal64:  return get<DecimalField<Decimal64>>()  < rhs.get<DecimalField<Decimal64>>();
             case Types::Decimal128: return get<DecimalField<Decimal128>>() < rhs.get<DecimalField<Decimal128>>();
@@ -499,6 +506,7 @@ public:
             case Types::String:  return get<String>()  <= rhs.get<String>();
             case Types::Array:   return get<Array>()   <= rhs.get<Array>();
             case Types::Tuple:   return get<Tuple>()   <= rhs.get<Tuple>();
+            case Types::Map:     return get<Map>()     <= rhs.get<Map>();
             case Types::Decimal32:  return get<DecimalField<Decimal32>>()  <= rhs.get<DecimalField<Decimal32>>();
             case Types::Decimal64:  return get<DecimalField<Decimal64>>()  <= rhs.get<DecimalField<Decimal64>>();
             case Types::Decimal128: return get<DecimalField<Decimal128>>() <= rhs.get<DecimalField<Decimal128>>();
@@ -536,6 +544,7 @@ public:
             case Types::String:  return get<String>()  == rhs.get<String>();
             case Types::Array:   return get<Array>()   == rhs.get<Array>();
             case Types::Tuple:   return get<Tuple>()   == rhs.get<Tuple>();
+            case Types::Map:     return get<Map>()     == rhs.get<Map>();
             case Types::UInt128: return get<UInt128>() == rhs.get<UInt128>();
             case Types::Int128:  return get<Int128>()  == rhs.get<Int128>();
             case Types::Decimal32:  return get<DecimalField<Decimal32>>()  == rhs.get<DecimalField<Decimal32>>();
@@ -575,6 +584,7 @@ public:
             case Types::String:  return f(field.template get<String>());
             case Types::Array:   return f(field.template get<Array>());
             case Types::Tuple:   return f(field.template get<Tuple>());
+            case Types::Map:     return f(field.template get<Map>());
             case Types::Decimal32:  return f(field.template get<DecimalField<Decimal32>>());
             case Types::Decimal64:  return f(field.template get<DecimalField<Decimal64>>());
             case Types::Decimal128: return f(field.template get<DecimalField<Decimal128>>());
@@ -600,7 +610,7 @@ public:
 
 private:
     std::aligned_union_t<DBMS_MIN_FIELD_SIZE - sizeof(Types::Which),
-        Null, UInt64, UInt128, Int64, Int128, Float64, String, Array, Tuple,
+        Null, UInt64, UInt128, Int64, Int128, Float64, String, Array, Tuple, Map,
         DecimalField<Decimal32>, DecimalField<Decimal64>, DecimalField<Decimal128>, DecimalField<Decimal256>,
         AggregateFunctionStateData,
         UInt256, Int256
@@ -699,6 +709,9 @@ private:
             case Types::Tuple:
                 destroy<Tuple>();
                 break;
+            case Types::Map:
+                destroy<Map>();
+                break;
             case Types::AggregateFunctionState:
                 destroy<AggregateFunctionStateData>();
                 break;
@@ -729,6 +742,7 @@ template <> struct Field::TypeToEnum<Float64> { static const Types::Which value
 template <> struct Field::TypeToEnum<String>  { static const Types::Which value = Types::String; };
 template <> struct Field::TypeToEnum<Array>   { static const Types::Which value = Types::Array; };
 template <> struct Field::TypeToEnum<Tuple>   { static const Types::Which value = Types::Tuple; };
+template <> struct Field::TypeToEnum<Map>     { static const Types::Which value = Types::Map; };
 template <> struct Field::TypeToEnum<DecimalField<Decimal32>>{ static const Types::Which value = Types::Decimal32; };
 template <> struct Field::TypeToEnum<DecimalField<Decimal64>>{ static const Types::Which value = Types::Decimal64; };
 template <> struct Field::TypeToEnum<DecimalField<Decimal128>>{ static const Types::Which value = Types::Decimal128; };
@@ -747,6 +761,7 @@ template <> struct Field::EnumToType<Field::Types::Float64> { using Type = Float
 template <> struct Field::EnumToType<Field::Types::String>  { using Type = String; };
 template <> struct Field::EnumToType<Field::Types::Array>   { using Type = Array; };
 template <> struct Field::EnumToType<Field::Types::Tuple>   { using Type = Tuple; };
+template <> struct Field::EnumToType<Field::Types::Map>     { using Type = Map; };
 template <> struct Field::EnumToType<Field::Types::Decimal32> { using Type = DecimalField<Decimal32>; };
 template <> struct Field::EnumToType<Field::Types::Decimal64> { using Type = DecimalField<Decimal64>; };
 template <> struct Field::EnumToType<Field::Types::Decimal128> { using Type = DecimalField<Decimal128>; };
@@ -814,6 +829,7 @@ T safeGet(Field & field)
 
 template <> struct TypeName<Array> { static std::string get() { return "Array"; } };
 template <> struct TypeName<Tuple> { static std::string get() { return "Tuple"; } };
+template <> struct TypeName<Map>   { static std::string get() { return "Map"; } };
 template <> struct TypeName<AggregateFunctionStateData> { static std::string get() { return "AggregateFunctionState"; } };
 
 template <typename T>
@@ -900,6 +916,12 @@ void writeBinary(const Tuple & x, WriteBuffer & buf);
 
 void writeText(const Tuple & x, WriteBuffer & buf);
 
+void readBinary(Map & x, ReadBuffer & buf);
+[[noreturn]] inline void readText(Map &, ReadBuffer &) { throw Exception("Cannot read Map.", ErrorCodes::NOT_IMPLEMENTED); }
+[[noreturn]] inline void readQuoted(Map &, ReadBuffer &) { throw Exception("Cannot read Map.", ErrorCodes::NOT_IMPLEMENTED); }
+void writeBinary(const Map & x, WriteBuffer & buf);
+void writeText(const Map & x, WriteBuffer & buf);
+[[noreturn]] inline void writeQuoted(const Map &, WriteBuffer &) { throw Exception("Cannot write Map quoted.", ErrorCodes::NOT_IMPLEMENTED); }
 
 __attribute__ ((noreturn)) inline void writeText(const AggregateFunctionStateData &, WriteBuffer &)
 {
diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp
index 4e59293c88d..2a8dd2acdc0 100644
--- a/src/Core/Settings.cpp
+++ b/src/Core/Settings.cpp
@@ -16,10 +16,8 @@ namespace ErrorCodes
     extern const int UNKNOWN_ELEMENT_IN_CONFIG;
 }
 
-
 IMPLEMENT_SETTINGS_TRAITS(SettingsTraits, LIST_OF_SETTINGS)
 
-
 /** Set the settings from the profile (in the server configuration, many settings can be listed in one profile).
     * The profile can also be set using the `set` functions, like the `profile` setting.
     */
diff --git a/src/Core/Settings.h b/src/Core/Settings.h
index 2cbe0c16cae..0555326022f 100644
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@@ -65,6 +65,7 @@ class IColumn;
     M(UInt64, distributed_connections_pool_size, DBMS_DEFAULT_DISTRIBUTED_CONNECTIONS_POOL_SIZE, "Maximum number of connections with one remote server in the pool.", 0) \
     M(UInt64, connections_with_failover_max_tries, DBMS_CONNECTION_POOL_WITH_FAILOVER_DEFAULT_MAX_TRIES, "The maximum number of attempts to connect to replicas.", 0) \
     M(UInt64, s3_min_upload_part_size, 512*1024*1024, "The minimum size of part to upload during multipart upload to S3.", 0) \
+    M(UInt64, s3_max_single_part_upload_size, 64*1024*1024, "The maximum size of object to upload using singlepart upload to S3.", 0) \
     M(UInt64, s3_max_redirects, 10, "Max number of S3 redirects hops allowed.", 0) \
     M(Bool, extremes, false, "Calculate minimums and maximums of the result columns. They can be output in JSON-formats.", IMPORTANT) \
     M(Bool, use_uncompressed_cache, true, "Whether to use the cache of uncompressed blocks.", 0) \
@@ -165,7 +166,7 @@ class IColumn;
     \
     M(UInt64, insert_quorum, 0, "For INSERT queries in the replicated table, wait writing for the specified number of replicas and linearize the addition of the data. 0 - disabled.", 0) \
     M(Milliseconds, insert_quorum_timeout, 600000, "", 0) \
-    M(Bool, insert_quorum_parallel, false, "For quorum INSERT queries - enable to make parallel inserts without linearizability", 0) \
+    M(Bool, insert_quorum_parallel, true, "For quorum INSERT queries - enable to make parallel inserts without linearizability", 0) \
     M(UInt64, select_sequential_consistency, 0, "For SELECT queries from the replicated table, throw an exception if the replica does not have a chunk written with the quorum; do not read the parts that have not yet been written with the quorum.", 0) \
     M(UInt64, table_function_remote_max_addresses, 1000, "The maximum number of different shards and the maximum number of replicas of one shard in the `remote` function.", 0) \
     M(Milliseconds, read_backoff_min_latency_ms, 1000, "Setting to reduce the number of threads in case of slow reads. Pay attention only to reads that took at least that much time.", 0) \
@@ -238,6 +239,8 @@ class IColumn;
       * Almost all limits apply to each stream individually. \
       */ \
     \
+    M(UInt64, limit, 0, "Limit on read rows from the most 'end' result for select query, default 0 means no limit length", 0) \
+    M(UInt64, offset, 0, "Offset on read rows from the most 'end' result for select query", 0) \
     M(UInt64, max_rows_to_read, 0, "Limit on read rows from the most 'deep' sources. That is, only in the deepest subquery. When reading from a remote server, it is only checked on a remote server.", 0) \
     M(UInt64, max_bytes_to_read, 0, "Limit on read bytes (after decompression) from the most 'deep' sources. That is, only in the deepest subquery. When reading from a remote server, it is only checked on a remote server.", 0) \
     M(OverflowMode, read_overflow_mode, OverflowMode::THROW, "What to do when the limit is exceeded.", 0) \
@@ -255,6 +258,7 @@ class IColumn;
     M(OverflowMode, sort_overflow_mode, OverflowMode::THROW, "What to do when the limit is exceeded.", 0) \
     M(UInt64, max_bytes_before_external_sort, 0, "", 0) \
     M(UInt64, max_bytes_before_remerge_sort, 1000000000, "In case of ORDER BY with LIMIT, when memory usage is higher than specified threshold, perform additional steps of merging blocks before final merge to keep just top LIMIT rows.", 0) \
+    M(Float, remerge_sort_lowered_memory_bytes_ratio, 2., "If memory usage after remerge does not reduced by this ratio, remerge will be disabled.", 0) \
     \
     M(UInt64, max_result_rows, 0, "Limit on result size in rows. Also checked for intermediate data sent from remote servers.", 0) \
     M(UInt64, max_result_bytes, 0, "Limit on result size in bytes (uncompressed). Also checked for intermediate data sent from remote servers.", 0) \
@@ -399,6 +403,8 @@ class IColumn;
     M(Bool, enable_global_with_statement, false, "Propagate WITH statements to UNION queries and all subqueries", 0) \
     M(Bool, aggregate_functions_null_for_empty, false, "Rewrite all aggregate functions in a query, adding -OrNull suffix to them", 0) \
     M(Bool, optimize_skip_merged_partitions, false, "Skip partitions with one part with level > 0 in optimize final", 0) \
+    M(Bool, optimize_on_insert, true, "Do the same transformation for inserted block of data as if merge was done on this block.", 0) \
+    M(Bool, allow_experimental_map_type, false, "Allow data type Map", 0) \
     \
     M(Bool, use_antlr_parser, false, "Parse incoming queries using ANTLR-generated parser", 0) \
     \
@@ -406,8 +412,6 @@ class IColumn;
     \
     M(UInt64, max_memory_usage_for_all_queries, 0, "Obsolete. Will be removed after 2020-10-20", 0) \
     M(UInt64, multiple_joins_rewriter_version, 0, "Obsolete setting, does nothing. Will be removed after 2021-03-31", 0) \
-    M(Bool, experimental_use_processors, true, "Obsolete setting, does nothing. Will be removed after 2020-11-29.", 0) \
-    M(Bool, force_optimize_skip_unused_shards_no_nested, false, "Obsolete setting, does nothing. Will be removed after 2020-12-01. Use force_optimize_skip_unused_shards_nesting instead.", 0) \
     M(Bool, enable_debug_queries, false, "Enabled debug queries, but now is obsolete", 0) \
     M(Bool, allow_experimental_database_atomic, true, "Obsolete setting, does nothing. Will be removed after 2021-02-12", 0) \
     M(UnionMode, union_default_mode, UnionMode::DISTINCT, "Set default Union Mode in SelectWithUnion query. Possible values: empty string, 'ALL', 'DISTINCT'. If empty, query without Union Mode will throw exception.", 0)
diff --git a/src/Core/SortCursor.h b/src/Core/SortCursor.h
index 7a222f70199..f383c3ded8e 100644
--- a/src/Core/SortCursor.h
+++ b/src/Core/SortCursor.h
@@ -30,7 +30,6 @@ struct SortCursorImpl
     ColumnRawPtrs all_columns;
     SortDescription desc;
     size_t sort_columns_size = 0;
-    size_t pos = 0;
     size_t rows = 0;
 
     /** Determines order if comparing columns are equal.
@@ -49,15 +48,20 @@ struct SortCursorImpl
     /** Is there at least one column with Collator. */
     bool has_collation = false;
 
+    /** We could use SortCursorImpl in case when columns aren't sorted
+      *  but we have their sorted permutation
+      */
+    IColumn::Permutation * permutation = nullptr;
+
     SortCursorImpl() {}
 
-    SortCursorImpl(const Block & block, const SortDescription & desc_, size_t order_ = 0)
+    SortCursorImpl(const Block & block, const SortDescription & desc_, size_t order_ = 0, IColumn::Permutation * perm = nullptr)
         : desc(desc_), sort_columns_size(desc.size()), order(order_), need_collation(desc.size())
     {
-        reset(block);
+        reset(block, perm);
     }
 
-    SortCursorImpl(const Columns & columns, const SortDescription & desc_, size_t order_ = 0)
+    SortCursorImpl(const Columns & columns, const SortDescription & desc_, size_t order_ = 0, IColumn::Permutation * perm = nullptr)
         : desc(desc_), sort_columns_size(desc.size()), order(order_), need_collation(desc.size())
     {
         for (auto & column_desc : desc)
@@ -66,19 +70,19 @@ struct SortCursorImpl
                 throw Exception("SortDescription should contain column position if SortCursor was used without header.",
                         ErrorCodes::LOGICAL_ERROR);
         }
-        reset(columns, {});
+        reset(columns, {}, perm);
     }
 
     bool empty() const { return rows == 0; }
 
     /// Set the cursor to the beginning of the new block.
-    void reset(const Block & block)
+    void reset(const Block & block, IColumn::Permutation * perm = nullptr)
     {
-        reset(block.getColumns(), block);
+        reset(block.getColumns(), block, perm);
     }
 
     /// Set the cursor to the beginning of the new block.
-    void reset(const Columns & columns, const Block & block)
+    void reset(const Columns & columns, const Block & block, IColumn::Permutation * perm = nullptr)
     {
         all_columns.clear();
         sort_columns.clear();
@@ -96,18 +100,33 @@ struct SortCursorImpl
                                    : column_desc.column_number;
             sort_columns.push_back(columns[column_number].get());
 
-            need_collation[j] = desc[j].collator != nullptr && sort_columns.back()->isCollationSupported();    /// TODO Nullable(String)
+            need_collation[j] = desc[j].collator != nullptr && sort_columns.back()->isCollationSupported();
             has_collation |= need_collation[j];
         }
 
         pos = 0;
         rows = all_columns[0]->size();
+        permutation = perm;
     }
 
+    size_t getRow() const
+    {
+        if (permutation)
+            return (*permutation)[pos];
+        return pos;
+    }
+
+    /// We need a possibility to change pos (see MergeJoin).
+    size_t & getPosRef() { return pos; }
+
     bool isFirst() const { return pos == 0; }
     bool isLast() const { return pos + 1 >= rows; }
     bool isValid() const { return pos < rows; }
     void next() { ++pos; }
+
+/// Prevent using pos instead of getRow()
+private:
+    size_t pos;
 };
 
 using SortCursorImpls = std::vector<SortCursorImpl>;
@@ -127,7 +146,7 @@ struct SortCursorHelper
 
     bool ALWAYS_INLINE greater(const SortCursorHelper & rhs) const
     {
-        return derived().greaterAt(rhs.derived(), impl->pos, rhs.impl->pos);
+        return derived().greaterAt(rhs.derived(), impl->getRow(), rhs.impl->getRow());
     }
 
     /// Inverted so that the priority queue elements are removed in ascending order.
diff --git a/src/Core/Types.h b/src/Core/Types.h
index 84c676b2204..0d578e17517 100644
--- a/src/Core/Types.h
+++ b/src/Core/Types.h
@@ -56,6 +56,7 @@ enum class TypeIndex
     Function,
     AggregateFunction,
     LowCardinality,
+    Map,
 };
 #if !__clang__
 #pragma GCC diagnostic pop
@@ -267,6 +268,7 @@ inline constexpr const char * getTypeName(TypeIndex idx)
         case TypeIndex::Function:   return "Function";
         case TypeIndex::AggregateFunction: return "AggregateFunction";
         case TypeIndex::LowCardinality: return "LowCardinality";
+        case TypeIndex::Map:        return "Map";
     }
 
     __builtin_unreachable();
diff --git a/src/DataStreams/MergingSortedBlockInputStream.cpp b/src/DataStreams/MergingSortedBlockInputStream.cpp
index ee6b93dc8e4..b7396a23d6a 100644
--- a/src/DataStreams/MergingSortedBlockInputStream.cpp
+++ b/src/DataStreams/MergingSortedBlockInputStream.cpp
@@ -222,7 +222,7 @@ void MergingSortedBlockInputStream::merge(MutableColumns & merged_columns, TSort
 //        std::cerr << "total_merged_rows: " << total_merged_rows << ", merged_rows: " << merged_rows << "\n";
 //        std::cerr << "Inserting row\n";
         for (size_t i = 0; i < num_columns; ++i)
-            merged_columns[i]->insertFrom(*current->all_columns[i], current->pos);
+            merged_columns[i]->insertFrom(*current->all_columns[i], current->getRow());
 
         if (out_row_sources_buf)
         {
diff --git a/src/DataStreams/RemoteBlockInputStream.cpp b/src/DataStreams/RemoteBlockInputStream.cpp
index c7c5ce2d00a..a62f7fca0b7 100644
--- a/src/DataStreams/RemoteBlockInputStream.cpp
+++ b/src/DataStreams/RemoteBlockInputStream.cpp
@@ -6,27 +6,27 @@ namespace DB
 
 RemoteBlockInputStream::RemoteBlockInputStream(
         Connection & connection,
-        const String & query_, const Block & header_, const Context & context_, const Settings * settings,
+        const String & query_, const Block & header_, const Context & context_,
         const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_, QueryProcessingStage::Enum stage_)
-    : query_executor(connection, query_, header_, context_, settings, throttler, scalars_, external_tables_, stage_)
+    : query_executor(connection, query_, header_, context_, throttler, scalars_, external_tables_, stage_)
 {
     init();
 }
 
 RemoteBlockInputStream::RemoteBlockInputStream(
         std::vector<IConnectionPool::Entry> && connections,
-        const String & query_, const Block & header_, const Context & context_, const Settings * settings,
+        const String & query_, const Block & header_, const Context & context_,
         const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_, QueryProcessingStage::Enum stage_)
-    : query_executor(std::move(connections), query_, header_, context_, settings, throttler, scalars_, external_tables_, stage_)
+    : query_executor(std::move(connections), query_, header_, context_, throttler, scalars_, external_tables_, stage_)
 {
     init();
 }
 
 RemoteBlockInputStream::RemoteBlockInputStream(
         const ConnectionPoolWithFailoverPtr & pool,
-        const String & query_, const Block & header_, const Context & context_, const Settings * settings,
+        const String & query_, const Block & header_, const Context & context_,
         const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_, QueryProcessingStage::Enum stage_)
-    : query_executor(pool, query_, header_, context_, settings, throttler, scalars_, external_tables_, stage_)
+    : query_executor(pool, query_, header_, context_, throttler, scalars_, external_tables_, stage_)
 {
     init();
 }
diff --git a/src/DataStreams/RemoteBlockInputStream.h b/src/DataStreams/RemoteBlockInputStream.h
index 628feb0ab80..5ef05ee99eb 100644
--- a/src/DataStreams/RemoteBlockInputStream.h
+++ b/src/DataStreams/RemoteBlockInputStream.h
@@ -6,7 +6,6 @@
 
 #include <DataStreams/IBlockInputStream.h>
 #include <Common/Throttler.h>
-#include <Interpreters/Context.h>
 #include <Client/ConnectionPool.h>
 #include <Client/MultiplexedConnections.h>
 #include <Interpreters/Cluster.h>
@@ -16,32 +15,31 @@
 namespace DB
 {
 
+class Context;
+
 /** This class allows one to launch queries on remote replicas of one shard and get results
   */
 class RemoteBlockInputStream : public IBlockInputStream
 {
 public:
     /// Takes already set connection.
-    /// If `settings` is nullptr, settings will be taken from context.
     RemoteBlockInputStream(
             Connection & connection,
-            const String & query_, const Block & header_, const Context & context_, const Settings * settings = nullptr,
+            const String & query_, const Block & header_, const Context & context_,
             const ThrottlerPtr & throttler = nullptr, const Scalars & scalars_ = Scalars(), const Tables & external_tables_ = Tables(),
             QueryProcessingStage::Enum stage_ = QueryProcessingStage::Complete);
 
     /// Accepts several connections already taken from pool.
-    /// If `settings` is nullptr, settings will be taken from context.
     RemoteBlockInputStream(
             std::vector<IConnectionPool::Entry> && connections,
-            const String & query_, const Block & header_, const Context & context_, const Settings * settings = nullptr,
+            const String & query_, const Block & header_, const Context & context_,
             const ThrottlerPtr & throttler = nullptr, const Scalars & scalars_ = Scalars(), const Tables & external_tables_ = Tables(),
             QueryProcessingStage::Enum stage_ = QueryProcessingStage::Complete);
 
     /// Takes a pool and gets one or several connections from it.
-    /// If `settings` is nullptr, settings will be taken from context.
     RemoteBlockInputStream(
             const ConnectionPoolWithFailoverPtr & pool,
-            const String & query_, const Block & header_, const Context & context_, const Settings * settings = nullptr,
+            const String & query_, const Block & header_, const Context & context_,
             const ThrottlerPtr & throttler = nullptr, const Scalars & scalars_ = Scalars(), const Tables & external_tables_ = Tables(),
             QueryProcessingStage::Enum stage_ = QueryProcessingStage::Complete);
 
diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 9abce0edef1..c38f42893af 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -8,7 +8,9 @@
 #include <Storages/SelectQueryInfo.h>
 #include <Interpreters/castColumn.h>
 #include <Interpreters/Cluster.h>
+#include <Interpreters/Context.h>
 #include <Interpreters/InternalTextLogsQueue.h>
+#include <IO/ConnectionTimeoutsContext.h>
 
 namespace DB
 {
@@ -20,14 +22,11 @@ namespace ErrorCodes
 
 RemoteQueryExecutor::RemoteQueryExecutor(
     Connection & connection,
-    const String & query_, const Block & header_, const Context & context_, const Settings * settings,
+    const String & query_, const Block & header_, const Context & context_,
     ThrottlerPtr throttler, const Scalars & scalars_, const Tables & external_tables_, QueryProcessingStage::Enum stage_)
     : header(header_), query(query_), context(context_)
     , scalars(scalars_), external_tables(external_tables_), stage(stage_)
 {
-    if (settings)
-        context.setSettings(*settings);
-
     create_multiplexed_connections = [this, &connection, throttler]()
     {
         return std::make_unique<MultiplexedConnections>(connection, context.getSettingsRef(), throttler);
@@ -36,14 +35,11 @@ RemoteQueryExecutor::RemoteQueryExecutor(
 
 RemoteQueryExecutor::RemoteQueryExecutor(
     std::vector<IConnectionPool::Entry> && connections,
-    const String & query_, const Block & header_, const Context & context_, const Settings * settings,
+    const String & query_, const Block & header_, const Context & context_,
     const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_, QueryProcessingStage::Enum stage_)
     : header(header_), query(query_), context(context_)
     , scalars(scalars_), external_tables(external_tables_), stage(stage_)
 {
-    if (settings)
-        context.setSettings(*settings);
-
     create_multiplexed_connections = [this, connections, throttler]() mutable
     {
         return std::make_unique<MultiplexedConnections>(
@@ -53,14 +49,11 @@ RemoteQueryExecutor::RemoteQueryExecutor(
 
 RemoteQueryExecutor::RemoteQueryExecutor(
     const ConnectionPoolWithFailoverPtr & pool,
-    const String & query_, const Block & header_, const Context & context_, const Settings * settings,
+    const String & query_, const Block & header_, const Context & context_,
     const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_, QueryProcessingStage::Enum stage_)
     : header(header_), query(query_), context(context_)
     , scalars(scalars_), external_tables(external_tables_), stage(stage_)
 {
-    if (settings)
-        context.setSettings(*settings);
-
     create_multiplexed_connections = [this, pool, throttler]()
     {
         const Settings & current_settings = context.getSettingsRef();
@@ -147,7 +140,7 @@ void RemoteQueryExecutor::sendQuery()
 
     multiplexed_connections = create_multiplexed_connections();
 
-    const auto& settings = context.getSettingsRef();
+    const auto & settings = context.getSettingsRef();
     if (settings.skip_unavailable_shards && 0 == multiplexed_connections->size())
         return;
 
diff --git a/src/DataStreams/RemoteQueryExecutor.h b/src/DataStreams/RemoteQueryExecutor.h
index 0db0e0218be..cec92a5f0e9 100644
--- a/src/DataStreams/RemoteQueryExecutor.h
+++ b/src/DataStreams/RemoteQueryExecutor.h
@@ -1,12 +1,16 @@
 #pragma once
 
-#include <Interpreters/Context.h>
 #include <Client/ConnectionPool.h>
 #include <Client/MultiplexedConnections.h>
+#include <Storages/IStorage_fwd.h>
+#include <Interpreters/Context.h>
+#include <Interpreters/StorageID.h>
 
 namespace DB
 {
 
+class Context;
+
 class Throttler;
 using ThrottlerPtr = std::shared_ptr<Throttler>;
 
@@ -21,26 +25,23 @@ class RemoteQueryExecutor
 {
 public:
     /// Takes already set connection.
-    /// If `settings` is nullptr, settings will be taken from context.
     RemoteQueryExecutor(
         Connection & connection,
-        const String & query_, const Block & header_, const Context & context_, const Settings * settings = nullptr,
+        const String & query_, const Block & header_, const Context & context_,
         ThrottlerPtr throttler_ = nullptr, const Scalars & scalars_ = Scalars(), const Tables & external_tables_ = Tables(),
         QueryProcessingStage::Enum stage_ = QueryProcessingStage::Complete);
 
     /// Accepts several connections already taken from pool.
-    /// If `settings` is nullptr, settings will be taken from context.
     RemoteQueryExecutor(
         std::vector<IConnectionPool::Entry> && connections,
-        const String & query_, const Block & header_, const Context & context_, const Settings * settings = nullptr,
+        const String & query_, const Block & header_, const Context & context_,
         const ThrottlerPtr & throttler = nullptr, const Scalars & scalars_ = Scalars(), const Tables & external_tables_ = Tables(),
         QueryProcessingStage::Enum stage_ = QueryProcessingStage::Complete);
 
     /// Takes a pool and gets one or several connections from it.
-    /// If `settings` is nullptr, settings will be taken from context.
     RemoteQueryExecutor(
         const ConnectionPoolWithFailoverPtr & pool,
-        const String & query_, const Block & header_, const Context & context_, const Settings * settings = nullptr,
+        const String & query_, const Block & header_, const Context & context_,
         const ThrottlerPtr & throttler = nullptr, const Scalars & scalars_ = Scalars(), const Tables & external_tables_ = Tables(),
         QueryProcessingStage::Enum stage_ = QueryProcessingStage::Complete);
 
diff --git a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp
index d82adf33108..6b2f94aa7b5 100644
--- a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp
+++ b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.cpp
@@ -25,10 +25,19 @@ namespace ErrorCodes
     extern const int LOGICAL_ERROR;
 }
 
-static const std::vector<String> supported_functions{"any", "anyLast", "min",
-    "max", "sum", "sumWithOverflow", "groupBitAnd", "groupBitOr", "groupBitXor",
-    "sumMap", "minMap", "maxMap", "groupArrayArray", "groupUniqArrayArray"};
+void DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(const AggregateFunctionPtr & function)
+{
+    static const std::vector<String> supported_functions{"any", "anyLast", "min",
+        "max", "sum", "sumWithOverflow", "groupBitAnd", "groupBitOr", "groupBitXor",
+        "sumMap", "minMap", "maxMap", "groupArrayArray", "groupUniqArrayArray"};
 
+    // check function
+    if (std::find(std::begin(supported_functions), std::end(supported_functions), function->getName()) == std::end(supported_functions))
+    {
+        throw Exception("Unsupported aggregate function " + function->getName() + ", supported functions are " + boost::algorithm::join(supported_functions, ","),
+                ErrorCodes::BAD_ARGUMENTS);
+    }
+}
 
 String DataTypeCustomSimpleAggregateFunction::getName() const
 {
@@ -114,12 +123,7 @@ static std::pair<DataTypePtr, DataTypeCustomDescPtr> create(const ASTPtr & argum
     AggregateFunctionProperties properties;
     function = AggregateFunctionFactory::instance().get(function_name, argument_types, params_row, properties);
 
-    // check function
-    if (std::find(std::begin(supported_functions), std::end(supported_functions), function->getName()) == std::end(supported_functions))
-    {
-        throw Exception("Unsupported aggregate function " + function->getName() + ", supported functions are " + boost::algorithm::join(supported_functions, ","),
-                        ErrorCodes::BAD_ARGUMENTS);
-    }
+    DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(function);
 
     DataTypePtr storage_type = DataTypeFactory::instance().get(argument_types[0]->getName());
 
diff --git a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h
index 0b75cd0ef88..44b007ab07f 100644
--- a/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h
+++ b/src/DataTypes/DataTypeCustomSimpleAggregateFunction.h
@@ -37,6 +37,7 @@ public:
 
     const AggregateFunctionPtr getFunction() const { return function; }
     String getName() const override;
+    static void checkSupportedFunctions(const AggregateFunctionPtr & function);
 };
 
 }
diff --git a/src/DataTypes/DataTypeFactory.cpp b/src/DataTypes/DataTypeFactory.cpp
index 5052a065163..1ff00d97e84 100644
--- a/src/DataTypes/DataTypeFactory.cpp
+++ b/src/DataTypes/DataTypeFactory.cpp
@@ -180,6 +180,7 @@ DataTypeFactory::DataTypeFactory()
     registerDataTypeDomainIPv4AndIPv6(*this);
     registerDataTypeDomainSimpleAggregateFunction(*this);
     registerDataTypeDomainGeo(*this);
+    registerDataTypeMap(*this);
 }
 
 DataTypeFactory & DataTypeFactory::instance()
diff --git a/src/DataTypes/DataTypeFactory.h b/src/DataTypes/DataTypeFactory.h
index ea77c50170c..192b1beae5d 100644
--- a/src/DataTypes/DataTypeFactory.h
+++ b/src/DataTypes/DataTypeFactory.h
@@ -73,6 +73,7 @@ void registerDataTypeFixedString(DataTypeFactory & factory);
 void registerDataTypeEnum(DataTypeFactory & factory);
 void registerDataTypeArray(DataTypeFactory & factory);
 void registerDataTypeTuple(DataTypeFactory & factory);
+void registerDataTypeMap(DataTypeFactory & factory);
 void registerDataTypeNullable(DataTypeFactory & factory);
 void registerDataTypeNothing(DataTypeFactory & factory);
 void registerDataTypeUUID(DataTypeFactory & factory);
diff --git a/src/DataTypes/DataTypeMap.cpp b/src/DataTypes/DataTypeMap.cpp
new file mode 100644
index 00000000000..2047928721a
--- /dev/null
+++ b/src/DataTypes/DataTypeMap.cpp
@@ -0,0 +1,375 @@
+#include <Common/StringUtils/StringUtils.h>
+#include <Columns/ColumnMap.h>
+#include <Columns/ColumnArray.h>
+#include <Core/Field.h>
+#include <Formats/FormatSettings.h>
+#include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypeArray.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeFactory.h>
+#include <Parsers/IAST.h>
+#include <Parsers/ASTNameTypePair.h>
+#include <Common/typeid_cast.h>
+#include <Common/assert_cast.h>
+#include <Common/quoteString.h>
+#include <IO/WriteHelpers.h>
+#include <IO/ReadHelpers.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/ReadBufferFromString.h>
+#include <IO/Operators.h>
+
+#include <ext/map.h>
+#include <ext/enumerate.h>
+
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+    extern const int CANNOT_READ_MAP_FROM_TEXT;
+}
+
+
+DataTypeMap::DataTypeMap(const DataTypes & elems_)
+{
+    assert(elems_.size() == 2);
+    key_type = elems_[0];
+    value_type = elems_[1];
+
+    nested = std::make_shared<DataTypeArray>(
+        std::make_shared<DataTypeTuple>(DataTypes{key_type, value_type}, Names{"keys", "values"}));
+}
+
+DataTypeMap::DataTypeMap(const DataTypePtr & key_type_, const DataTypePtr & value_type_)
+    : key_type(key_type_), value_type(value_type_)
+    , nested(std::make_shared<DataTypeArray>(
+        std::make_shared<DataTypeTuple>(DataTypes{key_type_, value_type_}, Names{"keys", "values"}))) {}
+
+std::string DataTypeMap::doGetName() const
+{
+    WriteBufferFromOwnString s;
+    s << "Map(" << key_type->getName() << "," << value_type->getName() << ")";
+
+    return s.str();
+}
+
+static const IColumn & extractNestedColumn(const IColumn & column)
+{
+    return assert_cast<const ColumnMap &>(column).getNestedColumn();
+}
+
+static IColumn & extractNestedColumn(IColumn & column)
+{
+    return assert_cast<ColumnMap &>(column).getNestedColumn();
+}
+
+void DataTypeMap::serializeBinary(const Field & field, WriteBuffer & ostr) const
+{
+    const auto & map = get<const Map &>(field);
+    writeVarUInt(map.size(), ostr);
+    for (const auto & elem : map)
+    {
+        const auto & tuple = elem.safeGet<const Tuple>();
+        assert(tuple.size() == 2);
+        key_type->serializeBinary(tuple[0], ostr);
+        value_type->serializeBinary(tuple[1], ostr);
+    }
+}
+
+void DataTypeMap::deserializeBinary(Field & field, ReadBuffer & istr) const
+{
+    size_t size;
+    readVarUInt(size, istr);
+    field = Map(size);
+    for (auto & elem : field.get<Map &>())
+    {
+        Tuple tuple(2);
+        key_type->deserializeBinary(tuple[0], istr);
+        value_type->deserializeBinary(tuple[1], istr);
+        elem = std::move(tuple);
+    }
+}
+
+void DataTypeMap::serializeBinary(const IColumn & column, size_t row_num, WriteBuffer & ostr) const
+{
+    nested->serializeBinary(extractNestedColumn(column), row_num, ostr);
+}
+
+void DataTypeMap::deserializeBinary(IColumn & column, ReadBuffer & istr) const
+{
+    nested->deserializeBinary(extractNestedColumn(column), istr);
+}
+
+
+template <typename Writer>
+void DataTypeMap::serializeTextImpl(const IColumn & column, size_t row_num, WriteBuffer & ostr, Writer && writer) const
+{
+    const auto & column_map = assert_cast<const ColumnMap &>(column);
+
+    const auto & nested_array = column_map.getNestedColumn();
+    const auto & nested_tuple = column_map.getNestedData();
+    const auto & offsets = nested_array.getOffsets();
+
+    size_t offset = offsets[row_num - 1];
+    size_t next_offset = offsets[row_num];
+
+    writeChar('{', ostr);
+    for (size_t i = offset; i < next_offset; ++i)
+    {
+        if (i != offset)
+            writeChar(',', ostr);
+        writer(key_type, nested_tuple.getColumn(0), i);
+        writeChar(':', ostr);
+        writer(value_type, nested_tuple.getColumn(1), i);
+    }
+    writeChar('}', ostr);
+}
+
+template <typename Reader>
+void DataTypeMap::deserializeTextImpl(IColumn & column, ReadBuffer & istr, bool need_safe_get_int_key, Reader && reader) const
+{
+    auto & column_map = assert_cast<ColumnMap &>(column);
+
+    auto & nested_array = column_map.getNestedColumn();
+    auto & nested_tuple = column_map.getNestedData();
+    auto & offsets = nested_array.getOffsets();
+
+    auto & key_column = nested_tuple.getColumn(0);
+    auto & value_column = nested_tuple.getColumn(1);
+
+    size_t size = 0;
+    assertChar('{', istr);
+
+    try
+    {
+        bool first = true;
+        while (!istr.eof() && *istr.position() != '}')
+        {
+            if (!first)
+            {
+                if (*istr.position() == ',')
+                    ++istr.position();
+                else
+                    throw Exception("Cannot read Map from text", ErrorCodes::CANNOT_READ_MAP_FROM_TEXT);
+            }
+
+            first = false;
+
+            skipWhitespaceIfAny(istr);
+
+            if (*istr.position() == '}')
+                break;
+
+            if (need_safe_get_int_key)
+            {
+                ReadBuffer::Position tmp = istr.position();
+                while (*tmp != ':' && *tmp != '}')
+                    ++tmp;
+                *tmp = ' ';
+                reader(key_type, key_column);
+            }
+            else
+            {
+                reader(key_type, key_column);
+                skipWhitespaceIfAny(istr);
+                assertChar(':', istr);
+            }
+
+            ++size;
+            skipWhitespaceIfAny(istr);
+            reader(value_type, value_column);
+
+            skipWhitespaceIfAny(istr);
+        }
+
+        offsets.push_back(offsets.back() + size);
+        assertChar('}', istr);
+    }
+    catch (...)
+    {
+        throw;
+    }
+}
+
+void DataTypeMap::serializeText(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    serializeTextImpl(column, row_num, ostr,
+        [&](const DataTypePtr & subcolumn_type, const IColumn & subcolumn, size_t pos)
+        {
+            subcolumn_type->serializeAsTextQuoted(subcolumn, pos, ostr, settings);
+        });
+}
+
+void DataTypeMap::deserializeText(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    // need_safe_get_int_key is set for Interger to prevent to readIntTextUnsafe
+    bool need_safe_get_int_key = isInteger(key_type);
+
+    deserializeTextImpl(column, istr, need_safe_get_int_key,
+        [&](const DataTypePtr & subcolumn_type, IColumn & subcolumn)
+        {
+            subcolumn_type->deserializeAsTextQuoted(subcolumn, istr, settings);
+        });
+}
+
+
+void DataTypeMap::serializeTextJSON(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    serializeTextImpl(column, row_num, ostr,
+        [&](const DataTypePtr & subcolumn_type, const IColumn & subcolumn, size_t pos)
+        {
+            subcolumn_type->serializeAsTextJSON(subcolumn, pos, ostr, settings);
+        });
+}
+
+void DataTypeMap::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    // need_safe_get_int_key is set for Interger to prevent to readIntTextUnsafe
+    bool need_safe_get_int_key = isInteger(key_type);
+
+    deserializeTextImpl(column, istr, need_safe_get_int_key,
+        [&](const DataTypePtr & subcolumn_type, IColumn & subcolumn)
+        {
+            subcolumn_type->deserializeAsTextJSON(subcolumn, istr, settings);
+        });
+}
+
+void DataTypeMap::serializeTextXML(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    const auto & column_map = assert_cast<const ColumnMap &>(column);
+    const auto & offsets = column_map.getNestedColumn().getOffsets();
+
+    size_t offset = offsets[row_num - 1];
+    size_t next_offset = offsets[row_num];
+
+    const auto & nested_data = column_map.getNestedData();
+
+    writeCString("<map>", ostr);
+    for (size_t i = offset; i < next_offset; ++i)
+    {
+        writeCString("<elem>", ostr);
+        writeCString("<key>", ostr);
+        key_type->serializeAsTextXML(nested_data.getColumn(0), i, ostr, settings);
+        writeCString("</key>", ostr);
+
+        writeCString("<value>", ostr);
+        value_type->serializeAsTextXML(nested_data.getColumn(1), i, ostr, settings);
+        writeCString("</value>", ostr);
+        writeCString("</elem>", ostr);
+    }
+    writeCString("</map>", ostr);
+}
+
+void DataTypeMap::serializeTextCSV(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings & settings) const
+{
+    WriteBufferFromOwnString wb;
+    serializeText(column, row_num, wb, settings);
+    writeCSV(wb.str(), ostr);
+}
+
+void DataTypeMap::deserializeTextCSV(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
+{
+    String s;
+    readCSV(s, istr, settings.csv);
+    ReadBufferFromString rb(s);
+    deserializeText(column, rb, settings);
+}
+
+
+void DataTypeMap::enumerateStreams(const StreamCallback & callback, SubstreamPath & path) const
+{
+    nested->enumerateStreams(callback, path);
+}
+
+void DataTypeMap::serializeBinaryBulkStatePrefix(
+    SerializeBinaryBulkSettings & settings,
+    SerializeBinaryBulkStatePtr & state) const
+{
+    nested->serializeBinaryBulkStatePrefix(settings, state);
+}
+
+void DataTypeMap::serializeBinaryBulkStateSuffix(
+    SerializeBinaryBulkSettings & settings,
+    SerializeBinaryBulkStatePtr & state) const
+{
+    nested->serializeBinaryBulkStateSuffix(settings, state);
+}
+
+void DataTypeMap::deserializeBinaryBulkStatePrefix(
+        DeserializeBinaryBulkSettings & settings,
+        DeserializeBinaryBulkStatePtr & state) const
+{
+    nested->deserializeBinaryBulkStatePrefix(settings, state);
+}
+
+
+void DataTypeMap::serializeBinaryBulkWithMultipleStreams(
+    const IColumn & column,
+    size_t offset,
+    size_t limit,
+    SerializeBinaryBulkSettings & settings,
+    SerializeBinaryBulkStatePtr & state) const
+{
+    nested->serializeBinaryBulkWithMultipleStreams(extractNestedColumn(column), offset, limit, settings, state);
+}
+
+void DataTypeMap::deserializeBinaryBulkWithMultipleStreams(
+    IColumn & column,
+    size_t limit,
+    DeserializeBinaryBulkSettings & settings,
+    DeserializeBinaryBulkStatePtr & state) const
+{
+    nested->deserializeBinaryBulkWithMultipleStreams(extractNestedColumn(column), limit, settings, state);
+}
+
+void DataTypeMap::serializeProtobuf(const IColumn & column, size_t row_num, ProtobufWriter & protobuf, size_t & value_index) const
+{
+    nested->serializeProtobuf(extractNestedColumn(column), row_num, protobuf, value_index);
+}
+
+void DataTypeMap::deserializeProtobuf(IColumn & column, ProtobufReader & protobuf, bool allow_add_row, bool & row_added) const
+{
+    nested->deserializeProtobuf(extractNestedColumn(column), protobuf, allow_add_row, row_added);
+}
+
+MutableColumnPtr DataTypeMap::createColumn() const
+{
+    return ColumnMap::create(nested->createColumn());
+}
+
+Field DataTypeMap::getDefault() const
+{
+    return Map();
+}
+
+bool DataTypeMap::equals(const IDataType & rhs) const
+{
+    if (typeid(rhs) != typeid(*this))
+        return false;
+
+    const DataTypeMap & rhs_map = static_cast<const DataTypeMap &>(rhs);
+    return nested->equals(*rhs_map.nested);
+}
+
+static DataTypePtr create(const ASTPtr & arguments)
+{
+    if (!arguments || arguments->children.size() != 2)
+        throw Exception("Map data type family must have two arguments: key and value types", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+    DataTypes nested_types;
+    nested_types.reserve(arguments->children.size());
+
+    for (const ASTPtr & child : arguments->children)
+        nested_types.emplace_back(DataTypeFactory::instance().get(child));
+
+    return std::make_shared<DataTypeMap>(nested_types);
+}
+
+
+void registerDataTypeMap(DataTypeFactory & factory)
+{
+    factory.registerDataType("Map", create);
+}
+}
diff --git a/src/DataTypes/DataTypeMap.h b/src/DataTypes/DataTypeMap.h
new file mode 100644
index 00000000000..a52969428f4
--- /dev/null
+++ b/src/DataTypes/DataTypeMap.h
@@ -0,0 +1,103 @@
+#pragma once
+
+#include <DataTypes/DataTypeWithSimpleSerialization.h>
+
+
+namespace DB
+{
+
+/** Map data type.
+  * Map is implemented as two arrays of keys and values.
+  * Serialization of type 'Map(K, V)' is similar to serialization.
+  * of 'Array(Tuple(keys K, values V))' or in other words of 'Nested(keys K, valuev V)'.
+  */
+class DataTypeMap final : public DataTypeWithSimpleSerialization
+{
+private:
+    DataTypePtr key_type;
+    DataTypePtr value_type;
+
+    /// 'nested' is an Array(Tuple(key_type, value_type))
+    DataTypePtr nested;
+
+public:
+    static constexpr bool is_parametric = true;
+
+    DataTypeMap(const DataTypes & elems);
+    DataTypeMap(const DataTypePtr & key_type_, const DataTypePtr & value_type_);
+
+    TypeIndex getTypeId() const override { return TypeIndex::Map; }
+    std::string doGetName() const override;
+    const char * getFamilyName() const override { return "Map"; }
+
+    bool canBeInsideNullable() const override { return false; }
+
+    void serializeBinary(const Field & field, WriteBuffer & ostr) const override;
+    void deserializeBinary(Field & field, ReadBuffer & istr) const override;
+    void serializeBinary(const IColumn & column, size_t row_num, WriteBuffer & ostr) const override;
+    void deserializeBinary(IColumn & column, ReadBuffer & istr) const override;
+    void serializeText(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+    void deserializeText(IColumn & column, ReadBuffer & istr, const FormatSettings &) const override;
+    void serializeTextJSON(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+    void deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings &) const override;
+    void serializeTextXML(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+
+    void serializeTextCSV(const IColumn & column, size_t row_num, WriteBuffer & ostr, const FormatSettings &) const override;
+    void deserializeTextCSV(IColumn & column, ReadBuffer & istr, const FormatSettings &) const override;
+
+
+    /** Each sub-column in a map is serialized in separate stream.
+      */
+    void enumerateStreams(const StreamCallback & callback, SubstreamPath & path) const override;
+
+    void serializeBinaryBulkStatePrefix(
+           SerializeBinaryBulkSettings & settings,
+           SerializeBinaryBulkStatePtr & state) const override;
+
+    void serializeBinaryBulkStateSuffix(
+           SerializeBinaryBulkSettings & settings,
+           SerializeBinaryBulkStatePtr & state) const override;
+
+    void deserializeBinaryBulkStatePrefix(
+           DeserializeBinaryBulkSettings & settings,
+           DeserializeBinaryBulkStatePtr & state) const override;
+
+    void serializeBinaryBulkWithMultipleStreams(
+           const IColumn & column,
+           size_t offset,
+           size_t limit,
+           SerializeBinaryBulkSettings & settings,
+           SerializeBinaryBulkStatePtr & state) const override;
+
+    void deserializeBinaryBulkWithMultipleStreams(
+           IColumn & column,
+           size_t limit,
+           DeserializeBinaryBulkSettings & settings,
+           DeserializeBinaryBulkStatePtr & state) const override;
+
+    void serializeProtobuf(const IColumn & column, size_t row_num, ProtobufWriter & protobuf, size_t & value_index) const override;
+    void deserializeProtobuf(IColumn & column, ProtobufReader & protobuf, bool allow_add_row, bool & row_added) const override;
+
+    MutableColumnPtr createColumn() const override;
+
+    Field getDefault() const override;
+
+    bool equals(const IDataType & rhs) const override;
+    bool isComparable() const override { return key_type->isComparable() && value_type->isComparable(); }
+    bool isParametric() const override { return true; }
+    bool haveSubtypes() const override { return true; }
+
+    const DataTypePtr & getKeyType() const { return key_type; }
+    const DataTypePtr & getValueType() const { return value_type; }
+    DataTypes getKeyValueTypes() const { return {key_type, value_type}; }
+
+private:
+    template <typename Writer>
+    void serializeTextImpl(const IColumn & column, size_t row_num, WriteBuffer & ostr, Writer && writer) const;
+
+    template <typename Reader>
+    void deserializeTextImpl(IColumn & column, ReadBuffer & istr, bool need_safe_get_int_key, Reader && reader) const;
+};
+
+}
+
diff --git a/src/DataTypes/DataTypesDecimal.h b/src/DataTypes/DataTypesDecimal.h
index 079812c2e74..3f7b4e2ac63 100644
--- a/src/DataTypes/DataTypesDecimal.h
+++ b/src/DataTypes/DataTypesDecimal.h
@@ -96,22 +96,29 @@ inline UInt32 getDecimalScale(const DataTypeDecimal<T> & data_type)
     return data_type.getScale();
 }
 
-template <typename FromDataType, typename ToDataType>
-inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
-convertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to)
+template <typename FromDataType, typename ToDataType, typename ReturnType = void>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, ReturnType>
+convertDecimalsImpl(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to, typename ToDataType::FieldType& result)
 {
     using FromFieldType = typename FromDataType::FieldType;
     using ToFieldType = typename ToDataType::FieldType;
     using MaxFieldType = std::conditional_t<(sizeof(FromFieldType) > sizeof(ToFieldType)), FromFieldType, ToFieldType>;
     using MaxNativeType = typename MaxFieldType::NativeType;
 
+    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
+
     MaxNativeType converted_value;
     if (scale_to > scale_from)
     {
         converted_value = DecimalUtils::scaleMultiplier<MaxNativeType>(scale_to - scale_from);
         if (common::mulOverflow(static_cast<MaxNativeType>(value.value), converted_value, converted_value))
-            throw Exception(std::string(ToDataType::family_name) + " convert overflow",
-                            ErrorCodes::DECIMAL_OVERFLOW);
+        {
+            if constexpr (throw_exception)
+                throw Exception(std::string(ToDataType::family_name) + " convert overflow",
+                                ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return ReturnType(false);
+        }
     }
     else
         converted_value = value.value / DecimalUtils::scaleMultiplier<MaxNativeType>(scale_from - scale_to);
@@ -120,35 +127,87 @@ convertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_fro
     {
         if (converted_value < std::numeric_limits<typename ToFieldType::NativeType>::min() ||
             converted_value > std::numeric_limits<typename ToFieldType::NativeType>::max())
-            throw Exception(std::string(ToDataType::family_name) + " convert overflow",
-                            ErrorCodes::DECIMAL_OVERFLOW);
+        {
+            if constexpr (throw_exception)
+                throw Exception(std::string(ToDataType::family_name) + " convert overflow",
+                                ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return ReturnType(false);
+        }
     }
 
-    return static_cast<typename ToFieldType::NativeType>(converted_value);
+    result = static_cast<typename ToFieldType::NativeType>(converted_value);
+
+    return ReturnType(true);
+}
+
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
+convertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to)
+{
+    using ToFieldType = typename ToDataType::FieldType;
+    ToFieldType result;
+
+    convertDecimalsImpl<FromDataType, ToDataType, void>(value, scale_from, scale_to, result);
+
+    return result;
+}
+
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>, bool>
+tryConvertDecimals(const typename FromDataType::FieldType & value, UInt32 scale_from, UInt32 scale_to, typename ToDataType::FieldType& result)
+{
+    return convertDecimalsImpl<FromDataType, ToDataType, bool>(value, scale_from, scale_to, result);
+}
+
+template <typename FromDataType, typename ToDataType, typename ReturnType>
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsNumber<typename ToDataType::FieldType>, ReturnType>
+convertFromDecimalImpl(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
+{
+    using FromFieldType = typename FromDataType::FieldType;
+    using ToFieldType = typename ToDataType::FieldType;
+
+    return DecimalUtils::convertToImpl<ToFieldType, FromFieldType, ReturnType>(value, scale, result);
 }
 
 template <typename FromDataType, typename ToDataType>
 inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsNumber<typename ToDataType::FieldType>, typename ToDataType::FieldType>
 convertFromDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
 {
-    using ToFieldType = typename ToDataType::FieldType;
+    typename ToDataType::FieldType result;
 
-    return DecimalUtils::convertTo<ToFieldType>(value, scale);
+    convertFromDecimalImpl<FromDataType, ToDataType, void>(value, scale, result);
+
+    return result;
 }
 
 template <typename FromDataType, typename ToDataType>
-inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
-convertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
+inline std::enable_if_t<IsDataTypeDecimal<FromDataType> && IsNumber<typename ToDataType::FieldType>, bool>
+tryConvertFromDecimal(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
+{
+    return convertFromDecimalImpl<FromDataType, ToDataType, bool>(value, scale, result);
+}
+
+template <typename FromDataType, typename ToDataType, typename ReturnType>
+inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, ReturnType>
+convertToDecimalImpl(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
 {
     using FromFieldType = typename FromDataType::FieldType;
     using ToFieldType = typename ToDataType::FieldType;
     using ToNativeType = typename ToFieldType::NativeType;
 
+    static constexpr bool throw_exception = std::is_same_v<ReturnType, void>;
+
     if constexpr (std::is_floating_point_v<FromFieldType>)
     {
         if (!std::isfinite(value))
-            throw Exception(std::string(ToDataType::family_name) + " convert overflow. Cannot convert infinity or NaN to decimal",
-                            ErrorCodes::DECIMAL_OVERFLOW);
+        {
+            if constexpr (throw_exception)
+                throw Exception(std::string(ToDataType::family_name) + " convert overflow. Cannot convert infinity or NaN to decimal",
+                                ErrorCodes::DECIMAL_OVERFLOW);
+            else
+                return false;
+        }
 
         auto out = value * static_cast<FromFieldType>(DecimalUtils::scaleMultiplier<ToNativeType>(scale));
         if constexpr (std::is_same_v<ToNativeType, Int128>)
@@ -157,29 +216,60 @@ convertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
             static constexpr Int128 max_int128 = maxInt128();
 
             if (out <= static_cast<ToNativeType>(min_int128) || out >= static_cast<ToNativeType>(max_int128))
-                throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
-                                ErrorCodes::DECIMAL_OVERFLOW);
+            {
+                if constexpr (throw_exception)
+                    throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
+                                    ErrorCodes::DECIMAL_OVERFLOW);
+                else
+                    return ReturnType(false);
+            }
         }
         else
         {
             if (out <= static_cast<FromFieldType>(std::numeric_limits<ToNativeType>::min()) ||
                 out >= static_cast<FromFieldType>(std::numeric_limits<ToNativeType>::max()))
-                throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
-                                ErrorCodes::DECIMAL_OVERFLOW);
+            {
+                if constexpr (throw_exception)
+                    throw Exception(std::string(ToDataType::family_name) + " convert overflow. Float is out of Decimal range",
+                                    ErrorCodes::DECIMAL_OVERFLOW);
+                else
+                    return ReturnType(false);
+            }
         }
-        return static_cast<ToNativeType>(out);
+
+        result = static_cast<ToNativeType>(out);
+
+        return ReturnType(true);
     }
     else
     {
         if constexpr (is_big_int_v<FromFieldType>)
-            return convertDecimals<DataTypeDecimal<Decimal256>, ToDataType>(static_cast<Int256>(value), 0, scale);
+            return ReturnType(convertDecimalsImpl<DataTypeDecimal<Decimal256>, ToDataType, ReturnType>(static_cast<Int256>(value), 0, scale, result));
         else if constexpr (std::is_same_v<FromFieldType, UInt64>)
-            return convertDecimals<DataTypeDecimal<Decimal128>, ToDataType>(value, 0, scale);
+            return ReturnType(convertDecimalsImpl<DataTypeDecimal<Decimal128>, ToDataType, ReturnType>(value, 0, scale, result));
         else
-            return convertDecimals<DataTypeDecimal<Decimal64>, ToDataType>(value, 0, scale);
+            return ReturnType(convertDecimalsImpl<DataTypeDecimal<Decimal64>, ToDataType, ReturnType>(value, 0, scale, result));
     }
 }
 
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, typename ToDataType::FieldType>
+convertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale)
+{
+    typename ToDataType::FieldType result;
+
+    convertToDecimalImpl<FromDataType, ToDataType, void>(value, scale, result);
+
+    return result;
+}
+
+template <typename FromDataType, typename ToDataType>
+inline std::enable_if_t<IsNumber<typename FromDataType::FieldType> && IsDataTypeDecimal<ToDataType>, bool>
+tryConvertToDecimal(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType& result)
+{
+    return convertToDecimalImpl<FromDataType, ToDataType, bool>(value, scale, result);
+}
+
 template <typename T>
 inline DataTypePtr createDecimalMaxPrecision(UInt64 scale)
 {
diff --git a/src/DataTypes/FieldToDataType.cpp b/src/DataTypes/FieldToDataType.cpp
index e01b264df81..95f03385121 100644
--- a/src/DataTypes/FieldToDataType.cpp
+++ b/src/DataTypes/FieldToDataType.cpp
@@ -1,6 +1,7 @@
 #include <Common/FieldVisitors.h>
 #include <DataTypes/FieldToDataType.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypesDecimal.h>
 #include <DataTypes/DataTypeString.h>
@@ -118,6 +119,24 @@ DataTypePtr FieldToDataType::operator() (const Tuple & tuple) const
     return std::make_shared<DataTypeTuple>(element_types);
 }
 
+DataTypePtr FieldToDataType::operator() (const Map & map) const
+{
+    DataTypes key_types;
+    DataTypes value_types;
+    key_types.reserve(map.size());
+    value_types.reserve(map.size());
+
+    for (const auto & elem : map)
+    {
+        const auto & tuple = elem.safeGet<const Tuple &>();
+        assert(tuple.size() == 2);
+        key_types.push_back(applyVisitor(FieldToDataType(), tuple[0]));
+        value_types.push_back(applyVisitor(FieldToDataType(), tuple[1]));
+    }
+
+    return std::make_shared<DataTypeMap>(getLeastSupertype(key_types), getLeastSupertype(value_types));
+}
+
 DataTypePtr FieldToDataType::operator() (const AggregateFunctionStateData & x) const
 {
     const auto & name = static_cast<const AggregateFunctionStateData &>(x).name;
diff --git a/src/DataTypes/FieldToDataType.h b/src/DataTypes/FieldToDataType.h
index 0f8cbfd266c..39457564a2f 100644
--- a/src/DataTypes/FieldToDataType.h
+++ b/src/DataTypes/FieldToDataType.h
@@ -26,6 +26,7 @@ public:
     DataTypePtr operator() (const String & x) const;
     DataTypePtr operator() (const Array & x) const;
     DataTypePtr operator() (const Tuple & tuple) const;
+    DataTypePtr operator() (const Map & map) const;
     DataTypePtr operator() (const DecimalField<Decimal32> & x) const;
     DataTypePtr operator() (const DecimalField<Decimal64> & x) const;
     DataTypePtr operator() (const DecimalField<Decimal128> & x) const;
diff --git a/src/DataTypes/IDataType.h b/src/DataTypes/IDataType.h
index 4601378e8c2..060572ee997 100644
--- a/src/DataTypes/IDataType.h
+++ b/src/DataTypes/IDataType.h
@@ -91,6 +91,8 @@ public:
 
             TupleElement,
 
+            MapElement,
+
             DictionaryKeys,
             DictionaryIndexes,
         };
@@ -449,6 +451,7 @@ public:
     static bool isSpecialCompressionAllowed(const SubstreamPath & path);
 private:
     friend class DataTypeFactory;
+    friend class AggregateFunctionSimpleState;
     /// Customize this DataType
     void setCustomization(DataTypeCustomDescPtr custom_desc_) const;
 
@@ -517,6 +520,7 @@ struct WhichDataType
     constexpr bool isUUID() const { return idx == TypeIndex::UUID; }
     constexpr bool isArray() const { return idx == TypeIndex::Array; }
     constexpr bool isTuple() const { return idx == TypeIndex::Tuple; }
+    constexpr bool isMap() const {return idx == TypeIndex::Map; }
     constexpr bool isSet() const { return idx == TypeIndex::Set; }
     constexpr bool isInterval() const { return idx == TypeIndex::Interval; }
 
diff --git a/src/DataTypes/ya.make b/src/DataTypes/ya.make
index b570adbdc7a..db6793db6ca 100644
--- a/src/DataTypes/ya.make
+++ b/src/DataTypes/ya.make
@@ -27,6 +27,7 @@ SRCS(
     DataTypeInterval.cpp
     DataTypeLowCardinality.cpp
     DataTypeLowCardinalityHelpers.cpp
+    DataTypeMap.cpp
     DataTypeNothing.cpp
     DataTypeNullable.cpp
     DataTypeNumberBase.cpp
diff --git a/src/Databases/DatabaseAtomic.cpp b/src/Databases/DatabaseAtomic.cpp
index af8b751e787..17a91a1fff9 100644
--- a/src/Databases/DatabaseAtomic.cpp
+++ b/src/Databases/DatabaseAtomic.cpp
@@ -35,8 +35,8 @@ public:
 };
 
 
-DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, Context & context_)
-    : DatabaseOrdinary(name_, std::move(metadata_path_), "store/", "DatabaseAtomic (" + name_ + ")", context_)
+DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const String & logger_name, const Context & context_)
+    : DatabaseOrdinary(name_, std::move(metadata_path_), "store/", logger_name, context_)
     , path_to_table_symlinks(global_context.getPath() + "data/" + escapeForFileName(name_) + "/")
     , path_to_metadata_symlink(global_context.getPath() + "metadata/" + escapeForFileName(name_))
     , db_uuid(uuid)
@@ -46,6 +46,11 @@ DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, C
     tryCreateMetadataSymlink();
 }
 
+DatabaseAtomic::DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const Context & context_)
+    : DatabaseAtomic(name_, std::move(metadata_path_), uuid, "DatabaseAtomic (" + name_ + ")", context_)
+{
+}
+
 String DatabaseAtomic::getTableDataPath(const String & table_name) const
 {
     std::lock_guard lock(mutex);
@@ -212,6 +217,9 @@ void DatabaseAtomic::renameTable(const Context & context, const String & table_n
     if (is_dictionary && !inside_database)
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Cannot move dictionary to other database");
 
+    if (!exchange)
+        other_db.checkMetadataFilenameAvailabilityUnlocked(to_table_name, inside_database ? db_lock : other_db_lock);
+
     StoragePtr table = getTableUnlocked(table_name, db_lock);
     table->checkTableCanBeRenamed();
     assert_can_move_mat_view(table);
diff --git a/src/Databases/DatabaseAtomic.h b/src/Databases/DatabaseAtomic.h
index 82408ff3ab3..1b1c0cd4353 100644
--- a/src/Databases/DatabaseAtomic.h
+++ b/src/Databases/DatabaseAtomic.h
@@ -20,8 +20,8 @@ namespace DB
 class DatabaseAtomic : public DatabaseOrdinary
 {
 public:
-
-    DatabaseAtomic(String name_, String metadata_path_, UUID uuid, Context & context_);
+    DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const String & logger_name, const Context & context_);
+    DatabaseAtomic(String name_, String metadata_path_, UUID uuid, const Context & context_);
 
     String getEngineName() const override { return "Atomic"; }
     UUID getUUID() const override { return db_uuid; }
@@ -51,14 +51,14 @@ public:
     void loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach) override;
 
     /// Atomic database cannot be detached if there is detached table which still in use
-    void assertCanBeDetached(bool cleanup);
+    void assertCanBeDetached(bool cleanup) override;
 
     UUID tryGetTableUUID(const String & table_name) const override;
 
     void tryCreateSymlink(const String & table_name, const String & actual_data_path, bool if_data_path_exist = false);
     void tryRemoveSymlink(const String & table_name);
 
-    void waitDetachedTableNotInUse(const UUID & uuid);
+    void waitDetachedTableNotInUse(const UUID & uuid) override;
 
 private:
     void commitAlterTable(const StorageID & table_id, const String & table_metadata_tmp_path, const String & table_metadata_path) override;
diff --git a/src/Databases/DatabaseFactory.cpp b/src/Databases/DatabaseFactory.cpp
index 2bb4c595c05..0ac80706bd8 100644
--- a/src/Databases/DatabaseFactory.cpp
+++ b/src/Databases/DatabaseFactory.cpp
@@ -120,27 +120,32 @@ DatabasePtr DatabaseFactory::getImpl(const ASTCreateQuery & create, const String
             const auto & [remote_host_name, remote_port] = parseAddress(host_name_and_port, 3306);
             auto mysql_pool = mysqlxx::Pool(mysql_database_name, remote_host_name, mysql_user_name, mysql_user_password, remote_port);
 
-            if (engine_name == "MaterializeMySQL")
+            if (engine_name == "MySQL")
             {
-                MySQLClient client(remote_host_name, remote_port, mysql_user_name, mysql_user_password);
+                auto mysql_database_settings = std::make_unique<ConnectionMySQLSettings>();
 
-                auto materialize_mode_settings = std::make_unique<MaterializeMySQLSettings>();
+                mysql_database_settings->loadFromQueryContext(context);
+                mysql_database_settings->loadFromQuery(*engine_define); /// higher priority
 
-                if (engine_define->settings)
-                    materialize_mode_settings->loadFromQuery(*engine_define);
-
-                return std::make_shared<DatabaseMaterializeMySQL>(
-                    context, database_name, metadata_path, engine_define, mysql_database_name, std::move(mysql_pool), std::move(client)
-                    , std::move(materialize_mode_settings));
+                return std::make_shared<DatabaseConnectionMySQL>(
+                    context, database_name, metadata_path, engine_define, mysql_database_name, std::move(mysql_database_settings), std::move(mysql_pool));
             }
 
-            auto mysql_database_settings = std::make_unique<ConnectionMySQLSettings>();
+            MySQLClient client(remote_host_name, remote_port, mysql_user_name, mysql_user_password);
 
-            mysql_database_settings->loadFromQueryContext(context);
-            mysql_database_settings->loadFromQuery(*engine_define); /// higher priority
+            auto materialize_mode_settings = std::make_unique<MaterializeMySQLSettings>();
 
-            return std::make_shared<DatabaseConnectionMySQL>(
-                context, database_name, metadata_path, engine_define, mysql_database_name, std::move(mysql_database_settings), std::move(mysql_pool));
+            if (engine_define->settings)
+                materialize_mode_settings->loadFromQuery(*engine_define);
+
+            if (create.uuid == UUIDHelpers::Nil)
+                return std::make_shared<DatabaseMaterializeMySQL<DatabaseOrdinary>>(
+                    context, database_name, metadata_path, uuid, mysql_database_name, std::move(mysql_pool), std::move(client)
+                    , std::move(materialize_mode_settings));
+            else
+                return std::make_shared<DatabaseMaterializeMySQL<DatabaseAtomic>>(
+                    context, database_name, metadata_path, uuid, mysql_database_name, std::move(mysql_pool), std::move(client)
+                    , std::move(materialize_mode_settings));
         }
         catch (...)
         {
diff --git a/src/Databases/DatabaseLazy.cpp b/src/Databases/DatabaseLazy.cpp
index 0119f17f843..f297bf2c82f 100644
--- a/src/Databases/DatabaseLazy.cpp
+++ b/src/Databases/DatabaseLazy.cpp
@@ -42,6 +42,14 @@ void DatabaseLazy::loadStoredObjects(
     iterateMetadataFiles(context, [this](const String & file_name)
     {
         const std::string table_name = file_name.substr(0, file_name.size() - 4);
+
+        auto detached_permanently_flag = Poco::File(getMetadataPath() + "/" + file_name + detached_suffix);
+        if (detached_permanently_flag.exists())
+        {
+            LOG_DEBUG(log, "Skipping permanently detached table {}.", backQuote(table_name));
+            return;
+        }
+
         attachTable(table_name, nullptr, {});
     });
 }
diff --git a/src/Databases/DatabaseOnDisk.cpp b/src/Databases/DatabaseOnDisk.cpp
index 1e6b4019c4b..2a8276ec6db 100644
--- a/src/Databases/DatabaseOnDisk.cpp
+++ b/src/Databases/DatabaseOnDisk.cpp
@@ -164,20 +164,38 @@ void DatabaseOnDisk::createTable(
     /// But there is protection from it - see using DDLGuard in InterpreterCreateQuery.
 
     if (isDictionaryExist(table_name))
-        throw Exception("Dictionary " + backQuote(getDatabaseName()) + "." + backQuote(table_name) + " already exists.",
-            ErrorCodes::DICTIONARY_ALREADY_EXISTS);
+        throw Exception(ErrorCodes::DICTIONARY_ALREADY_EXISTS, "Dictionary {}.{} already exists", backQuote(getDatabaseName()), backQuote(table_name));
 
     if (isTableExist(table_name, global_context))
-        throw Exception("Table " + backQuote(getDatabaseName()) + "." + backQuote(table_name) + " already exists.", ErrorCodes::TABLE_ALREADY_EXISTS);
+        throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists", backQuote(getDatabaseName()), backQuote(table_name));
+
+    String table_metadata_path = getObjectMetadataPath(table_name);
 
     if (create.attach_short_syntax)
     {
         /// Metadata already exists, table was detached
         attachTable(table_name, table, getTableDataPath(create));
+        removeDetachedPermanentlyFlag(table_name, table_metadata_path);
         return;
     }
 
-    String table_metadata_path = getObjectMetadataPath(table_name);
+    if (!create.attach)
+        checkMetadataFilenameAvailability(table_name);
+
+    if (create.attach && Poco::File(table_metadata_path).exists())
+    {
+        ASTPtr ast_detached = parseQueryFromMetadata(log, context, table_metadata_path);
+        auto & create_detached = ast_detached->as<ASTCreateQuery &>();
+
+        // either both should be Nil, either values should be equal
+        if (create.uuid != create_detached.uuid)
+            throw Exception(
+                    ErrorCodes::TABLE_ALREADY_EXISTS,
+                    "Table {}.{} already exist (detached permanently). To attach it back "
+                    "you need to use short ATTACH syntax or a full statement with the same UUID",
+                    backQuote(getDatabaseName()), backQuote(table_name));
+    }
+
     String table_metadata_tmp_path = table_metadata_path + create_suffix;
     String statement;
 
@@ -194,6 +212,26 @@ void DatabaseOnDisk::createTable(
     }
 
     commitCreateTable(create, table, table_metadata_tmp_path, table_metadata_path);
+
+    removeDetachedPermanentlyFlag(table_name, table_metadata_path);
+}
+
+/// If the table was detached permanently we will have a flag file with
+/// .sql.detached extension, is not needed anymore since we attached the table back
+void DatabaseOnDisk::removeDetachedPermanentlyFlag(const String & table_name, const String & table_metadata_path) const
+{
+    try
+    {
+        auto detached_permanently_flag = Poco::File(table_metadata_path + detached_suffix);
+
+        if (detached_permanently_flag.exists())
+            detached_permanently_flag.remove();
+    }
+    catch (Exception & e)
+    {
+        e.addMessage("while trying to remove permanenty detached flag. Table {}.{} may still be marked as permanently detached, and will not be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
+        throw;
+    }
 }
 
 void DatabaseOnDisk::commitCreateTable(const ASTCreateQuery & query, const StoragePtr & table,
@@ -215,6 +253,22 @@ void DatabaseOnDisk::commitCreateTable(const ASTCreateQuery & query, const Stora
     }
 }
 
+void DatabaseOnDisk::detachTablePermanently(const String & table_name)
+{
+    auto table = detachTable(table_name);
+
+    Poco::File detached_permanently_flag(getObjectMetadataPath(table_name) + detached_suffix);
+    try
+    {
+        detached_permanently_flag.createFile();
+    }
+    catch (Exception & e)
+    {
+        e.addMessage("while trying to set permanenty detached flag. Table {}.{} may be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
+        throw;
+    }
+}
+
 void DatabaseOnDisk::dropTable(const Context & context, const String & table_name, bool /*no_delay*/)
 {
     String table_metadata_path = getObjectMetadataPath(table_name);
@@ -253,6 +307,27 @@ void DatabaseOnDisk::dropTable(const Context & context, const String & table_nam
     Poco::File(table_metadata_path_drop).remove();
 }
 
+void DatabaseOnDisk::checkMetadataFilenameAvailability(const String & to_table_name) const
+{
+    std::unique_lock lock(mutex);
+    checkMetadataFilenameAvailabilityUnlocked(to_table_name, lock);
+}
+
+void DatabaseOnDisk::checkMetadataFilenameAvailabilityUnlocked(const String & to_table_name, std::unique_lock<std::mutex> &) const
+{
+    String table_metadata_path = getObjectMetadataPath(to_table_name);
+
+    if (Poco::File(table_metadata_path).exists())
+    {
+        auto detached_permanently_flag = Poco::File(table_metadata_path + detached_suffix);
+
+        if (detached_permanently_flag.exists())
+            throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists (detached permanently)", backQuote(database_name), backQuote(to_table_name));
+        else
+            throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists (detached)", backQuote(database_name), backQuote(to_table_name));
+    }
+}
+
 void DatabaseOnDisk::renameTable(
         const Context & context,
         const String & table_name,
@@ -299,6 +374,9 @@ void DatabaseOnDisk::renameTable(
         if (from_atomic_to_ordinary)
             create.uuid = UUIDHelpers::Nil;
 
+        if (auto * target_db = dynamic_cast<DatabaseOnDisk *>(&to_database))
+            target_db->checkMetadataFilenameAvailability(to_table_name);
+
         /// Notify the table that it is renamed. It will move data to new path (if it stores data on disk) and update StorageID
         table->rename(to_database.getTableDataPath(create), StorageID(create));
     }
@@ -328,6 +406,8 @@ void DatabaseOnDisk::renameTable(
     }
 }
 
+
+/// It returns create table statement (even if table is detached)
 ASTPtr DatabaseOnDisk::getCreateTableQueryImpl(const String & table_name, const Context &, bool throw_on_error) const
 {
     ASTPtr ast;
@@ -400,7 +480,7 @@ void DatabaseOnDisk::iterateMetadataFiles(const Context & context, const Iterati
 {
     auto process_tmp_drop_metadata_file = [&](const String & file_name)
     {
-        assert(getEngineName() != "Atomic");
+        assert(getUUID() == UUIDHelpers::Nil);
         static const char * tmp_drop_ext = ".sql.tmp_drop";
         const std::string object_name = file_name.substr(0, file_name.size() - strlen(tmp_drop_ext));
         if (Poco::File(context.getPath() + getDataPath() + '/' + object_name).exists())
@@ -430,8 +510,11 @@ void DatabaseOnDisk::iterateMetadataFiles(const Context & context, const Iterati
         if (endsWith(dir_it.name(), ".sql.bak"))
             continue;
 
-        static const char * tmp_drop_ext = ".sql.tmp_drop";
-        if (endsWith(dir_it.name(), tmp_drop_ext))
+        /// Permanently detached table flag
+        if (endsWith(dir_it.name(), ".sql.detached"))
+            continue;
+
+        if (endsWith(dir_it.name(), ".sql.tmp_drop"))
         {
             /// There are files that we tried to delete previously
             metadata_files.emplace(dir_it.name(), false);
diff --git a/src/Databases/DatabaseOnDisk.h b/src/Databases/DatabaseOnDisk.h
index 586491d4d29..fff2a259911 100644
--- a/src/Databases/DatabaseOnDisk.h
+++ b/src/Databases/DatabaseOnDisk.h
@@ -39,6 +39,8 @@ public:
         const StoragePtr & table,
         const ASTPtr & query) override;
 
+    void detachTablePermanently(const String & table_name) override;
+
     void dropTable(
         const Context & context,
         const String & table_name,
@@ -67,9 +69,14 @@ public:
 
     static ASTPtr parseQueryFromMetadata(Poco::Logger * log, const Context & context, const String & metadata_file_path, bool throw_on_error = true, bool remove_empty = false);
 
+    /// will throw when the table we want to attach already exists (in active / detached / detached permanently form)
+    void checkMetadataFilenameAvailability(const String & to_table_name) const;
+    void checkMetadataFilenameAvailabilityUnlocked(const String & to_table_name, std::unique_lock<std::mutex> &) const;
+
 protected:
     static constexpr const char * create_suffix = ".tmp";
     static constexpr const char * drop_suffix = ".tmp_drop";
+    static constexpr const char * detached_suffix = ".detached";
 
     using IteratingFunction = std::function<void(const String &)>;
 
@@ -87,6 +94,9 @@ protected:
 
     const String metadata_path;
     const String data_path;
+
+private:
+    void removeDetachedPermanentlyFlag(const String & table_name, const String & table_metadata_path) const;
 };
 
 }
diff --git a/src/Databases/DatabaseOrdinary.cpp b/src/Databases/DatabaseOrdinary.cpp
index 24e2bdcd6b2..58b4ad72a1b 100644
--- a/src/Databases/DatabaseOrdinary.cpp
+++ b/src/Databases/DatabaseOrdinary.cpp
@@ -135,6 +135,19 @@ void DatabaseOrdinary::loadStoredObjects(Context & context, bool has_force_resto
             {
                 auto * create_query = ast->as<ASTCreateQuery>();
                 create_query->database = database_name;
+
+                auto detached_permanently_flag = Poco::File(full_path.string() + detached_suffix);
+                if (detached_permanently_flag.exists())
+                {
+                    /// FIXME: even if we don't load the table we can still mark the uuid of it as taken.
+                    /// if (create_query->uuid != UUIDHelpers::Nil)
+                    ///     DatabaseCatalog::instance().addUUIDMapping(create_query->uuid);
+
+                    const std::string table_name = file_name.substr(0, file_name.size() - 4);
+                    LOG_DEBUG(log, "Skipping permanently detached table {}.", backQuote(table_name));
+                    return;
+                }
+
                 std::lock_guard lock{file_names_mutex};
                 file_names[file_name] = ast;
                 total_dictionaries += create_query->is_dictionary;
diff --git a/src/Databases/DatabasesCommon.cpp b/src/Databases/DatabasesCommon.cpp
index c5df954c2da..2d3d206162b 100644
--- a/src/Databases/DatabasesCommon.cpp
+++ b/src/Databases/DatabasesCommon.cpp
@@ -72,7 +72,7 @@ StoragePtr DatabaseWithOwnTablesBase::detachTableUnlocked(const String & table_n
 
     auto it = tables.find(table_name);
     if (it == tables.end())
-        throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist.",
+        throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist",
                         backQuote(database_name), backQuote(table_name));
     res = it->second;
     tables.erase(it);
@@ -80,7 +80,7 @@ StoragePtr DatabaseWithOwnTablesBase::detachTableUnlocked(const String & table_n
     auto table_id = res->getStorageID();
     if (table_id.hasUUID())
     {
-        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getEngineName() == "Atomic");
+        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getUUID() != UUIDHelpers::Nil);
         DatabaseCatalog::instance().removeUUIDMapping(table_id.uuid);
     }
 
@@ -102,7 +102,7 @@ void DatabaseWithOwnTablesBase::attachTableUnlocked(const String & table_name, c
 
     if (table_id.hasUUID())
     {
-        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getEngineName() == "Atomic");
+        assert(database_name == DatabaseCatalog::TEMPORARY_DATABASE || getUUID() != UUIDHelpers::Nil);
         DatabaseCatalog::instance().addUUIDMapping(table_id.uuid, shared_from_this(), table);
     }
 
@@ -131,7 +131,7 @@ void DatabaseWithOwnTablesBase::shutdown()
         kv.second->shutdown();
         if (table_id.hasUUID())
         {
-            assert(getDatabaseName() == DatabaseCatalog::TEMPORARY_DATABASE || getEngineName() == "Atomic");
+            assert(getDatabaseName() == DatabaseCatalog::TEMPORARY_DATABASE || getUUID() != UUIDHelpers::Nil);
             DatabaseCatalog::instance().removeUUIDMapping(table_id.uuid);
         }
     }
@@ -157,7 +157,7 @@ StoragePtr DatabaseWithOwnTablesBase::getTableUnlocked(const String & table_name
     auto it = tables.find(table_name);
     if (it != tables.end())
         return it->second;
-    throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist.",
+    throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist",
                     backQuote(database_name), backQuote(table_name));
 }
 
diff --git a/src/Databases/IDatabase.h b/src/Databases/IDatabase.h
index fadec5fe7a9..fc821fcab30 100644
--- a/src/Databases/IDatabase.h
+++ b/src/Databases/IDatabase.h
@@ -221,6 +221,8 @@ public:
     }
 
     /// Add a table to the database, but do not add it to the metadata. The database may not support this method.
+    ///
+    /// Note: ATTACH TABLE statement actually uses createTable method.
     virtual void attachTable(const String & /*name*/, const StoragePtr & /*table*/, [[maybe_unused]] const String & relative_table_path = {})
     {
         throw Exception("There is no ATTACH TABLE query for Database" + getEngineName(), ErrorCodes::NOT_IMPLEMENTED);
@@ -245,6 +247,13 @@ public:
         throw Exception("There is no DETACH DICTIONARY query for Database" + getEngineName(), ErrorCodes::NOT_IMPLEMENTED);
     }
 
+    /// Forget about the table without deleting it's data, but rename metadata file to prevent reloading it
+    /// with next restart. The database may not support this method.
+    virtual void detachTablePermanently(const String & /*name*/)
+    {
+        throw Exception("There is no DETACH TABLE PERMANENTLY query for Database" + getEngineName(), ErrorCodes::NOT_IMPLEMENTED);
+    }
+
     /// Rename the table and possibly move the table to another database.
     virtual void renameTable(
         const Context & /*context*/,
@@ -334,6 +343,10 @@ public:
     /// All tables and dictionaries should be detached before detaching the database.
     virtual bool shouldBeEmptyOnDetach() const { return true; }
 
+    virtual void assertCanBeDetached(bool /*cleanup*/) {}
+
+    virtual void waitDetachedTableNotInUse(const UUID & /*uuid*/) { assert(false); }
+
     /// Ask all tables to complete the background threads they are using and delete all table objects.
     virtual void shutdown() = 0;
 
diff --git a/src/Databases/MySQL/DatabaseConnectionMySQL.cpp b/src/Databases/MySQL/DatabaseConnectionMySQL.cpp
index 45483055739..35b016f255b 100644
--- a/src/Databases/MySQL/DatabaseConnectionMySQL.cpp
+++ b/src/Databases/MySQL/DatabaseConnectionMySQL.cpp
@@ -395,7 +395,7 @@ void DatabaseConnectionMySQL::loadStoredObjects(Context &, bool, bool /*force_at
     }
 }
 
-void DatabaseConnectionMySQL::dropTable(const Context &, const String & table_name, bool /*no_delay*/)
+void DatabaseConnectionMySQL::detachTablePermanently(const String & table_name)
 {
     std::lock_guard<std::mutex> lock{mutex};
 
@@ -429,6 +429,11 @@ void DatabaseConnectionMySQL::dropTable(const Context &, const String & table_na
     table_iter->second.second->is_dropped = true;
 }
 
+void DatabaseConnectionMySQL::dropTable(const Context &, const String & table_name, bool /*no_delay*/)
+{
+    detachTablePermanently(table_name);
+}
+
 DatabaseConnectionMySQL::~DatabaseConnectionMySQL()
 {
     try
diff --git a/src/Databases/MySQL/DatabaseConnectionMySQL.h b/src/Databases/MySQL/DatabaseConnectionMySQL.h
index d8694e71db2..3e305fcb20d 100644
--- a/src/Databases/MySQL/DatabaseConnectionMySQL.h
+++ b/src/Databases/MySQL/DatabaseConnectionMySQL.h
@@ -72,6 +72,8 @@ public:
 
     StoragePtr detachTable(const String & table_name) override;
 
+    void detachTablePermanently(const String & table_name) override;
+
     void dropTable(const Context &, const String & table_name, bool no_delay) override;
 
     void attachTable(const String & table_name, const StoragePtr & storage, const String & relative_table_path) override;
diff --git a/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp b/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp
index b5231a23d7e..6a9f1e37f8e 100644
--- a/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp
+++ b/src/Databases/MySQL/DatabaseMaterializeMySQL.cpp
@@ -8,6 +8,7 @@
 
 #    include <Interpreters/Context.h>
 #    include <Databases/DatabaseOrdinary.h>
+#    include <Databases/DatabaseAtomic.h>
 #    include <Databases/MySQL/DatabaseMaterializeTablesIterator.h>
 #    include <Databases/MySQL/MaterializeMySQLSyncThread.h>
 #    include <Parsers/ASTCreateQuery.h>
@@ -22,21 +23,37 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int NOT_IMPLEMENTED;
+    extern const int LOGICAL_ERROR;
 }
 
-DatabaseMaterializeMySQL::DatabaseMaterializeMySQL(
-    const Context & context, const String & database_name_, const String & metadata_path_, const IAST * database_engine_define_
-    , const String & mysql_database_name_, mysqlxx::Pool && pool_, MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_)
-    : IDatabase(database_name_), global_context(context.getGlobalContext()), engine_define(database_engine_define_->clone())
-    , nested_database(std::make_shared<DatabaseOrdinary>(database_name_, metadata_path_, context))
-    , settings(std::move(settings_)), log(&Poco::Logger::get("DatabaseMaterializeMySQL"))
+template<>
+DatabaseMaterializeMySQL<DatabaseOrdinary>::DatabaseMaterializeMySQL(
+    const Context & context, const String & database_name_, const String & metadata_path_, UUID /*uuid*/,
+    const String & mysql_database_name_, mysqlxx::Pool && pool_, MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_)
+    : DatabaseOrdinary(database_name_
+                      , metadata_path_
+                      , "data/" + escapeForFileName(database_name_) + "/"
+                      , "DatabaseMaterializeMySQL<Ordinary> (" + database_name_ + ")", context
+                      )
+    , settings(std::move(settings_))
     , materialize_thread(context, database_name_, mysql_database_name_, std::move(pool_), std::move(client_), settings.get())
 {
 }
 
-void DatabaseMaterializeMySQL::rethrowExceptionIfNeed() const
+template<>
+DatabaseMaterializeMySQL<DatabaseAtomic>::DatabaseMaterializeMySQL(
+    const Context & context, const String & database_name_, const String & metadata_path_, UUID uuid,
+    const String & mysql_database_name_, mysqlxx::Pool && pool_, MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_)
+    : DatabaseAtomic(database_name_, metadata_path_, uuid, "DatabaseMaterializeMySQL<Atomic> (" + database_name_ + ")", context)
+    , settings(std::move(settings_))
+    , materialize_thread(context, database_name_, mysql_database_name_, std::move(pool_), std::move(client_), settings.get())
 {
-    std::unique_lock<std::mutex> lock(mutex);
+}
+
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::rethrowExceptionIfNeed() const
+{
+    std::unique_lock<std::mutex> lock(Base::mutex);
 
     if (!settings->allows_query_when_mysql_lost && exception)
     {
@@ -46,129 +63,71 @@ void DatabaseMaterializeMySQL::rethrowExceptionIfNeed() const
         }
         catch (Exception & ex)
         {
+            /// This method can be called from multiple threads
+            /// and Exception can be modified concurrently by calling addMessage(...),
+            /// so we rethrow a copy.
             throw Exception(ex);
         }
     }
 }
 
-void DatabaseMaterializeMySQL::setException(const std::exception_ptr & exception_)
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::setException(const std::exception_ptr & exception_)
 {
-    std::unique_lock<std::mutex> lock(mutex);
+    std::unique_lock<std::mutex> lock(Base::mutex);
     exception = exception_;
 }
 
-ASTPtr DatabaseMaterializeMySQL::getCreateDatabaseQuery() const
-{
-    const auto & create_query = std::make_shared<ASTCreateQuery>();
-    create_query->database = database_name;
-    create_query->set(create_query->storage, engine_define);
-    return create_query;
-}
-
-void DatabaseMaterializeMySQL::loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach)
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach)
 {
+    Base::loadStoredObjects(context, has_force_restore_data_flag, force_attach);
     try
     {
-        std::unique_lock<std::mutex> lock(mutex);
-        nested_database->loadStoredObjects(context, has_force_restore_data_flag, force_attach);
         materialize_thread.startSynchronization();
+        started_up = true;
     }
     catch (...)
     {
-        tryLogCurrentException(log, "Cannot load MySQL nested database stored objects.");
+        tryLogCurrentException(Base::log, "Cannot load MySQL nested database stored objects.");
 
         if (!force_attach)
             throw;
     }
 }
 
-void DatabaseMaterializeMySQL::shutdown()
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::createTable(const Context & context, const String & name, const StoragePtr & table, const ASTPtr & query)
 {
-    materialize_thread.stopSynchronization();
-
-    auto iterator = nested_database->getTablesIterator(global_context, {});
-
-    /// We only shutdown the table, The tables is cleaned up when destructed database
-    for (; iterator->isValid(); iterator->next())
-        iterator->table()->shutdown();
+    assertCalledFromSyncThreadOrDrop("create table");
+    Base::createTable(context, name, table, query);
 }
 
-bool DatabaseMaterializeMySQL::empty() const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::dropTable(const Context & context, const String & name, bool no_delay)
 {
-    return nested_database->empty();
+    assertCalledFromSyncThreadOrDrop("drop table");
+    Base::dropTable(context, name, no_delay);
 }
 
-String DatabaseMaterializeMySQL::getDataPath() const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::attachTable(const String & name, const StoragePtr & table, const String & relative_table_path)
 {
-    return nested_database->getDataPath();
+    assertCalledFromSyncThreadOrDrop("attach table");
+    Base::attachTable(name, table, relative_table_path);
 }
 
-String DatabaseMaterializeMySQL::getMetadataPath() const
+template<typename Base>
+StoragePtr DatabaseMaterializeMySQL<Base>::detachTable(const String & name)
 {
-    return nested_database->getMetadataPath();
+    assertCalledFromSyncThreadOrDrop("detach table");
+    return Base::detachTable(name);
 }
 
-String DatabaseMaterializeMySQL::getTableDataPath(const String & table_name) const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::renameTable(const Context & context, const String & name, IDatabase & to_database, const String & to_name, bool exchange, bool dictionary)
 {
-    return nested_database->getTableDataPath(table_name);
-}
-
-String DatabaseMaterializeMySQL::getTableDataPath(const ASTCreateQuery & query) const
-{
-    return nested_database->getTableDataPath(query);
-}
-
-String DatabaseMaterializeMySQL::getObjectMetadataPath(const String & table_name) const
-{
-    return nested_database->getObjectMetadataPath(table_name);
-}
-
-UUID DatabaseMaterializeMySQL::tryGetTableUUID(const String & table_name) const
-{
-    return nested_database->tryGetTableUUID(table_name);
-}
-
-time_t DatabaseMaterializeMySQL::getObjectMetadataModificationTime(const String & name) const
-{
-    return nested_database->getObjectMetadataModificationTime(name);
-}
-
-void DatabaseMaterializeMySQL::createTable(const Context & context, const String & name, const StoragePtr & table, const ASTPtr & query)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support create table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->createTable(context, name, table, query);
-}
-
-void DatabaseMaterializeMySQL::dropTable(const Context & context, const String & name, bool no_delay)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support drop table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->dropTable(context, name, no_delay);
-}
-
-void DatabaseMaterializeMySQL::attachTable(const String & name, const StoragePtr & table, const String & relative_table_path)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support attach table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->attachTable(name, table, relative_table_path);
-}
-
-StoragePtr DatabaseMaterializeMySQL::detachTable(const String & name)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support detach table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    return nested_database->detachTable(name);
-}
-
-void DatabaseMaterializeMySQL::renameTable(const Context & context, const String & name, IDatabase & to_database, const String & to_name, bool exchange, bool dictionary)
-{
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support rename table.", ErrorCodes::NOT_IMPLEMENTED);
+    assertCalledFromSyncThreadOrDrop("rename table");
 
     if (exchange)
         throw Exception("MaterializeMySQL database not support exchange table.", ErrorCodes::NOT_IMPLEMENTED);
@@ -176,57 +135,37 @@ void DatabaseMaterializeMySQL::renameTable(const Context & context, const String
     if (dictionary)
         throw Exception("MaterializeMySQL database not support rename dictionary.", ErrorCodes::NOT_IMPLEMENTED);
 
-    if (to_database.getDatabaseName() != getDatabaseName())
+    if (to_database.getDatabaseName() != Base::getDatabaseName())
         throw Exception("Cannot rename with other database for MaterializeMySQL database.", ErrorCodes::NOT_IMPLEMENTED);
 
-    nested_database->renameTable(context, name, *nested_database, to_name, exchange, dictionary);
+    Base::renameTable(context, name, *this, to_name, exchange, dictionary);
 }
 
-void DatabaseMaterializeMySQL::alterTable(const Context & context, const StorageID & table_id, const StorageInMemoryMetadata & metadata)
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::alterTable(const Context & context, const StorageID & table_id, const StorageInMemoryMetadata & metadata)
 {
-    if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
-        throw Exception("MaterializeMySQL database not support alter table.", ErrorCodes::NOT_IMPLEMENTED);
-
-    nested_database->alterTable(context, table_id, metadata);
+    assertCalledFromSyncThreadOrDrop("alter table");
+    Base::alterTable(context, table_id, metadata);
 }
 
-bool DatabaseMaterializeMySQL::shouldBeEmptyOnDetach() const
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::drop(const Context & context)
 {
-    return false;
+    /// Remove metadata info
+    Poco::File metadata(Base::getMetadataPath() + "/.metadata");
+
+    if (metadata.exists())
+        metadata.remove(false);
+
+    Base::drop(context);
 }
 
-void DatabaseMaterializeMySQL::drop(const Context & context)
-{
-    if (nested_database->shouldBeEmptyOnDetach())
-    {
-        for (auto iterator = nested_database->getTablesIterator(context, {}); iterator->isValid(); iterator->next())
-        {
-            TableExclusiveLockHolder table_lock = iterator->table()->lockExclusively(
-                context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
-
-            nested_database->dropTable(context, iterator->name(), true);
-        }
-
-        /// Remove metadata info
-        Poco::File metadata(getMetadataPath() + "/.metadata");
-
-        if (metadata.exists())
-            metadata.remove(false);
-    }
-
-    nested_database->drop(context);
-}
-
-bool DatabaseMaterializeMySQL::isTableExist(const String & name, const Context & context) const
-{
-    return nested_database->isTableExist(name, context);
-}
-
-StoragePtr DatabaseMaterializeMySQL::tryGetTable(const String & name, const Context & context) const
+template<typename Base>
+StoragePtr DatabaseMaterializeMySQL<Base>::tryGetTable(const String & name, const Context & context) const
 {
     if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
     {
-        StoragePtr nested_storage = nested_database->tryGetTable(name, context);
+        StoragePtr nested_storage = Base::tryGetTable(name, context);
 
         if (!nested_storage)
             return {};
@@ -234,20 +173,71 @@ StoragePtr DatabaseMaterializeMySQL::tryGetTable(const String & name, const Cont
         return std::make_shared<StorageMaterializeMySQL>(std::move(nested_storage), this);
     }
 
-    return nested_database->tryGetTable(name, context);
+    return Base::tryGetTable(name, context);
 }
 
-DatabaseTablesIteratorPtr DatabaseMaterializeMySQL::getTablesIterator(const Context & context, const FilterByNameFunction & filter_by_table_name)
+template<typename Base>
+DatabaseTablesIteratorPtr DatabaseMaterializeMySQL<Base>::getTablesIterator(const Context & context, const DatabaseOnDisk::FilterByNameFunction & filter_by_table_name)
 {
     if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
     {
-        DatabaseTablesIteratorPtr iterator = nested_database->getTablesIterator(context, filter_by_table_name);
+        DatabaseTablesIteratorPtr iterator = Base::getTablesIterator(context, filter_by_table_name);
         return std::make_unique<DatabaseMaterializeTablesIterator>(std::move(iterator), this);
     }
 
-    return nested_database->getTablesIterator(context, filter_by_table_name);
+    return Base::getTablesIterator(context, filter_by_table_name);
 }
 
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::assertCalledFromSyncThreadOrDrop(const char * method) const
+{
+    if (!MaterializeMySQLSyncThread::isMySQLSyncThread() && started_up)
+        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "MaterializeMySQL database not support {}", method);
+}
+
+template<typename Base>
+void DatabaseMaterializeMySQL<Base>::shutdownSynchronizationThread()
+{
+    materialize_thread.stopSynchronization();
+    started_up = false;
+}
+
+template<typename Database, template<class> class Helper, typename... Args>
+auto castToMaterializeMySQLAndCallHelper(Database * database, Args && ... args)
+{
+    using Ordinary = DatabaseMaterializeMySQL<DatabaseOrdinary>;
+    using Atomic = DatabaseMaterializeMySQL<DatabaseAtomic>;
+    using ToOrdinary = typename std::conditional_t<std::is_const_v<Database>, const Ordinary *, Ordinary *>;
+    using ToAtomic = typename std::conditional_t<std::is_const_v<Database>, const Atomic *, Atomic *>;
+    if (auto * database_materialize = typeid_cast<ToOrdinary>(database))
+        return (database_materialize->*Helper<Ordinary>::v)(std::forward<Args>(args)...);
+    if (auto * database_materialize = typeid_cast<ToAtomic>(database))
+        return (database_materialize->*Helper<Atomic>::v)(std::forward<Args>(args)...);
+
+    throw Exception("LOGICAL_ERROR: cannot cast to DatabaseMaterializeMySQL, it is a bug.", ErrorCodes::LOGICAL_ERROR);
+}
+
+template<typename T> struct HelperSetException { static constexpr auto v = &T::setException; };
+void setSynchronizationThreadException(const DatabasePtr & materialize_mysql_db, const std::exception_ptr & exception)
+{
+    castToMaterializeMySQLAndCallHelper<IDatabase, HelperSetException>(materialize_mysql_db.get(), exception);
+}
+
+template<typename T> struct HelperStopSync { static constexpr auto v = &T::shutdownSynchronizationThread; };
+void stopDatabaseSynchronization(const DatabasePtr & materialize_mysql_db)
+{
+    castToMaterializeMySQLAndCallHelper<IDatabase, HelperStopSync>(materialize_mysql_db.get());
+}
+
+template<typename T> struct HelperRethrow { static constexpr auto v = &T::rethrowExceptionIfNeed; };
+void rethrowSyncExceptionIfNeed(const IDatabase * materialize_mysql_db)
+{
+    castToMaterializeMySQLAndCallHelper<const IDatabase, HelperRethrow>(materialize_mysql_db);
+}
+
+template class DatabaseMaterializeMySQL<DatabaseOrdinary>;
+template class DatabaseMaterializeMySQL<DatabaseAtomic>;
+
 }
 
 #endif
diff --git a/src/Databases/MySQL/DatabaseMaterializeMySQL.h b/src/Databases/MySQL/DatabaseMaterializeMySQL.h
index 799db65b481..e1229269a33 100644
--- a/src/Databases/MySQL/DatabaseMaterializeMySQL.h
+++ b/src/Databases/MySQL/DatabaseMaterializeMySQL.h
@@ -17,48 +17,34 @@ namespace DB
  *
  *  All table structure and data will be written to the local file system
  */
-class DatabaseMaterializeMySQL : public IDatabase
+template<typename Base>
+class DatabaseMaterializeMySQL : public Base
 {
 public:
+
     DatabaseMaterializeMySQL(
-        const Context & context, const String & database_name_, const String & metadata_path_,
-        const IAST * database_engine_define_, const String & mysql_database_name_, mysqlxx::Pool && pool_,
+        const Context & context, const String & database_name_, const String & metadata_path_, UUID uuid,
+        const String & mysql_database_name_, mysqlxx::Pool && pool_,
         MySQLClient && client_, std::unique_ptr<MaterializeMySQLSettings> settings_);
 
     void rethrowExceptionIfNeed() const;
 
     void setException(const std::exception_ptr & exception);
 protected:
-    const Context & global_context;
 
-    ASTPtr engine_define;
-    DatabasePtr nested_database;
     std::unique_ptr<MaterializeMySQLSettings> settings;
 
-    Poco::Logger * log;
     MaterializeMySQLSyncThread materialize_thread;
 
     std::exception_ptr exception;
 
+    std::atomic_bool started_up{false};
+
 public:
     String getEngineName() const override { return "MaterializeMySQL"; }
 
-    ASTPtr getCreateDatabaseQuery() const override;
-
     void loadStoredObjects(Context & context, bool has_force_restore_data_flag, bool force_attach) override;
 
-    void shutdown() override;
-
-    bool empty() const override;
-
-    String getDataPath() const override;
-
-    String getTableDataPath(const String & table_name) const override;
-
-    String getTableDataPath(const ASTCreateQuery & query) const override;
-
-    UUID tryGetTableUUID(const String & table_name) const override;
-
     void createTable(const Context & context, const String & name, const StoragePtr & table, const ASTPtr & query) override;
 
     void dropTable(const Context & context, const String & name, bool no_delay) override;
@@ -71,23 +57,22 @@ public:
 
     void alterTable(const Context & context, const StorageID & table_id, const StorageInMemoryMetadata & metadata) override;
 
-    time_t getObjectMetadataModificationTime(const String & name) const override;
-
-    String getMetadataPath() const override;
-
-    String getObjectMetadataPath(const String & table_name) const override;
-
-    bool shouldBeEmptyOnDetach() const override;
-
     void drop(const Context & context) override;
 
-    bool isTableExist(const String & name, const Context & context) const override;
-
     StoragePtr tryGetTable(const String & name, const Context & context) const override;
 
-    DatabaseTablesIteratorPtr getTablesIterator(const Context & context, const FilterByNameFunction & filter_by_table_name) override;
+    DatabaseTablesIteratorPtr getTablesIterator(const Context & context, const DatabaseOnDisk::FilterByNameFunction & filter_by_table_name) override;
+
+    void assertCalledFromSyncThreadOrDrop(const char * method) const;
+
+    void shutdownSynchronizationThread();
 };
 
+
+void setSynchronizationThreadException(const DatabasePtr & materialize_mysql_db, const std::exception_ptr & exception);
+void stopDatabaseSynchronization(const DatabasePtr & materialize_mysql_db);
+void rethrowSyncExceptionIfNeed(const IDatabase * materialize_mysql_db);
+
 }
 
 #endif
diff --git a/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h b/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h
index 86a5cbf8206..54031de40a2 100644
--- a/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h
+++ b/src/Databases/MySQL/DatabaseMaterializeTablesIterator.h
@@ -2,7 +2,6 @@
 
 #include <Databases/IDatabase.h>
 #include <Storages/StorageMaterializeMySQL.h>
-#include <Databases/MySQL/DatabaseMaterializeMySQL.h>
 
 namespace DB
 {
@@ -30,7 +29,7 @@ public:
 
     UUID uuid() const override { return nested_iterator->uuid(); }
 
-    DatabaseMaterializeTablesIterator(DatabaseTablesIteratorPtr nested_iterator_, DatabaseMaterializeMySQL * database_)
+    DatabaseMaterializeTablesIterator(DatabaseTablesIteratorPtr nested_iterator_, const IDatabase * database_)
         : nested_iterator(std::move(nested_iterator_)), database(database_)
     {
     }
@@ -38,8 +37,7 @@ public:
 private:
     mutable std::vector<StoragePtr> tables;
     DatabaseTablesIteratorPtr nested_iterator;
-    DatabaseMaterializeMySQL * database;
-
+    const IDatabase * database;
 };
 
 }
diff --git a/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp b/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp
index bc7c9e4c554..1da033fa4b3 100644
--- a/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp
+++ b/src/Databases/MySQL/MaterializeMySQLSyncThread.cpp
@@ -71,15 +71,6 @@ static BlockIO tryToExecuteQuery(const String & query_to_execute, Context & quer
     }
 }
 
-static inline DatabaseMaterializeMySQL & getDatabase(const String & database_name)
-{
-    DatabasePtr database = DatabaseCatalog::instance().getDatabase(database_name);
-
-    if (DatabaseMaterializeMySQL * database_materialize = typeid_cast<DatabaseMaterializeMySQL *>(database.get()))
-        return *database_materialize;
-
-    throw Exception("LOGICAL_ERROR: cannot cast to DatabaseMaterializeMySQL, it is a bug.", ErrorCodes::LOGICAL_ERROR);
-}
 
 MaterializeMySQLSyncThread::~MaterializeMySQLSyncThread()
 {
@@ -190,7 +181,8 @@ void MaterializeMySQLSyncThread::synchronization()
     {
         client.disconnect();
         tryLogCurrentException(log);
-        getDatabase(database_name).setException(std::current_exception());
+        auto db = DatabaseCatalog::instance().getDatabase(database_name);
+        setSynchronizationThreadException(db, std::current_exception());
     }
 }
 
@@ -343,7 +335,7 @@ std::optional<MaterializeMetadata> MaterializeMySQLSyncThread::prepareSynchroniz
             opened_transaction = false;
 
             MaterializeMetadata metadata(
-                connection, getDatabase(database_name).getMetadataPath() + "/.metadata", mysql_database_name, opened_transaction);
+                connection, DatabaseCatalog::instance().getDatabase(database_name)->getMetadataPath() + "/.metadata", mysql_database_name, opened_transaction);
 
             if (!metadata.need_dumping_tables.empty())
             {
diff --git a/src/Dictionaries/HTTPDictionarySource.cpp b/src/Dictionaries/HTTPDictionarySource.cpp
index 18a97f34486..67bd8462036 100644
--- a/src/Dictionaries/HTTPDictionarySource.cpp
+++ b/src/Dictionaries/HTTPDictionarySource.cpp
@@ -2,6 +2,7 @@
 #include <DataStreams/IBlockOutputStream.h>
 #include <DataStreams/OwningBlockInputStream.h>
 #include <IO/ConnectionTimeouts.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <IO/ReadWriteBufferFromHTTP.h>
 #include <IO/WriteBufferFromOStream.h>
 #include <IO/WriteBufferFromString.h>
diff --git a/src/Dictionaries/XDBCDictionarySource.cpp b/src/Dictionaries/XDBCDictionarySource.cpp
index 832c30ed4b7..89df4b606fe 100644
--- a/src/Dictionaries/XDBCDictionarySource.cpp
+++ b/src/Dictionaries/XDBCDictionarySource.cpp
@@ -6,6 +6,7 @@
 #include <Formats/FormatFactory.h>
 #include <IO/ReadWriteBufferFromHTTP.h>
 #include <IO/WriteHelpers.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <Interpreters/Context.h>
 #include <Poco/Net/HTTPRequest.h>
 #include <Poco/Util/AbstractConfiguration.h>
diff --git a/src/Disks/S3/DiskS3.cpp b/src/Disks/S3/DiskS3.cpp
index 3f3d23938ba..4786c05f8b0 100644
--- a/src/Disks/S3/DiskS3.cpp
+++ b/src/Disks/S3/DiskS3.cpp
@@ -17,6 +17,8 @@
 #include <Common/createHardLink.h>
 #include <Common/quoteString.h>
 #include <Common/thread_local_rng.h>
+#include <Common/ThreadPool.h>
+#include <common/logger_useful.h>
 
 #include <aws/s3/model/CopyObjectRequest.h>
 #include <aws/s3/model/DeleteObjectsRequest.h>
@@ -346,11 +348,11 @@ public:
         DiskS3::Metadata metadata_,
         const String & s3_path_,
         std::optional<DiskS3::ObjectMetadata> object_metadata_,
-        bool is_multipart,
         size_t min_upload_part_size,
+        size_t max_single_part_upload_size,
         size_t buf_size_)
         : WriteBufferFromFileBase(buf_size_, nullptr, 0)
-        , impl(WriteBufferFromS3(client_ptr_, bucket_, metadata_.s3_root_path + s3_path_, min_upload_part_size, is_multipart,std::move(object_metadata_), buf_size_))
+        , impl(WriteBufferFromS3(client_ptr_, bucket_, metadata_.s3_root_path + s3_path_, min_upload_part_size, max_single_part_upload_size,std::move(object_metadata_), buf_size_))
         , metadata(std::move(metadata_))
         , s3_path(s3_path_)
     {
@@ -540,7 +542,7 @@ DiskS3::DiskS3(
     String s3_root_path_,
     String metadata_path_,
     size_t min_upload_part_size_,
-    size_t min_multi_part_upload_size_,
+    size_t max_single_part_upload_size_,
     size_t min_bytes_for_seek_,
     bool send_metadata_)
     : IDisk(std::make_unique<AsyncExecutor>())
@@ -551,7 +553,7 @@ DiskS3::DiskS3(
     , s3_root_path(std::move(s3_root_path_))
     , metadata_path(std::move(metadata_path_))
     , min_upload_part_size(min_upload_part_size_)
-    , min_multi_part_upload_size(min_multi_part_upload_size_)
+    , max_single_part_upload_size(max_single_part_upload_size_)
     , min_bytes_for_seek(min_bytes_for_seek_)
     , send_metadata(send_metadata_)
 {
@@ -663,7 +665,7 @@ std::unique_ptr<ReadBufferFromFileBase> DiskS3::readFile(const String & path, si
     return std::make_unique<SeekAvoidingReadBuffer>(std::move(reader), min_bytes_for_seek);
 }
 
-std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path, size_t buf_size, WriteMode mode, size_t estimated_size, size_t)
+std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path, size_t buf_size, WriteMode mode, size_t, size_t)
 {
     bool exist = exists(path);
     if (exist && readMeta(path).read_only)
@@ -672,7 +674,6 @@ std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path,
     /// Path to store new S3 object.
     auto s3_path = getRandomName();
     auto object_metadata = createObjectMetadata(path);
-    bool is_multipart = estimated_size >= min_multi_part_upload_size;
     if (!exist || mode == WriteMode::Rewrite)
     {
         /// If metadata file exists - remove and create new.
@@ -685,7 +686,8 @@ std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path,
 
         LOG_DEBUG(&Poco::Logger::get("DiskS3"), "Write to file by path: {}. New S3 path: {}", backQuote(metadata_path + path), s3_root_path + s3_path);
 
-        return std::make_unique<WriteIndirectBufferFromS3>(client, bucket, metadata, s3_path, object_metadata, is_multipart, min_upload_part_size, buf_size);
+        return std::make_unique<WriteIndirectBufferFromS3>(
+            client, bucket, metadata, s3_path, object_metadata, min_upload_part_size, max_single_part_upload_size, buf_size);
     }
     else
     {
@@ -694,7 +696,8 @@ std::unique_ptr<WriteBufferFromFileBase> DiskS3::writeFile(const String & path,
         LOG_DEBUG(&Poco::Logger::get("DiskS3"), "Append to file by path: {}. New S3 path: {}. Existing S3 objects: {}.",
             backQuote(metadata_path + path), s3_root_path + s3_path, metadata.s3_objects.size());
 
-        return std::make_unique<WriteIndirectBufferFromS3>(client, bucket, metadata, s3_path, object_metadata, is_multipart, min_upload_part_size, buf_size);
+        return std::make_unique<WriteIndirectBufferFromS3>(
+            client, bucket, metadata, s3_path, object_metadata, min_upload_part_size, max_single_part_upload_size, buf_size);
     }
 }
 
diff --git a/src/Disks/S3/DiskS3.h b/src/Disks/S3/DiskS3.h
index 5bc8d51abb2..f62c603adda 100644
--- a/src/Disks/S3/DiskS3.h
+++ b/src/Disks/S3/DiskS3.h
@@ -34,7 +34,7 @@ public:
         String s3_root_path_,
         String metadata_path_,
         size_t min_upload_part_size_,
-        size_t min_multi_part_upload_size_,
+        size_t max_single_part_upload_size_,
         size_t min_bytes_for_seek_,
         bool send_metadata_);
 
@@ -133,7 +133,7 @@ private:
     const String s3_root_path;
     const String metadata_path;
     size_t min_upload_part_size;
-    size_t min_multi_part_upload_size;
+    size_t max_single_part_upload_size;
     size_t min_bytes_for_seek;
     bool send_metadata;
 
diff --git a/src/Disks/S3/registerDiskS3.cpp b/src/Disks/S3/registerDiskS3.cpp
index ff92c295fe4..fd658d95327 100644
--- a/src/Disks/S3/registerDiskS3.cpp
+++ b/src/Disks/S3/registerDiskS3.cpp
@@ -148,7 +148,7 @@ void registerDiskS3(DiskFactory & factory)
             uri.key,
             metadata_path,
             context.getSettingsRef().s3_min_upload_part_size,
-            config.getUInt64(config_prefix + ".min_multi_part_upload_size", 10 * 1024 * 1024),
+            context.getSettingsRef().s3_max_single_part_upload_size,
             config.getUInt64(config_prefix + ".min_bytes_for_seek", 1024 * 1024),
             config.getBool(config_prefix + ".send_object_metadata", false));
 
diff --git a/src/Formats/FormatFactory.h b/src/Formats/FormatFactory.h
index 3b0811d579a..0fe6f19f0b7 100644
--- a/src/Formats/FormatFactory.h
+++ b/src/Formats/FormatFactory.h
@@ -16,7 +16,6 @@ namespace DB
 
 class Block;
 class Context;
-struct FormatSettings;
 struct Settings;
 struct FormatFactorySettings;
 
diff --git a/src/Functions/FunctionBase64Conversion.h b/src/Functions/FunctionBase64Conversion.h
index 4bc2a779cf4..4887e9fe8cc 100644
--- a/src/Functions/FunctionBase64Conversion.h
+++ b/src/Functions/FunctionBase64Conversion.h
@@ -158,7 +158,9 @@ public:
             __msan_unpoison(dst_pos, outlen);
 
             source += srclen + 1;
-            dst_pos += outlen + 1;
+            dst_pos += outlen;
+            *dst_pos = '\0';
+            dst_pos += 1;
 
             dst_offsets[row] = dst_pos - dst;
             src_offset_prev = src_offsets[row];
diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 8c684d3578f..1d1711b79d2 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -3,6 +3,8 @@
 // Include this first, because `#define _asan_poison_address` from
 // llvm/Support/Compiler.h conflicts with its forward declaration in
 // sanitizer/asan_interface.h
+#include <memory>
+#include <type_traits>
 #include <Common/Arena.h>
 
 #include <DataTypes/DataTypesNumber.h>
@@ -20,6 +22,7 @@
 #include <Columns/ColumnFixedString.h>
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnAggregateFunction.h>
+#include "Core/DecimalFunctions.h"
 #include "IFunctionImpl.h"
 #include "FunctionHelpers.h"
 #include "IsOperation.h"
@@ -43,6 +46,8 @@
 #    pragma GCC diagnostic pop
 #endif
 
+#include <cassert>
+
 
 namespace DB
 {
@@ -181,8 +186,7 @@ inline constexpr const auto & undec(const T & x)
 }
 
 /// Binary operations for Decimals need scale args
-/// +|- scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
-/// *   no agrs scale. ScaleR = Scale1 + Scale2;
+/// +|-* scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
 /// /   first arg scale. ScaleR = Scale1 (scale_a = DecimalType<B>::getScale()).
 template <template <typename, typename> typename Operation, typename ResultType_, bool check_overflow = true>
 struct DecimalBinaryOperation
@@ -205,13 +209,16 @@ struct DecimalBinaryOperation
         DivideIntegralImpl<NativeResultType, NativeResultType>, /// substitute divide by intDiv (throw on division by zero)
         Operation<NativeResultType, NativeResultType>>;
 
-    using ArrayC = typename ColumnDecimal<ResultType>::Container;
+    using ArrayC = typename std::conditional_t<IsDecimalNumber<ResultType>,
+        ColumnDecimal<ResultType>,
+        ColumnVector<ResultType>>::Container;
 
     template <bool is_decimal_a, bool is_decimal_b, typename ArrayA, typename ArrayB>
     static void NO_INLINE vectorVector(const ArrayA & a, const ArrayB & b, ArrayC & c,
                                        NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
     {
         size_t size = a.size();
+
         if constexpr (is_plus_minus_compare)
         {
             if (scale_a != 1)
@@ -227,6 +234,28 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -261,6 +290,28 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -295,6 +346,29 @@ struct DecimalBinaryOperation
                 return;
             }
         }
+        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
+        {
+            if (scale_a != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
+                    else
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    if constexpr(IsDecimalNumber<ResultType>)
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
+                    else
+                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
+
+                return;
+            }
+        }
+
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -343,37 +417,36 @@ private:
     template <bool scale_left>
     static NO_SANITIZE_UNDEFINED NativeResultType applyScaled(NativeResultType a, NativeResultType b, NativeResultType scale)
     {
-        if constexpr (is_plus_minus_compare)
+        static_assert(is_plus_minus_compare);
+        NativeResultType res;
+
+        if constexpr (check_overflow)
         {
-            NativeResultType res;
+            bool overflow = false;
 
-            if constexpr (check_overflow)
-            {
-                bool overflow = false;
-                if constexpr (scale_left)
-                    overflow |= common::mulOverflow(a, scale, a);
-                else
-                    overflow |= common::mulOverflow(b, scale, b);
-
-                if constexpr (can_overflow)
-                    overflow |= Op::template apply<NativeResultType>(a, b, res);
-                else
-                    res = Op::template apply<NativeResultType>(a, b);
-
-                if (overflow)
-                    throw Exception("Decimal math overflow", ErrorCodes::DECIMAL_OVERFLOW);
-            }
+            if constexpr (scale_left)
+                overflow |= common::mulOverflow(a, scale, a);
             else
-            {
-                if constexpr (scale_left)
-                    a *= scale;
-                else
-                    b *= scale;
-                res = Op::template apply<NativeResultType>(a, b);
-            }
+                overflow |= common::mulOverflow(b, scale, b);
 
-            return res;
+            if constexpr (can_overflow)
+                overflow |= Op::template apply<NativeResultType>(a, b, res);
+            else
+                res = Op::template apply<NativeResultType>(a, b);
+
+            if (overflow)
+                throw Exception("Decimal math overflow", ErrorCodes::DECIMAL_OVERFLOW);
         }
+        else
+        {
+            if constexpr (scale_left)
+                a *= scale;
+            else
+                b *= scale;
+            res = Op::template apply<NativeResultType>(a, b);
+        }
+
+        return res;
     }
 
     template <bool is_decimal_a>
@@ -467,26 +540,36 @@ public:
         Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>, RightDataType>,
         Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
         Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
+
+        /// e.g Decimal * Float64 = Float64
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>,
+
         /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
         Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
         Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
+
         /// number <op> number -> see corresponding impl
         Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
             DataTypeFromFieldType<typename Op::ResultType>>,
+
         /// Date + Integral -> Date
         /// Integral + Date -> Date
         Case<IsOperation<Operation>::plus, Switch<
             Case<IsIntegral<RightDataType>, LeftDataType>,
             Case<IsIntegral<LeftDataType>, RightDataType>>>,
+
         /// Date - Date     -> Int32
         /// Date - Integral -> Date
         Case<IsOperation<Operation>::minus, Switch<
             Case<std::is_same_v<LeftDataType, RightDataType>, DataTypeInt32>,
             Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, LeftDataType>>>,
+
         /// least(Date, Date) -> Date
         /// greatest(Date, Date) -> Date
         Case<std::is_same_v<LeftDataType, RightDataType> && (IsOperation<Operation>::least || IsOperation<Operation>::greatest),
             LeftDataType>,
+
         /// Date % Int32 -> Int32
         /// Date % Float -> Float64
         Case<IsOperation<Operation>::modulo, Switch<
@@ -494,7 +577,6 @@ public:
             Case<IsDateOrDateTime<LeftDataType> && IsFloatingPoint<RightDataType>, DataTypeFloat64>>>>;
 };
 
-
 template <template <typename, typename> class Op, typename Name, bool valid_on_default_arguments = true>
 class FunctionBinaryArithmetic : public IFunction
 {
@@ -536,7 +618,13 @@ class FunctionBinaryArithmetic : public IFunction
     template <typename F>
     static bool castBothTypes(const IDataType * left, const IDataType * right, F && f)
     {
-        return castType(left, [&](const auto & left_) { return castType(right, [&](const auto & right_) { return f(left_, right_); }); });
+        return castType(left, [&](const auto & left_)
+        {
+            return castType(right, [&](const auto & right_)
+            {
+                return f(left_, right_);
+            });
+        });
     }
 
     static FunctionOverloadResolverPtr
@@ -790,10 +878,12 @@ public:
         }
 
         DataTypePtr type_res;
-        bool valid = castBothTypes(arguments[0].get(), arguments[1].get(), [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(arguments[0].get(), arguments[1].get(), [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
@@ -810,6 +900,7 @@ public:
             else
             {
                 using ResultDataType = typename BinaryOperationTraits<Op, LeftDataType, RightDataType>::ResultDataType;
+
                 if constexpr (!std::is_same_v<ResultDataType, InvalidType>)
                 {
                     if constexpr (IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>)
@@ -817,6 +908,10 @@ public:
                         ResultDataType result_type = decimalResultType<is_multiply, is_division>(left, right);
                         type_res = std::make_shared<ResultDataType>(result_type.getPrecision(), result_type.getScale());
                     }
+                    else if constexpr ((IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>) ||
+                        (IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>))
+                        type_res = std::make_shared<std::conditional_t<IsFloatingPoint<LeftDataType>,
+                            LeftDataType, RightDataType>>();
                     else if constexpr (IsDataTypeDecimal<LeftDataType>)
                         type_res = std::make_shared<LeftDataType>(left.getPrecision(), left.getScale());
                     else if constexpr (IsDataTypeDecimal<RightDataType>)
@@ -841,10 +936,15 @@ public:
             }
             return false;
         });
-        if (!valid)
-            throw Exception("Illegal types " + arguments[0]->getName() + " and " + arguments[1]->getName() + " of arguments of function " + String(name),
-                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-        return type_res;
+
+        if (valid)
+            return type_res;
+
+        throw Exception(
+            "Illegal types " + arguments[0]->getName() +
+            " and " + arguments[1]->getName() +
+            " of arguments of function " + String(name),
+            ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
     }
 
     ColumnPtr executeFixedString(const ColumnsWithTypeAndName & arguments) const
@@ -922,6 +1022,18 @@ public:
         return nullptr;
     }
 
+    template <class T, class ResultDataType>
+    static auto helperGetOrConvert(const auto & col_const, const auto & col)
+    {
+        using ResultType = typename ResultDataType::FieldType;
+        using NativeResultType = typename NativeType<ResultType>::Type;
+
+        if constexpr (IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
+            return DecimalUtils::convertTo<NativeResultType>(col_const->template getValue<T>(), col.getScale());
+        else
+            return col_const->template getValue<T>();
+    }
+
     template <typename A, typename B>
     ColumnPtr executeNumeric(const ColumnsWithTypeAndName & arguments, const A & left, const B & right) const
     {
@@ -954,59 +1066,97 @@ public:
                 using NativeResultType = typename NativeType<ResultType>::Type;
                 using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
                 using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
+                using FieldType = typename ResultDataType::FieldType;
 
-                ResultDataType type = decimalResultType<is_multiply, is_division>(left, right);
+                static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
+                static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
+                static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
 
-                static constexpr const bool dec_a = IsDecimalNumber<T0>;
-                static constexpr const bool dec_b = IsDecimalNumber<T1>;
+                const ResultDataType type = [&] {
+                    if constexpr (left_is_decimal && IsFloatingPoint<RightDataType>)
+                        return RightDataType();
+                    else if constexpr (right_is_decimal && IsFloatingPoint<LeftDataType>)
+                        return LeftDataType();
+                    else
+                        return decimalResultType<is_multiply, is_division>(left, right);
+                }();
+
+                FieldType left_scale;
+                FieldType right_scale;
 
-                typename ResultDataType::FieldType scale_a = type.scaleFactorFor(left, is_multiply);
-                typename ResultDataType::FieldType scale_b = type.scaleFactorFor(right, is_multiply || is_division);
                 if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                    scale_a = right.getScaleMultiplier();
+                    left_scale = right.getScaleMultiplier();
+                else if constexpr (result_is_decimal)
+                    left_scale = type.scaleFactorFor(left, is_multiply);
+                else if constexpr(left_is_decimal) // BUG precision loss
+                    left_scale = left.getScale();
+                else
+                    left_scale = 1; //won't be used, just to silence the warning
+
+                if constexpr (result_is_decimal)
+                    right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+                else if constexpr(right_is_decimal)
+                    right_scale = right.getScale();
+                else
+                    right_scale = 1; //same
 
                 /// non-vector result
                 if (col_left_const && col_right_const)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
                     auto res = check_decimal_overflow ?
-                        OpImplCheck::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b) :
-                        OpImpl::template constantConstant<dec_a, dec_b>(const_a, const_b, scale_a, scale_b);
+                        OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
+                            const_a, const_b, left_scale, right_scale) :
+                        OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
+                            const_a, const_b, left_scale, right_scale);
 
-                    return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
+                    if constexpr (result_is_decimal)
+                        return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
                             col_left_const->size(), toField(res, type.getScale()));
+                    else
+                         return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
                 }
 
-                col_res = ColVecResult::create(0, type.getScale());
+                if constexpr (result_is_decimal)
+                    col_res = ColVecResult::create(0, type.getScale());
+                else
+                    col_res = ColVecResult::create(0);
+
                 auto & vec_res = col_res->getData();
                 vec_res.resize(col_left_raw->size());
 
                 if (col_left && col_right)
                 {
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorVector<dec_a, dec_b>(col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template vectorVector<dec_a, dec_b>(col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left_const && col_right)
                 {
-                    NativeResultType const_a = col_left_const->template getValue<T0>();
+                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
+                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template constantVector<dec_a, dec_b>(const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                        OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
+                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
                 }
                 else if (col_left && col_right_const)
                 {
-                    NativeResultType const_b = col_right_const->template getValue<T1>();
+                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
                     if (check_decimal_overflow)
-                        OpImplCheck::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                        OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
                     else
-                        OpImpl::template vectorConstant<dec_a, dec_b>(col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                        OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
+                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
                 }
                 else
                     return nullptr;
@@ -1073,10 +1223,12 @@ public:
         const auto * left_generic = left_argument.type.get();
         const auto * right_generic = right_argument.type.get();
         ColumnPtr res;
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+
+        const bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
         {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
+
             if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
diff --git a/src/Functions/FunctionFactory.cpp b/src/Functions/FunctionFactory.cpp
index fbc8e11a9c9..945a78c2734 100644
--- a/src/Functions/FunctionFactory.cpp
+++ b/src/Functions/FunctionFactory.cpp
@@ -8,6 +8,8 @@
 
 #include <IO/WriteHelpers.h>
 
+#include <AggregateFunctions/AggregateFunctionFactory.h>
+
 namespace DB
 {
 
@@ -46,12 +48,15 @@ FunctionOverloadResolverImplPtr FunctionFactory::getImpl(
     auto res = tryGetImpl(name, context);
     if (!res)
     {
+        String extra_info;
+        if (AggregateFunctionFactory::instance().hasNameOrAlias(name))
+            extra_info = ". There is an aggregate function with the same name, but ordinary function is expected here";
+
         auto hints = this->getHints(name);
         if (!hints.empty())
-            throw Exception("Unknown function " + name + ". Maybe you meant: " + toString(hints),
-                            ErrorCodes::UNKNOWN_FUNCTION);
+            throw Exception(ErrorCodes::UNKNOWN_FUNCTION, "Unknown function {}{}. Maybe you meant: {}", name, extra_info, toString(hints));
         else
-            throw Exception("Unknown function " + name, ErrorCodes::UNKNOWN_FUNCTION);
+            throw Exception(ErrorCodes::UNKNOWN_FUNCTION, "Unknown function {}{}", name, extra_info);
     }
     return res;
 }
diff --git a/src/Functions/FunctionFactory.h b/src/Functions/FunctionFactory.h
index 7872c192b41..7990e78daf8 100644
--- a/src/Functions/FunctionFactory.h
+++ b/src/Functions/FunctionFactory.h
@@ -2,7 +2,6 @@
 
 #include <Functions/IFunctionAdaptors.h>
 #include <Common/IFactoryWithAliases.h>
-#include <Interpreters/Context.h>
 
 #include <functional>
 #include <memory>
diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp
index df962800385..257b852ecd8 100644
--- a/src/Functions/FunctionsConversion.cpp
+++ b/src/Functions/FunctionsConversion.cpp
@@ -1,16 +1,10 @@
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionsConversion.h>
-#include <Interpreters/Context.h>
 
 
 namespace DB
 {
 
-FunctionOverloadResolverImplPtr CastOverloadResolver::create(const Context & context)
-{
-    return createImpl(context.getSettingsRef().cast_keep_nullable);
-}
-
 void registerFunctionFixedString(FunctionFactory & factory);
 
 void registerFunctionsConversion(FunctionFactory & factory)
@@ -44,7 +38,10 @@ void registerFunctionsConversion(FunctionFactory & factory)
     registerFunctionFixedString(factory);
 
     factory.registerFunction<FunctionToUnixTimestamp>();
-    factory.registerFunction<CastOverloadResolver>(FunctionFactory::CaseInsensitive);
+
+    factory.registerFunction<CastOverloadResolver<CastType::nonAccurate>>(FunctionFactory::CaseInsensitive);
+    factory.registerFunction<CastOverloadResolver<CastType::accurate>>();
+    factory.registerFunction<CastOverloadResolver<CastType::accurateOrNull>>();
 
     factory.registerFunction<FunctionToUInt8OrZero>();
     factory.registerFunction<FunctionToUInt16OrZero>();
diff --git a/src/Functions/FunctionsConversion.h b/src/Functions/FunctionsConversion.h
index 1e4bd5c26d6..92607f718d5 100644
--- a/src/Functions/FunctionsConversion.h
+++ b/src/Functions/FunctionsConversion.h
@@ -20,6 +20,7 @@
 #include <DataTypes/DataTypeEnum.h>
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeNothing.h>
 #include <DataTypes/DataTypeUUID.h>
@@ -32,17 +33,20 @@
 #include <Columns/ColumnArray.h>
 #include <Columns/ColumnNullable.h>
 #include <Columns/ColumnTuple.h>
+#include <Columns/ColumnMap.h>
 #include <Columns/ColumnsCommon.h>
 #include <Common/FieldVisitors.h>
 #include <Common/assert_cast.h>
 #include <Common/quoteString.h>
+#include <Core/AccurateComparison.h>
 #include <Functions/IFunctionAdaptors.h>
 #include <Functions/FunctionsMiscellaneous.h>
 #include <Functions/FunctionHelpers.h>
 #include <Functions/DateTimeTransforms.h>
+#include <Functions/toFixedString.h>
 #include <DataTypes/DataTypeLowCardinality.h>
 #include <Columns/ColumnLowCardinality.h>
-#include <Functions/toFixedString.h>
+#include <Interpreters/Context.h>
 
 
 namespace DB
@@ -94,6 +98,15 @@ inline UInt32 extractToDecimalScale(const ColumnWithTypeAndName & named_column)
 /// Function toUnixTimestamp has exactly the same implementation as toDateTime of String type.
 struct NameToUnixTimestamp { static constexpr auto name = "toUnixTimestamp"; };
 
+struct AccurateConvertStrategyAdditions
+{
+    UInt32 scale { 0 };
+};
+
+struct AccurateOrNullConvertStrategyAdditions
+{
+    UInt32 scale { 0 };
+};
 
 /** Conversion of number types to each other, enums to numbers, dates and datetimes to numbers and back: done by straight assignment.
   *  (Date is represented internally as number of days from some day; DateTime - as unix timestamp)
@@ -106,7 +119,7 @@ struct ConvertImpl
 
     template <typename Additions = void *>
     static ColumnPtr NO_SANITIZE_UNDEFINED execute(
-        const ColumnsWithTypeAndName & arguments, const DataTypePtr & /*result_type*/, size_t /*input_rows_count*/,
+        const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type [[maybe_unused]], size_t /*input_rows_count*/,
         Additions additions [[maybe_unused]] = Additions())
     {
         const ColumnWithTypeAndName & named_from = arguments[0];
@@ -136,7 +149,17 @@ struct ConvertImpl
             typename ColVecTo::MutablePtr col_to = nullptr;
             if constexpr (IsDataTypeDecimal<ToDataType>)
             {
-                UInt32 scale = additions;
+                UInt32 scale;
+                if constexpr (std::is_same_v<Additions, AccurateConvertStrategyAdditions>
+                    || std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                {
+                    scale = additions.scale;
+                }
+                else
+                {
+                    scale = additions;
+                }
+
                 col_to = ColVecTo::create(0, scale);
             }
             else
@@ -147,36 +170,106 @@ struct ConvertImpl
             size_t size = vec_from.size();
             vec_to.resize(size);
 
-            for (size_t i = 0; i < size; ++i)
+            ColumnUInt8::MutablePtr col_null_map_to;
+            ColumnUInt8::Container * vec_null_map_to [[maybe_unused]] = nullptr;
+            if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
             {
-                if constexpr (IsDataTypeDecimal<FromDataType> || IsDataTypeDecimal<ToDataType>)
-                {
-                    if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>)
-                        vec_to[i] = convertDecimals<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), vec_to.getScale());
-                    else if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeNumber<ToDataType>)
-                        vec_to[i] = convertFromDecimal<FromDataType, ToDataType>(vec_from[i], vec_from.getScale());
-                    else if constexpr (IsDataTypeNumber<FromDataType> && IsDataTypeDecimal<ToDataType>)
-                        vec_to[i] = convertToDecimal<FromDataType, ToDataType>(vec_from[i], vec_to.getScale());
-                    else
-                        throw Exception("Unsupported data type in conversion function", ErrorCodes::CANNOT_CONVERT_TYPE);
-                }
-                else if constexpr (is_big_int_v<FromFieldType> || is_big_int_v<ToFieldType>)
-                {
-                    if constexpr (std::is_same_v<FromFieldType, UInt128> || std::is_same_v<ToFieldType, UInt128>)
-                        throw Exception("Unexpected UInt128 to big int conversion", ErrorCodes::NOT_IMPLEMENTED);
-                    /// If From Data is Nan or Inf, throw exception
-                    else if (!isFinite(vec_from[i]))
-                        throw Exception("Unexpected inf or nan to big int conversion", ErrorCodes::NOT_IMPLEMENTED);
-                    else
-                        vec_to[i] = bigint_cast<ToFieldType>(vec_from[i]);
-                }
-                else if constexpr (std::is_same_v<ToFieldType, UInt128> && sizeof(FromFieldType) <= sizeof(UInt64))
-                    vec_to[i] = static_cast<ToFieldType>(static_cast<UInt64>(vec_from[i]));
-                else
-                    vec_to[i] = static_cast<ToFieldType>(vec_from[i]);
+                col_null_map_to = ColumnUInt8::create(size, false);
+                vec_null_map_to = &col_null_map_to->getData();
             }
 
-            return col_to;
+            for (size_t i = 0; i < size; ++i)
+            {
+                if constexpr ((is_big_int_v<FromFieldType> || is_big_int_v<ToFieldType>) &&
+                    (std::is_same_v<FromFieldType, UInt128> || std::is_same_v<ToFieldType, UInt128>))
+                {
+                    if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                        (*vec_null_map_to)[i] = true;
+                    else
+                        throw Exception("Unexpected UInt128 to big int conversion", ErrorCodes::NOT_IMPLEMENTED);
+                }
+                else
+                {
+                    if constexpr (IsDataTypeDecimal<FromDataType> || IsDataTypeDecimal<ToDataType>)
+                    {
+                        if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                        {
+                            ToFieldType result;
+                            bool convert_result = false;
+
+                            if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                convert_result = tryConvertDecimals<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), vec_to.getScale(), result);
+                            else if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeNumber<ToDataType>)
+                                convert_result = tryConvertFromDecimal<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), result);
+                            else if constexpr (IsDataTypeNumber<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                convert_result = tryConvertToDecimal<FromDataType, ToDataType>(vec_from[i], vec_to.getScale(), result);
+
+                            if (convert_result)
+                                vec_to[i] = result;
+                            else
+                                (*vec_null_map_to)[i] = true;
+                        }
+                        else
+                        {
+                            if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                vec_to[i] = convertDecimals<FromDataType, ToDataType>(vec_from[i], vec_from.getScale(), vec_to.getScale());
+                            else if constexpr (IsDataTypeDecimal<FromDataType> && IsDataTypeNumber<ToDataType>)
+                                vec_to[i] = convertFromDecimal<FromDataType, ToDataType>(vec_from[i], vec_from.getScale());
+                            else if constexpr (IsDataTypeNumber<FromDataType> && IsDataTypeDecimal<ToDataType>)
+                                vec_to[i] = convertToDecimal<FromDataType, ToDataType>(vec_from[i], vec_to.getScale());
+                            else
+                                throw Exception("Unsupported data type in conversion function", ErrorCodes::CANNOT_CONVERT_TYPE);
+                        }
+                    }
+                    else
+                    {
+                        /// If From Data is Nan or Inf and we convert to integer type, throw exception
+                        if constexpr (std::is_floating_point_v<FromFieldType> && !std::is_floating_point_v<ToFieldType>)
+                        {
+                            if (!isFinite(vec_from[i]))
+                            {
+                                if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                                {
+                                    (*vec_null_map_to)[i] = true;
+                                    continue;
+                                }
+                                else
+                                    throw Exception("Unexpected inf or nan to integer conversion", ErrorCodes::CANNOT_CONVERT_TYPE);
+                            }
+                        }
+
+                        if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>
+                                || std::is_same_v<Additions, AccurateConvertStrategyAdditions>)
+                        {
+                            bool convert_result = accurate::convertNumeric(vec_from[i], vec_to[i]);
+
+                            if (!convert_result)
+                            {
+                                if (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                                {
+                                    (*vec_null_map_to)[i] = true;
+                                }
+                                else
+                                {
+                                    throw Exception(
+                                        "Value in column " + named_from.column->getName() + " cannot be safely converted into type "
+                                            + result_type->getName(),
+                                        ErrorCodes::CANNOT_CONVERT_TYPE);
+                                }
+                            }
+                        }
+                        else
+                        {
+                            vec_to[i] = static_cast<ToFieldType>(vec_from[i]);
+                        }
+                    }
+                }
+            }
+
+            if constexpr (std::is_same_v<Additions, AccurateOrNullConvertStrategyAdditions>)
+                return ColumnNullable::create(std::move(col_to), std::move(col_null_map_to));
+            else
+                return col_to;
         }
         else
             throw Exception("Illegal column " + named_from.column->getName() + " of first argument of function " + Name::name,
@@ -943,7 +1036,9 @@ struct ConvertImpl<DataTypeString, DataTypeUInt32, NameToUnixTimestamp>
 template <typename T, typename Name>
 struct ConvertImpl<std::enable_if_t<!T::is_parametric, T>, T, Name>
 {
-    static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/)
+    template <typename Additions = void *>
+    static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/,
+        Additions additions [[maybe_unused]] = Additions())
     {
         return arguments[0].column;
     }
@@ -1929,9 +2024,15 @@ private:
     std::optional<Diagnostic> diagnostic;
 };
 
-
 struct NameCast { static constexpr auto name = "CAST"; };
 
+enum class CastType
+{
+    nonAccurate,
+    accurate,
+    accurateOrNull
+};
+
 class FunctionCast final : public IFunctionBaseImpl
 {
 public:
@@ -1940,9 +2041,11 @@ public:
     using Diagnostic = ExecutableFunctionCast::Diagnostic;
 
     FunctionCast(const char * name_, MonotonicityForRange && monotonicity_for_range_
-        , const DataTypes & argument_types_, const DataTypePtr & return_type_, std::optional<Diagnostic> diagnostic_)
-        : name(name_), monotonicity_for_range(monotonicity_for_range_)
+        , const DataTypes & argument_types_, const DataTypePtr & return_type_
+        , std::optional<Diagnostic> diagnostic_, CastType cast_type_)
+        : name(name_), monotonicity_for_range(std::move(monotonicity_for_range_))
         , argument_types(argument_types_), return_type(return_type_), diagnostic(std::move(diagnostic_))
+        , cast_type(cast_type_)
     {
     }
 
@@ -1989,70 +2092,118 @@ private:
     DataTypePtr return_type;
 
     std::optional<Diagnostic> diagnostic;
+    CastType cast_type;
 
-    template <typename DataType>
-    WrapperType createWrapper(const DataTypePtr & from_type, const DataType * const, bool requested_result_is_nullable) const
+    WrapperType createFunctionAdaptor(FunctionPtr function, const DataTypePtr & from_type) const
     {
-        FunctionPtr function;
+        auto function_adaptor = FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
+                                    .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
+
+        return [function_adaptor]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        {
+            return function_adaptor->execute(arguments, result_type, input_rows_count);
+        };
+    }
+
+    static WrapperType createToNullableColumnWrapper()
+    {
+        return [] (ColumnsWithTypeAndName &, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        {
+            ColumnPtr res = result_type->createColumn();
+            ColumnUInt8::Ptr col_null_map_to = ColumnUInt8::create(input_rows_count, true);
+            return ColumnNullable::create(res->cloneResized(input_rows_count), std::move(col_null_map_to));
+        };
+    }
+
+    template <typename ToDataType>
+    WrapperType createWrapper(const DataTypePtr & from_type, const ToDataType * const to_type, bool requested_result_is_nullable) const
+    {
+        TypeIndex from_type_index = from_type->getTypeId();
+        WhichDataType which(from_type_index);
+        bool can_apply_accurate_cast = (cast_type == CastType::accurate || cast_type == CastType::accurateOrNull)
+            && (which.isInt() || which.isUInt() || which.isFloat());
 
         if (requested_result_is_nullable && checkAndGetDataType<DataTypeString>(from_type.get()))
         {
             /// In case when converting to Nullable type, we apply different parsing rule,
             /// that will not throw an exception but return NULL in case of malformed input.
-            function = FunctionConvertFromString<DataType, NameCast, ConvertFromStringExceptionMode::Null>::create();
+            FunctionPtr function = FunctionConvertFromString<ToDataType, NameCast, ConvertFromStringExceptionMode::Null>::create();
+            return createFunctionAdaptor(function, from_type);
         }
-        else
-            function = FunctionTo<DataType>::Type::create();
-
-        auto function_adaptor =
-                FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
-
-        return [function_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        else if (!can_apply_accurate_cast)
         {
-            return function_adaptor->execute(arguments, result_type, input_rows_count);
+            FunctionPtr function = FunctionTo<ToDataType>::Type::create();
+            return createFunctionAdaptor(function, from_type);
+        }
+
+        auto wrapper_cast_type = cast_type;
+
+        return [wrapper_cast_type, from_type_index, to_type]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *column_nullable, size_t input_rows_count)
+        {
+            ColumnPtr result_column;
+            auto res = callOnIndexAndDataType<ToDataType>(from_type_index, [&](const auto & types) -> bool {
+                using Types = std::decay_t<decltype(types)>;
+                using LeftDataType = typename Types::LeftType;
+                using RightDataType = typename Types::RightType;
+
+                if constexpr (IsDataTypeNumber<LeftDataType> && IsDataTypeNumber<RightDataType>)
+                {
+                    if (wrapper_cast_type == CastType::accurate)
+                    {
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, AccurateConvertStrategyAdditions());
+                    }
+                    else
+                    {
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, AccurateOrNullConvertStrategyAdditions());
+                    }
+
+                    return true;
+                }
+
+                return false;
+            });
+
+            /// Additionally check if callOnIndexAndDataType wasn't called at all.
+            if (!res)
+            {
+                if (wrapper_cast_type == CastType::accurateOrNull)
+                {
+                    auto nullable_column_wrapper = FunctionCast::createToNullableColumnWrapper();
+                    return nullable_column_wrapper(arguments, result_type, column_nullable, input_rows_count);
+                }
+                else
+                {
+                    throw Exception{"Conversion from " + std::string(getTypeName(from_type_index)) + " to " + to_type->getName() + " is not supported",
+                        ErrorCodes::CANNOT_CONVERT_TYPE};
+                }
+            }
+
+            return result_column;
         };
     }
 
-    static WrapperType createStringWrapper(const DataTypePtr & from_type)
+    WrapperType createStringWrapper(const DataTypePtr & from_type) const
     {
         FunctionPtr function = FunctionToString::create();
-
-        auto function_adaptor =
-                FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
-
-        return [function_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
-        {
-            return function_adaptor->execute(arguments, result_type, input_rows_count);
-        };
+        return createFunctionAdaptor(function, from_type);
     }
 
-    static WrapperType createFixedStringWrapper(const DataTypePtr & from_type, const size_t N)
+    WrapperType createFixedStringWrapper(const DataTypePtr & from_type, const size_t N) const
     {
         if (!isStringOrFixedString(from_type))
             throw Exception{"CAST AS FixedString is only implemented for types String and FixedString", ErrorCodes::NOT_IMPLEMENTED};
 
-        return [N] (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable *, size_t /*input_rows_count*/)
+        bool exception_mode_null = cast_type == CastType::accurateOrNull;
+        return [exception_mode_null, N] (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable *, size_t /*input_rows_count*/)
         {
-            return FunctionToFixedString::executeForN(arguments, N);
-        };
-    }
-
-    static WrapperType createUUIDWrapper(const DataTypePtr & from_type, const DataTypeUUID * const, bool requested_result_is_nullable)
-    {
-        if (requested_result_is_nullable)
-            throw Exception{"CAST AS Nullable(UUID) is not implemented", ErrorCodes::NOT_IMPLEMENTED};
-
-        FunctionPtr function = FunctionTo<DataTypeUUID>::Type::create();
-
-        auto function_adaptor =
-                FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                .build({ColumnWithTypeAndName{nullptr, from_type, ""}});
-
-        return [function_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
-        {
-            return function_adaptor->execute(arguments, result_type, input_rows_count);
+            if (exception_mode_null)
+                return FunctionToFixedString::executeForN<ConvertToFixedStringExceptionMode::Null>(arguments, N);
+            else
+                return FunctionToFixedString::executeForN<ConvertToFixedStringExceptionMode::Throw>(arguments, N);
         };
     }
 
@@ -2064,41 +2215,73 @@ private:
         UInt32 scale = to_type->getScale();
 
         WhichDataType which(type_index);
-        bool ok = which.isNativeInt() ||
-            which.isNativeUInt() ||
-            which.isDecimal() ||
-            which.isFloat() ||
-            which.isDateOrDateTime() ||
-            which.isStringOrFixedString();
+        bool ok = which.isNativeInt() || which.isNativeUInt() || which.isDecimal() || which.isFloat() || which.isDateOrDateTime()
+            || which.isStringOrFixedString();
         if (!ok)
-            throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
-                ErrorCodes::CANNOT_CONVERT_TYPE};
+        {
+            if (cast_type == CastType::accurateOrNull)
+                return createToNullableColumnWrapper();
+            else
+                throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
+                    ErrorCodes::CANNOT_CONVERT_TYPE};
+        }
 
-        return [type_index, scale, to_type] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
+        auto wrapper_cast_type = cast_type;
+
+        return [wrapper_cast_type, type_index, scale, to_type]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *column_nullable, size_t input_rows_count)
         {
             ColumnPtr result_column;
-            auto res = callOnIndexAndDataType<ToDataType>(type_index, [&](const auto & types) -> bool
-            {
+            auto res = callOnIndexAndDataType<ToDataType>(type_index, [&](const auto & types) -> bool {
                 using Types = std::decay_t<decltype(types)>;
                 using LeftDataType = typename Types::LeftType;
                 using RightDataType = typename Types::RightType;
 
+                if constexpr (IsDataTypeDecimalOrNumber<LeftDataType> && IsDataTypeDecimalOrNumber<RightDataType>)
+                {
+                    if (wrapper_cast_type == CastType::accurate)
+                    {
+                        AccurateConvertStrategyAdditions additions;
+                        additions.scale = scale;
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, additions);
+
+                        return true;
+                    }
+                    else if (wrapper_cast_type == CastType::accurateOrNull)
+                    {
+                        AccurateOrNullConvertStrategyAdditions additions;
+                        additions.scale = scale;
+                        result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(
+                            arguments, result_type, input_rows_count, additions);
+
+                        return true;
+                    }
+                }
+
                 result_column = ConvertImpl<LeftDataType, RightDataType, NameCast>::execute(arguments, result_type, input_rows_count, scale);
+
                 return true;
             });
 
             /// Additionally check if callOnIndexAndDataType wasn't called at all.
             if (!res)
             {
-                throw Exception{"Conversion from " + std::string(getTypeName(type_index)) + " to " + to_type->getName() +
-                                " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+                if (wrapper_cast_type == CastType::accurateOrNull)
+                {
+                    auto nullable_column_wrapper = FunctionCast::createToNullableColumnWrapper();
+                    return nullable_column_wrapper(arguments, result_type, column_nullable, input_rows_count);
+                }
+                else
+                    throw Exception{"Conversion from " + std::string(getTypeName(type_index)) + " to " + to_type->getName() + " is not supported",
+                        ErrorCodes::CANNOT_CONVERT_TYPE};
             }
 
             return result_column;
         };
     }
 
-    static WrapperType createAggregateFunctionWrapper(const DataTypePtr & from_type_untyped, const DataTypeAggregateFunction * to_type)
+    WrapperType createAggregateFunctionWrapper(const DataTypePtr & from_type_untyped, const DataTypeAggregateFunction * to_type) const
     {
         /// Conversion from String through parsing.
         if (checkAndGetDataType<DataTypeString>(from_type_untyped.get()))
@@ -2109,8 +2292,13 @@ private:
             };
         }
         else
-            throw Exception{"Conversion from " + from_type_untyped->getName() + " to " + to_type->getName() +
-                " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+        {
+            if (cast_type == CastType::accurateOrNull)
+                return createToNullableColumnWrapper();
+            else
+                throw Exception{"Conversion from " + from_type_untyped->getName() + " to " + to_type->getName() +
+                    " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+        }
     }
 
     WrapperType createArrayWrapper(const DataTypePtr & from_type_untyped, const DataTypeArray * to_type) const
@@ -2167,6 +2355,20 @@ private:
         };
     }
 
+    using ElementWrappers = std::vector<WrapperType>;
+
+    ElementWrappers getElementWrappers(const DataTypes & from_element_types, const DataTypes & to_element_types) const
+    {
+        ElementWrappers element_wrappers;
+        element_wrappers.reserve(from_element_types.size());
+
+        /// Create conversion wrapper for each element in tuple
+        for (const auto idx_type : ext::enumerate(from_element_types))
+            element_wrappers.push_back(prepareUnpackDictionaries(idx_type.second, to_element_types[idx_type.first]));
+
+        return element_wrappers;
+    }
+
     WrapperType createTupleWrapper(const DataTypePtr & from_type_untyped, const DataTypeTuple * to_type) const
     {
         /// Conversion from String through parsing.
@@ -2189,12 +2391,7 @@ private:
 
         const auto & from_element_types = from_type->getElements();
         const auto & to_element_types = to_type->getElements();
-        std::vector<WrapperType> element_wrappers;
-        element_wrappers.reserve(from_element_types.size());
-
-        /// Create conversion wrapper for each element in tuple
-        for (const auto idx_type : ext::enumerate(from_type->getElements()))
-            element_wrappers.push_back(prepareUnpackDictionaries(idx_type.second, to_element_types[idx_type.first]));
+        auto element_wrappers = getElementWrappers(from_element_types, to_element_types);
 
         return [element_wrappers, from_element_types, to_element_types]
             (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
@@ -2217,6 +2414,121 @@ private:
         };
     }
 
+    /// The case of: tuple([key1, key2, ..., key_n], [value1, value2, ..., value_n])
+    WrapperType createTupleToMapWrapper(const DataTypes & from_kv_types, const DataTypes & to_kv_types) const
+    {
+        return [element_wrappers = getElementWrappers(from_kv_types, to_kv_types), from_kv_types, to_kv_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
+        {
+            const auto * col = arguments.front().column.get();
+            const auto & column_tuple = assert_cast<const ColumnTuple &>(*col);
+
+            if (column_tuple.getColumn(0).size() != column_tuple.getColumn(1).size())
+                throw Exception(ErrorCodes::TYPE_MISMATCH,
+                    "CAST AS Map can only be performed from tuple of arrays with equal sizes."
+                    " Size of keys: {}. Size of values: {}", column_tuple.getColumn(0).size(), column_tuple.getColumn(1).size());
+
+            ColumnPtr offsets;
+            Columns converted_columns(2);
+            for (size_t i = 0; i < 2; ++i)
+            {
+                const auto & column_array = assert_cast<const ColumnArray &>(column_tuple.getColumn(i));
+                ColumnsWithTypeAndName element = {{column_array.getDataPtr(), from_kv_types[i], ""}};
+                converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
+
+                if (!offsets)
+                    offsets = column_array.getOffsetsPtr();
+            }
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], offsets);
+        };
+    }
+
+    WrapperType createMapToMapWrrapper(const DataTypes & from_kv_types, const DataTypes & to_kv_types) const
+    {
+        return [element_wrappers = getElementWrappers(from_kv_types, to_kv_types), from_kv_types, to_kv_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
+        {
+            const auto * col = arguments.front().column.get();
+            const auto & column_map = typeid_cast<const ColumnMap &>(*col);
+            const auto & nested_data = column_map.getNestedData();
+
+            Columns converted_columns(2);
+            for (size_t i = 0; i < 2; ++i)
+            {
+                ColumnsWithTypeAndName element = {{nested_data.getColumnPtr(i), from_kv_types[i], ""}};
+                converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
+            }
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], column_map.getNestedColumn().getOffsetsPtr());
+        };
+    }
+
+    /// The case of: [(key1, value1), (key2, value2), ...]
+    WrapperType createArrayToMapWrrapper(const DataTypes & from_kv_types, const DataTypes & to_kv_types) const
+    {
+        return [element_wrappers = getElementWrappers(from_kv_types, to_kv_types), from_kv_types, to_kv_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count) -> ColumnPtr
+        {
+            const auto * col = arguments.front().column.get();
+            const auto & column_array = typeid_cast<const ColumnArray &>(*col);
+            const auto & nested_data = typeid_cast<const ColumnTuple &>(column_array.getData());
+
+            Columns converted_columns(2);
+            for (size_t i = 0; i < 2; ++i)
+            {
+                ColumnsWithTypeAndName element = {{nested_data.getColumnPtr(i), from_kv_types[i], ""}};
+                converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
+            }
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], column_array.getOffsetsPtr());
+        };
+    }
+
+
+    WrapperType createMapWrapper(const DataTypePtr & from_type_untyped, const DataTypeMap * to_type) const
+    {
+        if (const auto * from_tuple = checkAndGetDataType<DataTypeTuple>(from_type_untyped.get()))
+        {
+            if (from_tuple->getElements().size() != 2)
+                throw Exception{"CAST AS Map from tuple requeires 2 elements.\n"
+                    "Left type: " + from_tuple->getName() + ", right type: " + to_type->getName(), ErrorCodes::TYPE_MISMATCH};
+
+            DataTypes from_kv_types;
+            const auto & to_kv_types = to_type->getKeyValueTypes();
+
+            for (const auto & elem : from_tuple->getElements())
+            {
+                const auto * type_array = checkAndGetDataType<DataTypeArray>(elem.get());
+                if (!type_array)
+                    throw Exception(ErrorCodes::TYPE_MISMATCH,
+                        "CAST AS Map can only be performed from tuples of array. Got: {}", from_tuple->getName());
+
+                from_kv_types.push_back(type_array->getNestedType());
+            }
+
+            return createTupleToMapWrapper(from_kv_types, to_kv_types);
+        }
+        else if (const auto * from_array = typeid_cast<const DataTypeArray *>(from_type_untyped.get()))
+        {
+            const auto * nested_tuple = typeid_cast<const DataTypeTuple *>(from_array->getNestedType().get());
+            if (!nested_tuple || nested_tuple->getElements().size() != 2)
+                throw Exception{"CAST AS Map from array requeires nested tuple of 2 elements.\n"
+                    "Left type: " + from_array->getName() + ", right type: " + to_type->getName(), ErrorCodes::TYPE_MISMATCH};
+
+            return createArrayToMapWrrapper(nested_tuple->getElements(), to_type->getKeyValueTypes());
+        }
+        else if (const auto * from_type = checkAndGetDataType<DataTypeMap>(from_type_untyped.get()))
+        {
+            return createMapToMapWrrapper(from_type->getKeyValueTypes(), to_type->getKeyValueTypes());
+        }
+        else
+        {
+            throw Exception{"Unsupported types to CAST AS Map\n"
+                "Left type: " + from_type_untyped->getName() + ", right type: " + to_type->getName(), ErrorCodes::TYPE_MISMATCH};
+        }
+    }
+
     template <typename FieldType>
     WrapperType createEnumWrapper(const DataTypePtr & from_type, const DataTypeEnum<FieldType> * to_type) const
     {
@@ -2235,17 +2547,16 @@ private:
         else if (isNativeNumber(from_type) || isEnum(from_type))
         {
             auto function = Function::create();
-            auto func_or_adaptor = FunctionOverloadResolverAdaptor(std::make_unique<DefaultOverloadResolver>(function))
-                    .build(ColumnsWithTypeAndName{{nullptr, from_type, "" }});
-
-            return [func_or_adaptor] (ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
-            {
-                return func_or_adaptor->execute(arguments, result_type, input_rows_count);
-            };
+            return createFunctionAdaptor(function, from_type);
         }
         else
-            throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() +
-                " is not supported", ErrorCodes::CANNOT_CONVERT_TYPE};
+        {
+            if (cast_type == CastType::accurateOrNull)
+                return createToNullableColumnWrapper();
+            else
+                throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
+                    ErrorCodes::CANNOT_CONVERT_TYPE};
+        }
     }
 
     template <typename EnumTypeFrom, typename EnumTypeTo>
@@ -2346,7 +2657,16 @@ private:
         if (from_type->onlyNull())
         {
             if (!to_nested->isNullable())
-                throw Exception{"Cannot convert NULL to a non-nullable type", ErrorCodes::CANNOT_CONVERT_TYPE};
+            {
+                if (cast_type == CastType::accurateOrNull)
+                {
+                    return createToNullableColumnWrapper();
+                }
+                else
+                {
+                    throw Exception{"Cannot convert NULL to a non-nullable type", ErrorCodes::CANNOT_CONVERT_TYPE};
+                }
+            }
 
             return [](ColumnsWithTypeAndName &, const DataTypePtr & result_type, const ColumnNullable *, size_t input_rows_count)
             {
@@ -2536,7 +2856,8 @@ private:
                 std::is_same_v<ToDataType, DataTypeFloat32> ||
                 std::is_same_v<ToDataType, DataTypeFloat64> ||
                 std::is_same_v<ToDataType, DataTypeDate> ||
-                std::is_same_v<ToDataType, DataTypeDateTime>)
+                std::is_same_v<ToDataType, DataTypeDateTime> ||
+                std::is_same_v<ToDataType, DataTypeUUID>)
             {
                 ret = createWrapper(from_type, checkAndGetDataType<ToDataType>(to_type.get()), requested_result_is_nullable);
                 return true;
@@ -2558,14 +2879,6 @@ private:
                 ret = createDecimalWrapper(from_type, checkAndGetDataType<ToDataType>(to_type.get()));
                 return true;
             }
-            if constexpr (std::is_same_v<ToDataType, DataTypeUUID>)
-            {
-                if (isStringOrFixedString(from_type))
-                {
-                    ret = createUUIDWrapper(from_type, checkAndGetDataType<ToDataType>(to_type.get()), requested_result_is_nullable);
-                    return true;
-                }
-            }
 
             return false;
         };
@@ -2584,6 +2897,8 @@ private:
                 return createArrayWrapper(from_type, checkAndGetDataType<DataTypeArray>(to_type.get()));
             case TypeIndex::Tuple:
                 return createTupleWrapper(from_type, checkAndGetDataType<DataTypeTuple>(to_type.get()));
+            case TypeIndex::Map:
+                return createMapWrapper(from_type, checkAndGetDataType<DataTypeMap>(to_type.get()));
 
             case TypeIndex::AggregateFunction:
                 return createAggregateFunctionWrapper(from_type, checkAndGetDataType<DataTypeAggregateFunction>(to_type.get()));
@@ -2591,20 +2906,91 @@ private:
                 break;
         }
 
-        throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
-            ErrorCodes::CANNOT_CONVERT_TYPE};
+        if (cast_type == CastType::accurateOrNull)
+            return createToNullableColumnWrapper();
+        else
+            throw Exception{"Conversion from " + from_type->getName() + " to " + to_type->getName() + " is not supported",
+                ErrorCodes::CANNOT_CONVERT_TYPE};
     }
 };
 
+class MonotonicityHelper
+{
+public:
+    using MonotonicityForRange = FunctionCast::MonotonicityForRange;
+
+    template <typename DataType>
+    static auto monotonicityForType(const DataType * const)
+    {
+        return FunctionTo<DataType>::Type::Monotonic::get;
+    }
+
+    static MonotonicityForRange getMonotonicityInformation(const DataTypePtr & from_type, const IDataType * to_type)
+    {
+        if (const auto type = checkAndGetDataType<DataTypeUInt8>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt16>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt32>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt64>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeUInt256>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt8>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt16>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt32>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt64>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt128>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeInt256>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeFloat32>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeFloat64>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeDate>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeDateTime>(to_type))
+            return monotonicityForType(type);
+        if (const auto type = checkAndGetDataType<DataTypeString>(to_type))
+            return monotonicityForType(type);
+        if (isEnum(from_type))
+        {
+            if (const auto type = checkAndGetDataType<DataTypeEnum8>(to_type))
+                return monotonicityForType(type);
+            if (const auto type = checkAndGetDataType<DataTypeEnum16>(to_type))
+                return monotonicityForType(type);
+        }
+        /// other types like Null, FixedString, Array and Tuple have no monotonicity defined
+        return {};
+    }
+};
+
+template<CastType cast_type>
 class CastOverloadResolver : public IFunctionOverloadResolverImpl
 {
 public:
     using MonotonicityForRange = FunctionCast::MonotonicityForRange;
     using Diagnostic = FunctionCast::Diagnostic;
 
-    static constexpr auto name = "CAST";
+    static constexpr auto accurate_cast_name = "accurateCast";
+    static constexpr auto accurate_cast_or_null_name = "accurateCastOrNull";
+    static constexpr auto cast_name = "CAST";
+
+    static constexpr auto name = cast_type == CastType::accurate
+        ? accurate_cast_name
+        : (cast_type == CastType::accurateOrNull ? accurate_cast_or_null_name : cast_name);
+
+    static FunctionOverloadResolverImplPtr create(const Context & context)
+    {
+        return createImpl(context.getSettingsRef().cast_keep_nullable);
+    }
 
-    static FunctionOverloadResolverImplPtr create(const Context & context);
     static FunctionOverloadResolverImplPtr createImpl(bool keep_nullable, std::optional<Diagnostic> diagnostic = {})
     {
         return std::make_unique<CastOverloadResolver>(keep_nullable, std::move(diagnostic));
@@ -2630,8 +3016,8 @@ protected:
         for (size_t i = 0; i < arguments.size(); ++i)
             data_types[i] = arguments[i].type;
 
-        auto monotonicity = getMonotonicityInformation(arguments.front().type, return_type.get());
-        return std::make_unique<FunctionCast>(name, std::move(monotonicity), data_types, return_type, diagnostic);
+        auto monotonicity = MonotonicityHelper::getMonotonicityInformation(arguments.front().type, return_type.get());
+        return std::make_unique<FunctionCast>(name, std::move(monotonicity), data_types, return_type, diagnostic, cast_type);
     }
 
     DataTypePtr getReturnType(const ColumnsWithTypeAndName & arguments) const override
@@ -2649,9 +3035,17 @@ protected:
                 ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
         DataTypePtr type = DataTypeFactory::instance().get(type_col->getValue<String>());
-        if (keep_nullable && arguments.front().type->isNullable())
+
+        if constexpr (cast_type == CastType::accurateOrNull)
+        {
             return makeNullable(type);
-        return type;
+        }
+        else
+        {
+            if (keep_nullable && arguments.front().type->isNullable())
+                return makeNullable(type);
+            return type;
+        }
     }
 
     bool useDefaultImplementationForNulls() const override { return false; }
@@ -2660,57 +3054,6 @@ protected:
 private:
     bool keep_nullable;
     std::optional<Diagnostic> diagnostic;
-
-    template <typename DataType>
-    static auto monotonicityForType(const DataType * const)
-    {
-        return FunctionTo<DataType>::Type::Monotonic::get;
-    }
-
-    static MonotonicityForRange getMonotonicityInformation(const DataTypePtr & from_type, const IDataType * to_type)
-    {
-        if (const auto * type = checkAndGetDataType<DataTypeUInt8>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt16>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt32>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt64>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeUInt256>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt8>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt16>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt32>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt64>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt128>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeInt256>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeFloat32>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeFloat64>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeDate>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeDateTime>(to_type))
-            return monotonicityForType(type);
-        if (const auto * type = checkAndGetDataType<DataTypeString>(to_type))
-            return monotonicityForType(type);
-        if (isEnum(from_type))
-        {
-            if (const auto * type = checkAndGetDataType<DataTypeEnum8>(to_type))
-                return monotonicityForType(type);
-            if (const auto * type = checkAndGetDataType<DataTypeEnum16>(to_type))
-                return monotonicityForType(type);
-        }
-        /// other types like Null, FixedString, Array and Tuple have no monotonicity defined
-        return {};
-    }
 };
 
 }
diff --git a/src/Functions/array/arrayAggregation.cpp b/src/Functions/array/arrayAggregation.cpp
new file mode 100644
index 00000000000..992a331d05b
--- /dev/null
+++ b/src/Functions/array/arrayAggregation.cpp
@@ -0,0 +1,318 @@
+#include <DataTypes/DataTypesNumber.h>
+#include <DataTypes/DataTypesDecimal.h>
+#include <DataTypes/DataTypeDateTime64.h>
+#include <Columns/ColumnsNumber.h>
+#include <Columns/ColumnDecimal.h>
+#include "FunctionArrayMapped.h"
+#include <Functions/FunctionFactory.h>
+
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+    extern const int ILLEGAL_COLUMN;
+}
+
+enum class AggregateOperation
+{
+    min,
+    max,
+    sum,
+    average
+};
+
+/**
+ * During array aggregation we derive result type from operation.
+ * For array min or array max we use array element as result type.
+ * For array average we use Float64.
+ * For array sum for for big integers, we use same type representation, decimal numbers we use Decimal128,
+ * for floating point numbers Float64, for numeric unsigned Int64, and for numeric signed UInt64.
+ */
+
+template <typename ArrayElement, AggregateOperation operation>
+struct ArrayAggregateResultImpl;
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::min>
+{
+    using Result = ArrayElement;
+};
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::max>
+{
+    using Result = ArrayElement;
+};
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::average>
+{
+    using Result = Float64;
+};
+
+template <typename ArrayElement>
+struct ArrayAggregateResultImpl<ArrayElement, AggregateOperation::sum>
+{
+    using Result =
+        std::conditional_t<std::is_same_v<ArrayElement, Int128>, Int128,
+            std::conditional_t<std::is_same_v<ArrayElement, Int256>, Int256,
+                std::conditional_t<std::is_same_v<ArrayElement, UInt256>, UInt256,
+                    std::conditional_t<IsDecimalNumber<ArrayElement>, Decimal128,
+                        std::conditional_t<std::is_floating_point_v<ArrayElement>, Float64,
+                            std::conditional_t<std::is_signed_v<ArrayElement>, Int64,
+                                UInt64>>>>>>;
+};
+
+template <typename ArrayElement, AggregateOperation operation>
+using ArrayAggregateResult = typename ArrayAggregateResultImpl<ArrayElement, operation>::Result;
+
+template<AggregateOperation aggregate_operation>
+struct ArrayAggregateImpl
+{
+    static bool needBoolean() { return false; }
+    static bool needExpression() { return false; }
+    static bool needOneArray() { return false; }
+
+    static DataTypePtr getReturnType(const DataTypePtr & expression_return, const DataTypePtr & /*array_element*/)
+    {
+        DataTypePtr result;
+
+        auto call = [&](const auto & types)
+        {
+            using Types = std::decay_t<decltype(types)>;
+            using DataType = typename Types::LeftType;
+
+            if constexpr (aggregate_operation == AggregateOperation::average)
+            {
+                result = std::make_shared<DataTypeFloat64>();
+
+                return true;
+            }
+            else if constexpr (IsDataTypeNumber<DataType>)
+            {
+                using NumberReturnType = ArrayAggregateResult<typename DataType::FieldType, aggregate_operation>;
+                result = std::make_shared<DataTypeNumber<NumberReturnType>>();
+
+                return true;
+            }
+            else if constexpr (IsDataTypeDecimal<DataType> && !IsDataTypeDateOrDateTime<DataType>)
+            {
+                using DecimalReturnType = ArrayAggregateResult<typename DataType::FieldType, aggregate_operation>;
+                UInt32 scale = getDecimalScale(*expression_return);
+                result = std::make_shared<DataTypeDecimal<DecimalReturnType>>(DecimalUtils::maxPrecision<DecimalReturnType>(), scale);
+
+                return true;
+            }
+
+            return false;
+        };
+
+        if (!callOnIndexAndDataType<void>(expression_return->getTypeId(), call))
+        {
+            throw Exception(
+                "array aggregation function cannot be performed on type " + expression_return->getName(),
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        }
+
+        return result;
+    }
+
+    template <typename Element>
+    static bool executeType(const ColumnPtr & mapped, const ColumnArray::Offsets & offsets, ColumnPtr & res_ptr)
+    {
+        using Result = ArrayAggregateResult<Element, aggregate_operation>;
+        using ColVecType = std::conditional_t<IsDecimalNumber<Element>, ColumnDecimal<Element>, ColumnVector<Element>>;
+        using ColVecResult = std::conditional_t<IsDecimalNumber<Result>, ColumnDecimal<Result>, ColumnVector<Result>>;
+
+        /// For average of array we return Float64 as result, but we want to keep precision
+        /// so we convert to Float64 as last step, but intermediate sum is represented as result of sum operation
+        static constexpr bool is_average_operation = aggregate_operation == AggregateOperation::average;
+        using SummAggregationType = ArrayAggregateResult<Element, AggregateOperation::sum>;
+
+        using AggregationType = std::conditional_t<is_average_operation, SummAggregationType, Result>;
+
+
+        const ColVecType * column = checkAndGetColumn<ColVecType>(&*mapped);
+
+        /// Constant case.
+        if (!column)
+        {
+            const ColumnConst * column_const = checkAndGetColumnConst<ColVecType>(&*mapped);
+
+            if (!column_const)
+                return false;
+
+            const AggregationType x = column_const->template getValue<Element>(); // NOLINT
+            const typename ColVecType::Container & data
+                = checkAndGetColumn<ColVecType>(&column_const->getDataColumn())->getData();
+
+            typename ColVecResult::MutablePtr res_column;
+            if constexpr (IsDecimalNumber<Element>)
+            {
+                res_column = ColVecResult::create(offsets.size(), data.getScale());
+            }
+            else
+                res_column = ColVecResult::create(offsets.size());
+
+            typename ColVecResult::Container & res = res_column->getData();
+
+            size_t pos = 0;
+            for (size_t i = 0; i < offsets.size(); ++i)
+            {
+                if constexpr (aggregate_operation == AggregateOperation::sum)
+                {
+                    size_t array_size = offsets[i] - pos;
+                    /// Just multiply the value by array size.
+                    res[i] = x * array_size;
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::min ||
+                                aggregate_operation == AggregateOperation::max)
+                {
+                    res[i] = x;
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::average)
+                {
+                    if constexpr (IsDecimalNumber<Element>)
+                    {
+                        res[i] = DecimalUtils::convertTo<Result>(x, data.getScale());
+                    }
+                    else
+                    {
+                        res[i] = x;
+                    }
+                }
+
+                pos = offsets[i];
+            }
+
+            res_ptr = std::move(res_column);
+            return true;
+        }
+
+        const typename ColVecType::Container & data = column->getData();
+
+        typename ColVecResult::MutablePtr res_column;
+        if constexpr (IsDecimalNumber<Element>)
+            res_column = ColVecResult::create(offsets.size(), data.getScale());
+        else
+            res_column = ColVecResult::create(offsets.size());
+
+        typename ColVecResult::Container & res = res_column->getData();
+
+        size_t pos = 0;
+        for (size_t i = 0; i < offsets.size(); ++i)
+        {
+            AggregationType s = 0;
+
+            /// Array is empty
+            if (offsets[i] == pos)
+            {
+                res[i] = s;
+                continue;
+            }
+
+            size_t count = 1;
+            s = data[pos]; // NOLINT
+            ++pos;
+
+            for (; pos < offsets[i]; ++pos)
+            {
+                auto element = data[pos];
+
+                if constexpr (aggregate_operation == AggregateOperation::sum ||
+                            aggregate_operation == AggregateOperation::average)
+                {
+                    s += element;
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::min)
+                {
+                    if (element < s)
+                    {
+                        s = element;
+                    }
+                }
+                else if constexpr (aggregate_operation == AggregateOperation::max)
+                {
+                    if (element > s)
+                    {
+                        s = element;
+                    }
+                }
+
+                ++count;
+            }
+
+            if constexpr (aggregate_operation == AggregateOperation::average)
+            {
+                if constexpr (IsDecimalNumber<Element>)
+                {
+                    s = s / count;
+                    res[i] = DecimalUtils::convertTo<Result>(s, data.getScale());
+                }
+                else
+                {
+                    res[i] = static_cast<Result>(s) / count;
+                }
+            }
+            else
+            {
+                res[i] = s;
+            }
+        }
+
+        res_ptr = std::move(res_column);
+        return true;
+    }
+
+    static ColumnPtr execute(const ColumnArray & array, ColumnPtr mapped)
+    {
+        const IColumn::Offsets & offsets = array.getOffsets();
+        ColumnPtr res;
+
+        if (executeType<UInt8>(mapped, offsets, res) ||
+            executeType<UInt16>(mapped, offsets, res) ||
+            executeType<UInt32>(mapped, offsets, res) ||
+            executeType<UInt64>(mapped, offsets, res) ||
+            executeType<UInt256>(mapped, offsets, res) ||
+            executeType<Int8>(mapped, offsets, res) ||
+            executeType<Int16>(mapped, offsets, res) ||
+            executeType<Int32>(mapped, offsets, res) ||
+            executeType<Int64>(mapped, offsets, res) ||
+            executeType<Int128>(mapped, offsets, res) ||
+            executeType<Int256>(mapped, offsets, res) ||
+            executeType<Float32>(mapped, offsets, res) ||
+            executeType<Float64>(mapped, offsets, res) ||
+            executeType<Decimal32>(mapped, offsets, res) ||
+            executeType<Decimal64>(mapped, offsets, res) ||
+            executeType<Decimal128>(mapped, offsets, res))
+            return res;
+        else
+            throw Exception("Unexpected column for arraySum: " + mapped->getName(), ErrorCodes::ILLEGAL_COLUMN);
+    }
+};
+
+struct NameArrayMin { static constexpr auto name = "arrayMin"; };
+using FunctionArrayMin = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::min>, NameArrayMin>;
+
+struct NameArrayMax { static constexpr auto name = "arrayMax"; };
+using FunctionArrayMax = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::max>, NameArrayMax>;
+
+struct NameArraySum { static constexpr auto name = "arraySum"; };
+using FunctionArraySum = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::sum>, NameArraySum>;
+
+struct NameArrayAverage { static constexpr auto name = "arrayAvg"; };
+using FunctionArrayAverage = FunctionArrayMapped<ArrayAggregateImpl<AggregateOperation::average>, NameArrayAverage>;
+
+void registerFunctionArrayAggregation(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionArrayMin>();
+    factory.registerFunction<FunctionArrayMax>();
+    factory.registerFunction<FunctionArraySum>();
+    factory.registerFunction<FunctionArrayAverage>();
+}
+
+}
+
diff --git a/src/Functions/array/arrayElement.cpp b/src/Functions/array/arrayElement.cpp
index 46a384db449..310b6b5b450 100644
--- a/src/Functions/array/arrayElement.cpp
+++ b/src/Functions/array/arrayElement.cpp
@@ -4,12 +4,16 @@
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypesNumber.h>
 #include <Core/ColumnNumbers.h>
 #include <Columns/ColumnArray.h>
+#include <Core/Field.h>
 #include <Columns/ColumnNullable.h>
 #include <Columns/ColumnsNumber.h>
 #include <Columns/ColumnString.h>
 #include <Columns/ColumnTuple.h>
+#include <Columns/ColumnMap.h>
 #include <Common/typeid_cast.h>
 #include <Common/assert_cast.h>
 
@@ -80,6 +84,38 @@ private:
     /** For a tuple array, the function is evaluated component-wise for each element of the tuple.
       */
     ColumnPtr executeTuple(const ColumnsWithTypeAndName & arguments, size_t input_rows_count) const;
+
+    /** For a map the function finds the matched value for a key.
+     *  Currently implemented just as linear seach in array.
+     *  However, optimizations are possible.
+     */
+    ColumnPtr executeMap(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const;
+
+    using Offsets = ColumnArray::Offsets;
+
+    static bool matchKeyToIndex(const IColumn & data, const Offsets & offsets,
+        const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs);
+
+    static bool matchKeyToIndexConst(const IColumn & data, const Offsets & offsets,
+        const Field & index, PaddedPODArray<UInt64> & matched_idxs);
+
+    template <typename DataType>
+    static bool matchKeyToIndexNumber(const IColumn & data, const Offsets & offsets,
+        const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs);
+
+    template <typename DataType>
+    static bool matchKeyToIndexNumberConst(const IColumn & data, const Offsets & offsets,
+        const Field & index, PaddedPODArray<UInt64> & matched_idxs);
+
+    static bool matchKeyToIndexString(const IColumn & data, const Offsets & offsets,
+        const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs);
+
+    static bool matchKeyToIndexStringConst(const IColumn & data, const Offsets & offsets,
+        const Field & index, PaddedPODArray<UInt64> & matched_idxs);
+
+    template <typename Matcher>
+    static void executeMatchKeyToIndex(const Offsets & offsets,
+        PaddedPODArray<UInt64> & matched_idxs, const Matcher & matcher);
 };
 
 
@@ -626,10 +662,8 @@ ColumnPtr FunctionArrayElement::executeArgument(
     const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, ArrayImpl::NullMapBuilder & builder, size_t input_rows_count) const
 {
     auto index = checkAndGetColumn<ColumnVector<IndexType>>(arguments[1].column.get());
-
     if (!index)
         return nullptr;
-
     const auto & index_data = index->getData();
 
     if (builder)
@@ -704,6 +738,236 @@ ColumnPtr FunctionArrayElement::executeTuple(const ColumnsWithTypeAndName & argu
     return ColumnTuple::create(result_tuple_columns);
 }
 
+namespace
+{
+
+struct MatcherString
+{
+    const ColumnString & data;
+    const ColumnString & index;
+
+    bool match(size_t row_data, size_t row_index) const
+    {
+        auto data_ref = data.getDataAt(row_data);
+        auto index_ref = index.getDataAt(row_index);
+        return memequalSmallAllowOverflow15(index_ref.data, index_ref.size, data_ref.data, data_ref.size);
+    }
+};
+
+struct MatcherStringConst
+{
+    const ColumnString & data;
+    const String & index;
+
+    bool match(size_t row_data, size_t /* row_index */) const
+    {
+        auto data_ref = data.getDataAt(row_data);
+        return index.size() == data_ref.size && memcmp(index.data(), data_ref.data, data_ref.size) == 0;
+    }
+};
+
+template <typename T>
+struct MatcherNumber
+{
+    const PaddedPODArray<T> & data;
+    const PaddedPODArray<T> & index;
+
+    bool match(size_t row_data, size_t row_index) const
+    {
+        return data[row_data] == index[row_index];
+    }
+};
+
+template <typename T>
+struct MatcherNumberConst
+{
+    const PaddedPODArray<T> & data;
+    T index;
+
+    bool match(size_t row_data, size_t /* row_index */) const
+    {
+        return data[row_data] == index;
+    }
+};
+
+}
+
+template <typename Matcher>
+void FunctionArrayElement::executeMatchKeyToIndex(
+    const Offsets & offsets, PaddedPODArray<UInt64> & matched_idxs, const Matcher & matcher)
+{
+    size_t rows = offsets.size();
+    for (size_t i = 0; i < rows; ++i)
+    {
+        bool matched = false;
+        size_t begin = offsets[i - 1];
+        size_t end = offsets[i];
+        for (size_t j = begin; j < end; ++j)
+        {
+            if (matcher.match(j, i))
+            {
+                matched_idxs.push_back(j - begin + 1);
+                matched = true;
+                break;
+            }
+        }
+
+        if (!matched)
+            matched_idxs.push_back(0);
+    }
+}
+
+bool FunctionArrayElement::matchKeyToIndexStringConst(
+    const IColumn & data, const Offsets & offsets,
+    const Field & index, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * data_string = checkAndGetColumn<ColumnString>(&data);
+    if (!data_string)
+        return false;
+
+    if (index.getType() != Field::Types::String)
+        return false;
+
+    MatcherStringConst matcher{*data_string, get<const String &>(index)};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+bool FunctionArrayElement::matchKeyToIndexString(
+    const IColumn & data, const Offsets & offsets,
+    const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * index_string = checkAndGetColumn<ColumnString>(arguments[1].column.get());
+    if (!index_string)
+        return false;
+
+    const auto * data_string = checkAndGetColumn<ColumnString>(&data);
+    if (!data_string)
+        return false;
+
+    MatcherString matcher{*data_string, *index_string};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+template <typename DataType>
+bool FunctionArrayElement::matchKeyToIndexNumberConst(
+    const IColumn & data, const Offsets & offsets,
+    const Field & index, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * data_numeric = checkAndGetColumn<ColumnVector<DataType>>(&data);
+    if (!data_numeric)
+        return false;
+
+    if (index.getType() != Field::Types::UInt64 && index.getType() != Field::Types::Int64)
+        return false;
+
+    MatcherNumberConst<DataType> matcher{data_numeric->getData(), get<DataType>(index)};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+template <typename DataType>
+bool FunctionArrayElement::matchKeyToIndexNumber(
+    const IColumn & data, const Offsets & offsets,
+    const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs)
+{
+    const auto * index_numeric = checkAndGetColumn<ColumnVector<DataType>>(arguments[1].column.get());
+    if (!index_numeric)
+        return false;
+
+    const auto * data_numeric = checkAndGetColumn<ColumnVector<DataType>>(&data);
+    if (!data_numeric)
+        return false;
+
+    MatcherNumber<DataType> matcher{data_numeric->getData(), index_numeric->getData()};
+    executeMatchKeyToIndex(offsets, matched_idxs, matcher);
+    return true;
+}
+
+bool FunctionArrayElement::matchKeyToIndex(
+    const IColumn & data, const Offsets & offsets,
+    const ColumnsWithTypeAndName & arguments, PaddedPODArray<UInt64> & matched_idxs)
+{
+    return matchKeyToIndexNumber<UInt8>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<UInt16>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<UInt32>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<UInt64>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int8>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int16>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int32>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexNumber<Int64>(data, offsets, arguments, matched_idxs)
+        || matchKeyToIndexString(data, offsets, arguments, matched_idxs);
+}
+
+bool FunctionArrayElement::matchKeyToIndexConst(
+    const IColumn & data, const Offsets & offsets,
+    const Field & index, PaddedPODArray<UInt64> & matched_idxs)
+{
+    return matchKeyToIndexNumberConst<UInt8>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<UInt16>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<UInt32>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<UInt64>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int8>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int16>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int32>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexNumberConst<Int64>(data, offsets, index, matched_idxs)
+        || matchKeyToIndexStringConst(data, offsets, index, matched_idxs);
+}
+
+ColumnPtr FunctionArrayElement::executeMap(
+    const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const
+{
+    const ColumnMap * col_map = typeid_cast<const ColumnMap *>(arguments[0].column.get());
+    if (!col_map)
+        return nullptr;
+
+    const auto & nested_column = col_map->getNestedColumn();
+    const auto & keys_data = col_map->getNestedData().getColumn(0);
+    const auto & values_data = col_map->getNestedData().getColumn(1);
+    const auto & offsets = nested_column.getOffsets();
+
+    /// At first step calculate indices in array of values for requested keys.
+    auto indices_column = DataTypeNumber<UInt64>().createColumn();
+    indices_column->reserve(input_rows_count);
+    auto & indices_data = assert_cast<ColumnVector<UInt64> &>(*indices_column).getData();
+
+    if (!isColumnConst(*arguments[1].column))
+    {
+        if (input_rows_count > 0 && !matchKeyToIndex(keys_data, offsets, arguments, indices_data))
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                "Illegal types of arguments: {}, {} for function ",
+                arguments[0].type->getName(), arguments[1].type->getName(), getName());
+    }
+    else
+    {
+        Field index = (*arguments[1].column)[0];
+
+        // Get Matched key's value
+        if (input_rows_count > 0 && !matchKeyToIndexConst(keys_data, offsets, index, indices_data))
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                "Illegal types of arguments: {}, {} for function {}",
+                arguments[0].type->getName(), arguments[1].type->getName(), getName());
+    }
+
+    /// Prepare arguments to call arrayElement for array with values and calculated indices at previous step.
+    ColumnsWithTypeAndName new_arguments =
+    {
+        {
+            ColumnArray::create(values_data.getPtr(), nested_column.getOffsetsPtr()),
+            std::make_shared<DataTypeArray>(result_type),
+            ""
+        },
+        {
+            std::move(indices_column),
+            std::make_shared<DataTypeNumber<UInt64>>(),
+            ""
+        }
+    };
+
+    return executeImpl(new_arguments, result_type, input_rows_count);
+}
+
 String FunctionArrayElement::getName() const
 {
     return name;
@@ -711,7 +975,10 @@ String FunctionArrayElement::getName() const
 
 DataTypePtr FunctionArrayElement::getReturnTypeImpl(const DataTypes & arguments) const
 {
-    const DataTypeArray * array_type = checkAndGetDataType<DataTypeArray>(arguments[0].get());
+    if (const auto * map_type = checkAndGetDataType<DataTypeMap>(arguments[0].get()))
+        return map_type->getValueType();
+
+    const auto * array_type = checkAndGetDataType<DataTypeArray>(arguments[0].get());
     if (!array_type)
     {
         throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
@@ -734,6 +1001,10 @@ ColumnPtr FunctionArrayElement::executeImpl(const ColumnsWithTypeAndName & argum
     /// Check nullability.
     bool is_array_of_nullable = false;
 
+    const ColumnMap * col_map = checkAndGetColumn<ColumnMap>(arguments[0].column.get());
+    if (col_map)
+        return executeMap(arguments, result_type, input_rows_count);
+
     const ColumnArray * col_array = nullptr;
     const ColumnArray * col_const_array = nullptr;
 
@@ -826,7 +1097,7 @@ ColumnPtr FunctionArrayElement::perform(const ColumnsWithTypeAndName & arguments
             || (res = executeArgument<Int16>(arguments, result_type, builder, input_rows_count))
             || (res = executeArgument<Int32>(arguments, result_type, builder, input_rows_count))
             || (res = executeArgument<Int64>(arguments, result_type, builder, input_rows_count))))
-        throw Exception("Second argument for function " + getName() + " must must have UInt or Int type.",
+        throw Exception("Second argument for function " + getName() + " must have UInt or Int type.",
                         ErrorCodes::ILLEGAL_COLUMN);
     }
     else
diff --git a/src/Functions/array/arrayEnumerateRanked.h b/src/Functions/array/arrayEnumerateRanked.h
index a0d865f08fe..ce82b8bcd86 100644
--- a/src/Functions/array/arrayEnumerateRanked.h
+++ b/src/Functions/array/arrayEnumerateRanked.h
@@ -304,10 +304,10 @@ void FunctionArrayEnumerateRankedExtended<Derived>::executeMethodImpl(
     const size_t depth_to_look = arrays_depths.max_array_depth;
     const auto & offsets = *offsets_by_depth[depth_to_look - 1];
 
-    using Map = ClearableHashMapWithStackMemory<UInt128, UInt32,
+    using Container = ClearableHashMapWithStackMemory<UInt128, UInt32,
         UInt128TrivialHash, INITIAL_SIZE_DEGREE>;
 
-    Map indices;
+    Container indices;
 
     std::vector<size_t> indices_by_depth(depth_to_look);
     std::vector<size_t> current_offset_n_by_depth(depth_to_look);
diff --git a/src/Functions/array/arrayIntersect.cpp b/src/Functions/array/arrayIntersect.cpp
index 1eb3e7ac6c3..736069d6060 100644
--- a/src/Functions/array/arrayIntersect.cpp
+++ b/src/Functions/array/arrayIntersect.cpp
@@ -438,22 +438,22 @@ ColumnPtr FunctionArrayIntersect::executeImpl(const ColumnsWithTypeAndName & arg
 template <typename T, size_t>
 void FunctionArrayIntersect::NumberExecutor::operator()()
 {
-    using Map = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
+    using Container = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
         INITIAL_SIZE_DEGREE>;
 
     if (!result && typeid_cast<const DataTypeNumber<T> *>(data_type.get()))
-        result = execute<Map, ColumnVector<T>, true>(arrays, ColumnVector<T>::create());
+        result = execute<Container, ColumnVector<T>, true>(arrays, ColumnVector<T>::create());
 }
 
 template <typename T, size_t>
 void FunctionArrayIntersect::DecimalExecutor::operator()()
 {
-    using Map = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
+    using Container = ClearableHashMapWithStackMemory<T, size_t, DefaultHash<T>,
         INITIAL_SIZE_DEGREE>;
 
     if (!result)
         if (auto * decimal = typeid_cast<const DataTypeDecimal<T> *>(data_type.get()))
-            result = execute<Map, ColumnDecimal<T>, true>(arrays, ColumnDecimal<T>::create(0, decimal->getScale()));
+            result = execute<Container, ColumnDecimal<T>, true>(arrays, ColumnDecimal<T>::create(0, decimal->getScale()));
 }
 
 template <typename Map, typename ColumnType, bool is_numeric_column>
diff --git a/src/Functions/array/arraySum.cpp b/src/Functions/array/arraySum.cpp
deleted file mode 100644
index 1aedcb6ef92..00000000000
--- a/src/Functions/array/arraySum.cpp
+++ /dev/null
@@ -1,146 +0,0 @@
-#include <DataTypes/DataTypesNumber.h>
-#include <DataTypes/DataTypesDecimal.h>
-#include <Columns/ColumnsNumber.h>
-#include <Columns/ColumnDecimal.h>
-#include "FunctionArrayMapped.h"
-#include <Functions/FunctionFactory.h>
-
-
-namespace DB
-{
-
-namespace ErrorCodes
-{
-    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
-    extern const int ILLEGAL_COLUMN;
-}
-
-struct ArraySumImpl
-{
-    static bool needBoolean() { return false; }
-    static bool needExpression() { return false; }
-    static bool needOneArray() { return false; }
-
-    static DataTypePtr getReturnType(const DataTypePtr & expression_return, const DataTypePtr & /*array_element*/)
-    {
-        WhichDataType which(expression_return);
-
-        if (which.isNativeUInt())
-            return std::make_shared<DataTypeUInt64>();
-
-        if (which.isNativeInt())
-            return std::make_shared<DataTypeInt64>();
-
-        if (which.isFloat())
-            return std::make_shared<DataTypeFloat64>();
-
-        if (which.isDecimal())
-        {
-            UInt32 scale = getDecimalScale(*expression_return);
-            return std::make_shared<DataTypeDecimal<Decimal128>>(DecimalUtils::maxPrecision<Decimal128>(), scale);
-        }
-
-        throw Exception("arraySum cannot add values of type " + expression_return->getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-    }
-
-    template <typename Element, typename Result>
-    static bool executeType(const ColumnPtr & mapped, const ColumnArray::Offsets & offsets, ColumnPtr & res_ptr)
-    {
-        using ColVecType = std::conditional_t<IsDecimalNumber<Element>, ColumnDecimal<Element>, ColumnVector<Element>>;
-        using ColVecResult = std::conditional_t<IsDecimalNumber<Result>, ColumnDecimal<Result>, ColumnVector<Result>>;
-
-        const ColVecType * column = checkAndGetColumn<ColVecType>(&*mapped);
-
-        /// Constant case.
-        if (!column)
-        {
-            const ColumnConst * column_const = checkAndGetColumnConst<ColVecType>(&*mapped);
-
-            if (!column_const)
-                return false;
-
-            const Result x = column_const->template getValue<Element>(); // NOLINT
-
-            typename ColVecResult::MutablePtr res_column;
-            if constexpr (IsDecimalNumber<Element>)
-            {
-                const typename ColVecType::Container & data =
-                    checkAndGetColumn<ColVecType>(&column_const->getDataColumn())->getData();
-                res_column = ColVecResult::create(offsets.size(), data.getScale());
-            }
-            else
-                res_column = ColVecResult::create(offsets.size());
-
-            typename ColVecResult::Container & res = res_column->getData();
-
-            size_t pos = 0;
-            for (size_t i = 0; i < offsets.size(); ++i)
-            {
-                /// Just multiply the value by array size.
-                res[i] = x * (offsets[i] - pos);
-                pos = offsets[i];
-            }
-
-            res_ptr = std::move(res_column);
-            return true;
-        }
-
-        const typename ColVecType::Container & data = column->getData();
-
-        typename ColVecResult::MutablePtr res_column;
-        if constexpr (IsDecimalNumber<Element>)
-            res_column = ColVecResult::create(offsets.size(), data.getScale());
-        else
-            res_column = ColVecResult::create(offsets.size());
-
-        typename ColVecResult::Container & res = res_column->getData();
-
-        size_t pos = 0;
-        for (size_t i = 0; i < offsets.size(); ++i)
-        {
-            Result s = 0;
-            for (; pos < offsets[i]; ++pos)
-            {
-                s += data[pos];
-            }
-            res[i] = s;
-        }
-
-        res_ptr = std::move(res_column);
-        return true;
-    }
-
-    static ColumnPtr execute(const ColumnArray & array, ColumnPtr mapped)
-    {
-        const IColumn::Offsets & offsets = array.getOffsets();
-        ColumnPtr res;
-
-        if (executeType< UInt8 , UInt64>(mapped, offsets, res) ||
-            executeType< UInt16, UInt64>(mapped, offsets, res) ||
-            executeType< UInt32, UInt64>(mapped, offsets, res) ||
-            executeType< UInt64, UInt64>(mapped, offsets, res) ||
-            executeType<  Int8 ,  Int64>(mapped, offsets, res) ||
-            executeType<  Int16,  Int64>(mapped, offsets, res) ||
-            executeType<  Int32,  Int64>(mapped, offsets, res) ||
-            executeType<  Int64,  Int64>(mapped, offsets, res) ||
-            executeType<Float32,Float64>(mapped, offsets, res) ||
-            executeType<Float64,Float64>(mapped, offsets, res) ||
-            executeType<Decimal32, Decimal128>(mapped, offsets, res) ||
-            executeType<Decimal64, Decimal128>(mapped, offsets, res) ||
-            executeType<Decimal128, Decimal128>(mapped, offsets, res))
-            return res;
-        else
-            throw Exception("Unexpected column for arraySum: " + mapped->getName(), ErrorCodes::ILLEGAL_COLUMN);
-    }
-};
-
-struct NameArraySum { static constexpr auto name = "arraySum"; };
-using FunctionArraySum = FunctionArrayMapped<ArraySumImpl, NameArraySum>;
-
-void registerFunctionArraySum(FunctionFactory & factory)
-{
-    factory.registerFunction<FunctionArraySum>();
-}
-
-}
-
diff --git a/src/Functions/encodeXML.cpp b/src/Functions/encodeXMLComponent.cpp
similarity index 93%
rename from src/Functions/encodeXML.cpp
rename to src/Functions/encodeXMLComponent.cpp
index 75c5d35668f..d839a3a9264 100644
--- a/src/Functions/encodeXML.cpp
+++ b/src/Functions/encodeXMLComponent.cpp
@@ -3,6 +3,7 @@
 #include <Functions/FunctionStringToString.h>
 #include <common/find_symbols.h>
 
+
 namespace DB
 {
 namespace ErrorCodes
@@ -12,12 +13,12 @@ namespace ErrorCodes
 
 namespace
 {
-    struct EncodeXMLName
+    struct EncodeXMLComponentName
     {
         static constexpr auto name = "encodeXMLComponent";
     };
 
-    class FunctionEncodeXMLImpl
+    class FunctionEncodeXMLComponentImpl
     {
     public:
         static void vector(
@@ -132,12 +133,12 @@ namespace
         }
     };
 
-    using FunctionEncodeXML = FunctionStringToString<FunctionEncodeXMLImpl, EncodeXMLName>;
+    using FunctionEncodeXMLComponent = FunctionStringToString<FunctionEncodeXMLComponentImpl, EncodeXMLComponentName>;
 
 }
 
-void registerFunctionEncodeXML(FunctionFactory & factory)
+void registerFunctionEncodeXMLComponent(FunctionFactory & factory)
 {
-    factory.registerFunction<FunctionEncodeXML>();
+    factory.registerFunction<FunctionEncodeXMLComponent>();
 }
 }
diff --git a/src/Functions/map.cpp b/src/Functions/map.cpp
new file mode 100644
index 00000000000..5993ab3706e
--- /dev/null
+++ b/src/Functions/map.cpp
@@ -0,0 +1,140 @@
+#include <Functions/IFunctionImpl.h>
+#include <Functions/FunctionFactory.h>
+#include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Columns/ColumnMap.h>
+#include <Columns/ColumnArray.h>
+#include <DataTypes/getLeastSupertype.h>
+#include <Interpreters/castColumn.h>
+#include <memory>
+
+
+namespace DB
+{
+namespace ErrorCodes
+{
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+}
+
+namespace
+{
+
+// map(x, y, ...) is a function that allows you to make key-value pair
+class FunctionMap : public IFunction
+{
+public:
+    static constexpr auto name = "map";
+
+    static FunctionPtr create(const Context &)
+    {
+        return std::make_shared<FunctionMap>();
+    }
+
+    String getName() const override
+    {
+        return name;
+    }
+
+    bool isVariadic() const override
+    {
+        return true;
+    }
+
+    size_t getNumberOfArguments() const override
+    {
+        return 0;
+    }
+
+    bool isInjective(const ColumnsWithTypeAndName &) const override
+    {
+        return true;
+    }
+
+    bool useDefaultImplementationForNulls() const override { return false; }
+    bool useDefaultImplementationForConstants() const override { return true; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (arguments.size() % 2 != 0)
+            throw Exception("Function " + getName() + " even number of arguments.", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+
+        DataTypes keys, values;
+        for (size_t i = 0; i < arguments.size(); i += 2)
+        {
+            keys.emplace_back(arguments[i]);
+            values.emplace_back(arguments[i + 1]);
+        }
+
+        DataTypes tmp;
+        tmp.emplace_back(getLeastSupertype(keys));
+        tmp.emplace_back(getLeastSupertype(values));
+        return std::make_shared<DataTypeMap>(tmp);
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
+    {
+        size_t num_elements = arguments.size();
+
+        if (num_elements == 0)
+            return result_type->createColumnConstWithDefaultValue(input_rows_count);
+
+        const auto & result_type_map = static_cast<const DataTypeMap &>(*result_type);
+        const DataTypePtr & key_type = result_type_map.getKeyType();
+        const DataTypePtr & value_type = result_type_map.getValueType();
+
+        Columns columns_holder(num_elements);
+        ColumnRawPtrs column_ptrs(num_elements);
+
+        for (size_t i = 0; i < num_elements; ++i)
+        {
+            const auto & arg = arguments[i];
+            const auto to_type = i % 2 == 0 ? key_type : value_type;
+
+            ColumnPtr preprocessed_column = castColumn(arg, to_type);
+            preprocessed_column = preprocessed_column->convertToFullColumnIfConst();
+
+            columns_holder[i] = std::move(preprocessed_column);
+            column_ptrs[i] = columns_holder[i].get();
+        }
+
+        /// Create and fill the result map.
+
+        MutableColumnPtr keys_data = key_type->createColumn();
+        MutableColumnPtr values_data = value_type->createColumn();
+        MutableColumnPtr offsets = DataTypeNumber<IColumn::Offset>().createColumn();
+
+        size_t total_elements = input_rows_count * num_elements / 2;
+        keys_data->reserve(total_elements);
+        values_data->reserve(total_elements);
+        offsets->reserve(input_rows_count);
+
+        IColumn::Offset current_offset = 0;
+        for (size_t i = 0; i < input_rows_count; ++i)
+        {
+            for (size_t j = 0; j < num_elements; j += 2)
+            {
+                keys_data->insertFrom(*column_ptrs[j], i);
+                values_data->insertFrom(*column_ptrs[j + 1], i);
+            }
+
+            current_offset += num_elements / 2;
+            offsets->insert(current_offset);
+        }
+
+        auto nested_column = ColumnArray::create(
+            ColumnTuple::create(Columns{std::move(keys_data), std::move(values_data)}),
+            std::move(offsets));
+
+        return ColumnMap::create(nested_column);
+    }
+};
+
+}
+
+void registerFunctionsMap(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionMap>();
+}
+
+}
diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp
index 7552af7dbf1..54695a1087d 100644
--- a/src/Functions/multiply.cpp
+++ b/src/Functions/multiply.cpp
@@ -1,3 +1,4 @@
+#include <type_traits>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionBinaryArithmetic.h>
 #include <common/arithmeticOverflow.h>
@@ -25,11 +26,17 @@ struct MultiplyImpl
             return static_cast<Result>(a) * b;
     }
 
-    /// Apply operation and check overflow. It's used for Deciamal operations. @returns true if overflowed, false otherwise.
+    /// Apply operation and check overflow. It's used for Decimal operations. @returns true if overflowed, false otherwise.
     template <typename Result = ResultType>
     static inline bool apply(A a, B b, Result & c)
     {
-        return common::mulOverflow(static_cast<Result>(a), b, c);
+        if constexpr (std::is_same_v<Result, float> || std::is_same_v<Result, double>)
+        {
+            c = static_cast<Result>(a) * b;
+            return false;
+        }
+        else
+            return common::mulOverflow(static_cast<Result>(a), b, c);
     }
 
 #if USE_EMBEDDED_COMPILER
diff --git a/src/Functions/registerFunctions.cpp b/src/Functions/registerFunctions.cpp
index 6dfebc13665..3f75746f861 100644
--- a/src/Functions/registerFunctions.cpp
+++ b/src/Functions/registerFunctions.cpp
@@ -10,6 +10,7 @@ namespace DB
 void registerFunctionsArithmetic(FunctionFactory &);
 void registerFunctionsArray(FunctionFactory &);
 void registerFunctionsTuple(FunctionFactory &);
+void registerFunctionsMap(FunctionFactory &);
 void registerFunctionsBitmap(FunctionFactory &);
 void registerFunctionsCoding(FunctionFactory &);
 void registerFunctionsComparison(FunctionFactory &);
@@ -64,6 +65,7 @@ void registerFunctions()
     registerFunctionsArithmetic(factory);
     registerFunctionsArray(factory);
     registerFunctionsTuple(factory);
+    registerFunctionsMap(factory);
 #if !defined(ARCADIA_BUILD)
     registerFunctionsBitmap(factory);
 #endif
diff --git a/src/Functions/registerFunctionsHigherOrder.cpp b/src/Functions/registerFunctionsHigherOrder.cpp
index 08938ef6534..d3621a03ecd 100644
--- a/src/Functions/registerFunctionsHigherOrder.cpp
+++ b/src/Functions/registerFunctionsHigherOrder.cpp
@@ -9,7 +9,7 @@ void registerFunctionArrayCount(FunctionFactory & factory);
 void registerFunctionArrayExists(FunctionFactory & factory);
 void registerFunctionArrayAll(FunctionFactory & factory);
 void registerFunctionArrayCompact(FunctionFactory & factory);
-void registerFunctionArraySum(FunctionFactory & factory);
+void registerFunctionArrayAggregation(FunctionFactory & factory);
 void registerFunctionArrayFirst(FunctionFactory & factory);
 void registerFunctionArrayFirstIndex(FunctionFactory & factory);
 void registerFunctionsArrayFill(FunctionFactory & factory);
@@ -27,7 +27,7 @@ void registerFunctionsHigherOrder(FunctionFactory & factory)
     registerFunctionArrayExists(factory);
     registerFunctionArrayAll(factory);
     registerFunctionArrayCompact(factory);
-    registerFunctionArraySum(factory);
+    registerFunctionArrayAggregation(factory);
     registerFunctionArrayFirst(factory);
     registerFunctionArrayFirstIndex(factory);
     registerFunctionsArrayFill(factory);
diff --git a/src/Functions/registerFunctionsString.cpp b/src/Functions/registerFunctionsString.cpp
index d24a9e1aa9d..426cc8f8d56 100644
--- a/src/Functions/registerFunctionsString.cpp
+++ b/src/Functions/registerFunctionsString.cpp
@@ -33,7 +33,7 @@ void registerFunctionRegexpQuoteMeta(FunctionFactory &);
 void registerFunctionNormalizeQuery(FunctionFactory &);
 void registerFunctionNormalizedQueryHash(FunctionFactory &);
 void registerFunctionCountMatches(FunctionFactory &);
-void registerFunctionEncodeXML(FunctionFactory & factory);
+void registerFunctionEncodeXMLComponent(FunctionFactory & factory);
 
 #if USE_BASE64
 void registerFunctionBase64Encode(FunctionFactory &);
@@ -69,7 +69,7 @@ void registerFunctionsString(FunctionFactory & factory)
     registerFunctionNormalizeQuery(factory);
     registerFunctionNormalizedQueryHash(factory);
     registerFunctionCountMatches(factory);
-    registerFunctionEncodeXML(factory);
+    registerFunctionEncodeXMLComponent(factory);
 #if USE_BASE64
     registerFunctionBase64Encode(factory);
     registerFunctionBase64Decode(factory);
diff --git a/src/Functions/tcpPort.cpp b/src/Functions/tcpPort.cpp
index 52acf0ade54..26991c900ab 100644
--- a/src/Functions/tcpPort.cpp
+++ b/src/Functions/tcpPort.cpp
@@ -1,5 +1,6 @@
 #include <DataTypes/DataTypesNumber.h>
 #include <Functions/FunctionFactory.h>
+#include <Interpreters/Context.h>
 
 
 namespace DB
diff --git a/src/Functions/toFixedString.h b/src/Functions/toFixedString.h
index de1ff7e8084..23c9e92faaf 100644
--- a/src/Functions/toFixedString.h
+++ b/src/Functions/toFixedString.h
@@ -5,6 +5,8 @@
 #include <DataTypes/DataTypesNumber.h>
 #include <Columns/ColumnString.h>
 #include <Columns/ColumnFixedString.h>
+#include <Columns/ColumnsNumber.h>
+#include <Columns/ColumnNullable.h>
 #include <IO/WriteHelpers.h>
 
 
@@ -18,6 +20,11 @@ namespace ErrorCodes
     extern const int NOT_IMPLEMENTED;
 }
 
+enum class ConvertToFixedStringExceptionMode
+{
+    Throw,
+    Null
+};
 
 /** Conversion to fixed string is implemented only for strings.
   */
@@ -55,13 +62,22 @@ public:
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
     {
         const auto n = arguments[1].column->getUInt(0);
-        return executeForN(arguments, n);
+        return executeForN<ConvertToFixedStringExceptionMode::Throw>(arguments, n);
     }
 
+    template<ConvertToFixedStringExceptionMode exception_mode>
     static ColumnPtr executeForN(const ColumnsWithTypeAndName & arguments, const size_t n)
     {
         const auto & column = arguments[0].column;
 
+        ColumnUInt8::MutablePtr col_null_map_to;
+        ColumnUInt8::Container * vec_null_map_to [[maybe_unused]] = nullptr;
+        if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Null)
+        {
+            col_null_map_to = ColumnUInt8::create(column->size(), false);
+            vec_null_map_to = &col_null_map_to->getData();
+        }
+
         if (const auto * column_string = checkAndGetColumn<ColumnString>(column.get()))
         {
             auto column_fixed = ColumnFixedString::create(n);
@@ -77,18 +93,42 @@ public:
                 const size_t off = i ? in_offsets[i - 1] : 0;
                 const size_t len = in_offsets[i] - off - 1;
                 if (len > n)
-                    throw Exception("String too long for type FixedString(" + toString(n) + ")",
-                        ErrorCodes::TOO_LARGE_STRING_SIZE);
+                {
+                    if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Throw)
+                    {
+                        throw Exception("String too long for type FixedString(" + toString(n) + ")",
+                            ErrorCodes::TOO_LARGE_STRING_SIZE);
+                    }
+                    else
+                    {
+                        (*vec_null_map_to)[i] = true;
+                        continue;
+                    }
+                }
                 memcpy(&out_chars[i * n], &in_chars[off], len);
             }
 
-            return column_fixed;
+            if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Null)
+                return ColumnNullable::create(std::move(column_fixed), std::move(col_null_map_to));
+            else
+                return column_fixed;
         }
         else if (const auto * column_fixed_string = checkAndGetColumn<ColumnFixedString>(column.get()))
         {
             const auto src_n = column_fixed_string->getN();
             if (src_n > n)
-                throw Exception{"String too long for type FixedString(" + toString(n) + ")", ErrorCodes::TOO_LARGE_STRING_SIZE};
+            {
+                if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Throw)
+                {
+                    throw Exception{"String too long for type FixedString(" + toString(n) + ")", ErrorCodes::TOO_LARGE_STRING_SIZE};
+                }
+                else
+                {
+                    auto column_fixed = ColumnFixedString::create(n);
+                    std::fill(vec_null_map_to->begin(), vec_null_map_to->end(), true);
+                    return ColumnNullable::create(column_fixed->cloneResized(column->size()), std::move(col_null_map_to));
+                }
+            }
 
             auto column_fixed = ColumnFixedString::create(n);
 
@@ -103,7 +143,16 @@ public:
             return column_fixed;
         }
         else
-            throw Exception("Unexpected column: " + column->getName(), ErrorCodes::ILLEGAL_COLUMN);
+        {
+            if constexpr (exception_mode == ConvertToFixedStringExceptionMode::Throw)
+                throw Exception("Unexpected column: " + column->getName(), ErrorCodes::ILLEGAL_COLUMN);
+            else
+            {
+                auto column_fixed = ColumnFixedString::create(n);
+                std::fill(vec_null_map_to->begin(), vec_null_map_to->end(), true);
+                return ColumnNullable::create(column_fixed->cloneResized(column->size()), std::move(col_null_map_to));
+            }
+        }
     }
 };
 
diff --git a/src/Functions/toModifiedJulianDay.cpp b/src/Functions/toModifiedJulianDay.cpp
index df3066a5005..32e9eb23987 100644
--- a/src/Functions/toModifiedJulianDay.cpp
+++ b/src/Functions/toModifiedJulianDay.cpp
@@ -62,11 +62,11 @@ namespace DB
             typename ColVecTo::Container & vec_to = col_to->getData();
 
             ColumnUInt8::MutablePtr col_null_map_to;
-            ColumnUInt8::Container * vec_null_map_to [[maybe_unused]] = nullptr;
+            UInt8 * vec_null_map_to [[maybe_unused]] = nullptr;
             if constexpr (nullOnErrors)
             {
                 col_null_map_to = ColumnUInt8::create(input_rows_count);
-                vec_null_map_to = &col_null_map_to->getData();
+                vec_null_map_to = col_null_map_to->getData().data();
             }
 
             size_t current_offset = 0;
@@ -83,12 +83,15 @@ namespace DB
                     {
                         const GregorianDate<> date(read_buffer);
                         vec_to[i] = date.toModifiedJulianDay<typename ToDataType::FieldType>();
-                        (*vec_null_map_to)[i] = false;
+                        vec_null_map_to[i] = false;
                     }
                     catch (const Exception & e)
                     {
                         if (e.code() == ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED || e.code() == ErrorCodes::CANNOT_PARSE_DATE)
-                            (*vec_null_map_to)[i] = true;
+                        {
+                            vec_to[i] = static_cast<Int32>(0);
+                            vec_null_map_to[i] = true;
+                        }
                         else
                             throw;
                     }
diff --git a/src/Functions/trap.cpp b/src/Functions/trap.cpp
index 34553f4b266..38cf53948e6 100644
--- a/src/Functions/trap.cpp
+++ b/src/Functions/trap.cpp
@@ -61,9 +61,10 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
-    [[clang::optnone]] void executeImpl(Block & columns, const ColumnNumbers & arguments, size_t result, size_t input_rows_count) const override
+    [[clang::optnone]]
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & block, const DataTypePtr & result_type, size_t input_rows_count) const override
     {
-        if (const ColumnConst * column = checkAndGetColumnConst<ColumnString>(columns[arguments[0]].column.get()))
+        if (const ColumnConst * column = checkAndGetColumnConst<ColumnString>(block[0].column.get()))
         {
             String mode = column->getValue<String>();
 
@@ -135,6 +136,10 @@ public:
             {
                 (void)context.getCurrentQueryId();
             }
+            else if (mode == "stack overflow")
+            {
+                executeImpl(block, result_type, input_rows_count);
+            }
             else if (mode == "mmap many")
             {
                 std::vector<void *> maps;
@@ -160,7 +165,7 @@ public:
         else
             throw Exception("The only argument for function " + getName() + " must be constant String", ErrorCodes::ILLEGAL_COLUMN);
 
-        columns[result].column = columns[result].type->createColumnConst(input_rows_count, 0ULL);
+        return result_type->createColumnConst(input_rows_count, 0ULL);
     }
 };
 
diff --git a/src/Functions/ya.make b/src/Functions/ya.make
index 201b41593f6..6e5d832db77 100644
--- a/src/Functions/ya.make
+++ b/src/Functions/ya.make
@@ -120,6 +120,7 @@ SRCS(
     appendTrailingCharIfAbsent.cpp
     array/array.cpp
     array/arrayAUC.cpp
+    array/arrayAggregation.cpp
     array/arrayAll.cpp
     array/arrayCompact.cpp
     array/arrayConcat.cpp
@@ -155,7 +156,6 @@ SRCS(
     array/arraySlice.cpp
     array/arraySort.cpp
     array/arraySplit.cpp
-    array/arraySum.cpp
     array/arrayUniq.cpp
     array/arrayWithConstant.cpp
     array/arrayZip.cpp
@@ -226,7 +226,7 @@ SRCS(
     dumpColumnStructure.cpp
     e.cpp
     empty.cpp
-    encodeXML.cpp
+    encodeXMLComponent.cpp
     encrypt.cpp
     endsWith.cpp
     equals.cpp
@@ -325,6 +325,7 @@ SRCS(
     lowCardinalityKeys.cpp
     lower.cpp
     lowerUTF8.cpp
+    map.cpp
     match.cpp
     materialize.cpp
     minus.cpp
diff --git a/src/IO/ConnectionTimeouts.h b/src/IO/ConnectionTimeouts.h
index 9e87dee4fc3..e5efabee6e2 100644
--- a/src/IO/ConnectionTimeouts.h
+++ b/src/IO/ConnectionTimeouts.h
@@ -1,14 +1,13 @@
 #pragma once
 
 #include <Poco/Timespan.h>
-#include <Core/Settings.h>
-
-#include <Interpreters/Context.h>
-#include <Poco/Util/AbstractConfiguration.h>
 
 namespace DB
 {
 
+class Context;
+struct Settings;
+
 struct ConnectionTimeouts
 {
     Poco::Timespan connection_timeout;
@@ -92,24 +91,10 @@ struct ConnectionTimeouts
     }
 
     /// Timeouts for the case when we have just single attempt to connect.
-    static ConnectionTimeouts getTCPTimeoutsWithoutFailover(const Settings & settings)
-    {
-        return ConnectionTimeouts(settings.connect_timeout, settings.send_timeout, settings.receive_timeout, settings.tcp_keep_alive_timeout);
-    }
-
+    static ConnectionTimeouts getTCPTimeoutsWithoutFailover(const Settings & settings);
     /// Timeouts for the case when we will try many addresses in a loop.
-    static ConnectionTimeouts getTCPTimeoutsWithFailover(const Settings & settings)
-    {
-        return ConnectionTimeouts(settings.connect_timeout_with_failover_ms, settings.send_timeout, settings.receive_timeout, settings.tcp_keep_alive_timeout, 0, settings.connect_timeout_with_failover_secure_ms);
-    }
-
-    static ConnectionTimeouts getHTTPTimeouts(const Context & context)
-    {
-        const auto & settings = context.getSettingsRef();
-        const auto & config = context.getConfigRef();
-        Poco::Timespan http_keep_alive_timeout{config.getUInt("keep_alive_timeout", 10), 0};
-        return ConnectionTimeouts(settings.http_connection_timeout, settings.http_send_timeout, settings.http_receive_timeout, settings.tcp_keep_alive_timeout, http_keep_alive_timeout);
-    }
+    static ConnectionTimeouts getTCPTimeoutsWithFailover(const Settings & settings);
+    static ConnectionTimeouts getHTTPTimeouts(const Context & context);
 };
 
 }
diff --git a/src/IO/ConnectionTimeoutsContext.h b/src/IO/ConnectionTimeoutsContext.h
new file mode 100644
index 00000000000..ce19738f507
--- /dev/null
+++ b/src/IO/ConnectionTimeoutsContext.h
@@ -0,0 +1,30 @@
+#pragma once
+
+#include <IO/ConnectionTimeouts.h>
+#include <Poco/Util/AbstractConfiguration.h>
+#include <Interpreters/Context.h>
+
+namespace DB
+{
+
+/// Timeouts for the case when we have just single attempt to connect.
+inline ConnectionTimeouts ConnectionTimeouts::getTCPTimeoutsWithoutFailover(const Settings & settings)
+{
+    return ConnectionTimeouts(settings.connect_timeout, settings.send_timeout, settings.receive_timeout, settings.tcp_keep_alive_timeout);
+}
+
+/// Timeouts for the case when we will try many addresses in a loop.
+inline ConnectionTimeouts ConnectionTimeouts::getTCPTimeoutsWithFailover(const Settings & settings)
+{
+    return ConnectionTimeouts(settings.connect_timeout_with_failover_ms, settings.send_timeout, settings.receive_timeout, settings.tcp_keep_alive_timeout, 0, settings.connect_timeout_with_failover_secure_ms);
+}
+
+inline ConnectionTimeouts ConnectionTimeouts::getHTTPTimeouts(const Context & context)
+{
+    const auto & settings = context.getSettingsRef();
+    const auto & config = context.getConfigRef();
+    Poco::Timespan http_keep_alive_timeout{config.getUInt("keep_alive_timeout", 10), 0};
+    return ConnectionTimeouts(settings.http_connection_timeout, settings.http_send_timeout, settings.http_receive_timeout, settings.tcp_keep_alive_timeout, http_keep_alive_timeout);
+}
+
+}
diff --git a/src/IO/HDFSCommon.cpp b/src/IO/HDFSCommon.cpp
deleted file mode 100644
index 4789dd2b0f4..00000000000
--- a/src/IO/HDFSCommon.cpp
+++ /dev/null
@@ -1,62 +0,0 @@
-#include <IO/HDFSCommon.h>
-#include <Poco/URI.h>
-
-#if USE_HDFS
-#include <Common/Exception.h>
-
-namespace DB
-{
-namespace ErrorCodes
-{
-extern const int BAD_ARGUMENTS;
-extern const int NETWORK_ERROR;
-}
-
-HDFSBuilderPtr createHDFSBuilder(const std::string & uri_str)
-{
-    const Poco::URI uri(uri_str);
-    const auto & host = uri.getHost();
-    auto port = uri.getPort();
-    const std::string path = "//";
-    if (host.empty())
-        throw Exception("Illegal HDFS URI: " + uri.toString(), ErrorCodes::BAD_ARGUMENTS);
-
-    HDFSBuilderPtr builder(hdfsNewBuilder());
-    if (builder == nullptr)
-        throw Exception("Unable to create builder to connect to HDFS: " + uri.toString() + " " + std::string(hdfsGetLastError()),
-            ErrorCodes::NETWORK_ERROR);
-    hdfsBuilderConfSetStr(builder.get(), "input.read.timeout", "60000"); // 1 min
-    hdfsBuilderConfSetStr(builder.get(), "input.write.timeout", "60000"); // 1 min
-    hdfsBuilderConfSetStr(builder.get(), "input.connect.timeout", "60000"); // 1 min
-
-    std::string user_info = uri.getUserInfo();
-    if (!user_info.empty() && user_info.front() != ':')
-    {
-        std::string user;
-        size_t delim_pos = user_info.find(':');
-        if (delim_pos != std::string::npos)
-            user = user_info.substr(0, delim_pos);
-        else
-            user = user_info;
-
-        hdfsBuilderSetUserName(builder.get(), user.c_str());
-    }
-    hdfsBuilderSetNameNode(builder.get(), host.c_str());
-    if (port != 0)
-    {
-        hdfsBuilderSetNameNodePort(builder.get(), port);
-    }
-    return builder;
-}
-
-HDFSFSPtr createHDFSFS(hdfsBuilder * builder)
-{
-    HDFSFSPtr fs(hdfsBuilderConnect(builder));
-    if (fs == nullptr)
-        throw Exception("Unable to connect to HDFS: " + std::string(hdfsGetLastError()),
-            ErrorCodes::NETWORK_ERROR);
-
-    return fs;
-}
-}
-#endif
diff --git a/src/IO/HDFSCommon.h b/src/IO/HDFSCommon.h
deleted file mode 100644
index 3d212ddc347..00000000000
--- a/src/IO/HDFSCommon.h
+++ /dev/null
@@ -1,58 +0,0 @@
-#pragma once
-#include <Common/config.h>
-#include <memory>
-#include <type_traits>
-
-#if USE_HDFS
-#include <hdfs/hdfs.h>
-
-namespace DB
-{
-namespace detail
-{
-struct HDFSBuilderDeleter
-{
-    void operator()(hdfsBuilder * builder_ptr)
-    {
-        hdfsFreeBuilder(builder_ptr);
-    }
-};
-struct HDFSFsDeleter
-{
-    void operator()(hdfsFS fs_ptr)
-    {
-        hdfsDisconnect(fs_ptr);
-    }
-};
-
-}
-
-struct HDFSFileInfo
-{
-    hdfsFileInfo * file_info;
-    int length;
-
-    HDFSFileInfo()
-        : file_info(nullptr)
-        , length(0)
-    {
-    }
-    HDFSFileInfo(const HDFSFileInfo & other) = delete;
-    HDFSFileInfo(HDFSFileInfo && other) = default;
-    HDFSFileInfo & operator=(const HDFSFileInfo & other) = delete;
-    HDFSFileInfo & operator=(HDFSFileInfo && other) = default;
-
-    ~HDFSFileInfo()
-    {
-        hdfsFreeFileInfo(file_info, length);
-    }
-};
-using HDFSBuilderPtr = std::unique_ptr<hdfsBuilder, detail::HDFSBuilderDeleter>;
-using HDFSFSPtr = std::unique_ptr<std::remove_pointer_t<hdfsFS>, detail::HDFSFsDeleter>;
-
-// set read/connect timeout, default value in libhdfs3 is about 1 hour, and too large
-/// TODO Allow to tune from query Settings.
-HDFSBuilderPtr createHDFSBuilder(const std::string & uri_str);
-HDFSFSPtr createHDFSFS(hdfsBuilder * builder);
-}
-#endif
diff --git a/src/IO/MMapReadBufferFromFileDescriptor.cpp b/src/IO/MMapReadBufferFromFileDescriptor.cpp
index bbeec8959b4..2034a116e0c 100644
--- a/src/IO/MMapReadBufferFromFileDescriptor.cpp
+++ b/src/IO/MMapReadBufferFromFileDescriptor.cpp
@@ -6,6 +6,7 @@
 #include <Common/ProfileEvents.h>
 #include <Common/formatReadable.h>
 #include <Common/Exception.h>
+#include <common/getPageSize.h>
 #include <IO/WriteHelpers.h>
 #include <IO/MMapReadBufferFromFileDescriptor.h>
 
@@ -38,7 +39,9 @@ void MMapReadBufferFromFileDescriptor::init(int fd_, size_t offset, size_t lengt
                 ErrorCodes::CANNOT_ALLOCATE_MEMORY);
 
         BufferBase::set(static_cast<char *>(buf), length, 0);
-        ReadBuffer::padded = (length % 4096) > 0 && (length % 4096) <= (4096 - 15); /// TODO determine page size
+
+        size_t page_size = static_cast<size_t>(::getPageSize());
+        ReadBuffer::padded = (length % page_size) > 0 && (length % page_size) <= (page_size - 15);
     }
 }
 
diff --git a/src/IO/S3Common.cpp b/src/IO/S3Common.cpp
index bc49c2641a0..06c51e058a0 100644
--- a/src/IO/S3Common.cpp
+++ b/src/IO/S3Common.cpp
@@ -19,6 +19,7 @@
 #    include <IO/S3/PocoHTTPClient.h>
 #    include <Poco/URI.h>
 #    include <re2/re2.h>
+#    include <boost/algorithm/string/case_conv.hpp>
 #    include <common/logger_useful.h>
 
 namespace
@@ -367,8 +368,12 @@ namespace S3
                 throw Exception(
                     "Bucket name length is out of bounds in virtual hosted style S3 URI: " + bucket + " (" + uri.toString() + ")", ErrorCodes::BAD_ARGUMENTS);
 
-            /// Remove leading '/' from path to extract key.
-            key = uri.getPath().substr(1);
+            if (!uri.getPath().empty())
+            {
+                /// Remove leading '/' from path to extract key.
+                key = uri.getPath().substr(1);
+            }
+
             if (key.empty() || key == "/")
                 throw Exception("Key name is empty in virtual hosted style S3 URI: " + key + " (" + uri.toString() + ")", ErrorCodes::BAD_ARGUMENTS);
             boost::to_upper(name);
diff --git a/src/IO/WriteBufferFromS3.cpp b/src/IO/WriteBufferFromS3.cpp
index e4fb0634bd0..09aabb1b21d 100644
--- a/src/IO/WriteBufferFromS3.cpp
+++ b/src/IO/WriteBufferFromS3.cpp
@@ -5,12 +5,9 @@
 #    include <IO/WriteBufferFromS3.h>
 #    include <IO/WriteHelpers.h>
 
-#    include <aws/core/utils/memory/stl/AWSStreamFwd.h>
-#    include <aws/core/utils/memory/stl/AWSStringStream.h>
 #    include <aws/s3/S3Client.h>
-#    include <aws/s3/model/CompleteMultipartUploadRequest.h>
-#    include <aws/s3/model/AbortMultipartUploadRequest.h>
 #    include <aws/s3/model/CreateMultipartUploadRequest.h>
+#    include <aws/s3/model/CompleteMultipartUploadRequest.h>
 #    include <aws/s3/model/PutObjectRequest.h>
 #    include <aws/s3/model/UploadPartRequest.h>
 #    include <common/logger_useful.h>
@@ -42,22 +39,19 @@ WriteBufferFromS3::WriteBufferFromS3(
     const String & bucket_,
     const String & key_,
     size_t minimum_upload_part_size_,
-    bool is_multipart_,
+    size_t max_single_part_upload_size_,
     std::optional<std::map<String, String>> object_metadata_,
     size_t buffer_size_)
     : BufferWithOwnMemory<WriteBuffer>(buffer_size_, nullptr, 0)
-    , is_multipart(is_multipart_)
     , bucket(bucket_)
     , key(key_)
     , object_metadata(std::move(object_metadata_))
     , client_ptr(std::move(client_ptr_))
-    , minimum_upload_part_size{minimum_upload_part_size_}
-    , temporary_buffer{std::make_unique<WriteBufferFromOwnString>()}
-    , last_part_size{0}
-{
-    if (is_multipart)
-        initiate();
-}
+    , minimum_upload_part_size(minimum_upload_part_size_)
+    , max_single_part_upload_size(max_single_part_upload_size_)
+    , temporary_buffer(Aws::MakeShared<Aws::StringStream>("temporary buffer"))
+    , last_part_size(0)
+{ }
 
 void WriteBufferFromS3::nextImpl()
 {
@@ -68,16 +62,17 @@ void WriteBufferFromS3::nextImpl()
 
     ProfileEvents::increment(ProfileEvents::S3WriteBytes, offset());
 
-    if (is_multipart)
-    {
-        last_part_size += offset();
+    last_part_size += offset();
 
-        if (last_part_size > minimum_upload_part_size)
-        {
-            writePart(temporary_buffer->str());
-            last_part_size = 0;
-            temporary_buffer->restart();
-        }
+    /// Data size exceeds singlepart upload threshold, need to use multipart upload.
+    if (multipart_upload_id.empty() && last_part_size > max_single_part_upload_size)
+        createMultipartUpload();
+
+    if (!multipart_upload_id.empty() && last_part_size > minimum_upload_part_size)
+    {
+        writePart();
+        last_part_size = 0;
+        temporary_buffer = Aws::MakeShared<Aws::StringStream>("temporary buffer");
     }
 }
 
@@ -88,17 +83,23 @@ void WriteBufferFromS3::finalize()
 
 void WriteBufferFromS3::finalizeImpl()
 {
-    if (!finalized)
+    if (finalized)
+        return;
+
+    next();
+
+    if (multipart_upload_id.empty())
     {
-        next();
-
-        if (is_multipart)
-            writePart(temporary_buffer->str());
-
-        complete();
-
-        finalized = true;
+        makeSinglepartUpload();
     }
+    else
+    {
+        /// Write rest of the data as last part.
+        writePart();
+        completeMultipartUpload();
+    }
+
+    finalized = true;
 }
 
 WriteBufferFromS3::~WriteBufferFromS3()
@@ -113,7 +114,7 @@ WriteBufferFromS3::~WriteBufferFromS3()
     }
 }
 
-void WriteBufferFromS3::initiate()
+void WriteBufferFromS3::createMultipartUpload()
 {
     Aws::S3::Model::CreateMultipartUploadRequest req;
     req.SetBucket(bucket);
@@ -125,17 +126,17 @@ void WriteBufferFromS3::initiate()
 
     if (outcome.IsSuccess())
     {
-        upload_id = outcome.GetResult().GetUploadId();
-        LOG_DEBUG(log, "Multipart upload initiated. Upload id: {}", upload_id);
+        multipart_upload_id = outcome.GetResult().GetUploadId();
+        LOG_DEBUG(log, "Multipart upload has created. Upload id: {}", multipart_upload_id);
     }
     else
         throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
 
-void WriteBufferFromS3::writePart(const String & data)
+void WriteBufferFromS3::writePart()
 {
-    if (data.empty())
+    if (temporary_buffer->tellp() <= 0)
         return;
 
     if (part_tags.size() == S3_WARN_MAX_PARTS)
@@ -149,93 +150,71 @@ void WriteBufferFromS3::writePart(const String & data)
     req.SetBucket(bucket);
     req.SetKey(key);
     req.SetPartNumber(part_tags.size() + 1);
-    req.SetUploadId(upload_id);
-    req.SetContentLength(data.size());
-    req.SetBody(std::make_shared<Aws::StringStream>(data));
+    req.SetUploadId(multipart_upload_id);
+    req.SetContentLength(temporary_buffer->tellp());
+    req.SetBody(temporary_buffer);
 
     auto outcome = client_ptr->UploadPart(req);
 
-    LOG_TRACE(log, "Writing part. Bucket: {}, Key: {}, Upload_id: {}, Data size: {}", bucket, key, upload_id, data.size());
+    LOG_TRACE(log, "Writing part. Bucket: {}, Key: {}, Upload_id: {}, Data size: {}", bucket, key, multipart_upload_id, temporary_buffer->tellp());
 
     if (outcome.IsSuccess())
     {
         auto etag = outcome.GetResult().GetETag();
         part_tags.push_back(etag);
-        LOG_DEBUG(log, "Writing part finished. Total parts: {}, Upload_id: {}, Etag: {}", part_tags.size(), upload_id, etag);
+        LOG_DEBUG(log, "Writing part finished. Total parts: {}, Upload_id: {}, Etag: {}", part_tags.size(), multipart_upload_id, etag);
     }
     else
         throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
-
-void WriteBufferFromS3::complete()
+void WriteBufferFromS3::completeMultipartUpload()
 {
-    if (is_multipart)
+    LOG_DEBUG(log, "Completing multipart upload. Bucket: {}, Key: {}, Upload_id: {}", bucket, key, multipart_upload_id);
+
+    Aws::S3::Model::CompleteMultipartUploadRequest req;
+    req.SetBucket(bucket);
+    req.SetKey(key);
+    req.SetUploadId(multipart_upload_id);
+
+    Aws::S3::Model::CompletedMultipartUpload multipart_upload;
+    for (size_t i = 0; i < part_tags.size(); ++i)
     {
-        if (part_tags.empty())
-        {
-            LOG_DEBUG(log, "Multipart upload has no data. Aborting it. Bucket: {}, Key: {}, Upload_id: {}", bucket, key, upload_id);
-
-            Aws::S3::Model::AbortMultipartUploadRequest req;
-            req.SetBucket(bucket);
-            req.SetKey(key);
-            req.SetUploadId(upload_id);
-
-            auto outcome = client_ptr->AbortMultipartUpload(req);
-
-            if (outcome.IsSuccess())
-                LOG_DEBUG(log, "Aborting multipart upload completed. Upload_id: {}", upload_id);
-            else
-                throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
-
-            return;
-        }
-
-        LOG_DEBUG(log, "Completing multipart upload. Bucket: {}, Key: {}, Upload_id: {}", bucket, key, upload_id);
-
-        Aws::S3::Model::CompleteMultipartUploadRequest req;
-        req.SetBucket(bucket);
-        req.SetKey(key);
-        req.SetUploadId(upload_id);
-
-        Aws::S3::Model::CompletedMultipartUpload multipart_upload;
-        for (size_t i = 0; i < part_tags.size(); ++i)
-        {
-            Aws::S3::Model::CompletedPart part;
-            multipart_upload.AddParts(part.WithETag(part_tags[i]).WithPartNumber(i + 1));
-        }
-
-        req.SetMultipartUpload(multipart_upload);
-
-        auto outcome = client_ptr->CompleteMultipartUpload(req);
-
-        if (outcome.IsSuccess())
-            LOG_DEBUG(log, "Multipart upload completed. Upload_id: {}", upload_id);
-        else
-            throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
+        Aws::S3::Model::CompletedPart part;
+        multipart_upload.AddParts(part.WithETag(part_tags[i]).WithPartNumber(i + 1));
     }
+
+    req.SetMultipartUpload(multipart_upload);
+
+    auto outcome = client_ptr->CompleteMultipartUpload(req);
+
+    if (outcome.IsSuccess())
+        LOG_DEBUG(log, "Multipart upload has completed. Upload_id: {}", multipart_upload_id);
     else
-    {
-        LOG_DEBUG(log, "Making single part upload. Bucket: {}, Key: {}", bucket, key);
+        throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
+}
 
-        Aws::S3::Model::PutObjectRequest req;
-        req.SetBucket(bucket);
-        req.SetKey(key);
-        if (object_metadata.has_value())
-            req.SetMetadata(object_metadata.value());
+void WriteBufferFromS3::makeSinglepartUpload()
+{
+    if (temporary_buffer->tellp() <= 0)
+        return;
 
-        /// This could be improved using an adapter to WriteBuffer.
-        const std::shared_ptr<Aws::IOStream> input_data = Aws::MakeShared<Aws::StringStream>("temporary buffer", temporary_buffer->str());
-        temporary_buffer = std::make_unique<WriteBufferFromOwnString>();
-        req.SetBody(input_data);
+    LOG_DEBUG(log, "Making single part upload. Bucket: {}, Key: {}", bucket, key);
 
-        auto outcome = client_ptr->PutObject(req);
+    Aws::S3::Model::PutObjectRequest req;
+    req.SetBucket(bucket);
+    req.SetKey(key);
+    req.SetContentLength(temporary_buffer->tellp());
+    req.SetBody(temporary_buffer);
+    if (object_metadata.has_value())
+        req.SetMetadata(object_metadata.value());
 
-        if (outcome.IsSuccess())
-            LOG_DEBUG(log, "Single part upload has completed. Bucket: {}, Key: {}", bucket, key);
-        else
-            throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
-    }
+    auto outcome = client_ptr->PutObject(req);
+
+    if (outcome.IsSuccess())
+        LOG_DEBUG(log, "Single part upload has completed. Bucket: {}, Key: {}", bucket, key);
+    else
+        throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
 }
diff --git a/src/IO/WriteBufferFromS3.h b/src/IO/WriteBufferFromS3.h
index 9bd7b896d7f..9e4b056603a 100644
--- a/src/IO/WriteBufferFromS3.h
+++ b/src/IO/WriteBufferFromS3.h
@@ -6,11 +6,13 @@
 
 #    include <memory>
 #    include <vector>
+#    include <common/logger_useful.h>
 #    include <common/types.h>
+
 #    include <IO/BufferWithOwnMemory.h>
-#    include <IO/HTTPCommon.h>
 #    include <IO/WriteBuffer.h>
-#    include <IO/WriteBufferFromString.h>
+
+#    include <aws/core/utils/memory/stl/AWSStringStream.h>
 
 namespace Aws::S3
 {
@@ -19,24 +21,30 @@ class S3Client;
 
 namespace DB
 {
-/* Perform S3 HTTP PUT request.
+
+/**
+ * Buffer to write a data to a S3 object with specified bucket and key.
+ * If data size written to the buffer is less than 'max_single_part_upload_size' write is performed using singlepart upload.
+ * In another case multipart upload is used:
+ * Data is divided on chunks with size greater than 'minimum_upload_part_size'. Last chunk can be less than this threshold.
+ * Each chunk is written as a part to S3.
  */
 class WriteBufferFromS3 : public BufferWithOwnMemory<WriteBuffer>
 {
 private:
-    bool is_multipart;
-
     String bucket;
     String key;
     std::optional<std::map<String, String>> object_metadata;
     std::shared_ptr<Aws::S3::S3Client> client_ptr;
     size_t minimum_upload_part_size;
-    std::unique_ptr<WriteBufferFromOwnString> temporary_buffer;
+    size_t max_single_part_upload_size;
+    /// Buffer to accumulate data.
+    std::shared_ptr<Aws::StringStream> temporary_buffer;
     size_t last_part_size;
 
     /// Upload in S3 is made in parts.
     /// We initiate upload, then upload each part and get ETag as a response, and then finish upload with listing all our parts.
-    String upload_id;
+    String multipart_upload_id;
     std::vector<String> part_tags;
 
     Poco::Logger * log = &Poco::Logger::get("WriteBufferFromS3");
@@ -47,7 +55,7 @@ public:
         const String & bucket_,
         const String & key_,
         size_t minimum_upload_part_size_,
-        bool is_multipart,
+        size_t max_single_part_upload_size_,
         std::optional<std::map<String, String>> object_metadata_ = std::nullopt,
         size_t buffer_size_ = DBMS_DEFAULT_BUFFER_SIZE);
 
@@ -61,9 +69,11 @@ public:
 private:
     bool finalized = false;
 
-    void initiate();
-    void writePart(const String & data);
-    void complete();
+    void createMultipartUpload();
+    void writePart();
+    void completeMultipartUpload();
+
+    void makeSinglepartUpload();
 
     void finalizeImpl();
 };
diff --git a/src/IO/WriteHelpers.h b/src/IO/WriteHelpers.h
index e5c6f5b4ab8..f89422f9715 100644
--- a/src/IO/WriteHelpers.h
+++ b/src/IO/WriteHelpers.h
@@ -29,11 +29,14 @@
 #include <IO/DoubleConverter.h>
 #include <IO/WriteBufferFromString.h>
 
-/// There is no dragonbox in Arcadia
-#if !defined(ARCADIA_BUILD)
-#   include <dragonbox/dragonbox_to_chars.h>
-#else
-#   include <ryu/ryu.h>
+#ifdef __clang__
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunused-parameter"
+#pragma clang diagnostic ignored "-Wsign-compare"
+#endif
+#include <dragonbox/dragonbox_to_chars.h>
+#ifdef __clang__
+#pragma clang diagnostic pop
 #endif
 
 #include <Formats/FormatSettings.h>
@@ -233,22 +236,14 @@ inline size_t writeFloatTextFastPath(T x, char * buffer)
         if (DecomposedFloat64(x).is_inside_int64())
             result = itoa(Int64(x), buffer) - buffer;
         else
-#if !defined(ARCADIA_BUILD)
             result = jkj::dragonbox::to_chars_n(x, buffer) - buffer;
-#else
-            result = d2s_buffered_n(x, buffer);
-#endif
     }
     else
     {
         if (DecomposedFloat32(x).is_inside_int32())
             result = itoa(Int32(x), buffer) - buffer;
         else
-#if !defined(ARCADIA_BUILD)
             result = jkj::dragonbox::to_chars_n(x, buffer) - buffer;
-#else
-            result = f2s_buffered_n(x, buffer);
-#endif
     }
 
     if (result <= 0)
diff --git a/src/IO/readFloatText.h b/src/IO/readFloatText.h
index a721642a185..5647d574c62 100644
--- a/src/IO/readFloatText.h
+++ b/src/IO/readFloatText.h
@@ -5,8 +5,15 @@
 #include <common/shift10.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <double-conversion/double-conversion.h>
-#include <fast_float/fast_float.h>
 
+#ifdef __clang__
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunneeded-internal-declaration"
+#endif
+#include <fast_float/fast_float.h>
+#ifdef __clang__
+#pragma clang diagnostic pop
+#endif
 
 /** Methods for reading floating point numbers from text with decimal representation.
   * There are "precise", "fast" and "simple" implementations.
@@ -249,6 +256,19 @@ ReturnType readFloatTextPreciseImpl(T & x, ReadBuffer & buf)
 }
 
 
+// credit: https://johnnylee-sde.github.io/Fast-numeric-string-to-int/
+static inline bool is_made_of_eight_digits_fast(uint64_t val) noexcept
+{
+    return (((val & 0xF0F0F0F0F0F0F0F0) | (((val + 0x0606060606060606) & 0xF0F0F0F0F0F0F0F0) >> 4)) == 0x3333333333333333);
+}
+
+static inline bool is_made_of_eight_digits_fast(const char * chars) noexcept
+{
+    uint64_t val;
+    ::memcpy(&val, chars, 8);
+    return is_made_of_eight_digits_fast(val);
+}
+
 template <size_t N, typename T>
 static inline void readUIntTextUpToNSignificantDigits(T & x, ReadBuffer & buf)
 {
@@ -266,9 +286,6 @@ static inline void readUIntTextUpToNSignificantDigits(T & x, ReadBuffer & buf)
             else
                 return;
         }
-
-        while (!buf.eof() && isNumericASCII(*buf.position()))
-            ++buf.position();
     }
     else
     {
@@ -283,10 +300,16 @@ static inline void readUIntTextUpToNSignificantDigits(T & x, ReadBuffer & buf)
             else
                 return;
         }
-
-        while (!buf.eof() && isNumericASCII(*buf.position()))
-            ++buf.position();
     }
+
+    while (!buf.eof() && (buf.position() + 8 <= buf.buffer().end()) &&
+         is_made_of_eight_digits_fast(buf.position()))
+    {
+        buf.position() += 8;
+    }
+
+    while (!buf.eof() && isNumericASCII(*buf.position()))
+        ++buf.position();
 }
 
 
@@ -319,7 +342,6 @@ ReturnType readFloatTextFastImpl(T & x, ReadBuffer & in)
         ++in.position();
     }
 
-
     auto count_after_sign = in.count();
 
     constexpr int significant_digits = std::numeric_limits<UInt64>::digits10;
diff --git a/src/IO/tests/o_direct_and_dirty_pages.cpp b/src/IO/tests/o_direct_and_dirty_pages.cpp
index 327d6357e9b..c60141d286f 100644
--- a/src/IO/tests/o_direct_and_dirty_pages.cpp
+++ b/src/IO/tests/o_direct_and_dirty_pages.cpp
@@ -3,6 +3,7 @@
 #include <iostream>
 
 #include <common/types.h>
+#include <common/getPageSize.h>
 #include <IO/WriteHelpers.h>
 #include <IO/ReadHelpers.h>
 #include <IO/WriteBufferFromFile.h>
@@ -16,6 +17,7 @@ int main(int, char **)
     {
         static const size_t N = 100000;
         static const size_t BUF_SIZE = 1048576;
+        size_t page_size = static_cast<size_t>(::getPageSize());
 
         ReadBufferFromFile rand_in("/dev/urandom");
         unsigned rand = 0;
@@ -33,7 +35,7 @@ int main(int, char **)
         }
 
         {
-            ReadBufferFromFile rb("test1", BUF_SIZE, O_RDONLY | O_DIRECT, nullptr, 4096);
+            ReadBufferFromFile rb("test1", BUF_SIZE, O_RDONLY | O_DIRECT, nullptr, page_size);
             String res;
             for (size_t i = 0; i < N; ++i)
                 readStringBinary(res, rb);
@@ -44,14 +46,14 @@ int main(int, char **)
         /// Write to file with O_DIRECT, read as usual.
 
         {
-            WriteBufferFromFile wb("test2", BUF_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, 4096);
+            WriteBufferFromFile wb("test2", BUF_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, page_size);
 
             for (size_t i = 0; i < N; ++i)
                 writeStringBinary(test, wb);
 
-            if (wb.offset() % 4096 != 0)
+            if (wb.offset() % page_size != 0)
             {
-                size_t pad = 4096 - wb.offset() % 4096;
+                size_t pad = page_size - wb.offset() % page_size;
                 memset(wb.position(), 0, pad);
                 wb.position() += pad;
             }
diff --git a/src/IO/ya.make b/src/IO/ya.make
index c2913587bb6..4dc3afb2f11 100644
--- a/src/IO/ya.make
+++ b/src/IO/ya.make
@@ -5,6 +5,7 @@ LIBRARY()
 
 ADDINCL(
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 PEERDIR(
@@ -13,6 +14,7 @@ PEERDIR(
     contrib/libs/brotli/enc
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 
diff --git a/src/IO/ya.make.in b/src/IO/ya.make.in
index 9b824228b92..252e102a819 100644
--- a/src/IO/ya.make.in
+++ b/src/IO/ya.make.in
@@ -4,6 +4,7 @@ LIBRARY()
 
 ADDINCL(
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 PEERDIR(
@@ -12,6 +13,7 @@ PEERDIR(
     contrib/libs/brotli/enc
     contrib/libs/poco/NetSSL_OpenSSL
     contrib/libs/zstd
+    contrib/restricted/fast_float
 )
 
 
diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index 75915bdad1a..62cb6047704 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -662,10 +662,10 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
             auto * right_arg = const_cast<Node *>(&actions_dag->addColumn(std::move(column), true));
             auto * left_arg = src_node;
 
-            CastOverloadResolver::Diagnostic diagnostic = {src_node->result_name, res_elem.name};
+            FunctionCast::Diagnostic diagnostic = {src_node->result_name, res_elem.name};
             FunctionOverloadResolverPtr func_builder_cast =
                     std::make_shared<FunctionOverloadResolverAdaptor>(
-                            CastOverloadResolver::createImpl(false, std::move(diagnostic)));
+                            CastOverloadResolver<CastType::nonAccurate>::createImpl(false, std::move(diagnostic)));
 
             Inputs children = { left_arg, right_arg };
             src_node = &actions_dag->addFunction(func_builder_cast, std::move(children), {}, true);
diff --git a/src/Interpreters/Cluster.cpp b/src/Interpreters/Cluster.cpp
index 218502e7f43..c9c56c96cbe 100644
--- a/src/Interpreters/Cluster.cpp
+++ b/src/Interpreters/Cluster.cpp
@@ -5,6 +5,7 @@
 #include <Common/isLocalAddress.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/parseAddress.h>
+#include <Core/Settings.h>
 #include <IO/HexWriteBuffer.h>
 #include <IO/WriteHelpers.h>
 #include <IO/ReadHelpers.h>
diff --git a/src/Interpreters/Cluster.h b/src/Interpreters/Cluster.h
index 4b6ee35efd5..c64d52724e5 100644
--- a/src/Interpreters/Cluster.h
+++ b/src/Interpreters/Cluster.h
@@ -1,13 +1,23 @@
 #pragma once
 
 #include <map>
-#include <Core/Settings.h>
 #include <Client/ConnectionPool.h>
 #include <Client/ConnectionPoolWithFailover.h>
 #include <Poco/Net/SocketAddress.h>
 
+namespace Poco
+{
+    namespace Util
+    {
+        class AbstractConfiguration;
+    }
+}
+
 namespace DB
 {
+
+struct Settings;
+
 namespace ErrorCodes
 {
     extern const int LOGICAL_ERROR;
diff --git a/src/Interpreters/ClusterProxy/IStreamFactory.h b/src/Interpreters/ClusterProxy/IStreamFactory.h
index c0b887e0489..3cf100cd85c 100644
--- a/src/Interpreters/ClusterProxy/IStreamFactory.h
+++ b/src/Interpreters/ClusterProxy/IStreamFactory.h
@@ -32,7 +32,7 @@ public:
     virtual void createForShard(
             const Cluster::ShardInfo & shard_info,
             const String & query, const ASTPtr & query_ast,
-            const Context & context, const ThrottlerPtr & throttler,
+            const std::shared_ptr<Context> & context_ptr, const ThrottlerPtr & throttler,
             const SelectQueryInfo & query_info,
             std::vector<QueryPlanPtr> & res,
             Pipes & remote_pipes,
diff --git a/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp b/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp
index 56f306595ac..e2a7c5b55dc 100644
--- a/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp
+++ b/src/Interpreters/ClusterProxy/SelectStreamFactory.cpp
@@ -7,6 +7,7 @@
 #include <Common/ProfileEvents.h>
 #include <Common/checkStackSize.h>
 #include <TableFunctions/TableFunctionFactory.h>
+#include <IO/ConnectionTimeoutsContext.h>
 
 #include <common/logger_useful.h>
 #include <Processors/Pipe.h>
@@ -113,13 +114,15 @@ String formattedAST(const ASTPtr & ast)
 void SelectStreamFactory::createForShard(
     const Cluster::ShardInfo & shard_info,
     const String &, const ASTPtr & query_ast,
-    const Context & context, const ThrottlerPtr & throttler,
+    const std::shared_ptr<Context> & context_ptr, const ThrottlerPtr & throttler,
     const SelectQueryInfo &,
     std::vector<QueryPlanPtr> & plans,
     Pipes & remote_pipes,
     Pipes & delayed_pipes,
     Poco::Logger * log)
 {
+    const auto & context = *context_ptr;
+
     bool add_agg_info = processed_stage == QueryProcessingStage::WithMergeableState;
     bool add_totals = false;
     bool add_extremes = false;
@@ -143,7 +146,7 @@ void SelectStreamFactory::createForShard(
     auto emplace_remote_stream = [&]()
     {
         auto remote_query_executor = std::make_shared<RemoteQueryExecutor>(
-            shard_info.pool, modified_query, header, context, nullptr, throttler, scalars, external_tables, processed_stage);
+            shard_info.pool, modified_query, header, context, throttler, scalars, external_tables, processed_stage);
         remote_query_executor->setLogger(log);
 
         remote_query_executor->setPoolMode(PoolMode::GET_MANY);
@@ -151,6 +154,7 @@ void SelectStreamFactory::createForShard(
             remote_query_executor->setMainTable(main_table);
 
         remote_pipes.emplace_back(createRemoteSourcePipe(remote_query_executor, add_agg_info, add_totals, add_extremes));
+        remote_pipes.back().addInterpreterContext(context_ptr);
     };
 
     const auto & settings = context.getSettingsRef();
@@ -242,7 +246,8 @@ void SelectStreamFactory::createForShard(
         /// Do it lazily to avoid connecting in the main thread.
 
         auto lazily_create_stream = [
-                pool = shard_info.pool, shard_num = shard_info.shard_num, modified_query, header = header, modified_query_ast, context, throttler,
+                pool = shard_info.pool, shard_num = shard_info.shard_num, modified_query, header = header, modified_query_ast,
+                &context, context_ptr, throttler,
                 main_table = main_table, table_func_ptr = table_func_ptr, scalars = scalars, external_tables = external_tables,
                 stage = processed_stage, local_delay, add_agg_info, add_totals, add_extremes]()
             -> Pipe
@@ -288,13 +293,14 @@ void SelectStreamFactory::createForShard(
                     connections.emplace_back(std::move(try_result.entry));
 
                 auto remote_query_executor = std::make_shared<RemoteQueryExecutor>(
-                    std::move(connections), modified_query, header, context, nullptr, throttler, scalars, external_tables, stage);
+                    std::move(connections), modified_query, header, context, throttler, scalars, external_tables, stage);
 
                 return createRemoteSourcePipe(remote_query_executor, add_agg_info, add_totals, add_extremes);
             }
         };
 
         delayed_pipes.emplace_back(createDelayedPipe(header, lazily_create_stream, add_totals, add_extremes));
+        delayed_pipes.back().addInterpreterContext(context_ptr);
     }
     else
         emplace_remote_stream();
diff --git a/src/Interpreters/ClusterProxy/SelectStreamFactory.h b/src/Interpreters/ClusterProxy/SelectStreamFactory.h
index b51ac109a11..596e99b8a18 100644
--- a/src/Interpreters/ClusterProxy/SelectStreamFactory.h
+++ b/src/Interpreters/ClusterProxy/SelectStreamFactory.h
@@ -37,7 +37,7 @@ public:
     void createForShard(
         const Cluster::ShardInfo & shard_info,
         const String & query, const ASTPtr & query_ast,
-        const Context & context, const ThrottlerPtr & throttler,
+        const std::shared_ptr<Context> & context_ptr, const ThrottlerPtr & throttler,
         const SelectQueryInfo & query_info,
         std::vector<QueryPlanPtr> & plans,
         Pipes & remote_pipes,
diff --git a/src/Interpreters/ClusterProxy/executeQuery.cpp b/src/Interpreters/ClusterProxy/executeQuery.cpp
index c79b17eac2a..59cbae67770 100644
--- a/src/Interpreters/ClusterProxy/executeQuery.cpp
+++ b/src/Interpreters/ClusterProxy/executeQuery.cpp
@@ -19,7 +19,7 @@ namespace DB
 namespace ClusterProxy
 {
 
-Context updateSettingsForCluster(const Cluster & cluster, const Context & context, const Settings & settings, Poco::Logger * log)
+std::shared_ptr<Context> updateSettingsForCluster(const Cluster & cluster, const Context & context, const Settings & settings, Poco::Logger * log)
 {
     Settings new_settings = settings;
     new_settings.queue_max_wait_ms = Cluster::saturate(new_settings.queue_max_wait_ms, settings.max_execution_time);
@@ -78,9 +78,8 @@ Context updateSettingsForCluster(const Cluster & cluster, const Context & contex
         }
     }
 
-    Context new_context(context);
-    new_context.setSettings(new_settings);
-
+    auto new_context = std::make_shared<Context>(context);
+    new_context->setSettings(new_settings);
     return new_context;
 }
 
@@ -99,7 +98,7 @@ void executeQuery(
 
     const std::string query = queryToString(query_ast);
 
-    Context new_context = updateSettingsForCluster(*query_info.cluster, context, settings, log);
+    auto new_context = updateSettingsForCluster(*query_info.cluster, context, settings, log);
 
     ThrottlerPtr user_level_throttler;
     if (auto * process_list_element = context.getProcessListElement())
diff --git a/src/Interpreters/ClusterProxy/executeQuery.h b/src/Interpreters/ClusterProxy/executeQuery.h
index 0b40c1412a1..8840b82d5b2 100644
--- a/src/Interpreters/ClusterProxy/executeQuery.h
+++ b/src/Interpreters/ClusterProxy/executeQuery.h
@@ -27,7 +27,7 @@ class IStreamFactory;
 ///   - optimize_skip_unused_shards_nesting
 ///
 /// @return new Context with adjusted settings
-Context updateSettingsForCluster(const Cluster & cluster, const Context & context, const Settings & settings, Poco::Logger * log = nullptr);
+std::shared_ptr<Context> updateSettingsForCluster(const Cluster & cluster, const Context & context, const Settings & settings, Poco::Logger * log = nullptr);
 
 /// Execute a distributed query, creating a vector of BlockInputStreams, from which the result can be read.
 /// `stream_factory` object encapsulates the logic of creating streams for a different type of query
diff --git a/src/Interpreters/Context.cpp b/src/Interpreters/Context.cpp
index 9955c1ca11f..337753ba0b8 100644
--- a/src/Interpreters/Context.cpp
+++ b/src/Interpreters/Context.cpp
@@ -445,6 +445,8 @@ struct ContextShared
 
         /// Stop trace collector if any
         trace_collector.reset();
+        /// Stop zookeeper connection
+        zookeeper.reset();
         /// Stop test_keeper storage
         test_keeper_storage.reset();
     }
diff --git a/src/Interpreters/DDLWorker.cpp b/src/Interpreters/DDLWorker.cpp
index 5046667fb33..cce62b1a6c4 100644
--- a/src/Interpreters/DDLWorker.cpp
+++ b/src/Interpreters/DDLWorker.cpp
@@ -41,6 +41,11 @@
 
 namespace fs = std::filesystem;
 
+namespace CurrentMetrics
+{
+    extern const Metric MaxDDLEntryID;
+}
+
 namespace DB
 {
 
@@ -312,6 +317,7 @@ DDLWorker::DDLWorker(int pool_size_, const std::string & zk_root_dir, Context &
     , pool_size(pool_size_)
     , worker_pool(pool_size_)
 {
+    CurrentMetrics::set(CurrentMetrics::MaxDDLEntryID, 0);
     last_tasks.reserve(pool_size);
 
     queue_dir = zk_root_dir;
@@ -806,6 +812,22 @@ void DDLWorker::processTask(DDLTask & task)
         task.was_executed = true;
     }
 
+    {
+        DB::ReadBufferFromString in(task.entry_name);
+        DB::assertString("query-", in);
+        UInt64 id;
+        readText(id, in);
+        auto prev_id = max_id.load(std::memory_order_relaxed);
+        while (prev_id < id)
+        {
+            if (max_id.compare_exchange_weak(prev_id, id))
+            {
+                CurrentMetrics::set(CurrentMetrics::MaxDDLEntryID, id);
+                break;
+            }
+        }
+    }
+
     /// FIXME: if server fails right here, the task will be executed twice. We need WAL here.
 
     /// Delete active flag and create finish flag
diff --git a/src/Interpreters/DDLWorker.h b/src/Interpreters/DDLWorker.h
index 7dd9c38e9da..1c365a0d9ba 100644
--- a/src/Interpreters/DDLWorker.h
+++ b/src/Interpreters/DDLWorker.h
@@ -1,6 +1,7 @@
 #pragma once
 
 #include <Interpreters/Cluster.h>
+#include <Interpreters/Context.h>
 #include <DataStreams/BlockIO.h>
 #include <Common/CurrentThread.h>
 #include <Common/ThreadPool.h>
@@ -137,6 +138,8 @@ private:
 
     ThreadGroupStatusPtr thread_group;
 
+    std::atomic<UInt64> max_id = 0;
+
     friend class DDLQueryStatusInputStream;
     friend struct DDLTask;
 };
diff --git a/src/Interpreters/DatabaseCatalog.cpp b/src/Interpreters/DatabaseCatalog.cpp
index ea1984c96fb..e9caaab4ef9 100644
--- a/src/Interpreters/DatabaseCatalog.cpp
+++ b/src/Interpreters/DatabaseCatalog.cpp
@@ -4,7 +4,7 @@
 #include <Storages/IStorage.h>
 #include <Databases/IDatabase.h>
 #include <Databases/DatabaseMemory.h>
-#include <Databases/DatabaseAtomic.h>
+#include <Databases/DatabaseOnDisk.h>
 #include <Poco/File.h>
 #include <Common/quoteString.h>
 #include <Storages/StorageMemory.h>
@@ -16,6 +16,15 @@
 #include <Common/CurrentMetrics.h>
 #include <common/logger_useful.h>
 
+#if !defined(ARCADIA_BUILD)
+#    include "config_core.h"
+#endif
+
+#if USE_MYSQL
+#    include <Databases/MySQL/MaterializeMySQLSyncThread.h>
+#    include <Storages/StorageMaterializeMySQL.h>
+#endif
+
 #include <filesystem>
 
 
@@ -202,7 +211,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
     if (!table_id)
     {
         if (exception)
-            exception->emplace("Cannot find table: StorageID is empty", ErrorCodes::UNKNOWN_TABLE);
+            exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Cannot find table: StorageID is empty");
         return {};
     }
 
@@ -214,9 +223,18 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         {
             assert(!db_and_table.first && !db_and_table.second);
             if (exception)
-                exception->emplace("Table " + table_id.getNameForLogs() + " doesn't exist.", ErrorCodes::UNKNOWN_TABLE);
+                exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Table {} doesn't exist", table_id.getNameForLogs());
             return {};
         }
+
+#if USE_MYSQL
+        /// It's definitely not the best place for this logic, but behaviour must be consistent with DatabaseMaterializeMySQL::tryGetTable(...)
+        if (db_and_table.first->getEngineName() == "MaterializeMySQL")
+        {
+            if (!MaterializeMySQLSyncThread::isMySQLSyncThread())
+                db_and_table.second = std::make_shared<StorageMaterializeMySQL>(std::move(db_and_table.second), db_and_table.first.get());
+        }
+#endif
         return db_and_table;
     }
 
@@ -226,7 +244,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         /// If table_id has no UUID, then the name of database was specified by user and table_id was not resolved through context.
         /// Do not allow access to TEMPORARY_DATABASE because it contains all temporary tables of all contexts and users.
         if (exception)
-            exception->emplace("Direct access to `" + String(TEMPORARY_DATABASE) + "` database is not allowed.", ErrorCodes::DATABASE_ACCESS_DENIED);
+            exception->emplace(ErrorCodes::DATABASE_ACCESS_DENIED, "Direct access to `{}` database is not allowed", String(TEMPORARY_DATABASE));
         return {};
     }
 
@@ -237,8 +255,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         if (databases.end() == it)
         {
             if (exception)
-                exception->emplace("Database " + backQuoteIfNeed(table_id.getDatabaseName()) + " doesn't exist",
-                                   ErrorCodes::UNKNOWN_DATABASE);
+                exception->emplace(ErrorCodes::UNKNOWN_DATABASE, "Database {} doesn't exist", backQuoteIfNeed(table_id.getDatabaseName()));
             return {};
         }
         database = it->second;
@@ -246,7 +263,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
 
     auto table = database->tryGetTable(table_id.table_name, context);
     if (!table && exception)
-            exception->emplace("Table " + table_id.getNameForLogs() + " doesn't exist.", ErrorCodes::UNKNOWN_TABLE);
+            exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Table {} doesn't exist", table_id.getNameForLogs());
     if (!table)
         database = nullptr;
 
@@ -286,7 +303,6 @@ void DatabaseCatalog::attachDatabase(const String & database_name, const Databas
     assertDatabaseDoesntExistUnlocked(database_name);
     databases.emplace(database_name, database);
     UUID db_uuid = database->getUUID();
-    assert((db_uuid != UUIDHelpers::Nil) ^ (dynamic_cast<DatabaseAtomic *>(database.get()) == nullptr));
     if (db_uuid != UUIDHelpers::Nil)
         db_uuid_map.emplace(db_uuid, database);
 }
@@ -313,9 +329,8 @@ DatabasePtr DatabaseCatalog::detachDatabase(const String & database_name, bool d
             if (!db->empty())
                 throw Exception("New table appeared in database being dropped or detached. Try again.",
                                 ErrorCodes::DATABASE_NOT_EMPTY);
-            auto * database_atomic = typeid_cast<DatabaseAtomic *>(db.get());
-            if (!drop && database_atomic)
-                database_atomic->assertCanBeDetached(false);
+            if (!drop)
+                db->assertCanBeDetached(false);
         }
         catch (...)
         {
diff --git a/src/Interpreters/ExpressionActions.h b/src/Interpreters/ExpressionActions.h
index 2b1aa5e2456..5104f1e8e72 100644
--- a/src/Interpreters/ExpressionActions.h
+++ b/src/Interpreters/ExpressionActions.h
@@ -2,7 +2,6 @@
 
 #include <Core/Block.h>
 #include <Core/ColumnNumbers.h>
-#include <Core/Settings.h>
 #include <Interpreters/ActionsDAG.h>
 
 #include <variant>
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 72d48efdd80..1b93c090842 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -112,6 +112,11 @@ bool sanitizeBlock(Block & block, bool throw_if_cannot_create_column)
     return true;
 }
 
+ExpressionAnalyzer::ExtractedSettings::ExtractedSettings(const Settings & settings_)
+    : use_index_for_in_with_subqueries(settings_.use_index_for_in_with_subqueries)
+    , size_limits_for_set(settings_.max_rows_in_set, settings_.max_bytes_in_set, settings_.set_overflow_mode)
+{}
+
 
 ExpressionAnalyzer::ExpressionAnalyzer(
     const ASTPtr & query_,
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index 6e87d0b7250..2567b32e37e 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -1,6 +1,5 @@
 #pragma once
 
-#include <Core/Settings.h>
 #include <DataStreams/IBlockStream_fwd.h>
 #include <Columns/FilterDescription.h>
 #include <Interpreters/AggregateDescription.h>
@@ -16,6 +15,7 @@ namespace DB
 
 class Block;
 class Context;
+struct Settings;
 
 struct ExpressionActionsChain;
 class ExpressionActions;
@@ -80,10 +80,7 @@ private:
         const bool use_index_for_in_with_subqueries;
         const SizeLimits size_limits_for_set;
 
-        ExtractedSettings(const Settings & settings_)
-        :   use_index_for_in_with_subqueries(settings_.use_index_for_in_with_subqueries),
-            size_limits_for_set(settings_.max_rows_in_set, settings_.max_bytes_in_set, settings_.set_overflow_mode)
-        {}
+        ExtractedSettings(const Settings & settings_);
     };
 
 public:
diff --git a/src/Interpreters/IInterpreterUnionOrSelectQuery.h b/src/Interpreters/IInterpreterUnionOrSelectQuery.h
index 67e3cf25e5a..aefa78260c7 100644
--- a/src/Interpreters/IInterpreterUnionOrSelectQuery.h
+++ b/src/Interpreters/IInterpreterUnionOrSelectQuery.h
@@ -34,6 +34,8 @@ protected:
     Block result_header;
     SelectQueryOptions options;
     size_t max_streams = 1;
+    bool settings_limit_offset_needed = false;
+    bool settings_limit_offset_done = false;
 };
 }
 
diff --git a/src/Interpreters/InterpreterCreateQuery.cpp b/src/Interpreters/InterpreterCreateQuery.cpp
index d57a14a61dc..2446075d38a 100644
--- a/src/Interpreters/InterpreterCreateQuery.cpp
+++ b/src/Interpreters/InterpreterCreateQuery.cpp
@@ -157,6 +157,35 @@ BlockIO InterpreterCreateQuery::createDatabase(ASTCreateQuery & create)
         if (!create.attach && fs::exists(metadata_path))
             throw Exception(ErrorCodes::DATABASE_ALREADY_EXISTS, "Metadata directory {} already exists", metadata_path.string());
     }
+    else if (create.storage->engine->name == "MaterializeMySQL")
+    {
+        /// It creates nested database with Ordinary or Atomic engine depending on UUID in query and default engine setting.
+        /// Do nothing if it's an internal ATTACH on server startup or short-syntax ATTACH query from user,
+        /// because we got correct query from the metadata file in this case.
+        /// If we got query from user, then normalize it first.
+        bool attach_from_user = create.attach && !internal && !create.attach_short_syntax;
+        bool create_from_user = !create.attach;
+
+        if (create_from_user)
+        {
+            const auto & default_engine = context.getSettingsRef().default_database_engine.value;
+            if (create.uuid == UUIDHelpers::Nil && default_engine == DefaultDatabaseEngine::Atomic)
+                create.uuid = UUIDHelpers::generateV4();    /// Will enable Atomic engine for nested database
+        }
+        else if (attach_from_user && create.uuid == UUIDHelpers::Nil)
+        {
+            /// Ambiguity is possible: should we attach nested database as Ordinary
+            /// or throw "UUID must be specified" for Atomic? So we suggest short syntax for Ordinary.
+            throw Exception("Use short attach syntax ('ATTACH DATABASE name;' without engine) to attach existing database "
+                            "or specify UUID to attach new database with Atomic engine", ErrorCodes::INCORRECT_QUERY);
+        }
+
+        /// Set metadata path according to nested engine
+        if (create.uuid == UUIDHelpers::Nil)
+            metadata_path = metadata_path / "metadata" / database_name_escaped;
+        else
+            metadata_path = metadata_path / "store" / DatabaseCatalog::getPathForUUID(create.uuid);
+    }
     else
     {
         bool is_on_cluster = context.getClientInfo().query_kind == ClientInfo::QueryKind::SECONDARY_QUERY;
@@ -586,6 +615,23 @@ void InterpreterCreateQuery::validateTableStructure(const ASTCreateQuery & creat
             }
         }
     }
+
+    if (!create.attach && !settings.allow_experimental_map_type)
+    {
+        for (const auto & name_and_type_pair : properties.columns.getAllPhysical())
+        {
+            WhichDataType which(*name_and_type_pair.type);
+            if (which.isMap())
+            {
+                const auto & type_name = name_and_type_pair.type->getName();
+                String message = "Cannot create table with column '" + name_and_type_pair.name + "' which type is '"
+                                 + type_name + "' because experimental Map type is not allowed. "
+                                 + "Set 'allow_experimental_map_type = 1' setting to enable";
+                throw Exception(message, ErrorCodes::ILLEGAL_COLUMN);
+            }
+        }
+
+    }
 }
 
 void InterpreterCreateQuery::setEngine(ASTCreateQuery & create) const
@@ -655,7 +701,7 @@ void InterpreterCreateQuery::assertOrSetUUID(ASTCreateQuery & create, const Data
 
     bool from_path = create.attach_from_path.has_value();
 
-    if (database->getEngineName() == "Atomic")
+    if (database->getUUID() != UUIDHelpers::Nil)
     {
         if (create.attach && !from_path && create.uuid == UUIDHelpers::Nil)
         {
@@ -705,7 +751,7 @@ BlockIO InterpreterCreateQuery::createTable(ASTCreateQuery & create)
         auto database = DatabaseCatalog::instance().getDatabase(database_name);
         bool if_not_exists = create.if_not_exists;
 
-        // Table SQL definition is available even if the table is detached
+        // Table SQL definition is available even if the table is detached (even permanently)
         auto query = database->getCreateTableQuery(create.table, context);
         create = query->as<ASTCreateQuery &>(); // Copy the saved create query, but use ATTACH instead of CREATE
         if (create.is_dictionary)
diff --git a/src/Interpreters/InterpreterDropQuery.cpp b/src/Interpreters/InterpreterDropQuery.cpp
index 5f7f70fdda4..47bafdbdcb4 100644
--- a/src/Interpreters/InterpreterDropQuery.cpp
+++ b/src/Interpreters/InterpreterDropQuery.cpp
@@ -11,7 +11,14 @@
 #include <Common/escapeForFileName.h>
 #include <Common/quoteString.h>
 #include <Common/typeid_cast.h>
-#include <Databases/DatabaseAtomic.h>
+
+#if !defined(ARCADIA_BUILD)
+#    include "config_core.h"
+#endif
+
+#if USE_MYSQL
+#   include <Databases/MySQL/DatabaseMaterializeMySQL.h>
+#endif
 
 
 namespace DB
@@ -23,6 +30,7 @@ namespace ErrorCodes
     extern const int SYNTAX_ERROR;
     extern const int UNKNOWN_TABLE;
     extern const int UNKNOWN_DICTIONARY;
+    extern const int NOT_IMPLEMENTED;
 }
 
 
@@ -48,6 +56,8 @@ BlockIO InterpreterDropQuery::execute()
     {
         if (!drop.is_dictionary)
             return executeToTable(drop);
+        else if (drop.permanently && drop.kind == ASTDropQuery::Kind::Detach)
+            throw Exception("DETACH PERMANENTLY is not implemented for dictionaries", ErrorCodes::NOT_IMPLEMENTED);
         else
             return executeToDictionary(drop.database, drop.table, drop.kind, drop.if_exists, drop.temporary, drop.no_ddl_lock);
     }
@@ -66,10 +76,7 @@ void InterpreterDropQuery::waitForTableToBeActuallyDroppedOrDetached(const ASTDr
     if (query.kind == ASTDropQuery::Kind::Drop)
         DatabaseCatalog::instance().waitTableFinallyDropped(uuid_to_wait);
     else if (query.kind == ASTDropQuery::Kind::Detach)
-    {
-        if (auto * atomic = typeid_cast<DatabaseAtomic *>(db.get()))
-            atomic->waitDetachedTableNotInUse(uuid_to_wait);
-    }
+        db->waitDetachedTableNotInUse(uuid_to_wait);
 }
 
 BlockIO InterpreterDropQuery::executeToTable(const ASTDropQuery & query)
@@ -122,10 +129,20 @@ BlockIO InterpreterDropQuery::executeToTableImpl(const ASTDropQuery & query, Dat
             table->checkTableCanBeDetached();
             table->shutdown();
             TableExclusiveLockHolder table_lock;
-            if (database->getEngineName() != "Atomic")
+            if (database->getUUID() == UUIDHelpers::Nil)
                 table_lock = table->lockExclusively(context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
-            /// Drop table from memory, don't touch data and metadata
-            database->detachTable(table_id.table_name);
+
+            if (query.permanently)
+            {
+                /// Drop table from memory, don't touch data, metadata file renamed and will be skipped during server restart
+                database->detachTablePermanently(table_id.table_name);
+            }
+            else
+            {
+                /// Drop table from memory, don't touch data and metadata
+                database->detachTable(table_id.table_name);
+            }
+
         }
         else if (query.kind == ASTDropQuery::Kind::Truncate)
         {
@@ -145,7 +162,7 @@ BlockIO InterpreterDropQuery::executeToTableImpl(const ASTDropQuery & query, Dat
             table->shutdown();
 
             TableExclusiveLockHolder table_lock;
-            if (database->getEngineName() != "Atomic")
+            if (database->getUUID() == UUIDHelpers::Nil)
                 table_lock = table->lockExclusively(context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
 
             database->dropTable(context, table_id.table_name, query.no_delay);
@@ -282,6 +299,14 @@ BlockIO InterpreterDropQuery::executeToDatabaseImpl(const ASTDropQuery & query,
             bool drop = query.kind == ASTDropQuery::Kind::Drop;
             context.checkAccess(AccessType::DROP_DATABASE, database_name);
 
+            if (query.kind == ASTDropQuery::Kind::Detach && query.permanently)
+                throw Exception("DETACH PERMANENTLY is not implemented for databases", ErrorCodes::NOT_IMPLEMENTED);
+
+#if USE_MYSQL
+            if (database->getEngineName() == "MaterializeMySQL")
+                stopDatabaseSynchronization(database);
+#endif
+
             if (database->shouldBeEmptyOnDetach())
             {
                 /// DETACH or DROP all tables and dictionaries inside database.
@@ -312,9 +337,8 @@ BlockIO InterpreterDropQuery::executeToDatabaseImpl(const ASTDropQuery & query,
             /// Protects from concurrent CREATE TABLE queries
             auto db_guard = DatabaseCatalog::instance().getExclusiveDDLGuardForDatabase(database_name);
 
-            auto * database_atomic = typeid_cast<DatabaseAtomic *>(database.get());
-            if (!drop && database_atomic)
-                database_atomic->assertCanBeDetached(true);
+            if (!drop)
+                database->assertCanBeDetached(true);
 
             /// DETACH or DROP database itself
             DatabaseCatalog::instance().detachDatabase(database_name, drop, database->shouldBeEmptyOnDetach());
diff --git a/src/Interpreters/InterpreterInsertQuery.cpp b/src/Interpreters/InterpreterInsertQuery.cpp
index 39381bf0241..e3fc67b432c 100644
--- a/src/Interpreters/InterpreterInsertQuery.cpp
+++ b/src/Interpreters/InterpreterInsertQuery.cpp
@@ -13,6 +13,7 @@
 #include <DataStreams/SquashingBlockOutputStream.h>
 #include <DataStreams/copyData.h>
 #include <IO/ConcatReadBuffer.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <Interpreters/InterpreterSelectWithUnionQuery.h>
 #include <Interpreters/InterpreterWatchQuery.h>
 #include <Interpreters/JoinedTables.h>
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index f5c71570de3..aee4eae8782 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1552,7 +1552,12 @@ void InterpreterSelectQuery::executeFetchColumns(
         throw Exception("Logical error in InterpreterSelectQuery: nowhere to read", ErrorCodes::LOGICAL_ERROR);
 
     /// Specify the number of threads only if it wasn't specified in storage.
-    if (!query_plan.getMaxThreads())
+    ///
+    /// But in case of remote query and prefer_localhost_replica=1 (default)
+    /// The inner local query (that is done in the same process, without
+    /// network interaction), it will setMaxThreads earlier and distributed
+    /// query will not update it.
+    if (!query_plan.getMaxThreads() || is_remote)
         query_plan.setMaxThreads(max_threads_execute_query);
 
     /// Aliases in table declaration.
@@ -1784,7 +1789,7 @@ void InterpreterSelectQuery::executeOrder(QueryPlan & query_plan, InputOrderInfo
     auto merge_sorting_step = std::make_unique<MergeSortingStep>(
             query_plan.getCurrentDataStream(),
             output_order_descr, settings.max_block_size, limit,
-            settings.max_bytes_before_remerge_sort,
+            settings.max_bytes_before_remerge_sort, settings.remerge_sort_lowered_memory_bytes_ratio,
             settings.max_bytes_before_external_sort, context->getTemporaryVolume(),
             settings.min_free_disk_space_for_temporary_data);
 
diff --git a/src/Interpreters/InterpreterSelectWithUnionQuery.cpp b/src/Interpreters/InterpreterSelectWithUnionQuery.cpp
index a836dc8b271..e6610df43ff 100644
--- a/src/Interpreters/InterpreterSelectWithUnionQuery.cpp
+++ b/src/Interpreters/InterpreterSelectWithUnionQuery.cpp
@@ -9,10 +9,14 @@
 #include <Processors/QueryPlan/IQueryPlanStep.h>
 #include <Processors/QueryPlan/QueryPlan.h>
 #include <Processors/QueryPlan/UnionStep.h>
+#include <Processors/QueryPlan/LimitStep.h>
+#include <Processors/QueryPlan/OffsetStep.h>
 #include <Common/typeid_cast.h>
 
 #include <Interpreters/InDepthNodeVisitor.h>
 
+#include <algorithm>
+
 namespace DB
 {
 
@@ -128,24 +132,28 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
     const ASTPtr & query_ptr_, const Context & context_, const SelectQueryOptions & options_, const Names & required_result_column_names)
     : IInterpreterUnionOrSelectQuery(query_ptr_, context_, options_)
 {
-    auto & ast = query_ptr->as<ASTSelectWithUnionQuery &>();
+    ASTSelectWithUnionQuery * ast = query_ptr->as<ASTSelectWithUnionQuery>();
+
+    const Settings & settings = context->getSettingsRef();
+    if (options.subquery_depth == 0 && (settings.limit > 0 || settings.offset > 0))
+        settings_limit_offset_needed = true;
 
     /// Normalize AST Tree
-    if (!ast.is_normalized)
+    if (!ast->is_normalized)
     {
-        CustomizeASTSelectWithUnionQueryNormalizeVisitor::Data union_default_mode{context->getSettingsRef().union_default_mode};
+        CustomizeASTSelectWithUnionQueryNormalizeVisitor::Data union_default_mode{settings.union_default_mode};
         CustomizeASTSelectWithUnionQueryNormalizeVisitor(union_default_mode).visit(query_ptr);
 
         /// After normalization, if it only has one ASTSelectWithUnionQuery child,
         /// we can lift it up, this can reduce one unnecessary recursion later.
-        if (ast.list_of_selects->children.size() == 1 && ast.list_of_selects->children.at(0)->as<ASTSelectWithUnionQuery>())
+        if (ast->list_of_selects->children.size() == 1 && ast->list_of_selects->children.at(0)->as<ASTSelectWithUnionQuery>())
         {
-            query_ptr = std::move(ast.list_of_selects->children.at(0));
-            ast = query_ptr->as<ASTSelectWithUnionQuery &>();
+            query_ptr = std::move(ast->list_of_selects->children.at(0));
+            ast = query_ptr->as<ASTSelectWithUnionQuery>();
         }
     }
 
-    size_t num_children = ast.list_of_selects->children.size();
+    size_t num_children = ast->list_of_selects->children.size();
     if (!num_children)
         throw Exception("Logical error: no children in ASTSelectWithUnionQuery", ErrorCodes::LOGICAL_ERROR);
 
@@ -161,7 +169,7 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
         /// Result header if there are no filtering by 'required_result_column_names'.
         /// We use it to determine positions of 'required_result_column_names' in SELECT clause.
 
-        Block full_result_header = getCurrentChildResultHeader(ast.list_of_selects->children.at(0), required_result_column_names);
+        Block full_result_header = getCurrentChildResultHeader(ast->list_of_selects->children.at(0), required_result_column_names);
 
         std::vector<size_t> positions_of_required_result_columns(required_result_column_names.size());
 
@@ -171,7 +179,7 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
         for (size_t query_num = 1; query_num < num_children; ++query_num)
         {
             Block full_result_header_for_current_select
-                = getCurrentChildResultHeader(ast.list_of_selects->children.at(query_num), required_result_column_names);
+                = getCurrentChildResultHeader(ast->list_of_selects->children.at(query_num), required_result_column_names);
 
             if (full_result_header_for_current_select.columns() != full_result_header.columns())
                 throw Exception("Different number of columns in UNION ALL elements:\n"
@@ -186,13 +194,59 @@ InterpreterSelectWithUnionQuery::InterpreterSelectWithUnionQuery(
         }
     }
 
+    if (num_children == 1 && settings_limit_offset_needed)
+    {
+        const ASTPtr first_select_ast = ast->list_of_selects->children.at(0);
+        ASTSelectQuery * select_query = first_select_ast->as<ASTSelectQuery>();
+
+        if (!select_query->withFill() && !select_query->limit_with_ties)
+        {
+            UInt64 limit_length = 0;
+            UInt64 limit_offset = 0;
+
+            const ASTPtr limit_offset_ast = select_query->limitOffset();
+            if (limit_offset_ast)
+            {
+                limit_offset = limit_offset_ast->as<ASTLiteral &>().value.safeGet<UInt64>();
+                UInt64 new_limit_offset = settings.offset + limit_offset;
+                limit_offset_ast->as<ASTLiteral &>().value = Field(new_limit_offset);
+            }
+            else if (settings.offset)
+            {
+                ASTPtr new_limit_offset_ast = std::make_shared<ASTLiteral>(Field(UInt64(settings.offset)));
+                select_query->setExpression(ASTSelectQuery::Expression::LIMIT_OFFSET, std::move(new_limit_offset_ast));
+            }
+
+            const ASTPtr limit_length_ast = select_query->limitLength();
+            if (limit_length_ast)
+            {
+                limit_length = limit_length_ast->as<ASTLiteral &>().value.safeGet<UInt64>();
+
+                UInt64 new_limit_length = 0;
+                if (settings.offset == 0)
+                    new_limit_length = std::min(limit_length, UInt64(settings.limit));
+                else if (settings.offset < limit_length)
+                    new_limit_length =  settings.limit ? std::min(UInt64(settings.limit), limit_length - settings.offset) : (limit_length - settings.offset);
+
+                limit_length_ast->as<ASTLiteral &>().value = Field(new_limit_length);
+            }
+            else if (settings.limit)
+            {
+                ASTPtr new_limit_length_ast = std::make_shared<ASTLiteral>(Field(UInt64(settings.limit)));
+                select_query->setExpression(ASTSelectQuery::Expression::LIMIT_LENGTH, std::move(new_limit_length_ast));
+            }
+
+            settings_limit_offset_done = true;
+        }
+    }
+
     for (size_t query_num = 0; query_num < num_children; ++query_num)
     {
         const Names & current_required_result_column_names
             = query_num == 0 ? required_result_column_names : required_result_column_names_for_other_selects[query_num];
 
         nested_interpreters.emplace_back(
-            buildCurrentChildInterpreter(ast.list_of_selects->children.at(query_num), current_required_result_column_names));
+            buildCurrentChildInterpreter(ast->list_of_selects->children.at(query_num), current_required_result_column_names));
     }
 
     /// Determine structure of the result.
@@ -293,39 +347,57 @@ void InterpreterSelectWithUnionQuery::buildQueryPlan(QueryPlan & query_plan)
 {
     // auto num_distinct_union = optimizeUnionList();
     size_t num_plans = nested_interpreters.size();
+    const Settings & settings = context->getSettingsRef();
 
     /// Skip union for single interpreter.
     if (num_plans == 1)
     {
         nested_interpreters.front()->buildQueryPlan(query_plan);
-        return;
+    }
+    else
+    {
+        std::vector<std::unique_ptr<QueryPlan>> plans(num_plans);
+        DataStreams data_streams(num_plans);
+
+        for (size_t i = 0; i < num_plans; ++i)
+        {
+            plans[i] = std::make_unique<QueryPlan>();
+            nested_interpreters[i]->buildQueryPlan(*plans[i]);
+            data_streams[i] = plans[i]->getCurrentDataStream();
+        }
+
+        auto max_threads = context->getSettingsRef().max_threads;
+        auto union_step = std::make_unique<UnionStep>(std::move(data_streams), result_header, max_threads);
+
+        query_plan.unitePlans(std::move(union_step), std::move(plans));
+
+        const auto & query = query_ptr->as<ASTSelectWithUnionQuery &>();
+        if (query.union_mode == ASTSelectWithUnionQuery::Mode::DISTINCT)
+        {
+            /// Add distinct transform
+            SizeLimits limits(settings.max_rows_in_distinct, settings.max_bytes_in_distinct, settings.distinct_overflow_mode);
+
+            auto distinct_step
+                = std::make_unique<DistinctStep>(query_plan.getCurrentDataStream(), limits, 0, result_header.getNames(), false);
+
+            query_plan.addStep(std::move(distinct_step));
+        }
     }
 
-    std::vector<std::unique_ptr<QueryPlan>> plans(num_plans);
-    DataStreams data_streams(num_plans);
-
-    for (size_t i = 0; i < num_plans; ++i)
+    if (settings_limit_offset_needed && !settings_limit_offset_done)
     {
-        plans[i] = std::make_unique<QueryPlan>();
-        nested_interpreters[i]->buildQueryPlan(*plans[i]);
-        data_streams[i] = plans[i]->getCurrentDataStream();
-    }
-
-    auto max_threads = context->getSettingsRef().max_threads;
-    auto union_step = std::make_unique<UnionStep>(std::move(data_streams), result_header, max_threads);
-
-    query_plan.unitePlans(std::move(union_step), std::move(plans));
-
-    const auto & query = query_ptr->as<ASTSelectWithUnionQuery &>();
-    if (query.union_mode == ASTSelectWithUnionQuery::Mode::DISTINCT)
-    {
-        /// Add distinct transform
-        const Settings & settings = context->getSettingsRef();
-        SizeLimits limits(settings.max_rows_in_distinct, settings.max_bytes_in_distinct, settings.distinct_overflow_mode);
-
-        auto distinct_step = std::make_unique<DistinctStep>(query_plan.getCurrentDataStream(), limits, 0, result_header.getNames(), false);
-
-        query_plan.addStep(std::move(distinct_step));
+        if (settings.limit > 0)
+        {
+            auto limit = std::make_unique<LimitStep>(query_plan.getCurrentDataStream(), settings.limit, settings.offset);
+            limit->setStepDescription("LIMIT OFFSET for SETTINGS");
+            query_plan.addStep(std::move(limit));
+        }
+        else
+        {
+            auto offset = std::make_unique<OffsetStep>(query_plan.getCurrentDataStream(), settings.offset);
+            offset->setStepDescription("OFFSET for SETTINGS");
+            query_plan.addStep(std::move(offset));
+        }
     }
 
 }
diff --git a/src/Interpreters/InterpreterShowTablesQuery.cpp b/src/Interpreters/InterpreterShowTablesQuery.cpp
index 861fa8c3505..49c55fa3a93 100644
--- a/src/Interpreters/InterpreterShowTablesQuery.cpp
+++ b/src/Interpreters/InterpreterShowTablesQuery.cpp
@@ -79,6 +79,26 @@ String InterpreterShowTablesQuery::getRewrittenQuery()
         return rewritten_query.str();
     }
 
+    /// SHOW SETTINGS
+    if (query.m_settings)
+    {
+        WriteBufferFromOwnString rewritten_query;
+        rewritten_query << "SELECT name, type, value FROM system.settings";
+
+        if (query.changed)
+            rewritten_query << " WHERE changed = 1";
+
+        if (!query.like.empty())
+        {
+            rewritten_query
+                << (query.changed ? " AND name " : " WHERE name ")
+                << (query.case_insensitive_like ? "ILIKE " : "LIKE ")
+                << DB::quote << query.like;
+        }
+
+        return rewritten_query.str();
+    }
+
     if (query.temporary && !query.from.empty())
         throw Exception("The `FROM` and `TEMPORARY` cannot be used together in `SHOW TABLES`", ErrorCodes::SYNTAX_ERROR);
 
diff --git a/src/Interpreters/MergeJoin.cpp b/src/Interpreters/MergeJoin.cpp
index 665ec4d60f3..4839980eac7 100644
--- a/src/Interpreters/MergeJoin.cpp
+++ b/src/Interpreters/MergeJoin.cpp
@@ -180,12 +180,16 @@ class MergeJoinCursor
 public:
     MergeJoinCursor(const Block & block, const SortDescription & desc_)
         : impl(SortCursorImpl(block, desc_))
-    {}
+    {
+        /// SortCursorImpl can work with permutation, but MergeJoinCursor can't.
+        if (impl.permutation)
+            throw Exception("Logical error: MergeJoinCursor doesn't support permutation", ErrorCodes::LOGICAL_ERROR);
+    }
 
-    size_t position() const { return impl.pos; }
+    size_t position() const { return impl.getRow(); }
     size_t end() const { return impl.rows; }
-    bool atEnd() const { return impl.pos >= impl.rows; }
-    void nextN(size_t num) { impl.pos += num; }
+    bool atEnd() const { return impl.getRow() >= impl.rows; }
+    void nextN(size_t num) { impl.getPosRef() += num; }
 
     void setCompareNullability(const MergeJoinCursor & rhs)
     {
@@ -254,10 +258,10 @@ private:
             else if (cmp > 0)
                 rhs.impl.next();
             else if (!cmp)
-                return Range{impl.pos, rhs.impl.pos, getEqualLength(), rhs.getEqualLength()};
+                return Range{impl.getRow(), rhs.impl.getRow(), getEqualLength(), rhs.getEqualLength()};
         }
 
-        return Range{impl.pos, rhs.impl.pos, 0, 0};
+        return Range{impl.getRow(), rhs.impl.getRow(), 0, 0};
     }
 
     template <bool left_nulls, bool right_nulls>
@@ -268,7 +272,7 @@ private:
             const auto * left_column = impl.sort_columns[i];
             const auto * right_column = rhs.impl.sort_columns[i];
 
-            int res = nullableCompareAt<left_nulls, right_nulls>(*left_column, *right_column, impl.pos, rhs.impl.pos);
+            int res = nullableCompareAt<left_nulls, right_nulls>(*left_column, *right_column, impl.getRow(), rhs.impl.getRow());
             if (res)
                 return res;
         }
@@ -278,11 +282,11 @@ private:
     /// Expects !atEnd()
     size_t getEqualLength()
     {
-        size_t pos = impl.pos + 1;
+        size_t pos = impl.getRow() + 1;
         for (; pos < impl.rows; ++pos)
             if (!samePrev(pos))
                 break;
-        return pos - impl.pos;
+        return pos - impl.getRow();
     }
 
     /// Expects lhs_pos > 0
@@ -516,7 +520,7 @@ void MergeJoin::mergeInMemoryRightBlocks()
     pipeline.init(std::move(source));
 
     /// TODO: there should be no split keys by blocks for RIGHT|FULL JOIN
-    pipeline.addTransform(std::make_shared<MergeSortingTransform>(pipeline.getHeader(), right_sort_description, max_rows_in_right_block, 0, 0, 0, nullptr, 0));
+    pipeline.addTransform(std::make_shared<MergeSortingTransform>(pipeline.getHeader(), right_sort_description, max_rows_in_right_block, 0, 0, 0, 0, nullptr, 0));
 
     auto sorted_input = PipelineExecutingBlockInputStream(std::move(pipeline));
 
diff --git a/src/Interpreters/Set.cpp b/src/Interpreters/Set.cpp
index 08e6107590e..dcf7b3b3c24 100644
--- a/src/Interpreters/Set.cpp
+++ b/src/Interpreters/Set.cpp
@@ -23,6 +23,7 @@
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Interpreters/NullableUtils.h>
 #include <Interpreters/sortBlock.h>
+#include <Interpreters/castColumn.h>
 #include <Interpreters/Context.h>
 
 #include <Storages/MergeTree/KeyCondition.h>
@@ -30,7 +31,6 @@
 #include <ext/range.h>
 #include <DataTypes/DataTypeLowCardinality.h>
 
-
 namespace DB
 {
 
@@ -251,11 +251,26 @@ ColumnPtr Set::execute(const Block & block, bool negative) const
 
     /// The constant columns to the left of IN are not supported directly. For this, they first materialize.
     Columns materialized_columns;
+    materialized_columns.reserve(num_key_columns);
 
     for (size_t i = 0; i < num_key_columns; ++i)
     {
-        checkTypesEqual(i, block.safeGetByPosition(i).type);
-        materialized_columns.emplace_back(block.safeGetByPosition(i).column->convertToFullColumnIfConst());
+        ColumnPtr result;
+
+        const auto & column_before_cast = block.safeGetByPosition(i);
+        ColumnWithTypeAndName column_to_cast
+            = {column_before_cast.column->convertToFullColumnIfConst(), column_before_cast.type, column_before_cast.name};
+
+        if (!transform_null_in && data_types[i]->canBeInsideNullable())
+        {
+            result = castColumnAccurateOrNull(column_to_cast, data_types[i]);
+        }
+        else
+        {
+            result = castColumnAccurate(column_to_cast, data_types[i]);
+        }
+
+        materialized_columns.emplace_back() = result;
         key_columns.emplace_back() = materialized_columns.back().get();
     }
 
diff --git a/src/Interpreters/castColumn.cpp b/src/Interpreters/castColumn.cpp
index 249538ec2e0..e8f8bd3bed1 100644
--- a/src/Interpreters/castColumn.cpp
+++ b/src/Interpreters/castColumn.cpp
@@ -1,15 +1,12 @@
-#include <Core/Field.h>
 #include <Interpreters/castColumn.h>
-#include <Interpreters/ExpressionActions.h>
-#include <DataTypes/DataTypeString.h>
-#include <Functions/IFunctionAdaptors.h>
-#include <Functions/FunctionsConversion.h>
 
+#include <Functions/FunctionsConversion.h>
 
 namespace DB
 {
 
-ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+template <CastType cast_type = CastType::nonAccurate>
+static ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
 {
     if (arg.type->equals(*type))
         return arg.column;
@@ -25,10 +22,33 @@ ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type
     };
 
     FunctionOverloadResolverPtr func_builder_cast =
-        std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver::createImpl(false));
+        std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver<cast_type>::createImpl(false));
 
     auto func_cast = func_builder_cast->build(arguments);
-    return func_cast->execute(arguments, type, arg.column->size());
+
+    if constexpr (cast_type == CastType::accurateOrNull)
+    {
+        return func_cast->execute(arguments, makeNullable(type), arg.column->size());
+    }
+    else
+    {
+        return func_cast->execute(arguments, type, arg.column->size());
+    }
+}
+
+ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+{
+    return castColumn<CastType::nonAccurate>(arg, type);
+}
+
+ColumnPtr castColumnAccurate(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+{
+    return castColumn<CastType::accurate>(arg, type);
+}
+
+ColumnPtr castColumnAccurateOrNull(const ColumnWithTypeAndName & arg, const DataTypePtr & type)
+{
+    return castColumn<CastType::accurateOrNull>(arg, type);
 }
 
 }
diff --git a/src/Interpreters/castColumn.h b/src/Interpreters/castColumn.h
index 44c7f071c26..fcbea0f4646 100644
--- a/src/Interpreters/castColumn.h
+++ b/src/Interpreters/castColumn.h
@@ -2,8 +2,11 @@
 
 #include <Core/ColumnWithTypeAndName.h>
 
-
 namespace DB
 {
+
 ColumnPtr castColumn(const ColumnWithTypeAndName & arg, const DataTypePtr & type);
+ColumnPtr castColumnAccurate(const ColumnWithTypeAndName & arg, const DataTypePtr & type);
+ColumnPtr castColumnAccurateOrNull(const ColumnWithTypeAndName & arg, const DataTypePtr & type);
+
 }
diff --git a/src/Interpreters/convertFieldToType.cpp b/src/Interpreters/convertFieldToType.cpp
index 18a3eeecb46..73bf493fa65 100644
--- a/src/Interpreters/convertFieldToType.cpp
+++ b/src/Interpreters/convertFieldToType.cpp
@@ -5,6 +5,7 @@
 
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypesDecimal.h>
 #include <DataTypes/DataTypeString.h>
diff --git a/src/Interpreters/loadMetadata.cpp b/src/Interpreters/loadMetadata.cpp
index bb476999f8c..71d3c7e6e5b 100644
--- a/src/Interpreters/loadMetadata.cpp
+++ b/src/Interpreters/loadMetadata.cpp
@@ -108,6 +108,7 @@ void loadMetadata(Context & context, const String & default_database_name)
 
         if (!it->isDirectory())
         {
+            /// TODO: DETACH DATABASE PERMANENTLY ?
             if (endsWith(it.name(), ".sql"))
             {
                 String db_name = it.name().substr(0, it.name().size() - 4);
diff --git a/src/Interpreters/tests/gtest_cycle_aliases.cpp b/src/Interpreters/tests/gtest_cycle_aliases.cpp
index 593db93de3e..56e23c6a497 100644
--- a/src/Interpreters/tests/gtest_cycle_aliases.cpp
+++ b/src/Interpreters/tests/gtest_cycle_aliases.cpp
@@ -5,6 +5,7 @@
 #include <Parsers/queryToString.h>
 #include <Parsers/ExpressionListParsers.h>
 #include <Parsers/parseQuery.h>
+#include <Core/Settings.h>
 
 using namespace DB;
 
diff --git a/src/Interpreters/tests/internal_iotop.cpp b/src/Interpreters/tests/internal_iotop.cpp
index 3d17ca32f3f..6025a46e9b7 100644
--- a/src/Interpreters/tests/internal_iotop.cpp
+++ b/src/Interpreters/tests/internal_iotop.cpp
@@ -5,6 +5,7 @@
 #include <Common/TaskStatsInfoGetter.h>
 #include <Poco/File.h>
 #include <Common/Stopwatch.h>
+#include <common/getPageSize.h>
 #include <common/getThreadId.h>
 #include <IO/WriteBufferFromString.h>
 #include <linux/taskstats.h>
@@ -61,8 +62,9 @@ static void do_io(size_t id)
     std::string path_dst = "test_out_" + std::to_string(id);
 
     {
+        size_t page_size = static_cast<size_t>(::getPageSize());
         ReadBufferFromFile rb("/dev/urandom");
-        WriteBufferFromFile wb(path_dst, DBMS_DEFAULT_BUFFER_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, 4096);
+        WriteBufferFromFile wb(path_dst, DBMS_DEFAULT_BUFFER_SIZE, O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666, nullptr, page_size);
         copyData(rb, wb, copy_size);
         wb.close();
     }
diff --git a/src/Parsers/ASTDropQuery.cpp b/src/Parsers/ASTDropQuery.cpp
index b09b588ca6e..6ea6c81c3be 100644
--- a/src/Parsers/ASTDropQuery.cpp
+++ b/src/Parsers/ASTDropQuery.cpp
@@ -67,6 +67,9 @@ void ASTDropQuery::formatQueryImpl(const FormatSettings & settings, FormatState
 
     formatOnCluster(settings);
 
+    if (permanently)
+        settings.ostr << " PERMANENTLY";
+
     if (no_delay)
         settings.ostr << (settings.hilite ? hilite_keyword : "") << " NO DELAY" << (settings.hilite ? hilite_none : "");
 }
diff --git a/src/Parsers/ASTDropQuery.h b/src/Parsers/ASTDropQuery.h
index 5096975c219..b612618aaec 100644
--- a/src/Parsers/ASTDropQuery.h
+++ b/src/Parsers/ASTDropQuery.h
@@ -33,6 +33,9 @@ public:
 
     bool no_delay{false};
 
+    // We detach the object permanently, so it will not be reattached back during server restart.
+    bool permanently{false};
+
     /** Get the text that identifies this element. */
     String getID(char) const override;
     ASTPtr clone() const override;
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index 4cca8ad0807..c86a8e5c139 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -3,6 +3,7 @@
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ASTWithAlias.h>
 #include <Parsers/ASTSubquery.h>
+#include <Parsers/ASTExpressionList.h>
 #include <IO/WriteHelpers.h>
 #include <IO/WriteBufferFromString.h>
 #include <Common/SipHash.h>
@@ -395,6 +396,19 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
             settings.ostr << (settings.hilite ? hilite_operator : "") << ')' << (settings.hilite ? hilite_none : "");
             written = true;
         }
+
+        if (!written && 0 == strcmp(name.c_str(), "map"))
+        {
+            settings.ostr << (settings.hilite ? hilite_operator : "") << '{' << (settings.hilite ? hilite_none : "");
+            for (size_t i = 0; i < arguments->children.size(); ++i)
+            {
+                if (i != 0)
+                    settings.ostr << ", ";
+                arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
+            }
+            settings.ostr << (settings.hilite ? hilite_operator : "") << '}' << (settings.hilite ? hilite_none : "");
+            written = true;
+        }
     }
 
     if (!written)
diff --git a/src/Parsers/ASTShowTablesQuery.cpp b/src/Parsers/ASTShowTablesQuery.cpp
index cd83bae06d9..61d68c4a273 100644
--- a/src/Parsers/ASTShowTablesQuery.cpp
+++ b/src/Parsers/ASTShowTablesQuery.cpp
@@ -55,6 +55,12 @@ void ASTShowTablesQuery::formatQueryImpl(const FormatSettings & settings, Format
         settings.ostr << (settings.hilite ? hilite_keyword : "") << "SHOW CLUSTER" << (settings.hilite ? hilite_none : "");
         settings.ostr << " " << backQuoteIfNeed(cluster_str);
     }
+    else if (m_settings)
+    {
+        settings.ostr << (settings.hilite ? hilite_keyword : "") << "SHOW " << (changed ? "CHANGED " : "") << "SETTINGS" <<
+            (settings.hilite ? hilite_none : "");
+        formatLike(settings);
+    }
     else
     {
         settings.ostr << (settings.hilite ? hilite_keyword : "") << "SHOW " << (temporary ? "TEMPORARY " : "") <<
diff --git a/src/Parsers/ASTShowTablesQuery.h b/src/Parsers/ASTShowTablesQuery.h
index 43976e8a958..57383dff66f 100644
--- a/src/Parsers/ASTShowTablesQuery.h
+++ b/src/Parsers/ASTShowTablesQuery.h
@@ -18,6 +18,8 @@ public:
     bool clusters{false};
     bool cluster{false};
     bool dictionaries{false};
+    bool m_settings{false};
+    bool changed{false};
     bool temporary{false};
 
     String cluster_str;
diff --git a/src/Parsers/ExpressionElementParsers.cpp b/src/Parsers/ExpressionElementParsers.cpp
index 6fbe9abc96e..8706aecb7ac 100644
--- a/src/Parsers/ExpressionElementParsers.cpp
+++ b/src/Parsers/ExpressionElementParsers.cpp
@@ -1110,6 +1110,10 @@ bool ParserCollectionOfLiterals<Collection>::parseImpl(Pos & pos, ASTPtr & node,
             {
                 ++pos;
             }
+            else if (pos->type == TokenType::Colon && std::is_same_v<Collection, Map> && arr.size() % 2 == 1)
+            {
+                ++pos;
+            }
             else
             {
                 expected.add(pos, "comma or closing bracket");
@@ -1130,6 +1134,7 @@ bool ParserCollectionOfLiterals<Collection>::parseImpl(Pos & pos, ASTPtr & node,
 
 template bool ParserCollectionOfLiterals<Array>::parseImpl(Pos & pos, ASTPtr & node, Expected & expected);
 template bool ParserCollectionOfLiterals<Tuple>::parseImpl(Pos & pos, ASTPtr & node, Expected & expected);
+template bool ParserCollectionOfLiterals<Map>::parseImpl(Pos & pos, ASTPtr & node, Expected & expected);
 
 bool ParserLiteral::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
 {
@@ -1564,6 +1569,7 @@ bool ParserExpressionElement::parseImpl(Pos & pos, ASTPtr & node, Expected & exp
 {
     return ParserSubquery().parse(pos, node, expected)
         || ParserTupleOfLiterals().parse(pos, node, expected)
+        || ParserMapOfLiterals().parse(pos, node, expected)
         || ParserParenthesisExpression().parse(pos, node, expected)
         || ParserArrayOfLiterals().parse(pos, node, expected)
         || ParserArray().parse(pos, node, expected)
diff --git a/src/Parsers/ExpressionElementParsers.h b/src/Parsers/ExpressionElementParsers.h
index 4713a7ea8a6..4033a6e4f61 100644
--- a/src/Parsers/ExpressionElementParsers.h
+++ b/src/Parsers/ExpressionElementParsers.h
@@ -270,6 +270,18 @@ protected:
     }
 };
 
+class ParserMapOfLiterals : public IParserBase
+{
+public:
+    ParserCollectionOfLiterals<Map> map_parser{TokenType::OpeningCurlyBrace, TokenType::ClosingCurlyBrace};
+protected:
+    const char * getName() const override { return "map"; }
+    bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override
+    {
+        return map_parser.parse(pos, node, expected);
+    }
+};
+
 class ParserArrayOfLiterals : public IParserBase
 {
 public:
diff --git a/src/Parsers/ExpressionListParsers.cpp b/src/Parsers/ExpressionListParsers.cpp
index dda1417ca97..f50cf71f54d 100644
--- a/src/Parsers/ExpressionListParsers.cpp
+++ b/src/Parsers/ExpressionListParsers.cpp
@@ -100,6 +100,7 @@ bool ParserList::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
     auto list = std::make_shared<ASTExpressionList>(result_separator);
     list->children = std::move(elements);
     node = list;
+
     return true;
 }
 
diff --git a/src/Parsers/IParser.h b/src/Parsers/IParser.h
index 05ceb8c900b..7dc31e4c1eb 100644
--- a/src/Parsers/IParser.h
+++ b/src/Parsers/IParser.h
@@ -5,7 +5,6 @@
 
 #include <Core/Defines.h>
 #include <common/types.h>
-#include <Core/Settings.h>
 #include <IO/WriteHelpers.h>
 #include <Parsers/IAST.h>
 #include <Parsers/TokenIterator.h>
diff --git a/src/Parsers/ParserCreateQuery.cpp b/src/Parsers/ParserCreateQuery.cpp
index b08646bcf9d..5fe6a4eee80 100644
--- a/src/Parsers/ParserCreateQuery.cpp
+++ b/src/Parsers/ParserCreateQuery.cpp
@@ -449,12 +449,22 @@ bool ParserCreateTableQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
         if (!s_rparen.ignore(pos, expected))
             return false;
 
-        if (!storage_p.parse(pos, storage, expected) && !is_temporary)
+        auto storage_parse_result = storage_p.parse(pos, storage, expected);
+
+        if (storage_parse_result && s_as.ignore(pos, expected))
+        {
+            if (!select_p.parse(pos, select, expected))
+                return false;
+        }
+
+        if (!storage_parse_result && !is_temporary)
         {
             if (!s_as.ignore(pos, expected))
                 return false;
             if (!table_function_p.parse(pos, as_table_function, expected))
+            {
                 return false;
+            }
         }
     }
     else
diff --git a/src/Parsers/ParserDropQuery.cpp b/src/Parsers/ParserDropQuery.cpp
index e09642e7b84..5400f33fbd9 100644
--- a/src/Parsers/ParserDropQuery.cpp
+++ b/src/Parsers/ParserDropQuery.cpp
@@ -11,7 +11,7 @@ namespace DB
 namespace
 {
 
-bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool optional_table_keyword = false)
+bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, const ASTDropQuery::Kind kind)
 {
     ParserKeyword s_temporary("TEMPORARY");
     ParserKeyword s_table("TABLE");
@@ -21,6 +21,7 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
     ParserToken s_dot(TokenType::Dot);
     ParserKeyword s_if_exists("IF EXISTS");
     ParserIdentifier name_p;
+    ParserKeyword s_permanently("PERMANENTLY");
     ParserKeyword s_no_delay("NO DELAY");
     ParserKeyword s_sync("SYNC");
 
@@ -32,6 +33,7 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
     bool is_dictionary = false;
     bool is_view = false;
     bool no_delay = false;
+    bool permanently = false;
 
     if (s_database.ignore(pos, expected))
     {
@@ -40,15 +42,6 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
 
         if (!name_p.parse(pos, database, expected))
             return false;
-
-        if (ParserKeyword{"ON"}.ignore(pos, expected))
-        {
-            if (!ASTQueryWithOnCluster::parse(pos, cluster_str, expected))
-                return false;
-        }
-
-        if (s_no_delay.ignore(pos, expected) || s_sync.ignore(pos, expected))
-            no_delay = true;
     }
     else
     {
@@ -59,7 +52,8 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
         else if (s_temporary.ignore(pos, expected))
             temporary = true;
 
-        if (!is_view && !is_dictionary && (!s_table.ignore(pos, expected) && !optional_table_keyword))
+        /// for TRUNCATE queries TABLE keyword is assumed as default and can be skipped
+        if (!is_view && !is_dictionary && (!s_table.ignore(pos, expected) && kind != ASTDropQuery::Kind::Truncate))
         {
             return false;
         }
@@ -76,26 +70,32 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
             if (!name_p.parse(pos, table, expected))
                 return false;
         }
-
-        if (ParserKeyword{"ON"}.ignore(pos, expected))
-        {
-            if (!ASTQueryWithOnCluster::parse(pos, cluster_str, expected))
-                return false;
-        }
-
-        if (s_no_delay.ignore(pos, expected) || s_sync.ignore(pos, expected))
-            no_delay = true;
     }
 
+    /// common for tables / dictionaries / databases
+    if (ParserKeyword{"ON"}.ignore(pos, expected))
+    {
+        if (!ASTQueryWithOnCluster::parse(pos, cluster_str, expected))
+            return false;
+    }
+
+    if (kind == ASTDropQuery::Kind::Detach && s_permanently.ignore(pos, expected))
+        permanently = true;
+
+    /// actually for TRUNCATE NO DELAY / SYNC means nothing
+    if (s_no_delay.ignore(pos, expected) || s_sync.ignore(pos, expected))
+        no_delay = true;
+
     auto query = std::make_shared<ASTDropQuery>();
     node = query;
 
-    query->kind = ASTDropQuery::Kind::Drop;
+    query->kind = kind;
     query->if_exists = if_exists;
     query->temporary = temporary;
     query->is_dictionary = is_dictionary;
     query->is_view = is_view;
     query->no_delay = no_delay;
+    query->permanently = permanently;
 
     tryGetIdentifierNameInto(database, query->database);
     tryGetIdentifierNameInto(table, query->table);
@@ -105,28 +105,6 @@ bool parseDropQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected, bool
     return true;
 }
 
-bool parseDetachQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected)
-{
-    if (parseDropQuery(pos, node, expected))
-    {
-        auto * drop_query = node->as<ASTDropQuery>();
-        drop_query->kind = ASTDropQuery::Kind::Detach;
-        return true;
-    }
-    return false;
-}
-
-bool parseTruncateQuery(IParser::Pos & pos, ASTPtr & node, Expected & expected)
-{
-    if (parseDropQuery(pos, node, expected, true))
-    {
-        auto * drop_query = node->as<ASTDropQuery>();
-        drop_query->kind = ASTDropQuery::Kind::Truncate;
-        return true;
-    }
-    return false;
-}
-
 }
 
 bool ParserDropQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
@@ -136,11 +114,11 @@ bool ParserDropQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
     ParserKeyword s_truncate("TRUNCATE");
 
     if (s_drop.ignore(pos, expected))
-        return parseDropQuery(pos, node, expected);
+        return parseDropQuery(pos, node, expected, ASTDropQuery::Kind::Drop);
     else if (s_detach.ignore(pos, expected))
-        return parseDetachQuery(pos, node, expected);
+        return parseDropQuery(pos, node, expected, ASTDropQuery::Kind::Detach);
     else if (s_truncate.ignore(pos, expected))
-        return parseTruncateQuery(pos, node, expected);
+        return parseDropQuery(pos, node, expected, ASTDropQuery::Kind::Truncate);
     else
         return false;
 }
diff --git a/src/Parsers/ParserDropQuery.h b/src/Parsers/ParserDropQuery.h
index 069b9c34ddd..39ff5b7c4fd 100644
--- a/src/Parsers/ParserDropQuery.h
+++ b/src/Parsers/ParserDropQuery.h
@@ -8,7 +8,7 @@ namespace DB
 {
 
 /** Query like this:
-  * DROP|DETACH|TRUNCATE TABLE [IF EXISTS] [db.]name
+  * DROP|DETACH|TRUNCATE TABLE [IF EXISTS] [db.]name [PERMANENTLY]
   *
   * Or:
   * DROP DATABASE [IF EXISTS] db
diff --git a/src/Parsers/ParserExplainQuery.cpp b/src/Parsers/ParserExplainQuery.cpp
index fe48addd0cd..7a7abba94a9 100644
--- a/src/Parsers/ParserExplainQuery.cpp
+++ b/src/Parsers/ParserExplainQuery.cpp
@@ -5,6 +5,7 @@
 #include <Parsers/ParserCreateQuery.h>
 #include <Parsers/ParserSelectWithUnionQuery.h>
 #include <Parsers/ParserSetQuery.h>
+#include <Parsers/ParserQuery.h>
 
 namespace DB
 {
@@ -51,7 +52,13 @@ bool ParserExplainQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected
     ParserCreateTableQuery create_p;
     ParserSelectWithUnionQuery select_p;
     ASTPtr query;
-    if (select_p.parse(pos, query, expected) ||
+    if (kind == ASTExplainQuery::ExplainKind::ParsedAST)
+    {
+        ParserQuery p(end);
+        if (p.parse(pos, query, expected))
+            explain_query->setExplainedQuery(std::move(query));
+    }
+    else if (select_p.parse(pos, query, expected) ||
         create_p.parse(pos, query, expected))
         explain_query->setExplainedQuery(std::move(query));
     else
diff --git a/src/Parsers/ParserExplainQuery.h b/src/Parsers/ParserExplainQuery.h
index b58597b9248..a1865e30239 100644
--- a/src/Parsers/ParserExplainQuery.h
+++ b/src/Parsers/ParserExplainQuery.h
@@ -9,8 +9,12 @@ namespace DB
 class ParserExplainQuery : public IParserBase
 {
 protected:
+    const char * end;
+
     const char * getName() const override { return "EXPLAIN"; }
     bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override;
+public:
+    ParserExplainQuery(const char* end_) : end(end_) {}
 };
 
 }
diff --git a/src/Parsers/ParserQuery.cpp b/src/Parsers/ParserQuery.cpp
index 89ddd1797c0..4550bdc8a75 100644
--- a/src/Parsers/ParserQuery.cpp
+++ b/src/Parsers/ParserQuery.cpp
@@ -26,7 +26,7 @@ namespace DB
 
 bool ParserQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
 {
-    ParserQueryWithOutput query_with_output_p;
+    ParserQueryWithOutput query_with_output_p(end);
     ParserInsertQuery insert_p(end);
     ParserUseQuery use_p;
     ParserSetQuery set_p;
diff --git a/src/Parsers/ParserQueryWithOutput.cpp b/src/Parsers/ParserQueryWithOutput.cpp
index 76a3e8b95f8..d5aa1e47533 100644
--- a/src/Parsers/ParserQueryWithOutput.cpp
+++ b/src/Parsers/ParserQueryWithOutput.cpp
@@ -48,7 +48,7 @@ bool ParserQueryWithOutput::parseImpl(Pos & pos, ASTPtr & node, Expected & expec
     ParserShowCreateAccessEntityQuery show_create_access_entity_p;
     ParserShowGrantsQuery show_grants_p;
     ParserShowPrivilegesQuery show_privileges_p;
-    ParserExplainQuery explain_p;
+    ParserExplainQuery explain_p(end);
 
     ASTPtr query;
 
diff --git a/src/Parsers/ParserQueryWithOutput.h b/src/Parsers/ParserQueryWithOutput.h
index 110cf923147..854d5a74ffd 100644
--- a/src/Parsers/ParserQueryWithOutput.h
+++ b/src/Parsers/ParserQueryWithOutput.h
@@ -11,8 +11,11 @@ namespace DB
 class ParserQueryWithOutput : public IParserBase
 {
 protected:
+    const char * end;
     const char * getName() const override { return "Query with output"; }
     bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override;
+public:
+    ParserQueryWithOutput(const char * end_) : end(end_) {}
 };
 
 }
diff --git a/src/Parsers/ParserSetQuery.cpp b/src/Parsers/ParserSetQuery.cpp
index 328d01f2f81..aac3a191a10 100644
--- a/src/Parsers/ParserSetQuery.cpp
+++ b/src/Parsers/ParserSetQuery.cpp
@@ -6,6 +6,7 @@
 #include <Parsers/ParserSetQuery.h>
 
 #include <Common/typeid_cast.h>
+#include <Common/SettingsChanges.h>
 
 
 namespace DB
diff --git a/src/Parsers/ParserSetQuery.h b/src/Parsers/ParserSetQuery.h
index 59a6109ea48..0bc1cec3093 100644
--- a/src/Parsers/ParserSetQuery.h
+++ b/src/Parsers/ParserSetQuery.h
@@ -7,6 +7,8 @@
 namespace DB
 {
 
+struct SettingChange;
+
 /** Query like this:
   * SET name1 = value1, name2 = value2, ...
   */
diff --git a/src/Parsers/ParserShowTablesQuery.cpp b/src/Parsers/ParserShowTablesQuery.cpp
index 4586e10a8a3..6022a31034a 100644
--- a/src/Parsers/ParserShowTablesQuery.cpp
+++ b/src/Parsers/ParserShowTablesQuery.cpp
@@ -24,6 +24,8 @@ bool ParserShowTablesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expec
     ParserKeyword s_clusters("CLUSTERS");
     ParserKeyword s_cluster("CLUSTER");
     ParserKeyword s_dictionaries("DICTIONARIES");
+    ParserKeyword s_settings("SETTINGS");
+    ParserKeyword s_changed("CHANGED");
     ParserKeyword s_from("FROM");
     ParserKeyword s_in("IN");
     ParserKeyword s_not("NOT");
@@ -99,6 +101,29 @@ bool ParserShowTablesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expec
 
         query->cluster_str = std::move(cluster_str);
     }
+    else if (bool changed = s_changed.ignore(pos); changed || s_settings.ignore(pos))
+    {
+        query->m_settings = true;
+
+        if (changed)
+        {
+            query->changed = true;
+            if (!s_settings.ignore(pos, expected))
+                return false;
+        }
+
+        /// Not expected due to "SHOW SETTINGS PROFILES"
+        if (bool insensitive = s_ilike.ignore(pos); insensitive || s_like.ignore(pos))
+        {
+            if (insensitive)
+                query->case_insensitive_like = true;
+
+            if (!like_p.parse(pos, like, expected))
+                return false;
+        }
+        else
+            return false;
+    }
     else
     {
         if (s_temporary.ignore(pos))
diff --git a/src/Parsers/tests/select_parser.cpp b/src/Parsers/tests/select_parser.cpp
index 7c18563659d..15295170c6b 100644
--- a/src/Parsers/tests/select_parser.cpp
+++ b/src/Parsers/tests/select_parser.cpp
@@ -22,7 +22,7 @@ try
         " INTO OUTFILE 'test.out'"
         " FORMAT TabSeparated";
 
-    ParserQueryWithOutput parser;
+    ParserQueryWithOutput parser(input.data() + input.size());
     ASTPtr ast = parseQuery(parser, input.data(), input.data() + input.size(), "", 0, 0);
 
     std::cout << "Success." << std::endl;
diff --git a/src/Parsers/tests/select_parser_fuzzer.cpp b/src/Parsers/tests/select_parser_fuzzer.cpp
index fc4670333fe..4848b285c07 100644
--- a/src/Parsers/tests/select_parser_fuzzer.cpp
+++ b/src/Parsers/tests/select_parser_fuzzer.cpp
@@ -11,7 +11,7 @@ try
 {
     std::string input = std::string(reinterpret_cast<const char*>(data), size);
 
-    DB::ParserQueryWithOutput parser;
+    DB::ParserQueryWithOutput parser(input.data() + input.size());
     DB::ASTPtr ast = parseQuery(parser, input.data(), input.data() + input.size(), "", 0, 0);
 
     DB::formatAST(*ast, std::cerr);
diff --git a/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp b/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp
index 68854e53a02..d7a65c2f15d 100644
--- a/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp
+++ b/src/Processors/Formats/Impl/ConstantExpressionTemplate.cpp
@@ -1,11 +1,13 @@
 #include <Columns/ColumnConst.h>
 #include <Columns/ColumnTuple.h>
+#include <Columns/ColumnMap.h>
 #include <Columns/ColumnsNumber.h>
 #include <DataTypes/DataTypesNumber.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeString.h>
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/DataTypeMap.h>
 #include <DataTypes/FieldToDataType.h>
 #include <Processors/Formats/IRowInputFormat.h>
 #include <Functions/FunctionFactory.h>
@@ -46,6 +48,7 @@ struct SpecialParserType
     bool is_nullable = false;
     bool is_array = false;
     bool is_tuple = false;
+    bool is_map = false;
     /// Type and nullability
     std::vector<std::pair<Field::Types::Which, bool>> nested_types;
 
@@ -119,6 +122,10 @@ static void fillLiteralInfo(DataTypes & nested_types, LiteralInfo & info)
         {
             field_type = Field::Types::Tuple;
         }
+        else if (type_info.isMap())
+        {
+            field_type = Field::Types::Map;
+        }
         else
             throw Exception("Unexpected literal type inside Array: " + nested_type->getName() + ". It's a bug",
                             ErrorCodes::LOGICAL_ERROR);
@@ -196,6 +203,13 @@ private:
                 if (not_null == array.end())
                     return true;
             }
+            else if (literal->value.getType() == Field::Types::Map)
+            {
+                const Map & map = literal->value.get<Map>();
+                if (map.size() % 2)
+                    return false;
+            }
+
             String column_name = "_dummy_" + std::to_string(replaced_literals.size());
             replaced_literals.emplace_back(literal, column_name, force_nullable);
             setDataType(replaced_literals.back());
@@ -243,6 +257,15 @@ private:
             fillLiteralInfo(nested_types, info);
             info.type = std::make_shared<DataTypeTuple>(nested_types);
         }
+        else if (field_type == Field::Types::Map)
+        {
+            info.special_parser.is_map = true;
+
+            info.type = applyVisitor(FieldToDataType(), info.literal->value);
+            auto nested_types = assert_cast<const DataTypeMap &>(*info.type).getKeyValueTypes();
+            fillLiteralInfo(nested_types, info);
+            info.type = std::make_shared<DataTypeMap>(nested_types);
+        }
         else
             throw Exception(String("Unexpected literal type ") + info.literal->value.getTypeName() + ". It's a bug",
                             ErrorCodes::LOGICAL_ERROR);
@@ -453,17 +476,19 @@ bool ConstantExpressionTemplate::parseLiteralAndAssertType(ReadBuffer & istr, co
     /// If literal does not fit entirely in the buffer, parsing error will happen.
     /// However, it's possible to deduce new template (or use template from cache) after error like it was template mismatch.
 
-    if (type_info.is_array || type_info.is_tuple)
+    if (type_info.is_array || type_info.is_tuple || type_info.is_map)
     {
         /// TODO faster way to check types without using Parsers
         ParserArrayOfLiterals parser_array;
         ParserTupleOfLiterals parser_tuple;
+        ParserMapOfLiterals parser_map;
 
         Tokens tokens_number(istr.position(), istr.buffer().end());
         IParser::Pos iterator(tokens_number, settings.max_parser_depth);
         Expected expected;
         ASTPtr ast;
-        if (!parser_array.parse(iterator, ast, expected) && !parser_tuple.parse(iterator, ast, expected))
+        if (!parser_array.parse(iterator, ast, expected) && !parser_tuple.parse(iterator, ast, expected)
+            && !parser_map.parse(iterator, ast, expected))
             return false;
 
         istr.position() = const_cast<char *>(iterator->begin);
@@ -474,8 +499,10 @@ bool ConstantExpressionTemplate::parseLiteralAndAssertType(ReadBuffer & istr, co
         DataTypes nested_types;
         if (type_info.is_array)
             nested_types = { assert_cast<const DataTypeArray &>(*collection_type).getNestedType() };
-        else
+        else if (type_info.is_tuple)
             nested_types = assert_cast<const DataTypeTuple &>(*collection_type).getElements();
+        else
+            nested_types = assert_cast<const DataTypeMap &>(*collection_type).getKeyValueTypes();
 
         for (size_t i = 0; i < nested_types.size(); ++i)
         {
diff --git a/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp
index 11d68372b10..6b2ee1c8039 100644
--- a/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/AggregatingSortedAlgorithm.cpp
@@ -257,12 +257,12 @@ void AggregatingSortedAlgorithm::AggregatingMergedData::addRow(SortCursor & curs
         throw Exception("Can't add a row to the group because it was not started.", ErrorCodes::LOGICAL_ERROR);
 
     for (auto & desc : def.columns_to_aggregate)
-        desc.column->insertMergeFrom(*cursor->all_columns[desc.column_number], cursor->pos);
+        desc.column->insertMergeFrom(*cursor->all_columns[desc.column_number], cursor->getRow());
 
     for (auto & desc : def.columns_to_simple_aggregate)
     {
         auto & col = cursor->all_columns[desc.column_number];
-        desc.add_function(desc.function.get(), desc.state.data(), &col, cursor->pos, arena.get());
+        desc.add_function(desc.function.get(), desc.state.data(), &col, cursor->getRow(), arena.get());
     }
 }
 
@@ -352,7 +352,7 @@ IMergingAlgorithm::Status AggregatingSortedAlgorithm::merge()
                 return Status(merged_data.pull());
             }
 
-            merged_data.startGroup(current->all_columns, current->pos);
+            merged_data.startGroup(current->all_columns, current->getRow());
         }
 
         merged_data.addRow(current);
diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp
index 8f6888b9a7a..ccb66259e2e 100644
--- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp
@@ -26,9 +26,9 @@ CollapsingSortedAlgorithm::CollapsingSortedAlgorithm(
     const String & sign_column,
     bool only_positive_sign_,
     size_t max_block_size,
+    Poco::Logger * log_,
     WriteBuffer * out_row_sources_buf_,
-    bool use_average_block_sizes,
-    Poco::Logger * log_)
+    bool use_average_block_sizes)
     : IMergingAlgorithmWithSharedChunks(num_inputs, std::move(description_), out_row_sources_buf_, max_row_refs)
     , merged_data(header.cloneEmptyColumns(), use_average_block_sizes, max_block_size)
     , sign_column_number(header.getPositionByName(sign_column))
@@ -123,7 +123,7 @@ IMergingAlgorithm::Status CollapsingSortedAlgorithm::merge()
             return Status(current.impl->order);
         }
 
-        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->pos];
+        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->getRow()];
 
         RowRef current_row;
         setRowRef(current_row, current);
diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h
index d95fac2f02b..028715f715b 100644
--- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h
+++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h
@@ -33,9 +33,9 @@ public:
         const String & sign_column,
         bool only_positive_sign_, /// For select final. Skip rows with sum(sign) < 0.
         size_t max_block_size,
-        WriteBuffer * out_row_sources_buf_,
-        bool use_average_block_sizes,
-        Poco::Logger * log_);
+        Poco::Logger * log_,
+        WriteBuffer * out_row_sources_buf_ = nullptr,
+        bool use_average_block_sizes = false);
 
     Status merge() override;
 
diff --git a/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp
index 0e704e5a05b..df10fb26d40 100644
--- a/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.cpp
@@ -164,12 +164,12 @@ IMergingAlgorithm::Status GraphiteRollupSortedAlgorithm::merge()
             return Status(current.impl->order);
         }
 
-        StringRef next_path = current->all_columns[columns_definition.path_column_num]->getDataAt(current->pos);
+        StringRef next_path = current->all_columns[columns_definition.path_column_num]->getDataAt(current->getRow());
         bool new_path = is_first || next_path != current_group_path;
 
         is_first = false;
 
-        time_t next_row_time = current->all_columns[columns_definition.time_column_num]->getUInt(current->pos);
+        time_t next_row_time = current->all_columns[columns_definition.time_column_num]->getUInt(current->getRow());
         /// Is new key before rounding.
         bool is_new_key = new_path || next_row_time != current_time;
 
@@ -227,7 +227,7 @@ IMergingAlgorithm::Status GraphiteRollupSortedAlgorithm::merge()
         /// and for rows with same maximum version - only last row.
         if (is_new_key
             || current->all_columns[columns_definition.version_column_num]->compareAt(
-                current->pos, current_subgroup_newest_row.row_num,
+                current->getRow(), current_subgroup_newest_row.row_num,
                 *(*current_subgroup_newest_row.all_columns)[columns_definition.version_column_num],
                 /* nan_direction_hint = */ 1) >= 0)
         {
@@ -263,7 +263,7 @@ IMergingAlgorithm::Status GraphiteRollupSortedAlgorithm::merge()
 
 void GraphiteRollupSortedAlgorithm::startNextGroup(SortCursor & cursor, Graphite::RollupRule next_rule)
 {
-    merged_data.startNextGroup(cursor->all_columns, cursor->pos, next_rule, columns_definition);
+    merged_data.startNextGroup(cursor->all_columns, cursor->getRow(), next_rule, columns_definition);
 }
 
 void GraphiteRollupSortedAlgorithm::finishCurrentGroup()
diff --git a/src/Processors/Merges/Algorithms/IMergingAlgorithm.h b/src/Processors/Merges/Algorithms/IMergingAlgorithm.h
index f86be2a7d1b..5c8d18875e7 100644
--- a/src/Processors/Merges/Algorithms/IMergingAlgorithm.h
+++ b/src/Processors/Merges/Algorithms/IMergingAlgorithm.h
@@ -29,6 +29,8 @@ public:
         /// between different algorithm objects in parallel FINAL.
         bool skip_last_row = false;
 
+        IColumn::Permutation * permutation = nullptr;
+
         void swap(Input & other)
         {
             chunk.swap(other.chunk);
diff --git a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp
index 0b13d689636..e4c60d7609c 100644
--- a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp
+++ b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithDelayedChunk.cpp
@@ -22,7 +22,7 @@ void IMergingAlgorithmWithDelayedChunk::initializeQueue(Inputs inputs)
         if (!current_inputs[source_num].chunk)
             continue;
 
-        cursors[source_num] = SortCursorImpl(current_inputs[source_num].chunk.getColumns(), description, source_num);
+        cursors[source_num] = SortCursorImpl(current_inputs[source_num].chunk.getColumns(), description, source_num, current_inputs[source_num].permutation);
     }
 
     queue = SortingHeap<SortCursor>(cursors);
@@ -37,7 +37,7 @@ void IMergingAlgorithmWithDelayedChunk::updateCursor(Input & input, size_t sourc
     last_chunk_sort_columns = std::move(cursors[source_num].sort_columns);
 
     current_input.swap(input);
-    cursors[source_num].reset(current_input.chunk.getColumns(), {});
+    cursors[source_num].reset(current_input.chunk.getColumns(), {}, current_input.permutation);
 
     queue.push(cursors[source_num]);
 }
diff --git a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp
index 39abe5c0ec7..97abffdc167 100644
--- a/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp
+++ b/src/Processors/Merges/Algorithms/IMergingAlgorithmWithSharedChunks.cpp
@@ -39,7 +39,7 @@ void IMergingAlgorithmWithSharedChunks::initialize(Inputs inputs)
 
         source.skip_last_row = inputs[source_num].skip_last_row;
         source.chunk = chunk_allocator.alloc(inputs[source_num].chunk);
-        cursors[source_num] = SortCursorImpl(source.chunk->getColumns(), description, source_num);
+        cursors[source_num] = SortCursorImpl(source.chunk->getColumns(), description, source_num, inputs[source_num].permutation);
 
         source.chunk->all_columns = cursors[source_num].all_columns;
         source.chunk->sort_columns = cursors[source_num].sort_columns;
@@ -55,7 +55,7 @@ void IMergingAlgorithmWithSharedChunks::consume(Input & input, size_t source_num
     auto & source = sources[source_num];
     source.skip_last_row = input.skip_last_row;
     source.chunk = chunk_allocator.alloc(input.chunk);
-    cursors[source_num].reset(source.chunk->getColumns(), {});
+    cursors[source_num].reset(source.chunk->getColumns(), {}, input.permutation);
 
     source.chunk->all_columns = cursors[source_num].all_columns;
     source.chunk->sort_columns = cursors[source_num].sort_columns;
diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp
index ee13ef70203..511bd9dd74f 100644
--- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp
@@ -139,7 +139,7 @@ IMergingAlgorithm::Status MergingSortedAlgorithm::mergeImpl(TSortingHeap & queue
 
         //std::cerr << "total_merged_rows: " << total_merged_rows << ", merged_rows: " << merged_rows << "\n";
         //std::cerr << "Inserting row\n";
-        merged_data.insertRow(current->all_columns, current->pos, current->rows);
+        merged_data.insertRow(current->all_columns, current->getRow(), current->rows);
 
         if (out_row_sources_buf)
         {
diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h
index 531b2636747..63dced26dd4 100644
--- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h
+++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h
@@ -18,9 +18,9 @@ public:
         size_t num_inputs,
         SortDescription description_,
         size_t max_block_size,
-        UInt64 limit_,
-        WriteBuffer * out_row_sources_buf_,
-        bool use_average_block_sizes);
+        UInt64 limit_ = 0,
+        WriteBuffer * out_row_sources_buf_ = nullptr,
+        bool use_average_block_sizes = false);
 
     void addInput();
 
diff --git a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp
index 3ee0df0efd8..132241844d7 100644
--- a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp
@@ -73,7 +73,7 @@ IMergingAlgorithm::Status ReplacingSortedAlgorithm::merge()
         if (version_column_number == -1
             || selected_row.empty()
             || current->all_columns[version_column_number]->compareAt(
-                current->pos, selected_row.row_num,
+                current->getRow(), selected_row.row_num,
                 *(*selected_row.all_columns)[version_column_number],
                 /* nan_direction_hint = */ 1) >= 0)
         {
diff --git a/src/Processors/Merges/Algorithms/RowRef.h b/src/Processors/Merges/Algorithms/RowRef.h
index 1b4da9781f8..e4610c88581 100644
--- a/src/Processors/Merges/Algorithms/RowRef.h
+++ b/src/Processors/Merges/Algorithms/RowRef.h
@@ -136,7 +136,7 @@ struct RowRef
     {
         sort_columns = cursor.impl->sort_columns.data();
         num_columns = cursor.impl->sort_columns.size();
-        row_num = cursor.impl->pos;
+        row_num = cursor.impl->getRow();
     }
 
     static bool checkEquals(size_t size, const IColumn ** lhs, size_t lhs_row, const IColumn ** rhs, size_t rhs_row)
@@ -192,7 +192,7 @@ struct RowRefWithOwnedChunk
     void set(SortCursor & cursor, SharedChunkPtr chunk)
     {
         owned_chunk = std::move(chunk);
-        row_num = cursor.impl->pos;
+        row_num = cursor.impl->getRow();
         all_columns = &owned_chunk->all_columns;
         sort_columns = &owned_chunk->sort_columns;
     }
diff --git a/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp
index 17e5e4364ff..f558a6f0677 100644
--- a/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/SummingSortedAlgorithm.cpp
@@ -688,10 +688,10 @@ IMergingAlgorithm::Status SummingSortedAlgorithm::merge()
                 return Status(merged_data.pull());
             }
 
-            merged_data.startGroup(current->all_columns, current->pos);
+            merged_data.startGroup(current->all_columns, current->getRow());
         }
         else
-            merged_data.addRow(current->all_columns, current->pos);
+            merged_data.addRow(current->all_columns, current->getRow());
 
         if (!current->isLast())
         {
diff --git a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp
index 6273bd28371..672242b253b 100644
--- a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp
+++ b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp
@@ -73,7 +73,7 @@ IMergingAlgorithm::Status VersionedCollapsingAlgorithm::merge()
 
         RowRef current_row;
 
-        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->pos];
+        Int8 sign = assert_cast<const ColumnInt8 &>(*current->all_columns[sign_column_number]).getData()[current->getRow()];
 
         setRowRef(current_row, current);
 
diff --git a/src/Processors/Merges/CollapsingSortedTransform.h b/src/Processors/Merges/CollapsingSortedTransform.h
index 4e65504a101..9e6bd306eee 100644
--- a/src/Processors/Merges/CollapsingSortedTransform.h
+++ b/src/Processors/Merges/CollapsingSortedTransform.h
@@ -27,9 +27,9 @@ public:
             sign_column,
             only_positive_sign,
             max_block_size,
+            &Poco::Logger::get("CollapsingSortedTransform"),
             out_row_sources_buf_,
-            use_average_block_sizes,
-            &Poco::Logger::get("CollapsingSortedTransform"))
+            use_average_block_sizes)
     {
     }
 
diff --git a/src/Processors/QueryPlan/MergeSortingStep.cpp b/src/Processors/QueryPlan/MergeSortingStep.cpp
index fb263cbcca1..b30286130b1 100644
--- a/src/Processors/QueryPlan/MergeSortingStep.cpp
+++ b/src/Processors/QueryPlan/MergeSortingStep.cpp
@@ -28,6 +28,7 @@ MergeSortingStep::MergeSortingStep(
     size_t max_merged_block_size_,
     UInt64 limit_,
     size_t max_bytes_before_remerge_,
+    double remerge_lowered_memory_bytes_ratio_,
     size_t max_bytes_before_external_sort_,
     VolumePtr tmp_volume_,
     size_t min_free_disk_space_)
@@ -36,6 +37,7 @@ MergeSortingStep::MergeSortingStep(
     , max_merged_block_size(max_merged_block_size_)
     , limit(limit_)
     , max_bytes_before_remerge(max_bytes_before_remerge_)
+    , remerge_lowered_memory_bytes_ratio(remerge_lowered_memory_bytes_ratio_)
     , max_bytes_before_external_sort(max_bytes_before_external_sort_), tmp_volume(tmp_volume_)
     , min_free_disk_space(min_free_disk_space_)
 {
@@ -64,6 +66,7 @@ void MergeSortingStep::transformPipeline(QueryPipeline & pipeline)
         return std::make_shared<MergeSortingTransform>(
                 header, description, max_merged_block_size, limit,
                 max_bytes_before_remerge / pipeline.getNumStreams(),
+                remerge_lowered_memory_bytes_ratio,
                 max_bytes_before_external_sort,
                 tmp_volume,
                 min_free_disk_space);
diff --git a/src/Processors/QueryPlan/MergeSortingStep.h b/src/Processors/QueryPlan/MergeSortingStep.h
index a54ea7ac365..a385a8a3e93 100644
--- a/src/Processors/QueryPlan/MergeSortingStep.h
+++ b/src/Processors/QueryPlan/MergeSortingStep.h
@@ -17,6 +17,7 @@ public:
             size_t max_merged_block_size_,
             UInt64 limit_,
             size_t max_bytes_before_remerge_,
+            double remerge_lowered_memory_bytes_ratio_,
             size_t max_bytes_before_external_sort_,
             VolumePtr tmp_volume_,
             size_t min_free_disk_space_);
@@ -36,6 +37,7 @@ private:
     UInt64 limit;
 
     size_t max_bytes_before_remerge;
+    double remerge_lowered_memory_bytes_ratio;
     size_t max_bytes_before_external_sort;
     VolumePtr tmp_volume;
     size_t min_free_disk_space;
diff --git a/src/Processors/Transforms/MergeSortingTransform.cpp b/src/Processors/Transforms/MergeSortingTransform.cpp
index 9be139e7875..ce6d0ad1f6c 100644
--- a/src/Processors/Transforms/MergeSortingTransform.cpp
+++ b/src/Processors/Transforms/MergeSortingTransform.cpp
@@ -25,7 +25,6 @@ namespace ErrorCodes
 {
     extern const int NOT_ENOUGH_SPACE;
 }
-class MergeSorter;
 
 
 class BufferingToFileTransform : public IAccumulatingTransform
@@ -96,10 +95,12 @@ MergeSortingTransform::MergeSortingTransform(
     const SortDescription & description_,
     size_t max_merged_block_size_, UInt64 limit_,
     size_t max_bytes_before_remerge_,
+    double remerge_lowered_memory_bytes_ratio_,
     size_t max_bytes_before_external_sort_, VolumePtr tmp_volume_,
     size_t min_free_disk_space_)
     : SortingTransform(header, description_, max_merged_block_size_, limit_)
     , max_bytes_before_remerge(max_bytes_before_remerge_)
+    , remerge_lowered_memory_bytes_ratio(remerge_lowered_memory_bytes_ratio_)
     , max_bytes_before_external_sort(max_bytes_before_external_sort_), tmp_volume(tmp_volume_)
     , min_free_disk_space(min_free_disk_space_) {}
 
@@ -268,9 +269,12 @@ void MergeSortingTransform::remerge()
 
     LOG_DEBUG(log, "Memory usage is lowered from {} to {}", ReadableSize(sum_bytes_in_blocks), ReadableSize(new_sum_bytes_in_blocks));
 
-    /// If the memory consumption was not lowered enough - we will not perform remerge anymore. 2 is a guess.
-    if (new_sum_bytes_in_blocks * 2 > sum_bytes_in_blocks)
+    /// If the memory consumption was not lowered enough - we will not perform remerge anymore.
+    if (remerge_lowered_memory_bytes_ratio && (new_sum_bytes_in_blocks * remerge_lowered_memory_bytes_ratio > sum_bytes_in_blocks))
+    {
         remerge_is_useful = false;
+        LOG_DEBUG(log, "Re-merging is not useful (memory usage was not lowered by remerge_sort_lowered_memory_bytes_ratio={})", remerge_lowered_memory_bytes_ratio);
+    }
 
     chunks = std::move(new_chunks);
     sum_rows_in_blocks = new_sum_rows_in_blocks;
diff --git a/src/Processors/Transforms/MergeSortingTransform.h b/src/Processors/Transforms/MergeSortingTransform.h
index d328cb818d5..c9ab804cdc8 100644
--- a/src/Processors/Transforms/MergeSortingTransform.h
+++ b/src/Processors/Transforms/MergeSortingTransform.h
@@ -22,6 +22,7 @@ public:
                           const SortDescription & description_,
                           size_t max_merged_block_size_, UInt64 limit_,
                           size_t max_bytes_before_remerge_,
+                          double remerge_lowered_memory_bytes_ratio_,
                           size_t max_bytes_before_external_sort_, VolumePtr tmp_volume_,
                           size_t min_free_disk_space_);
 
@@ -36,6 +37,7 @@ protected:
 
 private:
     size_t max_bytes_before_remerge;
+    double remerge_lowered_memory_bytes_ratio;
     size_t max_bytes_before_external_sort;
     VolumePtr tmp_volume;
     size_t min_free_disk_space;
diff --git a/src/Processors/Transforms/SortingTransform.cpp b/src/Processors/Transforms/SortingTransform.cpp
index 03c8e87ce7a..11f23530c9e 100644
--- a/src/Processors/Transforms/SortingTransform.cpp
+++ b/src/Processors/Transforms/SortingTransform.cpp
@@ -100,7 +100,7 @@ Chunk MergeSorter::mergeImpl(TSortingHeap & queue)
 
         /// Append a row from queue.
         for (size_t i = 0; i < num_columns; ++i)
-            merged_columns[i]->insertFrom(*current->all_columns[i], current->pos);
+            merged_columns[i]->insertFrom(*current->all_columns[i], current->getRow());
 
         ++total_merged_rows;
         ++merged_rows;
diff --git a/src/Server/HTTPHandler.cpp b/src/Server/HTTPHandler.cpp
index ed154ba65f2..472850950d4 100644
--- a/src/Server/HTTPHandler.cpp
+++ b/src/Server/HTTPHandler.cpp
@@ -36,6 +36,7 @@
 #include <DataStreams/IBlockInputStream.h>
 #include <Interpreters/executeQuery.h>
 #include <Interpreters/QueryParameterVisitor.h>
+#include <Interpreters/Context.h>
 #include <Common/typeid_cast.h>
 #include <Poco/Net/HTTPStream.h>
 
diff --git a/src/Server/InterserverIOHTTPHandler.cpp b/src/Server/InterserverIOHTTPHandler.cpp
index f4385e8ebc4..973759bedd1 100644
--- a/src/Server/InterserverIOHTTPHandler.cpp
+++ b/src/Server/InterserverIOHTTPHandler.cpp
@@ -11,6 +11,7 @@
 #include <IO/ReadBufferFromIStream.h>
 #include <IO/WriteBufferFromHTTPServerResponse.h>
 #include <Interpreters/InterserverIOHandler.h>
+#include <Interpreters/Context.h>
 #include "IServer.h"
 
 namespace DB
diff --git a/src/Server/StaticRequestHandler.cpp b/src/Server/StaticRequestHandler.cpp
index 7f63099c972..ad2c07ab0aa 100644
--- a/src/Server/StaticRequestHandler.cpp
+++ b/src/Server/StaticRequestHandler.cpp
@@ -10,6 +10,7 @@
 #include <IO/copyData.h>
 #include <IO/WriteHelpers.h>
 #include <IO/WriteBufferFromHTTPServerResponse.h>
+#include <Interpreters/Context.h>
 
 #include <Common/Exception.h>
 
diff --git a/src/Server/TestKeeperTCPHandler.cpp b/src/Server/TestKeeperTCPHandler.cpp
index ffb73985824..ad5c3898bb4 100644
--- a/src/Server/TestKeeperTCPHandler.cpp
+++ b/src/Server/TestKeeperTCPHandler.cpp
@@ -419,28 +419,40 @@ Coordination::OpNum TestKeeperTCPHandler::receiveRequest()
     Coordination::ZooKeeperRequestPtr request = Coordination::ZooKeeperRequestFactory::instance().get(opnum);
     request->xid = xid;
     request->readImpl(*in);
-    int response_fd = poll_wrapper->getResponseFD();
     auto promise = std::make_shared<std::promise<Coordination::ZooKeeperResponsePtr>>();
-    zkutil::ResponseCallback callback = [response_fd, promise] (const Coordination::ZooKeeperResponsePtr & response)
+    if (opnum != Coordination::OpNum::Close)
     {
-        promise->set_value(response);
-        [[maybe_unused]] int result = write(response_fd, &RESPONSE_BYTE, sizeof(RESPONSE_BYTE));
-    };
-
-    if (request->has_watch)
-    {
-        auto watch_promise = std::make_shared<std::promise<Coordination::ZooKeeperResponsePtr>>();
-        zkutil::ResponseCallback watch_callback = [response_fd, watch_promise] (const Coordination::ZooKeeperResponsePtr & response)
+        int response_fd = poll_wrapper->getResponseFD();
+        zkutil::ResponseCallback callback = [response_fd, promise] (const Coordination::ZooKeeperResponsePtr & response)
         {
-            watch_promise->set_value(response);
-            [[maybe_unused]] int result = write(response_fd, &WATCH_RESPONSE_BYTE, sizeof(WATCH_RESPONSE_BYTE));
+            promise->set_value(response);
+            [[maybe_unused]] int result = write(response_fd, &RESPONSE_BYTE, sizeof(RESPONSE_BYTE));
         };
-        test_keeper_storage->putRequest(request, session_id, callback, watch_callback);
-        responses.push(promise->get_future());
-        watch_responses.emplace_back(watch_promise->get_future());
+
+        if (request->has_watch)
+        {
+            auto watch_promise = std::make_shared<std::promise<Coordination::ZooKeeperResponsePtr>>();
+            zkutil::ResponseCallback watch_callback = [response_fd, watch_promise] (const Coordination::ZooKeeperResponsePtr & response)
+            {
+                watch_promise->set_value(response);
+                [[maybe_unused]] int result = write(response_fd, &WATCH_RESPONSE_BYTE, sizeof(WATCH_RESPONSE_BYTE));
+            };
+            test_keeper_storage->putRequest(request, session_id, callback, watch_callback);
+            responses.push(promise->get_future());
+            watch_responses.emplace_back(watch_promise->get_future());
+        }
+        else
+        {
+            test_keeper_storage->putRequest(request, session_id, callback);
+            responses.push(promise->get_future());
+        }
     }
     else
     {
+        zkutil::ResponseCallback callback = [promise] (const Coordination::ZooKeeperResponsePtr & response)
+        {
+            promise->set_value(response);
+        };
         test_keeper_storage->putRequest(request, session_id, callback);
         responses.push(promise->get_future());
     }
diff --git a/src/Storages/AlterCommands.cpp b/src/Storages/AlterCommands.cpp
index 559121d6911..59b40ffdfbb 100644
--- a/src/Storages/AlterCommands.cpp
+++ b/src/Storages/AlterCommands.cpp
@@ -793,6 +793,7 @@ void AlterCommands::apply(StorageInMemoryMetadata & metadata, const Context & co
         throw DB::Exception("Alter commands is not prepared. Cannot apply. It's a bug", ErrorCodes::LOGICAL_ERROR);
 
     auto metadata_copy = metadata;
+
     for (const AlterCommand & command : *this)
         if (!command.ignore)
             command.apply(metadata_copy, context);
@@ -823,6 +824,7 @@ void AlterCommands::apply(StorageInMemoryMetadata & metadata, const Context & co
 
     /// Changes in columns may lead to changes in TTL expressions.
     auto column_ttl_asts = metadata_copy.columns.getColumnTTLs();
+    metadata_copy.column_ttls_by_name.clear();
     for (const auto & [name, ast] : column_ttl_asts)
     {
         auto new_ttl_entry = TTLDescription::getTTLFromAST(ast, metadata_copy.columns, context, metadata_copy.primary_key);
@@ -830,7 +832,7 @@ void AlterCommands::apply(StorageInMemoryMetadata & metadata, const Context & co
     }
 
     if (metadata_copy.table_ttl.definition_ast != nullptr)
-        metadata.table_ttl = TTLTableDescription::getTTLForTableFromAST(
+        metadata_copy.table_ttl = TTLTableDescription::getTTLForTableFromAST(
             metadata_copy.table_ttl.definition_ast, metadata_copy.columns, context, metadata_copy.primary_key);
 
     metadata = std::move(metadata_copy);
diff --git a/src/Storages/Distributed/DirectoryMonitor.cpp b/src/Storages/Distributed/DirectoryMonitor.cpp
index 47da0a10d9e..5d089eb9f80 100644
--- a/src/Storages/Distributed/DirectoryMonitor.cpp
+++ b/src/Storages/Distributed/DirectoryMonitor.cpp
@@ -17,6 +17,7 @@
 #include <IO/WriteBufferFromFile.h>
 #include <Compression/CompressedReadBuffer.h>
 #include <IO/ConnectionTimeouts.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <IO/Operators.h>
 
 #include <boost/algorithm/string/find_iterator.hpp>
diff --git a/src/Storages/Distributed/DistributedBlockOutputStream.cpp b/src/Storages/Distributed/DistributedBlockOutputStream.cpp
index 2248c489679..d24967256a0 100644
--- a/src/Storages/Distributed/DistributedBlockOutputStream.cpp
+++ b/src/Storages/Distributed/DistributedBlockOutputStream.cpp
@@ -10,6 +10,7 @@
 #include <Compression/CompressedWriteBuffer.h>
 #include <IO/Operators.h>
 #include <IO/WriteBufferFromString.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <DataStreams/NativeBlockOutputStream.h>
 #include <DataStreams/RemoteBlockOutputStream.h>
 #include <DataStreams/ConvertingBlockInputStream.h>
diff --git a/src/Storages/HDFS/HDFSCommon.cpp b/src/Storages/HDFS/HDFSCommon.cpp
new file mode 100644
index 00000000000..e5ec8a06139
--- /dev/null
+++ b/src/Storages/HDFS/HDFSCommon.cpp
@@ -0,0 +1,194 @@
+#include <Storages/HDFS/HDFSCommon.h>
+#include <Poco/URI.h>
+#include <boost/algorithm/string/replace.hpp>
+
+#if USE_HDFS
+#include <Common/ShellCommand.h>
+#include <Common/Exception.h>
+#include <IO/WriteBufferFromString.h>
+#include <IO/Operators.h>
+#include <common/logger_useful.h>
+
+namespace DB
+{
+namespace ErrorCodes
+{
+extern const int BAD_ARGUMENTS;
+extern const int NETWORK_ERROR;
+extern const int EXCESSIVE_ELEMENT_IN_CONFIG;
+extern const int NO_ELEMENTS_IN_CONFIG;
+}
+
+const String HDFSBuilderWrapper::CONFIG_PREFIX = "hdfs";
+
+void HDFSBuilderWrapper::loadFromConfig(const Poco::Util::AbstractConfiguration & config,
+    const String & config_path, bool isUser)
+{
+    Poco::Util::AbstractConfiguration::Keys keys;
+
+    config.keys(config_path, keys);
+    for (const auto & key : keys)
+    {
+        const String key_path = config_path + "." + key;
+
+        String key_name;
+        if (key == "hadoop_kerberos_keytab")
+        {
+            need_kinit = true;
+            hadoop_kerberos_keytab = config.getString(key_path);
+            continue;
+        }
+        else if (key == "hadoop_kerberos_principal")
+        {
+            need_kinit = true;
+            hadoop_kerberos_principal = config.getString(key_path);
+
+#if USE_INTERNAL_HDFS3_LIBRARY
+            hdfsBuilderSetPrincipal(hdfs_builder, hadoop_kerberos_principal.c_str());
+#endif
+
+            continue;
+        }
+        else if (key == "hadoop_kerberos_kinit_command")
+        {
+            need_kinit = true;
+            hadoop_kerberos_kinit_command = config.getString(key_path);
+            continue;
+        }
+        else if (key == "hadoop_security_kerberos_ticket_cache_path")
+        {
+            if (isUser)
+            {
+                throw Exception("hadoop.security.kerberos.ticket.cache.path cannot be set per user",
+                    ErrorCodes::EXCESSIVE_ELEMENT_IN_CONFIG);
+            }
+
+            hadoop_security_kerberos_ticket_cache_path = config.getString(key_path);
+            // standard param - pass further
+        }
+
+        key_name = boost::replace_all_copy(key, "_", ".");
+
+        const auto & [k,v] = keep(key_name, config.getString(key_path));
+        hdfsBuilderConfSetStr(hdfs_builder, k.c_str(), v.c_str());
+    }
+}
+
+String HDFSBuilderWrapper::getKinitCmd()
+{
+
+    if (hadoop_kerberos_keytab.empty() || hadoop_kerberos_principal.empty())
+    {
+        throw Exception("Not enough parameters to run kinit",
+            ErrorCodes::NO_ELEMENTS_IN_CONFIG);
+    }
+
+    WriteBufferFromOwnString ss;
+
+    String cache_name =  hadoop_security_kerberos_ticket_cache_path.empty() ?
+        String() :
+        (String(" -c \"") + hadoop_security_kerberos_ticket_cache_path + "\"");
+
+    // command to run looks like
+    // kinit -R -t /keytab_dir/clickhouse.keytab -k somebody@TEST.CLICKHOUSE.TECH || ..
+    ss << hadoop_kerberos_kinit_command << cache_name <<
+        " -R -t \"" << hadoop_kerberos_keytab << "\" -k " << hadoop_kerberos_principal <<
+        "|| " << hadoop_kerberos_kinit_command << cache_name << " -t \"" <<
+        hadoop_kerberos_keytab << "\" -k " << hadoop_kerberos_principal;
+    return ss.str();
+}
+
+void HDFSBuilderWrapper::runKinit()
+{
+    String cmd = getKinitCmd();
+    LOG_DEBUG(&Poco::Logger::get("HDFSClient"), "running kinit: {}", cmd);
+
+    std::unique_lock<std::mutex> lck(kinit_mtx);
+
+    auto command = ShellCommand::execute(cmd);
+    auto status = command->tryWait();
+    if (status)
+    {
+        throw Exception("kinit failure: " + cmd, ErrorCodes::BAD_ARGUMENTS);
+    }
+}
+
+HDFSBuilderWrapper createHDFSBuilder(const String & uri_str, const Poco::Util::AbstractConfiguration & config)
+{
+    const Poco::URI uri(uri_str);
+    const auto & host = uri.getHost();
+    auto port = uri.getPort();
+    const String path = "//";
+    if (host.empty())
+        throw Exception("Illegal HDFS URI: " + uri.toString(), ErrorCodes::BAD_ARGUMENTS);
+
+    HDFSBuilderWrapper builder;
+    if (builder.get() == nullptr)
+        throw Exception("Unable to create builder to connect to HDFS: " +
+            uri.toString() + " " + String(hdfsGetLastError()),
+            ErrorCodes::NETWORK_ERROR);
+
+    hdfsBuilderConfSetStr(builder.get(), "input.read.timeout", "60000"); // 1 min
+    hdfsBuilderConfSetStr(builder.get(), "input.write.timeout", "60000"); // 1 min
+    hdfsBuilderConfSetStr(builder.get(), "input.connect.timeout", "60000"); // 1 min
+
+    String user_info = uri.getUserInfo();
+    String user;
+    if (!user_info.empty() && user_info.front() != ':')
+    {
+        size_t delim_pos = user_info.find(':');
+        if (delim_pos != String::npos)
+            user = user_info.substr(0, delim_pos);
+        else
+            user = user_info;
+
+        hdfsBuilderSetUserName(builder.get(), user.c_str());
+    }
+    hdfsBuilderSetNameNode(builder.get(), host.c_str());
+    if (port != 0)
+    {
+        hdfsBuilderSetNameNodePort(builder.get(), port);
+    }
+
+    if (config.has(HDFSBuilderWrapper::CONFIG_PREFIX))
+    {
+        builder.loadFromConfig(config, HDFSBuilderWrapper::CONFIG_PREFIX);
+    }
+
+    if (!user.empty())
+    {
+        String user_config_prefix = HDFSBuilderWrapper::CONFIG_PREFIX + "_" + user;
+        if (config.has(user_config_prefix))
+        {
+#if USE_INTERNAL_HDFS3_LIBRARY
+            builder.loadFromConfig(config, user_config_prefix, true);
+#else
+            throw Exception("Multi user HDFS configuration required internal libhdfs3",
+                ErrorCodes::EXCESSIVE_ELEMENT_IN_CONFIG);
+#endif
+        }
+    }
+
+    if (builder.need_kinit)
+    {
+        builder.runKinit();
+    }
+
+    return builder;
+}
+
+std::mutex HDFSBuilderWrapper::kinit_mtx;
+
+HDFSFSPtr createHDFSFS(hdfsBuilder * builder)
+{
+    HDFSFSPtr fs(hdfsBuilderConnect(builder));
+    if (fs == nullptr)
+        throw Exception("Unable to connect to HDFS: " + String(hdfsGetLastError()),
+            ErrorCodes::NETWORK_ERROR);
+
+    return fs;
+}
+
+}
+
+#endif
diff --git a/src/Storages/HDFS/HDFSCommon.h b/src/Storages/HDFS/HDFSCommon.h
new file mode 100644
index 00000000000..fa1ca88464e
--- /dev/null
+++ b/src/Storages/HDFS/HDFSCommon.h
@@ -0,0 +1,114 @@
+#pragma once
+
+#include <Common/config.h>
+
+#if USE_HDFS
+#include <memory>
+#include <type_traits>
+#include <vector>
+
+#include <hdfs/hdfs.h>
+#include <common/types.h>
+#include <mutex>
+
+#include <Interpreters/Context.h>
+#include <Poco/Util/AbstractConfiguration.h>
+
+
+namespace DB
+{
+namespace detail
+{
+    struct HDFSFsDeleter
+    {
+        void operator()(hdfsFS fs_ptr)
+        {
+            hdfsDisconnect(fs_ptr);
+        }
+    };
+}
+
+struct HDFSFileInfo
+{
+    hdfsFileInfo * file_info;
+    int length;
+
+    HDFSFileInfo()
+        : file_info(nullptr)
+        , length(0)
+    {
+    }
+    HDFSFileInfo(const HDFSFileInfo & other) = delete;
+    HDFSFileInfo(HDFSFileInfo && other) = default;
+    HDFSFileInfo & operator=(const HDFSFileInfo & other) = delete;
+    HDFSFileInfo & operator=(HDFSFileInfo && other) = default;
+
+    ~HDFSFileInfo()
+    {
+        hdfsFreeFileInfo(file_info, length);
+    }
+};
+
+class HDFSBuilderWrapper
+{
+    hdfsBuilder * hdfs_builder;
+    String hadoop_kerberos_keytab;
+    String hadoop_kerberos_principal;
+    String hadoop_kerberos_kinit_command = "kinit";
+    String hadoop_security_kerberos_ticket_cache_path;
+
+    static std::mutex kinit_mtx;
+
+    std::vector<std::pair<String, String>> config_stor;
+
+    // hdfs builder relies on an external config data storage
+    std::pair<String, String>& keep(const String & k, const String & v)
+    {
+        return config_stor.emplace_back(std::make_pair(k, v));
+    }
+
+    bool need_kinit{false};
+
+    static const String CONFIG_PREFIX;
+
+private:
+
+    void loadFromConfig(const Poco::Util::AbstractConfiguration & config, const String & config_path, bool isUser = false);
+
+    String getKinitCmd();
+
+    void runKinit();
+
+public:
+
+    hdfsBuilder *
+    get()
+    {
+        return hdfs_builder;
+    }
+
+    HDFSBuilderWrapper()
+        : hdfs_builder(hdfsNewBuilder())
+    {
+    }
+
+    ~HDFSBuilderWrapper()
+    {
+        hdfsFreeBuilder(hdfs_builder);
+
+    }
+
+    HDFSBuilderWrapper(const HDFSBuilderWrapper &) = delete;
+    HDFSBuilderWrapper(HDFSBuilderWrapper &&) = default;
+
+    friend HDFSBuilderWrapper createHDFSBuilder(const String & uri_str, const Poco::Util::AbstractConfiguration &);
+};
+
+using HDFSFSPtr = std::unique_ptr<std::remove_pointer_t<hdfsFS>, detail::HDFSFsDeleter>;
+
+// set read/connect timeout, default value in libhdfs3 is about 1 hour, and too large
+/// TODO Allow to tune from query Settings.
+HDFSBuilderWrapper createHDFSBuilder(const String & uri_str, const Poco::Util::AbstractConfiguration &);
+HDFSFSPtr createHDFSFS(hdfsBuilder * builder);
+}
+#endif
diff --git a/src/IO/ReadBufferFromHDFS.cpp b/src/Storages/HDFS/ReadBufferFromHDFS.cpp
similarity index 81%
rename from src/IO/ReadBufferFromHDFS.cpp
rename to src/Storages/HDFS/ReadBufferFromHDFS.cpp
index 0b9f75d8c46..f3b0e3022f1 100644
--- a/src/IO/ReadBufferFromHDFS.cpp
+++ b/src/Storages/HDFS/ReadBufferFromHDFS.cpp
@@ -1,7 +1,7 @@
 #include "ReadBufferFromHDFS.h"
 
 #if USE_HDFS
-#include <IO/HDFSCommon.h>
+#include <Storages/HDFS/HDFSCommon.h>
 #include <hdfs/hdfs.h>
 #include <mutex>
 
@@ -23,15 +23,16 @@ struct ReadBufferFromHDFS::ReadBufferFromHDFSImpl
 
     std::string hdfs_uri;
     hdfsFile fin;
-    HDFSBuilderPtr builder;
+    HDFSBuilderWrapper builder;
     HDFSFSPtr fs;
 
-    explicit ReadBufferFromHDFSImpl(const std::string & hdfs_name_)
-        : hdfs_uri(hdfs_name_)
+    explicit ReadBufferFromHDFSImpl(const std::string & hdfs_name_,
+        const Poco::Util::AbstractConfiguration & config_)
+        : hdfs_uri(hdfs_name_),
+          builder(createHDFSBuilder(hdfs_uri, config_))
     {
         std::lock_guard lock(hdfs_init_mutex);
 
-        builder = createHDFSBuilder(hdfs_uri);
         fs = createHDFSFS(builder.get());
         const size_t begin_of_path = hdfs_uri.find('/', hdfs_uri.find("//") + 2);
         const std::string path = hdfs_uri.substr(begin_of_path);
@@ -58,11 +59,14 @@ struct ReadBufferFromHDFS::ReadBufferFromHDFSImpl
     }
 };
 
+
 std::mutex ReadBufferFromHDFS::ReadBufferFromHDFSImpl::hdfs_init_mutex;
 
-ReadBufferFromHDFS::ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size)
-    : BufferWithOwnMemory<ReadBuffer>(buf_size)
-    , impl(std::make_unique<ReadBufferFromHDFSImpl>(hdfs_name_))
+ReadBufferFromHDFS::ReadBufferFromHDFS(const std::string & hdfs_name_,
+    const Poco::Util::AbstractConfiguration & config_,
+    size_t buf_size_)
+    : BufferWithOwnMemory<ReadBuffer>(buf_size_)
+    , impl(std::make_unique<ReadBufferFromHDFSImpl>(hdfs_name_, config_))
 {
 }
 
diff --git a/src/IO/ReadBufferFromHDFS.h b/src/Storages/HDFS/ReadBufferFromHDFS.h
similarity index 69%
rename from src/IO/ReadBufferFromHDFS.h
rename to src/Storages/HDFS/ReadBufferFromHDFS.h
index 58479d45660..8d26c001b2e 100644
--- a/src/IO/ReadBufferFromHDFS.h
+++ b/src/Storages/HDFS/ReadBufferFromHDFS.h
@@ -8,6 +8,13 @@
 #include <string>
 #include <memory>
 
+#include <hdfs/hdfs.h>
+
+#include <common/types.h>
+
+#include <Interpreters/Context.h>
+
+
 namespace DB
 {
 /** Accepts HDFS path to file and opens it.
@@ -18,7 +25,7 @@ class ReadBufferFromHDFS : public BufferWithOwnMemory<ReadBuffer>
     struct ReadBufferFromHDFSImpl;
     std::unique_ptr<ReadBufferFromHDFSImpl> impl;
 public:
-    ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE);
+    ReadBufferFromHDFS(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration &, size_t buf_size_ = DBMS_DEFAULT_BUFFER_SIZE);
     ~ReadBufferFromHDFS() override;
 
     bool nextImpl() override;
diff --git a/src/Storages/StorageHDFS.cpp b/src/Storages/HDFS/StorageHDFS.cpp
similarity index 95%
rename from src/Storages/StorageHDFS.cpp
rename to src/Storages/HDFS/StorageHDFS.cpp
index 0a4729b1c06..f50b4722ce1 100644
--- a/src/Storages/StorageHDFS.cpp
+++ b/src/Storages/HDFS/StorageHDFS.cpp
@@ -3,15 +3,15 @@
 #if USE_HDFS
 
 #include <Storages/StorageFactory.h>
-#include <Storages/StorageHDFS.h>
+#include <Storages/HDFS/StorageHDFS.h>
 #include <Interpreters/Context.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Parsers/ASTLiteral.h>
 #include <IO/ReadHelpers.h>
-#include <IO/ReadBufferFromHDFS.h>
-#include <IO/WriteBufferFromHDFS.h>
+#include <Storages/HDFS/ReadBufferFromHDFS.h>
+#include <Storages/HDFS/WriteBufferFromHDFS.h>
 #include <IO/WriteHelpers.h>
-#include <IO/HDFSCommon.h>
+#include <Storages/HDFS/HDFSCommon.h>
 #include <Formats/FormatFactory.h>
 #include <DataTypes/DataTypeString.h>
 #include <DataStreams/IBlockOutputStream.h>
@@ -121,7 +121,7 @@ public:
                 current_path = uri + path;
 
                 auto compression = chooseCompressionMethod(path, compression_method);
-                auto read_buf = wrapReadBufferWithCompressionMethod(std::make_unique<ReadBufferFromHDFS>(current_path), compression);
+                auto read_buf = wrapReadBufferWithCompressionMethod(std::make_unique<ReadBufferFromHDFS>(current_path, context.getGlobalContext().getConfigRef()), compression);
                 auto input_stream = FormatFactory::instance().getInput(format, *read_buf, sample_block, context, max_block_size);
 
                 reader = std::make_shared<OwningBlockInputStream<ReadBuffer>>(input_stream, std::move(read_buf));
@@ -180,7 +180,7 @@ public:
         const CompressionMethod compression_method)
         : sample_block(sample_block_)
     {
-        write_buf = wrapWriteBufferWithCompressionMethod(std::make_unique<WriteBufferFromHDFS>(uri), compression_method, 3);
+        write_buf = wrapWriteBufferWithCompressionMethod(std::make_unique<WriteBufferFromHDFS>(uri, context.getGlobalContext().getConfigRef()), compression_method, 3);
         writer = FormatFactory::instance().getOutput(format, *write_buf, sample_block, context);
     }
 
@@ -274,7 +274,7 @@ Pipe StorageHDFS::read(
     const String path_from_uri = uri.substr(begin_of_path);
     const String uri_without_path = uri.substr(0, begin_of_path);
 
-    HDFSBuilderPtr builder = createHDFSBuilder(uri_without_path + "/");
+    HDFSBuilderWrapper builder = createHDFSBuilder(uri_without_path + "/", context_.getGlobalContext().getConfigRef());
     HDFSFSPtr fs = createHDFSFS(builder.get());
 
     auto sources_info = std::make_shared<HDFSSource::SourcesInfo>();
diff --git a/src/Storages/StorageHDFS.h b/src/Storages/HDFS/StorageHDFS.h
similarity index 100%
rename from src/Storages/StorageHDFS.h
rename to src/Storages/HDFS/StorageHDFS.h
diff --git a/src/IO/WriteBufferFromHDFS.cpp b/src/Storages/HDFS/WriteBufferFromHDFS.cpp
similarity index 84%
rename from src/IO/WriteBufferFromHDFS.cpp
rename to src/Storages/HDFS/WriteBufferFromHDFS.cpp
index 2176c77b914..3873eace10d 100644
--- a/src/IO/WriteBufferFromHDFS.cpp
+++ b/src/Storages/HDFS/WriteBufferFromHDFS.cpp
@@ -2,8 +2,9 @@
 
 #if USE_HDFS
 
-#include <IO/WriteBufferFromHDFS.h>
-#include <IO/HDFSCommon.h>
+#include <Interpreters/Context.h>
+#include <Storages/HDFS/WriteBufferFromHDFS.h>
+#include <Storages/HDFS/HDFSCommon.h>
 #include <hdfs/hdfs.h>
 
 
@@ -23,12 +24,12 @@ struct WriteBufferFromHDFS::WriteBufferFromHDFSImpl
 {
     std::string hdfs_uri;
     hdfsFile fout;
-    HDFSBuilderPtr builder;
+    HDFSBuilderWrapper builder;
     HDFSFSPtr fs;
 
-    explicit WriteBufferFromHDFSImpl(const std::string & hdfs_name_)
+    explicit WriteBufferFromHDFSImpl(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration & config_)
         : hdfs_uri(hdfs_name_)
-        , builder(createHDFSBuilder(hdfs_uri))
+        , builder(createHDFSBuilder(hdfs_uri,config_))
         , fs(createHDFSFS(builder.get()))
     {
         const size_t begin_of_path = hdfs_uri.find('/', hdfs_uri.find("//") + 2);
@@ -72,9 +73,9 @@ struct WriteBufferFromHDFS::WriteBufferFromHDFSImpl
     }
 };
 
-WriteBufferFromHDFS::WriteBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size)
-    : BufferWithOwnMemory<WriteBuffer>(buf_size)
-    , impl(std::make_unique<WriteBufferFromHDFSImpl>(hdfs_name_))
+WriteBufferFromHDFS::WriteBufferFromHDFS(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration & config_, size_t buf_size_)
+    : BufferWithOwnMemory<WriteBuffer>(buf_size_)
+    , impl(std::make_unique<WriteBufferFromHDFSImpl>(hdfs_name_, config_))
 {
 }
 
diff --git a/src/IO/WriteBufferFromHDFS.h b/src/Storages/HDFS/WriteBufferFromHDFS.h
similarity index 80%
rename from src/IO/WriteBufferFromHDFS.h
rename to src/Storages/HDFS/WriteBufferFromHDFS.h
index 22eab5c4d5d..339ec0c2a7c 100644
--- a/src/IO/WriteBufferFromHDFS.h
+++ b/src/Storages/HDFS/WriteBufferFromHDFS.h
@@ -18,7 +18,7 @@ class WriteBufferFromHDFS : public BufferWithOwnMemory<WriteBuffer>
     struct WriteBufferFromHDFSImpl;
     std::unique_ptr<WriteBufferFromHDFSImpl> impl;
 public:
-    WriteBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE);
+    WriteBufferFromHDFS(const std::string & hdfs_name_, const Poco::Util::AbstractConfiguration &, size_t buf_size_ = DBMS_DEFAULT_BUFFER_SIZE);
     WriteBufferFromHDFS(WriteBufferFromHDFS &&) = default;
 
     void nextImpl() override;
diff --git a/src/Storages/Kafka/KafkaSettings.h b/src/Storages/Kafka/KafkaSettings.h
index 53fbe8adc6b..1df10d16339 100644
--- a/src/Storages/Kafka/KafkaSettings.h
+++ b/src/Storages/Kafka/KafkaSettings.h
@@ -20,7 +20,7 @@ class ASTStorage;
     M(Milliseconds, kafka_poll_timeout_ms, 0, "Timeout for single poll from Kafka.", 0) \
     /* default is min(max_block_size, kafka_max_block_size)*/ \
     M(UInt64, kafka_poll_max_batch_size, 0, "Maximum amount of messages to be polled in a single Kafka poll.", 0) \
-    /* default is = min_insert_block_size / kafka_num_consumers  */ \
+    /* default is = max_insert_block_size / kafka_num_consumers  */ \
     M(UInt64, kafka_max_block_size, 0, "Number of row collected by poll(s) for flushing data from Kafka.", 0) \
     /* default is stream_flush_interval_ms */ \
     M(Milliseconds, kafka_flush_interval_ms, 0, "Timeout for flushing data from Kafka.", 0) \
diff --git a/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp b/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
index 143a9920b93..df905215df1 100644
--- a/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
+++ b/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
@@ -3,33 +3,53 @@
 namespace DB
 {
 
-IMergeTreeDataPartWriter::IMergeTreeDataPartWriter(
-    const MergeTreeData::DataPartPtr & data_part_,
-    const NamesAndTypesList & columns_list_,
-    const StorageMetadataPtr & metadata_snapshot_,
-    const MergeTreeWriterSettings & settings_)
-    : data_part(data_part_)
-    , storage(data_part_->storage)
-    , metadata_snapshot(metadata_snapshot_)
-    , columns_list(columns_list_)
-    , settings(settings_)
-    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity){}
+Block getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation)
+{
+    Block result;
+    for (size_t i = 0, size = names.size(); i < size; ++i)
+    {
+        const auto & name = names[i];
+        result.insert(i, block.getByName(name));
+
+        /// Reorder primary key columns in advance and add them to `primary_key_columns`.
+        if (permutation)
+        {
+            auto & column = result.getByPosition(i);
+            column.column = column.column->permute(*permutation, 0);
+        }
+    }
+
+    return result;
+}
+
+Block permuteBlockIfNeeded(const Block & block, const IColumn::Permutation * permutation)
+{
+    Block result;
+    for (size_t i = 0; i < block.columns(); ++i)
+    {
+        result.insert(i, block.getByPosition(i));
+        if (permutation)
+        {
+            auto & column = result.getByPosition(i);
+            column.column = column.column->permute(*permutation, 0);
+        }
+    }
+    return result;
+}
 
 IMergeTreeDataPartWriter::IMergeTreeDataPartWriter(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
     const StorageMetadataPtr & metadata_snapshot_,
-    const MergeTreeIndices & skip_indices_,
-    const MergeTreeIndexGranularity & index_granularity_,
-    const MergeTreeWriterSettings & settings_)
+    const MergeTreeWriterSettings & settings_,
+    const MergeTreeIndexGranularity & index_granularity_)
     : data_part(data_part_)
     , storage(data_part_->storage)
     , metadata_snapshot(metadata_snapshot_)
     , columns_list(columns_list_)
-    , skip_indices(skip_indices_)
-    , index_granularity(index_granularity_)
     , settings(settings_)
-    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity) {}
+    , index_granularity(index_granularity_)
+    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity){}
 
 Columns IMergeTreeDataPartWriter::releaseIndexColumns()
 {
@@ -38,12 +58,6 @@ Columns IMergeTreeDataPartWriter::releaseIndexColumns()
         std::make_move_iterator(index_columns.end()));
 }
 
-void IMergeTreeDataPartWriter::next()
-{
-    current_mark = next_mark;
-    index_offset = next_index_offset;
-}
-
 IMergeTreeDataPartWriter::~IMergeTreeDataPartWriter() = default;
 
 }
diff --git a/src/Storages/MergeTree/IMergeTreeDataPartWriter.h b/src/Storages/MergeTree/IMergeTreeDataPartWriter.h
index 4a42a58a65b..5e12d5da678 100644
--- a/src/Storages/MergeTree/IMergeTreeDataPartWriter.h
+++ b/src/Storages/MergeTree/IMergeTreeDataPartWriter.h
@@ -13,6 +13,9 @@
 namespace DB
 {
 
+Block getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation);
+
+Block permuteBlockIfNeeded(const Block & block, const IColumn::Permutation * permutation);
 
 /// Writes data part to disk in different formats.
 /// Calculates and serializes primary and skip indices if needed.
@@ -23,68 +26,29 @@ public:
         const MergeTreeData::DataPartPtr & data_part_,
         const NamesAndTypesList & columns_list_,
         const StorageMetadataPtr & metadata_snapshot_,
-        const MergeTreeWriterSettings & settings_);
-
-    IMergeTreeDataPartWriter(
-        const MergeTreeData::DataPartPtr & data_part_,
-        const NamesAndTypesList & columns_list_,
-        const StorageMetadataPtr & metadata_snapshot_,
-        const MergeTreeIndices & skip_indices_,
-        const MergeTreeIndexGranularity & index_granularity_,
-        const MergeTreeWriterSettings & settings_);
+        const MergeTreeWriterSettings & settings_,
+        const MergeTreeIndexGranularity & index_granularity_ = {});
 
     virtual ~IMergeTreeDataPartWriter();
 
-    virtual void write(
-        const Block & block, const IColumn::Permutation * permutation = nullptr,
-        /* Blocks with already sorted index columns */
-        const Block & primary_key_block = {}, const Block & skip_indexes_block = {}) = 0;
+    virtual void write(const Block & block, const IColumn::Permutation * permutation) = 0;
 
-    virtual void calculateAndSerializePrimaryIndex(const Block & /* primary_index_block */) {}
-    virtual void calculateAndSerializeSkipIndices(const Block & /* skip_indexes_block */) {}
-
-    /// Shift mark and offset to prepare read next mark.
-    /// You must call it after calling write method and optionally
-    ///  calling calculations of primary and skip indices.
-    void next();
-
-    virtual void initSkipIndices() {}
-    virtual void initPrimaryIndex() {}
-
-    virtual void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) = 0;
-    virtual void finishPrimaryIndexSerialization(MergeTreeData::DataPart::Checksums & /* checksums */, bool /* sync */) {}
-    virtual void finishSkipIndicesSerialization(MergeTreeData::DataPart::Checksums & /* checksums */, bool /* sync */) {}
+    virtual void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) = 0;
 
     Columns releaseIndexColumns();
     const MergeTreeIndexGranularity & getIndexGranularity() const { return index_granularity; }
-    const MergeTreeIndices & getSkipIndices() { return skip_indices; }
 
 protected:
-    size_t getCurrentMark() const { return current_mark; }
-    size_t getIndexOffset() const { return index_offset; }
 
-    using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
-    using SerializationStates = std::unordered_map<String, SerializationState>;
-
-    MergeTreeData::DataPartPtr data_part;
+    const MergeTreeData::DataPartPtr data_part;
     const MergeTreeData & storage;
-    StorageMetadataPtr metadata_snapshot;
-    NamesAndTypesList columns_list;
-    MergeTreeIndices skip_indices;
+    const StorageMetadataPtr metadata_snapshot;
+    const NamesAndTypesList columns_list;
+    const MergeTreeWriterSettings settings;
     MergeTreeIndexGranularity index_granularity;
-    MergeTreeWriterSettings settings;
-    bool with_final_mark;
-
-    size_t next_mark = 0;
-    size_t next_index_offset = 0;
+    const bool with_final_mark;
 
     MutableColumns index_columns;
-
-private:
-    /// Data is already written up to this mark.
-    size_t current_mark = 0;
-    /// The offset to the first row of the block for which you want to write the index.
-    size_t index_offset = 0;
 };
 
 }
diff --git a/src/Storages/MergeTree/IMergedBlockOutputStream.cpp b/src/Storages/MergeTree/IMergedBlockOutputStream.cpp
index e5cc6d6ecbf..2a52d7d53f0 100644
--- a/src/Storages/MergeTree/IMergedBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/IMergedBlockOutputStream.cpp
@@ -15,25 +15,6 @@ IMergedBlockOutputStream::IMergedBlockOutputStream(
 {
 }
 
-Block IMergedBlockOutputStream::getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation)
-{
-    Block result;
-    for (size_t i = 0, size = names.size(); i < size; ++i)
-    {
-        const auto & name = names[i];
-        result.insert(i, block.getByName(name));
-
-        /// Reorder primary key columns in advance and add them to `primary_key_columns`.
-        if (permutation)
-        {
-            auto & column = result.getByPosition(i);
-            column.column = column.column->permute(*permutation, 0);
-        }
-    }
-
-    return result;
-}
-
 NameSet IMergedBlockOutputStream::removeEmptyColumnsFromPart(
     const MergeTreeDataPartPtr & data_part,
     NamesAndTypesList & columns,
diff --git a/src/Storages/MergeTree/IMergedBlockOutputStream.h b/src/Storages/MergeTree/IMergedBlockOutputStream.h
index 7ec9f85ff28..ed8da4d334b 100644
--- a/src/Storages/MergeTree/IMergedBlockOutputStream.h
+++ b/src/Storages/MergeTree/IMergedBlockOutputStream.h
@@ -42,8 +42,6 @@ protected:
     VolumePtr volume;
     String part_path;
 
-    static Block getBlockAndPermute(const Block & block, const Names & names, const IColumn::Permutation * permutation);
-
     IMergeTreeDataPart::MergeTreeWriterPtr writer;
 };
 
diff --git a/src/Storages/MergeTree/KeyCondition.cpp b/src/Storages/MergeTree/KeyCondition.cpp
index 5903dc89e9b..ead93a6e1d9 100644
--- a/src/Storages/MergeTree/KeyCondition.cpp
+++ b/src/Storages/MergeTree/KeyCondition.cpp
@@ -1135,7 +1135,7 @@ bool KeyCondition::tryParseAtomFromAST(const ASTPtr & node, const Context & cont
                         ColumnsWithTypeAndName arguments{
                             {nullptr, key_expr_type, ""}, {DataTypeString().createColumnConst(1, common_type->getName()), common_type, ""}};
                         FunctionOverloadResolverPtr func_builder_cast
-                                = std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver::createImpl(false));
+                                = std::make_shared<FunctionOverloadResolverAdaptor>(CastOverloadResolver<CastType::nonAccurate>::createImpl(false));
                         auto func_cast = func_builder_cast->build(arguments);
 
                         /// If we know the given range only contains one value, then we treat all functions as positive monotonic.
diff --git a/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp b/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp
index 40714e5af31..904081cc1df 100644
--- a/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/MergeTreeBlockOutputStream.cpp
@@ -13,16 +13,22 @@ Block MergeTreeBlockOutputStream::getHeader() const
 }
 
 
+void MergeTreeBlockOutputStream::writePrefix()
+{
+    /// Only check "too many parts" before write,
+    /// because interrupting long-running INSERT query in the middle is not convenient for users.
+    storage.delayInsertOrThrowIfNeeded();
+}
+
+
 void MergeTreeBlockOutputStream::write(const Block & block)
 {
-    storage.delayInsertOrThrowIfNeeded();
-
     auto part_blocks = storage.writer.splitBlockIntoParts(block, max_parts_per_block, metadata_snapshot);
     for (auto & current_block : part_blocks)
     {
         Stopwatch watch;
 
-        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot);
+        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot, optimize_on_insert);
         storage.renameTempPartAndAdd(part, &storage.increment);
 
         PartLog::addNewPart(storage.global_context, part, watch.elapsed());
diff --git a/src/Storages/MergeTree/MergeTreeBlockOutputStream.h b/src/Storages/MergeTree/MergeTreeBlockOutputStream.h
index 8aae7f3e625..8caa53382dc 100644
--- a/src/Storages/MergeTree/MergeTreeBlockOutputStream.h
+++ b/src/Storages/MergeTree/MergeTreeBlockOutputStream.h
@@ -14,20 +14,23 @@ class StorageMergeTree;
 class MergeTreeBlockOutputStream : public IBlockOutputStream
 {
 public:
-    MergeTreeBlockOutputStream(StorageMergeTree & storage_, const StorageMetadataPtr metadata_snapshot_, size_t max_parts_per_block_)
+    MergeTreeBlockOutputStream(StorageMergeTree & storage_, const StorageMetadataPtr metadata_snapshot_, size_t max_parts_per_block_, bool optimize_on_insert_)
         : storage(storage_)
         , metadata_snapshot(metadata_snapshot_)
         , max_parts_per_block(max_parts_per_block_)
+        , optimize_on_insert(optimize_on_insert_)
     {
     }
 
     Block getHeader() const override;
     void write(const Block & block) override;
+    void writePrefix() override;
 
 private:
     StorageMergeTree & storage;
     StorageMetadataPtr metadata_snapshot;
     size_t max_parts_per_block;
+    bool optimize_on_insert;
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeData.cpp b/src/Storages/MergeTree/MergeTreeData.cpp
index c3a599665bb..9fa19859c7f 100644
--- a/src/Storages/MergeTree/MergeTreeData.cpp
+++ b/src/Storages/MergeTree/MergeTreeData.cpp
@@ -2380,8 +2380,6 @@ void MergeTreeData::delayInsertOrThrowIfNeeded(Poco::Event * until) const
     }
 
     const size_t parts_count_in_partition = getMaxPartsCountForPartition();
-    if (parts_count_in_partition < settings->parts_to_delay_insert)
-        return;
 
     if (parts_count_in_partition >= settings->parts_to_throw_insert)
     {
@@ -2389,6 +2387,9 @@ void MergeTreeData::delayInsertOrThrowIfNeeded(Poco::Event * until) const
         throw Exception("Too many parts (" + toString(parts_count_in_partition) + "). Merges are processing significantly slower than inserts.", ErrorCodes::TOO_MANY_PARTS);
     }
 
+    if (parts_count_in_partition < settings->parts_to_delay_insert)
+        return;
+
     const size_t max_k = settings->parts_to_throw_insert - settings->parts_to_delay_insert; /// always > 0
     const size_t k = 1 + parts_count_in_partition - settings->parts_to_delay_insert; /// from 1 to max_k
     const double delay_milliseconds = ::pow(settings->max_delay_to_insert * 1000, static_cast<double>(k) / max_k);
@@ -2406,24 +2407,6 @@ void MergeTreeData::delayInsertOrThrowIfNeeded(Poco::Event * until) const
         std::this_thread::sleep_for(std::chrono::milliseconds(static_cast<size_t>(delay_milliseconds)));
 }
 
-void MergeTreeData::throwInsertIfNeeded() const
-{
-    const auto settings = getSettings();
-    const size_t parts_count_in_total = getPartsCount();
-    if (parts_count_in_total >= settings->max_parts_in_total)
-    {
-        ProfileEvents::increment(ProfileEvents::RejectedInserts);
-        throw Exception("Too many parts (" + toString(parts_count_in_total) + ") in all partitions in total. This indicates wrong choice of partition key. The threshold can be modified with 'max_parts_in_total' setting in <merge_tree> element in config.xml or with per-table setting.", ErrorCodes::TOO_MANY_PARTS);
-    }
-
-    const size_t parts_count_in_partition = getMaxPartsCountForPartition();
-
-    if (parts_count_in_partition >= settings->parts_to_throw_insert)
-    {
-        ProfileEvents::increment(ProfileEvents::RejectedInserts);
-        throw Exception("Too many parts (" + toString(parts_count_in_partition) + "). Merges are processing significantly slower than inserts.", ErrorCodes::TOO_MANY_PARTS);
-    }
-}
 
 MergeTreeData::DataPartPtr MergeTreeData::getActiveContainingPart(
     const MergeTreePartInfo & part_info, MergeTreeData::DataPartState state, DataPartsLock & /*lock*/) const
@@ -2451,6 +2434,7 @@ MergeTreeData::DataPartPtr MergeTreeData::getActiveContainingPart(
     return nullptr;
 }
 
+
 void MergeTreeData::swapActivePart(MergeTreeData::DataPartPtr part_copy)
 {
     auto lock = lockParts();
diff --git a/src/Storages/MergeTree/MergeTreeData.h b/src/Storages/MergeTree/MergeTreeData.h
index 38e6c34e232..382f53f40a4 100644
--- a/src/Storages/MergeTree/MergeTreeData.h
+++ b/src/Storages/MergeTree/MergeTreeData.h
@@ -423,7 +423,6 @@ public:
     /// If the table contains too many active parts, sleep for a while to give them time to merge.
     /// If until is non-null, wake up from the sleep earlier if the event happened.
     void delayInsertOrThrowIfNeeded(Poco::Event * until = nullptr) const;
-    void throwInsertIfNeeded() const;
 
     /// Renames temporary part to a permanent part and adds it to the parts set.
     /// It is assumed that the part does not intersect with existing parts.
diff --git a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
index 1065b992396..e375e653f8f 100644
--- a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
@@ -235,6 +235,7 @@ SelectPartsDecision MergeTreeDataMergerMutator::selectPartsToMerge(
     /// Previous part only in boundaries of partition frame
     const MergeTreeData::DataPartPtr * prev_part = nullptr;
 
+    size_t parts_selected_precondition = 0;
     for (const MergeTreeData::DataPartPtr & part : data_parts)
     {
         const String & partition_id = part->info.partition_id;
@@ -282,6 +283,8 @@ SelectPartsDecision MergeTreeDataMergerMutator::selectPartsToMerge(
         part_info.compression_codec_desc = part->default_codec->getFullCodecDesc();
         part_info.shall_participate_in_merges = has_volumes_with_disabled_merges ? part->shallParticipateInMerges(storage_policy) : true;
 
+        ++parts_selected_precondition;
+
         parts_ranges.back().emplace_back(part_info);
 
         /// Check for consistency of data parts. If assertion is failed, it requires immediate investigation.
@@ -294,6 +297,13 @@ SelectPartsDecision MergeTreeDataMergerMutator::selectPartsToMerge(
         prev_part = &part;
     }
 
+    if (parts_selected_precondition == 0)
+    {
+        if (out_disable_reason)
+            *out_disable_reason = "No parts satisfy preconditions for merge";
+        return SelectPartsDecision::CANNOT_SELECT;
+    }
+
     IMergeSelector::PartsRange parts_to_merge;
 
     if (metadata_snapshot->hasAnyTTL() && merge_with_ttl_allowed && !ttl_merges_blocker.isCancelled())
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
index 0b22f1271e3..389d2950f65 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.cpp
@@ -4,6 +4,11 @@
 namespace DB
 {
 
+namespace ErrorCodes
+{
+    extern const int LOGICAL_ERROR;
+}
+
 MergeTreeDataPartWriterCompact::MergeTreeDataPartWriterCompact(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
@@ -64,9 +69,64 @@ void MergeTreeDataPartWriterCompact::addStreams(const String & name, const IData
     type.enumerateStreams(callback, stream_path);
 }
 
-void MergeTreeDataPartWriterCompact::write(
-    const Block & block, const IColumn::Permutation * permutation,
-    const Block & primary_key_block, const Block & skip_indexes_block)
+namespace
+{
+
+/// Get granules for block using index_granularity
+Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity, size_t block_rows, size_t current_mark, bool last_block)
+{
+    if (current_mark >= index_granularity.getMarksCount())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Request to get granules from mark {} but index granularity size is {}", current_mark, index_granularity.getMarksCount());
+
+    Granules result;
+    size_t current_row = 0;
+    while (current_row < block_rows)
+    {
+        size_t expected_rows_in_mark = index_granularity.getMarkRows(current_mark);
+        size_t rows_left_in_block = block_rows - current_row;
+        if (rows_left_in_block < expected_rows_in_mark && !last_block)
+        {
+            /// Invariant: we always have equal amount of rows for block in compact parts because we accumulate them in buffer.
+            /// The only exclusion is the last block, when we cannot accumulate more rows.
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Required to write {} rows, but only {} rows was written for the non last granule", expected_rows_in_mark, rows_left_in_block);
+        }
+
+        result.emplace_back(Granule{
+            .start_row = current_row,
+            .granularity_rows = expected_rows_in_mark,
+            .block_rows = std::min(rows_left_in_block, expected_rows_in_mark),
+            .mark_number = current_mark,
+            .mark_on_start = true
+        });
+        current_row += expected_rows_in_mark;
+        current_mark++;
+    }
+
+    return result;
+}
+
+/// Write single granule of one column (rows between 2 marks)
+void writeColumnSingleGranule(
+    const ColumnWithTypeAndName & column,
+    IDataType::OutputStreamGetter stream_getter,
+    size_t from_row,
+    size_t number_of_rows)
+{
+    IDataType::SerializeBinaryBulkStatePtr state;
+    IDataType::SerializeBinaryBulkSettings serialize_settings;
+
+    serialize_settings.getter = stream_getter;
+    serialize_settings.position_independent_encoding = true;
+    serialize_settings.low_cardinality_max_dictionary_size = 0;
+
+    column.type->serializeBinaryBulkStatePrefix(serialize_settings, state);
+    column.type->serializeBinaryBulkWithMultipleStreams(*column.column, from_row, number_of_rows, serialize_settings, state);
+    column.type->serializeBinaryBulkStateSuffix(serialize_settings, state);
+}
+
+}
+
+void MergeTreeDataPartWriterCompact::write(const Block & block, const IColumn::Permutation * permutation)
 {
     /// Fill index granularity for this block
     /// if it's unknown (in case of insert data or horizontal merge,
@@ -77,58 +137,43 @@ void MergeTreeDataPartWriterCompact::write(
         fillIndexGranularity(index_granularity_for_block, block.rows());
     }
 
-    Block result_block;
-
-    if (permutation)
-    {
-        for (const auto & it : columns_list)
-        {
-            if (primary_key_block.has(it.name))
-                result_block.insert(primary_key_block.getByName(it.name));
-            else if (skip_indexes_block.has(it.name))
-                result_block.insert(skip_indexes_block.getByName(it.name));
-            else
-            {
-                auto column = block.getByName(it.name);
-                column.column = column.column->permute(*permutation, 0);
-                result_block.insert(column);
-            }
-        }
-    }
-    else
-    {
-        result_block = block;
-    }
+    Block result_block = permuteBlockIfNeeded(block, permutation);
 
     if (!header)
         header = result_block.cloneEmpty();
 
     columns_buffer.add(result_block.mutateColumns());
-    size_t last_mark_rows = index_granularity.getLastMarkRows();
+    size_t current_mark_rows = index_granularity.getMarkRows(getCurrentMark());
     size_t rows_in_buffer = columns_buffer.size();
 
-    if (rows_in_buffer < last_mark_rows)
+    if (rows_in_buffer >= current_mark_rows)
     {
-        /// If it's not enough rows for granule, accumulate blocks
-        ///  and save how much rows we already have.
-        next_index_offset = last_mark_rows - rows_in_buffer;
-        return;
+        Block flushed_block = header.cloneWithColumns(columns_buffer.releaseColumns());
+        auto granules_to_write = getGranulesToWrite(index_granularity, flushed_block.rows(), getCurrentMark(), /* last_block = */ false);
+        writeDataBlockPrimaryIndexAndSkipIndices(flushed_block, granules_to_write);
+        setCurrentMark(getCurrentMark() + granules_to_write.size());
     }
-
-    writeBlock(header.cloneWithColumns(columns_buffer.releaseColumns()));
 }
 
-void MergeTreeDataPartWriterCompact::writeBlock(const Block & block)
+void MergeTreeDataPartWriterCompact::writeDataBlockPrimaryIndexAndSkipIndices(const Block & block, const Granules & granules_to_write)
 {
-    size_t total_rows = block.rows();
-    size_t from_mark = getCurrentMark();
-    size_t current_row = 0;
+    writeDataBlock(block, granules_to_write);
 
-    while (current_row < total_rows)
+    if (settings.rewrite_primary_key)
     {
-        size_t rows_to_write = index_granularity.getMarkRows(from_mark);
+        Block primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), nullptr);
+        calculateAndSerializePrimaryIndex(primary_key_block, granules_to_write);
+    }
 
-        if (rows_to_write)
+    Block skip_indices_block = getBlockAndPermute(block, getSkipIndicesColumns(), nullptr);
+    calculateAndSerializeSkipIndices(skip_indices_block, granules_to_write);
+}
+
+void MergeTreeDataPartWriterCompact::writeDataBlock(const Block & block, const Granules & granules)
+{
+    for (const auto & granule : granules)
+    {
+        if (granule.granularity_rows)
             data_written = true;
 
         auto name_and_type = columns_list.begin();
@@ -161,53 +206,30 @@ void MergeTreeDataPartWriterCompact::writeBlock(const Block & block)
             writeIntBinary(plain_hashing.count(), marks);
             writeIntBinary(UInt64(0), marks);
 
-            writeColumnSingleGranule(block.getByName(name_and_type->name), stream_getter, current_row, rows_to_write);
+            writeColumnSingleGranule(block.getByName(name_and_type->name), stream_getter, granule.start_row, granule.granularity_rows);
 
             /// Each type always have at least one substream
             prev_stream->hashing_buf.next(); //-V522
         }
 
-        ++from_mark;
-        size_t rows_written = total_rows - current_row;
-        current_row += rows_to_write;
-
-        /// Correct last mark as it should contain exact amount of rows.
-        if (current_row >= total_rows && rows_written != rows_to_write)
-        {
-            rows_to_write = rows_written;
-            index_granularity.popMark();
-            index_granularity.appendMark(rows_written);
-        }
-
-        writeIntBinary(rows_to_write, marks);
+        writeIntBinary(granule.block_rows, marks);
     }
-
-    next_index_offset = 0;
-    next_mark = from_mark;
-}
-
-void MergeTreeDataPartWriterCompact::writeColumnSingleGranule(
-    const ColumnWithTypeAndName & column,
-    IDataType::OutputStreamGetter stream_getter,
-    size_t from_row,
-    size_t number_of_rows)
-{
-    IDataType::SerializeBinaryBulkStatePtr state;
-    IDataType::SerializeBinaryBulkSettings serialize_settings;
-
-    serialize_settings.getter = stream_getter;
-    serialize_settings.position_independent_encoding = true;
-    serialize_settings.low_cardinality_max_dictionary_size = 0;
-
-    column.type->serializeBinaryBulkStatePrefix(serialize_settings, state);
-    column.type->serializeBinaryBulkWithMultipleStreams(*column.column, from_row, number_of_rows, serialize_settings, state);
-    column.type->serializeBinaryBulkStateSuffix(serialize_settings, state);
 }
 
 void MergeTreeDataPartWriterCompact::finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync)
 {
     if (columns_buffer.size() != 0)
-        writeBlock(header.cloneWithColumns(columns_buffer.releaseColumns()));
+    {
+        auto block = header.cloneWithColumns(columns_buffer.releaseColumns());
+        auto granules_to_write = getGranulesToWrite(index_granularity, block.rows(), getCurrentMark(), /* last_block = */ true);
+        if (!granules_to_write.back().isCompleted())
+        {
+            /// Correct last mark as it should contain exact amount of rows.
+            index_granularity.popMark();
+            index_granularity.appendMark(granules_to_write.back().block_rows);
+        }
+        writeDataBlockPrimaryIndexAndSkipIndices(block, granules_to_write);
+    }
 
 #ifndef NDEBUG
     /// Offsets should be 0, because compressed block is written for every granule.
@@ -269,9 +291,13 @@ static void fillIndexGranularityImpl(
 
 void MergeTreeDataPartWriterCompact::fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block)
 {
+    size_t index_offset = 0;
+    if (index_granularity.getMarksCount() > getCurrentMark())
+        index_offset = index_granularity.getMarkRows(getCurrentMark()) - columns_buffer.size();
+
     fillIndexGranularityImpl(
         index_granularity,
-        getIndexOffset(),
+        index_offset,
         index_granularity_for_block,
         rows_in_block);
 }
@@ -328,4 +354,14 @@ size_t MergeTreeDataPartWriterCompact::ColumnsBuffer::size() const
     return accumulated_columns.at(0)->size();
 }
 
+void MergeTreeDataPartWriterCompact::finish(IMergeTreeDataPart::Checksums & checksums, bool sync)
+{
+    finishDataSerialization(checksums, sync);
+
+    if (settings.rewrite_primary_key)
+        finishPrimaryIndexSerialization(checksums, sync);
+
+    finishSkipIndicesSerialization(checksums, sync);
+}
+
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h
index a6187676dde..2913960fd35 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterCompact.h
@@ -18,16 +18,22 @@ public:
         const MergeTreeWriterSettings & settings,
         const MergeTreeIndexGranularity & index_granularity);
 
-    void write(const Block & block, const IColumn::Permutation * permutation,
-        const Block & primary_key_block, const Block & skip_indexes_block) override;
+    void write(const Block & block, const IColumn::Permutation * permutation) override;
 
-    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
-
-protected:
-    void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) override;
+    void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
 
 private:
-    void writeBlock(const Block & block);
+    /// Finish serialization of the data. Flush rows in buffer to disk, compute checksums.
+    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync);
+
+    void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) override;
+
+    /// Write block of rows into .bin file and marks in .mrk files
+    void writeDataBlock(const Block & block, const Granules & granules);
+
+    /// Write block of rows into .bin file and marks in .mrk files, primary index in .idx file
+    /// and skip indices in their corresponding files.
+    void writeDataBlockPrimaryIndexAndSkipIndices(const Block & block, const Granules & granules);
 
     void addToChecksums(MergeTreeDataPartChecksums & checksums);
 
@@ -78,13 +84,6 @@ private:
     /// marks -> marks_file
     std::unique_ptr<WriteBufferFromFileBase> marks_file;
     HashingWriteBuffer marks;
-
-    /// Write single granule of one column (rows between 2 marks)
-    static void writeColumnSingleGranule(
-        const ColumnWithTypeAndName & column,
-        IDataType::OutputStreamGetter stream_getter,
-        size_t from_row,
-        size_t number_of_rows);
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp
index f0738a1130a..5efcc63b2fc 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.cpp
@@ -19,12 +19,15 @@ MergeTreeDataPartWriterInMemory::MergeTreeDataPartWriterInMemory(
     , part_in_memory(part_) {}
 
 void MergeTreeDataPartWriterInMemory::write(
-    const Block & block, const IColumn::Permutation * permutation,
-    const Block & primary_key_block, const Block & /* skip_indexes_block */)
+    const Block & block, const IColumn::Permutation * permutation)
 {
     if (part_in_memory->block)
         throw Exception("DataPartWriterInMemory supports only one write", ErrorCodes::LOGICAL_ERROR);
 
+    Block primary_key_block;
+    if (settings.rewrite_primary_key)
+        primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), permutation);
+
     Block result_block;
     if (permutation)
     {
@@ -50,6 +53,9 @@ void MergeTreeDataPartWriterInMemory::write(
     if (with_final_mark)
         index_granularity.appendMark(0);
     part_in_memory->block = std::move(result_block);
+
+    if (settings.rewrite_primary_key)
+        calculateAndSerializePrimaryIndex(primary_key_block);
 }
 
 void MergeTreeDataPartWriterInMemory::calculateAndSerializePrimaryIndex(const Block & primary_index_block)
@@ -70,7 +76,7 @@ void MergeTreeDataPartWriterInMemory::calculateAndSerializePrimaryIndex(const Bl
     }
 }
 
-void MergeTreeDataPartWriterInMemory::finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool)
+void MergeTreeDataPartWriterInMemory::finish(IMergeTreeDataPart::Checksums & checksums, bool /* sync */)
 {
     /// If part is empty we still need to initialize block by empty columns.
     if (!part_in_memory->block)
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h
index 990661aad1b..495585ad4d2 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterInMemory.h
@@ -16,14 +16,13 @@ public:
         const MergeTreeWriterSettings & settings_);
 
     /// You can write only one block. In-memory part can be written only at INSERT.
-    void write(const Block & block, const IColumn::Permutation * permutation,
-        const Block & primary_key_block, const Block & skip_indexes_block) override;
+    void write(const Block & block, const IColumn::Permutation * permutation) override;
 
-    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
-
-    void calculateAndSerializePrimaryIndex(const Block & primary_index_block) override;
+    void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
 
 private:
+    void calculateAndSerializePrimaryIndex(const Block & primary_index_block);
+
     DataPartInMemoryPtr part_in_memory;
 };
 
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
index 0ec10f9db76..5190b439809 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp
@@ -70,14 +70,14 @@ MergeTreeDataPartWriterOnDisk::MergeTreeDataPartWriterOnDisk(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
     const StorageMetadataPtr & metadata_snapshot_,
-    const std::vector<MergeTreeIndexPtr> & indices_to_recalc_,
+    const MergeTreeIndices & indices_to_recalc_,
     const String & marks_file_extension_,
     const CompressionCodecPtr & default_codec_,
     const MergeTreeWriterSettings & settings_,
     const MergeTreeIndexGranularity & index_granularity_)
     : IMergeTreeDataPartWriter(data_part_,
-        columns_list_, metadata_snapshot_, indices_to_recalc_,
-        index_granularity_, settings_)
+        columns_list_, metadata_snapshot_, settings_, index_granularity_)
+    , skip_indices(indices_to_recalc_)
     , part_path(data_part_->getFullRelativePath())
     , marks_file_extension(marks_file_extension_)
     , default_codec(default_codec_)
@@ -89,6 +89,10 @@ MergeTreeDataPartWriterOnDisk::MergeTreeDataPartWriterOnDisk(
     auto disk = data_part->volume->getDisk();
     if (!disk->exists(part_path))
         disk->createDirectories(part_path);
+
+    if (settings.rewrite_primary_key)
+        initPrimaryIndex();
+    initSkipIndices();
 }
 
 // Implementation is split into static functions for ability
@@ -129,17 +133,7 @@ static size_t computeIndexGranularityImpl(
     return index_granularity_for_block;
 }
 
-static void fillIndexGranularityImpl(
-    MergeTreeIndexGranularity & index_granularity,
-    size_t index_offset,
-    size_t index_granularity_for_block,
-    size_t rows_in_block)
-{
-    for (size_t current_row = index_offset; current_row < rows_in_block; current_row += index_granularity_for_block)
-        index_granularity.appendMark(index_granularity_for_block);
-}
-
-size_t MergeTreeDataPartWriterOnDisk::computeIndexGranularity(const Block & block)
+size_t MergeTreeDataPartWriterOnDisk::computeIndexGranularity(const Block & block) const
 {
     const auto storage_settings = storage.getSettings();
     return computeIndexGranularityImpl(
@@ -150,15 +144,6 @@ size_t MergeTreeDataPartWriterOnDisk::computeIndexGranularity(const Block & bloc
             settings.can_use_adaptive_granularity);
 }
 
-void MergeTreeDataPartWriterOnDisk::fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block)
-{
-    fillIndexGranularityImpl(
-        index_granularity,
-        getIndexOffset(),
-        index_granularity_for_block,
-        rows_in_block);
-}
-
 void MergeTreeDataPartWriterOnDisk::initPrimaryIndex()
 {
     if (metadata_snapshot->hasPrimaryKey())
@@ -166,8 +151,6 @@ void MergeTreeDataPartWriterOnDisk::initPrimaryIndex()
         index_file_stream = data_part->volume->getDisk()->writeFile(part_path + "primary.idx", DBMS_DEFAULT_BUFFER_SIZE, WriteMode::Rewrite);
         index_stream = std::make_unique<HashingWriteBuffer>(*index_file_stream);
     }
-
-    primary_index_initialized = true;
 }
 
 void MergeTreeDataPartWriterOnDisk::initSkipIndices()
@@ -184,18 +167,12 @@ void MergeTreeDataPartWriterOnDisk::initSkipIndices()
                         default_codec, settings.max_compress_block_size,
                         0, settings.aio_threshold));
         skip_indices_aggregators.push_back(index_helper->createIndexAggregator());
-        skip_index_filling.push_back(0);
+        skip_index_accumulated_marks.push_back(0);
     }
-
-    skip_indices_initialized = true;
 }
 
-void MergeTreeDataPartWriterOnDisk::calculateAndSerializePrimaryIndex(const Block & primary_index_block)
+void MergeTreeDataPartWriterOnDisk::calculateAndSerializePrimaryIndex(const Block & primary_index_block, const Granules & granules_to_write)
 {
-    if (!primary_index_initialized)
-        throw Exception("Primary index is not initialized", ErrorCodes::LOGICAL_ERROR);
-
-    size_t rows = primary_index_block.rows();
     size_t primary_columns_num = primary_index_block.columns();
     if (index_columns.empty())
     {
@@ -215,98 +192,62 @@ void MergeTreeDataPartWriterOnDisk::calculateAndSerializePrimaryIndex(const Bloc
     MemoryTracker::BlockerInThread temporarily_disable_memory_tracker;
 
     /// Write index. The index contains Primary Key value for each `index_granularity` row.
-
-    size_t current_row = getIndexOffset();
-    size_t total_marks = index_granularity.getMarksCount();
-
-    while (index_mark < total_marks && current_row < rows)
+    for (const auto & granule : granules_to_write)
     {
-        if (metadata_snapshot->hasPrimaryKey())
+        if (metadata_snapshot->hasPrimaryKey() && granule.mark_on_start)
         {
             for (size_t j = 0; j < primary_columns_num; ++j)
             {
                 const auto & primary_column = primary_index_block.getByPosition(j);
-                index_columns[j]->insertFrom(*primary_column.column, current_row);
-                primary_column.type->serializeBinary(*primary_column.column, current_row, *index_stream);
+                index_columns[j]->insertFrom(*primary_column.column, granule.start_row);
+                primary_column.type->serializeBinary(*primary_column.column, granule.start_row, *index_stream);
             }
         }
-
-        current_row += index_granularity.getMarkRows(index_mark++);
     }
-
     /// store last index row to write final mark at the end of column
     for (size_t j = 0; j < primary_columns_num; ++j)
         last_block_index_columns[j] = primary_index_block.getByPosition(j).column;
 }
 
-void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block & skip_indexes_block)
+void MergeTreeDataPartWriterOnDisk::calculateAndSerializeSkipIndices(const Block & skip_indexes_block, const Granules & granules_to_write)
 {
-    if (!skip_indices_initialized)
-        throw Exception("Skip indices are not initialized", ErrorCodes::LOGICAL_ERROR);
-
-    size_t rows = skip_indexes_block.rows();
-    size_t skip_index_current_data_mark = 0;
-
     /// Filling and writing skip indices like in MergeTreeDataPartWriterWide::writeColumn
     for (size_t i = 0; i < skip_indices.size(); ++i)
     {
         const auto index_helper = skip_indices[i];
         auto & stream = *skip_indices_streams[i];
-        size_t prev_pos = 0;
-        skip_index_current_data_mark = skip_index_data_mark;
-        while (prev_pos < rows)
+        for (const auto & granule : granules_to_write)
         {
-            UInt64 limit = 0;
-            size_t current_index_offset = getIndexOffset();
-            if (prev_pos == 0 && current_index_offset != 0)
+            if (skip_indices_aggregators[i]->empty() && granule.mark_on_start)
             {
-                limit = current_index_offset;
-            }
-            else if (skip_index_current_data_mark == index_granularity.getMarksCount())
-            {
-                /// Case, when last granule was exceeded and no new granule was created.
-                limit = rows - prev_pos;
-            }
-            else
-            {
-                limit = index_granularity.getMarkRows(skip_index_current_data_mark);
-                if (skip_indices_aggregators[i]->empty())
-                {
-                    skip_indices_aggregators[i] = index_helper->createIndexAggregator();
-                    skip_index_filling[i] = 0;
+                skip_indices_aggregators[i] = index_helper->createIndexAggregator();
 
-                    if (stream.compressed.offset() >= settings.min_compress_block_size)
-                        stream.compressed.next();
+                if (stream.compressed.offset() >= settings.min_compress_block_size)
+                    stream.compressed.next();
 
-                    writeIntBinary(stream.plain_hashing.count(), stream.marks);
-                    writeIntBinary(stream.compressed.offset(), stream.marks);
-                    /// Actually this numbers is redundant, but we have to store them
-                    /// to be compatible with normal .mrk2 file format
-                    if (settings.can_use_adaptive_granularity)
-                        writeIntBinary(1UL, stream.marks);
-                }
-                /// this mark is aggregated, go to the next one
-                skip_index_current_data_mark++;
+                writeIntBinary(stream.plain_hashing.count(), stream.marks);
+                writeIntBinary(stream.compressed.offset(), stream.marks);
+                /// Actually this numbers is redundant, but we have to store them
+                /// to be compatible with normal .mrk2 file format
+                if (settings.can_use_adaptive_granularity)
+                    writeIntBinary(1UL, stream.marks);
             }
 
-            size_t pos = prev_pos;
-            skip_indices_aggregators[i]->update(skip_indexes_block, &pos, limit);
-
-            if (pos == prev_pos + limit)
+            size_t pos = granule.start_row;
+            skip_indices_aggregators[i]->update(skip_indexes_block, &pos, granule.granularity_rows);
+            if (granule.isCompleted())
             {
-                ++skip_index_filling[i];
+                ++skip_index_accumulated_marks[i];
 
                 /// write index if it is filled
-                if (skip_index_filling[i] == index_helper->index.granularity)
+                if (skip_index_accumulated_marks[i] == index_helper->index.granularity)
                 {
                     skip_indices_aggregators[i]->getGranuleAndReset()->serializeBinary(stream.compressed);
-                    skip_index_filling[i] = 0;
+                    skip_index_accumulated_marks[i] = 0;
                 }
             }
-            prev_pos = pos;
         }
     }
-    skip_index_data_mark = skip_index_current_data_mark;
 }
 
 void MergeTreeDataPartWriterOnDisk::finishPrimaryIndexSerialization(
@@ -360,7 +301,16 @@ void MergeTreeDataPartWriterOnDisk::finishSkipIndicesSerialization(
 
     skip_indices_streams.clear();
     skip_indices_aggregators.clear();
-    skip_index_filling.clear();
+    skip_index_accumulated_marks.clear();
+}
+
+Names MergeTreeDataPartWriterOnDisk::getSkipIndicesColumns() const
+{
+    std::unordered_set<String> skip_indexes_column_names_set;
+    for (const auto & index : skip_indices)
+        std::copy(index->index.column_names.cbegin(), index->index.column_names.cend(),
+                  std::inserter(skip_indexes_column_names_set, skip_indexes_column_names_set.end()));
+    return Names(skip_indexes_column_names_set.begin(), skip_indexes_column_names_set.end());
 }
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
index 55da413c9fb..bc09e0e61e1 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h
@@ -14,6 +14,35 @@
 namespace DB
 {
 
+/// Single unit for writing data to disk. Contains information about
+/// amount of rows to write and marks.
+struct Granule
+{
+    /// Start row in block for granule
+    size_t start_row;
+    /// Amount of rows which granule have to contain according to index
+    /// granularity.
+    /// NOTE: Sometimes it's not equal to actually written rows, for example
+    /// for the last granule if it's smaller than computed granularity.
+    size_t granularity_rows;
+    /// Amount of rows from block which have to be written to disk from start_row
+    size_t block_rows;
+    /// Global mark number in the list of all marks (index_granularity) for this part
+    size_t mark_number;
+    /// Should writer write mark for the first of this granule to disk.
+    /// NOTE: Sometimes we don't write mark for the start row, because
+    /// this granule can be continuation of the previous one.
+    bool mark_on_start;
+
+    /// Is this granule contain amout of rows equal to the value in index granularity
+    bool isCompleted() const
+    {
+        return granularity_rows == block_rows;
+    }
+};
+
+/// Multiple granules to write for concrete block.
+using Granules = std::vector<Granule>;
 
 /// Writes data part to disk in different formats.
 /// Calculates and serializes primary and skip indices if needed.
@@ -71,42 +100,46 @@ public:
         const MergeTreeWriterSettings & settings,
         const MergeTreeIndexGranularity & index_granularity);
 
-    void calculateAndSerializePrimaryIndex(const Block & primary_index_block) final;
-    void calculateAndSerializeSkipIndices(const Block & skip_indexes_block) final;
-
-     /// Count index_granularity for block and store in `index_granularity`
-    size_t computeIndexGranularity(const Block & block);
-    virtual void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block);
-
-    void initSkipIndices() final;
-    void initPrimaryIndex() final;
-
-    void finishPrimaryIndexSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync) final;
-    void finishSkipIndicesSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync) final;
-
     void setWrittenOffsetColumns(WrittenOffsetColumns * written_offset_columns_)
     {
         written_offset_columns = written_offset_columns_;
     }
 
 protected:
-    using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
-    using SerializationStates = std::unordered_map<String, SerializationState>;
+     /// Count index_granularity for block and store in `index_granularity`
+    size_t computeIndexGranularity(const Block & block) const;
 
-    String part_path;
+    /// Write primary index according to granules_to_write
+    void calculateAndSerializePrimaryIndex(const Block & primary_index_block, const Granules & granules_to_write);
+    /// Write skip indices according to granules_to_write. Skip indices also have their own marks
+    /// and one skip index granule can contain multiple "normal" marks. So skip indices serialization
+    /// require additional state: skip_indices_aggregators and skip_index_accumulated_marks
+    void calculateAndSerializeSkipIndices(const Block & skip_indexes_block, const Granules & granules_to_write);
+
+    /// Finishes primary index serialization: write final primary index row (if required) and compute checksums
+    void finishPrimaryIndexSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync);
+    /// Finishes skip indices serialization: write all accumulated data to disk and compute checksums
+    void finishSkipIndicesSerialization(MergeTreeData::DataPart::Checksums & checksums, bool sync);
+
+    /// Get global number of the current which we are writing (or going to start to write)
+    size_t getCurrentMark() const { return current_mark; }
+
+    void setCurrentMark(size_t mark) { current_mark = mark; }
+
+    /// Get unique non ordered skip indices column.
+    Names getSkipIndicesColumns() const;
+
+    const MergeTreeIndices skip_indices;
+
+    const String part_path;
     const String marks_file_extension;
-    CompressionCodecPtr default_codec;
+    const CompressionCodecPtr default_codec;
 
-    bool compute_granularity;
-    bool need_finish_last_granule;
-
-    /// Number of marsk in data from which skip indices have to start
-    /// aggregation. I.e. it's data mark number, not skip indices mark.
-    size_t skip_index_data_mark = 0;
+    const bool compute_granularity;
 
     std::vector<StreamPtr> skip_indices_streams;
     MergeTreeIndexAggregators skip_indices_aggregators;
-    std::vector<size_t> skip_index_filling;
+    std::vector<size_t> skip_index_accumulated_marks;
 
     std::unique_ptr<WriteBufferFromFileBase> index_file_stream;
     std::unique_ptr<HashingWriteBuffer> index_stream;
@@ -116,15 +149,18 @@ protected:
     Columns last_block_index_columns;
 
     bool data_written = false;
-    bool primary_index_initialized = false;
-    bool skip_indices_initialized = false;
 
     /// To correctly write Nested elements column-by-column.
     WrittenOffsetColumns * written_offset_columns = nullptr;
 
+    /// Data is already written up to this mark.
+    size_t current_mark = 0;
+
 private:
-    /// Index is already serialized up to this mark.
-    size_t index_mark = 0;
+    void initSkipIndices();
+    void initPrimaryIndex();
+
+    virtual void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) = 0;
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index c15c39e7b7f..1127c3a7dd7 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -1,6 +1,7 @@
 #include <Storages/MergeTree/MergeTreeDataPartWriterWide.h>
 #include <Interpreters/Context.h>
 #include <Compression/CompressionFactory.h>
+#include <Compression/CompressedReadBufferFromFile.h>
 
 namespace DB
 {
@@ -14,6 +15,56 @@ namespace
     constexpr auto DATA_FILE_EXTENSION = ".bin";
 }
 
+namespace
+{
+
+/// Get granules for block using index_granularity
+Granules getGranulesToWrite(const MergeTreeIndexGranularity & index_granularity, size_t block_rows, size_t current_mark, size_t rows_written_in_last_mark)
+{
+    if (current_mark >= index_granularity.getMarksCount())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Request to get granules from mark {} but index granularity size is {}", current_mark, index_granularity.getMarksCount());
+
+    Granules result;
+    size_t current_row = 0;
+    /// When our last mark is not finished yet and we have to write in rows into it
+    if (rows_written_in_last_mark > 0)
+    {
+        size_t rows_left_in_last_mark = index_granularity.getMarkRows(current_mark) - rows_written_in_last_mark;
+        size_t rows_left_in_block = block_rows - current_row;
+        result.emplace_back(Granule{
+            .start_row = current_row,
+            .granularity_rows = rows_left_in_last_mark,
+            .block_rows = std::min(rows_left_in_block, rows_left_in_last_mark),
+            .mark_number = current_mark,
+            .mark_on_start = false, /// Don't mark this granule because we have already marked it
+        });
+        current_row += rows_left_in_last_mark;
+        current_mark++;
+    }
+
+    /// Calculating normal granules for block
+    while (current_row < block_rows)
+    {
+        size_t expected_rows_in_mark = index_granularity.getMarkRows(current_mark);
+        size_t rows_left_in_block  = block_rows - current_row;
+        /// If we have less rows in block than expected in granularity
+        /// save incomplete granule
+        result.emplace_back(Granule{
+            .start_row = current_row,
+            .granularity_rows = expected_rows_in_mark,
+            .block_rows = std::min(rows_left_in_block, expected_rows_in_mark),
+            .mark_number = current_mark,
+            .mark_on_start = true,
+        });
+        current_row += expected_rows_in_mark;
+        current_mark++;
+    }
+
+    return result;
+}
+
+}
+
 MergeTreeDataPartWriterWide::MergeTreeDataPartWriterWide(
     const MergeTreeData::DataPartPtr & data_part_,
     const NamesAndTypesList & columns_list_,
@@ -69,7 +120,7 @@ void MergeTreeDataPartWriterWide::addStreams(
 
 
 IDataType::OutputStreamGetter MergeTreeDataPartWriterWide::createStreamGetter(
-        const String & name, WrittenOffsetColumns & offset_columns)
+        const String & name, WrittenOffsetColumns & offset_columns) const
 {
     return [&, this] (const IDataType::SubstreamPath & substream_path) -> WriteBuffer *
     {
@@ -81,13 +132,34 @@ IDataType::OutputStreamGetter MergeTreeDataPartWriterWide::createStreamGetter(
         if (is_offsets && offset_columns.count(stream_name))
             return nullptr;
 
-        return &column_streams[stream_name]->compressed;
+        return &column_streams.at(stream_name)->compressed;
     };
 }
 
-void MergeTreeDataPartWriterWide::write(const Block & block,
-    const IColumn::Permutation * permutation,
-    const Block & primary_key_block, const Block & skip_indexes_block)
+void MergeTreeDataPartWriterWide::shiftCurrentMark(const Granules & granules_written)
+{
+    auto last_granule = granules_written.back();
+    /// If we didn't finished last granule than we will continue to write it from new block
+    if (!last_granule.isCompleted())
+    {
+        /// Shift forward except last granule
+        setCurrentMark(getCurrentMark() + granules_written.size() - 1);
+        bool still_in_the_same_granule = granules_written.size() == 1;
+        /// We wrote whole block in the same granule, but didn't finished it.
+        /// So add written rows to rows written in last_mark
+        if (still_in_the_same_granule)
+            rows_written_in_last_mark += last_granule.block_rows;
+        else
+            rows_written_in_last_mark = last_granule.block_rows;
+    }
+    else
+    {
+        setCurrentMark(getCurrentMark() + granules_written.size());
+        rows_written_in_last_mark = 0;
+    }
+}
+
+void MergeTreeDataPartWriterWide::write(const Block & block, const IColumn::Permutation * permutation)
 {
     /// Fill index granularity for this block
     /// if it's unknown (in case of insert data or horizontal merge,
@@ -98,7 +170,14 @@ void MergeTreeDataPartWriterWide::write(const Block & block,
         fillIndexGranularity(index_granularity_for_block, block.rows());
     }
 
+    auto granules_to_write = getGranulesToWrite(index_granularity, block.rows(), getCurrentMark(), rows_written_in_last_mark);
+
     auto offset_columns = written_offset_columns ? *written_offset_columns : WrittenOffsetColumns{};
+    Block primary_key_block;
+    if (settings.rewrite_primary_key)
+        primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), permutation);
+
+    Block skip_indexes_block = getBlockAndPermute(block, getSkipIndicesColumns(), permutation);
 
     auto it = columns_list.begin();
     for (size_t i = 0; i < columns_list.size(); ++i, ++it)
@@ -110,25 +189,32 @@ void MergeTreeDataPartWriterWide::write(const Block & block,
             if (primary_key_block.has(it->name))
             {
                 const auto & primary_column = *primary_key_block.getByName(it->name).column;
-                writeColumn(column.name, *column.type, primary_column, offset_columns);
+                writeColumn(column.name, *column.type, primary_column, offset_columns, granules_to_write);
             }
             else if (skip_indexes_block.has(it->name))
             {
                 const auto & index_column = *skip_indexes_block.getByName(it->name).column;
-                writeColumn(column.name, *column.type, index_column, offset_columns);
+                writeColumn(column.name, *column.type, index_column, offset_columns, granules_to_write);
             }
             else
             {
                 /// We rearrange the columns that are not included in the primary key here; Then the result is released - to save RAM.
                 ColumnPtr permuted_column = column.column->permute(*permutation, 0);
-                writeColumn(column.name, *column.type, *permuted_column, offset_columns);
+                writeColumn(column.name, *column.type, *permuted_column, offset_columns, granules_to_write);
             }
         }
         else
         {
-            writeColumn(column.name, *column.type, *column.column, offset_columns);
+            writeColumn(column.name, *column.type, *column.column, offset_columns, granules_to_write);
         }
     }
+
+    if (settings.rewrite_primary_key)
+        calculateAndSerializePrimaryIndex(primary_key_block, granules_to_write);
+
+    calculateAndSerializeSkipIndices(skip_indexes_block, granules_to_write);
+
+    shiftCurrentMark(granules_to_write);
 }
 
 void MergeTreeDataPartWriterWide::writeSingleMark(
@@ -138,44 +224,67 @@ void MergeTreeDataPartWriterWide::writeSingleMark(
     size_t number_of_rows,
     DB::IDataType::SubstreamPath & path)
 {
-    type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
-     {
-         bool is_offsets = !substream_path.empty() && substream_path.back().type == IDataType::Substream::ArraySizes;
-
-         String stream_name = IDataType::getFileNameForStream(name, substream_path);
-
-         /// Don't write offsets more than one time for Nested type.
-         if (is_offsets && offset_columns.count(stream_name))
-             return;
-
-         Stream & stream = *column_streams[stream_name];
-
-         /// There could already be enough data to compress into the new block.
-         if (stream.compressed.offset() >= settings.min_compress_block_size)
-             stream.compressed.next();
-
-         writeIntBinary(stream.plain_hashing.count(), stream.marks);
-         writeIntBinary(stream.compressed.offset(), stream.marks);
-         if (settings.can_use_adaptive_granularity)
-             writeIntBinary(number_of_rows, stream.marks);
-     }, path);
+    StreamsWithMarks marks = getCurrentMarksForColumn(name, type, offset_columns, path);
+    for (const auto & mark : marks)
+        flushMarkToFile(mark, number_of_rows);
 }
 
-size_t MergeTreeDataPartWriterWide::writeSingleGranule(
+void MergeTreeDataPartWriterWide::flushMarkToFile(const StreamNameAndMark & stream_with_mark, size_t rows_in_mark)
+{
+    Stream & stream = *column_streams[stream_with_mark.stream_name];
+    writeIntBinary(stream_with_mark.mark.offset_in_compressed_file, stream.marks);
+    writeIntBinary(stream_with_mark.mark.offset_in_decompressed_block, stream.marks);
+    if (settings.can_use_adaptive_granularity)
+        writeIntBinary(rows_in_mark, stream.marks);
+}
+
+StreamsWithMarks MergeTreeDataPartWriterWide::getCurrentMarksForColumn(
+    const String & name,
+    const IDataType & type,
+    WrittenOffsetColumns & offset_columns,
+    DB::IDataType::SubstreamPath & path)
+{
+    StreamsWithMarks result;
+    type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
+    {
+        bool is_offsets = !substream_path.empty() && substream_path.back().type == IDataType::Substream::ArraySizes;
+
+        String stream_name = IDataType::getFileNameForStream(name, substream_path);
+
+        /// Don't write offsets more than one time for Nested type.
+        if (is_offsets && offset_columns.count(stream_name))
+            return;
+
+        Stream & stream = *column_streams[stream_name];
+
+        /// There could already be enough data to compress into the new block.
+        if (stream.compressed.offset() >= settings.min_compress_block_size)
+            stream.compressed.next();
+
+        StreamNameAndMark stream_with_mark;
+        stream_with_mark.stream_name = stream_name;
+        stream_with_mark.mark.offset_in_compressed_file = stream.plain_hashing.count();
+        stream_with_mark.mark.offset_in_decompressed_block = stream.compressed.offset();
+
+        result.push_back(stream_with_mark);
+    }, path);
+
+    return result;
+}
+
+void MergeTreeDataPartWriterWide::writeSingleGranule(
     const String & name,
     const IDataType & type,
     const IColumn & column,
     WrittenOffsetColumns & offset_columns,
     IDataType::SerializeBinaryBulkStatePtr & serialization_state,
     IDataType::SerializeBinaryBulkSettings & serialize_settings,
-    size_t from_row,
-    size_t number_of_rows,
-    bool write_marks)
+    const Granule & granule)
 {
-    if (write_marks)
-        writeSingleMark(name, type, offset_columns, number_of_rows, serialize_settings.path);
+    if (granule.mark_on_start)
+        writeSingleMark(name, type, offset_columns, granule.granularity_rows, serialize_settings.path);
 
-    type.serializeBinaryBulkWithMultipleStreams(column, from_row, number_of_rows, serialize_settings, serialization_state);
+    type.serializeBinaryBulkWithMultipleStreams(column, granule.start_row, granule.granularity_rows, serialize_settings, serialization_state);
 
     /// So that instead of the marks pointing to the end of the compressed block, there were marks pointing to the beginning of the next one.
     type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
@@ -190,8 +299,6 @@ size_t MergeTreeDataPartWriterWide::writeSingleGranule(
 
         column_streams[stream_name]->compressed.nextIfAtEnd();
     }, serialize_settings.path);
-
-    return from_row + number_of_rows;
 }
 
 /// Column must not be empty. (column.size() !== 0)
@@ -199,9 +306,11 @@ void MergeTreeDataPartWriterWide::writeColumn(
     const String & name,
     const IDataType & type,
     const IColumn & column,
-    WrittenOffsetColumns & offset_columns)
+    WrittenOffsetColumns & offset_columns,
+    const Granules & granules)
 {
     auto [it, inserted] = serialization_states.emplace(name, nullptr);
+
     if (inserted)
     {
         IDataType::SerializeBinaryBulkSettings serialize_settings;
@@ -215,48 +324,20 @@ void MergeTreeDataPartWriterWide::writeColumn(
     serialize_settings.low_cardinality_max_dictionary_size = global_settings.low_cardinality_max_dictionary_size;
     serialize_settings.low_cardinality_use_single_dictionary_for_part = global_settings.low_cardinality_use_single_dictionary_for_part != 0;
 
-    size_t total_rows = column.size();
-    size_t current_row = 0;
-    size_t current_column_mark = getCurrentMark();
-    size_t current_index_offset = getIndexOffset();
-    while (current_row < total_rows)
+    for (const auto & granule : granules)
     {
-        size_t rows_to_write;
-        bool write_marks = true;
-
-        /// If there is `index_offset`, then the first mark goes not immediately, but after this number of rows.
-        if (current_row == 0 && current_index_offset != 0)
-        {
-            write_marks = false;
-            rows_to_write = current_index_offset;
-        }
-        else
-        {
-            if (index_granularity.getMarksCount() <= current_column_mark)
-                throw Exception(
-                    "Incorrect size of index granularity expect mark " + toString(current_column_mark) + " totally have marks " + toString(index_granularity.getMarksCount()),
-                    ErrorCodes::LOGICAL_ERROR);
-
-            rows_to_write = index_granularity.getMarkRows(current_column_mark);
-        }
-
-        if (rows_to_write != 0)
+        if (granule.granularity_rows > 0)
             data_written = true;
 
-        current_row = writeSingleGranule(
-            name,
-            type,
-            column,
-            offset_columns,
-            it->second,
-            serialize_settings,
-            current_row,
-            rows_to_write,
-            write_marks
+        writeSingleGranule(
+           name,
+           type,
+           column,
+           offset_columns,
+           it->second,
+           serialize_settings,
+           granule
         );
-
-        if (write_marks)
-            current_column_mark++;
     }
 
     type.enumerateStreams([&] (const IDataType::SubstreamPath & substream_path, const IDataType & /* substream_type */)
@@ -268,9 +349,93 @@ void MergeTreeDataPartWriterWide::writeColumn(
             offset_columns.insert(stream_name);
         }
     }, serialize_settings.path);
+}
+
+
+void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name, const IDataType & type)
+{
+    if (!type.isValueRepresentedByNumber() || type.haveSubtypes())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot validate column of non fixed type {}", type.getName());
+
+    auto disk = data_part->volume->getDisk();
+    String mrk_path = fullPath(disk, part_path + name + marks_file_extension);
+    String bin_path = fullPath(disk, part_path + name + DATA_FILE_EXTENSION);
+    DB::ReadBufferFromFile mrk_in(mrk_path);
+    DB::CompressedReadBufferFromFile bin_in(bin_path, 0, 0, 0);
+    bool must_be_last = false;
+    UInt64 offset_in_compressed_file = 0;
+    UInt64 offset_in_decompressed_block = 0;
+    UInt64 index_granularity_rows = 0;
+
+    size_t mark_num;
+
+    for (mark_num = 0; !mrk_in.eof(); ++mark_num)
+    {
+        if (mark_num > index_granularity.getMarksCount())
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Incorrect number of marks in memory {}, on disk (at least) {}", index_granularity.getMarksCount(), mark_num + 1);
+
+        DB::readBinary(offset_in_compressed_file, mrk_in);
+        DB::readBinary(offset_in_decompressed_block, mrk_in);
+        if (settings.can_use_adaptive_granularity)
+            DB::readBinary(index_granularity_rows, mrk_in);
+        else
+            index_granularity_rows = storage.getSettings()->index_granularity;
+
+        if (must_be_last)
+        {
+            if (index_granularity_rows != 0)
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "We ran out of binary data but still have non empty mark #{} with rows number {}", mark_num, index_granularity_rows);
+
+            if (!mrk_in.eof())
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "Mark #{} must be last, but we still have some to read", mark_num);
+
+            break;
+        }
+
+        if (index_granularity_rows == 0)
+        {
+            auto column = type.createColumn();
+
+            type.deserializeBinaryBulk(*column, bin_in, 1000000000, 0.0);
+
+            throw Exception(ErrorCodes::LOGICAL_ERROR,
+                        "Still have {} rows in bin stream, last mark #{} index granularity size {}, last rows {}", column->size(), mark_num, index_granularity.getMarksCount(), index_granularity_rows);
+        }
+
+        if (index_granularity_rows != index_granularity.getMarkRows(mark_num))
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), in-memory {}, on disk {}",
+                mark_num, offset_in_compressed_file, offset_in_decompressed_block, index_granularity.getMarkRows(mark_num), index_granularity_rows);
+
+        auto column = type.createColumn();
+
+        type.deserializeBinaryBulk(*column, bin_in, index_granularity_rows, 0.0);
+
+        if (bin_in.eof())
+        {
+            must_be_last = true;
+        }
+        else if (column->size() != index_granularity_rows)
+        {
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), actually in bin file {}, in mrk file {}",
+                mark_num, offset_in_compressed_file, offset_in_decompressed_block, column->size(), index_granularity.getMarkRows(mark_num));
+        }
+    }
+
+    if (!mrk_in.eof())
+        throw Exception(ErrorCodes::LOGICAL_ERROR,
+                        "Still have something in marks stream, last mark #{} index granularity size {}, last rows {}", mark_num, index_granularity.getMarksCount(), index_granularity_rows);
+    if (!bin_in.eof())
+    {
+        auto column = type.createColumn();
+
+        type.deserializeBinaryBulk(*column, bin_in, 1000000000, 0.0);
+
+        throw Exception(ErrorCodes::LOGICAL_ERROR,
+                        "Still have {} rows in bin stream, last mark #{} index granularity size {}, last rows {}", column->size(), mark_num, index_granularity.getMarksCount(), index_granularity_rows);
+    }
 
-    next_mark = current_column_mark;
-    next_index_offset = current_row - total_rows;
 }
 
 void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync)
@@ -294,12 +459,9 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
             }
 
             if (write_final_mark)
-            {
                 writeFinalMark(it->name, it->type, offset_columns, serialize_settings.path);
-            }
         }
     }
-
     for (auto & stream : column_streams)
     {
         stream.second->finalize();
@@ -310,6 +472,24 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
 
     column_streams.clear();
     serialization_states.clear();
+
+#ifndef NDEBUG
+    for (const auto & column : columns_list)
+    {
+        if (column.type->isValueRepresentedByNumber() && !column.type->haveSubtypes())
+            validateColumnOfFixedSize(column.name, *column.type);
+    }
+#endif
+
+}
+
+void MergeTreeDataPartWriterWide::finish(IMergeTreeDataPart::Checksums & checksums, bool sync)
+{
+    finishDataSerialization(checksums, sync);
+    if (settings.rewrite_primary_key)
+        finishPrimaryIndexSerialization(checksums, sync);
+
+    finishSkipIndicesSerialization(checksums, sync);
 }
 
 void MergeTreeDataPartWriterWide::writeFinalMark(
@@ -331,4 +511,30 @@ void MergeTreeDataPartWriterWide::writeFinalMark(
     }, path);
 }
 
+static void fillIndexGranularityImpl(
+    MergeTreeIndexGranularity & index_granularity,
+    size_t index_offset,
+    size_t index_granularity_for_block,
+    size_t rows_in_block)
+{
+    for (size_t current_row = index_offset; current_row < rows_in_block; current_row += index_granularity_for_block)
+        index_granularity.appendMark(index_granularity_for_block);
+}
+
+void MergeTreeDataPartWriterWide::fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block)
+{
+    if (getCurrentMark() < index_granularity.getMarksCount() && getCurrentMark() != index_granularity.getMarksCount() - 1)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Trying to add marks, while current mark {}, but total marks {}", getCurrentMark(), index_granularity.getMarksCount());
+
+    size_t index_offset = 0;
+    if (rows_written_in_last_mark != 0)
+        index_offset = index_granularity.getLastMarkRows() - rows_written_in_last_mark;
+
+    fillIndexGranularityImpl(
+        index_granularity,
+        index_offset,
+        index_granularity_for_block,
+        rows_in_block);
+}
+
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
index aeb686a70a1..cffedd66a32 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
@@ -4,13 +4,19 @@
 namespace DB
 {
 
+struct StreamNameAndMark
+{
+    String stream_name;
+    MarkInCompressedFile mark;
+};
+
+using StreamsWithMarks = std::vector<StreamNameAndMark>;
+using ColumnNameToMark = std::unordered_map<String, StreamsWithMarks>;
+
 /// Writes data part in wide format.
 class MergeTreeDataPartWriterWide : public MergeTreeDataPartWriterOnDisk
 {
 public:
-
-    using ColumnToSize = std::map<std::string, UInt64>;
-
     MergeTreeDataPartWriterWide(
         const MergeTreeData::DataPartPtr & data_part,
         const NamesAndTypesList & columns_list,
@@ -21,14 +27,15 @@ public:
         const MergeTreeWriterSettings & settings,
         const MergeTreeIndexGranularity & index_granularity);
 
-    void write(const Block & block, const IColumn::Permutation * permutation,
-        const Block & primary_key_block, const Block & skip_indexes_block) override;
+    void write(const Block & block, const IColumn::Permutation * permutation) override;
 
-    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync) override;
-
-    IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns);
+    void finish(IMergeTreeDataPart::Checksums & checksums, bool sync) final;
 
 private:
+    /// Finish serialization of data: write final mark if required and compute checksums
+    /// Also validate written data in debug mode
+    void finishDataSerialization(IMergeTreeDataPart::Checksums & checksums, bool sync);
+
     /// Write data of one column.
     /// Return how many marks were written and
     /// how many rows were written for last mark
@@ -36,21 +43,32 @@ private:
         const String & name,
         const IDataType & type,
         const IColumn & column,
-        WrittenOffsetColumns & offset_columns);
+        WrittenOffsetColumns & offset_columns,
+        const Granules & granules);
 
-    /// Write single granule of one column (rows between 2 marks)
-    size_t writeSingleGranule(
+    /// Write single granule of one column.
+    void writeSingleGranule(
         const String & name,
         const IDataType & type,
         const IColumn & column,
         WrittenOffsetColumns & offset_columns,
         IDataType::SerializeBinaryBulkStatePtr & serialization_state,
         IDataType::SerializeBinaryBulkSettings & serialize_settings,
-        size_t from_row,
-        size_t number_of_rows,
-        bool write_marks);
+        const Granule & granule);
 
-    /// Write mark for column
+    /// Take offsets from column and return as MarkInCompressed file with stream name
+    StreamsWithMarks getCurrentMarksForColumn(
+        const String & name,
+        const IDataType & type,
+        WrittenOffsetColumns & offset_columns,
+        DB::IDataType::SubstreamPath & path);
+
+    /// Write mark to disk using stream and rows count
+    void flushMarkToFile(
+        const StreamNameAndMark & stream_with_mark,
+        size_t rows_in_mark);
+
+    /// Write mark for column taking offsets from column stream
     void writeSingleMark(
         const String & name,
         const IDataType & type,
@@ -70,10 +88,30 @@ private:
         const ASTPtr & effective_codec_desc,
         size_t estimated_size);
 
+    /// Method for self check (used in debug-build only). Checks that written
+    /// data and corresponding marks are consistent. Otherwise throws logical
+    /// errors.
+    void validateColumnOfFixedSize(const String & name, const IDataType & type);
+
+    void fillIndexGranularity(size_t index_granularity_for_block, size_t rows_in_block) override;
+
+    /// Use information from just written granules to shift current mark
+    /// in our index_granularity array.
+    void shiftCurrentMark(const Granules & granules_written);
+
+    IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns) const;
+
+    using SerializationState = IDataType::SerializeBinaryBulkStatePtr;
+    using SerializationStates = std::unordered_map<String, SerializationState>;
+
     SerializationStates serialization_states;
 
     using ColumnStreams = std::map<String, StreamPtr>;
     ColumnStreams column_streams;
+
+    /// How many rows we have already written in the current mark.
+    /// More than zero when incoming blocks are smaller then their granularity.
+    size_t rows_written_in_last_mark = 0;
 };
 
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataWriter.cpp b/src/Storages/MergeTree/MergeTreeDataWriter.cpp
index e65a8e22f32..c93d4bceba0 100644
--- a/src/Storages/MergeTree/MergeTreeDataWriter.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataWriter.cpp
@@ -16,6 +16,14 @@
 
 #include <Parsers/queryToString.h>
 
+#include <Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/MergingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/SummingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/AggregatingSortedAlgorithm.h>
+#include <Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h>
+#include <Processors/Merges/Algorithms/GraphiteRollupSortedAlgorithm.h>
+
 namespace ProfileEvents
 {
     extern const Event MergeTreeDataWriterBlocks;
@@ -194,7 +202,74 @@ BlocksWithPartition MergeTreeDataWriter::splitBlockIntoParts(const Block & block
     return result;
 }
 
-MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPartition & block_with_partition, const StorageMetadataPtr & metadata_snapshot)
+Block MergeTreeDataWriter::mergeBlock(const Block & block, SortDescription sort_description, Names & partition_key_columns, IColumn::Permutation *& permutation)
+{
+    size_t block_size = block.rows();
+
+    auto get_merging_algorithm = [&]() -> std::shared_ptr<IMergingAlgorithm>
+    {
+        switch (data.merging_params.mode)
+        {
+            /// There is nothing to merge in single block in ordinary MergeTree
+            case MergeTreeData::MergingParams::Ordinary:
+                return nullptr;
+            case MergeTreeData::MergingParams::Replacing:
+                return std::make_shared<ReplacingSortedAlgorithm>(
+                    block, 1, sort_description, data.merging_params.version_column, block_size + 1);
+            case MergeTreeData::MergingParams::Collapsing:
+                return std::make_shared<CollapsingSortedAlgorithm>(
+                    block, 1, sort_description, data.merging_params.sign_column,
+                    false, block_size + 1, &Poco::Logger::get("MergeTreeBlockOutputStream"));
+            case MergeTreeData::MergingParams::Summing:
+                return std::make_shared<SummingSortedAlgorithm>(
+                    block, 1, sort_description, data.merging_params.columns_to_sum,
+                    partition_key_columns, block_size + 1);
+            case MergeTreeData::MergingParams::Aggregating:
+                return std::make_shared<AggregatingSortedAlgorithm>(block, 1, sort_description, block_size + 1);
+            case MergeTreeData::MergingParams::VersionedCollapsing:
+                return std::make_shared<VersionedCollapsingAlgorithm>(
+                    block, 1, sort_description, data.merging_params.sign_column, block_size + 1);
+            case MergeTreeData::MergingParams::Graphite:
+                return std::make_shared<GraphiteRollupSortedAlgorithm>(
+                    block, 1, sort_description, block_size + 1, data.merging_params.graphite_params, time(nullptr));
+        }
+
+        __builtin_unreachable();
+    };
+
+    auto merging_algorithm = get_merging_algorithm();
+    if (!merging_algorithm)
+        return block;
+
+    Chunk chunk(block.getColumns(), block_size);
+
+    IMergingAlgorithm::Input input;
+    input.set(std::move(chunk));
+    input.permutation = permutation;
+
+    IMergingAlgorithm::Inputs inputs;
+    inputs.push_back(std::move(input));
+    merging_algorithm->initialize(std::move(inputs));
+
+    IMergingAlgorithm::Status status = merging_algorithm->merge();
+
+    /// Check that after first merge merging_algorithm is waiting for data from input 0.
+    if (status.required_source != 0)
+        throw Exception("Logical error: required source after the first merge is not 0.", ErrorCodes::LOGICAL_ERROR);
+
+    status = merging_algorithm->merge();
+
+    /// Check that merge is finished.
+    if (!status.is_finished)
+        throw Exception("Logical error: merge is not finished after the second merge.", ErrorCodes::LOGICAL_ERROR);
+
+    /// Merged Block is sorted and we don't need to use permutation anymore
+    permutation = nullptr;
+
+    return block.cloneWithColumns(status.chunk.getColumns());
+}
+
+MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPartition & block_with_partition, const StorageMetadataPtr & metadata_snapshot, bool optimize_on_insert)
 {
     Block & block = block_with_partition.block;
 
@@ -228,6 +303,38 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPa
     else
         part_name = new_part_info.getPartName();
 
+    /// If we need to calculate some columns to sort.
+    if (metadata_snapshot->hasSortingKey() || metadata_snapshot->hasSecondaryIndices())
+        data.getSortingKeyAndSkipIndicesExpression(metadata_snapshot)->execute(block);
+
+    Names sort_columns = metadata_snapshot->getSortingKeyColumns();
+    SortDescription sort_description;
+    size_t sort_columns_size = sort_columns.size();
+    sort_description.reserve(sort_columns_size);
+
+    for (size_t i = 0; i < sort_columns_size; ++i)
+        sort_description.emplace_back(block.getPositionByName(sort_columns[i]), 1, 1);
+
+    ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocks);
+
+    /// Sort
+    IColumn::Permutation * perm_ptr = nullptr;
+    IColumn::Permutation perm;
+    if (!sort_description.empty())
+    {
+        if (!isAlreadySorted(block, sort_description))
+        {
+            stableGetPermutation(block, sort_description, perm);
+            perm_ptr = &perm;
+        }
+        else
+            ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocksAlreadySorted);
+    }
+
+    Names partition_key_columns = metadata_snapshot->getPartitionKey().column_names;
+    if (optimize_on_insert)
+        block = mergeBlock(block, sort_description, partition_key_columns, perm_ptr);
+
     /// Size of part would not be greater than block.bytes() + epsilon
     size_t expected_size = block.bytes();
 
@@ -274,34 +381,6 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataWriter::writeTempPart(BlockWithPa
             sync_guard.emplace(disk, full_path);
     }
 
-    /// If we need to calculate some columns to sort.
-    if (metadata_snapshot->hasSortingKey() || metadata_snapshot->hasSecondaryIndices())
-        data.getSortingKeyAndSkipIndicesExpression(metadata_snapshot)->execute(block);
-
-    Names sort_columns = metadata_snapshot->getSortingKeyColumns();
-    SortDescription sort_description;
-    size_t sort_columns_size = sort_columns.size();
-    sort_description.reserve(sort_columns_size);
-
-    for (size_t i = 0; i < sort_columns_size; ++i)
-        sort_description.emplace_back(block.getPositionByName(sort_columns[i]), 1, 1);
-
-    ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocks);
-
-    /// Sort
-    IColumn::Permutation * perm_ptr = nullptr;
-    IColumn::Permutation perm;
-    if (!sort_description.empty())
-    {
-        if (!isAlreadySorted(block, sort_description))
-        {
-            stableGetPermutation(block, sort_description, perm);
-            perm_ptr = &perm;
-        }
-        else
-            ProfileEvents::increment(ProfileEvents::MergeTreeDataWriterBlocksAlreadySorted);
-    }
-
     if (metadata_snapshot->hasRowsTTL())
         updateTTL(metadata_snapshot->getRowsTTL(), new_data_part->ttl_infos, new_data_part->ttl_infos.table_ttl, block, true);
 
diff --git a/src/Storages/MergeTree/MergeTreeDataWriter.h b/src/Storages/MergeTree/MergeTreeDataWriter.h
index b4ad936672c..685d1adf947 100644
--- a/src/Storages/MergeTree/MergeTreeDataWriter.h
+++ b/src/Storages/MergeTree/MergeTreeDataWriter.h
@@ -45,7 +45,9 @@ public:
     /** All rows must correspond to same partition.
       * Returns part with unique name starting with 'tmp_', yet not added to MergeTreeData.
       */
-    MergeTreeData::MutableDataPartPtr writeTempPart(BlockWithPartition & block, const StorageMetadataPtr & metadata_snapshot);
+    MergeTreeData::MutableDataPartPtr writeTempPart(BlockWithPartition & block, const StorageMetadataPtr & metadata_snapshot, bool optimize_on_insert);
+
+    Block mergeBlock(const Block & block, SortDescription sort_description, Names & partition_key_columns, IColumn::Permutation *& permutation);
 
 private:
     MergeTreeData & data;
diff --git a/src/Storages/MergeTree/MergeTreeIOSettings.h b/src/Storages/MergeTree/MergeTreeIOSettings.h
index 41d9be07c70..7d8dda0c3ca 100644
--- a/src/Storages/MergeTree/MergeTreeIOSettings.h
+++ b/src/Storages/MergeTree/MergeTreeIOSettings.h
@@ -25,6 +25,7 @@ struct MergeTreeWriterSettings
         const MergeTreeSettingsPtr & storage_settings,
         bool can_use_adaptive_granularity_,
         size_t aio_threshold_,
+        bool rewrite_primary_key_,
         bool blocks_are_granules_size_ = false)
         : min_compress_block_size(
             storage_settings->min_compress_block_size ? storage_settings->min_compress_block_size : global_settings.min_compress_block_size)
@@ -33,6 +34,7 @@ struct MergeTreeWriterSettings
                                                         : global_settings.max_compress_block_size)
         , aio_threshold(aio_threshold_)
         , can_use_adaptive_granularity(can_use_adaptive_granularity_)
+        , rewrite_primary_key(rewrite_primary_key_)
         , blocks_are_granules_size(blocks_are_granules_size_)
     {
     }
@@ -41,6 +43,7 @@ struct MergeTreeWriterSettings
     size_t max_compress_block_size;
     size_t aio_threshold;
     bool can_use_adaptive_granularity;
+    bool rewrite_primary_key;
     bool blocks_are_granules_size;
 
     /// true if we write temporary files during alter.
diff --git a/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp b/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp
index bca0d0cb883..8dc353714cb 100644
--- a/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp
+++ b/src/Storages/MergeTree/MergeTreeIndexGranularity.cpp
@@ -7,6 +7,7 @@ namespace DB
 {
 namespace ErrorCodes
 {
+    extern const int LOGICAL_ERROR;
 }
 
 MergeTreeIndexGranularity::MergeTreeIndexGranularity(const std::vector<size_t> & marks_rows_partial_sums_)
@@ -20,6 +21,17 @@ MergeTreeIndexGranularity::MergeTreeIndexGranularity(size_t marks_count, size_t
 {
 }
 
+/// Rows after mark to next mark
+size_t MergeTreeIndexGranularity::getMarkRows(size_t mark_index) const
+{
+    if (mark_index >= getMarksCount())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Trying to get non existing mark {}, while size is {}", mark_index, getMarksCount());
+    if (mark_index == 0)
+        return marks_rows_partial_sums[0];
+    else
+        return marks_rows_partial_sums[mark_index] - marks_rows_partial_sums[mark_index - 1];
+}
+
 size_t MergeTreeIndexGranularity::getMarkStartingRow(size_t mark_index) const
 {
     if (mark_index == 0)
diff --git a/src/Storages/MergeTree/MergeTreeIndexGranularity.h b/src/Storages/MergeTree/MergeTreeIndexGranularity.h
index 5aefd0f102b..0923a496174 100644
--- a/src/Storages/MergeTree/MergeTreeIndexGranularity.h
+++ b/src/Storages/MergeTree/MergeTreeIndexGranularity.h
@@ -47,13 +47,7 @@ public:
     size_t getMarksCountWithoutFinal() const { return getMarksCount() - hasFinalMark(); }
 
     /// Rows after mark to next mark
-    inline size_t getMarkRows(size_t mark_index) const
-    {
-        if (mark_index == 0)
-            return marks_rows_partial_sums[0];
-        else
-            return marks_rows_partial_sums[mark_index] - marks_rows_partial_sums[mark_index - 1];
-    }
+    size_t getMarkRows(size_t mark_index) const;
 
     /// Return amount of rows before mark
     size_t getMarkStartingRow(size_t mark_index) const;
diff --git a/src/Storages/MergeTree/MergeTreeSettings.cpp b/src/Storages/MergeTree/MergeTreeSettings.cpp
index 15ff62e0aa6..e77668e8900 100644
--- a/src/Storages/MergeTree/MergeTreeSettings.cpp
+++ b/src/Storages/MergeTree/MergeTreeSettings.cpp
@@ -4,6 +4,7 @@
 #include <Parsers/ASTSetQuery.h>
 #include <Parsers/ASTFunction.h>
 #include <Common/Exception.h>
+#include <Core/Settings.h>
 
 
 namespace DB
diff --git a/src/Storages/MergeTree/MergedBlockOutputStream.cpp b/src/Storages/MergeTree/MergedBlockOutputStream.cpp
index 00a4c37c60d..e45aaac6169 100644
--- a/src/Storages/MergeTree/MergedBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/MergedBlockOutputStream.cpp
@@ -51,6 +51,7 @@ MergedBlockOutputStream::MergedBlockOutputStream(
         storage.getSettings(),
         data_part->index_granularity_info.is_adaptive,
         aio_threshold,
+        /* rewrite_primary_key = */ true,
         blocks_are_granules_size);
 
     if (aio_threshold > 0 && !merged_column_to_size.empty())
@@ -67,8 +68,6 @@ MergedBlockOutputStream::MergedBlockOutputStream(
         volume->getDisk()->createDirectories(part_path);
 
     writer = data_part->getWriter(columns_list, metadata_snapshot, skip_indices, default_codec, writer_settings);
-    writer->initPrimaryIndex();
-    writer->initSkipIndices();
 }
 
 /// If data is pre-sorted.
@@ -103,9 +102,7 @@ void MergedBlockOutputStream::writeSuffixAndFinalizePart(
         checksums = std::move(*additional_column_checksums);
 
     /// Finish columns serialization.
-    writer->finishDataSerialization(checksums, sync);
-    writer->finishPrimaryIndexSerialization(checksums, sync);
-    writer->finishSkipIndicesSerialization(checksums, sync);
+    writer->finish(checksums, sync);
 
     NamesAndTypesList part_columns;
     if (!total_columns_list)
@@ -219,19 +216,7 @@ void MergedBlockOutputStream::writeImpl(const Block & block, const IColumn::Perm
     if (!rows)
         return;
 
-    std::unordered_set<String> skip_indexes_column_names_set;
-    for (const auto & index : metadata_snapshot->getSecondaryIndices())
-        std::copy(index.column_names.cbegin(), index.column_names.cend(),
-                std::inserter(skip_indexes_column_names_set, skip_indexes_column_names_set.end()));
-    Names skip_indexes_column_names(skip_indexes_column_names_set.begin(), skip_indexes_column_names_set.end());
-
-    Block primary_key_block = getBlockAndPermute(block, metadata_snapshot->getPrimaryKeyColumns(), permutation);
-    Block skip_indexes_block = getBlockAndPermute(block, skip_indexes_column_names, permutation);
-
-    writer->write(block, permutation, primary_key_block, skip_indexes_block);
-    writer->calculateAndSerializeSkipIndices(skip_indexes_block);
-    writer->calculateAndSerializePrimaryIndex(primary_key_block);
-    writer->next();
+    writer->write(block, permutation);
 
     rows_count += rows;
 }
diff --git a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp
index 12b14d13656..87f67ee92a5 100644
--- a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp
+++ b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.cpp
@@ -14,7 +14,7 @@ MergedColumnOnlyOutputStream::MergedColumnOnlyOutputStream(
     const StorageMetadataPtr & metadata_snapshot_,
     const Block & header_,
     CompressionCodecPtr default_codec,
-    const std::vector<MergeTreeIndexPtr> & indices_to_recalc,
+    const MergeTreeIndices & indices_to_recalc,
     WrittenOffsetColumns * offset_columns_,
     const MergeTreeIndexGranularity & index_granularity,
     const MergeTreeIndexGranularityInfo * index_granularity_info)
@@ -28,7 +28,8 @@ MergedColumnOnlyOutputStream::MergedColumnOnlyOutputStream(
         global_settings,
         storage_settings,
         index_granularity_info ? index_granularity_info->is_adaptive : data_part->storage.canUseAdaptiveGranularity(),
-        global_settings.min_bytes_to_use_direct_io);
+        global_settings.min_bytes_to_use_direct_io,
+        /* rewrite_primary_key = */false);
 
     writer = data_part->getWriter(
         header.getNamesAndTypesList(),
@@ -43,25 +44,14 @@ MergedColumnOnlyOutputStream::MergedColumnOnlyOutputStream(
         throw Exception("MergedColumnOnlyOutputStream supports only parts stored on disk", ErrorCodes::NOT_IMPLEMENTED);
 
     writer_on_disk->setWrittenOffsetColumns(offset_columns_);
-    writer_on_disk->initSkipIndices();
 }
 
 void MergedColumnOnlyOutputStream::write(const Block & block)
 {
-    std::unordered_set<String> skip_indexes_column_names_set;
-    for (const auto & index : writer->getSkipIndices())
-        std::copy(index->index.column_names.cbegin(), index->index.column_names.cend(),
-                  std::inserter(skip_indexes_column_names_set, skip_indexes_column_names_set.end()));
-    Names skip_indexes_column_names(skip_indexes_column_names_set.begin(), skip_indexes_column_names_set.end());
-
-    Block skip_indexes_block = getBlockAndPermute(block, skip_indexes_column_names, nullptr);
-
     if (!block.rows())
         return;
 
-    writer->write(block);
-    writer->calculateAndSerializeSkipIndices(skip_indexes_block);
-    writer->next();
+    writer->write(block, nullptr);
 }
 
 void MergedColumnOnlyOutputStream::writeSuffix()
@@ -77,8 +67,7 @@ MergedColumnOnlyOutputStream::writeSuffixAndGetChecksums(
 {
     /// Finish columns serialization.
     MergeTreeData::DataPart::Checksums checksums;
-    writer->finishDataSerialization(checksums, sync);
-    writer->finishSkipIndicesSerialization(checksums, sync);
+    writer->finish(checksums, sync);
 
     auto columns = new_part->getColumns();
 
diff --git a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h
index 507a964ede0..393058dc82e 100644
--- a/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h
+++ b/src/Storages/MergeTree/MergedColumnOnlyOutputStream.h
@@ -18,7 +18,7 @@ public:
         const StorageMetadataPtr & metadata_snapshot_,
         const Block & header_,
         CompressionCodecPtr default_codec_,
-        const std::vector<MergeTreeIndexPtr> & indices_to_recalc_,
+        const MergeTreeIndices & indices_to_recalc_,
         WrittenOffsetColumns * offset_columns_ = nullptr,
         const MergeTreeIndexGranularity & index_granularity = {},
         const MergeTreeIndexGranularityInfo * index_granularity_info_ = nullptr);
diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp
index 9bf1de5d03c..54248e6032c 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.cpp
@@ -40,7 +40,8 @@ ReplicatedMergeTreeBlockOutputStream::ReplicatedMergeTreeBlockOutputStream(
     size_t quorum_timeout_ms_,
     size_t max_parts_per_block_,
     bool quorum_parallel_,
-    bool deduplicate_)
+    bool deduplicate_,
+    bool optimize_on_insert_)
     : storage(storage_)
     , metadata_snapshot(metadata_snapshot_)
     , quorum(quorum_)
@@ -49,6 +50,7 @@ ReplicatedMergeTreeBlockOutputStream::ReplicatedMergeTreeBlockOutputStream(
     , quorum_parallel(quorum_parallel_)
     , deduplicate(deduplicate_)
     , log(&Poco::Logger::get(storage.getLogName() + " (Replicated OutputStream)"))
+    , optimize_on_insert(optimize_on_insert_)
 {
     /// The quorum value `1` has the same meaning as if it is disabled.
     if (quorum == 1)
@@ -121,8 +123,6 @@ void ReplicatedMergeTreeBlockOutputStream::write(const Block & block)
 {
     last_block_is_duplicate = false;
 
-    storage.delayInsertOrThrowIfNeeded(&storage.partial_shutdown_event);
-
     auto zookeeper = storage.getZooKeeper();
     assertSessionIsNotExpired(zookeeper);
 
@@ -142,7 +142,7 @@ void ReplicatedMergeTreeBlockOutputStream::write(const Block & block)
 
         /// Write part to the filesystem under temporary name. Calculate a checksum.
 
-        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot);
+        MergeTreeData::MutableDataPartPtr part = storage.writer.writeTempPart(current_block, metadata_snapshot, optimize_on_insert);
 
         String block_id;
 
@@ -491,6 +491,8 @@ void ReplicatedMergeTreeBlockOutputStream::commitPart(
                 if (!zookeeper->tryGet(quorum_info.status_path, value, nullptr, event))
                     break;
 
+                LOG_TRACE(log, "Quorum node {} still exists, will wait for updates", quorum_info.status_path);
+
                 ReplicatedMergeTreeQuorumEntry quorum_entry(value);
 
                 /// If the node has time to disappear, and then appear again for the next insert.
@@ -499,6 +501,8 @@ void ReplicatedMergeTreeBlockOutputStream::commitPart(
 
                 if (!event->tryWait(quorum_timeout_ms))
                     throw Exception("Timeout while waiting for quorum", ErrorCodes::TIMEOUT_EXCEEDED);
+
+                LOG_TRACE(log, "Quorum {} updated, will check quorum node still exists", quorum_info.status_path);
             }
 
             /// And what if it is possible that the current replica at this time has ceased to be active
@@ -522,7 +526,9 @@ void ReplicatedMergeTreeBlockOutputStream::commitPart(
 
 void ReplicatedMergeTreeBlockOutputStream::writePrefix()
 {
-    storage.throwInsertIfNeeded();
+    /// Only check "too many parts" before write,
+    /// because interrupting long-running INSERT query in the middle is not convenient for users.
+    storage.delayInsertOrThrowIfNeeded(&storage.partial_shutdown_event);
 }
 
 
diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h
index 97c094c1128..3ac2c4bcfcb 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeBlockOutputStream.h
@@ -29,7 +29,8 @@ public:
         size_t quorum_timeout_ms_,
         size_t max_parts_per_block_,
         bool quorum_parallel_,
-        bool deduplicate_);
+        bool deduplicate_,
+        bool optimize_on_insert);
 
     Block getHeader() const override;
     void writePrefix() override;
@@ -71,6 +72,8 @@ private:
 
     using Logger = Poco::Logger;
     Poco::Logger * log;
+
+    bool optimize_on_insert;
 };
 
 }
diff --git a/src/Storages/MergeTree/SimpleMergeSelector.h b/src/Storages/MergeTree/SimpleMergeSelector.h
index fe57c40320a..4f277ad74cd 100644
--- a/src/Storages/MergeTree/SimpleMergeSelector.h
+++ b/src/Storages/MergeTree/SimpleMergeSelector.h
@@ -3,6 +3,83 @@
 #include <Storages/MergeTree/MergeSelector.h>
 
 
+/**
+We have a set of data parts that is dynamically changing - new data parts are added and there is background merging process.
+Background merging process periodically selects continuous range of data parts to merge.
+
+It tries to optimize the following metrics:
+1. Write amplification: total amount of data written on disk (source data + merges) relative to the amount of source data.
+It can be also considered as the total amount of work for merges.
+2. The number of data parts in the set at the random moment of time (average + quantiles).
+
+Also taking the following considerations:
+1. Older data parts should be merged less frequently than newer data parts.
+2. Larger data parts should be merged less frequently than smaller data parts.
+3. If no new parts arrive, we should continue to merge existing data parts to eventually optimize the table.
+4. Never allow too many parts, because it will slow down SELECT queries significantly.
+5. Multiple background merges can run concurrently but not too many.
+
+It is not possible to optimize both metrics, because they contradict to each other.
+To lower the number of parts we can merge eagerly but write amplification will increase.
+Then we need some balance between optimization of these two metrics.
+
+But some optimizations may improve both metrics.
+
+For example, we can look at the "merge tree" - the tree of data parts that were merged.
+If the tree is perfectly balanced then its depth is proportonal to the log(data size),
+the total amount of work is proportional to data_size * log(data_size)
+and the write amplification is proportional to log(data_size).
+If it's not balanced (e.g. every new data part is always merged with existing data parts),
+its depth is proportional to the data size, total amount of work is proportional to data_size^2.
+
+We can also control the "base of the logarithm" - you can look it as the number of data parts
+that are merged at once (the tree "arity"). But as the data parts are of different size, we should generalize it:
+calculate the ratio between total size of merged parts to the size of the largest part participated in merge.
+For example, if we merge 4 parts of size 5, 3, 1, 1 - then "base" will be 2 (10 / 5).
+
+Base of the logarithm (simply called `base` in `SimpleMergeSelector`) is the main knob to control the write amplification.
+The more it is, the less is write amplification but we will have more data parts on average.
+
+To fit all the considerations, we also adjust `base` depending on total parts count,
+parts size and parts age, with linear interpolation (then `base` is not a constant but a function of multiple variables,
+looking like a section of hyperplane).
+
+
+Then we apply the algorithm to select the optimal range of data parts to merge.
+There is a proof that this algorithm is optimal if we look in the future only by single step.
+
+The best range of data parts is selected.
+
+We also apply some tunes:
+- there is a fixed const of merging small parts (that is added to the size of data part before all estimations);
+- there are some heuristics to "stick" ranges to large data parts.
+
+It's still unclear if this algorithm is good or optimal at all. It's unclear if this algorithm is using the optimal coefficients.
+
+To test and optimize SimpleMergeSelector, we apply the following methods:
+- insert/merge simulator: a model simulating parts insertion and merging;
+merge selecting algorithm is applied and the relevant metrics are calculated to allow to tune the algorithm;
+- insert/merge simulator on real `system.part_log` from production - it gives realistic information about inserted data parts:
+their sizes, at what time intervals they are inserted;
+
+There is a research thesis dedicated to optimization of merge algorithm:
+https://presentations.clickhouse.tech/hse_2019/merge_algorithm.pptx
+
+This work made attempt to variate the coefficients in SimpleMergeSelector and to solve the optimization task:
+maybe some change in coefficients will give a clear win on all metrics. Surprisingly enough, it has found
+that our selection of coefficients is near optimal. It has found slightly more optimal coefficients,
+but I decided not to use them, because the representativeness of the test data is in question.
+
+This work did not make any attempt to propose any other algorithm.
+This work did not make any attempt to analyze the task with analytical methods.
+That's why I still believe that there are many opportunities to optimize the merge selection algorithm.
+
+Please do not mix the task with a similar task in other LSM-based systems (like RocksDB).
+Their problem statement is subtly different. Our set of data parts is consisted of data parts
+that are completely independent in stored data. Ranges of primary keys in data parts can intersect.
+When doing SELECT we read from all data parts. INSERTed data parts comes with unknown size...
+*/
+
 namespace DB
 {
 
diff --git a/src/Storages/RabbitMQ/StorageRabbitMQ.cpp b/src/Storages/RabbitMQ/StorageRabbitMQ.cpp
index 1c3b1bbd99c..f41c4805d24 100644
--- a/src/Storages/RabbitMQ/StorageRabbitMQ.cpp
+++ b/src/Storages/RabbitMQ/StorageRabbitMQ.cpp
@@ -9,6 +9,7 @@
 #include <DataTypes/DataTypeString.h>
 #include <Interpreters/InterpreterInsertQuery.h>
 #include <Interpreters/evaluateConstantExpression.h>
+#include <Interpreters/Context.h>
 #include <Parsers/ASTCreateQuery.h>
 #include <Parsers/ASTExpressionList.h>
 #include <Parsers/ASTInsertQuery.h>
diff --git a/src/Storages/RabbitMQ/StorageRabbitMQ.h b/src/Storages/RabbitMQ/StorageRabbitMQ.h
index f4bb3215b55..a46da6072af 100644
--- a/src/Storages/RabbitMQ/StorageRabbitMQ.h
+++ b/src/Storages/RabbitMQ/StorageRabbitMQ.h
@@ -2,7 +2,6 @@
 
 #include <Core/BackgroundSchedulePool.h>
 #include <Storages/IStorage.h>
-#include <Interpreters/Context.h>
 #include <Poco/Semaphore.h>
 #include <ext/shared_ptr_helper.h>
 #include <mutex>
@@ -19,6 +18,8 @@
 namespace DB
 {
 
+class Context;
+
 using ChannelPtr = std::shared_ptr<AMQP::TcpChannel>;
 
 class StorageRabbitMQ final: public ext::shared_ptr_helper<StorageRabbitMQ>, public IStorage
diff --git a/src/Storages/SetSettings.h b/src/Storages/SetSettings.h
index c5926e409c2..8e4706f2b9e 100644
--- a/src/Storages/SetSettings.h
+++ b/src/Storages/SetSettings.h
@@ -10,7 +10,8 @@ class ASTStorage;
 
 
 #define SET_RELATED_SETTINGS(M) \
-    M(Bool, persistent, true, "Disable setting to avoid the overhead of writing to disk for StorageSet", 0)
+    M(Bool, persistent, true, "Disable setting to avoid the overhead of writing to disk for StorageSet", 0) \
+    M(String, disk, "default", "Name of the disk used to persist set data", 0)
 
 #define LIST_OF_SET_SETTINGS(M) \
     SET_RELATED_SETTINGS(M) \
diff --git a/src/Storages/StorageDictionary.cpp b/src/Storages/StorageDictionary.cpp
index 4896ace4e7a..32fe7b4c026 100644
--- a/src/Storages/StorageDictionary.cpp
+++ b/src/Storages/StorageDictionary.cpp
@@ -121,9 +121,9 @@ StorageDictionary::StorageDictionary(
 void StorageDictionary::checkTableCanBeDropped() const
 {
     if (location == Location::SameDatabaseAndNameAsDictionary)
-        throw Exception("Cannot detach dictionary " + backQuote(dictionary_name) + " as table, use DETACH DICTIONARY query", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
+        throw Exception("Cannot drop/detach dictionary " + backQuote(dictionary_name) + " as table, use DROP DICTIONARY or DETACH DICTIONARY query instead", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
     if (location == Location::DictionaryDatabase)
-        throw Exception("Cannot detach table " + getStorageID().getFullTableName() + " from a database with DICTIONARY engine", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
+        throw Exception("Cannot drop/detach table " + getStorageID().getFullTableName() + " from a database with DICTIONARY engine", ErrorCodes::CANNOT_DETACH_DICTIONARY_AS_TABLE);
 }
 
 void StorageDictionary::checkTableCanBeDetached() const
diff --git a/src/Storages/StorageJoin.cpp b/src/Storages/StorageJoin.cpp
index 728f16ec6d8..8d4f0b3b3be 100644
--- a/src/Storages/StorageJoin.cpp
+++ b/src/Storages/StorageJoin.cpp
@@ -8,6 +8,7 @@
 #include <Core/ColumnNumbers.h>
 #include <DataStreams/IBlockInputStream.h>
 #include <DataTypes/NestedUtils.h>
+#include <Disks/IDisk.h>
 #include <Interpreters/joinDispatch.h>
 #include <Interpreters/TableJoin.h>
 #include <Interpreters/castColumn.h>
@@ -35,6 +36,7 @@ namespace ErrorCodes
 }
 
 StorageJoin::StorageJoin(
+    DiskPtr disk_,
     const String & relative_path_,
     const StorageID & table_id_,
     const Names & key_names_,
@@ -45,9 +47,8 @@ StorageJoin::StorageJoin(
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
     bool overwrite_,
-    const Context & context_,
     bool persistent_)
-    : StorageSetOrJoinBase{relative_path_, table_id_, columns_, constraints_, context_, persistent_}
+    : StorageSetOrJoinBase{disk_, relative_path_, table_id_, columns_, constraints_, persistent_}
     , key_names(key_names_)
     , use_nulls(use_nulls_)
     , limits(limits_)
@@ -69,9 +70,9 @@ StorageJoin::StorageJoin(
 void StorageJoin::truncate(
     const ASTPtr &, const StorageMetadataPtr & metadata_snapshot, const Context &, TableExclusiveLockHolder&)
 {
-    Poco::File(path).remove(true);
-    Poco::File(path).createDirectories();
-    Poco::File(path + "tmp/").createDirectories();
+    disk->removeRecursive(path);
+    disk->createDirectories(path);
+    disk->createDirectories(path + "tmp/");
 
     increment = 0;
     join = std::make_shared<HashJoin>(table_join, metadata_snapshot->getSampleBlock().sortColumns(), overwrite);
@@ -124,6 +125,7 @@ void registerStorageJoin(StorageFactory & factory)
         auto join_any_take_last_row = settings.join_any_take_last_row;
         auto old_any_join = settings.any_join_distinct_right_table_keys;
         bool persistent = true;
+        String disk_name = "default";
 
         if (args.storage_def && args.storage_def->settings)
         {
@@ -141,6 +143,8 @@ void registerStorageJoin(StorageFactory & factory)
                     join_any_take_last_row = setting.value;
                 else if (setting.name == "any_join_distinct_right_table_keys")
                     old_any_join = setting.value;
+                else if (setting.name == "disk")
+                    disk_name = setting.value.get<String>();
                 else if (setting.name == "persistent")
                 {
                     auto join_settings = std::make_unique<JoinSettings>();
@@ -148,12 +152,12 @@ void registerStorageJoin(StorageFactory & factory)
                     persistent = join_settings->persistent;
                 }
                 else
-                    throw Exception(
-                        "Unknown setting " + setting.name + " for storage " + args.engine_name,
-                        ErrorCodes::BAD_ARGUMENTS);
+                    throw Exception("Unknown setting " + setting.name + " for storage " + args.engine_name, ErrorCodes::BAD_ARGUMENTS);
             }
         }
 
+        DiskPtr disk = args.context.getDisk(disk_name);
+
         if (engine_args.size() < 3)
             throw Exception(
                 "Storage Join requires at least 3 parameters: Join(ANY|ALL|SEMI|ANTI, LEFT|INNER|RIGHT, keys...).",
@@ -219,6 +223,7 @@ void registerStorageJoin(StorageFactory & factory)
         }
 
         return StorageJoin::create(
+            disk,
             args.relative_data_path,
             args.table_id,
             key_names,
@@ -229,7 +234,6 @@ void registerStorageJoin(StorageFactory & factory)
             args.columns,
             args.constraints,
             join_any_take_last_row,
-            args.context,
             persistent);
     };
 
diff --git a/src/Storages/StorageJoin.h b/src/Storages/StorageJoin.h
index df250b89cb1..c453c036b65 100644
--- a/src/Storages/StorageJoin.h
+++ b/src/Storages/StorageJoin.h
@@ -67,6 +67,7 @@ private:
 
 protected:
     StorageJoin(
+        DiskPtr disk_,
         const String & relative_path_,
         const StorageID & table_id_,
         const Names & key_names_,
@@ -76,7 +77,6 @@ protected:
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
         bool overwrite,
-        const Context & context_,
         bool persistent_);
 };
 
diff --git a/src/Storages/StorageMaterializeMySQL.cpp b/src/Storages/StorageMaterializeMySQL.cpp
index 9be285adc69..721221e3fdc 100644
--- a/src/Storages/StorageMaterializeMySQL.cpp
+++ b/src/Storages/StorageMaterializeMySQL.cpp
@@ -21,12 +21,13 @@
 #include <Processors/Pipe.h>
 #include <Processors/Transforms/FilterTransform.h>
 
+#include <Databases/MySQL/DatabaseMaterializeMySQL.h>
 #include <Storages/SelectQueryInfo.h>
 
 namespace DB
 {
 
-StorageMaterializeMySQL::StorageMaterializeMySQL(const StoragePtr & nested_storage_, const DatabaseMaterializeMySQL * database_)
+StorageMaterializeMySQL::StorageMaterializeMySQL(const StoragePtr & nested_storage_, const IDatabase * database_)
     : StorageProxy(nested_storage_->getStorageID()), nested_storage(nested_storage_), database(database_)
 {
     auto nested_memory_metadata = nested_storage->getInMemoryMetadata();
@@ -45,7 +46,7 @@ Pipe StorageMaterializeMySQL::read(
     unsigned int num_streams)
 {
     /// If the background synchronization thread has exception.
-    database->rethrowExceptionIfNeed();
+    rethrowSyncExceptionIfNeed(database);
 
     NameSet column_names_set = NameSet(column_names.begin(), column_names.end());
     auto lock = nested_storage->lockForShare(context.getCurrentQueryId(), context.getSettingsRef().lock_acquire_timeout);
@@ -106,7 +107,7 @@ Pipe StorageMaterializeMySQL::read(
 NamesAndTypesList StorageMaterializeMySQL::getVirtuals() const
 {
     /// If the background synchronization thread has exception.
-    database->rethrowExceptionIfNeed();
+    rethrowSyncExceptionIfNeed(database);
     return nested_storage->getVirtuals();
 }
 
diff --git a/src/Storages/StorageMaterializeMySQL.h b/src/Storages/StorageMaterializeMySQL.h
index ff86c7abdc2..ea90c1ffc9e 100644
--- a/src/Storages/StorageMaterializeMySQL.h
+++ b/src/Storages/StorageMaterializeMySQL.h
@@ -5,7 +5,6 @@
 #if USE_MYSQL
 
 #include <Storages/StorageProxy.h>
-#include <Databases/MySQL/DatabaseMaterializeMySQL.h>
 
 namespace DB
 {
@@ -21,7 +20,7 @@ class StorageMaterializeMySQL final : public ext::shared_ptr_helper<StorageMater
 public:
     String getName() const override { return "MaterializeMySQL"; }
 
-    StorageMaterializeMySQL(const StoragePtr & nested_storage_, const DatabaseMaterializeMySQL * database_);
+    StorageMaterializeMySQL(const StoragePtr & nested_storage_, const IDatabase * database_);
 
     Pipe read(
         const Names & column_names, const StorageMetadataPtr & metadata_snapshot, SelectQueryInfo & query_info,
@@ -32,15 +31,18 @@ public:
     NamesAndTypesList getVirtuals() const override;
     ColumnSizeByName getColumnSizes() const override;
 
-private:
     StoragePtr getNested() const override { return nested_storage; }
+
+    void drop() override { nested_storage->drop(); }
+
+private:
     [[noreturn]] void throwNotAllowed() const
     {
         throw Exception("This method is not allowed for MaterializeMySQL", ErrorCodes::NOT_IMPLEMENTED);
     }
 
     StoragePtr nested_storage;
-    const DatabaseMaterializeMySQL * database;
+    const IDatabase * database;
 };
 
 }
diff --git a/src/Storages/StorageMemory.cpp b/src/Storages/StorageMemory.cpp
index 3c45ef84dd6..93f00206e6b 100644
--- a/src/Storages/StorageMemory.cpp
+++ b/src/Storages/StorageMemory.cpp
@@ -23,7 +23,7 @@ namespace ErrorCodes
 
 class MemorySource : public SourceWithProgress
 {
-    using InitializerFunc = std::function<void(BlocksList::const_iterator &, size_t &, std::shared_ptr<const BlocksList> &)>;
+    using InitializerFunc = std::function<void(std::shared_ptr<const Blocks> &)>;
 public:
     /// Blocks are stored in std::list which may be appended in another thread.
     /// We use pointer to the beginning of the list and its current size.
@@ -32,17 +32,15 @@ public:
 
     MemorySource(
         Names column_names_,
-        BlocksList::const_iterator first_,
-        size_t num_blocks_,
         const StorageMemory & storage,
         const StorageMetadataPtr & metadata_snapshot,
-        std::shared_ptr<const BlocksList> data_,
-        InitializerFunc initializer_func_ = [](BlocksList::const_iterator &, size_t &, std::shared_ptr<const BlocksList> &) {})
+        std::shared_ptr<const Blocks> data_,
+        std::shared_ptr<std::atomic<size_t>> parallel_execution_index_,
+        InitializerFunc initializer_func_ = {})
         : SourceWithProgress(metadata_snapshot->getSampleBlockForColumns(column_names_, storage.getVirtuals(), storage.getStorageID()))
         , column_names(std::move(column_names_))
-        , current_it(first_)
-        , num_blocks(num_blocks_)
         , data(data_)
+        , parallel_execution_index(parallel_execution_index_)
         , initializer_func(std::move(initializer_func_))
     {
     }
@@ -52,16 +50,20 @@ public:
 protected:
     Chunk generate() override
     {
-        if (!postponed_init_done)
+        if (initializer_func)
         {
-            initializer_func(current_it, num_blocks, data);
-            postponed_init_done = true;
+            initializer_func(data);
+            initializer_func = {};
         }
 
-        if (current_block_idx == num_blocks)
-            return {};
+        size_t current_index = getAndIncrementExecutionIndex();
 
-        const Block & src = *current_it;
+        if (current_index >= data->size())
+        {
+            return {};
+        }
+
+        const Block & src = (*data)[current_index];
         Columns columns;
         columns.reserve(column_names.size());
 
@@ -69,20 +71,26 @@ protected:
         for (const auto & name : column_names)
             columns.push_back(src.getByName(name).column);
 
-        if (++current_block_idx < num_blocks)
-            ++current_it;
-
         return Chunk(std::move(columns), src.rows());
     }
 
 private:
-    const Names column_names;
-    BlocksList::const_iterator current_it;
-    size_t num_blocks;
-    size_t current_block_idx = 0;
+    size_t getAndIncrementExecutionIndex()
+    {
+        if (parallel_execution_index)
+        {
+            return (*parallel_execution_index)++;
+        }
+        else
+        {
+            return execution_index++;
+        }
+    }
 
-    std::shared_ptr<const BlocksList> data;
-    bool postponed_init_done = false;
+    const Names column_names;
+    size_t execution_index = 0;
+    std::shared_ptr<const Blocks> data;
+    std::shared_ptr<std::atomic<size_t>> parallel_execution_index;
     InitializerFunc initializer_func;
 };
 
@@ -107,7 +115,7 @@ public:
         metadata_snapshot->check(block, true);
         {
             std::lock_guard lock(storage.mutex);
-            auto new_data = std::make_unique<BlocksList>(*(storage.data.get()));
+            auto new_data = std::make_unique<Blocks>(*(storage.data.get()));
             new_data->push_back(block);
             storage.data.set(std::move(new_data));
 
@@ -123,7 +131,7 @@ private:
 
 
 StorageMemory::StorageMemory(const StorageID & table_id_, ColumnsDescription columns_description_, ConstraintsDescription constraints_)
-    : IStorage(table_id_), data(std::make_unique<const BlocksList>())
+    : IStorage(table_id_), data(std::make_unique<const Blocks>())
 {
     StorageInMemoryMetadata storage_metadata;
     storage_metadata.setColumns(std::move(columns_description_));
@@ -155,21 +163,17 @@ Pipe StorageMemory::read(
 
         return Pipe(std::make_shared<MemorySource>(
             column_names,
-            data.get()->end(),
-            0,
             *this,
             metadata_snapshot,
-            data.get(),
-            [this](BlocksList::const_iterator & current_it, size_t & num_blocks, std::shared_ptr<const BlocksList> & current_data)
+            nullptr /* data */,
+            nullptr /* parallel execution index */,
+            [this](std::shared_ptr<const Blocks> & data_to_initialize)
             {
-                current_data = data.get();
-                current_it = current_data->begin();
-                num_blocks = current_data->size();
+                data_to_initialize = data.get();
             }));
     }
 
     auto current_data = data.get();
-
     size_t size = current_data->size();
 
     if (num_streams > size)
@@ -177,23 +181,11 @@ Pipe StorageMemory::read(
 
     Pipes pipes;
 
-    BlocksList::const_iterator it = current_data->begin();
+    auto parallel_execution_index = std::make_shared<std::atomic<size_t>>(0);
 
-    size_t offset = 0;
     for (size_t stream = 0; stream < num_streams; ++stream)
     {
-        size_t next_offset = (stream + 1) * size / num_streams;
-        size_t num_blocks = next_offset - offset;
-
-        assert(num_blocks > 0);
-
-        pipes.emplace_back(std::make_shared<MemorySource>(column_names, it, num_blocks, *this, metadata_snapshot, current_data));
-
-        while (offset < next_offset)
-        {
-            ++it;
-            ++offset;
-        }
+        pipes.emplace_back(std::make_shared<MemorySource>(column_names, *this, metadata_snapshot, current_data, parallel_execution_index));
     }
 
     return Pipe::unitePipes(std::move(pipes));
@@ -208,7 +200,7 @@ BlockOutputStreamPtr StorageMemory::write(const ASTPtr & /*query*/, const Storag
 
 void StorageMemory::drop()
 {
-    data.set(std::make_unique<BlocksList>());
+    data.set(std::make_unique<Blocks>());
     total_size_bytes.store(0, std::memory_order_relaxed);
     total_size_rows.store(0, std::memory_order_relaxed);
 }
@@ -233,7 +225,7 @@ void StorageMemory::mutate(const MutationCommands & commands, const Context & co
     auto in = interpreter->execute();
 
     in->readPrefix();
-    BlocksList out;
+    Blocks out;
     Block block;
     while ((block = in->read()))
     {
@@ -241,17 +233,17 @@ void StorageMemory::mutate(const MutationCommands & commands, const Context & co
     }
     in->readSuffix();
 
-    std::unique_ptr<BlocksList> new_data;
+    std::unique_ptr<Blocks> new_data;
 
     // all column affected
     if (interpreter->isAffectingAllColumns())
     {
-        new_data = std::make_unique<BlocksList>(out);
+        new_data = std::make_unique<Blocks>(out);
     }
     else
     {
         /// just some of the column affected, we need update it with new column
-        new_data = std::make_unique<BlocksList>(*(data.get()));
+        new_data = std::make_unique<Blocks>(*(data.get()));
         auto data_it = new_data->begin();
         auto out_it = out.begin();
 
@@ -284,7 +276,7 @@ void StorageMemory::mutate(const MutationCommands & commands, const Context & co
 void StorageMemory::truncate(
     const ASTPtr &, const StorageMetadataPtr &, const Context &, TableExclusiveLockHolder &)
 {
-    data.set(std::make_unique<BlocksList>());
+    data.set(std::make_unique<Blocks>());
     total_size_bytes.store(0, std::memory_order_relaxed);
     total_size_rows.store(0, std::memory_order_relaxed);
 }
diff --git a/src/Storages/StorageMemory.h b/src/Storages/StorageMemory.h
index e3730ca104f..6453e6a53e2 100644
--- a/src/Storages/StorageMemory.h
+++ b/src/Storages/StorageMemory.h
@@ -91,7 +91,7 @@ public:
 
 private:
     /// MultiVersion data storage, so that we can copy the list of blocks to readers.
-    MultiVersion<BlocksList> data;
+    MultiVersion<Blocks> data;
 
     mutable std::mutex mutex;
 
diff --git a/src/Storages/StorageMerge.cpp b/src/Storages/StorageMerge.cpp
index 97f4ccd0bba..63d28eaff77 100644
--- a/src/Storages/StorageMerge.cpp
+++ b/src/Storages/StorageMerge.cpp
@@ -72,11 +72,27 @@ StorageMerge::StorageMerge(
     const StorageID & table_id_,
     const ColumnsDescription & columns_,
     const String & source_database_,
-    const String & table_name_regexp_,
+    const Strings & source_tables_,
     const Context & context_)
     : IStorage(table_id_)
     , source_database(source_database_)
-    , table_name_regexp(table_name_regexp_)
+    , source_tables(std::in_place, source_tables_.begin(), source_tables_.end())
+    , global_context(context_.getGlobalContext())
+{
+    StorageInMemoryMetadata storage_metadata;
+    storage_metadata.setColumns(columns_);
+    setInMemoryMetadata(storage_metadata);
+}
+
+StorageMerge::StorageMerge(
+    const StorageID & table_id_,
+    const ColumnsDescription & columns_,
+    const String & source_database_,
+    const String & source_table_regexp_,
+    const Context & context_)
+    : IStorage(table_id_)
+    , source_database(source_database_)
+    , source_table_regexp(source_table_regexp_)
     , global_context(context_.getGlobalContext())
 {
     StorageInMemoryMetadata storage_metadata;
@@ -439,8 +455,17 @@ DatabaseTablesIteratorPtr StorageMerge::getDatabaseIterator(const Context & cont
         e.addMessage("while getting table iterator of Merge table. Maybe caused by two Merge tables that will endlessly try to read each other's data");
         throw;
     }
+
     auto database = DatabaseCatalog::instance().getDatabase(source_database);
-    auto table_name_match = [this](const String & table_name_) { return table_name_regexp.match(table_name_); };
+
+    auto table_name_match = [this](const String & table_name_) -> bool
+    {
+        if (source_tables)
+            return source_tables->count(table_name_);
+        else
+            return source_table_regexp->match(table_name_);
+    };
+
     return database->getTablesIterator(context, table_name_match);
 }
 
diff --git a/src/Storages/StorageMerge.h b/src/Storages/StorageMerge.h
index 681ea7015e7..223383a6975 100644
--- a/src/Storages/StorageMerge.h
+++ b/src/Storages/StorageMerge.h
@@ -48,7 +48,8 @@ public:
 
 private:
     String source_database;
-    OptimizedRegularExpression table_name_regexp;
+    std::optional<std::unordered_set<String>> source_tables;
+    std::optional<OptimizedRegularExpression> source_table_regexp;
     const Context & global_context;
 
     using StorageWithLockAndName = std::tuple<StoragePtr, TableLockHolder, String>;
@@ -72,7 +73,14 @@ protected:
         const StorageID & table_id_,
         const ColumnsDescription & columns_,
         const String & source_database_,
-        const String & table_name_regexp_,
+        const Strings & source_tables_,
+        const Context & context_);
+
+    StorageMerge(
+        const StorageID & table_id_,
+        const ColumnsDescription & columns_,
+        const String & source_database_,
+        const String & source_table_regexp_,
         const Context & context_);
 
     Pipe createSources(
diff --git a/src/Storages/StorageMergeTree.cpp b/src/Storages/StorageMergeTree.cpp
index eb28ccfa3d5..591f9840896 100644
--- a/src/Storages/StorageMergeTree.cpp
+++ b/src/Storages/StorageMergeTree.cpp
@@ -233,7 +233,7 @@ BlockOutputStreamPtr StorageMergeTree::write(const ASTPtr & /*query*/, const Sto
 
     const auto & settings = context.getSettingsRef();
     return std::make_shared<MergeTreeBlockOutputStream>(
-        *this, metadata_snapshot, settings.max_partitions_per_insert_block);
+        *this, metadata_snapshot, settings.max_partitions_per_insert_block, context.getSettingsRef().optimize_on_insert);
 }
 
 void StorageMergeTree::checkTableCanBeDropped() const
diff --git a/src/Storages/StorageReplicatedMergeTree.cpp b/src/Storages/StorageReplicatedMergeTree.cpp
index e2bf1592659..f91d83b6811 100644
--- a/src/Storages/StorageReplicatedMergeTree.cpp
+++ b/src/Storages/StorageReplicatedMergeTree.cpp
@@ -42,6 +42,7 @@
 #include <IO/ReadBufferFromString.h>
 #include <IO/Operators.h>
 #include <IO/ConnectionTimeouts.h>
+#include <IO/ConnectionTimeoutsContext.h>
 
 #include <Interpreters/InterpreterAlterQuery.h>
 #include <Interpreters/PartLog.h>
@@ -3234,6 +3235,8 @@ void StorageReplicatedMergeTree::updateQuorum(const String & part_name, bool is_
         ReplicatedMergeTreeQuorumEntry quorum_entry(value);
         if (quorum_entry.part_name != part_name)
         {
+            LOG_TRACE(log, "Quorum {}, already achieved for part {} current part {}",
+                      quorum_status_path, part_name, quorum_entry.part_name);
             /// The quorum has already been achieved. Moreover, another INSERT with a quorum has already started.
             break;
         }
@@ -3243,6 +3246,8 @@ void StorageReplicatedMergeTree::updateQuorum(const String & part_name, bool is_
         if (quorum_entry.replicas.size() >= quorum_entry.required_number_of_replicas)
         {
             /// The quorum is reached. Delete the node, and update information about the last part that was successfully written with quorum.
+            LOG_TRACE(log, "Got {} replicas confirmed quorum {}, going to remove node",
+                      quorum_entry.replicas.size(), quorum_status_path);
 
             Coordination::Requests ops;
             Coordination::Responses responses;
@@ -3290,6 +3295,8 @@ void StorageReplicatedMergeTree::updateQuorum(const String & part_name, bool is_
         }
         else
         {
+            LOG_TRACE(log, "Quorum {} still not satisfied (have only {} replicas), updating node",
+                      quorum_status_path, quorum_entry.replicas.size());
             /// We update the node, registering there one more replica.
             auto code = zookeeper->trySet(quorum_status_path, quorum_entry.toString(), stat.version);
 
@@ -3860,7 +3867,8 @@ BlockOutputStreamPtr StorageReplicatedMergeTree::write(const ASTPtr & /*query*/,
         query_settings.insert_quorum_timeout.totalMilliseconds(),
         query_settings.max_partitions_per_insert_block,
         query_settings.insert_quorum_parallel,
-        deduplicate);
+        deduplicate,
+        context.getSettingsRef().optimize_on_insert);
 }
 
 
@@ -4443,7 +4451,7 @@ PartitionCommandsResultInfo StorageReplicatedMergeTree::attachPartition(
     PartsTemporaryRename renamed_parts(*this, "detached/");
     MutableDataPartsVector loaded_parts = tryLoadPartsToAttach(partition, attach_part, query_context, renamed_parts);
 
-    ReplicatedMergeTreeBlockOutputStream output(*this, metadata_snapshot, 0, 0, 0, false, false);   /// TODO Allow to use quorum here.
+    ReplicatedMergeTreeBlockOutputStream output(*this, metadata_snapshot, 0, 0, 0, false, false, false);   /// TODO Allow to use quorum here.
     for (size_t i = 0; i < loaded_parts.size(); ++i)
     {
         String old_name = loaded_parts[i]->name;
diff --git a/src/Storages/StorageReplicatedMergeTree.h b/src/Storages/StorageReplicatedMergeTree.h
index f1094793944..d396f32dcca 100644
--- a/src/Storages/StorageReplicatedMergeTree.h
+++ b/src/Storages/StorageReplicatedMergeTree.h
@@ -39,7 +39,8 @@ namespace DB
   * - the structure of the table (/metadata, /columns)
   * - action log with data (/log/log-...,/replicas/replica_name/queue/queue-...);
   * - a replica list (/replicas), and replica activity tag (/replicas/replica_name/is_active), replica addresses (/replicas/replica_name/host);
-  * - select the leader replica (/leader_election) - this is the replica that assigns the merge;
+  * - select the leader replica (/leader_election) - these are the replicas that assigning merges, mutations and partition manipulations
+  *   (after ClickHouse version 20.5 we allow multiple leaders to act concurrently);
   * - a set of parts of data on each replica (/replicas/replica_name/parts);
   * - list of the last N blocks of data with checksum, for deduplication (/blocks);
   * - the list of incremental block numbers (/block_numbers) that we are about to insert,
diff --git a/src/Storages/StorageS3.cpp b/src/Storages/StorageS3.cpp
index fff44bb1f4c..0465cfdffba 100644
--- a/src/Storages/StorageS3.cpp
+++ b/src/Storages/StorageS3.cpp
@@ -141,17 +141,18 @@ namespace
     public:
         StorageS3BlockOutputStream(
             const String & format,
-            UInt64 min_upload_part_size,
             const Block & sample_block_,
             const Context & context,
             const CompressionMethod compression_method,
             const std::shared_ptr<Aws::S3::S3Client> & client,
             const String & bucket,
-            const String & key)
+            const String & key,
+            size_t min_upload_part_size,
+            size_t max_single_part_upload_size)
             : sample_block(sample_block_)
         {
             write_buf = wrapWriteBufferWithCompressionMethod(
-                std::make_unique<WriteBufferFromS3>(client, bucket, key, min_upload_part_size, true), compression_method, 3);
+                std::make_unique<WriteBufferFromS3>(client, bucket, key, min_upload_part_size, max_single_part_upload_size), compression_method, 3);
             writer = FormatFactory::instance().getOutput(format, *write_buf, sample_block, context);
         }
 
@@ -192,6 +193,7 @@ StorageS3::StorageS3(
     const StorageID & table_id_,
     const String & format_name_,
     UInt64 min_upload_part_size_,
+    UInt64 max_single_part_upload_size_,
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
     const Context & context_,
@@ -201,6 +203,7 @@ StorageS3::StorageS3(
     , global_context(context_.getGlobalContext())
     , format_name(format_name_)
     , min_upload_part_size(min_upload_part_size_)
+    , max_single_part_upload_size(max_single_part_upload_size_)
     , compression_method(compression_method_)
     , name(uri_.storage_name)
 {
@@ -331,9 +334,15 @@ Pipe StorageS3::read(
 BlockOutputStreamPtr StorageS3::write(const ASTPtr & /*query*/, const StorageMetadataPtr & metadata_snapshot, const Context & /*context*/)
 {
     return std::make_shared<StorageS3BlockOutputStream>(
-        format_name, min_upload_part_size, metadata_snapshot->getSampleBlock(),
-        global_context, chooseCompressionMethod(uri.endpoint, compression_method),
-        client, uri.bucket, uri.key);
+        format_name,
+        metadata_snapshot->getSampleBlock(),
+        global_context,
+        chooseCompressionMethod(uri.endpoint, compression_method),
+        client,
+        uri.bucket,
+        uri.key,
+        min_upload_part_size,
+        max_single_part_upload_size);
 }
 
 void registerStorageS3Impl(const String & name, StorageFactory & factory)
@@ -362,6 +371,7 @@ void registerStorageS3Impl(const String & name, StorageFactory & factory)
         }
 
         UInt64 min_upload_part_size = args.local_context.getSettingsRef().s3_min_upload_part_size;
+        UInt64 max_single_part_upload_size = args.local_context.getSettingsRef().s3_max_single_part_upload_size;
 
         String compression_method;
         String format_name;
@@ -383,6 +393,7 @@ void registerStorageS3Impl(const String & name, StorageFactory & factory)
             args.table_id,
             format_name,
             min_upload_part_size,
+            max_single_part_upload_size,
             args.columns,
             args.constraints,
             args.context,
diff --git a/src/Storages/StorageS3.h b/src/Storages/StorageS3.h
index 96f0cf02e88..f436fb85c90 100644
--- a/src/Storages/StorageS3.h
+++ b/src/Storages/StorageS3.h
@@ -31,6 +31,7 @@ public:
         const StorageID & table_id_,
         const String & format_name_,
         UInt64 min_upload_part_size_,
+        UInt64 max_single_part_upload_size_,
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
         const Context & context_,
@@ -59,7 +60,8 @@ private:
     const Context & global_context;
 
     String format_name;
-    UInt64 min_upload_part_size;
+    size_t min_upload_part_size;
+    size_t max_single_part_upload_size;
     String compression_method;
     std::shared_ptr<Aws::S3::S3Client> client;
     String name;
diff --git a/src/Storages/StorageSet.cpp b/src/Storages/StorageSet.cpp
index d691a185e1e..e518c7da0e4 100644
--- a/src/Storages/StorageSet.cpp
+++ b/src/Storages/StorageSet.cpp
@@ -6,12 +6,11 @@
 #include <Compression/CompressedWriteBuffer.h>
 #include <DataStreams/NativeBlockOutputStream.h>
 #include <DataStreams/NativeBlockInputStream.h>
+#include <Disks/IDisk.h>
 #include <Common/formatReadable.h>
-#include <Common/escapeForFileName.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <Interpreters/Set.h>
 #include <Interpreters/Context.h>
-#include <Poco/DirectoryIterator.h>
 #include <Parsers/ASTCreateQuery.h>
 #include <Parsers/ASTLiteral.h>
 
@@ -49,7 +48,7 @@ private:
     String backup_path;
     String backup_tmp_path;
     String backup_file_name;
-    WriteBufferFromFile backup_buf;
+    std::unique_ptr<WriteBufferFromFileBase> backup_buf;
     CompressedWriteBuffer compressed_backup_buf;
     NativeBlockOutputStream backup_stream;
     bool persistent;
@@ -68,8 +67,8 @@ SetOrJoinBlockOutputStream::SetOrJoinBlockOutputStream(
     , backup_path(backup_path_)
     , backup_tmp_path(backup_tmp_path_)
     , backup_file_name(backup_file_name_)
-    , backup_buf(backup_tmp_path + backup_file_name)
-    , compressed_backup_buf(backup_buf)
+    , backup_buf(table_.disk->writeFile(backup_tmp_path + backup_file_name))
+    , compressed_backup_buf(*backup_buf)
     , backup_stream(compressed_backup_buf, 0, metadata_snapshot->getSampleBlock())
     , persistent(persistent_)
 {
@@ -92,9 +91,10 @@ void SetOrJoinBlockOutputStream::writeSuffix()
     {
         backup_stream.flush();
         compressed_backup_buf.next();
-        backup_buf.next();
+        backup_buf->next();
+        backup_buf->finalize();
 
-        Poco::File(backup_tmp_path + backup_file_name).renameTo(backup_path + backup_file_name);
+        table.disk->replaceFile(backup_tmp_path + backup_file_name, backup_path + backup_file_name);
     }
 }
 
@@ -107,13 +107,14 @@ BlockOutputStreamPtr StorageSetOrJoinBase::write(const ASTPtr & /*query*/, const
 
 
 StorageSetOrJoinBase::StorageSetOrJoinBase(
+    DiskPtr disk_,
     const String & relative_path_,
     const StorageID & table_id_,
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
-    const Context & context_,
     bool persistent_)
     : IStorage(table_id_),
+    disk(disk_),
     persistent(persistent_)
 {
     StorageInMemoryMetadata storage_metadata;
@@ -125,19 +126,18 @@ StorageSetOrJoinBase::StorageSetOrJoinBase(
     if (relative_path_.empty())
         throw Exception("Join and Set storages require data path", ErrorCodes::INCORRECT_FILE_NAME);
 
-    base_path = context_.getPath();
-    path = base_path + relative_path_;
+    path = relative_path_;
 }
 
 
 StorageSet::StorageSet(
+    DiskPtr disk_,
     const String & relative_path_,
     const StorageID & table_id_,
     const ColumnsDescription & columns_,
     const ConstraintsDescription & constraints_,
-    const Context & context_,
     bool persistent_)
-    : StorageSetOrJoinBase{relative_path_, table_id_, columns_, constraints_, context_, persistent_},
+    : StorageSetOrJoinBase{disk_, relative_path_, table_id_, columns_, constraints_, persistent_},
     set(std::make_shared<Set>(SizeLimits(), false, true))
 {
 
@@ -158,9 +158,9 @@ std::optional<UInt64> StorageSet::totalBytes(const Settings &) const { return se
 
 void StorageSet::truncate(const ASTPtr &, const StorageMetadataPtr & metadata_snapshot, const Context &, TableExclusiveLockHolder &)
 {
-    Poco::File(path).remove(true);
-    Poco::File(path).createDirectories();
-    Poco::File(path + "tmp/").createDirectories();
+    disk->removeRecursive(path);
+    disk->createDirectories(path);
+    disk->createDirectories(path + "tmp/");
 
     Block header = metadata_snapshot->getSampleBlock();
     header = header.sortColumns();
@@ -173,24 +173,23 @@ void StorageSet::truncate(const ASTPtr &, const StorageMetadataPtr & metadata_sn
 
 void StorageSetOrJoinBase::restore()
 {
-    Poco::File tmp_dir(path + "tmp/");
-    if (!tmp_dir.exists())
+    if (!disk->exists(path + "tmp/"))
     {
-        tmp_dir.createDirectories();
+        disk->createDirectories(path + "tmp/");
         return;
     }
 
     static const char * file_suffix = ".bin";
     static const auto file_suffix_size = strlen(".bin");
 
-    Poco::DirectoryIterator dir_end;
-    for (Poco::DirectoryIterator dir_it(path); dir_end != dir_it; ++dir_it)
+    for (auto dir_it{disk->iterateDirectory(path)}; dir_it->isValid(); dir_it->next())
     {
-        const auto & name = dir_it.name();
+        const auto & name = dir_it->name();
+        const auto & file_path = dir_it->path();
 
-        if (dir_it->isFile()
+        if (disk->isFile(file_path)
             && endsWith(name, file_suffix)
-            && dir_it->getSize() > 0)
+            && disk->getFileSize(file_path) > 0)
         {
             /// Calculate the maximum number of available files with a backup to add the following files with large numbers.
             UInt64 file_num = parse<UInt64>(name.substr(0, name.size() - file_suffix_size));
@@ -205,8 +204,8 @@ void StorageSetOrJoinBase::restore()
 
 void StorageSetOrJoinBase::restoreFromFile(const String & file_path)
 {
-    ReadBufferFromFile backup_buf(file_path);
-    CompressedReadBuffer compressed_backup_buf(backup_buf);
+    auto backup_buf = disk->readFile(file_path);
+    CompressedReadBuffer compressed_backup_buf(*backup_buf);
     NativeBlockInputStream backup_stream(compressed_backup_buf, 0);
 
     backup_stream.readPrefix();
@@ -226,10 +225,9 @@ void StorageSetOrJoinBase::restoreFromFile(const String & file_path)
 void StorageSetOrJoinBase::rename(const String & new_path_to_table_data, const StorageID & new_table_id)
 {
     /// Rename directory with data.
-    String new_path = base_path + new_path_to_table_data;
-    Poco::File(path).renameTo(new_path);
+    disk->replaceFile(path, new_path_to_table_data);
 
-    path = new_path;
+    path = new_path_to_table_data;
     renameInMemory(new_table_id);
 }
 
@@ -251,7 +249,8 @@ void registerStorageSet(StorageFactory & factory)
             set_settings->loadFromQuery(*args.storage_def);
         }
 
-        return StorageSet::create(args.relative_data_path, args.table_id, args.columns, args.constraints, args.context, set_settings->persistent);
+        DiskPtr disk = args.context.getDisk(set_settings->disk);
+        return StorageSet::create(disk, args.relative_data_path, args.table_id, args.columns, args.constraints, set_settings->persistent);
     }, StorageFactory::StorageFeatures{ .supports_settings = true, });
 }
 
diff --git a/src/Storages/StorageSet.h b/src/Storages/StorageSet.h
index 0f976fce17d..9b9078f7dd5 100644
--- a/src/Storages/StorageSet.h
+++ b/src/Storages/StorageSet.h
@@ -2,6 +2,7 @@
 
 #include <ext/shared_ptr_helper.h>
 
+#include <Interpreters/Context.h>
 #include <Storages/IStorage.h>
 #include <Storages/SetSettings.h>
 
@@ -29,14 +30,14 @@ public:
 
 protected:
     StorageSetOrJoinBase(
+        DiskPtr disk_,
         const String & relative_path_,
         const StorageID & table_id_,
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
-        const Context & context_,
         bool persistent_);
 
-    String base_path;
+    DiskPtr disk;
     String path;
     bool persistent;
 
@@ -85,11 +86,11 @@ private:
 
 protected:
     StorageSet(
+        DiskPtr disk_,
         const String & relative_path_,
         const StorageID & table_id_,
         const ColumnsDescription & columns_,
         const ConstraintsDescription & constraints_,
-        const Context & context_,
         bool persistent_);
 };
 
diff --git a/src/Storages/StorageStripeLog.cpp b/src/Storages/StorageStripeLog.cpp
index a265df856c1..bc6afffddeb 100644
--- a/src/Storages/StorageStripeLog.cpp
+++ b/src/Storages/StorageStripeLog.cpp
@@ -40,8 +40,6 @@
 namespace DB
 {
 
-#define INDEX_BUFFER_SIZE 4096
-
 namespace ErrorCodes
 {
     extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
@@ -319,7 +317,7 @@ Pipe StorageStripeLog::read(
         return Pipe(std::make_shared<NullSource>(metadata_snapshot->getSampleBlockForColumns(column_names, getVirtuals(), getStorageID())));
     }
 
-    CompressedReadBufferFromFile index_in(disk->readFile(index_file, INDEX_BUFFER_SIZE));
+    CompressedReadBufferFromFile index_in(disk->readFile(index_file, 4096));
     std::shared_ptr<const IndexForNativeFormat> index{std::make_shared<IndexForNativeFormat>(index_in, column_names_set)};
 
     size_t size = index->blocks.size();
diff --git a/src/Storages/StorageURL.cpp b/src/Storages/StorageURL.cpp
index 8dcd549f9c8..00903abee59 100644
--- a/src/Storages/StorageURL.cpp
+++ b/src/Storages/StorageURL.cpp
@@ -10,6 +10,8 @@
 #include <IO/ReadWriteBufferFromHTTP.h>
 #include <IO/WriteBufferFromHTTP.h>
 #include <IO/WriteHelpers.h>
+#include <IO/ConnectionTimeouts.h>
+#include <IO/ConnectionTimeoutsContext.h>
 
 #include <Formats/FormatFactory.h>
 
diff --git a/src/Storages/StorageURL.h b/src/Storages/StorageURL.h
index 78d972c6e7e..21b2e3e27a1 100644
--- a/src/Storages/StorageURL.h
+++ b/src/Storages/StorageURL.h
@@ -4,12 +4,15 @@
 #include <Poco/URI.h>
 #include <ext/shared_ptr_helper.h>
 #include <DataStreams/IBlockOutputStream.h>
-#include <IO/ConnectionTimeouts.h>
+#include <Formats/FormatSettings.h>
 #include <IO/CompressionMethod.h>
 
 
 namespace DB
 {
+
+struct ConnectionTimeouts;
+
 /**
  * This class represents table engine for external urls.
  * It sends HTTP GET to server when select is called and
diff --git a/src/Storages/StorageXDBC.cpp b/src/Storages/StorageXDBC.cpp
index 3aca884d15a..f2f8cdb23f5 100644
--- a/src/Storages/StorageXDBC.cpp
+++ b/src/Storages/StorageXDBC.cpp
@@ -3,6 +3,7 @@
 #include <DataStreams/IBlockOutputStream.h>
 #include <Formats/FormatFactory.h>
 #include <IO/ReadHelpers.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <Interpreters/Context.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Parsers/ASTLiteral.h>
diff --git a/src/Storages/System/StorageSystemColumns.cpp b/src/Storages/System/StorageSystemColumns.cpp
index b0371fa12de..6726d502071 100644
--- a/src/Storages/System/StorageSystemColumns.cpp
+++ b/src/Storages/System/StorageSystemColumns.cpp
@@ -272,17 +272,28 @@ Pipe StorageSystemColumns::read(
     Pipes pipes;
 
     {
-        Databases databases = DatabaseCatalog::instance().getDatabases();
-
         /// Add `database` column.
         MutableColumnPtr database_column_mut = ColumnString::create();
-        for (const auto & database : databases)
+
+        const auto databases = DatabaseCatalog::instance().getDatabases();
+        for (const auto & [database_name, database] : databases)
         {
+            if (database_name == DatabaseCatalog::TEMPORARY_DATABASE)
+                continue; /// We don't want to show the internal database for temporary tables in system.columns
+
             /// We are skipping "Lazy" database because we cannot afford initialization of all its tables.
             /// This should be documented.
 
-            if (database.second->getEngineName() != "Lazy")
-                database_column_mut->insert(database.first);
+            if (database->getEngineName() != "Lazy")
+                database_column_mut->insert(database_name);
+        }
+
+        Tables external_tables;
+        if (context.hasSessionContext())
+        {
+            external_tables = context.getSessionContext().getExternalTables();
+            if (!external_tables.empty())
+                database_column_mut->insertDefault(); /// Empty database for external tables.
         }
 
         block_to_filter.insert(ColumnWithTypeAndName(std::move(database_column_mut), std::make_shared<DataTypeString>(), "database"));
@@ -297,29 +308,36 @@ Pipe StorageSystemColumns::read(
         }
 
         ColumnPtr & database_column = block_to_filter.getByName("database").column;
-        size_t rows = database_column->size();
 
         /// Add `table` column.
         MutableColumnPtr table_column_mut = ColumnString::create();
-        IColumn::Offsets offsets(rows);
-        for (size_t i = 0; i < rows; ++i)
+        IColumn::Offsets offsets(database_column->size());
+
+        for (size_t i = 0; i < database_column->size(); ++i)
         {
             const std::string database_name = (*database_column)[i].get<std::string>();
-            const DatabasePtr database = databases.at(database_name);
-            offsets[i] = i ? offsets[i - 1] : 0;
-
-            for (auto iterator = database->getTablesIterator(context); iterator->isValid(); iterator->next())
+            if (database_name.empty())
             {
-                if (const auto & table = iterator->table())
+                for (auto & [table_name, table] : external_tables)
                 {
-                    const String & table_name = iterator->name();
-                    storages.emplace(std::piecewise_construct,
-                        std::forward_as_tuple(database_name, table_name),
-                        std::forward_as_tuple(table));
+                    storages[{"", table_name}] = table;
                     table_column_mut->insert(table_name);
-                    ++offsets[i];
                 }
             }
+            else
+            {
+                const DatabasePtr & database = databases.at(database_name);
+                for (auto iterator = database->getTablesIterator(context); iterator->isValid(); iterator->next())
+                {
+                    if (const auto & table = iterator->table())
+                    {
+                        const String & table_name = iterator->name();
+                        storages[{database_name, table_name}] = table;
+                        table_column_mut->insert(table_name);
+                    }
+                }
+            }
+            offsets[i] = table_column_mut->size();
         }
 
         database_column = database_column->replicate(offsets);
diff --git a/src/Storages/System/StorageSystemDatabases.cpp b/src/Storages/System/StorageSystemDatabases.cpp
index 2a89bf34186..88ac987014d 100644
--- a/src/Storages/System/StorageSystemDatabases.cpp
+++ b/src/Storages/System/StorageSystemDatabases.cpp
@@ -25,17 +25,20 @@ void StorageSystemDatabases::fillData(MutableColumns & res_columns, const Contex
     const auto access = context.getAccess();
     const bool check_access_for_databases = !access->isGranted(AccessType::SHOW_DATABASES);
 
-    auto databases = DatabaseCatalog::instance().getDatabases();
-    for (const auto & database : databases)
+    const auto databases = DatabaseCatalog::instance().getDatabases();
+    for (const auto & [database_name, database] : databases)
     {
-        if (check_access_for_databases && !access->isGranted(AccessType::SHOW_DATABASES, database.first))
+        if (check_access_for_databases && !access->isGranted(AccessType::SHOW_DATABASES, database_name))
             continue;
 
-        res_columns[0]->insert(database.first);
-        res_columns[1]->insert(database.second->getEngineName());
-        res_columns[2]->insert(context.getPath() + database.second->getDataPath());
-        res_columns[3]->insert(database.second->getMetadataPath());
-        res_columns[4]->insert(database.second->getUUID());
+        if (database_name == DatabaseCatalog::TEMPORARY_DATABASE)
+            continue; /// We don't want to show the internal database for temporary tables in system.databases
+
+        res_columns[0]->insert(database_name);
+        res_columns[1]->insert(database->getEngineName());
+        res_columns[2]->insert(context.getPath() + database->getDataPath());
+        res_columns[3]->insert(database->getMetadataPath());
+        res_columns[4]->insert(database->getUUID());
    }
 }
 
diff --git a/src/Storages/System/StorageSystemDistributionQueue.cpp b/src/Storages/System/StorageSystemDistributionQueue.cpp
index edba7c13b1c..db649e7e1ba 100644
--- a/src/Storages/System/StorageSystemDistributionQueue.cpp
+++ b/src/Storages/System/StorageSystemDistributionQueue.cpp
@@ -9,6 +9,7 @@
 #include <Storages/VirtualColumnUtils.h>
 #include <Access/ContextAccess.h>
 #include <Common/typeid_cast.h>
+#include <Interpreters/Context.h>
 #include <Databases/IDatabase.h>
 
 namespace DB
diff --git a/src/Storages/System/StorageSystemTables.cpp b/src/Storages/System/StorageSystemTables.cpp
index c9b3223ee38..363a2a20828 100644
--- a/src/Storages/System/StorageSystemTables.cpp
+++ b/src/Storages/System/StorageSystemTables.cpp
@@ -65,8 +65,15 @@ StorageSystemTables::StorageSystemTables(const StorageID & table_id_)
 static ColumnPtr getFilteredDatabases(const ASTPtr & query, const Context & context)
 {
     MutableColumnPtr column = ColumnString::create();
-    for (const auto & db : DatabaseCatalog::instance().getDatabases())
-        column->insert(db.first);
+
+    const auto databases = DatabaseCatalog::instance().getDatabases();
+    for (const auto & database_name : databases | boost::adaptors::map_keys)
+    {
+        if (database_name == DatabaseCatalog::TEMPORARY_DATABASE)
+            continue; /// We don't want to show the internal database for temporary tables in system.tables
+
+        column->insert(database_name);
+    }
 
     Block block { ColumnWithTypeAndName(std::move(column), std::make_shared<DataTypeString>(), "database") };
     VirtualColumnUtils::filterBlockWithQuery(query, block, context);
@@ -192,7 +199,11 @@ protected:
 
                         // create_table_query
                         if (columns_mask[src_index++])
-                            res_columns[res_index++]->insertDefault();
+                        {
+                            auto temp_db = DatabaseCatalog::instance().getDatabaseForTemporaryTables();
+                            ASTPtr ast = temp_db ? temp_db->tryGetCreateTableQuery(table.second->getStorageID().getTableName(), context) : nullptr;
+                            res_columns[res_index++]->insert(ast ? queryToString(ast) : "");
+                        }
 
                         // engine_full
                         if (columns_mask[src_index++])
diff --git a/src/Storages/getStructureOfRemoteTable.cpp b/src/Storages/getStructureOfRemoteTable.cpp
index a987e3d4e8a..de5f3924ca9 100644
--- a/src/Storages/getStructureOfRemoteTable.cpp
+++ b/src/Storages/getStructureOfRemoteTable.cpp
@@ -71,7 +71,7 @@ ColumnsDescription getStructureOfRemoteTableInShard(
     };
 
     /// Execute remote query without restrictions (because it's not real user query, but part of implementation)
-    auto input = std::make_shared<RemoteBlockInputStream>(shard_info.pool, query, sample_block, new_context);
+    auto input = std::make_shared<RemoteBlockInputStream>(shard_info.pool, query, sample_block, *new_context);
     input->setPoolMode(PoolMode::GET_ONE);
     if (!table_func_ptr)
         input->setMainTable(table_id);
diff --git a/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp b/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp
index 4e7bff0ef41..0a3b3bd328f 100644
--- a/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp
+++ b/src/Storages/tests/gtest_aux_funcs_for_adaptive_granularity.cpp
@@ -4,6 +4,7 @@
 
 // I know that inclusion of .cpp is not good at all
 #include <Storages/MergeTree/MergeTreeDataPartWriterOnDisk.cpp> // NOLINT
+#include <Storages/MergeTree/MergeTreeDataPartWriterWide.cpp> // NOLINT
 
 using namespace DB;
 static Block getBlockWithSize(size_t required_size_in_bytes, size_t size_of_row_in_bytes)
diff --git a/src/TableFunctions/ITableFunctionXDBC.cpp b/src/TableFunctions/ITableFunctionXDBC.cpp
index 67d1257fe4c..e04a86b5abf 100644
--- a/src/TableFunctions/ITableFunctionXDBC.cpp
+++ b/src/TableFunctions/ITableFunctionXDBC.cpp
@@ -5,6 +5,7 @@
 #include <DataTypes/DataTypeFactory.h>
 #include <IO/ReadHelpers.h>
 #include <IO/ReadWriteBufferFromHTTP.h>
+#include <IO/ConnectionTimeoutsContext.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ASTLiteral.h>
diff --git a/src/TableFunctions/TableFunctionFile.h b/src/TableFunctions/TableFunctionFile.h
index 91e2225a6d0..02704e4bf7f 100644
--- a/src/TableFunctions/TableFunctionFile.h
+++ b/src/TableFunctions/TableFunctionFile.h
@@ -6,8 +6,8 @@
 namespace DB
 {
 class Context;
+
 /* file(path, format, structure) - creates a temporary storage from file
- *
  *
  * The file must be in the clickhouse data directory.
  * The relative path begins with the clickhouse data directory.
diff --git a/src/TableFunctions/TableFunctionHDFS.cpp b/src/TableFunctions/TableFunctionHDFS.cpp
index e2f227ef7b5..700cb93ca06 100644
--- a/src/TableFunctions/TableFunctionHDFS.cpp
+++ b/src/TableFunctions/TableFunctionHDFS.cpp
@@ -2,7 +2,7 @@
 #include "registerTableFunctions.h"
 
 #if USE_HDFS
-#include <Storages/StorageHDFS.h>
+#include <Storages/HDFS/StorageHDFS.h>
 #include <Storages/ColumnsDescription.h>
 #include <TableFunctions/TableFunctionFactory.h>
 #include <TableFunctions/TableFunctionHDFS.h>
diff --git a/src/TableFunctions/TableFunctionMerge.cpp b/src/TableFunctions/TableFunctionMerge.cpp
index a878909e29d..c5fb9a7686d 100644
--- a/src/TableFunctions/TableFunctionMerge.cpp
+++ b/src/TableFunctions/TableFunctionMerge.cpp
@@ -6,6 +6,7 @@
 #include <TableFunctions/ITableFunction.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Interpreters/Context.h>
+#include <Access/ContextAccess.h>
 #include <TableFunctions/TableFunctionMerge.h>
 #include <TableFunctions/TableFunctionFactory.h>
 #include <TableFunctions/registerTableFunctions.h>
@@ -14,37 +15,24 @@
 namespace DB
 {
 
-
 namespace ErrorCodes
 {
     extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
     extern const int UNKNOWN_TABLE;
 }
 
-
-static NamesAndTypesList chooseColumns(const String & source_database, const String & table_name_regexp_, const Context & context)
+namespace
 {
-    OptimizedRegularExpression table_name_regexp(table_name_regexp_);
-    auto table_name_match = [&](const String & table_name) { return table_name_regexp.match(table_name); };
-
-    StoragePtr any_table;
-
+    [[noreturn]] void throwNoTablesMatchRegexp(const String & source_database, const String & source_table_regexp)
     {
-        auto database = DatabaseCatalog::instance().getDatabase(source_database);
-        auto iterator = database->getTablesIterator(context, table_name_match);
-
-        if (iterator->isValid())
-            if (const auto & table = iterator->table())
-                any_table = table;
+        throw Exception(
+            "Error while executing table function merge. In database " + source_database
+                + " no one matches regular expression: " + source_table_regexp,
+            ErrorCodes::UNKNOWN_TABLE);
     }
-
-    if (!any_table)
-        throw Exception("Error while executing table function merge. In database " + source_database + " no one matches regular expression: "
-            + table_name_regexp_, ErrorCodes::UNKNOWN_TABLE);
-
-    return any_table->getInMemoryMetadataPtr()->getColumns().getAllPhysical();
 }
 
+
 void TableFunctionMerge::parseArguments(const ASTPtr & ast_function, const Context & context)
 {
     ASTs & args_func = ast_function->children;
@@ -65,21 +53,64 @@ void TableFunctionMerge::parseArguments(const ASTPtr & ast_function, const Conte
     args[1] = evaluateConstantExpressionAsLiteral(args[1], context);
 
     source_database = args[0]->as<ASTLiteral &>().value.safeGet<String>();
-    table_name_regexp = args[1]->as<ASTLiteral &>().value.safeGet<String>();
+    source_table_regexp = args[1]->as<ASTLiteral &>().value.safeGet<String>();
 }
 
+
+const Strings & TableFunctionMerge::getSourceTables(const Context & context) const
+{
+    if (source_tables)
+        return *source_tables;
+
+    auto database = DatabaseCatalog::instance().getDatabase(source_database);
+
+    OptimizedRegularExpression re(source_table_regexp);
+    auto table_name_match = [&](const String & table_name_) { return re.match(table_name_); };
+
+    auto access = context.getAccess();
+    bool granted_show_on_all_tables = access->isGranted(AccessType::SHOW_TABLES, source_database);
+    bool granted_select_on_all_tables = access->isGranted(AccessType::SELECT, source_database);
+
+    source_tables.emplace();
+    for (auto it = database->getTablesIterator(context, table_name_match); it->isValid(); it->next())
+    {
+        if (!it->table())
+            continue;
+        bool granted_show = granted_show_on_all_tables || access->isGranted(AccessType::SHOW_TABLES, source_database, it->name());
+        if (!granted_show)
+            continue;
+        if (!granted_select_on_all_tables)
+            access->checkAccess(AccessType::SELECT, source_database, it->name());
+        source_tables->emplace_back(it->name());
+    }
+
+    if (source_tables->empty())
+        throwNoTablesMatchRegexp(source_database, source_table_regexp);
+
+    return *source_tables;
+}
+
+
 ColumnsDescription TableFunctionMerge::getActualTableStructure(const Context & context) const
 {
-    return ColumnsDescription{chooseColumns(source_database, table_name_regexp, context)};
+    for (const auto & table_name : getSourceTables(context))
+    {
+        auto storage = DatabaseCatalog::instance().tryGetTable(StorageID{source_database, table_name}, context);
+        if (storage)
+            return ColumnsDescription{storage->getInMemoryMetadataPtr()->getColumns().getAllPhysical()};
+    }
+
+    throwNoTablesMatchRegexp(source_database, source_table_regexp);
 }
 
+
 StoragePtr TableFunctionMerge::executeImpl(const ASTPtr & /*ast_function*/, const Context & context, const std::string & table_name, ColumnsDescription /*cached_columns*/) const
 {
     auto res = StorageMerge::create(
         StorageID(getDatabaseName(), table_name),
         getActualTableStructure(context),
         source_database,
-        table_name_regexp,
+        getSourceTables(context),
         context);
 
     res->startup();
diff --git a/src/TableFunctions/TableFunctionMerge.h b/src/TableFunctions/TableFunctionMerge.h
index a9c4dd6b038..8f9f4522d17 100644
--- a/src/TableFunctions/TableFunctionMerge.h
+++ b/src/TableFunctions/TableFunctionMerge.h
@@ -19,11 +19,13 @@ private:
     StoragePtr executeImpl(const ASTPtr & ast_function, const Context & context, const std::string & table_name, ColumnsDescription cached_columns) const override;
     const char * getStorageTypeName() const override { return "Merge"; }
 
+    const Strings & getSourceTables(const Context & context) const;
     ColumnsDescription getActualTableStructure(const Context & context) const override;
     void parseArguments(const ASTPtr & ast_function, const Context & context) override;
 
     String source_database;
-    String table_name_regexp;
+    String source_table_regexp;
+    mutable std::optional<Strings> source_tables;
 };
 
 
diff --git a/src/TableFunctions/TableFunctionS3.cpp b/src/TableFunctions/TableFunctionS3.cpp
index dfe1cf6e792..cc7877b204e 100644
--- a/src/TableFunctions/TableFunctionS3.cpp
+++ b/src/TableFunctions/TableFunctionS3.cpp
@@ -67,6 +67,7 @@ StoragePtr TableFunctionS3::executeImpl(const ASTPtr & /*ast_function*/, const C
     Poco::URI uri (filename);
     S3::URI s3_uri (uri);
     UInt64 min_upload_part_size = context.getSettingsRef().s3_min_upload_part_size;
+    UInt64 max_single_part_upload_size = context.getSettingsRef().s3_max_single_part_upload_size;
 
     StoragePtr storage = StorageS3::create(
             s3_uri,
@@ -75,6 +76,7 @@ StoragePtr TableFunctionS3::executeImpl(const ASTPtr & /*ast_function*/, const C
             StorageID(getDatabaseName(), table_name),
             format,
             min_upload_part_size,
+            max_single_part_upload_size,
             getActualTableStructure(context),
             ConstraintsDescription{},
             const_cast<Context &>(context),
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index 9e5a2e29dc9..3ef09e5658f 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -15,16 +15,6 @@ install (
     COMPONENT clickhouse
     PATTERN "CMakeLists.txt" EXCLUDE
     PATTERN ".gitignore" EXCLUDE
-    PATTERN "top_level_domains" EXCLUDE
-)
-
-# Dereference symlink
-get_filename_component(TOP_LEVEL_DOMAINS_ABS_DIR config/top_level_domains REALPATH)
-install (
-    DIRECTORY "${TOP_LEVEL_DOMAINS_ABS_DIR}"
-    DESTINATION ${CMAKE_INSTALL_DATAROOTDIR}/clickhouse-test/config
-    USE_SOURCE_PERMISSIONS
-    COMPONENT clickhouse
 )
 
 install (FILES server-test.xml DESTINATION  ${CLICKHOUSE_ETC_DIR}/clickhouse-server COMPONENT clickhouse)
diff --git a/tests/clickhouse-test b/tests/clickhouse-test
index 3702268819b..45cde00037f 100755
--- a/tests/clickhouse-test
+++ b/tests/clickhouse-test
@@ -15,6 +15,7 @@ from subprocess import check_call
 from subprocess import Popen
 from subprocess import PIPE
 from subprocess import CalledProcessError
+from subprocess import TimeoutExpired
 from datetime import datetime
 from time import time, sleep
 from errno import ESRCH
@@ -114,6 +115,7 @@ def get_db_engine(args):
 def run_single_test(args, ext, server_logs_level, client_options, case_file, stdout_file, stderr_file):
     # print(client_options)
 
+    start_time = datetime.now()
     if args.database:
         database = args.database
         os.environ.setdefault("CLICKHOUSE_DATABASE", database)
@@ -129,7 +131,11 @@ def run_single_test(args, ext, server_logs_level, client_options, case_file, std
         database = 'test_{suffix}'.format(suffix=random_str())
 
         clickhouse_proc_create = Popen(shlex.split(args.client), stdin=PIPE, stdout=PIPE, stderr=PIPE, universal_newlines=True)
-        clickhouse_proc_create.communicate(("CREATE DATABASE " + database + get_db_engine(args)))
+        try:
+            clickhouse_proc_create.communicate(("CREATE DATABASE " + database + get_db_engine(args)), timeout=args.timeout)
+        except TimeoutExpired:
+            total_time = (datetime.now() - start_time).total_seconds()
+            return clickhouse_proc_create, "", "Timeout creating database {} before test".format(database), total_time
 
         os.environ["CLICKHOUSE_DATABASE"] = database
 
@@ -152,14 +158,26 @@ def run_single_test(args, ext, server_logs_level, client_options, case_file, std
     # print(command)
 
     proc = Popen(command, shell=True, env=os.environ)
-    start_time = datetime.now()
 
     while (datetime.now() - start_time).total_seconds() < args.timeout and proc.poll() is None:
         sleep(0.01)
 
     if not args.database:
         clickhouse_proc_create = Popen(shlex.split(args.client), stdin=PIPE, stdout=PIPE, stderr=PIPE, universal_newlines=True)
-        clickhouse_proc_create.communicate(("DROP DATABASE " + database))
+        seconds_left = max(args.timeout - (datetime.now() - start_time).total_seconds(), 10)
+        try:
+            clickhouse_proc_create.communicate(("DROP DATABASE " + database), timeout=seconds_left)
+        except TimeoutExpired:
+            # kill test process because it can also hung
+            if proc.returncode is None:
+                try:
+                    proc.kill()
+                except OSError as e:
+                    if e.errno != ESRCH:
+                        raise
+
+            total_time = (datetime.now() - start_time).total_seconds()
+            return clickhouse_proc_create, "", "Timeout dropping database {} after test".format(database), total_time
 
     total_time = (datetime.now() - start_time).total_seconds()
 
@@ -305,7 +323,7 @@ def run_tests_array(all_tests_with_params):
 
                     if args.testname:
                         clickhouse_proc = Popen(shlex.split(args.client), stdin=PIPE, stdout=PIPE, stderr=PIPE, universal_newlines=True)
-                        clickhouse_proc.communicate(("SELECT 'Running test {suite}/{case} from pid={pid}';".format(pid = os.getpid(), case = case, suite = suite)))
+                        clickhouse_proc.communicate(("SELECT 'Running test {suite}/{case} from pid={pid}';".format(pid = os.getpid(), case = case, suite = suite)), timeout=10)
 
                         if clickhouse_proc.returncode != 0:
                             failures += 1
@@ -330,6 +348,8 @@ def run_tests_array(all_tests_with_params):
                         print(MSG_FAIL, end='')
                         print_test_time(total_time)
                         print(" - Timeout!")
+                        if stderr:
+                            print(stderr)
                     else:
                         counter = 1
                         while proc.returncode != 0 and need_retry(stderr):
@@ -562,6 +582,9 @@ def main(args):
     else:
         args.skip = tests_to_skip_from_list
 
+    if args.use_skip_list and not args.sequential:
+        args.sequential = collect_sequential_list(args.skip_list_path)
+
     base_dir = os.path.abspath(args.queries)
     tmp_dir = os.path.abspath(args.tmp)
 
@@ -675,10 +698,19 @@ def main(args):
                 all_tests = [t for t in all_tests if any([re.search(r, t) for r in args.test])]
             all_tests.sort(key=key_func)
 
+            parallel_tests = []
+            sequential_tests = []
+            for test in all_tests:
+                if any(s in test for s in args.sequential):
+                    sequential_tests.append(test)
+                else:
+                    parallel_tests.append(test)
+
+            print("Found", len(parallel_tests), "parallel tests and", len(sequential_tests), "sequential tests")
             run_n, run_total = args.parallel.split('/')
             run_n = float(run_n)
             run_total = float(run_total)
-            tests_n = len(all_tests)
+            tests_n = len(parallel_tests)
             if run_total > tests_n:
                 run_total = tests_n
             if run_n > run_total:
@@ -690,18 +722,20 @@ def main(args):
             if jobs > run_total:
                 run_total = jobs
 
-            batch_size = len(all_tests) // jobs
-            all_tests_array = []
-            for i in range(0, len(all_tests), batch_size):
-                all_tests_array.append((all_tests[i:i+batch_size], suite, suite_dir, suite_tmp_dir, run_total))
+            batch_size = len(parallel_tests) // jobs
+            parallel_tests_array = []
+            for i in range(0, len(parallel_tests), batch_size):
+                parallel_tests_array.append((parallel_tests[i:i+batch_size], suite, suite_dir, suite_tmp_dir, run_total))
 
             if jobs > 1:
                 with closing(multiprocessing.Pool(processes=jobs)) as pool:
-                    pool.map(run_tests_array, all_tests_array)
-            else:
-                run_tests_array(all_tests_array[int(run_n)-1])
+                    pool.map(run_tests_array, parallel_tests_array)
 
-            total_tests_run += tests_n
+                run_tests_array((sequential_tests, suite, suite_dir, suite_tmp_dir, run_total))
+                total_tests_run += len(sequential_tests) + len(parallel_tests)
+            else:
+                run_tests_array((all_tests, suite, suite_dir, suite_tmp_dir, run_total))
+                total_tests_run += len(all_tests)
 
     if args.hung_check:
 
@@ -792,6 +826,20 @@ def collect_tests_to_skip(skip_list_path, build_flags):
 
     return result
 
+
+def collect_sequential_list(skip_list_path):
+    if not os.path.exists(skip_list_path):
+        return set([])
+
+    with open(skip_list_path, 'r') as skip_list_file:
+        content = skip_list_file.read()
+        # allows to have comments in skip_list.json
+        skip_dict = json.loads(json_minify(content))
+        if 'parallel' in skip_dict:
+            return skip_dict['parallel']
+        return set([])
+
+
 if __name__ == '__main__':
     parser=ArgumentParser(description='ClickHouse functional tests')
     parser.add_argument('-q', '--queries', help='Path to queries dir')
@@ -828,6 +876,7 @@ if __name__ == '__main__':
     parser.add_argument('--no-stateless', action='store_true', help='Disable all stateless tests')
     parser.add_argument('--no-stateful', action='store_true', help='Disable all stateful tests')
     parser.add_argument('--skip', nargs='+', help="Skip these tests")
+    parser.add_argument('--sequential', nargs='+', help="Run these tests sequentially even if --parallel specified")
     parser.add_argument('--no-long', action='store_false', dest='no_long', help='Do not run long tests')
     parser.add_argument('--client-option', nargs='+', help='Specify additional client argument')
     parser.add_argument('--print-time', action='store_true', dest='print_time', help='Print test time')
@@ -871,6 +920,9 @@ if __name__ == '__main__':
     if args.skip_list_path is None:
         args.skip_list_path = os.path.join(args.queries, 'skip_list.json')
 
+    if args.sequential is None:
+        args.sequential = set([])
+
     if args.tmp is None:
         args.tmp = args.queries
     if args.client is None:
diff --git a/tests/config/config.d/max_concurrent_queries.xml b/tests/config/config.d/max_concurrent_queries.xml
new file mode 100644
index 00000000000..f326c293244
--- /dev/null
+++ b/tests/config/config.d/max_concurrent_queries.xml
@@ -0,0 +1,4 @@
+<!-- Ignore limit of concurrent queries in tests -->
+<yandex>
+    <max_concurrent_queries>0</max_concurrent_queries>
+</yandex>
diff --git a/tests/config/install.sh b/tests/config/install.sh
index 416cc21893b..9965e1fb1ad 100755
--- a/tests/config/install.sh
+++ b/tests/config/install.sh
@@ -27,6 +27,7 @@ ln -sf $SRC_PATH/config.d/secure_ports.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/clusters.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/graphite.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/database_atomic.xml $DEST_SERVER_PATH/config.d/
+ln -sf $SRC_PATH/config.d/max_concurrent_queries.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/test_cluster_with_incorrect_pw.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/test_keeper_port.xml $DEST_SERVER_PATH/config.d/
 ln -sf $SRC_PATH/config.d/logging_no_rotate.xml $DEST_SERVER_PATH/config.d/
diff --git a/tests/config/top_level_domains b/tests/config/top_level_domains
deleted file mode 120000
index 7e12ab4ba2c..00000000000
--- a/tests/config/top_level_domains
+++ /dev/null
@@ -1 +0,0 @@
-../../docker/test/performance-comparison/config/top_level_domains
\ No newline at end of file
diff --git a/docker/test/performance-comparison/config/top_level_domains/public_suffix_list.dat b/tests/config/top_level_domains/public_suffix_list.dat
similarity index 100%
rename from docker/test/performance-comparison/config/top_level_domains/public_suffix_list.dat
rename to tests/config/top_level_domains/public_suffix_list.dat
diff --git a/tests/instructions/easy_tasks_sorted_ru.md b/tests/instructions/easy_tasks_sorted_ru.md
index 2dd60f97db3..2c38dcc2c8f 100644
--- a/tests/instructions/easy_tasks_sorted_ru.md
+++ b/tests/instructions/easy_tasks_sorted_ru.md
@@ -16,7 +16,7 @@
 
 `context.setSetting` - для выставления `max_memory_usage` и других.
 
-## Битовые операции для FixedString.
+## + Битовые операции для FixedString.
 
 bitAnd, bitOr, bitNot, bitXor для значения типа FixedString, интерпретируемого как набор бит.
 
@@ -26,7 +26,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 ```
 Потом используйте их в вашей функции.
 
-## Добавить generic вариант функций least, greatest.
+## + Добавить generic вариант функций least, greatest.
 
 `SELECT least(123, 456)` - работает.
 
@@ -38,7 +38,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 Смотрим, что все файлы в прикрепляемых кусках от правильного пользователя.
 
-## COLLATE должно работать для Nullable(String).
+## + COLLATE должно работать для Nullable(String).
 
 В ClickHouse есть возможность указать collation для сортировки строк. Это не работает для `Nullable(String)`.
 
@@ -46,13 +46,13 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 Состояния агрегатных функций могут быть записаны в дамп и считаны из него. Но десериализация состояний агрегатных функций небезопасна. Аккуратно выбранные пользовательские данные могут привести к segfault или порче памяти. Поэтому нужно просто сделать настройку, которая запрещает читать AggregateFunction из пользовательских данных.
 
-## В статистику jemalloc добавить информацию по arenas.
+## + В статистику jemalloc добавить информацию по arenas.
 
 В `system.asynchronous_metrics` - суммарный размер арен.
 
 # Более сложные задачи
 
-## Layout внешних словарей "direct".
+## + Layout внешних словарей "direct".
 
 Как cache, но без кэша — всегда прямой запрос в источник.
 
@@ -60,7 +60,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 ## Агрегатные функции для статистических тестов (e.g. тест нормальности распределения) и статистик.
 
-## Функции создания и обновления состояния агрегатной функции по одному кортежу аргументов.
+## + Функции создания и обновления состояния агрегатной функции по одному кортежу аргументов.
 
 В ClickHouse есть понятие - состояние вычисления агрегатной функции. Состояния агрегатных функций можно записывать в таблицы, складывать, финализировать и т. п. https://clickhouse.yandex/docs/ru/data_types/nested_data_structures/aggregatefunction/
 
@@ -85,7 +85,7 @@ void memoryBitAnd(const char * a, const char * b, char * result, size_t size);
 
 `changeYear(datetime, 2019)`
 
-## Исправить мерцание прогресс-бара в clickhouse-client.
+## + Исправить мерцание прогресс-бара в clickhouse-client.
 
 Это заметно при работе с серверами с большим пингом.
 Прогресс бар не должен мерцать.
@@ -107,15 +107,15 @@ https://dev.mysql.com/doc/refman/8.0/en/date-and-time-functions.html#function_da
 
 position с конца строки.
 
-## Функция indexOf должна поддерживать Enum-ы без cast-а.
+## + Функция indexOf должна поддерживать Enum-ы без cast-а.
 
 `indexOf(arr, 'hello')`, `indexOf(arr, 1)` должны работать, если arr имеет тип `Array(Enum8('hello' = 1, 'world' = 2))`
 
-## Комбинатор агрегатных функций Distinct.
+## + Комбинатор агрегатных функций Distinct.
 
 Пример: `avgDistinct(x)` - вычислить среднее по всем различным переданным значениям.
 
-## Метрики количества ошибок.
+## + Метрики количества ошибок.
 
 Добавляем счётчики всех ошибок (ErrorCodes) по аналогии с ProfileEvents. Кроме количества запоминаем также время последней ошибки, стек трейс, сообщение. Добавляем системную таблицу system.errors. Отправка в Graphite.
 
@@ -125,9 +125,9 @@ position с конца строки.
 
 ## Запрос CREATE OR REPLACE TABLE
 
-Атомарно (под блокировкой) удаляет таблицу перед созданием новой, если такая была.
+Атомарно удаляет таблицу перед созданием новой, если такая была.
 
-## Приведение типов для IN (subquery).
+## * Приведение типов для IN (subquery).
 
 `SELECT 1 IN (SELECT -1 UNION ALL SELECT 1)`
 
@@ -155,7 +155,7 @@ https://clickhouse.yandex/docs/en/query_language/create/#create-table
 
 Подключить одну из key-value БД как источник.
 
-## Движок таблиц Mongo, табличная функция mongo.
+## + Движок таблиц Mongo, табличная функция mongo.
 
 Возможность легко импортировать данные из MongoDB.
 
@@ -201,7 +201,7 @@ https://clickhouse.yandex/docs/en/operations/table_engines/external_data/
 
 ## Настройка rollup_use_nulls.
 
-## Настройка cast_keep_nullable.
+## + Настройка cast_keep_nullable.
 
 ## Функция bitEquals для сравнения произвольных типов данных побитово.
 
diff --git a/tests/integration/README.md b/tests/integration/README.md
index 0886dc2cfac..cdfb6b1a70a 100644
--- a/tests/integration/README.md
+++ b/tests/integration/README.md
@@ -38,7 +38,8 @@ sudo -H pip install \
     pytest-timeout \
     redis \
     tzlocal \
-    urllib3
+    urllib3 \
+    requests-kerberos
 ```
 
 (highly not recommended) If you really want to use OS packages on modern debian/ubuntu instead of "pip": `sudo apt install -y docker docker-compose python3-pytest python3-dicttoxml python3-docker python3-pymysql python3-pymongo python3-tzlocal python3-kazoo python3-psycopg2 kafka-python python3-pytest-timeout python3-minio`
@@ -135,3 +136,13 @@ named `test.py` containing tests in it. All functions with names starting with `
 To assert that two TSV files must be equal, wrap them in the `TSV` class and use the regular `assert`
 statement. Example: `assert TSV(result) == TSV(reference)`. In case the assertion fails, `pytest`
 will automagically detect the types of variables and only the small diff of two files is printed.
+
+### Troubleshooting
+
+If tests failing for misterious reasons, this may help:
+
+```
+sudo service docker stop
+sudo bash -c 'rm -rf /var/lib/docker/*'
+sudo service docker start
+```
diff --git a/tests/integration/helpers/cluster.py b/tests/integration/helpers/cluster.py
index 3967d016591..a65a420cd5b 100644
--- a/tests/integration/helpers/cluster.py
+++ b/tests/integration/helpers/cluster.py
@@ -137,6 +137,7 @@ class ClickHouseCluster:
         self.with_rabbitmq = False
         self.with_odbc_drivers = False
         self.with_hdfs = False
+        self.with_kerberized_hdfs = False
         self.with_mongo = False
         self.with_net_trics = False
         self.with_redis = False
@@ -172,7 +173,7 @@ class ClickHouseCluster:
                      macros=None,
                      with_zookeeper=False, with_mysql=False, with_kafka=False, with_kerberized_kafka=False, with_rabbitmq=False,
                      clickhouse_path_dir=None,
-                     with_odbc_drivers=False, with_postgres=False, with_hdfs=False, with_mongo=False,
+                     with_odbc_drivers=False, with_postgres=False, with_hdfs=False, with_kerberized_hdfs=False, with_mongo=False,
                      with_redis=False, with_minio=False, with_cassandra=False,
                      hostname=None, env_variables=None, image="yandex/clickhouse-integration-test", tag=None,
                      stay_alive=False, ipv4_address=None, ipv6_address=None, with_installed_binary=False, tmpfs=None,
@@ -210,6 +211,7 @@ class ClickHouseCluster:
             with_kafka=with_kafka,
             with_kerberized_kafka=with_kerberized_kafka,
             with_rabbitmq=with_rabbitmq,
+            with_kerberized_hdfs=with_kerberized_hdfs,
             with_mongo=with_mongo,
             with_redis=with_redis,
             with_minio=with_minio,
@@ -315,6 +317,14 @@ class ClickHouseCluster:
                                   p.join(docker_compose_yml_dir, 'docker_compose_hdfs.yml')]
             cmds.append(self.base_hdfs_cmd)
 
+        if with_kerberized_hdfs and not self.with_kerberized_hdfs:
+            self.with_kerberized_hdfs = True
+            self.base_cmd.extend(['--file', p.join(docker_compose_yml_dir, 'docker_compose_kerberized_hdfs.yml')])
+            self.base_kerberized_hdfs_cmd = ['docker-compose', '--project-directory', self.base_dir, '--project-name',
+                                             self.project_name, '--file',
+                                             p.join(docker_compose_yml_dir, 'docker_compose_kerberized_hdfs.yml')]
+            cmds.append(self.base_kerberized_hdfs_cmd)
+
         if with_mongo and not self.with_mongo:
             self.with_mongo = True
             self.base_cmd.extend(['--file', p.join(docker_compose_yml_dir, 'docker_compose_mongo.yml')])
@@ -476,12 +486,35 @@ class ClickHouseCluster:
 
         raise Exception("Cannot wait ZooKeeper container")
 
+    def make_hdfs_api(self, timeout=60, kerberized=False):
+        if kerberized:
+            keytab = p.abspath(p.join(self.instances['node1'].path, "secrets/clickhouse.keytab"))
+            krb_conf = p.abspath(p.join(self.instances['node1'].path, "secrets/krb_long.conf"))
+            hdfs_ip = self.get_instance_ip('kerberizedhdfs1')
+            # print("kerberizedhdfs1 ip ", hdfs_ip)
+            kdc_ip = self.get_instance_ip('hdfskerberos')
+            # print("kdc_ip ", kdc_ip)
+            self.hdfs_api = HDFSApi(user="root",
+                               timeout=timeout,
+                               kerberized=True,
+                               principal="root@TEST.CLICKHOUSE.TECH",
+                               keytab=keytab,
+                               krb_conf=krb_conf,
+                               host="kerberizedhdfs1",
+                               protocol="http",
+                               proxy_port=50070,
+                               data_port=1006,
+                               hdfs_ip=hdfs_ip,
+                               kdc_ip=kdc_ip)
+        else:
+            self.hdfs_api = HDFSApi(user="root", host="hdfs1")
+
+
     def wait_hdfs_to_start(self, timeout=60):
-        hdfs_api = HDFSApi("root")
         start = time.time()
         while time.time() - start < timeout:
             try:
-                hdfs_api.write_data("/somefilewithrandomname222", "1")
+                self.hdfs_api.write_data("/somefilewithrandomname222", "1")
                 print("Connected to HDFS and SafeMode disabled! ")
                 return
             except Exception as ex:
@@ -620,6 +653,7 @@ class ClickHouseCluster:
                 self.wait_schema_registry_to_start(120)
 
             if self.with_kerberized_kafka and self.base_kerberized_kafka_cmd:
+                print('Setup kerberized kafka')
                 env = os.environ.copy()
                 env['KERBERIZED_KAFKA_DIR'] = instance.path + '/'
                 subprocess.check_call(self.base_kerberized_kafka_cmd + common_opts + ['--renew-anon-volumes'], env=env)
@@ -631,8 +665,17 @@ class ClickHouseCluster:
             if self.with_hdfs and self.base_hdfs_cmd:
                 print('Setup HDFS')
                 subprocess_check_call(self.base_hdfs_cmd + common_opts)
+                self.make_hdfs_api()
                 self.wait_hdfs_to_start(120)
 
+            if self.with_kerberized_hdfs and self.base_kerberized_hdfs_cmd:
+                print('Setup kerberized HDFS')
+                env = os.environ.copy()
+                env['KERBERIZED_HDFS_DIR'] = instance.path + '/'
+                subprocess.check_call(self.base_kerberized_hdfs_cmd + common_opts, env=env)
+                self.make_hdfs_api(kerberized=True)
+                self.wait_hdfs_to_start(timeout=300)
+
             if self.with_mongo and self.base_mongo_cmd:
                 print('Setup Mongo')
                 subprocess_check_call(self.base_mongo_cmd + common_opts)
@@ -840,8 +883,8 @@ class ClickHouseInstance:
     def __init__(
             self, cluster, base_path, name, base_config_dir, custom_main_configs, custom_user_configs,
             custom_dictionaries,
-            macros, with_zookeeper, zookeeper_config_path, with_mysql, with_kafka, with_kerberized_kafka, with_rabbitmq, with_mongo,
-            with_redis, with_minio,
+            macros, with_zookeeper, zookeeper_config_path, with_mysql, with_kafka, with_kerberized_kafka, with_rabbitmq, with_kerberized_hdfs,
+            with_mongo, with_redis, with_minio,
             with_cassandra, server_bin_path, odbc_bridge_bin_path, clickhouse_path_dir, with_odbc_drivers,
             hostname=None, env_variables=None,
             image="yandex/clickhouse-integration-test", tag="latest",
@@ -871,6 +914,7 @@ class ClickHouseInstance:
         self.with_kafka = with_kafka
         self.with_kerberized_kafka = with_kerberized_kafka
         self.with_rabbitmq = with_rabbitmq
+        self.with_kerberized_hdfs = with_kerberized_hdfs
         self.with_mongo = with_mongo
         self.with_redis = with_redis
         self.with_minio = with_minio
@@ -885,7 +929,7 @@ class ClickHouseInstance:
         else:
             self.odbc_ini_path = ""
 
-        if with_kerberized_kafka:
+        if with_kerberized_kafka or with_kerberized_hdfs:
             self.keytab_path = '- ' + os.path.dirname(self.docker_compose_path) + "/secrets:/tmp/keytab"
             self.krb5_conf = '- ' + os.path.dirname(self.docker_compose_path) + "/secrets/krb.conf:/etc/krb5.conf:ro"
         else:
@@ -1231,7 +1275,7 @@ class ClickHouseInstance:
         if self.with_zookeeper:
             shutil.copy(self.zookeeper_config_path, conf_d_dir)
 
-        if self.with_kerberized_kafka:
+        if self.with_kerberized_kafka or self.with_kerberized_hdfs:
             shutil.copytree(self.kerberos_secrets_dir, p.abspath(p.join(self.path, 'secrets')))
 
         # Copy config.d configs
@@ -1272,6 +1316,9 @@ class ClickHouseInstance:
         if self.with_kerberized_kafka:
             depends_on.append("kerberized_kafka1")
 
+        if self.with_kerberized_hdfs:
+            depends_on.append("kerberizedhdfs1")
+
         if self.with_rabbitmq:
             depends_on.append("rabbitmq1")
 
diff --git a/tests/integration/helpers/hdfs_api.py b/tests/integration/helpers/hdfs_api.py
index d3afa9be837..cb742662855 100644
--- a/tests/integration/helpers/hdfs_api.py
+++ b/tests/integration/helpers/hdfs_api.py
@@ -2,45 +2,145 @@
 import io
 import gzip
 import subprocess
+import time
 from tempfile import NamedTemporaryFile
-
 import requests
+import requests_kerberos as reqkerb
+import socket
+import tempfile
+import logging
+import os
 
+g_dns_hook = None
+
+def custom_getaddrinfo(*args):
+    # print("from custom_getaddrinfo g_dns_hook is None ", g_dns_hook is None)
+    ret = g_dns_hook.custom_getaddrinfo(*args)
+    # print("g_dns_hook.custom_getaddrinfo result", ret)
+    return ret
+
+
+class mk_krb_conf(object):
+    def __init__(self, krb_conf, kdc_ip):
+        self.krb_conf = krb_conf
+        self.kdc_ip = kdc_ip
+        self.amended_krb_conf = None
+    def __enter__(self):
+        with open(self.krb_conf) as f:
+            content = f.read()
+        amended_content = content.replace('hdfskerberos', self.kdc_ip)
+        self.amended_krb_conf = tempfile.NamedTemporaryFile(delete=False, mode="w+")
+        self.amended_krb_conf.write(amended_content)
+        self.amended_krb_conf.close()
+        return self.amended_krb_conf.name
+    def __exit__(self, type, value, traceback):
+        if self.amended_krb_conf is not None:
+            self.amended_krb_conf.close()
+
+# tweak dns resolution to connect to localhost where api_host is in URL
+class dns_hook(object):
+    def __init__(self, hdfs_api):
+        # print("dns_hook.init ", hdfs_api.kerberized, hdfs_api.host, hdfs_api.data_port, hdfs_api.proxy_port)
+        self.hdfs_api = hdfs_api
+    def __enter__(self):
+        global g_dns_hook
+        g_dns_hook = self
+        # print("g_dns_hook is None ", g_dns_hook is None)
+        self.original_getaddrinfo = socket.getaddrinfo
+        socket.getaddrinfo = custom_getaddrinfo
+        return self
+    def __exit__(self, type, value, traceback):
+        global g_dns_hook
+        g_dns_hook = None
+        socket.getaddrinfo = self.original_getaddrinfo
+    def custom_getaddrinfo(self, *args):
+        (hostname, port) = args[:2]
+        # print("top of custom_getaddrinfo", hostname, port)
+
+        if hostname == self.hdfs_api.host and (port == self.hdfs_api.data_port or port == self.hdfs_api.proxy_port):
+            # print("dns_hook substitute")
+            return [(socket.AF_INET, 1, 6, '', ("127.0.0.1", port))]
+        else:
+            return self.original_getaddrinfo(*args)
 
 class HDFSApi(object):
-    def __init__(self, user):
-        self.host = "localhost"
-        self.http_proxy_port = "50070"
-        self.http_data_port = "50075"
+    def __init__(self, user, timeout=100, kerberized=False, principal=None,
+                 keytab=None, krb_conf=None,
+                 host = "localhost", protocol = "http",
+                 proxy_port = 50070, data_port = 50075, hdfs_ip = None, kdc_ip = None):
+        self.host = host
+        self.protocol = protocol
+        self.proxy_port = proxy_port
+        self.data_port = data_port
         self.user = user
+        self.kerberized = kerberized
+        self.principal = principal
+        self.keytab = keytab
+        self.timeout = timeout
+        self.hdfs_ip = hdfs_ip
+        self.kdc_ip = kdc_ip
+        self.krb_conf = krb_conf
+
+        # logging.basicConfig(level=logging.DEBUG)
+        # logging.getLogger().setLevel(logging.DEBUG)
+        # requests_log = logging.getLogger("requests.packages.urllib3")
+        # requests_log.setLevel(logging.DEBUG)
+        # requests_log.propagate = True
+
+        if kerberized:
+            self._run_kinit()
+            self.kerberos_auth = reqkerb.HTTPKerberosAuth(mutual_authentication=reqkerb.DISABLED, hostname_override=self.host, principal=self.principal)
+                #principal=self.principal,
+                #hostname_override=self.host, principal=self.principal)
+                # , mutual_authentication=reqkerb.REQUIRED, force_preemptive=True)
+        else:
+            self.kerberos_auth = None
+
+    def _run_kinit(self):
+        if self.principal is None or self.keytab is None:
+            raise Exception("kerberos principal and keytab are required")
+
+        with mk_krb_conf(self.krb_conf, self.kdc_ip) as instantiated_krb_conf:
+            # print("instantiated_krb_conf ", instantiated_krb_conf)
+
+            os.environ["KRB5_CONFIG"] = instantiated_krb_conf
+
+            cmd = "(kinit -R -t {keytab} -k {principal} || (sleep 5 && kinit -R -t {keytab} -k {principal})) ; klist".format(instantiated_krb_conf=instantiated_krb_conf, keytab=self.keytab, principal=self.principal)
+
+            # print(cmd)
+
+            start = time.time()
+
+            while time.time() - start < self.timeout:
+                try:
+                    subprocess.call(cmd, shell=True)
+                    print("KDC started, kinit successfully run")
+                    return
+                except Exception as ex:
+                    print("Can't run kinit ... waiting {}".format(str(ex)))
+                    time.sleep(1)
+
+        raise Exception("Kinit running failure")
 
     def read_data(self, path, universal_newlines=True):
-        response = requests.get(
-            "http://{host}:{port}/webhdfs/v1{path}?op=OPEN".format(host=self.host, port=self.http_proxy_port,
-                                                                   path=path), allow_redirects=False)
+        with dns_hook(self):
+            response = requests.get("{protocol}://{host}:{port}/webhdfs/v1{path}?op=OPEN".format(protocol=self.protocol, host=self.host, port=self.proxy_port, path=path), headers={'host': 'localhost'}, allow_redirects=False, verify=False, auth=self.kerberos_auth)
         if response.status_code != 307:
             response.raise_for_status()
-        additional_params = '&'.join(response.headers['Location'].split('&')[1:2])
-        response_data = requests.get(
-            "http://{host}:{port}/webhdfs/v1{path}?op=OPEN&{params}".format(host=self.host, port=self.http_data_port,
-                                                                            path=path, params=additional_params))
+        # additional_params = '&'.join(response.headers['Location'].split('&')[1:2])
+        url = "{location}".format(location=response.headers['Location'])
+        # print("redirected to ", url)
+        with dns_hook(self):
+            response_data = requests.get(url,
+                                         headers={'host': 'localhost'},
+                                         verify=False, auth=self.kerberos_auth)
         if response_data.status_code != 200:
             response_data.raise_for_status()
-
         if universal_newlines:
             return response_data.text
         else:
             return response_data.content
 
-    # Requests can't put file
-    def _curl_to_put(self, filename, path, params):
-        url = "http://{host}:{port}/webhdfs/v1{path}?op=CREATE&{params}".format(host=self.host,
-                                                                                port=self.http_data_port, path=path,
-                                                                                params=params)
-        cmd = "curl -s -i -X PUT -T {fname} '{url}'".format(fname=filename, url=url)
-        output = subprocess.check_output(cmd, shell=True, universal_newlines=True)
-        return output
-
     def write_data(self, path, content):
         named_file = NamedTemporaryFile(mode='wb+')
         fpath = named_file.name
@@ -48,19 +148,44 @@ class HDFSApi(object):
             content = content.encode()
         named_file.write(content)
         named_file.flush()
-        response = requests.put(
-            "http://{host}:{port}/webhdfs/v1{path}?op=CREATE".format(host=self.host, port=self.http_proxy_port,
-                                                                     path=path, user=self.user),
-            allow_redirects=False
-        )
+
+
+        if self.kerberized:
+            self._run_kinit()
+            self.kerberos_auth = reqkerb.HTTPKerberosAuth(mutual_authentication=reqkerb.DISABLED, hostname_override=self.host, principal=self.principal)
+            # print(self.kerberos_auth)
+
+        with dns_hook(self):
+            response = requests.put(
+                "{protocol}://{host}:{port}/webhdfs/v1{path}?op=CREATE".format(protocol=self.protocol, host=self.host,
+                                                                               port=self.proxy_port,
+                                                                               path=path, user=self.user),
+                allow_redirects=False,
+                headers={'host': 'localhost'},
+                params={'overwrite' : 'true'},
+                verify=False, auth=self.kerberos_auth
+            )
         if response.status_code != 307:
+            # print(response.headers)
             response.raise_for_status()
 
         additional_params = '&'.join(
             response.headers['Location'].split('&')[1:2] + ["user.name={}".format(self.user), "overwrite=true"])
-        output = self._curl_to_put(fpath, path, additional_params)
-        if "201 Created" not in output:
-            raise Exception("Can't create file on hdfs:\n {}".format(output))
+
+        with dns_hook(self), open(fpath, mode="rb") as fh:
+            file_data = fh.read()
+            protocol = "http" # self.protocol
+            response = requests.put(
+                "{location}".format(location=response.headers['Location']),
+                data=file_data,
+                headers={'content-type':'text/plain', 'host': 'localhost'},
+                params={'file': path, 'user.name' : self.user},
+                allow_redirects=False, verify=False, auth=self.kerberos_auth
+            )
+            # print(response)
+            if response.status_code != 201:
+                response.raise_for_status()
+
 
     def write_gzip_data(self, path, content):
         if isinstance(content, str):
diff --git a/tests/integration/test_allowed_url_from_config/test.py b/tests/integration/test_allowed_url_from_config/test.py
index 44715d92121..6442937c8f4 100644
--- a/tests/integration/test_allowed_url_from_config/test.py
+++ b/tests/integration/test_allowed_url_from_config/test.py
@@ -1,6 +1,5 @@
 import pytest
 from helpers.cluster import ClickHouseCluster
-from helpers.hdfs_api import HDFSApi
 
 cluster = ClickHouseCluster(__file__)
 node1 = cluster.add_instance('node1', main_configs=['configs/config_with_hosts.xml'])
@@ -101,9 +100,8 @@ def test_table_function_remote(start_cluster):
 
 
 def test_redirect(start_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\t\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\t\n"
+    start_cluster.hdfs_api.write_data("/simple_storage", "1\t\n")
+    assert start_cluster.hdfs_api.read_data("/simple_storage") == "1\t\n"
     node7.query(
         "CREATE TABLE table_test_7_1 (word String) ENGINE=URL('http://hdfs1:50070/webhdfs/v1/simple_storage?op=OPEN&namenoderpcaddress=hdfs1:9000&offset=0', CSV)")
     assert "not allowed" in node7.query_and_get_error("SET max_http_get_redirects=1; SELECT * from table_test_7_1")
diff --git a/tests/integration/test_graphite_merge_tree/configs/users.xml b/tests/integration/test_graphite_merge_tree/configs/users.xml
new file mode 100644
index 00000000000..cdd437797ce
--- /dev/null
+++ b/tests/integration/test_graphite_merge_tree/configs/users.xml
@@ -0,0 +1,8 @@
+<?xml version="1.0"?>
+<yandex>
+    <profiles>
+        <default>
+            <optimize_on_insert>0</optimize_on_insert>
+        </default>
+    </profiles>
+</yandex>
diff --git a/tests/integration/test_graphite_merge_tree/test.py b/tests/integration/test_graphite_merge_tree/test.py
index 502004d2dfe..ee8a18f1ca5 100644
--- a/tests/integration/test_graphite_merge_tree/test.py
+++ b/tests/integration/test_graphite_merge_tree/test.py
@@ -8,7 +8,8 @@ from helpers.test_tools import TSV
 
 cluster = ClickHouseCluster(__file__)
 instance = cluster.add_instance('instance',
-                                main_configs=['configs/graphite_rollup.xml'])
+                                main_configs=['configs/graphite_rollup.xml'],
+                                user_configs=["configs/users.xml"])
 q = instance.query
 
 
diff --git a/tests/integration/test_join_set_family_s3/__init__.py b/tests/integration/test_join_set_family_s3/__init__.py
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml b/tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml
new file mode 100644
index 00000000000..318a6bca95d
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/config.d/log_conf.xml
@@ -0,0 +1,12 @@
+<yandex>
+    <shutdown_wait_unfinished>3</shutdown_wait_unfinished>
+    <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/log.log</log>
+        <errorlog>/var/log/clickhouse-server/log.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+        <stderr>/var/log/clickhouse-server/stderr.log</stderr>
+        <stdout>/var/log/clickhouse-server/stdout.log</stdout>
+    </logger>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/config.xml b/tests/integration/test_join_set_family_s3/configs/config.xml
new file mode 100644
index 00000000000..5b9b5b5843a
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/config.xml
@@ -0,0 +1,28 @@
+<?xml version="1.0"?>
+<yandex>
+    <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/clickhouse-server.log</log>
+        <errorlog>/var/log/clickhouse-server/clickhouse-server.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+    </logger>
+
+    <tcp_port>9000</tcp_port>
+    <listen_host>127.0.0.1</listen_host>
+
+    <openSSL>
+        <client>
+            <cacheSessions>true</cacheSessions>
+            <verificationMode>none</verificationMode>
+            <invalidCertificateHandler>
+                <name>AcceptCertificateHandler</name>
+            </invalidCertificateHandler>
+        </client>
+    </openSSL>
+
+    <max_concurrent_queries>500</max_concurrent_queries>
+    <mark_cache_size>5368709120</mark_cache_size>
+    <path>./clickhouse/</path>
+    <users_config>users.xml</users_config>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/minio.xml b/tests/integration/test_join_set_family_s3/configs/minio.xml
new file mode 100644
index 00000000000..7337be1ad94
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/minio.xml
@@ -0,0 +1,14 @@
+<?xml version="1.0"?>
+<yandex>
+    <storage_configuration>
+        <disks>
+            <s3>
+                <type>s3</type>
+                <endpoint>http://minio1:9001/root/data/</endpoint>
+                <access_key_id>minio</access_key_id>
+                <secret_access_key>minio123</secret_access_key>
+                <send_object_metadata>true</send_object_metadata>
+            </s3>
+        </disks>
+    </storage_configuration>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/ssl.xml b/tests/integration/test_join_set_family_s3/configs/ssl.xml
new file mode 100644
index 00000000000..95cdc918bd0
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/ssl.xml
@@ -0,0 +1,12 @@
+<?xml version="1.0"?>
+<yandex>
+    <openSSL>
+        <client>
+            <cacheSessions>true</cacheSessions>
+            <verificationMode>none</verificationMode>
+            <invalidCertificateHandler>
+                <name>AcceptCertificateHandler</name>
+            </invalidCertificateHandler>
+        </client>
+    </openSSL>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/configs/users.xml b/tests/integration/test_join_set_family_s3/configs/users.xml
new file mode 100644
index 00000000000..6061af8e33d
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/configs/users.xml
@@ -0,0 +1,23 @@
+<?xml version="1.0"?>
+<yandex>
+    <profiles>
+        <default>
+        </default>
+    </profiles>
+
+    <users>
+        <default>
+            <password></password>
+            <networks incl="networks" replace="replace">
+                <ip>::/0</ip>
+            </networks>
+            <profile>default</profile>
+            <quota>default</quota>
+        </default>
+    </users>
+
+    <quotas>
+        <default>
+        </default>
+    </quotas>
+</yandex>
diff --git a/tests/integration/test_join_set_family_s3/test.py b/tests/integration/test_join_set_family_s3/test.py
new file mode 100644
index 00000000000..2a20c8eade9
--- /dev/null
+++ b/tests/integration/test_join_set_family_s3/test.py
@@ -0,0 +1,97 @@
+import logging
+import sys
+
+import pytest
+from helpers.cluster import ClickHouseCluster
+
+logging.getLogger().setLevel(logging.INFO)
+logging.getLogger().addHandler(logging.StreamHandler())
+
+
+@pytest.fixture(scope="module")
+def cluster():
+    try:
+        cluster = ClickHouseCluster(__file__)
+        cluster.add_instance("node",
+                             main_configs=["configs/minio.xml", "configs/ssl.xml", "configs/config.d/log_conf.xml"],
+                             with_minio=True, stay_alive=True)
+        logging.info("Starting cluster...")
+        cluster.start()
+
+        logging.info("Cluster started")
+
+        yield cluster
+    finally:
+        cluster.shutdown()
+
+
+def assert_objects_count(cluster, objects_count, path='data/'):
+    minio = cluster.minio_client
+    s3_objects = list(minio.list_objects(cluster.minio_bucket, path))
+    if objects_count != len(s3_objects):
+        for s3_object in s3_objects:
+            object_meta = minio.stat_object(cluster.minio_bucket, s3_object.object_name)
+            logging.info("Existing S3 object: %s", str(object_meta))
+        assert objects_count == len(s3_objects)
+
+
+def test_set_s3(cluster):
+    node = cluster.instances["node"]
+
+    node.query("CREATE TABLE testLocalSet (n UInt64) Engine = Set")
+    node.query("CREATE TABLE testS3Set (n UInt64) Engine = Set SETTINGS disk='s3'")
+
+    node.query("INSERT INTO TABLE testLocalSet VALUES (1)")
+    node.query("INSERT INTO TABLE testS3Set VALUES (1)")
+
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n1\t1\n0\t0\n"
+    assert_objects_count(cluster, 1)
+
+    node.query("INSERT INTO TABLE testLocalSet VALUES (2)")
+    node.query("INSERT INTO TABLE testS3Set VALUES (2)")
+
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n1\t1\n1\t1\n"
+    assert_objects_count(cluster, 2)
+
+    node.restart_clickhouse()
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n1\t1\n1\t1\n"
+
+    node.query("TRUNCATE TABLE testLocalSet")
+    node.query("TRUNCATE TABLE testS3Set")
+
+    assert node.query("SELECT number in testLocalSet, number in testS3Set FROM system.numbers LIMIT 3") == "0\t0\n0\t0\n0\t0\n"
+    assert_objects_count(cluster, 0)
+
+    node.query("DROP TABLE testLocalSet")
+    node.query("DROP TABLE testS3Set")
+
+
+def test_join_s3(cluster):
+    node = cluster.instances["node"]
+
+    node.query("CREATE TABLE testLocalJoin(`id` UInt64, `val` String) ENGINE = Join(ANY, LEFT, id)")
+    node.query("CREATE TABLE testS3Join(`id` UInt64, `val` String) ENGINE = Join(ANY, LEFT, id) SETTINGS disk='s3'")
+
+    node.query("INSERT INTO testLocalJoin VALUES (1, 'a')")
+    node.query("INSERT INTO testS3Join VALUES (1, 'a')")
+
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\na\ta\n\t\n"
+    assert_objects_count(cluster, 1)
+
+    node.query("INSERT INTO testLocalJoin VALUES (2, 'b')")
+    node.query("INSERT INTO testS3Join VALUES (2, 'b')")
+
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\na\ta\nb\tb\n"
+    assert_objects_count(cluster, 2)
+
+    node.restart_clickhouse()
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\na\ta\nb\tb\n"
+
+    node.query("TRUNCATE TABLE testLocalJoin")
+    node.query("TRUNCATE TABLE testS3Join")
+
+    assert node.query("SELECT joinGet('testLocalJoin', 'val', number) as local, joinGet('testS3Join', 'val', number) as s3 FROM system.numbers LIMIT 3") == "\t\n\t\n\t\n"
+    assert_objects_count(cluster, 0)
+
+    node.query("DROP TABLE testLocalJoin")
+    node.query("DROP TABLE testS3Join")
diff --git a/tests/integration/test_log_family_s3/configs/minio.xml b/tests/integration/test_log_family_s3/configs/minio.xml
index edcf1225142..7337be1ad94 100644
--- a/tests/integration/test_log_family_s3/configs/minio.xml
+++ b/tests/integration/test_log_family_s3/configs/minio.xml
@@ -2,13 +2,13 @@
 <yandex>
     <storage_configuration>
         <disks>
-            <default>
+            <s3>
                 <type>s3</type>
                 <endpoint>http://minio1:9001/root/data/</endpoint>
                 <access_key_id>minio</access_key_id>
                 <secret_access_key>minio123</secret_access_key>
                 <send_object_metadata>true</send_object_metadata>
-            </default>
+            </s3>
         </disks>
     </storage_configuration>
 </yandex>
diff --git a/tests/integration/test_log_family_s3/test.py b/tests/integration/test_log_family_s3/test.py
index 4ee370a7703..c23e7545b27 100644
--- a/tests/integration/test_log_family_s3/test.py
+++ b/tests/integration/test_log_family_s3/test.py
@@ -27,7 +27,6 @@ def cluster():
 def assert_objects_count(cluster, objects_count, path='data/'):
     minio = cluster.minio_client
     s3_objects = list(minio.list_objects(cluster.minio_bucket, path))
-    print(s3_objects, file=sys.stderr)
     if objects_count != len(s3_objects):
         for s3_object in s3_objects:
             object_meta = minio.stat_object(cluster.minio_bucket, s3_object.object_name)
@@ -41,7 +40,7 @@ def assert_objects_count(cluster, objects_count, path='data/'):
 def test_log_family_s3(cluster, log_engine, files_overhead, files_overhead_per_insert):
     node = cluster.instances["node"]
 
-    node.query("CREATE TABLE s3_test (id UInt64) Engine={}".format(log_engine))
+    node.query("CREATE TABLE s3_test (id UInt64) ENGINE={} SETTINGS disk = 's3'".format(log_engine))
 
     node.query("INSERT INTO s3_test SELECT number FROM numbers(5)")
     assert node.query("SELECT * FROM s3_test") == "0\n1\n2\n3\n4\n"
diff --git a/tests/integration/test_materialize_mysql_database/configs/users.xml b/tests/integration/test_materialize_mysql_database/configs/users.xml
index f6df1c30fc4..196ea5cfb99 100644
--- a/tests/integration/test_materialize_mysql_database/configs/users.xml
+++ b/tests/integration/test_materialize_mysql_database/configs/users.xml
@@ -4,6 +4,8 @@
         <default>
             <allow_experimental_database_materialize_mysql>1</allow_experimental_database_materialize_mysql>
             <allow_introspection_functions>1</allow_introspection_functions>
+            <optimize_on_insert>0</optimize_on_insert>
+            <default_database_engine>Ordinary</default_database_engine>
         </default>
     </profiles>
 
diff --git a/tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml b/tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml
new file mode 100644
index 00000000000..3add72ec554
--- /dev/null
+++ b/tests/integration/test_materialize_mysql_database/configs/users_db_atomic.xml
@@ -0,0 +1,19 @@
+<?xml version="1.0"?>
+<yandex>
+    <profiles>
+        <default>
+            <allow_experimental_database_materialize_mysql>1</allow_experimental_database_materialize_mysql>
+            <default_database_engine>Atomic</default_database_engine>
+        </default>
+    </profiles>
+
+    <users>
+        <default>
+            <password></password>
+            <networks incl="networks" replace="replace">
+                <ip>::/0</ip>
+            </networks>
+            <profile>default</profile>
+        </default>
+    </users>
+</yandex>
diff --git a/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py b/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
index 56b6d8b3a15..47e325d0e96 100644
--- a/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
+++ b/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
@@ -15,6 +15,7 @@ from multiprocessing.dummy import Pool
 
 def check_query(clickhouse_node, query, result_set, retry_count=60, interval_seconds=3):
     lastest_result = ''
+
     for i in range(retry_count):
         try:
             lastest_result = clickhouse_node.query(query)
@@ -35,6 +36,7 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
     clickhouse_node.query("DROP DATABASE IF EXISTS test_database")
     mysql_node.query("CREATE DATABASE test_database DEFAULT CHARACTER SET 'utf8'")
     # existed before the mapping was created
+
     mysql_node.query("CREATE TABLE test_database.test_table_1 ("
                      "`key` INT NOT NULL PRIMARY KEY, "
                      "unsigned_tiny_int TINYINT UNSIGNED, tiny_int TINYINT, "
@@ -51,9 +53,10 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
                      "_date Date, _datetime DateTime, _timestamp TIMESTAMP, _bool BOOLEAN) ENGINE = InnoDB;")
 
     # it already has some data
-    mysql_node.query(
-        "INSERT INTO test_database.test_table_1 VALUES(1, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', "
-        "'2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', true);")
+    mysql_node.query("""
+        INSERT INTO test_database.test_table_1 VALUES(1, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 
+        '2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', true);
+        """)
 
     clickhouse_node.query(
         "CREATE DATABASE test_database ENGINE = MaterializeMySQL('{}:3306', 'test_database', 'root', 'clickhouse')".format(
@@ -65,9 +68,10 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
                 "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
                 "2020-01-01 00:00:00\t2020-01-01 00:00:00\t1\n")
 
-    mysql_node.query(
-        "INSERT INTO test_database.test_table_1 VALUES(2, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', "
-        "'2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', false);")
+    mysql_node.query("""
+        INSERT INTO test_database.test_table_1 VALUES(2, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 
+        '2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', false);
+        """)
 
     check_query(clickhouse_node, "SELECT * FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
                 "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
@@ -76,14 +80,16 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
 
     mysql_node.query("UPDATE test_database.test_table_1 SET unsigned_tiny_int = 2 WHERE `key` = 1")
 
-    check_query(clickhouse_node, "SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,"
-                                 " small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, "
-                                 " unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, "
-                                 " _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ "
-                                 " _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
-                "1\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
-                "2020-01-01 00:00:00\t1\n2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
-                "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t0\n")
+    check_query(clickhouse_node, """
+        SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,
+         small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, 
+         unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, 
+         _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ 
+         _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV
+        """,
+        "1\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
+        "2020-01-01 00:00:00\t1\n2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
+        "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t0\n")
 
     # update primary key
     mysql_node.query("UPDATE test_database.test_table_1 SET `key` = 3 WHERE `unsigned_tiny_int` = 2")
@@ -472,10 +478,16 @@ def select_without_columns(clickhouse_node, mysql_node, service_name):
     clickhouse_node.query("CREATE VIEW v AS SELECT * FROM db.t")
     mysql_node.query("INSERT INTO db.t VALUES (1, 1), (2, 2)")
     mysql_node.query("DELETE FROM db.t WHERE a=2;")
-    check_query(clickhouse_node, "SELECT count((_sign, _version)) FROM db.t FORMAT TSV", "3\n")
 
-    assert clickhouse_node.query("SELECT count(_sign) FROM db.t FORMAT TSV") == "2\n"
-    assert clickhouse_node.query("SELECT count(_version) FROM db.t FORMAT TSV") == "2\n"
+    optimize_on_insert = clickhouse_node.query("SELECT value FROM system.settings WHERE name='optimize_on_insert'").strip()
+    if optimize_on_insert == "0":
+        res = ["3\n", "2\n", "2\n"]
+    else:
+        res = ["2\n", "2\n", "1\n"]
+    check_query(clickhouse_node, "SELECT count((_sign, _version)) FROM db.t FORMAT TSV", res[0])
+
+    assert clickhouse_node.query("SELECT count(_sign) FROM db.t FORMAT TSV") == res[1]
+    assert clickhouse_node.query("SELECT count(_version) FROM db.t FORMAT TSV") == res[2]
 
     assert clickhouse_node.query("SELECT count() FROM db.t FORMAT TSV") == "1\n"
     assert clickhouse_node.query("SELECT count(*) FROM db.t FORMAT TSV") == "1\n"
@@ -556,6 +568,12 @@ def err_sync_user_privs_with_materialize_mysql_database(clickhouse_node, mysql_n
     assert 'MySQL SYNC USER ACCESS ERR:' in str(exception.value)
     assert "priv_err_db" not in clickhouse_node.query("SHOW DATABASES")
 
+    mysql_node.query("GRANT SELECT ON priv_err_db.* TO 'test'@'%'")
+    time.sleep(3)
+    clickhouse_node.query("ATTACH DATABASE priv_err_db")
+    clickhouse_node.query("DROP DATABASE priv_err_db")
+    mysql_node.query("REVOKE SELECT ON priv_err_db.* FROM 'test'@'%'")
+
     mysql_node.query("DROP DATABASE priv_err_db;")
     mysql_node.query("DROP USER 'test'@'%'")
 
diff --git a/tests/integration/test_materialize_mysql_database/test.py b/tests/integration/test_materialize_mysql_database/test.py
index ed37e3d7502..1b7aa041540 100644
--- a/tests/integration/test_materialize_mysql_database/test.py
+++ b/tests/integration/test_materialize_mysql_database/test.py
@@ -14,7 +14,10 @@ from . import materialize_with_ddl
 DOCKER_COMPOSE_PATH = get_docker_compose_path()
 
 cluster = ClickHouseCluster(__file__)
-clickhouse_node = cluster.add_instance('node1', user_configs=["configs/users.xml"], with_mysql=False, stay_alive=True)
+
+node_db_ordinary = cluster.add_instance('node1', user_configs=["configs/users.xml"], with_mysql=False, stay_alive=True)
+node_db_atomic = cluster.add_instance('node2', user_configs=["configs/users_db_atomic.xml"], with_mysql=False, stay_alive=True)
+
 
 @pytest.fixture(scope="module")
 def started_cluster():
@@ -119,39 +122,30 @@ def started_mysql_8_0():
                                '--remove-orphans'])
 
 
-def test_materialize_database_dml_with_mysql_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_dml_with_mysql_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.dml_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
     materialize_with_ddl.materialize_mysql_database_with_datetime_and_decimal(clickhouse_node, started_mysql_5_7, "mysql1")
 
-
-def test_materialize_database_dml_with_mysql_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_dml_with_mysql_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.dml_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
     materialize_with_ddl.materialize_mysql_database_with_datetime_and_decimal(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_mysql_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
+    materialize_with_ddl.drop_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.create_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_add_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_drop_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    # mysql 5.7 cannot support alter rename column
+    # materialize_with_ddl.alter_rename_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
+    materialize_with_ddl.alter_modify_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
 
-
-def test_materialize_database_ddl_with_mysql_5_7(started_cluster, started_mysql_5_7):
-    try:
-        materialize_with_ddl.drop_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.create_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.alter_add_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                              "mysql1")
-        materialize_with_ddl.alter_drop_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                               "mysql1")
-        # mysql 5.7 cannot support alter rename column
-        # materialize_with_ddl.alter_rename_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
-        materialize_with_ddl.alter_rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                                "mysql1")
-        materialize_with_ddl.alter_modify_column_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7,
-                                                                                 "mysql1")
-    except:
-        print((clickhouse_node.query(
-            "select '\n', thread_id, query_id, arrayStringConcat(arrayMap(x -> concat(demangle(addressToSymbol(x)), '\n    ', addressToLine(x)), trace), '\n') AS sym from system.stack_trace format TSVRaw")))
-        raise
-
-
-def test_materialize_database_ddl_with_mysql_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_mysql_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.drop_table_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
     materialize_with_ddl.create_table_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
     materialize_with_ddl.rename_table_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
@@ -166,61 +160,72 @@ def test_materialize_database_ddl_with_mysql_8_0(started_cluster, started_mysql_
     materialize_with_ddl.alter_modify_column_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0,
                                                                              "mysql8_0")
 
-
-def test_materialize_database_ddl_with_empty_transaction_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_empty_transaction_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.query_event_with_empty_transaction(clickhouse_node, started_mysql_5_7, "mysql1")
 
-
-def test_materialize_database_ddl_with_empty_transaction_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_ddl_with_empty_transaction_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.query_event_with_empty_transaction(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
 
-def test_select_without_columns_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_select_without_columns_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.select_without_columns(clickhouse_node, started_mysql_5_7, "mysql1")
 
 
-def test_select_without_columns_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_select_without_columns_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.select_without_columns(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
 
-def test_insert_with_modify_binlog_checksum_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_insert_with_modify_binlog_checksum_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.insert_with_modify_binlog_checksum(clickhouse_node, started_mysql_5_7, "mysql1")
 
 
-def test_insert_with_modify_binlog_checksum_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_insert_with_modify_binlog_checksum_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.insert_with_modify_binlog_checksum(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
 
-def test_materialize_database_err_sync_user_privs_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_err_sync_user_privs_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.err_sync_user_privs_with_materialize_mysql_database(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_materialize_database_err_sync_user_privs_8_0(started_cluster, started_mysql_8_0):
+
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_materialize_database_err_sync_user_privs_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.err_sync_user_privs_with_materialize_mysql_database(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_network_partition_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_network_partition_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.network_partition_test(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_network_partition_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_network_partition_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.network_partition_test(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_mysql_kill_sync_thread_restore_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_kill_sync_thread_restore_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.mysql_kill_sync_thread_restore_test(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_mysql_kill_sync_thread_restore_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_kill_sync_thread_restore_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.mysql_kill_sync_thread_restore_test(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_mysql_killed_while_insert_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_killed_while_insert_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.mysql_killed_while_insert(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_mysql_killed_while_insert_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_mysql_killed_while_insert_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.mysql_killed_while_insert(clickhouse_node, started_mysql_8_0, "mysql8_0")
 
-
-def test_clickhouse_killed_while_insert_5_7(started_cluster, started_mysql_5_7):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_clickhouse_killed_while_insert_5_7(started_cluster, started_mysql_5_7, clickhouse_node):
     materialize_with_ddl.clickhouse_killed_while_insert(clickhouse_node, started_mysql_5_7, "mysql1")
 
-def test_clickhouse_killed_while_insert_8_0(started_cluster, started_mysql_8_0):
+@pytest.mark.parametrize(('clickhouse_node'), [node_db_ordinary, node_db_atomic])
+def test_clickhouse_killed_while_insert_8_0(started_cluster, started_mysql_8_0, clickhouse_node):
     materialize_with_ddl.clickhouse_killed_while_insert(clickhouse_node, started_mysql_8_0, "mysql8_0")
diff --git a/tests/integration/test_postgresql_protocol/test.py b/tests/integration/test_postgresql_protocol/test.py
index 52c911cb939..3436fc5bace 100644
--- a/tests/integration/test_postgresql_protocol/test.py
+++ b/tests/integration/test_postgresql_protocol/test.py
@@ -120,7 +120,7 @@ def test_python_client(server_address):
         cur.execute('select name from tables;')
 
     assert exc_info.value.args == (
-        "Query execution failed.\nDB::Exception: Table default.tables doesn't exist.\nSSL connection has been closed unexpectedly\n",)
+        "Query execution failed.\nDB::Exception: Table default.tables doesn't exist\nSSL connection has been closed unexpectedly\n",)
 
     ch = py_psql.connect(host=server_address, port=server_port, user='default', password='123', database='')
     cur = ch.cursor()
diff --git a/tests/integration/test_quorum_inserts/configs/users.d/settings.xml b/tests/integration/test_quorum_inserts/configs/users.d/settings.xml
index 29b762656aa..62d0d1ec39d 100644
--- a/tests/integration/test_quorum_inserts/configs/users.d/settings.xml
+++ b/tests/integration/test_quorum_inserts/configs/users.d/settings.xml
@@ -5,6 +5,7 @@
             <max_memory_usage>10000000000</max_memory_usage>
             <use_uncompressed_cache>0</use_uncompressed_cache>
             <insert_quorum>2</insert_quorum>
+            <insert_quorum_parallel>0</insert_quorum_parallel>
             <select_sequential_consistency>1</select_sequential_consistency>
         </default>
     </profiles>
diff --git a/tests/integration/test_redirect_url_storage/test.py b/tests/integration/test_redirect_url_storage/test.py
index f93548af0db..f2731794d43 100644
--- a/tests/integration/test_redirect_url_storage/test.py
+++ b/tests/integration/test_redirect_url_storage/test.py
@@ -17,9 +17,8 @@ def started_cluster():
 
 
 def test_url_without_redirect(started_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    started_cluster.hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
 
     # access datanode port directly
     node1.query(
@@ -28,9 +27,8 @@ def test_url_without_redirect(started_cluster):
 
 
 def test_url_with_redirect_not_allowed(started_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    started_cluster.hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
 
     # access proxy port without allowing redirects
     node1.query(
@@ -40,9 +38,8 @@ def test_url_with_redirect_not_allowed(started_cluster):
 
 
 def test_url_with_redirect_allowed(started_cluster):
-    hdfs_api = HDFSApi("root")
-    hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    started_cluster.hdfs_api.write_data("/simple_storage", "1\tMark\t72.53\n")
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
 
     # access proxy port with allowing redirects
     # http://localhost:50070/webhdfs/v1/b?op=OPEN&namenoderpcaddress=hdfs1:9000&offset=0
diff --git a/tests/integration/test_row_policy/configs/users.d/another_user.xml b/tests/integration/test_row_policy/configs/users.d/another_user.xml
index fb9608e5313..89e16e94c83 100644
--- a/tests/integration/test_row_policy/configs/users.d/another_user.xml
+++ b/tests/integration/test_row_policy/configs/users.d/another_user.xml
@@ -1,5 +1,10 @@
 <?xml version="1.0"?>
 <yandex>
+    <profiles>
+        <default>
+            <optimize_on_insert>0</optimize_on_insert>
+        </default>
+    </profiles>
     <users>
         <another>
             <password/>
@@ -10,4 +15,4 @@
             <quota>default</quota>
         </another>
     </users>
-</yandex>
\ No newline at end of file
+</yandex>
diff --git a/tests/integration/test_settings_profile/test.py b/tests/integration/test_settings_profile/test.py
index 5345ef9a474..3ceef9f25cf 100644
--- a/tests/integration/test_settings_profile/test.py
+++ b/tests/integration/test_settings_profile/test.py
@@ -207,34 +207,26 @@ def test_show_profiles():
 
 
 def test_allow_ddl():
-    assert "Not enough privileges" in instance.query_and_get_error("CREATE TABLE tbl(a Int32) ENGINE=Log", user="robin")
-    assert "DDL queries are prohibited" in instance.query_and_get_error("CREATE TABLE tbl(a Int32) ENGINE=Log",
-                                                                        settings={"allow_ddl": 0})
-
-    assert "Not enough privileges" in instance.query_and_get_error("GRANT CREATE ON tbl TO robin", user="robin")
-    assert "DDL queries are prohibited" in instance.query_and_get_error("GRANT CREATE ON tbl TO robin",
-                                                                        settings={"allow_ddl": 0})
-
+    assert "it's necessary to have grant" in instance.query_and_get_error("CREATE TABLE tbl(a Int32) ENGINE=Log", user="robin")
+    assert "it's necessary to have grant" in instance.query_and_get_error("GRANT CREATE ON tbl TO robin", user="robin")
+    assert "DDL queries are prohibited" in instance.query_and_get_error("CREATE TABLE tbl(a Int32) ENGINE=Log", settings={"allow_ddl": 0})
+    
     instance.query("GRANT CREATE ON tbl TO robin")
     instance.query("CREATE TABLE tbl(a Int32) ENGINE=Log", user="robin")
     instance.query("DROP TABLE tbl")
 
 
 def test_allow_introspection():
-    assert "Introspection functions are disabled" in instance.query_and_get_error("SELECT demangle('a')")
-    assert "Not enough privileges" in instance.query_and_get_error("SELECT demangle('a')", user="robin")
-    assert "Not enough privileges" in instance.query_and_get_error("SELECT demangle('a')", user="robin",
-                                                                   settings={"allow_introspection_functions": 1})
-
-    assert "Introspection functions are disabled" in instance.query_and_get_error("GRANT demangle ON *.* TO robin")
-    assert "Not enough privileges" in instance.query_and_get_error("GRANT demangle ON *.* TO robin", user="robin")
-    assert "Not enough privileges" in instance.query_and_get_error("GRANT demangle ON *.* TO robin", user="robin",
-                                                                   settings={"allow_introspection_functions": 1})
-
     assert instance.query("SELECT demangle('a')", settings={"allow_introspection_functions": 1}) == "signed char\n"
-    instance.query("GRANT demangle ON *.* TO robin", settings={"allow_introspection_functions": 1})
 
+    assert "Introspection functions are disabled" in instance.query_and_get_error("SELECT demangle('a')")
+    assert "it's necessary to have grant" in instance.query_and_get_error("SELECT demangle('a')", user="robin")
+    assert "it's necessary to have grant" in instance.query_and_get_error("SELECT demangle('a')", user="robin", settings={"allow_introspection_functions": 1})
+
+    instance.query("GRANT demangle ON *.* TO robin")
     assert "Introspection functions are disabled" in instance.query_and_get_error("SELECT demangle('a')", user="robin")
+    assert instance.query("SELECT demangle('a')", user="robin", settings={"allow_introspection_functions": 1}) == "signed char\n"
+
     instance.query("ALTER USER robin SETTINGS allow_introspection_functions=1")
     assert instance.query("SELECT demangle('a')", user="robin") == "signed char\n"
 
@@ -248,4 +240,4 @@ def test_allow_introspection():
     assert "Introspection functions are disabled" in instance.query_and_get_error("SELECT demangle('a')", user="robin")
 
     instance.query("REVOKE demangle ON *.* FROM robin", settings={"allow_introspection_functions": 1})
-    assert "Not enough privileges" in instance.query_and_get_error("SELECT demangle('a')", user="robin")
+    assert "it's necessary to have grant" in instance.query_and_get_error("SELECT demangle('a')", user="robin")
diff --git a/tests/integration/test_storage_hdfs/test.py b/tests/integration/test_storage_hdfs/test.py
index 59944a3725e..a6c8b7e1ee9 100644
--- a/tests/integration/test_storage_hdfs/test.py
+++ b/tests/integration/test_storage_hdfs/test.py
@@ -24,18 +24,14 @@ def started_cluster():
 
 
 def test_read_write_storage(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table SimpleHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/simple_storage', 'TSV')")
     node1.query("insert into SimpleHDFSStorage values (1, 'Mark', 72.53)")
-    assert hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
+    assert started_cluster.hdfs_api.read_data("/simple_storage") == "1\tMark\t72.53\n"
     assert node1.query("select * from SimpleHDFSStorage") == "1\tMark\t72.53\n"
 
 
 def test_read_write_storage_with_globs(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table HDFSStorageWithRange (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/storage{1..5}', 'TSV')")
     node1.query(
@@ -46,8 +42,8 @@ def test_read_write_storage_with_globs(started_cluster):
         "create table HDFSStorageWithAsterisk (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/storage*', 'TSV')")
 
     for i in ["1", "2", "3"]:
-        hdfs_api.write_data("/storage" + i, i + "\tMark\t72.53\n")
-        assert hdfs_api.read_data("/storage" + i) == i + "\tMark\t72.53\n"
+        started_cluster.hdfs_api.write_data("/storage" + i, i + "\tMark\t72.53\n")
+        assert started_cluster.hdfs_api.read_data("/storage" + i) == i + "\tMark\t72.53\n"
 
     assert node1.query("select count(*) from HDFSStorageWithRange") == "3\n"
     assert node1.query("select count(*) from HDFSStorageWithEnum") == "3\n"
@@ -77,25 +73,23 @@ def test_read_write_storage_with_globs(started_cluster):
 
 
 def test_read_write_table(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
-    hdfs_api.write_data("/simple_table_function", data)
+    started_cluster.hdfs_api.write_data("/simple_table_function", data)
 
-    assert hdfs_api.read_data("/simple_table_function") == data
+    assert started_cluster.hdfs_api.read_data("/simple_table_function") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64')") == data
 
 
 def test_write_table(started_cluster):
-    hdfs_api = HDFSApi("root")
 
     node1.query(
         "create table OtherHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/other_storage', 'TSV')")
     node1.query("insert into OtherHDFSStorage values (10, 'tomas', 55.55), (11, 'jack', 32.54)")
 
     result = "10\ttomas\t55.55\n11\tjack\t32.54\n"
-    assert hdfs_api.read_data("/other_storage") == result
+    assert started_cluster.hdfs_api.read_data("/other_storage") == result
     assert node1.query("select * from OtherHDFSStorage order by id") == result
 
 
@@ -120,15 +114,14 @@ def test_bad_hdfs_uri(started_cluster):
         print(ex)
         assert "Unable to open HDFS file" in str(ex)
 
-
+@pytest.mark.timeout(800)
 def test_globs_in_read_table(started_cluster):
-    hdfs_api = HDFSApi("root")
     some_data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
     globs_dir = "/dir_for_test_with_globs/"
     files = ["dir1/dir_dir/file1", "dir2/file2", "simple_table_function", "dir/file", "some_dir/dir1/file",
              "some_dir/dir2/file", "some_dir/file", "table1_function", "table2_function", "table3_function"]
     for filename in files:
-        hdfs_api.write_data(globs_dir + filename, some_data)
+        started_cluster.hdfs_api.write_data(globs_dir + filename, some_data)
 
     test_requests = [("dir{1..5}/dir_dir/file1", 1, 1),
                      ("*_table_functio?", 1, 1),
@@ -145,6 +138,7 @@ def test_globs_in_read_table(started_cluster):
 
     for pattern, paths_amount, files_amount in test_requests:
         inside_table_func = "'hdfs://hdfs1:9000" + globs_dir + pattern + "', 'TSV', 'id UInt64, text String, number Float64'"
+        print("inside_table_func ", inside_table_func)
         assert node1.query("select * from hdfs(" + inside_table_func + ")") == paths_amount * some_data
         assert node1.query("select count(distinct _path) from hdfs(" + inside_table_func + ")").rstrip() == str(
             paths_amount)
@@ -153,64 +147,61 @@ def test_globs_in_read_table(started_cluster):
 
 
 def test_read_write_gzip_table(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_gzip_data("/simple_table_function.gz", data)
+    started_cluster.hdfs_api.write_gzip_data("/simple_table_function.gz", data)
 
-    assert hdfs_api.read_gzip_data("/simple_table_function.gz") == data
+    assert started_cluster.hdfs_api.read_gzip_data("/simple_table_function.gz") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function.gz', 'TSV', 'id UInt64, text String, number Float64')") == data
 
 
 def test_read_write_gzip_table_with_parameter_gzip(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_gzip_data("/simple_table_function", data)
+    started_cluster.hdfs_api.write_gzip_data("/simple_table_function", data)
 
-    assert hdfs_api.read_gzip_data("/simple_table_function") == data
+    assert started_cluster.hdfs_api.read_gzip_data("/simple_table_function") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64', 'gzip')") == data
 
 
 def test_read_write_table_with_parameter_none(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_data("/simple_table_function.gz", data)
+    started_cluster.hdfs_api.write_data("/simple_table_function.gz", data)
 
-    assert hdfs_api.read_data("/simple_table_function.gz") == data
+    assert started_cluster.hdfs_api.read_data("/simple_table_function.gz") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function.gz', 'TSV', 'id UInt64, text String, number Float64', 'none')") == data
 
 
 def test_read_write_gzip_table_with_parameter_auto_gz(started_cluster):
-    hdfs_api = HDFSApi("root")
     data = "1\tHello Jessica\t555.222\n2\tI rolled a joint\t777.333\n"
-    hdfs_api.write_gzip_data("/simple_table_function.gz", data)
+    started_cluster.hdfs_api.write_gzip_data("/simple_table_function.gz", data)
 
-    assert hdfs_api.read_gzip_data("/simple_table_function.gz") == data
+    assert started_cluster.hdfs_api.read_gzip_data("/simple_table_function.gz") == data
 
     assert node1.query(
         "select * from hdfs('hdfs://hdfs1:9000/simple_table_function.gz', 'TSV', 'id UInt64, text String, number Float64', 'auto')") == data
 
 
 def test_write_gz_storage(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table GZHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/storage.gz', 'TSV')")
     node1.query("insert into GZHDFSStorage values (1, 'Mark', 72.53)")
-    assert hdfs_api.read_gzip_data("/storage.gz") == "1\tMark\t72.53\n"
+    assert started_cluster.hdfs_api.read_gzip_data("/storage.gz") == "1\tMark\t72.53\n"
     assert node1.query("select * from GZHDFSStorage") == "1\tMark\t72.53\n"
 
 
 def test_write_gzip_storage(started_cluster):
-    hdfs_api = HDFSApi("root")
-
     node1.query(
         "create table GZIPHDFSStorage (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://hdfs1:9000/gzip_storage', 'TSV', 'gzip')")
     node1.query("insert into GZIPHDFSStorage values (1, 'Mark', 72.53)")
-    assert hdfs_api.read_gzip_data("/gzip_storage") == "1\tMark\t72.53\n"
+    assert started_cluster.hdfs_api.read_gzip_data("/gzip_storage") == "1\tMark\t72.53\n"
     assert node1.query("select * from GZIPHDFSStorage") == "1\tMark\t72.53\n"
+
+if __name__ == '__main__':
+    cluster.start()
+    input("Cluster created, press any key to destroy...")
+    cluster.shutdown()
diff --git a/tests/integration/test_storage_kafka/test.py b/tests/integration/test_storage_kafka/test.py
index 06290053efa..07d2bcb60c0 100644
--- a/tests/integration/test_storage_kafka/test.py
+++ b/tests/integration/test_storage_kafka/test.py
@@ -9,6 +9,7 @@ import time
 import avro.schema
 from confluent_kafka.avro.cached_schema_registry_client import CachedSchemaRegistryClient
 from confluent_kafka.avro.serializer.message_serializer import MessageSerializer
+from confluent_kafka import admin
 
 import kafka.errors
 import pytest
@@ -1161,6 +1162,66 @@ def test_kafka_materialized_view(kafka_cluster):
 
     kafka_check_result(result, True)
 
+@pytest.mark.timeout(180)
+def test_librdkafka_snappy_regression(kafka_cluster):
+    """
+    Regression for UB in snappy-c (that is used in librdkafka),
+    backport pr is [1].
+
+      [1]: https://github.com/ClickHouse-Extras/librdkafka/pull/3
+
+    Example of corruption:
+
+        2020.12.10 09:59:56.831507 [ 20 ] {} <Error> void DB::StorageKafka::threadFunc(size_t): Code: 27, e.displayText() = DB::Exception: Cannot parse input: expected '"' before: 'foo"}': (while reading the value of key value): (at row 1)
+, Stack trace (when copying this message, always include the lines below):
+    """
+
+    # create topic with snappy compression
+    admin_client = admin.AdminClient({'bootstrap.servers': 'localhost:9092'})
+    topic_snappy = admin.NewTopic(topic='snappy_regression', num_partitions=1, replication_factor=1, config={
+        'compression.type': 'snappy',
+    })
+    admin_client.create_topics(new_topics=[topic_snappy], validate_only=False)
+
+    instance.query('''
+        CREATE TABLE test.kafka (key UInt64, value String)
+            ENGINE = Kafka
+            SETTINGS kafka_broker_list = 'kafka1:19092',
+                     kafka_topic_list = 'snappy_regression',
+                     kafka_group_name = 'ch_snappy_regression',
+                     kafka_format = 'JSONEachRow';
+    ''')
+
+    messages = []
+    expected = []
+    # To trigger this regression there should duplicated messages
+    # Orignal reproducer is:
+    #
+    #     $ gcc --version |& fgrep gcc
+    #     gcc (GCC) 10.2.0
+    #     $ yes foobarbaz | fold -w 80 | head -n10 >| in-…
+    #     $ make clean && make CFLAGS='-Wall -g -O2 -ftree-loop-vectorize -DNDEBUG=1 -DSG=1 -fPIC'
+    #     $ ./verify in
+    #     final comparision of in failed at 20 of 100
+    value = 'foobarbaz'*10
+    number_of_messages = 50
+    for i in range(number_of_messages):
+        messages.append(json.dumps({'key': i, 'value': value}))
+        expected.append(f'{i}\t{value}')
+    kafka_produce('snappy_regression', messages)
+
+    expected = '\n'.join(expected)
+
+    while True:
+        result = instance.query('SELECT * FROM test.kafka')
+        rows = len(result.strip('\n').split('\n'))
+        print(rows)
+        if rows == number_of_messages:
+            break
+
+    assert TSV(result) == TSV(expected)
+
+    instance.query('DROP TABLE test.kafka')
 
 @pytest.mark.timeout(180)
 def test_kafka_materialized_view_with_subquery(kafka_cluster):
diff --git a/tests/integration/test_storage_kerberized_hdfs/__init__.py b/tests/integration/test_storage_kerberized_hdfs/__init__.py
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml b/tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml
new file mode 100644
index 00000000000..3a93fa603c7
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/configs/hdfs.xml
@@ -0,0 +1,13 @@
+<yandex>
+  <hdfs>
+		<hadoop_kerberos_keytab>/tmp/keytab/clickhouse.keytab</hadoop_kerberos_keytab>
+		<hadoop_kerberos_principal>root@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+		<hadoop_security_authentication>kerberos</hadoop_security_authentication>
+  </hdfs>
+  <hdfs_suser>
+		<hadoop_kerberos_principal>specuser@TEST.CLICKHOUSE.TECH</hadoop_kerberos_principal>
+  </hdfs_suser>
+  <hdfs_dedicatedcachepath>
+		<hadoop_security_kerberos_ticket_cache_path>/tmp/kerb_cache</hadoop_security_kerberos_ticket_cache_path>
+  </hdfs_dedicatedcachepath>
+</yandex>
diff --git a/tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml b/tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml
new file mode 100644
index 00000000000..0de2745ca4c
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/configs/log_conf.xml
@@ -0,0 +1,11 @@
+<yandex>
+  <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/log.log</log>
+        <errorlog>/var/log/clickhouse-server/log.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+        <stderr>/var/log/clickhouse-server/stderr.log</stderr>
+        <stdout>/var/log/clickhouse-server/stdout.log</stdout>
+    </logger>
+</yandex>
diff --git a/tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh b/tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh
new file mode 100755
index 00000000000..38f098ae1e1
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/hdfs_configs/bootstrap.sh
@@ -0,0 +1,262 @@
+#!/bin/bash
+
+: "${HADOOP_PREFIX:=/usr/local/hadoop}"
+
+cat >> $HADOOP_PREFIX/etc/hadoop/hadoop-env.sh <<EOF
+export HADOOP_SECURE_DN_USER=hdfs
+export HADOOP_SECURE_DN_PID_DIR=$HADOOP_PREFIX/pid
+export HADOOP_SECURE_DN_LOG_DIR=$HADOOP_PREFIX/logs/hdfs
+export JSVC_HOME=$HADOOP_PREFIX/sbin
+EOF
+$HADOOP_PREFIX/etc/hadoop/hadoop-env.sh
+
+mkdir -p "${HADOOP_SECURE_DN_PID_DIR}"
+mkdir -p "${HADOOP_SECURE_DN_LOG_DIR}"
+
+rm /tmp/*.pid
+
+# installing libraries if any - (resource urls added comma separated to the ACP system variable)
+cd "${HADOOP_PREFIX}/share/hadoop/common" || exit
+for cp in ${ACP//,/ }; do  echo "== ${cp}"; curl -LO "${cp}" ; done;
+cd - || exit
+
+# altering the core-site configuration
+sed "s/HOSTNAME/${HOSTNAME}/" /usr/local/hadoop/etc/hadoop/core-site.xml.template | grep -v '/configuration' > /usr/local/hadoop/etc/hadoop/core-site.xml
+
+cat >> /usr/local/hadoop/etc/hadoop/core-site.xml << EOF
+  <property>
+    <name>hadoop.security.authentication</name>
+    <value>kerberos</value> <!-- A value of "simple" would disable security. -->
+  </property>
+   <property>
+     <name>hadoop.security.authorization</name>
+     <value>true</value>
+   </property>
+  <property>
+    <name>fs.defaultFS</name>
+    <value>hdfs://kerberizedhdfs1:9000</value>
+  </property>
+  <property>
+    <name>fs.default.name</name>
+    <value>hdfs://kerberizedhdfs1:9000</value>
+  </property>
+  <!--
+  <property>
+      <name>hadoop.rpc.protection</name>
+      <value>privacy</value>
+  </property>
+  -->
+</configuration>
+EOF
+
+
+cat > /usr/local/hadoop/etc/hadoop/hdfs-site.xml << EOF
+<configuration>
+<!--
+<property>
+	<name>dfs.permissions.enabled</name>
+	<value>false</value>
+</property>
+-->
+<property>
+  <name>dfs.replication</name>
+  <value>1</value>
+</property>
+<!-- General HDFS security config -->
+<property>
+  <name>dfs.block.access.token.enable</name>
+  <value>true</value>
+</property>
+<!-- NameNode security config -->
+<property>
+  <name>dfs.namenode.keytab.file</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+<property>
+  <name>dfs.namenode.kerberos.principal</name>
+  <value>hdfs/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<property>
+  <name>dfs.namenode.kerberos.internal.spnego.principal</name>
+  <value>HTTP/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<!-- Secondary NameNode security config -->
+<property>
+  <name>dfs.secondary.namenode.keytab.file</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+<property>
+  <name>dfs.secondary.namenode.kerberos.principal</name>
+  <value>hdfs/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<property>
+  <name>dfs.secondary.namenode.kerberos.internal.spnego.principal</name>
+  <value>HTTP/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<!-- DataNode security config
+<property>
+   <name>dfs.data.transfer.protectionл</name>
+   <value>integrity</value>
+</property>
+-->
+<property>
+  <name>dfs.datanode.data.dir.perm</name>
+  <value>700</value>
+</property>
+<property>
+  <name>dfs.datanode.address</name>
+  <value>0.0.0.0:1004</value>
+</property>
+<property>
+  <name>dfs.datanode.http.address</name>
+  <value>0.0.0.0:1006</value>
+</property>
+<!--
+<property>
+  <name>dfs.http.policy</name>
+  <value>HTTPS_ONLY</value>
+</property>
+-->
+
+
+<property>
+  <name>dfs.datanode.keytab.file</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+<property>
+  <name>dfs.datanode.kerberos.principal</name>
+  <value>hdfs/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+
+<!-- Web Authentication config -->
+<property>
+    <name>dfs.webhdfs.enabled</name>
+    <value>true</value>
+</property>
+<property>
+<name>dfs.encrypt.data.transfer</name>
+<value>false</value>
+</property>
+<property>
+  <name>dfs.web.authentication.kerberos.principal</name>
+  <value>HTTP/_HOST@TEST.CLICKHOUSE.TECH</value>
+</property>
+<property>
+    <name>dfs.web.authentication.kerberos.keytab</name>
+  <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.keytab</value> <!-- path to the HDFS keytab -->
+</property>
+</configuration>
+EOF
+
+
+
+# cat > /usr/local/hadoop/etc/hadoop/ssl-server.xml << EOF
+# <configuration>
+# <property>
+#   <name>ssl.server.truststore.location</name>
+#   <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.jks</value>
+# </property>
+# <property>
+#   <name>ssl.server.truststore.password</name>
+#   <value>masterkey</value>
+# </property>
+# <property>
+#   <name>ssl.server.keystore.location</name>
+#   <value>/usr/local/hadoop/etc/hadoop/conf/hdfs.jks</value>
+# </property>
+# <property>
+#   <name>ssl.server.keystore.password</name>
+#   <value>masterkey</value>
+# </property>
+# <property>
+#   <name>ssl.server.keystore.keypassword</name>
+#   <value>masterkey</value>
+# </property>
+# </configuration>
+# EOF
+
+cat > /usr/local/hadoop/etc/hadoop/log4j.properties << EOF
+# Set everything to be logged to the console
+log4j.rootCategory=DEBUG, console
+log4j.appender.console=org.apache.log4j.ConsoleAppender
+log4j.appender.console.target=System.err
+log4j.appender.console.layout=org.apache.log4j.PatternLayout
+log4j.appender.console.layout.ConversionPattern=%d{yy/MM/dd HH:mm:ss} %p %c{1}: %m%n
+
+# Set the default spark-shell log level to WARN. When running the spark-shell, the
+# log level for this class is used to overwrite the root logger's log level, so that
+# the user can have different defaults for the shell and regular Spark apps.
+log4j.logger.org.apache.spark.repl.Main=INFO
+
+# Settings to quiet third party logs that are too verbose
+log4j.logger.org.spark_project.jetty=DEBUG
+log4j.logger.org.spark_project.jetty.util.component.AbstractLifeCycle=ERROR
+log4j.logger.org.apache.spark.repl.SparkIMain$exprTyper=INFO
+log4j.logger.org.apache.spark.repl.SparkILoop$SparkILoopInterpreter=INFO
+log4j.logger.org.apache.parquet=ERROR
+log4j.logger.parquet=ERROR
+
+# SPARK-9183: Settings to avoid annoying messages when looking up nonexistent UDFs in SparkSQL with Hive support
+log4j.logger.org.apache.hadoop.hive.metastore.RetryingHMSHandler=FATAL
+log4j.logger.org.apache.hadoop.hive.ql.exec.FunctionRegistry=ERROR
+
+log4j.logger.org.apache.spark.deploy=DEBUG
+log4j.logger.org.apache.spark.executor=DEBUG
+log4j.logger.org.apache.spark.scheduler=DEBUG
+EOF
+
+useradd -u 1098 hdfs
+
+# keytool -genkey -alias kerberized_hdfs1.test.clickhouse.tech -keyalg rsa -keysize 1024 -dname "CN=kerberized_hdfs1.test.clickhouse.tech" -keypass masterkey -keystore /usr/local/hadoop/etc/hadoop/conf/hdfs.jks -storepass masterkey
+keytool -genkey -alias kerberizedhdfs1 -keyalg rsa -keysize 1024 -dname "CN=kerberizedhdfs1" -keypass masterkey -keystore /usr/local/hadoop/etc/hadoop/conf/hdfs.jks -storepass masterkey
+
+chmod g+r /usr/local/hadoop/etc/hadoop/conf/hdfs.jks
+
+
+service sshd start
+
+# yum --quiet --assumeyes install krb5-workstation.x86_64
+# yum --quiet --assumeyes install tcpdump
+
+# cd /tmp
+# curl http://archive.apache.org/dist/commons/daemon/source/commons-daemon-1.0.15-src.tar.gz   -o  commons-daemon-1.0.15-src.tar.gz
+# tar xzf commons-daemon-1.0.15-src.tar.gz
+# cd commons-daemon-1.0.15-src/src/native/unix
+# ./configure && make
+# cp ./jsvc /usr/local/hadoop/sbin
+
+
+until kinit -kt /usr/local/hadoop/etc/hadoop/conf/hdfs.keytab hdfs/kerberizedhdfs1@TEST.CLICKHOUSE.TECH; do sleep 2; done
+echo "KDC is up and ready to go... starting up"
+
+$HADOOP_PREFIX/sbin/start-dfs.sh
+$HADOOP_PREFIX/sbin/start-yarn.sh
+$HADOOP_PREFIX/sbin/mr-jobhistory-daemon.sh start historyserver
+
+chmod a+r /usr/local/hadoop/etc/hadoop/conf/hdfs.keytab   # create dedicated keytab for hdfsuser
+
+$HADOOP_PREFIX/sbin/start-secure-dns.sh
+sleep 3
+
+/usr/local/hadoop/bin/hdfs dfsadmin -safemode leave
+
+/usr/local/hadoop/bin/hdfs dfs -mkdir /user/specuser
+/usr/local/hadoop/bin/hdfs dfs -chown specuser /user/specuser
+
+kdestroy
+
+
+
+# adduser --groups hdfs hdfsuser
+
+# /usr/local/hadoop/sbin/hadoop-daemon.sh --config /usr/local/hadoop/etc/hadoop/ --script /usr/local/hadoop/sbin/hdfs start namenode
+# /usr/local/hadoop/sbin/hadoop-daemon.sh --config /usr/local/hadoop/etc/hadoop/ --script /usr/local/hadoop/sbin/hdfs start datanode
+
+
+if [[ $1 == "-d" ]]; then
+  while true; do sleep 1000; done
+fi
+
+if [[ $1 == "-bash" ]]; then
+  /bin/bash
+fi
diff --git a/tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh b/tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh
new file mode 100644
index 00000000000..daa060cbe29
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/kerberos_image_config.sh
@@ -0,0 +1,139 @@
+#!/bin/bash
+
+
+set -x  #  trace
+
+: "${REALM:=TEST.CLICKHOUSE.TECH}"
+: "${DOMAIN_REALM:=test.clickhouse.tech}"
+: "${KERB_MASTER_KEY:=masterkey}"
+: "${KERB_ADMIN_USER:=admin}"
+: "${KERB_ADMIN_PASS:=admin}"
+
+create_config() {
+  : "${KDC_ADDRESS:=$(hostname -f)}"
+
+  cat>/etc/krb5.conf<<EOF
+[logging]
+ default = FILE:/var/log/kerberos/krb5libs.log
+ kdc = FILE:/var/log/kerberos/krb5kdc.log
+ admin_server = FILE:/var/log/kerberos/kadmind.log
+
+[libdefaults]
+ default_realm = $REALM
+ dns_lookup_realm = false
+ dns_lookup_kdc = false
+ ticket_lifetime = 15d
+ renew_lifetime = 15d
+ forwardable = true
+ # WARNING: We use weaker key types to simplify testing as stronger key types
+ # require the enhanced security JCE policy file to be installed. You should
+ # NOT run with this configuration in production or any real environment. You
+ # have been warned.
+ default_tkt_enctypes = des-cbc-md5 des-cbc-crc des3-cbc-sha1
+ default_tgs_enctypes = des-cbc-md5 des-cbc-crc des3-cbc-sha1
+ permitted_enctypes = des-cbc-md5 des-cbc-crc des3-cbc-sha1
+
+[realms]
+ $REALM = {
+  kdc = $KDC_ADDRESS
+  admin_server = $KDC_ADDRESS
+ }
+
+[domain_realm]
+ .$DOMAIN_REALM = $REALM
+ $DOMAIN_REALM = $REALM
+EOF
+
+cat>/var/kerberos/krb5kdc/kdc.conf<<EOF
+[kdcdefaults]
+ kdc_ports = 88
+ kdc_tcp_ports = 88
+
+[realms]
+ $REALM = {
+  acl_file = /var/kerberos/krb5kdc/kadm5.acl
+  dict_file = /usr/share/dict/words
+  admin_keytab = /var/kerberos/krb5kdc/kadm5.keytab
+  # WARNING: We use weaker key types to simplify testing as stronger key types
+  # require the enhanced security JCE policy file to be installed. You should
+  # NOT run with this configuration in production or any real environment. You
+  # have been warned.
+  master_key_type = des3-hmac-sha1
+  supported_enctypes = arcfour-hmac:normal des3-hmac-sha1:normal des-cbc-crc:normal des:normal des:v4 des:norealm des:onlyrealm des:afs3
+  default_principal_flags = +preauth
+ }
+EOF
+}
+
+create_db() {
+  /usr/sbin/kdb5_util -P $KERB_MASTER_KEY -r $REALM create -s
+}
+
+start_kdc() {
+  mkdir -p /var/log/kerberos
+
+  /etc/rc.d/init.d/krb5kdc start
+  /etc/rc.d/init.d/kadmin start
+
+  chkconfig krb5kdc on
+  chkconfig kadmin on
+}
+
+restart_kdc() {
+  /etc/rc.d/init.d/krb5kdc restart
+  /etc/rc.d/init.d/kadmin restart
+}
+
+create_admin_user() {
+  kadmin.local -q "addprinc -pw $KERB_ADMIN_PASS $KERB_ADMIN_USER/admin"
+  echo "*/admin@$REALM *" > /var/kerberos/krb5kdc/kadm5.acl
+}
+
+create_keytabs() {
+
+
+  # kadmin.local -q "addprinc -randkey hdfs/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+  # kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab hdfs/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+
+  # kadmin.local -q "addprinc -randkey HTTP/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+  # kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab HTTP/kerberizedhdfs1.${DOMAIN_REALM}@${REALM}"
+  kadmin.local -q "addprinc -randkey hdfs/kerberizedhdfs1@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab hdfs/kerberizedhdfs1@${REALM}"
+
+  kadmin.local -q "addprinc -randkey HTTP/kerberizedhdfs1@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/hdfs.keytab HTTP/kerberizedhdfs1@${REALM}"
+
+  kadmin.local -q "addprinc -randkey hdfsuser/node1@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab hdfsuser/node1@${REALM}"
+  kadmin.local -q "addprinc -randkey hdfsuser@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab hdfsuser@${REALM}"
+  kadmin.local -q "addprinc -randkey root@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab root@${REALM}"
+  kadmin.local -q "addprinc -randkey specuser@${REALM}"
+  kadmin.local -q "ktadd -norandkey -k /tmp/keytab/clickhouse.keytab specuser@${REALM}"
+
+  chmod g+r /tmp/keytab/clickhouse.keytab
+}
+
+main() {
+
+  if [ ! -f /kerberos_initialized ]; then
+    create_config
+    create_db
+    create_admin_user
+    start_kdc
+
+    touch /kerberos_initialized
+  fi
+
+  if [ ! -f /var/kerberos/krb5kdc/principal ]; then
+    while true; do sleep 1000; done
+  else
+    start_kdc
+		create_keytabs
+    tail -F /var/log/kerberos/krb5kdc.log
+  fi
+
+}
+
+[[ "$0" == "${BASH_SOURCE[0]}" ]] && main "$@"
diff --git a/tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf b/tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf
new file mode 100644
index 00000000000..2c1e6f15f77
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/secrets/krb.conf
@@ -0,0 +1,24 @@
+[logging]
+ default = FILE:/var/log/krb5libs.log
+ kdc = FILE:/var/log/krb5kdc.log
+ admin_server = FILE:/var/log/kadmind.log
+
+[libdefaults]
+ default_realm = TEST.CLICKHOUSE.TECH
+ dns_lookup_realm = false
+ dns_lookup_kdc = false
+ ticket_lifetime = 15s
+ forwardable = true
+ default_tgs_enctypes = des3-hmac-sha1
+ default_tkt_enctypes = des3-hmac-sha1
+ permitted_enctypes = des3-hmac-sha1
+
+[realms]
+ TEST.CLICKHOUSE.TECH = {
+  kdc = hdfskerberos
+  admin_server = hdfskerberos
+ }
+
+[domain_realm]
+ .test.clickhouse.tech = TEST.CLICKHOUSE.TECH
+ test.clickhouse.tech = TEST.CLICKHOUSE.TECH
diff --git a/tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf b/tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf
new file mode 100644
index 00000000000..ec1c54240e8
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/secrets/krb_long.conf
@@ -0,0 +1,24 @@
+[logging]
+ default = FILE:/var/log/krb5libs.log
+ kdc = FILE:/var/log/krb5kdc.log
+ admin_server = FILE:/var/log/kadmind.log
+
+[libdefaults]
+ default_realm = TEST.CLICKHOUSE.TECH
+ dns_lookup_realm = false
+ dns_lookup_kdc = false
+ ticket_lifetime = 15d
+ forwardable = true
+ default_tgs_enctypes = des3-hmac-sha1
+ default_tkt_enctypes = des3-hmac-sha1
+ permitted_enctypes = des3-hmac-sha1
+
+[realms]
+ TEST.CLICKHOUSE.TECH = {
+  kdc = hdfskerberos
+  admin_server = hdfskerberos
+ }
+
+[domain_realm]
+ .test.clickhouse.tech = TEST.CLICKHOUSE.TECH
+ test.clickhouse.tech = TEST.CLICKHOUSE.TECH
diff --git a/tests/integration/test_storage_kerberized_hdfs/test.py b/tests/integration/test_storage_kerberized_hdfs/test.py
new file mode 100644
index 00000000000..a2a2a4ef88e
--- /dev/null
+++ b/tests/integration/test_storage_kerberized_hdfs/test.py
@@ -0,0 +1,108 @@
+import time
+import pytest
+
+import os
+
+from helpers.cluster import ClickHouseCluster
+import subprocess
+
+cluster = ClickHouseCluster(__file__)
+node1 = cluster.add_instance('node1', with_kerberized_hdfs=True, user_configs=[], main_configs=['configs/log_conf.xml', 'configs/hdfs.xml'])
+
+@pytest.fixture(scope="module")
+def started_cluster():
+    try:
+        cluster.start()
+
+        yield cluster
+
+    except Exception as ex:
+        print(ex)
+        raise ex
+    finally:
+        cluster.shutdown()
+
+def test_read_table(started_cluster):
+    data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
+    started_cluster.hdfs_api.write_data("/simple_table_function", data)
+
+    api_read = started_cluster.hdfs_api.read_data("/simple_table_function")
+    assert api_read == data
+
+    select_read = node1.query("select * from hdfs('hdfs://kerberizedhdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64')")
+    assert select_read == data
+
+
+def test_read_write_storage(started_cluster):
+    node1.query("create table SimpleHDFSStorage2 (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://kerberizedhdfs1:9000/simple_storage1', 'TSV')")
+    node1.query("insert into SimpleHDFSStorage2 values (1, 'Mark', 72.53)")
+
+    api_read = started_cluster.hdfs_api.read_data("/simple_storage1")
+    assert api_read == "1\tMark\t72.53\n"
+
+    select_read = node1.query("select * from SimpleHDFSStorage2")
+    assert select_read == "1\tMark\t72.53\n"
+
+
+def test_write_storage_not_expired(started_cluster):
+    node1.query("create table SimpleHDFSStorageNotExpired (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://kerberizedhdfs1:9000/simple_storage_not_expired', 'TSV')")
+
+    time.sleep(45)   # wait for ticket expiration
+    node1.query("insert into SimpleHDFSStorageNotExpired values (1, 'Mark', 72.53)")
+
+    api_read = started_cluster.hdfs_api.read_data("/simple_storage_not_expired")
+    assert api_read == "1\tMark\t72.53\n"
+
+    select_read = node1.query("select * from SimpleHDFSStorageNotExpired")
+    assert select_read == "1\tMark\t72.53\n"
+
+
+def test_two_users(started_cluster):
+    node1.query("create table HDFSStorOne (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://kerberizedhdfs1:9000/storage_user_one', 'TSV')")
+    node1.query("insert into HDFSStorOne values (1, 'Real', 86.00)")
+
+    node1.query("create table HDFSStorTwo (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://suser@kerberizedhdfs1:9000/user/specuser/storage_user_two', 'TSV')")
+    node1.query("insert into HDFSStorTwo values (1, 'Ideal', 74.00)")
+
+    select_read_1 = node1.query("select * from hdfs('hdfs://kerberizedhdfs1:9000/user/specuser/storage_user_two', 'TSV', 'id UInt64, text String, number Float64')")
+
+    select_read_2 = node1.query("select * from hdfs('hdfs://suser@kerberizedhdfs1:9000/storage_user_one', 'TSV', 'id UInt64, text String, number Float64')")
+
+def test_read_table_expired(started_cluster):
+    data = "1\tSerialize\t555.222\n2\tData\t777.333\n"
+    started_cluster.hdfs_api.write_data("/simple_table_function_relogin", data)
+
+    started_cluster.pause_container('hdfskerberos')
+    time.sleep(45)
+
+    try:
+        select_read = node1.query("select * from hdfs('hdfs://reloginuser&kerberizedhdfs1:9000/simple_table_function', 'TSV', 'id UInt64, text String, number Float64')")
+        assert False, "Exception have to be thrown"
+    except Exception as ex:
+        assert "DB::Exception: kinit failure:" in str(ex)
+
+    started_cluster.unpause_container('hdfskerberos')
+
+
+def test_prohibited(started_cluster):
+    node1.query("create table HDFSStorTwoProhibited (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://suser@kerberizedhdfs1:9000/storage_user_two_prohibited', 'TSV')")
+    try:
+        node1.query("insert into HDFSStorTwoProhibited values (1, 'SomeOne', 74.00)")
+        assert False, "Exception have to be thrown"
+    except Exception as ex:
+        assert "Unable to open HDFS file: /storage_user_two_prohibited error: Permission denied: user=specuser, access=WRITE" in str(ex)
+
+
+def test_cache_path(started_cluster):
+    node1.query("create table HDFSStorCachePath (id UInt32, name String, weight Float64) ENGINE = HDFS('hdfs://dedicatedcachepath@kerberizedhdfs1:9000/storage_dedicated_cache_path', 'TSV')")
+    try:
+        node1.query("insert into HDFSStorCachePath values (1, 'FatMark', 92.53)")
+        assert False, "Exception have to be thrown"
+    except Exception as ex:
+        assert "DB::Exception: hadoop.security.kerberos.ticket.cache.path cannot be set per user" in str(ex)
+
+
+if __name__ == '__main__':
+    cluster.start()
+    input("Cluster created, press any key to destroy...")
+    cluster.shutdown()
diff --git a/tests/integration/test_storage_s3/test.py b/tests/integration/test_storage_s3/test.py
index ca7206ca3b5..9e49c5ebf8f 100644
--- a/tests/integration/test_storage_s3/test.py
+++ b/tests/integration/test_storage_s3/test.py
@@ -306,7 +306,8 @@ def test_multipart_put(cluster, maybe_auth, positive):
         cluster.minio_redirect_host, cluster.minio_redirect_port, bucket, filename, maybe_auth, table_format)
 
     try:
-        run_query(instance, put_query, stdin=csv_data, settings={'s3_min_upload_part_size': min_part_size_bytes})
+        run_query(instance, put_query, stdin=csv_data, settings={'s3_min_upload_part_size': min_part_size_bytes,
+                                                                 's3_max_single_part_upload_size': 0})
     except helpers.client.QueryRuntimeException:
         if positive:
             raise
diff --git a/tests/integration/test_table_functions_access_rights/__init__.py b/tests/integration/test_table_functions_access_rights/__init__.py
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/integration/test_table_functions_access_rights/test.py b/tests/integration/test_table_functions_access_rights/test.py
new file mode 100644
index 00000000000..bd2f767413a
--- /dev/null
+++ b/tests/integration/test_table_functions_access_rights/test.py
@@ -0,0 +1,55 @@
+import pytest
+from helpers.cluster import ClickHouseCluster
+from helpers.test_tools import TSV
+
+cluster = ClickHouseCluster(__file__)
+instance = cluster.add_instance('instance')
+
+
+@pytest.fixture(scope="module", autouse=True)
+def started_cluster():
+    try:
+        cluster.start()
+
+        instance.query("CREATE TABLE table1(x UInt32) ENGINE = MergeTree ORDER BY tuple()")
+        instance.query("CREATE TABLE table2(x UInt32) ENGINE = MergeTree ORDER BY tuple()")
+        instance.query("INSERT INTO table1 VALUES (1)")
+        instance.query("INSERT INTO table2 VALUES (2)")
+
+        yield cluster
+
+    finally:
+        cluster.shutdown()
+
+
+@pytest.fixture(autouse=True)
+def cleanup_after_test():
+    try:
+        yield
+    finally:
+        instance.query("DROP USER IF EXISTS A")
+
+
+def test_merge():
+    select_query = "SELECT * FROM merge('default', 'table[0-9]+') ORDER BY x"
+    assert instance.query(select_query) == "1\n2\n"
+
+    instance.query("CREATE USER A")
+    assert "it's necessary to have grant CREATE TEMPORARY TABLE ON *.*" in instance.query_and_get_error(select_query, user = 'A')
+
+    instance.query("GRANT CREATE TEMPORARY TABLE ON *.* TO A")
+    assert "no one matches regular expression" in instance.query_and_get_error(select_query, user = 'A')
+    
+    instance.query("GRANT SELECT ON default.table1 TO A")
+    assert instance.query(select_query, user = 'A') == "1\n"
+
+    instance.query("GRANT SELECT ON default.* TO A")
+    assert instance.query(select_query, user = 'A') == "1\n2\n"
+
+    instance.query("REVOKE SELECT ON default.table1 FROM A")
+    assert instance.query(select_query, user = 'A') == "2\n"
+
+    instance.query("REVOKE ALL ON default.* FROM A")
+    instance.query("GRANT SELECT ON default.table1 TO A")
+    instance.query("GRANT INSERT ON default.table2 TO A")
+    assert "it's necessary to have grant SELECT ON default.table2" in instance.query_and_get_error(select_query, user = 'A')
diff --git a/tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml b/tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml
new file mode 100644
index 00000000000..318a6bca95d
--- /dev/null
+++ b/tests/integration/test_testkeeper_back_to_back/configs/logs_conf.xml
@@ -0,0 +1,12 @@
+<yandex>
+    <shutdown_wait_unfinished>3</shutdown_wait_unfinished>
+    <logger>
+        <level>trace</level>
+        <log>/var/log/clickhouse-server/log.log</log>
+        <errorlog>/var/log/clickhouse-server/log.err.log</errorlog>
+        <size>1000M</size>
+        <count>10</count>
+        <stderr>/var/log/clickhouse-server/stderr.log</stderr>
+        <stdout>/var/log/clickhouse-server/stdout.log</stdout>
+    </logger>
+</yandex>
diff --git a/tests/integration/test_testkeeper_back_to_back/test.py b/tests/integration/test_testkeeper_back_to_back/test.py
index 04a2150e7f8..f74d6a4c646 100644
--- a/tests/integration/test_testkeeper_back_to_back/test.py
+++ b/tests/integration/test_testkeeper_back_to_back/test.py
@@ -4,9 +4,10 @@ import random
 import string
 import os
 import time
+from multiprocessing.dummy import Pool
 
 cluster = ClickHouseCluster(__file__)
-node = cluster.add_instance('node', main_configs=['configs/enable_test_keeper.xml'], with_zookeeper=True)
+node = cluster.add_instance('node', main_configs=['configs/enable_test_keeper.xml', 'configs/logs_conf.xml'], with_zookeeper=True)
 from kazoo.client import KazooClient
 
 _genuine_zk_instance = None
@@ -400,3 +401,125 @@ def test_end_of_session(started_cluster):
                     zk.close()
         except:
             pass
+
+def test_end_of_watches_session(started_cluster):
+    fake_zk1 = None
+    fake_zk2 = None
+    try:
+        fake_zk1 = KazooClient(hosts=cluster.get_instance_ip("node") + ":9181")
+        fake_zk1.start()
+
+        fake_zk2 = KazooClient(hosts=cluster.get_instance_ip("node") + ":9181")
+        fake_zk2.start()
+
+        fake_zk1.create("/test_end_of_watches_session")
+
+        dummy_set = 0
+        def dummy_callback(event):
+            nonlocal dummy_set
+            dummy_set += 1
+            print(event)
+
+        for child_node in range(100):
+            fake_zk1.create("/test_end_of_watches_session/" + str(child_node))
+            fake_zk1.get_children("/test_end_of_watches_session/" + str(child_node), watch=dummy_callback)
+
+        fake_zk2.get_children("/test_end_of_watches_session/" + str(0), watch=dummy_callback)
+        fake_zk2.get_children("/test_end_of_watches_session/" + str(1), watch=dummy_callback)
+
+        fake_zk1.stop()
+        fake_zk1.close()
+
+        for child_node in range(100):
+            fake_zk2.create("/test_end_of_watches_session/" + str(child_node) + "/" + str(child_node), b"somebytes")
+
+        assert dummy_set == 2
+    finally:
+        try:
+            for zk in [fake_zk1, fake_zk2]:
+                if zk:
+                    zk.stop()
+                    zk.close()
+        except:
+            pass
+
+def test_concurrent_watches(started_cluster):
+    fake_zk = get_fake_zk()
+    fake_zk.restart()
+    global_path = "/test_concurrent_watches_0"
+    fake_zk.create(global_path)
+
+    dumb_watch_triggered_counter = 0
+    all_paths_triggered = []
+
+    existing_path = []
+    all_paths_created = []
+    watches_created = 0
+    def create_path_and_watch(i):
+        nonlocal watches_created
+        nonlocal all_paths_created
+        fake_zk.ensure_path(global_path + "/" + str(i))
+        # new function each time
+        def dumb_watch(event):
+            nonlocal dumb_watch_triggered_counter
+            dumb_watch_triggered_counter += 1
+            nonlocal all_paths_triggered
+            all_paths_triggered.append(event.path)
+
+        fake_zk.get(global_path + "/" + str(i), watch=dumb_watch)
+        all_paths_created.append(global_path + "/" + str(i))
+        watches_created += 1
+        existing_path.append(i)
+
+    trigger_called = 0
+    def trigger_watch(i):
+        nonlocal trigger_called
+        trigger_called += 1
+        fake_zk.set(global_path + "/" + str(i), b"somevalue")
+        try:
+            existing_path.remove(i)
+        except:
+            pass
+
+    def call(total):
+        for i in range(total):
+            create_path_and_watch(random.randint(0, 1000))
+            time.sleep(random.random() % 0.5)
+            try:
+                rand_num = random.choice(existing_path)
+                trigger_watch(rand_num)
+            except:
+                pass
+        while existing_path:
+            try:
+                rand_num = random.choice(existing_path)
+                trigger_watch(rand_num)
+            except:
+                pass
+
+    p = Pool(10)
+    arguments = [100] * 10
+    watches_must_be_created = sum(arguments)
+    watches_trigger_must_be_called = sum(arguments)
+    watches_must_be_triggered = sum(arguments)
+    p.map(call, arguments)
+    p.close()
+
+    # waiting for late watches
+    for i in range(50):
+        if dumb_watch_triggered_counter == watches_must_be_triggered:
+            break
+
+        time.sleep(0.1)
+
+    assert watches_created == watches_must_be_created
+    assert trigger_called >= watches_trigger_must_be_called
+    assert len(existing_path) == 0
+    if dumb_watch_triggered_counter != watches_must_be_triggered:
+        print("All created paths", all_paths_created)
+        print("All triggerred paths", all_paths_triggered)
+        print("All paths len", len(all_paths_created))
+        print("All triggered len", len(all_paths_triggered))
+        print("Diff", list(set(all_paths_created) - set(all_paths_triggered)))
+
+    assert dumb_watch_triggered_counter == watches_must_be_triggered
diff --git a/tests/performance/encodeXMLComponent.xml b/tests/performance/encodeXMLComponent.xml
new file mode 100644
index 00000000000..45241941ac3
--- /dev/null
+++ b/tests/performance/encodeXMLComponent.xml
@@ -0,0 +1,7 @@
+<test>
+    <preconditions>
+        <table_exists>test.hits</table_exists>
+    </preconditions>
+
+    <query>SELECT count() FROM test.hits WHERE NOT ignore(encodeXMLComponent(URL))</query>
+</test>
diff --git a/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql b/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql
index 120d599bd35..998a4517163 100644
--- a/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql
+++ b/tests/queries/0_stateless/00083_create_merge_tree_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS merge_tree;
 DROP TABLE IF EXISTS collapsing_merge_tree;
 DROP TABLE IF EXISTS versioned_collapsing_merge_tree;
diff --git a/tests/queries/0_stateless/00327_summing_composite_nested.sql b/tests/queries/0_stateless/00327_summing_composite_nested.sql
index b61bc71b892..9be21e87abf 100644
--- a/tests/queries/0_stateless/00327_summing_composite_nested.sql
+++ b/tests/queries/0_stateless/00327_summing_composite_nested.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS summing_composite_key;
 CREATE TABLE summing_composite_key (d Date, k UInt64, FirstMap Nested(k1 UInt32, k2ID Int8, s Float64), SecondMap Nested(k1ID UInt64, k2Key String, k3Type Int32, s Int64)) ENGINE = SummingMergeTree(d, k, 1);
 
diff --git a/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh b/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh
index 3cea3a65cdf..cc19a608c20 100755
--- a/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh
+++ b/tests/queries/0_stateless/00443_optimize_final_vertical_merge.sh
@@ -42,13 +42,13 @@ $CLICKHOUSE_CLIENT -q "INSERT INTO $name (date, Sign, ki) SELECT
 toDate(0) AS date,
 toInt8(1) AS Sign,
 toUInt64(0) AS ki
-FROM system.numbers LIMIT 9000"
+FROM system.numbers LIMIT 9000" --server_logs_file=/dev/null
 
 $CLICKHOUSE_CLIENT -q "INSERT INTO $name (date, Sign, ki) SELECT
 toDate(0) AS date,
 toInt8(1) AS Sign,
 number AS ki
-FROM system.numbers LIMIT 9000, 9000"
+FROM system.numbers LIMIT 9000, 9000" --server_logs_file=/dev/null
 
 $CLICKHOUSE_CLIENT -q "INSERT INTO $name SELECT
 toDate(0) AS date,
@@ -67,7 +67,7 @@ number AS di09,
 number AS di10,
 [number, number+1] AS \`n.i\`,
 [hex(number), hex(number+1)] AS \`n.s\`
-FROM system.numbers LIMIT $res_rows"
+FROM system.numbers LIMIT $res_rows" --server_logs_file=/dev/null
 
 while [[ $(get_num_parts) -ne 1 ]] ; do $CLICKHOUSE_CLIENT -q "OPTIMIZE TABLE $name PARTITION 197001" --server_logs_file=/dev/null; done
 
diff --git a/tests/queries/0_stateless/00453_top_k.reference b/tests/queries/0_stateless/00453_top_k.reference
index 1a768b03965..14beb3273fa 100644
--- a/tests/queries/0_stateless/00453_top_k.reference
+++ b/tests/queries/0_stateless/00453_top_k.reference
@@ -1 +1,8 @@
 [0,1,2,3,4,5,6,7,8,9]
+0	[[],[[],[NULL],[NULL,'1'],[NULL,'1','2'],[NULL,'1','2','3'],[NULL,'1','2','3','4'],[NULL,'1','2','3','4','5']]]
+1	[[[]],[[],[NULL],[NULL,'1'],[NULL,'1','2'],[NULL,'1','2','3'],[NULL,'1','2','3','4'],[NULL,'1','2','3','4','5'],[NULL,'1','2','3','4','5','6']]]
+2	[[[],[NULL]],[[],[NULL],[NULL,'1'],[NULL,'1','2'],[NULL,'1','2','3'],[NULL,'1','2','3','4'],[NULL,'1','2','3','4','5'],[NULL,'1','2','3','4','5','6'],[NULL,'1','2','3','4','5','6','7']]]
+3	[[[],[NULL],[NULL,'1']]]
+4	[[[],[NULL],[NULL,'1'],[NULL,'1','2']]]
+5	[[[],[NULL],[NULL,'1'],[NULL,'1','2'],[NULL,'1','2','3']]]
+6	[[[],[NULL],[NULL,'1'],[NULL,'1','2'],[NULL,'1','2','3'],[NULL,'1','2','3','4']]]
diff --git a/tests/queries/0_stateless/00453_top_k.sql b/tests/queries/0_stateless/00453_top_k.sql
index 1f79a8c5393..fb3b88e29e4 100644
--- a/tests/queries/0_stateless/00453_top_k.sql
+++ b/tests/queries/0_stateless/00453_top_k.sql
@@ -1 +1,15 @@
-SELECT topK(10)(n) FROM (SELECT if(number % 100 < 10, number % 10, number) AS n FROM system.numbers LIMIT 100000);
\ No newline at end of file
+SELECT topK(10)(n) FROM (SELECT if(number % 100 < 10, number % 10, number) AS n FROM system.numbers LIMIT 100000);
+
+SELECT
+    k,
+    topK(v)
+FROM
+(
+    SELECT
+        number % 7 AS k,
+        arrayMap(x -> arrayMap(x -> if(x = 0, NULL, toString(x)), range(x)), range(intDiv(number, 1))) AS v
+    FROM system.numbers
+    LIMIT 10
+)
+GROUP BY k
+ORDER BY k ASC
diff --git a/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql b/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql
index 4ed053f5953..ef8655a1861 100644
--- a/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql
+++ b/tests/queries/0_stateless/00509_extended_storage_definition_syntax_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 SELECT '*** Replicated with sampling ***';
 
 DROP TABLE IF EXISTS replicated_with_sampling;
diff --git a/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql b/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql
index b7824e7efdc..634b9781c7a 100644
--- a/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql
+++ b/tests/queries/0_stateless/00564_versioned_collapsing_merge_tree.sql
@@ -1,3 +1,5 @@
+set optimize_on_insert = 0;
+
 drop table if exists mult_tab;
 create table mult_tab (date Date, value String, version UInt64, sign Int8) engine = VersionedCollapsingMergeTree(date, (date), 8192, sign, version);
 insert into mult_tab select '2018-01-31', 'str_' || toString(number), 0, if(number % 2, 1, -1) from system.numbers limit 10;
diff --git a/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql b/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql
index 1bfbdaf75c5..8c51a6f34da 100644
--- a/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql
+++ b/tests/queries/0_stateless/00577_replacing_merge_tree_vertical_merge.sql
@@ -1,3 +1,5 @@
+set optimize_on_insert = 0;
+
 drop table if exists tab_00577;
 create table tab_00577 (date Date, version UInt64, val UInt64) engine = ReplacingMergeTree(version) partition by date order by date settings enable_vertical_merge_algorithm = 1, vertical_merge_algorithm_min_rows_to_activate = 1, vertical_merge_algorithm_min_columns_to_activate = 0;
 insert into tab_00577 values ('2018-01-01', 2, 2), ('2018-01-01', 1, 1);
diff --git a/tests/queries/0_stateless/00616_final_single_part.sql b/tests/queries/0_stateless/00616_final_single_part.sql
index df65123e29b..6618d0b1252 100644
--- a/tests/queries/0_stateless/00616_final_single_part.sql
+++ b/tests/queries/0_stateless/00616_final_single_part.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS test_00616;
 DROP TABLE IF EXISTS replacing_00616;
 
diff --git a/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql b/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql
index f488502b13b..715129cc699 100644
--- a/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql
+++ b/tests/queries/0_stateless/00620_optimize_on_nonleader_replica_zookeeper.sql
@@ -9,8 +9,8 @@ CREATE TABLE rename2 (p Int64, i Int64, v UInt64) ENGINE = ReplicatedReplacingMe
 INSERT INTO rename1 VALUES (0, 1, 0);
 INSERT INTO rename1 VALUES (0, 1, 1);
 
-OPTIMIZE TABLE rename1;
-OPTIMIZE TABLE rename2;
+OPTIMIZE TABLE rename1 FINAL;
+OPTIMIZE TABLE rename2 FINAL;
 SELECT * FROM rename1;
 
 RENAME TABLE rename2 TO rename3;
diff --git a/tests/queries/0_stateless/00660_optimize_final_without_partition.sql b/tests/queries/0_stateless/00660_optimize_final_without_partition.sql
index 8c1f2ebd361..6545ad6e85b 100644
--- a/tests/queries/0_stateless/00660_optimize_final_without_partition.sql
+++ b/tests/queries/0_stateless/00660_optimize_final_without_partition.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS partitioned_by_tuple;
 
 CREATE TABLE partitioned_by_tuple (d Date, x UInt8, w String, y UInt8) ENGINE SummingMergeTree (y) PARTITION BY (d, x) ORDER BY (d, x, w);
diff --git a/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql b/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql
index 77747bc0383..033202e04aa 100644
--- a/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql
+++ b/tests/queries/0_stateless/00661_optimize_final_replicated_without_partition_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS partitioned_by_tuple_replica1_00661;
 DROP TABLE IF EXISTS partitioned_by_tuple_replica2_00661;
 CREATE TABLE partitioned_by_tuple_replica1_00661(d Date, x UInt8, w String, y UInt8) ENGINE = ReplicatedSummingMergeTree('/clickhouse/tables/test/partitioned_by_tuple_00661', '1') PARTITION BY (d, x) ORDER BY (d, x, w);
diff --git a/tests/queries/0_stateless/00682_empty_parts_merge.sh b/tests/queries/0_stateless/00682_empty_parts_merge.sh
index 03915e3f5e4..1a336580b3a 100755
--- a/tests/queries/0_stateless/00682_empty_parts_merge.sh
+++ b/tests/queries/0_stateless/00682_empty_parts_merge.sh
@@ -4,7 +4,7 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 . "$CURDIR"/../shell_config.sh
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS ordinary_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE ordinary_00682(k UInt32) ENGINE MergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE ordinary_00682(k UInt32) ENGINE MergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO ordinary_00682(k) VALUES (1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO ordinary_00682(k) VALUES (1)"
@@ -21,7 +21,7 @@ ${CLICKHOUSE_CLIENT} --query="SELECT '*** Vertical merge ***'"
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS vertical_00682"
 ${CLICKHOUSE_CLIENT} --query="CREATE TABLE vertical_00682(k UInt32, v UInt32) ENGINE MergeTree ORDER BY k \
-    SETTINGS enable_vertical_merge_algorithm=1, \
+    SETTINGS enable_vertical_merge_algorithm=1, remove_empty_parts=0, \
              vertical_merge_algorithm_min_rows_to_activate=0, \
              vertical_merge_algorithm_min_columns_to_activate=0"
 
@@ -37,7 +37,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE vertical_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS summing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE summing_00682(k UInt32, v UInt32) ENGINE SummingMergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE summing_00682(k UInt32, v UInt32) ENGINE SummingMergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO summing_00682(k, v) VALUES (1, 1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO summing_00682(k, v) VALUES (1, 2)"
@@ -51,7 +51,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE summing_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS aggregating_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE aggregating_00682(k UInt32, v AggregateFunction(count)) ENGINE AggregatingMergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE aggregating_00682(k UInt32, v AggregateFunction(count)) ENGINE AggregatingMergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO aggregating_00682(k) VALUES (1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO aggregating_00682(k) VALUES (1)"
@@ -65,7 +65,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE aggregating_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS replacing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE replacing_00682(k UInt32, v String) ENGINE ReplacingMergeTree ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE replacing_00682(k UInt32, v String) ENGINE ReplacingMergeTree ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO replacing_00682(k, v) VALUES (1, 'a')"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO replacing_00682(k, v) VALUES (1, 'b')"
@@ -79,7 +79,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE replacing_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS collapsing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE collapsing_00682(k UInt32, v String, s Int8) ENGINE CollapsingMergeTree(s) ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE collapsing_00682(k UInt32, v String, s Int8) ENGINE CollapsingMergeTree(s) ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO collapsing_00682(k, v, s) VALUES (1, 'a', 1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO collapsing_00682(k, v, s) VALUES (2, 'b', 1)"
@@ -93,7 +93,7 @@ ${CLICKHOUSE_CLIENT} --query="DROP TABLE collapsing_00682"
 
 
 ${CLICKHOUSE_CLIENT} --query="DROP TABLE IF EXISTS versioned_collapsing_00682"
-${CLICKHOUSE_CLIENT} --query="CREATE TABLE versioned_collapsing_00682(k UInt32, val String, ver UInt32, s Int8) ENGINE VersionedCollapsingMergeTree(s, ver) ORDER BY k"
+${CLICKHOUSE_CLIENT} --query="CREATE TABLE versioned_collapsing_00682(k UInt32, val String, ver UInt32, s Int8) ENGINE VersionedCollapsingMergeTree(s, ver) ORDER BY k SETTINGS remove_empty_parts=0"
 
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO versioned_collapsing_00682(k, val, ver, s) VALUES (1, 'a', 0, 1)"
 ${CLICKHOUSE_CLIENT} --query="INSERT INTO versioned_collapsing_00682(k, val, ver, s) VALUES (2, 'b', 0, 1)"
diff --git a/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference b/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference
index f8fe62bc722..724d74d19c6 100644
--- a/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference
+++ b/tests/queries/0_stateless/00693_max_block_size_system_tables_columns.reference
@@ -2,7 +2,7 @@
 1
 1
 1
-	t_00693	Memory	1	[]		1970-01-01 00:00:00	[]	[]		Memory						\N	\N
+	t_00693	Memory	1	[]		1970-01-01 00:00:00	[]	[]	CREATE TEMPORARY TABLE t_00693 (`x` UInt8) ENGINE = Memory	Memory						\N	\N
 1
 1
 1
diff --git a/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql
index 3483e575fbb..a1fc25fbf0b 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_have_data_before_quorum_zookeeper.sql
@@ -17,7 +17,7 @@ SET select_sequential_consistency=1;
 SELECT x FROM quorum1 ORDER BY x;
 SELECT x FROM quorum2 ORDER BY x;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 INSERT INTO quorum1 VALUES (4, '1990-11-15');
 INSERT INTO quorum1 VALUES (5, '1990-11-15');
diff --git a/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql
index 42d1c9060fe..914d98ff3f5 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_lost_part_and_alive_part_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost_alive', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost_alive', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 INSERT INTO quorum1 VALUES (1, '2018-11-15');
@@ -30,7 +30,7 @@ SET select_sequential_consistency=1;
 
 SYSTEM START FETCHES quorum1;
 SYSTEM SYNC REPLICA quorum1;
- 
+
 SELECT x FROM quorum1 ORDER BY x;
 SELECT x FROM quorum2 ORDER BY x;
 
diff --git a/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql
index 9de73ac98e1..18cad0dda41 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_lost_part_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum_lost', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 SET insert_quorum_timeout=0;
@@ -27,7 +27,7 @@ SET insert_quorum_timeout=100;
 
 SYSTEM START FETCHES quorum1;
 SYSTEM SYNC REPLICA quorum1;
- 
+
 SELECT x FROM quorum1 ORDER BY x;
 SELECT x FROM quorum2 ORDER BY x;
 
diff --git a/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql
index 74f05d26b88..11c8324ac25 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_select_with_old_data_and_without_quorum_zookeeper.sql
@@ -13,7 +13,7 @@ INSERT INTO quorum1 VALUES (3, '2020-12-16');
 SYSTEM SYNC REPLICA quorum2;
 
 SET select_sequential_consistency=1;
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 SET insert_quorum_timeout=0;
 
diff --git a/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql
index fe64b9bc150..a4e262ce3c4 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_simple_test_1_parts_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum1', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum1', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 INSERT INTO quorum1 VALUES (1, '2018-11-15');
diff --git a/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql b/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql
index 07738ed0abd..40c08fd926f 100644
--- a/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql
+++ b/tests/queries/0_stateless/00732_quorum_insert_simple_test_2_parts_zookeeper.sql
@@ -6,7 +6,7 @@ DROP TABLE IF EXISTS quorum2;
 CREATE TABLE quorum1(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum2', '1') ORDER BY x PARTITION BY y;
 CREATE TABLE quorum2(x UInt32, y Date) ENGINE ReplicatedMergeTree('/clickhouse/tables/test_00732/quorum2', '2') ORDER BY x PARTITION BY y;
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 SET select_sequential_consistency=1;
 
 INSERT INTO quorum1 VALUES (1, '2018-11-15');
diff --git a/tests/queries/0_stateless/00754_alter_modify_order_by.sql b/tests/queries/0_stateless/00754_alter_modify_order_by.sql
index f8c584ed052..a09d824c928 100644
--- a/tests/queries/0_stateless/00754_alter_modify_order_by.sql
+++ b/tests/queries/0_stateless/00754_alter_modify_order_by.sql
@@ -1,4 +1,5 @@
 SET send_logs_level = 'fatal';
+SET optimize_on_insert = 0;
 
 DROP TABLE IF EXISTS old_style;
 CREATE TABLE old_style(d Date, x UInt32) ENGINE MergeTree(d, x, 8192);
diff --git a/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql b/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql
index 4b150fb3826..809adfaa498 100644
--- a/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql
+++ b/tests/queries/0_stateless/00754_alter_modify_order_by_replicated_zookeeper.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 SET send_logs_level = 'fatal';
 
 DROP TABLE IF EXISTS old_style;
diff --git a/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh b/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh
index d4344e6e8bd..0c13ad2d573 100755
--- a/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh
+++ b/tests/queries/0_stateless/00993_system_parts_race_condition_drop_zookeeper.sh
@@ -98,6 +98,6 @@ timeout $TIMEOUT bash -c thread6 2>&1 | grep "was not completely removed from Zo
 wait
 
 for i in {0..9}; do
-    $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS alter_table_$i" &
+    $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS alter_table_$i" 2>&1 | grep "was not completely removed from ZooKeeper" &
 done
 wait
diff --git a/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql b/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql
index 0b5845d3b04..90b1660e546 100644
--- a/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql
+++ b/tests/queries/0_stateless/01030_incorrect_count_summing_merge_tree.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 select '-- SummingMergeTree with Nullable column without duplicates.';  
 
 drop table if exists tst;
diff --git a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference
index 1fca8dab675..e362d692f42 100644
--- a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference
+++ b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.reference
@@ -1,5 +1,5 @@
 122
 
-Table dictdb.dict_invalidate doesn\'t exist.
+Table dictdb.dict_invalidate doesn\'t exist
 
 133
diff --git a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh
index 3dd3c0ee363..18f04032933 100755
--- a/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh
+++ b/tests/queries/0_stateless/01040_dictionary_invalidate_query_switchover_long.sh
@@ -51,7 +51,7 @@ function check_exception_detected()
 export -f check_exception_detected;
 timeout 30 bash -c check_exception_detected 2> /dev/null
 
-$CLICKHOUSE_CLIENT --query "SELECT last_exception FROM system.dictionaries WHERE database = 'dictdb' AND name = 'invalidate'" 2>&1 | grep -Eo "Table dictdb.dict_invalidate .* exist."
+$CLICKHOUSE_CLIENT --query "SELECT last_exception FROM system.dictionaries WHERE database = 'dictdb' AND name = 'invalidate'" 2>&1 | grep -Eo "Table dictdb.dict_invalidate .* exist"
 
 $CLICKHOUSE_CLIENT --query "
 CREATE TABLE dictdb.dict_invalidate
diff --git a/tests/queries/0_stateless/01056_create_table_as.reference b/tests/queries/0_stateless/01056_create_table_as.reference
index e69de29bb2d..c6e14cb675e 100644
--- a/tests/queries/0_stateless/01056_create_table_as.reference
+++ b/tests/queries/0_stateless/01056_create_table_as.reference
@@ -0,0 +1 @@
+1	String
diff --git a/tests/queries/0_stateless/01056_create_table_as.sql b/tests/queries/0_stateless/01056_create_table_as.sql
index bf2a143fa5a..c27f30b61d5 100644
--- a/tests/queries/0_stateless/01056_create_table_as.sql
+++ b/tests/queries/0_stateless/01056_create_table_as.sql
@@ -49,3 +49,7 @@ DROP DICTIONARY dict;
 DROP TABLE test_01056_dict_data.dict_data;
 
 DROP DATABASE test_01056_dict_data;
+
+CREATE TABLE t1 (x String) ENGINE = Memory AS SELECT 1;
+SELECT x, toTypeName(x) FROM t1;
+DROP TABLE t1;
diff --git a/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql b/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql
index 146107d6f3d..768a20c8ca4 100644
--- a/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql
+++ b/tests/queries/0_stateless/01081_PartialSortingTransform_full_column.sql
@@ -9,8 +9,6 @@ select 1 from remote('127.{1,2}', currentDatabase(), test_01081) lhs join system
 --     Code: 171. DB::Exception: Received from localhost:9000. DB::Exception: Received from 127.2:9000. DB::Exception: Block structure mismatch in  function connect between PartialSortingTransform and LazyOutputFormat stream: different columns:
 --     _dummy Int Int32(size = 0), 1 UInt8 UInt8(size = 0)
 --     _dummy Int Int32(size = 0), 1 UInt8 Const(size = 0, UInt8(size = 1)).
---
--- With experimental_use_processors=1 (default at the time of writing).
 
 insert into test_01081 select * from system.numbers limit 10;
 select 1 from remote('127.{1,2}', currentDatabase(), test_01081) lhs join system.one as rhs on rhs.dummy = 1 order by 1;
diff --git a/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql b/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql
index 8c4ba1fd318..5597bf523e4 100644
--- a/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql
+++ b/tests/queries/0_stateless/01090_zookeeper_mutations_and_insert_quorum.sql
@@ -4,7 +4,7 @@ DROP TABLE IF EXISTS mutations_and_quorum2;
 CREATE TABLE mutations_and_quorum1 (`server_date` Date, `something` String) ENGINE = ReplicatedMergeTree('/clickhouse/tables/test_01090/mutations_and_quorum', '1') PARTITION BY toYYYYMM(server_date) ORDER BY (server_date, something);
 CREATE TABLE mutations_and_quorum2 (`server_date` Date, `something` String) ENGINE = ReplicatedMergeTree('/clickhouse/tables/test_01090/mutations_and_quorum', '2') PARTITION BY toYYYYMM(server_date) ORDER BY (server_date, something);
 
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 INSERT INTO mutations_and_quorum1 VALUES ('2019-01-01', 'test1'), ('2019-02-01', 'test2'), ('2019-03-01', 'test3'), ('2019-04-01', 'test4'), ('2019-05-01', 'test1'), ('2019-06-01', 'test2'), ('2019-07-01', 'test3'), ('2019-08-01', 'test4'), ('2019-09-01', 'test1'), ('2019-10-01', 'test2'), ('2019-11-01', 'test3'), ('2019-12-01', 'test4');
 
diff --git a/tests/queries/0_stateless/01098_temporary_and_external_tables.reference b/tests/queries/0_stateless/01098_temporary_and_external_tables.reference
index c3e51f64b2d..1ddb10581a8 100644
--- a/tests/queries/0_stateless/01098_temporary_and_external_tables.reference
+++ b/tests/queries/0_stateless/01098_temporary_and_external_tables.reference
@@ -1,2 +1,9 @@
+CREATE TEMPORARY TABLE tmp_table\n(\n    `n` UInt64\n)\nENGINE = Memory AS\nSELECT number AS n\nFROM numbers(42)
+CREATE TEMPORARY TABLE tmp_table\n(\n    `n` UInt64\n)\nENGINE = Memory AS\nSELECT number AS n\nFROM numbers(42)
+CREATE TEMPORARY TABLE tmp_table\n(\n    `n` UInt64\n)\nENGINE = Memory AS\nSELECT number AS n\nFROM numbers(42)
+42
+OK
 OK
 3	9
+0
+1
diff --git a/tests/queries/0_stateless/01098_temporary_and_external_tables.sh b/tests/queries/0_stateless/01098_temporary_and_external_tables.sh
index 67f57567298..54834107728 100755
--- a/tests/queries/0_stateless/01098_temporary_and_external_tables.sh
+++ b/tests/queries/0_stateless/01098_temporary_and_external_tables.sh
@@ -6,20 +6,23 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 url_without_session="https://${CLICKHOUSE_HOST}:${CLICKHOUSE_PORT_HTTPS}/?"
 url="${url_without_session}session_id=test_01098"
 
-${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "CREATE TEMPORARY TABLE tmp_table AS SELECT number AS n FROM numbers(42)" > /dev/null;
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "DROP TEMPORARY TABLE IF EXISTS tmp_table"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "CREATE TEMPORARY TABLE tmp_table AS SELECT number AS n FROM numbers(42)"
 
-name_expr="'\`' || database || '\`.\`' || name || '\`'"
-full_tmp_name=$(echo "SELECT $name_expr FROM system.tables WHERE database='_temporary_and_external_tables' AND create_table_query LIKE '%tmp_table%'" | ${CLICKHOUSE_CURL} -m 30 -sSgk "$url" -d @-)
+id=$(echo "SELECT uuid FROM system.tables WHERE name='tmp_table' AND is_temporary" | ${CLICKHOUSE_CURL} -m 31 -sSgk "$url" -d @-)
+internal_table_name="_temporary_and_external_tables.\`_tmp_$id\`"
 
-echo "SELECT * FROM $full_tmp_name" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url" -d @- | grep -F "Code: 291" > /dev/null && echo "OK"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "SHOW CREATE TEMPORARY TABLE tmp_table"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "SHOW CREATE TABLE $internal_table_name"
+${CLICKHOUSE_CURL} -m 30 -sSk "$url_without_session" --data "SHOW CREATE TABLE $internal_table_name"
 
-echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+sum((number+GLOBAL+IN+(SELECT+number+AS+n+FROM+remote('127.0.0.2',+numbers(5))+WHERE+n+GLOBAL+IN+(SELECT+*+FROM+tmp_table)+AND+n+GLOBAL+NOT+IN+(SELECT+*+FROM+file)+))+AS+res),+sum(number*res)+FROM+remote('127.0.0.2',+numbers(10))";
+echo "SELECT COUNT() FROM tmp_table" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url" -d @-
+echo "SELECT COUNT() FROM $internal_table_name" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url" -d @- | grep -F "Code: 291" > /dev/null && echo "OK"
+echo "SELECT COUNT() FROM $internal_table_name" | ${CLICKHOUSE_CURL} -m 60 -sSgk "$url_without_session" -d @- | grep -F "Code: 291" > /dev/null && echo "OK"
+
+echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+sum((number+GLOBAL+IN+(SELECT+number+AS+n+FROM+remote('127.0.0.2',+numbers(5))+WHERE+n+GLOBAL+IN+(SELECT+*+FROM+tmp_table)+AND+n+GLOBAL+NOT+IN+(SELECT+*+FROM+file)+))+AS+res),+sum(number*res)+FROM+remote('127.0.0.2',+numbers(10))"
+
+echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+_1%2BsleepEachRow(3)+FROM+file" &
 
-echo -ne '0\n1\n' | ${CLICKHOUSE_CURL} -m 30 -sSkF 'file=@-' "$url&file_format=CSV&file_types=UInt64&query=SELECT+sleepEachRow(3)+FROM+file" > /dev/null &
-sleep 1
-full_tmp_names=$(echo "SELECT $name_expr FROM system.tables WHERE database='_temporary_and_external_tables' FORMAT TSV" | ${CLICKHOUSE_CURL} -m 30 -sSgk "$url_without_session" -d @-)
-for name in $full_tmp_names
-do
-  ${CLICKHOUSE_CURL} -m 30 -sSk "${url_without_session}query=SHOW+CREATE+TABLE+$name" 1>/dev/null 2>/dev/null
-done;
 wait
+${CLICKHOUSE_CURL} -m 30 -sSk "$url" --data "DROP TEMPORARY TABLE tmp_table"
diff --git a/tests/queries/0_stateless/01187_set_profile_as_setting.reference b/tests/queries/0_stateless/01187_set_profile_as_setting.reference
new file mode 100644
index 00000000000..ae1cc97819b
--- /dev/null
+++ b/tests/queries/0_stateless/01187_set_profile_as_setting.reference
@@ -0,0 +1,8 @@
+0	0
+0	0
+OK
+1	1
+2	1
+2	1
+OK
+1	1
diff --git a/tests/queries/0_stateless/01187_set_profile_as_setting.sh b/tests/queries/0_stateless/01187_set_profile_as_setting.sh
new file mode 100755
index 00000000000..8247ab8870a
--- /dev/null
+++ b/tests/queries/0_stateless/01187_set_profile_as_setting.sh
@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+$CLICKHOUSE_CLIENT -n -m -q "select value, changed from system.settings where name='readonly';"
+$CLICKHOUSE_CLIENT -n -m -q "set profile='default'; select value, changed from system.settings where name='readonly';"
+$CLICKHOUSE_CLIENT -n -m -q "set profile='readonly'; select value, changed from system.settings where name='readonly';" 2>&1| grep -Fa "Cannot modify 'send_logs_level' setting in readonly mode" > /dev/null && echo "OK"
+CLICKHOUSE_CLIENT=$(echo ${CLICKHOUSE_CLIENT} | sed 's/'"--send_logs_level=${CLICKHOUSE_CLIENT_SERVER_LOGS_LEVEL}"'/--send_logs_level=fatal/g')
+$CLICKHOUSE_CLIENT -n -m -q "set profile='readonly'; select value, changed from system.settings where name='readonly';"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=select+value,changed+from+system.settings+where+name='readonly'"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&profile=default&query=select+value,changed+from+system.settings+where+name='readonly'"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&profile=readonly&query=select+value,changed+from+system.settings+where+name='readonly'" 2>&1| grep -Fa "Cannot modify 'readonly' setting in readonly mode" > /dev/null && echo "OK"
+echo "select value, changed from system.settings where name='readonly';" | ${CLICKHOUSE_CURL} -sSg "${CLICKHOUSE_URL}&profile=readonly" -d @-
diff --git a/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql b/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql
index 575b4dfa4a5..dba734454c0 100644
--- a/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql
+++ b/tests/queries/0_stateless/01285_data_skip_index_over_aggregation.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS data_01285;
 
 SET max_threads=1;
diff --git a/tests/queries/0_stateless/01290_max_execution_speed_distributed.sql b/tests/queries/0_stateless/01290_max_execution_speed_distributed.sql
index 8282390ca90..b0f545838e6 100644
--- a/tests/queries/0_stateless/01290_max_execution_speed_distributed.sql
+++ b/tests/queries/0_stateless/01290_max_execution_speed_distributed.sql
@@ -1,4 +1,8 @@
-SET max_execution_speed = 1000000, timeout_before_checking_execution_speed = 0.001, max_block_size = 100;
+SET max_execution_speed = 1000000;
+SET timeout_before_checking_execution_speed = 0.001;
+SET max_block_size = 100;
+
+SET log_queries=1;
 
 CREATE TEMPORARY TABLE times (t DateTime);
 
@@ -10,4 +14,10 @@ SELECT max(t) - min(t) >= 1 FROM times;
 
 -- Check that the query was also throttled on "remote" servers.
 SYSTEM FLUSH LOGS;
-SELECT DISTINCT query_duration_ms >= 500 FROM system.query_log WHERE event_date >= yesterday() AND query LIKE '%special query for 01290_max_execution_speed_distributed%' AND type = 2;
+SELECT DISTINCT query_duration_ms >= 500
+FROM system.query_log
+WHERE
+    event_date >= yesterday() AND
+    query LIKE '%special query for 01290_max_execution_speed_distributed%' AND
+    query NOT LIKE '%system.query_log%' AND
+    type = 2;
diff --git a/tests/queries/0_stateless/01293_show_settings.reference b/tests/queries/0_stateless/01293_show_settings.reference
new file mode 100644
index 00000000000..feebc9e473d
--- /dev/null
+++ b/tests/queries/0_stateless/01293_show_settings.reference
@@ -0,0 +1,5 @@
+send_timeout	Seconds	300
+connect_timeout	Seconds	10
+connect_timeout_with_failover_ms	Milliseconds	50
+connect_timeout_with_failover_secure_ms	Milliseconds	100
+max_memory_usage	UInt64	10000000000
diff --git a/tests/queries/0_stateless/01293_show_settings.sql b/tests/queries/0_stateless/01293_show_settings.sql
new file mode 100644
index 00000000000..08f00ed201c
--- /dev/null
+++ b/tests/queries/0_stateless/01293_show_settings.sql
@@ -0,0 +1,3 @@
+show settings like 'send_timeout';
+SHOW SETTINGS ILIKE '%CONNECT_timeout%';
+SHOW CHANGED SETTINGS ILIKE '%MEMORY%';
diff --git a/tests/queries/0_stateless/01322_cast_keep_nullable.reference b/tests/queries/0_stateless/01322_cast_keep_nullable.reference
index 8ad99a10170..cfe80a7c1b1 100644
--- a/tests/queries/0_stateless/01322_cast_keep_nullable.reference
+++ b/tests/queries/0_stateless/01322_cast_keep_nullable.reference
@@ -4,6 +4,7 @@
 1	Nullable(Int32)
 2	Nullable(Float32)
 2	Nullable(UInt8)
+00000000-0000-0002-0000-000000000000	Nullable(UUID)
 3	Nullable(Int32)
 \N	Nullable(Int32)
 42	Nullable(Int32)
diff --git a/tests/queries/0_stateless/01322_cast_keep_nullable.sql b/tests/queries/0_stateless/01322_cast_keep_nullable.sql
index 10918717469..ef5f5747a7d 100644
--- a/tests/queries/0_stateless/01322_cast_keep_nullable.sql
+++ b/tests/queries/0_stateless/01322_cast_keep_nullable.sql
@@ -10,7 +10,7 @@ SELECT CAST(toNullable(toInt8(1)) AS Int32) as x, toTypeName(x);
 
 SELECT CAST(toNullable(toFloat32(2)), 'Float32') as x, toTypeName(x);
 SELECT CAST(toNullable(toFloat32(2)), 'UInt8') as x, toTypeName(x);
-SELECT CAST(toNullable(toFloat32(2)), 'UUID') as x, toTypeName(x); -- { serverError 70 }
+SELECT CAST(toNullable(toFloat32(2)), 'UUID') as x, toTypeName(x);
 
 SELECT CAST(if(1 = 1, toNullable(toInt8(3)), NULL) AS Int32) as x, toTypeName(x);
 SELECT CAST(if(1 = 0, toNullable(toInt8(3)), NULL) AS Int32) as x, toTypeName(x);
diff --git a/tests/queries/0_stateless/01323_add_scalars_in_time.sql b/tests/queries/0_stateless/01323_add_scalars_in_time.sql
index 2d7cf270017..2ee5603f760 100644
--- a/tests/queries/0_stateless/01323_add_scalars_in_time.sql
+++ b/tests/queries/0_stateless/01323_add_scalars_in_time.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS tags;
 
 CREATE TABLE tags (
diff --git a/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql b/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql
index 790fbca6b73..c5a874efe09 100644
--- a/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql
+++ b/tests/queries/0_stateless/01373_summing_merge_tree_exclude_partition_key.sql
@@ -1,3 +1,5 @@
+SET optimize_on_insert = 0;
+
 DROP TABLE IF EXISTS tt_01373;
 
 CREATE TABLE tt_01373
diff --git a/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql b/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql
index fca14b81e27..72e588c5c02 100644
--- a/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql
+++ b/tests/queries/0_stateless/01451_replicated_detach_drop_and_quorum.sql
@@ -22,7 +22,7 @@ SELECT * FROM replica2;
 -- drop of empty partition works
 ALTER TABLE replica2 DROP PARTITION ID 'all';
 
-SET insert_quorum=2;
+SET insert_quorum = 2, insert_quorum_parallel = 0;
 
 INSERT INTO replica2 VALUES (1);
 
diff --git a/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh b/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh
index 48cf034a4c3..e033e0d5b72 100755
--- a/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh
+++ b/tests/queries/0_stateless/01459_manual_write_to_replicas_quorum.sh
@@ -17,7 +17,7 @@ done
 function thread {
     for x in {0..99}; do
         while true; do
-            $CLICKHOUSE_CLIENT --insert_quorum 5 --query "INSERT INTO r$1 SELECT $x" 2>&1 | grep -qF 'Quorum for previous write has not been satisfied yet' || break
+            $CLICKHOUSE_CLIENT --insert_quorum 5 --insert_quorum_parallel 0 --query "INSERT INTO r$1 SELECT $x" 2>&1 | grep -qF 'Quorum for previous write has not been satisfied yet' || break
         done
     done
 }
diff --git a/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql b/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql
index 6a6be32278d..0eb9f20ad91 100644
--- a/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql
+++ b/tests/queries/0_stateless/01513_count_without_select_sequence_consistency_zookeeper.sql
@@ -17,7 +17,7 @@ SYSTEM SYNC REPLICA quorum3;
 
 SET select_sequential_consistency=0;
 SET optimize_trivial_count_query=1;
-SET insert_quorum=2;
+SET insert_quorum=2, insert_quorum_parallel=0;
 
 SYSTEM STOP FETCHES quorum1;
 
diff --git a/tests/queries/0_stateless/01550_create_map_type.reference b/tests/queries/0_stateless/01550_create_map_type.reference
new file mode 100644
index 00000000000..877199e92d5
--- /dev/null
+++ b/tests/queries/0_stateless/01550_create_map_type.reference
@@ -0,0 +1,29 @@
+zhangsan
+lisi
+1111	2222
+1112	2224
+1113	2226
+female
+zhangsan
+gender
+1116
+1117
+1118
+1119
+[]
+[]
+[]
+[]
+[]
+[]
+[0,2,0]
+[1,2,3]
+[1,3,2]
+[2,4,4]
+[3,5,6]
+[4,6,8]
+[5,7,10]
+[100,20,90]
+{1:'1',2:'2',3:'foo'}	1
+200000	560000	0
+200000	560000	0
diff --git a/tests/queries/0_stateless/01550_create_map_type.sql b/tests/queries/0_stateless/01550_create_map_type.sql
new file mode 100644
index 00000000000..7a08886a5d8
--- /dev/null
+++ b/tests/queries/0_stateless/01550_create_map_type.sql
@@ -0,0 +1,65 @@
+set allow_experimental_map_type = 1;
+
+-- String type
+drop table if exists table_map;
+create table table_map (a Map(String, String)) engine = Memory;
+insert into table_map values ({'name':'zhangsan', 'gender':'male'}), ({'name':'lisi', 'gender':'female'});
+select a['name'] from table_map;
+drop table if exists table_map;
+
+
+drop table if exists table_map;
+create table table_map (a Map(String, UInt64)) engine = MergeTree() order by a;
+insert into table_map select map('key1', number, 'key2', number * 2) from numbers(1111, 3);
+select a['key1'], a['key2'] from table_map;
+drop table if exists table_map;
+
+-- MergeTree Engine
+drop table if exists table_map;
+create table table_map (a Map(String, String), b String) engine = MergeTree() order by a;
+insert into table_map values ({'name':'zhangsan', 'gender':'male'}, 'name'), ({'name':'lisi', 'gender':'female'}, 'gender');
+select a[b] from table_map;
+select b from table_map where a = map('name','lisi', 'gender', 'female');
+drop table if exists table_map;
+
+-- Int type
+drop table if exists table_map;
+create table table_map(a Map(UInt8, UInt64), b UInt8) Engine = MergeTree() order by b;
+insert into table_map select map(number, number+5), number from numbers(1111,4);
+select a[b] from table_map;
+drop table if exists table_map;
+
+
+-- Array Type
+drop table if exists table_map;
+create table table_map(a Map(String, Array(UInt8))) Engine = MergeTree() order by a;
+insert into table_map values(map('k1', [1,2,3], 'k2', [4,5,6])), (map('k0', [], 'k1', [100,20,90]));
+insert into table_map select map('k1', [number, number + 2, number * 2]) from numbers(6);
+insert into table_map select map('k2', [number, number + 2, number * 2]) from numbers(6);
+select a['k1'] as col1 from table_map order by col1;
+drop table if exists table_map;
+
+SELECT CAST(([1, 2, 3], ['1', '2', 'foo']), 'Map(UInt8, String)') AS map, map[1];
+
+CREATE TABLE table_map (n UInt32, m Map(String, Int))
+ENGINE = MergeTree ORDER BY n SETTINGS min_bytes_for_wide_part = 0;
+
+-- coversion from Tuple(Array(K), Array(V))
+INSERT INTO table_map SELECT number, (arrayMap(x -> toString(x), range(number % 10 + 2)), range(number % 10 + 2)) FROM numbers(100000);
+-- coversion from Array(Tuple(K, V))
+INSERT INTO table_map SELECT number, arrayMap(x -> (toString(x), x), range(number % 10 + 2)) FROM numbers(100000);
+SELECT sum(m['1']), sum(m['7']), sum(m['100']) FROM table_map;
+
+DROP TABLE IF EXISTS table_map;
+
+CREATE TABLE table_map (n UInt32, m Map(String, Int))
+ENGINE = MergeTree ORDER BY n;
+
+-- coversion from Tuple(Array(K), Array(V))
+INSERT INTO table_map SELECT number, (arrayMap(x -> toString(x), range(number % 10 + 2)), range(number % 10 + 2)) FROM numbers(100000);
+-- coversion from Array(Tuple(K, V))
+INSERT INTO table_map SELECT number, arrayMap(x -> (toString(x), x), range(number % 10 + 2)) FROM numbers(100000);
+SELECT sum(m['1']), sum(m['7']), sum(m['100']) FROM table_map;
+
+DROP TABLE IF EXISTS table_map;
+
diff --git a/tests/queries/0_stateless/01550_type_map_formats.reference b/tests/queries/0_stateless/01550_type_map_formats.reference
new file mode 100644
index 00000000000..ca081db75a2
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats.reference
@@ -0,0 +1,46 @@
+JSON
+{
+	"meta":
+	[
+		{
+			"name": "m",
+			"type": "Map(String,UInt32)"
+		},
+		{
+			"name": "m1",
+			"type": "Map(String,Date)"
+		},
+		{
+			"name": "m2",
+			"type": "Map(String,Array(UInt32))"
+		}
+	],
+
+	"data":
+	[
+		{
+			"m": {"k1":1,"k2":2,"k3":3},
+			"m1": {"k1":"2020-05-05"},
+			"m2": {"k1":[],"k2":[7,8]}
+		},
+		{
+			"m": {"k1":10,"k3":30},
+			"m1": {"k2":"2020-06-06"},
+			"m2": {}
+		}
+	],
+
+	"rows": 2
+}
+JSONEachRow
+{"m":{"k1":1,"k2":2,"k3":3},"m1":{"k1":"2020-05-05"},"m2":{"k1":[],"k2":[7,8]}}
+{"m":{"k1":10,"k3":30},"m1":{"k2":"2020-06-06"},"m2":{}}
+CSV
+"{'k1':1,'k2':2,'k3':3}","{'k1':'2020-05-05'}","{'k1':[],'k2':[7,8]}"
+"{'k1':10,'k3':30}","{'k2':'2020-06-06'}","{}"
+TSV
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':10,'k3':30}	{'k2':'2020-06-06'}	{}
+TSKV
+m={'k1':1,'k2':2,'k3':3}	m1={'k1':'2020-05-05'}	m2={'k1':[],'k2':[7,8]}
+m={'k1':10,'k3':30}	m1={'k2':'2020-06-06'}	m2={}
diff --git a/tests/queries/0_stateless/01550_type_map_formats.sql b/tests/queries/0_stateless/01550_type_map_formats.sql
new file mode 100644
index 00000000000..6a09fca9398
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats.sql
@@ -0,0 +1,19 @@
+SET allow_experimental_map_type = 1;
+SET output_format_write_statistics = 0;
+
+DROP TABLE IF EXISTS map_formats;
+CREATE TABLE map_formats (m Map(String, UInt32), m1 Map(String, Date), m2 Map(String, Array(UInt32))) ENGINE = Log;
+
+INSERT INTO map_formats VALUES(map('k1', 1, 'k2', 2, 'k3', 3), map('k1', toDate('2020-05-05')), map('k1', [], 'k2', [7, 8]));
+INSERT INTO map_formats VALUES(map('k1', 10, 'k3', 30), map('k2', toDate('2020-06-06')), map());
+
+SELECT 'JSON';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT JSON;
+SELECT 'JSONEachRow';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT JSONEachRow;
+SELECT 'CSV';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT CSV;
+SELECT 'TSV';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT TSV;
+SELECT 'TSKV';
+SELECT * FROM map_formats ORDER BY m['k1'] FORMAT TSKV;
diff --git a/tests/queries/0_stateless/01550_type_map_formats_input.reference b/tests/queries/0_stateless/01550_type_map_formats_input.reference
new file mode 100644
index 00000000000..a360e6971d9
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats_input.reference
@@ -0,0 +1,5 @@
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
+{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}
diff --git a/tests/queries/0_stateless/01550_type_map_formats_input.sh b/tests/queries/0_stateless/01550_type_map_formats_input.sh
new file mode 100755
index 00000000000..e62002dc6ce
--- /dev/null
+++ b/tests/queries/0_stateless/01550_type_map_formats_input.sh
@@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+$CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS map_formats_input"
+$CLICKHOUSE_CLIENT -q "CREATE TABLE map_formats_input (m Map(String, UInt32), m1 Map(String, Date), m2 Map(String, Array(UInt32))) ENGINE = Log;" --allow_experimental_map_type 1
+
+$CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT JSONEachRow" <<< '{"m":{"k1":1,"k2":2,"k3":3},"m1":{"k1":"2020-05-05"},"m2":{"k1":[],"k2":[7,8]}}'
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
+$CLICKHOUSE_CLIENT -q "TRUNCATE TABLE map_formats_input"
+
+$CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT CSV" <<< "\"{'k1':1,'k2':2,'k3':3}\",\"{'k1':'2020-05-05'}\",\"{'k1':[],'k2':[7,8]}\""
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
+$CLICKHOUSE_CLIENT -q "TRUNCATE TABLE map_formats_input"
+
+$CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT TSV" <<< "{'k1':1,'k2':2,'k3':3}	{'k1':'2020-05-05'}	{'k1':[],'k2':[7,8]}"
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
+
+$CLICKHOUSE_CLIENT -q 'SELECT * FROM map_formats_input FORMAT Native' | $CLICKHOUSE_CLIENT -q "INSERT INTO map_formats_input FORMAT Native"
+$CLICKHOUSE_CLIENT -q "SELECT * FROM map_formats_input"
diff --git a/tests/queries/0_stateless/01556_accurate_cast_or_null.reference b/tests/queries/0_stateless/01556_accurate_cast_or_null.reference
new file mode 100644
index 00000000000..4287c755c3f
--- /dev/null
+++ b/tests/queries/0_stateless/01556_accurate_cast_or_null.reference
@@ -0,0 +1,30 @@
+\N
+5
+\N
+\N
+5
+\N
+\N
+5
+\N
+\N
+5
+\N
+5
+\N
+5
+\N
+\N
+1.000000000
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
+\N
diff --git a/tests/queries/0_stateless/01556_accurate_cast_or_null.sql b/tests/queries/0_stateless/01556_accurate_cast_or_null.sql
new file mode 100644
index 00000000000..b11ed5939fa
--- /dev/null
+++ b/tests/queries/0_stateless/01556_accurate_cast_or_null.sql
@@ -0,0 +1,35 @@
+SELECT accurateCastOrNull(-1, 'UInt8');
+SELECT accurateCastOrNull(5, 'UInt8');
+SELECT accurateCastOrNull(257, 'UInt8');
+SELECT accurateCastOrNull(-1, 'UInt16');
+SELECT accurateCastOrNull(5, 'UInt16');
+SELECT accurateCastOrNull(65536, 'UInt16');
+SELECT accurateCastOrNull(-1, 'UInt32');
+SELECT accurateCastOrNull(5, 'UInt32');
+SELECT accurateCastOrNull(4294967296, 'UInt32');
+SELECT accurateCastOrNull(-1, 'UInt64');
+SELECT accurateCastOrNull(5, 'UInt64');
+SELECT accurateCastOrNull(-1, 'UInt256');
+SELECT accurateCastOrNull(5, 'UInt256');
+
+SELECT accurateCastOrNull(-129, 'Int8');
+SELECT accurateCastOrNull(5, 'Int8');
+SELECT accurateCastOrNull(128, 'Int8');
+
+SELECT accurateCastOrNull(10, 'Decimal32(9)');
+SELECT accurateCastOrNull(1, 'Decimal32(9)');
+SELECT accurateCastOrNull(-10, 'Decimal32(9)');
+
+SELECT accurateCastOrNull('123', 'FixedString(2)');
+
+SELECT accurateCastOrNull(inf, 'Int64');
+SELECT accurateCastOrNull(inf, 'Int128');
+SELECT accurateCastOrNull(inf, 'Int256');
+SELECT accurateCastOrNull(nan, 'Int64');
+SELECT accurateCastOrNull(nan, 'Int128');
+SELECT accurateCastOrNull(nan, 'Int256');
+
+SELECT accurateCastOrNull(inf, 'UInt64');
+SELECT accurateCastOrNull(inf, 'UInt256');
+SELECT accurateCastOrNull(nan, 'UInt64');
+SELECT accurateCastOrNull(nan, 'UInt256');
diff --git a/tests/queries/0_stateless/01560_optimize_on_insert.reference b/tests/queries/0_stateless/01560_optimize_on_insert.reference
new file mode 100644
index 00000000000..7ace2043be0
--- /dev/null
+++ b/tests/queries/0_stateless/01560_optimize_on_insert.reference
@@ -0,0 +1,13 @@
+Replacing Merge Tree
+1	2020-01-01 00:00:00
+2	2020-01-02 00:00:00
+Collapsing Merge Tree
+1	1	2020-01-01 00:00:00
+Versioned Collapsing Merge Tree
+1	1	2	2020-01-01 00:00:00
+Summing Merge Tree
+1	6	2020-01-01 00:00:00
+2	6	2020-01-02 00:00:00
+Aggregating Merge Tree
+1	5	2020-01-01 00:00:00
+2	5	2020-01-02 00:00:00
diff --git a/tests/queries/0_stateless/01560_optimize_on_insert.sql b/tests/queries/0_stateless/01560_optimize_on_insert.sql
new file mode 100644
index 00000000000..9f6dac686bb
--- /dev/null
+++ b/tests/queries/0_stateless/01560_optimize_on_insert.sql
@@ -0,0 +1,35 @@
+SELECT 'Replacing Merge Tree';
+DROP TABLE IF EXISTS replacing_merge_tree;
+CREATE TABLE replacing_merge_tree (key UInt32, date Datetime) ENGINE=ReplacingMergeTree() PARTITION BY date ORDER BY key;
+INSERT INTO replacing_merge_tree VALUES (1, '2020-01-01'), (2, '2020-01-02'), (1, '2020-01-01'), (2, '2020-01-02');
+SELECT * FROM replacing_merge_tree ORDER BY key;
+DROP TABLE replacing_merge_tree;
+
+SELECT 'Collapsing Merge Tree';
+DROP TABLE IF EXISTS collapsing_merge_tree;
+CREATE TABLE collapsing_merge_tree (key UInt32, sign Int8, date Datetime) ENGINE=CollapsingMergeTree(sign) PARTITION BY date ORDER BY key;
+INSERT INTO collapsing_merge_tree VALUES (1, 1, '2020-01-01'), (2, 1, '2020-01-02'), (1, -1, '2020-01-01'), (2, -1, '2020-01-02'), (1, 1, '2020-01-01');
+SELECT * FROM collapsing_merge_tree ORDER BY key;
+DROP TABLE collapsing_merge_tree;
+
+SELECT 'Versioned Collapsing Merge Tree';
+DROP TABLE IF EXISTS versioned_collapsing_merge_tree;
+CREATE TABLE versioned_collapsing_merge_tree (key UInt32, sign Int8, version Int32, date Datetime) ENGINE=VersionedCollapsingMergeTree(sign, version) PARTITION BY date ORDER BY (key, version);
+INSERT INTO versioned_collapsing_merge_tree VALUES (1, 1, 1, '2020-01-01'), (1, -1, 1, '2020-01-01'), (1, 1, 2, '2020-01-01');
+SELECT * FROM versioned_collapsing_merge_tree ORDER BY key;
+DROP TABLE versioned_collapsing_merge_tree;
+
+SELECT 'Summing Merge Tree';
+DROP TABLE IF EXISTS summing_merge_tree;
+CREATE TABLE summing_merge_tree (key UInt32, val UInt32, date Datetime) ENGINE=SummingMergeTree(val) PARTITION BY date ORDER BY key;
+INSERT INTO summing_merge_tree VALUES (1, 1, '2020-01-01'), (2, 1, '2020-01-02'), (1, 5, '2020-01-01'), (2, 5, '2020-01-02');
+SELECT * FROM summing_merge_tree ORDER BY key;
+DROP TABLE summing_merge_tree;
+
+SELECT 'Aggregating Merge Tree';
+DROP TABLE IF EXISTS aggregating_merge_tree;
+CREATE TABLE aggregating_merge_tree (key UInt32, val SimpleAggregateFunction(max, UInt32), date Datetime) ENGINE=AggregatingMergeTree() PARTITION BY date ORDER BY key;
+INSERT INTO aggregating_merge_tree VALUES (1, 1, '2020-01-01'), (2, 1, '2020-01-02'), (1, 5, '2020-01-01'), (2, 5, '2020-01-02');
+SELECT * FROM aggregating_merge_tree ORDER BY key;
+DROP TABLE aggregating_merge_tree;
+
diff --git a/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference
new file mode 100644
index 00000000000..1c7908bf830
--- /dev/null
+++ b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.reference
@@ -0,0 +1,14 @@
+SimpleAggregateFunction(any, UInt64)	0
+SimpleAggregateFunction(anyLast, UInt64)	0
+SimpleAggregateFunction(min, UInt64)	0
+SimpleAggregateFunction(max, UInt64)	0
+SimpleAggregateFunction(sum, UInt64)	0
+SimpleAggregateFunction(sumWithOverflow, UInt64)	0
+SimpleAggregateFunction(groupBitAnd, UInt64)	0
+SimpleAggregateFunction(groupBitOr, UInt64)	0
+SimpleAggregateFunction(groupBitXor, UInt64)	0
+SimpleAggregateFunction(sumMap, Tuple(Array(UInt64), Array(UInt64)))	([],[])
+SimpleAggregateFunction(minMap, Tuple(Array(UInt64), Array(UInt64)))	([0],[0])
+SimpleAggregateFunction(maxMap, Tuple(Array(UInt64), Array(UInt64)))	([0],[0])
+SimpleAggregateFunction(groupArrayArray, Array(UInt64))	[0]
+SimpleAggregateFunction(groupUniqArrayArray, Array(UInt64))	[0]
diff --git a/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql
new file mode 100644
index 00000000000..00a12a69d16
--- /dev/null
+++ b/tests/queries/0_stateless/01570_aggregator_combinator_simple_state.sql
@@ -0,0 +1,17 @@
+with anySimpleState(number) as c select toTypeName(c), c from numbers(1);
+with anyLastSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with minSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with maxSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with sumSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with sumWithOverflowSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with groupBitAndSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with groupBitOrSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with groupBitXorSimpleState(number) as c select toTypeName(c), c from numbers(1);
+with sumMapSimpleState(([number], [number])) as c select toTypeName(c), c from numbers(1);
+with minMapSimpleState(([number], [number])) as c select toTypeName(c), c from numbers(1);
+with maxMapSimpleState(([number], [number])) as c select toTypeName(c), c from numbers(1);
+with groupArrayArraySimpleState([number]) as c select toTypeName(c), c from numbers(1);
+with groupUniqArrayArraySimpleState([number]) as c select toTypeName(c), c from numbers(1);
+
+-- non-SimpleAggregateFunction
+with countSimpleState(number) as c select toTypeName(c), c from numbers(1); -- { serverError 36 }
diff --git a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference b/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference
deleted file mode 100644
index 37883cc6c47..00000000000
--- a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.reference
+++ /dev/null
@@ -1,6 +0,0 @@
--9223372036854775808
-170141183460469231731687303715884105727
--9223372036854775808
-170141183460469231731687303715884105727
-0
-0
diff --git a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql b/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql
deleted file mode 100644
index b28fc80774f..00000000000
--- a/tests/queries/0_stateless/01581_to_int256_uint256_inf_nan.sql
+++ /dev/null
@@ -1,10 +0,0 @@
-SELECT toInt64(inf);
-SELECT toInt128(inf);
-SELECT toInt256(inf); -- { serverError 48 }
-SELECT toInt64(nan);
-SELECT toInt128(nan);
-SELECT toInt256(nan); -- { serverError 48 }
-SELECT toUInt64(inf);
-SELECT toUInt256(inf); -- { serverError 48 }
-SELECT toUInt64(nan);
-SELECT toUInt256(nan); -- { serverError 48 }
diff --git a/tests/queries/0_stateless/01581_to_int_inf_nan.reference b/tests/queries/0_stateless/01581_to_int_inf_nan.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01581_to_int_inf_nan.sql b/tests/queries/0_stateless/01581_to_int_inf_nan.sql
new file mode 100644
index 00000000000..4959b4d61e9
--- /dev/null
+++ b/tests/queries/0_stateless/01581_to_int_inf_nan.sql
@@ -0,0 +1,10 @@
+SELECT toInt64(inf); -- { serverError 70 }
+SELECT toInt128(inf); -- { serverError 70 }
+SELECT toInt256(inf); -- { serverError 70 }
+SELECT toInt64(nan); -- { serverError 70 }
+SELECT toInt128(nan); -- { serverError 70 }
+SELECT toInt256(nan); -- { serverError 70 }
+SELECT toUInt64(inf); -- { serverError 70 }
+SELECT toUInt256(inf); -- { serverError 70 }
+SELECT toUInt64(nan); -- { serverError 70 }
+SELECT toUInt256(nan); -- { serverError 70 }
diff --git a/tests/queries/0_stateless/01596_setting_limit_offset.reference b/tests/queries/0_stateless/01596_setting_limit_offset.reference
new file mode 100644
index 00000000000..96483268d43
--- /dev/null
+++ b/tests/queries/0_stateless/01596_setting_limit_offset.reference
@@ -0,0 +1,63 @@
+0
+1
+2
+3
+4
+15
+15
+16
+16
+17
+30
+30
+31
+31
+32
+102
+103
+104
+105
+105
+106
+107
+108
+109
+105
+106
+107
+108
+109
+60
+60
+61
+61
+62
+62
+63
+63
+64
+64
+60
+35
+35
+36
+36
+37
+37
+38
+38
+39
+39
+105
+106
+107
+108
+109
+12
+13
+13
+14
+14
+15
+15
+16
diff --git a/tests/queries/0_stateless/01596_setting_limit_offset.sql b/tests/queries/0_stateless/01596_setting_limit_offset.sql
new file mode 100644
index 00000000000..7c009a95837
--- /dev/null
+++ b/tests/queries/0_stateless/01596_setting_limit_offset.sql
@@ -0,0 +1,30 @@
+DROP TABLE IF EXISTS test;
+CREATE TABLE test (i UInt64) Engine = MergeTree() order by i;
+INSERT INTO test SELECT number FROM numbers(100);
+INSERT INTO test SELECT number FROM numbers(10,100);
+OPTIMIZE TABLE test FINAL;
+
+-- Only set limit
+SET limit = 5;
+SELECT * FROM test; -- 5 rows 
+SELECT * FROM test OFFSET 20; -- 5 rows 
+SELECT * FROM (SELECT i FROM test LIMIT 10 OFFSET 50) TMP; -- 5 rows 
+SELECT * FROM test LIMIT 4 OFFSET 192; -- 4 rows 
+SELECT * FROM test LIMIT 10 OFFSET 195; -- 5 rows
+
+-- Only set offset
+SET limit = 0;
+SET offset = 195;
+SELECT * FROM test; -- 5 rows
+SELECT * FROM test OFFSET 20; -- no result
+SELECT * FROM test LIMIT 100; -- no result
+SET offset = 10;
+SELECT * FROM test LIMIT 20 OFFSET 100; -- 10 rows
+SELECT * FROM test LIMIT 11 OFFSET 100; -- 1 rows
+
+-- offset and limit together
+SET limit = 10;
+SELECT * FROM test LIMIT 50 OFFSET 50; -- 10 rows 
+SELECT * FROM test LIMIT 50 OFFSET 190; -- 0 rows
+SELECT * FROM test LIMIT 50 OFFSET 185; -- 5 rows 
+SELECT * FROM test LIMIT 18 OFFSET 5; -- 8 rows 
diff --git a/tests/queries/0_stateless/01600_detach_permanently.reference b/tests/queries/0_stateless/01600_detach_permanently.reference
new file mode 100644
index 00000000000..98ed3b6762d
--- /dev/null
+++ b/tests/queries/0_stateless/01600_detach_permanently.reference
@@ -0,0 +1,102 @@
+##################
+setup env
+##################
+db_ordinary.log_table 1
+>table detached!
+db_ordinary.log_table 2
+>Table is back after restart
+10
+db_ordinary.log_table 3
+>table detached (permanently)!
+db_ordinary.log_table 4
+>table is still detached (after restart)!
+db_ordinary.log_table 5
+>Table is back after attach
+db_ordinary.log_table 6
+10
+##################
+db_ordinary.mt_table 1
+>table detached!
+db_ordinary.mt_table 2
+>Table is back after restart
+10
+db_ordinary.mt_table 3
+>table detached (permanently)!
+db_ordinary.mt_table 4
+>table is still detached (after restart)!
+db_ordinary.mt_table 5
+>Table is back after attach
+db_ordinary.mt_table 6
+10
+##################
+db_ordinary.null_table 1
+>table detached!
+db_ordinary.null_table 2
+>Table is back after restart
+0
+db_ordinary.null_table 3
+>table detached (permanently)!
+db_ordinary.null_table 4
+>table is still detached (after restart)!
+db_ordinary.null_table 5
+>Table is back after attach
+db_ordinary.null_table 6
+0
+##################
+db_atomic.log_table 1
+>table detached!
+db_atomic.log_table 2
+>Table is back after restart
+10
+db_atomic.log_table 3
+>table detached (permanently)!
+db_atomic.log_table 4
+>table is still detached (after restart)!
+db_atomic.log_table 5
+>Table is back after attach
+db_atomic.log_table 6
+10
+##################
+db_atomic.mt_table 1
+>table detached!
+db_atomic.mt_table 2
+>Table is back after restart
+10
+db_atomic.mt_table 3
+>table detached (permanently)!
+db_atomic.mt_table 4
+>table is still detached (after restart)!
+db_atomic.mt_table 5
+>Table is back after attach
+db_atomic.mt_table 6
+10
+##################
+db_atomic.null_table 1
+>table detached!
+db_atomic.null_table 2
+>Table is back after restart
+0
+db_atomic.null_table 3
+>table detached (permanently)!
+db_atomic.null_table 4
+>table is still detached (after restart)!
+db_atomic.null_table 5
+>Table is back after attach
+db_atomic.null_table 6
+0
+##################
+test for MV
+MV is working
+Usual detach works immediately till restart
+Usual detach activates after restart
+Permanent detach works immediately
+Permanent detach still works after restart
+View can be reattached
+##################
+test for MV with inner table
+MV is working
+1
+View can be reattached
+##################
+DETACH DATABASE is not implemented (proper error)
+1
diff --git a/tests/queries/0_stateless/01600_detach_permanently.sh b/tests/queries/0_stateless/01600_detach_permanently.sh
new file mode 100755
index 00000000000..e897f80a33f
--- /dev/null
+++ b/tests/queries/0_stateless/01600_detach_permanently.sh
@@ -0,0 +1,125 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+## tests with real clickhouse restart would be a bit to heavy,
+## to ensure the table will not reappear back clickhose-local is enough.
+
+WORKING_FOLDER_01600="${CLICKHOUSE_TMP}/01600_detach_permanently"
+rm -rf "${WORKING_FOLDER_01600}"
+mkdir -p "${WORKING_FOLDER_01600}"
+
+clickhouse_local() {
+    local query="$1"
+    shift
+    ${CLICKHOUSE_LOCAL} --query "$query" $@ -- --path="${WORKING_FOLDER_01600}"
+}
+
+test_detach_attach_sequence() {
+    local db="$1"
+    local table="$2"
+    echo "##################"
+
+    echo "${db}.${table} 1"
+    # normal DETACH - while process is running (clickhouse-local here, same for server) table is detached.
+    clickhouse_local "DETACH TABLE ${db}.${table}; SELECT if( count() = 0, '>table detached!', '>Fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    # but once we restart the precess (either clickhouse-local either clickhouse server) the table is back.
+    echo "${db}.${table} 2"
+    clickhouse_local "SELECT if(name='${table}', '>Table is back after restart', '>fail') FROM system.tables WHERE database='${db}' AND name='${table}'; SELECT count() FROM ${db}.${table};"
+
+    # permanent DETACH - table is detached, and metadata file renamed, prevening further attach
+    echo "${db}.${table} 3"
+    clickhouse_local "DETACH TABLE ${db}.${table} PERMANENTLY; SELECT if( count() = 0, '>table detached (permanently)!', '>Fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    # still detached after restart
+    echo "${db}.${table} 4"
+    clickhouse_local "SELECT if( count() = 0, '>table is still detached (after restart)!', '>Fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    # but can be reattached
+    echo "${db}.${table} 5"
+    clickhouse_local "ATTACH TABLE ${db}.${table}; SELECT if(name='${table}', '>Table is back after attach', '>fail') FROM system.tables WHERE database='${db}' AND name='${table}';"
+
+    echo "${db}.${table} 6"
+    clickhouse_local "SELECT count() FROM ${db}.${table};"
+}
+
+echo "##################"
+echo "setup env"
+
+clickhouse_local "DROP DATABASE IF EXISTS db_ordinary SYNC;"
+clickhouse_local "DROP DATABASE IF EXISTS db_atomic SYNC;"
+
+clickhouse_local "CREATE DATABASE db_ordinary Engine=Ordinary"
+clickhouse_local "CREATE DATABASE db_atomic Engine=Atomic"
+
+clickhouse_local "CREATE TABLE db_ordinary.log_table Engine=Log AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_ordinary.mt_table Engine=MergeTree ORDER BY tuple() AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_ordinary.null_table Engine=Null AS SELECT * FROM numbers(10)"
+
+clickhouse_local "CREATE TABLE db_atomic.log_table Engine=Log AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_atomic.mt_table Engine=MergeTree ORDER BY tuple() AS SELECT * FROM numbers(10)"
+clickhouse_local "CREATE TABLE db_atomic.null_table Engine=Null AS SELECT * FROM numbers(10)"
+
+test_detach_attach_sequence "db_ordinary" "log_table"
+test_detach_attach_sequence "db_ordinary" "mt_table"
+test_detach_attach_sequence "db_ordinary" "null_table"
+
+test_detach_attach_sequence "db_atomic" "log_table"
+test_detach_attach_sequence "db_atomic" "mt_table"
+test_detach_attach_sequence "db_atomic" "null_table"
+
+echo "##################"
+echo "test for MV"
+clickhouse_local "CREATE TABLE db_ordinary.src Engine=Null AS system.numbers"
+clickhouse_local "CREATE TABLE db_ordinary.dst Engine=Log AS system.numbers"
+clickhouse_local "CREATE MATERIALIZED VIEW db_ordinary.src2dst_mv_to TO db_ordinary.dst AS SELECT * FROM db_ordinary.src"
+
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 10, 'MV is working', 'MV failed') from db_ordinary.dst;"
+
+clickhouse_local "DETACH VIEW db_ordinary.src2dst_mv_to; INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 10, 'Usual detach works immediately till restart', 'Usual detach failed') from db_ordinary.dst;"
+
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'Usual detach activates after restart', 'Usual detach reactivation failed') from db_ordinary.dst;"
+
+clickhouse_local "DETACH VIEW db_ordinary.src2dst_mv_to PERMANENTLY; INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'Permanent detach works immediately', 'Permanent detach failed') from db_ordinary.dst;"
+
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'Permanent detach still works after restart', 'Permanent detach reactivated!') from db_ordinary.dst;"
+
+## Quite silly: ATTACH MATERIALIZED VIEW don't work with short syntax (w/o select), but i can attach it using ATTACH TABLE ...
+clickhouse_local "ATTACH TABLE db_ordinary.src2dst_mv_to"
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 30, 'View can be reattached', 'can not reattach permanently detached view') from db_ordinary.dst;"
+
+clickhouse_local "DROP VIEW db_ordinary.src2dst_mv_to SYNC"
+
+echo "##################"
+echo "test for MV with inner table"
+clickhouse_local "CREATE MATERIALIZED VIEW db_ordinary.src_mv_with_inner Engine=Log AS SELECT * FROM db_ordinary.src"
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+
+clickhouse_local "SELECT if(count() = 10, 'MV is working', 'MV failed') FROM db_ordinary.src_mv_with_inner"
+
+clickhouse_local "DETACH VIEW db_ordinary.src_mv_with_inner PERMANENTLY; INSERT INTO db_ordinary.src SELECT * FROM numbers(10)" --stacktrace
+clickhouse_local "SELECT if(count() = 10, 'MV can be detached permanently', 'MV detach failed') FROM db_ordinary.src_mv_with_inner" 2>&1 | grep -c "db_ordinary.src_mv_with_inner doesn't exist"
+## Quite silly: ATTACH MATERIALIZED VIEW don't work with short syntax (w/o select), but i can attach it using ATTACH TABLE ...
+clickhouse_local "ATTACH TABLE db_ordinary.src_mv_with_inner"
+clickhouse_local "INSERT INTO db_ordinary.src SELECT * FROM numbers(10)"
+clickhouse_local "SELECT if(count() = 20, 'View can be reattached', 'can not reattach permanently detached view') from db_ordinary.src_mv_with_inner;"
+
+## clickhouse_local can't work with dicts...
+# mkdir -p "${WORKING_FOLDER_01600}/user_files"
+# echo "1" > "${WORKING_FOLDER_01600}/user_files/dummy_dict.tsv"
+# clickhouse_local "DROP DICTIONARY db_ordinary.dummy; CREATE DICTIONARY db_ordinary.dummy (id UInt64) PRIMARY KEY id LAYOUT(FLAT()) SOURCE(FILE(path 'dummy_dict.tsv' format 'TabSeparated')) LIFETIME(MIN 1 MAX 10); DETACH DICTIONARY db_ordinary.dummy PERMANENTLY; SELECT dictGet('db_ordinary.dummy','val',toUInt64(1));"
+
+echo "##################"
+echo "DETACH DATABASE is not implemented (proper error)"
+clickhouse_local "DETACH DATABASE db_ordinary PERMANENTLY;" 2>&1 | grep -c 'DETACH PERMANENTLY is not implemented'
+
+# clean up
+rm -rf "${WORKING_FOLDER_01600}"
diff --git a/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.reference b/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql b/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql
new file mode 100644
index 00000000000..b33b74c918d
--- /dev/null
+++ b/tests/queries/0_stateless/01600_remerge_sort_lowered_memory_bytes_ratio.sql
@@ -0,0 +1,29 @@
+-- Check remerge_sort_lowered_memory_bytes_ratio setting
+
+set max_memory_usage='300Mi';
+-- enter remerge once limit*2 is reached
+set max_bytes_before_remerge_sort='10Mi';
+-- more blocks
+set max_block_size=40960;
+
+-- remerge_sort_lowered_memory_bytes_ratio default 2, slightly not enough
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 819200 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 186.25 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging is not useful (memory usage was not lowered by remerge_sort_lowered_memory_bytes_ratio=2.0)
+select number k, repeat(toString(number), 11) v1, repeat(toString(number), 12) v2 from numbers(toUInt64(3e6)) order by k limit 400e3 format Null; -- { serverError 241 }
+select number k, repeat(toString(number), 11) v1, repeat(toString(number), 12) v2 from numbers(toUInt64(3e6)) order by k limit 400e3 settings remerge_sort_lowered_memory_bytes_ratio=2. format Null; -- { serverError 241 }
+
+-- remerge_sort_lowered_memory_bytes_ratio 1.9 is good (need at least 1.91/0.98=1.94)
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 819200 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 186.25 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+--     MergeSortingTransform: Re-merging intermediate ORDER BY data (20 blocks with 809600 rows) to save memory consumption
+--     MergeSortingTransform: Memory usage is lowered from 188.13 MiB to 95.00 MiB
+select number k, repeat(toString(number), 11) v1, repeat(toString(number), 12) v2 from numbers(toUInt64(3e6)) order by k limit 400e3 settings remerge_sort_lowered_memory_bytes_ratio=1.9 format Null;
diff --git a/tests/queries/0_stateless/01600_select_in_different_types.reference b/tests/queries/0_stateless/01600_select_in_different_types.reference
new file mode 100644
index 00000000000..626fb152744
--- /dev/null
+++ b/tests/queries/0_stateless/01600_select_in_different_types.reference
@@ -0,0 +1,11 @@
+1
+0
+1
+1
+2
+2
+1
+1
+1
+0
+0
diff --git a/tests/queries/0_stateless/01600_select_in_different_types.sql b/tests/queries/0_stateless/01600_select_in_different_types.sql
new file mode 100644
index 00000000000..25d37c122e0
--- /dev/null
+++ b/tests/queries/0_stateless/01600_select_in_different_types.sql
@@ -0,0 +1,35 @@
+SELECT 1 IN (SELECT 1);
+SELECT -1 IN (SELECT 1);
+
+DROP TABLE IF EXISTS select_in_test;
+
+CREATE TABLE select_in_test(value UInt8) ENGINE=TinyLog;
+INSERT INTO select_in_test VALUES (1), (2), (3);
+
+SELECT value FROM select_in_test WHERE value IN (-1);
+SELECT value FROM select_in_test WHERE value IN (SELECT -1);
+
+SELECT value FROM select_in_test WHERE value IN (1);
+SELECT value FROM select_in_test WHERE value IN (SELECT 1);
+
+DROP TABLE select_in_test;
+
+CREATE TABLE select_in_test(value Int8) ENGINE=TinyLog;
+INSERT INTO select_in_test VALUES (-1), (2), (3);
+
+SELECT value FROM select_in_test WHERE value IN (1);
+SELECT value FROM select_in_test WHERE value IN (SELECT 1);
+
+SELECT value FROM select_in_test WHERE value IN (2);
+SELECT value FROM select_in_test WHERE value IN (SELECT 2);
+
+DROP TABLE select_in_test;
+
+SELECT 1 IN (1);
+SELECT '1' IN (SELECT 1);
+
+SELECT 1 IN (SELECT 1) SETTINGS transform_null_in = 1;
+SELECT 1 IN (SELECT 'a') SETTINGS transform_null_in = 1;
+SELECT 'a' IN (SELECT 1) SETTINGS transform_null_in = 1; -- { serverError 6 }
+SELECT 1 IN (SELECT -1) SETTINGS transform_null_in = 1;
+SELECT -1 IN (SELECT 1) SETTINGS transform_null_in = 1; -- { serverError 70 }
diff --git a/tests/queries/0_stateless/01601_accurate_cast.reference b/tests/queries/0_stateless/01601_accurate_cast.reference
new file mode 100644
index 00000000000..6ccedbdcfea
--- /dev/null
+++ b/tests/queries/0_stateless/01601_accurate_cast.reference
@@ -0,0 +1,8 @@
+5
+5
+5
+5
+5
+5
+1.000000000
+12
diff --git a/tests/queries/0_stateless/01601_accurate_cast.sql b/tests/queries/0_stateless/01601_accurate_cast.sql
new file mode 100644
index 00000000000..b5fd4fb04a4
--- /dev/null
+++ b/tests/queries/0_stateless/01601_accurate_cast.sql
@@ -0,0 +1,24 @@
+SELECT accurateCast(-1, 'UInt8'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt8');
+SELECT accurateCast(257, 'UInt8'); -- { serverError 70 }
+SELECT accurateCast(-1, 'UInt16'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt16');
+SELECT accurateCast(65536, 'UInt16'); -- { serverError 70 }
+SELECT accurateCast(-1, 'UInt32'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt32');
+SELECT accurateCast(4294967296, 'UInt32'); -- { serverError 70 }
+SELECT accurateCast(-1, 'UInt64'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt64');
+SELECT accurateCast(-1, 'UInt256'); -- { serverError 70 }
+SELECT accurateCast(5, 'UInt256');
+
+SELECT accurateCast(-129, 'Int8'); -- { serverError 70 }
+SELECT accurateCast(5, 'Int8');
+SELECT accurateCast(128, 'Int8'); -- { serverError 70 }
+
+SELECT accurateCast(10, 'Decimal32(9)'); -- { serverError 407 }
+SELECT accurateCast(1, 'Decimal32(9)');
+SELECT accurateCast(-10, 'Decimal32(9)'); -- { serverError 407 }
+
+SELECT accurateCast('123', 'FixedString(2)'); -- { serverError 131 }
+SELECT accurateCast('12', 'FixedString(2)');
diff --git a/tests/queries/0_stateless/01601_detach_permanently.reference b/tests/queries/0_stateless/01601_detach_permanently.reference
new file mode 100644
index 00000000000..6683603c972
--- /dev/null
+++ b/tests/queries/0_stateless/01601_detach_permanently.reference
@@ -0,0 +1,79 @@
+database atomic tests
+can not create table with same name as detached permanently
+can not detach twice
+can not drop detached
+can not replace with the other table
+can still show the create statement
+Row 1:
+──────
+statement: CREATE TABLE test1601_detach_permanently_atomic.test_name_reuse
+(
+    `number` UInt64
+)
+ENGINE = MergeTree
+ORDER BY tuple()
+SETTINGS index_granularity = 8192
+can not attach with bad uuid
+can attach with short syntax
+100
+can not detach permanently the table which is already detached (temporary)
+100
+After database reattachement the table is back (it was detached temporary)
+And we can detach it permanently
+After database reattachement the table is still absent (it was detached permamently)
+And we can not detach it permanently
+But we can attach it back
+And detach permanently again to check how database drop will behave
+DROP database
+-----------------------
+database ordinary tests
+can not create table with same name as detached permanently
+can not detach twice
+can not drop detached
+can not replace with the other table
+can still show the create statement
+Row 1:
+──────
+statement: CREATE TABLE test1601_detach_permanently_ordinary.test_name_reuse
+(
+    `number` UInt64
+)
+ENGINE = MergeTree
+ORDER BY tuple()
+SETTINGS index_granularity = 8192
+can attach with full syntax
+can attach with short syntax
+can not detach permanently the table which is already detached (temporary)
+After database reattachement the table is back (it was detached temporary)
+And we can detach it permanently
+After database reattachement the table is still absent (it was detached permamently)
+And we can not detach it permanently
+But we can attach it back
+And detach permanently again to check how database drop will behave
+DROP database - Directory not empty error, but database deteched
+DROP database - now success
+-----------------------
+database lazy tests
+can not create table with same name as detached permanently
+can not detach twice
+can not drop detached
+can not replace with the other table
+can still show the create statement
+Row 1:
+──────
+statement: CREATE TABLE test1601_detach_permanently_lazy.test_name_reuse
+(
+    `number` UInt64
+)
+ENGINE = Log
+can attach with full syntax
+can attach with short syntax
+can not detach permanently the table which is already detached (temporary)
+After database reattachement the table is back (it was detached temporary)
+And we can detach it permanently
+After database reattachement the table is still absent (it was detached permamently)
+And we can not detach it permanently
+But we can attach it back
+And detach permanently again to check how database drop will behave
+DROP database - Directory not empty error, but database deteched
+DROP database - now success
diff --git a/tests/queries/0_stateless/01601_detach_permanently.sql b/tests/queries/0_stateless/01601_detach_permanently.sql
new file mode 100644
index 00000000000..0e21fd8f463
--- /dev/null
+++ b/tests/queries/0_stateless/01601_detach_permanently.sql
@@ -0,0 +1,214 @@
+SELECT 'database atomic tests';
+
+DROP DATABASE IF EXISTS test1601_detach_permanently_atomic;
+CREATE DATABASE test1601_detach_permanently_atomic Engine=Atomic;
+
+create table test1601_detach_permanently_atomic.test_name_reuse (number UInt64) engine=MergeTree order by tuple();
+
+INSERT INTO test1601_detach_permanently_atomic.test_name_reuse SELECT * FROM numbers(100);
+
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY;
+
+SELECT 'can not create table with same name as detached permanently';
+create table test1601_detach_permanently_atomic.test_name_reuse (number UInt64) engine=MergeTree order by tuple(); -- { serverError 57 }
+
+SELECT 'can not detach twice';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY; -- { serverError 60 }
+DETACH table test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can not drop detached';
+drop table test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 60 }
+
+create table test1601_detach_permanently_atomic.test_name_rename_attempt (number UInt64) engine=MergeTree order by tuple();
+
+SELECT 'can not replace with the other table';
+RENAME TABLE test1601_detach_permanently_atomic.test_name_rename_attempt TO test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 57 }
+EXCHANGE TABLES test1601_detach_permanently_atomic.test_name_rename_attempt AND test1601_detach_permanently_atomic.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can still show the create statement';
+SHOW CREATE TABLE test1601_detach_permanently_atomic.test_name_reuse FORMAT Vertical;
+
+SELECT 'can not attach with bad uuid';
+ATTACH TABLE test1601_detach_permanently_atomic.test_name_reuse UUID '00000000-0000-0000-0000-000000000001'　(`number` UInt64　)　ENGINE = MergeTree　ORDER BY tuple()　SETTINGS index_granularity = 8192 ;  -- { serverError 57 }
+
+SELECT 'can attach with short syntax';
+ATTACH TABLE test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT count() FROM test1601_detach_permanently_atomic.test_name_reuse;
+
+DETACH table test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT 'can not detach permanently the table which is already detached (temporary)';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+DETACH DATABASE test1601_detach_permanently_atomic;
+ATTACH DATABASE test1601_detach_permanently_atomic;
+
+SELECT count() FROM test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT 'After database reattachement the table is back (it was detached temporary)';
+SELECT 'And we can detach it permanently';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY;
+
+DETACH DATABASE test1601_detach_permanently_atomic;
+ATTACH DATABASE test1601_detach_permanently_atomic;
+
+SELECT 'After database reattachement the table is still absent (it was detached permamently)';
+SELECT 'And we can not detach it permanently';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+SELECT 'But we can attach it back';
+ATTACH TABLE test1601_detach_permanently_atomic.test_name_reuse;
+
+SELECT 'And detach permanently again to check how database drop will behave';
+DETACH table test1601_detach_permanently_atomic.test_name_reuse PERMANENTLY;
+
+SELECT 'DROP database';
+DROP DATABASE test1601_detach_permanently_atomic SYNC;
+
+SELECT '-----------------------';
+SELECT 'database ordinary tests';
+
+DROP DATABASE IF EXISTS test1601_detach_permanently_ordinary;
+CREATE DATABASE test1601_detach_permanently_ordinary Engine=Ordinary;
+
+create table test1601_detach_permanently_ordinary.test_name_reuse (number UInt64) engine=MergeTree order by tuple();
+
+INSERT INTO test1601_detach_permanently_ordinary.test_name_reuse SELECT * FROM numbers(100);
+
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+SELECT 'can not create table with same name as detached permanently';
+create table test1601_detach_permanently_ordinary.test_name_reuse (number UInt64) engine=MergeTree order by tuple(); -- { serverError 57 }
+
+SELECT 'can not detach twice';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY; -- { serverError 60 }
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can not drop detached';
+drop table test1601_detach_permanently_ordinary.test_name_reuse; -- { serverError 60 }
+
+create table test1601_detach_permanently_ordinary.test_name_rename_attempt (number UInt64) engine=MergeTree order by tuple();
+
+SELECT 'can not replace with the other table';
+RENAME TABLE test1601_detach_permanently_ordinary.test_name_rename_attempt TO test1601_detach_permanently_ordinary.test_name_reuse; -- { serverError 57 }
+
+SELECT 'can still show the create statement';
+SHOW CREATE TABLE test1601_detach_permanently_ordinary.test_name_reuse FORMAT Vertical;
+
+SELECT 'can attach with full syntax';
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse (`number` UInt64　)　ENGINE = MergeTree　ORDER BY tuple()　SETTINGS index_granularity = 8192;
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+SELECT 'can attach with short syntax';
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse;
+
+SELECT 'can not detach permanently the table which is already detached (temporary)';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+DETACH DATABASE test1601_detach_permanently_ordinary;
+ATTACH DATABASE test1601_detach_permanently_ordinary;
+
+SELECT 'After database reattachement the table is back (it was detached temporary)';
+SELECT 'And we can detach it permanently';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+DETACH DATABASE test1601_detach_permanently_ordinary;
+ATTACH DATABASE test1601_detach_permanently_ordinary;
+
+SELECT 'After database reattachement the table is still absent (it was detached permamently)';
+SELECT 'And we can not detach it permanently';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+SELECT 'But we can attach it back';
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+
+SELECT 'And detach permanently again to check how database drop will behave';
+DETACH table test1601_detach_permanently_ordinary.test_name_reuse PERMANENTLY;
+
+SELECT 'DROP database - Directory not empty error, but database deteched';
+DROP DATABASE test1601_detach_permanently_ordinary; -- { serverError 1000 }
+
+ATTACH DATABASE test1601_detach_permanently_ordinary;
+
+ATTACH TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+DROP TABLE test1601_detach_permanently_ordinary.test_name_reuse;
+
+SELECT 'DROP database - now success';
+DROP DATABASE test1601_detach_permanently_ordinary;
+
+
+SELECT '-----------------------';
+SELECT 'database lazy tests';
+
+DROP DATABASE IF EXISTS test1601_detach_permanently_lazy;
+CREATE DATABASE test1601_detach_permanently_lazy Engine=Lazy(10);
+
+create table test1601_detach_permanently_lazy.test_name_reuse (number UInt64) engine=Log;
+
+INSERT INTO test1601_detach_permanently_lazy.test_name_reuse SELECT * FROM numbers(100);
+
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+SELECT 'can not create table with same name as detached permanently';
+create table test1601_detach_permanently_lazy.test_name_reuse (number UInt64) engine=Log; -- { serverError 57 }
+
+SELECT 'can not detach twice';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY; -- { serverError 60 }
+DETACH table test1601_detach_permanently_lazy.test_name_reuse; -- { serverError 60 }
+
+SELECT 'can not drop detached';
+drop table test1601_detach_permanently_lazy.test_name_reuse; -- { serverError 60 }
+
+create table test1601_detach_permanently_lazy.test_name_rename_attempt (number UInt64) engine=Log;
+
+SELECT 'can not replace with the other table';
+RENAME TABLE test1601_detach_permanently_lazy.test_name_rename_attempt TO test1601_detach_permanently_lazy.test_name_reuse; -- { serverError 57 }
+
+SELECT 'can still show the create statement';
+SHOW CREATE TABLE test1601_detach_permanently_lazy.test_name_reuse FORMAT Vertical;
+
+SELECT 'can attach with full syntax';
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse (`number` UInt64　)　ENGINE = Log;
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+SELECT 'can attach with short syntax';
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse;
+
+DETACH table test1601_detach_permanently_lazy.test_name_reuse;
+
+SELECT 'can not detach permanently the table which is already detached (temporary)';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+DETACH DATABASE test1601_detach_permanently_lazy;
+ATTACH DATABASE test1601_detach_permanently_lazy;
+
+SELECT 'After database reattachement the table is back (it was detached temporary)';
+SELECT 'And we can detach it permanently';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+DETACH DATABASE test1601_detach_permanently_lazy;
+ATTACH DATABASE test1601_detach_permanently_lazy;
+
+SELECT 'After database reattachement the table is still absent (it was detached permamently)';
+SELECT 'And we can not detach it permanently';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY; -- { serverError 60 }
+
+SELECT 'But we can attach it back';
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse;
+
+SELECT 'And detach permanently again to check how database drop will behave';
+DETACH table test1601_detach_permanently_lazy.test_name_reuse PERMANENTLY;
+
+SELECT 'DROP database - Directory not empty error, but database deteched';
+DROP DATABASE test1601_detach_permanently_lazy; -- { serverError 1000 }
+
+ATTACH DATABASE test1601_detach_permanently_lazy;
+
+ATTACH TABLE test1601_detach_permanently_lazy.test_name_reuse;
+DROP TABLE test1601_detach_permanently_lazy.test_name_reuse;
+
+SELECT 'DROP database - now success';
+DROP DATABASE test1601_detach_permanently_lazy;
diff --git a/tests/queries/0_stateless/01601_temporary_table_session_scope.reference b/tests/queries/0_stateless/01601_temporary_table_session_scope.reference
new file mode 100644
index 00000000000..84b736f86c9
--- /dev/null
+++ b/tests/queries/0_stateless/01601_temporary_table_session_scope.reference
@@ -0,0 +1,5 @@
+CREATE TEMPORARY TABLE tmptable (`x` UInt32) ENGINE = Memory
+CREATE TEMPORARY TABLE tmptable (`y` Float64, `z` String) ENGINE = Memory
+x	UInt32	1
+y	Float64	1
+z	String	2
diff --git a/tests/queries/0_stateless/01601_temporary_table_session_scope.sh b/tests/queries/0_stateless/01601_temporary_table_session_scope.sh
new file mode 100755
index 00000000000..fbc45aec216
--- /dev/null
+++ b/tests/queries/0_stateless/01601_temporary_table_session_scope.sh
@@ -0,0 +1,19 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+IF+EXISTS+tmptable&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+IF+EXISTS+tmptable&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=CREATE+TEMPORARY+TABLE+tmptable(x+UInt32)&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=CREATE+TEMPORARY+TABLE+tmptable(y+Float64,+z+String)&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+create_table_query+FROM+system.tables+WHERE+database=''+AND+name='tmptable'&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+create_table_query+FROM+system.tables+WHERE+database=''+AND+name='tmptable'&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+name,type,position+FROM+system.columns+WHERE+database=''+AND+table='tmptable'&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=SELECT+name,type,position+FROM+system.columns+WHERE+database=''+AND+table='tmptable'&session_id=session_1601b"
+
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+tmptable&session_id=session_1601a"
+${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&query=DROP+TEMPORARY+TABLE+tmptable&session_id=session_1601b"
diff --git a/tests/queries/0_stateless/01602_array_aggregation.reference b/tests/queries/0_stateless/01602_array_aggregation.reference
new file mode 100644
index 00000000000..4baef982f29
--- /dev/null
+++ b/tests/queries/0_stateless/01602_array_aggregation.reference
@@ -0,0 +1,60 @@
+Array min 	1
+Array max 	6
+Array sum 	21
+Array avg 	3.5
+Table array int min
+1
+0
+1
+Table array int max
+6
+0
+3
+Table array int sum
+21
+0
+6
+Table array int avg
+3.5
+0
+2
+Table array decimal min
+1.00000000
+0.00000000
+1.00000000
+Table array decimal max
+6.00000000
+0.00000000
+3.00000000
+Table array decimal sum
+21.00000000
+0.00000000
+6.00000000
+Table array decimal avg
+3.5
+0
+2
+Types of aggregation result array min
+Int8	Int16	Int32	Int64
+UInt8	UInt16	UInt32	UInt64
+Int128	Int256	UInt256
+Float32	Float64
+Decimal(9, 8)	Decimal(18, 8)	Decimal(38, 8)
+Types of aggregation result array max
+Int8	Int16	Int32	Int64
+UInt8	UInt16	UInt32	UInt64
+Int128	Int256	UInt256
+Float32	Float64
+Decimal(9, 8)	Decimal(18, 8)	Decimal(38, 8)
+Types of aggregation result array summ
+Int64	Int64	Int64	Int64
+UInt64	UInt64	UInt64	UInt64
+Int128	Int256	UInt256
+Float64	Float64
+Decimal(38, 8)	Decimal(38, 8)	Decimal(38, 8)
+Types of aggregation result array avg
+Float64	Float64	Float64	Float64
+Float64	Float64	Float64	Float64
+Float64	Float64	Float64
+Float64	Float64
+Float64	Float64	Float64
diff --git a/tests/queries/0_stateless/01602_array_aggregation.sql b/tests/queries/0_stateless/01602_array_aggregation.sql
new file mode 100644
index 00000000000..7159eb1442b
--- /dev/null
+++ b/tests/queries/0_stateless/01602_array_aggregation.sql
@@ -0,0 +1,60 @@
+SELECT 'Array min ', (arrayMin(array(1,2,3,4,5,6)));
+SELECT 'Array max ', (arrayMax(array(1,2,3,4,5,6)));
+SELECT 'Array sum ', (arraySum(array(1,2,3,4,5,6)));
+SELECT 'Array avg ', (arrayAvg(array(1,2,3,4,5,6)));
+
+DROP TABLE IF EXISTS test_aggregation;
+CREATE TABLE test_aggregation (x Array(Int)) ENGINE=TinyLog;
+
+INSERT INTO test_aggregation VALUES ([1,2,3,4,5,6]), ([]), ([1,2,3]);
+
+SELECT 'Table array int min';
+SELECT arrayMin(x) FROM test_aggregation;
+SELECT 'Table array int max';
+SELECT arrayMax(x) FROM test_aggregation;
+SELECT 'Table array int sum';
+SELECT arraySum(x) FROM test_aggregation;
+SELECT 'Table array int avg';
+SELECT arrayAvg(x) FROM test_aggregation;
+
+DROP TABLE test_aggregation;
+
+CREATE TABLE test_aggregation (x Array(Decimal64(8))) ENGINE=TinyLog;
+
+INSERT INTO test_aggregation VALUES ([1,2,3,4,5,6]), ([]), ([1,2,3]);
+
+SELECT 'Table array decimal min';
+SELECT arrayMin(x) FROM test_aggregation;
+SELECT 'Table array decimal max';
+SELECT arrayMax(x) FROM test_aggregation;
+SELECT 'Table array decimal sum';
+SELECT arraySum(x) FROM test_aggregation;
+SELECT 'Table array decimal avg';
+SELECT arrayAvg(x) FROM test_aggregation;
+
+DROP TABLE test_aggregation;
+
+SELECT 'Types of aggregation result array min';
+SELECT toTypeName(arrayMin([toInt8(0)])), toTypeName(arrayMin([toInt16(0)])), toTypeName(arrayMin([toInt32(0)])), toTypeName(arrayMin([toInt64(0)]));
+SELECT toTypeName(arrayMin([toUInt8(0)])), toTypeName(arrayMin([toUInt16(0)])), toTypeName(arrayMin([toUInt32(0)])), toTypeName(arrayMin([toUInt64(0)]));
+SELECT toTypeName(arrayMin([toInt128(0)])), toTypeName(arrayMin([toInt256(0)])), toTypeName(arrayMin([toUInt256(0)]));
+SELECT toTypeName(arrayMin([toFloat32(0)])), toTypeName(arrayMin([toFloat64(0)]));
+SELECT toTypeName(arrayMin([toDecimal32(0, 8)])), toTypeName(arrayMin([toDecimal64(0, 8)])), toTypeName(arrayMin([toDecimal128(0, 8)]));
+SELECT 'Types of aggregation result array max';
+SELECT toTypeName(arrayMax([toInt8(0)])), toTypeName(arrayMax([toInt16(0)])), toTypeName(arrayMax([toInt32(0)])), toTypeName(arrayMax([toInt64(0)]));
+SELECT toTypeName(arrayMax([toUInt8(0)])), toTypeName(arrayMax([toUInt16(0)])), toTypeName(arrayMax([toUInt32(0)])), toTypeName(arrayMax([toUInt64(0)]));
+SELECT toTypeName(arrayMax([toInt128(0)])), toTypeName(arrayMax([toInt256(0)])), toTypeName(arrayMax([toUInt256(0)]));
+SELECT toTypeName(arrayMax([toFloat32(0)])), toTypeName(arrayMax([toFloat64(0)]));
+SELECT toTypeName(arrayMax([toDecimal32(0, 8)])), toTypeName(arrayMax([toDecimal64(0, 8)])), toTypeName(arrayMax([toDecimal128(0, 8)]));
+SELECT 'Types of aggregation result array summ';
+SELECT toTypeName(arraySum([toInt8(0)])), toTypeName(arraySum([toInt16(0)])), toTypeName(arraySum([toInt32(0)])), toTypeName(arraySum([toInt64(0)]));
+SELECT toTypeName(arraySum([toUInt8(0)])), toTypeName(arraySum([toUInt16(0)])), toTypeName(arraySum([toUInt32(0)])), toTypeName(arraySum([toUInt64(0)]));
+SELECT toTypeName(arraySum([toInt128(0)])), toTypeName(arraySum([toInt256(0)])), toTypeName(arraySum([toUInt256(0)]));
+SELECT toTypeName(arraySum([toFloat32(0)])), toTypeName(arraySum([toFloat64(0)]));
+SELECT toTypeName(arraySum([toDecimal32(0, 8)])), toTypeName(arraySum([toDecimal64(0, 8)])), toTypeName(arraySum([toDecimal128(0, 8)]));
+SELECT 'Types of aggregation result array avg';
+SELECT toTypeName(arrayAvg([toInt8(0)])), toTypeName(arrayAvg([toInt16(0)])), toTypeName(arrayAvg([toInt32(0)])), toTypeName(arrayAvg([toInt64(0)]));
+SELECT toTypeName(arrayAvg([toUInt8(0)])), toTypeName(arrayAvg([toUInt16(0)])), toTypeName(arrayAvg([toUInt32(0)])), toTypeName(arrayAvg([toUInt64(0)]));
+SELECT toTypeName(arrayAvg([toInt128(0)])), toTypeName(arrayAvg([toInt256(0)])), toTypeName(arrayAvg([toUInt256(0)]));
+SELECT toTypeName(arrayAvg([toFloat32(0)])), toTypeName(arrayAvg([toFloat64(0)]));
+SELECT toTypeName(arrayAvg([toDecimal32(0, 8)])), toTypeName(arrayAvg([toDecimal64(0, 8)])), toTypeName(arrayAvg([toDecimal128(0, 8)]));
diff --git a/tests/queries/0_stateless/01602_max_distributed_connections.reference b/tests/queries/0_stateless/01602_max_distributed_connections.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01602_max_distributed_connections.sh b/tests/queries/0_stateless/01602_max_distributed_connections.sh
new file mode 100755
index 00000000000..8c19b6f5bb7
--- /dev/null
+++ b/tests/queries/0_stateless/01602_max_distributed_connections.sh
@@ -0,0 +1,15 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+common_opts=(
+    "--format=Null"
+
+    "--max_threads=1"
+    "--max_distributed_connections=3"
+)
+
+# NOTE: the test use higher timeout to avoid flakiness.
+timeout 9s ${CLICKHOUSE_CLIENT} "$@" "${common_opts[@]}" -q "select sleep(3) from remote('127.{1,2,3,4,5}', system.one)" --prefer_localhost_replica=0
+timeout 9s ${CLICKHOUSE_CLIENT} "$@" "${common_opts[@]}" -q "select sleep(3) from remote('127.{1,2,3,4,5}', system.one)" --prefer_localhost_replica=1
diff --git a/tests/queries/0_stateless/01602_modified_julian_day_msan.reference b/tests/queries/0_stateless/01602_modified_julian_day_msan.reference
new file mode 100644
index 00000000000..587579af915
--- /dev/null
+++ b/tests/queries/0_stateless/01602_modified_julian_day_msan.reference
@@ -0,0 +1 @@
+Ok.
diff --git a/tests/queries/0_stateless/01602_modified_julian_day_msan.sql b/tests/queries/0_stateless/01602_modified_julian_day_msan.sql
new file mode 100644
index 00000000000..d18665f0bcf
--- /dev/null
+++ b/tests/queries/0_stateless/01602_modified_julian_day_msan.sql
@@ -0,0 +1,4 @@
+SELECT tryBase64Decode(( SELECT countSubstrings(toModifiedJulianDayOrNull('\0'), '') ) AS n, ( SELECT regionIn('l. ') ) AS srocpnuv); -- { serverError 43 }
+SELECT countSubstrings(toModifiedJulianDayOrNull('\0'), ''); -- { serverError 43 }
+SELECT countSubstrings(toInt32OrNull('123qwe123'), ''); -- { serverError 43 }
+SELECT 'Ok.';
diff --git a/tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference
new file mode 100644
index 00000000000..5b321d6c87a
--- /dev/null
+++ b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.reference
@@ -0,0 +1,10 @@
+	test_01602a	CREATE TEMPORARY TABLE test_01602a (`x` UInt32) ENGINE = Memory	Memory	Memory	1
+	test_01602b	CREATE TEMPORARY TABLE test_01602b (`y` Float64, `z` String) ENGINE = Memory	Memory	Memory	1
+	test_01602a	x	UInt32	1			0	0	0		0	0	0	0	
+	test_01602b	y	Float64	1			0	0	0		0	0	0	0	
+	test_01602b	z	String	2			0	0	0		0	0	0	0	
+CREATE TEMPORARY TABLE test_01602a\n(\n    `x` UInt32\n)\nENGINE = Memory
+CREATE TEMPORARY TABLE test_01602b\n(\n    `y` Float64,\n    `z` String\n)\nENGINE = Memory
+0
+0
+0
diff --git a/tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql
new file mode 100644
index 00000000000..4eb679d7307
--- /dev/null
+++ b/tests/queries/0_stateless/01602_temporary_table_in_system_tables.sql
@@ -0,0 +1,18 @@
+DROP TEMPORARY TABLE IF EXISTS test_01602a;
+DROP TEMPORARY TABLE IF EXISTS test_01602b;
+
+CREATE TEMPORARY TABLE test_01602a(x UInt32);
+CREATE TEMPORARY TABLE test_01602b(y Float64, z String);
+
+SELECT database, name, create_table_query, engine, engine_full, is_temporary FROM system.tables WHERE name LIKE 'test_01602%' ORDER BY name;
+SELECT * FROM system.columns WHERE table LIKE 'test_01602%' ORDER BY table, name;
+
+SHOW CREATE TEMPORARY TABLE test_01602a;
+SHOW CREATE TEMPORARY TABLE test_01602b;
+
+SELECT COUNT() FROM system.databases WHERE name='_temporary_and_external_tables';
+SELECT COUNT() FROM system.tables WHERE database='_temporary_and_external_tables';
+SELECT COUNT() FROM system.columns WHERE database='_temporary_and_external_tables';
+
+DROP TEMPORARY TABLE test_01602a;
+DROP TEMPORARY TABLE test_01602b;
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
new file mode 100644
index 00000000000..c25beb42a27
--- /dev/null
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -0,0 +1,12 @@
+2.4
+10.165
+0.00012000000000000002
+150.16500000000002
+7.775900000000001
+56622.689999999995
+5.988376688440276
+2.9941884322906844
+1.641386318314034
+1.641386318314034
+1.6413863192433382
+1.6413863192433382
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
new file mode 100644
index 00000000000..407fe3649ec
--- /dev/null
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -0,0 +1,26 @@
+SELECT toDecimal32(2, 2) * 1.2;
+SELECT toDecimal64(0.5, 2) * 20.33;
+SELECT 0.00001 * toDecimal32(12, 2);
+SELECT 30.033 * toDecimal32(5, 1);
+
+CREATE TABLE IF NOT EXISTS test01603 (
+    f64 Float64,
+    d Decimal64(3) DEFAULT toDecimal32(f64, 3),
+    f32 Float32 DEFAULT f64
+) ENGINE=MergeTree() ORDER BY f32;
+
+INSERT INTO test01603(f64) SELECT 1 / (number + 1) FROM system.numbers LIMIT 1000;
+
+SELECT sum(d * 1.1) FROM test01603;
+SELECT sum(8.01 * d) FROM test01603;
+SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
+SELECT sum(toDecimal64(40, 2) * f32) FROM test01603;
+SELECT sum(f64 * toDecimal64(0.1, 2)) FROM test01603;
+SELECT sum(toDecimal64(0.3, 2) * f32) FROM test01603;
+
+SELECT sum(f64 * d) FROM test01603;
+SELECT sum(d * f64) FROM test01603;
+SELECT sum(f32 * d) FROM test01603;
+SELECT sum(d * f32) FROM test01603;
+
+DROP TABLE IF EXISTS test01603;
diff --git a/tests/queries/0_stateless/01603_insert_select_too_many_parts.reference b/tests/queries/0_stateless/01603_insert_select_too_many_parts.reference
new file mode 100644
index 00000000000..f599e28b8ab
--- /dev/null
+++ b/tests/queries/0_stateless/01603_insert_select_too_many_parts.reference
@@ -0,0 +1 @@
+10
diff --git a/tests/queries/0_stateless/01603_insert_select_too_many_parts.sql b/tests/queries/0_stateless/01603_insert_select_too_many_parts.sql
new file mode 100644
index 00000000000..a56b680e212
--- /dev/null
+++ b/tests/queries/0_stateless/01603_insert_select_too_many_parts.sql
@@ -0,0 +1,14 @@
+DROP TABLE IF EXISTS too_many_parts;
+CREATE TABLE too_many_parts (x UInt64) ENGINE = MergeTree ORDER BY tuple() SETTINGS parts_to_delay_insert = 5, parts_to_throw_insert = 5;
+
+SYSTEM STOP MERGES too_many_parts;
+SET max_block_size = 1, min_insert_block_size_rows = 0, min_insert_block_size_bytes = 0;
+
+-- exception is not thrown if threshold is exceeded when multi-block INSERT is already started.
+INSERT INTO too_many_parts SELECT * FROM numbers(10);
+SELECT count() FROM too_many_parts;
+
+-- exception is thrown if threshold is exceeded on new INSERT.
+INSERT INTO too_many_parts SELECT * FROM numbers(10); -- { serverError 252 }
+
+DROP TABLE too_many_parts;
diff --git a/tests/queries/0_stateless/01603_remove_column_ttl.reference b/tests/queries/0_stateless/01603_remove_column_ttl.reference
new file mode 100644
index 00000000000..bb0d0b8aafe
--- /dev/null
+++ b/tests/queries/0_stateless/01603_remove_column_ttl.reference
@@ -0,0 +1,6 @@
+1	32
+2	0
+CREATE TABLE default.table_with_column_ttl\n(\n    `EventTime` DateTime,\n    `UserID` UInt64,\n    `Age` UInt8\n)\nENGINE = MergeTree\nORDER BY tuple()\nSETTINGS min_bytes_for_wide_part = 0, index_granularity = 8192
+1	32
+2	0
+3	27
diff --git a/tests/queries/0_stateless/01603_remove_column_ttl.sql b/tests/queries/0_stateless/01603_remove_column_ttl.sql
new file mode 100644
index 00000000000..556d968d270
--- /dev/null
+++ b/tests/queries/0_stateless/01603_remove_column_ttl.sql
@@ -0,0 +1,30 @@
+DROP TABLE IF EXISTS table_with_column_ttl;
+CREATE TABLE table_with_column_ttl
+(
+    EventTime DateTime,
+    UserID UInt64,
+    Age UInt8 TTL EventTime + INTERVAL 3 MONTH
+)
+ENGINE MergeTree()
+ORDER BY tuple()
+SETTINGS min_bytes_for_wide_part = 0; -- column TTL doesn't work for compact parts
+
+INSERT INTO table_with_column_ttl VALUES (now(), 1, 32);
+
+INSERT INTO table_with_column_ttl VALUES (now() - INTERVAL 4 MONTH, 2, 45);
+
+OPTIMIZE TABLE table_with_column_ttl FINAL;
+
+SELECT UserID, Age FROM table_with_column_ttl ORDER BY UserID;
+
+ALTER TABLE table_with_column_ttl MODIFY COLUMN Age REMOVE TTL;
+
+SHOW CREATE TABLE table_with_column_ttl;
+
+INSERT INTO table_with_column_ttl VALUES (now() - INTERVAL 10 MONTH, 3, 27);
+
+OPTIMIZE TABLE table_with_column_ttl FINAL;
+
+SELECT UserID, Age FROM table_with_column_ttl ORDER BY UserID;
+
+DROP TABLE table_with_column_ttl;
diff --git a/tests/queries/0_stateless/01603_rename_overwrite_bug.reference b/tests/queries/0_stateless/01603_rename_overwrite_bug.reference
new file mode 100644
index 00000000000..2c98ac3d71a
--- /dev/null
+++ b/tests/queries/0_stateless/01603_rename_overwrite_bug.reference
@@ -0,0 +1,4 @@
+0
+200
+0
+200
diff --git a/tests/queries/0_stateless/01603_rename_overwrite_bug.sql b/tests/queries/0_stateless/01603_rename_overwrite_bug.sql
new file mode 100644
index 00000000000..c857e08b243
--- /dev/null
+++ b/tests/queries/0_stateless/01603_rename_overwrite_bug.sql
@@ -0,0 +1,22 @@
+DROP database IF EXISTS test_1603_rename_bug_ordinary;
+create database test_1603_rename_bug_ordinary engine=Ordinary;
+create table test_1603_rename_bug_ordinary.foo engine=Memory as select * from numbers(100);
+create table test_1603_rename_bug_ordinary.bar engine=Log as select * from numbers(200);
+detach table test_1603_rename_bug_ordinary.foo;
+rename table test_1603_rename_bug_ordinary.bar to test_1603_rename_bug_ordinary.foo; -- { serverError 57 }
+attach table test_1603_rename_bug_ordinary.foo;
+SELECT count() from test_1603_rename_bug_ordinary.foo;
+SELECT count() from test_1603_rename_bug_ordinary.bar;
+DROP DATABASE test_1603_rename_bug_ordinary;
+
+-- was not broken, adding just in case.
+DROP database IF EXISTS test_1603_rename_bug_atomic;
+create database test_1603_rename_bug_atomic engine=Atomic;
+create table test_1603_rename_bug_atomic.foo engine=Memory as select * from numbers(100);
+create table test_1603_rename_bug_atomic.bar engine=Log as select * from numbers(200);
+detach table test_1603_rename_bug_atomic.foo;
+rename table test_1603_rename_bug_atomic.bar to test_1603_rename_bug_atomic.foo; -- { serverError 57 }
+attach table test_1603_rename_bug_atomic.foo;
+SELECT count() from test_1603_rename_bug_atomic.foo;
+SELECT count() from test_1603_rename_bug_atomic.bar;
+DROP DATABASE test_1603_rename_bug_atomic;
diff --git a/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference
new file mode 100644
index 00000000000..e674ac5fabb
--- /dev/null
+++ b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.reference
@@ -0,0 +1,8 @@
+AlterQuery  t1 (children 1)
+ ExpressionList (children 1)
+  AlterCommand 25 (children 1)
+   Function equals (children 1)
+    ExpressionList (children 2)
+     Identifier date
+     Function today (children 1)
+      ExpressionList
diff --git a/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql
new file mode 100644
index 00000000000..413acb789e0
--- /dev/null
+++ b/tests/queries/0_stateless/01604_explain_ast_of_nonselect_query.sql
@@ -0,0 +1 @@
+explain ast alter table t1 delete where date = today()
diff --git a/tests/queries/0_stateless/arcadia_skip_list.txt b/tests/queries/0_stateless/arcadia_skip_list.txt
index 172dd579481..c297698641b 100644
--- a/tests/queries/0_stateless/arcadia_skip_list.txt
+++ b/tests/queries/0_stateless/arcadia_skip_list.txt
@@ -174,5 +174,7 @@
 01584_distributed_buffer_cannot_find_column
 01018_ip_dictionary
 01582_distinct_subquery_groupby
+01558_ttest
 01558_ttest_scipy
 01561_mann_whitney_scipy
+01601_custom_tld
diff --git a/tests/queries/skip_list.json b/tests/queries/skip_list.json
index 62921b17597..f3cc522be97 100644
--- a/tests/queries/skip_list.json
+++ b/tests/queries/skip_list.json
@@ -106,5 +106,177 @@
     "polymorphic-parts": [
         "01508_partition_pruning", /// bug, shoud be fixed
         "01482_move_to_prewhere_and_cast" /// bug, shoud be fixed
+    ],
+    "parallel":
+    [
+        /// Pessimistic list of tests which work badly in parallel.
+        /// Probably they need better investigation.
+        "00062_replicated_merge_tree_alter_zookeeper",
+        "00109_shard_totals_after_having",
+        "00110_external_sort",
+        "00116_storage_set",
+        "00121_drop_column_zookeeper",
+        "00133_long_shard_memory_tracker_and_exception_safety",
+        "00180_attach_materialized_view",
+        "00226_zookeeper_deduplication_and_unexpected_parts",
+        "00236_replicated_drop_on_non_leader_zookeeper",
+        "00305_http_and_readonly",
+        "00311_array_primary_key",
+        "00417_kill_query",
+        "00423_storage_log_single_thread",
+        "00429_long_http_bufferization",
+        "00446_clear_column_in_partition_concurrent_zookeeper",
+        "00446_clear_column_in_partition_zookeeper",
+        "00463_long_sessions_in_http_interface",
+        "00505_shard_secure",
+        "00508_materialized_view_to",
+        "00516_deduplication_after_drop_partition_zookeeper",
+        "00534_functions_bad_arguments10",
+        "00552_or_nullable",
+        "00564_versioned_collapsing_merge_tree",
+        "00571_non_exist_database_when_create_materializ_view",
+        "00575_illegal_column_exception_when_drop_depen_column",
+        "00599_create_view_with_subquery",
+        "00612_http_max_query_size",
+        "00619_union_highlite",
+        "00620_optimize_on_nonleader_replica_zookeeper",
+        "00625_arrays_in_nested",
+        "00626_replace_partition_from_table",
+        "00626_replace_partition_from_table_zookeeper",
+        "00633_materialized_view_and_too_many_parts_zookeeper",
+        "00652_mergetree_mutations",
+        "00652_replicated_mutations_zookeeper",
+        "00682_empty_parts_merge",
+        "00688_low_cardinality_serialization",
+        "00693_max_block_size_system_tables_columns",
+        "00699_materialized_view_mutations",
+        "00701_rollup",
+        "00715_fetch_merged_or_mutated_part_zookeeper",
+        "00751_default_databasename_for_view",
+        "00753_alter_attach",
+        "00754_alter_modify_column_partitions",
+        "00754_alter_modify_order_by_replicated_zookeeper",
+        "00763_long_lock_buffer_alter_destination_table",
+        "00804_test_alter_compression_codecs",
+        "00804_test_custom_compression_codecs",
+        "00804_test_custom_compression_codes_log_storages",
+        "00804_test_delta_codec_compression",
+        "00834_cancel_http_readonly_queries_on_client_close",
+        "00834_kill_mutation",
+        "00834_kill_mutation_replicated_zookeeper",
+        "00840_long_concurrent_select_and_drop_deadlock",
+        "00899_long_attach_memory_limit",
+        "00910_zookeeper_custom_compression_codecs_replicated",
+        "00926_adaptive_index_granularity_merge_tree",
+        "00926_adaptive_index_granularity_pk",
+        "00926_adaptive_index_granularity_replacing_merge_tree",
+        "00926_zookeeper_adaptive_index_granularity_replicated_merge_tree",
+        "00933_alter_ttl",
+        "00933_reserved_word",
+        "00933_test_fix_extra_seek_on_compressed_cache",
+        "00933_ttl_replicated_zookeeper",
+        "00933_ttl_with_default",
+        "00955_test_final_mark",
+        "00976_ttl_with_old_parts",
+        "00980_merge_alter_settings",
+        "00980_zookeeper_merge_tree_alter_settings",
+        "00988_constraints_replication_zookeeper",
+        "00989_parallel_parts_loading",
+        "00993_system_parts_race_condition_drop_zookeeper",
+        "01013_sync_replica_timeout_zookeeper",
+        "01014_lazy_database_concurrent_recreate_reattach_and_show_tables",
+        "01015_attach_part",
+        "01018_ddl_dictionaries_concurrent_requrests",
+        "01018_ddl_dictionaries_create",
+        "01018_ddl_dictionaries_select",
+        "01021_only_tuple_columns",
+        "01031_mutations_interpreter_and_context",
+        "01033_dictionaries_lifetime",
+        "01035_concurrent_move_partition_from_table_zookeeper",
+        "01045_zookeeper_system_mutations_with_parts_names",
+        "01053_ssd_dictionary",
+        "01055_compact_parts_1",
+        "01060_avro",
+        "01060_shutdown_table_after_detach",
+        "01070_materialize_ttl",
+        "01070_modify_ttl",
+        "01070_mutations_with_dependencies",
+        "01071_live_view_detach_dependency",
+        "01071_prohibition_secondary_index_with_old_format_merge_tree",
+        "01073_attach_if_not_exists",
+        "01076_parallel_alter_replicated_zookeeper",
+        "01079_parallel_alter_add_drop_column_zookeeper",
+        "01079_parallel_alter_detach_table_zookeeper",
+        "01083_expressions_in_engine_arguments",
+        "01085_max_distributed_connections_http",
+        "01092_memory_profiler",
+        "01098_temporary_and_external_tables",
+        "01107_atomic_db_detach_attach",
+        "01108_restart_replicas_rename_deadlock_zookeeper",
+        "01110_dictionary_layout_without_arguments",
+        "01114_database_atomic",
+        "01127_month_partitioning_consistency_select",
+        "01130_in_memory_parts_partitons",
+        "01135_default_and_alter_zookeeper",
+        "01148_zookeeper_path_macros_unfolding",
+        "01190_full_attach_syntax",
+        "01193_metadata_loading",
+        "01200_mutations_memory_consumption",
+        "01238_http_memory_tracking",
+        "01249_bad_arguments_for_bloom_filter",
+        "01251_dict_is_in_infinite_loop",
+        "01254_dict_load_after_detach_attach",
+        "01259_dictionary_custom_settings_ddl",
+        "01267_alter_default_key_columns_zookeeper",
+        "01268_dictionary_direct_layout",
+        "01269_alias_type_differs",
+        "01272_suspicious_codecs",
+        "01277_alter_rename_column_constraint_zookeeper",
+        "01280_ssd_complex_key_dictionary",
+        "01280_ttl_where_group_by",
+        "01281_group_by_limit_memory_tracking",
+        "01281_unsucceeded_insert_select_queries_counter",
+        "01293_system_distribution_queue",
+        "01294_lazy_database_concurrent",
+        "01294_lazy_database_concurrent_recreate_reattach_and_show_tables",
+        "01305_replica_create_drop_zookeeper",
+        "01307_multiple_leaders_zookeeper",
+        "01318_long_unsuccessful_mutation_zookeeper",
+        "01319_manual_write_to_replicas",
+        "01338_long_select_and_alter",
+        "01338_long_select_and_alter_zookeeper",
+        "01355_alter_column_with_order",
+        "01355_ilike",
+        "01357_version_collapsing_attach_detach_zookeeper",
+        "01375_compact_parts_codecs",
+        "01378_alter_rename_with_ttl_zookeeper",
+        "01388_clear_all_columns",
+        "01396_inactive_replica_cleanup_nodes_zookeeper",
+        "01412_cache_dictionary_race",
+        "01414_mutations_and_errors_zookeeper",
+        "01415_inconsistent_merge_tree_settings",
+        "01415_sticking_mutations",
+        "01417_freeze_partition_verbose",
+        "01417_freeze_partition_verbose_zookeeper",
+        "01430_modify_sample_by_zookeeper",
+        "01454_storagememory_data_race_challenge",
+        "01456_modify_column_type_via_add_drop_update",
+        "01457_create_as_table_function_structure",
+        "01459_manual_write_to_replicas",
+        "01460_DistributedFilesToInsert",
+        "01465_ttl_recompression",
+        "01471_calculate_ttl_during_merge",
+        "01493_alter_remove_properties_zookeeper",
+        "01493_storage_set_persistency",
+        "01494_storage_join_persistency",
+        "01516_drop_table_stress",
+        "01541_max_memory_usage_for_user",
+        "attach",
+        "ddl_dictionaries",
+        "dictionary",
+        "limit_memory",
+        "live_view",
+        "memory_leak",
+        "memory_limit"
     ]
 }
diff --git a/tests/testflows/rbac/tests/privileges/create/create_table.py b/tests/testflows/rbac/tests/privileges/create/create_table.py
index 3985679a08a..c39b0c9a5e4 100644
--- a/tests/testflows/rbac/tests/privileges/create/create_table.py
+++ b/tests/testflows/rbac/tests/privileges/create/create_table.py
@@ -706,6 +706,9 @@ def create_as_merge(self, node=None):
                 with When("I grant CREATE TABLE privilege to a user"):
                     node.query(f"GRANT CREATE TABLE ON {table_name} TO {user_name}")
 
+                with And("I grant SELECT privilege to a user to allow executing the table function merge()"):
+                    node.query(f"GRANT SELECT ON {source_table_name} TO {user_name}")
+
                 with Then("I try to create a table as another table"):
                     node.query(f"CREATE TABLE {table_name} AS merge(default,'{source_table_name}')", settings = [("user", f"{user_name}")])
 
diff --git a/utils/convert-month-partitioned-parts/main.cpp b/utils/convert-month-partitioned-parts/main.cpp
index af8e221a10b..bce1e08077c 100644
--- a/utils/convert-month-partitioned-parts/main.cpp
+++ b/utils/convert-month-partitioned-parts/main.cpp
@@ -1,4 +1,5 @@
 #include <DataTypes/DataTypeDate.h>
+#include <DataTypes/DataTypesNumber.h>
 #include <IO/HashingWriteBuffer.h>
 #include <IO/ReadBufferFromFile.h>
 #include <IO/WriteBufferFromFile.h>
diff --git a/utils/db-generator/query_db_generator.cpp b/utils/db-generator/query_db_generator.cpp
index fd310fe1897..b6e287278ff 100644
--- a/utils/db-generator/query_db_generator.cpp
+++ b/utils/db-generator/query_db_generator.cpp
@@ -1255,10 +1255,10 @@ void parseSelectQuery(DB::ASTPtr ast, TableList & all_tables)
 
 TableList getTablesFromSelect(std::vector<std::string> queries)
 {
-    DB::ParserQueryWithOutput parser;
     TableList result;
     for (std::string & query : queries)
     {
+        DB::ParserQueryWithOutput parser(query.data() + query.size());
         DB::ASTPtr ast = parseQuery(parser, query.data(), query.data() + query.size(), "", 0, 0);
         for (auto & select : getSelect(ast))
         {
diff --git a/utils/iotest/iotest.cpp b/utils/iotest/iotest.cpp
index e578a539bcd..5b28dbc95f9 100644
--- a/utils/iotest/iotest.cpp
+++ b/utils/iotest/iotest.cpp
@@ -6,6 +6,7 @@
 #include <Common/Stopwatch.h>
 #include <Common/ThreadPool.h>
 #include <Common/randomSeed.h>
+#include <common/getPageSize.h>
 
 #include <cstdlib>
 #include <iomanip>
@@ -46,7 +47,7 @@ void thread(int fd, int mode, size_t min_offset, size_t max_offset, size_t block
 {
     using namespace DB;
 
-    Memory<> direct_buf(block_size, sysconf(_SC_PAGESIZE));
+    Memory<> direct_buf(block_size, ::getPageSize());
     std::vector<char> simple_buf(block_size);
 
     char * buf;
diff --git a/utils/iotest/iotest_aio.cpp b/utils/iotest/iotest_aio.cpp
index 24508c1dd9f..66de4713893 100644
--- a/utils/iotest/iotest_aio.cpp
+++ b/utils/iotest/iotest_aio.cpp
@@ -14,6 +14,7 @@ int main(int, char **) { return 0; }
 #include <Common/ThreadPool.h>
 #include <Common/Stopwatch.h>
 #include <Common/randomSeed.h>
+#include <common/getPageSize.h>
 #include <pcg_random.hpp>
 #include <IO/BufferWithOwnMemory.h>
 #include <IO/ReadHelpers.h>
@@ -52,7 +53,7 @@ void thread(int fd, int mode, size_t min_offset, size_t max_offset, size_t block
 
     std::vector<Memory<>> buffers(buffers_count);
     for (size_t i = 0; i < buffers_count; ++i)
-        buffers[i] = Memory<>(block_size, sysconf(_SC_PAGESIZE));
+        buffers[i] = Memory<>(block_size, ::getPageSize());
 
     pcg64_fast rng(randomSeed());
 
diff --git a/utils/make_changelog.py b/utils/make_changelog.py
deleted file mode 100755
index 4f703108d38..00000000000
--- a/utils/make_changelog.py
+++ /dev/null
@@ -1,507 +0,0 @@
-#!/usr/bin/env python3
-# Note: should work with python 2 and 3
-
-
-import requests
-import json
-import subprocess
-import re
-import os
-import time
-import logging
-import codecs
-import argparse
-
-
-GITHUB_API_URL = 'https://api.github.com/'
-SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
-
-
-def http_get_json(url, token, max_retries, retry_timeout):
-
-    for t in range(max_retries):
-
-        if token:
-            resp = requests.get(url, headers={"Authorization": "token {}".format(token)})
-        else:
-            resp = requests.get(url)
-
-        if resp.status_code != 200:
-            msg = "Request {} failed with code {}.\n{}\n".format(url, resp.status_code, resp.text)
-
-            if resp.status_code == 403 or resp.status_code >= 500:
-                try:
-                    if (resp.json()['message'].startswith('API rate limit exceeded') or resp.status_code >= 500) and t + 1 < max_retries:
-                        logging.warning(msg)
-                        time.sleep(retry_timeout)
-                        continue
-                except Exception:
-                    pass
-
-            raise Exception(msg)
-
-        return resp.json()
-
-
-def github_api_get_json(query, token, max_retries, retry_timeout):
-    return http_get_json(GITHUB_API_URL + query, token, max_retries, retry_timeout)
-
-
-def check_sha(sha):
-    if not (re.match('^[a-hA-H0-9]+$', sha) and len(sha) >= 7):
-        raise Exception("String " + sha + " doesn't look like git sha.")
-
-
-def get_merge_base(first, second, project_root):
-    try:
-        command = "git merge-base {} {}".format(first, second)
-        text = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, cwd=project_root).stdout.read()
-        text = text.decode('utf-8', 'ignore')
-        sha = tuple(filter(len, text.split()))[0]
-        check_sha(sha)
-        return sha
-    except Exception:
-        logging.error('Cannot find merge base for %s and %s', first, second)
-        raise
-
-def rev_parse(rev, project_root):
-    try:
-        command = "git rev-parse {}".format(rev)
-        text = subprocess.check_output(command, shell=True, cwd=project_root)
-        text = text.decode('utf-8', 'ignore')
-        sha = tuple(filter(len, text.split()))[0]
-        check_sha(sha)
-        return sha
-    except Exception:
-        logging.error('Cannot find revision %s', rev)
-        raise
-
-
-# Get list of commits from branch to base_sha. Update commits_info.
-def get_commits_from_branch(repo, branch, base_sha, commits_info, max_pages, token, max_retries, retry_timeout):
-
-    def get_commits_from_page(page):
-        query = 'repos/{}/commits?sha={}&page={}'.format(repo, branch, page)
-        resp = github_api_get_json(query, token, max_retries, retry_timeout)
-        for commit in resp:
-            sha = commit['sha']
-            if sha not in commits_info:
-                commits_info[sha] = commit
-
-        return [commit['sha'] for commit in resp]
-
-    commits = []
-    found_base_commit = False
-
-    for page in range(max_pages):
-        page_commits = get_commits_from_page(page)
-        for commit in page_commits:
-            if commit == base_sha:
-                found_base_commit = True
-                break
-            commits.append(commit)
-
-        if found_base_commit:
-            break
-
-    if not found_base_commit:
-        raise Exception("Can't found base commit sha {} in branch {}. Checked {} commits on {} pages.\nCommits: {}"
-                        .format(base_sha, branch, len(commits), max_pages, ' '.join(commits)))
-    return commits
-
-
-# Get list of commits a specified commit is cherry-picked from. Can return an empty list.
-def parse_original_commits_from_cherry_pick_message(commit_message):
-    prefix = '(cherry picked from commits'
-    pos = commit_message.find(prefix)
-    if pos == -1:
-        prefix = '(cherry picked from commit'
-        pos = commit_message.find(prefix)
-        if pos == -1:
-            return []
-    pos += len(prefix)
-    endpos = commit_message.find(')', pos)
-    if endpos == -1:
-        return []
-    lst = [x.strip() for x in commit_message[pos:endpos].split(',')]
-    lst = [x for x in lst if x]
-    return lst
-
-
-# Use GitHub search api to check if commit from any pull request. Update pull_requests info.
-def find_pull_request_for_commit(commit_info, pull_requests, token, max_retries, retry_timeout):
-    commits = [commit_info['sha']] + parse_original_commits_from_cherry_pick_message(commit_info['commit']['message'])
-
-    # Special case for cherry-picked merge commits without -x option. Parse pr number from commit message and search it.
-    if commit_info['commit']['message'].startswith('Merge pull request'):
-        tokens = commit_info['commit']['message'][len('Merge pull request'):].split()
-        if len(tokens) > 0 and tokens[0].startswith('#'):
-            pr_number = tokens[0][1:]
-            if len(pr_number) > 0 and pr_number.isdigit():
-                commits = [pr_number]
-
-    query = 'search/issues?q={}+type:pr+repo:{}&sort=created&order=asc'.format(' '.join(commits), repo)
-    resp = github_api_get_json(query, token, max_retries, retry_timeout)
-
-    found = False
-    for item in resp['items']:
-        if 'pull_request' in item:
-            found = True
-            number = item['number']
-            if number not in pull_requests:
-                pull_requests[number] = {
-                    'title': item['title'],
-                    'description': item['body'],
-                    'user': item['user']['login'],
-                }
-
-    return found
-
-
-# Find pull requests from list of commits. If no pull request found, add commit to not_found_commits list.
-def find_pull_requests(commits, commits_info, token, max_retries, retry_timeout):
-    not_found_commits = []
-    pull_requests = {}
-
-    for i, commit in enumerate(commits):
-        if (i + 1) % 10 == 0:
-            logging.info('Processed %d commits', i + 1)
-        if not find_pull_request_for_commit(commits_info[commit], pull_requests, token, max_retries, retry_timeout):
-            not_found_commits.append(commit)
-
-    return not_found_commits, pull_requests
-
-
-# Find pull requests by list of numbers
-def find_pull_requests_by_num(pull_requests_nums, token, max_retries, retry_timeout):
-    pull_requests = {}
-
-    for pr in pull_requests_nums:
-        item = github_api_get_json('repos/{}/pulls/{}'.format(repo, pr), token, max_retries, retry_timeout)
-
-        number = item['number']
-        if number not in pull_requests:
-            pull_requests[number] = {
-                'title': item['title'],
-                'description': item['body'],
-                'user': item['user']['login'],
-            }
-
-    return pull_requests
-
-
-# Get users for all unknown commits and pull requests.
-def get_users_info(pull_requests, commits_info, token, max_retries, retry_timeout):
-
-    users = {}
-
-    def update_user(user):
-        if user not in users:
-            query = 'users/{}'.format(user)
-            resp = github_api_get_json(query, token, max_retries, retry_timeout)
-            users[user] = resp
-
-    for pull_request in list(pull_requests.values()):
-        update_user(pull_request['user'])
-
-    for commit_info in list(commits_info.values()):
-        if 'committer' in commit_info and commit_info['committer'] is not None and 'login' in commit_info['committer']:
-            update_user(commit_info['committer']['login'])
-        else:
-            logging.warning('Not found author for commit %s.', commit_info['html_url'])
-
-    return users
-
-
-# List of unknown commits -> text description.
-def process_unknown_commits(commits, commits_info, users):
-
-    pattern = 'Commit: [{}]({})\nAuthor: {}\nMessage: {}'
-
-    texts = []
-
-    for commit in commits:
-        info = commits_info[commit]
-        html_url = info['html_url']
-        msg = info['commit']['message']
-
-        name = None
-        login = None
-        author = None
-
-        if not info['author']:
-            author = 'Unknown'
-        else:
-            # GitHub login
-            if 'login' in info['author']:
-                login = info['author']['login']
-
-                # First, try get name from github user
-                try:
-                    name = users[login]['name']
-                except KeyError:
-                    pass
-            else:
-                login = 'Unknown'
-
-            # Then, try get name from commit
-            if not name:
-                try:
-                    name = info['commit']['author']['name']
-                except KeyError:
-                    pass
-
-            author = '[{}]({})'.format(name or login, info['author']['html_url'])
-
-        texts.append(pattern.format(commit, html_url, author, msg))
-
-    text = 'Commits which are not from any pull request:\n\n'
-    return text + '\n\n'.join(texts)
-
-# This function mirrors the PR description checks in ClickhousePullRequestTrigger.
-# Returns False if the PR should not be mentioned changelog.
-def parse_one_pull_request(item):
-    description = item['description']
-    lines = [line for line in [x.strip() for x in description.split('\n')] if line] if description else []
-    lines = [re.sub(r'\s+', ' ', l) for l in lines]
-
-    cat_pos = None
-    short_descr_pos = None
-    long_descr_pos = None
-
-    if lines:
-        for i in range(len(lines) - 1):
-            if re.match(r'(?i).*category.*:$', lines[i]):
-                cat_pos = i
-            if re.match(r'(?i)^\**\s*(Short description|Change\s*log entry)', lines[i]):
-                short_descr_pos = i
-            if re.match(r'(?i)^\**\s*Detailed description', lines[i]):
-                long_descr_pos = i
-
-    if cat_pos is None:
-        return False
-    cat = lines[cat_pos + 1]
-    cat = re.sub(r'^[-*\s]*', '', cat)
-
-    # Filter out the PR categories that are not for changelog.
-    if re.match(r'(?i)doc|((non|in|not|un)[-\s]*significant)|(not[ ]*for[ ]*changelog)', cat):
-        return False
-
-    short_descr = ''
-    if short_descr_pos:
-        short_descr_end = long_descr_pos or len(lines)
-        short_descr = lines[short_descr_pos + 1]
-        if short_descr_pos + 2 != short_descr_end:
-            short_descr += ' ...'
-
-    # If we have nothing meaningful
-    if not re.match('\w', short_descr):
-        short_descr = item['title']
-
-    # TODO: Add detailed description somewhere
-
-    item['entry'] = short_descr
-    item['category'] = cat
-
-    return True
-
-
-# List of pull requests -> text description.
-def process_pull_requests(pull_requests, users, repo):
-    groups = {}
-
-    for id, item in list(pull_requests.items()):
-        if not parse_one_pull_request(item):
-            continue
-
-        pattern = "{} [#{}]({}) ({})"
-        link = 'https://github.com/{}/pull/{}'.format(repo, id)
-        author = 'author not found'
-        if item['user'] in users:
-            # TODO get user from any commit if no user name on github
-            user = users[item['user']]
-            author = '[{}]({})'.format(user['name'] or user['login'], user['html_url'])
-
-        cat = item['category']
-        if cat not in groups:
-            groups[cat] = []
-        groups[cat].append(pattern.format(item['entry'], id, link, author))
-
-    categories_preferred_order = ['Backward Incompatible Change', 'New Feature', 'Bug Fix', 'Improvement', 'Performance Improvement', 'Build/Testing/Packaging Improvement', 'Other']
-
-    def categories_sort_key(name):
-        if name in categories_preferred_order:
-            return str(categories_preferred_order.index(name)).zfill(3)
-        else:
-            return name.lower()
-
-    texts = []
-    for group, text in sorted(list(groups.items()), key = lambda kv: categories_sort_key(kv[0])):
-        items = ['* {}'.format(pr) for pr in text]
-        texts.append('### {}\n{}'.format(group if group else '[No category]', '\n'.join(items)))
-
-    return '\n\n'.join(texts)
-
-
-# Load inner state. For debug purposes.
-def load_state(state_file, base_sha, new_tag, prev_tag):
-
-    state = {}
-
-    if state_file:
-        try:
-            if os.path.exists(state_file):
-                logging.info('Reading state from %s', state_file)
-                with codecs.open(state_file, encoding='utf-8') as f:
-                    state = json.loads(f.read())
-            else:
-                logging.info('State file does not exist. Will create new one.')
-        except Exception as e:
-            logging.warning('Cannot load state from %s. Reason: %s', state_file, str(e))
-
-    if state:
-        if 'base_sha' not in state or 'new_tag' not in state or 'prev_tag' not in state:
-            logging.warning('Invalid state. Will create new one.')
-        elif state['base_sha'] == base_sha and state['new_tag'] == new_tag and state['prev_tag'] == prev_tag:
-            logging.info('State loaded.')
-        else:
-            logging.info('Loaded state has different tags or merge base sha. Will create new state.')
-            state = {}
-
-    return state
-
-
-# Save inner state. For debug purposes.
-def save_state(state_file, state):
-    with codecs.open(state_file, 'w', encoding='utf-8') as f:
-        f.write(json.dumps(state, indent=4, separators=(',', ': ')))
-
-
-def make_changelog(new_tag, prev_tag, pull_requests_nums, repo, repo_folder, state_file, token, max_retries, retry_timeout):
-
-    base_sha = None
-    if new_tag and prev_tag:
-        base_sha = get_merge_base(new_tag, prev_tag, repo_folder)
-        logging.info('Base sha: %s', base_sha)
-
-    # Step 1. Get commits from merge_base to new_tag HEAD.
-    # Result is a list of commits + map with commits info (author, message)
-    commits_info = {}
-    commits = []
-    is_commits_loaded = False
-
-    # Step 2. For each commit check if it is from any pull request (using github search api).
-    # Result is a list of unknown commits + map with pull request info (author, description).
-    unknown_commits = []
-    pull_requests = {}
-    is_pull_requests_loaded = False
-
-    # Step 3. Map users with their info (Name)
-    users = {}
-    is_users_loaded = False
-
-    # Step 4. Make changelog text from data above.
-
-    state = load_state(state_file, base_sha, new_tag, prev_tag)
-
-    if state:
-
-        if 'commits' in state and 'commits_info' in state:
-            logging.info('Loading commits from %s', state_file)
-            commits_info = state['commits_info']
-            commits = state['commits']
-            is_commits_loaded = True
-
-        if 'pull_requests' in state and 'unknown_commits' in state:
-            logging.info('Loading pull requests from %s', state_file)
-            unknown_commits = state['unknown_commits']
-            pull_requests = state['pull_requests']
-            is_pull_requests_loaded = True
-
-        if 'users' in state:
-            logging.info('Loading users requests from %s', state_file)
-            users = state['users']
-            is_users_loaded = True
-
-    if base_sha:
-        state['base_sha'] = base_sha
-        state['new_tag'] = new_tag
-        state['prev_tag'] = prev_tag
-
-        if not is_commits_loaded:
-            logging.info('Getting commits using github api.')
-            commits = get_commits_from_branch(repo, new_tag, base_sha, commits_info, 100, token, max_retries, retry_timeout)
-            state['commits'] = commits
-            state['commits_info'] = commits_info
-
-        logging.info('Found %d commits from %s to %s.\n', len(commits), new_tag, base_sha)
-        save_state(state_file, state)
-
-        if not is_pull_requests_loaded:
-            logging.info('Searching for pull requests using github api.')
-            unknown_commits, pull_requests = find_pull_requests(commits, commits_info, token, max_retries, retry_timeout)
-            state['unknown_commits'] = unknown_commits
-            state['pull_requests'] = pull_requests
-    else:
-        pull_requests = find_pull_requests_by_num(pull_requests_nums.split(','), token, max_retries, retry_timeout)
-
-    logging.info('Found %d pull requests and %d unknown commits.\n', len(pull_requests), len(unknown_commits))
-    save_state(state_file, state)
-
-    if not is_users_loaded:
-        logging.info('Getting users info using github api.')
-        users = get_users_info(pull_requests, commits_info, token, max_retries, retry_timeout)
-        state['users'] = users
-
-    logging.info('Found %d users.', len(users))
-    save_state(state_file, state)
-
-    changelog = '{}\n\n{}'.format(process_pull_requests(pull_requests, users, repo), process_unknown_commits(unknown_commits, commits_info, users))
-
-    # Substitute links to issues
-    changelog = re.sub(r'(?<!\[)#(\d{4,})(?!\])', r'[#\1](https://github.com/{}/issues/\1)'.format(repo), changelog)
-
-    # Remove double whitespaces and trailing whitespaces
-    changelog = re.sub(r' {2,}| +$', r''.format(repo), changelog)
-
-    print(changelog.encode('utf-8'))
-
-
-if __name__ == '__main__':
-
-    parser = argparse.ArgumentParser(description='Make changelog.')
-    parser.add_argument('prev_release_tag', nargs='?', help='Git tag from previous release.')
-    parser.add_argument('new_release_tag', nargs='?', help='Git tag for new release.')
-    parser.add_argument('--pull-requests', help='Process the specified list of pull-request numbers (comma separated) instead of commits between tags.')
-    parser.add_argument('--token', help='Github token. Use it to increase github api query limit.')
-    parser.add_argument('--directory', help='ClickHouse repo directory. Script dir by default.')
-    parser.add_argument('--state', help='File to dump inner states result.', default='changelog_state.json')
-    parser.add_argument('--repo', help='ClickHouse repo on GitHub.', default='ClickHouse/ClickHouse')
-    parser.add_argument('--max_retry', default=100, type=int,
-                        help='Max number of retries pre api query in case of API rate limit exceeded error.')
-    parser.add_argument('--retry_timeout', help='Timeout after retry in seconds.', type=int, default=5)
-
-    args = parser.parse_args()
-    prev_release_tag = args.prev_release_tag
-    new_release_tag = args.new_release_tag
-    token = args.token or ''
-    repo_folder = args.directory or SCRIPT_DIR
-    state_file = args.state
-    repo = args.repo
-    max_retry = args.max_retry
-    retry_timeout = args.retry_timeout
-    pull_requests = args.pull_requests
-
-    if (not prev_release_tag or not new_release_tag) and not pull_requests:
-        raise Exception('Either release tags or --pull-requests must be specified')
-
-    if prev_release_tag and new_release_tag and pull_requests:
-        raise Exception('Either release tags or --pull-requests must be specified')
-
-    logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s')
-
-    repo_folder = os.path.expanduser(repo_folder)
-    new_release_tag = rev_parse(new_release_tag, repo_folder)
-    prev_release_tag = rev_parse(prev_release_tag, repo_folder)
-
-    make_changelog(new_release_tag, prev_release_tag, pull_requests, repo, repo_folder, state_file, token, max_retry, retry_timeout)
diff --git a/utils/zookeeper-test/main.cpp b/utils/zookeeper-test/main.cpp
index fe7bf93fd15..8f8aac00866 100644
--- a/utils/zookeeper-test/main.cpp
+++ b/utils/zookeeper-test/main.cpp
@@ -6,10 +6,15 @@
 #include <common/LineReader.h>
 #include <common/logger_useful.h>
 #include <fmt/format.h>
+#include <random>
+#include <iterator>
+#include <algorithm>
+#include <chrono>
 
 #include <iostream>
 #include <sstream>
 #include <exception>
+#include <future>
 
 using namespace std;
 
@@ -144,9 +149,98 @@ void testMultiRequest(zkutil::ZooKeeper & zk)
     checkEq(zk, "/data/multirequest", "bbb");
 }
 
+std::mutex elements_mutex;
+std::vector<int> current_elements;
+std::atomic<int> watches_triggered = 0;
+
+void triggerWatch(const Coordination::WatchResponse &)
+{
+    watches_triggered++;
+}
+
+template<typename Iter, typename RandomGenerator>
+Iter select_randomly(Iter start, Iter end, RandomGenerator& g)
+{
+    std::uniform_int_distribution<> dis(0, std::distance(start, end) - 1);
+    std::advance(start, dis(g));
+    return start;
+}
+
+template<typename Iter>
+Iter select_randomly(Iter start, Iter end)
+{
+    static std::random_device rd;
+    static std::mt19937 gen(rd());
+    return select_randomly(start, end, gen);
+}
+
+std::atomic<int> element_counter = 0;
+std::atomic<int> failed_setup_counter = 0;
+
+void createPathAndSetWatch(zkutil::ZooKeeper & zk, const String & path_prefix, size_t total)
+{
+    for (size_t i = 0; i < total; ++i)
+    {
+        int element = element_counter++;
+        zk.createIfNotExists(path_prefix + "/" + std::to_string(element), "");
+
+        std::string result;
+        if (!zk.tryGetWatch(path_prefix + "/" + std::to_string(element), result, nullptr, triggerWatch))
+            failed_setup_counter++;
+
+        {
+            std::lock_guard lock(elements_mutex);
+            current_elements.push_back(element);
+        }
+
+        std::this_thread::sleep_for(std::chrono::milliseconds(200));
+
+        {
+            std::lock_guard lock(elements_mutex);
+            if (current_elements.empty())
+                continue;
+            element = *select_randomly(current_elements.begin(), current_elements.end());
+            current_elements.erase(std::remove(current_elements.begin(), current_elements.end(), element), current_elements.end());
+        }
+        zk.tryRemove(path_prefix + "/" + std::to_string(element));
+    }
+
+}
+
+void tryConcurrentWatches(zkutil::ZooKeeper & zk)
+{
+    std::string path_prefix = "/concurrent_watches";
+    std::vector<std::future<void>> asyncs;
+    zk.createIfNotExists(path_prefix, "");
+    for (size_t i = 0; i < 100; ++i)
+    {
+        auto callback = [&zk, path_prefix] ()
+        {
+            createPathAndSetWatch(zk, path_prefix, 100);
+        };
+        asyncs.push_back(std::async(std::launch::async, callback));
+    }
+
+    for (auto & async : asyncs)
+    {
+        async.wait();
+    }
+
+    size_t counter = 0;
+    while (watches_triggered != 100 * 100)
+    {
+        std::this_thread::sleep_for(std::chrono::milliseconds(100));
+        if (counter++ > 20)
+            break;
+    }
+
+    std::cerr << "Failed setup counter:" << failed_setup_counter << std::endl;
+    std::cerr << "Current elements size:" << current_elements.size() << std::endl;
+    std::cerr << "WatchesTriggered:" << watches_triggered << std::endl;
+}
+
 int main(int argc, char *argv[])
 {
-
     if (argc != 2)
     {
         std::cerr << "usage: " << argv[0] << " hosts" << std::endl;
@@ -168,6 +262,7 @@ int main(int argc, char *argv[])
         testMultiRequest(zk);
         testCreateSetWatchEvent(zk);
         testCreateListWatchEvent(zk);
+        tryConcurrentWatches(zk);
     }
     catch (...)
     {
diff --git a/website/benchmark/hardware/index.html b/website/benchmark/hardware/index.html
index 41a87a4e49f..7fbb3b6d505 100644
--- a/website/benchmark/hardware/index.html
+++ b/website/benchmark/hardware/index.html
@@ -71,7 +71,8 @@ Results for Dell XPS laptop and Google Pixel phone is from <b>Alexander Kuzmenko
 Results for Android phones for "cold cache" are done without cache flushing, so they are not "cold" and cannot be compared.<br/>
 Results for Digital Ocean are from <b>Zimin Aleksey</b>.<br/>
 Results for 2x EPYC 7642 w/ 512 GB RAM (192 Cores) + 12X 1TB SSD (RAID6) are from <b>Yiğit Konur</b> and <b>Metehan Çetinkaya</b> of seo.do.<br/>
-Results for Raspberry Pi and Digital Ocean CPU-optimized are from <b>Fritz Wijaya</b>.
+Results for Raspberry Pi and Digital Ocean CPU-optimized are from <b>Fritz Wijaya</b>.<br/>
+Results for Digitalocean (Storage-intesinve VMs) + (CPU/GP) are from <b>Yiğit Konur</b> and <b>Metehan Çetinkaya</b> of seo.do.
 </p>
     </div>
 </div>
diff --git a/website/benchmark/hardware/results/do_storage_optimized.json b/website/benchmark/hardware/results/do_storage_optimized.json
new file mode 100644
index 00000000000..6c6cee5423b
--- /dev/null
+++ b/website/benchmark/hardware/results/do_storage_optimized.json
@@ -0,0 +1,380 @@
+[
+    {
+        "system":       "DigitalOcean Storage-opt 8",
+        "system_full":  "DigitalOcean 8 CPUs, 64 GB RAM, 1.17 TB NVM SSD, Storage Intensive, Frankfurt, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.003, 0.002, 0.002],
+[0.020, 0.014, 0.014],
+[0.060, 0.048, 0.049],
+[0.089, 0.064, 0.062],
+[0.194, 0.177, 0.178],
+[0.521, 0.502, 0.497],
+[0.031, 0.026, 0.026],
+[0.020, 0.018, 0.018],
+[0.782, 0.738, 0.730],
+[0.874, 0.830, 0.834],
+[0.288, 0.265, 0.261],
+[0.340, 0.316, 0.316],
+[1.144, 1.016, 1.000],
+[1.423, 1.359, 1.348],
+[1.296, 1.234, 1.217],
+[1.458, 1.393, 1.392],
+[3.573, 3.374, 3.335],
+[2.143, 2.065, 2.083],
+[6.179, 6.153, 6.110],
+[0.101, 0.069, 0.069],
+[1.611, 1.383, 1.362],
+[1.659, 1.409, 1.388],
+[3.642, 3.124, 3.104],
+[2.399, 1.690, 1.652],
+[0.501, 0.434, 0.412],
+[0.382, 0.345, 0.346],
+[0.490, 0.425, 0.420],
+[1.401, 1.192, 1.165],
+[2.132, 1.956, 1.979],
+[1.777, 1.775, 1.770],
+[1.134, 1.033, 1.043],
+[1.611, 1.475, 1.451],
+[9.319, 8.830, 8.826],
+[5.214, 5.071, 4.960],
+[5.464, 5.079, 5.113],
+[1.862, 1.810, 1.800],
+[0.223, 0.173, 0.167],
+[0.074, 0.058, 0.060],
+[0.039, 0.037, 0.038],
+[0.430, 0.354, 0.355],
+[0.034, 0.015, 0.013],
+[0.016, 0.010, 0.011],
+[0.006, 0.006, 0.006]
+        ]
+    },
+    {
+        "system":       "DigitalOcean Storage-opt 16",
+        "system_full":  "DigitalOcean 16 CPUs, 128 GB RAM, 2.34 TB NVM SSD, Storage Intensive, Frankfurt, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.002, 0.002, 0.002],
+[0.017, 0.010, 0.010],
+[0.039, 0.029, 0.027],
+[0.055, 0.039, 0.038],
+[0.125, 0.107, 0.112],
+[0.302, 0.281, 0.271],
+[0.021, 0.016, 0.016],
+[0.012, 0.011, 0.011],
+[0.418, 0.398, 0.397],
+[0.473, 0.446, 0.442],
+[0.172, 0.153, 0.155],
+[0.202, 0.184, 0.185],
+[0.605, 0.543, 0.538],
+[0.753, 0.693, 0.682],
+[0.687, 0.645, 0.635],
+[0.708, 0.701, 0.693],
+[1.703, 1.642, 1.655],
+[1.037, 0.994, 0.997],
+[3.563, 3.314, 3.364],
+[0.085, 0.047, 0.039],
+[0.857, 0.745, 0.742],
+[0.900, 0.770, 0.752],
+[1.931, 1.679, 1.672],
+[1.537, 0.937, 0.914],
+[0.272, 0.230, 0.219],
+[0.212, 0.192, 0.187],
+[0.271, 0.237, 0.225],
+[0.782, 0.659, 0.661],
+[1.169, 1.063, 1.068],
+[1.193, 1.182, 1.180],
+[0.625, 0.590, 0.572],
+[0.936, 0.822, 0.800],
+[4.796, 4.587, 4.589],
+[3.061, 2.715, 2.700],
+[2.819, 2.709, 2.708],
+[1.078, 1.048, 1.064],
+[0.210, 0.163, 0.150],
+[0.073, 0.057, 0.058],
+[0.037, 0.036, 0.035],
+[0.387, 0.357, 0.324],
+[0.030, 0.015, 0.017],
+[0.014, 0.010, 0.015],
+[0.012, 0.006, 0.005]
+        ]
+    },
+    {
+        "system":       "DigitalOcean Storage-opt 24",
+        "system_full":  "DigitalOcean, 24 CPUs, 192 GB RAM, 3.52 TB NVM SSD, Storage Intensive, Frankfurt, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.002, 0.002, 0.001],
+[0.017, 0.009, 0.009],
+[0.033, 0.022, 0.022],
+[0.046, 0.029, 0.028],
+[0.114, 0.101, 0.113],
+[0.228, 0.215, 0.209],
+[0.018, 0.013, 0.013],
+[0.012, 0.010, 0.010],
+[0.316, 0.294, 0.294],
+[0.350, 0.328, 0.330],
+[0.142, 0.129, 0.124],
+[0.157, 0.141, 0.138],
+[0.452, 0.401, 0.403],
+[0.550, 0.508, 0.503],
+[0.883, 0.455, 0.451],
+[0.539, 0.535, 0.533],
+[1.278, 1.206, 1.207],
+[0.771, 0.746, 0.736],
+[2.776, 2.543, 2.571],
+[0.084, 0.065, 0.029],
+[0.666, 0.539, 0.586],
+[0.703, 0.559, 0.566],
+[1.404, 1.232, 1.222],
+[1.243, 0.695, 0.701],
+[0.215, 0.173, 0.164],
+[0.162, 0.143, 0.140],
+[0.207, 0.169, 0.169],
+[0.606, 0.506, 0.505],
+[0.849, 0.788, 0.774],
+[1.085, 1.074, 1.072],
+[0.426, 0.392, 0.394],
+[0.718, 0.616, 0.600],
+[3.579, 3.372, 3.428],
+[2.161, 2.006, 1.980],
+[2.491, 2.026, 1.991],
+[0.839, 0.818, 0.817],
+[0.220, 0.154, 0.163],
+[0.074, 0.058, 0.063],
+[0.037, 0.038, 0.038],
+[0.391, 0.353, 0.329],
+[0.034, 0.012, 0.014],
+[0.014, 0.012, 0.011],
+[0.005, 0.005, 0.005]
+        ]
+    },
+    {
+        "system":       "DigitalOcean Storage-opt 32",
+        "system_full":  "DigitalOcean, 32 CPUs, 256 GB RAM, 4.56 TB NVM SSD, Storage Intensive, Frankfurt, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.002, 0.002, 0.001],
+[0.016, 0.011, 0.010],
+[0.029, 0.020, 0.019],
+[0.044, 0.025, 0.025],
+[0.103, 0.094, 0.091],
+[0.202, 0.181, 0.182],
+[0.021, 0.014, 0.012],
+[0.012, 0.010, 0.010],
+[0.259, 0.240, 0.237],
+[0.291, 0.268, 0.267],
+[0.128, 0.114, 0.113],
+[0.137, 0.121, 0.120],
+[0.351, 0.315, 0.315],
+[0.435, 0.403, 0.401],
+[0.389, 0.377, 0.370],
+[0.473, 0.459, 0.450],
+[1.076, 1.019, 1.034],
+[0.673, 0.632, 0.653],
+[1.804, 1.773, 1.779],
+[0.055, 0.048, 0.030],
+[0.555, 0.472, 0.462],
+[0.575, 0.465, 0.478],
+[1.245, 1.087, 1.073],
+[1.206, 0.561, 0.572],
+[0.178, 0.137, 0.137],
+[0.133, 0.116, 0.116],
+[0.168, 0.139, 0.133],
+[0.533, 0.434, 0.440],
+[0.727, 0.676, 0.668],
+[1.037, 1.029, 1.029],
+[0.343, 0.308, 0.308],
+[0.503, 0.448, 0.449],
+[3.128, 2.888, 2.937],
+[1.772, 1.605, 1.586],
+[1.653, 1.591, 1.606],
+[0.754, 0.718, 0.703],
+[0.198, 0.156, 0.143],
+[0.074, 0.056, 0.057],
+[0.040, 0.040, 0.035],
+[0.395, 0.333, 0.320],
+[0.031, 0.017, 0.014],
+[0.013, 0.011, 0.017],
+[0.008, 0.009, 0.006]
+        ]
+    },
+    {
+        "system":       "DigitalOcean Memory-opt 32",
+        "system_full":  "DigitalOcean, 32 CPUs, 256 GB RAM, 4.8 TB SSD, Memory Intensive, Frankfurt, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.003, 0.002, 0.002],
+[0.016, 0.009, 0.009],
+[0.030, 0.019, 0.019],
+[0.045, 0.026, 0.025],
+[0.103, 0.091, 0.090],
+[0.200, 0.180, 0.175],
+[0.017, 0.012, 0.012],
+[0.012, 0.011, 0.010],
+[0.264, 0.237, 0.234],
+[0.291, 0.270, 0.259],
+[0.124, 0.111, 0.111],
+[0.136, 0.118, 0.117],
+[0.352, 0.316, 0.303],
+[0.439, 0.394, 0.392],
+[0.391, 0.370, 0.372],
+[0.466, 0.450, 0.449],
+[1.075, 1.014, 1.008],
+[0.654, 0.628, 0.619],
+[1.810, 1.801, 1.787],
+[0.053, 0.032, 0.030],
+[0.550, 0.443, 0.461],
+[0.553, 0.461, 0.465],
+[1.237, 1.064, 1.065],
+[1.163, 0.589, 0.548],
+[0.178, 0.132, 0.135],
+[0.131, 0.114, 0.112],
+[0.164, 0.133, 0.132],
+[0.530, 0.442, 0.433],
+[0.720, 0.664, 0.659],
+[1.032, 1.024, 1.023],
+[0.346, 0.311, 0.314],
+[0.502, 0.439, 0.438],
+[3.146, 2.869, 2.891],
+[1.757, 1.615, 1.589],
+[1.642, 1.571, 1.580],
+[0.743, 0.711, 0.722],
+[0.208, 0.151, 0.140],
+[0.077, 0.057, 0.056],
+[0.037, 0.037, 0.040],
+[0.378, 0.310, 0.315],
+[0.031, 0.014, 0.017],
+[0.019, 0.010, 0.010],
+[0.005, 0.005, 0.005]
+        ]
+    },
+    {
+        "system":       "DigitalOcean CPU-opt, 32",
+        "system_full":  "DigitalOcean, 32 CPUs, 64 GB RAM, 3.52 TB SSD, CPU Intensive, Frankfurt, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.002, 0.002, 0.001],
+[0.016, 0.009, 0.009],
+[0.026, 0.018, 0.018],
+[0.040, 0.025, 0.024],
+[0.097, 0.083, 0.088],
+[0.190, 0.167, 0.171],
+[0.020, 0.011, 0.011],
+[0.011, 0.009, 0.009],
+[0.253, 0.228, 0.231],
+[0.277, 0.257, 0.258],
+[0.117, 0.104, 0.107],
+[0.127, 0.111, 0.111],
+[0.346, 0.313, 0.310],
+[0.435, 0.396, 0.392],
+[0.391, 0.366, 0.361],
+[0.469, 0.458, 0.455],
+[1.070, 1.026, 1.033],
+[0.649, 0.626, 0.627],
+[1.859, 1.808, 1.803],
+[0.065, 0.032, 0.025],
+[0.513, 0.402, 0.420],
+[0.525, 0.418, 0.417],
+[1.167, 1.004, 0.988],
+[1.118, 0.531, 0.538],
+[0.171, 0.126, 0.122],
+[0.124, 0.108, 0.105],
+[0.158, 0.126, 0.124],
+[0.501, 0.407, 0.409],
+[0.691, 0.624, 0.619],
+[1.106, 1.095, 1.094],
+[0.337, 0.301, 0.302],
+[0.504, 0.447, 0.453],
+[3.187, 2.934, 2.963],
+[1.771, 1.583, 1.570],
+[1.661, 1.584, 1.573],
+[0.748, 0.733, 0.736],
+[0.198, 0.152, 0.145],
+[0.072, 0.055, 0.053],
+[0.039, 0.035, 0.035],
+[0.395, 0.332, 0.322],
+[0.032, 0.012, 0.012],
+[0.019, 0.014, 0.009],
+[0.005, 0.005, 0.005]
+        ]
+    },
+    {
+        "system":       "DigitalOcean General 40",
+        "system_full":  "DigitalOcean, 40 CPUs, 64 GB RAM, 1 TB SSD, General Purpose, San Francisco, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz",
+        "cpu_vendor":   "Intel",
+        "cpu_model":    "Xeon Gold 6140",
+        "time":         "2020-12-13 00:00:00",
+        "kind":         "cloud",
+        "result":
+        [
+[0.003, 0.001, 0.001],
+[0.019, 0.009, 0.009],
+[0.045, 0.017, 0.017],
+[0.103, 0.022, 0.021],
+[0.131, 0.086, 0.088],
+[0.210, 0.167, 0.163],
+[0.018, 0.011, 0.011],
+[0.014, 0.010, 0.010],
+[0.246, 0.204, 0.204],
+[0.297, 0.231, 0.238],
+[0.138, 0.103, 0.102],
+[0.148, 0.106, 0.107],
+[0.352, 0.313, 0.309],
+[0.453, 0.394, 0.392],
+[0.392, 0.349, 0.349],
+[0.452, 0.429, 0.429],
+[1.060, 0.972, 0.966],
+[0.668, 0.591, 0.606],
+[1.721, 1.645, 1.640],
+[0.102, 0.039, 0.024],
+[0.973, 0.386, 0.381],
+[1.105, 0.405, 0.388],
+[1.996, 0.986, 0.969],
+[2.381, 0.475, 0.463],
+[0.294, 0.112, 0.111],
+[0.167, 0.096, 0.099],
+[0.296, 0.114, 0.114],
+[0.961, 0.432, 0.424],
+[0.840, 0.611, 0.610],
+[1.100, 1.090, 1.088],
+[0.374, 0.292, 0.283],
+[0.691, 0.432, 0.429],
+[3.111, 2.863, 2.922],
+[1.857, 1.579, 1.553],
+[1.762, 1.548, 1.542],
+[0.573, 0.552, 0.542],
+[0.221, 0.140, 0.144],
+[0.079, 0.059, 0.058],
+[0.038, 0.034, 0.035],
+[0.405, 0.342, 0.354],
+[0.039, 0.018, 0.013],
+[0.015, 0.010, 0.010],
+[0.005, 0.005, 0.005]
+        ]
+    }
+]
diff --git a/website/benchmark/hardware/results/core_i7_macbook_pro_2018.json b/website/benchmark/hardware/results/macbook_pro_core_i7_2018.json
similarity index 100%
rename from website/benchmark/hardware/results/core_i7_macbook_pro_2018.json
rename to website/benchmark/hardware/results/macbook_pro_core_i7_2018.json
diff --git a/website/benchmark/hardware/results/macbook_pro_core_i7_2020.json b/website/benchmark/hardware/results/macbook_pro_core_i7_2020.json
new file mode 100644
index 00000000000..8250c5f3c7e
--- /dev/null
+++ b/website/benchmark/hardware/results/macbook_pro_core_i7_2020.json
@@ -0,0 +1,54 @@
+[
+    {
+        "system":       "MacBook Pro 2020",
+        "system_full":  "MacBook Pro 2020, 2.3 GHz Quad-Core Intel Core i7 (10th gen), 32 GiB RAM, 2TB SSD",
+        "time":         "2020-06-25 00:00:00",
+        "kind":         "laptop",
+        "result":
+        [
+[0.003, 0.002, 0.002],
+[0.046, 0.025, 0.015],
+[0.042, 0.042, 0.041],
+[0.069, 0.068, 0.069],
+[0.164, 0.166, 0.165],
+[0.559, 0.550, 0.552],
+[0.026, 0.026, 0.026],
+[0.016, 0.015, 0.016],
+[0.891, 0.880, 0.879],
+[1.029, 1.031, 1.022],
+[0.355, 0.358, 0.369],
+[0.415, 0.416, 0.417],
+[1.406, 1.355, 1.349],
+[1.775, 1.787, 1.788],
+[1.615, 1.633, 1.619],
+[1.532, 1.525, 1.482],
+[4.284, 4.203, 4.180],
+[2.825, 2.782, 2.756],
+[8.516, 8.328, 8.408],
+[0.097, 0.073, 0.074],
+[1.541, 1.557, 1.557],
+[1.945, 1.920, 1.911],
+[4.410, 4.375, 4.353],
+[null, null, null],
+[0.613, 0.616, 0.621],
+[0.535, 0.557, 0.541],
+[0.622, 0.622, 0.623],
+[1.536, 1.548, 1.535],
+[2.436, 2.430, 2.435],
+[2.551, 2.542, 2.589],
+[1.470, 1.426, 1.430],
+[2.377, 2.248, 2.227],
+[15.628, null, null],
+[6.155, 6.397, 6.368],
+[6.439, 6.392, 6.412],
+[2.643, 2.337, 2.316],
+[0.163, 0.174, 0.155],
+[0.060, 0.064, 0.063],
+[0.057, 0.057, 0.053],
+[0.331, 0.314, 0.314],
+[0.016, 0.023, 0.020],
+[0.014, 0.014, 0.013],
+[0.005, 0.006, 0.007]
+        ]
+    }
+]

From 259ef9ecf4cc6e0b6206c8d4a2dd53970c6d7be2 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 18 Dec 2020 19:00:51 +0300
Subject: [PATCH 200/504] corrected the test

---
 .../0_stateless/01603_decimal_mult_float.reference        | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
index c25beb42a27..2c567fac2cd 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -3,9 +3,11 @@
 0.00012000000000000002
 150.16500000000002
 7.775900000000001
-56622.689999999995
-5.988376688440276
-2.9941884322906844
+56.62268999999999
+598.8376688440276
+299.41883813298773
+0.7485470860550346
+2.2456413519648777
 1.641386318314034
 1.641386318314034
 1.6413863192433382

From 132a0b42eb010344563daf79f74c51a00c02475a Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 18 Dec 2020 20:13:28 +0300
Subject: [PATCH 201/504] cleanup

---
 src/Interpreters/AggregateDescription.cpp     | 28 ++++++++++++
 src/Interpreters/AggregateDescription.h       | 15 ++++++-
 src/Interpreters/ExpressionAnalyzer.cpp       | 45 ++++---------------
 .../ExtractExpressionInfoVisitor.cpp          |  3 ++
 .../RewriteAnyFunctionVisitor.cpp             |  9 +++-
 src/Interpreters/TreeRewriter.cpp             |  4 +-
 src/Parsers/ASTFunction.cpp                   |  8 ----
 src/Parsers/ExpressionElementParsers.cpp      |  4 +-
 src/Parsers/ExpressionElementParsers.h        |  6 +--
 src/Processors/QueryPlan/WindowStep.cpp       | 45 ++++++++++++++++---
 src/Processors/QueryPlan/WindowStep.h         |  4 +-
 .../Transforms/ExpressionTransform.h          |  5 +--
 src/Processors/Transforms/WindowTransform.cpp | 17 ++-----
 src/Processors/Transforms/WindowTransform.h   | 13 +++---
 src/Processors/ya.make                        |  2 +
 src/Storages/IStorage.cpp                     | 18 ++++----
 .../01591_window_functions.reference          |  3 ++
 .../0_stateless/01591_window_functions.sql    |  3 ++
 18 files changed, 138 insertions(+), 94 deletions(-)

diff --git a/src/Interpreters/AggregateDescription.cpp b/src/Interpreters/AggregateDescription.cpp
index e483eb1b7a1..985789ca676 100644
--- a/src/Interpreters/AggregateDescription.cpp
+++ b/src/Interpreters/AggregateDescription.cpp
@@ -1,6 +1,7 @@
 #include <Interpreters/AggregateDescription.h>
 #include <Common/FieldVisitors.h>
 #include <IO/Operators.h>
+#include <Parsers/ASTFunction.h>
 
 namespace DB
 {
@@ -99,4 +100,31 @@ void AggregateDescription::explain(WriteBuffer & out, size_t indent) const
     }
 }
 
+std::string WindowFunctionDescription::dump() const
+{
+    WriteBufferFromOwnString ss;
+
+    ss << "window function '" << column_name << "' over '" << window_name <<"\n";
+    ss << "function node " << function_node->dumpTree() << "\n";
+    ss << "aggregate function '" << aggregate_function->getName() << "'\n";
+    if (!function_parameters.empty())
+    {
+        ss << "parameters " << toString(function_parameters) << "\n";
+    }
+
+    return ss.str();
+}
+
+std::string WindowDescription::dump() const
+{
+    WriteBufferFromOwnString ss;
+
+    ss << "window '" << window_name << "'\n";
+    ss << "partition_by " << dumpSortDescription(partition_by) << "\n";
+    ss << "order_by " << dumpSortDescription(order_by) << "\n";
+    ss << "full_sort_description " << dumpSortDescription(full_sort_description) << "\n";
+
+    return ss.str();
+}
+
 }
diff --git a/src/Interpreters/AggregateDescription.h b/src/Interpreters/AggregateDescription.h
index 61e684a1458..3cbba101baa 100644
--- a/src/Interpreters/AggregateDescription.h
+++ b/src/Interpreters/AggregateDescription.h
@@ -29,7 +29,12 @@ using AggregateDescriptions = std::vector<AggregateDescription>;
 
 struct WindowFunctionDescription
 {
+    // According to the standard, a window function can refer to a window declared
+    // elsewhere, using its name. We haven't implemented this yet: all windows
+    // are declared in OVER clause, but they still get an auto-generated name,
+    // and this name is used to find the window that corresponds to a function.
     std::string window_name;
+
     std::string column_name;
     const ASTFunction * function_node;
     AggregateFunctionPtr aggregate_function;
@@ -43,10 +48,18 @@ struct WindowFunctionDescription
 struct WindowDescription
 {
     std::string window_name;
-    // Always ASC for now.
+
+    // We don't care about the particular order of keys for PARTITION BY, only
+    // that they are sorted. For now we always require ASC, but we could be more
+    // flexible and match any direction, or even different order of columns.
     SortDescription partition_by;
+
     SortDescription order_by;
+
+    // To calculate the window function, we sort input data first by PARTITION BY,
+    // then by ORDER BY. This field holds this combined sort order.
     SortDescription full_sort_description;
+
     // No frame info as of yet.
 
     std::string dump() const;
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 61b8dc86350..af6ee4a1732 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1,7 +1,5 @@
 #include <Core/Block.h>
 
-#include <iostream>
-
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ASTIdentifier.h>
 #include <Parsers/ASTLiteral.h>
@@ -52,6 +50,9 @@
 #include <Interpreters/GlobalSubqueriesVisitor.h>
 #include <Interpreters/GetAggregatesVisitor.h>
 
+#include <IO/Operators.h>
+#include <IO/WriteBufferFromString.h>
+
 namespace DB
 {
 
@@ -473,7 +474,7 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
 bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 {
     // Convenient to check here because at least we have the Context.
-    if (windowFunctions().size() > 0 &&
+    if (!windowFunctions().empty() &&
         !context.getSettingsRef().allow_experimental_window_functions)
     {
         throw Exception(ErrorCodes::NOT_IMPLEMENTED,
@@ -481,15 +482,14 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
             windowFunctions()[0]->formatForErrorMessage());
     }
 
+    int next_window_index = 1;
     for (const ASTFunction * function_node : windowFunctions())
     {
         assert(function_node->is_window_function);
 
-        // FIXME not thread-safe, should use a per-query counter.
-        static int window_index = 1;
-
         WindowDescription window_description;
-        window_description.window_name = fmt::format("window_{}", window_index++);
+        window_description.window_name = fmt::format("window_{}",
+            next_window_index++);
 
         if (function_node->window_partition_by)
         {
@@ -548,6 +548,7 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 
         // Requiring a constant reference to a shared pointer to non-const AST
         // doesn't really look sane, but the visitor does indeed require it.
+        // Hence we clone the node (not very sane either, I know).
         getRootActionsNoMakeSet(window_function.function_node->clone(),
             true, actions);
 
@@ -572,7 +573,6 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
             window_function.argument_names[i] = name;
         }
 
-
         AggregateFunctionProperties properties;
         window_function.aggregate_function
             = AggregateFunctionFactory::instance().get(
@@ -583,8 +583,6 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
         window_descriptions.insert({window_description.window_name,
             window_description});
         window_functions.push_back(window_function);
-
-        fmt::print(stderr, "{}\n", window_function.dump());
     }
 
     return !windowFunctions().empty();
@@ -1502,7 +1500,7 @@ void ExpressionAnalysisResult::checkActions() const
 
 std::string ExpressionAnalysisResult::dump() const
 {
-    std::stringstream ss;
+    WriteBufferFromOwnString ss;
 
     ss << "need_aggregate " << need_aggregate << "\n";
     ss << "has_order_by " << has_order_by << "\n";
@@ -1571,29 +1569,4 @@ std::string ExpressionAnalysisResult::dump() const
     return ss.str();
 }
 
-std::string WindowFunctionDescription::dump() const
-{
-    std::stringstream ss;
-    ss << "window function '" << column_name << "' over '" << window_name <<"\n";
-    ss << "function node " << function_node->dumpTree() << "\n";
-    ss << "aggregate function '" << aggregate_function->getName() << "'\n";
-    if (function_parameters.size())
-    {
-        ss << "parameters " << toString(function_parameters) << "\n";
-    }
-
-    return ss.str();
-}
-
-std::string WindowDescription::dump() const
-{
-    std::stringstream ss;
-    ss << "window '" << window_name << "'\n";
-    ss << "partition_by " << dumpSortDescription(partition_by) << "\n";
-    ss << "order_by " << dumpSortDescription(order_by) << "\n";
-    ss << "full_sort_description " << dumpSortDescription(full_sort_description) << "\n";
-
-    return ss.str();
-}
-
 }
diff --git a/src/Interpreters/ExtractExpressionInfoVisitor.cpp b/src/Interpreters/ExtractExpressionInfoVisitor.cpp
index 579fd13642f..c7bd433a45b 100644
--- a/src/Interpreters/ExtractExpressionInfoVisitor.cpp
+++ b/src/Interpreters/ExtractExpressionInfoVisitor.cpp
@@ -22,6 +22,9 @@ void ExpressionInfoMatcher::visit(const ASTFunction & ast_function, const ASTPtr
     {
         data.is_array_join = true;
     }
+    // "is_aggregate_function" doesn't mean much by itself. Apparently here it is
+    // used to move filters from HAVING to WHERE, and probably for this purpose
+    // an aggregate function calculated as a window function is not relevant.
     else if (!ast_function.is_window_function
         && AggregateFunctionFactory::instance().isAggregateFunctionName(
             ast_function.name))
diff --git a/src/Interpreters/RewriteAnyFunctionVisitor.cpp b/src/Interpreters/RewriteAnyFunctionVisitor.cpp
index 3e5d0b693cf..ed3663a3b86 100644
--- a/src/Interpreters/RewriteAnyFunctionVisitor.cpp
+++ b/src/Interpreters/RewriteAnyFunctionVisitor.cpp
@@ -38,11 +38,16 @@ bool extractIdentifiers(const ASTFunction & func, std::unordered_set<ASTPtr *> &
             if (arg_func->name == "lambda")
                 return false;
 
-            if ( arg_func->is_window_function
+            // We are looking for identifiers inside a function calculated inside
+            // the aggregate function `any()`. Window or aggregate function can't
+            // be inside `any` (checked by GetAggregatesMatcher).
+            if (arg_func->is_window_function
                 || AggregateFunctionFactory::instance().isAggregateFunctionName(
                     arg_func->name))
             {
-                return false;
+                throw Exception(ErrorCodes::LOGICAL_ERROR,
+                    "Found an aggregate function '{}' inside another aggregate"
+                    " function", arg_func->formatForErrorMessage());
             }
 
             if (!extractIdentifiers(*arg_func, identifiers))
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index befae896d2f..f72d36400ef 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -434,7 +434,7 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
             for (auto & arg : node->arguments->children)
             {
                 assertNoAggregates(arg, "inside another aggregate function");
-                assertNoWindows(arg, "inside another window function");
+                assertNoWindows(arg, "inside an aggregate function");
             }
         }
     }
@@ -459,7 +459,7 @@ std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSel
         {
             for (auto & arg : node->arguments->children)
             {
-                //assertNoAggregates(arg, "inside another aggregate function");
+                assertNoAggregates(arg, "inside a window function");
                 assertNoWindows(arg, "inside another window function");
             }
         }
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index e2939b2bd89..569c710db03 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -39,14 +39,6 @@ void ASTFunction::appendColumnNameImpl(WriteBuffer & ostr) const
             (*it)->appendColumnName(ostr);
         }
     writeChar(')', ostr);
-
-    /*
-    if (window_name)
-    {
-        writeCString(" over ", ostr);
-        window_name->appendColumnName(ostr);
-    }
-    */
 }
 
 /** Get the text that identifies this element. */
diff --git a/src/Parsers/ExpressionElementParsers.cpp b/src/Parsers/ExpressionElementParsers.cpp
index cd9d8e921ef..fde0e129459 100644
--- a/src/Parsers/ExpressionElementParsers.cpp
+++ b/src/Parsers/ExpressionElementParsers.cpp
@@ -388,7 +388,9 @@ bool ParserFunction::parseImpl(Pos & pos, ASTPtr & node, Expected & expected)
         function_node->is_window_function = true;
 
         // We are slightly breaking the parser interface by parsing the window
-        // definition into an existing ASTFunction, hence this weirdness.
+        // definition into an existing ASTFunction. Normally it would take a
+        // reference to ASTPtr and assign it the new node. We only have a pointer
+        // of a different type, hence this workaround with a temporary pointer.
         ASTPtr function_node_as_iast = function_node;
 
         ParserWindowDefinition window_definition;
diff --git a/src/Parsers/ExpressionElementParsers.h b/src/Parsers/ExpressionElementParsers.h
index 0c15eeb08e6..e91bc21979a 100644
--- a/src/Parsers/ExpressionElementParsers.h
+++ b/src/Parsers/ExpressionElementParsers.h
@@ -124,13 +124,11 @@ protected:
     bool allow_function_parameters;
 };
 
-struct ParserWindowDefinition : public IParserBase
+// Window definition (the thing that goes after OVER) for window function.
+class ParserWindowDefinition : public IParserBase
 {
-
-protected:
     const char * getName() const override { return "window definition"; }
     bool parseImpl(Pos & pos, ASTPtr & node, Expected & expected) override;
-
 };
 
 class ParserCodecDeclarationList : public IParserBase
diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
index 8b235eca39e..82c589b8b20 100644
--- a/src/Processors/QueryPlan/WindowStep.cpp
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -25,7 +25,7 @@ static ITransformingStep::Traits getTraits()
     };
 }
 
-static Block addWindowFunctionColumns(const Block & block,
+static Block addWindowFunctionResultColumns(const Block & block,
     std::vector<WindowFunctionDescription> window_functions)
 {
     auto result = block;
@@ -48,21 +48,22 @@ WindowStep::WindowStep(const DataStream & input_stream_,
         const std::vector<WindowFunctionDescription> & window_functions_)
     : ITransformingStep(
         input_stream_,
-            addWindowFunctionColumns(input_stream_.header, window_functions_),
+            addWindowFunctionResultColumns(input_stream_.header,
+                window_functions_),
         getTraits())
     , window_description(window_description_)
     , window_functions(window_functions_)
     , input_header(input_stream_.header)
 {
-    /// Some columns may be removed by expression.
-    updateDistinctColumns(output_stream->header, output_stream->distinct_columns);
+    // We don't remove any columns, only add, so probably we don't have to update
+    // the output DataStream::distinct_columns.
 }
 
 void WindowStep::transformPipeline(QueryPipeline & pipeline)
 {
     pipeline.addSimpleTransform([&](const Block & /*header*/)
     {
-        return std::make_shared<Transform>(input_header,
+        return std::make_shared<WindowTransform>(input_header,
             output_stream->header, window_description, window_functions);
     });
 
@@ -73,8 +74,38 @@ void WindowStep::transformPipeline(QueryPipeline & pipeline)
 void WindowStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
-    (void) prefix;
-    /// FIXME add some printing
+    settings.out << prefix << "Window: (";
+    if (!window_description.partition_by.empty())
+    {
+        settings.out << "PARTITION BY ";
+        for (size_t i = 0; i < window_description.partition_by.size(); ++i)
+        {
+            if (i > 0)
+            {
+                settings.out << ", ";
+            }
+
+            settings.out << window_description.partition_by[i].column_name;
+        }
+    }
+    if (!window_description.partition_by.empty()
+        && !window_description.order_by.empty())
+    {
+        settings.out << " ";
+    }
+    if (!window_description.order_by.empty())
+    {
+        settings.out << "ORDER BY "
+            << dumpSortDescription(window_description.order_by);
+    }
+    settings.out << ")\n";
+
+    for (size_t i = 0; i < window_functions.size(); ++i)
+    {
+        settings.out << prefix << (i == 0 ? "Functions: "
+                                          : "           ");
+        settings.out << window_functions[i].column_name << "\n";
+    }
 }
 
 }
diff --git a/src/Processors/QueryPlan/WindowStep.h b/src/Processors/QueryPlan/WindowStep.h
index 5bdb14ae5fe..069d02c655c 100644
--- a/src/Processors/QueryPlan/WindowStep.h
+++ b/src/Processors/QueryPlan/WindowStep.h
@@ -14,13 +14,11 @@ class WindowTransform;
 class WindowStep : public ITransformingStep
 {
 public:
-    using Transform = WindowTransform;
-
     explicit WindowStep(const DataStream & input_stream_,
             const WindowDescription & window_description_,
             const std::vector<WindowFunctionDescription> & window_functions_);
 
-    String getName() const override { return "Expression"; }
+    String getName() const override { return "Window"; }
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
diff --git a/src/Processors/Transforms/ExpressionTransform.h b/src/Processors/Transforms/ExpressionTransform.h
index c0c0a63a6a8..525273bad81 100644
--- a/src/Processors/Transforms/ExpressionTransform.h
+++ b/src/Processors/Transforms/ExpressionTransform.h
@@ -19,10 +19,7 @@ public:
             const Block & header_,
             ExpressionActionsPtr expression_);
 
-    String getName() const override
-    {
-        return "ExpressionTransform";
-    }
+    String getName() const override { return "ExpressionTransform"; }
 
     static Block transformHeader(Block header, const ExpressionActionsPtr & expression);
 
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index 07eb8118a3b..9ffad3ae367 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -88,7 +88,6 @@ void WindowTransform::transform(Chunk & chunk)
         const auto & f = ws.window_function;
         const auto * a = f.aggregate_function.get();
 
-        //*
         // Create the resulting column.
         ColumnWithTypeAndName column_with_type;
         column_with_type.name = f.column_name;
@@ -98,6 +97,9 @@ void WindowTransform::transform(Chunk & chunk)
 
         for (size_t row = 0; row < num_rows; row++)
         {
+            // THIS IS BROKEN when we have multiple window functions.
+            // Have to switch the loops or track partition per-function.
+            //
             // Check whether the new partition has started and reinitialize the
             // aggregate function states.
             assert(partition_start_columns.size() == partition_by_indices.size());
@@ -120,7 +122,7 @@ void WindowTransform::transform(Chunk & chunk)
                 // Check whether the new partition started, by comparing all the
                 // PARTITION BY columns.
                 size_t first_inequal_column = 0;
-                for ( ; first_inequal_column < partition_start_columns.size();
+                for (; first_inequal_column < partition_start_columns.size();
                       ++first_inequal_column)
                 {
                     const auto * current_column = block.getColumns()[
@@ -165,17 +167,6 @@ void WindowTransform::transform(Chunk & chunk)
         }
 
         block.insert(column_with_type);
-        /*/
-        auto & column_with_type = block.getByName(f.column_name);
-        auto c = IColumn::mutate(std::move(column_with_type.column));
-
-        for (size_t i = 0; i < num_rows; i++)
-        {
-            c->insert(UInt64(i));
-        }
-
-        column_with_type.column.reset(c.get());
-        //*/
     }
 
     chunk.setColumns(block.getColumns(), num_rows);
diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
index 5e1444b9949..93af208ed4f 100644
--- a/src/Processors/Transforms/WindowTransform.h
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -13,6 +13,7 @@ using ExpressionActionsPtr = std::shared_ptr<ExpressionActions>;
 
 class Arena;
 
+// Runtime data for computing one window function
 struct WindowFunctionWorkspace
 {
     WindowFunctionDescription window_function;
@@ -22,11 +23,10 @@ struct WindowFunctionWorkspace
     std::vector<const IColumn *> argument_columns;
 };
 
-/** Executes a certain expression over the block.
-  * The expression consists of column identifiers from the block, constants, common functions.
-  * For example: hits * 2 + 3, url LIKE '%yandex%'
-  * The expression processes each row independently of the others.
-  */
+/*
+ * Computes several window functions that share the same window. The input must
+ * be sorted correctly for this window (PARTITION BY, then ORDER BY).
+ */
 class WindowTransform : public ISimpleTransform
 {
 public:
@@ -51,8 +51,10 @@ public:
     Block input_header;
 
     WindowDescription window_description;
+
     // Indices of the PARTITION BY columns in block.
     std::vector<size_t> partition_by_indices;
+
     // The columns for PARTITION BY and the row in these columns where the
     // current partition started. They might be in some of the previous blocks,
     // so we have to keep the shared ownership of the columns. We don't keep the
@@ -63,6 +65,7 @@ public:
     std::vector<ColumnPtr> partition_start_columns;
     size_t partition_start_row = 0;
 
+    // Data for computing the window functions.
     std::vector<WindowFunctionWorkspace> workspaces;
 
     std::unique_ptr<Arena> arena;
diff --git a/src/Processors/ya.make b/src/Processors/ya.make
index eacc0c44d24..e382f32bfdb 100644
--- a/src/Processors/ya.make
+++ b/src/Processors/ya.make
@@ -123,6 +123,7 @@ SRCS(
     QueryPlan/SettingQuotaAndLimitsStep.cpp
     QueryPlan/TotalsHavingStep.cpp
     QueryPlan/UnionStep.cpp
+    QueryPlan/WindowStep.cpp
     ResizeProcessor.cpp
     Sources/DelayedSource.cpp
     Sources/RemoteSource.cpp
@@ -155,6 +156,7 @@ SRCS(
     Transforms/RollupTransform.cpp
     Transforms/SortingTransform.cpp
     Transforms/TotalsHavingTransform.cpp
+    Transforms/WindowTransform.cpp
     printPipeline.cpp
 
 )
diff --git a/src/Storages/IStorage.cpp b/src/Storages/IStorage.cpp
index 76c975c4654..0bd576a858b 100644
--- a/src/Storages/IStorage.cpp
+++ b/src/Storages/IStorage.cpp
@@ -3,16 +3,18 @@
 #include <sparsehash/dense_hash_map>
 #include <sparsehash/dense_hash_set>
 
-#include <Storages/AlterCommands.h>
+#include <Common/StringUtils/StringUtils.h>
+#include <Common/quoteString.h>
+#include <IO/Operators.h>
+#include <IO/WriteBufferFromString.h>
+#include <Interpreters/Context.h>
+#include <Interpreters/ExpressionActions.h>
+#include <Interpreters/InterpreterSelectQuery.h>
 #include <Parsers/ASTCreateQuery.h>
 #include <Parsers/ASTSetQuery.h>
 #include <Processors/Pipe.h>
 #include <Processors/QueryPlan/ReadFromPreparedSource.h>
-#include <Interpreters/Context.h>
-#include <Common/StringUtils/StringUtils.h>
-#include <Common/quoteString.h>
-#include <Interpreters/ExpressionActions.h>
-#include <Interpreters/InterpreterSelectQuery.h>
+#include <Storages/AlterCommands.h>
 
 
 namespace DB
@@ -170,7 +172,7 @@ NamesAndTypesList IStorage::getVirtuals() const
 
 std::string PrewhereDAGInfo::dump() const
 {
-    std::stringstream ss;
+    WriteBufferFromOwnString ss;
     ss << "PrewhereDagInfo\n";
 
     if (alias_actions)
@@ -196,7 +198,7 @@ std::string PrewhereDAGInfo::dump() const
 
 std::string FilterInfo::dump() const
 {
-    std::stringstream ss;
+    WriteBufferFromOwnString ss;
     ss << "FilterInfo for column '" << column_name <<"', do_remove_column "
        << do_remove_column << "\n";
     if (actions_dag)
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 67d37653965..427d7a24e3a 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -83,6 +83,9 @@ select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 
 select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
+-- this one doesn't work yet -- looks like the column names clash, and the
+-- window count() is overwritten with aggregate count()
+-- select number, count(), count() over (partition by intDiv(number, 3)) from numbers(10) group by number order by count() desc;
 
 9	9
 6	8
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index fcb3da896dc..dce7069ed5f 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -27,3 +27,6 @@ select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 -- should work in ORDER BY though
 select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
+-- this one doesn't work yet -- looks like the column names clash, and the
+-- window count() is overwritten with aggregate count()
+-- select number, count(), count() over (partition by intDiv(number, 3)) from numbers(10) group by number order by count() desc;

From 32cbfabfaf75b51ad60166c11f2b7554c80efad6 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Fri, 18 Dec 2020 23:04:51 +0300
Subject: [PATCH 202/504] Update settings.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Выполнил перевод на русский язык и внес поправки в английскую версию.
---
 .../mergetree-family/replication.md           |  3 +--
 docs/en/operations/settings/settings.md       | 10 +++++++---
 docs/ru/operations/settings/settings.md       | 19 +++++++++++++++++++
 3 files changed, 27 insertions(+), 5 deletions(-)

diff --git a/docs/en/engines/table-engines/mergetree-family/replication.md b/docs/en/engines/table-engines/mergetree-family/replication.md
index 4758ff321e8..03319fd7575 100644
--- a/docs/en/engines/table-engines/mergetree-family/replication.md
+++ b/docs/en/engines/table-engines/mergetree-family/replication.md
@@ -113,8 +113,6 @@ You can have any number of replicas of the same data. Yandex.Metrica uses double
 
 The system monitors data synchronicity on replicas and is able to recover after a failure. Failover is automatic (for small differences in data) or semi-automatic (when data differs too much, which may indicate a configuration error).
 
-You can use a special logic for performing merges (for more information, see the documentation for [execute_merges_on_single_replica_time_threshold](../../../operations/settings/settings.md#execute-merges-on-single-replica-time-threshold) setting). 
-
 ## Creating Replicated Tables {#creating-replicated-tables}
 
 The `Replicated` prefix is added to the table engine name. For example:`ReplicatedMergeTree`.
@@ -286,5 +284,6 @@ If the data in ZooKeeper was lost or damaged, you can save data by moving it to
 **See also**
 
 -   [background_schedule_pool_size](../../../operations/settings/settings.md#background_schedule_pool_size)
+-   [execute_merges_on_single_replica_time_threshold setting](../../../operations/settings/settings.md#execute-merges-on-single-replica-time-threshold)
 
 [Original article](https://clickhouse.tech/docs/en/operations/table_engines/replication/) <!--hide-->
diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 6e34d76fb97..a9c26d4c06f 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2465,12 +2465,16 @@ Enables special logic to perform merges on replicas.
 Possible values:
 
 -   Positive integer (in seconds).
--   0 — Special merges logic is not executed.
+-   0 — Special merges logic is not used. Merges happen in the usual way on all the replicas.
 
 Default value: `0`.
 
-**Special logic to perform merges**
+**Usage**
 
-Selects one replica to perform the merge on. Sets the time threshold from the start of the merge. Other replicas wait for the merge to finish, then download the result. If the time threshold passes, and the selected replica does not perform merge, then the merge is performed manually.
+It can be useful when merges are CPU bounded not IO bounded (heavy compression is in use, calculation expensive aggregate functions or default expressions or just very high number of tiny merges).
+
+Selects one replica to perform the merge on. Sets the time threshold from the start of the merge. Other replicas wait for the merge to finish, then download the result. If the time threshold passes and the selected replica does not perform the merge, then the merge is performed on other replicas as usual.
+
+High values for that threshold may lead to replication delays.
 
 [Original article](https://clickhouse.tech/docs/en/operations/settings/settings/) <!-- hide -->
diff --git a/docs/ru/operations/settings/settings.md b/docs/ru/operations/settings/settings.md
index 0a8094231c2..2cf6c5e5795 100644
--- a/docs/ru/operations/settings/settings.md
+++ b/docs/ru/operations/settings/settings.md
@@ -2324,4 +2324,23 @@ SELECT number FROM numbers(3) FORMAT JSONEachRow;
 
 Значение по умолчанию: `0`.
 
+## execute_merges_on_single_replica_time_threshold {#execute-merges-on-single-replica-time-threshold}
+
+Включает особую логику выполнения слияний на репликах.
+
+Возможные значения:
+
+-   Положительное целое число (в секундах).
+-   0 — не используется особая логика выполнения слияний. Слияния происходят обычным образом на всех репликах.
+
+Значение по умолчанию: `0`.
+
+**Использование**
+
+Это может быть полезно, когда слияния ограничены процессором, а не вводом-выводом (используется тяжелое сжатие, вычисление дорогостоящих агрегатных функций или выражений по умолчанию или просто очень большое количество мелких слияний).
+
+Выбирается одна реплика для выполнения слияния. Устанавливается порог времени с момента начала слияния. Другие реплики ждут завершения слияния, а затем скачивают результат. Если время выполнения слияния превышает установленный порог и выбранная реплика не выполняет слияние, тогда слияние выполняется на других репликах как обычно.
+
+Большие значения этой настройки могут привести к задержкам репликации.
+
 [Оригинальная статья](https://clickhouse.tech/docs/ru/operations/settings/settings/) <!--hide-->

From b2d53d802d99ada0f73017a9631d18f98bc266ef Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 18 Dec 2020 23:32:52 +0300
Subject: [PATCH 203/504] Fix for non-adaptive granularity

---
 .../MergeTree/IMergeTreeDataPartWriter.cpp    |  4 ++-
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 28 +++++++++++++------
 2 files changed, 23 insertions(+), 9 deletions(-)

diff --git a/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp b/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
index df905215df1..83fd9692e49 100644
--- a/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
+++ b/src/Storages/MergeTree/IMergeTreeDataPartWriter.cpp
@@ -49,7 +49,9 @@ IMergeTreeDataPartWriter::IMergeTreeDataPartWriter(
     , columns_list(columns_list_)
     , settings(settings_)
     , index_granularity(index_granularity_)
-    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity){}
+    , with_final_mark(storage.getSettings()->write_final_mark && settings.can_use_adaptive_granularity)
+{
+}
 
 Columns IMergeTreeDataPartWriter::releaseIndexColumns()
 {
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index ca0e4e014a5..3769fd6daf2 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -429,8 +429,8 @@ void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name,
 
         if (index_granularity_rows != index_granularity.getMarkRows(mark_num))
             throw Exception(
-                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), in-memory {}, on disk {}, total marks {}",
-                mark_num, offset_in_compressed_file, offset_in_decompressed_block, index_granularity.getMarkRows(mark_num), index_granularity_rows, index_granularity.getMarksCount());
+                ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for part {} for mark #{} (compressed offset {}, decompressed offset {}), in-memory {}, on disk {}, total marks {}",
+                data_part->getFullPath(), mark_num, offset_in_compressed_file, offset_in_decompressed_block, index_granularity.getMarkRows(mark_num), index_granularity_rows, index_granularity.getMarksCount());
 
         auto column = type.createColumn();
 
@@ -444,6 +444,9 @@ void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name,
         /// Now they must be equal
         if (column->size() != index_granularity_rows)
         {
+            if (must_be_last && !settings.can_use_adaptive_granularity)
+                break;
+
             throw Exception(
                 ErrorCodes::LOGICAL_ERROR, "Incorrect mark rows for mark #{} (compressed offset {}, decompressed offset {}), actually in bin file {}, in mrk file {}",
                 mark_num, offset_in_compressed_file, offset_in_decompressed_block, column->size(), index_granularity.getMarkRows(mark_num));
@@ -575,19 +578,28 @@ void MergeTreeDataPartWriterWide::adjustLastMarkAndFlushToDisk()
     if (last_non_written_marks.empty())
         throw Exception(ErrorCodes::LOGICAL_ERROR, "No saved marks for last mark {} having rows offset {}, total marks {}", getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
 
-    index_granularity.popMark();
-    index_granularity.appendMark(rows_written_in_last_mark);
+    if (settings.can_use_adaptive_granularity)
+    {
+        index_granularity.popMark();
+        index_granularity.appendMark(rows_written_in_last_mark);
+    }
+
     for (const auto & [name, marks] : last_non_written_marks)
+    {
         for (const auto & mark : marks)
             flushMarkToFile(mark, index_granularity.getMarkRows(getCurrentMark()));
+    }
 
     last_non_written_marks.clear();
 
-    for (size_t i = 0; i < skip_indices.size(); ++i)
-        ++skip_index_accumulated_marks[i];
+    if (settings.can_use_adaptive_granularity)
+    {
+        for (size_t i = 0; i < skip_indices.size(); ++i)
+            ++skip_index_accumulated_marks[i];
 
-    setCurrentMark(getCurrentMark() + 1);
-    rows_written_in_last_mark = 0;
+        setCurrentMark(getCurrentMark() + 1);
+        rows_written_in_last_mark = 0;
+    }
 }
 
 }

From 5705fa5a34c3c172c6bca00f19040ad49e5fcec9 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Sat, 19 Dec 2020 00:08:13 +0300
Subject: [PATCH 204/504] Update settings.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Внес поправки в английскую и русскую версии.
---
 .../en/engines/table-engines/mergetree-family/replication.md | 4 ++--
 docs/en/operations/settings/settings.md                      | 4 ++--
 .../ru/engines/table-engines/mergetree-family/replication.md | 5 +++++
 docs/ru/operations/settings/settings.md                      | 4 ++--
 4 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/docs/en/engines/table-engines/mergetree-family/replication.md b/docs/en/engines/table-engines/mergetree-family/replication.md
index 03319fd7575..04a239e0b06 100644
--- a/docs/en/engines/table-engines/mergetree-family/replication.md
+++ b/docs/en/engines/table-engines/mergetree-family/replication.md
@@ -281,9 +281,9 @@ After this, you can launch the server, create a `MergeTree` table, move the data
 
 If the data in ZooKeeper was lost or damaged, you can save data by moving it to an unreplicated table as described above.
 
-**See also**
+**See Also**
 
 -   [background_schedule_pool_size](../../../operations/settings/settings.md#background_schedule_pool_size)
--   [execute_merges_on_single_replica_time_threshold setting](../../../operations/settings/settings.md#execute-merges-on-single-replica-time-threshold)
+-   [execute_merges_on_single_replica_time_threshold](../../../operations/settings/settings.md#execute-merges-on-single-replica-time-threshold)
 
 [Original article](https://clickhouse.tech/docs/en/operations/table_engines/replication/) <!--hide-->
diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index a9c26d4c06f..092c76af298 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2471,10 +2471,10 @@ Default value: `0`.
 
 **Usage**
 
-It can be useful when merges are CPU bounded not IO bounded (heavy compression is in use, calculation expensive aggregate functions or default expressions or just very high number of tiny merges).
-
 Selects one replica to perform the merge on. Sets the time threshold from the start of the merge. Other replicas wait for the merge to finish, then download the result. If the time threshold passes and the selected replica does not perform the merge, then the merge is performed on other replicas as usual.
 
 High values for that threshold may lead to replication delays.
 
+Heavy compression is in use, calculation expensive aggregate functions or default expressions or just very high number of tiny merges when using the settings. It can be useful when merges are CPU bounded not IO bounded.
+
 [Original article](https://clickhouse.tech/docs/en/operations/settings/settings/) <!-- hide -->
diff --git a/docs/ru/engines/table-engines/mergetree-family/replication.md b/docs/ru/engines/table-engines/mergetree-family/replication.md
index 6d3930e33ce..f17e1b035d4 100644
--- a/docs/ru/engines/table-engines/mergetree-family/replication.md
+++ b/docs/ru/engines/table-engines/mergetree-family/replication.md
@@ -246,4 +246,9 @@ $ sudo -u clickhouse touch /var/lib/clickhouse/flags/force_restore_data
 
 Если данные в ZooKeeper оказались утеряны или повреждены, то вы можете сохранить данные, переместив их в нереплицируемую таблицу, как описано в пункте выше.
 
+**Смотрите также**
+
+-   [background_schedule_pool_size](../../../operations/settings/settings.md#background_schedule_pool_size)
+-   [execute_merges_on_single_replica_time_threshold](../../../operations/settings/settings.md#execute-merges-on-single-replica-time-threshold)
+
 [Оригинальная статья](https://clickhouse.tech/docs/ru/operations/table_engines/replication/) <!--hide-->
diff --git a/docs/ru/operations/settings/settings.md b/docs/ru/operations/settings/settings.md
index 2cf6c5e5795..8777dba6609 100644
--- a/docs/ru/operations/settings/settings.md
+++ b/docs/ru/operations/settings/settings.md
@@ -2337,10 +2337,10 @@ SELECT number FROM numbers(3) FORMAT JSONEachRow;
 
 **Использование**
 
-Это может быть полезно, когда слияния ограничены процессором, а не вводом-выводом (используется тяжелое сжатие, вычисление дорогостоящих агрегатных функций или выражений по умолчанию или просто очень большое количество мелких слияний).
-
 Выбирается одна реплика для выполнения слияния. Устанавливается порог времени с момента начала слияния. Другие реплики ждут завершения слияния, а затем скачивают результат. Если время выполнения слияния превышает установленный порог и выбранная реплика не выполняет слияние, тогда слияние выполняется на других репликах как обычно.
 
 Большие значения этой настройки могут привести к задержкам репликации.
 
+При использовании настройки происходит тяжелое сжатие, вычисление слишком затратных агрегатных функций или выражений по умолчанию или очень большое количество мелких слияний. Эта настройка полезна, когда слияния ограничены процессором, а не вводом-выводом.
+
 [Оригинальная статья](https://clickhouse.tech/docs/ru/operations/settings/settings/) <!--hide-->

From ca6708e5945e983625469f1e878659a2670d7e70 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Sat, 19 Dec 2020 21:59:20 +0300
Subject: [PATCH 205/504] Fix check and granule adjust for non adaptive parts

---
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 38 ++++++++++++-------
 .../MergeTree/MergeTreeDataPartWriterWide.h   |  2 +-
 2 files changed, 26 insertions(+), 14 deletions(-)

diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index 3769fd6daf2..84deccbf23e 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -172,7 +172,7 @@ void MergeTreeDataPartWriterWide::write(const Block & block, const IColumn::Perm
         {
             size_t rows_left_in_last_mark = index_granularity.getMarkRows(getCurrentMark()) - rows_written_in_last_mark;
             if (rows_left_in_last_mark > index_granularity_for_block)
-                adjustLastMarkAndFlushToDisk();
+                adjustLastMarkIfNeedAndFlushToDisk();
         }
 
         fillIndexGranularity(index_granularity_for_block, block.rows());
@@ -390,7 +390,7 @@ void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name,
     bool must_be_last = false;
     UInt64 offset_in_compressed_file = 0;
     UInt64 offset_in_decompressed_block = 0;
-    UInt64 index_granularity_rows = 0;
+    UInt64 index_granularity_rows = data_part->index_granularity_info.fixed_index_granularity;
 
     size_t mark_num;
 
@@ -404,7 +404,7 @@ void MergeTreeDataPartWriterWide::validateColumnOfFixedSize(const String & name,
         if (settings.can_use_adaptive_granularity)
             DB::readBinary(index_granularity_rows, mrk_in);
         else
-            index_granularity_rows = storage.getSettings()->index_granularity;
+            index_granularity_rows = data_part->index_granularity_info.fixed_index_granularity;
 
         if (must_be_last)
         {
@@ -476,7 +476,7 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
     serialize_settings.low_cardinality_use_single_dictionary_for_part = global_settings.low_cardinality_use_single_dictionary_for_part != 0;
     WrittenOffsetColumns offset_columns;
     if (rows_written_in_last_mark > 0)
-        adjustLastMarkAndFlushToDisk();
+        adjustLastMarkIfNeedAndFlushToDisk();
 
     bool write_final_mark = (with_final_mark && data_written);
 
@@ -506,6 +506,8 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
     serialization_states.clear();
 
 #ifndef NDEBUG
+    /// Heavy weight validation of written data. Checks that we are able to read
+    /// data according to marks. Otherwise throws LOGICAL_ERROR (equal to about in debug mode)
     for (const auto & column : columns_list)
     {
         if (column.type->isValueRepresentedByNumber() && !column.type->haveSubtypes())
@@ -570,20 +572,26 @@ void MergeTreeDataPartWriterWide::fillIndexGranularity(size_t index_granularity_
 }
 
 
-void MergeTreeDataPartWriterWide::adjustLastMarkAndFlushToDisk()
+void MergeTreeDataPartWriterWide::adjustLastMarkIfNeedAndFlushToDisk()
 {
-    if (getCurrentMark() != index_granularity.getMarksCount() - 1)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Non last mark {} having rows offset {}, total marks {}", getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
-
-    if (last_non_written_marks.empty())
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "No saved marks for last mark {} having rows offset {}, total marks {}", getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
-
-    if (settings.can_use_adaptive_granularity)
+    /// We can adjust marks only if we computed granularity for blocks.
+    /// Otherwise we cannot change granularity because it will differ from
+    /// other columns
+    if (compute_granularity && settings.can_use_adaptive_granularity)
     {
+        if (getCurrentMark() != index_granularity.getMarksCount() - 1)
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Non last mark {} (with {} rows) having rows offset {}, total marks {}",
+                            getCurrentMark(), index_granularity.getMarkRows(getCurrentMark()), rows_written_in_last_mark, index_granularity.getMarksCount());
+
         index_granularity.popMark();
         index_granularity.appendMark(rows_written_in_last_mark);
     }
 
+    /// Last mark should be filled, otherwise it's a bug
+    if (last_non_written_marks.empty())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "No saved marks for last mark {} having rows offset {}, total marks {}",
+                        getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
+
     for (const auto & [name, marks] : last_non_written_marks)
     {
         for (const auto & mark : marks)
@@ -592,12 +600,16 @@ void MergeTreeDataPartWriterWide::adjustLastMarkAndFlushToDisk()
 
     last_non_written_marks.clear();
 
-    if (settings.can_use_adaptive_granularity)
+    if (compute_granularity && settings.can_use_adaptive_granularity)
     {
+        /// Also we add mark to each skip index because all of them
+        /// already accumulated all rows from current adjusting mark
         for (size_t i = 0; i < skip_indices.size(); ++i)
             ++skip_index_accumulated_marks[i];
 
+        /// This mark completed, go further
         setCurrentMark(getCurrentMark() + 1);
+        /// Without offset
         rows_written_in_last_mark = 0;
     }
 }
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
index 3fc3e3bfa3d..b350e590d07 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
@@ -104,7 +104,7 @@ private:
     ///
     /// This function used when blocks change granularity drastically and we have unfinished mark.
     /// Also useful to have exact amount of rows in last (non-final) mark.
-    void adjustLastMarkAndFlushToDisk();
+    void adjustLastMarkIfNeedAndFlushToDisk();
 
     IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns) const;
 

From 0041c293abdc00c6b1f20bcb8efad84ff4e1d1c2 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Sun, 20 Dec 2020 11:01:39 +0300
Subject: [PATCH 206/504] Fix gcc-10 build

---
 src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index 84deccbf23e..beb3646df17 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -355,13 +355,13 @@ void MergeTreeDataPartWriterWide::writeColumn(
 
         if (granule.is_complete)
         {
-            auto it = last_non_written_marks.find(name);
-            if (it == last_non_written_marks.end())
+            auto marks_it = last_non_written_marks.find(name);
+            if (marks_it == last_non_written_marks.end())
                 throw Exception(ErrorCodes::LOGICAL_ERROR, "No mark was saved for incomplete granule for column {}", backQuoteIfNeed(name));
 
-            for (const auto & mark : it->second)
+            for (const auto & mark : marks_it->second)
                 flushMarkToFile(mark, index_granularity.getMarkRows(granule.mark_number));
-            last_non_written_marks.erase(it);
+            last_non_written_marks.erase(marks_it);
         }
     }
 

From 906fd18f4aa23814079c9958a85caa7a8561111b Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Sun, 20 Dec 2020 14:21:04 +0300
Subject: [PATCH 207/504] Update settings.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Внес поправки в английскую и русскую версии.
---
 docs/en/operations/settings/settings.md | 2 +-
 docs/ru/operations/settings/settings.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 092c76af298..8b9790b4d5d 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2475,6 +2475,6 @@ Selects one replica to perform the merge on. Sets the time threshold from the st
 
 High values for that threshold may lead to replication delays.
 
-Heavy compression is in use, calculation expensive aggregate functions or default expressions or just very high number of tiny merges when using the settings. It can be useful when merges are CPU bounded not IO bounded.
+It can be useful when merges are CPU bounded not IO bounded (performing heavy data compression, calculating aggregate functions or default expressions that require a large amount of calculations, or just very high number of tiny merges).
 
 [Original article](https://clickhouse.tech/docs/en/operations/settings/settings/) <!-- hide -->
diff --git a/docs/ru/operations/settings/settings.md b/docs/ru/operations/settings/settings.md
index 8777dba6609..2f1b89fe645 100644
--- a/docs/ru/operations/settings/settings.md
+++ b/docs/ru/operations/settings/settings.md
@@ -2341,6 +2341,6 @@ SELECT number FROM numbers(3) FORMAT JSONEachRow;
 
 Большие значения этой настройки могут привести к задержкам репликации.
 
-При использовании настройки происходит тяжелое сжатие, вычисление слишком затратных агрегатных функций или выражений по умолчанию или очень большое количество мелких слияний. Эта настройка полезна, когда слияния ограничены процессором, а не вводом-выводом.
+Эта настройка полезна, когда скорость слияния ограничивается мощностью процессора, а не скоростью операций ввода-вывода (при выполнении "тяжелого" сжатия данных, при расчете агрегатных функций или выражений по умолчанию, требующих большого объема вычислений, или просто при большом количестве мелких слияний).
 
 [Оригинальная статья](https://clickhouse.tech/docs/ru/operations/settings/settings/) <!--hide-->

From 07b5c03c542e835ad8c1b0b905d9532b3dab4b20 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 11:24:52 +0300
Subject: [PATCH 208/504] Better adjustment of the last mark

---
 .../MergeTree/MergeTreeDataPartWriterWide.cpp | 51 ++++++++++++-------
 .../MergeTree/MergeTreeDataPartWriterWide.h   |  7 +--
 .../0_stateless/00955_test_final_mark.sql     | 16 +++---
 3 files changed, 44 insertions(+), 30 deletions(-)

diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
index beb3646df17..be735104e99 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.cpp
@@ -171,8 +171,18 @@ void MergeTreeDataPartWriterWide::write(const Block & block, const IColumn::Perm
         if (rows_written_in_last_mark > 0)
         {
             size_t rows_left_in_last_mark = index_granularity.getMarkRows(getCurrentMark()) - rows_written_in_last_mark;
+            /// Previous granularity was much bigger than our new block's
+            /// granularity let's adjust it, because we want add new
+            /// heavy-weight blocks into small old granule.
             if (rows_left_in_last_mark > index_granularity_for_block)
-                adjustLastMarkIfNeedAndFlushToDisk();
+            {
+                /// We have already written more rows than granularity of our block.
+                /// adjust last mark rows and flush to disk.
+                if (rows_written_in_last_mark >= index_granularity_for_block)
+                    adjustLastMarkIfNeedAndFlushToDisk(rows_written_in_last_mark);
+                else /// We still can write some rows from new block into previous granule.
+                    adjustLastMarkIfNeedAndFlushToDisk(index_granularity_for_block - rows_written_in_last_mark);
+            }
         }
 
         fillIndexGranularity(index_granularity_for_block, block.rows());
@@ -476,7 +486,7 @@ void MergeTreeDataPartWriterWide::finishDataSerialization(IMergeTreeDataPart::Ch
     serialize_settings.low_cardinality_use_single_dictionary_for_part = global_settings.low_cardinality_use_single_dictionary_for_part != 0;
     WrittenOffsetColumns offset_columns;
     if (rows_written_in_last_mark > 0)
-        adjustLastMarkIfNeedAndFlushToDisk();
+        adjustLastMarkIfNeedAndFlushToDisk(rows_written_in_last_mark);
 
     bool write_final_mark = (with_final_mark && data_written);
 
@@ -572,7 +582,7 @@ void MergeTreeDataPartWriterWide::fillIndexGranularity(size_t index_granularity_
 }
 
 
-void MergeTreeDataPartWriterWide::adjustLastMarkIfNeedAndFlushToDisk()
+void MergeTreeDataPartWriterWide::adjustLastMarkIfNeedAndFlushToDisk(size_t new_rows_in_last_mark)
 {
     /// We can adjust marks only if we computed granularity for blocks.
     /// Otherwise we cannot change granularity because it will differ from
@@ -584,7 +594,7 @@ void MergeTreeDataPartWriterWide::adjustLastMarkIfNeedAndFlushToDisk()
                             getCurrentMark(), index_granularity.getMarkRows(getCurrentMark()), rows_written_in_last_mark, index_granularity.getMarksCount());
 
         index_granularity.popMark();
-        index_granularity.appendMark(rows_written_in_last_mark);
+        index_granularity.appendMark(new_rows_in_last_mark);
     }
 
     /// Last mark should be filled, otherwise it's a bug
@@ -592,25 +602,28 @@ void MergeTreeDataPartWriterWide::adjustLastMarkIfNeedAndFlushToDisk()
         throw Exception(ErrorCodes::LOGICAL_ERROR, "No saved marks for last mark {} having rows offset {}, total marks {}",
                         getCurrentMark(), rows_written_in_last_mark, index_granularity.getMarksCount());
 
-    for (const auto & [name, marks] : last_non_written_marks)
+    if (rows_written_in_last_mark == new_rows_in_last_mark)
     {
-        for (const auto & mark : marks)
-            flushMarkToFile(mark, index_granularity.getMarkRows(getCurrentMark()));
-    }
+        for (const auto & [name, marks] : last_non_written_marks)
+        {
+            for (const auto & mark : marks)
+                flushMarkToFile(mark, index_granularity.getMarkRows(getCurrentMark()));
+        }
 
-    last_non_written_marks.clear();
+        last_non_written_marks.clear();
 
-    if (compute_granularity && settings.can_use_adaptive_granularity)
-    {
-        /// Also we add mark to each skip index because all of them
-        /// already accumulated all rows from current adjusting mark
-        for (size_t i = 0; i < skip_indices.size(); ++i)
-            ++skip_index_accumulated_marks[i];
+        if (compute_granularity && settings.can_use_adaptive_granularity)
+        {
+            /// Also we add mark to each skip index because all of them
+            /// already accumulated all rows from current adjusting mark
+            for (size_t i = 0; i < skip_indices.size(); ++i)
+                ++skip_index_accumulated_marks[i];
 
-        /// This mark completed, go further
-        setCurrentMark(getCurrentMark() + 1);
-        /// Without offset
-        rows_written_in_last_mark = 0;
+            /// This mark completed, go further
+            setCurrentMark(getCurrentMark() + 1);
+            /// Without offset
+            rows_written_in_last_mark = 0;
+        }
     }
 }
 
diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
index b350e590d07..8c76c10abef 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartWriterWide.h
@@ -99,12 +99,13 @@ private:
     /// in our index_granularity array.
     void shiftCurrentMark(const Granules & granules_written);
 
-    /// Change rows in the last mark in index_granularity to rows_written_in_last_mark.
-    /// Flush all marks from last_non_written_marks to disk and increment current mark.
+    /// Change rows in the last mark in index_granularity to new_rows_in_last_mark.
+    /// Flush all marks from last_non_written_marks to disk and increment current mark if already written rows
+    /// (rows_written_in_last_granule) equal to new_rows_in_last_mark.
     ///
     /// This function used when blocks change granularity drastically and we have unfinished mark.
     /// Also useful to have exact amount of rows in last (non-final) mark.
-    void adjustLastMarkIfNeedAndFlushToDisk();
+    void adjustLastMarkIfNeedAndFlushToDisk(size_t new_rows_in_last_mark);
 
     IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenOffsetColumns & offset_columns) const;
 
diff --git a/tests/queries/0_stateless/00955_test_final_mark.sql b/tests/queries/0_stateless/00955_test_final_mark.sql
index 50ca3d008f9..e020f10b71a 100644
--- a/tests/queries/0_stateless/00955_test_final_mark.sql
+++ b/tests/queries/0_stateless/00955_test_final_mark.sql
@@ -18,7 +18,7 @@ INSERT INTO mt_with_pk (d, x, y, z, `n.Age`, `n.Name`) VALUES (toDate('2018-10-0
 
 SELECT COUNT(*) FROM mt_with_pk WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1 AND database = currentDatabase();
 
 SELECT '===test merge===';
 INSERT INTO mt_with_pk (d, x, y, z, `n.Age`, `n.Name`) VALUES (toDate('2018-10-01'), toDateTime('2018-10-01 07:57:57'), [4, 4, 4], 14, [111, 222], ['Lui', 'Dave']), (toDate('2018-10-01'), toDateTime('2018-10-01 08:57:57'), [5, 5, 5], 15, [333, 444], ['John', 'Mike']), (toDate('2018-10-01'), toDateTime('2018-10-01 09:57:57'), [6, 6, 6], 16, [555, 666, 777], ['Alex', 'Jim', 'Tom']);
@@ -27,7 +27,7 @@ OPTIMIZE TABLE mt_with_pk FINAL;
 
 SELECT COUNT(*) FROM mt_with_pk WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1 AND database = currentDatabase();
 
 SELECT '===test alter===';
 ALTER TABLE mt_with_pk MODIFY COLUMN y Array(String);
@@ -38,7 +38,7 @@ OPTIMIZE TABLE mt_with_pk FINAL;
 
 SELECT COUNT(*) FROM mt_with_pk WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1 AND database = currentDatabase();
 
 SELECT '===test mutation===';
 ALTER TABLE mt_with_pk UPDATE w = 0 WHERE 1 SETTINGS mutations_sync = 2;
@@ -58,7 +58,7 @@ OPTIMIZE TABLE mt_with_pk FINAL;
 
 SELECT COUNT(*) FROM mt_with_pk WHERE z + w > 5000;
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_pk' AND database = currentDatabase() AND active=1 AND database = currentDatabase();
 
 DROP TABLE IF EXISTS mt_with_pk;
 
@@ -119,7 +119,7 @@ INSERT INTO mt_without_pk (d, x, y, z, `n.Age`, `n.Name`) VALUES (toDate('2018-1
 
 SELECT COUNT(*) FROM mt_without_pk WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_without_pk' AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_without_pk' AND active=1 AND database = currentDatabase();
 
 INSERT INTO mt_without_pk (d, x, y, z, `n.Age`, `n.Name`) VALUES (toDate('2018-10-01'), toDateTime('2018-10-01 07:57:57'), [4, 4, 4], 14, [111, 222], ['Lui', 'Dave']), (toDate('2018-10-01'), toDateTime('2018-10-01 08:57:57'), [5, 5, 5], 15, [333, 444], ['John', 'Mike']), (toDate('2018-10-01'), toDateTime('2018-10-01 09:57:57'), [6, 6, 6], 16, [555, 666, 777], ['Alex', 'Jim', 'Tom']);
 
@@ -127,7 +127,7 @@ OPTIMIZE TABLE mt_without_pk FINAL;
 
 SELECT COUNT(*) FROM mt_without_pk WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_without_pk' AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_without_pk' AND active=1 AND database = currentDatabase();
 
 DROP TABLE IF EXISTS mt_without_pk;
 
@@ -149,7 +149,7 @@ INSERT INTO mt_with_small_granularity (d, x, y, z, `n.Age`, `n.Name`) VALUES (to
 
 SELECT COUNT(*) FROM mt_with_small_granularity WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_small_granularity' AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_small_granularity' AND active=1 AND database = currentDatabase();
 
 INSERT INTO mt_with_small_granularity (d, x, y, z, `n.Age`, `n.Name`) VALUES (toDate('2018-10-01'), toDateTime('2018-10-01 07:57:57'), [4, 4, 4], 14, [111, 222], ['Lui', 'Dave']), (toDate('2018-10-01'), toDateTime('2018-10-01 08:57:57'), [5, 5, 5], 15, [333, 444], ['John', 'Mike']), (toDate('2018-10-01'), toDateTime('2018-10-01 09:57:57'), [6, 6, 6], 16, [555, 666, 777], ['Alex', 'Jim', 'Tom']);
 
@@ -157,6 +157,6 @@ OPTIMIZE TABLE mt_with_small_granularity FINAL;
 
 SELECT COUNT(*) FROM mt_with_small_granularity WHERE x > toDateTime('2018-10-01 23:57:57');
 
-SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_small_granularity' AND active=1;
+SELECT sum(marks) FROM system.parts WHERE table = 'mt_with_small_granularity' AND active=1 AND database = currentDatabase();
 
 DROP TABLE IF EXISTS mt_with_small_granularity;

From bc0e91b5868bcbd90292afe8f65c1ec7281c9ebc Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 21 Dec 2020 12:23:23 +0300
Subject: [PATCH 209/504] Fix special build

---
 src/Processors/Executors/PollingQueue.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Processors/Executors/PollingQueue.h b/src/Processors/Executors/PollingQueue.h
index 5588bb87436..9d37bf0a2cc 100644
--- a/src/Processors/Executors/PollingQueue.h
+++ b/src/Processors/Executors/PollingQueue.h
@@ -49,7 +49,7 @@ public:
     void finish();
 };
 #else
-class AsyncTaskQueue
+class PollingQueue
 {
 public:
     bool empty() { return true; }

From ac9fc1550921b8e458de5b2dc6373b535ceaef07 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Mon, 21 Dec 2020 12:59:11 +0300
Subject: [PATCH 210/504] cleanup

---
 src/Interpreters/ExpressionAnalyzer.cpp        |  1 +
 src/Interpreters/GetAggregatesVisitor.h        |  8 +++++---
 src/Interpreters/RewriteAnyFunctionVisitor.cpp |  7 +++----
 src/Processors/Transforms/WindowTransform.cpp  | 12 ++++++------
 4 files changed, 15 insertions(+), 13 deletions(-)

diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index af6ee4a1732..42f9e5598fd 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -66,6 +66,7 @@ namespace ErrorCodes
     extern const int ILLEGAL_TYPE_OF_ARGUMENT;
     extern const int ILLEGAL_TYPE_OF_COLUMN_FOR_FILTER;
     extern const int LOGICAL_ERROR;
+    extern const int NOT_IMPLEMENTED;
     extern const int UNKNOWN_IDENTIFIER;
     extern const int UNKNOWN_TYPE_OF_AST_NODE;
 }
diff --git a/src/Interpreters/GetAggregatesVisitor.h b/src/Interpreters/GetAggregatesVisitor.h
index 94510f0aaa2..d416a5f240e 100644
--- a/src/Interpreters/GetAggregatesVisitor.h
+++ b/src/Interpreters/GetAggregatesVisitor.h
@@ -20,9 +20,11 @@ public:
     {
         const char * assert_no_aggregates = nullptr;
         const char * assert_no_windows = nullptr;
-        std::unordered_set<String> uniq_names;
-        std::vector<const ASTFunction *> aggregates;
-        std::vector<const ASTFunction *> window_functions;
+        // Explicit empty initializers are needed to make designated initializers
+        // work on GCC 10.
+        std::unordered_set<String> uniq_names {};
+        std::vector<const ASTFunction *> aggregates {};
+        std::vector<const ASTFunction *> window_functions {};
     };
 
     static bool needChildVisit(const ASTPtr & node, const ASTPtr & child)
diff --git a/src/Interpreters/RewriteAnyFunctionVisitor.cpp b/src/Interpreters/RewriteAnyFunctionVisitor.cpp
index ed3663a3b86..e8f05962862 100644
--- a/src/Interpreters/RewriteAnyFunctionVisitor.cpp
+++ b/src/Interpreters/RewriteAnyFunctionVisitor.cpp
@@ -40,14 +40,13 @@ bool extractIdentifiers(const ASTFunction & func, std::unordered_set<ASTPtr *> &
 
             // We are looking for identifiers inside a function calculated inside
             // the aggregate function `any()`. Window or aggregate function can't
-            // be inside `any` (checked by GetAggregatesMatcher).
+            // be inside `any`, but this check in GetAggregatesMatcher happens
+            // later, so we have to explicitly skip these nested functions here.
             if (arg_func->is_window_function
                 || AggregateFunctionFactory::instance().isAggregateFunctionName(
                     arg_func->name))
             {
-                throw Exception(ErrorCodes::LOGICAL_ERROR,
-                    "Found an aggregate function '{}' inside another aggregate"
-                    " function", arg_func->formatForErrorMessage());
+                return false;
             }
 
             if (!extractIdentifiers(*arg_func, identifiers))
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index 9ffad3ae367..b39926a0335 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -18,10 +18,10 @@ WindowTransform::WindowTransform(const Block & input_header_,
     , window_description(window_description_)
 {
     workspaces.reserve(window_function_descriptions.size());
-    for (size_t i = 0; i < window_function_descriptions.size(); ++i)
+    for (const auto & f : window_function_descriptions)
     {
         WindowFunctionWorkspace workspace;
-        workspace.window_function = window_function_descriptions[i];
+        workspace.window_function = f;
 
         const auto & aggregate_function
             = workspace.window_function.aggregate_function;
@@ -61,10 +61,10 @@ WindowTransform::WindowTransform(const Block & input_header_,
 WindowTransform::~WindowTransform()
 {
     // Some states may be not created yet if the creation failed.
-    for (size_t i = 0; i < workspaces.size(); i++)
+    for (auto & w : workspaces)
     {
-        workspaces[i].window_function.aggregate_function->destroy(
-            workspaces[i].aggregate_function_state.data());
+        w.window_function.aggregate_function->destroy(
+            w.aggregate_function_state.data());
     }
 }
 
@@ -103,7 +103,7 @@ void WindowTransform::transform(Chunk & chunk)
             // Check whether the new partition has started and reinitialize the
             // aggregate function states.
             assert(partition_start_columns.size() == partition_by_indices.size());
-            if (partition_start_columns.size() == 0)
+            if (partition_start_columns.empty())
             {
                 // No PARTITION BY at all, do nothing.
             }

From 643739df6b934da138e041106cf9e25891473507 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 13:06:36 +0300
Subject: [PATCH 211/504] Trying to enable thread fuzzer in flaky check

---
 docker/test/stateless/run.sh | 13 ++++++++++++-
 src/Common/ThreadFuzzer.cpp  |  5 +++++
 2 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/docker/test/stateless/run.sh b/docker/test/stateless/run.sh
index 8793f3d22b4..cf401fe8849 100755
--- a/docker/test/stateless/run.sh
+++ b/docker/test/stateless/run.sh
@@ -12,12 +12,23 @@ dpkg -i package_folder/clickhouse-test_*.deb
 # install test configs
 /usr/share/clickhouse-test/config/install.sh
 
-service clickhouse-server start && sleep 5
+# For flaky check we also enable thread fuzzer
+if [ "$NUM_TRIES" -gt "1" ]; then
+    export THREAD_FUZZER_CPU_TIME_PERIOD_US=1000
+    export THREAD_FUZZER_SLEEP_PROBABILITY=0.1
+    export THREAD_FUZZER_SLEEP_TIME_US=100000
+    # simpliest way to forward env variables to server
+    /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml --daemon
+    sleep 5
+else
+    service clickhouse-server start && sleep 5
+fi
 
 if grep -q -- "--use-skip-list" /usr/bin/clickhouse-test; then
     SKIP_LIST_OPT="--use-skip-list"
 fi
 
+
 function run_tests()
 {
     # We can have several additional options so we path them as array because it's
diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index a32e50c44b2..76cae4bb0ac 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -10,6 +10,7 @@
 #include <common/sleep.h>
 
 #include <IO/ReadHelpers.h>
+#include <common/logger_useful.h>
 
 #include <Common/Exception.h>
 #include <Common/thread_local_rng.h>
@@ -102,6 +103,10 @@ void ThreadFuzzer::initConfiguration()
     initFromEnv(sleep_probability, "THREAD_FUZZER_SLEEP_PROBABILITY");
     initFromEnv(sleep_time_us, "THREAD_FUZZER_SLEEP_TIME_US");
 
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep probability {}", sleep_probability);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Cpu time period {}", cpu_time_period_us);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep time us {}", sleep_time_us);
+
 #if THREAD_FUZZER_WRAP_PTHREAD
 #    define INIT_WRAPPER_PARAMS(RET, NAME, ...) \
         initFromEnv(NAME##_before_yield_probability, "THREAD_FUZZER_" #NAME "_BEFORE_YIELD_PROBABILITY"); \

From fdf18ceec0d446862ac49799327530349065d587 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 13:08:06 +0300
Subject: [PATCH 212/504] Remove new line

---
 docker/test/stateless/run.sh | 1 -
 1 file changed, 1 deletion(-)

diff --git a/docker/test/stateless/run.sh b/docker/test/stateless/run.sh
index cf401fe8849..7aa4d5e74ab 100755
--- a/docker/test/stateless/run.sh
+++ b/docker/test/stateless/run.sh
@@ -28,7 +28,6 @@ if grep -q -- "--use-skip-list" /usr/bin/clickhouse-test; then
     SKIP_LIST_OPT="--use-skip-list"
 fi
 
-
 function run_tests()
 {
     # We can have several additional options so we path them as array because it's

From 21533258fa8071fa301b05fcdcf2d51442d23aa4 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 13:09:07 +0300
Subject: [PATCH 213/504] Add some test to check

---
 tests/queries/0_stateless/01415_sticking_mutations.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/queries/0_stateless/01415_sticking_mutations.sh b/tests/queries/0_stateless/01415_sticking_mutations.sh
index 9ae1ef03d03..ce34cd09ca3 100755
--- a/tests/queries/0_stateless/01415_sticking_mutations.sh
+++ b/tests/queries/0_stateless/01415_sticking_mutations.sh
@@ -43,7 +43,7 @@ function check_sticky_mutations()
 
     $CLICKHOUSE_CLIENT --query "SYSTEM START MERGES sticking_mutations"
 
-    # just to be sure, that previous mutations finished
+    # Just to be sure, that previous mutations finished
     $CLICKHOUSE_CLIENT --query "ALTER TABLE sticking_mutations DELETE WHERE value2 % 31 == 0 SETTINGS mutations_sync = 1"
 
     $CLICKHOUSE_CLIENT --query "OPTIMIZE TABLE sticking_mutations FINAL"

From 24d01d2460cac741723424064c65088ee21ff5b9 Mon Sep 17 00:00:00 2001
From: qianmoQ <shicheng@ttxit.com>
Date: Mon, 21 Dec 2020 18:31:32 +0800
Subject: [PATCH 214/504] translate documentation for interfaces/formats

---
 docs/zh/interfaces/formats.md           | 845 ++++++++++++++++--------
 docs/zh/interfaces/third-party/index.md |  14 +-
 2 files changed, 574 insertions(+), 285 deletions(-)

diff --git a/docs/zh/interfaces/formats.md b/docs/zh/interfaces/formats.md
index 58d06916ed8..808bff42335 100644
--- a/docs/zh/interfaces/formats.md
+++ b/docs/zh/interfaces/formats.md
@@ -1,179 +1,237 @@
-# 输入输出格式 {#formats}
+---
+toc_priority: 21
+toc_title: 输入/输出格式
+---
 
-ClickHouse 可以接受多种数据格式，可以在 (`INSERT`) 以及 (`SELECT`) 请求中使用。
+# 输入/输出格式 {#formats}
 
-下列表格列出了支持的数据格式以及在 (`INSERT`) 以及 (`SELECT`) 请求中使用它们的方式。
+ClickHouse可以接受和返回各种格式的数据。输入支持的格式可以用来解析提供给`INSERT`的数据，可以从文件备份表(如File, URL或HDFS)执行`SELECT`，或者读取外部字典。输出支持的格式可用于获取`SELECT`的结果，并支持执行`INSERT`文件的表中。
 
-| 格式                                                            | INSERT | SELECT |
-|-----------------------------------------------------------------|--------|--------|
-| [TabSeparated](#tabseparated)                                   | ✔      | ✔      |
-| [TabSeparatedRaw](#tabseparatedraw)                             | ✗      | ✔      |
-| [TabSeparatedWithNames](#tabseparatedwithnames)                 | ✔      | ✔      |
-| [TabSeparatedWithNamesAndTypes](#tabseparatedwithnamesandtypes) | ✔      | ✔      |
-| [模板](#format-template)                                        | ✔      | ✔      |
-| [TemplateIgnoreSpaces](#templateignorespaces)                   | ✔      | ✗      |
-| [CSV](#csv)                                                     | ✔      | ✔      |
-| [CSVWithNames](#csvwithnames)                                   | ✔      | ✔      |
-| [自定义分离](#format-customseparated)                           | ✔      | ✔      |
-| [值](#data-format-values)                                       | ✔      | ✔      |
-| [垂直](#vertical)                                               | ✗      | ✔      |
-| VerticalRaw                                                     | ✗      | ✔      |
-| [JSON](#json)                                                   | ✗      | ✔      |
-| [JSONCompact](#jsoncompact)                                     | ✗      | ✔      |
-| [JSONEachRow](#jsoneachrow)                                     | ✔      | ✔      |
-| [TSKV](#tskv)                                                   | ✔      | ✔      |
-| [漂亮](#pretty)                                                 | ✗      | ✔      |
-| [PrettyCompact](#prettycompact)                                 | ✗      | ✔      |
-| [PrettyCompactMonoBlock](#prettycompactmonoblock)               | ✗      | ✔      |
-| [PrettyNoEscapes](#prettynoescapes)                             | ✗      | ✔      |
-| [PrettySpace](#prettyspace)                                     | ✗      | ✔      |
-| [Protobuf](#protobuf)                                           | ✔      | ✔      |
-| [Avro](#data-format-avro)                                       | ✔      | ✔      |
-| [AvroConfluent](#data-format-avro-confluent)                    | ✔      | ✗      |
-| [镶木地板](#data-format-parquet)                                | ✔      | ✔      |
-| [ORC](#data-format-orc)                                         | ✔      | ✗      |
-| [RowBinary](#rowbinary)                                         | ✔      | ✔      |
-| [RowBinaryWithNamesAndTypes](#rowbinarywithnamesandtypes)       | ✔      | ✔      |
-| [本地人](#native)                                               | ✔      | ✔      |
-| [Null](#null)                                                   | ✗      | ✔      |
-| [XML](#xml)                                                     | ✗      | ✔      |
-| [CapnProto](#capnproto)                                         | ✔      | ✔      |
+以下是支持的格式:
+
+| 格式                                                                                    | 输入   | 输出   |
+|-----------------------------------------------------------------------------------------|-------|--------|
+| [TabSeparated](#tabseparated)                                                           | ✔     | ✔      |
+| [TabSeparatedRaw](#tabseparatedraw)                                                     | ✔     | ✔      |
+| [TabSeparatedWithNames](#tabseparatedwithnames)                                         | ✔     | ✔      |
+| [TabSeparatedWithNamesAndTypes](#tabseparatedwithnamesandtypes)                         | ✔     | ✔      |
+| [Template](#format-template)                                                            | ✔     | ✔      |
+| [TemplateIgnoreSpaces](#templateignorespaces)                                           | ✔     | ✗      |
+| [CSV](#csv)                                                                             | ✔     | ✔      |
+| [CSVWithNames](#csvwithnames)                                                           | ✔     | ✔      |
+| [CustomSeparated](#format-customseparated)                                              | ✔     | ✔      |
+| [Values](#data-format-values)                                                           | ✔     | ✔      |
+| [Vertical](#vertical)                                                                   | ✗     | ✔      |
+| [VerticalRaw](#verticalraw)                                                             | ✗     | ✔      |
+| [JSON](#json)                                                                           | ✗     | ✔      |
+| [JSONAsString](#jsonasstring)                                                           | ✔     | ✗      |
+| [JSONString](#jsonstring)                                                               | ✗     | ✔      |
+| [JSONCompact](#jsoncompact)                                                             | ✗     | ✔      |
+| [JSONCompactString](#jsoncompactstring)                                                 | ✗     | ✔      |
+| [JSONEachRow](#jsoneachrow)                                                             | ✔     | ✔      |
+| [JSONEachRowWithProgress](#jsoneachrowwithprogress)                                     | ✗     | ✔      |
+| [JSONStringEachRow](#jsonstringeachrow)                                                 | ✔     | ✔      |
+| [JSONStringEachRowWithProgress](#jsonstringeachrowwithprogress)                         | ✗     | ✔      |
+| [JSONCompactEachRow](#jsoncompacteachrow)                                               | ✔     | ✔      |
+| [JSONCompactEachRowWithNamesAndTypes](#jsoncompacteachrowwithnamesandtypes)             | ✔     | ✔      |
+| [JSONCompactStringEachRow](#jsoncompactstringeachrow)                                   | ✔     | ✔      |
+| [JSONCompactStringEachRowWithNamesAndTypes](#jsoncompactstringeachrowwithnamesandtypes) | ✔     | ✔      |
+| [TSKV](#tskv)                                                                           | ✔     | ✔      |
+| [Pretty](#pretty)                                                                       | ✗     | ✔      |
+| [PrettyCompact](#prettycompact)                                                         | ✗     | ✔      |
+| [PrettyCompactMonoBlock](#prettycompactmonoblock)                                       | ✗     | ✔      |
+| [PrettyNoEscapes](#prettynoescapes)                                                     | ✗     | ✔      |
+| [PrettySpace](#prettyspace)                                                             | ✗     | ✔      |
+| [Protobuf](#protobuf)                                                                   | ✔     | ✔      |
+| [ProtobufSingle](#protobufsingle)                                                       | ✔     | ✔      |
+| [Avro](#data-format-avro)                                                               | ✔     | ✔      |
+| [AvroConfluent](#data-format-avro-confluent)                                            | ✔     | ✗      |
+| [Parquet](#data-format-parquet)                                                         | ✔     | ✔      |
+| [Arrow](#data-format-arrow)                                                             | ✔     | ✔      |
+| [ArrowStream](#data-format-arrow-stream)                                                | ✔     | ✔      |
+| [ORC](#data-format-orc)                                                                 | ✔     | ✗      |
+| [RowBinary](#rowbinary)                                                                 | ✔     | ✔      |
+| [RowBinaryWithNamesAndTypes](#rowbinarywithnamesandtypes)                               | ✔     | ✔      |
+| [Native](#native)                                                                       | ✔     | ✔      |
+| [Null](#null)                                                                           | ✗     | ✔      |
+| [XML](#xml)                                                                             | ✗     | ✔      |
+| [CapnProto](#capnproto)                                                                 | ✔     | ✗      |
+| [LineAsString](#lineasstring)                                                           | ✔     | ✗      |
+
+您可以使用ClickHouse设置控制一些格式处理参数。更多详情设置请参考[设置](../operations/settings/settings.md) 
 
 ## TabSeparated {#tabseparated}
 
-在 TabSeparated 格式中，数据按行写入。每行包含由制表符分隔的值。除了行中的最后一个值（后面紧跟换行符）之外，每个值都跟随一个制表符。 在任何地方都可以使用严格的 Unix 命令行。最后一行还必须在最后包含换行符。值以文本格式编写，不包含引号，并且要转义特殊字符。
+在TabSeparated分隔格式中，数据按行写入。每行包含由制表符分隔的值。每个值后跟一个制表符，除了行中最后一个值后跟换行。在任何地方都采用严格的Unix换行。最后一行还必须在末尾包含换行。值以文本格式编写，不包含引号，并使用转义的特殊字符。
 
-这种格式也可以用 `TSV` 来表示。
+这种格式也可以用`TSV`来表示。
 
-TabSeparated 格式非常方便用于自定义程序或脚本处理数据。HTTP 客户端接口默认会用这种格式，命令行客户端批量模式下也会用这种格式。这种格式允许在不同数据库之间传输数据。例如，从 MYSQL 中导出数据然后导入到 ClickHouse 中，反之亦然。
+`TabSeparated`格式便于使用自定义程序和脚本处理数据。默认情况下，它在HTTP接口和命令行客户端的批处理模式中使用。这种格式还允许在不同dbms之间传输数据。例如，您可以从MySQL获取转储并将其上传到ClickHouse，反之亦然。
 
-TabSeparated 格式支持输出数据总值（当使用 WITH TOTALS） 以及极值（当 ‘extremes’ 设置是1）。这种情况下，总值和极值输出在主数据的后面。主要的数据，总值，极值会以一个空行隔开，例如：
+`TabSeparated`格式支持输出total值(与TOTALS一起使用时)和extreme值(当`extreme`被设置为1时)。在这种情况下，total值和extreme值会在主数据后输出。主要结果、总值和极值之间用空行分隔。示例:
 
 ``` sql
 SELECT EventDate, count() AS c FROM test.hits GROUP BY EventDate WITH TOTALS ORDER BY EventDate FORMAT TabSeparated``
 ```
 
-    2014-03-17      1406958
-    2014-03-18      1383658
-    2014-03-19      1405797
-    2014-03-20      1353623
-    2014-03-21      1245779
-    2014-03-22      1031592
-    2014-03-23      1046491
+``` text
+2014-03-17      1406958
+2014-03-18      1383658
+2014-03-19      1405797
+2014-03-20      1353623
+2014-03-21      1245779
+2014-03-22      1031592
+2014-03-23      1046491
 
-    1970-01-01      8873898
+1970-01-01      8873898
 
-    2014-03-17      1031592
-    2014-03-23      1406958
+2014-03-17      1031592
+2014-03-23      1406958
+```
 
-### 数据解析方式 {#shu-ju-jie-xi-fang-shi}
+### 数据格式化 {#data-formatting}
 
-整数以十进制形式写入。数字在开头可以包含额外的 `+` 字符（解析时忽略，格式化时不记录）。非负数不能包含负号。 读取时，允许将空字符串解析为零，或者（对于带符号的类型）将仅包含负号的字符串解析为零。 不符合相应数据类型的数字可能会被解析为不同的数字，而不会显示错误消息。
+整数是用十进制形式写的。数字可以在开头包含一个额外的`+`字符(解析时忽略，格式化时不记录)。非负数不能包含负号。在读取时，允许将空字符串解析为零，或者(对于有符号类型)将仅由一个负号组成的字符串解析为零。不符合相应数据类型的数字可以被解析为不同的数字，而不会出现错误消息。
 
-浮点数以十进制形式写入。点号用作小数点分隔符。支持指数等符号，如’inf’，‘+ inf’，‘-inf’和’nan’。 浮点数的输入可以以小数点开始或结束。
-格式化的时候，浮点数的精确度可能会丢失。
-解析的时候，没有严格需要去读取与机器可以表示的最接近的数值。
+浮点数以十进制形式书写。`.`号用作十进制分隔符。支持指数符号，如`inf`、`+inf`、`-inf`和`nan`。浮点数的条目可以以小数点开始或结束。
+在格式化期间，浮点数可能会丢失准确性。
+在解析期间，并不严格要求读取与机器可以表示的最接近的数值。
 
-日期会以 YYYY-MM-DD 格式写入和解析，但会以任何字符作为分隔符。
-带时间的日期会以 YYYY-MM-DD hh:mm:ss 格式写入和解析，但会以任何字符作为分隔符。
-这一切都发生在客户端或服务器启动时的系统时区（取决于哪一种格式的数据）。对于具有时间的日期，夏时制时间未指定。 因此，如果转储在夏令时中有时间，则转储不会明确地匹配数据，解析将选择两者之一。
-在读取操作期间，不正确的日期和具有时间的日期可以使用自然溢出或空日期和时间进行分析，而不会出现错误消息。
+日期以YYYY-MM-DD格式编写，并以相同的格式解析，但使用任何字符作为分隔符。
+日期和时间以`YYYY-MM-DD hh:mm:ss`的格式书写，并以相同的格式解析，但使用任何字符作为分隔符。
+这一切都发生在客户端或服务器启动时的系统时区(取决于它们对数据的格式)。对于带有时间的日期，夏时制时间未指定。因此，如果转储在夏令时有时间，则转储不会明确地与数据匹配，解析将选择这两次中的一次。
+在读取操作期间，不正确的日期和具有时间的日期可以使用自然溢出或null日期和时间进行分析，而不会出现错误消息。
 
-有个例外情况，Unix 时间戳格式（10个十进制数字）也支持使用时间解析日期。结果不是时区相关的。格式 YYYY-MM-DD hh:mm:ss和 NNNNNNNNNN 会自动区分。
+有个例外情况，Unix时间戳格式也支持用时间解析日期(如果它恰好由10个十进制数字组成)。其结果与时间区域无关。格式`YYYY-MM-DD hh:mm:ss`和`NNNNNNNNNN`是自动区分的。
 
-字符串以反斜线转义的特殊字符输出。 以下转义序列用于输出：`\b`，`\f`，`\r`，`\n`，`\t`，`\0`，`\'`，`\\`。 解析还支持`\a`，`\v`和`\xHH`（十六进制转义字符）和任何`\c`字符，其中`c`是任何字符（这些序列被转换为`c`）。 因此，读取数据支持可以将换行符写为`\n`或`\`的格式，或者换行。例如，字符串 `Hello world` 在单词之间换行而不是空格可以解析为以下任何形式：
+字符串以反斜杠转义的特殊字符输出。下面的转义序列用于输出:`\b`, `\f`, `\r`, `\n`, `\t`, `\0`, `\'`, `\\`。解析还支持`\a`、`\v`和`\xHH`(十六进制转义字符)和任何`\c`字符，其中`c`是任何字符(这些序列被转换为`c`)。因此，读取数据支持这样一种格式，即可以将换行符写成`\n`或`\`，或者写成换行符。例如，字符串`Hello world`在单词之间有换行符，而不是空格，可以用以下语法进行解析:
 
-    Hello\nworld
+``` text
+Hello\nworld
 
-    Hello\
-    world
+Hello\
+world
+```
 
-第二种形式是支持的，因为 MySQL 读取 tab-separated 格式数据集的时候也会使用它。
+第二种形式是支持的，因为MySQL读取tab-separated格式数据集的时候也会使用它。
 
-在 TabSeparated 格式中传递数据时需要转义的最小字符集为：Tab，换行符（LF）和反斜杠。
+在TabSeparated分隔格式传递数据时需要转义的最小字符集:`Tab`、换行符(LF)和反斜杠。
 
-只有一小组符号会被转义。你可以轻易地找到一个字符串值，但这不会正常在你的终端显示。
+只有一小部分符号被转义。您可以很容易地找到一个字符串值，而您的终端将在输出中不显示它。
 
-数组写在方括号内的逗号分隔值列表中。 通常情况下，数组中的数字项目会被拼凑，但日期，带时间的日期以及字符串将使用与上面相同的转义规则用单引号引起来。
+数组写在方括号内的逗号分隔值列表中。数组中的数字项按正常格式进行格式化。`Date`和`DateTime`类型用单引号表示。字符串使用与上面相同的转义规则在单引号中编写。
 
-[NULL](../sql-reference/syntax.md) 将输出为 `\N`。
+[NULL](../sql-reference/syntax.md)将输出为`\N`。
+
+[Nested](../sql-reference/data-types/nested-data-structures/nested.md)结构的每个元素都表示为数组。
+
+示例：
+
+``` sql
+CREATE TABLE nestedt
+(
+    `id` UInt8,
+    `aux` Nested(
+        a UInt8,
+        b String
+    )
+)
+ENGINE = TinyLog
+```
+
+``` sql
+INSERT INTO nestedt Values ( 1, [1], ['a'])
+```
+
+``` sql
+SELECT * FROM nestedt FORMAT TSV
+```
+
+``` text
+1  [1]    ['a']
+```
 
 ## TabSeparatedRaw {#tabseparatedraw}
 
-与 `TabSeparated` 格式不一样的是，行数据是不会被转义的。
-该格式仅适用于输出查询结果，但不适用于解析输入（将数据插入到表中）。
+与`TabSeparated`格式的不同之处在于，写入的行没有转义。
+使用这种格式解析时，每个字段中不允许使用制表符或换行符。
 
-这种格式也可以使用名称 `TSVRaw` 来表示。
+这种格式也可以使用名称`TSVRaw`来表示。
 
 ## TabSeparatedWithNames {#tabseparatedwithnames}
 
-与 `TabSeparated` 格式不一样的是，第一行会显示列的名称。
-在解析过程中，第一行完全被忽略。您不能使用列名来确定其位置或检查其正确性。
-（未来可能会加入解析头行的功能）
+与`TabSeparated`格式不同的是列名写在第一行。
+在解析过程中，第一行被完全忽略。不能使用列名来确定它们的位置或检查它们的正确性。
+(将来可能会添加对头行解析的支持。)
 
-这种格式也可以使用名称 `TSVWithNames` 来表示。
+这种格式也可以使用名称`TSVWithNames`来表示。
 
 ## TabSeparatedWithNamesAndTypes {#tabseparatedwithnamesandtypes}
 
-与 `TabSeparated` 格式不一样的是，第一行会显示列的名称，第二行会显示列的类型。
-在解析过程中，第一行和第二行完全被忽略。
+与`TabSeparated`格式不同的是列名写在第一行，而列类型写在第二行。
+在解析过程中，将完全忽略第一行和第二行。
 
-这种格式也可以使用名称 `TSVWithNamesAndTypes` 来表示。
+这种格式也可以使用名称`TSVWithNamesAndTypes`来表示。
 
-## 模板 {#format-template}
+## Template {#format-template}
 
-此格式允许为具有指定转义规则的值指定带有占位符的自定义格式字符串。
+此格式允许指定带有占位符的自定义格式字符串，这些占位符用于指定转义规则。
 
-它使用设置 `format_schema`, `format_schema_rows`, `format_schema_rows_between_delimiter` and some settings of other formats (e.g. `output_format_json_quote_64bit_integers` 使用时 `JSON` 逃跑，进一步查看)
+它使用设置`format_schema`, `format_schema_rows`, `format_schema_rows_between_delimiter`以及其他格式的一些设置(例如转义`JSON`时使用`output_format_json_quote_64bit_integers`)
 
-格式字符串 `format_schema_rows` 使用以下语法指定行格式:
+设置`format_template_row`指定文件的路径，该文件包含以下语法的行格式字符串:
 
 `delimiter_1${column_1:serializeAs_1}delimiter_2${column_2:serializeAs_2} ... delimiter_N`,
 
-    where `delimiter_i` is a delimiter between values (`$` symbol can be escaped as `$$`),
-    `column_i` is a name of a column whose values are to be selected or inserted (if empty, then column will be skipped),
-    `serializeAs_i` is an escaping rule for the column values. The following escaping rules are supported:
+其中，`delimiter_i`是值之间的分隔符(`$`符号可以转义为`$$`)，
+`column_i`是要选择或插入其值的列的名称或索引(如果为空，则跳过该列)，
+`serializeAs_i`是列值的转义规则。支持以下转义规则:
 
-    - `CSV`, `JSON`, `XML` (similarly to the formats of the same names)
-    - `Escaped` (similarly to `TSV`)
-    - `Quoted` (similarly to `Values`)
-    - `Raw` (without escaping, similarly to `TSVRaw`)
-    - `None` (no escaping rule, see further)
+-   `CSV`, `JSON`, `XML` (类似于相同名称的格式)
+-   `Escaped` (类似于`TSV`)
+-   `Quoted` (类似于`Values`)
+-   `Raw` (类似于`TSVRaw`)
+-   `None`
 
-    If escaping rule is omitted, then`None` will be used. `XML` and `Raw` are suitable only for output.
+如果省略了转义规则，那么将使用`None`。`XML`和`Raw`只适用于输出。
 
-    So, for the following format string:
+对于下面的格式字符串:
 
       `Search phrase: ${SearchPhrase:Quoted}, count: ${c:Escaped}, ad price: $$${price:JSON};`
 
-    the values of `SearchPhrase`, `c` and `price` columns, which are escaped as `Quoted`, `Escaped` and `JSON` will be printed (for select) or will be expected (for insert) between `Search phrase: `, `, count: `, `, ad price: $` and `;` delimiters respectively. For example:
+`SearchPhrase`、`c`和`price`列的值被转义为`quotation`、`Escaped`和`JSON`将分别在`Search phrase:`， `， count: `， `， ad price: $`和`;`分隔符之间打印(用于选择)或expected(用于插入)。例如:
 
-    `Search phrase: 'bathroom interior design', count: 2166, ad price: $3;`
+`Search phrase: 'bathroom interior design', count: 2166, ad price: $3;`
 
-该 `format_schema_rows_between_delimiter` setting指定行之间的分隔符，该分隔符在除最后一行之外的每一行之后打印（或预期） (`\n` 默认情况下)
+`format_template_rows_between_delimiter`设置指定行之间的分隔符，它将打印(或expected)在每一行之后，最后一行除外(默认为`\n`)。
 
-格式字符串 `format_schema` 具有相同的语法 `format_schema_rows` 并允许指定前缀，后缀和打印一些附加信息的方式。 它包含以下占位符而不是列名:
+设置`format_template_resultset`指定文件路径，该文件包含resultset的格式字符串。resultset的格式字符串与row的格式字符串具有相同的语法，允许指定前缀、后缀和打印一些附加信息的方法。它包含以下占位符而不是列名:
 
--   `data` 包含数据的行 `format_schema_rows` 格式，由分隔 `format_schema_rows_between_delimiter`. 此占位符必须是格式字符串中的第一个占位符。
--   `totals` 是包含总值的行 `format_schema_rows` 格式（与总计一起使用时)
--   `min` 是具有最小值的行 `format_schema_rows` 格式（当极值设置为1时)
--   `max` 是具有最大值的行 `format_schema_rows` 格式（当极值设置为1时)
--   `rows` 输出行总数
--   `rows_before_limit` 是没有限制的最小行数。 仅当查询包含LIMIT时输出。 如果查询包含GROUP BY，则rows_before_limit_at_least是没有限制的确切行数。
--   `time` 请求执行时间以秒为单位
+-   `data` `format_template_row`格式的数据行，由`format_template_rows_between_delimiter`分隔。此占位符必须是格式字符串中的第一个占位符。
+-   `totals` `format_template_row`格式的总值(和WITH TOTALS一起使用)
+-   `min` `format_template_row`格式的最小值(当极值设置为1时)
+-   `max` `format_template_row`格式的最大值(当极值设置为1时)
+-   `rows` 输出行的总数
+-   `rows_before_limit` 没有LIMIT的最小行数。仅当查询包含LIMIT时输出。如果查询包含GROUP BY，那么rows_before_limit_at_least就是没有LIMIT的确切行数。
+-   `time` 请求执行时间（秒）
 -   `rows_read` 已读取的行数
--   `bytes_read` 被读取的字节数（未压缩）
+-   `bytes_read` 已读取（未压缩）的字节数
 
-占位符 `data`, `totals`, `min` 和 `max` 必须没有指定转义规则（或 `None` 必须明确指定）。 其余的占位符可能具有指定的任何转义规则。
-如果 `format_schema` 设置为空字符串, `${data}` 用作默认值。
-对于插入查询格式允许跳过一些列或一些字段，如果前缀或后缀（见示例）。
+占位符`data`、`totals`、`min`和`max`必须没有指定转义规则(或者必须显式指定`None`)。其余占位符可以指定任何转义规则。
+如果`format_template_resultset`设置为空字符串，则使用`${data}`作为默认值。
+对于insert查询，格式允许跳过某些列或某些字段的前缀或后缀(参见示例)。
 
-`Select` 示例:
+Select示例:
 
 ``` sql
-SELECT SearchPhrase, count() AS c FROM test.hits GROUP BY SearchPhrase ORDER BY c DESC LIMIT 5
-FORMAT Template
-SETTINGS format_schema = '<!DOCTYPE HTML>
+SELECT SearchPhrase, count() AS c FROM test.hits GROUP BY SearchPhrase ORDER BY c DESC LIMIT 5 FORMAT Template SETTINGS
+format_template_resultset = '/some/path/resultset.format', format_template_row = '/some/path/row.format', format_template_rows_between_delimiter = '\n    '
+```
+
+`/some/path/resultset.format`:
+
+``` text
+<!DOCTYPE HTML>
 <html> <head> <title>Search phrases</title> </head>
  <body>
   <table border="1"> <caption>Search phrases</caption>
@@ -185,11 +243,17 @@ SETTINGS format_schema = '<!DOCTYPE HTML>
   </table>
   <b>Processed ${rows_read:XML} rows in ${time:XML} sec</b>
  </body>
-</html>',
-format_schema_rows = '<tr> <td>${SearchPhrase:XML}</td> <td>${с:XML}</td> </tr>',
-format_schema_rows_between_delimiter = '\n    '
+</html>
 ```
 
+`/some/path/row.format`:
+
+``` text
+<tr> <td>${0:XML}</td> <td>${1:XML}</td> </tr>
+```
+
+结果：
+
 ``` html
 <!DOCTYPE HTML>
 <html> <head> <title>Search phrases</title> </head>
@@ -210,92 +274,122 @@ format_schema_rows_between_delimiter = '\n    '
 </html>
 ```
 
-`Insert` 示例:
+Insert示例：
 
-    Some header
-    Page views: 5, User id: 4324182021466249494, Useless field: hello, Duration: 146, Sign: -1
-    Page views: 6, User id: 4324182021466249494, Useless field: world, Duration: 185, Sign: 1
-    Total rows: 2
+``` text
+Some header
+Page views: 5, User id: 4324182021466249494, Useless field: hello, Duration: 146, Sign: -1
+Page views: 6, User id: 4324182021466249494, Useless field: world, Duration: 185, Sign: 1
+Total rows: 2
+```
 
 ``` sql
 INSERT INTO UserActivity FORMAT Template SETTINGS
-format_schema = 'Some header\n${data}\nTotal rows: ${:CSV}\n',
-format_schema_rows = 'Page views: ${PageViews:CSV}, User id: ${UserID:CSV}, Useless field: ${:CSV}, Duration: ${Duration:CSV}, Sign: ${Sign:CSV}'
+format_template_resultset = '/some/path/resultset.format', format_template_row = '/some/path/row.format'
 ```
 
-`PageViews`, `UserID`, `Duration` 和 `Sign` 占位符内部是表中列的名称。 值后 `Useless field` 在行和之后 `\nTotal rows:` in后缀将被忽略。
+`/some/path/resultset.format`:
+
+``` text
+Some header\n${data}\nTotal rows: ${:CSV}\n
+```
+
+`/some/path/row.format`:
+
+``` text
+Page views: ${PageViews:CSV}, User id: ${UserID:CSV}, Useless field: ${:CSV}, Duration: ${Duration:CSV}, Sign: ${Sign:CSV}
+```
+
+`PageViews`, `UserID`, `Duration`和`Sign` 内部占位符是表中列的名称。将忽略行中`Useless field`后面和后缀中`\nTotal rows:`之后的值。
 输入数据中的所有分隔符必须严格等于指定格式字符串中的分隔符。
 
 ## TemplateIgnoreSpaces {#templateignorespaces}
 
-此格式仅适用于输入。
-类似于 `Template`，但跳过输入流中的分隔符和值之间的空格字符。 但是，如果格式字符串包含空格字符，则在输入流中将需要这些字符。 还允许指定空占位符 (`${}` 或 `${:None}`）将一些分隔符分成单独的部分，以忽略它们之间的空格。 此类占位符仅用于跳过空格字符。
-可以阅读 `JSON` 如果列的值在所有行中具有相同的顺序，则使用此格式。 例如，以下请求可用于从格式的输出示例中插入数据 [JSON](#json):
+这种格式只适用于输入。
+类似于`Template`，但跳过输入流中分隔符和值之间的空白字符。但是，如果格式字符串包含空格字符，这些字符将会出现在输入流中。还允许指定空占位符(`${}`或`${:None}`)来将一些分隔符分割为单独的部分，以忽略它们之间的空格。这种占位符仅用于跳过空白字符。
+如果列的值在所有行的顺序相同，那么可以使用这种格式读取`JSON`。可以使用以下请求从格式为[JSON](#json)的输出示例中插入数据：
 
 ``` sql
 INSERT INTO table_name FORMAT TemplateIgnoreSpaces SETTINGS
-format_schema = '{${}"meta"${}:${:JSON},${}"data"${}:${}[${data}]${},${}"totals"${}:${:JSON},${}"extremes"${}:${:JSON},${}"rows"${}:${:JSON},${}"rows_before_limit_at_least"${}:${:JSON}${}}',
-format_schema_rows = '{${}"SearchPhrase"${}:${}${phrase:JSON}${},${}"c"${}:${}${cnt:JSON}${}}',
-format_schema_rows_between_delimiter = ','
+format_template_resultset = '/some/path/resultset.format', format_template_row = '/some/path/row.format', format_template_rows_between_delimiter = ','
+```
+
+`/some/path/resultset.format`:
+
+``` text
+{${}"meta"${}:${:JSON},${}"data"${}:${}[${data}]${},${}"totals"${}:${:JSON},${}"extremes"${}:${:JSON},${}"rows"${}:${:JSON},${}"rows_before_limit_at_least"${}:${:JSON}${}}
+```
+
+`/some/path/row.format`:
+
+``` text
+{${}"SearchPhrase"${}:${}${phrase:JSON}${},${}"c"${}:${}${cnt:JSON}${}}
 ```
 
 ## TSKV {#tskv}
 
-与 `TabSeparated` 格式类似，但它输出的是 `name=value` 的格式。名称会和 `TabSeparated` 格式一样被转义，`=` 字符也会被转义。
+类似于TabSeparated，但是输出的值是name=value格式。名称的转义方式与TabSeparated格式相同，=符号也是转义的。
 
-    SearchPhrase=   count()=8267016
-    SearchPhrase=bathroom interior design    count()=2166
-    SearchPhrase=yandex     count()=1655
-    SearchPhrase=2014 spring fashion    count()=1549
-    SearchPhrase=freeform photos       count()=1480
-    SearchPhrase=angelina jolie    count()=1245
-    SearchPhrase=omsk       count()=1112
-    SearchPhrase=photos of dog breeds    count()=1091
-    SearchPhrase=curtain designs        count()=1064
-    SearchPhrase=baku       count()=1000
+``` text
+SearchPhrase=   count()=8267016
+SearchPhrase=bathroom interior design    count()=2166
+SearchPhrase=yandex     count()=1655
+SearchPhrase=2014 spring fashion    count()=1549
+SearchPhrase=freeform photos       count()=1480
+SearchPhrase=angelina jolie    count()=1245
+SearchPhrase=omsk       count()=1112
+SearchPhrase=photos of dog breeds    count()=1091
+SearchPhrase=curtain designs        count()=1064
+SearchPhrase=baku       count()=1000
+```
 
-[NULL](../sql-reference/syntax.md) 输出为 `\N`。
+[NULL](../sql-reference/syntax.md)格式为`\N`。
 
 ``` sql
 SELECT * FROM t_null FORMAT TSKV
 ```
 
-    x=1 y=\N
+``` text
+x=1    y=\N
+```
 
-当有大量的小列时，这种格式是低效的，通常没有理由使用它。它被用于 Yandex 公司的一些部门。
+当有大量的小列时，这种格式是无效的，并且通常没有理由使用它。不过，就效率而言，它并不比JSONEachRow差。
+这种格式支持数据输出和解析。对于解析，不同列的值支持任何顺序。省略某些值是可以接受的——它们被视为与其默认值相等。在这种情况下，0和空白行被用作默认值。不支持在表中指定的复杂值作为缺省值。
 
-数据的输出和解析都支持这种格式。对于解析，任何顺序都支持不同列的值。可以省略某些值，用 `-` 表示， 它们被视为等于它们的默认值。在这种情况下，零和空行被用作默认值。作为默认值，不支持表中指定的复杂值。
-
-对于不带等号或值，可以用附加字段 `tskv` 来表示，这种在解析上是被允许的。这样的话该字段被忽略。
+解析允许存在不带等号或值的附加字段`tskv`。此字段被忽略。
 
 ## CSV {#csv}
 
-按逗号分隔的数据格式([RFC](https://tools.ietf.org/html/rfc4180))。
+按`,`分隔的数据格式([RFC](https://tools.ietf.org/html/rfc4180))。
 
-格式化的时候，行是用双引号括起来的。字符串中的双引号会以两个双引号输出，除此之外没有其他规则来做字符转义了。日期和时间也会以双引号包括。数字的输出不带引号。值由一个单独的字符隔开，这个字符默认是 `,`。行使用 Unix 换行符（LF）分隔。 数组序列化成 CSV 规则如下：首先将数组序列化为 TabSeparated 格式的字符串，然后将结果字符串用双引号包括输出到 CSV。CSV 格式的元组被序列化为单独的列（即它们在元组中的嵌套关系会丢失）。
+格式化时，行是用双引号括起来的。字符串中的双引号会以两个双引号输出，除此之外没有其他规则来做字符转义了。日期和时间也会以双引号包括。数字的输出不带引号。值由一个单独的字符隔开，这个字符默认是`,`。行使用Unix换行符（LF）分隔。数组序列化成CSV规则如下：首先将数组序列化为`TabSeparated`格式的字符串，然后将结果字符串用双引号包括输出到`CSV`。`CSV`格式的元组被序列化为单独的列（即它们在元组中的嵌套关系会丢失）。
 
-    clickhouse-client --format_csv_delimiter="|" --query="INSERT INTO test.csv FORMAT CSV" < data.csv
+``` bash
+$ clickhouse-client --format_csv_delimiter="|" --query="INSERT INTO test.csv FORMAT CSV" < data.csv
+```
 
-\*默认情况下间隔符是 `,` ，在 [format_csv_delimiter](../operations/settings/settings.md#settings-format_csv_delimiter) 中可以了解更多间隔符配置。
+\* 默认情况下间隔符是`,` ，在[format_csv_delimiter](../operations/settings/settings.md#settings-format_csv_delimiter)中可以了解更多分隔符配置。
 
-解析的时候，可以使用或不使用引号来解析所有值。支持双引号和单引号。行也可以不用引号排列。 在这种情况下，它们被解析为逗号或换行符（CR 或 LF）。在解析不带引号的行时，若违反 RFC 规则，会忽略前导和尾随的空格和制表符。 对于换行，全部支持 Unix（LF），Windows（CR LF）和 Mac OS Classic（CR LF）。
+解析的时候，可以使用或不使用引号来解析所有值。支持双引号和单引号。行也可以不用引号排列。在这种情况下，它们被解析为逗号或换行符（`CR或`LF`）。在解析不带引号的行时，若违反`RFC`规则，会忽略前缀和结尾的空格和制表符。对于换行，全部支持Unix（LF），Windows（CR LF）和Mac OS Classic（CR LF）。
 
-`NULL` 将输出为 `\N`。
+如果启用[input_format_defaults_for_omitted_fields](../operations/settings/settings.md#session_settings-input_format_defaults_for_omitted_fields)，空的末尾加引号的输入值将替换为相应列的默认值。
 
-CSV 格式是和 TabSeparated 一样的方式输出总数和极值。
+`NULL`被格式化为`\N`或`NULL`或一个空的非引号字符串(详见配置[input_format_csv_unquoted_null_literal_as_null](../operations/settings/settings.md#settings-input_format_csv_unquoted_null_literal_as_null)或[input_format_defaults_for_omitted_fields](../operations/settings/settings.md#session_settings-input_format_defaults_for_omitted_fields))。
+
+`CSV`格式支持输出总数和极值的方式与`TabSeparated`相同。
 
 ## CSVWithNames {#csvwithnames}
 
-会输出带头部行，和 `TabSeparatedWithNames` 一样。
+会输出带头部的信息(字段列表)，和`TabSeparatedWithNames`一样。
 
-## 自定义分离 {#format-customseparated}
+## CustomSeparated {#format-customseparated}
 
-类似于 [模板](#format-template)，但它打印或读取所有列，并使用从设置转义规则 `format_custom_escaping_rule` 从设置和分隔符 `format_custom_field_delimiter`, `format_custom_row_before_delimiter`, `format_custom_row_after_delimiter`, `format_custom_row_between_delimiter`, `format_custom_result_before_delimiter` 和 `format_custom_result_after_delimiter`，而不是从格式字符串。
-也有 `CustomSeparatedIgnoreSpaces` 格式，这是类似于 `TemplateIgnoreSpaces`.
+类似于[Template](#format-template)， 但它打印或读取所有列和使用转义规则在设置`format_custom_escaping_rule`和分隔符设置`format_custom_field_delimiter`,`format_custom_row_before_delimiter`,`format_custom_row_after_delimiter`,`format_custom_row_between_delimiter`,`format_custom_result_before_delimiter`,`format_custom_result_after_delimiter`中,而不是从格式字符串。
+也有`CustomSeparatedIgnoreSpaces`格式，这是类似于`TemplateIgnoreSpaces`。
 
 ## JSON {#json}
 
-以 JSON 格式输出数据。除了数据表之外，它还输出列名称和类型以及一些附加信息：输出行的总数以及在没有 LIMIT 时可以输出的行数。 例：
+以JSON格式输出数据。除了数据表之外，它还输出列名和类型，以及一些附加信息: 输出行的总数，以及如果没有LIMIT的话可输出的行数。示例:
 
 ``` sql
 SELECT SearchPhrase, count() AS c FROM test.hits GROUP BY SearchPhrase WITH TOTALS ORDER BY c DESC LIMIT 5 FORMAT JSON
@@ -306,166 +400,326 @@ SELECT SearchPhrase, count() AS c FROM test.hits GROUP BY SearchPhrase WITH TOTA
         "meta":
         [
                 {
-                        "name": "SearchPhrase",
+                        "name": "'hello'",
                         "type": "String"
                 },
                 {
-                        "name": "c",
+                        "name": "multiply(42, number)",
                         "type": "UInt64"
+                },
+                {
+                        "name": "range(5)",
+                        "type": "Array(UInt8)"
                 }
         ],
 
         "data":
         [
                 {
-                        "SearchPhrase": "",
-                        "c": "8267016"
+                        "'hello'": "hello",
+                        "multiply(42, number)": "0",
+                        "range(5)": [0,1,2,3,4]
                 },
                 {
-                        "SearchPhrase": "bathroom interior design",
-                        "c": "2166"
+                        "'hello'": "hello",
+                        "multiply(42, number)": "42",
+                        "range(5)": [0,1,2,3,4]
                 },
                 {
-                        "SearchPhrase": "yandex",
-                        "c": "1655"
-                },
-                {
-                        "SearchPhrase": "spring 2014 fashion",
-                        "c": "1549"
-                },
-                {
-                        "SearchPhrase": "freeform photos",
-                        "c": "1480"
+                        "'hello'": "hello",
+                        "multiply(42, number)": "84",
+                        "range(5)": [0,1,2,3,4]
                 }
         ],
 
-        "totals":
-        {
-                "SearchPhrase": "",
-                "c": "8873898"
-        },
+        "rows": 3,
 
-        "extremes":
-        {
-                "min":
-                {
-                        "SearchPhrase": "",
-                        "c": "1480"
-                },
-                "max":
-                {
-                        "SearchPhrase": "",
-                        "c": "8267016"
-                }
-        },
-
-        "rows": 5,
-
-        "rows_before_limit_at_least": 141137
+        "rows_before_limit_at_least": 3
 }
 ```
 
-JSON 与 JavaScript 兼容。为了确保这一点，一些字符被另外转义：斜线`/`被转义为`\/`; 替代的换行符 `U+2028` 和 `U+2029` 会打断一些浏览器解析，它们会被转义为 `\uXXXX`。 ASCII 控制字符被转义：退格，换页，换行，回车和水平制表符被替换为`\b`，`\f`，`\n`，`\r`，`\t` 作为使用`\uXXXX`序列的00-1F范围内的剩余字节。 无效的 UTF-8 序列更改为替换字符 ，因此输出文本将包含有效的 UTF-8 序列。 为了与 JavaScript 兼容，默认情况下，Int64 和 UInt64 整数用双引号引起来。要除去引号，可以将配置参数 output_format_json_quote_64bit_integers 设置为0。
+JSON与JavaScript兼容。为了确保这一点，一些字符被另外转义：斜线`/`被转义为`\/`; 替代的换行符`U+2028`和`U+2029`会打断一些浏览器解析，它们会被转义为`\uXXXX`。 ASCII控制字符被转义：退格，换页，换行，回车和水平制表符被替换为`\b`，`\f`，`\n`，`\r`，`\t` 作为使用`\uXXXX`序列的00-1F范围内的剩余字节。 无效的UTF-8序列更改为替换字符，因此输出文本将包含有效的UTF-8序列。 为了与JavaScript兼容，默认情况下，Int64和UInt64整数用双引号引起来。要除去引号，可以将配置参数`output_format_json_quote_64bit_integers`设置为0。
 
 `rows` – 结果输出的行数。
 
-`rows_before_limit_at_least` 去掉 LIMIT 过滤后的最小行总数。 只会在查询包含 LIMIT 条件时输出。
-若查询包含 GROUP BY，rows_before_limit_at_least 就是去掉 LIMIT 后过滤后的准确行数。
+`rows_before_limit_at_least`去掉 LIMIT过滤后的最小行总数。 只会在查询包含LIMIT条件时输出。
+若查询包含 GROUP BY，`rows_before_limit_at_least`就是去掉LIMIT后过滤后的准确行数。
 
-`totals` – 总值 （当使用 TOTALS 条件时）。
+`totals` – 总值 （当使用TOTALS条件时）。
 
-`extremes` – 极值 （当 extremes 设置为 1时）。
+`extremes` – 极值（当extremes设置为1时）。
 
 该格式仅适用于输出查询结果，但不适用于解析输入（将数据插入到表中）。
 
-ClickHouse 支持 [NULL](../sql-reference/syntax.md), 在 JSON 格式中以 `null` 输出来表示.
+ClickHouse支持[NULL](../sql-reference/syntax.md), 在JSON输出中显示为`null`。若要在输出中启用`+nan`、`-nan`、`+inf`、`-inf`值，请设置[output_format_json_quote_denormals](../operations/settings/settings.md#settings-output_format_json_quote_denormals)为1。
 
-参考 JSONEachRow 格式。
+**参考**
 
-## JSONCompact {#jsoncompact}
+-   [JSONEachRow](#jsoneachrow)格式
+-   [output_format_json_array_of_rows](../operations/settings/settings.md#output-format-json-array-of-rows)设置
 
-与 JSON 格式不同的是它以数组的方式输出结果，而不是以结构体。
+## JSONString {#jsonstring}
+
+与JSON的不同之处在于数据字段以字符串输出，而不是以类型化JSON值输出。
 
 示例：
 
-``` json
+```json
 {
         "meta":
         [
                 {
-                        "name": "SearchPhrase",
+                        "name": "'hello'",
                         "type": "String"
                 },
                 {
-                        "name": "c",
+                        "name": "multiply(42, number)",
                         "type": "UInt64"
+                },
+                {
+                        "name": "range(5)",
+                        "type": "Array(UInt8)"
                 }
         ],
 
         "data":
         [
-                ["", "8267016"],
-                ["bathroom interior design", "2166"],
-                ["yandex", "1655"],
-                ["fashion trends spring 2014", "1549"],
-                ["freeform photo", "1480"]
+                {
+                        "'hello'": "hello",
+                        "multiply(42, number)": "0",
+                        "range(5)": "[0,1,2,3,4]"
+                },
+                {
+                        "'hello'": "hello",
+                        "multiply(42, number)": "42",
+                        "range(5)": "[0,1,2,3,4]"
+                },
+                {
+                        "'hello'": "hello",
+                        "multiply(42, number)": "84",
+                        "range(5)": "[0,1,2,3,4]"
+                }
         ],
 
-        "totals": ["","8873898"],
+        "rows": 3,
 
-        "extremes":
-        {
-                "min": ["","1480"],
-                "max": ["","8267016"]
-        },
-
-        "rows": 5,
-
-        "rows_before_limit_at_least": 141137
+        "rows_before_limit_at_least": 3
 }
 ```
 
-这种格式仅仅适用于输出结果集，而不适用于解析（将数据插入到表中）。
-参考 `JSONEachRow` 格式。
+## JSONAsString {#jsonasstring}
 
-## JSONEachRow {#jsoneachrow}
+在这种格式中，一个JSON对象被解释为一个值。如果输入有几个JSON对象(逗号分隔)，它们将被解释为独立的行。
 
-将数据结果每一行以 JSON 结构体输出（换行分割 JSON 结构体）。
+这种格式只能对具有单个字段类型的表进行解析[String](../sql-reference/data-types/string.md)。其余的列必须设置为[DEFAULT](../sql-reference/statements/create/table.md#default)或[MATERIALIZED](../sql-reference/statements/create/table.md#materialized)，或者忽略。一旦将整个JSON对象收集为字符串，就可以使用[JSON函数](../sql-reference/functions/json-functions.md)运行它。
 
-``` json
-{"SearchPhrase":"","count()":"8267016"}
-{"SearchPhrase": "bathroom interior design","count()": "2166"}
-{"SearchPhrase":"yandex","count()":"1655"}
-{"SearchPhrase":"2014 spring fashion","count()":"1549"}
-{"SearchPhrase":"freeform photo","count()":"1480"}
-{"SearchPhrase":"angelina jolie","count()":"1245"}
-{"SearchPhrase":"omsk","count()":"1112"}
-{"SearchPhrase":"photos of dog breeds","count()":"1091"}
-{"SearchPhrase":"curtain designs","count()":"1064"}
-{"SearchPhrase":"baku","count()":"1000"}
+**示例**
+
+查询：
+
+``` sql
+DROP TABLE IF EXISTS json_as_string;
+CREATE TABLE json_as_string (json String) ENGINE = Memory;
+INSERT INTO json_as_string FORMAT JSONAsString {"foo":{"bar":{"x":"y"},"baz":1}},{},{"any json stucture":1}
+SELECT * FROM json_as_string;
 ```
 
-与 JSON 格式不同的是，没有替换无效的UTF-8序列。任何一组字节都可以在行中输出。这是必要的，因为这样数据可以被格式化而不会丢失任何信息。值的转义方式与JSON相同。
+结果：
 
-对于解析，任何顺序都支持不同列的值。可以省略某些值 - 它们被视为等于它们的默认值。在这种情况下，零和空行被用作默认值。 作为默认值，不支持表中指定的复杂值。元素之间的空白字符被忽略。如果在对象之后放置逗号，它将被忽略。对象不一定必须用新行分隔。
+``` text
+┌─json──────────────────────────────┐
+│ {"foo":{"bar":{"x":"y"},"baz":1}} │
+│ {}                                │
+│ {"any json stucture":1}           │
+└───────────────────────────────────┘
+```
 
-### 嵌套结构的使用 {#jsoneachrow-nested}
+## JSONCompact {#jsoncompact}
+## JSONCompactString {#jsoncompactstring}
 
-如果你有一张桌子 [嵌套式](../sql-reference/data-types/nested-data-structures/nested.md) 数据类型列，可以插入具有相同结构的JSON数据。 启用此功能与 [input_format_import_nested_json](../operations/settings/settings.md#settings-input_format_import_nested_json) 设置。
+与JSON格式不同的是它以数组的方式输出结果，而不是以结构体。
 
-例如，请考虑下表:
+示例：
+
+``` json
+// JSONCompact
+{
+        "meta":
+        [
+                {
+                        "name": "'hello'",
+                        "type": "String"
+                },
+                {
+                        "name": "multiply(42, number)",
+                        "type": "UInt64"
+                },
+                {
+                        "name": "range(5)",
+                        "type": "Array(UInt8)"
+                }
+        ],
+
+        "data":
+        [
+                ["hello", "0", [0,1,2,3,4]],
+                ["hello", "42", [0,1,2,3,4]],
+                ["hello", "84", [0,1,2,3,4]]
+        ],
+
+        "rows": 3,
+
+        "rows_before_limit_at_least": 3
+}
+```
+
+```json
+// JSONCompactString
+{
+        "meta":
+        [
+                {
+                        "name": "'hello'",
+                        "type": "String"
+                },
+                {
+                        "name": "multiply(42, number)",
+                        "type": "UInt64"
+                },
+                {
+                        "name": "range(5)",
+                        "type": "Array(UInt8)"
+                }
+        ],
+
+        "data":
+        [
+                ["hello", "0", "[0,1,2,3,4]"],
+                ["hello", "42", "[0,1,2,3,4]"],
+                ["hello", "84", "[0,1,2,3,4]"]
+        ],
+
+        "rows": 3,
+
+        "rows_before_limit_at_least": 3
+}
+```
+
+## JSONEachRow {#jsoneachrow}
+## JSONStringEachRow {#jsonstringeachrow}
+## JSONCompactEachRow {#jsoncompacteachrow}
+## JSONCompactStringEachRow {#jsoncompactstringeachrow}
+
+使用这些格式时，ClickHouse会将行输出为分隔的、换行分隔的JSON值，但数据作为一个整体不是有效的JSON。
+
+``` json
+{"some_int":42,"some_str":"hello","some_tuple":[1,"a"]} // JSONEachRow
+[42,"hello",[1,"a"]] // JSONCompactEachRow
+["42","hello","(2,'a')"] // JSONCompactStringsEachRow
+```
+
+在插入数据时，应该为每一行提供一个单独的JSON值。
+
+## JSONEachRowWithProgress {#jsoneachrowwithprogress}
+## JSONStringEachRowWithProgress {#jsonstringeachrowwithprogress}
+
+与`JSONEachRow`/`JSONStringEachRow`不同的是，ClickHouse还将生成作为JSON值的进度信息。
+
+```json
+{"row":{"'hello'":"hello","multiply(42, number)":"0","range(5)":[0,1,2,3,4]}}
+{"row":{"'hello'":"hello","multiply(42, number)":"42","range(5)":[0,1,2,3,4]}}
+{"row":{"'hello'":"hello","multiply(42, number)":"84","range(5)":[0,1,2,3,4]}}
+{"progress":{"read_rows":"3","read_bytes":"24","written_rows":"0","written_bytes":"0","total_rows_to_read":"3"}}
+```
+
+## JSONCompactEachRowWithNamesAndTypes {#jsoncompacteachrowwithnamesandtypes}
+## JSONCompactStringEachRowWithNamesAndTypes {#jsoncompactstringeachrowwithnamesandtypes}
+
+与`JSONCompactEachRow`/`JSONCompactStringEachRow`不同的是，其中列名和类型被写入前两行。
+
+```json
+["'hello'", "multiply(42, number)", "range(5)"]
+["String", "UInt64", "Array(UInt8)"]
+["hello", "0", [0,1,2,3,4]]
+["hello", "42", [0,1,2,3,4]]
+["hello", "84", [0,1,2,3,4]]
+```
+
+### Inserting Data {#inserting-data}
+
+``` sql
+INSERT INTO UserActivity FORMAT JSONEachRow {"PageViews":5, "UserID":"4324182021466249494", "Duration":146,"Sign":-1} {"UserID":"4324182021466249494","PageViews":6,"Duration":185,"Sign":1}
+```
+
+ClickHouse允许:
+
+- 对象中key-value的任何顺序。
+- 省略一些值。
+
+ClickHouse忽略元素之间的空格和对象后面的逗号。您可以在一行中传递所有对象。你不需要用换行符把它们分开。
+
+**省略值处理**
+
+ClickHouse将省略的值替换为对应的[data types](../sql-reference/data-types/index.md)默认值。
+
+如果指定了`DEFAULT expr`，则ClickHouse根据属性使用不同的替换规则，详看[input_format_defaults_for_omitted_fields](../operations/settings/settings.md#session_settings-input_format_defaults_for_omitted_fields)设置。
+
+参考下表：
+
+``` sql
+CREATE TABLE IF NOT EXISTS example_table
+(
+    x UInt32,
+    a DEFAULT x * 2
+) ENGINE = Memory;
+```
+
+-   如果`input_format_defaults_for_omitted_fields = 0`, 那么`x`和`a`的默认值等于`0`(作为`UInt32`数据类型的默认值)。
+-   如果`input_format_defaults_for_omitted_fields = 1`, 那么`x`的默认值为`0`，但`a`的默认值为`x * 2`。
+
+!!! note "注意"
+当使用`insert_sample_with_metadata = 1`插入数据时，与使用`insert_sample_with_metadata = 0`插入数据相比，ClickHouse消耗更多的计算资源。
+
+### Selecting Data {#selecting-data}
+
+以`UserActivity`表为例:
+
+``` text
+┌──────────────UserID─┬─PageViews─┬─Duration─┬─Sign─┐
+│ 4324182021466249494 │         5 │      146 │   -1 │
+│ 4324182021466249494 │         6 │      185 │    1 │
+└─────────────────────┴───────────┴──────────┴──────┘
+```
+
+当查询`SELECT * FROM UserActivity FORMAT JSONEachRow`返回:
+
+``` text
+{"UserID":"4324182021466249494","PageViews":5,"Duration":146,"Sign":-1}
+{"UserID":"4324182021466249494","PageViews":6,"Duration":185,"Sign":1}
+```
+
+与[JSON](#json)格式不同，没有替换无效的UTF-8序列。值以与`JSON`相同的方式转义。
+
+!!! note "提示"
+字符串中可以输出任意一组字节。如果您确信表中的数据可以被格式化为JSON而不会丢失任何信息，那么就使用`JSONEachRow`格式。
+
+### Nested Structures {#jsoneachrow-nested}
+
+如果您有一个包含[Nested](../sql-reference/data-types/nested-data-structures/nested.md)数据类型列的表，您可以插入具有相同结构的JSON数据。使用[input_format_import_nested_json](../operations/settings/settings.md#settings-input_format_import_nested_json)设置启用该特性。
+
+例如，请参考下表:
 
 ``` sql
 CREATE TABLE json_each_row_nested (n Nested (s String, i Int32) ) ENGINE = Memory
 ```
 
-正如你可以在找到 `Nested` 数据类型说明，ClickHouse将嵌套结构的每个组件视为单独的列, `n.s` 和 `n.i` 为了我们的桌子 所以你可以通过以下方式插入数据:
+正如您在`Nested`数据类型描述中看到的，ClickHouse将嵌套结构的每个组件作为一个单独的列(`n.s`和`n.i`是我们的表)。您可以通过以下方式插入数据:
 
 ``` sql
 INSERT INTO json_each_row_nested FORMAT JSONEachRow {"n.s": ["abc", "def"], "n.i": [1, 23]}
 ```
 
-将数据作为分层JSON对象集插入 [input_format_import_nested_json=1](../operations/settings/settings.md#settings-input_format_import_nested_json).
+将数据作为分层JSON对象集插入[input_format_import_nested_json=1](../operations/settings/settings.md#settings-input_format_import_nested_json)。
 
 ``` json
 {
@@ -508,74 +762,93 @@ SELECT * FROM json_each_row_nested
 └───────────────┴────────┘
 ```
 
-## 本地人 {#native}
+## Native {#native}
 
-最高性能的格式。 据通过二进制格式的块进行写入和读取。对于每个块，该块中的行数，列数，列名称和类型以及列的部分将被相继记录。 换句话说，这种格式是 «列式»的 - 它不会将列转换为行。 这是用于在服务器之间进行交互的本地界面中使用的格式，用于使用命令行客户端和 C++ 客户端。
+最高性能的格式。通过二进制格式的块进行写入和读取。对于每个块，该中的行数，列数，列名称和类型以及列的部分将被相继记录。 换句话说，这种格式是`columnar`的 - 它不会将列转换为行。这是用于在服务器之间进行交互的本地界面中使用的格式，用于使用命令行客户端和C++客户端。
 
-您可以使用此格式快速生成只能由 ClickHouse DBMS 读取的格式。但自己处理这种格式是没有意义的。
+您可以使用此格式快速生成只能由ClickHouse DBMS读取的格式。但自己处理这种格式是没有意义的。
 
 ## Null {#null}
 
-没有输出。但是，查询已处理完毕，并且在使用命令行客户端时，数据将传输到客户端。这仅用于测试，包括生产力测试。
+没有输出。但是，查询已处理完毕，并且在使用命令行客户端时，数据将传输到客户端。这仅用于测试，包括性能测试。
 显然，这种格式只适用于输出，不适用于解析。
 
-## 漂亮 {#pretty}
+## Pretty {#pretty}
 
-将数据以表格形式输出，也可以使用 ANSI 转义字符在终端中设置颜色。
+将数据以表格形式输出，也可以使用ANSI转义字符在终端中设置颜色。
 它会绘制一个完整的表格，每行数据在终端中占用两行。
-每一个结果块都会以单独的表格输出。这是很有必要的，以便结果块不用缓冲结果输出（缓冲在可以预见结果集宽度的时候是很有必要的）。
+每个结果块作为一个单独的表输出。这是必要的，以便在输出块时不需要缓冲结果(为了预先计算所有值的可见宽度，缓冲是必要的)。
 
-[NULL](../sql-reference/syntax.md) 输出为 `ᴺᵁᴸᴸ`。
+[NULL](../sql-reference/syntax.md)输出为`ᴺᵁᴸᴸ`。
+
+示例(显示[PrettyCompact](#prettycompact)格式)
 
 ``` sql
 SELECT * FROM t_null
 ```
 
-    ┌─x─┬────y─┐
-    │ 1 │ ᴺᵁᴸᴸ │
-    └───┴──────┘
+``` text
+┌─x─┬────y─┐
+│ 1 │ ᴺᵁᴸᴸ │
+└───┴──────┘
+```
 
-为避免将太多数据传输到终端，只打印前10,000行。 如果行数大于或等于10,000，则会显示消息«Showed first 10 000»。
+行没有转义为Pretty\* 格式。示例显示了[PrettyCompact](#prettycompact)格式:
+
+``` sql
+SELECT 'String with \'quotes\' and \t character' AS Escaping_test
+```
+
+``` text
+┌─Escaping_test────────────────────────┐
+│ String with 'quotes' and      character │
+└──────────────────────────────────────┘
+```
+
+为避免将太多数据传输到终端，只打印前10,000行。 如果行数大于或等于10,000，则会显示消息`Showed first 10 000`。
 该格式仅适用于输出查询结果，但不适用于解析输入（将数据插入到表中）。
 
-Pretty格式支持输出总值（当使用 WITH TOTALS 时）和极值（当 `extremes` 设置为1时）。 在这些情况下，总数值和极值在主数据之后以单独的表格形式输出。 示例（以 PrettyCompact 格式显示）：
+Pretty格式支持输出合计值(当使用WITH TOTALS时)和极值(当`extremes`设置为1时)。在这些情况下，合计值和极值将输出在主要数据之后，在单独的表中。示例(显示为[PrettyCompact](#prettycompact)格式):
 
 ``` sql
 SELECT EventDate, count() AS c FROM test.hits GROUP BY EventDate WITH TOTALS ORDER BY EventDate FORMAT PrettyCompact
 ```
 
-    ┌──EventDate─┬───────c─┐
-    │ 2014-03-17 │ 1406958 │
-    │ 2014-03-18 │ 1383658 │
-    │ 2014-03-19 │ 1405797 │
-    │ 2014-03-20 │ 1353623 │
-    │ 2014-03-21 │ 1245779 │
-    │ 2014-03-22 │ 1031592 │
-    │ 2014-03-23 │ 1046491 │
-    └────────────┴─────────┘
+``` text
+┌──EventDate─┬───────c─┐
+│ 2014-03-17 │ 1406958 │
+│ 2014-03-18 │ 1383658 │
+│ 2014-03-19 │ 1405797 │
+│ 2014-03-20 │ 1353623 │
+│ 2014-03-21 │ 1245779 │
+│ 2014-03-22 │ 1031592 │
+│ 2014-03-23 │ 1046491 │
+└────────────┴─────────┘
 
-    Totals:
-    ┌──EventDate─┬───────c─┐
-    │ 1970-01-01 │ 8873898 │
-    └────────────┴─────────┘
+Totals:
+┌──EventDate─┬───────c─┐
+│ 1970-01-01 │ 8873898 │
+└────────────┴─────────┘
 
-    Extremes:
-    ┌──EventDate─┬───────c─┐
-    │ 2014-03-17 │ 1031592 │
-    │ 2014-03-23 │ 1406958 │
-    └────────────┴─────────┘
+Extremes:
+┌──EventDate─┬───────c─┐
+│ 2014-03-17 │ 1031592 │
+│ 2014-03-23 │ 1406958 │
+└────────────┴─────────┘
+```
 
 ## PrettyCompact {#prettycompact}
 
-与 `Pretty` 格式不一样的是，`PrettyCompact` 去掉了行之间的表格分割线，这样使得结果更加紧凑。这种格式会在交互命令行客户端下默认使用。
+与[Pretty](#pretty)格式不一样的是`PrettyCompact`去掉了行之间的表格分割线，这样使得结果更加紧凑。
+这种格式会在交互命令行客户端下默认使用。
 
 ## PrettyCompactMonoBlock {#prettycompactmonoblock}
 
-与 `PrettyCompact` 格式不一样的是，它支持 10,000 行数据缓冲，然后输出在一个表格中，不会按照块来区分
+与[PrettyCompact](#prettycompact)格式不一样的是，它支持10,000行数据缓冲，然后输出在一个表格中，不会按照块来区分。
 
 ## PrettyNoEscapes {#prettynoescapes}
 
-与 `Pretty` 格式不一样的是，它不使用 ANSI 字符转义， 这在浏览器显示数据以及在使用 `watch` 命令行工具是有必要的。
+与`Pretty`格式不一样的是，它不使用ANSI字符转义，这在浏览器显示数据以及在使用`watch`命令行工具是有必要的。
 
 示例：
 
@@ -583,7 +856,7 @@ SELECT EventDate, count() AS c FROM test.hits GROUP BY EventDate WITH TOTALS ORD
 watch -n1 "clickhouse-client --query='SELECT event, value FROM system.events FORMAT PrettyCompactNoEscapes'"
 ```
 
-您可以使用 HTTP 接口来获取数据，显示在浏览器中。
+您可以使用HTTP接口来获取数据，显示在浏览器中。
 
 ### PrettyCompactNoEscapes {#prettycompactnoescapes}
 
@@ -593,9 +866,17 @@ watch -n1 "clickhouse-client --query='SELECT event, value FROM system.events FOR
 
 用法类似上述。
 
+### PrettyCompactNoEscapes {#prettycompactnoescapes}
+
+与前面的设置相同。
+
+### PrettySpaceNoEscapes {#prettyspacenoescapes}
+
+与前面的设置相同。
+
 ## PrettySpace {#prettyspace}
 
-与 `PrettyCompact`(#prettycompact) 格式不一样的是，它使用空格来代替网格来显示数据。
+与[PrettyCompact](#prettycompact)格式不一样的是，它使用空格来代替网格来显示数据。
 
 ## RowBinary {#rowbinary}
 
diff --git a/docs/zh/interfaces/third-party/index.md b/docs/zh/interfaces/third-party/index.md
index b55897073da..40206f6b57b 100644
--- a/docs/zh/interfaces/third-party/index.md
+++ b/docs/zh/interfaces/third-party/index.md
@@ -1,8 +1,16 @@
 ---
-machine_translated: true
-machine_translated_rev: 72537a2d527c63c07aa5d2361a8829f3895cf2bd
-toc_folder_title: "\u7B2C\u4E09\u65B9"
+toc_folder_title: 第三方工具
 toc_priority: 24
 ---
 
+# 第三方工具 {#third-party-interfaces}
 
+这是第三方工具的链接集合，它们提供了一些ClickHouse的接口。它可以是可视化界面、命令行界面或API:
+
+-   [Client libraries](../../interfaces/third-party/client-libraries.md)
+-   [Integrations](../../interfaces/third-party/integrations.md)
+-   [GUI](../../interfaces/third-party/gui.md)
+-   [Proxies](../../interfaces/third-party/proxy.md)
+
+!!! note "注意"
+支持通用API的通用工具[ODBC](../../interfaces/odbc.md)或[JDBC](../../interfaces/jdbc.md)，通常也适用于ClickHouse，但这里没有列出，因为它们实在太多了。

From 1e3951ebc83ac1d066b57e4fbad8c910ea347391 Mon Sep 17 00:00:00 2001
From: qianmoQ <shicheng@ttxit.com>
Date: Mon, 21 Dec 2020 18:39:54 +0800
Subject: [PATCH 215/504] translate documentation for interfaces/thied-party

---
 .../third-party/client-libraries.md           | 65 ++++++++-----
 .../zh/interfaces/third-party/integrations.md | 94 ++++++++++---------
 docs/zh/interfaces/third-party/proxy.md       | 49 +++++-----
 3 files changed, 119 insertions(+), 89 deletions(-)

diff --git a/docs/zh/interfaces/third-party/client-libraries.md b/docs/zh/interfaces/third-party/client-libraries.md
index 77c929b9730..e94eb8bcfc0 100644
--- a/docs/zh/interfaces/third-party/client-libraries.md
+++ b/docs/zh/interfaces/third-party/client-libraries.md
@@ -1,51 +1,66 @@
-# 第三方开发的库 {#di-san-fang-kai-fa-de-ku}
+---
+toc_priority: 26
+toc_title: 客户端开发库
+---
 
-!!! warning "放弃"
-    Yandex不维护下面列出的库，也没有进行任何广泛的测试以确保其质量。
+# 第三方开发库 {#client-libraries-from-third-party-developers}
+
+!!! warning "声明"
+Yandex**没有**维护下面列出的库，也没有做过任何广泛的测试来确保它们的质量。
 
 -   Python
     -   [infi.clickhouse_orm](https://github.com/Infinidat/infi.clickhouse_orm)
-    -   [ﾂ环板driverｮﾂ嘉ｯﾂ偲](https://github.com/mymarilyn/clickhouse-driver)
-    -   [ﾂ环板clientｮﾂ嘉ｯﾂ偲](https://github.com/yurial/clickhouse-client)
+    -   [clickhouse-driver](https://github.com/mymarilyn/clickhouse-driver)
+    -   [clickhouse-client](https://github.com/yurial/clickhouse-client)
+    -   [aiochclient](https://github.com/maximdanilchenko/aiochclient)
 -   PHP
     -   [smi2/phpclickhouse](https://packagist.org/packages/smi2/phpClickHouse)
-    -   [8bitov/clickhouse-php客户端](https://packagist.org/packages/8bitov/clickhouse-php-client)
-    -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人](https://packagist.org/packages/bozerkins/clickhouse-client)
-    -   [ﾂ环板clientｮﾂ嘉ｯﾂ偲](https://packagist.org/packages/simpod/clickhouse-client)
+    -   [8bitov/clickhouse-php-client](https://packagist.org/packages/8bitov/clickhouse-php-client)
+    -   [bozerkins/clickhouse-client](https://packagist.org/packages/bozerkins/clickhouse-client)
+    -   [simpod/clickhouse-client](https://packagist.org/packages/simpod/clickhouse-client)
     -   [seva-code/php-click-house-client](https://packagist.org/packages/seva-code/php-click-house-client)
-    -   [ﾂ环板clientｮﾂ嘉ｯﾂ偲](https://github.com/SeasX/SeasClick)
--   走吧
+    -   [SeasClick C++ client](https://github.com/SeasX/SeasClick)
+    -   [one-ck](https://github.com/lizhichao/one-ck)
+    -   [glushkovds/phpclickhouse-laravel](https://packagist.org/packages/glushkovds/phpclickhouse-laravel)
+-   Go
     -   [clickhouse](https://github.com/kshvakov/clickhouse/)
-    -   [ﾂ环板-ｮﾂ嘉ｯﾂ偲](https://github.com/roistat/go-clickhouse)
-    -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人](https://github.com/mailru/go-clickhouse)
+    -   [go-clickhouse](https://github.com/roistat/go-clickhouse)
+    -   [mailrugo-clickhouse](https://github.com/mailru/go-clickhouse)
     -   [golang-clickhouse](https://github.com/leprosus/golang-clickhouse)
+-   Swift
+    -   [ClickHouseNIO](https://github.com/patrick-zippenfenig/ClickHouseNIO)
+    -   [ClickHouseVapor ORM](https://github.com/patrick-zippenfenig/ClickHouseVapor)
 -   NodeJs
-    -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人)](https://github.com/TimonKK/clickhouse)
-    -   [ﾂ环板-ｮﾂ嘉ｯﾂ偲](https://github.com/apla/node-clickhouse)
+    -   [clickhouse (NodeJs)](https://github.com/TimonKK/clickhouse)
+    -   [node-clickhouse](https://github.com/apla/node-clickhouse)
 -   Perl
     -   [perl-DBD-ClickHouse](https://github.com/elcamlost/perl-DBD-ClickHouse)
     -   [HTTP-ClickHouse](https://metacpan.org/release/HTTP-ClickHouse)
-    -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人](https://metacpan.org/release/AnyEvent-ClickHouse)
+    -   [AnyEvent-ClickHouse](https://metacpan.org/release/AnyEvent-ClickHouse)
 -   Ruby
-    -   [ﾂ暗ｪﾂ氾环催ﾂ団)](https://github.com/shlima/click_house)
-    -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人](https://github.com/PNixx/clickhouse-activerecord)
+    -   [ClickHouse (Ruby)](https://github.com/shlima/click_house)
+    -   [clickhouse-activerecord](https://github.com/PNixx/clickhouse-activerecord)
 -   R
     -   [clickhouse-r](https://github.com/hannesmuehleisen/clickhouse-r)
-    -   [RClickhouse](https://github.com/IMSMWU/RClickhouse)
+    -   [RClickHouse](https://github.com/IMSMWU/RClickHouse)
 -   Java
     -   [clickhouse-client-java](https://github.com/VirtusAI/clickhouse-client-java)
--   斯卡拉
-    -   [掳胫client-禄脢鹿脷露胫鲁隆鹿-client酶](https://github.com/crobox/clickhouse-scala-client)
+    -   [clickhouse-client](https://github.com/Ecwid/clickhouse-client)
+-   Scala
+    -   [clickhouse-scala-client](https://github.com/crobox/clickhouse-scala-client)
 -   Kotlin
     -   [AORM](https://github.com/TanVD/AORM)
 -   C#
     -   [Octonica.ClickHouseClient](https://github.com/Octonica/ClickHouseClient)
-    -   [克莱克豪斯Ado](https://github.com/killwort/ClickHouse-Net)
+    -   [ClickHouse.Ado](https://github.com/killwort/ClickHouse-Net)
+    -   [ClickHouse.Client](https://github.com/DarkWanderer/ClickHouse.Client)
     -   [ClickHouse.Net](https://github.com/ilyabreev/ClickHouse.Net)
-    -   [克莱克豪斯客户](https://github.com/DarkWanderer/ClickHouse.Client)
--   仙丹
+-   Elixir
     -   [clickhousex](https://github.com/appodeal/clickhousex/)
--   尼姆
+    -   [pillar](https://github.com/sofakingworld/pillar)
+-   Nim
     -   [nim-clickhouse](https://github.com/leonardoce/nim-clickhouse)
+-   Haskell
+    -   [hdbc-clickhouse](https://github.com/zaneli/hdbc-clickhouse)
 
-[来源文章](https://clickhouse.tech/docs/zh/interfaces/third-party/client_libraries/) <!--hide-->
+[来源文章](https://clickhouse.tech/docs/en/interfaces/third-party/client_libraries/) <!--hide-->
diff --git a/docs/zh/interfaces/third-party/integrations.md b/docs/zh/interfaces/third-party/integrations.md
index 2c57e23b724..403ef994bb9 100644
--- a/docs/zh/interfaces/third-party/integrations.md
+++ b/docs/zh/interfaces/third-party/integrations.md
@@ -1,100 +1,108 @@
-# 第三方集成库 {#di-san-fang-ji-cheng-ku}
+---
+toc_priority: 27
+toc_title: 第三方集成库
+---
+
+# 第三方集成库 {#integration-libraries-from-third-party-developers}
 
 !!! warning "声明"
-    Yandex不维护下面列出的库，也没有进行任何广泛的测试以确保其质量。
+Yandex**没有**维护下面列出的库，也没有做过任何广泛的测试来确保它们的质量。
 
-## 基建产品 {#ji-jian-chan-pin}
+## 基础设施 {#infrastructure-products}
 
--   关系数据库管理系统
+-   关系数据库
     -   [MySQL](https://www.mysql.com)
         -   [mysql2ch](https://github.com/long2ice/mysql2ch)
         -   [ProxySQL](https://github.com/sysown/proxysql/wiki/ClickHouse-Support)
         -   [clickhouse-mysql-data-reader](https://github.com/Altinity/clickhouse-mysql-data-reader)
-        -   [horgh-复制器](https://github.com/larsnovikov/horgh-replicator)
+        -   [horgh-replicator](https://github.com/larsnovikov/horgh-replicator)
     -   [PostgreSQL](https://www.postgresql.org)
         -   [clickhousedb_fdw](https://github.com/Percona-Lab/clickhousedb_fdw)
-        -   [infi.clickhouse_fdw](https://github.com/Infinidat/infi.clickhouse_fdw) (使用 [infi.clickhouse_orm](https://github.com/Infinidat/infi.clickhouse_orm))
+        -   [infi.clickhouse_fdw](https://github.com/Infinidat/infi.clickhouse_fdw) (uses [infi.clickhouse_orm](https://github.com/Infinidat/infi.clickhouse_orm))
         -   [pg2ch](https://github.com/mkabilov/pg2ch)
+        -   [clickhouse_fdw](https://github.com/adjust/clickhouse_fdw)
     -   [MSSQL](https://en.wikipedia.org/wiki/Microsoft_SQL_Server)
-        -   [ClickHouseMightrator](https://github.com/zlzforever/ClickHouseMigrator)
+        -   [ClickHouseMigrator](https://github.com/zlzforever/ClickHouseMigrator)
 -   消息队列
-    -   [卡夫卡](https://kafka.apache.org)
-        -   [clickhouse_sinker](https://github.com/housepower/clickhouse_sinker) (使用 [去客户](https://github.com/ClickHouse/clickhouse-go/))
+    -   [Kafka](https://kafka.apache.org)
+        -   [clickhouse_sinker](https://github.com/housepower/clickhouse_sinker) (uses [Go client](https://github.com/ClickHouse/clickhouse-go/))
         -   [stream-loader-clickhouse](https://github.com/adform/stream-loader)
 -   流处理
     -   [Flink](https://flink.apache.org)
         -   [flink-clickhouse-sink](https://github.com/ivi-ru/flink-clickhouse-sink)
 -   对象存储
     -   [S3](https://en.wikipedia.org/wiki/Amazon_S3)
-    -   [ﾂ环板backupｮﾂ嘉ｯﾂ偲](https://github.com/AlexAkulov/clickhouse-backup)
+        -   [clickhouse-backup](https://github.com/AlexAkulov/clickhouse-backup)
 -   容器编排
     -   [Kubernetes](https://kubernetes.io)
-    -   [clickhouse-操](https://github.com/Altinity/clickhouse-operator)
+        -   [clickhouse-operator](https://github.com/Altinity/clickhouse-operator)
 -   配置管理
-    -   [木偶](https://puppet.com)
-    -   [ﾂ环板/ｮﾂ嘉ｯﾂ偲](https://forge.puppet.com/innogames/clickhouse)
-    -   [mfedotov/clickhouse](https://forge.puppet.com/mfedotov/clickhouse)
--   监控
-    -   [石墨](https://graphiteapp.org)
+    -   [puppet](https://puppet.com)
+        -   [innogames/clickhouse](https://forge.puppet.com/innogames/clickhouse)
+        -   [mfedotov/clickhouse](https://forge.puppet.com/mfedotov/clickhouse)
+-   Monitoring
+    -   [Graphite](https://graphiteapp.org)
         -   [graphouse](https://github.com/yandex/graphouse)
-        -   [ﾂ暗ｪﾂ氾环催ﾂ団](https://github.com/lomik/carbon-clickhouse) +
-        -   [ﾂ环板-ｮﾂ嘉ｯﾂ偲](https://github.com/lomik/graphite-clickhouse)
-        -   [石墨-ch-optimizer](https://github.com/innogames/graphite-ch-optimizer) -优化静态分区 [\*GraphiteMergeTree](../../engines/table-engines/mergetree-family/graphitemergetree.md#graphitemergetree) 如果从规则 [汇总配置](../../engines/table-engines/mergetree-family/graphitemergetree.md#rollup-configuration) 可以应用
+        -   [carbon-clickhouse](https://github.com/lomik/carbon-clickhouse) +
+        -   [graphite-clickhouse](https://github.com/lomik/graphite-clickhouse)
+        -   [graphite-ch-optimizer](https://github.com/innogames/graphite-ch-optimizer) - optimizes staled partitions in [\*GraphiteMergeTree](../../engines/table-engines/mergetree-family/graphitemergetree.md#graphitemergetree) if rules from [rollup configuration](../../engines/table-engines/mergetree-family/graphitemergetree.md#rollup-configuration) could be applied
     -   [Grafana](https://grafana.com/)
         -   [clickhouse-grafana](https://github.com/Vertamedia/clickhouse-grafana)
-    -   [普罗米修斯号](https://prometheus.io/)
+    -   [Prometheus](https://prometheus.io/)
         -   [clickhouse_exporter](https://github.com/f1yegor/clickhouse_exporter)
         -   [PromHouse](https://github.com/Percona-Lab/PromHouse)
-        -   [clickhouse_exporter](https://github.com/hot-wifi/clickhouse_exporter) （用途 [去客户](https://github.com/kshvakov/clickhouse/))
+        -   [clickhouse_exporter](https://github.com/hot-wifi/clickhouse_exporter) (uses [Go client](https://github.com/kshvakov/clickhouse/))
     -   [Nagios](https://www.nagios.org/)
         -   [check_clickhouse](https://github.com/exogroup/check_clickhouse/)
         -   [check_clickhouse.py](https://github.com/innogames/igmonplugins/blob/master/src/check_clickhouse.py)
     -   [Zabbix](https://www.zabbix.com)
-        -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人](https://github.com/Altinity/clickhouse-zabbix-template)
+        -   [clickhouse-zabbix-template](https://github.com/Altinity/clickhouse-zabbix-template)
     -   [Sematext](https://sematext.com/)
-        -   [clickhouse积分](https://github.com/sematext/sematext-agent-integrations/tree/master/clickhouse)
--   记录
+        -   [clickhouse integration](https://github.com/sematext/sematext-agent-integrations/tree/master/clickhouse)
+-   Logging
     -   [rsyslog](https://www.rsyslog.com/)
-        -   [鹿茫house omhousee酶](https://www.rsyslog.com/doc/master/configuration/modules/omclickhouse.html)
+        -   [omclickhouse](https://www.rsyslog.com/doc/master/configuration/modules/omclickhouse.html)
     -   [fluentd](https://www.fluentd.org)
-        -   [loghouse](https://github.com/flant/loghouse) (对于 [Kubernetes](https://kubernetes.io))
-    -   [Sematext](https://www.sematext.com/logagent)
-        -   [logagent输出-插件-clickhouse](https://sematext.com/docs/logagent/output-plugin-clickhouse/)
--   地理
+        -   [loghouse](https://github.com/flant/loghouse) (for [Kubernetes](https://kubernetes.io))
+    -   [logagent](https://www.sematext.com/logagent)
+        -   [logagent output-plugin-clickhouse](https://sematext.com/docs/logagent/output-plugin-clickhouse/)
+-   Geo
     -   [MaxMind](https://dev.maxmind.com/geoip/)
-        -   [ﾂ环板-ｮﾂ嘉ｯﾂ偲青clickｼﾂ氾ｶﾂ鉄ﾂ工ﾂ渉](https://github.com/AlexeyKupershtokh/clickhouse-maxmind-geoip)
+        -   [clickhouse-maxmind-geoip](https://github.com/AlexeyKupershtokh/clickhouse-maxmind-geoip)
 
-## 编程语言生态系统 {#bian-cheng-yu-yan-sheng-tai-xi-tong}
+## 编程语言 {#programming-language-ecosystems}
 
 -   Python
     -   [SQLAlchemy](https://www.sqlalchemy.org)
-        -   [ﾂ暗ｪﾂ氾环催ﾂ団ﾂ法ﾂ人](https://github.com/cloudflare/sqlalchemy-clickhouse) (使用 [infi.clickhouse_orm](https://github.com/Infinidat/infi.clickhouse_orm))
-    -   [熊猫](https://pandas.pydata.org)
+        -   [sqlalchemy-clickhouse](https://github.com/cloudflare/sqlalchemy-clickhouse) (uses [infi.clickhouse_orm](https://github.com/Infinidat/infi.clickhouse_orm))
+    -   [pandas](https://pandas.pydata.org)
         -   [pandahouse](https://github.com/kszucs/pandahouse)
 -   PHP
     -   [Doctrine](https://www.doctrine-project.org/)
         -   [dbal-clickhouse](https://packagist.org/packages/friendsofdoctrine/dbal-clickhouse)
 -   R
     -   [dplyr](https://db.rstudio.com/dplyr/)
-    -   [RClickhouse](https://github.com/IMSMWU/RClickhouse) (使用 [ﾂ暗ｪﾂ氾环催ﾂ団](https://github.com/artpaul/clickhouse-cpp))
+        -   [RClickHouse](https://github.com/IMSMWU/RClickHouse) (uses [clickhouse-cpp](https://github.com/artpaul/clickhouse-cpp))
 -   Java
     -   [Hadoop](http://hadoop.apache.org)
-    -   [clickhouse-hdfs-装载机](https://github.com/jaykelin/clickhouse-hdfs-loader) (使用 [JDBC](../../sql-reference/table-functions/jdbc.md))
--   斯卡拉
+        -   [clickhouse-hdfs-loader](https://github.com/jaykelin/clickhouse-hdfs-loader) (uses [JDBC](../../sql-reference/table-functions/jdbc.md))
+-   Scala
     -   [Akka](https://akka.io)
-    -   [掳胫client-禄脢鹿脷露胫鲁隆鹿-client酶](https://github.com/crobox/clickhouse-scala-client)
+        -   [clickhouse-scala-client](https://github.com/crobox/clickhouse-scala-client)
 -   C#
     -   [ADO.NET](https://docs.microsoft.com/en-us/dotnet/framework/data/adonet/ado-net-overview)
-    -   [克莱克豪斯Ado](https://github.com/killwort/ClickHouse-Net)
-    -   [ClickHouse.Net](https://github.com/ilyabreev/ClickHouse.Net)
-    -   [ClickHouse.Net.Migrations](https://github.com/ilyabreev/ClickHouse.Net.Migrations)
--   仙丹
+        -   [ClickHouse.Ado](https://github.com/killwort/ClickHouse-Net)
+        -   [ClickHouse.Client](https://github.com/DarkWanderer/ClickHouse.Client)
+        -   [ClickHouse.Net](https://github.com/ilyabreev/ClickHouse.Net)
+        -   [ClickHouse.Net.Migrations](https://github.com/ilyabreev/ClickHouse.Net.Migrations)
+-   Elixir
     -   [Ecto](https://github.com/elixir-ecto/ecto)
-    -   [clickhouse_ecto](https://github.com/appodeal/clickhouse_ecto)
+        -   [clickhouse_ecto](https://github.com/appodeal/clickhouse_ecto)
 -   Ruby
     -   [Ruby on Rails](https://rubyonrails.org/)
         -   [activecube](https://github.com/bitquery/activecube)
+        -   [ActiveRecord](https://github.com/PNixx/clickhouse-activerecord)
     -   [GraphQL](https://github.com/graphql)
         -   [activecube-graphql](https://github.com/bitquery/activecube-graphql)
 
-[来源文章](https://clickhouse.tech/docs/zh/interfaces/third-party/integrations/) <!--hide-->
+[源文章](https://clickhouse.tech/docs/en/interfaces/third-party/integrations/) <!--hide-->
diff --git a/docs/zh/interfaces/third-party/proxy.md b/docs/zh/interfaces/third-party/proxy.md
index 798717c0602..a2050863f30 100644
--- a/docs/zh/interfaces/third-party/proxy.md
+++ b/docs/zh/interfaces/third-party/proxy.md
@@ -1,37 +1,44 @@
-# 来自第三方开发人员的代理服务器 {#lai-zi-di-san-fang-kai-fa-ren-yuan-de-dai-li-fu-wu-qi}
+---
+toc_priority: 29
+toc_title: 第三方代理
+---
 
-[chproxy](https://github.com/Vertamedia/chproxy) 是ClickHouse数据库的http代理和负载均衡器。
+# 第三方代理 {#proxy-servers-from-third-party-developers}
 
-特征
+## chproxy {#chproxy}
 
-*每用户路由和响应缓存。
-*灵活的限制。
-\*自动SSL证书续订。
+[chproxy](https://github.com/Vertamedia/chproxy), 是一个用于ClickHouse数据库的HTTP代理和负载均衡器。
 
-在Go中实现。
+特性:
+
+-   用户路由和响应缓存。
+-   灵活的限制。
+-   自动SSL证书续订。
+
+使用go语言实现。
 
 ## KittenHouse {#kittenhouse}
 
-[KittenHouse](https://github.com/VKCOM/kittenhouse) 设计为ClickHouse和应用程序服务器之间的本地代理，以防在应用程序端缓冲INSERT数据是不可能或不方便的。
+[KittenHouse](https://github.com/VKCOM/kittenhouse)被设计为ClickHouse和应用服务器之间的本地代理，以防不可能或不方便在应用程序端缓冲插入数据。
 
-特征：
+特性:
 
-*内存和磁盘数据缓冲。
-*每表路由。
-\*负载平衡和健康检查。
+-   内存和磁盘上的数据缓冲。
+-   表路由。
+-   负载平衡和运行状况检查。
 
-在Go中实现。
+使用go语言实现。
 
-## ﾂ环板-ｮﾂ嘉ｯﾂ偲 {#clickhouse-bulk}
+## ClickHouse-Bulk {#clickhouse-bulk}
 
-[ﾂ环板-ｮﾂ嘉ｯﾂ偲](https://github.com/nikepan/clickhouse-bulk) 是一个简单的ClickHouse插入收集器。
+[ClickHouse-Bulk](https://github.com/nikepan/clickhouse-bulk)是一个简单的ClickHouse收集器。
 
-特征：
+特性:
 
-*分组请求并按阈值或间隔发送。
-*多个远程服务器。
-\*基本身份验证。
+-   按阈值或间隔对请求进行分组并发送。
+-   多个远程服务器。
+-   基本身份验证。
 
-在Go中实现。
+使用go语言实现。
 
-[来源文章](https://clickhouse.tech/docs/zh/interfaces/third-party/proxy/) <!--hide-->
+[Original article](https://clickhouse.tech/docs/en/interfaces/third-party/proxy/) <!--hide-->

From 7a81e0a0f5f438050631b6e02978f0a82cfce40d Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Mon, 21 Dec 2020 18:45:39 +0800
Subject: [PATCH 216/504] Extend mergetree setting alter restriction

---
 src/Storages/MergeTree/MergeTreeSettings.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/src/Storages/MergeTree/MergeTreeSettings.h b/src/Storages/MergeTree/MergeTreeSettings.h
index 2f3931786a6..d14e1f8a91b 100644
--- a/src/Storages/MergeTree/MergeTreeSettings.h
+++ b/src/Storages/MergeTree/MergeTreeSettings.h
@@ -137,7 +137,8 @@ struct MergeTreeSettings : public BaseSettings<MergeTreeSettingsTraits>
     /// We check settings after storage creation
     static bool isReadonlySetting(const String & name)
     {
-        return name == "index_granularity" || name == "index_granularity_bytes";
+        return name == "index_granularity" || name == "index_granularity_bytes" || name == "write_final_mark"
+            || name == "enable_mixed_granularity_parts" || name == "in_memory_parts_enable_wal";
     }
 
     static bool isPartFormatSetting(const String & name)

From c3a99e21bd3fc1b7783da7af30d93ee997a66161 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 21 Dec 2020 14:39:17 +0300
Subject: [PATCH 217/504] Refactor some code.

---
 src/Functions/FunctionsStringHash.cpp | 164 +++++++++++++++-----------
 src/Functions/FunctionsStringHash.h   |  28 +++--
 2 files changed, 108 insertions(+), 84 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index f8c78a808b3..b27d8601f3a 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -18,52 +18,94 @@
 
 namespace DB
 {
+
+namespace ErrorCodes
+{
+    extern const int NOT_IMPLEMENTED;
+}
+
 struct Hash
 {
     static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points)
     {
-        return intHashCRC32(unalignedLoad<UInt32>(code_points));
+#ifdef __SSE4_2__
+        return _mm_crc32_u64(-1ULL, unalignedLoad<UInt32>(code_points));
+#else
+        throw Exception("ngramASCIIHash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
     }
 
     static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points)
-    {
-        UInt64 combined = (static_cast<UInt64>(code_points[0]) << 32) | code_points[1];
-#ifdef __SSE4_2__
-        return _mm_crc32_u64(code_points[2], combined);
-#else
-        return (intHashCRC32(combined) ^ intHashCRC32(code_points[2]));
-#endif
-    }
-
-    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
     {
         UInt64 crc = -1ULL;
 #ifdef __SSE4_2__
-        for (size_t i = offset; i < size; ++i)
-            crc = _mm_crc32_u64(crc, hashes[i]);
-        for (size_t i = 0; i < offset; ++i)
-            crc = _mm_crc32_u64(crc, hashes[i]);
+        crc = _mm_crc32_u64(crc, code_points[0]);
+        crc = _mm_crc32_u64(crc, code_points[1]);
+        crc = _mm_crc32_u64(crc, code_points[2]);
 #else
-        for (size_t i = offset; i < size; ++i)
-            crc = intHashCRC32(crc) ^ intHashCRC32(hashes[i]);
-        for (size_t i = 0; i < offset; ++i)
-            crc = intHashCRC32(crc) ^ intHashCRC32(hashes[i]);
+        throw Exception("ngramUTF8Hash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
 #endif
         return crc;
     }
 
-    template <typename CodePoint>
-    static ALWAYS_INLINE inline UInt64 hashSum(const CodePoint * hashes, size_t K)
+    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt32 * hashes, size_t size, size_t offset)
     {
-        UInt64 crc = -1ULL;
+        UInt64 crc1 = -1ULL;
+        UInt64 crc2 = -1ULL;
 #ifdef __SSE4_2__
-        for (size_t i = 0; i < K; ++i)
-            crc = _mm_crc32_u64(crc, hashes[i]);
+        for (size_t i = offset; i < size; i += 2)
+            crc1 = _mm_crc32_u64(crc1, hashes[i]);
+        for (size_t i = offset + 1; i < size; i += 2)
+            crc2 = _mm_crc32_u64(crc2, hashes[i]);
+
+        if ((size - offset) & 1)
+        {
+            for (size_t i = 0; i < offset; i += 2)
+                crc2 = _mm_crc32_u64(crc2, hashes[i]);
+            for (size_t i = 1; i < offset; i += 2)
+                crc1 = _mm_crc32_u64(crc1, hashes[i]);
+        }
+        else
+        {
+            for (size_t i = 0; i < offset; i += 2)
+                crc1 = _mm_crc32_u64(crc1, hashes[i]);
+            for (size_t i = 1; i < offset; i += 2)
+                crc2 = _mm_crc32_u64(crc2, hashes[i]);
+        }
 #else
-        for (size_t i = 0; i < K; ++i)
-            crc = intHashCRC32(crc) ^ intHashCRC32(hashes[i]);
+        throw Exception("wordShinglesHash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
 #endif
-        return crc;
+        return crc1 | (crc2 << 32u);
+    }
+
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt8 * hashes, size_t K)
+    {
+        UInt64 crc1 = -1ULL;
+        UInt64 crc2 = -1ULL;
+#ifdef __SSE4_2__
+        for (size_t i = 0; i < K; i += 2)
+            crc1 = _mm_crc32_u8(crc1, hashes[i]);
+        for (size_t i = 1; i < K; i += 2)
+            crc2 = _mm_crc32_u8(crc2, hashes[i]);
+#else
+        throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+        return crc1 | (crc2 << 32u);
+    }
+
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt32 * hashes, size_t K)
+    {
+        UInt64 crc1 = -1ULL;
+        UInt64 crc2 = -1ULL;
+#ifdef __SSE4_2__
+        for (size_t i = 0; i < K; i += 2)
+            crc1 = _mm_crc32_u32(crc1, hashes[i]);
+        for (size_t i = 1; i < K; i += 2)
+            crc2 = _mm_crc32_u32(crc2, hashes[i]);
+#else
+        throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+        return crc1 | (crc2 << 32u);
     }
 };
 
@@ -76,7 +118,6 @@ struct Hash
 template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimhashImpl
 {
-    using ResultType = UInt64;
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
     // we made an assumption that the size of one word cann't exceed 128, which may not true
     // if some word's size exceed 128, it would be cut up to several word
@@ -204,25 +245,6 @@ struct SimhashImpl
         return res_bit.to_ullong();
     }
 
-    template <typename CalcFunc, typename... Args>
-    static ALWAYS_INLINE inline auto dispatch(CalcFunc calc_func, Args &&... args)
-    {
-        if constexpr (Ngram)
-        {
-            if constexpr (!UTF8)
-                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
-            else
-                return calc_func(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
-        }
-        else
-        {
-            if constexpr (!UTF8)
-                return calc_func(std::forward<Args>(args)..., StrOp::readOneASCIIWord, Hash::wordShinglesHash);
-            else
-                return calc_func(std::forward<Args>(args)..., StrOp::readOneUTF8Word, Hash::wordShinglesHash);
-        }
-    }
-
     static void apply(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, PaddedPODArray<UInt64> & res)
     {
         for (size_t i = 0; i < offsets.size(); ++i)
@@ -232,9 +254,19 @@ struct SimhashImpl
             if (data_size <= max_string_size)
             {
                 if constexpr (Ngram)
-                    res[i] = dispatch(ngramCalculateHashValue, one_data, data_size);
+                {
+                    if constexpr (!UTF8)
+                        res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
+                    else
+                        res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
+                }
                 else
-                    res[i] = dispatch(wordShinglesCalculateHashValue, one_data, data_size);
+                {
+                    if constexpr (!UTF8)
+                        res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
+                    else
+                        res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+                }
             }
             else
                 res[i] = -1ull;
@@ -289,7 +321,6 @@ struct MinhashImpl
     using Greater = std::greater<size_t>;
     using MaxHeap = FixedHeap<std::less<size_t>, K, -1ULL>;
     using MinHeap = FixedHeap<std::greater<size_t>, K, 0>;
-    using ResultType = UInt64;
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
     static constexpr size_t max_string_size = 1u << 15;
     static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
@@ -383,25 +414,6 @@ struct MinhashImpl
         return std::make_tuple(res1, res2);
     }
 
-    template <typename CalcFunc, typename... Args>
-    static ALWAYS_INLINE inline auto dispatch(CalcFunc calc_func, Args &&... args)
-    {
-        if constexpr (Ngram)
-        {
-            if constexpr (!UTF8)
-                return calc_func(std::forward<Args>(args)..., StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
-            else
-                return calc_func(std::forward<Args>(args)..., StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
-        }
-        else
-        {
-            if constexpr (!UTF8)
-                return calc_func(std::forward<Args>(args)..., StrOp::readOneASCIIWord, Hash::wordShinglesHash);
-            else
-                return calc_func(std::forward<Args>(args)..., StrOp::readOneUTF8Word, Hash::wordShinglesHash);
-        }
-    }
-
     static void apply(
         const ColumnString::Chars & data,
         const ColumnString::Offsets & offsets,
@@ -415,9 +427,19 @@ struct MinhashImpl
             if (data_size <= max_string_size)
             {
                 if constexpr (Ngram)
-                    std::tie(res1[i], res2[i]) = dispatch(ngramCalculateHashValue, one_data, data_size);
+                {
+                    if constexpr (!UTF8)
+                        std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
+                    else
+                        std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
+                }
                 else
-                    std::tie(res1[i], res2[i]) = dispatch(wordShinglesCalculateHashValue, one_data, data_size);
+                {
+                    if constexpr (!UTF8)
+                        std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
+                    else
+                        std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+                }
             }
             else
                 std::tie(res1[i], res2[i]) = std::make_tuple(-1ull, -1ull);
diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index 93795d1eaab..979f2bd8a9d 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -35,12 +35,14 @@ public:
     DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
     {
         if (!isString(arguments[0]))
-            throw Exception(
-                "Illegal type " + arguments[0]->getName() + " of argument of function " + getName(), ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                            "Function {} expect single String argument, got {}", getName(), arguments[0]->getName());
+
+        auto type = std::make_shared<DataTypeUInt64>();
         if constexpr (is_simhash)
-            return std::make_shared<DataTypeNumber<typename Impl::ResultType>>();
-        auto element = DataTypeFactory::instance().get("UInt64");
-        return std::make_shared<DataTypeTuple>(DataTypes{element, element});
+            return type;
+
+        return std::make_shared<DataTypeTuple>(DataTypes{type, type});
     }
 
     bool useDefaultImplementationForConstants() const override { return true; }
@@ -48,24 +50,24 @@ public:
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t) const override
     {
         const ColumnPtr & column = arguments[0].column;
-        using ResultType = typename Impl::ResultType;
+
         if constexpr (is_simhash)
         {
             // non const string, const case is handled by useDefaultImplementationForConstants.
-            auto col_res = ColumnVector<ResultType>::create();
-            typename ColumnVector<ResultType>::Container & vec_res = col_res->getData();
+            auto col_res = ColumnVector<UInt64>::create();
+            auto & vec_res = col_res->getData();
             vec_res.resize(column->size());
             const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
             Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_res);
-            return std::move(col_res);
+            return col_res;
         }
         else // Min hash
         {
             // non const string
-            auto col_h1 = ColumnVector<ResultType>::create();
-            auto col_h2 = ColumnVector<ResultType>::create();
-            typename ColumnVector<ResultType>::Container & vec_h1 = col_h1->getData();
-            typename ColumnVector<ResultType>::Container & vec_h2 = col_h2->getData();
+            auto col_h1 = ColumnVector<UInt64>::create();
+            auto col_h2 = ColumnVector<UInt64>::create();
+            auto & vec_h1 = col_h1->getData();
+            auto & vec_h2 = col_h2->getData();
             vec_h1.resize(column->size());
             vec_h2.resize(column->size());
             const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);

From 321ca2654eb5b41500a7fbacaf16894a3c90591a Mon Sep 17 00:00:00 2001
From: qianmoQ <shicheng@ttxit.com>
Date: Mon, 21 Dec 2020 19:40:55 +0800
Subject: [PATCH 218/504] translate documentation for whats-new

---
 docs/zh/whats-new/index.md              |  8 +--
 docs/zh/whats-new/roadmap.md            | 15 ++----
 docs/zh/whats-new/security-changelog.md | 65 +++++++++++++++++++------
 3 files changed, 57 insertions(+), 31 deletions(-)

diff --git a/docs/zh/whats-new/index.md b/docs/zh/whats-new/index.md
index 0f248773402..1c28a538eed 100644
--- a/docs/zh/whats-new/index.md
+++ b/docs/zh/whats-new/index.md
@@ -1,8 +1,8 @@
 ---
-machine_translated: true
-machine_translated_rev: 72537a2d527c63c07aa5d2361a8829f3895cf2bd
-toc_folder_title: "\u65B0\u589E\u5185\u5BB9"
-toc_priority: 72
+toc_folder_title: ClickHouse事迹
+toc_priority: 82
 ---
 
+# ClickHouse变更及改动? {#whats-new-in-clickhouse}
 
+对于已经发布的版本，有一个[roadmap](../whats-new/roadmap.md)和一个详细的[changelog](../whats-new/changelog/index.md)。
diff --git a/docs/zh/whats-new/roadmap.md b/docs/zh/whats-new/roadmap.md
index 377746efcb7..df4b027f7da 100644
--- a/docs/zh/whats-new/roadmap.md
+++ b/docs/zh/whats-new/roadmap.md
@@ -1,17 +1,10 @@
 ---
 toc_priority: 74
-toc_title: 路线图
+toc_title: Roadmap
 ---
 
-# 路线图 {#roadmap}
+# Roadmap {#roadmap}
 
-## Q2 2020 {#q2-2020}
-
--  和外部认证服务集成
-
-## Q3 2020 {#q3-2020}
-
--   资源池，为用户提供更精准的集群资源分配
-
-{## [原始文档](https://clickhouse.tech/docs/en/roadmap/) ##}
+`2021年Roadmap`已公布供公开讨论查看[这里](https://github.com/ClickHouse/ClickHouse/issues/17623).
 
+{## [源文章](https://clickhouse.tech/docs/en/roadmap/) ##}
diff --git a/docs/zh/whats-new/security-changelog.md b/docs/zh/whats-new/security-changelog.md
index f21158c7aed..9cf0ca0b08c 100644
--- a/docs/zh/whats-new/security-changelog.md
+++ b/docs/zh/whats-new/security-changelog.md
@@ -1,41 +1,74 @@
-## 修复于 ClickHouse Release 18.12.13, 2018-09-10 {#xiu-fu-yu-clickhouse-release-18-12-13-2018-09-10}
+---
+toc_priority: 76
+toc_title: 安全更新日志
+---
+
+## 修复于ClickHouse Release 19.14.3.3, 2019-09-10 {#fixed-in-clickhouse-release-19-14-3-3-2019-09-10}
+
+### CVE-2019-15024 {#cve-2019-15024}
+
+对ZooKeeper具有写访问权限并且可以运行ClickHouse所在网络上可用的自定义服务器的攻击者可以创建一个自定义的恶意服务器，该服务器将充当ClickHouse副本并在ZooKeeper中注册。当另一个副本将从恶意副本获取数据部分时，它可以强制clickhouse服务器写入文件系统上的任意路径。
+
+作者：Yandex信息安全团队Eldar Zaitov
+
+### CVE-2019-16535 {#cve-2019-16535}
+
+解压算法中的OOB-read、OOB-write和整数下溢可以通过本机协议实现RCE或DoS。
+
+作者: Yandex信息安全团队Eldar Zaitov
+
+### CVE-2019-16536 {#cve-2019-16536}
+
+恶意的经过身份验证的客户端可能会触发导致DoS的堆栈溢出。
+
+作者: Yandex信息安全团队Eldar Zaitov
+
+## 修复于ClickHouse Release 19.13.6.1, 2019-09-20 {#fixed-in-clickhouse-release-19-13-6-1-2019-09-20}
+
+### CVE-2019-18657 {#cve-2019-18657}
+
+表函数`url`存在允许攻击者在请求中插入任意HTTP标头的漏洞。
+
+作者: [Nikita Tikhomirov](https://github.com/NSTikhomirov)
+
+## 修复于ClickHouse Release 18.12.13, 2018-09-10 {#fixed-in-clickhouse-release-18-12-13-2018-09-10}
 
 ### CVE-2018-14672 {#cve-2018-14672}
 
-加载CatBoost模型的功能，允许遍历路径并通过错误消息读取任意文件。
+加载CatBoost模型的函数允许路径遍历和通过错误消息读取任意文件。
 
-来源: Yandex信息安全团队的Andrey Krasichkov
+作者：Yandex信息安全团队Andrey Krasichkov
 
-## 修复于 ClickHouse Release 18.10.3, 2018-08-13 {#xiu-fu-yu-clickhouse-release-18-10-3-2018-08-13}
+## 修复于Release 18.10.3, 2018-08-13 {#fixed-in-clickhouse-release-18-10-3-2018-08-13}
 
 ### CVE-2018-14671 {#cve-2018-14671}
 
-unixODBC允许从文件系统加载任意共享对象，从而导致«远程执行代码»漏洞。
+unixODBC允许从文件系统加载任意共享对象，从而导致远程代码执行漏洞。
 
-来源：Yandex信息安全团队的Andrey Krasichkov和Evgeny Sidorov
+作者：Yandex信息安全团队Andrey Krasichkov和Evgeny Sidorov
 
-## 修复于 ClickHouse Release 1.1.54388, 2018-06-28 {#xiu-fu-yu-clickhouse-release-1-1-54388-2018-06-28}
+## 修复于ClickHouse Release 1.1.54388, 2018-06-28 {#fixed-in-clickhouse-release-1-1-54388-2018-06-28}
 
 ### CVE-2018-14668 {#cve-2018-14668}
 
-远程表函数功能允许在 «user», «password» 及 «default_database» 字段中使用任意符号，从而导致跨协议请求伪造攻击。
+`remote`表函数允许在`user`，`password`和`default_database`字段中使用任意符号，从而导致跨协议请求伪造攻击。
 
-来源：Yandex信息安全团队的Andrey Krasichkov
+者：Yandex信息安全团队Andrey Krasichkov
 
-## 修复于 ClickHouse Release 1.1.54390, 2018-07-06 {#xiu-fu-yu-clickhouse-release-1-1-54390-2018-07-06}
+## 修复于ClickHouse Release 1.1.54390, 2018-07-06 {#fixed-in-clickhouse-release-1-1-54390-2018-07-06}
 
 ### CVE-2018-14669 {#cve-2018-14669}
 
-ClickHouse MySQL客户端启用了 «LOAD DATA LOCAL INFILE» 功能，该功能允许恶意MySQL数据库从连接的ClickHouse服务器读取任意文件。
+ClickHouse MySQL客户端启用了`LOAD DATA LOCAL INFILE`功能，允许恶意MySQL数据库从连接的ClickHouse服务器读取任意文件。
 
-来源：Yandex信息安全团队的Andrey Krasichkov和Evgeny Sidorov
+作者：Yandex信息安全团队Andrey Krasichkov和Evgeny Sidorov
 
-## 修复于 ClickHouse Release 1.1.54131, 2017-01-10 {#xiu-fu-yu-clickhouse-release-1-1-54131-2017-01-10}
+## 修复于ClickHouse Release 1.1.54131, 2017-01-10 {#fixed-in-clickhouse-release-1-1-54131-2017-01-10}
 
 ### CVE-2018-14670 {#cve-2018-14670}
 
-deb软件包中的错误配置可能导致使用未经授权的数据库。
+deb包中的错误配置可能导致未经授权使用数据库。
 
-来源：英国国家网络安全中心（NCSC）
+作者：英国国家网络安全中心（NCSC）
 
-[来源文章](https://clickhouse.tech/docs/en/security_changelog/) <!--hide-->
+{## [Original article](https://clickhouse.tech/docs/en/security_changelog/) ##}

From 24cb824bfe79f671794816492b69a01f04d7dd74 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 15:09:27 +0300
Subject: [PATCH 219/504] Simplify coverage images

---
 docker/images.json                            | 21 ++----
 docker/test/coverage/Dockerfile               | 22 +++---
 .../run.sh                                    | 11 ++-
 .../s3downloader                              |  0
 docker/test/stateful_with_coverage/Dockerfile | 15 ----
 .../test/stateless_with_coverage/Dockerfile   | 47 ------------
 docker/test/stateless_with_coverage/run.sh    | 75 -------------------
 7 files changed, 26 insertions(+), 165 deletions(-)
 rename docker/test/{stateful_with_coverage => coverage}/run.sh (82%)
 rename docker/test/{stateful_with_coverage => coverage}/s3downloader (100%)
 delete mode 100644 docker/test/stateful_with_coverage/Dockerfile
 delete mode 100644 docker/test/stateless_with_coverage/Dockerfile
 delete mode 100755 docker/test/stateless_with_coverage/run.sh

diff --git a/docker/images.json b/docker/images.json
index 16747cc3725..e7300972b6d 100644
--- a/docker/images.json
+++ b/docker/images.json
@@ -17,10 +17,6 @@
         "name": "yandex/clickhouse-unbundled-builder",
         "dependent": []
     },
-    "docker/test/coverage": {
-        "name": "yandex/clickhouse-coverage",
-        "dependent": []
-    },
     "docker/test/compatibility/centos": {
         "name": "yandex/clickhouse-test-old-centos",
         "dependent": []
@@ -48,27 +44,22 @@
     "docker/test/stateless": {
         "name": "yandex/clickhouse-stateless-test",
         "dependent": [
-            "docker/test/stateful"
+            "docker/test/stateful",
+            "docker/test/coverage"
         ]
     },
     "docker/test/stateless_pytest": {
         "name": "yandex/clickhouse-stateless-pytest",
         "dependent": []
     },
-    "docker/test/stateless_with_coverage": {
-        "name": "yandex/clickhouse-stateless-test-with-coverage",
-        "dependent": [
-            "docker/test/stateful_with_coverage"
-        ]
-    },
     "docker/test/stateful": {
         "name": "yandex/clickhouse-stateful-test",
         "dependent": [
             "docker/test/stress"
         ]
     },
-    "docker/test/stateful_with_coverage": {
-        "name": "yandex/clickhouse-stateful-test-with-coverage",
+    "docker/test/coverage": {
+        "name": "yandex/clickhouse-test-coverage",
         "dependent": []
     },
     "docker/test/unit": {
@@ -143,9 +134,7 @@
          "name": "yandex/clickhouse-test-base",
          "dependent": [
             "docker/test/stateless",
-            "docker/test/stateless_with_coverage",
-            "docker/test/stateless_pytest",
-            "docker/test/coverage"
+            "docker/test/stateless_pytest"
          ]
     },
     "docker/packager/unbundled": {
diff --git a/docker/test/coverage/Dockerfile b/docker/test/coverage/Dockerfile
index cea1a63cf6f..681f65e0f6f 100644
--- a/docker/test/coverage/Dockerfile
+++ b/docker/test/coverage/Dockerfile
@@ -1,16 +1,18 @@
-# docker build -t yandex/clickhouse-coverage .
-ARG PARENT_TAG=latest
-FROM yandex/clickhouse-test-base:${PARENT_TAG}
+# docker build -t yandex/clickhouse-test-coverage .
+FROM yandex/clickhouse-stateless-test
 
+RUN apt-get update -y \
+    && env DEBIAN_FRONTEND=noninteractive \
+        apt-get install --yes --no-install-recommends \
+            cmake
+
+COPY s3downloader /s3downloader
+COPY run.sh /run.sh
+
+ENV DATASETS="hits visits"
 ENV COVERAGE_DIR=/coverage_reports
 ENV SOURCE_DIR=/build
 ENV OUTPUT_DIR=/output
 ENV IGNORE='.*contrib.*'
 
-RUN apt-get update && apt-get install cmake --yes --no-install-recommends
-
-CMD mkdir -p /build/obj-x86_64-linux-gnu && cd /build/obj-x86_64-linux-gnu && CC=clang-11 CXX=clang++-11 cmake .. && cd /; \
-    dpkg -i /package_folder/clickhouse-common-static_*.deb; \
-    llvm-profdata-11 merge -sparse ${COVERAGE_DIR}/* -o clickhouse.profdata && \
-    llvm-cov-11 export /usr/bin/clickhouse -instr-profile=clickhouse.profdata -j=16 -format=lcov -skip-functions -ignore-filename-regex $IGNORE > output.lcov && \
-    genhtml output.lcov --ignore-errors source --output-directory ${OUTPUT_DIR}
+CMD ["/bin/bash", "/run.sh"]
diff --git a/docker/test/stateful_with_coverage/run.sh b/docker/test/coverage/run.sh
similarity index 82%
rename from docker/test/stateful_with_coverage/run.sh
rename to docker/test/coverage/run.sh
index 5fc6350fad8..41da390e06b 100755
--- a/docker/test/stateful_with_coverage/run.sh
+++ b/docker/test/coverage/run.sh
@@ -89,10 +89,17 @@ fi
 # more idiologically correct.
 read -ra ADDITIONAL_OPTIONS <<< "${ADDITIONAL_OPTIONS:-}"
 
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test --testname --shard --zookeeper --no-stateless --hung-check --print-time "$SKIP_LIST_OPT" "${ADDITIONAL_OPTIONS[@]}" "$SKIP_TESTS_OPTION" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee test_output/test_result.txt
+LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test --testname --shard --zookeeper --print-time "$SKIP_LIST_OPT" "${ADDITIONAL_OPTIONS[@]}" "$SKIP_TESTS_OPTION" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee test_output/test_result.txt
 
 kill_clickhouse
 
 sleep 3
 
-cp /*.profraw /profraw ||:
+mkdir -p $COVERAGE_DIR
+mv /*.profraw $COVERAGE_DIR
+
+mkdir -p $SOURCE_DIR/obj-x86_64-linux-gnu
+cd $SOURCE_DIR/obj-x86_64-linux-gnu && CC=clang-11 CXX=clang++-11 cmake .. && cd /
+llvm-profdata-11 merge -sparse ${COVERAGE_DIR}/* -o clickhouse.profdata
+llvm-cov-11 export /usr/bin/clickhouse -instr-profile=clickhouse.profdata -j=16 -format=lcov -skip-functions -ignore-filename-regex $IGNORE > output.lcov
+genhtml output.lcov --ignore-errors source --output-directory ${OUTPUT_DIR}
diff --git a/docker/test/stateful_with_coverage/s3downloader b/docker/test/coverage/s3downloader
similarity index 100%
rename from docker/test/stateful_with_coverage/s3downloader
rename to docker/test/coverage/s3downloader
diff --git a/docker/test/stateful_with_coverage/Dockerfile b/docker/test/stateful_with_coverage/Dockerfile
deleted file mode 100644
index 07acf4ed4e6..00000000000
--- a/docker/test/stateful_with_coverage/Dockerfile
+++ /dev/null
@@ -1,15 +0,0 @@
-# docker build -t yandex/clickhouse-stateful-test-with-coverage .
-FROM yandex/clickhouse-stateless-test-with-coverage
-
-RUN apt-get update -y \
-    && env DEBIAN_FRONTEND=noninteractive \
-        apt-get install --yes --no-install-recommends \
-        python3-requests procps psmisc
-
-COPY s3downloader /s3downloader
-COPY run.sh /run.sh
-
-ENV DATASETS="hits visits"
-
-CMD ["/bin/bash", "/run.sh"]
-
diff --git a/docker/test/stateless_with_coverage/Dockerfile b/docker/test/stateless_with_coverage/Dockerfile
deleted file mode 100644
index e31d2d1e009..00000000000
--- a/docker/test/stateless_with_coverage/Dockerfile
+++ /dev/null
@@ -1,47 +0,0 @@
-# docker build -t yandex/clickhouse-stateless-test-with-coverage .
-# TODO: that can be based on yandex/clickhouse-stateless-test (llvm version and CMD differs)
-FROM yandex/clickhouse-test-base
-
-ARG odbc_driver_url="https://github.com/ClickHouse/clickhouse-odbc/releases/download/v1.1.4.20200302/clickhouse-odbc-1.1.4-Linux.tar.gz"
-
-RUN apt-get update -y \
-    && env DEBIAN_FRONTEND=noninteractive \
-        apt-get install --yes --no-install-recommends \
-            bash \
-            tzdata \
-            fakeroot \
-            debhelper \
-            expect \
-            python3 \
-            python3-lxml \
-            python3-termcolor \
-            python3-requests \
-            sudo \
-            openssl \
-            ncdu \
-            netcat-openbsd \
-            telnet \
-            tree \
-            moreutils \
-            brotli \
-            zstd \
-            gdb \
-            lsof \
-            unixodbc \
-            wget \
-            qemu-user-static \
-            procps \
-            psmisc
-
-RUN mkdir -p /tmp/clickhouse-odbc-tmp \
-   && wget -nv -O - ${odbc_driver_url} | tar --strip-components=1 -xz -C /tmp/clickhouse-odbc-tmp \
-   && cp /tmp/clickhouse-odbc-tmp/lib64/*.so /usr/local/lib/ \
-   && odbcinst -i -d -f /tmp/clickhouse-odbc-tmp/share/doc/clickhouse-odbc/config/odbcinst.ini.sample \
-   && odbcinst -i -s -l -f /tmp/clickhouse-odbc-tmp/share/doc/clickhouse-odbc/config/odbc.ini.sample \
-   && rm -rf /tmp/clickhouse-odbc-tmp
-
-ENV TZ=Europe/Moscow
-RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone
-COPY run.sh /run.sh
-
-CMD ["/bin/bash", "/run.sh"]
diff --git a/docker/test/stateless_with_coverage/run.sh b/docker/test/stateless_with_coverage/run.sh
deleted file mode 100755
index 4e4d9430a11..00000000000
--- a/docker/test/stateless_with_coverage/run.sh
+++ /dev/null
@@ -1,75 +0,0 @@
-#!/bin/bash
-
-kill_clickhouse () {
-    echo "clickhouse pids $(pgrep -u clickhouse)" | ts '%Y-%m-%d %H:%M:%S'
-    pkill -f "clickhouse-server" 2>/dev/null
-
-
-    for _ in {1..120}
-    do
-        if ! pkill -0 -f "clickhouse-server" ; then break ; fi
-        echo "ClickHouse still alive" | ts '%Y-%m-%d %H:%M:%S'
-        sleep 1
-    done
-
-    if pkill -0 -f "clickhouse-server"
-    then
-        pstree -apgT
-        jobs
-        echo "Failed to kill the ClickHouse server"  | ts '%Y-%m-%d %H:%M:%S'
-        return 1
-    fi
-}
-
-start_clickhouse () {
-    LLVM_PROFILE_FILE='server_%h_%p_%m.profraw' sudo -Eu clickhouse /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml &
-    counter=0
-    until clickhouse-client --query "SELECT 1"
-    do
-        if [ "$counter" -gt 120 ]
-        then
-            echo "Cannot start clickhouse-server"
-            cat /var/log/clickhouse-server/stdout.log
-            tail -n1000 /var/log/clickhouse-server/stderr.log
-            tail -n1000 /var/log/clickhouse-server/clickhouse-server.log
-            break
-        fi
-        sleep 0.5
-        counter=$((counter + 1))
-    done
-}
-
-chmod 777 /
-
-dpkg -i package_folder/clickhouse-common-static_*.deb; \
-    dpkg -i package_folder/clickhouse-common-static-dbg_*.deb; \
-    dpkg -i package_folder/clickhouse-server_*.deb;  \
-    dpkg -i package_folder/clickhouse-client_*.deb; \
-    dpkg -i package_folder/clickhouse-test_*.deb
-
-
-mkdir -p /var/lib/clickhouse
-mkdir -p /var/log/clickhouse-server
-chmod 777 -R /var/lib/clickhouse
-chmod 777 -R /var/log/clickhouse-server/
-
-# install test configs
-/usr/share/clickhouse-test/config/install.sh
-
-start_clickhouse
-
-if grep -q -- "--use-skip-list" /usr/bin/clickhouse-test; then
-    SKIP_LIST_OPT="--use-skip-list"
-fi
-
-# We can have several additional options so we path them as array because it's
-# more idiologically correct.
-read -ra ADDITIONAL_OPTIONS <<< "${ADDITIONAL_OPTIONS:-}"
-
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test --testname --shard --zookeeper --hung-check --print-time "$SKIP_LIST_OPT" "${ADDITIONAL_OPTIONS[@]}" "$SKIP_TESTS_OPTION" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee test_output/test_result.txt
-
-kill_clickhouse
-
-sleep 3
-
-cp /*.profraw /profraw ||:

From 5a5fbf9eaefc20bf365783dc6f8489bdd3e8714d Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 15:22:07 +0300
Subject: [PATCH 220/504] Remove logs from thread fuzzer

---
 src/Common/ThreadFuzzer.cpp | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index 76cae4bb0ac..88ff53534e6 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -103,10 +103,6 @@ void ThreadFuzzer::initConfiguration()
     initFromEnv(sleep_probability, "THREAD_FUZZER_SLEEP_PROBABILITY");
     initFromEnv(sleep_time_us, "THREAD_FUZZER_SLEEP_TIME_US");
 
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep probability {}", sleep_probability);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Cpu time period {}", cpu_time_period_us);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep time us {}", sleep_time_us);
-
 #if THREAD_FUZZER_WRAP_PTHREAD
 #    define INIT_WRAPPER_PARAMS(RET, NAME, ...) \
         initFromEnv(NAME##_before_yield_probability, "THREAD_FUZZER_" #NAME "_BEFORE_YIELD_PROBABILITY"); \

From 0bc1479a48ed90296a8bbb2a2af04676eea8adda Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 14:19:12 +0300
Subject: [PATCH 221/504] Fix clickhouse-test

---
 tests/clickhouse-test | 42 ++++++++++++++++++++----------------------
 1 file changed, 20 insertions(+), 22 deletions(-)

diff --git a/tests/clickhouse-test b/tests/clickhouse-test
index 45cde00037f..07677d5b823 100755
--- a/tests/clickhouse-test
+++ b/tests/clickhouse-test
@@ -251,7 +251,7 @@ stop_time = None
 
 # def run_tests_array(all_tests, suite, suite_dir, suite_tmp_dir, run_total):
 def run_tests_array(all_tests_with_params):
-    all_tests, suite, suite_dir, suite_tmp_dir, run_total = all_tests_with_params
+    all_tests, suite, suite_dir, suite_tmp_dir = all_tests_with_params
     global exit_code
     global SERVER_DIED
     global stop_time
@@ -698,6 +698,7 @@ def main(args):
                 all_tests = [t for t in all_tests if any([re.search(r, t) for r in args.test])]
             all_tests.sort(key=key_func)
 
+            jobs = args.jobs
             parallel_tests = []
             sequential_tests = []
             for test in all_tests:
@@ -706,35 +707,32 @@ def main(args):
                 else:
                     parallel_tests.append(test)
 
-            print("Found", len(parallel_tests), "parallel tests and", len(sequential_tests), "sequential tests")
-            run_n, run_total = args.parallel.split('/')
-            run_n = float(run_n)
-            run_total = float(run_total)
-            tests_n = len(parallel_tests)
-            if run_total > tests_n:
-                run_total = tests_n
-            if run_n > run_total:
-                continue
+            if jobs > 1 and len(parallel_tests) > 0:
+                print("Found", len(parallel_tests), "parallel tests and", len(sequential_tests), "sequential tests")
+                run_n, run_total = args.parallel.split('/')
+                run_n = float(run_n)
+                run_total = float(run_total)
+                tests_n = len(parallel_tests)
+                if run_total > tests_n:
+                    run_total = tests_n
 
-            jobs = args.jobs
-            if jobs > tests_n:
-                jobs = tests_n
-            if jobs > run_total:
-                run_total = jobs
+                if jobs > tests_n:
+                    jobs = tests_n
+                if jobs > run_total:
+                    run_total = jobs
 
-            batch_size = len(parallel_tests) // jobs
-            parallel_tests_array = []
-            for i in range(0, len(parallel_tests), batch_size):
-                parallel_tests_array.append((parallel_tests[i:i+batch_size], suite, suite_dir, suite_tmp_dir, run_total))
+                batch_size = len(parallel_tests) // jobs
+                parallel_tests_array = []
+                for i in range(0, len(parallel_tests), batch_size):
+                    parallel_tests_array.append((parallel_tests[i:i+batch_size], suite, suite_dir, suite_tmp_dir))
 
-            if jobs > 1:
                 with closing(multiprocessing.Pool(processes=jobs)) as pool:
                     pool.map(run_tests_array, parallel_tests_array)
 
-                run_tests_array((sequential_tests, suite, suite_dir, suite_tmp_dir, run_total))
+                run_tests_array((sequential_tests, suite, suite_dir, suite_tmp_dir))
                 total_tests_run += len(sequential_tests) + len(parallel_tests)
             else:
-                run_tests_array((all_tests, suite, suite_dir, suite_tmp_dir, run_total))
+                run_tests_array((all_tests, suite, suite_dir, suite_tmp_dir))
                 total_tests_run += len(all_tests)
 
     if args.hung_check:

From b760d7d7d1c66e05a0f5637e4e96a24e479a14d8 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 15:22:57 +0300
Subject: [PATCH 222/504] Remove more changes

---
 src/Common/ThreadFuzzer.cpp | 1 -
 1 file changed, 1 deletion(-)

diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index 88ff53534e6..a32e50c44b2 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -10,7 +10,6 @@
 #include <common/sleep.h>
 
 #include <IO/ReadHelpers.h>
-#include <common/logger_useful.h>
 
 #include <Common/Exception.h>
 #include <Common/thread_local_rng.h>

From 110e76e71da4cc27115bbeaf5239a5158564d6bd Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 21 Dec 2020 15:42:57 +0300
Subject: [PATCH 223/504] Fix special build.

---
 src/DataStreams/RemoteQueryExecutor.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/DataStreams/RemoteQueryExecutor.cpp b/src/DataStreams/RemoteQueryExecutor.cpp
index 5123c763218..51d229a1126 100644
--- a/src/DataStreams/RemoteQueryExecutor.cpp
+++ b/src/DataStreams/RemoteQueryExecutor.cpp
@@ -199,7 +199,7 @@ Block RemoteQueryExecutor::read()
     }
 }
 
-std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext> & read_context)
+std::variant<Block, int> RemoteQueryExecutor::read(std::unique_ptr<ReadContext> & read_context [[maybe_unused]])
 {
 
 #if defined(OS_LINUX)

From e0400e335b28d1337a8dfd7eb0709747ad976f11 Mon Sep 17 00:00:00 2001
From: "Islam Israfilov (Islam93)" <israfilov93@gmail.com>
Date: Mon, 21 Dec 2020 16:21:44 +0300
Subject: [PATCH 224/504] see also block fixes for datetime rus doc

---
 docs/ru/sql-reference/data-types/datetime.md   | 12 +++++++-----
 docs/ru/sql-reference/data-types/datetime64.md | 11 ++++++-----
 2 files changed, 13 insertions(+), 10 deletions(-)

diff --git a/docs/ru/sql-reference/data-types/datetime.md b/docs/ru/sql-reference/data-types/datetime.md
index 87c5da68f35..f9b7c8e142c 100644
--- a/docs/ru/sql-reference/data-types/datetime.md
+++ b/docs/ru/sql-reference/data-types/datetime.md
@@ -116,12 +116,14 @@ FROM dt
 
 ## See Also {#see-also}
 
--   [Функции преобразования типов](../../sql-reference/data-types/datetime.md)
--   [Функции для работы с датой и временем](../../sql-reference/data-types/datetime.md)
--   [Функции для работы с массивами](../../sql-reference/data-types/datetime.md)
+-   [Функции преобразования типов](../../sql-reference/functions/type-conversion-functions.md)
+-   [Функции для работы с датой и временем](../../sql-reference/functions/date-time-functions.md)
+-   [Функции для работы с массивами](../../sql-reference/functions/array-functions.md)
 -   [Настройка `date_time_input_format`](../../operations/settings/settings.md#settings-date_time_input_format)
--   [Конфигурационный параметр сервера `timezone`](../../sql-reference/data-types/datetime.md#server_configuration_parameters-timezone)
--   [Операторы для работы с датой и временем](../../sql-reference/data-types/datetime.md#operators-datetime)
+-   [Настройка `date_time_output_format`](../../operations/settings/settings.md#settings-date_time_output_format)
+-   [Конфигурационный параметр сервера `timezone`](../../operations/server-configuration-parameters/settings.md#server_configuration_parameters-timezone)
+-   [Операторы для работы с датой и временем](../../sql-reference/operators/index.md#operators-datetime)
 -   [Тип данных `Date`](date.md)
+-   [Тип данных `DateTime64`](datetime64.md)
 
 [Оригинальная статья](https://clickhouse.tech/docs/ru/data_types/datetime/) <!--hide-->
diff --git a/docs/ru/sql-reference/data-types/datetime64.md b/docs/ru/sql-reference/data-types/datetime64.md
index 0a602e44636..b5e185c9132 100644
--- a/docs/ru/sql-reference/data-types/datetime64.md
+++ b/docs/ru/sql-reference/data-types/datetime64.md
@@ -92,11 +92,12 @@ FROM dt
 
 ## See Also {#see-also}
 
--   [Функции преобразования типов](../../sql-reference/data-types/datetime64.md)
--   [Функции для работы с датой и временем](../../sql-reference/data-types/datetime64.md)
--   [Функции для работы с массивами](../../sql-reference/data-types/datetime64.md)
+-   [Функции преобразования типов](../../sql-reference/functions/type-conversion-functions.md)
+-   [Функции для работы с датой и временем](../../sql-reference/functions/date-time-functions.md)
+-   [Функции для работы с массивами](../../sql-reference/functions/array-functions.md)
 -   [Настройка `date_time_input_format`](../../operations/settings/settings.md#settings-date_time_input_format)
--   [Конфигурационный параметр сервера `timezone`](../../sql-reference/data-types/datetime64.md#server_configuration_parameters-timezone)
--   [Операторы для работы с датой и временем](../../sql-reference/data-types/datetime64.md#operators-datetime)
+-   [Настройка `date_time_output_format`](../../operations/settings/settings.md#settings-date_time_output_format)
+-   [Конфигурационный параметр сервера `timezone`](../../operations/server-configuration-parameters/settings.md#server_configuration_parameters-timezone)
+-   [Операторы для работы с датой и временем](../../sql-reference/operators/index.md#operators-datetime)
 -   [Тип данных `Date`](date.md)
 -   [Тип данных `DateTime`](datetime.md)

From 69b40ffae5ab25563c5b94265b4710ae65974b78 Mon Sep 17 00:00:00 2001
From: Peng Jian <pengjian.uestc@gmail.com>
Date: Mon, 21 Dec 2020 21:46:55 +0800
Subject: [PATCH 225/504] fix nullptr

---
 src/Interpreters/TreeRewriter.cpp | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)

diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 2dc9a4ee818..125f8663afe 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -624,12 +624,23 @@ void TreeRewriterResult::collectUsedColumns(const ASTPtr & query, bool is_select
         for (const auto & name : columns_context.requiredColumns())
             ss << " '" << name << "'";
 
-        ss << ", maybe you meant: ";
-        for (const auto & name : columns_context.requiredColumns())
+        if (storage)
         {
-            auto hints = storage->getHints(name);
-            if (!hints.empty())
-                ss << " '" << toString(hints) << " '";
+            ss << ", maybe you meant: ";
+            for (const auto & name : columns_context.requiredColumns())
+            {
+                auto hints = storage->getHints(name);
+                if (!hints.empty())
+                    ss << " '" << toString(hints) << " '";
+            }
+        }
+        else
+        {
+            if (!source_column_names.empty())
+                for (const auto & name : columns_context.requiredColumns())
+                    ss << " '" << name << "'";
+            else
+                ss << ", no source columns";
         }
 
         if (columns_context.has_table_join)

From 29e0b4ec40e2a09c0c7f30f6918ddb53229067c3 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 21 Dec 2020 17:14:05 +0300
Subject: [PATCH 226/504] Refactor cose a little bit more.

---
 src/Functions/ExtractString.h         | 9 +++++++--
 src/Functions/FunctionsStringHash.cpp | 6 +++---
 2 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index 51d6f17380c..b659d072887 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -19,11 +19,16 @@ namespace DB
 template <size_t N, bool CaseInsensitive>
 struct ExtractStringImpl
 {
+    /// Padding form ColumnsString. It is a number of bytes we can always read starting from pos if pos < end.
     static constexpr size_t default_padding = 16;
 
-    // the length of code_points = default_padding + N -1
+    /// Functions are read `default_padding - (N - 1)` bytes into the buffer. Window of size N is used.
+    /// Read copies `N - 1` last bytes from buffer into beginning, and then reads new bytes.
+    static constexpr size_t buffer_size = default_padding + N - 1;
+
+    // the length of code_points = buffer_size
     // pos: the current beginning location that we want to copy data
-    // end: the end loction of the string
+    // end: the end location of the string
     static ALWAYS_INLINE size_t readASCIICodePoints(UInt8 * code_points, const char *& pos, const char * end)
     {
         /// Offset before which we copy some data.
diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index b27d8601f3a..70e524c5df8 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -48,7 +48,7 @@ struct Hash
         return crc;
     }
 
-    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt32 * hashes, size_t size, size_t offset)
+    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
@@ -122,7 +122,7 @@ struct SimhashImpl
     // we made an assumption that the size of one word cann't exceed 128, which may not true
     // if some word's size exceed 128, it would be cut up to several word
     static constexpr size_t max_string_size = 1u << 15;
-    static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
+    static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
 
     // Simhash ngram calculate function: String ->UInt64
     // this function extracting ngram from input string, and maintain a 64-dimensions vector
@@ -323,7 +323,7 @@ struct MinhashImpl
     using MinHeap = FixedHeap<std::greater<size_t>, K, 0>;
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
     static constexpr size_t max_string_size = 1u << 15;
-    static constexpr size_t simultaneously_codepoints_num = StrOp::default_padding + N - 1;
+    static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
 
     // Minhash ngram calculate function, String -> Tuple(UInt64, UInt64)
     // we extract ngram from input string, and calculate a hash value for each ngram

From 78429eca70b1bc5803a603fa41f8ebddaae2a802 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 21 Dec 2020 17:30:17 +0300
Subject: [PATCH 227/504] Refactor code a little bit more.

---
 src/Functions/FunctionsStringHash.cpp | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 70e524c5df8..ec0f5fea808 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -104,6 +104,21 @@ struct Hash
             crc2 = _mm_crc32_u32(crc2, hashes[i]);
 #else
         throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+        return crc1 | (crc2 << 32u);
+    }
+
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt64 * hashes, size_t K)
+    {
+        UInt64 crc1 = -1ULL;
+        UInt64 crc2 = -1ULL;
+#ifdef __SSE4_2__
+        for (size_t i = 0; i < K; i += 2)
+            crc1 = _mm_crc32_u64(crc1, hashes[i]);
+        for (size_t i = 1; i < K; i += 2)
+            crc2 = _mm_crc32_u64(crc2, hashes[i]);
+#else
+        throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
 #endif
         return crc1 | (crc2 << 32u);
     }

From 8f2d4f2d1ec02c80caac0ec4f73e39b244163dbd Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 18:46:36 +0300
Subject: [PATCH 228/504] Fix user

---
 docker/test/stateless/run.sh | 2 +-
 src/Common/ThreadFuzzer.cpp  | 5 +++++
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/docker/test/stateless/run.sh b/docker/test/stateless/run.sh
index 7aa4d5e74ab..fde699401b2 100755
--- a/docker/test/stateless/run.sh
+++ b/docker/test/stateless/run.sh
@@ -18,7 +18,7 @@ if [ "$NUM_TRIES" -gt "1" ]; then
     export THREAD_FUZZER_SLEEP_PROBABILITY=0.1
     export THREAD_FUZZER_SLEEP_TIME_US=100000
     # simpliest way to forward env variables to server
-    /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml --daemon
+    sudo -u clickhouse /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml --daemon
     sleep 5
 else
     service clickhouse-server start && sleep 5
diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index a32e50c44b2..76cae4bb0ac 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -10,6 +10,7 @@
 #include <common/sleep.h>
 
 #include <IO/ReadHelpers.h>
+#include <common/logger_useful.h>
 
 #include <Common/Exception.h>
 #include <Common/thread_local_rng.h>
@@ -102,6 +103,10 @@ void ThreadFuzzer::initConfiguration()
     initFromEnv(sleep_probability, "THREAD_FUZZER_SLEEP_PROBABILITY");
     initFromEnv(sleep_time_us, "THREAD_FUZZER_SLEEP_TIME_US");
 
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep probability {}", sleep_probability);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Cpu time period {}", cpu_time_period_us);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep time us {}", sleep_time_us);
+
 #if THREAD_FUZZER_WRAP_PTHREAD
 #    define INIT_WRAPPER_PARAMS(RET, NAME, ...) \
         initFromEnv(NAME##_before_yield_probability, "THREAD_FUZZER_" #NAME "_BEFORE_YIELD_PROBABILITY"); \

From 12f0f82d3d8d9282092f5afecd16fdc548a0ccba Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Mon, 21 Dec 2020 19:53:28 +0300
Subject: [PATCH 229/504] Update test.

---
 .../01016_simhash_minhash.reference           | 180 +++++++++++++-----
 .../0_stateless/01016_simhash_minhash.sql     |  64 +++++++
 2 files changed, 201 insertions(+), 43 deletions(-)

diff --git a/tests/queries/0_stateless/01016_simhash_minhash.reference b/tests/queries/0_stateless/01016_simhash_minhash.reference
index 2ababa29d1e..edd5afc1af7 100644
--- a/tests/queries/0_stateless/01016_simhash_minhash.reference
+++ b/tests/queries/0_stateless/01016_simhash_minhash.reference
@@ -1,50 +1,144 @@
 0
 2718169299
 2718169299
-3333471646
-26585365
-4124079607
-4124079607
-4124079607
-979945684
-(3700739653,3614688582)
-(2594676265,556335836)
-(2594676265,556335836)
-(3157724679,410999184)
-(1378962320,1336242123)
-(3277652371,1284714580)
-(3277652371,1284714580)
-(3277652371,1284714580)
-(3140472415,3787127930)
+1315333491
+1099965843
+5746351769509927967
+5746351769509927967
+8347269581771603092
+6041373934059725027
+(17178276249054052155,8864230932371215121)
+(14133097226001036899,7985237721476952807)
+(14133097226001036899,7985237721476952807)
+(4661257206578284012,15229878657590021759)
+(3087743741749030713,11631667950302077749)
+(11923981719512934676,1193672187225825732)
+(11923981719512934676,1193672187225825732)
+(17970606678134635272,3825545538448404526)
+(9422952829151664974,568010773615758889)
 2548869326
 2548869326
 401385678
 401385710
-4258739090
-4260836242
-718415633
-718681881
-2314703251
-1238864275
-3900085650
-3907425682
-2314703251
-1238864275
-3569207545
-3568143609
-(1525603924,509999509)
-(1525603924,3764233597)
-(1525603924,2706466536)
-(1525603924,1315689278)
-(3824755630,2122451089)
-(946380879,2122451089)
-(3295904092,4129673330)
-(3295904092,4129673330)
-(138351420,974287950)
-(824220170,974287950)
-(3300081739,2402902535)
-(3300081739,3993394872)
-(138351420,974287950)
-(824220170,974287950)
-(3083836461,957058619)
-(4120380459,90533100)
+2652202579
+2652235347
+2984455347
+2984488115
+12804820948382413807
+12804820948919350245
+11651601468065149391
+11651600368014488527
+18377198011227067677
+18233505035951822655
+5501050600367972694
+5501050600367972692
+(8590465925632898311,12699049311112305995)
+(8590465925632898311,15828587343885202011)
+(8590465925632898311,15824051019631343049)
+(8590465925632898311,12699049311222825283)
+(217966158370437743,14452995911556652133)
+(217966158370437743,14452995911556652133)
+(2170210914777151141,5341809779339553313)
+(12469866236432988845,5341809779339553313)
+(12271157076799061825,5514511977572226426)
+(11639913962681153226,2767634094725305612)
+(12271157075024394466,17994666970078080114)
+(12271157077109587702,13572452308677868240)
+(6252006845407214340,13538761942960976531)
+(13795977174459370328,6392395597500134035)
+(16118993428517222971,13602445809406467)
+(16118993428517222971,13602445809406467)
+uniqExact	6
+ngramSimhash
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	938403918
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	904817231
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	904849486
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	938469966
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	938404430
+ngramSimhashCaseInsensitive
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	938453071
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	938453599
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	938404430
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	636382047
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	938388046
+ngramSimhashUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2400625214
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2669060670
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	2671174174
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2669060798
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	2635506238
+ngramSimhashCaseInsensitiveUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2984307934
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	2967514366
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2715855070
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2967529694
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2984290526
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2984306910
+wordShingleSimhash
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384813566025024242
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2393820766427040734
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	2421405261516400471
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384883934767174398
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384813567165864670
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384813567098766070
+wordShingleSimhashCaseInsensitive
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	11635224793909957342
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617192803208139478
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617192803208151794
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617192803208151766
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	3006891407629799254
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617263171950236406
+wordShingleSimhashUTF8
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	9097818277104946605
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9084246141658271116
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9084247241171471628
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	9088752215857929613
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9093255814816009484
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9084247481822285196
+wordShingleSimhashCaseInsensitiveUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14788772559981154978
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	14497164445320454820
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14500537785782895266
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14787646625647636642
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14500016612976573090
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	14787956717160870888
+ngramMinhash
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(15568933215262012353,16287411738807860353)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(9473589826959436958,14264235017873782379)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(261441656340606110,13387826928927239258)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(261441656340606110,3305790294064680121)
+ngramMinhashCaseInsensitive
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(15568933215262012353,16287411738807860353)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(9473589826959436958,14264235017873782379)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(3051755284325985438,3305790294064680121)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(3051755284325985438,13387826928927239258)
+ngramMinhashUTF8
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	4	(309830857064065611,7476109060377919216)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(309830856946430871,7521913981442105351)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(309830857559697399,7476109060377919216)
+ngramMinhashCaseInsensitiveUTF8
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(13010809262502929096,2266175201446733829)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	5	(16827851229372179144,976408052548769549)
+wordShingleMinhash
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(14343822344862533053,11776483993821900250)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(18417749332128868312,11776483993821900250)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(18417749329907528200,14156831980621923226)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(4600092690178227586,11776483993821900250)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(12998011837685887081,1565093152297016105)
+wordShingleMinhashCaseInsensitive
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(12998011837880940480,1565093152297016105)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(1100751419997894255,15225006848401474458)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(1100751419777226283,12993805708561478711)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(1260401089202135898,12993805709529540523)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(1638964264353944555,12993805708561478711)
+wordShingleMinhashUTF8
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(742280067319112377,14237963017046410351)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(7237654052534217600,14400297883226437452)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(742280067319112377,17574811665615962276)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(3458625375707825328,17574811665615962276)
+wordShingleMinhashCaseInsensitiveUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7032848390598450936,5104668712725998486)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(15582670464629505464,13034678298246801511)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(9935434838523508980,7648038926638343017)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(7032848390598450936,16870743692447971238)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7302041809563941951,6856814412450461959)
diff --git a/tests/queries/0_stateless/01016_simhash_minhash.sql b/tests/queries/0_stateless/01016_simhash_minhash.sql
index 9e87216d26f..61b9ac14259 100644
--- a/tests/queries/0_stateless/01016_simhash_minhash.sql
+++ b/tests/queries/0_stateless/01016_simhash_minhash.sql
@@ -44,4 +44,68 @@ SELECT wordShingleMinhashCaseInsensitive(s) FROM defaults;
 SELECT wordShingleMinhashUTF8(s) FROM defaults;
 SELECT wordShingleMinhashCaseInsensitiveUTF8(s) FROM defaults;
 
+TRUNCATE TABLE defaults;
+INSERT INTO defaults SELECT arrayJoin(splitByString('\n\n',
+'ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.
+ClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.
+ClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.
+
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.
+ClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system''s read and write availability.
+ClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.
+
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.
+ClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system''s read / write availability.
+ClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.
+
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.
+ClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.
+ClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.
+
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.
+ClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.
+ClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.
+
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.
+ClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.
+ClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.'
+));
+
+SELECT 'uniqExact', uniqExact(s) FROM defaults;
+
+
+SELECT 'ngramSimhash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhash(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimhashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimhashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimhashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimhash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhash(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimhashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimhashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimhashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+
+SELECT 'ngramMinhash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhash(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinhashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinhashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinhashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinhash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhash(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinhashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinhashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinhashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+
 DROP TABLE defaults;

From 2222122f9f4bb998f133014d27d30997bddbc739 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Mon, 21 Dec 2020 20:02:59 +0300
Subject: [PATCH 230/504] Update ThreadFuzzer.cpp

---
 src/Common/ThreadFuzzer.cpp | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index 76cae4bb0ac..946e8590fc2 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -104,7 +104,9 @@ void ThreadFuzzer::initConfiguration()
     initFromEnv(sleep_time_us, "THREAD_FUZZER_SLEEP_TIME_US");
 
     LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep probability {}", sleep_probability);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Cpu time period {}", cpu_time_period_us);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Yield probability {}", yield_probability);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Migrate probability {}", migrate_probability);
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "CPU time period {}", cpu_time_period_us);
     LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep time us {}", sleep_time_us);
 
 #if THREAD_FUZZER_WRAP_PTHREAD

From 6a93e487c6e08de2180b76e6a68b07cf8b1db20e Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Mon, 21 Dec 2020 22:30:37 +0300
Subject: [PATCH 231/504] Document the countMatches function.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Задокументировал функцию countMatches.
---
 .../functions/string-search-functions.md      | 57 +++++++++++++++++--
 1 file changed, 53 insertions(+), 4 deletions(-)

diff --git a/docs/en/sql-reference/functions/string-search-functions.md b/docs/en/sql-reference/functions/string-search-functions.md
index 3cde7dd71d4..957866faea3 100644
--- a/docs/en/sql-reference/functions/string-search-functions.md
+++ b/docs/en/sql-reference/functions/string-search-functions.md
@@ -400,7 +400,8 @@ Result:
 └──────────────────────────────────────────────────────────────────────────────────────────┘
 ```
 
-**See also**
+**See Also**
+
 -   [extractAllGroupsVertical](#extractallgroups-vertical)
 
 ## extractAllGroupsVertical {#extractallgroups-vertical}
@@ -440,7 +441,8 @@ Result:
 └────────────────────────────────────────────────────────────────────────────────────────┘
 ```
 
-**See also**
+**See Also**
+
 -   [extractAllGroupsHorizontal](#extractallgroups-horizontal)
 
 ## like(haystack, pattern), haystack LIKE pattern operator {#function-like}
@@ -590,8 +592,55 @@ Result:
 └───────────────────────────────┘
 ```
 
-[Original article](https://clickhouse.tech/docs/en/query_language/functions/string_search_functions/) <!--hide-->
-
 ## countMatches(haystack, pattern) {#countmatcheshaystack-pattern}
 
 Returns the number of regular expression matches for a `pattern` in a `haystack`.
+
+**Syntax**
+
+``` sql
+countMatches(haystack, pattern)
+```
+
+**Parameters**
+
+-   `haystack` — The string to search in. [String](../../sql-reference/syntax.md#syntax-string-literal).
+-   `pattern` — The regular expression with [re2 syntax](https://github.com/google/re2/wiki/Syntax). [String](../../sql-reference/data-types/string.md).
+
+**Returned value**
+
+-   The number of matches.
+
+Type: [UInt64](../../sql-reference/data-types/int-uint.md).
+
+**Examples**
+
+Query:
+
+``` sql
+SELECT countMatches('foobar.com', 'o')
+```
+
+Result:
+
+``` text
+┌─countMatches('foobar.com', 'o')─┐
+│                               3 │
+└─────────────────────────────────┘
+```
+
+Query:
+
+``` sql
+SELECT countMatches('aaaa', 'aa')
+```
+
+Result:
+
+``` text
+┌─countMatches('aaaa', 'aa')─┐
+│                             2 │
+└───────────────────────────────┘
+```
+
+[Original article](https://clickhouse.tech/docs/en/query_language/functions/string_search_functions/) <!--hide-->

From 8bbb67875803a01acac9a2cd861a2b3ec2ce8f24 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 23:09:32 +0300
Subject: [PATCH 232/504] Correctly forward env

---
 docker/test/stateless/run.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docker/test/stateless/run.sh b/docker/test/stateless/run.sh
index fde699401b2..8b706729109 100755
--- a/docker/test/stateless/run.sh
+++ b/docker/test/stateless/run.sh
@@ -18,7 +18,7 @@ if [ "$NUM_TRIES" -gt "1" ]; then
     export THREAD_FUZZER_SLEEP_PROBABILITY=0.1
     export THREAD_FUZZER_SLEEP_TIME_US=100000
     # simpliest way to forward env variables to server
-    sudo -u clickhouse /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml --daemon
+    sudo -E -u clickhouse /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml --daemon
     sleep 5
 else
     service clickhouse-server start && sleep 5

From 302e8f697d4039b9887fecd8ff09e530de7cb638 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Mon, 21 Dec 2020 23:13:26 +0300
Subject: [PATCH 233/504] Document tcpPort function
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Задокументировал функцию tcp_port.
---
 .../functions/other-functions.md              | 40 +++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 51a1f6b4cd7..4d69edb3ec1 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1677,4 +1677,44 @@ Result:
 UNSUPPORTED_METHOD
 ```
 
+## tcpPort {#tcpPort}
+
+Returns TCP port number listened by this server.
+
+**Syntax**
+
+``` sql
+tcpPort()
+```
+
+**Parameters**
+
+-   None.
+
+**Returned value**
+
+-   The TCP port number.
+
+Type: [UInt16](../../sql-reference/data-types/int-uint.md).
+
+**Example**
+
+Query:
+
+``` sql
+SELECT tcpPort();
+```
+
+Result:
+
+``` text
+┌─tcpPort()─┐
+│      9000 │
+└───────────┘
+```
+
+**See Also**
+
+-   [tcp_port](../../operations/server-configuration-parameters/settings.md#server_configuration_parameters-tcp_port)
+
 [Original article](https://clickhouse.tech/docs/en/query_language/functions/other_functions/) <!--hide-->

From b10f13f34aed905ec3dbc0ab8447f2c1ad1c8834 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 21 Dec 2020 23:24:16 +0300
Subject: [PATCH 234/504] More thread fuzzer

---
 docker/test/stateless/run.sh | 15 +++++++++++++++
 src/Common/ThreadFuzzer.cpp  |  1 +
 2 files changed, 16 insertions(+)

diff --git a/docker/test/stateless/run.sh b/docker/test/stateless/run.sh
index 8b706729109..e6e987e1d94 100755
--- a/docker/test/stateless/run.sh
+++ b/docker/test/stateless/run.sh
@@ -17,6 +17,21 @@ if [ "$NUM_TRIES" -gt "1" ]; then
     export THREAD_FUZZER_CPU_TIME_PERIOD_US=1000
     export THREAD_FUZZER_SLEEP_PROBABILITY=0.1
     export THREAD_FUZZER_SLEEP_TIME_US=100000
+
+    export THREAD_FUZZER_pthread_mutex_lock_BEFORE_MIGRATE_PROBABILITY=1
+    export THREAD_FUZZER_pthread_mutex_lock_AFTER_MIGRATE_PROBABILITY=1
+    export THREAD_FUZZER_pthread_mutex_unlock_BEFORE_MIGRATE_PROBABILITY=1
+    export THREAD_FUZZER_pthread_mutex_unlock_AFTER_MIGRATE_PROBABILITY=1
+
+    export THREAD_FUZZER_pthread_mutex_lock_BEFORE_SLEEP_PROBABILITY=0.001
+    export THREAD_FUZZER_pthread_mutex_lock_AFTER_SLEEP_PROBABILITY=0.001
+    export THREAD_FUZZER_pthread_mutex_unlock_BEFORE_SLEEP_PROBABILITY=0.001
+    export THREAD_FUZZER_pthread_mutex_unlock_AFTER_SLEEP_PROBABILITY=0.001
+    export THREAD_FUZZER_pthread_mutex_lock_BEFORE_SLEEP_TIME_US=10000
+    export THREAD_FUZZER_pthread_mutex_lock_AFTER_SLEEP_TIME_US=10000
+    export THREAD_FUZZER_pthread_mutex_unlock_BEFORE_SLEEP_TIME_US=10000
+    export THREAD_FUZZER_pthread_mutex_unlock_AFTER_SLEEP_TIME_US=10000
+
     # simpliest way to forward env variables to server
     sudo -E -u clickhouse /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml --daemon
     sleep 5
diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index 946e8590fc2..ed155473061 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -125,6 +125,7 @@ void ThreadFuzzer::initConfiguration()
 
 #    undef INIT_WRAPPER_PARAMS
 #endif
+    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "pthread_mutex_lock_before_migrate_probability {}", pthread_mutex_lock_before_migrate_probability);
 }
 
 

From 356e9ad19567357a1062638c62aefeadd0d4924d Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Mon, 21 Dec 2020 23:51:44 +0300
Subject: [PATCH 235/504] Update datetime.md

---
 docs/ru/sql-reference/data-types/datetime.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/ru/sql-reference/data-types/datetime.md b/docs/ru/sql-reference/data-types/datetime.md
index f9b7c8e142c..9894fa2802b 100644
--- a/docs/ru/sql-reference/data-types/datetime.md
+++ b/docs/ru/sql-reference/data-types/datetime.md
@@ -119,8 +119,8 @@ FROM dt
 -   [Функции преобразования типов](../../sql-reference/functions/type-conversion-functions.md)
 -   [Функции для работы с датой и временем](../../sql-reference/functions/date-time-functions.md)
 -   [Функции для работы с массивами](../../sql-reference/functions/array-functions.md)
--   [Настройка `date_time_input_format`](../../operations/settings/settings.md#settings-date_time_input_format)
--   [Настройка `date_time_output_format`](../../operations/settings/settings.md#settings-date_time_output_format)
+-   [Настройка `date_time_input_format`](../../operations/settings/settings/#settings-date_time_input_format)
+-   [Настройка `date_time_output_format`](../../operations/settings/settings/)
 -   [Конфигурационный параметр сервера `timezone`](../../operations/server-configuration-parameters/settings.md#server_configuration_parameters-timezone)
 -   [Операторы для работы с датой и временем](../../sql-reference/operators/index.md#operators-datetime)
 -   [Тип данных `Date`](date.md)

From 575d5c69ef7c2e81436cddf6f166a59d4ae24fb4 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 00:23:45 +0300
Subject: [PATCH 236/504] fix partition boundaries for multiple window
 functions

---
 src/Processors/Transforms/WindowTransform.cpp | 157 ++++++++++--------
 src/Processors/Transforms/WindowTransform.h   |   4 +-
 2 files changed, 89 insertions(+), 72 deletions(-)

diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index b39926a0335..0b5dbc08654 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -38,7 +38,6 @@ WindowTransform::WindowTransform(const Block & input_header_,
         {
             workspace.argument_column_indices.push_back(
                 input_header.getPositionByName(argument_name));
-
         }
 
         workspace.aggregate_function_state.reset(aggregate_function->sizeOfData(),
@@ -61,10 +60,10 @@ WindowTransform::WindowTransform(const Block & input_header_,
 WindowTransform::~WindowTransform()
 {
     // Some states may be not created yet if the creation failed.
-    for (auto & w : workspaces)
+    for (auto & ws : workspaces)
     {
-        w.window_function.aggregate_function->destroy(
-            w.aggregate_function_state.data());
+        ws.window_function.aggregate_function->destroy(
+            ws.aggregate_function_state.data());
     }
 }
 
@@ -73,43 +72,76 @@ void WindowTransform::transform(Chunk & chunk)
     const size_t num_rows = chunk.getNumRows();
     auto block = getInputPort().getHeader().cloneWithColumns(chunk.detachColumns());
 
-    for (auto & workspace : workspaces)
-    {
-        workspace.argument_columns.clear();
-        for (const auto column_index : workspace.argument_column_indices)
-        {
-            workspace.argument_columns.push_back(
-                block.getColumns()[column_index].get());
-        }
-    }
-
     for (auto & ws : workspaces)
     {
-        const auto & f = ws.window_function;
-        const auto * a = f.aggregate_function.get();
+        ws.argument_columns.clear();
+        for (const auto column_index : ws.argument_column_indices)
+        {
+            ws.argument_columns.push_back(block.getColumns()[column_index].get());
+        }
 
         // Create the resulting column.
-        ColumnWithTypeAndName column_with_type;
-        column_with_type.name = f.column_name;
-        column_with_type.type = a->getReturnType();
-        auto c = column_with_type.type->createColumn();
-        column_with_type.column.reset(c.get());
+        ColumnWithTypeAndName column_description;
+        column_description.name = ws.window_function.column_name;
+        column_description.type
+            = ws.window_function.aggregate_function->getReturnType();
+        ws.result_column = column_description.type->createColumn();
+        column_description.column.reset(ws.result_column.get());
 
-        for (size_t row = 0; row < num_rows; row++)
+        block.insert(column_description);
+    }
+
+    // We loop for all window functions for each row. Switching the loops might
+    // be more efficient, because we would run less code and access less data in
+    // the inner loop. If you change this, don't forget to fix the calculation of
+    // partition boundaries. Probably it has to be precalculated and stored as a
+    // UInt8 array. An interesting optimization would be to pass it as an extra
+    // column from the previous sorting step -- that step might need to make
+    // similar comparison anyway, if it's sorting only by the PARTITION BY columns.
+    for (size_t row = 0; row < num_rows; row++)
+    {
+        // Check whether the new partition has started. We have to reset the
+        // aggregate functions when the new partition starts.
+        assert(partition_start_columns.size() == partition_by_indices.size());
+        bool new_partition = false;
+        if (partition_start_columns.empty())
         {
-            // THIS IS BROKEN when we have multiple window functions.
-            // Have to switch the loops or track partition per-function.
-            //
-            // Check whether the new partition has started and reinitialize the
-            // aggregate function states.
-            assert(partition_start_columns.size() == partition_by_indices.size());
-            if (partition_start_columns.empty())
+            // No PARTITION BY at all, do nothing.
+        }
+        else if (partition_start_columns[0] == nullptr)
+        {
+            // This is the first partition.
+            new_partition = true;
+            partition_start_columns.clear();
+            for (const auto i : partition_by_indices)
             {
-                // No PARTITION BY at all, do nothing.
+                partition_start_columns.push_back(block.getColumns()[i]);
             }
-            else if (partition_start_columns[0] == nullptr)
+            partition_start_row = row;
+        }
+        else
+        {
+            // Check whether the new partition started, by comparing all the
+            // PARTITION BY columns.
+            size_t first_inequal_column = 0;
+            for (; first_inequal_column < partition_start_columns.size();
+                  ++first_inequal_column)
             {
-                // This is the first partition.
+                const auto * current_column = block.getColumns()[
+                    partition_by_indices[first_inequal_column]].get();
+
+                if (current_column->compareAt(row, partition_start_row,
+                    *partition_start_columns[first_inequal_column],
+                    1 /* nan_direction_hint */) != 0)
+                {
+                    break;
+                }
+            }
+
+            if (first_inequal_column < partition_start_columns.size())
+            {
+                // The new partition has started. Remember where.
+                new_partition = true;
                 partition_start_columns.clear();
                 for (const auto i : partition_by_indices)
                 {
@@ -117,56 +149,39 @@ void WindowTransform::transform(Chunk & chunk)
                 }
                 partition_start_row = row;
             }
-            else
+        }
+
+        for (auto & ws : workspaces)
+        {
+            const auto & f = ws.window_function;
+            const auto * a = f.aggregate_function.get();
+            auto * buf = ws.aggregate_function_state.data();
+
+            if (new_partition)
             {
-                // Check whether the new partition started, by comparing all the
-                // PARTITION BY columns.
-                size_t first_inequal_column = 0;
-                for (; first_inequal_column < partition_start_columns.size();
-                      ++first_inequal_column)
-                {
-                    const auto * current_column = block.getColumns()[
-                        partition_by_indices[first_inequal_column]].get();
-
-                    if (current_column->compareAt(row, partition_start_row,
-                        *partition_start_columns[first_inequal_column],
-                        1 /* nan_direction_hint */) != 0)
-                    {
-                        break;
-                    }
-                }
-
-                if (first_inequal_column < partition_start_columns.size())
-                {
-                    // The new partition has started.
-                    // 1) Remember it.
-                    partition_start_columns.clear();
-                    for (const auto i : partition_by_indices)
-                    {
-                        partition_start_columns.push_back(block.getColumns()[i]);
-                    }
-                    partition_start_row = row;
-
-                    // 2) Reset the aggregate function states.
-                    ws.window_function.aggregate_function->destroy(
-                        ws.aggregate_function_state.data());
-                    ws.window_function.aggregate_function->create(
-                        ws.aggregate_function_state.data());
-                }
+                // Reset the aggregate function states.
+                a->destroy(buf);
+                a->create(buf);
             }
 
             // Update the aggregate function state and save the result.
-            a->add(ws.aggregate_function_state.data(),
+            a->add(buf,
                 ws.argument_columns.data(),
                 row,
                 arena.get());
 
-            a->insertResultInto(ws.aggregate_function_state.data(),
-                *c,
+            a->insertResultInto(buf,
+                *ws.result_column,
                 arena.get());
         }
+    }
 
-        block.insert(column_with_type);
+    // We have to release the mutable reference to the result column before we
+    // return this block, or else extra copying may occur when the subsequent
+    // processors modify the block. Workspaces live longer than individual blocks.
+    for (auto & ws : workspaces)
+    {
+        ws.result_column.detach();
     }
 
     chunk.setColumns(block.getColumns(), num_rows);
diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
index 93af208ed4f..89e09d15763 100644
--- a/src/Processors/Transforms/WindowTransform.h
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -19,8 +19,10 @@ struct WindowFunctionWorkspace
     WindowFunctionDescription window_function;
     AlignedBuffer aggregate_function_state;
     std::vector<size_t> argument_column_indices;
-    // Be careful, this is per-chunk.
+
+    // Argument and result columns. Be careful, they are per-chunk.
     std::vector<const IColumn *> argument_columns;
+    MutableColumnPtr result_column;
 };
 
 /*

From 881ce408bfbd74d82bca56ee9c5bc0672b0d7760 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Tue, 22 Dec 2020 00:48:18 +0300
Subject: [PATCH 237/504] rearranged some code parts, unified the
 impls::process method

---
 src/Functions/FunctionBinaryArithmetic.h | 997 +++++++++++------------
 1 file changed, 462 insertions(+), 535 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 1d1711b79d2..a9bff29fca6 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -62,6 +62,117 @@ namespace ErrorCodes
     extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
 }
 
+namespace traits_
+{
+struct InvalidType; /// Used to indicate undefined operation
+
+template <bool V, typename T> struct Case : std::bool_constant<V> { using type = T; };
+
+/// Switch<Case<C0, T0>, ...> -- select the first Ti for which Ci is true, InvalidType if none.
+template <typename... Ts> using Switch = typename std::disjunction<Ts..., Case<true, InvalidType>>::type;
+
+template <class T>
+using DataTypeFromFieldType = std::conditional_t<std::is_same_v<T, NumberTraits::Error>,
+    InvalidType, DataTypeNumber<T>>;
+
+template <typename DataType> constexpr bool IsIntegral = false;
+template <> inline constexpr bool IsIntegral<DataTypeUInt8> = true;
+template <> inline constexpr bool IsIntegral<DataTypeUInt16> = true;
+template <> inline constexpr bool IsIntegral<DataTypeUInt32> = true;
+template <> inline constexpr bool IsIntegral<DataTypeUInt64> = true;
+template <> inline constexpr bool IsIntegral<DataTypeInt8> = true;
+template <> inline constexpr bool IsIntegral<DataTypeInt16> = true;
+template <> inline constexpr bool IsIntegral<DataTypeInt32> = true;
+template <> inline constexpr bool IsIntegral<DataTypeInt64> = true;
+
+template <typename DataType> constexpr bool IsExtended = false;
+template <> inline constexpr bool IsExtended<DataTypeUInt256> = true;
+template <> inline constexpr bool IsExtended<DataTypeInt128> = true;
+template <> inline constexpr bool IsExtended<DataTypeInt256> = true;
+
+template <typename DataType> constexpr bool IsIntegralOrExtended = IsIntegral<DataType> || IsExtended<DataType>;
+template <typename DataType> constexpr bool IsIntegralOrExtendedOrDecimal =
+    IsIntegralOrExtended<DataType> ||
+    IsDataTypeDecimal<DataType>;
+
+template <typename DataType> constexpr bool IsFloatingPoint = false;
+template <> inline constexpr bool IsFloatingPoint<DataTypeFloat32> = true;
+template <> inline constexpr bool IsFloatingPoint<DataTypeFloat64> = true;
+
+template <typename DataType> constexpr bool IsDateOrDateTime = false;
+template <> inline constexpr bool IsDateOrDateTime<DataTypeDate> = true;
+template <> inline constexpr bool IsDateOrDateTime<DataTypeDateTime> = true;
+
+template <typename T0, typename T1> constexpr bool UseLeftDecimal = false;
+template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal256>, DataTypeDecimal<Decimal128>> = true;
+template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal256>, DataTypeDecimal<Decimal64>> = true;
+template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal256>, DataTypeDecimal<Decimal32>> = true;
+template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal128>, DataTypeDecimal<Decimal32>> = true;
+template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal128>, DataTypeDecimal<Decimal64>> = true;
+template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal64>, DataTypeDecimal<Decimal32>> = true;
+
+template <template <typename, typename> class Operation, typename LeftDataType, typename RightDataType>
+struct BinaryOperationTraits
+{
+    using T0 = typename LeftDataType::FieldType;
+    using T1 = typename RightDataType::FieldType;
+private: /// it's not correct for Decimal
+    using Op = Operation<T0, T1>;
+
+public:
+    static constexpr bool allow_decimal = IsOperation<Operation>::allow_decimal;
+
+    /// Appropriate result type for binary operator on numeric types. "Date" can also mean
+    /// DateTime, but if both operands are Dates, their type must be the same (e.g. Date - DateTime is invalid).
+    using ResultDataType = Switch<
+        /// Decimal cases
+        Case<!allow_decimal && (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>), InvalidType>,
+        Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType> && UseLeftDecimal<LeftDataType, RightDataType>, LeftDataType>,
+        Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>, RightDataType>,
+        Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
+        Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
+
+        /// e.g Decimal * Float64 = Float64
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>,
+            RightDataType>,
+        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>,
+            LeftDataType>,
+
+        /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
+        Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
+        Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
+
+        /// number <op> number -> see corresponding impl
+        Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
+            DataTypeFromFieldType<typename Op::ResultType>>,
+
+        /// Date + Integral -> Date
+        /// Integral + Date -> Date
+        Case<IsOperation<Operation>::plus, Switch<
+            Case<IsIntegral<RightDataType>, LeftDataType>,
+            Case<IsIntegral<LeftDataType>, RightDataType>>>,
+
+        /// Date - Date     -> Int32
+        /// Date - Integral -> Date
+        Case<IsOperation<Operation>::minus, Switch<
+            Case<std::is_same_v<LeftDataType, RightDataType>, DataTypeInt32>,
+            Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, LeftDataType>>>,
+
+        /// least(Date, Date) -> Date
+        /// greatest(Date, Date) -> Date
+        Case<std::is_same_v<LeftDataType, RightDataType> && (IsOperation<Operation>::least || IsOperation<Operation>::greatest),
+            LeftDataType>,
+
+        /// Date % Int32 -> Int32
+        /// Date % Float -> Float64
+        Case<IsOperation<Operation>::modulo, Switch<
+            Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, RightDataType>,
+            Case<IsDateOrDateTime<LeftDataType> && IsFloatingPoint<RightDataType>, DataTypeFloat64>>>>;
+};
+}
+
+namespace impl_
+{
 
 /** Arithmetic operations: +, -, *, /, %,
   * intDiv (integer division)
@@ -69,47 +180,58 @@ namespace ErrorCodes
   * Etc.
   */
 
-template <typename A, typename B, typename Op, typename ResultType_ = typename Op::ResultType>
+enum class OpCase { Vector, LeftConstant, RightConstant };
+
+template <class T>
+inline constexpr const auto & undec(const T & x)
+{
+    if constexpr (IsDecimalNumber<T>)
+        return x.value;
+    else
+        return x;
+}
+
+template <typename A, typename B, typename Op, typename OpResultType = typename Op::ResultType>
 struct BinaryOperation
 {
-    using ResultType = ResultType_;
+    using ResultType = OpResultType;
     static const constexpr bool allow_fixed_string = false;
 
-    static void NO_INLINE vectorVector(const A * __restrict a, const B * __restrict b, ResultType * __restrict c, size_t size)
+    template <OpCase op_case>
+    static void NO_INLINE process(const A * __restrict a, const B * __restrict b, ResultType * __restrict c, size_t size)
     {
         for (size_t i = 0; i < size; ++i)
-            c[i] = Op::template apply<ResultType>(a[i], b[i]);
+            if constexpr (op_case == OpCase::Vector)
+                c[i] = Op::template apply<ResultType>(a[i], b[i]);
+            else if constexpr (op_case == OpCase::LeftConstant)
+                c[i] = Op::template apply<ResultType>(*a, b[i]);
+            else
+                c[i] = Op::template apply<ResultType>(a[i], *b);
     }
 
-    static void NO_INLINE vectorConstant(const A * __restrict a, B b, ResultType * __restrict c, size_t size)
-    {
-        for (size_t i = 0; i < size; ++i)
-            c[i] = Op::template apply<ResultType>(a[i], b);
-    }
-
-    static void NO_INLINE constantVector(A a, const B * __restrict b, ResultType * __restrict c, size_t size)
-    {
-        for (size_t i = 0; i < size; ++i)
-            c[i] = Op::template apply<ResultType>(a, b[i]);
-    }
-
-    static ResultType constantConstant(A a, B b)
-    {
-        return Op::template apply<ResultType>(a, b);
-    }
+    static ResultType process(A a, B b) { return Op::template apply<ResultType>(a, b); }
 };
 
 template <typename Op>
 struct FixedStringOperationImpl
 {
-    static void NO_INLINE vectorVector(const UInt8 * __restrict a, const UInt8 * __restrict b, UInt8 * __restrict c, size_t size)
+    template <OpCase op_case>
+    static void NO_INLINE process(
+        const UInt8 * __restrict a, const UInt8 * __restrict b, UInt8 * __restrict result,
+        size_t size, [[maybe_unused]] size_t N)
     {
-        for (size_t i = 0; i < size; ++i)
-            c[i] = Op::template apply<UInt8>(a[i], b[i]);
+        if constexpr (op_case == OpCase::Vector)
+            for (size_t i = 0; i < size; ++i)
+                result[i] = Op::template apply<UInt8>(a[i], b[i]);
+        else if constexpr (op_case == OpCase::LeftConstant)
+            withConst<true>(b, a, result, size, N);
+        else
+            withConst<false>(a, b, result, size, N);
     }
 
+private:
     template <bool inverted>
-    static void NO_INLINE vector_constant_impl(const UInt8 * __restrict a, const UInt8 * __restrict b, UInt8 * __restrict c, size_t size, size_t N)
+    static void NO_INLINE withConst(const UInt8 * __restrict a, const UInt8 * __restrict b, UInt8 * __restrict c, size_t size, size_t N)
     {
         /// These complications are needed to avoid integer division in inner loop.
 
@@ -125,12 +247,14 @@ struct FixedStringOperationImpl
         ///      16 bytes starting from the last offset inside b.
 
         const size_t b_repeated_size = N + 15;
+
         UInt8 b_repeated[b_repeated_size];
+
         for (size_t i = 0; i < b_repeated_size; ++i)
             b_repeated[i] = b[i % N];
 
         size_t b_offset = 0;
-        size_t b_increment = 16 % N;
+        const size_t b_increment = 16 % N;
 
         /// Example:
         ///
@@ -154,43 +278,87 @@ struct FixedStringOperationImpl
                     : Op::template apply<UInt8>(b_repeated[b_offset + j], a[i + j]);
 
             b_offset += b_increment;
+
             if (b_offset >= N) /// This condition is easily predictable.
                 b_offset -= N;
         }
     }
-
-    static void vectorConstant(const UInt8 * __restrict a, const UInt8 * __restrict b, UInt8 * __restrict c, size_t size, size_t N)
-    {
-        vector_constant_impl<false>(a, b, c, size, N);
-    }
-
-    static void constantVector(const UInt8 * __restrict a, const UInt8 * __restrict b, UInt8 * __restrict c, size_t size, size_t N)
-    {
-        vector_constant_impl<true>(b, a, c, size, N);
-    }
 };
 
-
 template <typename A, typename B, typename Op, typename ResultType = typename Op::ResultType>
-struct BinaryOperationImpl : BinaryOperation<A, B, Op, ResultType>
-{
-};
+struct BinaryOperationImpl : BinaryOperation<A, B, Op, ResultType> { };
 
-template <typename T>
-inline constexpr const auto & undec(const T & x)
-{
-    if constexpr (IsDecimalNumber<T>)
-        return x.value;
-    else
-        return x;
-}
-
-/// Binary operations for Decimals need scale args
-/// +|-* scale one of args (which scale factor is not 1). ScaleR = oneof(Scale1, Scale2);
-/// /   first arg scale. ScaleR = Scale1 (scale_a = DecimalType<B>::getScale()).
-template <template <typename, typename> typename Operation, typename ResultType_, bool check_overflow = true>
+/**
+ * Binary operations with Decimals (either Decimal OP Decimal or Decimal Op Float) need to scale the args correctly.
+ *  - + (plus), - (minus), * (multiply), least and greatest operations scale one of the args (which scale factor is not 1).
+ *    The resulting scale is either left or the right scale.
+ *  - / (divide) operation scales the first argument.
+ *    The resulting scale is the first one's.
+ */
+template <template <typename, typename> typename Operation, class OpResultType, bool check_overflow = true>
 struct DecimalBinaryOperation
 {
+private:
+    using ResultType = OpResultType; // e.g. Decimal32
+    using NativeResultType = typename NativeType<ResultType>::Type; // e.g. UInt32 for Decimal32
+
+    using ResultContainerType = typename std::conditional_t<IsDecimalNumber<ResultType>,
+        ColumnDecimal<ResultType>,
+        ColumnVector<ResultType>>::Container;
+
+public:
+    template <OpCase op_case, bool is_decimal_a, bool is_decimal_b, bool scale_is_left_scale, class A, class B>
+    static void NO_INLINE process(const A & a, const B & b, ResultContainerType & c, NativeResultType scale)
+    {
+        if constexpr(op_case == OpCase::LeftConstant) static_assert(!IsDecimalNumber<A>);
+        if constexpr(op_case == OpCase::RightConstant) static_assert(!IsDecimalNumber<B>);
+
+        const size_t size = a.size();
+
+        if constexpr (is_plus_minus_compare)
+        {
+            if (scale != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    c[i] = applyScaled<scale_is_left_scale>(
+                        unwrap<op_case, OpCase::LeftConstant>(a, i),
+                        unwrap<op_case, OpCase::RightConstant>(b, i),
+                        scale);
+                return;
+            }
+        }
+        else if constexpr (is_division && is_decimal_b)
+        {
+            for (size_t i = 0; i < size; ++i)
+                c[i] = applyScaledDiv<is_decimal_a>(
+                    unwrap<op_case, OpCase::LeftConstant>(a, i),
+                    unwrap<op_case, OpCase::RightConstant>(b, i),
+                    scale);
+            return;
+        }
+
+        for (size_t i = 0; i < size; ++i)
+            c[i] = apply(
+                unwrap<op_case, OpCase::LeftConstant>(a, i),
+                unwrap<op_case, OpCase::RightConstant>(b, i));
+    }
+
+    template <bool is_decimal_a, bool is_decimal_b, bool scale_is_left_scale, class A, class B>
+    static ResultType process(A a, B b, NativeResultType scale)
+    {
+        static_assert(!IsDecimalNumber<A>);
+        static_assert(!IsDecimalNumber<B>);
+
+        if constexpr (is_division && is_decimal_b)
+            return applyScaledDiv<is_decimal_a>(a, b, scale);
+        else if constexpr (is_plus_minus_compare)
+            if (scale != 1)
+                return applyScaled<scale_is_left_scale>(a, b, scale);
+
+        return apply(a, b);
+    }
+
+private:
     static constexpr bool is_plus_minus =   IsOperation<Operation>::plus ||
                                             IsOperation<Operation>::minus;
     static constexpr bool is_multiply =     IsOperation<Operation>::multiply;
@@ -203,203 +371,18 @@ struct DecimalBinaryOperation
     static constexpr bool is_plus_minus_compare = is_plus_minus || is_compare;
     static constexpr bool can_overflow = is_plus_minus || is_multiply;
 
-    using ResultType = ResultType_;
-    using NativeResultType = typename NativeType<ResultType>::Type;
     using Op = std::conditional_t<is_float_division,
         DivideIntegralImpl<NativeResultType, NativeResultType>, /// substitute divide by intDiv (throw on division by zero)
         Operation<NativeResultType, NativeResultType>>;
 
-    using ArrayC = typename std::conditional_t<IsDecimalNumber<ResultType>,
-        ColumnDecimal<ResultType>,
-        ColumnVector<ResultType>>::Container;
-
-    template <bool is_decimal_a, bool is_decimal_b, typename ArrayA, typename ArrayB>
-    static void NO_INLINE vectorVector(const ArrayA & a, const ArrayB & b, ArrayC & c,
-                                       NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
+    template <OpCase op_case, OpCase target>
+    static auto unwrap(const auto& elem, int i)
     {
-        size_t size = a.size();
-
-        if constexpr (is_plus_minus_compare)
-        {
-            if (scale_a != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<true>(undec(a[i]), undec(b[i]), scale_a);
-                return;
-            }
-            else if (scale_b != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<false>(undec(a[i]), undec(b[i]), scale_b);
-                return;
-            }
-        }
-        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
-        {
-            if (scale_a != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
-                    else
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
-                return;
-            }
-            else if (scale_b != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
-                    else
-                        c[i] = apply(undec(a[i]), undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
-
-                return;
-            }
-        }
-        else if constexpr (is_division && is_decimal_b)
-        {
-            for (size_t i = 0; i < size; ++i)
-                c[i] = applyScaledDiv<is_decimal_a>(undec(a[i]), undec(b[i]), scale_a);
-            return;
-        }
-
-        /// default: use it if no return before
-        for (size_t i = 0; i < size; ++i)
-            c[i] = apply(undec(a[i]), undec(b[i]));
+        if constexpr(op_case == target)
+            return undec(elem);
+        return undec(elem[i]);
     }
 
-    template <bool is_decimal_a, bool is_decimal_b, typename ArrayA, typename B>
-    static void NO_INLINE vectorConstant(const ArrayA & a, B b, ArrayC & c,
-                                         NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
-    {
-        static_assert(!IsDecimalNumber<B>);
-
-        size_t size = a.size();
-        if constexpr (is_plus_minus_compare)
-        {
-            if (scale_a != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<true>(undec(a[i]), b, scale_a);
-                return;
-            }
-            else if (scale_b != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<false>(undec(a[i]), b, scale_b);
-                return;
-            }
-        }
-        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
-        {
-            if (scale_a != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
-                    else
-                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
-                return;
-            }
-            else if (scale_b != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
-                    else
-                        c[i] = apply(undec(a[i]), b) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
-
-                return;
-            }
-        }
-        else if constexpr (is_division && is_decimal_b)
-        {
-            for (size_t i = 0; i < size; ++i)
-                c[i] = applyScaledDiv<is_decimal_a>(undec(a[i]), b, scale_a);
-            return;
-        }
-
-        /// default: use it if no return before
-        for (size_t i = 0; i < size; ++i)
-            c[i] = apply(undec(a[i]), b);
-    }
-
-    template <bool is_decimal_a, bool is_decimal_b, typename A, typename ArrayB>
-    static void NO_INLINE constantVector(A a, const ArrayB & b, ArrayC & c,
-                                         NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
-    {
-        static_assert(!IsDecimalNumber<A>);
-
-        size_t size = b.size();
-        if constexpr (is_plus_minus_compare)
-        {
-            if (scale_a != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<true>(a, undec(b[i]), scale_a);
-                return;
-            }
-            else if (scale_b != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<false>(a, undec(b[i]), scale_b);
-                return;
-            }
-        }
-        else if constexpr(is_multiply && (!is_decimal_a || !is_decimal_b)) // BUG very bad solution
-        {
-            if (scale_a != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_a);
-                    else
-                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_a);
-                return;
-            }
-            else if (scale_b != 1)
-            {
-                for (size_t i = 0; i < size; ++i)
-                    if constexpr(IsDecimalNumber<ResultType>)
-                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<NativeResultType>(scale_b);
-                    else
-                        c[i] = apply(a, undec(b[i])) / DecimalUtils::scaleMultiplier<Int64>(scale_b);
-
-                return;
-            }
-        }
-
-        else if constexpr (is_division && is_decimal_b)
-        {
-            for (size_t i = 0; i < size; ++i)
-                c[i] = applyScaledDiv<is_decimal_a>(a, undec(b[i]), scale_a);
-            return;
-        }
-
-        /// default: use it if no return before
-        for (size_t i = 0; i < size; ++i)
-            c[i] = apply(a, undec(b[i]));
-    }
-
-    template <bool is_decimal_a, bool is_decimal_b, typename A, typename B>
-    static ResultType constantConstant(A a, B b, NativeResultType scale_a [[maybe_unused]], NativeResultType scale_b [[maybe_unused]])
-    {
-        static_assert(!IsDecimalNumber<A>);
-        static_assert(!IsDecimalNumber<B>);
-
-        if constexpr (is_plus_minus_compare)
-        {
-            if (scale_a != 1)
-                return applyScaled<true>(a, b, scale_a);
-            else if (scale_b != 1)
-                return applyScaled<false>(a, b, scale_b);
-        }
-        else if constexpr (is_division && is_decimal_b)
-            return applyScaledDiv<is_decimal_a>(a, b, scale_a);
-        return apply(a, b);
-    }
-
-private:
     /// there's implicit type conversion here
     static NativeResultType apply(NativeResultType a, NativeResultType b)
     {
@@ -474,108 +457,10 @@ private:
         }
     }
 };
+}
 
-
-/// Used to indicate undefined operation
-struct InvalidType;
-
-template <bool V, typename T> struct Case : std::bool_constant<V> { using type = T; };
-
-/// Switch<Case<C0, T0>, ...> -- select the first Ti for which Ci is true; InvalidType if none.
-template <typename... Ts> using Switch = typename std::disjunction<Ts..., Case<true, InvalidType>>::type;
-
-template <typename DataType> constexpr bool IsIntegral = false;
-template <> inline constexpr bool IsIntegral<DataTypeUInt8> = true;
-template <> inline constexpr bool IsIntegral<DataTypeUInt16> = true;
-template <> inline constexpr bool IsIntegral<DataTypeUInt32> = true;
-template <> inline constexpr bool IsIntegral<DataTypeUInt64> = true;
-template <> inline constexpr bool IsIntegral<DataTypeInt8> = true;
-template <> inline constexpr bool IsIntegral<DataTypeInt16> = true;
-template <> inline constexpr bool IsIntegral<DataTypeInt32> = true;
-template <> inline constexpr bool IsIntegral<DataTypeInt64> = true;
-
-template <typename DataType> constexpr bool IsExtended = false;
-template <> inline constexpr bool IsExtended<DataTypeUInt256> = true;
-template <> inline constexpr bool IsExtended<DataTypeInt128> = true;
-template <> inline constexpr bool IsExtended<DataTypeInt256> = true;
-
-template <typename DataType> constexpr bool IsIntegralOrExtended = IsIntegral<DataType> || IsExtended<DataType>;
-template <typename DataType> constexpr bool IsIntegralOrExtendedOrDecimal = IsIntegralOrExtended<DataType> || IsDataTypeDecimal<DataType>;
-
-template <typename DataType> constexpr bool IsFloatingPoint = false;
-template <> inline constexpr bool IsFloatingPoint<DataTypeFloat32> = true;
-template <> inline constexpr bool IsFloatingPoint<DataTypeFloat64> = true;
-
-template <typename DataType> constexpr bool IsDateOrDateTime = false;
-template <> inline constexpr bool IsDateOrDateTime<DataTypeDate> = true;
-template <> inline constexpr bool IsDateOrDateTime<DataTypeDateTime> = true;
-
-template <typename T0, typename T1> constexpr bool UseLeftDecimal = false;
-template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal256>, DataTypeDecimal<Decimal128>> = true;
-template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal256>, DataTypeDecimal<Decimal64>> = true;
-template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal256>, DataTypeDecimal<Decimal32>> = true;
-template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal128>, DataTypeDecimal<Decimal32>> = true;
-template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal128>, DataTypeDecimal<Decimal64>> = true;
-template <> inline constexpr bool UseLeftDecimal<DataTypeDecimal<Decimal64>, DataTypeDecimal<Decimal32>> = true;
-
-template <typename T> using DataTypeFromFieldType = std::conditional_t<std::is_same_v<T, NumberTraits::Error>, InvalidType, DataTypeNumber<T>>;
-
-template <template <typename, typename> class Operation, typename LeftDataType, typename RightDataType>
-struct BinaryOperationTraits
-{
-    using T0 = typename LeftDataType::FieldType;
-    using T1 = typename RightDataType::FieldType;
-private: /// it's not correct for Decimal
-    using Op = Operation<T0, T1>;
-public:
-
-    static constexpr bool allow_decimal = IsOperation<Operation>::allow_decimal;
-
-    /// Appropriate result type for binary operator on numeric types. "Date" can also mean
-    /// DateTime, but if both operands are Dates, their type must be the same (e.g. Date - DateTime is invalid).
-    using ResultDataType = Switch<
-        /// Decimal cases
-        Case<!allow_decimal && (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>), InvalidType>,
-        Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType> && UseLeftDecimal<LeftDataType, RightDataType>, LeftDataType>,
-        Case<IsDataTypeDecimal<LeftDataType> && IsDataTypeDecimal<RightDataType>, RightDataType>,
-        Case<IsDataTypeDecimal<LeftDataType> && IsIntegralOrExtended<RightDataType>, LeftDataType>,
-        Case<IsDataTypeDecimal<RightDataType> && IsIntegralOrExtended<LeftDataType>, RightDataType>,
-
-        /// e.g Decimal * Float64 = Float64
-        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<LeftDataType> && IsFloatingPoint<RightDataType>, RightDataType>,
-        Case<IsOperation<Operation>::multiply && IsDataTypeDecimal<RightDataType> && IsFloatingPoint<LeftDataType>, LeftDataType>,
-
-        /// Decimal <op> Real is not supported (traditional DBs convert Decimal <op> Real to Real)
-        Case<IsDataTypeDecimal<LeftDataType> && !IsIntegralOrExtendedOrDecimal<RightDataType>, InvalidType>,
-        Case<IsDataTypeDecimal<RightDataType> && !IsIntegralOrExtendedOrDecimal<LeftDataType>, InvalidType>,
-
-        /// number <op> number -> see corresponding impl
-        Case<!IsDateOrDateTime<LeftDataType> && !IsDateOrDateTime<RightDataType>,
-            DataTypeFromFieldType<typename Op::ResultType>>,
-
-        /// Date + Integral -> Date
-        /// Integral + Date -> Date
-        Case<IsOperation<Operation>::plus, Switch<
-            Case<IsIntegral<RightDataType>, LeftDataType>,
-            Case<IsIntegral<LeftDataType>, RightDataType>>>,
-
-        /// Date - Date     -> Int32
-        /// Date - Integral -> Date
-        Case<IsOperation<Operation>::minus, Switch<
-            Case<std::is_same_v<LeftDataType, RightDataType>, DataTypeInt32>,
-            Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, LeftDataType>>>,
-
-        /// least(Date, Date) -> Date
-        /// greatest(Date, Date) -> Date
-        Case<std::is_same_v<LeftDataType, RightDataType> && (IsOperation<Operation>::least || IsOperation<Operation>::greatest),
-            LeftDataType>,
-
-        /// Date % Int32 -> Int32
-        /// Date % Float -> Float64
-        Case<IsOperation<Operation>::modulo, Switch<
-            Case<IsDateOrDateTime<LeftDataType> && IsIntegral<RightDataType>, RightDataType>,
-            Case<IsDateOrDateTime<LeftDataType> && IsFloatingPoint<RightDataType>, DataTypeFloat64>>>>;
-};
+using namespace traits_;
+using namespace impl_;
 
 template <template <typename, typename> class Op, typename Name, bool valid_on_default_arguments = true>
 class FunctionBinaryArithmetic : public IFunction
@@ -817,6 +702,134 @@ class FunctionBinaryArithmetic : public IFunction
         return function->execute(new_arguments, result_type, input_rows_count);
     }
 
+    template <class T, class ResultDataType>
+    static auto helperGetOrConvert(const auto & col_const, const auto & col)
+    {
+        using ResultType = typename ResultDataType::FieldType;
+        using NativeResultType = typename NativeType<ResultType>::Type;
+
+        if constexpr (IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
+            return DecimalUtils::convertTo<NativeResultType>(col_const->template getValue<T>(), col.getScale());
+        else
+            return col_const->template getValue<T>();
+    }
+
+    template <class LeftDataType, class RightDataType, class ResultDataType>
+    ColumnPtr executeNumericWithDecimal(
+        const auto& left, const auto& right,
+        const ColumnConst * const col_left_const, const ColumnConst * const col_right_const,
+        const auto * const col_left, const auto * const col_right,
+        size_t col_left_size)
+    {
+        using T0 = typename LeftDataType::FieldType;
+        using T1 = typename RightDataType::FieldType;
+        using ResultType = typename ResultDataType::FieldType;
+
+        using NativeResultType = typename NativeType<ResultType>::Type;
+        using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
+        using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
+
+        using ColVecResult = std::conditional_t<IsDecimalNumber<ResultType>,
+            ColumnDecimal<ResultType>, ColumnVector<ResultType>>;
+
+        static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
+        static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
+        static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
+
+        typename ColVecResult::MutablePtr col_res = nullptr;
+
+        const ResultDataType type = [&] {
+            if constexpr (left_is_decimal && IsFloatingPoint<RightDataType>)
+                return RightDataType();
+            else if constexpr (right_is_decimal && IsFloatingPoint<LeftDataType>)
+                return LeftDataType();
+            else
+                return decimalResultType<is_multiply, is_division>(left, right);
+        }();
+
+        ResultType left_scale;
+        ResultType right_scale;
+
+        if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
+            left_scale = right.getScaleMultiplier();
+        else if constexpr (result_is_decimal)
+            left_scale = type.scaleFactorFor(left, is_multiply);
+        else if constexpr(left_is_decimal) // BUG precision loss
+            left_scale = left.getScale();
+        else
+            left_scale = 1; //won't be used, just to silence the warning
+
+        if constexpr (result_is_decimal)
+            right_scale = type.scaleFactorFor(right, is_multiply || is_division);
+        else if constexpr(right_is_decimal)
+            right_scale = right.getScale();
+        else
+            right_scale = 1; //same
+
+        /// non-vector result
+        if (col_left_const && col_right_const)
+        {
+            const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
+            const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
+
+            auto res = check_decimal_overflow ?
+                OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
+                    const_a, const_b, left_scale, right_scale) :
+                OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
+                    const_a, const_b, left_scale, right_scale);
+
+            if constexpr (result_is_decimal)
+                return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
+                    col_left_const->size(), toField(res, type.getScale()));
+            else
+                 return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
+        }
+
+        if constexpr (result_is_decimal)
+            col_res = ColVecResult::create(0, type.getScale());
+        else
+            col_res = ColVecResult::create(0);
+
+        auto & vec_res = col_res->getData();
+        vec_res.resize(col_left_size);
+
+        if (col_left && col_right)
+        {
+            if (check_decimal_overflow)
+                OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
+                    col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
+            else
+                OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
+                    col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
+        }
+        else if (col_left_const && col_right)
+        {
+            const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
+
+            if (check_decimal_overflow)
+                OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
+                    const_a, col_right->getData(), vec_res, left_scale, right_scale);
+            else
+                OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
+                    const_a, col_right->getData(), vec_res, left_scale, right_scale);
+        }
+        else if (col_left && col_right_const)
+        {
+            const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
+
+            if (check_decimal_overflow)
+                OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
+                    col_left->getData(), const_b, vec_res, left_scale, right_scale);
+            else
+                OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
+                    col_left->getData(), const_b, vec_res, left_scale, right_scale);
+        }
+        else
+            return nullptr;
+
+        return col_res;
+    }
+
 public:
     static constexpr auto name = Name::name;
     static FunctionPtr create(const Context & context) { return std::make_shared<FunctionBinaryArithmetic>(context); }
@@ -826,10 +839,7 @@ public:
         check_decimal_overflow(decimalCheckArithmeticOverflow(context))
     {}
 
-    String getName() const override
-    {
-        return name;
-    }
+    String getName() const override { return name; }
 
     size_t getNumberOfArguments() const override { return 2; }
 
@@ -884,7 +894,8 @@ public:
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
 
-            if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
+            if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> ||
+                          std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
                     return false;
@@ -951,35 +962,44 @@ public:
     {
         using OpImpl = FixedStringOperationImpl<Op<UInt8, UInt8>>;
 
-        const auto * col_left_raw = arguments[0].column.get();
-        const auto * col_right_raw = arguments[1].column.get();
+        const auto * const col_left_raw = arguments[0].column.get();
+        const auto * const col_right_raw = arguments[1].column.get();
+
         if (const auto * col_left_const = checkAndGetColumnConst<ColumnFixedString>(col_left_raw))
         {
             if (const auto * col_right_const = checkAndGetColumnConst<ColumnFixedString>(col_right_raw))
             {
                 const auto * col_left = checkAndGetColumn<ColumnFixedString>(col_left_const->getDataColumn());
                 const auto * col_right = checkAndGetColumn<ColumnFixedString>(col_right_const->getDataColumn());
+
                 if (col_left->getN() != col_right->getN())
                     return nullptr;
+
                 auto col_res = ColumnFixedString::create(col_left->getN());
                 auto & out_chars = col_res->getChars();
+
                 out_chars.resize(col_left->getN());
-                OpImpl::vectorVector(col_left->getChars().data(),
-                                      col_right->getChars().data(),
-                                      out_chars.data(),
-                                      out_chars.size());
+
+                OpImpl::process<OpCase::Vector>(
+                    col_left->getChars().data(),
+                    col_right->getChars().data(),
+                    out_chars.data(),
+                    out_chars.size());
+
                 return ColumnConst::create(std::move(col_res), col_left_raw->size());
             }
         }
 
-        bool is_left_column_const = checkAndGetColumnConst<ColumnFixedString>(col_left_raw) != nullptr;
-        bool is_right_column_const = checkAndGetColumnConst<ColumnFixedString>(col_right_raw) != nullptr;
+        const bool is_left_column_const = checkAndGetColumnConst<ColumnFixedString>(col_left_raw) != nullptr;
+        const bool is_right_column_const = checkAndGetColumnConst<ColumnFixedString>(col_right_raw) != nullptr;
 
         const auto * col_left = is_left_column_const
-                        ? checkAndGetColumn<ColumnFixedString>(checkAndGetColumnConst<ColumnFixedString>(col_left_raw)->getDataColumn())
+                        ? checkAndGetColumn<ColumnFixedString>(
+                            checkAndGetColumnConst<ColumnFixedString>(col_left_raw)->getDataColumn())
                         : checkAndGetColumn<ColumnFixedString>(col_left_raw);
         const auto * col_right = is_right_column_const
-                        ? checkAndGetColumn<ColumnFixedString>(checkAndGetColumnConst<ColumnFixedString>(col_right_raw)->getDataColumn())
+                        ? checkAndGetColumn<ColumnFixedString>(
+                            checkAndGetColumnConst<ColumnFixedString>(col_right_raw)->getDataColumn())
                         : checkAndGetColumn<ColumnFixedString>(col_right_raw);
 
         if (col_left && col_right)
@@ -993,7 +1013,7 @@ public:
 
             if (!is_left_column_const && !is_right_column_const)
             {
-                OpImpl::vectorVector(
+                OpImpl::process<OpCase::Vector>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
@@ -1001,7 +1021,7 @@ public:
             }
             else if (is_left_column_const)
             {
-                OpImpl::constantVector(
+                OpImpl::process<OpCase::LeftConstant>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
@@ -1010,30 +1030,19 @@ public:
             }
             else
             {
-                OpImpl::vectorConstant(
+                OpImpl::process<OpCase::RightConstant>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
                     out_chars.size(),
                     col_left->getN());
             }
+
             return col_res;
         }
         return nullptr;
     }
 
-    template <class T, class ResultDataType>
-    static auto helperGetOrConvert(const auto & col_const, const auto & col)
-    {
-        using ResultType = typename ResultDataType::FieldType;
-        using NativeResultType = typename NativeType<ResultType>::Type;
-
-        if constexpr (IsFloatingPoint<ResultDataType> && IsDecimalNumber<T>)
-            return DecimalUtils::convertTo<NativeResultType>(col_const->template getValue<T>(), col.getScale());
-        else
-            return col_const->template getValue<T>();
-    }
-
     template <typename A, typename B>
     ColumnPtr executeNumeric(const ColumnsWithTypeAndName & arguments, const A & left, const B & right) const
     {
@@ -1041,160 +1050,78 @@ public:
         using RightDataType = std::decay_t<decltype(right)>;
         using ResultDataType = typename BinaryOperationTraits<Op, LeftDataType, RightDataType>::ResultDataType;
 
-        if constexpr (!std::is_same_v<ResultDataType, InvalidType>)
+        if constexpr (std::is_same_v<ResultDataType, InvalidType>)
+            return nullptr;
+
+        using T0 = typename LeftDataType::FieldType;
+        using T1 = typename RightDataType::FieldType;
+        using ResultType = typename ResultDataType::FieldType;
+        using ColVecT0 = std::conditional_t<IsDecimalNumber<T0>, ColumnDecimal<T0>, ColumnVector<T0>>;
+        using ColVecT1 = std::conditional_t<IsDecimalNumber<T1>, ColumnDecimal<T1>, ColumnVector<T1>>;
+        using ColVecResult = std::conditional_t<IsDecimalNumber<ResultType>, ColumnDecimal<ResultType>, ColumnVector<ResultType>>;
+
+        const auto * const col_left_raw = arguments[0].column.get();
+        const auto * const col_right_raw = arguments[1].column.get();
+
+        const size_t col_left_size = col_left_raw->size();
+
+        const ColumnConst * const col_left_const = checkAndGetColumnConst<ColVecT0>(col_left_raw);
+        const ColumnConst * const col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw);
+
+        const ColVecT0 * const col_left = checkAndGetColumn<ColVecT0>(col_left_raw);
+        const ColVecT1 * const col_right = checkAndGetColumn<ColVecT1>(col_right_raw);
+
+        if constexpr (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>)
+            return executeNumericWithDecimal<LeftDataType, RightDataType, ResultDataType>(
+                left, right,
+                col_left_const, col_right_const,
+                col_left, col_right,
+                col_left_size);
+
+        using OpImpl = BinaryOperationImpl<T0, T1, Op<T0, T1>, ResultType>;
+
+        /// non-vector result
+        if (col_left_const && col_right_const)
         {
-            using T0 = typename LeftDataType::FieldType;
-            using T1 = typename RightDataType::FieldType;
-            using ResultType = typename ResultDataType::FieldType;
-            using ColVecT0 = std::conditional_t<IsDecimalNumber<T0>, ColumnDecimal<T0>, ColumnVector<T0>>;
-            using ColVecT1 = std::conditional_t<IsDecimalNumber<T1>, ColumnDecimal<T1>, ColumnVector<T1>>;
-            using ColVecResult = std::conditional_t<IsDecimalNumber<ResultType>, ColumnDecimal<ResultType>, ColumnVector<ResultType>>;
+            const auto res = OpImpl::process(
+                col_left_const->template getValue<T0>(),
+                col_right_const->template getValue<T1>());
 
-            const auto * col_left_raw = arguments[0].column.get();
-            const auto * col_right_raw = arguments[1].column.get();
-
-            auto col_left_const = checkAndGetColumnConst<ColVecT0>(col_left_raw);
-            auto col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw);
-
-            typename ColVecResult::MutablePtr col_res = nullptr;
-
-            auto col_left = checkAndGetColumn<ColVecT0>(col_left_raw);
-            auto col_right = checkAndGetColumn<ColVecT1>(col_right_raw);
-
-            if constexpr (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>)
-            {
-                using NativeResultType = typename NativeType<ResultType>::Type;
-                using OpImpl = DecimalBinaryOperation<Op, ResultType, false>;
-                using OpImplCheck = DecimalBinaryOperation<Op, ResultType, true>;
-                using FieldType = typename ResultDataType::FieldType;
-
-                static constexpr const bool left_is_decimal = IsDecimalNumber<T0>;
-                static constexpr const bool right_is_decimal = IsDecimalNumber<T1>;
-                static constexpr const bool result_is_decimal = IsDataTypeDecimal<ResultDataType>;
-
-                const ResultDataType type = [&] {
-                    if constexpr (left_is_decimal && IsFloatingPoint<RightDataType>)
-                        return RightDataType();
-                    else if constexpr (right_is_decimal && IsFloatingPoint<LeftDataType>)
-                        return LeftDataType();
-                    else
-                        return decimalResultType<is_multiply, is_division>(left, right);
-                }();
-
-                FieldType left_scale;
-                FieldType right_scale;
-
-                if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                    left_scale = right.getScaleMultiplier();
-                else if constexpr (result_is_decimal)
-                    left_scale = type.scaleFactorFor(left, is_multiply);
-                else if constexpr(left_is_decimal) // BUG precision loss
-                    left_scale = left.getScale();
-                else
-                    left_scale = 1; //won't be used, just to silence the warning
-
-                if constexpr (result_is_decimal)
-                    right_scale = type.scaleFactorFor(right, is_multiply || is_division);
-                else if constexpr(right_is_decimal)
-                    right_scale = right.getScale();
-                else
-                    right_scale = 1; //same
-
-                /// non-vector result
-                if (col_left_const && col_right_const)
-                {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
-
-                    auto res = check_decimal_overflow ?
-                        OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
-                            const_a, const_b, left_scale, right_scale) :
-                        OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
-                            const_a, const_b, left_scale, right_scale);
-
-                    if constexpr (result_is_decimal)
-                        return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
-                            col_left_const->size(), toField(res, type.getScale()));
-                    else
-                         return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
-                }
-
-                if constexpr (result_is_decimal)
-                    col_res = ColVecResult::create(0, type.getScale());
-                else
-                    col_res = ColVecResult::create(0);
-
-                auto & vec_res = col_res->getData();
-                vec_res.resize(col_left_raw->size());
-
-                if (col_left && col_right)
-                {
-                    if (check_decimal_overflow)
-                        OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
-                    else
-                        OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
-                }
-                else if (col_left_const && col_right)
-                {
-                    const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
-
-                    if (check_decimal_overflow)
-                        OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
-                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
-                    else
-                        OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
-                            const_a, col_right->getData(), vec_res, left_scale, right_scale);
-                }
-                else if (col_left && col_right_const)
-                {
-                    const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
-
-                    if (check_decimal_overflow)
-                        OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
-                    else
-                        OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
-                            col_left->getData(), const_b, vec_res, left_scale, right_scale);
-                }
-                else
-                    return nullptr;
-            }
-            else
-            {
-                using OpImpl = BinaryOperationImpl<T0, T1, Op<T0, T1>, ResultType>;
-
-                /// non-vector result
-                if (col_left_const && col_right_const)
-                {
-                    auto res = OpImpl::constantConstant(col_left_const->template getValue<T0>(), col_right_const->template getValue<T1>());
-                    return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
-                }
-
-                col_res = ColVecResult::create();
-                auto & vec_res = col_res->getData();
-                vec_res.resize(col_left_raw->size());
-
-                if (col_left && col_right)
-                {
-                    OpImpl::vectorVector(col_left->getData().data(), col_right->getData().data(), vec_res.data(), vec_res.size());
-                }
-                else if (col_left_const && col_right)
-                {
-                    OpImpl::constantVector(col_left_const->template getValue<T0>(), col_right->getData().data(), vec_res.data(), vec_res.size());
-                }
-                else if (col_left && col_right_const)
-                {
-                    OpImpl::vectorConstant(col_left->getData().data(), col_right_const->template getValue<T1>(), vec_res.data(), vec_res.size());
-                }
-                else
-                    return nullptr;
-            }
-
-            return col_res;
+            return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
         }
-        return nullptr;
+
+        typename ColVecResult::MutablePtr col_res = ColVecResult::create();
+        auto & vec_res = col_res->getData();
+        vec_res.resize(col_left_size);
+
+        if (col_left && col_right)
+        {
+            OpImpl::process<OpCase::Vector>(
+                col_left->getData().data(),
+                col_right->getData().data(),
+                vec_res.data(),
+                vec_res.size());
+        }
+        else if (col_left_const && col_right)
+        {
+            OpImpl::process<OpCase::LeftConstant>(
+                col_left_const->template getValue<T0>(),
+                col_right->getData().data(),
+                vec_res.data(),
+                vec_res.size());
+        }
+        else if (col_left && col_right_const)
+        {
+            OpImpl::process<OpCase::RightConstant>(
+                col_left->getData().data(),
+                col_right_const->template getValue<T1>(),
+                vec_res.data(),
+                vec_res.size());
+        }
+        else
+            return nullptr;
+
+        return col_res;
     }
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
@@ -1220,8 +1147,8 @@ public:
 
         const auto & left_argument = arguments[0];
         const auto & right_argument = arguments[1];
-        const auto * left_generic = left_argument.type.get();
-        const auto * right_generic = right_argument.type.get();
+        const auto * const left_generic = left_argument.type.get();
+        const auto * const right_generic = right_argument.type.get();
         ColumnPtr res;
 
         const bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
@@ -1229,7 +1156,8 @@ public:
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
 
-            if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> || std::is_same_v<DataTypeFixedString, RightDataType>)
+            if constexpr (std::is_same_v<DataTypeFixedString, LeftDataType> ||
+                std::is_same_v<DataTypeFixedString, RightDataType>)
             {
                 if constexpr (!Op<DataTypeFixedString, DataTypeFixedString>::allow_fixed_string)
                     return false;
@@ -1357,8 +1285,8 @@ public:
 
     bool hasInformationAboutMonotonicity() const override
     {
-        std::string_view name_ = Name::name;
-        return (name_ == "minus" || name_ == "plus" || name_ == "divide" || name_ == "intDiv");
+        const std::string_view name = Name::name;
+        return (name == "minus" || name == "plus" || name == "divide" || name == "intDiv");
     }
 
     Monotonicity getMonotonicityForRange(const IDataType &, const Field & left_point, const Field & right_point) const override
@@ -1371,8 +1299,9 @@ public:
         if (applyVisitor(FieldVisitorAccurateEquals(), left_point, right_point))
             return {true, true, false};
 
-        std::string_view name_ = Name::name;
-        if (name_ == "minus" || name_ == "plus")
+        const std::string_view name = Name::name;
+
+        if (name == "minus" || name == "plus")
         {
             // const +|- variable
             if (left.column && isColumnConst(*left.column))
@@ -1390,7 +1319,7 @@ public:
                 };
                 transform(left_point);
                 transform(right_point);
-                if (name_ == "plus")
+                if (name == "plus")
                 {
                     // Check if there is an overflow
                     if (applyVisitor(FieldVisitorAccurateLess(), left_point, right_point)
@@ -1432,7 +1361,7 @@ public:
                     return {false, true, false};
             }
         }
-        if (name_ == "divide" || name_ == "intDiv")
+        if (name == "divide" || name == "intDiv")
         {
             // const / variable
             if (left.column && isColumnConst(*left.column))
@@ -1475,7 +1404,6 @@ private:
     DataTypePtr return_type;
 };
 
-
 template <template <typename, typename> class Op, typename Name, bool valid_on_default_arguments = true>
 class BinaryArithmeticOverloadResolver : public IFunctionOverloadResolverImpl
 {
@@ -1524,5 +1452,4 @@ public:
 private:
     const Context & context;
 };
-
 }

From 2db721b647b9a413cf6d49324881f050652b8506 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Tue, 22 Dec 2020 01:37:07 +0300
Subject: [PATCH 238/504] further rearrangement

---
 src/Functions/FunctionBinaryArithmetic.h | 134 ++++++++++++-----------
 1 file changed, 73 insertions(+), 61 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index a9bff29fca6..ec5c9e8bd67 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -307,8 +307,9 @@ private:
         ColumnVector<ResultType>>::Container;
 
 public:
-    template <OpCase op_case, bool is_decimal_a, bool is_decimal_b, bool scale_is_left_scale, class A, class B>
-    static void NO_INLINE process(const A & a, const B & b, ResultContainerType & c, NativeResultType scale)
+    template <OpCase op_case, bool is_decimal_a, bool is_decimal_b, class A, class B>
+    static void NO_INLINE process(const A & a, const B & b, ResultContainerType & c,
+        NativeResultType scale_a, NativeResultType scale_b)
     {
         if constexpr(op_case == OpCase::LeftConstant) static_assert(!IsDecimalNumber<A>);
         if constexpr(op_case == OpCase::RightConstant) static_assert(!IsDecimalNumber<B>);
@@ -317,13 +318,22 @@ public:
 
         if constexpr (is_plus_minus_compare)
         {
-            if (scale != 1)
+            if (scale_a != 1)
             {
                 for (size_t i = 0; i < size; ++i)
-                    c[i] = applyScaled<scale_is_left_scale>(
+                    c[i] = applyScaled<true>(
                         unwrap<op_case, OpCase::LeftConstant>(a, i),
                         unwrap<op_case, OpCase::RightConstant>(b, i),
-                        scale);
+                        scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                for (size_t i = 0; i < size; ++i)
+                    c[i] = applyScaled<false>(
+                        unwrap<op_case, OpCase::LeftConstant>(a, i),
+                        unwrap<op_case, OpCase::RightConstant>(b, i),
+                        scale_b);
                 return;
             }
         }
@@ -333,7 +343,7 @@ public:
                 c[i] = applyScaledDiv<is_decimal_a>(
                     unwrap<op_case, OpCase::LeftConstant>(a, i),
                     unwrap<op_case, OpCase::RightConstant>(b, i),
-                    scale);
+                    scale_a);
             return;
         }
 
@@ -343,17 +353,21 @@ public:
                 unwrap<op_case, OpCase::RightConstant>(b, i));
     }
 
-    template <bool is_decimal_a, bool is_decimal_b, bool scale_is_left_scale, class A, class B>
-    static ResultType process(A a, B b, NativeResultType scale)
+    template <bool is_decimal_a, bool is_decimal_b, class A, class B>
+    static ResultType process(A a, B b, NativeResultType scale_a, NativeResultType scale_b)
     {
         static_assert(!IsDecimalNumber<A>);
         static_assert(!IsDecimalNumber<B>);
 
         if constexpr (is_division && is_decimal_b)
-            return applyScaledDiv<is_decimal_a>(a, b, scale);
+            return applyScaledDiv<is_decimal_a>(a, b, scale_a);
         else if constexpr (is_plus_minus_compare)
-            if (scale != 1)
-                return applyScaled<scale_is_left_scale>(a, b, scale);
+        {
+            if (scale_a != 1)
+                return applyScaled<true>(a, b, scale_a);
+            if (scale_b != 1)
+                return applyScaled<false>(a, b, scale_b);
+        }
 
         return apply(a, b);
     }
@@ -714,12 +728,21 @@ class FunctionBinaryArithmetic : public IFunction
             return col_const->template getValue<T>();
     }
 
+    template <OpCase op_case, bool left_decimal, bool right_decimal, class OpImpl, class OpImplCheck>
+    void helperInvokeEither(const auto& left, const auto& right, auto& vec_res, auto scale_a, auto scale_b) const
+    {
+        if (check_decimal_overflow)
+            OpImplCheck::template process<op_case, left_decimal, right_decimal>(left, right, vec_res, scale_a, scale_b);
+        else
+            OpImpl::template process<op_case, left_decimal, right_decimal>(left, right, vec_res, scale_a, scale_b);
+    }
+
     template <class LeftDataType, class RightDataType, class ResultDataType>
     ColumnPtr executeNumericWithDecimal(
         const auto& left, const auto& right,
         const ColumnConst * const col_left_const, const ColumnConst * const col_right_const,
         const auto * const col_left, const auto * const col_right,
-        size_t col_left_size)
+        size_t col_left_size) const
     {
         using T0 = typename LeftDataType::FieldType;
         using T1 = typename RightDataType::FieldType;
@@ -747,24 +770,25 @@ class FunctionBinaryArithmetic : public IFunction
                 return decimalResultType<is_multiply, is_division>(left, right);
         }();
 
-        ResultType left_scale;
-        ResultType right_scale;
+        const ResultType scale_a = [&] {
+            if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
+                return right.getScaleMultiplier();
+            else if constexpr (result_is_decimal)
+                return type.scaleFactorFor(left, is_multiply);
+            else if constexpr(left_is_decimal)
+                return left.getScale();
+            else
+                return 1; //won't be used, just to silence the warning
+        }();
 
-        if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-            left_scale = right.getScaleMultiplier();
-        else if constexpr (result_is_decimal)
-            left_scale = type.scaleFactorFor(left, is_multiply);
-        else if constexpr(left_is_decimal) // BUG precision loss
-            left_scale = left.getScale();
-        else
-            left_scale = 1; //won't be used, just to silence the warning
-
-        if constexpr (result_is_decimal)
-            right_scale = type.scaleFactorFor(right, is_multiply || is_division);
-        else if constexpr(right_is_decimal)
-            right_scale = right.getScale();
-        else
-            right_scale = 1; //same
+        const ResultType scale_b = [&] {
+            if constexpr (result_is_decimal)
+                return type.scaleFactorFor(right, is_multiply || is_division);
+            else if constexpr(right_is_decimal)
+                return right.getScale();
+            else
+                return 1; //same
+        }();
 
         /// non-vector result
         if (col_left_const && col_right_const)
@@ -772,11 +796,9 @@ class FunctionBinaryArithmetic : public IFunction
             const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
             const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
-            auto res = check_decimal_overflow ?
-                OpImplCheck::template constantConstant<left_is_decimal, right_is_decimal>(
-                    const_a, const_b, left_scale, right_scale) :
-                OpImpl::template constantConstant<left_is_decimal, right_is_decimal>(
-                    const_a, const_b, left_scale, right_scale);
+            const ResultType res = check_decimal_overflow
+                ? OpImplCheck::template process<left_is_decimal, right_is_decimal>(const_a, const_b, scale_a, scale_b)
+                : OpImpl::template process<left_is_decimal, right_is_decimal>(const_a, const_b, scale_a, scale_b);
 
             if constexpr (result_is_decimal)
                 return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
@@ -795,34 +817,22 @@ class FunctionBinaryArithmetic : public IFunction
 
         if (col_left && col_right)
         {
-            if (check_decimal_overflow)
-                OpImplCheck::template vectorVector<left_is_decimal, right_is_decimal>(
-                    col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
-            else
-                OpImpl::template vectorVector<left_is_decimal, right_is_decimal>(
-                    col_left->getData(), col_right->getData(), vec_res, left_scale, right_scale);
+            helperInvokeEither<OpCase::Vector, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
+                col_left->getData(), col_right->getData(), vec_res, scale_a, scale_b);
         }
         else if (col_left_const && col_right)
         {
             const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
 
-            if (check_decimal_overflow)
-                OpImplCheck::template constantVector<left_is_decimal, right_is_decimal>(
-                    const_a, col_right->getData(), vec_res, left_scale, right_scale);
-            else
-                OpImpl::template constantVector<left_is_decimal, right_is_decimal>(
-                    const_a, col_right->getData(), vec_res, left_scale, right_scale);
+            helperInvokeEither<OpCase::LeftConstant, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
+                const_a, col_right->getData(), vec_res, scale_a, scale_b);
         }
         else if (col_left && col_right_const)
         {
             const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
-            if (check_decimal_overflow)
-                OpImplCheck::template vectorConstant<left_is_decimal, right_is_decimal>(
-                    col_left->getData(), const_b, vec_res, left_scale, right_scale);
-            else
-                OpImpl::template vectorConstant<left_is_decimal, right_is_decimal>(
-                    col_left->getData(), const_b, vec_res, left_scale, right_scale);
+            helperInvokeEither<OpCase::RightConstant, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
+                col_left->getData(), const_b, vec_res, scale_a, scale_b);
         }
         else
             return nullptr;
@@ -980,11 +990,11 @@ public:
 
                 out_chars.resize(col_left->getN());
 
-                OpImpl::process<OpCase::Vector>(
+                OpImpl::template process<OpCase::Vector>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
-                    out_chars.size());
+                    out_chars.size(), {});
 
                 return ColumnConst::create(std::move(col_res), col_left_raw->size());
             }
@@ -1013,15 +1023,15 @@ public:
 
             if (!is_left_column_const && !is_right_column_const)
             {
-                OpImpl::process<OpCase::Vector>(
+                OpImpl::template process<OpCase::Vector>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
-                    out_chars.size());
+                    out_chars.size(), {});
             }
             else if (is_left_column_const)
             {
-                OpImpl::process<OpCase::LeftConstant>(
+                OpImpl::template process<OpCase::LeftConstant>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
@@ -1030,7 +1040,7 @@ public:
             }
             else
             {
-                OpImpl::process<OpCase::RightConstant>(
+                OpImpl::template process<OpCase::RightConstant>(
                     col_left->getChars().data(),
                     col_right->getChars().data(),
                     out_chars.data(),
@@ -1053,6 +1063,8 @@ public:
         if constexpr (std::is_same_v<ResultDataType, InvalidType>)
             return nullptr;
 
+        static_assert(!std::is_same_v<ResultDataType, InvalidType>);
+
         using T0 = typename LeftDataType::FieldType;
         using T1 = typename RightDataType::FieldType;
         using ResultType = typename ResultDataType::FieldType;
@@ -1096,7 +1108,7 @@ public:
 
         if (col_left && col_right)
         {
-            OpImpl::process<OpCase::Vector>(
+            OpImpl::template process<OpCase::Vector>(
                 col_left->getData().data(),
                 col_right->getData().data(),
                 vec_res.data(),
@@ -1104,7 +1116,7 @@ public:
         }
         else if (col_left_const && col_right)
         {
-            OpImpl::process<OpCase::LeftConstant>(
+            OpImpl::template process<OpCase::LeftConstant>(
                 col_left_const->template getValue<T0>(),
                 col_right->getData().data(),
                 vec_res.data(),
@@ -1112,7 +1124,7 @@ public:
         }
         else if (col_left && col_right_const)
         {
-            OpImpl::process<OpCase::RightConstant>(
+            OpImpl::template process<OpCase::RightConstant>(
                 col_left->getData().data(),
                 col_right_const->template getValue<T1>(),
                 vec_res.data(),

From af4d332999d5722c22df49380549eb8f77580c19 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Tue, 22 Dec 2020 02:17:02 +0300
Subject: [PATCH 239/504] Fix log message for memory tracking drift

---
 src/Interpreters/AsynchronousMetrics.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Interpreters/AsynchronousMetrics.cpp b/src/Interpreters/AsynchronousMetrics.cpp
index 40a3aa520fa..6388302be83 100644
--- a/src/Interpreters/AsynchronousMetrics.cpp
+++ b/src/Interpreters/AsynchronousMetrics.cpp
@@ -219,7 +219,7 @@ void AsynchronousMetrics::update()
                 ReadableSize(amount),
                 ReadableSize(peak),
                 ReadableSize(new_peak),
-                ReadableSize(new_peak - peak)
+                ReadableSize(new_peak - amount)
             );
 
             total_memory_tracker.set(new_peak);

From 77b816caea50a9636e04a047c413e9945d2f66ce Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 02:47:14 +0300
Subject: [PATCH 240/504] cleanup

---
 src/AggregateFunctions/IAggregateFunction.h | 9 ++++++---
 src/Interpreters/InterpreterSelectQuery.cpp | 1 -
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/src/AggregateFunctions/IAggregateFunction.h b/src/AggregateFunctions/IAggregateFunction.h
index b591bd3acd7..d5e931ccc73 100644
--- a/src/AggregateFunctions/IAggregateFunction.h
+++ b/src/AggregateFunctions/IAggregateFunction.h
@@ -104,9 +104,12 @@ public:
         return false;
     }
 
-    /// Inserts results into a column.
-    /// This method must be called once, from single thread.
-    /// After this method was called for state, you can't do anything with state but destroy.
+    /// Inserts results into a column. This method might modify the state (e.g.
+    /// sort an array), so must be called once, from single thread. The state
+    /// must remain valid though, and the subsequent calls to add/merge/
+    /// insertResultInto must work correctly. This kind of call sequence occurs
+    /// in `runningAccumulate`, or when calculating an aggregate function as a
+    /// window function.
     virtual void insertResultInto(AggregateDataPtr place, IColumn & to, Arena * arena) const = 0;
 
     /// Used for machine learning methods. Predict result from trained model.
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index e81447dec28..45f550eed5f 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1784,7 +1784,6 @@ void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
             + w.window_name + "'");
         query_plan.addStep(std::move(merge_sorting_step));
 
-
         // First MergeSorted, now MergingSorted......
         auto merging_sorted = std::make_unique<MergingSortedStep>(
             query_plan.getCurrentDataStream(),

From 52ace7a3d1cce636c2a1116824746f0b6cc9c1d5 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 04:37:45 +0300
Subject: [PATCH 241/504] one plan step per window

---
 src/Interpreters/AggregateDescription.h       |  5 ++
 src/Interpreters/ExpressionAnalyzer.cpp       | 38 +++++++++-----
 src/Interpreters/ExpressionAnalyzer.h         |  3 +-
 src/Interpreters/InterpreterSelectQuery.cpp   | 12 ++---
 src/Interpreters/TreeRewriter.cpp             |  2 +-
 src/Interpreters/TreeRewriter.h               |  2 +-
 src/Parsers/ASTFunction.cpp                   | 29 +++++++++--
 src/Parsers/ASTFunction.h                     |  5 ++
 src/Parsers/IAST.h                            |  1 +
 .../01591_window_functions.reference          | 50 +++++++++++++++++++
 .../0_stateless/01591_window_functions.sql    | 10 ++++
 11 files changed, 131 insertions(+), 26 deletions(-)

diff --git a/src/Interpreters/AggregateDescription.h b/src/Interpreters/AggregateDescription.h
index 3cbba101baa..400b63bc4dc 100644
--- a/src/Interpreters/AggregateDescription.h
+++ b/src/Interpreters/AggregateDescription.h
@@ -62,6 +62,11 @@ struct WindowDescription
 
     // No frame info as of yet.
 
+
+    // Reverse map to function descriptions, for convenience of building the
+    // plan. Just copy them because it's more convenient.
+    std::vector<WindowFunctionDescription> window_functions;
+
     std::string dump() const;
 };
 
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 42f9e5598fd..53b1dbfd9be 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -475,22 +475,20 @@ bool ExpressionAnalyzer::makeAggregateDescriptions(ActionsDAGPtr & actions)
 bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
 {
     // Convenient to check here because at least we have the Context.
-    if (!windowFunctions().empty() &&
+    if (!syntax->window_function_asts.empty() &&
         !context.getSettingsRef().allow_experimental_window_functions)
     {
         throw Exception(ErrorCodes::NOT_IMPLEMENTED,
             "Window functions are not implemented (while processing '{}')",
-            windowFunctions()[0]->formatForErrorMessage());
+            syntax->window_function_asts[0]->formatForErrorMessage());
     }
 
-    int next_window_index = 1;
-    for (const ASTFunction * function_node : windowFunctions())
+    for (const ASTFunction * function_node : syntax->window_function_asts)
     {
         assert(function_node->is_window_function);
 
         WindowDescription window_description;
-        window_description.window_name = fmt::format("window_{}",
-            next_window_index++);
+        window_description.window_name = function_node->getWindowDescription();
 
         if (function_node->window_partition_by)
         {
@@ -581,12 +579,28 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
                 window_function.argument_types,
                 window_function.function_parameters, properties);
 
-        window_descriptions.insert({window_description.window_name,
-            window_description});
-        window_functions.push_back(window_function);
+        window_function_descriptions.push_back(window_function);
+
+        if (auto it = window_descriptions.find(window_description.window_name);
+            it != window_descriptions.end())
+        {
+            assert(it->second.full_sort_description
+                == window_description.full_sort_description);
+        }
+        else
+        {
+            window_descriptions.insert({window_description.window_name,
+                window_description});
+        }
     }
 
-    return !windowFunctions().empty();
+    // Populate the reverse map.
+    for (const auto & f : window_function_descriptions)
+    {
+        window_descriptions[f.window_name].window_functions.push_back(f);
+    }
+
+    return !window_function_descriptions.empty();
 }
 
 
@@ -965,7 +979,7 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
 {
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
-    for (const auto & f : window_functions)
+    for (const auto & f : window_function_descriptions)
     {
         // Requiring a constant reference to a shared pointer to non-const AST
         // doesn't really look sane, but the visitor does indeed require it.
@@ -982,7 +996,7 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
     }
 
     // 2) mark the columns that are really required:
-    for (const auto & f : window_functions)
+    for (const auto & f : window_function_descriptions)
     {
         for (const auto & a : f.function_node->arguments->children)
         {
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index b90c74abb62..0d550b10668 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -62,7 +62,7 @@ struct ExpressionAnalyzerData
 
     bool has_window = false;
     WindowDescriptions window_descriptions;
-    WindowFunctionDescriptions window_functions;
+    WindowFunctionDescriptions window_function_descriptions;
     NamesAndTypesList window_columns;
 
     bool has_global_subqueries = false;
@@ -141,7 +141,6 @@ protected:
     const TableJoin & analyzedJoin() const { return *syntax->analyzed_join; }
     const NamesAndTypesList & sourceColumns() const { return syntax->required_source_columns; }
     const std::vector<const ASTFunction *> & aggregates() const { return syntax->aggregates; }
-    const std::vector<const ASTFunction *> & windowFunctions() const { return syntax->window_functions; }
     /// Find global subqueries in the GLOBAL IN/JOIN sections. Fills in external_tables.
     void initGlobalSubqueriesAndExternalTables(bool do_global);
 
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 45f550eed5f..13170726a11 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1754,10 +1754,8 @@ void InterpreterSelectQuery::executeExpression(QueryPlan & query_plan, const Act
 
 void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
 {
-    for (const auto & f : query_analyzer->window_functions)
+    for (const auto & [_, w] : query_analyzer->window_descriptions)
     {
-        const auto & w = query_analyzer->window_descriptions[f.window_name];
-
         const Settings & settings = context->getSettingsRef();
 
         auto partial_sorting = std::make_unique<PartialSortingStep>(
@@ -1784,7 +1782,7 @@ void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
             + w.window_name + "'");
         query_plan.addStep(std::move(merge_sorting_step));
 
-        // First MergeSorted, now MergingSorted......
+        // First MergeSorted, now MergingSorted.
         auto merging_sorted = std::make_unique<MergingSortedStep>(
             query_plan.getCurrentDataStream(),
             w.full_sort_description,
@@ -1797,9 +1795,9 @@ void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
         auto window_step = std::make_unique<WindowStep>(
             query_plan.getCurrentDataStream(),
             w,
-            std::vector<WindowFunctionDescription>(1, f));
-        window_step->setStepDescription("Window step for function '"
-            + f.column_name + "'");
+            w.window_functions);
+        window_step->setStepDescription("Window step for window '"
+            + w.window_name + "'");
 
         query_plan.addStep(std::move(window_step));
     }
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index f72d36400ef..0dee5a8648d 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -751,7 +751,7 @@ TreeRewriterResultPtr TreeRewriter::analyzeSelect(
     collectJoinedColumns(*result.analyzed_join, *select_query, tables_with_columns, result.aliases);
 
     result.aggregates = getAggregates(query, *select_query);
-    result.window_functions = getWindowFunctions(query, *select_query);
+    result.window_function_asts = getWindowFunctions(query, *select_query);
     result.collectUsedColumns(query, true);
     result.ast_join = select_query->join();
 
diff --git a/src/Interpreters/TreeRewriter.h b/src/Interpreters/TreeRewriter.h
index 7d801ef9d83..d9f98ee40bd 100644
--- a/src/Interpreters/TreeRewriter.h
+++ b/src/Interpreters/TreeRewriter.h
@@ -35,7 +35,7 @@ struct TreeRewriterResult
     Aliases aliases;
     std::vector<const ASTFunction *> aggregates;
 
-    std::vector<const ASTFunction *> window_functions;
+    std::vector<const ASTFunction *> window_function_asts;
 
     /// Which column is needed to be ARRAY-JOIN'ed to get the specified.
     /// For example, for `SELECT s.v ... ARRAY JOIN a AS s` will get "s.v" -> "a.v".
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index 569c710db03..d5d03a540c9 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -475,21 +475,44 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
     }
 
     settings.ostr << " OVER (";
+    appendWindowDescription(settings, state, nested_dont_need_parens);
+    settings.ostr << ")";
+}
+
+std::string ASTFunction::getWindowDescription() const
+{
+    WriteBufferFromOwnString ostr;
+    FormatSettings settings{ostr, true /* one_line */};
+    FormatState state;
+    FormatStateStacked frame;
+    appendWindowDescription(settings, state, frame);
+    return ostr.str();
+}
+
+void ASTFunction::appendWindowDescription(const FormatSettings & settings,
+    FormatState & state, FormatStateStacked frame) const
+{
+    if (!is_window_function)
+    {
+        return;
+    }
+
     if (window_partition_by)
     {
         settings.ostr << "PARTITION BY ";
-        window_partition_by->formatImpl(settings, state, nested_dont_need_parens);
+        window_partition_by->formatImpl(settings, state, frame);
     }
+
     if (window_partition_by && window_order_by)
     {
         settings.ostr << " ";
     }
+
     if (window_order_by)
     {
         settings.ostr << "ORDER BY ";
-        window_order_by->formatImpl(settings, state, nested_dont_need_parens);
+        window_order_by->formatImpl(settings, state, frame);
     }
-    settings.ostr << ")";
 }
 
 }
diff --git a/src/Parsers/ASTFunction.h b/src/Parsers/ASTFunction.h
index 02f6f46f020..38e5f3f095c 100644
--- a/src/Parsers/ASTFunction.h
+++ b/src/Parsers/ASTFunction.h
@@ -39,6 +39,11 @@ public:
 
     ASTPtr toLiteral() const;  // Try to convert functions like Array or Tuple to a literal form.
 
+    void appendWindowDescription(const FormatSettings & settings,
+        FormatState & state, FormatStateStacked frame) const;
+
+    std::string getWindowDescription() const;
+
 protected:
     void formatImplWithoutAlias(const FormatSettings & settings, FormatState & state, FormatStateStacked frame) const override;
     void appendColumnNameImpl(WriteBuffer & ostr) const override;
diff --git a/src/Parsers/IAST.h b/src/Parsers/IAST.h
index 47cbcfde4d3..bed6c5bcdf9 100644
--- a/src/Parsers/IAST.h
+++ b/src/Parsers/IAST.h
@@ -158,6 +158,7 @@ public:
         bool always_quote_identifiers = false;
         IdentifierQuotingStyle identifier_quoting_style = IdentifierQuotingStyle::Backticks;
 
+        // Newline or whitespace.
         char nl_or_ws;
 
         FormatSettings(WriteBuffer & ostr_, bool one_line_)
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 427d7a24e3a..bca5ade4672 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -87,6 +87,10 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 -- window count() is overwritten with aggregate count()
 -- select number, count(), count() over (partition by intDiv(number, 3)) from numbers(10) group by number order by count() desc;
 
+-- different windows
+-- an explain test would also be helpful, but it's too immature now and I don't
+-- want to change reference all the time
+
 9	9
 6	8
 7	8
@@ -97,3 +101,49 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 0	2
 1	2
 2	2
+select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 5) order by number) as m from numbers(31) order by number settings max_block_size = 2;
+
+-- two functions over the same window
+-- an explain test would also be helpful, but it's too immature now and I don't
+-- want to change reference all the time
+
+0	2	1
+1	2	2
+2	2	3
+3	5	4
+4	5	5
+5	5	1
+6	8	2
+7	8	3
+8	8	4
+9	11	5
+10	11	1
+11	11	2
+12	14	3
+13	14	4
+14	14	5
+15	17	1
+16	17	2
+17	17	3
+18	20	4
+19	20	5
+20	20	1
+21	23	2
+22	23	3
+23	23	4
+24	26	5
+25	26	1
+26	26	2
+27	29	3
+28	29	4
+29	29	5
+30	30	1
+select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 3) order by number desc) as m from numbers(7) order by number settings max_block_size = 2;
+
+0	2	3
+1	2	2
+2	2	1
+3	5	3
+4	5	2
+5	5	1
+6	6	1
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index dce7069ed5f..a28d435d3f8 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -30,3 +30,13 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 -- this one doesn't work yet -- looks like the column names clash, and the
 -- window count() is overwritten with aggregate count()
 -- select number, count(), count() over (partition by intDiv(number, 3)) from numbers(10) group by number order by count() desc;
+
+-- different windows
+-- an explain test would also be helpful, but it's too immature now and I don't
+-- want to change reference all the time
+select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 5) order by number) as m from numbers(31) order by number settings max_block_size = 2;
+
+-- two functions over the same window
+-- an explain test would also be helpful, but it's too immature now and I don't
+-- want to change reference all the time
+select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 3) order by number desc) as m from numbers(7) order by number settings max_block_size = 2;

From abad6b9e49dd18de6e06bee04ae2da43afdc143e Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 04:41:02 +0300
Subject: [PATCH 242/504] clang-tidy is so annoying

---
 src/Processors/Transforms/WindowTransform.h | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
index 89e09d15763..7fecbc75c86 100644
--- a/src/Processors/Transforms/WindowTransform.h
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -33,10 +33,11 @@ class WindowTransform : public ISimpleTransform
 {
 public:
     WindowTransform(
-            const Block & input_header,
-            const Block & output_header,
-            const WindowDescription & window_description,
-            const std::vector<WindowFunctionDescription> & window_functions);
+            const Block & input_header_,
+            const Block & output_header_,
+            const WindowDescription & window_description_,
+            const std::vector<WindowFunctionDescription> &
+                window_functions_descriptions);
 
     ~WindowTransform() override;
 

From f9012b12fbd74bb5e5534033a9405ffc34c0a38d Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 05:57:22 +0300
Subject: [PATCH 243/504] Proper implementation

---
 base/daemon/BaseDaemon.cpp | 216 +++++++++++++++++++++++++++++--------
 base/daemon/BaseDaemon.h   |  13 ++-
 programs/server/Server.cpp |   1 +
 3 files changed, 183 insertions(+), 47 deletions(-)

diff --git a/base/daemon/BaseDaemon.cpp b/base/daemon/BaseDaemon.cpp
index 331f9da56dd..f054be6d713 100644
--- a/base/daemon/BaseDaemon.cpp
+++ b/base/daemon/BaseDaemon.cpp
@@ -4,6 +4,8 @@
 #include <sys/stat.h>
 #include <sys/types.h>
 #include <sys/time.h>
+#include <sys/wait.h>
+#include <sys/resource.h>
 #include <fcntl.h>
 #include <errno.h>
 #include <string.h>
@@ -12,7 +14,6 @@
 #include <unistd.h>
 
 #include <typeinfo>
-#include <sys/resource.h>
 #include <iostream>
 #include <fstream>
 #include <sstream>
@@ -478,7 +479,7 @@ void BaseDaemon::terminate()
 void BaseDaemon::kill()
 {
     dumpCoverageReportIfPossible();
-    pid.reset();
+    pid_file.reset();
     if (::raise(SIGKILL) != 0)
         throw Poco::SystemException("cannot kill process");
 }
@@ -648,10 +649,6 @@ void BaseDaemon::initialize(Application & self)
             throw Poco::OpenFileException("Cannot attach stdout to " + stdout_path);
     }
 
-    /// Create pid file.
-    if (config().has("pid"))
-        pid.emplace(config().getString("pid"), DB::StatusFile::write_pid);
-
     /// Change path for logging.
     if (!log_path.empty())
     {
@@ -667,9 +664,17 @@ void BaseDaemon::initialize(Application & self)
             throw Poco::Exception("Cannot change directory to /tmp");
     }
 
-    // sensitive data masking rules are not used here
+    /// sensitive data masking rules are not used here
     buildLoggers(config(), logger(), self.commandName());
 
+    /// After initialized loggers but before initialized signal handling.
+    if (should_setup_watchdog)
+        setupWatchdog();
+
+    /// Create pid file.
+    if (config().has("pid"))
+        pid_file.emplace(config().getString("pid"), DB::StatusFile::write_pid);
+
     if (is_daemon)
     {
         /** Change working directory to the directory to write core dumps.
@@ -704,54 +709,65 @@ void BaseDaemon::initialize(Application & self)
 }
 
 
+static void addSignalHandler(const std::vector<int> & signals, signal_function handler, std::vector<int> * out_handled_signals)
+{
+    struct sigaction sa;
+    memset(&sa, 0, sizeof(sa));
+    sa.sa_sigaction = handler;
+    sa.sa_flags = SA_SIGINFO;
+
+#if defined(OS_DARWIN)
+    sigemptyset(&sa.sa_mask);
+    for (auto signal : signals)
+        sigaddset(&sa.sa_mask, signal);
+#else
+    if (sigemptyset(&sa.sa_mask))
+        throw Poco::Exception("Cannot set signal handler.");
+
+    for (auto signal : signals)
+        if (sigaddset(&sa.sa_mask, signal))
+            throw Poco::Exception("Cannot set signal handler.");
+#endif
+
+    for (auto signal : signals)
+        if (sigaction(signal, &sa, nullptr))
+            throw Poco::Exception("Cannot set signal handler.");
+
+    if (out_handled_signals)
+        std::copy(signals.begin(), signals.end(), std::back_inserter(*out_handled_signals));
+};
+
+
+static void blockSignals(const std::vector<int> & signals)
+{
+    sigset_t sig_set;
+
+    if (sigemptyset(&sig_set))
+        throw Poco::Exception("Cannot block signal.");
+
+    for (auto signal : signals)
+        if (sigaddset(&sig_set, signal))
+            throw Poco::Exception("Cannot block signal.");
+
+    if (pthread_sigmask(SIG_BLOCK, &sig_set, nullptr))
+        throw Poco::Exception("Cannot block signal.");
+};
+
+
 void BaseDaemon::initializeTerminationAndSignalProcessing()
 {
     SentryWriter::initialize(config());
     std::set_terminate(terminate_handler);
 
     /// We want to avoid SIGPIPE when working with sockets and pipes, and just handle return value/errno instead.
-    {
-        sigset_t sig_set;
-        if (sigemptyset(&sig_set) || sigaddset(&sig_set, SIGPIPE) || pthread_sigmask(SIG_BLOCK, &sig_set, nullptr))
-            throw Poco::Exception("Cannot block signal.");
-    }
+    blockSignals({SIGPIPE});
 
     /// Setup signal handlers.
-    auto add_signal_handler =
-        [this](const std::vector<int> & signals, signal_function handler)
-        {
-            struct sigaction sa;
-            memset(&sa, 0, sizeof(sa));
-            sa.sa_sigaction = handler;
-            sa.sa_flags = SA_SIGINFO;
-
-            {
-#if defined(OS_DARWIN)
-                sigemptyset(&sa.sa_mask);
-                for (auto signal : signals)
-                    sigaddset(&sa.sa_mask, signal);
-#else
-                if (sigemptyset(&sa.sa_mask))
-                    throw Poco::Exception("Cannot set signal handler.");
-
-                for (auto signal : signals)
-                    if (sigaddset(&sa.sa_mask, signal))
-                        throw Poco::Exception("Cannot set signal handler.");
-#endif
-
-                for (auto signal : signals)
-                    if (sigaction(signal, &sa, nullptr))
-                        throw Poco::Exception("Cannot set signal handler.");
-
-                std::copy(signals.begin(), signals.end(), std::back_inserter(handled_signals));
-            }
-        };
-
     /// SIGTSTP is added for debugging purposes. To output a stack trace of any running thread at anytime.
 
-    add_signal_handler({SIGABRT, SIGSEGV, SIGILL, SIGBUS, SIGSYS, SIGFPE, SIGPIPE, SIGTSTP}, signalHandler);
-    add_signal_handler({SIGHUP, SIGUSR1}, closeLogsSignalHandler);
-    add_signal_handler({SIGINT, SIGQUIT, SIGTERM}, terminateRequestedSignalHandler);
+    addSignalHandler({SIGABRT, SIGSEGV, SIGILL, SIGBUS, SIGSYS, SIGFPE, SIGPIPE, SIGTSTP}, signalHandler, &handled_signals);
+    addSignalHandler({SIGHUP, SIGUSR1}, closeLogsSignalHandler, &handled_signals);
+    addSignalHandler({SIGINT, SIGQUIT, SIGTERM}, terminateRequestedSignalHandler, &handled_signals);
 
 #if defined(SANITIZER)
     __sanitizer_set_death_callback(sanitizerDeathCallback);
@@ -863,7 +879,9 @@ void BaseDaemon::onInterruptSignals(int signal_id)
     if (sigint_signals_counter >= 2)
     {
         LOG_INFO(&logger(), "Received second signal Interrupt. Immediately terminate.");
-        kill();
+        call_default_signal_handler(signal_id);
+        /// If the above did not help.
+        _exit(128 + signal_id);
     }
 }
 
@@ -873,3 +891,109 @@ void BaseDaemon::waitForTerminationRequest()
     std::unique_lock<std::mutex> lock(signal_handler_mutex);
     signal_event.wait(lock, [this](){ return terminate_signals_counter > 0; });
 }
+
+
+void BaseDaemon::shouldSetupWatchdog(char * argv0_)
+{
+    should_setup_watchdog = true;
+    argv0 = argv0_;
+}
+
+
+void BaseDaemon::setupWatchdog()
+{
+    std::string original_process_name;
+    if (argv0)
+        original_process_name = argv0;
+
+    while (true)
+    {
+        pid_t pid = fork();
+
+        if (-1 == pid)
+            throw Poco::Exception("Cannot fork");
+
+        if (0 == pid)
+        {
+            logger().information("Forked a child process to watch");
+            return;
+        }
+
+        /// Change short thread name and process name.
+        setThreadName("clckhouse-watch");   /// 15 characters
+
+        if (argv0)
+        {
+            const char * new_process_name = "clickhouse-watchdog";
+            memset(argv0, 0, original_process_name.size());
+            memcpy(argv0, new_process_name, std::min(strlen(new_process_name), original_process_name.size()));
+        }
+
+        logger().information(fmt::format("Will watch for the process with pid {}", pid));
+
+        /// Ignore signals that only need to be delivered to the child process.
+        addSignalHandler(
+            {SIGHUP, SIGUSR1, SIGINT, SIGQUIT, SIGTERM},
+            [](int, siginfo_t *, void *) {}, nullptr);
+
+        int status = 0;
+        do
+        {
+            if (-1 != waitpid(pid, &status, WUNTRACED | WCONTINUED) || errno == ECHILD)
+            {
+                if (WIFSTOPPED(status))
+                    logger().warning(fmt::format("Child process was stopped by signal {}.", WSTOPSIG(status)));
+                else if (WIFCONTINUED(status))
+                    logger().warning(fmt::format("Child process was continued."));
+                else
+                    break;
+            }
+            else if (errno != EINTR)
+                throw Poco::Exception("Cannot waitpid, errno: " + std::string(strerror(errno)));
+        } while (true);
+
+        if (errno == ECHILD)
+        {
+            logger().information("Child process no longer exists.");
+            _exit(status);
+        }
+
+        if (WIFEXITED(status))
+        {
+            logger().information(fmt::format("Child process exited normally with code {}.", WEXITSTATUS(status)));
+            _exit(status);
+        }
+
+        if (WIFSIGNALED(status))
+        {
+            int sig = WTERMSIG(status);
+
+            if (sig == SIGKILL)
+            {
+                logger().fatal(fmt::format("Child process was terminated by signal {} (KILL)."
+                    " If it is not done by 'forcestop' command or manually,"
+                    " the possible cause is OOM Killer (see 'dmesg' and look at the '/var/log/kern.log' for the details).", sig));
+            }
+            else
+            {
+                logger().fatal(fmt::format("Child process was terminated by signal {}.", sig));
+
+                if (sig == SIGINT || sig == SIGTERM || sig == SIGQUIT)
+                    _exit(status);
+            }
+        }
+        else
+        {
+            logger().fatal("Child process was not exited normally by unknown reason.");
+        }
+
+        /// Automatic restart is not enabled but you can play with it.
+#if 1
+        _exit(status);
+#else
+        logger().information("Will restart.");
+        if (argv0)
+            memcpy(argv0, original_process_name.c_str(), original_process_name.size());
+#endif
+    }
+}
diff --git a/base/daemon/BaseDaemon.h b/base/daemon/BaseDaemon.h
index f4d3f3dfe98..1b3661fab24 100644
--- a/base/daemon/BaseDaemon.h
+++ b/base/daemon/BaseDaemon.h
@@ -131,6 +131,11 @@ public:
     /// also doesn't close global internal pipes for signal handling
     static void closeFDs();
 
+    /// If this method is called after initialization and before run,
+    /// will fork child process and setup watchdog that will print diagnostic info, if the child terminates.
+    /// argv0 is needed to change process name (consequently, it is needed for scripts involving "pgrep", "pidof" to work correctly).
+    void shouldSetupWatchdog(char * argv0_);
+
 protected:
     /// Возвращает TaskManager приложения
     /// все методы task_manager следует вызывать из одного потока
@@ -148,6 +153,9 @@ protected:
     /// initialize termination process and signal handlers
     virtual void initializeTerminationAndSignalProcessing();
 
+    /// fork the main process and watch if it was killed
+    void setupWatchdog();
+
     /// реализация обработки сигналов завершения через pipe не требует блокировки сигнала с помощью sigprocmask во всех потоках
     void waitForTerminationRequest()
 #if defined(POCO_CLICKHOUSE_PATCH) || POCO_VERSION >= 0x02000000 // in old upstream poco not vitrual
@@ -164,7 +172,7 @@ protected:
 
     std::unique_ptr<Poco::TaskManager> task_manager;
 
-    std::optional<DB::StatusFile> pid;
+    std::optional<DB::StatusFile> pid_file;
 
     std::atomic_bool is_cancelled{false};
 
@@ -194,6 +202,9 @@ protected:
     String build_id_info;
 
     std::vector<int> handled_signals;
+
+    bool should_setup_watchdog = false;
+    char * argv0 = nullptr;
 };
 
 
diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index 9da2cf70923..b5576f95c3d 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -104,6 +104,7 @@ namespace CurrentMetrics
 int mainEntryClickHouseServer(int argc, char ** argv)
 {
     DB::Server app;
+    app.shouldSetupWatchdog(argc ? argv[0] : nullptr);
     try
     {
         return app.run(argc, argv);

From f977184028595186ebe83a3e638ba06290efe977 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 06:17:31 +0300
Subject: [PATCH 244/504] Slightly better

---
 base/daemon/BaseDaemon.cpp | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)

diff --git a/base/daemon/BaseDaemon.cpp b/base/daemon/BaseDaemon.cpp
index f054be6d713..a5c282a41e3 100644
--- a/base/daemon/BaseDaemon.cpp
+++ b/base/daemon/BaseDaemon.cpp
@@ -6,6 +6,9 @@
 #include <sys/time.h>
 #include <sys/wait.h>
 #include <sys/resource.h>
+#if defined(__linux__)
+    #include <sys/prctl.h>
+#endif
 #include <fcntl.h>
 #include <errno.h>
 #include <string.h>
@@ -908,7 +911,8 @@ void BaseDaemon::setupWatchdog()
 
     while (true)
     {
-        pid_t pid = fork();
+        static pid_t pid = -1;
+        pid = fork();
 
         if (-1 == pid)
             throw Poco::Exception("Cannot fork");
@@ -916,6 +920,10 @@ void BaseDaemon::setupWatchdog()
         if (0 == pid)
         {
             logger().information("Forked a child process to watch");
+#if defined(__linux__)
+            if (0 != prctl(PR_SET_PDEATHSIG, SIGKILL))
+                logger().warning("Cannot do prctl to ask termination with parent.");
+#endif
             return;
         }
 
@@ -931,10 +939,20 @@ void BaseDaemon::setupWatchdog()
 
         logger().information(fmt::format("Will watch for the process with pid {}", pid));
 
-        /// Ignore signals that only need to be delivered to the child process.
+        /// Forward signals to the child process.
         addSignalHandler(
             {SIGHUP, SIGUSR1, SIGINT, SIGQUIT, SIGTERM},
-            [](int, siginfo_t *, void *) {}, nullptr);
+            [](int sig, siginfo_t *, void *)
+            {
+                /// Forward all signals except INT as it is send by shell to both processes when user press Ctrl+C.
+                if (sig == SIGINT)
+                    return;
+
+                const char * error_message = "Cannot forward signal to the child process.\n";
+                if (0 != ::kill(pid, sig))
+                    (void)write(STDERR_FILENO, error_message, strlen(error_message));
+            },
+            nullptr);
 
         int status = 0;
         do

From dadfc76f8fcdbc973b95ada5c259b84765d1823a Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 07:02:39 +0300
Subject: [PATCH 245/504] fix the fuzzer

---
 programs/client/QueryFuzzer.cpp             | 41 ++++++++++++++++++++-
 programs/client/QueryFuzzer.h               | 13 +++++++
 src/Processors/Transforms/WindowTransform.h |  2 +-
 3 files changed, 53 insertions(+), 3 deletions(-)

diff --git a/programs/client/QueryFuzzer.cpp b/programs/client/QueryFuzzer.cpp
index 99b6a977826..90f3addd916 100644
--- a/programs/client/QueryFuzzer.cpp
+++ b/programs/client/QueryFuzzer.cpp
@@ -327,11 +327,44 @@ void QueryFuzzer::fuzz(ASTs & asts)
     }
 }
 
+struct ScopedIncrement
+{
+    size_t & counter;
+
+    ScopedIncrement(size_t & counter_) : counter(counter_) { ++counter; }
+    ~ScopedIncrement() { --counter; }
+};
+
 void QueryFuzzer::fuzz(ASTPtr & ast)
 {
     if (!ast)
         return;
 
+    // Check for exceeding max depth.
+    ScopedIncrement depth_increment(current_ast_depth);
+    if (current_ast_depth > 1000)
+    {
+        // The AST is too deep (see the comment for current_ast_depth). Probably
+        // we don't have to print the tree because it's slow...
+        fmt::print(stderr, "The AST is too deep (depth {}, {} visited nodes)"
+            " :\n{}\n", current_ast_depth, debug_visited_nodes.size(),
+            (*debug_top_ast)->dumpTree());
+
+        return;
+    }
+
+    // Check for loops.
+    auto [_, inserted] = debug_visited_nodes.insert(ast.get());
+    if (!inserted)
+    {
+        fmt::print(stderr, "The AST node '{}' was already visited before."
+            " Depth {}, {} visited nodes, current top AST:\n{}\n",
+            static_cast<void *>(ast.get()), current_ast_depth,
+            debug_visited_nodes.size(), (*debug_top_ast)->dumpTree());
+        assert(false);
+    }
+
+    // The fuzzing.
     if (auto * with_union = typeid_cast<ASTSelectWithUnionQuery *>(ast.get()))
     {
         fuzz(with_union->list_of_selects);
@@ -365,13 +398,13 @@ void QueryFuzzer::fuzz(ASTPtr & ast)
         fuzzColumnLikeExpressionList(fn->arguments.get());
         fuzzColumnLikeExpressionList(fn->parameters.get());
 
-        fuzz(fn->children);
-
         if (fn->is_window_function)
         {
             fuzzColumnLikeExpressionList(fn->window_partition_by);
             fuzzOrderByList(fn->window_order_by);
         }
+
+        fuzz(fn->children);
     }
     else if (auto * select = typeid_cast<ASTSelectQuery *>(ast.get()))
     {
@@ -502,6 +535,10 @@ void QueryFuzzer::collectFuzzInfoRecurse(const ASTPtr ast)
 
 void QueryFuzzer::fuzzMain(ASTPtr & ast)
 {
+    current_ast_depth = 0;
+    debug_visited_nodes.clear();
+    debug_top_ast = &ast;
+
     collectFuzzInfoMain(ast);
     fuzz(ast);
 
diff --git a/programs/client/QueryFuzzer.h b/programs/client/QueryFuzzer.h
index ffec09c3883..e9d3f150283 100644
--- a/programs/client/QueryFuzzer.h
+++ b/programs/client/QueryFuzzer.h
@@ -26,6 +26,13 @@ struct QueryFuzzer
 {
     pcg64 fuzz_rand{randomSeed()};
 
+    // We add elements to expression lists with fixed probability. Some elements
+    // are so large, that the expected number of elements we add to them is
+    // one or higher, hence this process might never finish. Put some limit on the
+    // total depth of AST to prevent this.
+    // This field is reset for each fuzzMain() call.
+    size_t current_ast_depth = 0;
+
     // These arrays hold parts of queries that we can substitute into the query
     // we are currently fuzzing. We add some part from each new query we are asked
     // to fuzz, and keep this state between queries, so the fuzzing output becomes
@@ -39,6 +46,12 @@ struct QueryFuzzer
     std::unordered_map<std::string, ASTPtr> table_like_map;
     std::vector<ASTPtr> table_like;
 
+    // Some debug fields for detecting problematic ASTs with loops.
+    // These are reset for each fuzzMain call.
+    std::unordered_set<const IAST *> debug_visited_nodes;
+    ASTPtr * debug_top_ast;
+
+
     // This is the only function you have to call -- it will modify the passed
     // ASTPtr to point to new AST with some random changes.
     void fuzzMain(ASTPtr & ast);
diff --git a/src/Processors/Transforms/WindowTransform.h b/src/Processors/Transforms/WindowTransform.h
index 7fecbc75c86..3d284263171 100644
--- a/src/Processors/Transforms/WindowTransform.h
+++ b/src/Processors/Transforms/WindowTransform.h
@@ -37,7 +37,7 @@ public:
             const Block & output_header_,
             const WindowDescription & window_description_,
             const std::vector<WindowFunctionDescription> &
-                window_functions_descriptions);
+                window_function_descriptions);
 
     ~WindowTransform() override;
 

From 919172f9eeb641b1ab03bb82d04eaf09ba1624fa Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 07:16:56 +0300
Subject: [PATCH 246/504] more fuzzer fixes

---
 programs/client/QueryFuzzer.cpp | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/programs/client/QueryFuzzer.cpp b/programs/client/QueryFuzzer.cpp
index 90f3addd916..38f3a756b7a 100644
--- a/programs/client/QueryFuzzer.cpp
+++ b/programs/client/QueryFuzzer.cpp
@@ -29,6 +29,11 @@
 namespace DB
 {
 
+namespace ErrorCodes
+{
+    extern const int TOO_DEEP_RECURSION;
+}
+
 Field QueryFuzzer::getRandomField(int type)
 {
     switch (type)
@@ -342,15 +347,15 @@ void QueryFuzzer::fuzz(ASTPtr & ast)
 
     // Check for exceeding max depth.
     ScopedIncrement depth_increment(current_ast_depth);
-    if (current_ast_depth > 1000)
+    if (current_ast_depth > 500)
     {
-        // The AST is too deep (see the comment for current_ast_depth). Probably
-        // we don't have to print the tree because it's slow...
-        fmt::print(stderr, "The AST is too deep (depth {}, {} visited nodes)"
-            " :\n{}\n", current_ast_depth, debug_visited_nodes.size(),
-            (*debug_top_ast)->dumpTree());
-
-        return;
+        // The AST is too deep (see the comment for current_ast_depth). Throw
+        // an exception to fail fast and not use this query as an etalon, or we'll
+        // end up in a very slow and useless loop. It also makes sense to set it
+        // lower than the default max parse depth on the server (1000), so that
+        // we don't get the useless error about parse depth from the server either.
+        throw Exception(ErrorCodes::TOO_DEEP_RECURSION,
+            "AST depth exceeded while fuzzing ({})", current_ast_depth);
     }
 
     // Check for loops.

From 3dafd4bb49bc83f7de64886144cc0ca80cba055f Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 09:45:08 +0300
Subject: [PATCH 247/504] Minor changes

---
 base/daemon/BaseDaemon.cpp | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/base/daemon/BaseDaemon.cpp b/base/daemon/BaseDaemon.cpp
index a5c282a41e3..f6a904ac0a8 100644
--- a/base/daemon/BaseDaemon.cpp
+++ b/base/daemon/BaseDaemon.cpp
@@ -944,13 +944,17 @@ void BaseDaemon::setupWatchdog()
             {SIGHUP, SIGUSR1, SIGINT, SIGQUIT, SIGTERM},
             [](int sig, siginfo_t *, void *)
             {
-                /// Forward all signals except INT as it is send by shell to both processes when user press Ctrl+C.
+                /// Forward all signals except INT as it can be send by terminal to the process group when user press Ctrl+C,
+                /// and we process double delivery of this signal as immediate termination.
                 if (sig == SIGINT)
                     return;
 
                 const char * error_message = "Cannot forward signal to the child process.\n";
                 if (0 != ::kill(pid, sig))
-                    (void)write(STDERR_FILENO, error_message, strlen(error_message));
+                {
+                    auto res = write(STDERR_FILENO, error_message, strlen(error_message));
+                    (void)res;
+                }
             },
             nullptr);
 

From b81e067cd4f4d785bbb80c1ae37fa800ef5f3987 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 09:51:09 +0300
Subject: [PATCH 248/504] Update IStorage.h

---
 src/Storages/IStorage.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/src/Storages/IStorage.h b/src/Storages/IStorage.h
index 9d5a1fe85a9..7110510a6d0 100644
--- a/src/Storages/IStorage.h
+++ b/src/Storages/IStorage.h
@@ -87,7 +87,6 @@ public:
         : storage_id(std::move(storage_id_))
         , metadata(std::make_unique<StorageInMemoryMetadata>()) {} //-V730
 
-    virtual ~IStorage() override = default;
     IStorage(const IStorage &) = delete;
     IStorage & operator=(const IStorage &) = delete;
 

From dda0eb53c15cda4d9888e426a39f306f6e5b8e0a Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Tue, 22 Dec 2020 10:13:22 +0300
Subject: [PATCH 249/504] Fix variable names in AsynchronousMetrics::update()

---
 src/Interpreters/AsynchronousMetrics.cpp | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/Interpreters/AsynchronousMetrics.cpp b/src/Interpreters/AsynchronousMetrics.cpp
index 6388302be83..be56ae6700a 100644
--- a/src/Interpreters/AsynchronousMetrics.cpp
+++ b/src/Interpreters/AsynchronousMetrics.cpp
@@ -212,18 +212,18 @@ void AsynchronousMetrics::update()
         {
             Int64 amount = total_memory_tracker.get();
             Int64 peak = total_memory_tracker.getPeak();
-            Int64 new_peak = data.resident;
+            Int64 new_amount = data.resident;
 
             LOG_DEBUG(&Poco::Logger::get("AsynchronousMetrics"),
                 "MemoryTracking: was {}, peak {}, will set to {} (RSS), difference: {}",
                 ReadableSize(amount),
                 ReadableSize(peak),
-                ReadableSize(new_peak),
-                ReadableSize(new_peak - amount)
+                ReadableSize(new_amount),
+                ReadableSize(new_amount - amount)
             );
 
-            total_memory_tracker.set(new_peak);
-            CurrentMetrics::set(CurrentMetrics::MemoryTracking, new_peak);
+            total_memory_tracker.set(new_amount);
+            CurrentMetrics::set(CurrentMetrics::MemoryTracking, new_amount);
         }
     }
 #endif

From 3da533766ddb698b85c9f3bc1cab2358694d1aa5 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 10:25:45 +0300
Subject: [PATCH 250/504] clang-tidy.......

---
 programs/client/QueryFuzzer.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/programs/client/QueryFuzzer.cpp b/programs/client/QueryFuzzer.cpp
index 38f3a756b7a..53ede4a3d92 100644
--- a/programs/client/QueryFuzzer.cpp
+++ b/programs/client/QueryFuzzer.cpp
@@ -336,7 +336,7 @@ struct ScopedIncrement
 {
     size_t & counter;
 
-    ScopedIncrement(size_t & counter_) : counter(counter_) { ++counter; }
+    explicit ScopedIncrement(size_t & counter_) : counter(counter_) { ++counter; }
     ~ScopedIncrement() { --counter; }
 };
 

From 7f590090f6da8eaabe7359489e7de33e028430e1 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 10:28:02 +0300
Subject: [PATCH 251/504] Remove logs

---
 src/Common/ThreadFuzzer.cpp | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/src/Common/ThreadFuzzer.cpp b/src/Common/ThreadFuzzer.cpp
index ed155473061..88ff53534e6 100644
--- a/src/Common/ThreadFuzzer.cpp
+++ b/src/Common/ThreadFuzzer.cpp
@@ -103,12 +103,6 @@ void ThreadFuzzer::initConfiguration()
     initFromEnv(sleep_probability, "THREAD_FUZZER_SLEEP_PROBABILITY");
     initFromEnv(sleep_time_us, "THREAD_FUZZER_SLEEP_TIME_US");
 
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep probability {}", sleep_probability);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Yield probability {}", yield_probability);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Migrate probability {}", migrate_probability);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "CPU time period {}", cpu_time_period_us);
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "Sleep time us {}", sleep_time_us);
-
 #if THREAD_FUZZER_WRAP_PTHREAD
 #    define INIT_WRAPPER_PARAMS(RET, NAME, ...) \
         initFromEnv(NAME##_before_yield_probability, "THREAD_FUZZER_" #NAME "_BEFORE_YIELD_PROBABILITY"); \
@@ -125,7 +119,6 @@ void ThreadFuzzer::initConfiguration()
 
 #    undef INIT_WRAPPER_PARAMS
 #endif
-    LOG_DEBUG(&Poco::Logger::get("ThreadFuzzer"), "pthread_mutex_lock_before_migrate_probability {}", pthread_mutex_lock_before_migrate_probability);
 }
 
 

From 100f9d5de33a5137f316451739602399b4b8d72c Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 10:38:39 +0300
Subject: [PATCH 252/504] Fix error and remove useless code

---
 base/daemon/BaseDaemon.cpp | 23 +++--------------------
 base/daemon/BaseDaemon.h   | 20 --------------------
 2 files changed, 3 insertions(+), 40 deletions(-)

diff --git a/base/daemon/BaseDaemon.cpp b/base/daemon/BaseDaemon.cpp
index f6a904ac0a8..19bbca508bc 100644
--- a/base/daemon/BaseDaemon.cpp
+++ b/base/daemon/BaseDaemon.cpp
@@ -26,7 +26,6 @@
 #include <Poco/Observer.h>
 #include <Poco/AutoPtr.h>
 #include <Poco/PatternFormatter.h>
-#include <Poco/TaskManager.h>
 #include <Poco/File.h>
 #include <Poco/Path.h>
 #include <Poco/Message.h>
@@ -474,7 +473,6 @@ BaseDaemon::~BaseDaemon()
 
 void BaseDaemon::terminate()
 {
-    getTaskManager().cancelAll();
     if (::raise(SIGTERM) != 0)
         throw Poco::SystemException("cannot terminate process");
 }
@@ -568,7 +566,6 @@ void BaseDaemon::initialize(Application & self)
 {
     closeFDs();
 
-    task_manager = std::make_unique<Poco::TaskManager>();
     ServerApplication::initialize(self);
 
     /// now highest priority (lowest value) is PRIO_APPLICATION = -100, we want higher!
@@ -805,23 +802,6 @@ void BaseDaemon::logRevision() const
         + ", PID " + std::to_string(getpid()));
 }
 
-/// Makes server shutdown if at least one Poco::Task have failed.
-void BaseDaemon::exitOnTaskError()
-{
-    Poco::Observer<BaseDaemon, Poco::TaskFailedNotification> obs(*this, &BaseDaemon::handleNotification);
-    getTaskManager().addObserver(obs);
-}
-
-/// Used for exitOnTaskError()
-void BaseDaemon::handleNotification(Poco::TaskFailedNotification *_tfn)
-{
-    task_failed = true;
-    Poco::AutoPtr<Poco::TaskFailedNotification> fn(_tfn);
-    Poco::Logger * lg = &(logger());
-    LOG_ERROR(lg, "Task '{}' failed. Daemon is shutting down. Reason - {}", fn->task()->name(), fn->reason().displayText());
-    ServerApplication::terminate();
-}
-
 void BaseDaemon::defineOptions(Poco::Util::OptionSet & new_options)
 {
     new_options.addOption(
@@ -905,6 +885,9 @@ void BaseDaemon::shouldSetupWatchdog(char * argv0_)
 
 void BaseDaemon::setupWatchdog()
 {
+    /// Initialize in advance to avoid double initialization in forked processes.
+    DateLUT::instance();
+
     std::string original_process_name;
     if (argv0)
         original_process_name = argv0;
diff --git a/base/daemon/BaseDaemon.h b/base/daemon/BaseDaemon.h
index 1b3661fab24..0fb5bb47b2d 100644
--- a/base/daemon/BaseDaemon.h
+++ b/base/daemon/BaseDaemon.h
@@ -12,7 +12,6 @@
 #include <chrono>
 #include <Poco/Process.h>
 #include <Poco/ThreadPool.h>
-#include <Poco/TaskNotification.h>
 #include <Poco/Util/Application.h>
 #include <Poco/Util/ServerApplication.h>
 #include <Poco/Net/SocketAddress.h>
@@ -26,9 +25,6 @@
 #include <loggers/Loggers.h>
 
 
-namespace Poco { class TaskManager; }
-
-
 /// \brief Base class for applications that can run as daemons.
 ///
 /// \code
@@ -61,9 +57,6 @@ public:
     /// Определяет параметр командной строки
     void defineOptions(Poco::Util::OptionSet & new_options) override;
 
-    /// Заставляет демон завершаться, если хотя бы одна задача завершилась неудачно
-    void exitOnTaskError();
-
     /// Завершение демона ("мягкое")
     void terminate();
 
@@ -137,16 +130,8 @@ public:
     void shouldSetupWatchdog(char * argv0_);
 
 protected:
-    /// Возвращает TaskManager приложения
-    /// все методы task_manager следует вызывать из одного потока
-    /// иначе возможен deadlock, т.к. joinAll выполняется под локом, а любой метод тоже берет лок
-    Poco::TaskManager & getTaskManager() { return *task_manager; }
-
     virtual void logRevision() const;
 
-    /// Используется при exitOnTaskError()
-    void handleNotification(Poco::TaskFailedNotification *);
-
     /// thread safe
     virtual void handleSignal(int signal_id);
 
@@ -170,15 +155,10 @@ protected:
 
     virtual std::string getDefaultCorePath() const;
 
-    std::unique_ptr<Poco::TaskManager> task_manager;
-
     std::optional<DB::StatusFile> pid_file;
 
     std::atomic_bool is_cancelled{false};
 
-    /// Флаг устанавливается по сообщению из Task (при аварийном завершении).
-    bool task_failed = false;
-
     bool log_to_console = false;
 
     /// Событие, чтобы проснуться во время ожидания

From 74671bda4ae1f6ed8eda89200e9de65ed34555a7 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 22 Dec 2020 11:01:13 +0300
Subject: [PATCH 253/504] Fix special build

---
 src/Functions/FunctionsStringHash.cpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index ec0f5fea808..acb2feb87b8 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -295,12 +295,12 @@ class FixedHeap
 public:
     FixedHeap() = delete;
 
-    explicit FixedHeap(F f_) : f(f_), data_t(std::make_shared<std::vector<size_t>>(K, v))
+    explicit FixedHeap(F f_) : f(f_), data_t(std::make_shared<std::vector<UInt64>>(K, v))
     {
         std::make_heap(data_t->begin(), data_t->end(), f);
     }
 
-    void insertAndReplace(size_t new_v)
+    void insertAndReplace(UInt64 new_v)
     {
         data_t->push_back(new_v);
         std::push_heap(data_t->begin(), data_t->end(), f);
@@ -308,11 +308,11 @@ public:
         data_t->pop_back();
     }
 
-    const size_t * data() { return data_t->data(); }
+    const UInt64 * data() { return data_t->data(); }
 
 private:
     F f;
-    std::shared_ptr<std::vector<size_t>> data_t;
+    std::shared_ptr<std::vector<UInt64>> data_t;
 };
 
 

From 121f7ae833a7212cbe35954d783268f6cf64bcae Mon Sep 17 00:00:00 2001
From: spongedc <duchuangucas@live.com>
Date: Tue, 22 Dec 2020 16:06:33 +0800
Subject: [PATCH 254/504] Support builtin function isIPv4String && isIPv6String

---
 src/Functions/FunctionsCoding.cpp             |  2 +
 src/Functions/FunctionsCoding.h               | 93 +++++++++++++++++++
 .../00725_ipv4_ipv6_domains.reference         | 24 +++++
 .../0_stateless/00725_ipv4_ipv6_domains.sql   | 27 ++++++
 4 files changed, 146 insertions(+)

diff --git a/src/Functions/FunctionsCoding.cpp b/src/Functions/FunctionsCoding.cpp
index 97add9bf32a..c2c746aad9e 100644
--- a/src/Functions/FunctionsCoding.cpp
+++ b/src/Functions/FunctionsCoding.cpp
@@ -32,6 +32,8 @@ void registerFunctionsCoding(FunctionFactory & factory)
     factory.registerFunction<FunctionToIPv6>();
     factory.registerFunction<FunctionIPv6CIDRToRange>();
     factory.registerFunction<FunctionIPv4CIDRToRange>();
+    factory.registerFunction<FunctionIsIPv4String>();
+    factory.registerFunction<FunctionIsIPv6String>();
 }
 
 }
diff --git a/src/Functions/FunctionsCoding.h b/src/Functions/FunctionsCoding.h
index ac3262f5131..9e761ace6fb 100644
--- a/src/Functions/FunctionsCoding.h
+++ b/src/Functions/FunctionsCoding.h
@@ -1831,5 +1831,98 @@ public:
     }
 };
 
+class FunctionIsIPv4String : public FunctionIPv4StringToNum
+{
+public:
+    static constexpr auto name = "isIPv4String";
+
+    static FunctionPtr create(const Context &) { return std::make_shared<FunctionIsIPv4String>(); }
+
+    String getName() const override { return name; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (!isString(arguments[0]))
+            throw Exception("Illegal type " + arguments[0]->getName() + " of argument of function " + getName(),
+                            ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        return std::make_shared<DataTypeUInt8>();
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
+    {
+        const ColumnPtr & column = arguments[0].column;
+        if (const ColumnString * col = checkAndGetColumn<ColumnString>(column.get()))
+        {
+            auto col_res = ColumnUInt8::create();
+
+            ColumnUInt8::Container & vec_res = col_res->getData();
+            vec_res.resize(col->size());
+
+            const ColumnString::Chars & vec_src = col->getChars();
+            const ColumnString::Offsets & offsets_src = col->getOffsets();
+            size_t prev_offset = 0;
+            UInt32 result = 0;
+
+            for (size_t i = 0; i < vec_res.size(); ++i)
+            {
+                vec_res[i] = DB::parseIPv4(reinterpret_cast<const char *>(&vec_src[prev_offset]), reinterpret_cast<unsigned char*>(&result));
+                prev_offset = offsets_src[i];
+            }
+            return col_res;
+        }
+        else
+            throw Exception("Illegal column " + arguments[0].column->getName()
+                            + " of argument of function " + getName(),
+                            ErrorCodes::ILLEGAL_COLUMN);
+    }
+};
+
+class FunctionIsIPv6String : public FunctionIPv6StringToNum
+{
+public:
+    static constexpr auto name = "isIPv6String";
+
+    static FunctionPtr create(const Context &) { return std::make_shared<FunctionIsIPv6String>(); }
+
+    String getName() const override { return name; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (!isString(arguments[0]))
+            throw Exception("Illegal type " + arguments[0]->getName() + " of argument of function " + getName(),
+                            ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
+        return std::make_shared<DataTypeUInt8>();
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
+    {
+        const ColumnPtr & column = arguments[0].column;
+
+        if (const ColumnString * col = checkAndGetColumn<ColumnString>(column.get()))
+        {
+            auto col_res = ColumnUInt8::create();
+
+            ColumnUInt8::Container & vec_res = col_res->getData();
+            vec_res.resize(col->size());
+
+            const ColumnString::Chars & vec_src = col->getChars();
+            const ColumnString::Offsets & offsets_src = col->getOffsets();
+            size_t prev_offset = 0;
+            char v[IPV6_BINARY_LENGTH];
+
+            for (size_t i = 0; i < vec_res.size(); ++i)
+            {
+                vec_res[i] = DB::parseIPv6(reinterpret_cast<const char *>(&vec_src[prev_offset]), reinterpret_cast<unsigned char*>(v));
+                prev_offset = offsets_src[i];
+            }
+            return col_res;
+        }
+        else
+            throw Exception("Illegal column " + arguments[0].column->getName()
+                            + " of argument of function " + getName(),
+                            ErrorCodes::ILLEGAL_COLUMN);
+    }
+};
 
 }
diff --git a/tests/queries/0_stateless/00725_ipv4_ipv6_domains.reference b/tests/queries/0_stateless/00725_ipv4_ipv6_domains.reference
index 28051d15f65..5060b5253fe 100644
--- a/tests/queries/0_stateless/00725_ipv4_ipv6_domains.reference
+++ b/tests/queries/0_stateless/00725_ipv4_ipv6_domains.reference
@@ -25,3 +25,27 @@ ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff	FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
 > 127.0.0.1	2001:db8:ac10:fe01:feed:babe:cafe:f00d
 > 127.0.0.1	ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff
 = 127.0.0.1	::ffff:127.0.0.1
+0.0.0.0 is ipv4 string:                                 	1
+255.255.255.255 is ipv4 string:                         	1
+192.168.0.91 is ipv4 string:                            	1
+127.0.0.1 is ipv4 string:                               	1
+8.8.8.8 is ipv4 string:                                 	1
+hello is ipv4 string:                                   	0
+0:0:0:0:0:0:0:0 is ipv4 string:                         	0
+0000:0000:0000:0000:0000:FFFF:C1FC:110A is ipv4 string: 	0
+FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF is ipv4 string: 	0
+::ffff:127.0.0.1 is ipv4 string:                        	0
+::ffff:8.8.8.8 is ipv4 string:                          	0
+2001:0DB8:AC10:FE01:FEED:BABE:CAFE:F00D is ipv4 string: 	0
+0.0.0.0 is ipv6 string:                                 	0
+255.255.255.255 is ipv6 string:                         	0
+192.168.0.91 is ipv6 string:                            	0
+127.0.0.1 is ipv6 string:                               	0
+8.8.8.8 is ipv6 string:                                 	0
+hello is ipv6 string:                                   	0
+0:0:0:0:0:0:0:0 is ipv6 string:                         	1
+0000:0000:0000:0000:0000:FFFF:C1FC:110A is ipv6 string: 	1
+FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF is ipv6 string: 	1
+::ffff:127.0.0.1 is ipv6 string:                        	1
+::ffff:8.8.8.8 is ipv6 string:                          	1
+2001:0DB8:AC10:FE01:FEED:BABE:CAFE:F00D is ipv6 string: 	1
diff --git a/tests/queries/0_stateless/00725_ipv4_ipv6_domains.sql b/tests/queries/0_stateless/00725_ipv4_ipv6_domains.sql
index c40e3190ed8..099dc20762e 100644
--- a/tests/queries/0_stateless/00725_ipv4_ipv6_domains.sql
+++ b/tests/queries/0_stateless/00725_ipv4_ipv6_domains.sql
@@ -57,3 +57,30 @@ SELECT '= 127.0.0.1', ipv6_ FROM ipv6_test
 -- TODO: Assert that invalid values can't be inserted into IPv6 column.
 
 DROP TABLE IF EXISTS ipv6_test;
+
+SELECT '0.0.0.0 is ipv4 string:                                 ', isIPv4String('0.0.0.0');
+SELECT '255.255.255.255 is ipv4 string:                         ', isIPv4String('255.255.255.255');
+SELECT '192.168.0.91 is ipv4 string:                            ', isIPv4String('192.168.0.91');
+SELECT '127.0.0.1 is ipv4 string:                               ', isIPv4String('127.0.0.1');
+SELECT '8.8.8.8 is ipv4 string:                                 ', isIPv4String('8.8.8.8');
+SELECT 'hello is ipv4 string:                                   ', isIPv4String('hello');
+SELECT '0:0:0:0:0:0:0:0 is ipv4 string:                         ', isIPv4String('0:0:0:0:0:0:0:0');
+SELECT '0000:0000:0000:0000:0000:FFFF:C1FC:110A is ipv4 string: ', isIPv4String('0000:0000:0000:0000:0000:FFFF:C1FC:110A');
+SELECT 'FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF is ipv4 string: ', isIPv4String('FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF');
+SELECT '::ffff:127.0.0.1 is ipv4 string:                        ', isIPv4String('::ffff:127.0.0.1');
+SELECT '::ffff:8.8.8.8 is ipv4 string:                          ', isIPv4String('::ffff:8.8.8.8');
+SELECT '2001:0DB8:AC10:FE01:FEED:BABE:CAFE:F00D is ipv4 string: ', isIPv4String('2001:0DB8:AC10:FE01:FEED:BABE:CAFE:F00D');
+
+SELECT '0.0.0.0 is ipv6 string:                                 ', isIPv6String('0.0.0.0');
+SELECT '255.255.255.255 is ipv6 string:                         ', isIPv6String('255.255.255.255');
+SELECT '192.168.0.91 is ipv6 string:                            ', isIPv6String('192.168.0.91');
+SELECT '127.0.0.1 is ipv6 string:                               ', isIPv6String('127.0.0.1');
+SELECT '8.8.8.8 is ipv6 string:                                 ', isIPv6String('8.8.8.8');
+SELECT 'hello is ipv6 string:                                   ', isIPv6String('hello');
+SELECT '0:0:0:0:0:0:0:0 is ipv6 string:                         ', isIPv6String('0:0:0:0:0:0:0:0');
+SELECT '0000:0000:0000:0000:0000:FFFF:C1FC:110A is ipv6 string: ', isIPv6String('0000:0000:0000:0000:0000:FFFF:C1FC:110A');
+SELECT 'FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF is ipv6 string: ', isIPv6String('FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF');
+SELECT '::ffff:127.0.0.1 is ipv6 string:                        ', isIPv6String('::ffff:127.0.0.1');
+SELECT '::ffff:8.8.8.8 is ipv6 string:                          ', isIPv6String('::ffff:8.8.8.8');
+SELECT '2001:0DB8:AC10:FE01:FEED:BABE:CAFE:F00D is ipv6 string: ', isIPv6String('2001:0DB8:AC10:FE01:FEED:BABE:CAFE:F00D');
+

From b9df795595a9b9ade9596bdeb1bd07f234fa6ee7 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 11:16:36 +0300
Subject: [PATCH 255/504] Another code for TSan

---
 src/Storages/System/StorageSystemStackTrace.cpp | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index dd21495ed4b..e47852c6045 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -39,6 +39,7 @@ namespace
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.
     std::atomic<int> data_ready_num = 0;
+    std::atomic<bool> signal_latch = 0;   /// Only need for thread sanitizer.
 
     /** Notes:
       * Only one query from the table can be processed at the moment of time.
@@ -71,6 +72,10 @@ namespace
         if (notification_num != sequence_num.load(std::memory_order_acquire))
             return;
 
+        bool expected = false;
+        if (!signal_latch.compare_exchange_strong(expected, true, std::memory_order_acquire))
+            return;
+
         /// All these methods are signal-safe.
         const ucontext_t signal_context = *reinterpret_cast<ucontext_t *>(context);
         stack_trace = StackTrace(signal_context);
@@ -81,7 +86,7 @@ namespace
             memcpy(query_id_data, query_id.data, query_id_size);
 
         /// This is unneeded (because we synchronize through pipe) but makes TSan happy.
-        data_ready_num.store(sequence_num, std::memory_order_release);
+        data_ready_num.store(notification_num, std::memory_order_release);
 
         ssize_t res = ::write(notification_pipe.fds_rw[1], &notification_num, sizeof(notification_num));
 
@@ -89,6 +94,8 @@ namespace
         (void)res;
 
         errno = saved_errno;
+
+        signal_latch.store(false, std::memory_order_release);
     }
 
     /// Wait for data in pipe and read it.

From 2d28c8cb609ec423771bb02a8cebdc6b29cd1aff Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 11:20:57 +0300
Subject: [PATCH 256/504] Better runner

---
 docker/test/coverage/run.sh | 25 ++++++++++++++++---------
 1 file changed, 16 insertions(+), 9 deletions(-)

diff --git a/docker/test/coverage/run.sh b/docker/test/coverage/run.sh
index 41da390e06b..5626fbbd5d4 100755
--- a/docker/test/coverage/run.sh
+++ b/docker/test/coverage/run.sh
@@ -80,21 +80,28 @@ LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "RENAME TA
 LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "RENAME TABLE datasets.visits_v1 TO test.visits"
 LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "SHOW TABLES FROM test"
 
+LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
 
-if grep -q -- "--use-skip-list" /usr/bin/clickhouse-test; then
-    SKIP_LIST_OPT="--use-skip-list"
-fi
-
-# We can have several additional options so we path them as array because it's
-# more idiologically correct.
-read -ra ADDITIONAL_OPTIONS <<< "${ADDITIONAL_OPTIONS:-}"
-
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test --testname --shard --zookeeper --print-time "$SKIP_LIST_OPT" "${ADDITIONAL_OPTIONS[@]}" "$SKIP_TESTS_OPTION" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee test_output/test_result.txt
+readarray -t FAILED_TESTS < <(awk '/FAIL|TIMEOUT|ERROR/ { print substr($3, 1, length($3)-1) }' "/test_result.txt")
 
 kill_clickhouse
 
 sleep 3
 
+if [[ -n "${FAILED_TESTS[*]}" ]]
+then
+    # Clean the data so that there is no interference from the previous test run.
+    rm -rf /var/lib/clickhouse/{{meta,}data,user_files} ||:
+
+    start_clickhouse
+
+    echo "Going to run again: ${FAILED_TESTS[*]}"
+
+    LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test --order=random --testname --shard --zookeeper --use-skip-list "${FAILED_TESTS[@]}" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee -a "/test_result.txt"
+else
+    echo "No failed tests"
+fi
+
 mkdir -p $COVERAGE_DIR
 mv /*.profraw $COVERAGE_DIR
 

From c6d3f2722e16e155e9cee9f57f551f9a62cfeab7 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 11:21:18 +0300
Subject: [PATCH 257/504] Revert unneeded changes

---
 src/Common/StackTrace.cpp | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/Common/StackTrace.cpp b/src/Common/StackTrace.cpp
index 38a2235d516..b285a45bdc5 100644
--- a/src/Common/StackTrace.cpp
+++ b/src/Common/StackTrace.cpp
@@ -252,12 +252,12 @@ void StackTrace::symbolize(const StackTrace::FramePointers & frame_pointers, siz
 #endif
 }
 
-NO_SANITIZE_THREAD StackTrace::StackTrace()
+StackTrace::StackTrace()
 {
     tryCapture();
 }
 
-NO_SANITIZE_THREAD StackTrace::StackTrace(const ucontext_t & signal_context)
+StackTrace::StackTrace(const ucontext_t & signal_context)
 {
     tryCapture();
 
@@ -295,17 +295,17 @@ void StackTrace::tryCapture()
 #endif
 }
 
-size_t NO_SANITIZE_THREAD StackTrace::getSize() const
+size_t StackTrace::getSize() const
 {
     return size;
 }
 
-size_t NO_SANITIZE_THREAD StackTrace::getOffset() const
+size_t StackTrace::getOffset() const
 {
     return offset;
 }
 
-const NO_SANITIZE_THREAD StackTrace::FramePointers & StackTrace::getFramePointers() const
+const StackTrace::FramePointers & StackTrace::getFramePointers() const
 {
     return frame_pointers;
 }

From 5109c3420e8467e95324614e14fc8087bb70b35f Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 22 Dec 2020 11:42:09 +0300
Subject: [PATCH 258/504] Update comment.

---
 src/Processors/IProcessor.h | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/src/Processors/IProcessor.h b/src/Processors/IProcessor.h
index 08adc151424..c3abe40c3b7 100644
--- a/src/Processors/IProcessor.h
+++ b/src/Processors/IProcessor.h
@@ -195,14 +195,19 @@ public:
         throw Exception("Method 'work' is not implemented for " + getName() + " processor", ErrorCodes::NOT_IMPLEMENTED);
     }
 
-    /** You may call this method if 'prepare' returned Async.
+    /** Executor must call this method when 'prepare' returned Async.
       * This method cannot access any ports. It should use only data that was prepared by 'prepare' method.
       *
-      * This method should return instantly and fire an event (or many events) when asynchronous job will be done.
-      * When the job is not done, method 'prepare' will return Wait and the user may block and wait for next event before checking again.
+      * This method should instantly return epollable file descriptor which will be readable when asynchronous job is done.
+      * When descriptor is readable, method `work` is called to continue data processing.
       *
-      * Note that it can fire many events in EventCounter while doing its job,
-      *  and you have to wait for next event (or do something else) every time when 'prepare' returned Wait.
+      * NOTE: it would be more logical to let `work()` return ASYNC status instead of prepare. This will get
+      * prepare() -> work() -> schedule() -> work() -> schedule() -> .. -> work() -> prepare()
+      * chain instead of
+      * prepare() -> work() -> prepare() -> schedule() -> work() -> prepare() -> schedule() -> .. -> work() -> prepare()
+      *
+      * It is expected that executor epoll using level-triggered notifications.
+      * Read all available data from descriptor before returning ASYNC.
       */
     virtual int schedule()
     {

From dc450e9a83b99833451952375164767012d003ac Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 22 Dec 2020 11:55:21 +0300
Subject: [PATCH 259/504] Update comment.

---
 src/DataStreams/RemoteQueryExecutorReadContext.h | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/src/DataStreams/RemoteQueryExecutorReadContext.h b/src/DataStreams/RemoteQueryExecutorReadContext.h
index fc81132ea18..f8c64954b83 100644
--- a/src/DataStreams/RemoteQueryExecutorReadContext.h
+++ b/src/DataStreams/RemoteQueryExecutorReadContext.h
@@ -28,12 +28,18 @@ public:
     std::exception_ptr exception;
     FiberStack stack;
     boost::context::fiber fiber;
+    /// This mutex for fiber is needed because fiber could be destroyed in cancel method from another thread.
     std::mutex fiber_lock;
 
     Poco::Timespan receive_timeout;
     MultiplexedConnections & connections;
     Poco::Net::Socket * last_used_socket = nullptr;
 
+    /// Here we have three descriptors we are going to wait:
+    /// * socket_fd is a descriptor of connection. It may be changed in case of reading from several replicas.
+    /// * timer is a timerfd descriptor to manually check socket timeout
+    /// * pipe_fd is a pipe we use to cancel query and socket polling by executor.
+    /// We put those descriptors into our own epoll_fd which is used by external executor.
     TimerDescriptor timer{CLOCK_MONOTONIC, 0};
     int socket_fd = -1;
     int epoll_fd;
@@ -67,7 +73,6 @@ public:
         }
 
         auto routine = Routine{connections, *this};
-        // connection_lock = routine.connection_lock.get();
         fiber = boost::context::fiber(std::allocator_arg_t(), stack, std::move(routine));
     }
 
@@ -177,8 +182,10 @@ public:
     void cancel()
     {
         std::lock_guard guard(fiber_lock);
+        /// It is safe to just destroy fiber - we are not in the process of reading from socket.
         boost::context::fiber to_destroy = std::move(fiber);
 
+        /// Send something to pipe to cancel executor waiting.
         uint64_t buf = 0;
         while (-1 == write(pipe_fd[1], &buf, sizeof(buf)))
         {

From b184141e20bd825469d4bcd715a33445bccfd31b Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 12:22:49 +0300
Subject: [PATCH 260/504] Collect client coverage

---
 docker/test/coverage/run.sh | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/docker/test/coverage/run.sh b/docker/test/coverage/run.sh
index 5626fbbd5d4..e360fcd5aea 100755
--- a/docker/test/coverage/run.sh
+++ b/docker/test/coverage/run.sh
@@ -67,20 +67,20 @@ fi
 chmod 777 -R /var/lib/clickhouse
 
 
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "SHOW DATABASES"
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "ATTACH DATABASE datasets ENGINE = Ordinary"
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "CREATE DATABASE test"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "SHOW DATABASES"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "ATTACH DATABASE datasets ENGINE = Ordinary"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "CREATE DATABASE test"
 
 kill_clickhouse
 start_clickhouse
 
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "SHOW TABLES FROM datasets"
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "SHOW TABLES FROM test"
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "RENAME TABLE datasets.hits_v1 TO test.hits"
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "RENAME TABLE datasets.visits_v1 TO test.visits"
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-client --query "SHOW TABLES FROM test"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "SHOW TABLES FROM datasets"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "SHOW TABLES FROM test"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "RENAME TABLE datasets.hits_v1 TO test.hits"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "RENAME TABLE datasets.visits_v1 TO test.visits"
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "SHOW TABLES FROM test"
 
-LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
 
 readarray -t FAILED_TESTS < <(awk '/FAIL|TIMEOUT|ERROR/ { print substr($3, 1, length($3)-1) }' "/test_result.txt")
 
@@ -97,7 +97,7 @@ then
 
     echo "Going to run again: ${FAILED_TESTS[*]}"
 
-    LLVM_PROFILE_FILE='client_coverage.profraw' clickhouse-test --order=random --testname --shard --zookeeper --use-skip-list "${FAILED_TESTS[@]}" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee -a "/test_result.txt"
+    LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test --order=random --testname --shard --zookeeper --use-skip-list "${FAILED_TESTS[@]}" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee -a "/test_result.txt"
 else
     echo "No failed tests"
 fi

From c974fb8948ca6ab9719a850f3b072498cd4ead4b Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Tue, 22 Dec 2020 18:42:18 +0800
Subject: [PATCH 261/504] Remove obtrusive tests

---
 .../test.py                                   | 28 -------------------
 .../test_polymorphic_parts/test.py            | 11 --------
 2 files changed, 39 deletions(-)

diff --git a/tests/integration/test_adaptive_granularity_different_settings/test.py b/tests/integration/test_adaptive_granularity_different_settings/test.py
index 55b7e1c91b8..ce8f32b6ec6 100644
--- a/tests/integration/test_adaptive_granularity_different_settings/test.py
+++ b/tests/integration/test_adaptive_granularity_different_settings/test.py
@@ -51,31 +51,3 @@ def test_attach_detach(start_cluster):
 
     assert node1.query("SELECT COUNT() FROM  test") == "4\n"
     assert node2.query("SELECT COUNT() FROM  test") == "4\n"
-
-
-def test_mutate_with_mixed_granularity(start_cluster):
-    node3.query("""
-        CREATE TABLE test (date Date, key UInt64, value1 String, value2 String)
-        ENGINE = MergeTree
-        ORDER BY key PARTITION BY date""")
-
-    node3.query(
-        "INSERT INTO test SELECT toDate('2019-10-01') + number % 5, number, toString(number), toString(number * number) FROM numbers(500)")
-
-    assert node3.query("SELECT COUNT() FROM test") == "500\n"
-
-    node3.restart_with_latest_version()
-
-    assert node3.query("SELECT COUNT() FROM test") == "500\n"
-
-    node3.query("ALTER TABLE test MODIFY SETTING enable_mixed_granularity_parts = 1")
-
-    node3.query(
-        "INSERT INTO test SELECT toDate('2019-10-01') + number % 5, number, toString(number), toString(number * number) FROM numbers(500, 500)")
-
-    assert node3.query("SELECT COUNT() FROM test") == "1000\n"
-    assert node3.query("SELECT COUNT() FROM test WHERE key % 100 == 0") == "10\n"
-
-    node3.query("ALTER TABLE test DELETE WHERE key % 100 == 0", settings={"mutations_sync": "2"})
-
-    assert node3.query("SELECT COUNT() FROM test WHERE key % 100 == 0") == "0\n"
diff --git a/tests/integration/test_polymorphic_parts/test.py b/tests/integration/test_polymorphic_parts/test.py
index 50a8192fbc5..80cb59c6789 100644
--- a/tests/integration/test_polymorphic_parts/test.py
+++ b/tests/integration/test_polymorphic_parts/test.py
@@ -432,17 +432,6 @@ def test_in_memory_wal(start_cluster):
     # Check file exists
     node11.exec_in_container(['bash', '-c', 'test -f {}'.format(broken_wal_file)])
 
-    # Data is lost without WAL
-    node11.query("ALTER TABLE wal_table MODIFY SETTING in_memory_parts_enable_wal = 0")
-    with PartitionManager() as pm:
-        pm.partition_instances(node11, node12)
-
-        insert_random_data('wal_table', node11, 50)
-        check(node11, 350, 7)
-
-        node11.restart_clickhouse(kill=True)
-        check(node11, 300, 6)
-
 
 def test_in_memory_wal_rotate(start_cluster):
     # Write every part to single wal

From 57a16420353be05b84742c15c7ceef3ed91d4247 Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Tue, 22 Dec 2020 14:30:29 +0300
Subject: [PATCH 262/504] fix race in aggregation with combinator distinct

---
 src/Common/ThreadPool.cpp                     |  7 +++++
 src/Common/ThreadPool.h                       |  1 +
 src/Interpreters/Aggregator.cpp               | 31 ++++++++++++-------
 src/Interpreters/Aggregator.h                 |  1 +
 .../01605_dictinct_two_level.reference        | 20 ++++++++++++
 .../0_stateless/01605_dictinct_two_level.sql  | 25 +++++++++++++++
 6 files changed, 74 insertions(+), 11 deletions(-)
 create mode 100644 tests/queries/0_stateless/01605_dictinct_two_level.reference
 create mode 100644 tests/queries/0_stateless/01605_dictinct_two_level.sql

diff --git a/src/Common/ThreadPool.cpp b/src/Common/ThreadPool.cpp
index 7b2c2108629..7fc0d65aa5b 100644
--- a/src/Common/ThreadPool.cpp
+++ b/src/Common/ThreadPool.cpp
@@ -55,6 +55,13 @@ void ThreadPoolImpl<Thread>::setMaxThreads(size_t value)
     max_threads = value;
 }
 
+template <typename Thread>
+size_t ThreadPoolImpl<Thread>::getMaxThreads() const
+{
+    std::lock_guard lock(mutex);
+    return max_threads;
+}
+
 template <typename Thread>
 void ThreadPoolImpl<Thread>::setMaxFreeThreads(size_t value)
 {
diff --git a/src/Common/ThreadPool.h b/src/Common/ThreadPool.h
index 8dd6cbbe02c..0ae023e4ebd 100644
--- a/src/Common/ThreadPool.h
+++ b/src/Common/ThreadPool.h
@@ -71,6 +71,7 @@ public:
     void setMaxThreads(size_t value);
     void setMaxFreeThreads(size_t value);
     void setQueueSize(size_t value);
+    size_t getMaxThreads() const;
 
 private:
     mutable std::mutex mutex;
diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp
index 87abca4d7cd..f6f1f7c8d53 100644
--- a/src/Interpreters/Aggregator.cpp
+++ b/src/Interpreters/Aggregator.cpp
@@ -913,15 +913,15 @@ template <typename Method>
 Block Aggregator::convertOneBucketToBlock(
     AggregatedDataVariants & data_variants,
     Method & method,
+    Arena * arena,
     bool final,
     size_t bucket) const
 {
     Block block = prepareBlockAndFill(data_variants, final, method.data.impls[bucket].size(),
-        [bucket, &method, this] (
+        [bucket, &method, arena, this] (
             MutableColumns & key_columns,
             AggregateColumnsData & aggregate_columns,
             MutableColumns & final_aggregate_columns,
-            Arena * arena,
             bool final_)
         {
             convertToBlockImpl(method, method.data.impls[bucket],
@@ -950,7 +950,7 @@ Block Aggregator::mergeAndConvertOneBucketToBlock(
         mergeBucketImpl<decltype(merged_data.NAME)::element_type>(variants, bucket, arena); \
         if (is_cancelled && is_cancelled->load(std::memory_order_seq_cst)) \
             return {}; \
-        block = convertOneBucketToBlock(merged_data, *merged_data.NAME, final, bucket); \
+        block = convertOneBucketToBlock(merged_data, *merged_data.NAME, arena, final, bucket); \
     }
 
     APPLY_FOR_VARIANTS_TWO_LEVEL(M)
@@ -982,7 +982,7 @@ void Aggregator::writeToTemporaryFileImpl(
 
     for (size_t bucket = 0; bucket < Method::Data::NUM_BUCKETS; ++bucket)
     {
-        Block block = convertOneBucketToBlock(data_variants, method, false, bucket);
+        Block block = convertOneBucketToBlock(data_variants, method, data_variants.aggregates_pool, false, bucket);
         out.write(block);
         update_max_sizes(block);
     }
@@ -1285,7 +1285,7 @@ Block Aggregator::prepareBlockAndFill(
         }
     }
 
-    filler(key_columns, aggregate_columns_data, final_aggregate_columns, data_variants.aggregates_pool, final);
+    filler(key_columns, aggregate_columns_data, final_aggregate_columns, final);
 
     Block res = header.cloneEmpty();
 
@@ -1352,7 +1352,6 @@ Block Aggregator::prepareBlockAndFillWithoutKey(AggregatedDataVariants & data_va
         MutableColumns & key_columns,
         AggregateColumnsData & aggregate_columns,
         MutableColumns & final_aggregate_columns,
-        Arena * arena,
         bool final_)
     {
         if (data_variants.type == AggregatedDataVariants::Type::without_key || params.overflow_row)
@@ -1367,7 +1366,8 @@ Block Aggregator::prepareBlockAndFillWithoutKey(AggregatedDataVariants & data_va
             }
             else
             {
-                insertAggregatesIntoColumns(data, final_aggregate_columns, arena);
+                /// Always single-thread. It's safe to pass current arena from 'aggregates_pool'.
+                insertAggregatesIntoColumns(data, final_aggregate_columns, data_variants.aggregates_pool);
             }
 
             if (params.overflow_row)
@@ -1395,13 +1395,12 @@ Block Aggregator::prepareBlockAndFillSingleLevel(AggregatedDataVariants & data_v
         MutableColumns & key_columns,
         AggregateColumnsData & aggregate_columns,
         MutableColumns & final_aggregate_columns,
-        Arena * arena,
         bool final_)
     {
     #define M(NAME) \
         else if (data_variants.type == AggregatedDataVariants::Type::NAME) \
             convertToBlockImpl(*data_variants.NAME, data_variants.NAME->data, \
-                key_columns, aggregate_columns, final_aggregate_columns, arena, final_);
+                key_columns, aggregate_columns, final_aggregate_columns, data_variants.aggregates_pool, final_);
 
         if (false) {} // NOLINT
         APPLY_FOR_VARIANTS_SINGLE_LEVEL(M)
@@ -1435,11 +1434,21 @@ BlocksList Aggregator::prepareBlocksAndFillTwoLevelImpl(
     bool final,
     ThreadPool * thread_pool) const
 {
+    size_t max_threads = thread_pool->getMaxThreads();
+    if (max_threads > data_variants.aggregates_pools.size())
+        for (size_t i = data_variants.aggregates_pools.size(); i < max_threads; ++i)
+            data_variants.aggregates_pools.push_back(std::make_shared<Arena>());
+
     auto converter = [&](size_t bucket, ThreadGroupStatusPtr thread_group)
     {
         if (thread_group)
             CurrentThread::attachToIfDetached(thread_group);
-        return convertOneBucketToBlock(data_variants, method, final, bucket);
+
+        /// Select Arena to avoid race conditions
+        size_t thread_number = static_cast<size_t>(bucket) % max_threads;
+        Arena * arena = data_variants.aggregates_pools.at(thread_number).get();
+
+        return convertOneBucketToBlock(data_variants, method, arena, final, bucket);
     };
 
     /// packaged_task is used to ensure that exceptions are automatically thrown into the main stream.
@@ -1949,7 +1958,7 @@ private:
             else if (method == AggregatedDataVariants::Type::NAME) \
             { \
                 aggregator.mergeBucketImpl<decltype(merged_data.NAME)::element_type>(data, bucket_num, arena); \
-                block = aggregator.convertOneBucketToBlock(merged_data, *merged_data.NAME, final, bucket_num); \
+                block = aggregator.convertOneBucketToBlock(merged_data, *merged_data.NAME, arena, final, bucket_num); \
             }
 
             APPLY_FOR_VARIANTS_TWO_LEVEL(M)
diff --git a/src/Interpreters/Aggregator.h b/src/Interpreters/Aggregator.h
index c688da9d32d..86806b7fbad 100644
--- a/src/Interpreters/Aggregator.h
+++ b/src/Interpreters/Aggregator.h
@@ -1212,6 +1212,7 @@ protected:
     Block convertOneBucketToBlock(
         AggregatedDataVariants & data_variants,
         Method & method,
+        Arena * arena,
         bool final,
         size_t bucket) const;
 
diff --git a/tests/queries/0_stateless/01605_dictinct_two_level.reference b/tests/queries/0_stateless/01605_dictinct_two_level.reference
new file mode 100644
index 00000000000..50d1615e1aa
--- /dev/null
+++ b/tests/queries/0_stateless/01605_dictinct_two_level.reference
@@ -0,0 +1,20 @@
+['0']
+['1']
+['2']
+['3']
+['4']
+['5']
+['6']
+['7']
+['8']
+['9']
+test.com	['foo3223','foo6455','foo382','foo5566','foo1037']
+test.com0	['foo0']
+test.com0.0001	['foo1']
+test.com0.0002	['foo2']
+test.com0.0003	['foo3']
+test.com0.0004	['foo4']
+test.com0.0005	['foo5']
+test.com0.0006	['foo6']
+test.com0.0007	['foo7']
+test.com0.0008	['foo8']
diff --git a/tests/queries/0_stateless/01605_dictinct_two_level.sql b/tests/queries/0_stateless/01605_dictinct_two_level.sql
new file mode 100644
index 00000000000..5f20ae590c5
--- /dev/null
+++ b/tests/queries/0_stateless/01605_dictinct_two_level.sql
@@ -0,0 +1,25 @@
+SET group_by_two_level_threshold_bytes = 1;
+SET group_by_two_level_threshold = 1;
+
+SELECT groupArray(DISTINCT toString(number % 10)) FROM numbers_mt(50000) 
+    GROUP BY number ORDER BY number LIMIT 10
+    SETTINGS max_threads = 2, max_block_size = 2000;
+
+DROP TABLE IF EXISTS dictinct_two_level;
+
+CREATE TABLE dictinct_two_level (
+    time DateTime64(3),
+    domain String,
+    subdomain String
+) ENGINE = MergeTree ORDER BY time;
+
+INSERT INTO dictinct_two_level SELECT 1546300800000, 'test.com', concat('foo', toString(number % 10000)) from numbers(10000);
+INSERT INTO dictinct_two_level SELECT 1546300800000, concat('test.com', toString(number / 10000)) , concat('foo', toString(number % 10000)) from numbers(10000);
+
+SELECT
+    domain, groupArraySample(5, 11111)(DISTINCT subdomain) AS example_subdomains
+FROM dictinct_two_level
+GROUP BY domain ORDER BY domain, example_subdomains
+LIMIT 10;
+
+DROP TABLE IF EXISTS dictinct_two_level;

From 41229b8bc9be8b9d4ba659a090342f26c41a5918 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 14:52:48 +0300
Subject: [PATCH 263/504] Run small tests

---
 docker/test/coverage/run.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docker/test/coverage/run.sh b/docker/test/coverage/run.sh
index e360fcd5aea..cb972dc827d 100755
--- a/docker/test/coverage/run.sh
+++ b/docker/test/coverage/run.sh
@@ -80,7 +80,7 @@ LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "RENAM
 LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "RENAME TABLE datasets.visits_v1 TO test.visits"
 LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "SHOW TABLES FROM test"
 
-LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 00676_group_by_in 01533_optimize_skip_merged_partitions 01232_preparing_sets_race_condition 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
 
 readarray -t FAILED_TESTS < <(awk '/FAIL|TIMEOUT|ERROR/ { print substr($3, 1, length($3)-1) }' "/test_result.txt")
 

From 23156fe6e42fc42b849f30b3f268a7be7b4bdfe6 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 16:46:09 +0300
Subject: [PATCH 264/504] Fix accident stop merges for merge tree

---
 src/Storages/StorageMergeTree.cpp | 64 +++++++++++++++++++++++--------
 src/Storages/StorageMergeTree.h   |  4 ++
 2 files changed, 51 insertions(+), 17 deletions(-)

diff --git a/src/Storages/StorageMergeTree.cpp b/src/Storages/StorageMergeTree.cpp
index c81b3538042..8d1bee3e889 100644
--- a/src/Storages/StorageMergeTree.cpp
+++ b/src/Storages/StorageMergeTree.cpp
@@ -1124,41 +1124,71 @@ ActionLock StorageMergeTree::stopMergesAndWait()
 }
 
 
+MergeTreeDataPartPtr StorageMergeTree::outdatePart(const String & part_name, bool force)
+{
+
+    if (force)
+    {
+        /// Forcefully stop merges and make part outdated
+        auto merge_blocker = stopMergesAndWait();
+        auto part = getPartIfExists(part_name, {MergeTreeDataPartState::Committed});
+        if (!part)
+            throw Exception("Part " + part_name + " not found, won't try to drop it.", ErrorCodes::NO_SUCH_DATA_PART);
+        removePartsFromWorkingSet({part}, true);
+        return part;
+    }
+    else
+    {
+
+        /// Wait merges selector
+        std::unique_lock lock(currently_processing_in_background_mutex);
+
+        auto part = getPartIfExists(part_name, {MergeTreeDataPartState::Committed});
+        /// It's okay, part was already removed
+        if (!part)
+            return nullptr;
+
+        /// Part will be "removed" by merge or mutation, it's OK in case of some
+        /// background cleanup processes like removing of empty parts.
+        if (currently_merging_mutating_parts.count(part))
+            return nullptr;
+
+        removePartsFromWorkingSet({part}, true);
+        return part;
+    }
+}
+
 void StorageMergeTree::dropPartition(const ASTPtr & partition, bool detach, bool drop_part, const Context & context, bool throw_if_noop)
 {
     {
-        /// Asks to complete merges and does not allow them to start.
-        /// This protects against "revival" of data for a removed partition after completion of merge.
-        auto merge_blocker = stopMergesAndWait();
-
-        auto metadata_snapshot = getInMemoryMetadataPtr();
-
         MergeTreeData::DataPartsVector parts_to_remove;
+        auto metadata_snapshot = getInMemoryMetadataPtr();
 
         if (drop_part)
         {
-            String part_name = partition->as<ASTLiteral &>().value.safeGet<String>();
-            auto part = getPartIfExists(part_name, {MergeTreeDataPartState::Committed});
-
-            if (part)
-                parts_to_remove.push_back(part);
-            else if (throw_if_noop)
-                throw Exception("Part " + part_name + " not found, won't try to drop it.", ErrorCodes::NO_SUCH_DATA_PART);
-            else
+            auto part = outdatePart(partition->as<ASTLiteral &>().value.safeGet<String>(), throw_if_noop);
+            /// Nothing to do, part was removed in some different way
+            if (!part)
                 return;
+
+            parts_to_remove.push_back(part);
         }
         else
         {
+            /// Asks to complete merges and does not allow them to start.
+            /// This protects against "revival" of data for a removed partition after completion of merge.
+            auto merge_blocker = stopMergesAndWait();
             String partition_id = getPartitionIDFromQuery(partition, context);
             parts_to_remove = getDataPartsVectorInPartition(MergeTreeDataPartState::Committed, partition_id);
-        }
 
-        // TODO should we throw an exception if parts_to_remove is empty?
-        removePartsFromWorkingSet(parts_to_remove, true);
+            /// TODO should we throw an exception if parts_to_remove is empty?
+            removePartsFromWorkingSet(parts_to_remove, true);
+        }
 
         if (detach)
         {
             /// If DETACH clone parts to detached/ directory
+            /// NOTE: no race with background cleanup until we hold pointers to parts
             for (const auto & part : parts_to_remove)
             {
                 LOG_INFO(log, "Detaching {}", part->relative_path);
diff --git a/src/Storages/StorageMergeTree.h b/src/Storages/StorageMergeTree.h
index e6ebad19788..155967b8824 100644
--- a/src/Storages/StorageMergeTree.h
+++ b/src/Storages/StorageMergeTree.h
@@ -135,6 +135,10 @@ private:
       */
     bool merge(bool aggressive, const String & partition_id, bool final, bool deduplicate, const Names & deduplicate_by_columns, String * out_disable_reason = nullptr, bool optimize_skip_merged_partitions = false);
 
+    /// Make part state outdated and queue it to remove without timeout
+    /// If force, then stop merges and block them until part state became outdated. Throw exception if part doesn't exists
+    /// It not force, then take merges selector and check that part is not participating in background operations.
+    MergeTreeDataPartPtr outdatePart(const String & part_name, bool force);
     ActionLock stopMergesAndWait();
 
     /// Allocate block number for new mutation, write mutation to disk

From d1d139a168ddeb2dd4d9a2af24fdb33bf58e6bec Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 16:48:10 +0300
Subject: [PATCH 265/504] Fix comment

---
 src/Storages/StorageMergeTree.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/StorageMergeTree.h b/src/Storages/StorageMergeTree.h
index 155967b8824..3263f124afa 100644
--- a/src/Storages/StorageMergeTree.h
+++ b/src/Storages/StorageMergeTree.h
@@ -137,7 +137,7 @@ private:
 
     /// Make part state outdated and queue it to remove without timeout
     /// If force, then stop merges and block them until part state became outdated. Throw exception if part doesn't exists
-    /// It not force, then take merges selector and check that part is not participating in background operations.
+    /// If not force, then take merges selector and check that part is not participating in background operations.
     MergeTreeDataPartPtr outdatePart(const String & part_name, bool force);
     ActionLock stopMergesAndWait();
 

From 2c07516aeea0d5bdc92932685927232f8bd1c1b0 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 22 Dec 2020 18:17:23 +0300
Subject: [PATCH 266/504] Fix special build and style.

---
 src/Functions/FunctionsStringHash.cpp | 12 ++++++------
 src/Functions/bitHammingDistance.cpp  |  3 ++-
 src/Functions/ya.make                 |  3 +++
 3 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index acb2feb87b8..5fd46117ea3 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -26,7 +26,7 @@ namespace ErrorCodes
 
 struct Hash
 {
-    static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points)
+    static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points [[maybe_unused]])
     {
 #ifdef __SSE4_2__
         return _mm_crc32_u64(-1ULL, unalignedLoad<UInt32>(code_points));
@@ -35,7 +35,7 @@ struct Hash
 #endif
     }
 
-    static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points)
+    static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points [[maybe_unused]])
     {
         UInt64 crc = -1ULL;
 #ifdef __SSE4_2__
@@ -48,7 +48,7 @@ struct Hash
         return crc;
     }
 
-    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
+    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes [[maybe_unused]], size_t size [[maybe_unused]], size_t offset [[maybe_unused]])
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
@@ -78,7 +78,7 @@ struct Hash
         return crc1 | (crc2 << 32u);
     }
 
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt8 * hashes, size_t K)
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt8 * hashes [[maybe_unused]], size_t K [[maybe_unused]])
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
@@ -93,7 +93,7 @@ struct Hash
         return crc1 | (crc2 << 32u);
     }
 
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt32 * hashes, size_t K)
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt32 * hashes [[maybe_unused]], size_t K [[maybe_unused]])
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
@@ -108,7 +108,7 @@ struct Hash
         return crc1 | (crc2 << 32u);
     }
 
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt64 * hashes, size_t K)
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt64 * hashes [[maybe_unused]], size_t K [[maybe_unused]])
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
diff --git a/src/Functions/bitHammingDistance.cpp b/src/Functions/bitHammingDistance.cpp
index 6e07ca0ad7c..9b9ff5b6c07 100644
--- a/src/Functions/bitHammingDistance.cpp
+++ b/src/Functions/bitHammingDistance.cpp
@@ -98,7 +98,8 @@ public:
         const auto * left_generic = arguments[0].type.get();
         const auto * right_generic = arguments[1].type.get();
         ColumnPtr result_column;
-        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right) {
+        bool valid = castBothTypes(left_generic, right_generic, [&](const auto & left, const auto & right)
+        {
             using LeftDataType = std::decay_t<decltype(left)>;
             using RightDataType = std::decay_t<decltype(right)>;
             using T0 = typename LeftDataType::FieldType;
diff --git a/src/Functions/ya.make b/src/Functions/ya.make
index 6e5d832db77..7e64deef64d 100644
--- a/src/Functions/ya.make
+++ b/src/Functions/ya.make
@@ -53,6 +53,7 @@ SRCS(
     FunctionsRandom.cpp
     FunctionsRound.cpp
     FunctionsStringArray.cpp
+    FunctionsStringHash.cpp
     FunctionsStringSimilarity.cpp
     GatherUtils/concat.cpp
     GatherUtils/createArraySink.cpp
@@ -185,6 +186,7 @@ SRCS(
     bitBoolMaskAnd.cpp
     bitBoolMaskOr.cpp
     bitCount.cpp
+    bitHammingDistance.cpp
     bitNot.cpp
     bitOr.cpp
     bitRotateLeft.cpp
@@ -504,6 +506,7 @@ SRCS(
     tryBase64Decode.cpp
     tuple.cpp
     tupleElement.cpp
+    tupleHammingDistance.cpp
     upper.cpp
     upperUTF8.cpp
     uptime.cpp

From 396ea14f67aabe128d75f592b3b27ed19d15672f Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 22 Dec 2020 18:29:32 +0300
Subject: [PATCH 267/504] Fix special build.

---
 src/Functions/FunctionsStringHash.cpp | 99 ++++++++++++++++-----------
 1 file changed, 58 insertions(+), 41 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 5fd46117ea3..d57be67ef7f 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -26,55 +26,78 @@ namespace ErrorCodes
 
 struct Hash
 {
-    static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points [[maybe_unused]])
+    static UInt64 crc32u64(UInt64 crc [[maybe_unused]], UInt64 val [[maybe_unused]])
     {
 #ifdef __SSE4_2__
-        return _mm_crc32_u64(-1ULL, unalignedLoad<UInt32>(code_points));
+        return _mm_crc32_u64(crc, val);
+#elif defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)
+        return __crc32cd(crc, val);
 #else
-        throw Exception("ngramASCIIHash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+        throw Exception("String hash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
 #endif
     }
 
-    static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points [[maybe_unused]])
+    static UInt64 crc32u32(UInt64 crc [[maybe_unused]], UInt32 val [[maybe_unused]])
+    {
+#ifdef __SSE4_2__
+        return _mm_crc32_u32(crc, val);
+#elif defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)
+        return __crc32cw(crc, val);
+#else
+        throw Exception("String hash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+    }
+
+    static UInt64 crc32u8(UInt64 crc [[maybe_unused]], UInt8 val [[maybe_unused]])
+    {
+#ifdef __SSE4_2__
+        return _mm_crc32_u8(crc, val);
+#elif defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)
+        return __crc32cb(crc, val);
+#else
+        throw Exception("String hash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+    }
+
+    static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points)
+    {
+        return crc32u64(-1ULL, unalignedLoad<UInt32>(code_points));
+    }
+
+    static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points)
     {
         UInt64 crc = -1ULL;
-#ifdef __SSE4_2__
-        crc = _mm_crc32_u64(crc, code_points[0]);
-        crc = _mm_crc32_u64(crc, code_points[1]);
-        crc = _mm_crc32_u64(crc, code_points[2]);
-#else
-        throw Exception("ngramUTF8Hash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
-#endif
+        crc = crc32u64(crc, code_points[0]);
+        crc = crc32u64(crc, code_points[1]);
+        crc = crc32u64(crc, code_points[2]);
         return crc;
     }
 
-    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes [[maybe_unused]], size_t size [[maybe_unused]], size_t offset [[maybe_unused]])
+    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
-#ifdef __SSE4_2__
+
         for (size_t i = offset; i < size; i += 2)
-            crc1 = _mm_crc32_u64(crc1, hashes[i]);
+            crc1 = crc32u64(crc1, hashes[i]);
         for (size_t i = offset + 1; i < size; i += 2)
-            crc2 = _mm_crc32_u64(crc2, hashes[i]);
+            crc2 = crc32u64(crc2, hashes[i]);
 
         if ((size - offset) & 1)
         {
             for (size_t i = 0; i < offset; i += 2)
-                crc2 = _mm_crc32_u64(crc2, hashes[i]);
+                crc2 = crc32u64(crc2, hashes[i]);
             for (size_t i = 1; i < offset; i += 2)
-                crc1 = _mm_crc32_u64(crc1, hashes[i]);
+                crc1 = crc32u64(crc1, hashes[i]);
         }
         else
         {
             for (size_t i = 0; i < offset; i += 2)
-                crc1 = _mm_crc32_u64(crc1, hashes[i]);
+                crc1 = crc32u64(crc1, hashes[i]);
             for (size_t i = 1; i < offset; i += 2)
-                crc2 = _mm_crc32_u64(crc2, hashes[i]);
+                crc2 = crc32u64(crc2, hashes[i]);
         }
-#else
-        throw Exception("wordShinglesHash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
-#endif
+
         return crc1 | (crc2 << 32u);
     }
 
@@ -82,14 +105,12 @@ struct Hash
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
-#ifdef __SSE4_2__
+
         for (size_t i = 0; i < K; i += 2)
-            crc1 = _mm_crc32_u8(crc1, hashes[i]);
+            crc1 = crc32u8(crc1, hashes[i]);
         for (size_t i = 1; i < K; i += 2)
-            crc2 = _mm_crc32_u8(crc2, hashes[i]);
-#else
-        throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
-#endif
+            crc2 = crc32u8(crc2, hashes[i]);
+
         return crc1 | (crc2 << 32u);
     }
 
@@ -97,29 +118,25 @@ struct Hash
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
-#ifdef __SSE4_2__
+
         for (size_t i = 0; i < K; i += 2)
-            crc1 = _mm_crc32_u32(crc1, hashes[i]);
+            crc1 = crc32u32(crc1, hashes[i]);
         for (size_t i = 1; i < K; i += 2)
-            crc2 = _mm_crc32_u32(crc2, hashes[i]);
-#else
-        throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
-#endif
+            crc2 = crc32u32(crc2, hashes[i]);
+
         return crc1 | (crc2 << 32u);
     }
 
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt64 * hashes [[maybe_unused]], size_t K [[maybe_unused]])
+    static ALWAYS_INLINE inline UInt64 hashSum(const UInt64 * hashes, size_t K)
     {
         UInt64 crc1 = -1ULL;
         UInt64 crc2 = -1ULL;
-#ifdef __SSE4_2__
+
         for (size_t i = 0; i < K; i += 2)
-            crc1 = _mm_crc32_u64(crc1, hashes[i]);
+            crc1 = crc32u64(crc1, hashes[i]);
         for (size_t i = 1; i < K; i += 2)
-            crc2 = _mm_crc32_u64(crc2, hashes[i]);
-#else
-        throw Exception("hashSum is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
-#endif
+            crc2 = crc32u64(crc2, hashes[i]);
+
         return crc1 | (crc2 << 32u);
     }
 };

From e28ab8d68394604e8f3be18e482e293827fd572a Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 18:37:13 +0300
Subject: [PATCH 268/504] Better coverage build

---
 src/CMakeLists.txt | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 50e6d037b54..ced68213b94 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -193,6 +193,8 @@ if ((NOT USE_LIBCXX) AND COMPILER_CLANG AND OS_LINUX)
     list (APPEND DBMS_COMMON_LIBRARIES atomic)
 endif()
 
+set_source_files_properties(${dbms_sources} PROPERTIES COMPILE_FLAGS ${WITH_COVERAGE})
+
 if (MAKE_STATIC_LIBRARIES OR NOT SPLIT_SHARED_LIBRARIES)
     add_library (dbms STATIC ${dbms_headers} ${dbms_sources})
     target_link_libraries (dbms PRIVATE clickhouse_parsers_new jemalloc libdivide ${DBMS_COMMON_LIBRARIES})

From a25e00a3d92cd45266316de809189be638487238 Mon Sep 17 00:00:00 2001
From: Ilya Yatsishin <2159081+qoega@users.noreply.github.com>
Date: Tue, 22 Dec 2020 18:37:34 +0300
Subject: [PATCH 269/504] Update
 docs/en/sql-reference/functions/other-functions.md

---
 docs/en/sql-reference/functions/other-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 4d69edb3ec1..43f66913780 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1679,7 +1679,7 @@ UNSUPPORTED_METHOD
 
 ## tcpPort {#tcpPort}
 
-Returns TCP port number listened by this server.
+Returns [native interface](../../interfaces/tcp.md) TCP port number listened by this server.
 
 **Syntax**
 

From 2f4968c8da61a8e2d73e5a6bc320761a565bd436 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Tue, 22 Dec 2020 19:07:04 +0300
Subject: [PATCH 270/504] Fix style.

---
 src/Functions/ExtractString.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index b659d072887..c4251f8c4a6 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -1,3 +1,4 @@
+#pragma once
 #include <Common/PODArray.h>
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/UTF8Helpers.h>

From 57bb75b75baf337738e1fce4383f1bdad4fa6bf9 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Tue, 22 Dec 2020 19:45:03 +0300
Subject: [PATCH 271/504] finished rearranging the code

---
 src/Functions/FunctionBinaryArithmetic.h | 155 +++++++++++++----------
 1 file changed, 85 insertions(+), 70 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index ec5c9e8bd67..90eb73979b8 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -314,7 +314,12 @@ public:
         if constexpr(op_case == OpCase::LeftConstant) static_assert(!IsDecimalNumber<A>);
         if constexpr(op_case == OpCase::RightConstant) static_assert(!IsDecimalNumber<B>);
 
-        const size_t size = a.size();
+        size_t size;
+
+        if constexpr(op_case == OpCase::LeftConstant)
+            size = b.size();
+        else
+            size = a.size();
 
         if constexpr (is_plus_minus_compare)
         {
@@ -390,11 +395,12 @@ private:
         Operation<NativeResultType, NativeResultType>>;
 
     template <OpCase op_case, OpCase target>
-    static auto unwrap(const auto& elem, int i)
+    static auto unwrap(const auto& elem, size_t i)
     {
         if constexpr(op_case == target)
             return undec(elem);
-        return undec(elem[i]);
+        else
+            return undec(elem[i]);
     }
 
     /// there's implicit type conversion here
@@ -1062,78 +1068,87 @@ public:
 
         if constexpr (std::is_same_v<ResultDataType, InvalidType>)
             return nullptr;
-
-        static_assert(!std::is_same_v<ResultDataType, InvalidType>);
-
-        using T0 = typename LeftDataType::FieldType;
-        using T1 = typename RightDataType::FieldType;
-        using ResultType = typename ResultDataType::FieldType;
-        using ColVecT0 = std::conditional_t<IsDecimalNumber<T0>, ColumnDecimal<T0>, ColumnVector<T0>>;
-        using ColVecT1 = std::conditional_t<IsDecimalNumber<T1>, ColumnDecimal<T1>, ColumnVector<T1>>;
-        using ColVecResult = std::conditional_t<IsDecimalNumber<ResultType>, ColumnDecimal<ResultType>, ColumnVector<ResultType>>;
-
-        const auto * const col_left_raw = arguments[0].column.get();
-        const auto * const col_right_raw = arguments[1].column.get();
-
-        const size_t col_left_size = col_left_raw->size();
-
-        const ColumnConst * const col_left_const = checkAndGetColumnConst<ColVecT0>(col_left_raw);
-        const ColumnConst * const col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw);
-
-        const ColVecT0 * const col_left = checkAndGetColumn<ColVecT0>(col_left_raw);
-        const ColVecT1 * const col_right = checkAndGetColumn<ColVecT1>(col_right_raw);
-
-        if constexpr (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>)
-            return executeNumericWithDecimal<LeftDataType, RightDataType, ResultDataType>(
-                left, right,
-                col_left_const, col_right_const,
-                col_left, col_right,
-                col_left_size);
-
-        using OpImpl = BinaryOperationImpl<T0, T1, Op<T0, T1>, ResultType>;
-
-        /// non-vector result
-        if (col_left_const && col_right_const)
+        else // we can't avoid the else because otherwise the compiler may assume the ResultDataType may be Invalid
+             // and that would produce the compile error.
         {
-            const auto res = OpImpl::process(
-                col_left_const->template getValue<T0>(),
-                col_right_const->template getValue<T1>());
+            using T0 = typename LeftDataType::FieldType;
+            using T1 = typename RightDataType::FieldType;
+            using ResultType = typename ResultDataType::FieldType;
+            using ColVecT0 = std::conditional_t<IsDecimalNumber<T0>, ColumnDecimal<T0>, ColumnVector<T0>>;
+            using ColVecT1 = std::conditional_t<IsDecimalNumber<T1>, ColumnDecimal<T1>, ColumnVector<T1>>;
+            using ColVecResult = std::conditional_t<IsDecimalNumber<ResultType>, ColumnDecimal<ResultType>, ColumnVector<ResultType>>;
 
-            return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
-        }
+            const auto * const col_left_raw = arguments[0].column.get();
+            const auto * const col_right_raw = arguments[1].column.get();
 
-        typename ColVecResult::MutablePtr col_res = ColVecResult::create();
-        auto & vec_res = col_res->getData();
-        vec_res.resize(col_left_size);
+            const size_t col_left_size = col_left_raw->size();
 
-        if (col_left && col_right)
-        {
-            OpImpl::template process<OpCase::Vector>(
-                col_left->getData().data(),
-                col_right->getData().data(),
-                vec_res.data(),
-                vec_res.size());
-        }
-        else if (col_left_const && col_right)
-        {
-            OpImpl::template process<OpCase::LeftConstant>(
-                col_left_const->template getValue<T0>(),
-                col_right->getData().data(),
-                vec_res.data(),
-                vec_res.size());
-        }
-        else if (col_left && col_right_const)
-        {
-            OpImpl::template process<OpCase::RightConstant>(
-                col_left->getData().data(),
-                col_right_const->template getValue<T1>(),
-                vec_res.data(),
-                vec_res.size());
-        }
-        else
-            return nullptr;
+            const ColumnConst * const col_left_const = checkAndGetColumnConst<ColVecT0>(col_left_raw);
+            const ColumnConst * const col_right_const = checkAndGetColumnConst<ColVecT1>(col_right_raw);
 
-        return col_res;
+            const ColVecT0 * const col_left = checkAndGetColumn<ColVecT0>(col_left_raw);
+            const ColVecT1 * const col_right = checkAndGetColumn<ColVecT1>(col_right_raw);
+
+            if constexpr (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>)
+                return executeNumericWithDecimal<LeftDataType, RightDataType, ResultDataType>(
+                    left, right,
+                    col_left_const, col_right_const,
+                    col_left, col_right,
+                    col_left_size);
+            else //can't avoid else and another indentation level, otherwise the compiler would try to instantiate
+                 // ColVecResult for Decimals which would lead to a compile error.
+            {
+                using OpImpl = BinaryOperationImpl<T0, T1, Op<T0, T1>, ResultType>;
+
+                /// non-vector result
+                if (col_left_const && col_right_const)
+                {
+                    const auto res = OpImpl::process(
+                        col_left_const->template getValue<T0>(),
+                        col_right_const->template getValue<T1>());
+
+                    return ResultDataType().createColumnConst(col_left_const->size(), toField(res));
+                }
+
+                typename ColVecResult::MutablePtr col_res = ColVecResult::create();
+
+                auto & vec_res = col_res->getData();
+                vec_res.resize(col_left_size);
+
+                if (col_left && col_right)
+                {
+                    OpImpl::template process<OpCase::Vector>(
+                        col_left->getData().data(),
+                        col_right->getData().data(),
+                        vec_res.data(),
+                        vec_res.size());
+                }
+                else if (col_left_const && col_right)
+                {
+                    const T0 value = col_left_const->template getValue<T0>();
+
+                    OpImpl::template process<OpCase::LeftConstant>(
+                        &value,
+                        col_right->getData().data(),
+                        vec_res.data(),
+                        vec_res.size());
+                }
+                else if (col_left && col_right_const)
+                {
+                    const T1 value = col_right_const->template getValue<T1>();
+
+                    OpImpl::template process<OpCase::RightConstant>(
+                        col_left->getData().data(),
+                        &value,
+                        vec_res.data(),
+                        vec_res.size());
+                }
+                else
+                    return nullptr;
+
+                return col_res;
+            }
+        }
     }
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override

From 674d0083d94a62ef168c31e48627ea24a92ef0f4 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 20:10:10 +0300
Subject: [PATCH 272/504] Remove unused code; fix clang tidy; fix old comments

---
 base/daemon/BaseDaemon.cpp | 12 +-----------
 base/daemon/BaseDaemon.h   | 26 +++++++-------------------
 2 files changed, 8 insertions(+), 30 deletions(-)

diff --git a/base/daemon/BaseDaemon.cpp b/base/daemon/BaseDaemon.cpp
index 19bbca508bc..84140471879 100644
--- a/base/daemon/BaseDaemon.cpp
+++ b/base/daemon/BaseDaemon.cpp
@@ -485,17 +485,6 @@ void BaseDaemon::kill()
         throw Poco::SystemException("cannot kill process");
 }
 
-void BaseDaemon::sleep(double seconds)
-{
-    wakeup_event.reset();
-    wakeup_event.tryWait(seconds * 1000);
-}
-
-void BaseDaemon::wakeup()
-{
-    wakeup_event.set();
-}
-
 std::string BaseDaemon::getDefaultCorePath() const
 {
     return "/opt/cores/";
@@ -871,6 +860,7 @@ void BaseDaemon::onInterruptSignals(int signal_id)
 
 void BaseDaemon::waitForTerminationRequest()
 {
+    /// NOTE: as we already process signals via pipe, we don't have to block them with sigprocmask in threads
     std::unique_lock<std::mutex> lock(signal_handler_mutex);
     signal_event.wait(lock, [this](){ return terminate_signals_counter > 0; });
 }
diff --git a/base/daemon/BaseDaemon.h b/base/daemon/BaseDaemon.h
index 0fb5bb47b2d..090d4997606 100644
--- a/base/daemon/BaseDaemon.h
+++ b/base/daemon/BaseDaemon.h
@@ -48,28 +48,26 @@ public:
     BaseDaemon();
     ~BaseDaemon() override;
 
-    /// Загружает конфигурацию и "строит" логгеры на запись в файлы
+    /// Load configuration, prepare loggers, etc.
     void initialize(Poco::Util::Application &) override;
 
-    /// Читает конфигурацию
     void reloadConfiguration();
 
-    /// Определяет параметр командной строки
+    /// Process command line parameters
     void defineOptions(Poco::Util::OptionSet & new_options) override;
 
-    /// Завершение демона ("мягкое")
-    void terminate();
+    /// Graceful shutdown
+    static void terminate();
 
-    /// Завершение демона ("жёсткое")
+    /// Forceful shutdown
     void kill();
 
-    /// Получен ли сигнал на завершение?
+    /// Cancellation request has been received.
     bool isCancelled() const
     {
         return is_cancelled;
     }
 
-    /// Получение ссылки на экземпляр демона
     static BaseDaemon & instance()
     {
         return dynamic_cast<BaseDaemon &>(Poco::Util::Application::instance());
@@ -78,12 +76,6 @@ public:
     /// return none if daemon doesn't exist, reference to the daemon otherwise
     static std::optional<std::reference_wrapper<BaseDaemon>> tryGetInstance() { return tryGetInstance<BaseDaemon>(); }
 
-    /// Спит заданное количество секунд или до события wakeup
-    void sleep(double seconds);
-
-    /// Разбудить
-    void wakeup();
-
     /// В Graphite компоненты пути(папки) разделяются точкой.
     /// У нас принят путь формата root_path.hostname_yandex_ru.key
     /// root_path по умолчанию one_min
@@ -141,7 +133,6 @@ protected:
     /// fork the main process and watch if it was killed
     void setupWatchdog();
 
-    /// реализация обработки сигналов завершения через pipe не требует блокировки сигнала с помощью sigprocmask во всех потоках
     void waitForTerminationRequest()
 #if defined(POCO_CLICKHOUSE_PATCH) || POCO_VERSION >= 0x02000000 // in old upstream poco not vitrual
     override
@@ -161,10 +152,7 @@ protected:
 
     bool log_to_console = false;
 
-    /// Событие, чтобы проснуться во время ожидания
-    Poco::Event wakeup_event;
-
-    /// Поток, в котором принимается сигнал HUP/USR1 для закрытия логов.
+    /// A thread that acts on HUP and USR1 signal (close logs).
     Poco::Thread signal_listener_thread;
     std::unique_ptr<Poco::Runnable> signal_listener;
 

From 062c226498cb8f2c9ce5ad0c09da0d6f80a33edf Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Tue, 22 Dec 2020 20:13:18 +0300
Subject: [PATCH 273/504] Fix Darwin build

---
 base/daemon/BaseDaemon.cpp | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/base/daemon/BaseDaemon.cpp b/base/daemon/BaseDaemon.cpp
index 84140471879..f25bcdb91e1 100644
--- a/base/daemon/BaseDaemon.cpp
+++ b/base/daemon/BaseDaemon.cpp
@@ -731,12 +731,18 @@ static void blockSignals(const std::vector<int> & signals)
 {
     sigset_t sig_set;
 
+#if defined(OS_DARWIN)
+    sigemptyset(&sig_set);
+    for (auto signal : signals)
+        sigaddset(&sig_set, signal);
+#else
     if (sigemptyset(&sig_set))
         throw Poco::Exception("Cannot block signal.");
 
     for (auto signal : signals)
         if (sigaddset(&sig_set, signal))
             throw Poco::Exception("Cannot block signal.");
+#endif
 
     if (pthread_sigmask(SIG_BLOCK, &sig_set, nullptr))
         throw Poco::Exception("Cannot block signal.");

From 3f33e6b3e02ffd02186a167d9b5e4b9a3a2fd0b2 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 20:46:31 +0300
Subject: [PATCH 274/504] some review fixes

---
 src/AggregateFunctions/AggregateFunctionAvg.h |  4 +--
 src/Interpreters/ActionsDAG.cpp               |  2 +-
 src/Interpreters/ExpressionAnalyzer.cpp       |  2 +-
 src/Interpreters/InterpreterSelectQuery.cpp   |  2 +-
 src/Processors/Transforms/WindowTransform.cpp | 34 ++++++++-----------
 .../01591_window_functions.reference          | 10 +++---
 6 files changed, 24 insertions(+), 30 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionAvg.h b/src/AggregateFunctions/AggregateFunctionAvg.h
index d07ff5db2f2..a46796b9c8d 100644
--- a/src/AggregateFunctions/AggregateFunctionAvg.h
+++ b/src/AggregateFunctions/AggregateFunctionAvg.h
@@ -127,10 +127,10 @@ public:
     void insertResultInto(AggregateDataPtr place, IColumn & to, Arena *) const override
     {
         if constexpr (IsDecimalNumber<Numerator> || IsDecimalNumber<Denominator>)
-            static_cast<ColumnVector<Float64> &>(to).getData().push_back(
+            assert_cast<ColumnVector<Float64> &>(to).getData().push_back(
                 this->data(place).divideIfAnyDecimal(num_scale, denom_scale));
         else
-            static_cast<ColumnVector<Float64> &>(to).getData().push_back(this->data(place).divide());
+            assert_cast<ColumnVector<Float64> &>(to).getData().push_back(this->data(place).divide());
     }
 private:
     UInt32 num_scale;
diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index 72f852c2322..2fc78261f17 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -624,7 +624,7 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
             {
                 auto & input = inputs[res_elem.name];
                 if (input.empty())
-                    throw Exception("Cannot find column '" + backQuote(res_elem.name) + "' in source stream",
+                    throw Exception("Cannot find column " + backQuote(res_elem.name) + " in source stream",
                                     ErrorCodes::THERE_IS_NO_COLUMN);
 
                 src_node = actions_dag->inputs[input.front()];
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 53b1dbfd9be..d267edf43a7 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -988,7 +988,7 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
         // Add empty INPUT with window function name.
         // It is an aggregate function, so it won't be added by getRootActions.
         ColumnWithTypeAndName col;
-        col.type = std::make_shared<DataTypeInt64>();
+        col.type = f.aggregate_function->getReturnType();
         col.column = col.type->createColumn();
         col.name = f.column_name;
 
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 13170726a11..5a3058ce790 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -952,7 +952,7 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
 
             if (expressions.need_aggregate)
             {
-               executeAggregation(query_plan, expressions.before_aggregation, aggregate_overflow_row, aggregate_final, query_info.input_order_info);
+                executeAggregation(query_plan, expressions.before_aggregation, aggregate_overflow_row, aggregate_final, query_info.input_order_info);
                 /// We need to reset input order info, so that executeOrder can't use  it
                 query_info.input_order_info.reset();
             }
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index 0b5dbc08654..6e8b0ea8e39 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -70,34 +70,28 @@ WindowTransform::~WindowTransform()
 void WindowTransform::transform(Chunk & chunk)
 {
     const size_t num_rows = chunk.getNumRows();
-    auto block = getInputPort().getHeader().cloneWithColumns(chunk.detachColumns());
+    auto columns = chunk.detachColumns();
 
     for (auto & ws : workspaces)
     {
         ws.argument_columns.clear();
         for (const auto column_index : ws.argument_column_indices)
         {
-            ws.argument_columns.push_back(block.getColumns()[column_index].get());
+            ws.argument_columns.push_back(columns[column_index].get());
         }
 
-        // Create the resulting column.
-        ColumnWithTypeAndName column_description;
-        column_description.name = ws.window_function.column_name;
-        column_description.type
-            = ws.window_function.aggregate_function->getReturnType();
-        ws.result_column = column_description.type->createColumn();
-        column_description.column.reset(ws.result_column.get());
-
-        block.insert(column_description);
+        ws.result_column = ws.window_function.aggregate_function->getReturnType()
+            ->createColumn();
     }
 
     // We loop for all window functions for each row. Switching the loops might
     // be more efficient, because we would run less code and access less data in
     // the inner loop. If you change this, don't forget to fix the calculation of
-    // partition boundaries. Probably it has to be precalculated and stored as a
-    // UInt8 array. An interesting optimization would be to pass it as an extra
-    // column from the previous sorting step -- that step might need to make
-    // similar comparison anyway, if it's sorting only by the PARTITION BY columns.
+    // partition boundaries. Probably it has to be precalculated and stored as
+    // an array of offsets. An interesting optimization would be to pass it as
+    // an extra column from the previous sorting step -- that step might need to
+    // make similar comparison anyway, if it's sorting only by the PARTITION BY
+    // columns.
     for (size_t row = 0; row < num_rows; row++)
     {
         // Check whether the new partition has started. We have to reset the
@@ -115,7 +109,7 @@ void WindowTransform::transform(Chunk & chunk)
             partition_start_columns.clear();
             for (const auto i : partition_by_indices)
             {
-                partition_start_columns.push_back(block.getColumns()[i]);
+                partition_start_columns.push_back(columns[i]);
             }
             partition_start_row = row;
         }
@@ -127,7 +121,7 @@ void WindowTransform::transform(Chunk & chunk)
             for (; first_inequal_column < partition_start_columns.size();
                   ++first_inequal_column)
             {
-                const auto * current_column = block.getColumns()[
+                const auto * current_column = columns[
                     partition_by_indices[first_inequal_column]].get();
 
                 if (current_column->compareAt(row, partition_start_row,
@@ -145,7 +139,7 @@ void WindowTransform::transform(Chunk & chunk)
                 partition_start_columns.clear();
                 for (const auto i : partition_by_indices)
                 {
-                    partition_start_columns.push_back(block.getColumns()[i]);
+                    partition_start_columns.push_back(columns[i]);
                 }
                 partition_start_row = row;
             }
@@ -181,10 +175,10 @@ void WindowTransform::transform(Chunk & chunk)
     // processors modify the block. Workspaces live longer than individual blocks.
     for (auto & ws : workspaces)
     {
-        ws.result_column.detach();
+        columns.push_back(std::move(ws.result_column));
     }
 
-    chunk.setColumns(block.getColumns(), num_rows);
+    chunk.setColumns(std::move(columns), num_rows);
 }
 
 }
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index bca5ade4672..6c78108d734 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -39,15 +39,15 @@ select number, avg(number) over (order by number) from numbers(10);
 -- no order by
 
 0	0
-1	0
+1	0.5
 2	1
-3	1
+3	1.5
 4	2
-5	2
+5	2.5
 6	3
-7	3
+7	3.5
 8	4
-9	4
+9	4.5
 select number, quantileExact(number) over (partition by intDiv(number, 3)) from numbers(10);
 
 -- can add an alias after window spec

From eecd51bb75fa603d802a258f93c1718935bd3975 Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Tue, 22 Dec 2020 21:29:47 +0300
Subject: [PATCH 275/504] fix for uninitialized thread pool

---
 src/Interpreters/Aggregator.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp
index f6f1f7c8d53..de2c86d7c12 100644
--- a/src/Interpreters/Aggregator.cpp
+++ b/src/Interpreters/Aggregator.cpp
@@ -1434,7 +1434,7 @@ BlocksList Aggregator::prepareBlocksAndFillTwoLevelImpl(
     bool final,
     ThreadPool * thread_pool) const
 {
-    size_t max_threads = thread_pool->getMaxThreads();
+    size_t max_threads = thread_pool ? thread_pool->getMaxThreads() : 1;
     if (max_threads > data_variants.aggregates_pools.size())
         for (size_t i = data_variants.aggregates_pools.size(); i < max_threads; ++i)
             data_variants.aggregates_pools.push_back(std::make_shared<Arena>());

From a6d420d809e4adecb1fc4c6a3b9bb5f8fdd0822e Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Tue, 22 Dec 2020 21:32:31 +0300
Subject: [PATCH 276/504] Update other-functions.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Перевел на русский язык.
---
 .../functions/other-functions.md              | 40 +++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index 91bf0f5b3a0..4df16739f60 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1589,4 +1589,44 @@ SELECT countDigits(toDecimal32(1, 9)), countDigits(toDecimal32(-1, 9)),
 10	10	19	19	39	39
 ```
 
+## tcpPort {#tcpPort}
+
+Вовращает номер TCP порта, который использует сервер для [нативного протокола](../../interfaces/tcp.md).
+
+**Синтаксис**
+
+``` sql
+tcpPort()
+```
+
+**Параметры**
+
+-   Нет.
+
+**Возвращаемое значение**
+
+-   Номер TCP порта.
+
+Тип: [UInt16](../../sql-reference/data-types/int-uint.md).
+
+**Example**
+
+Запрос:
+
+``` sql
+SELECT tcpPort();
+```
+
+Результат:
+
+``` text
+┌─tcpPort()─┐
+│      9000 │
+└───────────┘
+```
+
+**Смотрите также**
+
+-   [tcp_port](../../operations/server-configuration-parameters/settings.md#server_configuration_parameters-tcp_port)
+
 [Оригинальная статья](https://clickhouse.tech/docs/ru/query_language/functions/other_functions/) <!--hide-->

From 14c11ac46f1616aef766061ca84c741d3d8bae18 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Tue, 22 Dec 2020 21:37:42 +0300
Subject: [PATCH 277/504] Update other-functions.md
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Перевел example.
---
 docs/ru/sql-reference/functions/other-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index 4df16739f60..14e09ce7295 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1609,7 +1609,7 @@ tcpPort()
 
 Тип: [UInt16](../../sql-reference/data-types/int-uint.md).
 
-**Example**
+**Пример**
 
 Запрос:
 

From f2b25c708c8f003aa95b5496ef757d474f497b9f Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Tue, 22 Dec 2020 22:01:15 +0300
Subject: [PATCH 278/504] review fixes

---
 src/Interpreters/AggregateDescription.cpp   |  2 +-
 src/Interpreters/AggregateDescription.h     | 10 +--
 src/Interpreters/ExpressionAnalyzer.cpp     | 96 +++++++++++----------
 src/Interpreters/ExpressionAnalyzer.h       |  6 +-
 src/Interpreters/InterpreterSelectQuery.cpp |  2 +-
 5 files changed, 58 insertions(+), 58 deletions(-)

diff --git a/src/Interpreters/AggregateDescription.cpp b/src/Interpreters/AggregateDescription.cpp
index 985789ca676..2748a2abe9d 100644
--- a/src/Interpreters/AggregateDescription.cpp
+++ b/src/Interpreters/AggregateDescription.cpp
@@ -104,7 +104,7 @@ std::string WindowFunctionDescription::dump() const
 {
     WriteBufferFromOwnString ss;
 
-    ss << "window function '" << column_name << "' over '" << window_name <<"\n";
+    ss << "window function '" << column_name << "\n";
     ss << "function node " << function_node->dumpTree() << "\n";
     ss << "aggregate function '" << aggregate_function->getName() << "'\n";
     if (!function_parameters.empty())
diff --git a/src/Interpreters/AggregateDescription.h b/src/Interpreters/AggregateDescription.h
index 400b63bc4dc..f1fc232d04d 100644
--- a/src/Interpreters/AggregateDescription.h
+++ b/src/Interpreters/AggregateDescription.h
@@ -29,12 +29,6 @@ using AggregateDescriptions = std::vector<AggregateDescription>;
 
 struct WindowFunctionDescription
 {
-    // According to the standard, a window function can refer to a window declared
-    // elsewhere, using its name. We haven't implemented this yet: all windows
-    // are declared in OVER clause, but they still get an auto-generated name,
-    // and this name is used to find the window that corresponds to a function.
-    std::string window_name;
-
     std::string column_name;
     const ASTFunction * function_node;
     AggregateFunctionPtr aggregate_function;
@@ -62,9 +56,7 @@ struct WindowDescription
 
     // No frame info as of yet.
 
-
-    // Reverse map to function descriptions, for convenience of building the
-    // plan. Just copy them because it's more convenient.
+    // The window functions that are calculated for this window.
     std::vector<WindowFunctionDescription> window_functions;
 
     std::string dump() const;
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index d267edf43a7..31c12490408 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -535,7 +535,6 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
             window_description.order_by.end());
 
         WindowFunctionDescription window_function;
-        window_function.window_name = window_description.window_name;
         window_function.function_node = function_node;
         window_function.column_name
             = window_function.function_node->getColumnName();
@@ -579,28 +578,19 @@ bool ExpressionAnalyzer::makeWindowDescriptions(ActionsDAGPtr & actions)
                 window_function.argument_types,
                 window_function.function_parameters, properties);
 
-        window_function_descriptions.push_back(window_function);
+        auto [it, inserted] = window_descriptions.insert(
+            {window_description.window_name, window_description});
 
-        if (auto it = window_descriptions.find(window_description.window_name);
-            it != window_descriptions.end())
+        if (!inserted)
         {
             assert(it->second.full_sort_description
                 == window_description.full_sort_description);
         }
-        else
-        {
-            window_descriptions.insert({window_description.window_name,
-                window_description});
-        }
+
+        it->second.window_functions.push_back(window_function);
     }
 
-    // Populate the reverse map.
-    for (const auto & f : window_function_descriptions)
-    {
-        window_descriptions[f.window_name].window_functions.push_back(f);
-    }
-
-    return !window_function_descriptions.empty();
+    return !syntax->window_function_asts.empty();
 }
 
 
@@ -979,37 +969,53 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
 {
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
-    for (const auto & f : window_function_descriptions)
-    {
-        // Requiring a constant reference to a shared pointer to non-const AST
-        // doesn't really look sane, but the visitor does indeed require it.
-        getRootActionsNoMakeSet(f.function_node->clone(), true, step.actions());
-
-        // Add empty INPUT with window function name.
-        // It is an aggregate function, so it won't be added by getRootActions.
-        ColumnWithTypeAndName col;
-        col.type = f.aggregate_function->getReturnType();
-        col.column = col.type->createColumn();
-        col.name = f.column_name;
-
-        step.actions()->addInput(col);
-    }
-
-    // 2) mark the columns that are really required:
-    for (const auto & f : window_function_descriptions)
-    {
-        for (const auto & a : f.function_node->arguments->children)
-        {
-            // 2.1) function arguments,
-            step.required_output.push_back(a->getColumnName());
-        }
-        // 2.2) function result,
-        step.required_output.push_back(f.column_name);
-    }
-
-    // 2.3) PARTITION BY and ORDER BY columns.
+    // 1) Add actions for window functions and their arguments;
+    // 2) Mark the columns that are really required.
     for (const auto & [_, w] : window_descriptions)
     {
+        for (const auto & f : w.window_functions)
+        {
+            // 1.1) arguments of window functions;
+            // Requiring a constant reference to a shared pointer to non-const AST
+            // doesn't really look sane, but the visitor does indeed require it.
+            getRootActionsNoMakeSet(f.function_node->clone(),
+                true /* no_subqueries */, step.actions());
+
+            // 1.2) result of window function: an empty INPUT.
+            // It is an aggregate function, so it won't be added by getRootActions.
+            // This is something of a hack. Other options:
+            //  a] do it like aggregate function -- break the chain of actions
+            //     and manually add window functions to the starting list of
+            //     input columns. Logically this is similar to what we're doing
+            //     now, but would require to split the window function processing
+            //     into a full-fledged step after plain functions. This would be
+            //     somewhat cumbersome. With INPUT hack we can avoid a separate
+            //     step and pretend that window functions are almost "normal"
+            //     select functions. The limitation of both these ways is that
+            //     we can't reference window functions in other SELECT
+            //     expressions.
+            //  b] add a WINDOW action type, then sort, then split the chain on
+            //     each WINDOW action and insert the Window pipeline between the
+            //     Expression pipelines. This is a "proper" way that would allow
+            //     us to depend on window functions in other functions. But it's
+            //     complicated so I avoid doing it for now.
+            ColumnWithTypeAndName col;
+            col.type = f.aggregate_function->getReturnType();
+            col.column = col.type->createColumn();
+            col.name = f.column_name;
+
+            step.actions()->addInput(col);
+
+            for (const auto & a : f.function_node->arguments->children)
+            {
+                // 2.1) function arguments;
+                step.required_output.push_back(a->getColumnName());
+            }
+            // 2.2) function result;
+            step.required_output.push_back(f.column_name);
+        }
+
+        // 2.3) PARTITION BY and ORDER BY columns.
         for (const auto & c : w.full_sort_description)
         {
             step.required_output.push_back(c.column_name);
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index 0d550b10668..fb0cb4ea4c3 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -62,7 +62,6 @@ struct ExpressionAnalyzerData
 
     bool has_window = false;
     WindowDescriptions window_descriptions;
-    WindowFunctionDescriptions window_function_descriptions;
     NamesAndTypesList window_columns;
 
     bool has_global_subqueries = false;
@@ -76,7 +75,7 @@ struct ExpressionAnalyzerData
   *
   * NOTE: if `ast` is a SELECT query from a table, the structure of this table should not change during the lifetime of ExpressionAnalyzer.
   */
-class ExpressionAnalyzer : public ExpressionAnalyzerData, private boost::noncopyable
+class ExpressionAnalyzer : protected ExpressionAnalyzerData, private boost::noncopyable
 {
 private:
     /// Extracts settings to enlight which are used (and avoid copy of others).
@@ -121,6 +120,9 @@ public:
     /// Get intermediates for tests
     const ExpressionAnalyzerData & getAnalyzedData() const { return *this; }
 
+    /// A list of windows for window functions.
+    const WindowDescriptions & windowDescriptions() const { return window_descriptions; }
+
 protected:
     ExpressionAnalyzer(
         const ASTPtr & query_,
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 5a3058ce790..38cc19a00d6 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1754,7 +1754,7 @@ void InterpreterSelectQuery::executeExpression(QueryPlan & query_plan, const Act
 
 void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
 {
-    for (const auto & [_, w] : query_analyzer->window_descriptions)
+    for (const auto & [_, w] : query_analyzer->windowDescriptions())
     {
         const Settings & settings = context->getSettingsRef();
 

From bb50329aa3ac06d52273b1c8215a574d3ba3c70d Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Tue, 22 Dec 2020 22:04:46 +0300
Subject: [PATCH 279/504] fixed the test and the code

---
 src/Functions/FunctionBinaryArithmetic.h      | 57 ++++++++++++++-----
 .../01603_decimal_mult_float.reference        | 14 ++---
 .../01603_decimal_mult_float.reference.orig   | 20 -------
 .../0_stateless/01603_decimal_mult_float.sql  |  1 +
 4 files changed, 52 insertions(+), 40 deletions(-)
 delete mode 100644 tests/queries/0_stateless/01603_decimal_mult_float.reference.orig

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 90eb73979b8..1adddb11620 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -342,6 +342,32 @@ public:
                 return;
             }
         }
+        else if constexpr(is_multiply)
+        {
+            if (scale_a != 1)
+            {
+                scale_a = 1 / scale_a; // see scale_a initialization for detail
+
+                for (size_t i = 0; i < size; ++i)
+                    c[i] = applyScaled<true, false>(
+                        unwrap<op_case, OpCase::LeftConstant>(a, i),
+                        unwrap<op_case, OpCase::RightConstant>(b, i),
+                        scale_a);
+                return;
+            }
+            else if (scale_b != 1)
+            {
+                scale_b = 1 / scale_b;
+
+                for (size_t i = 0; i < size; ++i)
+                    c[i] = applyScaled<false, false>(
+                        unwrap<op_case, OpCase::LeftConstant>(a, i),
+                        unwrap<op_case, OpCase::RightConstant>(b, i),
+                        scale_b);
+                return;
+            }
+
+        }
         else if constexpr (is_division && is_decimal_b)
         {
             for (size_t i = 0; i < size; ++i)
@@ -417,13 +443,13 @@ private:
             return Op::template apply<NativeResultType>(a, b);
     }
 
-    template <bool scale_left>
+    template <bool scale_left, bool may_check_overflow = true>
     static NO_SANITIZE_UNDEFINED NativeResultType applyScaled(NativeResultType a, NativeResultType b, NativeResultType scale)
     {
-        static_assert(is_plus_minus_compare);
+        static_assert(is_plus_minus_compare || is_multiply);
         NativeResultType res;
 
-        if constexpr (check_overflow)
+        if constexpr (check_overflow && may_check_overflow)
         {
             bool overflow = false;
 
@@ -778,22 +804,26 @@ class FunctionBinaryArithmetic : public IFunction
 
         const ResultType scale_a = [&] {
             if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
-                return right.getScaleMultiplier();
+                return right.getScaleMultiplier(); // the division impl uses only the scale_a
             else if constexpr (result_is_decimal)
                 return type.scaleFactorFor(left, is_multiply);
-            else if constexpr(left_is_decimal)
-                return left.getScale();
+            else if constexpr(left_is_decimal) // needed for e.g. multiply(Decimal, Float) to scale the arguments
+                // scale multiplier = 10^scale, so we need to get this value divided by 10^0 -> 2nd arg is 0 as
+                // the convertTo basically divides the given value on the second arg.
+                // Note we are using the scale factor instead of 1 / scale factor, it will be correctly processed in
+                // the impl.
+                return DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
             else
-                return 1; //won't be used, just to silence the warning
+                return 1; // the default value which won't cause any re-scale
         }();
 
         const ResultType scale_b = [&] {
             if constexpr (result_is_decimal)
                 return type.scaleFactorFor(right, is_multiply || is_division);
             else if constexpr(right_is_decimal)
-                return right.getScale();
+                return DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
             else
-                return 1; //same
+                return 1;
         }();
 
         /// non-vector result
@@ -803,8 +833,9 @@ class FunctionBinaryArithmetic : public IFunction
             const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
             const ResultType res = check_decimal_overflow
-                ? OpImplCheck::template process<left_is_decimal, right_is_decimal>(const_a, const_b, scale_a, scale_b)
-                : OpImpl::template process<left_is_decimal, right_is_decimal>(const_a, const_b, scale_a, scale_b);
+                // the arguments are already scaled after conversion
+                ? OpImplCheck::template process<left_is_decimal, right_is_decimal>(const_a, const_b, 1, 1)
+                : OpImpl::template process<left_is_decimal, right_is_decimal>(const_a, const_b, 1, 1);
 
             if constexpr (result_is_decimal)
                 return ResultDataType(type.getPrecision(), type.getScale()).createColumnConst(
@@ -831,14 +862,14 @@ class FunctionBinaryArithmetic : public IFunction
             const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
 
             helperInvokeEither<OpCase::LeftConstant, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
-                const_a, col_right->getData(), vec_res, scale_a, scale_b);
+                const_a, col_right->getData(), vec_res, 1, scale_b);
         }
         else if (col_left && col_right_const)
         {
             const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
             helperInvokeEither<OpCase::RightConstant, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
-                col_left->getData(), const_b, vec_res, scale_a, scale_b);
+                col_left->getData(), const_b, vec_res, scale_a, 1);
         }
         else
             return nullptr;
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference b/tests/queries/0_stateless/01603_decimal_mult_float.reference
index 2c567fac2cd..ee1eeb525ba 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.reference
@@ -3,12 +3,12 @@
 0.00012000000000000002
 150.16500000000002
 7.775900000000001
-56.62268999999999
-598.8376688440276
-299.41883813298773
-0.7485470860550346
-2.2456413519648777
+56.62269
+598.8376688440277
+299.41883723437786
+0.7485470860550345
+2.245641373854596
 1.641386318314034
 1.641386318314034
-1.6413863192433382
-1.6413863192433382
+1.641386334333447
+1.641386334333447
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.reference.orig b/tests/queries/0_stateless/01603_decimal_mult_float.reference.orig
deleted file mode 100644
index 0be5ff86018..00000000000
--- a/tests/queries/0_stateless/01603_decimal_mult_float.reference.orig
+++ /dev/null
@@ -1,20 +0,0 @@
-2.4
-10.165
-0.00012000000000000002
-150.16500000000002
-7.775900000000001
-<<<<<<< HEAD
-56.62268999999999
-598.8376688440276
-299.41883813298773
-0.7485470860550346
-2.2456413519648777
-=======
-56622.689999999995
-5.988376688440276
-2.9941884322906844
->>>>>>> 17331d2732d3de57870db222a2dc80441aa6febd
-1.641386318314034
-1.641386318314034
-1.6413863192433382
-1.6413863192433382
diff --git a/tests/queries/0_stateless/01603_decimal_mult_float.sql b/tests/queries/0_stateless/01603_decimal_mult_float.sql
index 407fe3649ec..99e4b775da7 100644
--- a/tests/queries/0_stateless/01603_decimal_mult_float.sql
+++ b/tests/queries/0_stateless/01603_decimal_mult_float.sql
@@ -13,6 +13,7 @@ INSERT INTO test01603(f64) SELECT 1 / (number + 1) FROM system.numbers LIMIT 100
 
 SELECT sum(d * 1.1) FROM test01603;
 SELECT sum(8.01 * d) FROM test01603;
+
 SELECT sum(f64 * toDecimal64(80, 2)) FROM test01603;
 SELECT sum(toDecimal64(40, 2) * f32) FROM test01603;
 SELECT sum(f64 * toDecimal64(0.1, 2)) FROM test01603;

From 5fc2c2169a6399ea94d0e9eed0aeedc43716d877 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Tue, 22 Dec 2020 22:07:07 +0300
Subject: [PATCH 280/504] Drop libnsl dependency (because of gRPC -> c-ares)

c-ares searching for gethostbyname in the libnsl library, however in the
version that shipped with gRPC it doing it wrong [1], since it uses
CHECK_LIBRARY_EXISTS(), which will return TRUE even if the function exists in
another dependent library. The upstream already contains correct macro [2],
but it is not included in gRPC (even upstream gRPC, not the one that is
shipped with clickhousee).

  [1]: https://github.com/c-ares/c-ares/blob/e982924acee7f7313b4baa4ee5ec000c5e373c30/CMakeLists.txt#L125
  [2]: https://github.com/c-ares/c-ares/blob/44fbc813685a1fa8aa3f27fcd7544faf612d376a/CMakeLists.txt#L146

And because if you by some reason have libnsl [3] installed, clickhouse will
reject to start w/o it. While this is completelly different library.

  [3]: https://packages.debian.org/bullseye/libnsl2
---
 contrib/grpc-cmake/CMakeLists.txt | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/contrib/grpc-cmake/CMakeLists.txt b/contrib/grpc-cmake/CMakeLists.txt
index efb0f1c4f43..97ca3fab4db 100644
--- a/contrib/grpc-cmake/CMakeLists.txt
+++ b/contrib/grpc-cmake/CMakeLists.txt
@@ -54,6 +54,26 @@ else ()
   set(CARES_SHARED ON CACHE BOOL "" FORCE)
 endif ()
 
+# Disable looking for libnsl on a platforms that has gethostbyname in glibc
+#
+# c-ares searching for gethostbyname in the libnsl library, however in the
+# version that shipped with gRPC it doing it wrong [1], since it uses
+# CHECK_LIBRARY_EXISTS(), which will return TRUE even if the function exists in
+# another dependent library. The upstream already contains correct macro [2],
+# but it is not included in gRPC (even upstream gRPC, not the one that is
+# shipped with clickhousee).
+#
+#   [1]: https://github.com/c-ares/c-ares/blob/e982924acee7f7313b4baa4ee5ec000c5e373c30/CMakeLists.txt#L125
+#   [2]: https://github.com/c-ares/c-ares/blob/44fbc813685a1fa8aa3f27fcd7544faf612d376a/CMakeLists.txt#L146
+#
+# And because if you by some reason have libnsl [3] installed, clickhouse will
+# reject to start w/o it. While this is completelly different library.
+#
+#   [3]: https://packages.debian.org/bullseye/libnsl2
+if (NOT CMAKE_SYSTEM_NAME STREQUAL "SunOS")
+  set(HAVE_LIBNSL OFF CACHE BOOL "" FORCE)
+endif()
+
 # We don't want to build C# extensions.
 set(gRPC_BUILD_CSHARP_EXT OFF)
 

From 46a8027947493f84e908fef1c5f4a1da6214a2f3 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Tue, 22 Dec 2020 22:10:03 +0300
Subject: [PATCH 281/504] Translation into Russian language
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Выполнил перевод на русский язык.
---
 .../functions/string-search-functions.md      |  8 +--
 .../functions/string-search-functions.md      | 51 +++++++++++++++++++
 2 files changed, 55 insertions(+), 4 deletions(-)

diff --git a/docs/en/sql-reference/functions/string-search-functions.md b/docs/en/sql-reference/functions/string-search-functions.md
index 957866faea3..ef7e378ad11 100644
--- a/docs/en/sql-reference/functions/string-search-functions.md
+++ b/docs/en/sql-reference/functions/string-search-functions.md
@@ -618,15 +618,15 @@ Type: [UInt64](../../sql-reference/data-types/int-uint.md).
 Query:
 
 ``` sql
-SELECT countMatches('foobar.com', 'o')
+SELECT countMatches('foobar.com', 'o+')
 ```
 
 Result:
 
 ``` text
-┌─countMatches('foobar.com', 'o')─┐
-│                               3 │
-└─────────────────────────────────┘
+┌─countMatches('foobar.com', 'o+')─┐
+│                                2 │
+└──────────────────────────────────┘
 ```
 
 Query:
diff --git a/docs/ru/sql-reference/functions/string-search-functions.md b/docs/ru/sql-reference/functions/string-search-functions.md
index d2f1119783b..8bc2bb12382 100644
--- a/docs/ru/sql-reference/functions/string-search-functions.md
+++ b/docs/ru/sql-reference/functions/string-search-functions.md
@@ -521,5 +521,56 @@ SELECT * FROM Months WHERE ilike(name, '%j%')
 
 !!! note "Примечание"
     Для случая UTF-8 мы используем триграммное расстояние. Вычисление n-граммного расстояния не совсем честное. Мы используем 2-х байтные хэши для хэширования n-грамм, а затем вычисляем (не)симметрическую разность между хэш таблицами – могут возникнуть коллизии. В формате UTF-8 без учета регистра мы не используем честную функцию `tolower` – мы обнуляем 5-й бит (нумерация с нуля) каждого байта кодовой точки, а также первый бит нулевого байта, если байтов больше 1 – это работает для латиницы и почти для всех кириллических букв.
+	
+## countMatches(haystack, pattern) {#countmatcheshaystack-pattern}
+
+Возвращает количество совпадений, найденных в строке `haystack`, для регулярного выражения `pattern`.
+
+**Синтаксис**
+
+``` sql
+countMatches(haystack, pattern)
+```
+
+**Параметры**
+
+-   `haystack` — строка, по которой выполняется поиск. [String](../../sql-reference/syntax.md#syntax-string-literal).
+-   `pattern` — регулярное выражение, построенное по синтаксическим правилам [re2](https://github.com/google/re2/wiki/Syntax). [String](../../sql-reference/data-types/string.md).
+
+**Возвращаемое значение**
+
+-   Количество совпадений.
+
+Тип: [UInt64](../../sql-reference/data-types/int-uint.md).
+
+**Примеры**
+
+Запрос:
+
+``` sql
+SELECT countMatches('foobar.com', 'o+')
+```
+
+Результат:
+
+``` text
+┌─countMatches('foobar.com', 'o+')─┐
+│                                2 │
+└──────────────────────────────────┘
+```
+
+Запрос:
+
+``` sql
+SELECT countMatches('aaaa', 'aa')
+```
+
+Результат:
+
+``` text
+┌─countMatches('aaaa', 'aa')─┐
+│                             2 │
+└───────────────────────────────┘
+```
 
 [Оригинальная статья](https://clickhouse.tech/docs/ru/query_language/functions/string_search_functions/) <!--hide-->

From e49021630e6b948769f9e3850c566659af29e8d1 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Tue, 22 Dec 2020 22:21:19 +0300
Subject: [PATCH 282/504] Do not override RULE_LAUNCH_COMPILE/RULE_LAUNCH_LINK
 in rocksdb

- it is already done in in find/ccache.cmake
- it does not respect disabling ccache
- it does not allow to override ccache using custom location (not in
  PATH)
---
 contrib/rocksdb-cmake/CMakeLists.txt | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/contrib/rocksdb-cmake/CMakeLists.txt b/contrib/rocksdb-cmake/CMakeLists.txt
index f99401ce75d..77a30776a4a 100644
--- a/contrib/rocksdb-cmake/CMakeLists.txt
+++ b/contrib/rocksdb-cmake/CMakeLists.txt
@@ -2,12 +2,6 @@
 set(ROCKSDB_SOURCE_DIR "${ClickHouse_SOURCE_DIR}/contrib/rocksdb")
 list(APPEND CMAKE_MODULE_PATH "${ROCKSDB_SOURCE_DIR}/cmake/modules/")
 
-find_program(CCACHE_FOUND ccache)
-if(CCACHE_FOUND)
-  set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
-  set_property(GLOBAL PROPERTY RULE_LAUNCH_LINK ccache)
-endif(CCACHE_FOUND)
-
 if (SANITIZE STREQUAL "undefined")
     set(WITH_UBSAN ON)
 elseif (SANITIZE STREQUAL "address")

From 8bdebe8ab647dace9d0a8f66021292e72548e64a Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 23:15:27 +0300
Subject: [PATCH 283/504] Disable coverage for all except clickhouse code

---
 CMakeLists.txt             | 6 +++---
 base/common/CMakeLists.txt | 2 ++
 src/CMakeLists.txt         | 2 --
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 52cb64fd763..33cf0bcb084 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -252,15 +252,15 @@ endif ()
 option(WITH_COVERAGE "Profile the resulting binary/binaries" OFF)
 
 if (WITH_COVERAGE AND COMPILER_CLANG)
-    set(COMPILER_FLAGS "${COMPILER_FLAGS} -fprofile-instr-generate -fcoverage-mapping")
+    SET(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
     # If we want to disable coverage for specific translation units
     set(WITHOUT_COVERAGE "-fno-profile-instr-generate -fno-coverage-mapping")
 endif()
 
 if (WITH_COVERAGE AND COMPILER_GCC)
-    set(COMPILER_FLAGS "${COMPILER_FLAGS} -fprofile-arcs -ftest-coverage")
     set(COVERAGE_OPTION "-lgcov")
-    set(WITHOUT_COVERAGE "-fno-profile-arcs -fno-test-coverage")
+    SET(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
+    set(WITH_COVERAGE "-fprofile-arcs -ftest-coverage")
 endif()
 
 set(COMPILER_FLAGS "${COMPILER_FLAGS}")
diff --git a/base/common/CMakeLists.txt b/base/common/CMakeLists.txt
index cea52b443dd..9a37f38ab14 100644
--- a/base/common/CMakeLists.txt
+++ b/base/common/CMakeLists.txt
@@ -37,6 +37,8 @@ add_library (common ${SRCS})
 
 if (WITH_COVERAGE)
     target_compile_definitions(common PUBLIC WITH_COVERAGE=1)
+    target_compile_options(common PUBLIC ${COMPILER_COVERAGE_OPTIONS})
+    target_link_options(common PUBLIC ${COMPILER_COVERAGE_OPTIONS})
 else ()
     target_compile_definitions(common PUBLIC WITH_COVERAGE=0)
 endif ()
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 86a7611642d..7596d85571c 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -193,8 +193,6 @@ if ((NOT USE_LIBCXX) AND COMPILER_CLANG AND OS_LINUX)
     list (APPEND DBMS_COMMON_LIBRARIES atomic)
 endif()
 
-set_source_files_properties(${dbms_sources} PROPERTIES COMPILE_FLAGS ${WITH_COVERAGE})
-
 if (MAKE_STATIC_LIBRARIES OR NOT SPLIT_SHARED_LIBRARIES)
     add_library (dbms STATIC ${dbms_headers} ${dbms_sources})
     target_link_libraries (dbms PRIVATE clickhouse_parsers_new jemalloc libdivide ${DBMS_COMMON_LIBRARIES})

From cc0c43ace53fe94c599a3ba9ec9417e184f7f15c Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Tue, 22 Dec 2020 23:16:59 +0300
Subject: [PATCH 284/504] Fix case

---
 CMakeLists.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 33cf0bcb084..4ff492bda8c 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -252,14 +252,14 @@ endif ()
 option(WITH_COVERAGE "Profile the resulting binary/binaries" OFF)
 
 if (WITH_COVERAGE AND COMPILER_CLANG)
-    SET(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
+    set(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
     # If we want to disable coverage for specific translation units
     set(WITHOUT_COVERAGE "-fno-profile-instr-generate -fno-coverage-mapping")
 endif()
 
 if (WITH_COVERAGE AND COMPILER_GCC)
     set(COVERAGE_OPTION "-lgcov")
-    SET(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
+    set(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
     set(WITH_COVERAGE "-fprofile-arcs -ftest-coverage")
 endif()
 

From 9e31800ebb0481270739595e302d8f298839fc95 Mon Sep 17 00:00:00 2001
From: Vitaly Baranov <vitbar@yandex-team.ru>
Date: Tue, 22 Dec 2020 23:16:25 +0300
Subject: [PATCH 285/504] Fix filling table system.settings_profile_elements.

---
 .../StorageSystemSettingsProfileElements.cpp  | 168 ++++++++----------
 ..._settings_profile_while_assigned.reference |   2 +
 ...5_drop_settings_profile_while_assigned.sql |   8 +
 3 files changed, 82 insertions(+), 96 deletions(-)
 create mode 100644 tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.reference
 create mode 100644 tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.sql

diff --git a/src/Storages/System/StorageSystemSettingsProfileElements.cpp b/src/Storages/System/StorageSystemSettingsProfileElements.cpp
index 6d6df4fe114..cf47416e188 100644
--- a/src/Storages/System/StorageSystemSettingsProfileElements.cpp
+++ b/src/Storages/System/StorageSystemSettingsProfileElements.cpp
@@ -68,121 +68,97 @@ void StorageSystemSettingsProfileElements::fillData(MutableColumns & res_columns
 
     auto add_rows_for_single_element = [&](const String & owner_name, EntityType owner_type, const SettingsProfileElement & element, size_t & index)
     {
-        switch (owner_type)
+        size_t old_num_rows = column_profile_name.size();
+        size_t new_num_rows = old_num_rows + 1;
+        size_t current_index = index++;
+
+        bool inserted_value = false;
+        if (!element.value.isNull() && !element.setting_name.empty())
         {
-            case EntityType::SETTINGS_PROFILE:
-            {
-                column_user_name.insertDefault();
-                column_user_name_null_map.push_back(true);
-                column_role_name.insertDefault();
-                column_role_name_null_map.push_back(true);
-                column_profile_name.insertData(owner_name.data(), owner_name.length());
-                column_profile_name_null_map.push_back(false);
-                break;
-            }
-            case EntityType::USER:
-            {
-                column_user_name.insertData(owner_name.data(), owner_name.length());
-                column_user_name_null_map.push_back(false);
-                column_profile_name.insertDefault();
-                column_profile_name_null_map.push_back(true);
-                column_role_name.insertDefault();
-                column_role_name_null_map.push_back(true);
-                break;
-            }
-            case EntityType::ROLE:
-            {
-                column_user_name.insertDefault();
-                column_user_name_null_map.push_back(true);
-                column_role_name.insertData(owner_name.data(), owner_name.length());
-                column_role_name_null_map.push_back(false);
-                column_profile_name.insertDefault();
-                column_profile_name_null_map.push_back(true);
-                break;
-            }
-            default:
-                assert(false);
+            String str = Settings::valueToStringUtil(element.setting_name, element.value);
+            column_value.insertData(str.data(), str.length());
+            column_value_null_map.push_back(false);
+            inserted_value = true;
         }
 
+        bool inserted_min = false;
+        if (!element.min_value.isNull() && !element.setting_name.empty())
+        {
+            String str = Settings::valueToStringUtil(element.setting_name, element.min_value);
+            column_min.insertData(str.data(), str.length());
+            column_min_null_map.push_back(false);
+            inserted_min = true;
+        }
+
+        bool inserted_max = false;
+        if (!element.max_value.isNull() && !element.setting_name.empty())
+        {
+            String str = Settings::valueToStringUtil(element.setting_name, element.max_value);
+            column_max.insertData(str.data(), str.length());
+            column_max_null_map.push_back(false);
+            inserted_max = true;
+        }
+
+        bool inserted_readonly = false;
+        if (element.readonly && !element.setting_name.empty())
+        {
+            column_readonly.push_back(*element.readonly);
+            column_readonly_null_map.push_back(false);
+            inserted_readonly = true;
+        }
+
+        bool inserted_setting_name = false;
+        if (inserted_value || inserted_min || inserted_max || inserted_readonly)
+        {
+            const auto & setting_name = element.setting_name;
+            column_setting_name.insertData(setting_name.data(), setting_name.size());
+            column_setting_name_null_map.push_back(false);
+            inserted_setting_name = true;
+        }
+
+        bool inserted_inherit_profile = false;
         if (element.parent_profile)
         {
             auto parent_profile = access_control.tryReadName(*element.parent_profile);
             if (parent_profile)
             {
-                column_index.push_back(index++);
-                column_setting_name.insertDefault();
-                column_setting_name_null_map.push_back(true);
-                column_value.insertDefault();
-                column_value_null_map.push_back(true);
-                column_min.insertDefault();
-                column_min_null_map.push_back(true);
-                column_max.insertDefault();
-                column_max_null_map.push_back(true);
-                column_readonly.push_back(0);
-                column_readonly_null_map.push_back(true);
                 const String & parent_profile_str = *parent_profile;
                 column_inherit_profile.insertData(parent_profile_str.data(), parent_profile_str.length());
                 column_inherit_profile_null_map.push_back(false);
+                inserted_inherit_profile = true;
             }
         }
 
-        if (!element.setting_name.empty()
-            && (!element.value.isNull() || !element.min_value.isNull() || !element.max_value.isNull() || element.readonly))
+        if (inserted_setting_name || inserted_inherit_profile)
         {
-            const auto & setting_name = element.setting_name;
-            column_index.push_back(index++);
-            column_setting_name.insertData(setting_name.data(), setting_name.size());
-            column_setting_name_null_map.push_back(false);
-
-            if (element.value.isNull())
+            switch (owner_type)
             {
-                column_value.insertDefault();
-                column_value_null_map.push_back(true);
-            }
-            else
-            {
-                String str = Settings::valueToStringUtil(setting_name, element.value);
-                column_value.insertData(str.data(), str.length());
-                column_value_null_map.push_back(false);
+                case EntityType::SETTINGS_PROFILE:
+                {
+                    column_profile_name.insertData(owner_name.data(), owner_name.length());
+                    column_profile_name_null_map.push_back(false);
+                    break;
+                }
+                case EntityType::USER:
+                {
+                    column_user_name.insertData(owner_name.data(), owner_name.length());
+                    column_user_name_null_map.push_back(false);
+                    break;
+                }
+                case EntityType::ROLE:
+                {
+                    column_role_name.insertData(owner_name.data(), owner_name.length());
+                    column_role_name_null_map.push_back(false);
+                    break;
+                }
+                default:
+                    assert(false);
             }
 
-            if (element.min_value.isNull())
-            {
-                column_min.insertDefault();
-                column_min_null_map.push_back(true);
-            }
-            else
-            {
-                String str = Settings::valueToStringUtil(setting_name, element.min_value);
-                column_min.insertData(str.data(), str.length());
-                column_min_null_map.push_back(false);
-            }
+            column_index.push_back(current_index);
 
-            if (element.max_value.isNull())
-            {
-                column_max.insertDefault();
-                column_max_null_map.push_back(true);
-            }
-            else
-            {
-                String str = Settings::valueToStringUtil(setting_name, element.max_value);
-                column_max.insertData(str.data(), str.length());
-                column_max_null_map.push_back(false);
-            }
-
-            if (element.readonly)
-            {
-                column_readonly.push_back(*element.readonly);
-                column_readonly_null_map.push_back(false);
-            }
-            else
-            {
-                column_readonly.push_back(0);
-                column_readonly_null_map.push_back(true);
-            }
-
-            column_inherit_profile.insertDefault();
-            column_inherit_profile_null_map.push_back(true);
+            for (auto & res_column : res_columns)
+                res_column->insertManyDefaults(new_num_rows - res_column->size());
         }
     };
 
diff --git a/tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.reference b/tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.reference
new file mode 100644
index 00000000000..47942812a11
--- /dev/null
+++ b/tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.reference
@@ -0,0 +1,2 @@
+\N	test_01605	\N	0	\N	\N	\N	\N	\N	test_01605
+PROFILE DROPPED
diff --git a/tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.sql b/tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.sql
new file mode 100644
index 00000000000..c9205d7fd89
--- /dev/null
+++ b/tests/queries/0_stateless/01605_drop_settings_profile_while_assigned.sql
@@ -0,0 +1,8 @@
+CREATE USER OR REPLACE 'test_01605';
+CREATE SETTINGS PROFILE OR REPLACE 'test_01605';
+ALTER USER 'test_01605' SETTINGS PROFILE 'test_01605';
+SELECT * FROM system.settings_profile_elements WHERE user_name='test_01605' OR profile_name='test_01605';
+DROP SETTINGS PROFILE 'test_01605';
+SELECT 'PROFILE DROPPED';
+SELECT * FROM system.settings_profile_elements WHERE user_name='test_01605' OR profile_name='test_01605';
+DROP USER 'test_01605';

From ba33b815b1a7720259709ac9d3e17cdd00f1e38a Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Tue, 22 Dec 2020 23:33:11 +0300
Subject: [PATCH 286/504] updated the tests

---
 tests/queries/0_stateless/00700_decimal_arithm.reference | 4 ++++
 tests/queries/0_stateless/00700_decimal_arithm.sql       | 8 ++++----
 2 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/tests/queries/0_stateless/00700_decimal_arithm.reference b/tests/queries/0_stateless/00700_decimal_arithm.reference
index d0040ff5ae2..74c83497df8 100644
--- a/tests/queries/0_stateless/00700_decimal_arithm.reference
+++ b/tests/queries/0_stateless/00700_decimal_arithm.reference
@@ -3,11 +3,13 @@
 84	0	1764	1	1	1
 84.840	0.000	1799.456400	1.000	1.000	1.000
 84.840000000	0.000000000
+84.840000000000000000	0.000000000000000000	0.000000000000000000000000000000000000	1.000000000000000000	1.000000000000000000	1.000000000000000000
 84.840000000000000000	0.000000000000000000
 84.84	0.00	1799.4564	1.00	1.00	1.00
 63	21	-42	882	-882	2	0	2	0
 63	21	-42	882	-882	2	0	2	0
 63	21	-42	882	-882	2	0	2	0
+0.420000000000000021	0.419999999999999979	0.419999999999999916	8.820000000000000000	-8.820000000000000000	0.020000000000000000	0.005000000000000000
 63.420	21.420	-41.580	890.820	-890.820	2.020	0.505	2.020	0.505
 63.420000000	21.420000000	-41.580000000	890.820000000	-890.820000000	2.020000000	0.505000000	2.020000000	0.505000000
 63.420000000000000000	21.420000000000000000	-41.580000000000000000	890.820000000000000000	-890.820000000000000000	2.020000000000000000	0.505000000000000000	2.020000000000000000	0.505000000000000000
@@ -15,7 +17,9 @@
 63	-21	42	882	-882	0	2	0	2
 63	-21	42	882	-882	0	2	0	2
 63	-21	42	882	-882	0	2	0	2
+0.420000000000000021	-0.419999999999999979	-0.419999999999999916	8.820000000000000000	-8.820000000000000000	0.000000000000000000	0.000000000000000000
 63.420	-21.420	41.580	890.820	-890.820	0.495	1.980	0.495	1.980
+42.420000021	-42.419999979	-42.419999916	890.820000000	-890.820000000	0.000000000	0.000000000
 63.420000000	-21.420000000	41.580000000	890.820000000	-890.820000000
 63.420000000000000000	-21.420000000000000000	41.580000000000000000	890.820000000000000000	-890.820000000000000000	0.495049504950495049	1.980198019801980198	0.495049504950495049	1.980198019801980198
 63.42	-21.42	41.58	890.82	-890.82	0.49	1.98	0.49	1.98
diff --git a/tests/queries/0_stateless/00700_decimal_arithm.sql b/tests/queries/0_stateless/00700_decimal_arithm.sql
index 3fa2aade743..627c1583eb4 100644
--- a/tests/queries/0_stateless/00700_decimal_arithm.sql
+++ b/tests/queries/0_stateless/00700_decimal_arithm.sql
@@ -31,14 +31,14 @@ SELECT f + f, f - f, f * f, f / f, intDiv(f, f), intDivOrZero(f, f) FROM decimal
 SELECT g + g, g - g, g * g, g / g, intDiv(g, g), intDivOrZero(g, g) FROM decimal WHERE g > 0;
 SELECT h + h, h - h, h * h, h / h, intDiv(h, h), intDivOrZero(h, h) FROM decimal WHERE h > 0; -- { serverError 407 }
 SELECT h + h, h - h FROM decimal WHERE h > 0;
-SELECT i + i, i - i, i * i, i / i, intDiv(i, i), intDivOrZero(i, i) FROM decimal WHERE i > 0; -- { serverError 407 }
+SELECT i + i, i - i, i * i, i / i, intDiv(i, i), intDivOrZero(i, i) FROM decimal WHERE i > 0;
 SELECT i + i, i - i FROM decimal WHERE i > 0;
 SELECT j + j, j - j, j * j, j / j, intDiv(j, j), intDivOrZero(j, j) FROM decimal WHERE j > 0;
 
 SELECT a + 21, a - 21, a - 84, a * 21, a * -21, a / 21, a / 84, intDiv(a, 21), intDivOrZero(a, 84) FROM decimal WHERE a = 42;
 SELECT b + 21, b - 21, b - 84, b * 21, b * -21, b / 21, b / 84, intDiv(b, 21), intDivOrZero(b, 84) FROM decimal WHERE b = 42;
 SELECT c + 21, c - 21, c - 84, c * 21, c * -21, c / 21, c / 84, intDiv(c, 21), intDivOrZero(c, 84) FROM decimal WHERE c = 42;
-SELECT e + 21, e - 21, e - 84, e * 21, e * -21, e / 21, e / 84 FROM decimal WHERE e > 0; -- { serverError 407 }
+SELECT e + 21, e - 21, e - 84, e * 21, e * -21, e / 21, e / 84 FROM decimal WHERE e > 0;
 SELECT f + 21, f - 21, f - 84, f * 21, f * -21, f / 21, f / 84 FROM decimal WHERE f > 0; -- { serverError 407 }
 SELECT g + 21, g - 21, g - 84, g * 21, g * -21, g / 21, g / 84, intDiv(g, 21), intDivOrZero(g, 84) FROM decimal WHERE g > 0;
 SELECT h + 21, h - 21, h - 84, h * 21, h * -21, h / 21, h / 84, intDiv(h, 21), intDivOrZero(h, 84) FROM decimal WHERE h > 0;
@@ -48,10 +48,10 @@ SELECT j + 21, j - 21, j - 84, j * 21, j * -21, j / 21, j / 84, intDiv(j, 21), i
 SELECT 21 + a, 21 - a, 84 - a, 21 * a, -21 * a, 21 / a, 84 / a, intDiv(21, a), intDivOrZero(84, a) FROM decimal WHERE a = 42;
 SELECT 21 + b, 21 - b, 84 - b, 21 * b, -21 * b, 21 / b, 84 / b, intDiv(21, b), intDivOrZero(84, b) FROM decimal WHERE b = 42;
 SELECT 21 + c, 21 - c, 84 - c, 21 * c, -21 * c, 21 / c, 84 / c, intDiv(21, c), intDivOrZero(84, c) FROM decimal WHERE c = 42;
-SELECT 21 + e, 21 - e, 84 - e, 21 * e, -21 * e, 21 / e, 84 / e FROM decimal WHERE e > 0; -- { serverError 407 }
+SELECT 21 + e, 21 - e, 84 - e, 21 * e, -21 * e, 21 / e, 84 / e FROM decimal WHERE e > 0;
 SELECT 21 + f, 21 - f, 84 - f, 21 * f, -21 * f, 21 / f, 84 / f FROM decimal WHERE f > 0; -- { serverError 407 }
 SELECT 21 + g, 21 - g, 84 - g, 21 * g, -21 * g, 21 / g, 84 / g, intDiv(21, g), intDivOrZero(84, g) FROM decimal WHERE g > 0;
-SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h, 21 / h, 84 / h FROM decimal WHERE h > 0; -- { serverError 407 }
+SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h, 21 / h, 84 / h FROM decimal WHERE h > 0;
 SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h FROM decimal WHERE h > 0;
 SELECT 21 + i, 21 - i, 84 - i, 21 * i, -21 * i, 21 / i, 84 / i, intDiv(21, i), intDivOrZero(84, i) FROM decimal WHERE i > 0;
 SELECT 21 + j, 21 - j, 84 - j, 21 * j, -21 * j, 21 / j, 84 / j, intDiv(21, j), intDivOrZero(84, j) FROM decimal WHERE j > 0;

From 27f647f93de5dc1400d748b1278066219f57e363 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Mon, 30 Nov 2020 19:42:41 +0300
Subject: [PATCH 287/504] done

---
 .../ParallelParsingBlockInputStream.cpp       | 26 +++++++++++++------
 .../ParallelParsingBlockInputStream.h         |  8 +++++-
 src/Formats/FormatFactory.cpp                 |  2 +-
 src/Formats/FormatFactory.h                   |  2 +-
 src/Formats/JSONEachRowUtils.cpp              |  8 ++++--
 src/Formats/JSONEachRowUtils.h                |  2 +-
 .../Formats/Impl/CSVRowInputFormat.cpp        |  9 +++++--
 .../Formats/Impl/RegexpRowInputFormat.cpp     |  7 ++---
 .../Impl/TabSeparatedRowInputFormat.cpp       |  8 ++++--
 ...el_parsing_exception_with_offset.reference |  1 +
 ..._parallel_parsing_exception_with_offset.sh | 14 ++++++++++
 11 files changed, 66 insertions(+), 21 deletions(-)
 create mode 100644 tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.reference
 create mode 100755 tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh

diff --git a/src/DataStreams/ParallelParsingBlockInputStream.cpp b/src/DataStreams/ParallelParsingBlockInputStream.cpp
index 19b04d36fc1..759a98de0a3 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.cpp
+++ b/src/DataStreams/ParallelParsingBlockInputStream.cpp
@@ -126,9 +126,13 @@ void ParallelParsingBlockInputStream::segmentatorThreadFunction(ThreadGroupStatu
             // Segmentating the original input.
             unit.segment.resize(0);
 
-            const bool have_more_data = file_segmentation_engine(original_buffer,
-                unit.segment, min_chunk_bytes);
+            auto [have_more_data, currently_read_rows] = file_segmentation_engine(
+                original_buffer, unit.segment, min_chunk_bytes);
 
+            unit.offset = successfully_read_rows_count;
+            successfully_read_rows_count += currently_read_rows;
+
+            
             unit.is_last = !have_more_data;
             unit.status = READY_TO_PARSE;
             scheduleParserThreadForUnitWithNumber(segmentator_ticket_number);
@@ -142,7 +146,7 @@ void ParallelParsingBlockInputStream::segmentatorThreadFunction(ThreadGroupStatu
     }
     catch (...)
     {
-        onBackgroundException();
+        onBackgroundException(successfully_read_rows_count);
     }
 }
 
@@ -157,11 +161,11 @@ void ParallelParsingBlockInputStream::parserThreadFunction(ThreadGroupStatusPtr
 
     setThreadName("ChunkParser");
 
+    const auto current_unit_number = current_ticket_number % processing_units.size();
+    auto & unit = processing_units[current_unit_number];
+
     try
     {
-        const auto current_unit_number = current_ticket_number % processing_units.size();
-        auto & unit = processing_units[current_unit_number];
-
         /*
          * This is kind of suspicious -- the input_process_creator contract with
          * respect to multithreaded use is not clear, but we hope that it is
@@ -195,11 +199,11 @@ void ParallelParsingBlockInputStream::parserThreadFunction(ThreadGroupStatusPtr
     }
     catch (...)
     {
-        onBackgroundException();
+        onBackgroundException(unit.offset);
     }
 }
 
-void ParallelParsingBlockInputStream::onBackgroundException()
+void ParallelParsingBlockInputStream::onBackgroundException(size_t offset)
 {
     tryLogCurrentException(__PRETTY_FUNCTION__);
 
@@ -207,6 +211,12 @@ void ParallelParsingBlockInputStream::onBackgroundException()
     if (!background_exception)
     {
         background_exception = std::current_exception();
+ 
+        if (Exception * e = exception_cast<Exception *>(background_exception))
+            e->addMessage(fmt::format(
+                "Offset: {}. P.S. You have to sum up offset and the row number from"
+                "the previous message to calculate the true row number where the"
+                "error occured due to parallel parsing algorithm specificity", offset));
     }
     finished = true;
     reader_condvar.notify_all();
diff --git a/src/DataStreams/ParallelParsingBlockInputStream.h b/src/DataStreams/ParallelParsingBlockInputStream.h
index c882acd9ddd..749de83b583 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.h
+++ b/src/DataStreams/ParallelParsingBlockInputStream.h
@@ -149,6 +149,8 @@ private:
         BlockExt block_ext;
         Memory<> segment;
         std::atomic<ProcessingUnitStatus> status;
+        /// Needed for better exception message.
+        size_t offset = 0;
         bool is_last{false};
     };
 
@@ -159,6 +161,10 @@ private:
     std::deque<ProcessingUnit> processing_units;
 
 
+    /// Compute it to have a more understandable error message.
+    size_t successfully_read_rows_count{0};
+
+
     void scheduleParserThreadForUnitWithNumber(size_t ticket_number);
     void finishAndWait();
 
@@ -169,7 +175,7 @@ private:
     // threads. This function is used by segmentator and parsed threads.
     // readImpl() is called from the main thread, so the exception handling
     // is different.
-    void onBackgroundException();
+    void onBackgroundException(size_t offset);
 };
 
 }
diff --git a/src/Formats/FormatFactory.cpp b/src/Formats/FormatFactory.cpp
index 926b790a4bd..877f62dfb34 100644
--- a/src/Formats/FormatFactory.cpp
+++ b/src/Formats/FormatFactory.cpp
@@ -163,7 +163,7 @@ BlockInputStreamPtr FormatFactory::getInput(
     // (segmentator + two parsers + reader).
     bool parallel_parsing = settings.input_format_parallel_parsing && file_segmentation_engine && settings.max_threads >= 4;
 
-    if (settings.min_chunk_bytes_for_parallel_parsing * settings.max_threads * 2 > settings.max_memory_usage)
+    if (settings.max_memory_usage && settings.min_chunk_bytes_for_parallel_parsing * settings.max_threads * 2 > settings.max_memory_usage)
         parallel_parsing = false;
 
     if (parallel_parsing && name == "JSONEachRow")
diff --git a/src/Formats/FormatFactory.h b/src/Formats/FormatFactory.h
index 0fe6f19f0b7..d78cc2e2740 100644
--- a/src/Formats/FormatFactory.h
+++ b/src/Formats/FormatFactory.h
@@ -54,7 +54,7 @@ public:
       * Reads at least min_chunk_bytes and some more until the end of the chunk, depends on the format.
       * Used in ParallelParsingBlockInputStream.
       */
-    using FileSegmentationEngine = std::function<bool(
+    using FileSegmentationEngine = std::function<std::pair<bool, size_t>(
         ReadBuffer & buf,
         DB::Memory<> & memory,
         size_t min_chunk_bytes)>;
diff --git a/src/Formats/JSONEachRowUtils.cpp b/src/Formats/JSONEachRowUtils.cpp
index a1d9b4a5fff..6017f3983c6 100644
--- a/src/Formats/JSONEachRowUtils.cpp
+++ b/src/Formats/JSONEachRowUtils.cpp
@@ -4,13 +4,14 @@
 namespace DB
 {
 
-bool fileSegmentationEngineJSONEachRowImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
+std::pair<bool, size_t> fileSegmentationEngineJSONEachRowImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
 {
     skipWhitespaceIfAny(in);
 
     char * pos = in.position();
     size_t balance = 0;
     bool quotes = false;
+    size_t number_of_rows = 0;
 
     while (loadAtPosition(in, memory, pos) && (balance || memory.size() + static_cast<size_t>(pos - in.position()) < min_chunk_size))
     {
@@ -57,11 +58,14 @@ bool fileSegmentationEngineJSONEachRowImpl(ReadBuffer & in, DB::Memory<> & memor
                 quotes = true;
                 ++pos;
             }
+
+            if (balance == 0)
+                ++number_of_rows;
         }
     }
 
     saveUpToPosition(in, memory, pos);
-    return loadAtPosition(in, memory, pos);
+    return {loadAtPosition(in, memory, pos), number_of_rows};
 }
 
 }
diff --git a/src/Formats/JSONEachRowUtils.h b/src/Formats/JSONEachRowUtils.h
index 92679fe3589..adf85f37a22 100644
--- a/src/Formats/JSONEachRowUtils.h
+++ b/src/Formats/JSONEachRowUtils.h
@@ -3,6 +3,6 @@
 namespace DB
 {
 
-bool fileSegmentationEngineJSONEachRowImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size);
+std::pair<bool, size_t> fileSegmentationEngineJSONEachRowImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size);
 
 }
diff --git a/src/Processors/Formats/Impl/CSVRowInputFormat.cpp b/src/Processors/Formats/Impl/CSVRowInputFormat.cpp
index 7bca5f2c5d9..89f0e0e5d2c 100644
--- a/src/Processors/Formats/Impl/CSVRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/CSVRowInputFormat.cpp
@@ -424,11 +424,12 @@ void registerInputFormatProcessorCSV(FormatFactory & factory)
     }
 }
 
-static bool fileSegmentationEngineCSVImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
+static std::pair<bool, size_t> fileSegmentationEngineCSVImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
 {
     char * pos = in.position();
     bool quotes = false;
     bool need_more_data = true;
+    size_t number_of_rows = 0;
 
     while (loadAtPosition(in, memory, pos) && need_more_data)
     {
@@ -458,6 +459,7 @@ static bool fileSegmentationEngineCSVImpl(ReadBuffer & in, DB::Memory<> & memory
             }
             else if (*pos == '\n')
             {
+                ++number_of_rows;
                 if (memory.size() + static_cast<size_t>(pos - in.position()) >= min_chunk_size)
                     need_more_data = false;
                 ++pos;
@@ -470,13 +472,16 @@ static bool fileSegmentationEngineCSVImpl(ReadBuffer & in, DB::Memory<> & memory
                     need_more_data = false;
                 ++pos;
                 if (loadAtPosition(in, memory, pos) && *pos == '\n')
+                {
                     ++pos;
+                    ++number_of_rows;
+                }  
             }
         }
     }
 
     saveUpToPosition(in, memory, pos);
-    return loadAtPosition(in, memory, pos);
+    return {loadAtPosition(in, memory, pos), number_of_rows};
 }
 
 void registerFileSegmentationEngineCSV(FormatFactory & factory)
diff --git a/src/Processors/Formats/Impl/RegexpRowInputFormat.cpp b/src/Processors/Formats/Impl/RegexpRowInputFormat.cpp
index c3f9d07b893..6e14a1dc3c8 100644
--- a/src/Processors/Formats/Impl/RegexpRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/RegexpRowInputFormat.cpp
@@ -173,10 +173,11 @@ void registerInputFormatProcessorRegexp(FormatFactory & factory)
     });
 }
 
-static bool fileSegmentationEngineRegexpImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
+static std::pair<bool, size_t> fileSegmentationEngineRegexpImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
 {
     char * pos = in.position();
     bool need_more_data = true;
+    size_t number_of_rows = 0;
 
     while (loadAtPosition(in, memory, pos) && need_more_data)
     {
@@ -196,12 +197,12 @@ static bool fileSegmentationEngineRegexpImpl(ReadBuffer & in, DB::Memory<> & mem
             need_more_data = false;
 
         ++pos;
-
+        ++number_of_rows;
     }
 
     saveUpToPosition(in, memory, pos);
 
-    return loadAtPosition(in, memory, pos);
+    return {loadAtPosition(in, memory, pos), number_of_rows};
 }
 
 void registerFileSegmentationEngineRegexp(FormatFactory & factory)
diff --git a/src/Processors/Formats/Impl/TabSeparatedRowInputFormat.cpp b/src/Processors/Formats/Impl/TabSeparatedRowInputFormat.cpp
index 529b70e4e09..69a5e61caf2 100644
--- a/src/Processors/Formats/Impl/TabSeparatedRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/TabSeparatedRowInputFormat.cpp
@@ -423,10 +423,11 @@ void registerInputFormatProcessorTabSeparated(FormatFactory & factory)
     }
 }
 
-static bool fileSegmentationEngineTabSeparatedImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
+static std::pair<bool, size_t> fileSegmentationEngineTabSeparatedImpl(ReadBuffer & in, DB::Memory<> & memory, size_t min_chunk_size)
 {
     bool need_more_data = true;
     char * pos = in.position();
+    size_t number_of_rows = 0;
 
     while (loadAtPosition(in, memory, pos) && need_more_data)
     {
@@ -443,6 +444,9 @@ static bool fileSegmentationEngineTabSeparatedImpl(ReadBuffer & in, DB::Memory<>
         }
         else if (*pos == '\n' || *pos == '\r')
         {
+            if (*pos == '\n')
+                ++number_of_rows;
+
             if (memory.size() + static_cast<size_t>(pos - in.position()) >= min_chunk_size)
                 need_more_data = false;
             ++pos;
@@ -451,7 +455,7 @@ static bool fileSegmentationEngineTabSeparatedImpl(ReadBuffer & in, DB::Memory<>
 
     saveUpToPosition(in, memory, pos);
 
-    return loadAtPosition(in, memory, pos);
+    return {loadAtPosition(in, memory, pos), number_of_rows};
 }
 
 void registerFileSegmentationEngineTabSeparated(FormatFactory & factory)
diff --git a/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.reference b/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.reference
new file mode 100644
index 00000000000..d86bac9de59
--- /dev/null
+++ b/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.reference
@@ -0,0 +1 @@
+OK
diff --git a/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh b/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh
new file mode 100755
index 00000000000..805f4267818
--- /dev/null
+++ b/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh
@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+. "$CURDIR"/../shell_config.sh
+
+CLICKHOUSE_CLIENT=$(echo ${CLICKHOUSE_CLIENT} | sed 's/'"--send_logs_level=${CLICKHOUSE_CLIENT_SERVER_LOGS_LEVEL}"'/--send_logs_level=none/g')
+
+$CLICKHOUSE_CLIENT --query="DROP TABLE IF EXISTS check;"
+
+$CLICKHOUSE_CLIENT --query="CREATE TABLE check (x UInt64) ENGINE = Memory;"
+
+(seq 1 2000000; echo 'hello'; seq 1 20000000) | $CLICKHOUSE_CLIENT --multiquery --query="SET input_format_parallel_parsing=1; SET min_chunk_bytes_for_parallel_parsing=100000; INSERT INTO check(x) FORMAT TSV " 2>&1 | grep -q "Offset: 1988984" && echo 'OK' || echo 'FAIL' ||:  
+
+$CLICKHOUSE_CLIENT --query="DROP TABLE check;"

From c60c161168771a8904c4205a1b0f16f85e60095c Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 10 Dec 2020 20:26:36 +0300
Subject: [PATCH 288/504] add ParsingException

---
 src/Common/CounterInFile.h                    |  2 +-
 src/Common/Exception.h                        | 32 +++++++++++++++++++
 src/DataStreams/NativeBlockInputStream.cpp    |  4 +--
 .../ParallelParsingBlockInputStream.cpp       | 13 ++++----
 src/DataTypes/DataTypeArray.cpp               |  4 +--
 src/DataTypes/DataTypeFixedString.cpp         |  2 +-
 src/DataTypes/DataTypeNullable.cpp            |  8 ++---
 src/IO/ReadBuffer.h                           |  2 +-
 src/IO/ReadHelpers.cpp                        | 12 +++----
 src/IO/ReadHelpers.h                          | 12 +++----
 src/IO/parseDateTimeBestEffort.cpp            |  2 +-
 src/IO/readDecimalText.h                      |  4 +--
 src/IO/readFloatText.h                        | 14 ++++----
 src/Processors/Formats/IRowInputFormat.cpp    | 20 ++++++++++++
 .../Formats/Impl/ArrowBlockInputFormat.cpp    |  4 +--
 .../Formats/Impl/AvroRowInputFormat.cpp       |  2 +-
 .../Impl/JSONCompactEachRowRowInputFormat.cpp |  2 +-
 .../Impl/JSONEachRowRowInputFormat.cpp        |  2 +-
 .../Formats/Impl/ORCBlockInputFormat.cpp      |  2 +-
 .../Formats/Impl/ParquetBlockInputFormat.cpp  |  2 +-
 .../Formats/Impl/TSKVRowInputFormat.cpp       |  4 +--
 .../Formats/Impl/TemplateRowInputFormat.cpp   |  2 +-
 ..._parallel_parsing_exception_with_offset.sh |  2 +-
 23 files changed, 102 insertions(+), 51 deletions(-)

diff --git a/src/Common/CounterInFile.h b/src/Common/CounterInFile.h
index 48414bd09cc..8cd4534d413 100644
--- a/src/Common/CounterInFile.h
+++ b/src/Common/CounterInFile.h
@@ -87,7 +87,7 @@ public:
                 {
                     /// A more understandable error message.
                     if (e.code() == DB::ErrorCodes::CANNOT_READ_ALL_DATA || e.code() == DB::ErrorCodes::ATTEMPT_TO_READ_AFTER_EOF)
-                        throw DB::Exception("File " + path + " is empty. You must fill it manually with appropriate value.", e.code());
+                        throw DB::ParsingException("File " + path + " is empty. You must fill it manually with appropriate value.", e.code());
                     else
                         throw;
                 }
diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index 3da2e2fb0d0..bff39fe454e 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -96,6 +96,38 @@ private:
 };
 
 
+/// Special class of exceptions, used mostly in ParallelParsingInputFormat for 
+/// more convinient calculation of problem line number.
+class ParsingException : public Exception 
+{
+public:
+    using Exception::Exception;
+
+    void formatInternalMessage() {
+        try
+        {
+            message(fmt::format(message(), line_number_));
+        }
+        catch (...) {}
+    }
+
+    int getLineNumber() {
+        return line_number_;
+    }
+
+    void setLineNumber(int line_number) {
+        line_number_ = line_number;
+    }
+
+private:
+    int line_number_;
+    mutable std::string formatted_message_;
+
+    const char * name() const throw() override { return "DB::ParsingException"; }
+    const char * className() const throw() override { return "DB::ParsingException"; }
+};
+
+
 using Exceptions = std::vector<std::exception_ptr>;
 
 
diff --git a/src/DataStreams/NativeBlockInputStream.cpp b/src/DataStreams/NativeBlockInputStream.cpp
index ae2d6886fa8..e9967e88638 100644
--- a/src/DataStreams/NativeBlockInputStream.cpp
+++ b/src/DataStreams/NativeBlockInputStream.cpp
@@ -83,7 +83,7 @@ void NativeBlockInputStream::readData(const IDataType & type, IColumn & column,
     type.deserializeBinaryBulkWithMultipleStreams(column, rows, settings, state);
 
     if (column.size() != rows)
-        throw Exception("Cannot read all data in NativeBlockInputStream. Rows read: " + toString(column.size()) + ". Rows expected: " + toString(rows) + ".",
+        throw ParsingException("Cannot read all data in NativeBlockInputStream. Rows read: " + toString(column.size()) + ". Rows expected: " + toString(rows) + ".",
             ErrorCodes::CANNOT_READ_ALL_DATA);
 }
 
@@ -106,7 +106,7 @@ Block NativeBlockInputStream::readImpl()
     if (istr.eof())
     {
         if (use_index)
-            throw Exception("Input doesn't contain all data for index.", ErrorCodes::CANNOT_READ_ALL_DATA);
+            throw ParsingException("Input doesn't contain all data for index.", ErrorCodes::CANNOT_READ_ALL_DATA);
 
         return res;
     }
diff --git a/src/DataStreams/ParallelParsingBlockInputStream.cpp b/src/DataStreams/ParallelParsingBlockInputStream.cpp
index 759a98de0a3..ec0652ae766 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.cpp
+++ b/src/DataStreams/ParallelParsingBlockInputStream.cpp
@@ -205,19 +205,18 @@ void ParallelParsingBlockInputStream::parserThreadFunction(ThreadGroupStatusPtr
 
 void ParallelParsingBlockInputStream::onBackgroundException(size_t offset)
 {
-    tryLogCurrentException(__PRETTY_FUNCTION__);
-
     std::unique_lock<std::mutex> lock(mutex);
     if (!background_exception)
     {
         background_exception = std::current_exception();
  
-        if (Exception * e = exception_cast<Exception *>(background_exception))
-            e->addMessage(fmt::format(
-                "Offset: {}. P.S. You have to sum up offset and the row number from"
-                "the previous message to calculate the true row number where the"
-                "error occured due to parallel parsing algorithm specificity", offset));
+        if (ParsingException * e = exception_cast<ParsingException *>(background_exception)) 
+        {
+            e->setLineNumber(e->getLineNumber() + offset);
+            e->formatInternalMessage();
+        }
     }
+    tryLogCurrentException(__PRETTY_FUNCTION__);
     finished = true;
     reader_condvar.notify_all();
     segmentator_condvar.notify_all();
diff --git a/src/DataTypes/DataTypeArray.cpp b/src/DataTypes/DataTypeArray.cpp
index 1a8130fb36d..9cd56d0e2b5 100644
--- a/src/DataTypes/DataTypeArray.cpp
+++ b/src/DataTypes/DataTypeArray.cpp
@@ -272,7 +272,7 @@ void DataTypeArray::deserializeBinaryBulkWithMultipleStreams(
     /// Check consistency between offsets and elements subcolumns.
     /// But if elements column is empty - it's ok for columns of Nested types that was added by ALTER.
     if (!nested_column.empty() && nested_column.size() != last_offset)
-        throw Exception("Cannot read all array values: read just " + toString(nested_column.size()) + " of " + toString(last_offset),
+        throw ParsingException("Cannot read all array values: read just " + toString(nested_column.size()) + " of " + toString(last_offset),
             ErrorCodes::CANNOT_READ_ALL_DATA);
 }
 
@@ -325,7 +325,7 @@ static void deserializeTextImpl(IColumn & column, ReadBuffer & istr, Reader && r
                 if (*istr.position() == ',')
                     ++istr.position();
                 else
-                    throw Exception(ErrorCodes::CANNOT_READ_ARRAY_FROM_TEXT,
+                    throw ParsingException(ErrorCodes::CANNOT_READ_ARRAY_FROM_TEXT,
                         "Cannot read array from text, expected comma or end of array, found '{}'",
                         *istr.position());
             }
diff --git a/src/DataTypes/DataTypeFixedString.cpp b/src/DataTypes/DataTypeFixedString.cpp
index 585c5709be7..c0b684cfcd9 100644
--- a/src/DataTypes/DataTypeFixedString.cpp
+++ b/src/DataTypes/DataTypeFixedString.cpp
@@ -103,7 +103,7 @@ void DataTypeFixedString::deserializeBinaryBulk(IColumn & column, ReadBuffer & i
     size_t read_bytes = istr.readBig(reinterpret_cast<char *>(&data[initial_size]), max_bytes);
 
     if (read_bytes % n != 0)
-        throw Exception("Cannot read all data of type FixedString. Bytes read:" + toString(read_bytes) + ". String size:" + toString(n) + ".",
+        throw ParsingException("Cannot read all data of type FixedString. Bytes read:" + toString(read_bytes) + ". String size:" + toString(n) + ".",
             ErrorCodes::CANNOT_READ_ALL_DATA);
 
     data.resize(initial_size + read_bytes);
diff --git a/src/DataTypes/DataTypeNullable.cpp b/src/DataTypes/DataTypeNullable.cpp
index ed501939901..a0fc8baaf7e 100644
--- a/src/DataTypes/DataTypeNullable.cpp
+++ b/src/DataTypes/DataTypeNullable.cpp
@@ -235,7 +235,7 @@ ReturnType DataTypeNullable::deserializeTextEscaped(IColumn & column, ReadBuffer
     /// Little tricky, because we cannot discriminate null from first character.
 
     if (istr.eof())
-        throw Exception("Unexpected end of stream, while parsing value of Nullable type", ErrorCodes::CANNOT_READ_ALL_DATA);
+        throw ParsingException("Unexpected end of stream, while parsing value of Nullable type", ErrorCodes::CANNOT_READ_ALL_DATA);
 
     /// This is not null, surely.
     if (*istr.position() != '\\')
@@ -250,7 +250,7 @@ ReturnType DataTypeNullable::deserializeTextEscaped(IColumn & column, ReadBuffer
         ++istr.position();
 
         if (istr.eof())
-            throw Exception("Unexpected end of stream, while parsing value of Nullable type, after backslash", ErrorCodes::CANNOT_READ_ALL_DATA);
+            throw ParsingException("Unexpected end of stream, while parsing value of Nullable type, after backslash", ErrorCodes::CANNOT_READ_ALL_DATA);
 
         return safeDeserialize<ReturnType>(column, *nested_data_type,
             [&istr]
@@ -405,11 +405,11 @@ ReturnType DataTypeNullable::deserializeTextCSV(IColumn & column, ReadBuffer & i
                 /// or if someone uses 'U' or 'L' as delimiter in CSV.
                 /// In the first case we cannot continue reading anyway. The second case seems to be unlikely.
                 if (settings.csv.delimiter == 'U' || settings.csv.delimiter == 'L')
-                    throw DB::Exception("Enabled setting input_format_csv_unquoted_null_literal_as_null may not work correctly "
+                    throw DB::ParsingException("Enabled setting input_format_csv_unquoted_null_literal_as_null may not work correctly "
                                         "with format_csv_delimiter = 'U' or 'L' for large input.", ErrorCodes::CANNOT_READ_ALL_DATA);
                 WriteBufferFromOwnString parsed_value;
                 nested_data_type->serializeAsTextCSV(nested, nested.size() - 1, parsed_value, settings);
-                throw DB::Exception("Error while parsing \"" + std::string(null_literal, null_prefix_len)
+                throw DB::ParsingException("Error while parsing \"" + std::string(null_literal, null_prefix_len)
                                     + std::string(istr.position(), std::min(size_t{10}, istr.available())) + "\" as Nullable(" + nested_data_type->getName()
                                     + ") at position " + std::to_string(istr.count()) + ": expected \"NULL\" or " + nested_data_type->getName()
                                     + ", got \"" + std::string(null_literal, buf.count()) + "\", which was deserialized as \""
diff --git a/src/IO/ReadBuffer.h b/src/IO/ReadBuffer.h
index 3d6eb6970ce..08f1bc2a3bb 100644
--- a/src/IO/ReadBuffer.h
+++ b/src/IO/ReadBuffer.h
@@ -165,7 +165,7 @@ public:
     {
         auto read_bytes = read(to, n);
         if (n != read_bytes)
-            throw Exception("Cannot read all data. Bytes read: " + std::to_string(read_bytes) + ". Bytes expected: " + std::to_string(n) + ".", ErrorCodes::CANNOT_READ_ALL_DATA);
+            throw ParsingException("Cannot read all data. Bytes read: " + std::to_string(read_bytes) + ". Bytes expected: " + std::to_string(n) + ".", ErrorCodes::CANNOT_READ_ALL_DATA);
     }
 
     /** A method that can be more efficiently implemented in derived classes, in the case of reading large enough blocks.
diff --git a/src/IO/ReadHelpers.cpp b/src/IO/ReadHelpers.cpp
index e290da39535..97a8d937d39 100644
--- a/src/IO/ReadHelpers.cpp
+++ b/src/IO/ReadHelpers.cpp
@@ -96,7 +96,7 @@ void NO_INLINE throwAtAssertionFailed(const char * s, ReadBuffer & buf)
     else
         out << " before: " << quote << String(buf.position(), std::min(SHOW_CHARS_ON_SYNTAX_ERROR, buf.buffer().end() - buf.position()));
 
-    throw Exception(out.str(), ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
+    throw ParsingException(out.str(), ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
 }
 
 
@@ -503,7 +503,7 @@ static void readAnyQuotedStringInto(Vector & s, ReadBuffer & buf)
 {
     if (buf.eof() || *buf.position() != quote)
     {
-        throw Exception(ErrorCodes::CANNOT_PARSE_QUOTED_STRING,
+        throw ParsingException(ErrorCodes::CANNOT_PARSE_QUOTED_STRING,
             "Cannot parse quoted string: expected opening quote '{}', got '{}'",
             std::string{quote}, buf.eof() ? "EOF" : std::string{*buf.position()});
     }
@@ -538,7 +538,7 @@ static void readAnyQuotedStringInto(Vector & s, ReadBuffer & buf)
             parseComplexEscapeSequence(s, buf);
     }
 
-    throw Exception("Cannot parse quoted string: expected closing quote",
+    throw ParsingException("Cannot parse quoted string: expected closing quote",
         ErrorCodes::CANNOT_PARSE_QUOTED_STRING);
 }
 
@@ -716,7 +716,7 @@ ReturnType readJSONStringInto(Vector & s, ReadBuffer & buf)
     auto error = [](const char * message [[maybe_unused]], int code [[maybe_unused]])
     {
         if constexpr (throw_exception)
-            throw Exception(message, code);
+            throw ParsingException(message, code);
         return ReturnType(false);
     };
 
@@ -861,7 +861,7 @@ ReturnType readDateTimeTextFallback(time_t & datetime, ReadBuffer & buf, const D
             s_pos[size] = 0;
 
             if constexpr (throw_exception)
-                throw Exception(std::string("Cannot parse datetime ") + s, ErrorCodes::CANNOT_PARSE_DATETIME);
+                throw ParsingException(std::string("Cannot parse datetime ") + s, ErrorCodes::CANNOT_PARSE_DATETIME);
             else
                 return false;
         }
@@ -899,7 +899,7 @@ ReturnType readDateTimeTextFallback(time_t & datetime, ReadBuffer & buf, const D
         else
         {
             if constexpr (throw_exception)
-                throw Exception("Cannot parse datetime", ErrorCodes::CANNOT_PARSE_DATETIME);
+                throw ParsingException("Cannot parse datetime", ErrorCodes::CANNOT_PARSE_DATETIME);
             else
                 return false;
         }
diff --git a/src/IO/ReadHelpers.h b/src/IO/ReadHelpers.h
index dbbaae0816f..fa6b1fc2d8a 100644
--- a/src/IO/ReadHelpers.h
+++ b/src/IO/ReadHelpers.h
@@ -300,7 +300,7 @@ ReturnType readIntTextImpl(T & x, ReadBuffer & buf)
                 else
                 {
                     if constexpr (throw_exception)
-                        throw Exception("Unsigned type must not contain '-' symbol", ErrorCodes::CANNOT_PARSE_NUMBER);
+                        throw ParsingException("Unsigned type must not contain '-' symbol", ErrorCodes::CANNOT_PARSE_NUMBER);
                     else
                         return ReturnType(false);
                 }
@@ -648,7 +648,7 @@ inline ReturnType readUUIDTextImpl(UUID & uuid, ReadBuffer & buf)
 
                 if constexpr (throw_exception)
                 {
-                    throw Exception(std::string("Cannot parse uuid ") + s, ErrorCodes::CANNOT_PARSE_UUID);
+                    throw ParsingException(std::string("Cannot parse uuid ") + s, ErrorCodes::CANNOT_PARSE_UUID);
                 }
                 else
                 {
@@ -669,7 +669,7 @@ inline ReturnType readUUIDTextImpl(UUID & uuid, ReadBuffer & buf)
 
         if constexpr (throw_exception)
         {
-            throw Exception(std::string("Cannot parse uuid ") + s, ErrorCodes::CANNOT_PARSE_UUID);
+            throw ParsingException(std::string("Cannot parse uuid ") + s, ErrorCodes::CANNOT_PARSE_UUID);
         }
         else
         {
@@ -824,7 +824,7 @@ inline void readDateTimeText(LocalDateTime & datetime, ReadBuffer & buf)
     if (19 != size)
     {
         s[size] = 0;
-        throw Exception(std::string("Cannot parse datetime ") + s, ErrorCodes::CANNOT_PARSE_DATETIME);
+        throw ParsingException(std::string("Cannot parse datetime ") + s, ErrorCodes::CANNOT_PARSE_DATETIME);
     }
 
     datetime.year((s[0] - '0') * 1000 + (s[1] - '0') * 100 + (s[2] - '0') * 10 + (s[3] - '0'));
@@ -1016,7 +1016,7 @@ void readQuoted(std::vector<T> & x, ReadBuffer & buf)
             if (*buf.position() == ',')
                 ++buf.position();
             else
-                throw Exception("Cannot read array from text", ErrorCodes::CANNOT_READ_ARRAY_FROM_TEXT);
+                throw ParsingException("Cannot read array from text", ErrorCodes::CANNOT_READ_ARRAY_FROM_TEXT);
         }
 
         first = false;
@@ -1039,7 +1039,7 @@ void readDoubleQuoted(std::vector<T> & x, ReadBuffer & buf)
             if (*buf.position() == ',')
                 ++buf.position();
             else
-                throw Exception("Cannot read array from text", ErrorCodes::CANNOT_READ_ARRAY_FROM_TEXT);
+                throw ParsingException("Cannot read array from text", ErrorCodes::CANNOT_READ_ARRAY_FROM_TEXT);
         }
 
         first = false;
diff --git a/src/IO/parseDateTimeBestEffort.cpp b/src/IO/parseDateTimeBestEffort.cpp
index 8a188d22236..063955cdd1e 100644
--- a/src/IO/parseDateTimeBestEffort.cpp
+++ b/src/IO/parseDateTimeBestEffort.cpp
@@ -99,7 +99,7 @@ ReturnType parseDateTimeBestEffortImpl(
     auto on_error = [](const std::string & message [[maybe_unused]], int code [[maybe_unused]])
     {
         if constexpr (std::is_same_v<ReturnType, void>)
-            throw Exception(message, code);
+            throw ParsingException(message, code);
         else
             return false;
     };
diff --git a/src/IO/readDecimalText.h b/src/IO/readDecimalText.h
index 727dd67c389..203d8e3963b 100644
--- a/src/IO/readDecimalText.h
+++ b/src/IO/readDecimalText.h
@@ -120,7 +120,7 @@ inline bool readDigits(ReadBuffer & buf, T & x, uint32_t & digits, int32_t & exp
                 if (!tryReadIntText(addition_exp, buf))
                 {
                     if constexpr (_throw_on_error)
-                        throw Exception("Cannot parse exponent while reading decimal", ErrorCodes::CANNOT_PARSE_NUMBER);
+                        throw ParsingException("Cannot parse exponent while reading decimal", ErrorCodes::CANNOT_PARSE_NUMBER);
                     else
                         return false;
                 }
@@ -133,7 +133,7 @@ inline bool readDigits(ReadBuffer & buf, T & x, uint32_t & digits, int32_t & exp
                 if (digits_only)
                 {
                     if constexpr (_throw_on_error)
-                        throw Exception("Unexpected symbol while reading decimal", ErrorCodes::CANNOT_PARSE_NUMBER);
+                        throw ParsingException("Unexpected symbol while reading decimal", ErrorCodes::CANNOT_PARSE_NUMBER);
                     return false;
                 }
                 stop = true;
diff --git a/src/IO/readFloatText.h b/src/IO/readFloatText.h
index 5647d574c62..eac6183e332 100644
--- a/src/IO/readFloatText.h
+++ b/src/IO/readFloatText.h
@@ -160,7 +160,7 @@ ReturnType readFloatTextPreciseImpl(T & x, ReadBuffer & buf)
         if (unlikely(res.ec != std::errc()))
         {
             if constexpr (throw_exception)
-                throw Exception("Cannot read floating point value", ErrorCodes::CANNOT_PARSE_NUMBER);
+                throw ParsingException("Cannot read floating point value", ErrorCodes::CANNOT_PARSE_NUMBER);
             else
                 return ReturnType(false);
         }
@@ -243,7 +243,7 @@ ReturnType readFloatTextPreciseImpl(T & x, ReadBuffer & buf)
         if (unlikely(res.ec != std::errc()))
         {
             if constexpr (throw_exception)
-                throw Exception("Cannot read floating point value", ErrorCodes::CANNOT_PARSE_NUMBER);
+                throw ParsingException("Cannot read floating point value", ErrorCodes::CANNOT_PARSE_NUMBER);
             else
                 return ReturnType(false);
         }
@@ -331,7 +331,7 @@ ReturnType readFloatTextFastImpl(T & x, ReadBuffer & in)
     if (in.eof())
     {
         if constexpr (throw_exception)
-            throw Exception("Cannot read floating point value", ErrorCodes::CANNOT_PARSE_NUMBER);
+            throw ParsingException("Cannot read floating point value", ErrorCodes::CANNOT_PARSE_NUMBER);
         else
             return false;
     }
@@ -387,7 +387,7 @@ ReturnType readFloatTextFastImpl(T & x, ReadBuffer & in)
         if (in.eof())
         {
             if constexpr (throw_exception)
-                throw Exception("Cannot read floating point value: nothing after exponent", ErrorCodes::CANNOT_PARSE_NUMBER);
+                throw ParsingException("Cannot read floating point value: nothing after exponent", ErrorCodes::CANNOT_PARSE_NUMBER);
             else
                 return false;
         }
@@ -425,7 +425,7 @@ ReturnType readFloatTextFastImpl(T & x, ReadBuffer & in)
         if (in.eof())
         {
             if constexpr (throw_exception)
-                throw Exception("Cannot read floating point value: no digits read", ErrorCodes::CANNOT_PARSE_NUMBER);
+                throw ParsingException("Cannot read floating point value: no digits read", ErrorCodes::CANNOT_PARSE_NUMBER);
             else
                 return false;
         }
@@ -436,14 +436,14 @@ ReturnType readFloatTextFastImpl(T & x, ReadBuffer & in)
             if (in.eof())
             {
                 if constexpr (throw_exception)
-                    throw Exception("Cannot read floating point value: nothing after plus sign", ErrorCodes::CANNOT_PARSE_NUMBER);
+                    throw ParsingException("Cannot read floating point value: nothing after plus sign", ErrorCodes::CANNOT_PARSE_NUMBER);
                 else
                     return false;
             }
             else if (negative)
             {
                 if constexpr (throw_exception)
-                    throw Exception("Cannot read floating point value: plus after minus sign", ErrorCodes::CANNOT_PARSE_NUMBER);
+                    throw ParsingException("Cannot read floating point value: plus after minus sign", ErrorCodes::CANNOT_PARSE_NUMBER);
                 else
                     return false;
             }
diff --git a/src/Processors/Formats/IRowInputFormat.cpp b/src/Processors/Formats/IRowInputFormat.cpp
index a6c66f1cacc..f65a590ab42 100644
--- a/src/Processors/Formats/IRowInputFormat.cpp
+++ b/src/Processors/Formats/IRowInputFormat.cpp
@@ -132,6 +132,26 @@ Chunk IRowInputFormat::generate()
             }
         }
     }
+    catch (ParsingException & e) 
+    {
+        String verbose_diagnostic;
+        try
+        {
+            verbose_diagnostic = getDiagnosticInfo();
+        }
+        catch (const Exception & exception)
+        {
+            verbose_diagnostic = "Cannot get verbose diagnostic: " + exception.message();
+        }
+        catch (...)
+        {
+            /// Error while trying to obtain verbose diagnostic. Ok to ignore.
+        }
+
+        e.setLineNumber(total_rows);
+        e.addMessage("(at row {})\n" + verbose_diagnostic);
+        throw;
+    }
     catch (Exception & e)
     {
         if (!isParseError(e.code()))
diff --git a/src/Processors/Formats/Impl/ArrowBlockInputFormat.cpp b/src/Processors/Formats/Impl/ArrowBlockInputFormat.cpp
index 64a9ce68ef7..4edef1f1365 100644
--- a/src/Processors/Formats/Impl/ArrowBlockInputFormat.cpp
+++ b/src/Processors/Formats/Impl/ArrowBlockInputFormat.cpp
@@ -48,12 +48,12 @@ Chunk ArrowBlockInputFormat::generate()
     }
 
     if (!batch_result.ok())
-        throw Exception(ErrorCodes::CANNOT_READ_ALL_DATA,
+        throw ParsingException(ErrorCodes::CANNOT_READ_ALL_DATA,
             "Error while reading batch of Arrow data: {}", batch_result.status().ToString());
 
     auto table_result = arrow::Table::FromRecordBatches({*batch_result});
     if (!table_result.ok())
-        throw Exception(ErrorCodes::CANNOT_READ_ALL_DATA,
+        throw ParsingException(ErrorCodes::CANNOT_READ_ALL_DATA,
             "Error while reading batch of Arrow data: {}", table_result.status().ToString());
 
     ++record_batch_current;
diff --git a/src/Processors/Formats/Impl/AvroRowInputFormat.cpp b/src/Processors/Formats/Impl/AvroRowInputFormat.cpp
index ccfe65bbba3..a8d71790f41 100644
--- a/src/Processors/Formats/Impl/AvroRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/AvroRowInputFormat.cpp
@@ -190,7 +190,7 @@ AvroDeserializer::DeserializeFn AvroDeserializer::createDeserializeFn(avro::Node
                 {
                     decoder.decodeString(tmp);
                     if (tmp.length() != 36)
-                        throw Exception(std::string("Cannot parse uuid ") + tmp, ErrorCodes::CANNOT_PARSE_UUID);
+                        throw ParsingException(std::string("Cannot parse uuid ") + tmp, ErrorCodes::CANNOT_PARSE_UUID);
 
                     UUID uuid;
                     parseUUID(reinterpret_cast<const UInt8 *>(tmp.data()), std::reverse_iterator<UInt8 *>(reinterpret_cast<UInt8 *>(&uuid) + 16));
diff --git a/src/Processors/Formats/Impl/JSONCompactEachRowRowInputFormat.cpp b/src/Processors/Formats/Impl/JSONCompactEachRowRowInputFormat.cpp
index f20e764dcfd..1fc5041b1f3 100644
--- a/src/Processors/Formats/Impl/JSONCompactEachRowRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/JSONCompactEachRowRowInputFormat.cpp
@@ -171,7 +171,7 @@ bool JSONCompactEachRowRowInputFormat::readRow(DB::MutableColumns &columns, DB::
 
         skipWhitespaceIfAny(in);
         if (in.eof())
-            throw Exception("Unexpected end of stream while parsing JSONCompactEachRow format", ErrorCodes::CANNOT_READ_ALL_DATA);
+            throw ParsingException("Unexpected end of stream while parsing JSONCompactEachRow format", ErrorCodes::CANNOT_READ_ALL_DATA);
         if (file_column + 1 != column_indexes_for_input_fields.size())
         {
             assertChar(',', in);
diff --git a/src/Processors/Formats/Impl/JSONEachRowRowInputFormat.cpp b/src/Processors/Formats/Impl/JSONEachRowRowInputFormat.cpp
index 234839b41f5..8a707ae6554 100644
--- a/src/Processors/Formats/Impl/JSONEachRowRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/JSONEachRowRowInputFormat.cpp
@@ -173,7 +173,7 @@ inline bool JSONEachRowRowInputFormat::advanceToNextKey(size_t key_index)
     skipWhitespaceIfAny(in);
 
     if (in.eof())
-        throw Exception("Unexpected end of stream while parsing JSONEachRow format", ErrorCodes::CANNOT_READ_ALL_DATA);
+        throw ParsingException("Unexpected end of stream while parsing JSONEachRow format", ErrorCodes::CANNOT_READ_ALL_DATA);
     else if (*in.position() == '}')
     {
         ++in.position();
diff --git a/src/Processors/Formats/Impl/ORCBlockInputFormat.cpp b/src/Processors/Formats/Impl/ORCBlockInputFormat.cpp
index a12ca09eec0..8bbf0fc089b 100644
--- a/src/Processors/Formats/Impl/ORCBlockInputFormat.cpp
+++ b/src/Processors/Formats/Impl/ORCBlockInputFormat.cpp
@@ -38,7 +38,7 @@ Chunk ORCBlockInputFormat::generate()
     std::shared_ptr<arrow::Table> table;
     arrow::Status read_status = file_reader->Read(&table);
     if (!read_status.ok())
-        throw Exception{"Error while reading ORC data: " + read_status.ToString(),
+        throw ParsingException{"Error while reading ORC data: " + read_status.ToString(),
                         ErrorCodes::CANNOT_READ_ALL_DATA};
 
     ArrowColumnToCHColumn::arrowTableToCHChunk(res, table, header, "ORC");
diff --git a/src/Processors/Formats/Impl/ParquetBlockInputFormat.cpp b/src/Processors/Formats/Impl/ParquetBlockInputFormat.cpp
index 031974dc357..bb55c71b7ca 100644
--- a/src/Processors/Formats/Impl/ParquetBlockInputFormat.cpp
+++ b/src/Processors/Formats/Impl/ParquetBlockInputFormat.cpp
@@ -47,7 +47,7 @@ Chunk ParquetBlockInputFormat::generate()
     std::shared_ptr<arrow::Table> table;
     arrow::Status read_status = file_reader->ReadRowGroup(row_group_current, column_indices, &table);
     if (!read_status.ok())
-        throw Exception{"Error while reading Parquet data: " + read_status.ToString(),
+        throw ParsingException{"Error while reading Parquet data: " + read_status.ToString(),
                         ErrorCodes::CANNOT_READ_ALL_DATA};
 
     ++row_group_current;
diff --git a/src/Processors/Formats/Impl/TSKVRowInputFormat.cpp b/src/Processors/Formats/Impl/TSKVRowInputFormat.cpp
index abb468741c5..8d769cab346 100644
--- a/src/Processors/Formats/Impl/TSKVRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/TSKVRowInputFormat.cpp
@@ -89,7 +89,7 @@ static bool readName(ReadBuffer & buf, StringRef & ref, String & tmp)
         }
     }
 
-    throw Exception("Unexpected end of stream while reading key name from TSKV format", ErrorCodes::CANNOT_READ_ALL_DATA);
+    throw ParsingException("Unexpected end of stream while reading key name from TSKV format", ErrorCodes::CANNOT_READ_ALL_DATA);
 }
 
 
@@ -157,7 +157,7 @@ bool TSKVRowInputFormat::readRow(MutableColumns & columns, RowReadExtension & ex
 
             if (in.eof())
             {
-                throw Exception("Unexpected end of stream after field in TSKV format: " + name_ref.toString(), ErrorCodes::CANNOT_READ_ALL_DATA);
+                throw ParsingException("Unexpected end of stream after field in TSKV format: " + name_ref.toString(), ErrorCodes::CANNOT_READ_ALL_DATA);
             }
             else if (*in.position() == '\t')
             {
diff --git a/src/Processors/Formats/Impl/TemplateRowInputFormat.cpp b/src/Processors/Formats/Impl/TemplateRowInputFormat.cpp
index ecb529a99af..6023b38e4de 100644
--- a/src/Processors/Formats/Impl/TemplateRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/TemplateRowInputFormat.cpp
@@ -489,7 +489,7 @@ void TemplateRowInputFormat::skipToNextDelimiterOrEof(const String & delimiter)
 
 void TemplateRowInputFormat::throwUnexpectedEof()
 {
-    throw Exception("Unexpected EOF while parsing row " + std::to_string(row_num) + ". "
+    throw ParsingException("Unexpected EOF while parsing row " + std::to_string(row_num) + ". "
                     "Maybe last row has wrong format or input doesn't contain specified suffix before EOF.",
                     ErrorCodes::CANNOT_READ_ALL_DATA);
 }
diff --git a/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh b/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh
index 805f4267818..caa180b1e0f 100755
--- a/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh
+++ b/tests/queries/0_stateless/01583_parallel_parsing_exception_with_offset.sh
@@ -9,6 +9,6 @@ $CLICKHOUSE_CLIENT --query="DROP TABLE IF EXISTS check;"
 
 $CLICKHOUSE_CLIENT --query="CREATE TABLE check (x UInt64) ENGINE = Memory;"
 
-(seq 1 2000000; echo 'hello'; seq 1 20000000) | $CLICKHOUSE_CLIENT --multiquery --query="SET input_format_parallel_parsing=1; SET min_chunk_bytes_for_parallel_parsing=100000; INSERT INTO check(x) FORMAT TSV " 2>&1 | grep -q "Offset: 1988984" && echo 'OK' || echo 'FAIL' ||:  
+(seq 1 2000000; echo 'hello'; seq 1 20000000) | $CLICKHOUSE_CLIENT --input_format_parallel_parsing=1 --min_chunk_bytes_for_parallel_parsing=1000 --query="INSERT INTO check(x) FORMAT TSV " 2>&1 | grep -q "(at row 2000001)" && echo 'OK' || echo 'FAIL' ||:  
 
 $CLICKHOUSE_CLIENT --query="DROP TABLE check;"

From a285cb83df5caea3592e13e9f2eede7a3f8dbc8c Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 10 Dec 2020 21:02:40 +0300
Subject: [PATCH 289/504] style

---
 src/Common/Exception.h                              | 13 +++++--------
 src/DataStreams/ParallelParsingBlockInputStream.cpp |  5 ++---
 src/Processors/Formats/IRowInputFormat.cpp          |  2 +-
 src/Processors/Formats/Impl/CSVRowInputFormat.cpp   |  2 +-
 4 files changed, 9 insertions(+), 13 deletions(-)

diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index bff39fe454e..0610af6bcc7 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -98,12 +98,13 @@ private:
 
 /// Special class of exceptions, used mostly in ParallelParsingInputFormat for 
 /// more convinient calculation of problem line number.
-class ParsingException : public Exception 
+class ParsingException : public Exception
 {
 public:
     using Exception::Exception;
 
-    void formatInternalMessage() {
+    void formatInternalMessage()
+    {
         try
         {
             message(fmt::format(message(), line_number_));
@@ -111,13 +112,9 @@ public:
         catch (...) {}
     }
 
-    int getLineNumber() {
-        return line_number_;
-    }
+    int getLineNumber() { return line_number_; }
 
-    void setLineNumber(int line_number) {
-        line_number_ = line_number;
-    }
+    void setLineNumber(int line_number) { line_number_ = line_number;}
 
 private:
     int line_number_;
diff --git a/src/DataStreams/ParallelParsingBlockInputStream.cpp b/src/DataStreams/ParallelParsingBlockInputStream.cpp
index ec0652ae766..c3daed21434 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.cpp
+++ b/src/DataStreams/ParallelParsingBlockInputStream.cpp
@@ -132,7 +132,6 @@ void ParallelParsingBlockInputStream::segmentatorThreadFunction(ThreadGroupStatu
             unit.offset = successfully_read_rows_count;
             successfully_read_rows_count += currently_read_rows;
 
-            
             unit.is_last = !have_more_data;
             unit.status = READY_TO_PARSE;
             scheduleParserThreadForUnitWithNumber(segmentator_ticket_number);
@@ -209,8 +208,8 @@ void ParallelParsingBlockInputStream::onBackgroundException(size_t offset)
     if (!background_exception)
     {
         background_exception = std::current_exception();
- 
-        if (ParsingException * e = exception_cast<ParsingException *>(background_exception)) 
+
+        if (ParsingException * e = exception_cast<ParsingException *>(background_exception))
         {
             e->setLineNumber(e->getLineNumber() + offset);
             e->formatInternalMessage();
diff --git a/src/Processors/Formats/IRowInputFormat.cpp b/src/Processors/Formats/IRowInputFormat.cpp
index f65a590ab42..85a7750b0b1 100644
--- a/src/Processors/Formats/IRowInputFormat.cpp
+++ b/src/Processors/Formats/IRowInputFormat.cpp
@@ -132,7 +132,7 @@ Chunk IRowInputFormat::generate()
             }
         }
     }
-    catch (ParsingException & e) 
+    catch (ParsingException & e)
     {
         String verbose_diagnostic;
         try
diff --git a/src/Processors/Formats/Impl/CSVRowInputFormat.cpp b/src/Processors/Formats/Impl/CSVRowInputFormat.cpp
index 89f0e0e5d2c..8422f09e364 100644
--- a/src/Processors/Formats/Impl/CSVRowInputFormat.cpp
+++ b/src/Processors/Formats/Impl/CSVRowInputFormat.cpp
@@ -475,7 +475,7 @@ static std::pair<bool, size_t> fileSegmentationEngineCSVImpl(ReadBuffer & in, DB
                 {
                     ++pos;
                     ++number_of_rows;
-                }  
+                }
             }
         }
     }

From 8d4465ddeb6c9a24b81aabb2531ce3ec6f4dcbb4 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 11 Dec 2020 00:12:54 +0300
Subject: [PATCH 290/504] better

---
 src/Common/Exception.h                        | 25 ++++++++++++++++---
 .../ParallelParsingBlockInputStream.cpp       |  1 -
 2 files changed, 21 insertions(+), 5 deletions(-)

diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index 0610af6bcc7..d03c0a19b6c 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -55,6 +55,11 @@ public:
         extendedMessage(fmt::format(format, std::forward<Args>(args)...));
     }
 
+    /// Note that base class has exactly the same, but not virtual function.
+    virtual std::string displayText() const {
+        return Poco::Exception::displayText();
+    }
+
     void addMessage(const std::string& message)
     {
         extendedMessage(message);
@@ -103,17 +108,29 @@ class ParsingException : public Exception
 public:
     using Exception::Exception;
 
-    void formatInternalMessage()
-    {
+    /// In a good way we have to mark this function virtual in the base class Poco::Exception. 
+    /// For now it is virtual only since Exception (above class) in the hierarchy.
+    std::string displayText() const override {
         try
         {
-            message(fmt::format(message(), line_number_));
+            formatted_message_ = fmt::format(message(), line_number_);
         }
         catch (...) {}
+        
+        if (!formatted_message_.empty())
+        {
+            std::string result = name();
+            result.append(": ");
+            result.append(formatted_message_);
+            return result;
+        }
+        else 
+        {
+            return Exception::displayText();
+        }
     }
 
     int getLineNumber() { return line_number_; }
-
     void setLineNumber(int line_number) { line_number_ = line_number;}
 
 private:
diff --git a/src/DataStreams/ParallelParsingBlockInputStream.cpp b/src/DataStreams/ParallelParsingBlockInputStream.cpp
index c3daed21434..49b7929d7d7 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.cpp
+++ b/src/DataStreams/ParallelParsingBlockInputStream.cpp
@@ -212,7 +212,6 @@ void ParallelParsingBlockInputStream::onBackgroundException(size_t offset)
         if (ParsingException * e = exception_cast<ParsingException *>(background_exception))
         {
             e->setLineNumber(e->getLineNumber() + offset);
-            e->formatInternalMessage();
         }
     }
     tryLogCurrentException(__PRETTY_FUNCTION__);

From 4bb52c062a010cc3728dcac7b648c62b0d4a8472 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 11 Dec 2020 17:22:37 +0300
Subject: [PATCH 291/504] fix tests

---
 tests/queries/0_stateless/00900_parquet_load.reference | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/queries/0_stateless/00900_parquet_load.reference b/tests/queries/0_stateless/00900_parquet_load.reference
index f93be897da8..52c5fbedcc6 100644
--- a/tests/queries/0_stateless/00900_parquet_load.reference
+++ b/tests/queries/0_stateless/00900_parquet_load.reference
@@ -52,7 +52,7 @@
 23.00
 24.00
 === Try load data from datapage_v2.snappy.parquet
-Code: 33. DB::Ex---tion: Error while reading Parquet data: IOError: Not yet implemented: Unsupported encoding.: data for INSERT was parsed from stdin
+Code: 33. DB::ParsingEx---tion: Error while reading Parquet data: IOError: Not yet implemented: Unsupported encoding.: data for INSERT was parsed from stdin
 
 === Try load data from dict-page-offset-zero.parquet
 1552
@@ -257,7 +257,7 @@ Code: 70. DB::Ex---tion: The type "list" of an input column "int64_list" is not
 Code: 70. DB::Ex---tion: The type "list" of an input column "a" is not supported for conversion from a Parquet data format: data for INSERT was parsed from stdin
 
 === Try load data from nested_maps.snappy.parquet
-Code: 70. DB::Ex---tion: The type "map" of an input column "a" is not supported for conversion from a Parquet data format: data for INSERT was parsed from stdin
+Code: 70. DB::ParsingEx---tion: The type "map" of an input column "a" is not supported for conversion from a Parquet data format: data for INSERT was parsed from stdin
 
 === Try load data from non_hadoop_lz4_compressed.parquet
 1593604800	abc	42

From 0dafcc38a5d2155a538c6fdcc828950c25d93cc6 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 11 Dec 2020 18:15:18 +0300
Subject: [PATCH 292/504] better

---
 contrib/poco           | 2 +-
 src/Common/Exception.h | 8 +-------
 2 files changed, 2 insertions(+), 8 deletions(-)

diff --git a/contrib/poco b/contrib/poco
index 08974cc024b..2c32e17c7df 160000
--- a/contrib/poco
+++ b/contrib/poco
@@ -1 +1 @@
-Subproject commit 08974cc024b2e748f5b1d45415396706b3521d0f
+Subproject commit 2c32e17c7dfee1f8bf24227b697cdef5fddf0823
diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index d03c0a19b6c..0129045abd3 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -55,11 +55,6 @@ public:
         extendedMessage(fmt::format(format, std::forward<Args>(args)...));
     }
 
-    /// Note that base class has exactly the same, but not virtual function.
-    virtual std::string displayText() const {
-        return Poco::Exception::displayText();
-    }
-
     void addMessage(const std::string& message)
     {
         extendedMessage(message);
@@ -108,8 +103,7 @@ class ParsingException : public Exception
 public:
     using Exception::Exception;
 
-    /// In a good way we have to mark this function virtual in the base class Poco::Exception. 
-    /// For now it is virtual only since Exception (above class) in the hierarchy.
+    /// We use additional field formatted_message_ to make this method const.
     std::string displayText() const override {
         try
         {

From 35dbd2bd77ff93c81fdf1cc821ecfcb9ce167948 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 11 Dec 2020 23:57:16 +0300
Subject: [PATCH 293/504] style

---
 src/Common/Exception.h | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index 0129045abd3..91270d9116b 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -96,7 +96,7 @@ private:
 };
 
 
-/// Special class of exceptions, used mostly in ParallelParsingInputFormat for 
+/// Special class of exceptions, used mostly in ParallelParsingInputFormat for
 /// more convinient calculation of problem line number.
 class ParsingException : public Exception
 {
@@ -104,13 +104,15 @@ public:
     using Exception::Exception;
 
     /// We use additional field formatted_message_ to make this method const.
-    std::string displayText() const override {
+    std::string displayText() const override
+    {
         try
         {
             formatted_message_ = fmt::format(message(), line_number_);
         }
-        catch (...) {}
-        
+        catch (...)
+        {}
+
         if (!formatted_message_.empty())
         {
             std::string result = name();
@@ -118,7 +120,7 @@ public:
             result.append(formatted_message_);
             return result;
         }
-        else 
+        else
         {
             return Exception::displayText();
         }

From a9bd177b2e0cdcba18371bd5dcf7e2adb932fad5 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 18 Dec 2020 03:17:26 +0300
Subject: [PATCH 294/504] fix

---
 src/Common/Exception.h                     | 24 ++++++++++++++++++++--
 src/Processors/Formats/IRowInputFormat.cpp |  2 +-
 2 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index 91270d9116b..e523aa515d7 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -102,13 +102,33 @@ class ParsingException : public Exception
 {
 public:
     using Exception::Exception;
+    
+    ParsingException()
+    {
+        Exception::addMessage("{}");
+    }
+    
+    ParsingException(const std::string & msg, int code)
+        : Exception(msg, code) 
+    {
+        Exception::addMessage("{}");
+    }
+
+    ParsingException(int code, const std::string & message)
+        : Exception(message, code)
+    {
+        Exception::addMessage("{}");
+    }
 
     /// We use additional field formatted_message_ to make this method const.
     std::string displayText() const override
     {
         try
         {
-            formatted_message_ = fmt::format(message(), line_number_);
+            if (line_number_ == -1)
+                formatted_message_ = fmt::format(message(), "");
+            else
+                formatted_message_ = fmt::format(message(), fmt::format("(at row {})\n", line_number_));
         }
         catch (...)
         {}
@@ -130,7 +150,7 @@ public:
     void setLineNumber(int line_number) { line_number_ = line_number;}
 
 private:
-    int line_number_;
+    ssize_t line_number_{-1};
     mutable std::string formatted_message_;
 
     const char * name() const throw() override { return "DB::ParsingException"; }
diff --git a/src/Processors/Formats/IRowInputFormat.cpp b/src/Processors/Formats/IRowInputFormat.cpp
index 85a7750b0b1..79090ae2b89 100644
--- a/src/Processors/Formats/IRowInputFormat.cpp
+++ b/src/Processors/Formats/IRowInputFormat.cpp
@@ -149,7 +149,7 @@ Chunk IRowInputFormat::generate()
         }
 
         e.setLineNumber(total_rows);
-        e.addMessage("(at row {})\n" + verbose_diagnostic);
+        e.addMessage(verbose_diagnostic);
         throw;
     }
     catch (Exception & e)

From 6f30fae34eb09416f4d7b9e2b037133dfd077d68 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 18 Dec 2020 04:47:24 +0300
Subject: [PATCH 295/504] final fix

---
 src/Common/Exception.cpp | 44 ++++++++++++++++++++++++++++++++++++
 src/Common/Exception.h   | 48 ++++++++--------------------------------
 2 files changed, 53 insertions(+), 39 deletions(-)

diff --git a/src/Common/Exception.cpp b/src/Common/Exception.cpp
index d9bbb170dcc..50a78cc1c83 100644
--- a/src/Common/Exception.cpp
+++ b/src/Common/Exception.cpp
@@ -450,5 +450,49 @@ ExecutionStatus ExecutionStatus::fromCurrentException(const std::string & start_
     return ExecutionStatus(getCurrentExceptionCode(), msg);
 }
 
+ParsingException::ParsingException()
+{
+    Exception::message(Exception::message() + "{}");
+}
+
+ParsingException::ParsingException(const std::string & msg, int code)
+    : Exception(msg, code) 
+{
+    Exception::message(Exception::message() + "{}");
+}
+
+ParsingException::ParsingException(int code, const std::string & message)
+    : Exception(message, code)
+{
+    Exception::message(Exception::message() + "{}");
+}
+
+
+/// We use additional field formatted_message_ to make this method const.
+std::string ParsingException::displayText() const
+{
+    try
+    {
+        if (line_number_ == -1)
+            formatted_message_ = fmt::format(message(), "");
+        else
+            formatted_message_ = fmt::format(message(), fmt::format(": (at row {})\n", line_number_));
+    }
+    catch (...)
+    {}
+
+    if (!formatted_message_.empty())
+    {
+        std::string result = name();
+        result.append(": ");
+        result.append(formatted_message_);
+        return result;
+    }
+    else
+    {
+        return Exception::displayText();
+    }
+}
+
 
 }
diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index e523aa515d7..4e5d8c8ad47 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -101,50 +101,20 @@ private:
 class ParsingException : public Exception
 {
 public:
-    using Exception::Exception;
+    ParsingException();
+    ParsingException(const std::string & msg, int code);
+    ParsingException(int code, const std::string & message);
     
-    ParsingException()
+    // Format message with fmt::format, like the logging functions.
+    template <typename ...Args>
+    ParsingException(int code, const std::string & fmt, Args&&... args)
+        : Exception(fmt::format(fmt, std::forward<Args>(args)...), code)
     {
-        Exception::addMessage("{}");
-    }
-    
-    ParsingException(const std::string & msg, int code)
-        : Exception(msg, code) 
-    {
-        Exception::addMessage("{}");
+        Exception::message(Exception::message() + "{}");
     }
 
-    ParsingException(int code, const std::string & message)
-        : Exception(message, code)
-    {
-        Exception::addMessage("{}");
-    }
 
-    /// We use additional field formatted_message_ to make this method const.
-    std::string displayText() const override
-    {
-        try
-        {
-            if (line_number_ == -1)
-                formatted_message_ = fmt::format(message(), "");
-            else
-                formatted_message_ = fmt::format(message(), fmt::format("(at row {})\n", line_number_));
-        }
-        catch (...)
-        {}
-
-        if (!formatted_message_.empty())
-        {
-            std::string result = name();
-            result.append(": ");
-            result.append(formatted_message_);
-            return result;
-        }
-        else
-        {
-            return Exception::displayText();
-        }
-    }
+    std::string displayText() const override;
 
     int getLineNumber() { return line_number_; }
     void setLineNumber(int line_number) { line_number_ = line_number;}

From 138b74aa3bbd95bfa5bb31c9e309e10c0587c8a5 Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@mtlog-perftest03j.yandex.ru>
Date: Wed, 23 Dec 2020 01:04:02 +0300
Subject: [PATCH 296/504] better

---
 src/DataStreams/NativeBlockInputStream.cpp | 2 +-
 src/DataTypes/DataTypeFixedString.cpp      | 2 +-
 src/IO/ReadBuffer.h                        | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/DataStreams/NativeBlockInputStream.cpp b/src/DataStreams/NativeBlockInputStream.cpp
index e9967e88638..b182d5e0588 100644
--- a/src/DataStreams/NativeBlockInputStream.cpp
+++ b/src/DataStreams/NativeBlockInputStream.cpp
@@ -83,7 +83,7 @@ void NativeBlockInputStream::readData(const IDataType & type, IColumn & column,
     type.deserializeBinaryBulkWithMultipleStreams(column, rows, settings, state);
 
     if (column.size() != rows)
-        throw ParsingException("Cannot read all data in NativeBlockInputStream. Rows read: " + toString(column.size()) + ". Rows expected: " + toString(rows) + ".",
+        throw Exception("Cannot read all data in NativeBlockInputStream. Rows read: " + toString(column.size()) + ". Rows expected: " + toString(rows) + ".",
             ErrorCodes::CANNOT_READ_ALL_DATA);
 }
 
diff --git a/src/DataTypes/DataTypeFixedString.cpp b/src/DataTypes/DataTypeFixedString.cpp
index c0b684cfcd9..585c5709be7 100644
--- a/src/DataTypes/DataTypeFixedString.cpp
+++ b/src/DataTypes/DataTypeFixedString.cpp
@@ -103,7 +103,7 @@ void DataTypeFixedString::deserializeBinaryBulk(IColumn & column, ReadBuffer & i
     size_t read_bytes = istr.readBig(reinterpret_cast<char *>(&data[initial_size]), max_bytes);
 
     if (read_bytes % n != 0)
-        throw ParsingException("Cannot read all data of type FixedString. Bytes read:" + toString(read_bytes) + ". String size:" + toString(n) + ".",
+        throw Exception("Cannot read all data of type FixedString. Bytes read:" + toString(read_bytes) + ". String size:" + toString(n) + ".",
             ErrorCodes::CANNOT_READ_ALL_DATA);
 
     data.resize(initial_size + read_bytes);
diff --git a/src/IO/ReadBuffer.h b/src/IO/ReadBuffer.h
index 08f1bc2a3bb..3d6eb6970ce 100644
--- a/src/IO/ReadBuffer.h
+++ b/src/IO/ReadBuffer.h
@@ -165,7 +165,7 @@ public:
     {
         auto read_bytes = read(to, n);
         if (n != read_bytes)
-            throw ParsingException("Cannot read all data. Bytes read: " + std::to_string(read_bytes) + ". Bytes expected: " + std::to_string(n) + ".", ErrorCodes::CANNOT_READ_ALL_DATA);
+            throw Exception("Cannot read all data. Bytes read: " + std::to_string(read_bytes) + ". Bytes expected: " + std::to_string(n) + ".", ErrorCodes::CANNOT_READ_ALL_DATA);
     }
 
     /** A method that can be more efficiently implemented in derived classes, in the case of reading large enough blocks.

From 84d0084c1a7314cada3c8f23613167451eca6c3c Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@mtlog-perftest03j.yandex.ru>
Date: Wed, 23 Dec 2020 01:33:07 +0300
Subject: [PATCH 297/504] style

---
 src/Common/Exception.cpp | 2 +-
 src/Common/Exception.h   | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Common/Exception.cpp b/src/Common/Exception.cpp
index 50a78cc1c83..b782471a4e8 100644
--- a/src/Common/Exception.cpp
+++ b/src/Common/Exception.cpp
@@ -456,7 +456,7 @@ ParsingException::ParsingException()
 }
 
 ParsingException::ParsingException(const std::string & msg, int code)
-    : Exception(msg, code) 
+    : Exception(msg, code)
 {
     Exception::message(Exception::message() + "{}");
 }
diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index 4e5d8c8ad47..4b04de5d8a2 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -104,7 +104,7 @@ public:
     ParsingException();
     ParsingException(const std::string & msg, int code);
     ParsingException(int code, const std::string & message);
-    
+
     // Format message with fmt::format, like the logging functions.
     template <typename ...Args>
     ParsingException(int code, const std::string & fmt, Args&&... args)

From ffce2a490bb049bfcd33977f5a3563e19ed5b0a3 Mon Sep 17 00:00:00 2001
From: nikitamikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 25 Sep 2020 19:42:02 +0300
Subject: [PATCH 298/504] done

---
 .../configs/conf.d/big_clusters.xml           |  71 ++++++++
 .../test_cluster_copier/task_many_to_one.xml  | 128 ++++++++++++++
 tests/integration/test_cluster_copier/test.py |  32 +---
 .../test_cluster_copier/test_big_cluster.py   | 161 ++++++++++++++++++
 .../{trivial_test.py => test_trivial.py}      |  66 +++----
 5 files changed, 392 insertions(+), 66 deletions(-)
 create mode 100644 tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml
 create mode 100644 tests/integration/test_cluster_copier/task_many_to_one.xml
 create mode 100644 tests/integration/test_cluster_copier/test_big_cluster.py
 rename tests/integration/test_cluster_copier/{trivial_test.py => test_trivial.py} (75%)

diff --git a/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml b/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml
new file mode 100644
index 00000000000..cba995af211
--- /dev/null
+++ b/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml
@@ -0,0 +1,71 @@
+<?xml version="1.0"?>
+<yandex>
+    <remote_servers>
+        <anime_source>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_0_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_0_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_1_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_1_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_2_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_2_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_3_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_3_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_4_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_4_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+        </anime_source>
+
+        <anime_destination>
+            <shard> 
+                <replica>
+                    <host>s1_0_0</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+        </anime_destination>
+    </remote_servers>
+</yandex>
\ No newline at end of file
diff --git a/tests/integration/test_cluster_copier/task_many_to_one.xml b/tests/integration/test_cluster_copier/task_many_to_one.xml
new file mode 100644
index 00000000000..ec28948e7ad
--- /dev/null
+++ b/tests/integration/test_cluster_copier/task_many_to_one.xml
@@ -0,0 +1,128 @@
+<yandex>
+    <!-- Configuration of clusters as in an ordinary server config -->
+    <remote_servers>
+        <anime_source>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_0_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_0_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_1_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_1_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_2_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_2_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_3_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_3_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+            <shard>
+                <internal_replication>true</internal_replication>    
+                <replica>
+                    <host>s0_4_0</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>s0_4_1</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+        </anime_source>
+
+        <anime_destination>
+            <shard> 
+                <replica>
+                    <host>s1_0_0</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+        </anime_destination>
+    </remote_servers>
+
+    <!-- How many simultaneously active workers are possible. If you run more workers superfluous workers will sleep. -->
+    <max_workers>3</max_workers>
+
+    <!-- Setting used to fetch (pull) data from source cluster tables -->
+    <settings_pull>
+        <readonly>1</readonly>
+    </settings_pull>
+
+    <!-- Setting used to insert (push) data to destination cluster tables -->
+    <settings_push>
+        <readonly>0</readonly>
+    </settings_push>
+
+    <!-- Common setting for fetch (pull) and insert (push) operations. Also, copier process context uses it.
+         They are overlaid by <settings_pull/> and <settings_push/> respectively. -->
+    <settings>
+        <connect_timeout>3</connect_timeout>
+        <!-- Sync insert is set forcibly, leave it here just in case. -->
+        <insert_distributed_sync>1</insert_distributed_sync>
+    </settings>
+
+    <!-- Copying tasks description.
+         You could specify several table task in the same task description (in the same ZooKeeper node), they will be performed
+         sequentially.
+    -->
+    <tables>
+        <!-- A table task, copies one table. -->
+        <table_first>
+            <!-- Source cluster name (from <remote_servers/> section) and tables in it that should be copied -->
+            <cluster_pull>anime_source</cluster_pull>
+            <database_pull>default</database_pull>
+            <table_pull>anime_all</table_pull>
+
+            <!-- Destination cluster name and tables in which the data should be inserted -->
+            <cluster_push>anime_destination</cluster_push>
+            <database_push>default</database_push>
+            <table_push>anime</table_push>
+			
+			<!-- Engine of destination tables.
+                 If destination tables have not be created, workers create them using columns definition from source tables and engine
+                 definition from here.
+                 NOTE: If the first worker starts insert data and detects that destination partition is not empty then the partition will
+                 be dropped and refilled, take it into account if you already have some data in destination tables. You could directly
+                 specify partitions that should be copied in <enabled_partitions/>, they should be in quoted format like partition column of
+                 system.parts table.
+            -->
+            <engine>
+				ENGINE = MergeTree()
+				ORDER BY (first_id, second_id, toSecond(datetime))
+                PARTITION BY toSecond(datetime)
+            </engine>
+
+            <!-- Sharding key used to insert data to destination cluster -->
+            <sharding_key>cityHash64(first_id, second_id)</sharding_key>
+        </table_first>
+    </tables>
+</yandex>
\ No newline at end of file
diff --git a/tests/integration/test_cluster_copier/test.py b/tests/integration/test_cluster_copier/test.py
index d87969630cd..00d464af98f 100644
--- a/tests/integration/test_cluster_copier/test.py
+++ b/tests/integration/test_cluster_copier/test.py
@@ -94,6 +94,8 @@ class Task1:
         instance.query("INSERT INTO hits_all SELECT * FROM system.numbers LIMIT 1002",
                        settings={"insert_distributed_sync": 1})
 
+        print instance.query("SELECT DISTINCT 'all' AS partition FROM hits_all ORDER BY partition DESC")
+
     def check(self):
         assert TSV(self.cluster.instances['s0_0_0'].query("SELECT count() FROM hits_all")) == TSV("1002\n")
         assert TSV(self.cluster.instances['s1_0_0'].query("SELECT count() FROM hits_all")) == TSV("1002\n")
@@ -310,13 +312,7 @@ def execute_task(task, cmd_options):
 
 # Tests
 
-@pytest.mark.parametrize(
-    ('use_sample_offset'),
-    [
-        False,
-        True
-    ]
-)
+@pytest.mark.parametrize(('use_sample_offset'),[False,True])
 def test_copy_simple(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(Task1(started_cluster), ['--experimental-use-sample-offset', '1'])
@@ -324,13 +320,7 @@ def test_copy_simple(started_cluster, use_sample_offset):
         execute_task(Task1(started_cluster), [])
 
 
-@pytest.mark.parametrize(
-    ('use_sample_offset'),
-    [
-        False,
-        True
-    ]
-)
+@pytest.mark.parametrize(('use_sample_offset'),[False,True])
 def test_copy_with_recovering(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(Task1(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
@@ -339,13 +329,7 @@ def test_copy_with_recovering(started_cluster, use_sample_offset):
         execute_task(Task1(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
 
 
-@pytest.mark.parametrize(
-    ('use_sample_offset'),
-    [
-        False,
-        True
-    ]
-)
+@pytest.mark.parametrize(('use_sample_offset'),[False,True])
 def test_copy_with_recovering_after_move_faults(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(Task1(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
@@ -380,11 +364,7 @@ def test_no_index(started_cluster):
 def test_no_arg(started_cluster):
     execute_task(Task_no_arg(started_cluster), [])
 
+
 def test_non_partitioned_table(started_cluster):
     execute_task(Task_non_partitioned_table(started_cluster), [])
 
-if __name__ == '__main__':
-    with contextmanager(started_cluster)() as cluster:
-        for name, instance in list(cluster.instances.items()):
-            print(name, instance.ip_address)
-        input("Cluster created, press any key to destroy...")
diff --git a/tests/integration/test_cluster_copier/test_big_cluster.py b/tests/integration/test_cluster_copier/test_big_cluster.py
new file mode 100644
index 00000000000..007b118703f
--- /dev/null
+++ b/tests/integration/test_cluster_copier/test_big_cluster.py
@@ -0,0 +1,161 @@
+import os
+import random
+import sys
+import time
+from contextlib import contextmanager
+
+import docker
+import kazoo
+import pytest
+from helpers.cluster import ClickHouseCluster
+from helpers.test_tools import TSV
+
+CURRENT_TEST_DIR = os.path.dirname(os.path.abspath(__file__))
+sys.path.insert(0, os.path.dirname(CURRENT_TEST_DIR))
+
+COPYING_FAIL_PROBABILITY = 0.2
+MOVING_FAIL_PROBABILITY = 0.2
+
+cluster = ClickHouseCluster(__file__)
+
+
+def check_all_hosts_sucesfully_executed(tsv_content, num_hosts):
+    M = TSV.toMat(tsv_content)
+    hosts = [(l[0], l[1]) for l in M]  # (host, port)
+    codes = [l[2] for l in M]
+
+    assert len(hosts) == num_hosts and len(set(hosts)) == num_hosts, "\n" + tsv_content
+    assert len(set(codes)) == 1, "\n" + tsv_content
+    assert codes[0] == "0", "\n" + tsv_content
+
+
+def ddl_check_query(instance, query, num_hosts=3):
+    contents = instance.query(query)
+    check_all_hosts_sucesfully_executed(contents, num_hosts)
+    return contents
+
+
+@pytest.fixture(scope="module")
+def started_cluster():
+    global cluster
+    try:
+        clusters_schema = {
+            "0": {
+                "0": ["0", "1"],
+                "1": ["0", "1"],
+                "2": ["0", "1"],
+                "3": ["0", "1"],
+                "4": ["0", "1"]
+            },
+            "1": {
+                "0": ["0"]
+            }
+        }
+
+        for cluster_name, shards in clusters_schema.iteritems():
+            for shard_name, replicas in shards.iteritems():
+                for replica_name in replicas:
+                    name = "s{}_{}_{}".format(cluster_name, shard_name, replica_name)
+                    cluster.add_instance(name,
+                                            main_configs=["configs/conf.d/query_log.xml", "configs/conf.d/ddl.xml",
+                                                        "configs/conf.d/big_clusters.xml"],
+                                            user_configs=["configs/users.xml"],
+                                            macros={"cluster": cluster_name, "shard": shard_name, "replica": replica_name},
+                                            with_zookeeper=True)
+
+        cluster.start()
+        yield cluster
+
+    finally:
+        cluster.shutdown()
+
+
+class Task_many_to_one:
+    def __init__(self, cluster):
+        self.cluster = cluster 
+        self.zk_task_path = "/clickhouse-copier/task_many_to_one"
+        self.copier_task_config = open(os.path.join(CURRENT_TEST_DIR, 'task_many_to_one.xml'), 'r').read()
+        self.rows = 1000
+
+    def start(self):
+        instance = cluster.instances['s0_0_0']
+
+        for cluster_num in ["anime_source", "anime_destination"]:
+            instance.query("DROP DATABASE IF EXISTS default ON CLUSTER {}".format(cluster_num))
+            instance.query("CREATE DATABASE IF NOT EXISTS default ON CLUSTER {}".format(cluster_num))
+
+        instance.query("CREATE TABLE anime ON CLUSTER anime_source (first_id UUID DEFAULT generateUUIDv4(), second_id UInt64, datetime DateTime DEFAULT now()) " +
+                       "ENGINE=ReplicatedMergeTree " +
+                       "PARTITION BY toSecond(datetime) " +  
+                       "ORDER BY (first_id, second_id, toSecond(datetime))")
+        instance.query("CREATE TABLE anime_all ON CLUSTER anime_source (first_id UUID, second_id UInt64, datetime DateTime) ENGINE=Distributed(anime_source, default, anime, rand() % 5)")
+        instance.query("INSERT INTO anime_all SELECT generateUUIDv4(), number, now() FROM system.numbers LIMIT 1002",
+                       settings={"insert_distributed_sync": 1})
+
+    def check(self):
+        assert TSV(self.cluster.instances['s0_0_0'].query("SELECT count() FROM anime_all")) == TSV("1002\n")
+        assert TSV(self.cluster.instances['s1_0_0'].query("SELECT count() FROM anime")) == TSV("1002\n")
+
+def execute_task(task, cmd_options):
+    task.start()
+
+    zk = cluster.get_kazoo_client('zoo1')
+    print "Use ZooKeeper server: {}:{}".format(zk.hosts[0][0], zk.hosts[0][1])
+
+    try:
+        zk.delete("/clickhouse-copier", recursive=True)
+    except kazoo.exceptions.NoNodeError:
+        print "No node /clickhouse-copier. It is Ok in first test."
+
+    zk_task_path = task.zk_task_path
+    zk.ensure_path(zk_task_path)
+    zk.create(zk_task_path + "/description", task.copier_task_config)
+
+    # Run cluster-copier processes on each node
+    docker_api = docker.from_env().api
+    copiers_exec_ids = []
+
+    cmd = ['/usr/bin/clickhouse', 'copier',
+           '--config', '/etc/clickhouse-server/config-copier.xml',
+           '--task-path', zk_task_path,
+           '--base-dir', '/var/log/clickhouse-server/copier']
+    cmd += cmd_options
+    
+    # copiers = []
+    copiers = random.sample(cluster.instances.keys(), 3)
+    copiers.append('s1_0_0')
+
+    for instance_name in copiers:
+        instance = cluster.instances[instance_name]
+        container = instance.get_docker_handle()
+        instance.copy_file_to_container(os.path.join(CURRENT_TEST_DIR, "configs/config-copier.xml"),
+                                        "/etc/clickhouse-server/config-copier.xml")
+        print "Copied copier config to {}".format(instance.name)
+        exec_id = docker_api.exec_create(container.id, cmd, stderr=True)
+        output = docker_api.exec_start(exec_id).decode('utf8')
+        print(output)
+        copiers_exec_ids.append(exec_id)
+        print "Copier for {} ({}) has started".format(instance.name, instance.ip_address)
+
+    # Wait for copiers stopping and check their return codes
+    for exec_id, instance_name in zip(copiers_exec_ids, copiers):
+        instance = cluster.instances[instance_name]
+        while True:
+            res = docker_api.exec_inspect(exec_id)
+            if not res['Running']:
+                break
+            time.sleep(0.5)
+
+        assert res['ExitCode'] == 0, "Instance: {} ({}). Info: {}".format(instance.name, instance.ip_address, repr(res))
+
+    try:
+        task.check()
+    finally:
+        zk.delete(zk_task_path, recursive=True)
+
+
+# Tests
+
+def test_copy_simple(started_cluster):
+    execute_task(Task_many_to_one(started_cluster), [])
+ 
\ No newline at end of file
diff --git a/tests/integration/test_cluster_copier/trivial_test.py b/tests/integration/test_cluster_copier/test_trivial.py
similarity index 75%
rename from tests/integration/test_cluster_copier/trivial_test.py
rename to tests/integration/test_cluster_copier/test_trivial.py
index 8a43440ac90..48d5aaddb21 100644
--- a/tests/integration/test_cluster_copier/trivial_test.py
+++ b/tests/integration/test_cluster_copier/test_trivial.py
@@ -11,8 +11,8 @@ sys.path.insert(0, os.path.dirname(CURRENT_TEST_DIR))
 from helpers.cluster import ClickHouseCluster
 from helpers.test_tools import TSV
 
-COPYING_FAIL_PROBABILITY = 0.33
-MOVING_FAIL_PROBABILITY = 0.1
+COPYING_FAIL_PROBABILITY = 0.2
+MOVING_FAIL_PROBABILITY = 0.2
 cluster = None
 
 
@@ -85,6 +85,11 @@ def execute_task(task, cmd_options):
     zk = cluster.get_kazoo_client('zoo1')
     print("Use ZooKeeper server: {}:{}".format(zk.hosts[0][0], zk.hosts[0][1]))
 
+    try:
+        zk.delete("/clickhouse-copier", recursive=True)
+    except kazoo.exceptions.NoNodeError:
+        print "No node /clickhouse-copier. It is Ok in first test."
+
     zk_task_path = task.zk_task_path
     zk.ensure_path(zk_task_path)
     zk.create(zk_task_path + "/description", task.copier_task_config)
@@ -99,23 +104,28 @@ def execute_task(task, cmd_options):
            '--base-dir', '/var/log/clickhouse-server/copier']
     cmd += cmd_options
 
-    print(cmd)
+    copiers = cluster.instances.keys()
 
-    for instance_name, instance in cluster.instances.items():
+    for instance_name in copiers:
+        instance = cluster.instances[instance_name]
         container = instance.get_docker_handle()
+        instance.copy_file_to_container(os.path.join(CURRENT_TEST_DIR, "configs/config-copier.xml"),
+                                        "/etc/clickhouse-server/config-copier.xml")
+        print "Copied copier config to {}".format(instance.name)
         exec_id = docker_api.exec_create(container.id, cmd, stderr=True)
-        docker_api.exec_start(exec_id, detach=True)
-
+        output = docker_api.exec_start(exec_id).decode('utf8')
+        print(output)
         copiers_exec_ids.append(exec_id)
         print("Copier for {} ({}) has started".format(instance.name, instance.ip_address))
 
     # Wait for copiers stopping and check their return codes
-    for exec_id, instance in zip(copiers_exec_ids, iter(cluster.instances.values())):
+    for exec_id, instance_name in zip(copiers_exec_ids, copiers):
+        instance = cluster.instances[instance_name]
         while True:
             res = docker_api.exec_inspect(exec_id)
             if not res['Running']:
                 break
-            time.sleep(1)
+            time.sleep(0.5)
 
         assert res['ExitCode'] == 0, "Instance: {} ({}). Info: {}".format(instance.name, instance.ip_address, repr(res))
 
@@ -125,56 +135,32 @@ def execute_task(task, cmd_options):
         zk.delete(zk_task_path, recursive=True)
 
 
+
 # Tests
 
 
-@pytest.mark.parametrize(
-    ('use_sample_offset'),
-    [
-        False,
-        True
-    ]
-)
+@pytest.mark.parametrize(('use_sample_offset'),[False,True])
 def test_trivial_copy(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--experimental-use-sample-offset', '1'])
     else:
-        print("AAAAA")
         execute_task(TaskTrivial(started_cluster, use_sample_offset), [])
 
 
-@pytest.mark.parametrize(
-    ('use_sample_offset'),
-    [
-        False,
-        True
-    ]
-)
+@pytest.mark.parametrize(('use_sample_offset'),[False,True])
 def test_trivial_copy_with_copy_fault(started_cluster, use_sample_offset):
     if use_sample_offset:
-        execute_task(TaskTrivial(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
                                                     '--experimental-use-sample-offset', '1'])
     else:
-        execute_task(TaskTrivial(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
 
 
-@pytest.mark.parametrize(
-    ('use_sample_offset'),
-    [
-        False,
-        True
-    ]
-)
+@pytest.mark.parametrize(('use_sample_offset'),[False,True])
 def test_trivial_copy_with_move_fault(started_cluster, use_sample_offset):
     if use_sample_offset:
-        execute_task(TaskTrivial(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
                                                     '--experimental-use-sample-offset', '1'])
     else:
-        execute_task(TaskTrivial(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY)])
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY)])
 
-
-if __name__ == '__main__':
-    with contextmanager(started_cluster)() as cluster:
-        for name, instance in list(cluster.instances.items()):
-            print(name, instance.ip_address)
-        input("Cluster created, press any key to destroy...")

From f91adc1207a702b104f15e5504484a4b9df728ac Mon Sep 17 00:00:00 2001
From: romanzhukov <romanzhukov@yandex-team.ru>
Date: Wed, 23 Dec 2020 01:58:02 +0300
Subject: [PATCH 299/504] DOCSUP-5043: Add RU translation for PR #16378

---
 docs/ru/commercial/cloud.md | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/docs/ru/commercial/cloud.md b/docs/ru/commercial/cloud.md
index e6b0309c456..8d02a6332a7 100644
--- a/docs/ru/commercial/cloud.md
+++ b/docs/ru/commercial/cloud.md
@@ -18,4 +18,16 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 -   Шифрование и изоляция
 -   Автоматизированное техническое обслуживание
 
+## Altinity.Cloud {#altinity.cloud}
+
+[Altinity.Cloud](https://altinity.com/cloud-database/) — это полностью управляемый ClickHouse-as-a-Service для публичного облака Amazon.
+-   Быстрое развертывание кластеров ClickHouse на ресурсах Amazon
+-   Easy scale-out/scale-in as well as vertical scaling of nodes
+-   Легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов
+-   Изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC
+-   Настраиваемые типы и объемы хранилищ
+-   Cross-AZ scaling for performance and high availability
+-   Cross-AZ масштабирование для повышения производительности и высокой доступности
+-   Встроенный мониторинг и редактор SQL-запросов
+
 {## [Оригинальная статья](https://clickhouse.tech/docs/ru/commercial/cloud/) ##}

From 41397b7455763a8f78a7391a5cda420313dabf65 Mon Sep 17 00:00:00 2001
From: romanzhukov <romanzhukov@yandex-team.ru>
Date: Wed, 23 Dec 2020 02:02:04 +0300
Subject: [PATCH 300/504] DOCSUP-5043: Add RU translation for PR # 15423

---
 docs/en/commercial/cloud.md                                    | 2 +-
 docs/ru/operations/server-configuration-parameters/settings.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/commercial/cloud.md b/docs/en/commercial/cloud.md
index 06216517db8..03c8b7645a3 100644
--- a/docs/en/commercial/cloud.md
+++ b/docs/en/commercial/cloud.md
@@ -20,7 +20,7 @@ toc_title: Cloud
 
 ## Altinity.Cloud {#altinity.cloud}
 
-[Altinity.Cloud](https://altinity.com/cloud-database/) is a fully managed ClickHouse-as-a-Service for the Amazon public cloud. 
+[Altinity.Cloud](https://altinity.com/cloud-database/) is a fully managed ClickHouse-as-a-Service for the Amazon public cloud.
 -   Fast deployment of ClickHouse clusters on Amazon resources
 -   Easy scale-out/scale-in as well as vertical scaling of nodes
 -   Isolated per-tenant VPCs with public endpoint or VPC peering
diff --git a/docs/ru/operations/server-configuration-parameters/settings.md b/docs/ru/operations/server-configuration-parameters/settings.md
index 58aae05f188..e07ae968177 100644
--- a/docs/ru/operations/server-configuration-parameters/settings.md
+++ b/docs/ru/operations/server-configuration-parameters/settings.md
@@ -420,7 +420,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 Возможные значения:
 
 -   Положительное целое число.
--   0 — объём используемой памяти не ограничен.
+-   0 — автоматически.
 
 Значение по умолчанию: `0`.
 

From 6af30a5dca8db4fbe7d25bc283663cb755ee5488 Mon Sep 17 00:00:00 2001
From: romanzhukov <romanzhukov@yandex-team.ru>
Date: Wed, 23 Dec 2020 02:23:10 +0300
Subject: [PATCH 301/504] DOCSUP-5043: Add translations and updates for PR
 #15416 (EN & RU)

---
 .../settings.md                               | 17 +++++++++++----
 docs/en/operations/settings/settings.md       |  2 +-
 .../settings.md                               | 21 +++++++++++++++++++
 docs/ru/operations/settings/settings.md       |  2 +-
 4 files changed, 36 insertions(+), 6 deletions(-)

diff --git a/docs/en/operations/server-configuration-parameters/settings.md b/docs/en/operations/server-configuration-parameters/settings.md
index 533fcea5500..1877e474555 100644
--- a/docs/en/operations/server-configuration-parameters/settings.md
+++ b/docs/en/operations/server-configuration-parameters/settings.md
@@ -82,16 +82,25 @@ List of prefixes for [custom settings](../../operations/settings/index.md#custom
 
 ## core_dump
 
-Configures soft limit for core dump file size, one gigabyte by default.
+Configures soft limit for core dump file size.
+
+Possible values:
+
+-   Positive integer.
+
+Default value: `1073741824`.
+
+!!! info "Note"
+    Hard limit is configured via system tools
+
+**Example**
+
 ```xml
 <core_dump>
     <size_limit>1073741824</size_limit>
 </core_dump> 
 ```
 
-(Hard limit is configured via system tools)
-
-
 ## default_database {#default-database}
 
 The default database.
diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 86e16fe4819..93d5bb9c67f 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -1312,7 +1312,7 @@ See also:
 
 Write to a quorum timeout in milliseconds. If the timeout has passed and no write has taken place yet, ClickHouse will generate an exception and the client must repeat the query to write the same block to the same or any other replica.
 
-Default value: 600000 milliseconds (ten minutes).
+Default value: 600 000 milliseconds (ten minutes).
 
 See also:
 
diff --git a/docs/ru/operations/server-configuration-parameters/settings.md b/docs/ru/operations/server-configuration-parameters/settings.md
index e07ae968177..affe09c7703 100644
--- a/docs/ru/operations/server-configuration-parameters/settings.md
+++ b/docs/ru/operations/server-configuration-parameters/settings.md
@@ -80,6 +80,27 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 
 -   [Пользовательские настройки](../../operations/settings/index.md#custom_settings)
 
+## core_dump
+
+Задает мягкое ограничение для размера файла дампа ядра.
+
+Возможные значения:
+
+-   Положительное целое число.
+
+Значение по умолчанию: `1073741824`.
+
+!!! info "Примечание"
+    Жесткое ограничение настраивается с помощью системных инструментов.
+
+**Пример**
+
+```xml
+<core_dump>
+    <size_limit>1073741824</size_limit>
+</core_dump> 
+```
+
 ## default\_database {#default-database}
 
 База данных по умолчанию.
diff --git a/docs/ru/operations/settings/settings.md b/docs/ru/operations/settings/settings.md
index 0a8094231c2..8e41fc421c6 100644
--- a/docs/ru/operations/settings/settings.md
+++ b/docs/ru/operations/settings/settings.md
@@ -1258,7 +1258,7 @@ ClickHouse генерирует исключение
 
 Время ожидания кворумной записи в миллисекундах. Если время прошло, а запись так не состоялась, то ClickHouse сгенерирует исключение и клиент должен повторить запрос на запись того же блока на эту же или любую другую реплику.
 
-Значение по умолчанию: 600000 миллисекунд (10 минут).
+Значение по умолчанию: 600 000 миллисекунд (10 минут).
 
 См. также:
 

From 7932f1ababbe722cfbb88f6df6496e4c9907c7e7 Mon Sep 17 00:00:00 2001
From: romanzhukov <romanzhukov@yandex-team.ru>
Date: Wed, 23 Dec 2020 02:27:53 +0300
Subject: [PATCH 302/504] DOCSUP-5043: Minor fix

---
 docs/ru/sql-reference/functions/tuple-map-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/sql-reference/functions/tuple-map-functions.md b/docs/ru/sql-reference/functions/tuple-map-functions.md
index 5d71443da7a..a2b25e68fe5 100644
--- a/docs/ru/sql-reference/functions/tuple-map-functions.md
+++ b/docs/ru/sql-reference/functions/tuple-map-functions.md
@@ -23,7 +23,7 @@ mapAdd(Tuple(Array, Array), Tuple(Array, Array) [, ...])
 
 **Возвращаемое значение**
 
--   Возвращает один [кортеж](../../sql-reference/data-types/tuple.md#tuplet1-t2), в котором первый массив содержит отсортированные ключи, а второй - значения.
+-   Возвращает один [кортеж](../../sql-reference/data-types/tuple.md#tuplet1-t2), в котором первый массив содержит отсортированные ключи, а второй — значения.
 
 **Пример**
 

From 78c4b9f2bca3fb3fb26fecaa17b9b675a84f43c8 Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Wed, 23 Dec 2020 03:11:32 +0300
Subject: [PATCH 303/504] do not allow merge from wide to compact parts

---
 .../MergeTree/MergeTreeDataMergerMutator.cpp  |  5 ++-
 .../MergeTree/MergeTreeDataPartType.h         | 12 ++++++-
 ...01606_merge_from_wide_to_compact.reference |  9 +++++
 .../01606_merge_from_wide_to_compact.sql      | 36 +++++++++++++++++++
 4 files changed, 60 insertions(+), 2 deletions(-)
 create mode 100644 tests/queries/0_stateless/01606_merge_from_wide_to_compact.reference
 create mode 100644 tests/queries/0_stateless/01606_merge_from_wide_to_compact.sql

diff --git a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
index 2365ef141b6..3f63d34b1c6 100644
--- a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
@@ -79,13 +79,16 @@ void FutureMergedMutatedPart::assign(MergeTreeData::DataPartsVector parts_)
 
     size_t sum_rows = 0;
     size_t sum_bytes_uncompressed = 0;
+    MergeTreeDataPartType future_part_type;
     for (const auto & part : parts_)
     {
         sum_rows += part->rows_count;
         sum_bytes_uncompressed += part->getTotalColumnsSize().data_uncompressed;
+        future_part_type = std::min(future_part_type, part->getType());
     }
 
-    auto future_part_type = parts_.front()->storage.choosePartTypeOnDisk(sum_bytes_uncompressed, sum_rows);
+    auto chosen_type = parts_.front()->storage.choosePartTypeOnDisk(sum_bytes_uncompressed, sum_rows);
+    future_part_type = std::min(future_part_type, chosen_type);
     assign(std::move(parts_), future_part_type);
 }
 
diff --git a/src/Storages/MergeTree/MergeTreeDataPartType.h b/src/Storages/MergeTree/MergeTreeDataPartType.h
index 44e170141f6..6aec9c97f44 100644
--- a/src/Storages/MergeTree/MergeTreeDataPartType.h
+++ b/src/Storages/MergeTree/MergeTreeDataPartType.h
@@ -18,7 +18,7 @@ public:
         /// Data of all columns is stored in one file. Marks are also stored in single file.
         COMPACT,
 
-        /// Format with buffering data in RAM. Not implemented yet.
+        /// Format with buffering data in RAM.
         IN_MEMORY,
 
         UNKNOWN,
@@ -37,6 +37,16 @@ public:
         return !(*this == other);
     }
 
+    bool operator<(const MergeTreeDataPartType & other) const
+    {
+        return value < other.value;
+    }
+
+    bool operator>(const MergeTreeDataPartType & other) const
+    {
+        return value > other.value;
+    }
+
     void fromString(const String & str);
     String toString() const;
 
diff --git a/tests/queries/0_stateless/01606_merge_from_wide_to_compact.reference b/tests/queries/0_stateless/01606_merge_from_wide_to_compact.reference
new file mode 100644
index 00000000000..d1d987f7ad0
--- /dev/null
+++ b/tests/queries/0_stateless/01606_merge_from_wide_to_compact.reference
@@ -0,0 +1,9 @@
+all_1_1_0	Wide
+all_2_2_0	Wide
+all_3_3_0	Wide
+all_1_3_1	Wide
+300000
+all_1_3_1	Wide
+all_4_4_0	Compact
+all_1_4_2	Wide
+400000
diff --git a/tests/queries/0_stateless/01606_merge_from_wide_to_compact.sql b/tests/queries/0_stateless/01606_merge_from_wide_to_compact.sql
new file mode 100644
index 00000000000..15596d3f6db
--- /dev/null
+++ b/tests/queries/0_stateless/01606_merge_from_wide_to_compact.sql
@@ -0,0 +1,36 @@
+DROP TABLE IF EXISTS wide_to_comp;
+
+CREATE TABLE wide_to_comp (a Int, b Int, c Int)
+    ENGINE = MergeTree ORDER BY a
+    settings vertical_merge_algorithm_min_rows_to_activate = 1,
+    vertical_merge_algorithm_min_columns_to_activate = 1,
+    min_bytes_for_wide_part = 0;
+
+SYSTEM STOP merges wide_to_comp;
+
+INSERT INTO wide_to_comp SELECT number, number, number FROM numbers(100000);
+INSERT INTO wide_to_comp SELECT number, number, number FROM numbers(100000);
+INSERT INTO wide_to_comp SELECT number, number, number FROM numbers(100000);
+
+SELECT name, part_type FROM system.parts WHERE table = 'wide_to_comp' AND database = currentDatabase() AND active ORDER BY name;
+
+ALTER TABLE wide_to_comp MODIFY setting min_rows_for_wide_part = 10000000;
+SYSTEM START merges wide_to_comp;
+OPTIMIZE TABLE wide_to_comp FINAL;
+
+SELECT name, part_type FROM system.parts WHERE table = 'wide_to_comp' AND database = currentDatabase() AND active ORDER BY name;
+SELECT count() FROM wide_to_comp WHERE not ignore(*);
+
+SYSTEM STOP merges wide_to_comp;
+INSERT INTO wide_to_comp SELECT number, number, number FROM numbers(100000);
+
+SELECT name, part_type FROM system.parts WHERE table = 'wide_to_comp' AND database = currentDatabase() AND active ORDER BY name;
+
+ALTER TABLE wide_to_comp MODIFY setting min_rows_for_wide_part = 10000000;
+SYSTEM START merges wide_to_comp;
+OPTIMIZE TABLE wide_to_comp FINAL;
+
+SELECT name, part_type FROM system.parts WHERE table = 'wide_to_comp' AND database = currentDatabase() AND active ORDER BY name;
+SELECT count() FROM wide_to_comp WHERE not ignore(*);
+
+DROP TABLE wide_to_comp;

From 92b2f5ae98319722516d8088dfc37cda0d04f92a Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Wed, 23 Dec 2020 03:44:23 +0300
Subject: [PATCH 304/504] Update 00900_parquet_load.reference

---
 tests/queries/0_stateless/00900_parquet_load.reference | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/queries/0_stateless/00900_parquet_load.reference b/tests/queries/0_stateless/00900_parquet_load.reference
index 52c5fbedcc6..1c890119486 100644
--- a/tests/queries/0_stateless/00900_parquet_load.reference
+++ b/tests/queries/0_stateless/00900_parquet_load.reference
@@ -257,7 +257,7 @@ Code: 70. DB::Ex---tion: The type "list" of an input column "int64_list" is not
 Code: 70. DB::Ex---tion: The type "list" of an input column "a" is not supported for conversion from a Parquet data format: data for INSERT was parsed from stdin
 
 === Try load data from nested_maps.snappy.parquet
-Code: 70. DB::ParsingEx---tion: The type "map" of an input column "a" is not supported for conversion from a Parquet data format: data for INSERT was parsed from stdin
+Code: 70. DB::Ex---tion: The type "map" of an input column "a" is not supported for conversion from a Parquet data format: data for INSERT was parsed from stdin
 
 === Try load data from non_hadoop_lz4_compressed.parquet
 1593604800	abc	42

From d1c0f5f157e2f423845d8af983967edd6f921ba8 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Wed, 23 Dec 2020 04:24:05 +0300
Subject: [PATCH 305/504] updated finalizeAggregation

---
 .../functions/other-functions.md              | 95 ++++++++++++++++++-
 1 file changed, 94 insertions(+), 1 deletion(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 51a1f6b4cd7..daf39e1e898 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1172,7 +1172,100 @@ Result:
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Takes state of aggregate function. Returns result of aggregation (finalized state).
+Takes [state](../../sql-reference/aggregate-functions/combinators/#agg-functions-combinator-state) of aggregate function. Returns result of aggregation (finalized state).
+
+
+**Syntax** 
+
+``` sql
+finalizeAggregation(state)
+```
+
+**Parameters**
+
+-   `state` — State of aggregation. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction).
+
+**Returned value(s)**
+
+-   Value/values that was aggregated.
+
+Type: Value of any types that was aggregated. [See](../../sql-reference/data-types/aggregatefunction/) 
+
+**Examples**
+
+Query:
+
+```sql
+SELECT finalizeAggregation(( SELECT countState(number) FROM numbers(10)));
+```
+
+Result:
+
+```text
+┌─finalizeAggregation(_subquery16)─┐
+│                               10 │
+└──────────────────────────────────┘
+```
+
+Query:
+
+```sql
+SELECT finalizeAggregation(( SELECT sumState(number) FROM numbers(10)));
+```
+
+Result:
+
+```text
+┌─finalizeAggregation(_subquery20)─┐
+│                               45 │
+└──────────────────────────────────┘
+```
+
+Note that NULL values are ignored. 
+
+Query:
+
+```sql
+SELECT finalizeAggregation(arrayReduce('anyState', [NULL, 2, 3]));
+```
+
+Result:
+
+```text
+┌─finalizeAggregation(arrayReduce('anyState', [NULL, 2, 3]))─┐
+│                                                          2 │
+└────────────────────────────────────────────────────────────┘
+```
+
+Combined example:
+
+Query:
+
+```sql
+WITH initializeAggregation('sumState', number) AS one_row_sum_state
+SELECT
+    number,
+    finalizeAggregation(one_row_sum_state) AS one_row_sum,
+    runningAccumulate(one_row_sum_state) AS cumulative_sum
+FROM numbers(10)
+```
+
+Result:
+
+```text
+┌─number─┬─one_row_sum─┬─cumulative_sum─┐
+│      0 │           0 │              0 │
+│      1 │           1 │              1 │
+│      2 │           2 │              3 │
+│      3 │           3 │              6 │
+│      4 │           4 │             10 │
+│      5 │           5 │             15 │
+│      6 │           6 │             21 │
+│      7 │           7 │             28 │
+│      8 │           8 │             36 │
+│      9 │           9 │             45 │
+└────────┴─────────────┴────────────────┘
+```
 
 ## runningAccumulate {#runningaccumulate}
 

From 1e12290e2d064c6638f2aaa877dc6e1a26fc8c15 Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Wed, 23 Dec 2020 04:26:23 +0300
Subject: [PATCH 306/504] Update task_many_to_one.xml

---
 tests/integration/test_cluster_copier/task_many_to_one.xml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/integration/test_cluster_copier/task_many_to_one.xml b/tests/integration/test_cluster_copier/task_many_to_one.xml
index ec28948e7ad..3ce862d7ff7 100644
--- a/tests/integration/test_cluster_copier/task_many_to_one.xml
+++ b/tests/integration/test_cluster_copier/task_many_to_one.xml
@@ -125,4 +125,4 @@
             <sharding_key>cityHash64(first_id, second_id)</sharding_key>
         </table_first>
     </tables>
-</yandex>
\ No newline at end of file
+</yandex>

From aedb1874aff23cfe99450b01e454ff583326fed7 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Wed, 23 Dec 2020 04:40:23 +0300
Subject: [PATCH 307/504] Fix clang-tidy

---
 src/Storages/System/StorageSystemStackTrace.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/System/StorageSystemStackTrace.cpp b/src/Storages/System/StorageSystemStackTrace.cpp
index e47852c6045..0b5e82a1f3d 100644
--- a/src/Storages/System/StorageSystemStackTrace.cpp
+++ b/src/Storages/System/StorageSystemStackTrace.cpp
@@ -39,7 +39,7 @@ namespace
 
     std::atomic<int> sequence_num = 0;    /// For messages sent via pipe.
     std::atomic<int> data_ready_num = 0;
-    std::atomic<bool> signal_latch = 0;   /// Only need for thread sanitizer.
+    std::atomic<bool> signal_latch = false;   /// Only need for thread sanitizer.
 
     /** Notes:
       * Only one query from the table can be processed at the moment of time.

From 9d8948657cef2f49bdcf1498d26b8361c1ef342d Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Wed, 23 Dec 2020 04:54:49 +0300
Subject: [PATCH 308/504] Update TreeRewriter.cpp

---
 src/Interpreters/TreeRewriter.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 125f8663afe..3a9b5c3ff9e 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -631,7 +631,7 @@ void TreeRewriterResult::collectUsedColumns(const ASTPtr & query, bool is_select
             {
                 auto hints = storage->getHints(name);
                 if (!hints.empty())
-                    ss << " '" << toString(hints) << " '";
+                    ss << " '" << toString(hints) << "'";
             }
         }
         else

From 54d2fe847ff8a6070dc12eb1b0723590e7a2c3e9 Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Wed, 23 Dec 2020 05:02:54 +0300
Subject: [PATCH 309/504] Update
 src/DataStreams/ParallelParsingBlockInputStream.cpp

Co-authored-by: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
---
 src/DataStreams/ParallelParsingBlockInputStream.cpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/src/DataStreams/ParallelParsingBlockInputStream.cpp b/src/DataStreams/ParallelParsingBlockInputStream.cpp
index 49b7929d7d7..2cd6fe8a8bf 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.cpp
+++ b/src/DataStreams/ParallelParsingBlockInputStream.cpp
@@ -209,7 +209,8 @@ void ParallelParsingBlockInputStream::onBackgroundException(size_t offset)
     {
         background_exception = std::current_exception();
 
-        if (ParsingException * e = exception_cast<ParsingException *>(background_exception))
+        if (ParsingException * e = exception_cast<ParsingException *>(background_exception)
+            && e->getLineNumber() != -1)
         {
             e->setLineNumber(e->getLineNumber() + offset);
         }

From e1fc9122cc0270b20f4567d91d81022d204e2b2a Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Wed, 23 Dec 2020 05:28:53 +0300
Subject: [PATCH 310/504] Update ParallelParsingBlockInputStream.cpp

---
 src/DataStreams/ParallelParsingBlockInputStream.cpp | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/src/DataStreams/ParallelParsingBlockInputStream.cpp b/src/DataStreams/ParallelParsingBlockInputStream.cpp
index 2cd6fe8a8bf..b7a0c3cab99 100644
--- a/src/DataStreams/ParallelParsingBlockInputStream.cpp
+++ b/src/DataStreams/ParallelParsingBlockInputStream.cpp
@@ -209,11 +209,9 @@ void ParallelParsingBlockInputStream::onBackgroundException(size_t offset)
     {
         background_exception = std::current_exception();
 
-        if (ParsingException * e = exception_cast<ParsingException *>(background_exception)
-            && e->getLineNumber() != -1)
-        {
-            e->setLineNumber(e->getLineNumber() + offset);
-        }
+        if (ParsingException * e = exception_cast<ParsingException *>(background_exception))
+            if (e->getLineNumber() != -1)
+                e->setLineNumber(e->getLineNumber() + offset);
     }
     tryLogCurrentException(__PRETTY_FUNCTION__);
     finished = true;

From d9180f1e3e9612689242c478ac98adacde7da4a9 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Wed, 23 Dec 2020 05:40:55 +0300
Subject: [PATCH 311/504] Make some perf tests faster on slower machines

---
 tests/performance/constant_column_search.xml  |  2 +-
 tests/performance/countMatches.xml            |  6 +++---
 tests/performance/general_purpose_hashes.xml  |  4 ++--
 tests/performance/generate_table_function.xml |  2 +-
 tests/performance/push_down_limit.xml         |  3 ++-
 tests/performance/string_sort.xml             |  2 +-
 tests/performance/url_hits.xml                | 17 +++++++++++++----
 tests/performance/website.xml                 |  2 +-
 8 files changed, 24 insertions(+), 14 deletions(-)

diff --git a/tests/performance/constant_column_search.xml b/tests/performance/constant_column_search.xml
index 94b41fbf907..cb76fd4cefb 100644
--- a/tests/performance/constant_column_search.xml
+++ b/tests/performance/constant_column_search.xml
@@ -43,7 +43,7 @@
 
     <query><![CDATA[SELECT DISTINCT Title, multiFuzzyMatchAny(Title, 2, ['^metrika\\.ry$']) AS distance FROM hits_100m_single WHERE distance = 1]]></query>
     <query><![CDATA[SELECT DISTINCT Title, multiFuzzyMatchAny(Title, 5, ['^metrika\\.ry$']) AS distance FROM hits_100m_single WHERE distance = 1]]></query>
-    <query><![CDATA[SELECT sum(multiFuzzyMatchAny(Title, 3, ['hello$', 'world$', '^hello'])) FROM hits_100m_single]]></query>
+    <query><![CDATA[SELECT sum(multiFuzzyMatchAny(Title, 3, ['hello$', 'world$', '^hello'])) FROM (SELECT * FROM hits_100m_single LIMIT 50000000) FORMAT Null]]></query>
     <query><![CDATA[SELECT count() FROM hits_10m_single WHERE multiFuzzyMatchAny(URL, 2, ['about/address', 'for_woman', '^https?://lm-company.ruy/$', 'ultimateguitar.com'])]]></query>
 
 
diff --git a/tests/performance/countMatches.xml b/tests/performance/countMatches.xml
index 72bd2b132c6..96905ba308f 100644
--- a/tests/performance/countMatches.xml
+++ b/tests/performance/countMatches.xml
@@ -9,14 +9,14 @@
 
     <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, 'yandex')) SETTINGS max_threads=2</query>
     <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, 'yandex|google')) SETTINGS max_threads=2</query>
-    <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, '(\\w+=\\w+)')) SETTINGS max_threads=2</query>
+    <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, '(\\w+=\\w+)')) SETTINGS max_threads=4</query>
 
     <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatchesCaseInsensitive(URL, 'yandex')) SETTINGS max_threads=2</query>
     <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatchesCaseInsensitive(URL, 'yandex|google')) SETTINGS max_threads=2</query>
-    <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatchesCaseInsensitive(URL, '(\\w+=\\w+)')) SETTINGS max_threads=2</query>
+    <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatchesCaseInsensitive(URL, '(\\w+=\\w+)')) SETTINGS max_threads=4</query>
 
     <!-- Another variant of case-insensitivity -->
     <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, '(?i)yandex')) SETTINGS max_threads=2</query>
     <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, '(?i)yandex|google')) SETTINGS max_threads=2</query>
-    <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, '(?i)(\\w+=\\w+)')) SETTINGS max_threads=2</query>
+    <query>SELECT count() FROM test.hits WHERE NOT ignore(countMatches(URL, '(?i)(\\w+=\\w+)')) SETTINGS max_threads=4</query>
 </test>
diff --git a/tests/performance/general_purpose_hashes.xml b/tests/performance/general_purpose_hashes.xml
index e37e1c34496..c54e1c2fb02 100644
--- a/tests/performance/general_purpose_hashes.xml
+++ b/tests/performance/general_purpose_hashes.xml
@@ -35,14 +35,14 @@
            <name>table_slow</name>
            <values>
                <value>zeros(1000000)</value>
-               <value>zeros_mt(10000000)</value>
+               <value>zeros_mt(5000000)</value>
            </values>
         </substitution>
         <substitution>
            <name>table</name>
            <values>
                <value>numbers(100000000)</value>
-               <value>numbers_mt(1000000000)</value>
+               <value>numbers_mt(500000000)</value>
            </values>
         </substitution>
     </substitutions>
diff --git a/tests/performance/generate_table_function.xml b/tests/performance/generate_table_function.xml
index 0628e120966..bc49a7de1bd 100644
--- a/tests/performance/generate_table_function.xml
+++ b/tests/performance/generate_table_function.xml
@@ -11,7 +11,7 @@
     <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Tuple(Int32, Int64)', 0, 10, 10) LIMIT 1000000000);</query>
     <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Array(Int8)', 0, 10, 10) LIMIT 100000000);</query>
     <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Array(Nullable(Int32))', 0, 10, 10) LIMIT 100000000);</query>
-    <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Tuple(Int32, Array(Int64))', 0, 10, 10) LIMIT 1000000000);</query>
+    <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Tuple(Int32, Array(Int64))', 0, 10, 10) LIMIT 500000000);</query>
     <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Nullable(String)', 0, 10, 10) LIMIT 100000000);</query>
     <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i Array(String)', 0, 10, 10) LIMIT 100000000);</query>
     <query>SELECT sum(NOT ignore(*)) FROM (SELECT * FROM generateRandom('i UUID', 0, 10, 10) LIMIT 1000000000);</query>
diff --git a/tests/performance/push_down_limit.xml b/tests/performance/push_down_limit.xml
index 0611216410d..7e51d7d0763 100644
--- a/tests/performance/push_down_limit.xml
+++ b/tests/performance/push_down_limit.xml
@@ -1,7 +1,8 @@
 <test>
     <create_query>CREATE VIEW numbers_view AS SELECT number from numbers_mt(100000000) order by number desc</create_query>
 
-    <query>select number from (select number from numbers(1500000000) order by -number) limit 10</query>
+    <!-- This must have been an EXPLAIN test. -->
+    <query>select number from (select number from numbers(500000000) order by -number) limit 10</query>
     <query>select number from (select number from numbers_mt(1500000000) order by -number) limit 10</query>
 
     <query short="1">select number from numbers_view limit 100</query>
diff --git a/tests/performance/string_sort.xml b/tests/performance/string_sort.xml
index 82a91892af4..23ec8cf3144 100644
--- a/tests/performance/string_sort.xml
+++ b/tests/performance/string_sort.xml
@@ -35,7 +35,7 @@
     <query><![CDATA[SELECT {str1} FROM hits_100m_single ORDER BY {str1} LIMIT 2000 format Null]]></query>
     <query><![CDATA[SELECT {str1} FROM hits_100m_single ORDER BY {str1} LIMIT 5000 format Null]]></query>
     <query><![CDATA[SELECT {str1} FROM hits_100m_single ORDER BY {str1} LIMIT 10000 format Null]]></query>
-    <query><![CDATA[SELECT {str1} FROM hits_100m_single ORDER BY {str1} LIMIT 65535 format Null]]></query>
+    <query><![CDATA[SELECT {str1} FROM hits_10m_single ORDER BY {str1} LIMIT 65535 format Null]]></query>
 
     <query><![CDATA[SELECT {str1} FROM hits_100m_single ORDER BY {str1}, CounterID LIMIT 10 format Null]]></query>
     <query><![CDATA[SELECT {str1} FROM hits_100m_single ORDER BY {str1}, CounterID LIMIT 300 format Null]]></query>
diff --git a/tests/performance/url_hits.xml b/tests/performance/url_hits.xml
index 072fb5b94e7..f0ad6a786e0 100644
--- a/tests/performance/url_hits.xml
+++ b/tests/performance/url_hits.xml
@@ -1,13 +1,22 @@
 <test>
     <preconditions>
         <table_exists>hits_100m_single</table_exists>
+        <table_exists>hits_10m_single</table_exists>
         <table_exists>test.hits</table_exists>
     </preconditions>
 
 
     <substitutions>
         <substitution>
-           <name>func</name>
+           <name>func_slow</name>
+           <values>
+               <value>URLHierarchy</value>
+               <value>URLPathHierarchy</value>
+           </values>
+       </substitution>
+
+        <substitution>
+           <name>func_fast</name>
            <values>
                <value>protocol</value>
                <value>domain</value>
@@ -22,8 +31,6 @@
                <value>queryStringAndFragment</value>
                <value>extractURLParameters</value>
                <value>extractURLParameterNames</value>
-               <value>URLHierarchy</value>
-               <value>URLPathHierarchy</value>
                <value>decodeURLComponent</value>
                <value>cutWWW</value>
                <value>cutQueryString</value>
@@ -31,7 +38,9 @@
            </values>
        </substitution>
     </substitutions>
-    <query>SELECT count() FROM hits_100m_single WHERE NOT ignore({func}(URL))</query>
+
+    <query>SELECT count() FROM hits_100m_single WHERE NOT ignore({func_fast}(URL))</query>
+    <query>SELECT count() FROM hits_10m_single  WHERE NOT ignore({func_slow}(URL))</query>
 
     <!-- firstSignificantSubdomain/firstSignificantSubdomainCustom -->
     <query>SELECT count() FROM test.hits WHERE NOT ignore(firstSignificantSubdomain(URL)) SETTINGS max_threads=1</query>
diff --git a/tests/performance/website.xml b/tests/performance/website.xml
index 9e7c8cdc536..ae497295ddb 100644
--- a/tests/performance/website.xml
+++ b/tests/performance/website.xml
@@ -36,7 +36,7 @@
 <query>SELECT SearchEngineID, SearchPhrase, count() AS c FROM {table} WHERE SearchPhrase != '' GROUP BY SearchEngineID, SearchPhrase ORDER BY c DESC LIMIT 10</query>
 <query>SELECT UserID, count() FROM {table} GROUP BY UserID ORDER BY count() DESC LIMIT 10</query>
 <query>SELECT UserID, SearchPhrase, count() FROM {table} GROUP BY UserID, SearchPhrase ORDER BY count() DESC LIMIT 10</query>
-<query>SELECT UserID, SearchPhrase, count() FROM {table} GROUP BY UserID, SearchPhrase LIMIT 10</query>
+<query>SELECT UserID, SearchPhrase, count() FROM hits_10m_single GROUP BY UserID, SearchPhrase LIMIT 10</query>
 <query>SELECT UserID, toMinute(EventTime) AS m, SearchPhrase, count() FROM hits_10m_single GROUP BY UserID, m, SearchPhrase ORDER BY count() DESC LIMIT 10</query>
 <query short="1">SELECT count() FROM hits_100m_single WHERE UserID = 12345678901234567890</query>
 <query>SELECT count() FROM hits_100m_single WHERE URL LIKE '%metrika%'</query>

From 4d34ea230970c2ba05151893bf9091d5ab46c11b Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Wed, 23 Dec 2020 06:32:55 +0300
Subject: [PATCH 312/504] Attempt to debug docs

---
 docs/tools/release.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/tools/release.sh b/docs/tools/release.sh
index a65827fc073..fd58b4d8305 100755
--- a/docs/tools/release.sh
+++ b/docs/tools/release.sh
@@ -32,7 +32,7 @@ then
     git add ".nojekyll"
 
     # Push to GitHub rewriting the existing contents.
-    git commit -a -m "Add new release at $(date)"
+    git commit --quiet -m "Add new release at $(date)"
     git push --force origin master
 
     if [[ ! -z "${CLOUDFLARE_TOKEN}" ]]

From 497a93095b6b43ccfdb8a2b1b3f1fb23aa13a56b Mon Sep 17 00:00:00 2001
From: filimonov <1549571+filimonov@users.noreply.github.com>
Date: Wed, 23 Dec 2020 06:30:45 +0100
Subject: [PATCH 313/504] Update Pipe.cpp

Fixes #18240
---
 src/Processors/Pipe.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Processors/Pipe.cpp b/src/Processors/Pipe.cpp
index 12c1f628e8d..e8943790e68 100644
--- a/src/Processors/Pipe.cpp
+++ b/src/Processors/Pipe.cpp
@@ -28,7 +28,7 @@ static void checkSource(const IProcessor & source)
                         ErrorCodes::LOGICAL_ERROR);
 
     if (source.getOutputs().size() > 1)
-        throw Exception("Source for pipe should have single or two outputs, but " + source.getName() + " has " +
+        throw Exception("Source for pipe should have single output, but " + source.getName() + " has " +
                         toString(source.getOutputs().size()) + " outputs.", ErrorCodes::LOGICAL_ERROR);
 }
 

From 22a0edc5967ae842ac82eb4e7186ef6849a50dc9 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <36882414+akuzm@users.noreply.github.com>
Date: Wed, 23 Dec 2020 09:44:44 +0300
Subject: [PATCH 314/504] Update IStorage.cpp

---
 src/Storages/IStorage.cpp | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/Storages/IStorage.cpp b/src/Storages/IStorage.cpp
index a0ef033870c..5f500518516 100644
--- a/src/Storages/IStorage.cpp
+++ b/src/Storages/IStorage.cpp
@@ -214,5 +214,6 @@ std::string FilterInfo::dump() const
     }
 
     return ss.str();
+}
 
 }

From cc73cd2e20466d6d2b24d904998773e9765fd2e4 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Wed, 23 Dec 2020 11:08:57 +0300
Subject: [PATCH 315/504] added a few details

---
 docs/en/sql-reference/functions/other-functions.md | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index daf39e1e898..a6389bdc485 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1172,8 +1172,7 @@ Result:
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Takes [state](../../sql-reference/aggregate-functions/combinators/#agg-functions-combinator-state) of aggregate function. Returns result of aggregation (finalized state).
-
+Takes state of aggregate function. Returns result of aggregation (finalized state). Useful with using [-State](../../sql-reference/aggregate-functions/combinators/#agg-functions-combinator-state) combinator.
 
 **Syntax** 
 
@@ -1267,6 +1266,12 @@ Result:
 └────────┴─────────────┴────────────────┘
 ```
 
+**See Also** 
+
+[arrayReduce](../../sql-reference/functions/array-functions/#arrayreduce)
+
+[initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation)
+
 ## runningAccumulate {#runningaccumulate}
 
 Accumulates states of an aggregate function for each row of a data block.

From 35a6adf598aa9fb7b09f555801de9f7021251c41 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Wed, 23 Dec 2020 11:22:49 +0300
Subject: [PATCH 316/504] fixed links

---
 docs/en/sql-reference/functions/other-functions.md | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index a6389bdc485..40208d15667 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1172,7 +1172,7 @@ Result:
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Takes state of aggregate function. Returns result of aggregation (finalized state). Useful with using [-State](../../sql-reference/aggregate-functions/combinators/#agg-functions-combinator-state) combinator.
+Takes state of aggregate function. Returns result of aggregation (finalized state). Useful with using [-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state) combinator.
 
 **Syntax** 
 
@@ -1182,13 +1182,13 @@ finalizeAggregation(state)
 
 **Parameters**
 
--   `state` — State of aggregation. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction).
+-   `state` — State of aggregation. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction.md).
 
 **Returned value(s)**
 
 -   Value/values that was aggregated.
 
-Type: Value of any types that was aggregated. [See](../../sql-reference/data-types/aggregatefunction/) 
+Type: Value of any types that was aggregated. [See](../../sql-reference/data-types/aggregatefunction.md) 
 
 **Examples**
 
@@ -1268,9 +1268,9 @@ Result:
 
 **See Also** 
 
-[arrayReduce](../../sql-reference/functions/array-functions/#arrayreduce)
+[arrayReduce](../../sql-reference/functions/array-functions.md/#arrayreduce)
 
-[initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation)
+[initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation.md)
 
 ## runningAccumulate {#runningaccumulate}
 

From 011b159f068bae363373495c0542debbcc63480d Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Wed, 23 Dec 2020 11:52:45 +0300
Subject: [PATCH 317/504] fixed links x2

---
 docs/en/sql-reference/functions/other-functions.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 40208d15667..6a0f6c7d86f 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1182,13 +1182,13 @@ finalizeAggregation(state)
 
 **Parameters**
 
--   `state` — State of aggregation. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction.md).
+-   `state` — State of aggregation. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction).
 
 **Returned value(s)**
 
 -   Value/values that was aggregated.
 
-Type: Value of any types that was aggregated. [See](../../sql-reference/data-types/aggregatefunction.md) 
+Type: Value of any types that was aggregated. 
 
 **Examples**
 
@@ -1268,7 +1268,7 @@ Result:
 
 **See Also** 
 
-[arrayReduce](../../sql-reference/functions/array-functions.md/#arrayreduce)
+[arrayReduce](../../sql-reference/functions/array-functions.md#arrayreduce)
 
 [initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation.md)
 

From 092ca61ea9204113bd59735bd1428ff33ca18e9c Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Wed, 23 Dec 2020 12:05:59 +0300
Subject: [PATCH 318/504] fixed typos

---
 docs/en/sql-reference/functions/other-functions.md | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 6a0f6c7d86f..b89b648bd8b 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1172,12 +1172,12 @@ Result:
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Takes state of aggregate function. Returns result of aggregation (finalized state). Useful with using [-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state) combinator.
+Takes state of aggregate function. Returns result of aggregation (finalized state). Useful with [-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state) combinator.
 
 **Syntax** 
 
 ``` sql
-finalizeAggregation(state)
+finalizeAggregation(state);
 ```
 
 **Parameters**
@@ -1268,9 +1268,8 @@ Result:
 
 **See Also** 
 
-[arrayReduce](../../sql-reference/functions/array-functions.md#arrayreduce)
-
-[initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation.md)
+-   [arrayReduce](../../sql-reference/functions/array-functions.md#arrayreduce)
+-   [initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation.md)
 
 ## runningAccumulate {#runningaccumulate}
 

From 518cac61503c58e43472783e3772a42ca9585553 Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Wed, 23 Dec 2020 13:11:24 +0300
Subject: [PATCH 319/504] fix for review

---
 src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
index 3f63d34b1c6..8ddf7cde387 100644
--- a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
@@ -79,7 +79,7 @@ void FutureMergedMutatedPart::assign(MergeTreeData::DataPartsVector parts_)
 
     size_t sum_rows = 0;
     size_t sum_bytes_uncompressed = 0;
-    MergeTreeDataPartType future_part_type;
+    MergeTreeDataPartType future_part_type = MergeTreeDataPartType::UNKNOWN;
     for (const auto & part : parts_)
     {
         sum_rows += part->rows_count;

From ddfe9f61e0f360aad55beceb282a411e4737c5c3 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Wed, 23 Dec 2020 13:57:07 +0300
Subject: [PATCH 320/504] Fix website

---
 docs/tools/purge_cache_for_changed_files.py | 78 ---------------------
 docs/tools/release.sh                       |  4 +-
 2 files changed, 3 insertions(+), 79 deletions(-)
 delete mode 100644 docs/tools/purge_cache_for_changed_files.py

diff --git a/docs/tools/purge_cache_for_changed_files.py b/docs/tools/purge_cache_for_changed_files.py
deleted file mode 100644
index 6cfc9d18a57..00000000000
--- a/docs/tools/purge_cache_for_changed_files.py
+++ /dev/null
@@ -1,78 +0,0 @@
-#!/usr/bin/env python3
-
-import subprocess
-import requests
-import os
-import time
-
-FNAME_START = "+++"
-
-CLOUDFLARE_URL = "https://api.cloudflare.com/client/v4/zones/4fc6fb1d46e87851605aa7fa69ca6fe0/purge_cache"
-
-# we have changes in revision and commit sha on all pages
-# so such changes have to be ignored
-MIN_CHANGED_WORDS = 4
-
-
-def collect_changed_files():
-    proc = subprocess.Popen("git diff HEAD~1 --word-diff=porcelain | grep -e '^+[^+]\|^\-[^\-]\|^\+\+\+'", stdout=subprocess.PIPE, shell=True)
-    changed_files = []
-    current_file_name = ""
-    changed_words = []
-    while True:
-        line = proc.stdout.readline().decode("utf-8").strip()
-        if not line:
-            break
-        if FNAME_START in line:
-            if changed_words:
-                if len(changed_words) > MIN_CHANGED_WORDS:
-                    changed_files.append(current_file_name)
-                changed_words = []
-            current_file_name = line[6:]
-        else:
-            changed_words.append(line)
-    return changed_files
-
-
-def filter_and_transform_changed_files(changed_files, base_domain):
-    result = []
-    for f in changed_files:
-        if f.endswith(".html"):
-            result.append(base_domain + f.replace("index.html", ""))
-    return result
-
-
-def convert_to_dicts(changed_files, batch_size):
-    result = []
-    current_batch = {"files": []}
-    for f in changed_files:
-        if len(current_batch["files"]) >= batch_size:
-            result.append(current_batch)
-            current_batch = {"files": []}
-        current_batch["files"].append(f)
-
-    if current_batch["files"]:
-        result.append(current_batch)
-    return result
-
-
-def post_data(prepared_batches, token):
-    headers = {"Authorization": "Bearer {}".format(token)}
-    for batch in prepared_batches:
-        print(("Pugring cache for", ", ".join(batch["files"])))
-        response = requests.post(CLOUDFLARE_URL, json=batch, headers=headers)
-        response.raise_for_status()
-        time.sleep(3)
-
-
-if __name__ == "__main__":
-    token = os.getenv("CLOUDFLARE_TOKEN")
-    if not token:
-        raise Exception("Env variable CLOUDFLARE_TOKEN is empty")
-    base_domain = os.getenv("BASE_DOMAIN", "https://content.clickhouse.tech/")
-    changed_files = collect_changed_files()
-    print(("Found", len(changed_files), "changed files"))
-    filtered_files = filter_and_transform_changed_files(changed_files, base_domain)
-    print(("Files rest after filtering", len(filtered_files)))
-    prepared_batches = convert_to_dicts(filtered_files, 25)
-    post_data(prepared_batches, token)
diff --git a/docs/tools/release.sh b/docs/tools/release.sh
index fd58b4d8305..9e529b25d0f 100755
--- a/docs/tools/release.sh
+++ b/docs/tools/release.sh
@@ -38,6 +38,8 @@ then
     if [[ ! -z "${CLOUDFLARE_TOKEN}" ]]
     then
         sleep 1m
-        python3 "${BASE_DIR}/purge_cache_for_changed_files.py"
+        # https://api.cloudflare.com/#zone-purge-files-by-cache-tags,-host-or-prefix
+        POST_DATA='{"hosts":"clickhouse.tech"}'
+        curl -X POST "https://api.cloudflare.com/client/v4/zones/4fc6fb1d46e87851605aa7fa69ca6fe0/purge_cache" -H "Authorization: Bearer ${CLOUDFLARE_TOKEN}" -H "Content-Type:application/json" --data "${POST_DATA}"
     fi
 fi

From 58d0e8901ee241a59a122858210c450b1a903eb7 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 14:05:59 +0300
Subject: [PATCH 321/504] Remove hardcoded tests

---
 docker/test/coverage/run.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docker/test/coverage/run.sh b/docker/test/coverage/run.sh
index cb972dc827d..e2369a28a9a 100755
--- a/docker/test/coverage/run.sh
+++ b/docker/test/coverage/run.sh
@@ -80,7 +80,7 @@ LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "RENAM
 LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "RENAME TABLE datasets.visits_v1 TO test.visits"
 LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-client --query "SHOW TABLES FROM test"
 
-LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 00676_group_by_in 01533_optimize_skip_merged_partitions 01232_preparing_sets_race_condition 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
+LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test -j 8 --testname --shard --zookeeper --print-time --use-skip-list 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee /test_result.txt
 
 readarray -t FAILED_TESTS < <(awk '/FAIL|TIMEOUT|ERROR/ { print substr($3, 1, length($3)-1) }' "/test_result.txt")
 
@@ -97,7 +97,7 @@ then
 
     echo "Going to run again: ${FAILED_TESTS[*]}"
 
-    LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test --order=random --testname --shard --zookeeper --use-skip-list "${FAILED_TESTS[@]}" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee -a "/test_result.txt"
+    LLVM_PROFILE_FILE='client_coverage_%5m.profraw' clickhouse-test --order=random --testname --shard --zookeeper --use-skip-list "${FAILED_TESTS[@]}" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee -a /test_result.txt
 else
     echo "No failed tests"
 fi

From 54455b4740d03b70bd66a498568b71785053b64b Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 14:53:49 +0300
Subject: [PATCH 322/504] Add test for already working code

---
 src/Storages/MergeTree/IMergeTreeDataPart.cpp |  7 +++++
 .../test_default_compression_codec/test.py    | 26 +++++++++++++++++++
 2 files changed, 33 insertions(+)

diff --git a/src/Storages/MergeTree/IMergeTreeDataPart.cpp b/src/Storages/MergeTree/IMergeTreeDataPart.cpp
index 6a98529966c..104eedf060e 100644
--- a/src/Storages/MergeTree/IMergeTreeDataPart.cpp
+++ b/src/Storages/MergeTree/IMergeTreeDataPart.cpp
@@ -549,6 +549,13 @@ CompressionCodecPtr IMergeTreeDataPart::detectDefaultCompressionCodec() const
         auto column_size = getColumnSize(part_column.name, *part_column.type);
         if (column_size.data_compressed != 0 && !storage_columns.hasCompressionCodec(part_column.name))
         {
+            String path_to_data_file = getFullRelativePath() + getFileNameForColumn(part_column) + ".bin";
+            if (!volume->getDisk()->exists(path_to_data_file))
+            {
+                LOG_WARNING(storage.log, "Part's {} column {} has non zero data compressed size, but data file {} doesn't exists", name, backQuoteIfNeed(part_column.name), path_to_data_file);
+                continue;
+            }
+
             result = getCompressionCodecForFile(volume->getDisk(), getFullRelativePath() + getFileNameForColumn(part_column) + ".bin");
             break;
         }
diff --git a/tests/integration/test_default_compression_codec/test.py b/tests/integration/test_default_compression_codec/test.py
index 116454cc31f..4d111db59ab 100644
--- a/tests/integration/test_default_compression_codec/test.py
+++ b/tests/integration/test_default_compression_codec/test.py
@@ -9,6 +9,7 @@ cluster = ClickHouseCluster(__file__)
 node1 = cluster.add_instance('node1', main_configs=['configs/default_compression.xml', 'configs/wide_parts_only.xml'], with_zookeeper=True)
 node2 = cluster.add_instance('node2', main_configs=['configs/default_compression.xml', 'configs/wide_parts_only.xml'], with_zookeeper=True)
 node3 = cluster.add_instance('node3', main_configs=['configs/default_compression.xml', 'configs/wide_parts_only.xml'], image='yandex/clickhouse-server', tag='20.3.16', stay_alive=True, with_installed_binary=True)
+node4 = cluster.add_instance('node4')
 
 @pytest.fixture(scope="module")
 def start_cluster():
@@ -228,3 +229,28 @@ def test_default_codec_version_update(start_cluster):
         "SELECT default_compression_codec FROM system.parts WHERE table = 'compression_table' and name = '2_2_2_1'") == "LZ4HC(5)\n"
     assert node3.query(
         "SELECT default_compression_codec FROM system.parts WHERE table = 'compression_table' and name = '3_3_3_1'") == "LZ4\n"
+
+def test_default_codec_for_compact_parts(start_cluster):
+    node4.query("""
+    CREATE TABLE compact_parts_table (
+        key UInt64,
+        data String
+    )
+    ENGINE MergeTree ORDER BY tuple()
+    """)
+
+    node4.query("INSERT INTO compact_parts_table VALUES (1, 'Hello world')")
+    assert node4.query("SELECT COUNT() FROM compact_parts_table") == "1\n"
+
+    node4.query("ALTER TABLE compact_parts_table DETACH PART 'all_1_1_0'")
+
+    node4.exec_in_container(["bash", "-c", "rm /var/lib/clickhouse/data/default/compact_parts_table/detached/all_1_1_0/default_compression_codec.txt"])
+
+    node4.query("ALTER TABLE compact_parts_table ATTACH PART 'all_1_1_0'")
+
+    assert node4.query("SELECT COUNT() FROM compact_parts_table") == "1\n"
+
+    node4.query("DETACH TABLE compact_parts_table")
+    node4.query("ATTACH TABLE compact_parts_table")
+
+    assert node4.query("SELECT COUNT() FROM compact_parts_table") == "1\n"

From fb1c6c4eb4e82775fe440712675d434a5081278f Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 15:03:55 +0300
Subject: [PATCH 323/504] Update MergeTreeSettings.h

---
 src/Storages/MergeTree/MergeTreeSettings.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Storages/MergeTree/MergeTreeSettings.h b/src/Storages/MergeTree/MergeTreeSettings.h
index d14e1f8a91b..7038d876156 100644
--- a/src/Storages/MergeTree/MergeTreeSettings.h
+++ b/src/Storages/MergeTree/MergeTreeSettings.h
@@ -138,7 +138,7 @@ struct MergeTreeSettings : public BaseSettings<MergeTreeSettingsTraits>
     static bool isReadonlySetting(const String & name)
     {
         return name == "index_granularity" || name == "index_granularity_bytes" || name == "write_final_mark"
-            || name == "enable_mixed_granularity_parts" || name == "in_memory_parts_enable_wal";
+            || name == "enable_mixed_granularity_parts";
     }
 
     static bool isPartFormatSetting(const String & name)

From 5fc3da74d0b08678ecab363082bcdda2b4fea1ad Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 15:04:29 +0300
Subject: [PATCH 324/504] Update test.py

---
 tests/integration/test_polymorphic_parts/test.py | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/tests/integration/test_polymorphic_parts/test.py b/tests/integration/test_polymorphic_parts/test.py
index 80cb59c6789..06b38ac2e59 100644
--- a/tests/integration/test_polymorphic_parts/test.py
+++ b/tests/integration/test_polymorphic_parts/test.py
@@ -431,7 +431,17 @@ def test_in_memory_wal(start_cluster):
     node11.exec_in_container(['bash', '-c', 'test -s {}'.format(wal_file)])
     # Check file exists
     node11.exec_in_container(['bash', '-c', 'test -f {}'.format(broken_wal_file)])
+    # Data is lost without WAL
+    node11.query("ALTER TABLE wal_table MODIFY SETTING in_memory_parts_enable_wal = 0")
+    with PartitionManager() as pm:
+        pm.partition_instances(node11, node12)
 
+        insert_random_data('wal_table', node11, 50)
+        check(node11, 350, 7)
+
+        node11.restart_clickhouse(kill=True)
+        check(node11, 300, 6)
+        
 
 def test_in_memory_wal_rotate(start_cluster):
     # Write every part to single wal

From af53559b4fae52c240b7377053bef44e0f0c6bba Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 15:05:25 +0300
Subject: [PATCH 325/504] Update test.py

---
 tests/integration/test_polymorphic_parts/test.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tests/integration/test_polymorphic_parts/test.py b/tests/integration/test_polymorphic_parts/test.py
index 06b38ac2e59..9d294e5948f 100644
--- a/tests/integration/test_polymorphic_parts/test.py
+++ b/tests/integration/test_polymorphic_parts/test.py
@@ -433,6 +433,7 @@ def test_in_memory_wal(start_cluster):
     node11.exec_in_container(['bash', '-c', 'test -f {}'.format(broken_wal_file)])
     # Data is lost without WAL
     node11.query("ALTER TABLE wal_table MODIFY SETTING in_memory_parts_enable_wal = 0")
+    
     with PartitionManager() as pm:
         pm.partition_instances(node11, node12)
 
@@ -441,7 +442,7 @@ def test_in_memory_wal(start_cluster):
 
         node11.restart_clickhouse(kill=True)
         check(node11, 300, 6)
-        
+
 
 def test_in_memory_wal_rotate(start_cluster):
     # Write every part to single wal

From 28aef2ddd482976c4bcd417b06c3c98294db305c Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 15:05:55 +0300
Subject: [PATCH 326/504] Update test.py

---
 tests/integration/test_polymorphic_parts/test.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/integration/test_polymorphic_parts/test.py b/tests/integration/test_polymorphic_parts/test.py
index 9d294e5948f..83b1be104ac 100644
--- a/tests/integration/test_polymorphic_parts/test.py
+++ b/tests/integration/test_polymorphic_parts/test.py
@@ -431,9 +431,9 @@ def test_in_memory_wal(start_cluster):
     node11.exec_in_container(['bash', '-c', 'test -s {}'.format(wal_file)])
     # Check file exists
     node11.exec_in_container(['bash', '-c', 'test -f {}'.format(broken_wal_file)])
+    
     # Data is lost without WAL
     node11.query("ALTER TABLE wal_table MODIFY SETTING in_memory_parts_enable_wal = 0")
-    
     with PartitionManager() as pm:
         pm.partition_instances(node11, node12)
 

From e3d9f4739db8ecf372a9fb26088bf6dd9b793003 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 15:06:21 +0300
Subject: [PATCH 327/504] Update test.py

---
 tests/integration/test_polymorphic_parts/test.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/integration/test_polymorphic_parts/test.py b/tests/integration/test_polymorphic_parts/test.py
index 83b1be104ac..50a8192fbc5 100644
--- a/tests/integration/test_polymorphic_parts/test.py
+++ b/tests/integration/test_polymorphic_parts/test.py
@@ -431,7 +431,7 @@ def test_in_memory_wal(start_cluster):
     node11.exec_in_container(['bash', '-c', 'test -s {}'.format(wal_file)])
     # Check file exists
     node11.exec_in_container(['bash', '-c', 'test -f {}'.format(broken_wal_file)])
-    
+
     # Data is lost without WAL
     node11.query("ALTER TABLE wal_table MODIFY SETTING in_memory_parts_enable_wal = 0")
     with PartitionManager() as pm:

From 7a9282ee7bc6884bad7ce19f575f1cdc2b30b15e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 23 Dec 2020 15:31:16 +0300
Subject: [PATCH 328/504] Try fix ya.make

---
 src/DataStreams/ya.make    | 1 +
 src/DataStreams/ya.make.in | 1 +
 2 files changed, 2 insertions(+)

diff --git a/src/DataStreams/ya.make b/src/DataStreams/ya.make
index 776578af131..858bf7081e7 100644
--- a/src/DataStreams/ya.make
+++ b/src/DataStreams/ya.make
@@ -6,6 +6,7 @@ LIBRARY()
 PEERDIR(
     clickhouse/src/Common
     contrib/libs/poco/MongoDB
+    contrib/restricted/boost/libs
 )
 
 NO_COMPILER_WARNINGS()
diff --git a/src/DataStreams/ya.make.in b/src/DataStreams/ya.make.in
index d6a683daa66..1624ddb799f 100644
--- a/src/DataStreams/ya.make.in
+++ b/src/DataStreams/ya.make.in
@@ -5,6 +5,7 @@ LIBRARY()
 PEERDIR(
     clickhouse/src/Common
     contrib/libs/poco/MongoDB
+    contrib/restricted/boost/libs
 )
 
 NO_COMPILER_WARNINGS()

From b6ff4713ee5d8428fcde09b39265ce24dce8eead Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Wed, 23 Dec 2020 15:31:18 +0300
Subject: [PATCH 329/504] Build contrib in release

---
 contrib/CMakeLists.txt | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/contrib/CMakeLists.txt b/contrib/CMakeLists.txt
index 9fccee89c1d..a8cd1f07c1a 100644
--- a/contrib/CMakeLists.txt
+++ b/contrib/CMakeLists.txt
@@ -14,6 +14,10 @@ unset (_current_dir_name)
 set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -w")
 set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -w")
 
+if (WITH_COVERAGE)
+  add_compile_options(-O3 -DNDEBUG -finline)
+endif()
+
 if (SANITIZE STREQUAL "undefined")
     # 3rd-party libraries usually not intended to work with UBSan.
     add_compile_options(-fno-sanitize=undefined)

From 34b6b8407ac14bc7abc338cbb34e79013979347e Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Wed, 23 Dec 2020 15:55:51 +0300
Subject: [PATCH 330/504] Fix website

---
 docs/tools/release.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/tools/release.sh b/docs/tools/release.sh
index 9e529b25d0f..5a838474324 100755
--- a/docs/tools/release.sh
+++ b/docs/tools/release.sh
@@ -39,7 +39,7 @@ then
     then
         sleep 1m
         # https://api.cloudflare.com/#zone-purge-files-by-cache-tags,-host-or-prefix
-        POST_DATA='{"hosts":"clickhouse.tech"}'
+        POST_DATA='{"hosts":["clickhouse.tech"]}'
         curl -X POST "https://api.cloudflare.com/client/v4/zones/4fc6fb1d46e87851605aa7fa69ca6fe0/purge_cache" -H "Authorization: Bearer ${CLOUDFLARE_TOKEN}" -H "Content-Type:application/json" --data "${POST_DATA}"
     fi
 fi

From 2e0a2257b8ab26220e34c01a1970918e88903fc8 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Wed, 23 Dec 2020 18:51:06 +0800
Subject: [PATCH 331/504] -SimpleState function arg type = return type

---
 .../AggregateFunctionSimpleState.h                  | 10 +++++++++-
 .../01620_fix_simple_state_arg_type.reference       |  6 ++++++
 .../0_stateless/01620_fix_simple_state_arg_type.sql | 13 +++++++++++++
 3 files changed, 28 insertions(+), 1 deletion(-)
 create mode 100644 tests/queries/0_stateless/01620_fix_simple_state_arg_type.reference
 create mode 100644 tests/queries/0_stateless/01620_fix_simple_state_arg_type.sql

diff --git a/src/AggregateFunctions/AggregateFunctionSimpleState.h b/src/AggregateFunctions/AggregateFunctionSimpleState.h
index eefe00e77c0..3de57bc2072 100644
--- a/src/AggregateFunctions/AggregateFunctionSimpleState.h
+++ b/src/AggregateFunctions/AggregateFunctionSimpleState.h
@@ -1,5 +1,6 @@
 #pragma once
 
+#include <AggregateFunctions/AggregateFunctionFactory.h>
 #include <AggregateFunctions/IAggregateFunction.h>
 #include <DataTypes/DataTypeCustomSimpleAggregateFunction.h>
 #include <DataTypes/DataTypeFactory.h>
@@ -32,10 +33,17 @@ public:
     DataTypePtr getReturnType() const override
     {
         DataTypeCustomSimpleAggregateFunction::checkSupportedFunctions(nested_func);
+
         // Need to make a clone because it'll be customized.
         auto storage_type = DataTypeFactory::instance().get(nested_func->getReturnType()->getName());
+
+        // Need to make a new function with promoted argument types because SimpleAggregates requires arg_type = return_type.
+        AggregateFunctionProperties properties;
+        auto function
+            = AggregateFunctionFactory::instance().get(nested_func->getName(), {storage_type}, nested_func->getParameters(), properties);
+
         DataTypeCustomNamePtr custom_name
-            = std::make_unique<DataTypeCustomSimpleAggregateFunction>(nested_func, DataTypes{nested_func->getReturnType()}, params);
+            = std::make_unique<DataTypeCustomSimpleAggregateFunction>(function, DataTypes{nested_func->getReturnType()}, params);
         storage_type->setCustomization(std::make_unique<DataTypeCustomDesc>(std::move(custom_name), nullptr));
         return storage_type;
     }
diff --git a/tests/queries/0_stateless/01620_fix_simple_state_arg_type.reference b/tests/queries/0_stateless/01620_fix_simple_state_arg_type.reference
new file mode 100644
index 00000000000..8f165d1f81d
--- /dev/null
+++ b/tests/queries/0_stateless/01620_fix_simple_state_arg_type.reference
@@ -0,0 +1,6 @@
+1	10
+20	30
+40	110
+1	10
+20	60
+40	120
diff --git a/tests/queries/0_stateless/01620_fix_simple_state_arg_type.sql b/tests/queries/0_stateless/01620_fix_simple_state_arg_type.sql
new file mode 100644
index 00000000000..e66168f5e62
--- /dev/null
+++ b/tests/queries/0_stateless/01620_fix_simple_state_arg_type.sql
@@ -0,0 +1,13 @@
+drop table if exists ay;
+
+create table ay engine AggregatingMergeTree order by i as select 1 i, sumSimpleState(10) group by i;
+insert into ay values(40, 60);
+insert into ay values(40, 50);
+insert into ay values(20, 30);
+optimize table ay;
+select * from ay;
+insert into ay values(20, 30), (40, 10);
+optimize table ay;
+select * from ay;
+
+drop table if exists ay;

From 473aef519d89581fce7c4bdcad63c6159fc700ff Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 16:11:20 +0300
Subject: [PATCH 332/504] reverted 1 scale to fix 3 tests

---
 src/Functions/FunctionBinaryArithmetic.h                 | 4 ++--
 tests/queries/0_stateless/00700_decimal_arithm.reference | 3 ---
 tests/queries/0_stateless/00700_decimal_arithm.sql       | 6 +++---
 3 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 1adddb11620..439e554566c 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -862,14 +862,14 @@ class FunctionBinaryArithmetic : public IFunction
             const NativeResultType const_a = helperGetOrConvert<T0, ResultDataType>(col_left_const, left);
 
             helperInvokeEither<OpCase::LeftConstant, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
-                const_a, col_right->getData(), vec_res, 1, scale_b);
+                const_a, col_right->getData(), vec_res, scale_a, scale_b);
         }
         else if (col_left && col_right_const)
         {
             const NativeResultType const_b = helperGetOrConvert<T1, ResultDataType>(col_right_const, right);
 
             helperInvokeEither<OpCase::RightConstant, left_is_decimal, right_is_decimal, OpImpl, OpImplCheck>(
-                col_left->getData(), const_b, vec_res, scale_a, 1);
+                col_left->getData(), const_b, vec_res, scale_a, scale_b);
         }
         else
             return nullptr;
diff --git a/tests/queries/0_stateless/00700_decimal_arithm.reference b/tests/queries/0_stateless/00700_decimal_arithm.reference
index 74c83497df8..63cc190a32c 100644
--- a/tests/queries/0_stateless/00700_decimal_arithm.reference
+++ b/tests/queries/0_stateless/00700_decimal_arithm.reference
@@ -9,7 +9,6 @@
 63	21	-42	882	-882	2	0	2	0
 63	21	-42	882	-882	2	0	2	0
 63	21	-42	882	-882	2	0	2	0
-0.420000000000000021	0.419999999999999979	0.419999999999999916	8.820000000000000000	-8.820000000000000000	0.020000000000000000	0.005000000000000000
 63.420	21.420	-41.580	890.820	-890.820	2.020	0.505	2.020	0.505
 63.420000000	21.420000000	-41.580000000	890.820000000	-890.820000000	2.020000000	0.505000000	2.020000000	0.505000000
 63.420000000000000000	21.420000000000000000	-41.580000000000000000	890.820000000000000000	-890.820000000000000000	2.020000000000000000	0.505000000000000000	2.020000000000000000	0.505000000000000000
@@ -17,9 +16,7 @@
 63	-21	42	882	-882	0	2	0	2
 63	-21	42	882	-882	0	2	0	2
 63	-21	42	882	-882	0	2	0	2
-0.420000000000000021	-0.419999999999999979	-0.419999999999999916	8.820000000000000000	-8.820000000000000000	0.000000000000000000	0.000000000000000000
 63.420	-21.420	41.580	890.820	-890.820	0.495	1.980	0.495	1.980
-42.420000021	-42.419999979	-42.419999916	890.820000000	-890.820000000	0.000000000	0.000000000
 63.420000000	-21.420000000	41.580000000	890.820000000	-890.820000000
 63.420000000000000000	-21.420000000000000000	41.580000000000000000	890.820000000000000000	-890.820000000000000000	0.495049504950495049	1.980198019801980198	0.495049504950495049	1.980198019801980198
 63.42	-21.42	41.58	890.82	-890.82	0.49	1.98	0.49	1.98
diff --git a/tests/queries/0_stateless/00700_decimal_arithm.sql b/tests/queries/0_stateless/00700_decimal_arithm.sql
index 627c1583eb4..d9d960e8c9c 100644
--- a/tests/queries/0_stateless/00700_decimal_arithm.sql
+++ b/tests/queries/0_stateless/00700_decimal_arithm.sql
@@ -38,7 +38,7 @@ SELECT j + j, j - j, j * j, j / j, intDiv(j, j), intDivOrZero(j, j) FROM decimal
 SELECT a + 21, a - 21, a - 84, a * 21, a * -21, a / 21, a / 84, intDiv(a, 21), intDivOrZero(a, 84) FROM decimal WHERE a = 42;
 SELECT b + 21, b - 21, b - 84, b * 21, b * -21, b / 21, b / 84, intDiv(b, 21), intDivOrZero(b, 84) FROM decimal WHERE b = 42;
 SELECT c + 21, c - 21, c - 84, c * 21, c * -21, c / 21, c / 84, intDiv(c, 21), intDivOrZero(c, 84) FROM decimal WHERE c = 42;
-SELECT e + 21, e - 21, e - 84, e * 21, e * -21, e / 21, e / 84 FROM decimal WHERE e > 0;
+SELECT e + 21, e - 21, e - 84, e * 21, e * -21, e / 21, e / 84 FROM decimal WHERE e > 0; -- { serverError 407 }
 SELECT f + 21, f - 21, f - 84, f * 21, f * -21, f / 21, f / 84 FROM decimal WHERE f > 0; -- { serverError 407 }
 SELECT g + 21, g - 21, g - 84, g * 21, g * -21, g / 21, g / 84, intDiv(g, 21), intDivOrZero(g, 84) FROM decimal WHERE g > 0;
 SELECT h + 21, h - 21, h - 84, h * 21, h * -21, h / 21, h / 84, intDiv(h, 21), intDivOrZero(h, 84) FROM decimal WHERE h > 0;
@@ -48,10 +48,10 @@ SELECT j + 21, j - 21, j - 84, j * 21, j * -21, j / 21, j / 84, intDiv(j, 21), i
 SELECT 21 + a, 21 - a, 84 - a, 21 * a, -21 * a, 21 / a, 84 / a, intDiv(21, a), intDivOrZero(84, a) FROM decimal WHERE a = 42;
 SELECT 21 + b, 21 - b, 84 - b, 21 * b, -21 * b, 21 / b, 84 / b, intDiv(21, b), intDivOrZero(84, b) FROM decimal WHERE b = 42;
 SELECT 21 + c, 21 - c, 84 - c, 21 * c, -21 * c, 21 / c, 84 / c, intDiv(21, c), intDivOrZero(84, c) FROM decimal WHERE c = 42;
-SELECT 21 + e, 21 - e, 84 - e, 21 * e, -21 * e, 21 / e, 84 / e FROM decimal WHERE e > 0;
+SELECT 21 + e, 21 - e, 84 - e, 21 * e, -21 * e, 21 / e, 84 / e FROM decimal WHERE e > 0; -- { serverError 407 }
 SELECT 21 + f, 21 - f, 84 - f, 21 * f, -21 * f, 21 / f, 84 / f FROM decimal WHERE f > 0; -- { serverError 407 }
 SELECT 21 + g, 21 - g, 84 - g, 21 * g, -21 * g, 21 / g, 84 / g, intDiv(21, g), intDivOrZero(84, g) FROM decimal WHERE g > 0;
-SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h, 21 / h, 84 / h FROM decimal WHERE h > 0;
+SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h, 21 / h, 84 / h FROM decimal WHERE h > 0;  -- { serverError 407 }
 SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h FROM decimal WHERE h > 0;
 SELECT 21 + i, 21 - i, 84 - i, 21 * i, -21 * i, 21 / i, 84 / i, intDiv(21, i), intDivOrZero(84, i) FROM decimal WHERE i > 0;
 SELECT 21 + j, 21 - j, 84 - j, 21 * j, -21 * j, 21 / j, 84 / j, intDiv(21, j), intDivOrZero(84, j) FROM decimal WHERE j > 0;

From fec0a590829d80f6f33260c3e7263bfc5281d8bf Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 16:57:11 +0300
Subject: [PATCH 333/504] transfered scale factor calc for multiply from the
 impl

---
 src/Functions/FunctionBinaryArithmetic.h | 24 +++++++++---------------
 1 file changed, 9 insertions(+), 15 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 439e554566c..64dfb9e2691 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -311,12 +311,12 @@ public:
     static void NO_INLINE process(const A & a, const B & b, ResultContainerType & c,
         NativeResultType scale_a, NativeResultType scale_b)
     {
-        if constexpr(op_case == OpCase::LeftConstant) static_assert(!IsDecimalNumber<A>);
-        if constexpr(op_case == OpCase::RightConstant) static_assert(!IsDecimalNumber<B>);
+        if constexpr (op_case == OpCase::LeftConstant) static_assert(!IsDecimalNumber<A>);
+        if constexpr (op_case == OpCase::RightConstant) static_assert(!IsDecimalNumber<B>);
 
         size_t size;
 
-        if constexpr(op_case == OpCase::LeftConstant)
+        if constexpr (op_case == OpCase::LeftConstant)
             size = b.size();
         else
             size = a.size();
@@ -342,12 +342,10 @@ public:
                 return;
             }
         }
-        else if constexpr(is_multiply)
+        else if constexpr (is_multiply)
         {
             if (scale_a != 1)
             {
-                scale_a = 1 / scale_a; // see scale_a initialization for detail
-
                 for (size_t i = 0; i < size; ++i)
                     c[i] = applyScaled<true, false>(
                         unwrap<op_case, OpCase::LeftConstant>(a, i),
@@ -357,8 +355,6 @@ public:
             }
             else if (scale_b != 1)
             {
-                scale_b = 1 / scale_b;
-
                 for (size_t i = 0; i < size; ++i)
                     c[i] = applyScaled<false, false>(
                         unwrap<op_case, OpCase::LeftConstant>(a, i),
@@ -423,7 +419,7 @@ private:
     template <OpCase op_case, OpCase target>
     static auto unwrap(const auto& elem, size_t i)
     {
-        if constexpr(op_case == target)
+        if constexpr (op_case == target)
             return undec(elem);
         else
             return undec(elem[i]);
@@ -807,12 +803,10 @@ class FunctionBinaryArithmetic : public IFunction
                 return right.getScaleMultiplier(); // the division impl uses only the scale_a
             else if constexpr (result_is_decimal)
                 return type.scaleFactorFor(left, is_multiply);
-            else if constexpr(left_is_decimal) // needed for e.g. multiply(Decimal, Float) to scale the arguments
+            else if constexpr (left_is_decimal) // needed for e.g. multiply(Decimal, Float) to scale the arguments
                 // scale multiplier = 10^scale, so we need to get this value divided by 10^0 -> 2nd arg is 0 as
                 // the convertTo basically divides the given value on the second arg.
-                // Note we are using the scale factor instead of 1 / scale factor, it will be correctly processed in
-                // the impl.
-                return DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
+                return 1 / DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
             else
                 return 1; // the default value which won't cause any re-scale
         }();
@@ -820,8 +814,8 @@ class FunctionBinaryArithmetic : public IFunction
         const ResultType scale_b = [&] {
             if constexpr (result_is_decimal)
                 return type.scaleFactorFor(right, is_multiply || is_division);
-            else if constexpr(right_is_decimal)
-                return DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
+            else if constexpr (right_is_decimal)
+                return 1 / DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
             else
                 return 1;
         }();

From fcfe76d7038f9177a7b419042eb7b37906c59895 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:18:38 +0300
Subject: [PATCH 334/504] trying to fix different types of *

---
 src/Functions/FunctionBinaryArithmetic.h | 20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 64dfb9e2691..bbc84f91dc4 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -802,20 +802,26 @@ class FunctionBinaryArithmetic : public IFunction
             if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                 return right.getScaleMultiplier(); // the division impl uses only the scale_a
             else if constexpr (result_is_decimal)
-                return type.scaleFactorFor(left, is_multiply);
-            else if constexpr (left_is_decimal) // needed for e.g. multiply(Decimal, Float) to scale the arguments
-                // scale multiplier = 10^scale, so we need to get this value divided by 10^0 -> 2nd arg is 0 as
-                // the convertTo basically divides the given value on the second arg.
-                return 1 / DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
+                // scale_a = 1 is result is decimal
+                return is_multiply ? 1 : type.scaleFactorFor(left, false);
+            else if constexpr (left_is_decimal)
+            {
+                const ResultType scale = DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
+                // if const, the scale needs to be inverted
+                return !col_left_const ? scale : (1 / scale);
+            }
             else
                 return 1; // the default value which won't cause any re-scale
         }();
 
         const ResultType scale_b = [&] {
             if constexpr (result_is_decimal)
-                return type.scaleFactorFor(right, is_multiply || is_division);
+                return is_multiply ? 1 : type.scaleFactorFor(right, true);
             else if constexpr (right_is_decimal)
-                return 1 / DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
+            {
+                const ResultType scale = DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
+                return !col_right_const ? scale : (1 / scale);
+            }
             else
                 return 1;
         }();

From e169b39fa78f4edca0983faf705118c977e35203 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Wed, 23 Dec 2020 17:19:51 +0300
Subject: [PATCH 335/504] Fix website

---
 docs/tools/release.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/tools/release.sh b/docs/tools/release.sh
index 5a838474324..389b63ace7f 100755
--- a/docs/tools/release.sh
+++ b/docs/tools/release.sh
@@ -39,7 +39,7 @@ then
     then
         sleep 1m
         # https://api.cloudflare.com/#zone-purge-files-by-cache-tags,-host-or-prefix
-        POST_DATA='{"hosts":["clickhouse.tech"]}'
+        POST_DATA='{"hosts":["content.clickhouse.tech"]}'
         curl -X POST "https://api.cloudflare.com/client/v4/zones/4fc6fb1d46e87851605aa7fa69ca6fe0/purge_cache" -H "Authorization: Bearer ${CLOUDFLARE_TOKEN}" -H "Content-Type:application/json" --data "${POST_DATA}"
     fi
 fi

From 0ab198238a7c9f986b0600518e83eedd7d4ded3d Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:21:12 +0300
Subject: [PATCH 336/504] fix the compile error

---
 src/Functions/FunctionBinaryArithmetic.h | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index bbc84f91dc4..723f06be468 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -802,8 +802,12 @@ class FunctionBinaryArithmetic : public IFunction
             if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                 return right.getScaleMultiplier(); // the division impl uses only the scale_a
             else if constexpr (result_is_decimal)
-                // scale_a = 1 is result is decimal
-                return is_multiply ? 1 : type.scaleFactorFor(left, false);
+            {
+                if constexpr (is_multiply)
+                    return 1;
+                else
+                    return type.scaleFactorFor(left, false);
+            }
             else if constexpr (left_is_decimal)
             {
                 const ResultType scale = DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
@@ -816,7 +820,12 @@ class FunctionBinaryArithmetic : public IFunction
 
         const ResultType scale_b = [&] {
             if constexpr (result_is_decimal)
-                return is_multiply ? 1 : type.scaleFactorFor(right, true);
+            {
+                if constexpr (is_multiply)
+                    return 1;
+                else
+                    return type.scaleFactorFor(right, is_division);
+            }
             else if constexpr (right_is_decimal)
             {
                 const ResultType scale = DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);

From 2516f8408eb1bf6a4255afe7bdd18430bbcf1d1c Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:23:20 +0300
Subject: [PATCH 337/504] fix initialization

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 723f06be468..c683604bca3 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -804,7 +804,7 @@ class FunctionBinaryArithmetic : public IFunction
             else if constexpr (result_is_decimal)
             {
                 if constexpr (is_multiply)
-                    return 1;
+                    return ResultType{1};
                 else
                     return type.scaleFactorFor(left, false);
             }
@@ -822,7 +822,7 @@ class FunctionBinaryArithmetic : public IFunction
             if constexpr (result_is_decimal)
             {
                 if constexpr (is_multiply)
-                    return 1;
+                    return ResultType{1};
                 else
                     return type.scaleFactorFor(right, is_division);
             }

From 4988e051346e43ea1384e68de84db2efd43163a3 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:26:03 +0300
Subject: [PATCH 338/504] fixing the Decimal * Float aggregate case

---
 src/Functions/FunctionBinaryArithmetic.h | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index c683604bca3..f15301173d2 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -811,8 +811,7 @@ class FunctionBinaryArithmetic : public IFunction
             else if constexpr (left_is_decimal)
             {
                 const ResultType scale = DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
-                // if const, the scale needs to be inverted
-                return !col_left_const ? scale : (1 / scale);
+                return col_left_const ? scale : (1 / scale);
             }
             else
                 return 1; // the default value which won't cause any re-scale
@@ -829,7 +828,7 @@ class FunctionBinaryArithmetic : public IFunction
             else if constexpr (right_is_decimal)
             {
                 const ResultType scale = DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
-                return !col_right_const ? scale : (1 / scale);
+                return col_right_const ? scale : (1 / scale);
             }
             else
                 return 1;

From 631a2bce79c549c6235f1b8c86fdec9c64512fa7 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:29:03 +0300
Subject: [PATCH 339/504] reverted the tests to initial state

---
 tests/queries/0_stateless/00700_decimal_arithm.reference | 1 -
 tests/queries/0_stateless/00700_decimal_arithm.sql       | 4 ++--
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/tests/queries/0_stateless/00700_decimal_arithm.reference b/tests/queries/0_stateless/00700_decimal_arithm.reference
index 63cc190a32c..d0040ff5ae2 100644
--- a/tests/queries/0_stateless/00700_decimal_arithm.reference
+++ b/tests/queries/0_stateless/00700_decimal_arithm.reference
@@ -3,7 +3,6 @@
 84	0	1764	1	1	1
 84.840	0.000	1799.456400	1.000	1.000	1.000
 84.840000000	0.000000000
-84.840000000000000000	0.000000000000000000	0.000000000000000000000000000000000000	1.000000000000000000	1.000000000000000000	1.000000000000000000
 84.840000000000000000	0.000000000000000000
 84.84	0.00	1799.4564	1.00	1.00	1.00
 63	21	-42	882	-882	2	0	2	0
diff --git a/tests/queries/0_stateless/00700_decimal_arithm.sql b/tests/queries/0_stateless/00700_decimal_arithm.sql
index d9d960e8c9c..3fa2aade743 100644
--- a/tests/queries/0_stateless/00700_decimal_arithm.sql
+++ b/tests/queries/0_stateless/00700_decimal_arithm.sql
@@ -31,7 +31,7 @@ SELECT f + f, f - f, f * f, f / f, intDiv(f, f), intDivOrZero(f, f) FROM decimal
 SELECT g + g, g - g, g * g, g / g, intDiv(g, g), intDivOrZero(g, g) FROM decimal WHERE g > 0;
 SELECT h + h, h - h, h * h, h / h, intDiv(h, h), intDivOrZero(h, h) FROM decimal WHERE h > 0; -- { serverError 407 }
 SELECT h + h, h - h FROM decimal WHERE h > 0;
-SELECT i + i, i - i, i * i, i / i, intDiv(i, i), intDivOrZero(i, i) FROM decimal WHERE i > 0;
+SELECT i + i, i - i, i * i, i / i, intDiv(i, i), intDivOrZero(i, i) FROM decimal WHERE i > 0; -- { serverError 407 }
 SELECT i + i, i - i FROM decimal WHERE i > 0;
 SELECT j + j, j - j, j * j, j / j, intDiv(j, j), intDivOrZero(j, j) FROM decimal WHERE j > 0;
 
@@ -51,7 +51,7 @@ SELECT 21 + c, 21 - c, 84 - c, 21 * c, -21 * c, 21 / c, 84 / c, intDiv(21, c), i
 SELECT 21 + e, 21 - e, 84 - e, 21 * e, -21 * e, 21 / e, 84 / e FROM decimal WHERE e > 0; -- { serverError 407 }
 SELECT 21 + f, 21 - f, 84 - f, 21 * f, -21 * f, 21 / f, 84 / f FROM decimal WHERE f > 0; -- { serverError 407 }
 SELECT 21 + g, 21 - g, 84 - g, 21 * g, -21 * g, 21 / g, 84 / g, intDiv(21, g), intDivOrZero(84, g) FROM decimal WHERE g > 0;
-SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h, 21 / h, 84 / h FROM decimal WHERE h > 0;  -- { serverError 407 }
+SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h, 21 / h, 84 / h FROM decimal WHERE h > 0; -- { serverError 407 }
 SELECT 21 + h, 21 - h, 84 - h, 21 * h, -21 * h FROM decimal WHERE h > 0;
 SELECT 21 + i, 21 - i, 84 - i, 21 * i, -21 * i, 21 / i, 84 / i, intDiv(21, i), intDivOrZero(84, i) FROM decimal WHERE i > 0;
 SELECT 21 + j, 21 - j, 84 - j, 21 * j, -21 * j, 21 / j, 84 / j, intDiv(21, j), intDivOrZero(84, j) FROM decimal WHERE j > 0;

From fdc8950fbe2fd8c6499405abd5bcf4f1c9f08d22 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Wed, 23 Dec 2020 17:31:10 +0300
Subject: [PATCH 340/504] Do not use watchdog when server is run from tty

---
 programs/server/Server.cpp | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/programs/server/Server.cpp b/programs/server/Server.cpp
index f87b4294587..76765c0374c 100644
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@@ -104,7 +104,12 @@ namespace CurrentMetrics
 int mainEntryClickHouseServer(int argc, char ** argv)
 {
     DB::Server app;
-    app.shouldSetupWatchdog(argc ? argv[0] : nullptr);
+
+    /// Do not fork separate process from watchdog if we attached to terminal.
+    /// Otherwise it breaks gdb usage.
+    if (argc > 0 && !isatty(STDIN_FILENO) && !isatty(STDOUT_FILENO) && !isatty(STDERR_FILENO))
+        app.shouldSetupWatchdog(argv[0]);
+
     try
     {
         return app.run(argc, argv);

From 01b311c5c4cced0b3c399a0ecaa24b343df1411b Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:42:19 +0300
Subject: [PATCH 341/504] over-scale fix

---
 src/Functions/FunctionBinaryArithmetic.h | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index f15301173d2..d9b9fd65bfd 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -804,14 +804,21 @@ class FunctionBinaryArithmetic : public IFunction
             else if constexpr (result_is_decimal)
             {
                 if constexpr (is_multiply)
+                    // the decimal impl uses scales, but if the result is decimal, both of the arguments are decimal,
+                    // so they would multiply correctly, so we need to scale the result to the neutral element (1).
+                    // The explicit type is needed as the int (in contrast with float) can't be implicitly converted
+                    // to decimal.
                     return ResultType{1};
                 else
                     return type.scaleFactorFor(left, false);
             }
             else if constexpr (left_is_decimal)
             {
-                const ResultType scale = DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
-                return col_left_const ? scale : (1 / scale);
+                if (col_left_const)
+                    // the column will be converted to native type later, no need to scale it twice.
+                    return 1;
+
+                return 1 / DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
             }
             else
                 return 1; // the default value which won't cause any re-scale
@@ -827,8 +834,10 @@ class FunctionBinaryArithmetic : public IFunction
             }
             else if constexpr (right_is_decimal)
             {
-                const ResultType scale = DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
-                return col_right_const ? scale : (1 / scale);
+                if (col_right_const)
+                    return 1;
+
+                return 1 / DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
             }
             else
                 return 1;

From b48370d4bd95ce54670d1b1b341a04ffb7fade20 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Wed, 23 Dec 2020 17:44:43 +0300
Subject: [PATCH 342/504] fixed explicit type

---
 src/Functions/FunctionBinaryArithmetic.h | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index d9b9fd65bfd..d3639556ab6 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -816,7 +816,8 @@ class FunctionBinaryArithmetic : public IFunction
             {
                 if (col_left_const)
                     // the column will be converted to native type later, no need to scale it twice.
-                    return 1;
+                    // the explicit type is needed to specify lambda return type
+                    return ResultType{1};
 
                 return 1 / DecimalUtils::convertTo<ResultType>(left.getScaleMultiplier(), 0);
             }
@@ -835,7 +836,7 @@ class FunctionBinaryArithmetic : public IFunction
             else if constexpr (right_is_decimal)
             {
                 if (col_right_const)
-                    return 1;
+                    return ResultType{1};
 
                 return 1 / DecimalUtils::convertTo<ResultType>(right.getScaleMultiplier(), 0);
             }

From e56a1c442c7116dba86a94556b0e81dc35a7d0d0 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 23 Dec 2020 18:20:52 +0300
Subject: [PATCH 343/504] Use ucontext fir arcadia build.

---
 base/common/defines.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/base/common/defines.h b/base/common/defines.h
index 1029becb971..c9cc6feaae8 100644
--- a/base/common/defines.h
+++ b/base/common/defines.h
@@ -71,6 +71,10 @@
 #    define BOOST_USE_UCONTEXT 1
 #endif
 
+#if defined(ARCADIA_BUILD)
+#    define BOOST_USE_UCONTEXT 1
+#endif
+
 /// TODO: Strange enough, there is no way to detect UB sanitizer.
 
 /// Explicitly allow undefined behaviour for certain functions. Use it as a function attribute.

From 51481c9c5fa5235d2afcfdc98470dfc3fdb86cae Mon Sep 17 00:00:00 2001
From: Ivan <5627721+abyss7@users.noreply.github.com>
Date: Wed, 23 Dec 2020 18:25:05 +0300
Subject: [PATCH 344/504] Add ANTLR test config to another json

---
 tests/ci/ci_config.json | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/tests/ci/ci_config.json b/tests/ci/ci_config.json
index 738c8e8482b..1efcf39601e 100644
--- a/tests/ci/ci_config.json
+++ b/tests/ci/ci_config.json
@@ -225,6 +225,18 @@
                 "with_coverage": false
             }
         },
+        "Functional stateless tests (ANTLR debug)": {
+            "required_build_properties": {
+                "compiler": "clang-11",
+                "package_type": "deb",
+                "build_type": "debug",
+                "sanitizer": "none",
+                "bundled": "bundled",
+                "splitted": "unsplitted",
+                "clang-tidy": "disable",
+                "with_coverage": false
+            }
+        },
         "Functional stateful tests (release)": {
             "required_build_properties": {
                 "compiler": "clang-11",

From 6fc225e676d3eaa4983ab03d9bbc2b789d1612f7 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Thu, 24 Dec 2020 00:04:05 +0800
Subject: [PATCH 345/504] Distributed insertion to one random shard (#18294)

* Distributed insertion to one random shard

* add some tests

* add some documentation

* Respect shards' weights

* fine locking

Co-authored-by: Ivan Lezhankin <ilezhankin@yandex-team.ru>
---
 docs/en/operations/settings/settings.md       | 12 +++++
 src/Core/Settings.h                           |  4 +-
 .../DistributedBlockOutputStream.cpp          | 44 +++++++++++++------
 .../DistributedBlockOutputStream.h            |  4 +-
 src/Storages/StorageDistributed.cpp           | 31 ++++++++++++-
 src/Storages/StorageDistributed.h             | 10 +++--
 ...01615_random_one_shard_insertion.reference |  8 ++++
 .../01615_random_one_shard_insertion.sql      | 22 ++++++++++
 8 files changed, 115 insertions(+), 20 deletions(-)
 create mode 100644 tests/queries/0_stateless/01615_random_one_shard_insertion.reference
 create mode 100644 tests/queries/0_stateless/01615_random_one_shard_insertion.sql

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 86e16fe4819..7f0a61ccd42 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -1855,6 +1855,18 @@ Default value: `0`.
 -   [Distributed Table Engine](../../engines/table-engines/special/distributed.md#distributed)
 -   [Managing Distributed Tables](../../sql-reference/statements/system.md#query-language-system-distributed)
 
+## insert_distributed_one_random_shard {#insert_distributed_one_random_shard}
+
+Enables or disables random shard insertion into a [Distributed](../../engines/table-engines/special/distributed.md#distributed) table when there is no distributed key.
+
+By default, when inserting data into a `Distributed` table with more than one shard, the ClickHouse server will any insertion request if there is no distributed key. When `insert_distributed_one_random_shard = 1`, insertions are allowed and data is forwarded randomly among all shards.
+
+Possible values:
+
+-   0 — Insertion is rejected if there are multiple shards and no distributed key is given.
+-   1 — Insertion is done randomly among all available shards when no distributed key is given.
+
+Default value: `0`.
 
 ## use_compact_format_in_distributed_parts_names {#use_compact_format_in_distributed_parts_names}
 
diff --git a/src/Core/Settings.h b/src/Core/Settings.h
index b1353c3009d..b09e960da36 100644
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@@ -490,7 +490,9 @@ class IColumn;
     \
     M(Bool, output_format_enable_streaming, false, "Enable streaming in output formats that support it.", 0) \
     M(Bool, output_format_write_statistics, true, "Write statistics about read rows, bytes, time elapsed in suitable output formats.", 0) \
-    M(Bool, output_format_pretty_row_numbers, false, "Add row numbers before each row for pretty output format", 0)
+    M(Bool, output_format_pretty_row_numbers, false, "Add row numbers before each row for pretty output format", 0) \
+    M(Bool, insert_distributed_one_random_shard, false, "If setting is enabled, inserting into distributed table will choose a random shard to write when there is no sharding key", 0) \
+
 
 // End of FORMAT_FACTORY_SETTINGS
 // Please add settings non-related to formats into the COMMON_SETTINGS above.
diff --git a/src/Storages/Distributed/DistributedBlockOutputStream.cpp b/src/Storages/Distributed/DistributedBlockOutputStream.cpp
index d24967256a0..040f33ea02e 100644
--- a/src/Storages/Distributed/DistributedBlockOutputStream.cpp
+++ b/src/Storages/Distributed/DistributedBlockOutputStream.cpp
@@ -138,11 +138,22 @@ void DistributedBlockOutputStream::write(const Block & block)
 
 void DistributedBlockOutputStream::writeAsync(const Block & block)
 {
-    if (storage.getShardingKeyExpr() && (cluster->getShardsInfo().size() > 1))
-        return writeSplitAsync(block);
+    const Settings & settings = context.getSettingsRef();
+    bool random_shard_insert = settings.insert_distributed_one_random_shard && !storage.has_sharding_key;
 
-    writeAsyncImpl(block);
-    ++inserted_blocks;
+    if (random_shard_insert)
+    {
+        writeAsyncImpl(block, storage.getRandomShardIndex(cluster->getShardsInfo()));
+    }
+    else
+    {
+
+        if (storage.getShardingKeyExpr() && (cluster->getShardsInfo().size() > 1))
+            return writeSplitAsync(block);
+
+        writeAsyncImpl(block);
+        ++inserted_blocks;
+    }
 }
 
 
@@ -175,18 +186,18 @@ std::string DistributedBlockOutputStream::getCurrentStateDescription()
 }
 
 
-void DistributedBlockOutputStream::initWritingJobs(const Block & first_block)
+void DistributedBlockOutputStream::initWritingJobs(const Block & first_block, size_t start, size_t end)
 {
     const Settings & settings = context.getSettingsRef();
     const auto & addresses_with_failovers = cluster->getShardsAddresses();
     const auto & shards_info = cluster->getShardsInfo();
-    size_t num_shards = shards_info.size();
+    size_t num_shards = end - start;
 
     remote_jobs_count = 0;
     local_jobs_count = 0;
     per_shard_jobs.resize(shards_info.size());
 
-    for (size_t shard_index : ext::range(0, shards_info.size()))
+    for (size_t shard_index : ext::range(start, end))
     {
         const auto & shard_info = shards_info[shard_index];
         auto & shard_jobs = per_shard_jobs[shard_index];
@@ -242,10 +253,11 @@ void DistributedBlockOutputStream::waitForJobs()
 }
 
 
-ThreadPool::Job DistributedBlockOutputStream::runWritingJob(DistributedBlockOutputStream::JobReplica & job, const Block & current_block)
+ThreadPool::Job
+DistributedBlockOutputStream::runWritingJob(DistributedBlockOutputStream::JobReplica & job, const Block & current_block, size_t num_shards)
 {
     auto thread_group = CurrentThread::getGroup();
-    return [this, thread_group, &job, &current_block]()
+    return [this, thread_group, &job, &current_block, num_shards]()
     {
         if (thread_group)
             CurrentThread::attachToIfDetached(thread_group);
@@ -262,7 +274,6 @@ ThreadPool::Job DistributedBlockOutputStream::runWritingJob(DistributedBlockOutp
         });
 
         const auto & shard_info = cluster->getShardsInfo()[job.shard_index];
-        size_t num_shards = cluster->getShardsInfo().size();
         auto & shard_job = per_shard_jobs[job.shard_index];
         const auto & addresses = cluster->getShardsAddresses();
 
@@ -356,12 +367,19 @@ void DistributedBlockOutputStream::writeSync(const Block & block)
 {
     const Settings & settings = context.getSettingsRef();
     const auto & shards_info = cluster->getShardsInfo();
-    size_t num_shards = shards_info.size();
+    bool random_shard_insert = settings.insert_distributed_one_random_shard && !storage.has_sharding_key;
+    size_t start = 0, end = shards_info.size();
+    if (random_shard_insert)
+    {
+        start = storage.getRandomShardIndex(shards_info);
+        end = start + 1;
+    }
+    size_t num_shards = end - start;
 
     if (!pool)
     {
         /// Deferred initialization. Only for sync insertion.
-        initWritingJobs(block);
+        initWritingJobs(block, start, end);
 
         pool.emplace(remote_jobs_count + local_jobs_count);
 
@@ -394,7 +412,7 @@ void DistributedBlockOutputStream::writeSync(const Block & block)
         finished_jobs_count = 0;
         for (size_t shard_index : ext::range(0, shards_info.size()))
             for (JobReplica & job : per_shard_jobs[shard_index].replicas_jobs)
-                pool->scheduleOrThrowOnError(runWritingJob(job, block));
+                pool->scheduleOrThrowOnError(runWritingJob(job, block, num_shards));
     }
     catch (...)
     {
diff --git a/src/Storages/Distributed/DistributedBlockOutputStream.h b/src/Storages/Distributed/DistributedBlockOutputStream.h
index 872776f0867..ef37776893a 100644
--- a/src/Storages/Distributed/DistributedBlockOutputStream.h
+++ b/src/Storages/Distributed/DistributedBlockOutputStream.h
@@ -73,10 +73,10 @@ private:
     /// Performs synchronous insertion to remote nodes. If timeout_exceeded flag was set, throws.
     void writeSync(const Block & block);
 
-    void initWritingJobs(const Block & first_block);
+    void initWritingJobs(const Block & first_block, size_t start, size_t end);
 
     struct JobReplica;
-    ThreadPool::Job runWritingJob(JobReplica & job, const Block & current_block);
+    ThreadPool::Job runWritingJob(DistributedBlockOutputStream::JobReplica & job, const Block & current_block, size_t num_shards);
 
     void waitForJobs();
 
diff --git a/src/Storages/StorageDistributed.cpp b/src/Storages/StorageDistributed.cpp
index a991103d33b..4ce7efb60b4 100644
--- a/src/Storages/StorageDistributed.cpp
+++ b/src/Storages/StorageDistributed.cpp
@@ -17,6 +17,7 @@
 #include <Common/escapeForFileName.h>
 #include <Common/typeid_cast.h>
 #include <Common/quoteString.h>
+#include <Common/randomSeed.h>
 
 #include <Parsers/ASTDropQuery.h>
 #include <Parsers/ASTExpressionList.h>
@@ -373,6 +374,7 @@ StorageDistributed::StorageDistributed(
     , cluster_name(global_context.getMacros()->expand(cluster_name_))
     , has_sharding_key(sharding_key_)
     , relative_data_path(relative_data_path_)
+    , rng(randomSeed())
 {
     StorageInMemoryMetadata storage_metadata;
     storage_metadata.setColumns(columns_);
@@ -543,7 +545,8 @@ BlockOutputStreamPtr StorageDistributed::write(const ASTPtr &, const StorageMeta
     }
 
     /// If sharding key is not specified, then you can only write to a shard containing only one shard
-    if (!has_sharding_key && ((cluster->getLocalShardCount() + cluster->getRemoteShardCount()) >= 2))
+    if (!settings.insert_distributed_one_random_shard && !has_sharding_key
+        && ((cluster->getLocalShardCount() + cluster->getRemoteShardCount()) >= 2))
     {
         throw Exception("Method write is not supported by storage " + getName() + " with more than one shard and no sharding key provided",
                         ErrorCodes::STORAGE_REQUIRES_PARAMETER);
@@ -890,6 +893,32 @@ void StorageDistributed::rename(const String & new_path_to_table_data, const Sto
 }
 
 
+size_t StorageDistributed::getRandomShardIndex(const Cluster::ShardsInfo & shards)
+{
+
+    UInt32 total_weight = 0;
+    for (const auto & shard : shards)
+        total_weight += shard.weight;
+
+    assert(total_weight > 0);
+
+    size_t res;
+    {
+        std::lock_guard lock(rng_mutex);
+        res = std::uniform_int_distribution<size_t>(0, total_weight - 1)(rng);
+    }
+
+    for (auto i = 0ul, s = shards.size(); i < s; ++i)
+    {
+        if (shards[i].weight > res)
+            return i;
+        res -= shards[i].weight;
+    }
+
+    __builtin_unreachable();
+}
+
+
 void StorageDistributed::renameOnDisk(const String & new_path_to_table_data)
 {
     for (const DiskPtr & disk : data_volume->getDisks())
diff --git a/src/Storages/StorageDistributed.h b/src/Storages/StorageDistributed.h
index 58ade73b4cf..ce7e48c85a9 100644
--- a/src/Storages/StorageDistributed.h
+++ b/src/Storages/StorageDistributed.h
@@ -10,7 +10,9 @@
 #include <Parsers/ASTFunction.h>
 #include <common/logger_useful.h>
 #include <Common/ActionBlocker.h>
+#include <Interpreters/Cluster.h>
 
+#include <pcg_random.hpp>
 
 namespace DB
 {
@@ -24,9 +26,6 @@ using VolumePtr = std::shared_ptr<IVolume>;
 class ExpressionActions;
 using ExpressionActionsPtr = std::shared_ptr<ExpressionActions>;
 
-class Cluster;
-using ClusterPtr = std::shared_ptr<Cluster>;
-
 /** A distributed table that resides on multiple servers.
   * Uses data from the specified database and tables on each server.
   *
@@ -126,6 +125,8 @@ public:
 
     NamesAndTypesList getVirtuals() const override;
 
+    size_t getRandomShardIndex(const Cluster::ShardsInfo & shards);
+
     String remote_database;
     String remote_table;
     ASTPtr remote_table_function_ptr;
@@ -198,6 +199,9 @@ protected:
     std::unordered_map<std::string, ClusterNodeData> cluster_nodes_data;
     mutable std::mutex cluster_nodes_mutex;
 
+    // For random shard index generation
+    mutable std::mutex rng_mutex;
+    pcg64 rng;
 };
 
 }
diff --git a/tests/queries/0_stateless/01615_random_one_shard_insertion.reference b/tests/queries/0_stateless/01615_random_one_shard_insertion.reference
new file mode 100644
index 00000000000..448a73c4789
--- /dev/null
+++ b/tests/queries/0_stateless/01615_random_one_shard_insertion.reference
@@ -0,0 +1,8 @@
+0
+0
+1
+1
+2
+2
+3
+3
diff --git a/tests/queries/0_stateless/01615_random_one_shard_insertion.sql b/tests/queries/0_stateless/01615_random_one_shard_insertion.sql
new file mode 100644
index 00000000000..7d07629feda
--- /dev/null
+++ b/tests/queries/0_stateless/01615_random_one_shard_insertion.sql
@@ -0,0 +1,22 @@
+drop table if exists shard;
+drop table if exists distr;
+
+create table shard (id Int32) engine = MergeTree order by cityHash64(id);
+create table distr as shard engine Distributed (test_cluster_two_shards_localhost, currentDatabase(), shard);
+
+insert into distr (id) values (0), (1);  -- { serverError 55; }
+
+set insert_distributed_sync = 1;
+
+insert into distr (id) values (0), (1);  -- { serverError 55; }
+
+set insert_distributed_sync = 0;
+set insert_distributed_one_random_shard = 1;
+
+insert into distr (id) values (0), (1);
+insert into distr (id) values (2), (3);
+
+select * from distr order by id;
+
+drop table if exists shard;
+drop table if exists distr;

From 8ffadac2a329449be45ef0828ec9275422deda93 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Wed, 23 Dec 2020 19:49:36 +0300
Subject: [PATCH 346/504] Try disable ucontect for arcadia

---
 base/common/defines.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/base/common/defines.h b/base/common/defines.h
index c9cc6feaae8..39df4698b88 100644
--- a/base/common/defines.h
+++ b/base/common/defines.h
@@ -71,8 +71,8 @@
 #    define BOOST_USE_UCONTEXT 1
 #endif
 
-#if defined(ARCADIA_BUILD)
-#    define BOOST_USE_UCONTEXT 1
+#if defined(ARCADIA_BUILD) && defined(BOOST_USE_UCONTEXT)
+#    undef BOOST_USE_UCONTEXT
 #endif
 
 /// TODO: Strange enough, there is no way to detect UB sanitizer.

From 05d45fcf61f78b61d071610485de27f9f6ae33ce Mon Sep 17 00:00:00 2001
From: vdimir <vdimirc@gmail.com>
Date: Wed, 23 Dec 2020 16:58:54 +0000
Subject: [PATCH 347/504] Raise an error if more than one key is provided to
 ip_dictionary

---
 src/Dictionaries/IPAddressDictionary.cpp | 26 ++++++++----------------
 1 file changed, 9 insertions(+), 17 deletions(-)

diff --git a/src/Dictionaries/IPAddressDictionary.cpp b/src/Dictionaries/IPAddressDictionary.cpp
index 1e6ac5f7783..53eb0c32e16 100644
--- a/src/Dictionaries/IPAddressDictionary.cpp
+++ b/src/Dictionaries/IPAddressDictionary.cpp
@@ -475,16 +475,10 @@ void IPAddressDictionary::loadData()
     auto stream = source_ptr->loadAll();
     stream->readPrefix();
 
-    /// created upfront to avoid excess allocations
-    const auto keys_size = dict_struct.key->size();
-
     const auto attributes_size = attributes.size();
 
     std::vector<IPRecord> ip_records;
 
-    row_idx.reserve(keys_size);
-    mask_column.reserve(keys_size);
-
     bool has_ipv6 = false;
 
     while (const auto block = stream->read())
@@ -492,19 +486,14 @@ void IPAddressDictionary::loadData()
         const auto rows = block.rows();
         element_count += rows;
 
-        const auto key_column_ptrs = ext::map<Columns>(
-            ext::range(0, keys_size), [&](const size_t attribute_idx) { return block.safeGetByPosition(attribute_idx).column; });
-
+        const ColumnPtr key_column_ptr = block.safeGetByPosition(0).column;
         const auto attribute_column_ptrs = ext::map<Columns>(ext::range(0, attributes_size), [&](const size_t attribute_idx)
         {
-            return block.safeGetByPosition(keys_size + attribute_idx).column;
+            return block.safeGetByPosition(attribute_idx + 1).column;
         });
 
         for (const auto row : ext::range(0, rows))
         {
-            /// calculate key once per row
-            const auto key_column = key_column_ptrs.front();
-
             for (const auto attribute_idx : ext::range(0, attributes_size))
             {
                 const auto & attribute_column = *attribute_column_ptrs[attribute_idx];
@@ -513,7 +502,7 @@ void IPAddressDictionary::loadData()
                 setAttributeValue(attribute, attribute_column[row]);
             }
 
-            const auto [addr, prefix] = parseIPFromString(std::string_view(key_column->getDataAt(row)));
+            const auto [addr, prefix] = parseIPFromString(std::string_view(key_column_ptr->getDataAt(row)));
             has_ipv6 = has_ipv6 || (addr.family() == Poco::Net::IPAddress::IPv6);
 
             size_t row_number = ip_records.size();
@@ -523,6 +512,9 @@ void IPAddressDictionary::loadData()
 
     stream->readSuffix();
 
+    row_idx.reserve(ip_records.size());
+    mask_column.reserve(ip_records.size());
+
     if (has_ipv6)
     {
         auto deleted_count = sortAndUnique(ip_records,
@@ -1175,13 +1167,13 @@ void registerDictionaryTrie(DictionaryFactory & factory)
                              const std::string & config_prefix,
                              DictionarySourcePtr source_ptr) -> DictionaryPtr
     {
-        if (!dict_struct.key)
-            throw Exception{"'key' is required for dictionary of layout 'ip_trie'", ErrorCodes::BAD_ARGUMENTS};
+        if (!dict_struct.key || dict_struct.key->size() != 1)
+            throw Exception{"Dictionary of layout 'ip_trie' has to have one 'key'", ErrorCodes::BAD_ARGUMENTS};
 
         const auto dict_id = StorageID::fromDictionaryConfig(config, config_prefix);
         const DictionaryLifetime dict_lifetime{config, config_prefix + ".lifetime"};
         const bool require_nonempty = config.getBool(config_prefix + ".require_nonempty", false);
-        // This is specialised trie for storing IPv4 and IPv6 prefixes.
+        // This is specialised dictionary for storing IPv4 and IPv6 prefixes.
         return std::make_unique<IPAddressDictionary>(dict_id, dict_struct, std::move(source_ptr), dict_lifetime, require_nonempty);
     };
     factory.registerLayout("ip_trie", create_layout, true);

From 204a20fe29f1ee552b7420e79bf2a947b66c9e3f Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Wed, 23 Dec 2020 20:34:00 +0300
Subject: [PATCH 348/504] remove unused code

---
 src/Interpreters/Aggregator.cpp               | 297 ------------------
 src/Interpreters/Aggregator.h                 |   6 -
 .../Transforms/AggregatingTransform.cpp       |   1 +
 .../Transforms/AggregatingTransform.h         |   6 +
 .../Transforms/MergingAggregatedTransform.cpp |   6 +
 5 files changed, 13 insertions(+), 303 deletions(-)

diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp
index 87abca4d7cd..15fc36d2364 100644
--- a/src/Interpreters/Aggregator.cpp
+++ b/src/Interpreters/Aggregator.cpp
@@ -13,9 +13,7 @@
 #include <Columns/ColumnArray.h>
 #include <Columns/ColumnTuple.h>
 #include <Columns/ColumnLowCardinality.h>
-#include <DataStreams/IBlockInputStream.h>
 #include <DataStreams/NativeBlockOutputStream.h>
-#include <DataStreams/NullBlockInputStream.h>
 #include <DataStreams/materializeBlock.h>
 #include <IO/WriteBufferFromFile.h>
 #include <Compression/CompressedWriteBuffer.h>
@@ -1026,56 +1024,6 @@ bool Aggregator::checkLimits(size_t result_size, bool & no_more_keys) const
 }
 
 
-void Aggregator::execute(const BlockInputStreamPtr & stream, AggregatedDataVariants & result)
-{
-    if (isCancelled())
-        return;
-
-    ColumnRawPtrs key_columns(params.keys_size);
-    AggregateColumns aggregate_columns(params.aggregates_size);
-
-    /** Used if there is a limit on the maximum number of rows in the aggregation,
-      *  and if group_by_overflow_mode == ANY.
-      * In this case, new keys are not added to the set, but aggregation is performed only by
-      *  keys that have already managed to get into the set.
-      */
-    bool no_more_keys = false;
-
-    LOG_TRACE(log, "Aggregating");
-
-    Stopwatch watch;
-
-    size_t src_rows = 0;
-    size_t src_bytes = 0;
-
-    /// Read all the data
-    while (Block block = stream->read())
-    {
-        if (isCancelled())
-            return;
-
-        src_rows += block.rows();
-        src_bytes += block.bytes();
-
-        if (!executeOnBlock(block, result, key_columns, aggregate_columns, no_more_keys))
-            break;
-    }
-
-    /// If there was no data, and we aggregate without keys, and we must return single row with the result of empty aggregation.
-    /// To do this, we pass a block with zero rows to aggregate.
-    if (result.empty() && params.keys_size == 0 && !params.empty_result_for_aggregation_by_empty_set)
-        executeOnBlock(stream->getHeader(), result, key_columns, aggregate_columns, no_more_keys);
-
-    double elapsed_seconds = watch.elapsedSeconds();
-    size_t rows = result.sizeWithoutOverflowRow();
-
-    LOG_TRACE(log, "Aggregated. {} to {} rows (from {}) in {} sec. ({} rows/sec., {}/sec.)",
-        src_rows, rows, ReadableSize(src_bytes),
-        elapsed_seconds, src_rows / elapsed_seconds,
-        ReadableSize(src_bytes / elapsed_seconds));
-}
-
-
 template <typename Method, typename Table>
 void Aggregator::convertToBlockImpl(
     Method & method,
@@ -1769,206 +1717,6 @@ void NO_INLINE Aggregator::mergeBucketImpl(
     }
 }
 
-
-/** Combines aggregation states together, turns them into blocks, and outputs streams.
-  * If the aggregation states are two-level, then it produces blocks strictly in order of 'bucket_num'.
-  * (This is important for distributed processing.)
-  * In doing so, it can handle different buckets in parallel, using up to `threads` threads.
-  */
-class MergingAndConvertingBlockInputStream : public IBlockInputStream
-{
-public:
-    /** The input is a set of non-empty sets of partially aggregated data,
-      *  which are all either single-level, or are two-level.
-      */
-    MergingAndConvertingBlockInputStream(const Aggregator & aggregator_, ManyAggregatedDataVariants & data_, bool final_, size_t threads_)
-        : aggregator(aggregator_), data(data_), final(final_), threads(threads_)
-    {
-        /// At least we need one arena in first data item per thread
-        if (!data.empty() && threads > data[0]->aggregates_pools.size())
-        {
-            Arenas & first_pool = data[0]->aggregates_pools;
-            for (size_t j = first_pool.size(); j < threads; j++)
-                first_pool.emplace_back(std::make_shared<Arena>());
-        }
-    }
-
-    String getName() const override { return "MergingAndConverting"; }
-
-    Block getHeader() const override { return aggregator.getHeader(final); }
-
-    ~MergingAndConvertingBlockInputStream() override
-    {
-        LOG_TRACE(&Poco::Logger::get(__PRETTY_FUNCTION__), "Waiting for threads to finish");
-
-        /// We need to wait for threads to finish before destructor of 'parallel_merge_data',
-        ///  because the threads access 'parallel_merge_data'.
-        if (parallel_merge_data)
-            parallel_merge_data->pool.wait();
-    }
-
-protected:
-    Block readImpl() override
-    {
-        if (data.empty())
-            return {};
-
-        if (current_bucket_num >= NUM_BUCKETS)
-            return {};
-
-        AggregatedDataVariantsPtr & first = data[0];
-
-        if (current_bucket_num == -1)
-        {
-            ++current_bucket_num;
-
-            if (first->type == AggregatedDataVariants::Type::without_key || aggregator.params.overflow_row)
-            {
-                aggregator.mergeWithoutKeyDataImpl(data);
-                return aggregator.prepareBlockAndFillWithoutKey(
-                    *first, final, first->type != AggregatedDataVariants::Type::without_key);
-            }
-        }
-
-        if (!first->isTwoLevel())
-        {
-            if (current_bucket_num > 0)
-                return {};
-
-            if (first->type == AggregatedDataVariants::Type::without_key)
-                return {};
-
-            ++current_bucket_num;
-
-        #define M(NAME) \
-            else if (first->type == AggregatedDataVariants::Type::NAME) \
-                aggregator.mergeSingleLevelDataImpl<decltype(first->NAME)::element_type>(data);
-            if (false) {} // NOLINT
-            APPLY_FOR_VARIANTS_SINGLE_LEVEL(M)
-        #undef M
-            else
-                throw Exception("Unknown aggregated data variant.", ErrorCodes::UNKNOWN_AGGREGATED_DATA_VARIANT);
-
-            return aggregator.prepareBlockAndFillSingleLevel(*first, final);
-        }
-        else
-        {
-            if (!parallel_merge_data)
-            {
-                parallel_merge_data = std::make_unique<ParallelMergeData>(threads);
-                for (size_t i = 0; i < threads; ++i)
-                    scheduleThreadForNextBucket();
-            }
-
-            Block res;
-
-            while (true)
-            {
-                std::unique_lock lock(parallel_merge_data->mutex);
-
-                if (parallel_merge_data->exception)
-                    std::rethrow_exception(parallel_merge_data->exception);
-
-                auto it = parallel_merge_data->ready_blocks.find(current_bucket_num);
-                if (it != parallel_merge_data->ready_blocks.end())
-                {
-                    ++current_bucket_num;
-                    scheduleThreadForNextBucket();
-
-                    if (it->second)
-                    {
-                        res.swap(it->second);
-                        break;
-                    }
-                    else if (current_bucket_num >= NUM_BUCKETS)
-                        break;
-                }
-
-                parallel_merge_data->condvar.wait(lock);
-            }
-
-            return res;
-        }
-    }
-
-private:
-    const Aggregator & aggregator;
-    ManyAggregatedDataVariants data;
-    bool final;
-    size_t threads;
-
-    Int32 current_bucket_num = -1;
-    Int32 max_scheduled_bucket_num = -1;
-    static constexpr Int32 NUM_BUCKETS = 256;
-
-    struct ParallelMergeData
-    {
-        std::map<Int32, Block> ready_blocks;
-        std::exception_ptr exception;
-        std::mutex mutex;
-        std::condition_variable condvar;
-        ThreadPool pool;
-
-        explicit ParallelMergeData(size_t threads_) : pool(threads_) {}
-    };
-
-    std::unique_ptr<ParallelMergeData> parallel_merge_data;
-
-    void scheduleThreadForNextBucket()
-    {
-        ++max_scheduled_bucket_num;
-        if (max_scheduled_bucket_num >= NUM_BUCKETS)
-            return;
-
-        parallel_merge_data->pool.scheduleOrThrowOnError(
-            [this, max_scheduled_bucket_num = max_scheduled_bucket_num, group = CurrentThread::getGroup()]
-            { return thread(max_scheduled_bucket_num, group); });
-    }
-
-    void thread(Int32 bucket_num, ThreadGroupStatusPtr thread_group)
-    {
-        try
-        {
-            setThreadName("MergingAggregtd");
-            if (thread_group)
-                CurrentThread::attachToIfDetached(thread_group);
-            CurrentMetrics::Increment metric_increment{CurrentMetrics::QueryThread};
-
-            /// TODO: add no_more_keys support maybe
-
-            auto & merged_data = *data[0];
-            auto method = merged_data.type;
-            Block block;
-
-            /// Select Arena to avoid race conditions
-            size_t thread_number = static_cast<size_t>(bucket_num) % threads;
-            Arena * arena = merged_data.aggregates_pools.at(thread_number).get();
-
-            if (false) {} // NOLINT
-        #define M(NAME) \
-            else if (method == AggregatedDataVariants::Type::NAME) \
-            { \
-                aggregator.mergeBucketImpl<decltype(merged_data.NAME)::element_type>(data, bucket_num, arena); \
-                block = aggregator.convertOneBucketToBlock(merged_data, *merged_data.NAME, final, bucket_num); \
-            }
-
-            APPLY_FOR_VARIANTS_TWO_LEVEL(M)
-        #undef M
-
-            std::lock_guard lock(parallel_merge_data->mutex);
-            parallel_merge_data->ready_blocks[bucket_num] = std::move(block);
-        }
-        catch (...)
-        {
-            std::lock_guard lock(parallel_merge_data->mutex);
-            if (!parallel_merge_data->exception)
-                parallel_merge_data->exception = std::current_exception();
-        }
-
-        parallel_merge_data->condvar.notify_all();
-    }
-};
-
 ManyAggregatedDataVariants Aggregator::prepareVariantsToMerge(ManyAggregatedDataVariants & data_variants) const
 {
     if (data_variants.empty())
@@ -2030,18 +1778,6 @@ ManyAggregatedDataVariants Aggregator::prepareVariantsToMerge(ManyAggregatedData
     return non_empty_data;
 }
 
-std::unique_ptr<IBlockInputStream> Aggregator::mergeAndConvertToBlocks(
-    ManyAggregatedDataVariants & data_variants, bool final, size_t max_threads) const
-{
-    ManyAggregatedDataVariants non_empty_data = prepareVariantsToMerge(data_variants);
-
-    if (non_empty_data.empty())
-        return std::make_unique<NullBlockInputStream>(getHeader(final));
-
-    return std::make_unique<MergingAndConvertingBlockInputStream>(*this, non_empty_data, final, max_threads);
-}
-
-
 template <bool no_more_keys, typename Method, typename Table>
 void NO_INLINE Aggregator::mergeStreamsImplCase(
     Block & block,
@@ -2159,38 +1895,6 @@ void NO_INLINE Aggregator::mergeWithoutKeyStreamsImpl(
 }
 
 
-void Aggregator::mergeStream(const BlockInputStreamPtr & stream, AggregatedDataVariants & result, size_t max_threads)
-{
-    if (isCancelled())
-        return;
-
-    /** If the remote servers used a two-level aggregation method,
-      *  then blocks will contain information about the number of the bucket.
-      * Then the calculations can be parallelized by buckets.
-      * We decompose the blocks to the bucket numbers indicated in them.
-      */
-    BucketToBlocks bucket_to_blocks;
-
-    /// Read all the data.
-    LOG_TRACE(log, "Reading blocks of partially aggregated data.");
-
-    size_t total_input_rows = 0;
-    size_t total_input_blocks = 0;
-    while (Block block = stream->read())
-    {
-        if (isCancelled())
-            return;
-
-        total_input_rows += block.rows();
-        ++total_input_blocks;
-        bucket_to_blocks[block.info.bucket_num].emplace_back(std::move(block));
-    }
-
-    LOG_TRACE(log, "Read {} blocks of partially aggregated data, total {} rows.", total_input_blocks, total_input_rows);
-
-    mergeBlocks(bucket_to_blocks, result, max_threads);
-}
-
 void Aggregator::mergeBlocks(BucketToBlocks bucket_to_blocks, AggregatedDataVariants & result, size_t max_threads)
 {
     if (bucket_to_blocks.empty())
@@ -2429,7 +2133,6 @@ Block Aggregator::mergeBlocks(BlocksList & blocks, bool final)
     return block;
 }
 
-
 template <typename Method>
 void NO_INLINE Aggregator::convertBlockToTwoLevelImpl(
     Method & method,
diff --git a/src/Interpreters/Aggregator.h b/src/Interpreters/Aggregator.h
index c688da9d32d..a07dbbb7e59 100644
--- a/src/Interpreters/Aggregator.h
+++ b/src/Interpreters/Aggregator.h
@@ -928,9 +928,6 @@ public:
 
     Aggregator(const Params & params_);
 
-    /// Aggregate the source. Get the result in the form of one of the data structures.
-    void execute(const BlockInputStreamPtr & stream, AggregatedDataVariants & result);
-
     using AggregateColumns = std::vector<ColumnRawPtrs>;
     using AggregateColumnsData = std::vector<ColumnAggregateFunction::Container *>;
     using AggregateColumnsConstData = std::vector<const ColumnAggregateFunction::Container *>;
@@ -954,9 +951,6 @@ public:
       */
     BlocksList convertToBlocks(AggregatedDataVariants & data_variants, bool final, size_t max_threads) const;
 
-    /** Merge several aggregation data structures and output the result as a block stream.
-      */
-    std::unique_ptr<IBlockInputStream> mergeAndConvertToBlocks(ManyAggregatedDataVariants & data_variants, bool final, size_t max_threads) const;
     ManyAggregatedDataVariants prepareVariantsToMerge(ManyAggregatedDataVariants & data_variants) const;
 
     /** Merge the stream of partially aggregated blocks into one data structure.
diff --git a/src/Processors/Transforms/AggregatingTransform.cpp b/src/Processors/Transforms/AggregatingTransform.cpp
index 0a97cc3d4cb..c6907202d31 100644
--- a/src/Processors/Transforms/AggregatingTransform.cpp
+++ b/src/Processors/Transforms/AggregatingTransform.cpp
@@ -379,6 +379,7 @@ private:
 
         for (size_t thread = 0; thread < num_threads; ++thread)
         {
+            /// Select Arena to avoid race conditions
             Arena * arena = first->aggregates_pools.at(thread).get();
             auto source = std::make_shared<ConvertingAggregatedToChunksSource>(
                     params, data, shared_data, arena);
diff --git a/src/Processors/Transforms/AggregatingTransform.h b/src/Processors/Transforms/AggregatingTransform.h
index ab18992151a..86f6ecb4a36 100644
--- a/src/Processors/Transforms/AggregatingTransform.h
+++ b/src/Processors/Transforms/AggregatingTransform.h
@@ -103,6 +103,12 @@ private:
 
     ColumnRawPtrs key_columns;
     Aggregator::AggregateColumns aggregate_columns;
+
+    /** Used if there is a limit on the maximum number of rows in the aggregation,
+     *   and if group_by_overflow_mode == ANY.
+     *  In this case, new keys are not added to the set, but aggregation is performed only by
+     *   keys that have already managed to get into the set.
+     */
     bool no_more_keys = false;
 
     ManyAggregatedDataPtr many_data;
diff --git a/src/Processors/Transforms/MergingAggregatedTransform.cpp b/src/Processors/Transforms/MergingAggregatedTransform.cpp
index f386cbe4bec..1a04f85fd9c 100644
--- a/src/Processors/Transforms/MergingAggregatedTransform.cpp
+++ b/src/Processors/Transforms/MergingAggregatedTransform.cpp
@@ -34,6 +34,12 @@ void MergingAggregatedTransform::consume(Chunk chunk)
     if (!agg_info)
         throw Exception("Chunk should have AggregatedChunkInfo in MergingAggregatedTransform.", ErrorCodes::LOGICAL_ERROR);
 
+    /** If the remote servers used a two-level aggregation method,
+      *  then blocks will contain information about the number of the bucket.
+      * Then the calculations can be parallelized by buckets.
+      * We decompose the blocks to the bucket numbers indicated in them.
+      */
+
     auto block = getInputPort().getHeader().cloneWithColumns(chunk.getColumns());
     block.info.is_overflows = agg_info->is_overflows;
     block.info.bucket_num = agg_info->bucket_num;

From 857e2ef5577c3fc02aae2f3cc1a58fa28c72ec36 Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Wed, 23 Dec 2020 21:01:49 +0300
Subject: [PATCH 349/504] remove unused code

---
 src/Interpreters/Aggregator.h | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/src/Interpreters/Aggregator.h b/src/Interpreters/Aggregator.h
index 9137e13c729..331d50672ab 100644
--- a/src/Interpreters/Aggregator.h
+++ b/src/Interpreters/Aggregator.h
@@ -953,11 +953,6 @@ public:
 
     ManyAggregatedDataVariants prepareVariantsToMerge(ManyAggregatedDataVariants & data_variants) const;
 
-    /** Merge the stream of partially aggregated blocks into one data structure.
-      * (Pre-aggregate several blocks that represent the result of independent aggregations from remote servers.)
-      */
-    void mergeStream(const BlockInputStreamPtr & stream, AggregatedDataVariants & result, size_t max_threads);
-
     using BucketToBlocks = std::map<Int32, BlocksList>;
     /// Merge partially aggregated blocks separated to buckets into one data structure.
     void mergeBlocks(BucketToBlocks bucket_to_blocks, AggregatedDataVariants & result, size_t max_threads);
@@ -1006,7 +1001,6 @@ public:
 
 protected:
     friend struct AggregatedDataVariants;
-    friend class MergingAndConvertingBlockInputStream;
     friend class ConvertingAggregatedToChunksTransform;
     friend class ConvertingAggregatedToChunksSource;
     friend class AggregatingInOrderTransform;

From 08beb3ad6427d812d6e42a84fcd02a767bb00fc6 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Tue, 22 Dec 2020 22:10:37 +0300
Subject: [PATCH 350/504] d/compat: bump compatibility level to 10

Official CI builds uses ubuntu:20.04, which has the following:
- debhelper: 12.10ubuntu1
- dpkg-dev: 1.19.7ubuntu3 (should not be relevant, but let's keep it)

While debian:testing has:
- debhelper: 13.3
- dpkg: 1.20.5 (should not be relevant, but let's keep it)

In debhelper 13 compat level <= 9 has been deprecated [1], hence the
warning:

    warning: Compatibility levels before 10 are deprecated (level 8 in use)

  [1]: https://salsa.debian.org/debian/debhelper/-/commit/427594c41330917401356ebed2a951e0e9143256

And this is not only warning but also compat level implicitly will be
changed to 10. For now this is only for debian:testing but in future it
will be relevant for newer ubuntu, and hence official builds.
---
 debian/compat | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/debian/compat b/debian/compat
index 45a4fb75db8..f599e28b8ab 100644
--- a/debian/compat
+++ b/debian/compat
@@ -1 +1 @@
-8
+10

From 0325203759df131811f79564ee042a51522ffe17 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Tue, 22 Dec 2020 22:56:44 +0300
Subject: [PATCH 351/504] d/rules: disable hardening (they were not added in
 official builds anyway)

After upgrading compat level to 10 (in previous commit to avoid warning
on debian:testing, and this is not just a warning, but the compat level
implicitly changed to 10), hardening became working, since it requires
compat level >= 9 [1]

  [1]: https://wiki.debian.org/Hardening

While hardening for clickhouse builds should not be used, so just
disable them.
---
 debian/rules | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/debian/rules b/debian/rules
index 30015ba0afd..8eb47e95389 100755
--- a/debian/rules
+++ b/debian/rules
@@ -5,7 +5,7 @@
 export DH_VERBOSE=1
 
 # -pie only for static mode
-export DEB_BUILD_MAINT_OPTIONS=hardening=+all,-pie
+export DEB_BUILD_MAINT_OPTIONS=hardening=-all
 
 # because copy_headers.sh have hardcoded path to build/include_directories.txt
 BUILDDIR = obj-$(DEB_HOST_GNU_TYPE)

From a35ef5ace3cb829fd381dbc5c2f1e9179738286f Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@jakalletti-build.sas.yp-c.yandex.net>
Date: Wed, 23 Dec 2020 21:59:27 +0300
Subject: [PATCH 352/504] better

---
 tests/integration/test_cluster_copier/test.py | 32 +++++++--
 .../test_cluster_copier/test_big_cluster.py   | 20 +++---
 .../test_cluster_copier/test_trivial.py       | 66 +++++++++++--------
 3 files changed, 75 insertions(+), 43 deletions(-)

diff --git a/tests/integration/test_cluster_copier/test.py b/tests/integration/test_cluster_copier/test.py
index 00d464af98f..d87969630cd 100644
--- a/tests/integration/test_cluster_copier/test.py
+++ b/tests/integration/test_cluster_copier/test.py
@@ -94,8 +94,6 @@ class Task1:
         instance.query("INSERT INTO hits_all SELECT * FROM system.numbers LIMIT 1002",
                        settings={"insert_distributed_sync": 1})
 
-        print instance.query("SELECT DISTINCT 'all' AS partition FROM hits_all ORDER BY partition DESC")
-
     def check(self):
         assert TSV(self.cluster.instances['s0_0_0'].query("SELECT count() FROM hits_all")) == TSV("1002\n")
         assert TSV(self.cluster.instances['s1_0_0'].query("SELECT count() FROM hits_all")) == TSV("1002\n")
@@ -312,7 +310,13 @@ def execute_task(task, cmd_options):
 
 # Tests
 
-@pytest.mark.parametrize(('use_sample_offset'),[False,True])
+@pytest.mark.parametrize(
+    ('use_sample_offset'),
+    [
+        False,
+        True
+    ]
+)
 def test_copy_simple(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(Task1(started_cluster), ['--experimental-use-sample-offset', '1'])
@@ -320,7 +324,13 @@ def test_copy_simple(started_cluster, use_sample_offset):
         execute_task(Task1(started_cluster), [])
 
 
-@pytest.mark.parametrize(('use_sample_offset'),[False,True])
+@pytest.mark.parametrize(
+    ('use_sample_offset'),
+    [
+        False,
+        True
+    ]
+)
 def test_copy_with_recovering(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(Task1(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
@@ -329,7 +339,13 @@ def test_copy_with_recovering(started_cluster, use_sample_offset):
         execute_task(Task1(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
 
 
-@pytest.mark.parametrize(('use_sample_offset'),[False,True])
+@pytest.mark.parametrize(
+    ('use_sample_offset'),
+    [
+        False,
+        True
+    ]
+)
 def test_copy_with_recovering_after_move_faults(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(Task1(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
@@ -364,7 +380,11 @@ def test_no_index(started_cluster):
 def test_no_arg(started_cluster):
     execute_task(Task_no_arg(started_cluster), [])
 
-
 def test_non_partitioned_table(started_cluster):
     execute_task(Task_non_partitioned_table(started_cluster), [])
 
+if __name__ == '__main__':
+    with contextmanager(started_cluster)() as cluster:
+        for name, instance in list(cluster.instances.items()):
+            print(name, instance.ip_address)
+        input("Cluster created, press any key to destroy...")
diff --git a/tests/integration/test_cluster_copier/test_big_cluster.py b/tests/integration/test_cluster_copier/test_big_cluster.py
index 007b118703f..998af8cbb40 100644
--- a/tests/integration/test_cluster_copier/test_big_cluster.py
+++ b/tests/integration/test_cluster_copier/test_big_cluster.py
@@ -52,8 +52,8 @@ def started_cluster():
             }
         }
 
-        for cluster_name, shards in clusters_schema.iteritems():
-            for shard_name, replicas in shards.iteritems():
+        for cluster_name, shards in clusters_schema.items():
+            for shard_name, replicas in shards.items():
                 for replica_name in replicas:
                     name = "s{}_{}_{}".format(cluster_name, shard_name, replica_name)
                     cluster.add_instance(name,
@@ -100,16 +100,16 @@ def execute_task(task, cmd_options):
     task.start()
 
     zk = cluster.get_kazoo_client('zoo1')
-    print "Use ZooKeeper server: {}:{}".format(zk.hosts[0][0], zk.hosts[0][1])
+    print("Use ZooKeeper server: {}:{}".format(zk.hosts[0][0], zk.hosts[0][1]))
 
     try:
         zk.delete("/clickhouse-copier", recursive=True)
     except kazoo.exceptions.NoNodeError:
-        print "No node /clickhouse-copier. It is Ok in first test."
+        print("No node /clickhouse-copier. It is Ok in first test.")
 
     zk_task_path = task.zk_task_path
     zk.ensure_path(zk_task_path)
-    zk.create(zk_task_path + "/description", task.copier_task_config)
+    zk.create(zk_task_path + "/description", task.copier_task_config.encode())
 
     # Run cluster-copier processes on each node
     docker_api = docker.from_env().api
@@ -120,22 +120,20 @@ def execute_task(task, cmd_options):
            '--task-path', zk_task_path,
            '--base-dir', '/var/log/clickhouse-server/copier']
     cmd += cmd_options
-    
-    # copiers = []
-    copiers = random.sample(cluster.instances.keys(), 3)
-    copiers.append('s1_0_0')
+
+    copiers = random.sample(list(cluster.instances.keys()), 3)
 
     for instance_name in copiers:
         instance = cluster.instances[instance_name]
         container = instance.get_docker_handle()
         instance.copy_file_to_container(os.path.join(CURRENT_TEST_DIR, "configs/config-copier.xml"),
                                         "/etc/clickhouse-server/config-copier.xml")
-        print "Copied copier config to {}".format(instance.name)
+        print("Copied copier config to {}".format(instance.name))
         exec_id = docker_api.exec_create(container.id, cmd, stderr=True)
         output = docker_api.exec_start(exec_id).decode('utf8')
         print(output)
         copiers_exec_ids.append(exec_id)
-        print "Copier for {} ({}) has started".format(instance.name, instance.ip_address)
+        print("Copier for {} ({}) has started".format(instance.name, instance.ip_address))
 
     # Wait for copiers stopping and check their return codes
     for exec_id, instance_name in zip(copiers_exec_ids, copiers):
diff --git a/tests/integration/test_cluster_copier/test_trivial.py b/tests/integration/test_cluster_copier/test_trivial.py
index 48d5aaddb21..ba6cefcec38 100644
--- a/tests/integration/test_cluster_copier/test_trivial.py
+++ b/tests/integration/test_cluster_copier/test_trivial.py
@@ -11,8 +11,8 @@ sys.path.insert(0, os.path.dirname(CURRENT_TEST_DIR))
 from helpers.cluster import ClickHouseCluster
 from helpers.test_tools import TSV
 
-COPYING_FAIL_PROBABILITY = 0.2
-MOVING_FAIL_PROBABILITY = 0.2
+COPYING_FAIL_PROBABILITY = 0.33
+MOVING_FAIL_PROBABILITY = 0.1
 cluster = None
 
 
@@ -85,11 +85,6 @@ def execute_task(task, cmd_options):
     zk = cluster.get_kazoo_client('zoo1')
     print("Use ZooKeeper server: {}:{}".format(zk.hosts[0][0], zk.hosts[0][1]))
 
-    try:
-        zk.delete("/clickhouse-copier", recursive=True)
-    except kazoo.exceptions.NoNodeError:
-        print "No node /clickhouse-copier. It is Ok in first test."
-
     zk_task_path = task.zk_task_path
     zk.ensure_path(zk_task_path)
     zk.create(zk_task_path + "/description", task.copier_task_config)
@@ -104,28 +99,23 @@ def execute_task(task, cmd_options):
            '--base-dir', '/var/log/clickhouse-server/copier']
     cmd += cmd_options
 
-    copiers = cluster.instances.keys()
+    print(cmd)
 
-    for instance_name in copiers:
-        instance = cluster.instances[instance_name]
+    for instance_name, instance in cluster.instances.items():
         container = instance.get_docker_handle()
-        instance.copy_file_to_container(os.path.join(CURRENT_TEST_DIR, "configs/config-copier.xml"),
-                                        "/etc/clickhouse-server/config-copier.xml")
-        print "Copied copier config to {}".format(instance.name)
         exec_id = docker_api.exec_create(container.id, cmd, stderr=True)
-        output = docker_api.exec_start(exec_id).decode('utf8')
-        print(output)
+        docker_api.exec_start(exec_id, detach=True)
+
         copiers_exec_ids.append(exec_id)
         print("Copier for {} ({}) has started".format(instance.name, instance.ip_address))
 
     # Wait for copiers stopping and check their return codes
-    for exec_id, instance_name in zip(copiers_exec_ids, copiers):
-        instance = cluster.instances[instance_name]
+    for exec_id, instance in zip(copiers_exec_ids, iter(cluster.instances.values())):
         while True:
             res = docker_api.exec_inspect(exec_id)
             if not res['Running']:
                 break
-            time.sleep(0.5)
+            time.sleep(1)
 
         assert res['ExitCode'] == 0, "Instance: {} ({}). Info: {}".format(instance.name, instance.ip_address, repr(res))
 
@@ -135,32 +125,56 @@ def execute_task(task, cmd_options):
         zk.delete(zk_task_path, recursive=True)
 
 
-
 # Tests
 
 
-@pytest.mark.parametrize(('use_sample_offset'),[False,True])
+@pytest.mark.parametrize(
+    ('use_sample_offset'),
+    [
+        False,
+        True
+    ]
+)
 def test_trivial_copy(started_cluster, use_sample_offset):
     if use_sample_offset:
         execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--experimental-use-sample-offset', '1'])
     else:
+        print("AAAAA")
         execute_task(TaskTrivial(started_cluster, use_sample_offset), [])
 
 
-@pytest.mark.parametrize(('use_sample_offset'),[False,True])
+@pytest.mark.parametrize(
+    ('use_sample_offset'),
+    [
+        False,
+        True
+    ]
+)
 def test_trivial_copy_with_copy_fault(started_cluster, use_sample_offset):
     if use_sample_offset:
-        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
+        execute_task(TaskTrivial(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
                                                     '--experimental-use-sample-offset', '1'])
     else:
-        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
+        execute_task(TaskTrivial(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
 
 
-@pytest.mark.parametrize(('use_sample_offset'),[False,True])
+@pytest.mark.parametrize(
+    ('use_sample_offset'),
+    [
+        False,
+        True
+    ]
+)
 def test_trivial_copy_with_move_fault(started_cluster, use_sample_offset):
     if use_sample_offset:
-        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
+        execute_task(TaskTrivial(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
                                                     '--experimental-use-sample-offset', '1'])
     else:
-        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY)])
+        execute_task(TaskTrivial(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY)])
 
+
+if __name__ == '__main__':
+    with contextmanager(started_cluster)() as cluster:
+        for name, instance in list(cluster.instances.items()):
+            print(name, instance.ip_address)
+        input("Cluster created, press any key to destroy...")
\ No newline at end of file

From 7ac7d991adcacb8957a0e06a3cfef282b65dab38 Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@jakalletti-build.sas.yp-c.yandex.net>
Date: Wed, 23 Dec 2020 22:02:53 +0300
Subject: [PATCH 353/504] replace word anime with word big

---
 .../configs/conf.d/big_clusters.xml              |  8 ++++----
 .../test_cluster_copier/task_many_to_one.xml     | 16 ++++++++--------
 .../test_cluster_copier/test_big_cluster.py      | 12 ++++++------
 3 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml b/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml
index cba995af211..da29534cf33 100644
--- a/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml
+++ b/tests/integration/test_cluster_copier/configs/conf.d/big_clusters.xml
@@ -1,7 +1,7 @@
 <?xml version="1.0"?>
 <yandex>
     <remote_servers>
-        <anime_source>
+        <big_source>
             <shard>
                 <internal_replication>true</internal_replication>    
                 <replica>
@@ -57,15 +57,15 @@
                     <port>9000</port>
                 </replica>
             </shard>
-        </anime_source>
+        </big_source>
 
-        <anime_destination>
+        <big_destination>
             <shard> 
                 <replica>
                     <host>s1_0_0</host>
                     <port>9000</port>
                 </replica>
             </shard>
-        </anime_destination>
+        </big_destination>
     </remote_servers>
 </yandex>
\ No newline at end of file
diff --git a/tests/integration/test_cluster_copier/task_many_to_one.xml b/tests/integration/test_cluster_copier/task_many_to_one.xml
index 3ce862d7ff7..038ef5df3ae 100644
--- a/tests/integration/test_cluster_copier/task_many_to_one.xml
+++ b/tests/integration/test_cluster_copier/task_many_to_one.xml
@@ -1,7 +1,7 @@
 <yandex>
     <!-- Configuration of clusters as in an ordinary server config -->
     <remote_servers>
-        <anime_source>
+        <big_source>
             <shard>
                 <internal_replication>true</internal_replication>    
                 <replica>
@@ -57,16 +57,16 @@
                     <port>9000</port>
                 </replica>
             </shard>
-        </anime_source>
+        </big_source>
 
-        <anime_destination>
+        <big_destination>
             <shard> 
                 <replica>
                     <host>s1_0_0</host>
                     <port>9000</port>
                 </replica>
             </shard>
-        </anime_destination>
+        </big_destination>
     </remote_servers>
 
     <!-- How many simultaneously active workers are possible. If you run more workers superfluous workers will sleep. -->
@@ -98,14 +98,14 @@
         <!-- A table task, copies one table. -->
         <table_first>
             <!-- Source cluster name (from <remote_servers/> section) and tables in it that should be copied -->
-            <cluster_pull>anime_source</cluster_pull>
+            <cluster_pull>big_source</cluster_pull>
             <database_pull>default</database_pull>
-            <table_pull>anime_all</table_pull>
+            <table_pull>big_all</table_pull>
 
             <!-- Destination cluster name and tables in which the data should be inserted -->
-            <cluster_push>anime_destination</cluster_push>
+            <cluster_push>big_destination</cluster_push>
             <database_push>default</database_push>
-            <table_push>anime</table_push>
+            <table_push>big</table_push>
 			
 			<!-- Engine of destination tables.
                  If destination tables have not be created, workers create them using columns definition from source tables and engine
diff --git a/tests/integration/test_cluster_copier/test_big_cluster.py b/tests/integration/test_cluster_copier/test_big_cluster.py
index 998af8cbb40..2424f0149e4 100644
--- a/tests/integration/test_cluster_copier/test_big_cluster.py
+++ b/tests/integration/test_cluster_copier/test_big_cluster.py
@@ -80,21 +80,21 @@ class Task_many_to_one:
     def start(self):
         instance = cluster.instances['s0_0_0']
 
-        for cluster_num in ["anime_source", "anime_destination"]:
+        for cluster_num in ["big_source", "big_destination"]:
             instance.query("DROP DATABASE IF EXISTS default ON CLUSTER {}".format(cluster_num))
             instance.query("CREATE DATABASE IF NOT EXISTS default ON CLUSTER {}".format(cluster_num))
 
-        instance.query("CREATE TABLE anime ON CLUSTER anime_source (first_id UUID DEFAULT generateUUIDv4(), second_id UInt64, datetime DateTime DEFAULT now()) " +
+        instance.query("CREATE TABLE big ON CLUSTER big_source (first_id UUID DEFAULT generateUUIDv4(), second_id UInt64, datetime DateTime DEFAULT now()) " +
                        "ENGINE=ReplicatedMergeTree " +
                        "PARTITION BY toSecond(datetime) " +  
                        "ORDER BY (first_id, second_id, toSecond(datetime))")
-        instance.query("CREATE TABLE anime_all ON CLUSTER anime_source (first_id UUID, second_id UInt64, datetime DateTime) ENGINE=Distributed(anime_source, default, anime, rand() % 5)")
-        instance.query("INSERT INTO anime_all SELECT generateUUIDv4(), number, now() FROM system.numbers LIMIT 1002",
+        instance.query("CREATE TABLE big_all ON CLUSTER big_source (first_id UUID, second_id UInt64, datetime DateTime) ENGINE=Distributed(big_source, default, big, rand() % 5)")
+        instance.query("INSERT INTO big_all SELECT generateUUIDv4(), number, now() FROM system.numbers LIMIT 1002",
                        settings={"insert_distributed_sync": 1})
 
     def check(self):
-        assert TSV(self.cluster.instances['s0_0_0'].query("SELECT count() FROM anime_all")) == TSV("1002\n")
-        assert TSV(self.cluster.instances['s1_0_0'].query("SELECT count() FROM anime")) == TSV("1002\n")
+        assert TSV(self.cluster.instances['s0_0_0'].query("SELECT count() FROM big_all")) == TSV("1002\n")
+        assert TSV(self.cluster.instances['s1_0_0'].query("SELECT count() FROM big")) == TSV("1002\n")
 
 def execute_task(task, cmd_options):
     task.start()

From d97455a94bc98345388b5c51acbe9635ac117ea6 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Wed, 23 Dec 2020 22:07:59 +0300
Subject: [PATCH 354/504] Ignore SOURCE_DATE_EPOCH for newer ccache (4+)

ccache 4+ starts respecting SOURCE_DATE_EPOCH [1], it will include it
into the hash of the manifest.

SOURCE_DATE_EPOCH will be passed from debhelpers, by extracting last
entry from d/changelog (or current time if there is entries)

And this will make ccache unusable.

Fix this by ignoring SOURCE_DATE_EPOCH, like current ccache version on
CI does (3.7.7)

  [1]: https://github.com/ccache/ccache/blob/238553a1c470e4cadc08c720a1ab915ac8d41e13/src/ccache.cpp#L1279
---
 cmake/find/ccache.cmake | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/cmake/find/ccache.cmake b/cmake/find/ccache.cmake
index 8e9fe4d84ce..2cdfed56de8 100644
--- a/cmake/find/ccache.cmake
+++ b/cmake/find/ccache.cmake
@@ -31,8 +31,20 @@ if (CCACHE_FOUND AND NOT COMPILER_MATCHES_CCACHE)
 
    if (CCACHE_VERSION VERSION_GREATER "3.2.0" OR NOT CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
       message(STATUS "Using ${CCACHE_FOUND} ${CCACHE_VERSION}")
-      set_property (GLOBAL PROPERTY RULE_LAUNCH_COMPILE ${CCACHE_FOUND})
-      set_property (GLOBAL PROPERTY RULE_LAUNCH_LINK ${CCACHE_FOUND})
+
+      # 4+ ccache respect SOURCE_DATE_EPOCH (always includes it into the hash
+      # of the manifest) and debian will extract these from d/changelog, and
+      # makes cache of ccache unusable
+      #
+      # FIXME: once sloppiness will be introduced for this this can be removed.
+      if (CCACHE_VERSION VERSION_GREATER "4.0")
+         message(STATUS "Ignore SOURCE_DATE_EPOCH for ccache")
+         set_property (GLOBAL PROPERTY RULE_LAUNCH_COMPILE "env -u SOURCE_DATE_EPOCH ${CCACHE_FOUND}")
+         set_property (GLOBAL PROPERTY RULE_LAUNCH_LINK "env -u SOURCE_DATE_EPOCH ${CCACHE_FOUND}")
+      else()
+         set_property (GLOBAL PROPERTY RULE_LAUNCH_COMPILE ${CCACHE_FOUND})
+         set_property (GLOBAL PROPERTY RULE_LAUNCH_LINK ${CCACHE_FOUND})
+      endif()
    else ()
       message(${RECONFIGURE_MESSAGE_LEVEL} "Not using ${CCACHE_FOUND} ${CCACHE_VERSION} bug: https://bugzilla.samba.org/show_bug.cgi?id=8118")
    endif ()

From 92bc52d76586147a1a6122d123bca0966f956467 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Wed, 23 Dec 2020 23:08:48 +0300
Subject: [PATCH 355/504] Fix Poco::Crypto library (missing linking with
 crypto)

---
 contrib/poco-cmake/Crypto/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/poco-cmake/Crypto/CMakeLists.txt b/contrib/poco-cmake/Crypto/CMakeLists.txt
index 468f1641510..1685e96728b 100644
--- a/contrib/poco-cmake/Crypto/CMakeLists.txt
+++ b/contrib/poco-cmake/Crypto/CMakeLists.txt
@@ -30,7 +30,7 @@ if (ENABLE_SSL)
 
         target_compile_options (_poco_crypto PRIVATE -Wno-newline-eof)
         target_include_directories (_poco_crypto SYSTEM PUBLIC ${LIBRARY_DIR}/Crypto/include)
-        target_link_libraries (_poco_crypto PUBLIC Poco::Foundation ssl)
+        target_link_libraries (_poco_crypto PUBLIC Poco::Foundation ssl crypto)
     else ()
         add_library (Poco::Crypto UNKNOWN IMPORTED GLOBAL)
 

From 7c76bd2def95cbe92e917575a51f51dff2bfc5af Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Wed, 23 Dec 2020 23:43:21 +0300
Subject: [PATCH 356/504] Fix ParsingException::displayText override (guard
 with POCO_CLICKHOUSE_PATCH)

---
 src/Common/Exception.h | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index 4b04de5d8a2..c26f3f41ab4 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -4,6 +4,7 @@
 #include <vector>
 #include <memory>
 
+#include <Poco/Version.h>
 #include <Poco/Exception.h>
 
 #include <Common/StackTrace.h>
@@ -114,7 +115,11 @@ public:
     }
 
 
-    std::string displayText() const override;
+    std::string displayText() const
+#if defined(POCO_CLICKHOUSE_PATCH)
+    override
+#endif
+    ;
 
     int getLineNumber() { return line_number_; }
     void setLineNumber(int line_number) { line_number_ = line_number;}

From 763844beaca81d0ddbaa5976c595be751c133450 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Wed, 23 Dec 2020 23:44:50 +0300
Subject: [PATCH 357/504] Make poco library path variable scope less global

---
 cmake/find/poco.cmake             | 4 +---
 contrib/poco-cmake/CMakeLists.txt | 2 ++
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/cmake/find/poco.cmake b/cmake/find/poco.cmake
index e749fc27925..99233728152 100644
--- a/cmake/find/poco.cmake
+++ b/cmake/find/poco.cmake
@@ -1,8 +1,6 @@
 option (USE_INTERNAL_POCO_LIBRARY "Use internal Poco library" ON)
 
-if (USE_INTERNAL_POCO_LIBRARY)
-    set (LIBRARY_DIR ${ClickHouse_SOURCE_DIR}/contrib/poco)
-else ()
+if (NOT USE_INTERNAL_POCO_LIBRARY)
     find_path (ROOT_DIR NAMES Foundation/include/Poco/Poco.h include/Poco/Poco.h)
     if (NOT ROOT_DIR)
         message (${RECONFIGURE_MESSAGE_LEVEL} "Can't find system poco")
diff --git a/contrib/poco-cmake/CMakeLists.txt b/contrib/poco-cmake/CMakeLists.txt
index fff9051f40f..1d2dc7b873e 100644
--- a/contrib/poco-cmake/CMakeLists.txt
+++ b/contrib/poco-cmake/CMakeLists.txt
@@ -1,3 +1,5 @@
+set (LIBRARY_DIR ${ClickHouse_SOURCE_DIR}/contrib/poco)
+
 add_subdirectory (Crypto)
 add_subdirectory (Data)
 add_subdirectory (Data/ODBC)

From bb4668fd6ea8054d384913dbc51f05fbf1212316 Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Thu, 24 Dec 2020 00:18:08 +0300
Subject: [PATCH 358/504] Fix bug: no newline after exception message in some
 tools

---
 programs/compressor/Compressor.cpp | 2 +-
 programs/format/Format.cpp         | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/programs/compressor/Compressor.cpp b/programs/compressor/Compressor.cpp
index dd906bffe7b..da92efdf251 100644
--- a/programs/compressor/Compressor.cpp
+++ b/programs/compressor/Compressor.cpp
@@ -155,7 +155,7 @@ int mainEntryClickHouseCompressor(int argc, char ** argv)
     }
     catch (...)
     {
-        std::cerr << getCurrentExceptionMessage(true);
+        std::cerr << getCurrentExceptionMessage(true) << '\n';
         return getCurrentExceptionCode();
     }
 
diff --git a/programs/format/Format.cpp b/programs/format/Format.cpp
index 986dc67a798..86a85d1d4a5 100644
--- a/programs/format/Format.cpp
+++ b/programs/format/Format.cpp
@@ -142,7 +142,7 @@ int mainEntryClickHouseFormat(int argc, char ** argv)
     }
     catch (...)
     {
-        std::cerr << getCurrentExceptionMessage(true);
+        std::cerr << getCurrentExceptionMessage(true) << '\n';
         return getCurrentExceptionCode();
     }
 

From 317df5f5d27e02d96b83a4fb46aec78e9913579c Mon Sep 17 00:00:00 2001
From: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date: Thu, 24 Dec 2020 00:57:33 +0300
Subject: [PATCH 359/504] try fix 'value is too short' when converting from
 String to Nullable(T)

---
 src/Core/callOnTypeIndex.h          |  54 +++++++--------
 src/Functions/FunctionsConversion.h | 102 ++++++++++++++++++++--------
 src/Functions/IFunction.cpp         |  35 ++++++----
 src/Functions/IFunctionAdaptors.h   |   2 +
 4 files changed, 124 insertions(+), 69 deletions(-)

diff --git a/src/Core/callOnTypeIndex.h b/src/Core/callOnTypeIndex.h
index 2ef049762f1..ccbccd7105a 100644
--- a/src/Core/callOnTypeIndex.h
+++ b/src/Core/callOnTypeIndex.h
@@ -162,43 +162,43 @@ template <typename T> class DataTypeNumber;
 template <typename T> class DataTypeDecimal;
 
 
-template <typename T, typename F>
-bool callOnIndexAndDataType(TypeIndex number, F && f)
+template <typename T, typename F, typename... ExtraArgs>
+bool callOnIndexAndDataType(TypeIndex number, F && f, ExtraArgs && ... args)
 {
     switch (number)
     {
-        case TypeIndex::UInt8:          return f(TypePair<DataTypeNumber<UInt8>, T>());
-        case TypeIndex::UInt16:         return f(TypePair<DataTypeNumber<UInt16>, T>());
-        case TypeIndex::UInt32:         return f(TypePair<DataTypeNumber<UInt32>, T>());
-        case TypeIndex::UInt64:         return f(TypePair<DataTypeNumber<UInt64>, T>());
-        case TypeIndex::UInt256:        return f(TypePair<DataTypeNumber<UInt256>, T>());
+        case TypeIndex::UInt8:          return f(TypePair<DataTypeNumber<UInt8>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::UInt16:         return f(TypePair<DataTypeNumber<UInt16>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::UInt32:         return f(TypePair<DataTypeNumber<UInt32>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::UInt64:         return f(TypePair<DataTypeNumber<UInt64>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::UInt256:        return f(TypePair<DataTypeNumber<UInt256>, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::Int8:           return f(TypePair<DataTypeNumber<Int8>, T>());
-        case TypeIndex::Int16:          return f(TypePair<DataTypeNumber<Int16>, T>());
-        case TypeIndex::Int32:          return f(TypePair<DataTypeNumber<Int32>, T>());
-        case TypeIndex::Int64:          return f(TypePair<DataTypeNumber<Int64>, T>());
-        case TypeIndex::Int128:         return f(TypePair<DataTypeNumber<Int128>, T>());
-        case TypeIndex::Int256:         return f(TypePair<DataTypeNumber<Int256>, T>());
+        case TypeIndex::Int8:           return f(TypePair<DataTypeNumber<Int8>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Int16:          return f(TypePair<DataTypeNumber<Int16>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Int32:          return f(TypePair<DataTypeNumber<Int32>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Int64:          return f(TypePair<DataTypeNumber<Int64>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Int128:         return f(TypePair<DataTypeNumber<Int128>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Int256:         return f(TypePair<DataTypeNumber<Int256>, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::Float32:        return f(TypePair<DataTypeNumber<Float32>, T>());
-        case TypeIndex::Float64:        return f(TypePair<DataTypeNumber<Float64>, T>());
+        case TypeIndex::Float32:        return f(TypePair<DataTypeNumber<Float32>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Float64:        return f(TypePair<DataTypeNumber<Float64>, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::Decimal32:      return f(TypePair<DataTypeDecimal<Decimal32>, T>());
-        case TypeIndex::Decimal64:      return f(TypePair<DataTypeDecimal<Decimal64>, T>());
-        case TypeIndex::Decimal128:     return f(TypePair<DataTypeDecimal<Decimal128>, T>());
-        case TypeIndex::Decimal256:     return f(TypePair<DataTypeDecimal<Decimal256>, T>());
+        case TypeIndex::Decimal32:      return f(TypePair<DataTypeDecimal<Decimal32>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Decimal64:      return f(TypePair<DataTypeDecimal<Decimal64>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Decimal128:     return f(TypePair<DataTypeDecimal<Decimal128>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Decimal256:     return f(TypePair<DataTypeDecimal<Decimal256>, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::Date:           return f(TypePair<DataTypeDate, T>());
-        case TypeIndex::DateTime:       return f(TypePair<DataTypeDateTime, T>());
-        case TypeIndex::DateTime64:     return f(TypePair<DataTypeDateTime64, T>());
+        case TypeIndex::Date:           return f(TypePair<DataTypeDate, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::DateTime:       return f(TypePair<DataTypeDateTime, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::DateTime64:     return f(TypePair<DataTypeDateTime64, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::String:         return f(TypePair<DataTypeString, T>());
-        case TypeIndex::FixedString:    return f(TypePair<DataTypeFixedString, T>());
+        case TypeIndex::String:         return f(TypePair<DataTypeString, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::FixedString:    return f(TypePair<DataTypeFixedString, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::Enum8:          return f(TypePair<DataTypeEnum<Int8>, T>());
-        case TypeIndex::Enum16:         return f(TypePair<DataTypeEnum<Int16>, T>());
+        case TypeIndex::Enum8:          return f(TypePair<DataTypeEnum<Int8>, T>(), std::forward<ExtraArgs>(args)...);
+        case TypeIndex::Enum16:         return f(TypePair<DataTypeEnum<Int16>, T>(), std::forward<ExtraArgs>(args)...);
 
-        case TypeIndex::UUID:           return f(TypePair<DataTypeUUID, T>());
+        case TypeIndex::UUID:           return f(TypePair<DataTypeUUID, T>(), std::forward<ExtraArgs>(args)...);
 
         default:
             break;
diff --git a/src/Functions/FunctionsConversion.h b/src/Functions/FunctionsConversion.h
index 92607f718d5..911496e676d 100644
--- a/src/Functions/FunctionsConversion.h
+++ b/src/Functions/FunctionsConversion.h
@@ -108,10 +108,14 @@ struct AccurateOrNullConvertStrategyAdditions
     UInt32 scale { 0 };
 };
 
+
+struct ConvertDefaultBehaviorTag {};
+struct ConvertReturnNullOnErrorTag {};
+
 /** Conversion of number types to each other, enums to numbers, dates and datetimes to numbers and back: done by straight assignment.
   *  (Date is represented internally as number of days from some day; DateTime - as unix timestamp)
   */
-template <typename FromDataType, typename ToDataType, typename Name>
+template <typename FromDataType, typename ToDataType, typename Name, typename SpecialTag = ConvertDefaultBehaviorTag>
 struct ConvertImpl
 {
     using FromFieldType = typename FromDataType::FieldType;
@@ -279,7 +283,7 @@ struct ConvertImpl
 
 /** Conversion of DateTime to Date: throw off time component.
   */
-template <typename Name> struct ConvertImpl<DataTypeDateTime, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeDateTime, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeDateTime, DataTypeDate, ToDateImpl> {};
 
 
@@ -301,7 +305,7 @@ struct ToDateTimeImpl
     }
 };
 
-template <typename Name> struct ConvertImpl<DataTypeDate, DataTypeDateTime, Name>
+template <typename Name> struct ConvertImpl<DataTypeDate, DataTypeDateTime, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeDate, DataTypeDateTime, ToDateTimeImpl> {};
 
 /// Implementation of toDate function.
@@ -357,21 +361,21 @@ struct ToDateTransform8Or16Signed
   *  when user write toDate(UInt32), expecting conversion of unix timestamp to Date.
   *  (otherwise such usage would be frequent mistake).
   */
-template <typename Name> struct ConvertImpl<DataTypeUInt32, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeUInt32, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeUInt32, DataTypeDate, ToDateTransform32Or64<UInt32, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeUInt64, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeUInt64, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeUInt64, DataTypeDate, ToDateTransform32Or64<UInt64, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeInt8, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeInt8, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeInt8, DataTypeDate, ToDateTransform8Or16Signed<Int8, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeInt16, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeInt16, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeInt16, DataTypeDate, ToDateTransform8Or16Signed<Int16, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeInt32, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeInt32, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeInt32, DataTypeDate, ToDateTransform32Or64Signed<Int32, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeInt64, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeInt64, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeInt64, DataTypeDate, ToDateTransform32Or64Signed<Int64, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeFloat32, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeFloat32, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeFloat32, DataTypeDate, ToDateTransform32Or64Signed<Float32, UInt16>> {};
-template <typename Name> struct ConvertImpl<DataTypeFloat64, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeFloat64, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeFloat64, DataTypeDate, ToDateTransform32Or64Signed<Float64, UInt16>> {};
 
 
@@ -456,9 +460,9 @@ struct ToDateTime64Transform
     }
 };
 
-template <typename Name> struct ConvertImpl<DataTypeDate, DataTypeDateTime64, Name>
+template <typename Name> struct ConvertImpl<DataTypeDate, DataTypeDateTime64, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeDate, DataTypeDateTime64, ToDateTime64Transform> {};
-template <typename Name> struct ConvertImpl<DataTypeDateTime, DataTypeDateTime64, Name>
+template <typename Name> struct ConvertImpl<DataTypeDateTime, DataTypeDateTime64, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeDateTime, DataTypeDateTime64, ToDateTime64Transform> {};
 
 /** Conversion of DateTime64 to Date or DateTime: discards fractional part.
@@ -481,9 +485,9 @@ struct FromDateTime64Transform
     }
 };
 
-template <typename Name> struct ConvertImpl<DataTypeDateTime64, DataTypeDate, Name>
+template <typename Name> struct ConvertImpl<DataTypeDateTime64, DataTypeDate, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeDateTime64, DataTypeDate, FromDateTime64Transform<ToDateImpl>> {};
-template <typename Name> struct ConvertImpl<DataTypeDateTime64, DataTypeDateTime, Name>
+template <typename Name> struct ConvertImpl<DataTypeDateTime64, DataTypeDateTime, Name, ConvertDefaultBehaviorTag>
     : DateTimeTransformImpl<DataTypeDateTime64, DataTypeDateTime, FromDateTime64Transform<ToDateTimeImpl>> {};
 
 
@@ -547,7 +551,7 @@ struct FormatImpl<DataTypeDecimal<FieldType>>
 
 /// DataTypeEnum<T> to DataType<T> free conversion
 template <typename FieldType, typename Name>
-struct ConvertImpl<DataTypeEnum<FieldType>, DataTypeNumber<FieldType>, Name>
+struct ConvertImpl<DataTypeEnum<FieldType>, DataTypeNumber<FieldType>, Name, ConvertDefaultBehaviorTag>
 {
     static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/)
     {
@@ -557,7 +561,7 @@ struct ConvertImpl<DataTypeEnum<FieldType>, DataTypeNumber<FieldType>, Name>
 
 
 template <typename FromDataType, typename Name>
-struct ConvertImpl<FromDataType, std::enable_if_t<!std::is_same_v<FromDataType, DataTypeString>, DataTypeString>, Name>
+struct ConvertImpl<FromDataType, std::enable_if_t<!std::is_same_v<FromDataType, DataTypeString>, DataTypeString>, Name, ConvertDefaultBehaviorTag>
 {
     using FromFieldType = typename FromDataType::FieldType;
     using ColVecType = std::conditional_t<IsDecimalNumber<FromFieldType>, ColumnDecimal<FromFieldType>, ColumnVector<FromFieldType>>;
@@ -973,14 +977,24 @@ struct ConvertThroughParsing
 };
 
 
+
+
 template <typename ToDataType, typename Name>
-struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeString>, DataTypeString>, ToDataType, Name>
+struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeString>, DataTypeString>, ToDataType, Name, ConvertDefaultBehaviorTag>
     : ConvertThroughParsing<DataTypeString, ToDataType, Name, ConvertFromStringExceptionMode::Throw, ConvertFromStringParsingMode::Normal> {};
 
 template <typename ToDataType, typename Name>
-struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeFixedString>, DataTypeFixedString>, ToDataType, Name>
+struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeFixedString>, DataTypeFixedString>, ToDataType, Name, ConvertDefaultBehaviorTag>
     : ConvertThroughParsing<DataTypeFixedString, ToDataType, Name, ConvertFromStringExceptionMode::Throw, ConvertFromStringParsingMode::Normal> {};
 
+template <typename ToDataType, typename Name>
+struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeString>, DataTypeString>, ToDataType, Name, ConvertReturnNullOnErrorTag>
+    : ConvertThroughParsing<DataTypeString, ToDataType, Name, ConvertFromStringExceptionMode::Null, ConvertFromStringParsingMode::Normal> {};
+
+template <typename ToDataType, typename Name>
+struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeFixedString>, DataTypeFixedString>, ToDataType, Name, ConvertReturnNullOnErrorTag>
+    : ConvertThroughParsing<DataTypeFixedString, ToDataType, Name, ConvertFromStringExceptionMode::Null, ConvertFromStringParsingMode::Normal> {};
+
 /// Generic conversion of any type from String. Used for complex types: Array and Tuple.
 struct ConvertImplGenericFromString
 {
@@ -1027,14 +1041,14 @@ struct ConvertImplGenericFromString
 
 
 template <>
-struct ConvertImpl<DataTypeString, DataTypeUInt32, NameToUnixTimestamp>
-    : ConvertImpl<DataTypeString, DataTypeDateTime, NameToUnixTimestamp> {};
+struct ConvertImpl<DataTypeString, DataTypeUInt32, NameToUnixTimestamp, ConvertDefaultBehaviorTag>
+    : ConvertImpl<DataTypeString, DataTypeDateTime, NameToUnixTimestamp, ConvertDefaultBehaviorTag> {};
 
 
 /** If types are identical, just take reference to column.
   */
 template <typename T, typename Name>
-struct ConvertImpl<std::enable_if_t<!T::is_parametric, T>, T, Name>
+struct ConvertImpl<std::enable_if_t<!T::is_parametric, T>, T, Name, ConvertDefaultBehaviorTag>
 {
     template <typename Additions = void *>
     static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/,
@@ -1049,7 +1063,7 @@ struct ConvertImpl<std::enable_if_t<!T::is_parametric, T>, T, Name>
   * Cutting sequences of zero bytes from end of strings.
   */
 template <typename Name>
-struct ConvertImpl<DataTypeFixedString, DataTypeString, Name>
+struct ConvertImpl<DataTypeFixedString, DataTypeString, Name, ConvertDefaultBehaviorTag>
 {
     static ColumnPtr execute(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/)
     {
@@ -1153,6 +1167,9 @@ public:
 
     static constexpr bool to_datetime64 = std::is_same_v<ToDataType, DataTypeDateTime64>;
 
+    static constexpr bool to_string_or_fixed_string = std::is_same_v<ToDataType, DataTypeFixedString> ||
+                                                      std::is_same_v<ToDataType, DataTypeString>;
+
     static FunctionPtr create(const Context &) { return std::make_shared<FunctionConvert>(); }
     static FunctionPtr create() { return std::make_shared<FunctionConvert>(); }
 
@@ -1166,6 +1183,14 @@ public:
     bool isInjective(const ColumnsWithTypeAndName &) const override { return std::is_same_v<Name, NameToString>; }
 
     DataTypePtr getReturnTypeImpl(const ColumnsWithTypeAndName & arguments) const override
+    {
+        auto getter = [&] (const auto & args) { return getReturnTypeImplRemovedNullable(args); };
+        auto res = FunctionOverloadResolverAdaptor::getReturnTypeDefaultImplementationForNulls(arguments, getter);
+        to_nullable = res->isNullable();
+        return res;
+    }
+
+    DataTypePtr getReturnTypeImplRemovedNullable(const ColumnsWithTypeAndName & arguments) const
     {
         FunctionArgumentDescriptors mandatory_args = {{"Value", nullptr, nullptr, nullptr}};
         FunctionArgumentDescriptors optional_args;
@@ -1243,11 +1268,17 @@ public:
         }
     }
 
+    bool useDefaultImplementationForNulls() const override { return to_nullable; }  // FIXME 
     bool useDefaultImplementationForConstants() const override { return true; }
     ColumnNumbers getArgumentsThatAreAlwaysConstant() const override { return {1}; }
     bool canBeExecutedOnDefaultArguments() const override { return false; }
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
+    {
+        return executeImplRemovedNullable(arguments, result_type, input_rows_count);
+    }
+
+    ColumnPtr executeImplRemovedNullable(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const
     {
         try
         {
@@ -1292,6 +1323,8 @@ public:
     }
 
 private:
+    mutable bool to_nullable = false;
+
     ColumnPtr executeInternal(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const
     {
         if (arguments.empty())
@@ -1301,11 +1334,12 @@ private:
         const IDataType * from_type = arguments[0].type.get();
         ColumnPtr result_column;
 
-        auto call = [&](const auto & types) -> bool
+        auto call = [&](const auto & types, const auto & tag) -> bool
         {
             using Types = std::decay_t<decltype(types)>;
             using LeftDataType = typename Types::LeftType;
             using RightDataType = typename Types::RightType;
+            using SpecialTag = std::decay_t<decltype(tag)>;
 
             if constexpr (IsDataTypeDecimal<RightDataType>)
             {
@@ -1349,10 +1383,10 @@ private:
                     throw Exception("Wrong UUID conversion", ErrorCodes::CANNOT_CONVERT_TYPE);
                 }
                 else
-                    result_column = ConvertImpl<LeftDataType, RightDataType, Name>::execute(arguments, result_type, input_rows_count);
+                    result_column = ConvertImpl<LeftDataType, RightDataType, Name, SpecialTag>::execute(arguments, result_type, input_rows_count);
             }
             else
-                result_column = ConvertImpl<LeftDataType, RightDataType, Name>::execute(arguments, result_type, input_rows_count);
+                result_column = ConvertImpl<LeftDataType, RightDataType, Name, SpecialTag>::execute(arguments, result_type, input_rows_count);
 
             return true;
         };
@@ -1365,7 +1399,7 @@ private:
 
             if (to_datetime64 || scale != 0) /// When scale = 0, the data type is DateTime otherwise the data type is DateTime64
             {
-                if (!callOnIndexAndDataType<DataTypeDateTime64>(from_type->getTypeId(), call))
+                if (!callOnIndexAndDataType<DataTypeDateTime64>(from_type->getTypeId(), call, ConvertDefaultBehaviorTag{}))
                     throw Exception("Illegal type " + arguments[0].type->getName() + " of argument of function " + getName(),
                                     ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
@@ -1373,7 +1407,19 @@ private:
             }
         }
 
-        bool done = callOnIndexAndDataType<ToDataType>(from_type->getTypeId(), call);
+        bool done;
+        if constexpr (to_string_or_fixed_string)
+        {
+            done = callOnIndexAndDataType<ToDataType>(from_type->getTypeId(), call, ConvertDefaultBehaviorTag{});
+        }
+        else
+        {
+            if (to_nullable && WhichDataType(from_type).isStringOrFixedString())
+                done = callOnIndexAndDataType<ToDataType>(from_type->getTypeId(), call, ConvertReturnNullOnErrorTag{});
+            else
+                done = callOnIndexAndDataType<ToDataType>(from_type->getTypeId(), call, ConvertDefaultBehaviorTag{});
+        }
+
         if (!done)
         {
             /// Generic conversion of any type to String.
diff --git a/src/Functions/IFunction.cpp b/src/Functions/IFunction.cpp
index f51c44cf904..e4a1adb8525 100644
--- a/src/Functions/IFunction.cpp
+++ b/src/Functions/IFunction.cpp
@@ -465,25 +465,32 @@ void FunctionOverloadResolverAdaptor::checkNumberOfArguments(size_t number_of_ar
                         ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
 }
 
+
+DataTypePtr FunctionOverloadResolverAdaptor::getReturnTypeDefaultImplementationForNulls(const ColumnsWithTypeAndName & arguments,
+                                                                                        const DefaultReturnTypeGetter & getter)
+{
+    NullPresence null_presence = getNullPresense(arguments);
+
+    if (null_presence.has_null_constant)
+    {
+        return makeNullable(std::make_shared<DataTypeNothing>());
+    }
+    if (null_presence.has_nullable)
+    {
+        Block nested_columns = createBlockWithNestedColumns(arguments);
+        auto return_type = getter(ColumnsWithTypeAndName(nested_columns.begin(), nested_columns.end()));
+        return makeNullable(return_type);
+    }
+
+    return getter(arguments);
+}
+
 DataTypePtr FunctionOverloadResolverAdaptor::getReturnTypeWithoutLowCardinality(const ColumnsWithTypeAndName & arguments) const
 {
     checkNumberOfArguments(arguments.size());
 
     if (!arguments.empty() && impl->useDefaultImplementationForNulls())
-    {
-        NullPresence null_presence = getNullPresense(arguments);
-
-        if (null_presence.has_null_constant)
-        {
-            return makeNullable(std::make_shared<DataTypeNothing>());
-        }
-        if (null_presence.has_nullable)
-        {
-            Block nested_columns = createBlockWithNestedColumns(arguments);
-            auto return_type = impl->getReturnType(ColumnsWithTypeAndName(nested_columns.begin(), nested_columns.end()));
-            return makeNullable(return_type);
-        }
-    }
+        return getReturnTypeDefaultImplementationForNulls(arguments, [&](const auto & args) { return impl->getReturnType(args); });
 
     return impl->getReturnType(arguments);
 }
diff --git a/src/Functions/IFunctionAdaptors.h b/src/Functions/IFunctionAdaptors.h
index 737d904f6c9..cbb71c99b1b 100644
--- a/src/Functions/IFunctionAdaptors.h
+++ b/src/Functions/IFunctionAdaptors.h
@@ -129,6 +129,8 @@ public:
         return impl->getArgumentsThatDontImplyNullableReturnType(number_of_arguments);
     }
 
+    using DefaultReturnTypeGetter = std::function<DataTypePtr(const ColumnsWithTypeAndName &)>;
+    static DataTypePtr getReturnTypeDefaultImplementationForNulls(const ColumnsWithTypeAndName & arguments, const DefaultReturnTypeGetter & getter);
 private:
     FunctionOverloadResolverImplPtr impl;
 

From 4a37db455d3664f93c36bcdf5d246790e4af855a Mon Sep 17 00:00:00 2001
From: robot-clickhouse <robot-clickhouse@yandex-team.ru>
Date: Thu, 24 Dec 2020 01:02:07 +0300
Subject: [PATCH 360/504] Update version_date.tsv after release 20.12.4.5

---
 utils/list-versions/version_date.tsv | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/list-versions/version_date.tsv b/utils/list-versions/version_date.tsv
index cec9dbf1b08..e76918dd368 100644
--- a/utils/list-versions/version_date.tsv
+++ b/utils/list-versions/version_date.tsv
@@ -1,3 +1,4 @@
+v20.12.4.5-stable	2020-12-24
 v20.12.3.3-stable	2020-12-09
 v20.12.2.1-stable	2020-12-09
 v20.11.5.18-stable	2020-12-06

From 42b76979fe51153c30f86cc893eade69bce56e50 Mon Sep 17 00:00:00 2001
From: robot-clickhouse <robot-clickhouse@yandex-team.ru>
Date: Thu, 24 Dec 2020 01:20:16 +0300
Subject: [PATCH 361/504] Update version_date.tsv after release 20.11.6.6

---
 utils/list-versions/version_date.tsv | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/list-versions/version_date.tsv b/utils/list-versions/version_date.tsv
index e76918dd368..be5209447c9 100644
--- a/utils/list-versions/version_date.tsv
+++ b/utils/list-versions/version_date.tsv
@@ -1,6 +1,7 @@
 v20.12.4.5-stable	2020-12-24
 v20.12.3.3-stable	2020-12-09
 v20.12.2.1-stable	2020-12-09
+v20.11.6.6-stable	2020-12-24
 v20.11.5.18-stable	2020-12-06
 v20.11.4.13-stable	2020-11-20
 v20.11.3.3-stable	2020-11-13

From c49a6302fa5f359c7096a532197e4ec7dcd0c132 Mon Sep 17 00:00:00 2001
From: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date: Thu, 24 Dec 2020 03:13:53 +0300
Subject: [PATCH 362/504] fix, add test

---
 src/Functions/FunctionsConversion.h           | 24 +++++++-----
 .../01186_conversion_to_nullable.reference    | 38 +++++++++++++++++++
 .../01186_conversion_to_nullable.sql          | 13 +++++++
 3 files changed, 65 insertions(+), 10 deletions(-)
 create mode 100644 tests/queries/0_stateless/01186_conversion_to_nullable.reference
 create mode 100644 tests/queries/0_stateless/01186_conversion_to_nullable.sql

diff --git a/src/Functions/FunctionsConversion.h b/src/Functions/FunctionsConversion.h
index 911496e676d..84a41a9c415 100644
--- a/src/Functions/FunctionsConversion.h
+++ b/src/Functions/FunctionsConversion.h
@@ -977,8 +977,6 @@ struct ConvertThroughParsing
 };
 
 
-
-
 template <typename ToDataType, typename Name>
 struct ConvertImpl<std::enable_if_t<!std::is_same_v<ToDataType, DataTypeString>, DataTypeString>, ToDataType, Name, ConvertDefaultBehaviorTag>
     : ConvertThroughParsing<DataTypeString, ToDataType, Name, ConvertFromStringExceptionMode::Throw, ConvertFromStringParsingMode::Normal> {};
@@ -1044,6 +1042,9 @@ template <>
 struct ConvertImpl<DataTypeString, DataTypeUInt32, NameToUnixTimestamp, ConvertDefaultBehaviorTag>
     : ConvertImpl<DataTypeString, DataTypeDateTime, NameToUnixTimestamp, ConvertDefaultBehaviorTag> {};
 
+template <>
+struct ConvertImpl<DataTypeString, DataTypeUInt32, NameToUnixTimestamp, ConvertReturnNullOnErrorTag>
+    : ConvertImpl<DataTypeString, DataTypeDateTime, NameToUnixTimestamp, ConvertReturnNullOnErrorTag> {};
 
 /** If types are identical, just take reference to column.
   */
@@ -1187,6 +1188,7 @@ public:
         auto getter = [&] (const auto & args) { return getReturnTypeImplRemovedNullable(args); };
         auto res = FunctionOverloadResolverAdaptor::getReturnTypeDefaultImplementationForNulls(arguments, getter);
         to_nullable = res->isNullable();
+        checked_return_type = true;
         return res;
     }
 
@@ -1268,17 +1270,16 @@ public:
         }
     }
 
-    bool useDefaultImplementationForNulls() const override { return to_nullable; }  // FIXME 
+    /// Function actually uses default implementation for nulls,
+    /// but we need to know if return type is Nullable or not,
+    /// so we use checked_return_type only to intercept the first call to getReturnTypeImpl(...).
+    bool useDefaultImplementationForNulls() const override { return checked_return_type; }
+
     bool useDefaultImplementationForConstants() const override { return true; }
     ColumnNumbers getArgumentsThatAreAlwaysConstant() const override { return {1}; }
     bool canBeExecutedOnDefaultArguments() const override { return false; }
 
     ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
-    {
-        return executeImplRemovedNullable(arguments, result_type, input_rows_count);
-    }
-
-    ColumnPtr executeImplRemovedNullable(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const
     {
         try
         {
@@ -1323,6 +1324,7 @@ public:
     }
 
 private:
+    mutable bool checked_return_type = false;
     mutable bool to_nullable = false;
 
     ColumnPtr executeInternal(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const
@@ -1359,12 +1361,12 @@ private:
                 const ColumnWithTypeAndName & scale_column = arguments[1];
                 UInt32 scale = extractToDecimalScale(scale_column);
 
-                result_column = ConvertImpl<LeftDataType, RightDataType, Name>::execute(arguments, result_type, input_rows_count, scale);
+                result_column = ConvertImpl<LeftDataType, RightDataType, Name, SpecialTag>::execute(arguments, result_type, input_rows_count, scale);
             }
             else if constexpr (IsDataTypeDateOrDateTime<RightDataType> && std::is_same_v<LeftDataType, DataTypeDateTime64>)
             {
                 const auto * dt64 = assert_cast<const DataTypeDateTime64 *>(arguments[0].type.get());
-                result_column = ConvertImpl<LeftDataType, RightDataType, Name>::execute(arguments, result_type, input_rows_count, dt64->getScale());
+                result_column = ConvertImpl<LeftDataType, RightDataType, Name, SpecialTag>::execute(arguments, result_type, input_rows_count, dt64->getScale());
             }
             else if constexpr (IsDataTypeDecimalOrNumber<LeftDataType> && IsDataTypeDecimalOrNumber<RightDataType>)
             {
@@ -1414,6 +1416,8 @@ private:
         }
         else
         {
+            /// We should use ConvertFromStringExceptionMode::Null mode when converting from String (or FixedString)
+            /// to Nullable type, to avoid 'value is too short' error on attempt to parse empty string from NULL values.
             if (to_nullable && WhichDataType(from_type).isStringOrFixedString())
                 done = callOnIndexAndDataType<ToDataType>(from_type->getTypeId(), call, ConvertReturnNullOnErrorTag{});
             else
diff --git a/tests/queries/0_stateless/01186_conversion_to_nullable.reference b/tests/queries/0_stateless/01186_conversion_to_nullable.reference
new file mode 100644
index 00000000000..04146644154
--- /dev/null
+++ b/tests/queries/0_stateless/01186_conversion_to_nullable.reference
@@ -0,0 +1,38 @@
+42
+\N
+0
+\N
+\N
+42
+\N
+0
+\N
+256
+2020-12-24
+\N
+1970-01-01
+\N
+1970-01-01
+2020-12-24 01:02:03
+\N
+1970-01-01 03:00:00
+\N
+2020-12-24 01:02:03.00
+\N
+1970-01-01 03:00:00.00
+1970-01-01 03:00:00.00
+946721532
+\N
+\N
+42.00
+\N
+\N
+42.00000000
+\N
+3.14159000
+42
+\N
+test
+42\0\0\0\0\0\0
+\N
+test\0\0\0\0
diff --git a/tests/queries/0_stateless/01186_conversion_to_nullable.sql b/tests/queries/0_stateless/01186_conversion_to_nullable.sql
new file mode 100644
index 00000000000..bf7df6234d2
--- /dev/null
+++ b/tests/queries/0_stateless/01186_conversion_to_nullable.sql
@@ -0,0 +1,13 @@
+select toUInt8(x) from values('x Nullable(String)', '42', NULL, '0', '', '256');
+select toInt64(x) from values('x Nullable(String)', '42', NULL, '0', '', '256');
+
+select toDate(x) from values('x Nullable(String)', '2020-12-24', NULL, '0000-00-00', '', '9999-01-01');
+select toDateTime(x) from values('x Nullable(String)', '2020-12-24 01:02:03', NULL, '0000-00-00 00:00:00', '');
+select toDateTime64(x, 2) from values('x Nullable(String)', '2020-12-24 01:02:03', NULL, '0000-00-00 00:00:00', '');
+select toUnixTimestamp(x) from values ('x Nullable(String)', '2000-01-01 13:12:12', NULL, '');
+
+select toDecimal32(x, 2) from values ('x Nullable(String)', '42', NULL, '3.14159');
+select toDecimal64(x, 8) from values ('x Nullable(String)', '42', NULL, '3.14159');
+
+select toString(x) from values ('x Nullable(String)', '42', NULL, 'test');
+select toFixedString(x, 8) from values ('x Nullable(String)', '42', NULL, 'test');

From 22c51cb89a03e8b696ed24d2be71d2aa946e2f0f Mon Sep 17 00:00:00 2001
From: robot-clickhouse <robot-clickhouse@yandex-team.ru>
Date: Thu, 24 Dec 2020 03:18:12 +0300
Subject: [PATCH 363/504] Update version_date.tsv after release 20.10.7.4

---
 utils/list-versions/version_date.tsv | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/list-versions/version_date.tsv b/utils/list-versions/version_date.tsv
index be5209447c9..d252d6ed7b4 100644
--- a/utils/list-versions/version_date.tsv
+++ b/utils/list-versions/version_date.tsv
@@ -6,6 +6,7 @@ v20.11.5.18-stable	2020-12-06
 v20.11.4.13-stable	2020-11-20
 v20.11.3.3-stable	2020-11-13
 v20.11.2.1-stable	2020-11-11
+v20.10.7.4-stable	2020-12-24
 v20.10.6.27-stable	2020-12-06
 v20.10.5.10-stable	2020-11-20
 v20.10.4.1-stable	2020-11-13

From a58acbae67430055a16d0e6a34d1d265d5727040 Mon Sep 17 00:00:00 2001
From: robot-clickhouse <robot-clickhouse@yandex-team.ru>
Date: Thu, 24 Dec 2020 03:54:43 +0300
Subject: [PATCH 364/504] Update version_date.tsv after release 20.8.10.13

---
 utils/list-versions/version_date.tsv | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/list-versions/version_date.tsv b/utils/list-versions/version_date.tsv
index d252d6ed7b4..e7e4b78c99b 100644
--- a/utils/list-versions/version_date.tsv
+++ b/utils/list-versions/version_date.tsv
@@ -18,6 +18,7 @@ v20.9.5.5-stable	2020-11-13
 v20.9.4.76-stable	2020-10-29
 v20.9.3.45-stable	2020-10-09
 v20.9.2.20-stable	2020-09-22
+v20.8.10.13-lts	2020-12-24
 v20.8.9.6-lts	2020-12-10
 v20.8.8.2-lts	2020-12-07
 v20.8.7.15-lts	2020-11-20

From f2a35b59e07bfd99980118ccd34b5bdd4571cb4b Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 03:55:46 +0300
Subject: [PATCH 365/504] Update CHANGELOG.md

---
 CHANGELOG.md | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 6d9174c7d07..7920fafa11c 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,19 @@
 ### ClickHouse release 20.12
 
+### ClickHouse release v20.12.4.5-stable, 2020-12-24
+
+#### Bug Fix
+
+* Fixed issue when `clickhouse-odbc-bridge` process is unreachable by server on machines with dual IPv4/IPv6 stack; - Fixed issue when ODBC dictionary updates are performed using malformed queries and/or cause crashes; Possibly closes [#14489](https://github.com/ClickHouse/ClickHouse/issues/14489). [#18278](https://github.com/ClickHouse/ClickHouse/pull/18278) ([Denis Glazachev](https://github.com/traceon)).
+* Fixed key comparison between Enum and Int types. This fixes [#17989](https://github.com/ClickHouse/ClickHouse/issues/17989). [#18214](https://github.com/ClickHouse/ClickHouse/pull/18214) ([Amos Bird](https://github.com/amosbird)).
+* Fixed unique key convert crash in `MaterializeMySQL` database engine. This fixes [#18186](https://github.com/ClickHouse/ClickHouse/issues/18186) and fixes [#16372](https://github.com/ClickHouse/ClickHouse/issues/16372)  [#18211](https://github.com/ClickHouse/ClickHouse/pull/18211) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed `std::out_of_range: basic_string` in S3 URL parsing. [#18059](https://github.com/ClickHouse/ClickHouse/pull/18059) ([Vladimir Chebotarev](https://github.com/excitoon)).
+* Fixed the issue when some tables not synchronized to ClickHouse from MySQL caused by the fact that convertion MySQL prefix index wasn't supported for MaterializeMySQL. This fixes [#15187](https://github.com/ClickHouse/ClickHouse/issues/15187) and fixes [#17912](https://github.com/ClickHouse/ClickHouse/issues/17912)  [#17944](https://github.com/ClickHouse/ClickHouse/pull/17944) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed the issue when query optimization was producing wrong result if query contains `ARRAY JOIN`. [#17887](https://github.com/ClickHouse/ClickHouse/pull/17887) ([sundyli](https://github.com/sundy-li)).
+* Fix possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).
+* Fix empty `system.stack_trace` table when server is running in daemon mode. [#17630](https://github.com/ClickHouse/ClickHouse/pull/17630) ([Amos Bird](https://github.com/amosbird)).
+
+
 ### ClickHouse release v20.12.3.3-stable, 2020-12-13
 
 #### Backward Incompatible Change

From ac8cc2b423fbc34af6b8b2676c352c012c319087 Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 03:56:43 +0300
Subject: [PATCH 366/504] Update CHANGELOG.md

---
 CHANGELOG.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 7920fafa11c..ae26f90c006 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -10,8 +10,8 @@
 * Fixed `std::out_of_range: basic_string` in S3 URL parsing. [#18059](https://github.com/ClickHouse/ClickHouse/pull/18059) ([Vladimir Chebotarev](https://github.com/excitoon)).
 * Fixed the issue when some tables not synchronized to ClickHouse from MySQL caused by the fact that convertion MySQL prefix index wasn't supported for MaterializeMySQL. This fixes [#15187](https://github.com/ClickHouse/ClickHouse/issues/15187) and fixes [#17912](https://github.com/ClickHouse/ClickHouse/issues/17912)  [#17944](https://github.com/ClickHouse/ClickHouse/pull/17944) ([Winter Zhang](https://github.com/zhang2014)).
 * Fixed the issue when query optimization was producing wrong result if query contains `ARRAY JOIN`. [#17887](https://github.com/ClickHouse/ClickHouse/pull/17887) ([sundyli](https://github.com/sundy-li)).
-* Fix possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).
-* Fix empty `system.stack_trace` table when server is running in daemon mode. [#17630](https://github.com/ClickHouse/ClickHouse/pull/17630) ([Amos Bird](https://github.com/amosbird)).
+* Fixed possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).
+* Fixed empty `system.stack_trace` table when server is running in daemon mode. [#17630](https://github.com/ClickHouse/ClickHouse/pull/17630) ([Amos Bird](https://github.com/amosbird)).
 
 
 ### ClickHouse release v20.12.3.3-stable, 2020-12-13

From 54a11d91aca118e4721713de31577044892b364a Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 04:02:54 +0300
Subject: [PATCH 367/504] Update CHANGELOG.md

---
 CHANGELOG.md | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index ae26f90c006..ed558b4dfd3 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -137,6 +137,20 @@
 
 ## ClickHouse release 20.11
 
+### ClickHouse release v20.11.6.6-stable, 2020-12-24
+
+#### Bug Fix
+
+* Fixed issue when `clickhouse-odbc-bridge` process is unreachable by server on machines with dual `IPv4/IPv6 stack` and fixed issue when ODBC dictionary updates are performed using malformed queries and/or cause crashes. This possibly fixes [#14489](https://github.com/ClickHouse/ClickHouse/issues/14489). [#18278](https://github.com/ClickHouse/ClickHouse/pull/18278) ([Denis Glazachev](https://github.com/traceon)).
+* Fixed key comparison between `Enum` and `Int` types. This fixes [#17989](https://github.com/ClickHouse/ClickHouse/issues/17989). [#18214](https://github.com/ClickHouse/ClickHouse/pull/18214) ([Amos Bird](https://github.com/amosbird)).
+* Fixed unique key convert crash in `MaterializeMySQL` database engine. This fixes [#18186](https://github.com/ClickHouse/ClickHouse/issues/18186) and fixes [#16372](https://github.com/ClickHouse/ClickHouse/issues/16372)  [#18211](https://github.com/ClickHouse/ClickHouse/pull/18211) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed `std::out_of_range: basic_string` in S3 URL parsing. [#18059](https://github.com/ClickHouse/ClickHouse/pull/18059) ([Vladimir Chebotarev](https://github.com/excitoon)).
+* Fixed the issue when some tables not synchronized to ClickHouse from MySQL caused by the fact that convertion MySQL prefix index wasn't supported for MaterializeMySQL. This fixes [#15187](https://github.com/ClickHouse/ClickHouse/issues/15187) and fixes [#17912](https://github.com/ClickHouse/ClickHouse/issues/17912)  [#17944](https://github.com/ClickHouse/ClickHouse/pull/17944) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed the issue when query optimization was producing wrong result if query contains `ARRAY JOIN`. [#17887](https://github.com/ClickHouse/ClickHouse/pull/17887) ([sundyli](https://github.com/sundy-li)).
+* Fixed possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).
+* Fixed empty `system.stack_trace` table when server is running in daemon mode. [#17630](https://github.com/ClickHouse/ClickHouse/pull/17630) ([Amos Bird](https://github.com/amosbird)).
+
+
 ### ClickHouse release v20.11.3.3-stable, 2020-11-13
 
 #### Bug Fix

From 3e03c2ae63ccdd35f17708c0408199b2ba2aa2df Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 04:14:49 +0300
Subject: [PATCH 368/504] Update CHANGELOG.md

---
 CHANGELOG.md | 38 +++++++++++++++++++++++++++++++++++---
 1 file changed, 35 insertions(+), 3 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index ed558b4dfd3..f9bbda56e40 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -141,14 +141,46 @@
 
 #### Bug Fix
 
-* Fixed issue when `clickhouse-odbc-bridge` process is unreachable by server on machines with dual `IPv4/IPv6 stack` and fixed issue when ODBC dictionary updates are performed using malformed queries and/or cause crashes. This possibly fixes [#14489](https://github.com/ClickHouse/ClickHouse/issues/14489). [#18278](https://github.com/ClickHouse/ClickHouse/pull/18278) ([Denis Glazachev](https://github.com/traceon)).
-* Fixed key comparison between `Enum` and `Int` types. This fixes [#17989](https://github.com/ClickHouse/ClickHouse/issues/17989). [#18214](https://github.com/ClickHouse/ClickHouse/pull/18214) ([Amos Bird](https://github.com/amosbird)).
+* Fixed issue when `clickhouse-odbc-bridge` process is unreachable by server on machines with dual `IPv4/IPv6 stack` and fixed issue when ODBC dictionary updates are performed using malformed queries and/or cause crashes. This possibly closes [#14489](https://github.com/ClickHouse/ClickHouse/issues/14489). [#18278](https://github.com/ClickHouse/ClickHouse/pull/18278) ([Denis Glazachev](https://github.com/traceon)).
+* Fixed key comparison between Enum and Int types. This fixes [#17989](https://github.com/ClickHouse/ClickHouse/issues/17989). [#18214](https://github.com/ClickHouse/ClickHouse/pull/18214) ([Amos Bird](https://github.com/amosbird)).
 * Fixed unique key convert crash in `MaterializeMySQL` database engine. This fixes [#18186](https://github.com/ClickHouse/ClickHouse/issues/18186) and fixes [#16372](https://github.com/ClickHouse/ClickHouse/issues/16372)  [#18211](https://github.com/ClickHouse/ClickHouse/pull/18211) ([Winter Zhang](https://github.com/zhang2014)).
 * Fixed `std::out_of_range: basic_string` in S3 URL parsing. [#18059](https://github.com/ClickHouse/ClickHouse/pull/18059) ([Vladimir Chebotarev](https://github.com/excitoon)).
 * Fixed the issue when some tables not synchronized to ClickHouse from MySQL caused by the fact that convertion MySQL prefix index wasn't supported for MaterializeMySQL. This fixes [#15187](https://github.com/ClickHouse/ClickHouse/issues/15187) and fixes [#17912](https://github.com/ClickHouse/ClickHouse/issues/17912)  [#17944](https://github.com/ClickHouse/ClickHouse/pull/17944) ([Winter Zhang](https://github.com/zhang2014)).
 * Fixed the issue when query optimization was producing wrong result if query contains `ARRAY JOIN`. [#17887](https://github.com/ClickHouse/ClickHouse/pull/17887) ([sundyli](https://github.com/sundy-li)).
-* Fixed possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).
+* Fix possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).
+* Do not restore parts from WAL if `in_memory_parts_enable_wal` is disabled. [#17802](https://github.com/ClickHouse/ClickHouse/pull/17802) ([detailyang](https://github.com/detailyang)).
+* Fixed problem when ClickHouse fails to resume connection to MySQL servers. [#17681](https://github.com/ClickHouse/ClickHouse/pull/17681) ([Alexander Kazakov](https://github.com/Akazz)).
+* Fixed inconsistent behaviour of `optimize_trivial_count_query` with partition predicate. [#17644](https://github.com/ClickHouse/ClickHouse/pull/17644) ([Azat Khuzhin](https://github.com/azat)).
 * Fixed empty `system.stack_trace` table when server is running in daemon mode. [#17630](https://github.com/ClickHouse/ClickHouse/pull/17630) ([Amos Bird](https://github.com/amosbird)).
+* Fixed the behaviour when xxception `fmt::v7::format_error` can be logged in background for MergeTree tables. This fixes [#17613](https://github.com/ClickHouse/ClickHouse/issues/17613). [#17615](https://github.com/ClickHouse/ClickHouse/pull/17615) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed the behaviour when clickhouse-client is used in interactive mode with multiline queries and single line comment was erronously extended till the end of query. This fixes [#13654](https://github.com/ClickHouse/ClickHouse/issues/13654). [#17565](https://github.com/ClickHouse/ClickHouse/pull/17565) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed the issue when server can stop accepting connections in very rare cases. [#17542](https://github.com/ClickHouse/ClickHouse/pull/17542) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed alter query hang when the corresponding mutation was killed on the different replica. This fixes [#16953](https://github.com/ClickHouse/ClickHouse/issues/16953). [#17499](https://github.com/ClickHouse/ClickHouse/pull/17499) ([alesapin](https://github.com/alesapin)).
+* Fixed bug when mark cache size was underestimated by clickhouse. It may happen when there are a lot of tiny files with marks. [#17496](https://github.com/ClickHouse/ClickHouse/pull/17496) ([alesapin](https://github.com/alesapin)).
+* Fixed `ORDER BY` with enabled setting `optimize_redundant_functions_in_order_by`. [#17471](https://github.com/ClickHouse/ClickHouse/pull/17471) ([Anton Popov](https://github.com/CurtizJ)).
+* Fixed duplicates after `DISTINCT` which were possible because of incorrect optimization. This fixes [#17294](https://github.com/ClickHouse/ClickHouse/issues/17294). [#17296](https://github.com/ClickHouse/ClickHouse/pull/17296) ([li chengxiang](https://github.com/chengxianglibra)). [#17439](https://github.com/ClickHouse/ClickHouse/pull/17439) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fixed crash while reading from `JOIN` table with `LowCardinality` types. This fixes [#17228](https://github.com/ClickHouse/ClickHouse/issues/17228). [#17397](https://github.com/ClickHouse/ClickHouse/pull/17397) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fixed set index invalidation when there are const columns in the subquery. This fixes [#17246](https://github.com/ClickHouse/ClickHouse/issues/17246) . [#17249](https://github.com/ClickHouse/ClickHouse/pull/17249) ([Amos Bird](https://github.com/amosbird)).
+* Fixed possible wrong index analysis when the types of the index comparison are different. This fixes [#17122](https://github.com/ClickHouse/ClickHouse/issues/17122). [#17145](https://github.com/ClickHouse/ClickHouse/pull/17145) ([Amos Bird](https://github.com/amosbird)).
+* Fixed `ColumnConst` comparison which leads to crash. This fixes [#17088](https://github.com/ClickHouse/ClickHouse/issues/17088) . [#17135](https://github.com/ClickHouse/ClickHouse/pull/17135) ([Amos Bird](https://github.com/amosbird)).
+* Fixed bug when `ON CLUSTER` queries may hang forever for non-leader `ReplicatedMergeTreeTables`. [#17089](https://github.com/ClickHouse/ClickHouse/pull/17089) ([alesapin](https://github.com/alesapin)).
+* Fixed fuzzer-found bug in funciton `fuzzBits`. This fixes [#16980](https://github.com/ClickHouse/ClickHouse/issues/16980). [#17051](https://github.com/ClickHouse/ClickHouse/pull/17051) ([hexiaoting](https://github.com/hexiaoting)).
+* Avoid unnecessary network errors for remote queries which may be cancelled while execution, like queries with `LIMIT`. [#17006](https://github.com/ClickHouse/ClickHouse/pull/17006) ([Azat Khuzhin](https://github.com/azat)).
+* Fixed wrong result in big integers (128, 256 bit) when casting from double. [#16986](https://github.com/ClickHouse/ClickHouse/pull/16986) ([Mike](https://github.com/myrrc)).
+* Reresolve the IP of the `format_avro_schema_registry_url` in case of errors. [#16985](https://github.com/ClickHouse/ClickHouse/pull/16985) ([filimonov](https://github.com/filimonov)).
+* Fixed possible server crash after `ALTER TABLE ... MODIFY COLUMN ... NewType` when `SELECT` have `WHERE` expression on altering column and alter doesn't finished yet. [#16968](https://github.com/ClickHouse/ClickHouse/pull/16968) ([Amos Bird](https://github.com/amosbird)).
+* Blame info was not calculated correctly in `clickhouse-git-import`. [#16959](https://github.com/ClickHouse/ClickHouse/pull/16959) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed order by optimization with monotonous functions. Fixes [#16107](https://github.com/ClickHouse/ClickHouse/issues/16107). [#16956](https://github.com/ClickHouse/ClickHouse/pull/16956) ([Anton Popov](https://github.com/CurtizJ)).
+* Fixed optimization of group by with enabled setting `optimize_aggregators_of_group_by_keys` and joins. This fixes [#12604](https://github.com/ClickHouse/ClickHouse/issues/12604). [#16951](https://github.com/ClickHouse/ClickHouse/pull/16951) ([Anton Popov](https://github.com/CurtizJ)).
+* Install script should always create subdirs in config folders. This is only relevant for Docker build with custom config. [#16936](https://github.com/ClickHouse/ClickHouse/pull/16936) ([filimonov](https://github.com/filimonov)).
+* Fixed possible error `Illegal type of argument` for queries with `ORDER BY`. This fixes [#16580](https://github.com/ClickHouse/ClickHouse/issues/16580). [#16928](https://github.com/ClickHouse/ClickHouse/pull/16928) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Abort multipart upload if no data was written to WriteBufferFromS3. [#16840](https://github.com/ClickHouse/ClickHouse/pull/16840) ([Pavel Kovalenko](https://github.com/Jokser)).
+* Fixed crash when using `any` without any arguments. This fixes [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803). [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
+* Fixed the behaviour when ClickHouse used to always return 0 insted of a number of affected rows for `INSERT` queries via MySQL protocol. This fixes [#16605](https://github.com/ClickHouse/ClickHouse/issues/16605). [#16715](https://github.com/ClickHouse/ClickHouse/pull/16715) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed uncontrolled growth of TDigest. [#16680](https://github.com/ClickHouse/ClickHouse/pull/16680) ([hrissan](https://github.com/hrissan)).
+* Fixed remote query failure when using suffix `if` in Aggregate function. This fixes [#16574](https://github.com/ClickHouse/ClickHouse/issues/16574) fixes [#16231](https://github.com/ClickHouse/ClickHouse/issues/16231) [#16610](https://github.com/ClickHouse/ClickHouse/pull/16610) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed inconsistent behavior caused by `select_sequential_consistency` for optimized trivial count query and system.tables. [#16309](https://github.com/ClickHouse/ClickHouse/pull/16309) ([Hao Chen](https://github.com/haoch)).
+* Throw error when use ColumnTransformer replace non exist column. [#16183](https://github.com/ClickHouse/ClickHouse/pull/16183) ([hexiaoting](https://github.com/hexiaoting)).
 
 
 ### ClickHouse release v20.11.3.3-stable, 2020-11-13

From 407a3bfcb9f1e00ce95f47ba49358ffec4e1528d Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 04:23:10 +0300
Subject: [PATCH 369/504] Update CHANGELOG.md

---
 CHANGELOG.md | 40 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index f9bbda56e40..40dac3468f3 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -312,6 +312,46 @@
 
 ## ClickHouse release 20.10
 
+### ClickHouse release v20.10.7.4-stable, 2020-12-24
+                                                                                                                                                                                                                                                              
+#### Bug Fix                                                                                                                                                                                                                                                  
+                                                                                                                                                                                                                                                              
+* Fixed issue when `clickhouse-odbc-bridge` process is unreachable by server on machines with dual `IPv4/IPv6` stack and fixed issue when ODBC dictionary updates are performed using malformed queries and/or cause crashes. This possibly closes [#14489](https://github.com/ClickHouse/ClickHouse/issues/14489). [#18278](https://github.com/ClickHouse/ClickHouse/pull/18278) ([Denis Glazachev](https://github.com/traceon)).                                                                                                
+* Fix key comparison between Enum and Int types. This fixes [#17989](https://github.com/ClickHouse/ClickHouse/issues/17989). [#18214](https://github.com/ClickHouse/ClickHouse/pull/18214) ([Amos Bird](https://github.com/amosbird)).                        
+* Fixed unique key convert crash in `MaterializeMySQL` database engine. This fixes [#18186](https://github.com/ClickHouse/ClickHouse/issues/18186) and fixes [#16372](https://github.com/ClickHouse/ClickHouse/issues/16372)  [#18211](https://github.com/ClickHouse/ClickHouse/pull/18211) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed `std::out_of_range: basic_string` in S3 URL parsing. [#18059](https://github.com/ClickHouse/ClickHouse/pull/18059) ([Vladimir Chebotarev](https://github.com/excitoon)).
+* Fixed the issue when some tables not synchronized to ClickHouse from MySQL caused by the fact that convertion MySQL prefix index wasn't supported for MaterializeMySQL. This fixes [#15187](https://github.com/ClickHouse/ClickHouse/issues/15187) and fixes [#17912](https://github.com/ClickHouse/ClickHouse/issues/17912)  [#17944](https://github.com/ClickHouse/ClickHouse/pull/17944) ([Winter Zhang](https://github.com/zhang2014)).                                                                                                                                                                                
+* Fix possible segfault in `topK` aggregate function. This closes [#17404](https://github.com/ClickHouse/ClickHouse/issues/17404). [#17845](https://github.com/ClickHouse/ClickHouse/pull/17845) ([Maksim Kita](https://github.com/kitaisreal)).              
+* Do not restore parts from `WAL` if `in_memory_parts_enable_wal` is disabled. [#17802](https://github.com/ClickHouse/ClickHouse/pull/17802) ([detailyang](https://github.com/detailyang)).                                                                     
+* Fixed problem when ClickHouse fails to resume connection to MySQL servers. [#17681](https://github.com/ClickHouse/ClickHouse/pull/17681) ([Alexander Kazakov](https://github.com/Akazz)).                                                                   
+* Fixed empty `system.stack_trace` table when server is running in daemon mode. [#17630](https://github.com/ClickHouse/ClickHouse/pull/17630) ([Amos Bird](https://github.com/amosbird)).                                                                       
+* Fixed the behaviour when `clickhouse-client` is used in interactive mode with multiline queries and single line comment was erronously extended till the end of query. This fixes [#13654](https://github.com/ClickHouse/ClickHouse/issues/13654). [#17565](https://github.com/ClickHouse/ClickHouse/pull/17565) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed the issue when server can stop accepting connections in very rare cases. [#17542](https://github.com/ClickHouse/ClickHouse/pull/17542) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed `ALTER` query hang when the corresponding mutation was killed on the different replica. This fixes [#16953](https://github.com/ClickHouse/ClickHouse/issues/16953). [#17499](https://github.com/ClickHouse/ClickHouse/pull/17499) ([alesapin](https://github.com/alesapin)).
+* Fixed bug when mark cache size was underestimated by clickhouse. It may happen when there are a lot of tiny files with marks. [#17496](https://github.com/ClickHouse/ClickHouse/pull/17496) ([alesapin](https://github.com/alesapin)).
+* Fixed `ORDER BY` with enabled setting `optimize_redundant_functions_in_order_by`. [#17471](https://github.com/ClickHouse/ClickHouse/pull/17471) ([Anton Popov](https://github.com/CurtizJ)).
+* Fixed duplicates after `DISTINCT` which were possible because of incorrect optimization. Fixes [#17294](https://github.com/ClickHouse/ClickHouse/issues/17294). [#17296](https://github.com/ClickHouse/ClickHouse/pull/17296) ([li chengxiang](https://github.com/chengxianglibra)). [#17439](https://github.com/ClickHouse/ClickHouse/pull/17439) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fixed crash while reading from `JOIN` table with `LowCardinality` types. This fixes [#17228](https://github.com/ClickHouse/ClickHouse/issues/17228). [#17397](https://github.com/ClickHouse/ClickHouse/pull/17397) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Fixed set index invalidation when there are const columns in the subquery. This fixes [#17246](https://github.com/ClickHouse/ClickHouse/issues/17246) . [#17249](https://github.com/ClickHouse/ClickHouse/pull/17249) ([Amos Bird](https://github.com/amosbird)).
+* Fixed `ColumnConst` comparison which leads to crash. This fixed [#17088](https://github.com/ClickHouse/ClickHouse/issues/17088) . [#17135](https://github.com/ClickHouse/ClickHouse/pull/17135) ([Amos Bird](https://github.com/amosbird)).
+* Fixed bug when `ON CLUSTER` queries may hang forever for non-leader `ReplicatedMergeTreeTables`. [#17089](https://github.com/ClickHouse/ClickHouse/pull/17089) ([alesapin](https://github.com/alesapin)).
+* Fixed fuzzer-found bug in function `fuzzBits`. This fixes [#16980](https://github.com/ClickHouse/ClickHouse/issues/16980). [#17051](https://github.com/ClickHouse/ClickHouse/pull/17051) ([hexiaoting](https://github.com/hexiaoting)).
+* Avoid unnecessary network errors for remote queries which may be cancelled while execution, like queries with `LIMIT`. [#17006](https://github.com/ClickHouse/ClickHouse/pull/17006) ([Azat Khuzhin](https://github.com/azat)).
+* Fixed wrong result in big integers (128, 256 bit) when casting from double. [#16986](https://github.com/ClickHouse/ClickHouse/pull/16986) ([Mike](https://github.com/myrrc)).
+* Reresolve the IP of the `format_avro_schema_registry_url` in case of errors. [#16985](https://github.com/ClickHouse/ClickHouse/pull/16985) ([filimonov](https://github.com/filimonov)).
+* Fixed possible server crash after `ALTER TABLE ... MODIFY COLUMN ... NewType` when `SELECT` have `WHERE` expression on altering column and alter doesn't finished yet. [#16968](https://github.com/ClickHouse/ClickHouse/pull/16968) ([Amos Bird](https://github.com/amosbird)).
+* Blame info was not calculated correctly in `clickhouse-git-import`. [#16959](https://github.com/ClickHouse/ClickHouse/pull/16959) ([alexey-milovidov](https://github.com/alexey-milovidov)).
+* Fixed order by optimization with monotonous functions. This fixes [#16107](https://github.com/ClickHouse/ClickHouse/issues/16107). [#16956](https://github.com/ClickHouse/ClickHouse/pull/16956) ([Anton Popov](https://github.com/CurtizJ)).
+* Fixrf optimization of group by with enabled setting `optimize_aggregators_of_group_by_keys` and joins. This fixes [#12604](https://github.com/ClickHouse/ClickHouse/issues/12604). [#16951](https://github.com/ClickHouse/ClickHouse/pull/16951) ([Anton Popov](https://github.com/CurtizJ)).
+* Install script should always create subdirs in config folders. This is only relevant for Docker build with custom config. [#16936](https://github.com/ClickHouse/ClickHouse/pull/16936) ([filimonov](https://github.com/filimonov)).
+* Fixrf possible error `Illegal type of argument` for queries with `ORDER BY`. This fixes [#16580](https://github.com/ClickHouse/ClickHouse/issues/16580). [#16928](https://github.com/ClickHouse/ClickHouse/pull/16928) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Abort multipart upload if no data was written to `WriteBufferFromS3`. [#16840](https://github.com/ClickHouse/ClickHouse/pull/16840) ([Pavel Kovalenko](https://github.com/Jokser)).
+* Fixed crash when using `any` without any arguments. This fixes [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803). [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
+* Fixed the behaviour when ClickHouse used to always return 0 insted of a number of affected rows for `INSERT` queries via MySQL protocol. This fixes [#16605](https://github.com/ClickHouse/ClickHouse/issues/16605). [#16715](https://github.com/ClickHouse/ClickHouse/pull/16715) ([Winter Zhang](https://github.com/zhang2014)).
+* Fixed uncontrolled growth of `TDigest`. [#16680](https://github.com/ClickHouse/ClickHouse/pull/16680) ([hrissan](https://github.com/hrissan)).
+* Fixed remote query failure when using suffix `if` in Aggregate function. This fixes [#16574](https://github.com/ClickHouse/ClickHouse/issues/16574) fixes [#16231](https://github.com/ClickHouse/ClickHouse/issues/16231) [#16610](https://github.com/ClickHouse/ClickHouse/pull/16610) ([Winter Zhang](https://github.com/zhang2014)).
+
+
 ### ClickHouse release v20.10.4.1-stable, 2020-11-13
 
 #### Bug Fix

From 9cf67db4bbf1e22a2e4f81b295a878599266befa Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 04:33:41 +0300
Subject: [PATCH 370/504] Update CHANGELOG.md

---
 CHANGELOG.md | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 40dac3468f3..d61aeb84b7f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -739,6 +739,31 @@
 
 ## ClickHouse release 20.8
 
+### ClickHouse release v20.8.10.13-lts, 2020-12-24
+
+#### Bug Fix
+
+* When server log rotation was configured using `logger.size` parameter with numeric value larger than 2^32, the logs were not rotated properly. [#17905](https://github.com/ClickHouse/ClickHouse/pull/17905) ([Alexander Kuzmenkov](https://github.com/akuzm)).
+* Fixed incorrect initialization of `max_compress_block_size` in MergeTreeWriterSettings with `min_compress_block_size`. [#17833](https://github.com/ClickHouse/ClickHouse/pull/17833) ([flynn](https://github.com/ucasFL)).
+* Fixed problem when ClickHouse fails to resume connection to MySQL servers. [#17681](https://github.com/ClickHouse/ClickHouse/pull/17681) ([Alexander Kazakov](https://github.com/Akazz)).
+* Fixed `ALTER` query hang when the corresponding mutation was killed on the different replica. This fixes [#16953](https://github.com/ClickHouse/ClickHouse/issues/16953). [#17499](https://github.com/ClickHouse/ClickHouse/pull/17499) ([alesapin](https://github.com/alesapin)).
+* Fixed a bug when mark cache size was underestimated by ClickHouse. It may happen when there are a lot of tiny files with marks. [#17496](https://github.com/ClickHouse/ClickHouse/pull/17496) ([alesapin](https://github.com/alesapin)).
+* Fixed `ORDER BY` with enabled setting `optimize_redundant_functions_in_order_by`. [#17471](https://github.com/ClickHouse/ClickHouse/pull/17471) ([Anton Popov](https://github.com/CurtizJ)).
+* Fixed `ColumnConst` comparison which leads to crash. This fixed [#17088](https://github.com/ClickHouse/ClickHouse/issues/17088) . [#17135](https://github.com/ClickHouse/ClickHouse/pull/17135) ([Amos Bird](https://github.com/amosbird)).
+* Fixed bug when `ON CLUSTER` queries may hang forever for non-leader ReplicatedMergeTreeTables. [#17089](https://github.com/ClickHouse/ClickHouse/pull/17089) ([alesapin](https://github.com/alesapin)).
+* Avoid unnecessary network errors for remote queries which may be cancelled while execution, like queries with `LIMIT`. [#17006](https://github.com/ClickHouse/ClickHouse/pull/17006) ([Azat Khuzhin](https://github.com/azat)).
+* Reresolve the IP of the `format_avro_schema_registry_url` in case of errors. [#16985](https://github.com/ClickHouse/ClickHouse/pull/16985) ([filimonov](https://github.com/filimonov)).
+* Fixed possible server crash after `ALTER TABLE ... MODIFY COLUMN ... NewType` when `SELECT` have `WHERE` expression on altering column and alter doesn't finished yet. [#16968](https://github.com/ClickHouse/ClickHouse/pull/16968) ([Amos Bird](https://github.com/amosbird)).
+* Install script should always create subdirs in config folders. This is only relevant for Docker build with custom config. [#16936](https://github.com/ClickHouse/ClickHouse/pull/16936) ([filimonov](https://github.com/filimonov)).
+* Fixed possible error `Illegal type of argument` for queries with `ORDER BY`. Fixes [#16580](https://github.com/ClickHouse/ClickHouse/issues/16580). [#16928](https://github.com/ClickHouse/ClickHouse/pull/16928) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
+* Abort multipart upload if no data was written to WriteBufferFromS3. [#16840](https://github.com/ClickHouse/ClickHouse/pull/16840) ([Pavel Kovalenko](https://github.com/Jokser)).
+* Fixed crash when using `any` without any arguments. This is for [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803) . cc @azat. [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
+* Fixed `IN` operator over several columns and tuples with enabled `transform_null_in` setting. Fixes [#15310](https://github.com/ClickHouse/ClickHouse/issues/15310). [#16722](https://github.com/ClickHouse/ClickHouse/pull/16722) ([Anton Popov](https://github.com/CurtizJ)).
+* Fixed inconsistent behaviour of `optimize_read_in_order/optimize_aggregation_in_order` with max_threads > 0 and expression in ORDER BY. [#16637](https://github.com/ClickHouse/ClickHouse/pull/16637) ([Azat Khuzhin](https://github.com/azat)).
+* Fixed the issue when query optimization was producing wrong result if query contains `ARRAY JOIN`. [#17887](https://github.com/ClickHouse/ClickHouse/pull/17887) ([sundyli](https://github.com/sundy-li)).
+* Query is finished faster in case of exception. Cancel execution on remote replicas if exception happens. [#15578](https://github.com/ClickHouse/ClickHouse/pull/15578) ([Azat Khuzhin](https://github.com/azat)).
+
+
 ### ClickHouse release v20.8.6.6-lts, 2020-11-13
 
 #### Bug Fix

From 90ee0188d07d65685224ba66923c211c198ba04a Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Thu, 24 Dec 2020 04:34:34 +0300
Subject: [PATCH 371/504] Update CHANGELOG.md

---
 CHANGELOG.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index d61aeb84b7f..2ec48bcd584 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -757,7 +757,7 @@
 * Install script should always create subdirs in config folders. This is only relevant for Docker build with custom config. [#16936](https://github.com/ClickHouse/ClickHouse/pull/16936) ([filimonov](https://github.com/filimonov)).
 * Fixed possible error `Illegal type of argument` for queries with `ORDER BY`. Fixes [#16580](https://github.com/ClickHouse/ClickHouse/issues/16580). [#16928](https://github.com/ClickHouse/ClickHouse/pull/16928) ([Nikolai Kochetov](https://github.com/KochetovNicolai)).
 * Abort multipart upload if no data was written to WriteBufferFromS3. [#16840](https://github.com/ClickHouse/ClickHouse/pull/16840) ([Pavel Kovalenko](https://github.com/Jokser)).
-* Fixed crash when using `any` without any arguments. This is for [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803) . cc @azat. [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
+* Fixed crash when using `any` without any arguments. This fixes [#16803](https://github.com/ClickHouse/ClickHouse/issues/16803). [#16826](https://github.com/ClickHouse/ClickHouse/pull/16826) ([Amos Bird](https://github.com/amosbird)).
 * Fixed `IN` operator over several columns and tuples with enabled `transform_null_in` setting. Fixes [#15310](https://github.com/ClickHouse/ClickHouse/issues/15310). [#16722](https://github.com/ClickHouse/ClickHouse/pull/16722) ([Anton Popov](https://github.com/CurtizJ)).
 * Fixed inconsistent behaviour of `optimize_read_in_order/optimize_aggregation_in_order` with max_threads > 0 and expression in ORDER BY. [#16637](https://github.com/ClickHouse/ClickHouse/pull/16637) ([Azat Khuzhin](https://github.com/azat)).
 * Fixed the issue when query optimization was producing wrong result if query contains `ARRAY JOIN`. [#17887](https://github.com/ClickHouse/ClickHouse/pull/17887) ([sundyli](https://github.com/sundy-li)).

From f39c9613284e976d8716d7962e3f3ecce58c735e Mon Sep 17 00:00:00 2001
From: vladimir golovchenko <golova77@hotmail.com>
Date: Wed, 23 Dec 2020 17:18:07 -0800
Subject: [PATCH 372/504] Added tests to cover cases with 12-hour clock time.

---
 ...1_parse_date_time_best_effort_us.reference |  4 +++
 .../01351_parse_date_time_best_effort_us.sql  |  4 +++
 .../01442_date_time_with_params.reference     | 30 ++++++++++++++++++
 .../01442_date_time_with_params.sql           | 31 ++++++++++++++++++-
 ..._besteffort_or_null_empty_string.reference |  4 +++
 ...tetime_besteffort_or_null_empty_string.sql |  4 +++
 6 files changed, 76 insertions(+), 1 deletion(-)

diff --git a/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.reference b/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.reference
index 22a778b715a..9d0a11776b9 100644
--- a/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.reference
+++ b/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.reference
@@ -7,6 +7,10 @@
  01/02/2017 03:04:05+300     2017-01-02 00:04:05 
  01.02.2017 03:04:05GMT      2017-01-02 03:04:05 
  01-02-2017 03:04:05 MSD     2017-01-01 23:04:05 
+ 01-02-2017 11:04:05 AM      2017-01-02 11:04:05 
+ 01-02-2017 11:04:05 PM      2017-01-02 23:04:05 
+ 01-02-2017 12:04:05 AM      2017-01-02 00:04:05 
+ 01-02-2017 12:04:05 PM      2017-01-02 12:04:05 
  01.02.17 03:04:05 MSD Feb   2017-02-01 23:04:05 
  01/02/2017 03:04:05 MSK     2017-01-02 00:04:05 
  12/13/2019                  2019-12-13 00:00:00 
diff --git a/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.sql b/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.sql
index 0c33b2de49f..3345ace8305 100644
--- a/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.sql
+++ b/tests/queries/0_stateless/01351_parse_date_time_best_effort_us.sql
@@ -11,6 +11,10 @@ FROM
 '01/02/2017 03:04:05+300',
 '01.02.2017 03:04:05GMT',
 '01-02-2017 03:04:05 MSD',
+'01-02-2017 11:04:05 AM',
+'01-02-2017 11:04:05 PM',
+'01-02-2017 12:04:05 AM',
+'01-02-2017 12:04:05 PM',
 '01.02.17 03:04:05 MSD Feb',
 '01/02/2017 03:04:05 MSK',
 '12/13/2019',
diff --git a/tests/queries/0_stateless/01442_date_time_with_params.reference b/tests/queries/0_stateless/01442_date_time_with_params.reference
index 94ffcffdd37..19f78c83f82 100644
--- a/tests/queries/0_stateless/01442_date_time_with_params.reference
+++ b/tests/queries/0_stateless/01442_date_time_with_params.reference
@@ -2,42 +2,72 @@
 2020-01-01 00:00:00	DateTime	2020-01-01 00:02:00.11	DateTime64(2)	2020-01-01 00:03:00	DateTime(\'Europe/Moscow\')	2020-01-01 00:04:00.220	DateTime64(3, \'Europe/Moscow\')	2020-01-01 00:05:00	DateTime
 2020-01-01 00:00:00	DateTime	2020-01-01 00:02:00.11	DateTime64(2)	2020-01-01 00:03:00	DateTime(\'Europe/Moscow\')	2020-01-01 00:04:00.220	DateTime64(3, \'Europe/Moscow\')	2020-01-01 00:05:00	DateTime
 2020-01-01 00:00:00	DateTime
+parseDateTimeBestEffort
 2020-05-14 03:37:03.000	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 11:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 23:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 00:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 12:37:03.000	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.253	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.253	DateTime64(3, \'UTC\')
 2020-05-14 06:37:03.253	DateTime64(3, \'Europe/Minsk\')
 2020-05-14 03:37:03.253	DateTime64(3, \'UTC\')
+parseDateTimeBestEffortOrNull
 \N	Nullable(DateTime64(3))
 2020-05-14 03:37:03.000	Nullable(DateTime64(3, \'UTC\'))
 2020-05-14 03:37:03.000	Nullable(DateTime64(3, \'UTC\'))
+2020-05-14 11:37:03.000	Nullable(DateTime64(3, \'UTC\'))
+2020-05-14 23:37:03.000	Nullable(DateTime64(3, \'UTC\'))
+2020-05-14 00:37:03.000	Nullable(DateTime64(3, \'UTC\'))
+2020-05-14 12:37:03.000	Nullable(DateTime64(3, \'UTC\'))
 2020-05-14 03:37:03.253	Nullable(DateTime64(3, \'UTC\'))
 2020-05-14 03:37:03.253	Nullable(DateTime64(3, \'UTC\'))
 2020-05-14 06:37:03.253	Nullable(DateTime64(3, \'Europe/Minsk\'))
 2020-05-14 03:37:03.253	Nullable(DateTime64(3, \'UTC\'))
+parseDateTimeBestEffortOrZero
 1970-01-01 00:00:00.000	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.000	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 11:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 23:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 00:37:03.000	DateTime64(3, \'UTC\')
+2020-05-14 12:37:03.000	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.253	DateTime64(3, \'UTC\')
 2020-05-14 03:37:03.253	DateTime64(3, \'UTC\')
 2020-05-14 06:37:03.253	DateTime64(3, \'Europe/Minsk\')
 2020-05-14 03:37:03.253	DateTime64(3, \'UTC\')
+parseDateTime32BestEffort
 2020-05-14 03:37:03	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
+2020-05-14 11:37:03	DateTime(\'UTC\')
+2020-05-14 23:37:03	DateTime(\'UTC\')
+2020-05-14 00:37:03	DateTime(\'UTC\')
+2020-05-14 12:37:03	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
 2020-05-14 06:37:03	DateTime(\'Europe/Minsk\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
+parseDateTime32BestEffortOrNull
 \N	Nullable(DateTime)
 2020-05-14 03:37:03	Nullable(DateTime(\'UTC\'))
 2020-05-14 03:37:03	Nullable(DateTime(\'UTC\'))
+2020-05-14 11:37:03	Nullable(DateTime(\'UTC\'))
+2020-05-14 23:37:03	Nullable(DateTime(\'UTC\'))
+2020-05-14 00:37:03	Nullable(DateTime(\'UTC\'))
+2020-05-14 12:37:03	Nullable(DateTime(\'UTC\'))
 2020-05-14 03:37:03	Nullable(DateTime(\'UTC\'))
 2020-05-14 03:37:03	Nullable(DateTime(\'UTC\'))
 2020-05-14 06:37:03	Nullable(DateTime(\'Europe/Minsk\'))
 2020-05-14 03:37:03	Nullable(DateTime(\'UTC\'))
+parseDateTime32BestEffortOrZero
 1970-01-01 00:00:00	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
+2020-05-14 11:37:03	DateTime(\'UTC\')
+2020-05-14 23:37:03	DateTime(\'UTC\')
+2020-05-14 00:37:03	DateTime(\'UTC\')
+2020-05-14 12:37:03	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
 2020-05-14 03:37:03	DateTime(\'UTC\')
 2020-05-14 06:37:03	DateTime(\'Europe/Minsk\')
diff --git a/tests/queries/0_stateless/01442_date_time_with_params.sql b/tests/queries/0_stateless/01442_date_time_with_params.sql
index 2eefa47ba8c..52815460245 100644
--- a/tests/queries/0_stateless/01442_date_time_with_params.sql
+++ b/tests/queries/0_stateless/01442_date_time_with_params.sql
@@ -12,50 +12,79 @@ SELECT CAST('2020-01-01 00:00:00', 'DateTime') AS a, toTypeName(a), CAST('2020-0
 
 SELECT toDateTime32('2020-01-01 00:00:00') AS a, toTypeName(a);
 
+SELECT 'parseDateTimeBestEffort';
 SELECT parseDateTimeBestEffort('<Empty>', 3) AS a, toTypeName(a); -- {serverError 41}
 SELECT parseDateTimeBestEffort('2020-05-14T03:37:03', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffort('2020-05-14 03:37:03', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffort('2020-05-14 11:37:03 AM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffort('2020-05-14 11:37:03 PM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffort('2020-05-14 12:37:03 AM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffort('2020-05-14 12:37:03 PM', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffort('2020-05-14T03:37:03.253184', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffort('2020-05-14T03:37:03.253184Z', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffort('2020-05-14T03:37:03.253184Z', 3, 'Europe/Minsk') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffort(materialize('2020-05-14T03:37:03.253184Z'), 3, 'UTC') AS a, toTypeName(a);
 
+SELECT 'parseDateTimeBestEffortOrNull';
 SELECT parseDateTimeBestEffortOrNull('<Empty>', 3) AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrNull('2020-05-14T03:37:03', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrNull('2020-05-14 03:37:03', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrNull('2020-05-14 11:37:03 AM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrNull('2020-05-14 11:37:03 PM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrNull('2020-05-14 12:37:03 AM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrNull('2020-05-14 12:37:03 PM', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrNull('2020-05-14T03:37:03.253184', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrNull('2020-05-14T03:37:03.253184Z', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrNull('2020-05-14T03:37:03.253184Z', 3, 'Europe/Minsk') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrNull(materialize('2020-05-14T03:37:03.253184Z'), 3, 'UTC') AS a, toTypeName(a);
 
+SELECT 'parseDateTimeBestEffortOrZero';
 SELECT parseDateTimeBestEffortOrZero('<Empty>', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrZero('2020-05-14T03:37:03', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrZero('2020-05-14 03:37:03', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrZero('2020-05-14 11:37:03 AM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrZero('2020-05-14 11:37:03 PM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrZero('2020-05-14 12:37:03 AM', 3, 'UTC') AS a, toTypeName(a);
+SELECT parseDateTimeBestEffortOrZero('2020-05-14 12:37:03 PM', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrZero('2020-05-14T03:37:03.253184', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrZero('2020-05-14T03:37:03.253184Z', 3, 'UTC') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrZero('2020-05-14T03:37:03.253184Z', 3, 'Europe/Minsk') AS a, toTypeName(a);
 SELECT parseDateTimeBestEffortOrZero(materialize('2020-05-14T03:37:03.253184Z'), 3, 'UTC') AS a, toTypeName(a);
 
-
+SELECT 'parseDateTime32BestEffort';
 SELECT parseDateTime32BestEffort('<Empty>') AS a, toTypeName(a); -- {serverError 41}
 SELECT parseDateTime32BestEffort('2020-05-14T03:37:03', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffort('2020-05-14 03:37:03', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffort('2020-05-14 11:37:03 AM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffort('2020-05-14 11:37:03 PM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffort('2020-05-14 12:37:03 AM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffort('2020-05-14 12:37:03 PM', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffort('2020-05-14T03:37:03.253184', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffort('2020-05-14T03:37:03.253184Z', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffort('2020-05-14T03:37:03.253184Z', 'Europe/Minsk') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffort(materialize('2020-05-14T03:37:03.253184Z'), 'UTC') AS a, toTypeName(a);
 
+SELECT 'parseDateTime32BestEffortOrNull';
 SELECT parseDateTime32BestEffortOrNull('<Empty>') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrNull('2020-05-14T03:37:03', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrNull('2020-05-14 03:37:03', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrNull('2020-05-14 11:37:03 AM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrNull('2020-05-14 11:37:03 PM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrNull('2020-05-14 12:37:03 AM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrNull('2020-05-14 12:37:03 PM', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrNull('2020-05-14T03:37:03.253184', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrNull('2020-05-14T03:37:03.253184Z', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrNull('2020-05-14T03:37:03.253184Z', 'Europe/Minsk') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrNull(materialize('2020-05-14T03:37:03.253184Z'), 'UTC') AS a, toTypeName(a);
 
+SELECT 'parseDateTime32BestEffortOrZero';
 SELECT parseDateTime32BestEffortOrZero('<Empty>', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrZero('2020-05-14T03:37:03', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrZero('2020-05-14 03:37:03', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrZero('2020-05-14 11:37:03 AM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrZero('2020-05-14 11:37:03 PM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrZero('2020-05-14 12:37:03 AM', 'UTC') AS a, toTypeName(a);
+SELECT parseDateTime32BestEffortOrZero('2020-05-14 12:37:03 PM', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrZero('2020-05-14T03:37:03.253184', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrZero('2020-05-14T03:37:03.253184Z', 'UTC') AS a, toTypeName(a);
 SELECT parseDateTime32BestEffortOrZero('2020-05-14T03:37:03.253184Z', 'Europe/Minsk') AS a, toTypeName(a);
diff --git a/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.reference b/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.reference
index 14be7c77e23..675d09bdccf 100644
--- a/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.reference
+++ b/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.reference
@@ -1,5 +1,9 @@
 2010-01-01 00:00:00
 2010-01-01 01:01:01
+2020-01-01 11:01:01
+2020-01-01 23:01:01
+2020-01-01 00:01:01
+2020-01-01 12:01:01
 2000-01-01 01:01:01
 \N
 \N
diff --git a/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.sql b/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.sql
index 66d28534b30..ad14c4ede06 100644
--- a/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.sql
+++ b/tests/queries/0_stateless/01543_parse_datetime_besteffort_or_null_empty_string.sql
@@ -1,5 +1,9 @@
 SELECT parseDateTimeBestEffortOrNull('2010-01-01');
 SELECT parseDateTimeBestEffortOrNull('2010-01-01 01:01:01');
+SELECT parseDateTimeBestEffortOrNull('2020-01-01 11:01:01 am');
+SELECT parseDateTimeBestEffortOrNull('2020-01-01 11:01:01 pm');
+SELECT parseDateTimeBestEffortOrNull('2020-01-01 12:01:01 am');
+SELECT parseDateTimeBestEffortOrNull('2020-01-01 12:01:01 pm');
 SELECT parseDateTimeBestEffortOrNull('01:01:01');
 SELECT parseDateTimeBestEffortOrNull('20100');
 SELECT parseDateTimeBestEffortOrNull('0100:0100:0000');

From ed4fe206aae2e2df5f3e5ff838ff0c33d88a3c55 Mon Sep 17 00:00:00 2001
From: vladimir golovchenko <golova77@hotmail.com>
Date: Wed, 23 Dec 2020 18:08:21 -0800
Subject: [PATCH 373/504] Fixed the parseDateTimeBestEffort*-functions to
 correctly handle '12 am' time.

---
 src/IO/parseDateTimeBestEffort.cpp | 5 +++++
 src/IO/parseDateTimeBestEffort.h   | 2 +-
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/src/IO/parseDateTimeBestEffort.cpp b/src/IO/parseDateTimeBestEffort.cpp
index 063955cdd1e..a3ea418dd9f 100644
--- a/src/IO/parseDateTimeBestEffort.cpp
+++ b/src/IO/parseDateTimeBestEffort.cpp
@@ -119,6 +119,7 @@ ReturnType parseDateTimeBestEffortImpl(
     UInt8 time_zone_offset_hour = 0;
     UInt8 time_zone_offset_minute = 0;
 
+    bool is_am = false;
     bool is_pm = false;
 
     auto read_alpha_month = [&month] (const auto & alpha)
@@ -486,6 +487,7 @@ ReturnType parseDateTimeBestEffortImpl(
                     {
                         if (alpha[0] == 'A' || alpha[0] == 'a')
                         {
+                            is_am = true;
                         }
                         else if (alpha[0] == 'P' || alpha[0] == 'p')
                         {
@@ -560,6 +562,9 @@ ReturnType parseDateTimeBestEffortImpl(
     if (!check_date(is_leap_year, month, day_of_month))
         return on_error("Cannot read DateTime: unexpected date: " + std::to_string(year) + "-" + std::to_string(month) + "-" + std::to_string(day_of_month), ErrorCodes::CANNOT_PARSE_DATETIME);
 
+    if (is_am && hour == 12)
+        hour = 0;
+
     if (is_pm && hour < 12)
         hour += 12;
 
diff --git a/src/IO/parseDateTimeBestEffort.h b/src/IO/parseDateTimeBestEffort.h
index fcc0195c4d9..61de31a0213 100644
--- a/src/IO/parseDateTimeBestEffort.h
+++ b/src/IO/parseDateTimeBestEffort.h
@@ -50,7 +50,7 @@ class ReadBuffer;
   *
   * single whitespace can be used as a separator
   *
-  * AM/PM - AM is ignored and PM means: add 12 hours if value is less than 12.
+  * AM/PM - AM means: subtract 12 hours if a value is 12 and PM means: add 12 hours if a value is less than 12.
   *
   * Jan/Feb/Mar/Apr/May/Jun/Jul/Aug/Sep/Oct/Nov/Dec - allowed to specify month
   * Mon/Tue/Wed/Thu/Fri/Sat/Sun - simply ignored.

From 844239b1c3af0ad970a05f3d1df83ddfeffca3af Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 07:03:33 +0300
Subject: [PATCH 374/504] tmp fixes (split me)

---
 src/Interpreters/ActionsDAG.cpp               |  3 +
 src/Interpreters/ExpressionAnalyzer.cpp       | 63 +++++++++++++++++--
 src/Interpreters/ExpressionAnalyzer.h         |  2 +
 .../ExtractExpressionInfoVisitor.cpp          | 28 +++++++--
 .../ExtractExpressionInfoVisitor.h            |  3 +-
 src/Interpreters/InterpreterSelectQuery.cpp   |  3 +
 .../PredicateExpressionsOptimizer.cpp         | 14 ++++-
 src/Interpreters/PredicateRewriteVisitor.cpp  |  2 +-
 src/Interpreters/TreeOptimizer.cpp            | 12 ++++
 src/Interpreters/TreeRewriter.cpp             |  2 +
 src/Processors/QueryPlan/AggregatingStep.cpp  |  4 +-
 src/Processors/QueryPlan/AggregatingStep.h    |  4 +-
 src/Processors/QueryPlan/ArrayJoinStep.cpp    |  2 +-
 src/Processors/QueryPlan/ArrayJoinStep.h      |  2 +-
 src/Processors/QueryPlan/CreatingSetsStep.cpp |  4 +-
 src/Processors/QueryPlan/CreatingSetsStep.h   |  4 +-
 src/Processors/QueryPlan/DistinctStep.cpp     |  2 +-
 src/Processors/QueryPlan/DistinctStep.h       |  2 +-
 src/Processors/QueryPlan/ExpressionStep.cpp   |  2 +-
 src/Processors/QueryPlan/ExpressionStep.h     |  2 +-
 src/Processors/QueryPlan/FillingStep.cpp      |  2 +-
 src/Processors/QueryPlan/FillingStep.h        |  2 +-
 src/Processors/QueryPlan/FilterStep.cpp       |  2 +-
 src/Processors/QueryPlan/FilterStep.h         |  2 +-
 .../QueryPlan/FinishSortingStep.cpp           |  2 +-
 src/Processors/QueryPlan/FinishSortingStep.h  |  2 +-
 src/Processors/QueryPlan/IQueryPlanStep.cpp   | 10 ++-
 src/Processors/QueryPlan/IQueryPlanStep.h     | 25 ++++----
 src/Processors/QueryPlan/ISourceStep.cpp      |  2 +-
 src/Processors/QueryPlan/ISourceStep.h        |  2 +-
 .../QueryPlan/ITransformingStep.cpp           |  2 +-
 src/Processors/QueryPlan/ITransformingStep.h  |  2 +-
 src/Processors/QueryPlan/LimitByStep.cpp      |  2 +-
 src/Processors/QueryPlan/LimitByStep.h        |  2 +-
 src/Processors/QueryPlan/LimitStep.cpp        |  2 +-
 src/Processors/QueryPlan/LimitStep.h          |  2 +-
 src/Processors/QueryPlan/MergeSortingStep.cpp |  2 +-
 src/Processors/QueryPlan/MergeSortingStep.h   |  2 +-
 .../QueryPlan/MergingAggregatedStep.cpp       |  2 +-
 .../QueryPlan/MergingAggregatedStep.h         |  2 +-
 src/Processors/QueryPlan/MergingFinal.cpp     |  2 +-
 src/Processors/QueryPlan/MergingFinal.h       |  2 +-
 .../QueryPlan/MergingSortedStep.cpp           |  2 +-
 src/Processors/QueryPlan/MergingSortedStep.h  |  2 +-
 src/Processors/QueryPlan/OffsetStep.cpp       |  2 +-
 src/Processors/QueryPlan/OffsetStep.h         |  2 +-
 .../QueryPlan/PartialSortingStep.cpp          |  2 +-
 src/Processors/QueryPlan/PartialSortingStep.h |  2 +-
 src/Processors/QueryPlan/QueryPlan.cpp        | 23 +++++--
 src/Processors/QueryPlan/QueryPlan.h          |  6 ++
 src/Processors/QueryPlan/TotalsHavingStep.cpp |  2 +-
 src/Processors/QueryPlan/TotalsHavingStep.h   |  2 +-
 src/Processors/QueryPlan/UnionStep.cpp        |  2 +-
 src/Processors/QueryPlan/UnionStep.h          |  2 +-
 src/Processors/QueryPlan/WindowStep.cpp       |  2 +-
 src/Processors/QueryPlan/WindowStep.h         |  2 +-
 src/Processors/Transforms/WindowTransform.cpp |  5 ++
 .../01591_window_functions.reference          | 38 +++++++----
 .../0_stateless/01591_window_functions.sql    | 15 ++++-
 59 files changed, 254 insertions(+), 92 deletions(-)

diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index 2fc78261f17..f9505cc4199 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -691,6 +691,9 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
 
 ActionsDAGPtr ActionsDAG::merge(ActionsDAG && first, ActionsDAG && second)
 {
+    fmt::print(stderr, "actiondag merge first '{}'\n, second '{}\n",
+        first.dumpDAG(), second.dumpDAG());
+
     /// first: x (1), x (2), y ==> x (2), z, x (3)
     /// second: x (1), x (2), x (3) ==> x (3), x (2), x (1)
     /// merge: x (1), x (2), x (3), y =(first)=> x (3), y, x (2), z, x (4) =(second)=> y, z, x (4), x (2), x (3)
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 31c12490408..e271f78b699 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1004,15 +1004,17 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
             col.column = col.type->createColumn();
             col.name = f.column_name;
 
-            step.actions()->addInput(col);
+//            step.actions()->addInput(col);
+            columns_after_window.push_back({f.column_name,
+                f.aggregate_function->getReturnType()});
 
             for (const auto & a : f.function_node->arguments->children)
             {
                 // 2.1) function arguments;
                 step.required_output.push_back(a->getColumnName());
             }
-            // 2.2) function result;
-            step.required_output.push_back(f.column_name);
+//            // 2.2) function result;
+//            step.required_output.push_back(f.column_name);
         }
 
         // 2.3) PARTITION BY and ORDER BY columns.
@@ -1030,6 +1032,8 @@ bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain,
     if (!select_query->having())
         return false;
 
+    fmt::print(stderr, "appendHaving:\n'{}'\n", select_query->dumpTree());
+
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
     getRootActionsForHaving(select_query->having(), only_types, step.actions());
@@ -1048,6 +1052,15 @@ void SelectQueryExpressionAnalyzer::appendSelect(ExpressionActionsChain & chain,
 
     for (const auto & child : select_query->select()->children)
     {
+        if (const auto * function = typeid_cast<const ASTFunction *>(child.get());
+            function
+            && function->is_window_function)
+        {
+            // Skip window function columns here -- they are calculated after
+            // other SELECT expressions by a special step.
+            continue;
+        }
+
         step.required_output.push_back(child->getColumnName());
     }
 }
@@ -1421,9 +1434,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         /// If there is aggregation, we execute expressions in SELECT and ORDER BY on the initiating server, otherwise on the source servers.
         query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
 
-        query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
-
-        selected_columns = chain.getLastStep().required_output;
         has_order_by = query.orderBy() != nullptr;
         before_order_and_select = query_analyzer.appendOrderBy(
                 chain,
@@ -1431,6 +1441,44 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
                 optimize_read_in_order,
                 order_by_elements_actions);
 
+        if (has_window)
+        {
+            fmt::print(stderr, "chain before window args: '{}'\n", chain.dumpChain());
+
+            query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
+
+            fmt::print(stderr, "after window args chain: '{}'\n", chain.dumpChain());
+
+            auto select_required_columns = chain.getLastStep().required_output;
+
+            for (const auto & x : chain.getLastActions()->getNamesAndTypesList())
+            {
+                query_analyzer.columns_after_window.push_back(x);
+            }
+
+            finalize_chain(chain);
+            //auto & window_step = chain.addStep();
+            //window_step.actions_dag.query_analyzer.columns_after_window);
+
+            auto & step = chain.lastStep(query_analyzer.columns_after_window);
+            step.required_output = select_required_columns;
+            const auto * select_query = query_analyzer.getSelectQuery();
+            for (const auto & child : select_query->select()->children)
+            {
+                if (const auto * function
+                        = typeid_cast<const ASTFunction *>(child.get());
+                    function && function->is_window_function)
+                {
+                    // See its counterpart in appendSelect()
+                    step.required_output.push_back(child->getColumnName());
+                }
+            }
+
+            fmt::print(stderr, "chain after window columns: '{}'\n", chain.dumpChain());
+        }
+
+        selected_columns = chain.getLastStep().required_output;
+
         if (query_analyzer.appendLimitBy(chain, only_types || !second_stage))
         {
             before_limit_by = chain.getLastActions();
@@ -1440,8 +1488,11 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         final_projection = query_analyzer.appendProjectResult(chain);
 
         finalize_chain(chain);
+
+        fmt::print(stderr, "final chain: '{}'\n", chain.dumpChain());
     }
 
+
     /// Before executing WHERE and HAVING, remove the extra columns from the block (mostly the aggregation keys).
     removeExtraColumns();
 
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index fb0cb4ea4c3..1095689a775 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -55,6 +55,8 @@ struct ExpressionAnalyzerData
     NamesAndTypesList columns_after_join;
     /// Columns after ARRAY JOIN, JOIN, and/or aggregation.
     NamesAndTypesList aggregated_columns;
+    /// Columns after window functions.
+    NamesAndTypesList columns_after_window;
 
     bool has_aggregation = false;
     NamesAndTypesList aggregation_keys;
diff --git a/src/Interpreters/ExtractExpressionInfoVisitor.cpp b/src/Interpreters/ExtractExpressionInfoVisitor.cpp
index 2d9339447b1..64c23cd4fd1 100644
--- a/src/Interpreters/ExtractExpressionInfoVisitor.cpp
+++ b/src/Interpreters/ExtractExpressionInfoVisitor.cpp
@@ -22,15 +22,22 @@ void ExpressionInfoMatcher::visit(const ASTFunction & ast_function, const ASTPtr
     {
         data.is_array_join = true;
     }
-    // "is_aggregate_function" doesn't mean much by itself. Apparently here it is
-    // used to move filters from HAVING to WHERE, and probably for this purpose
-    // an aggregate function calculated as a window function is not relevant.
+    // "is_aggregate_function" is used to determine whether we can move a filter
+    // (1) from HAVING to WHERE or (2) from WHERE of a parent query to HAVING of
+    // a subquery.
+    // For aggregate functions we can't do (1) but can do (2).
+    // For window functions both don't make sense -- they are not allowed in
+    // WHERE or HAVING.
     else if (!ast_function.is_window_function
         && AggregateFunctionFactory::instance().isAggregateFunctionName(
             ast_function.name))
     {
         data.is_aggregate_function = true;
     }
+    else if (ast_function.is_window_function)
+    {
+        data.is_window_function = true;
+    }
     else
     {
         const auto & function = FunctionFactory::instance().tryGet(ast_function.name, data.context);
@@ -75,15 +82,26 @@ bool ExpressionInfoMatcher::needChildVisit(const ASTPtr & node, const ASTPtr &)
     return !node->as<ASTSubquery>();
 }
 
-bool hasStatefulFunction(const ASTPtr & node, const Context & context)
+bool hasNonRewritableFunction(const ASTPtr & node, const Context & context)
 {
     for (const auto & select_expression : node->children)
     {
         ExpressionInfoVisitor::Data expression_info{.context = context, .tables = {}};
         ExpressionInfoVisitor(expression_info).visit(select_expression);
 
-        if (expression_info.is_stateful_function)
+        if (expression_info.is_stateful_function
+            || expression_info.is_window_function)
+        {
+            // If an outer query has a WHERE on window function, we can't move
+            // it into the subquery, because window functions are not allowed in
+            // WHERE and HAVING. Example:
+            // select * from (
+            //     select number,
+            //          count(*) over (partition by intDiv(number, 3)) c
+            //     from numbers(3)
+            // ) where c > 1;
             return true;
+        }
     }
 
     return false;
diff --git a/src/Interpreters/ExtractExpressionInfoVisitor.h b/src/Interpreters/ExtractExpressionInfoVisitor.h
index 0cb43e5b00a..d05415490e6 100644
--- a/src/Interpreters/ExtractExpressionInfoVisitor.h
+++ b/src/Interpreters/ExtractExpressionInfoVisitor.h
@@ -21,6 +21,7 @@ struct ExpressionInfoMatcher
         bool is_array_join = false;
         bool is_stateful_function = false;
         bool is_aggregate_function = false;
+        bool is_window_function = false;
         bool is_deterministic_function = true;
         std::unordered_set<size_t> unique_reference_tables_pos = {};
     };
@@ -36,6 +37,6 @@ struct ExpressionInfoMatcher
 
 using ExpressionInfoVisitor = ConstInDepthNodeVisitor<ExpressionInfoMatcher, true>;
 
-bool hasStatefulFunction(const ASTPtr & node, const Context & context);
+bool hasNonRewritableFunction(const ASTPtr & node, const Context & context);
 
 }
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 38cc19a00d6..9a6e6b7593b 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -521,6 +521,9 @@ Block InterpreterSelectQuery::getSampleBlockImpl()
             filter_info,
             source_header);
 
+    fmt::print(stderr, "ExpressionAnalysisResult: '{}'\nat\n{}\n",
+        analysis_result.dump(), StackTrace().toString());
+
     if (options.to_stage == QueryProcessingStage::Enum::FetchColumns)
     {
         auto header = source_header;
diff --git a/src/Interpreters/PredicateExpressionsOptimizer.cpp b/src/Interpreters/PredicateExpressionsOptimizer.cpp
index 614693b5ae0..00b47be408a 100644
--- a/src/Interpreters/PredicateExpressionsOptimizer.cpp
+++ b/src/Interpreters/PredicateExpressionsOptimizer.cpp
@@ -90,8 +90,12 @@ std::vector<ASTs> PredicateExpressionsOptimizer::extractTablesPredicates(const A
         ExpressionInfoVisitor::Data expression_info{.context = context, .tables = tables_with_columns};
         ExpressionInfoVisitor(expression_info).visit(predicate_expression);
 
-        if (expression_info.is_stateful_function || !expression_info.is_deterministic_function)
-            return {};   /// Not optimized when predicate contains stateful function or indeterministic function
+        if (expression_info.is_stateful_function
+            || !expression_info.is_deterministic_function
+            || expression_info.is_window_function)
+        {
+            return {};   /// Not optimized when predicate contains stateful function or indeterministic function or window functions
+        }
 
         if (!expression_info.is_array_join)
         {
@@ -190,6 +194,12 @@ bool PredicateExpressionsOptimizer::tryMovePredicatesFromHavingToWhere(ASTSelect
         if (expression_info.is_stateful_function)
             return false;
 
+        if (expression_info.is_window_function)
+        {
+            // Window functions are not allowed in either HAVING or WHERE.
+            return false;
+        }
+
         if (expression_info.is_aggregate_function)
             having_predicates.emplace_back(moving_predicate);
         else
diff --git a/src/Interpreters/PredicateRewriteVisitor.cpp b/src/Interpreters/PredicateRewriteVisitor.cpp
index 9795675bcc8..5773629d0d1 100644
--- a/src/Interpreters/PredicateRewriteVisitor.cpp
+++ b/src/Interpreters/PredicateRewriteVisitor.cpp
@@ -88,7 +88,7 @@ bool PredicateRewriteVisitorData::rewriteSubquery(ASTSelectQuery & subquery, con
         || (!optimize_with && subquery.with())
         || subquery.withFill()
         || subquery.limitBy() || subquery.limitLength()
-        || hasStatefulFunction(subquery.select(), context))
+        || hasNonRewritableFunction(subquery.select(), context))
         return false;
 
     for (const auto & predicate : predicates)
diff --git a/src/Interpreters/TreeOptimizer.cpp b/src/Interpreters/TreeOptimizer.cpp
index cee19c632fa..c02f68496e2 100644
--- a/src/Interpreters/TreeOptimizer.cpp
+++ b/src/Interpreters/TreeOptimizer.cpp
@@ -594,9 +594,21 @@ void TreeOptimizer::apply(ASTPtr & query, Aliases & aliases, const NameSet & sou
     if (settings.optimize_arithmetic_operations_in_aggregate_functions)
         optimizeAggregationFunctions(query);
 
+    const std::string before = select_query->dumpTree();
+    fmt::print(stderr, "before predicate pushdown:\n'{}'\n",
+        select_query->dumpTree());
     /// Push the predicate expression down to the subqueries.
     rewrite_subqueries = PredicateExpressionsOptimizer(context, tables_with_columns, settings).optimize(*select_query);
 
+    fmt::print(stderr, "after predicate pushdown:\n'{}'\n",
+        select_query->dumpTree());
+
+    if (before != select_query->dumpTree())
+    {
+        fmt::print(stderr, "predicate pushdown made changes at\n'{}'\n",
+            StackTrace().toString());
+    }
+
     /// GROUP BY injective function elimination.
     optimizeGroupBy(select_query, source_columns_set, context);
 
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 0dee5a8648d..9dfd15a0e77 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -443,6 +443,8 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
 
 std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSelectQuery & select_query)
 {
+    fmt::print(stderr, "AST at getWindowFunctions:\n'{}'\n", query->dumpTree());
+
     /// There can not be window functions inside the WHERE and PREWHERE.
     if (select_query.where())
         assertNoWindows(select_query.where(), "in WHERE");
diff --git a/src/Processors/QueryPlan/AggregatingStep.cpp b/src/Processors/QueryPlan/AggregatingStep.cpp
index e8d4a262366..8b787859467 100644
--- a/src/Processors/QueryPlan/AggregatingStep.cpp
+++ b/src/Processors/QueryPlan/AggregatingStep.cpp
@@ -156,12 +156,12 @@ void AggregatingStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void AggregatingStep::describeActions(FormatSettings & settings) const
+void AggregatingStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     params.explain(settings.out, settings.offset);
 }
 
-void AggregatingStep::describePipeline(FormatSettings & settings) const
+void AggregatingStep::describePipeline(QueryPlanStepFormatSettings & settings) const
 {
     if (!aggregating.empty())
         IQueryPlanStep::describePipeline(aggregating, settings);
diff --git a/src/Processors/QueryPlan/AggregatingStep.h b/src/Processors/QueryPlan/AggregatingStep.h
index 853173895b3..3d12e79a662 100644
--- a/src/Processors/QueryPlan/AggregatingStep.h
+++ b/src/Processors/QueryPlan/AggregatingStep.h
@@ -29,8 +29,8 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings &) const override;
-    void describePipeline(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings &) const override;
+    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     Aggregator::Params params;
diff --git a/src/Processors/QueryPlan/ArrayJoinStep.cpp b/src/Processors/QueryPlan/ArrayJoinStep.cpp
index b71d2176e52..01cfd7c561a 100644
--- a/src/Processors/QueryPlan/ArrayJoinStep.cpp
+++ b/src/Processors/QueryPlan/ArrayJoinStep.cpp
@@ -69,7 +69,7 @@ void ArrayJoinStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void ArrayJoinStep::describeActions(FormatSettings & settings) const
+void ArrayJoinStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     bool first = true;
diff --git a/src/Processors/QueryPlan/ArrayJoinStep.h b/src/Processors/QueryPlan/ArrayJoinStep.h
index 92c7e0a1304..338ec7f0583 100644
--- a/src/Processors/QueryPlan/ArrayJoinStep.h
+++ b/src/Processors/QueryPlan/ArrayJoinStep.h
@@ -15,7 +15,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     void updateInputStream(DataStream input_stream, Block result_header);
 
diff --git a/src/Processors/QueryPlan/CreatingSetsStep.cpp b/src/Processors/QueryPlan/CreatingSetsStep.cpp
index 5868a7045f7..c6b77bc46c3 100644
--- a/src/Processors/QueryPlan/CreatingSetsStep.cpp
+++ b/src/Processors/QueryPlan/CreatingSetsStep.cpp
@@ -47,7 +47,7 @@ void CreatingSetStep::transformPipeline(QueryPipeline & pipeline)
     pipeline.addCreatingSetsTransform(getOutputStream().header, std::move(subquery_for_set), network_transfer_limits, context);
 }
 
-void CreatingSetStep::describeActions(FormatSettings & settings) const
+void CreatingSetStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
 
@@ -102,7 +102,7 @@ QueryPipelinePtr CreatingSetsStep::updatePipeline(QueryPipelines pipelines)
     return main_pipeline;
 }
 
-void CreatingSetsStep::describePipeline(FormatSettings & settings) const
+void CreatingSetsStep::describePipeline(QueryPlanStepFormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/CreatingSetsStep.h b/src/Processors/QueryPlan/CreatingSetsStep.h
index ec13ab2052e..f610eeb64d9 100644
--- a/src/Processors/QueryPlan/CreatingSetsStep.h
+++ b/src/Processors/QueryPlan/CreatingSetsStep.h
@@ -22,7 +22,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     String description;
@@ -40,7 +40,7 @@ public:
 
     QueryPipelinePtr updatePipeline(QueryPipelines pipelines) override;
 
-    void describePipeline(FormatSettings & settings) const override;
+    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     Processors processors;
diff --git a/src/Processors/QueryPlan/DistinctStep.cpp b/src/Processors/QueryPlan/DistinctStep.cpp
index 60966b08beb..ed2bae002ea 100644
--- a/src/Processors/QueryPlan/DistinctStep.cpp
+++ b/src/Processors/QueryPlan/DistinctStep.cpp
@@ -79,7 +79,7 @@ void DistinctStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void DistinctStep::describeActions(FormatSettings & settings) const
+void DistinctStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Columns: ";
diff --git a/src/Processors/QueryPlan/DistinctStep.h b/src/Processors/QueryPlan/DistinctStep.h
index 4bfd73ce044..901e199ca39 100644
--- a/src/Processors/QueryPlan/DistinctStep.h
+++ b/src/Processors/QueryPlan/DistinctStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     SizeLimits set_size_limits;
diff --git a/src/Processors/QueryPlan/ExpressionStep.cpp b/src/Processors/QueryPlan/ExpressionStep.cpp
index f7ca0a972dc..028c9502c71 100644
--- a/src/Processors/QueryPlan/ExpressionStep.cpp
+++ b/src/Processors/QueryPlan/ExpressionStep.cpp
@@ -87,7 +87,7 @@ void ExpressionStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void ExpressionStep::describeActions(FormatSettings & settings) const
+void ExpressionStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     bool first = true;
diff --git a/src/Processors/QueryPlan/ExpressionStep.h b/src/Processors/QueryPlan/ExpressionStep.h
index 60f186688b0..2519765dbae 100644
--- a/src/Processors/QueryPlan/ExpressionStep.h
+++ b/src/Processors/QueryPlan/ExpressionStep.h
@@ -26,7 +26,7 @@ public:
 
     void updateInputStream(DataStream input_stream, bool keep_header);
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     const ActionsDAGPtr & getExpression() const { return actions_dag; }
 
diff --git a/src/Processors/QueryPlan/FillingStep.cpp b/src/Processors/QueryPlan/FillingStep.cpp
index 1a8fba97ee2..18c71b89601 100644
--- a/src/Processors/QueryPlan/FillingStep.cpp
+++ b/src/Processors/QueryPlan/FillingStep.cpp
@@ -43,7 +43,7 @@ void FillingStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void FillingStep::describeActions(FormatSettings & settings) const
+void FillingStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     settings.out << String(settings.offset, ' ');
     dumpSortDescription(sort_description, input_streams.front().header, settings.out);
diff --git a/src/Processors/QueryPlan/FillingStep.h b/src/Processors/QueryPlan/FillingStep.h
index 85736464a6c..a7836e4de68 100644
--- a/src/Processors/QueryPlan/FillingStep.h
+++ b/src/Processors/QueryPlan/FillingStep.h
@@ -15,7 +15,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     SortDescription sort_description;
diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp
index 8fb405e685b..98f111a1529 100644
--- a/src/Processors/QueryPlan/FilterStep.cpp
+++ b/src/Processors/QueryPlan/FilterStep.cpp
@@ -80,7 +80,7 @@ void FilterStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void FilterStep::describeActions(FormatSettings & settings) const
+void FilterStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Filter column: " << filter_column_name << '\n';
diff --git a/src/Processors/QueryPlan/FilterStep.h b/src/Processors/QueryPlan/FilterStep.h
index 72b02624faa..561fbf347da 100644
--- a/src/Processors/QueryPlan/FilterStep.h
+++ b/src/Processors/QueryPlan/FilterStep.h
@@ -22,7 +22,7 @@ public:
 
     void updateInputStream(DataStream input_stream, bool keep_header);
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     const ActionsDAGPtr & getExpression() const { return actions_dag; }
     const String & getFilterColumnName() const { return filter_column_name; }
diff --git a/src/Processors/QueryPlan/FinishSortingStep.cpp b/src/Processors/QueryPlan/FinishSortingStep.cpp
index d883bd0e0dd..7af8fad2980 100644
--- a/src/Processors/QueryPlan/FinishSortingStep.cpp
+++ b/src/Processors/QueryPlan/FinishSortingStep.cpp
@@ -85,7 +85,7 @@ void FinishSortingStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void FinishSortingStep::describeActions(FormatSettings & settings) const
+void FinishSortingStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
 
diff --git a/src/Processors/QueryPlan/FinishSortingStep.h b/src/Processors/QueryPlan/FinishSortingStep.h
index 4bb62037faa..25dc0b78bda 100644
--- a/src/Processors/QueryPlan/FinishSortingStep.h
+++ b/src/Processors/QueryPlan/FinishSortingStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/IQueryPlanStep.cpp b/src/Processors/QueryPlan/IQueryPlanStep.cpp
index 71c4caaa795..a864043ddcb 100644
--- a/src/Processors/QueryPlan/IQueryPlanStep.cpp
+++ b/src/Processors/QueryPlan/IQueryPlanStep.cpp
@@ -18,7 +18,8 @@ const DataStream & IQueryPlanStep::getOutputStream() const
     return *output_stream;
 }
 
-static void doDescribeHeader(const Block & header, size_t count, IQueryPlanStep::FormatSettings & settings)
+static void doDescribeHeader(const Block & header, size_t count,
+    QueryPlanStepFormatSettings & settings)
 {
     String prefix(settings.offset, settings.indent_char);
     prefix += "Header";
@@ -46,11 +47,14 @@ static void doDescribeHeader(const Block & header, size_t count, IQueryPlanStep:
 
         first = false;
         elem.dumpNameAndType(settings.out);
+        settings.out << ": ";
+        elem.dumpStructure(settings.out);
         settings.out << '\n';
     }
 }
 
-static void doDescribeProcessor(const IProcessor & processor, size_t count, IQueryPlanStep::FormatSettings & settings)
+static void doDescribeProcessor(const IProcessor & processor, size_t count,
+    QueryPlanStepFormatSettings & settings)
 {
     settings.out << String(settings.offset, settings.indent_char) << processor.getName();
     if (count > 1)
@@ -87,7 +91,7 @@ static void doDescribeProcessor(const IProcessor & processor, size_t count, IQue
     settings.offset += settings.indent;
 }
 
-void IQueryPlanStep::describePipeline(const Processors & processors, FormatSettings & settings)
+void IQueryPlanStep::describePipeline(const Processors & processors, QueryPlanStepFormatSettings & settings)
 {
     const IProcessor * prev = nullptr;
     size_t count = 0;
diff --git a/src/Processors/QueryPlan/IQueryPlanStep.h b/src/Processors/QueryPlan/IQueryPlanStep.h
index dfdbc9df442..1de1bbc438f 100644
--- a/src/Processors/QueryPlan/IQueryPlanStep.h
+++ b/src/Processors/QueryPlan/IQueryPlanStep.h
@@ -61,6 +61,16 @@ public:
 
 using DataStreams = std::vector<DataStream>;
 
+// Not a nested class so that we can forward declare it.
+struct QueryPlanStepFormatSettings
+{
+    WriteBuffer & out;
+    size_t offset = 0;
+    const size_t indent = 2;
+    const char indent_char = ' ';
+    const bool write_header = false;
+};
+
 /// Single step of query plan.
 class IQueryPlanStep
 {
@@ -86,20 +96,11 @@ public:
     const std::string & getStepDescription() const { return step_description; }
     void setStepDescription(std::string description) { step_description = std::move(description); }
 
-    struct FormatSettings
-    {
-        WriteBuffer & out;
-        size_t offset = 0;
-        const size_t indent = 2;
-        const char indent_char = ' ';
-        const bool write_header = false;
-    };
-
     /// Get detailed description of step actions. This is shown in EXPLAIN query with options `actions = 1`.
-    virtual void describeActions(FormatSettings & /*settings*/) const {}
+    virtual void describeActions(QueryPlanStepFormatSettings & /*settings*/) const {}
 
     /// Get description of processors added in current step. Should be called after updatePipeline().
-    virtual void describePipeline(FormatSettings & /*settings*/) const {}
+    virtual void describePipeline(QueryPlanStepFormatSettings & /*settings*/) const {}
 
 protected:
     DataStreams input_streams;
@@ -108,7 +109,7 @@ protected:
     /// Text description about what current step does.
     std::string step_description;
 
-    static void describePipeline(const Processors & processors, FormatSettings & settings);
+    static void describePipeline(const Processors & processors, QueryPlanStepFormatSettings & settings);
 };
 
 using QueryPlanStepPtr = std::unique_ptr<IQueryPlanStep>;
diff --git a/src/Processors/QueryPlan/ISourceStep.cpp b/src/Processors/QueryPlan/ISourceStep.cpp
index 1796f033896..367be8e8ea5 100644
--- a/src/Processors/QueryPlan/ISourceStep.cpp
+++ b/src/Processors/QueryPlan/ISourceStep.cpp
@@ -19,7 +19,7 @@ QueryPipelinePtr ISourceStep::updatePipeline(QueryPipelines)
     return pipeline;
 }
 
-void ISourceStep::describePipeline(FormatSettings & settings) const
+void ISourceStep::describePipeline(QueryPlanStepFormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/ISourceStep.h b/src/Processors/QueryPlan/ISourceStep.h
index fdb3dd566cb..050e39ac11b 100644
--- a/src/Processors/QueryPlan/ISourceStep.h
+++ b/src/Processors/QueryPlan/ISourceStep.h
@@ -14,7 +14,7 @@ public:
 
     virtual void initializePipeline(QueryPipeline & pipeline) = 0;
 
-    void describePipeline(FormatSettings & settings) const override;
+    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
 
 protected:
     /// We collect processors got after pipeline transformation.
diff --git a/src/Processors/QueryPlan/ITransformingStep.cpp b/src/Processors/QueryPlan/ITransformingStep.cpp
index cb27bf38278..5430bafa27a 100644
--- a/src/Processors/QueryPlan/ITransformingStep.cpp
+++ b/src/Processors/QueryPlan/ITransformingStep.cpp
@@ -65,7 +65,7 @@ void ITransformingStep::updateDistinctColumns(const Block & res_header, NameSet
     }
 }
 
-void ITransformingStep::describePipeline(FormatSettings & settings) const
+void ITransformingStep::describePipeline(QueryPlanStepFormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/ITransformingStep.h b/src/Processors/QueryPlan/ITransformingStep.h
index bd99478ec37..b1bfebd4571 100644
--- a/src/Processors/QueryPlan/ITransformingStep.h
+++ b/src/Processors/QueryPlan/ITransformingStep.h
@@ -55,7 +55,7 @@ public:
     const TransformTraits & getTransformTraits() const { return transform_traits; }
     const DataStreamTraits & getDataStreamTraits() const { return data_stream_traits; }
 
-    void describePipeline(FormatSettings & settings) const override;
+    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
 
 protected:
     /// Clear distinct_columns if res_header doesn't contain all of them.
diff --git a/src/Processors/QueryPlan/LimitByStep.cpp b/src/Processors/QueryPlan/LimitByStep.cpp
index 9fcf5b60164..15c64fcea44 100644
--- a/src/Processors/QueryPlan/LimitByStep.cpp
+++ b/src/Processors/QueryPlan/LimitByStep.cpp
@@ -46,7 +46,7 @@ void LimitByStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void LimitByStep::describeActions(FormatSettings & settings) const
+void LimitByStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
 
diff --git a/src/Processors/QueryPlan/LimitByStep.h b/src/Processors/QueryPlan/LimitByStep.h
index 9320735640c..01dcff5c912 100644
--- a/src/Processors/QueryPlan/LimitByStep.h
+++ b/src/Processors/QueryPlan/LimitByStep.h
@@ -16,7 +16,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     size_t group_length;
diff --git a/src/Processors/QueryPlan/LimitStep.cpp b/src/Processors/QueryPlan/LimitStep.cpp
index 565d05956e5..38ec9265d4e 100644
--- a/src/Processors/QueryPlan/LimitStep.cpp
+++ b/src/Processors/QueryPlan/LimitStep.cpp
@@ -50,7 +50,7 @@ void LimitStep::transformPipeline(QueryPipeline & pipeline)
     pipeline.addTransform(std::move(transform));
 }
 
-void LimitStep::describeActions(FormatSettings & settings) const
+void LimitStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Limit " << limit << '\n';
diff --git a/src/Processors/QueryPlan/LimitStep.h b/src/Processors/QueryPlan/LimitStep.h
index eea186fea4a..a95bf28f896 100644
--- a/src/Processors/QueryPlan/LimitStep.h
+++ b/src/Processors/QueryPlan/LimitStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     size_t getLimitForSorting() const
     {
diff --git a/src/Processors/QueryPlan/MergeSortingStep.cpp b/src/Processors/QueryPlan/MergeSortingStep.cpp
index b30286130b1..e2af023f598 100644
--- a/src/Processors/QueryPlan/MergeSortingStep.cpp
+++ b/src/Processors/QueryPlan/MergeSortingStep.cpp
@@ -73,7 +73,7 @@ void MergeSortingStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void MergeSortingStep::describeActions(FormatSettings & settings) const
+void MergeSortingStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/MergeSortingStep.h b/src/Processors/QueryPlan/MergeSortingStep.h
index a385a8a3e93..c41c209dd55 100644
--- a/src/Processors/QueryPlan/MergeSortingStep.h
+++ b/src/Processors/QueryPlan/MergeSortingStep.h
@@ -26,7 +26,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/MergingAggregatedStep.cpp b/src/Processors/QueryPlan/MergingAggregatedStep.cpp
index 85bbbeab59a..2fddc88bbc1 100644
--- a/src/Processors/QueryPlan/MergingAggregatedStep.cpp
+++ b/src/Processors/QueryPlan/MergingAggregatedStep.cpp
@@ -63,7 +63,7 @@ void MergingAggregatedStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void MergingAggregatedStep::describeActions(FormatSettings & settings) const
+void MergingAggregatedStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     return params->params.explain(settings.out, settings.offset);
 }
diff --git a/src/Processors/QueryPlan/MergingAggregatedStep.h b/src/Processors/QueryPlan/MergingAggregatedStep.h
index 5ffceb7f938..2d9d902a294 100644
--- a/src/Processors/QueryPlan/MergingAggregatedStep.h
+++ b/src/Processors/QueryPlan/MergingAggregatedStep.h
@@ -23,7 +23,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     AggregatingTransformParamsPtr params;
diff --git a/src/Processors/QueryPlan/MergingFinal.cpp b/src/Processors/QueryPlan/MergingFinal.cpp
index c1dec231f11..46431e0c767 100644
--- a/src/Processors/QueryPlan/MergingFinal.cpp
+++ b/src/Processors/QueryPlan/MergingFinal.cpp
@@ -153,7 +153,7 @@ void MergingFinal::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void MergingFinal::describeActions(FormatSettings & settings) const
+void MergingFinal::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/MergingFinal.h b/src/Processors/QueryPlan/MergingFinal.h
index c01f5c7f9a1..207c191c9cb 100644
--- a/src/Processors/QueryPlan/MergingFinal.h
+++ b/src/Processors/QueryPlan/MergingFinal.h
@@ -22,7 +22,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     size_t num_output_streams;
diff --git a/src/Processors/QueryPlan/MergingSortedStep.cpp b/src/Processors/QueryPlan/MergingSortedStep.cpp
index c59540b009f..a1249ef412f 100644
--- a/src/Processors/QueryPlan/MergingSortedStep.cpp
+++ b/src/Processors/QueryPlan/MergingSortedStep.cpp
@@ -62,7 +62,7 @@ void MergingSortedStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void MergingSortedStep::describeActions(FormatSettings & settings) const
+void MergingSortedStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/MergingSortedStep.h b/src/Processors/QueryPlan/MergingSortedStep.h
index 483cfa5e8a7..0c91d505f3d 100644
--- a/src/Processors/QueryPlan/MergingSortedStep.h
+++ b/src/Processors/QueryPlan/MergingSortedStep.h
@@ -21,7 +21,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/OffsetStep.cpp b/src/Processors/QueryPlan/OffsetStep.cpp
index 7ac3d3f2110..1a06f2ba05e 100644
--- a/src/Processors/QueryPlan/OffsetStep.cpp
+++ b/src/Processors/QueryPlan/OffsetStep.cpp
@@ -36,7 +36,7 @@ void OffsetStep::transformPipeline(QueryPipeline & pipeline)
     pipeline.addTransform(std::move(transform));
 }
 
-void OffsetStep::describeActions(FormatSettings & settings) const
+void OffsetStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     settings.out << String(settings.offset, ' ') << "Offset " << offset << '\n';
 }
diff --git a/src/Processors/QueryPlan/OffsetStep.h b/src/Processors/QueryPlan/OffsetStep.h
index 17949371edf..18ea58635af 100644
--- a/src/Processors/QueryPlan/OffsetStep.h
+++ b/src/Processors/QueryPlan/OffsetStep.h
@@ -15,7 +15,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     size_t offset;
diff --git a/src/Processors/QueryPlan/PartialSortingStep.cpp b/src/Processors/QueryPlan/PartialSortingStep.cpp
index ce34eca9112..072217b4d0d 100644
--- a/src/Processors/QueryPlan/PartialSortingStep.cpp
+++ b/src/Processors/QueryPlan/PartialSortingStep.cpp
@@ -70,7 +70,7 @@ void PartialSortingStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void PartialSortingStep::describeActions(FormatSettings & settings) const
+void PartialSortingStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/PartialSortingStep.h b/src/Processors/QueryPlan/PartialSortingStep.h
index 172ef25c300..6d9dbe2b9d8 100644
--- a/src/Processors/QueryPlan/PartialSortingStep.h
+++ b/src/Processors/QueryPlan/PartialSortingStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/QueryPlan.cpp b/src/Processors/QueryPlan/QueryPlan.cpp
index e88090b4819..ee7c5220b4c 100644
--- a/src/Processors/QueryPlan/QueryPlan.cpp
+++ b/src/Processors/QueryPlan/QueryPlan.cpp
@@ -204,7 +204,7 @@ void QueryPlan::addInterpreterContext(std::shared_ptr<Context> context)
 
 static void explainStep(
     const IQueryPlanStep & step,
-    IQueryPlanStep::FormatSettings & settings,
+    QueryPlanStepFormatSettings & settings,
     const QueryPlan::ExplainPlanOptions & options)
 {
     std::string prefix(settings.offset, ' ');
@@ -247,11 +247,20 @@ static void explainStep(
         step.describeActions(settings);
 }
 
+std::string debugExplainStep(const IQueryPlanStep & step)
+{
+    WriteBufferFromOwnString out;
+    QueryPlanStepFormatSettings settings{.out = out};
+    QueryPlan::ExplainPlanOptions options{.actions = true};
+    explainStep(step, settings, options);
+    return out.str();
+}
+
 void QueryPlan::explainPlan(WriteBuffer & buffer, const ExplainPlanOptions & options)
 {
     checkInitialized();
 
-    IQueryPlanStep::FormatSettings settings{.out = buffer, .write_header = options.header};
+    QueryPlanStepFormatSettings settings{.out = buffer, .write_header = options.header};
 
     struct Frame
     {
@@ -284,7 +293,7 @@ void QueryPlan::explainPlan(WriteBuffer & buffer, const ExplainPlanOptions & opt
     }
 }
 
-static void explainPipelineStep(IQueryPlanStep & step, IQueryPlanStep::FormatSettings & settings)
+static void explainPipelineStep(IQueryPlanStep & step, QueryPlanStepFormatSettings & settings)
 {
     settings.out << String(settings.offset, settings.indent_char) << "(" << step.getName() << ")\n";
     size_t current_offset = settings.offset;
@@ -297,7 +306,7 @@ void QueryPlan::explainPipeline(WriteBuffer & buffer, const ExplainPipelineOptio
 {
     checkInitialized();
 
-    IQueryPlanStep::FormatSettings settings{.out = buffer, .write_header = options.header};
+    QueryPlanStepFormatSettings settings{.out = buffer, .write_header = options.header};
 
     struct Frame
     {
@@ -488,6 +497,12 @@ static bool tryMergeExpressions(QueryPlan::Node * parent_node, QueryPlan::Node *
 {
     auto & parent = parent_node->step;
     auto & child = child_node->step;
+
+    fmt::print(stderr,
+        "try merge expressions: parent '{}'\n, child '{}'\n at {}\n",
+        debugExplainStep(*parent), debugExplainStep(*child),
+        StackTrace().toString());
+
     /// TODO: FilterStep
     auto * parent_expr = typeid_cast<ExpressionStep *>(parent.get());
     auto * child_expr = typeid_cast<ExpressionStep *>(child.get());
diff --git a/src/Processors/QueryPlan/QueryPlan.h b/src/Processors/QueryPlan/QueryPlan.h
index cbe487312cc..5ae991a328f 100644
--- a/src/Processors/QueryPlan/QueryPlan.h
+++ b/src/Processors/QueryPlan/QueryPlan.h
@@ -13,6 +13,7 @@ class DataStream;
 
 class IQueryPlanStep;
 using QueryPlanStepPtr = std::unique_ptr<IQueryPlanStep>;
+struct QueryPlanStepFormatSettings;
 
 class QueryPipeline;
 using QueryPipelinePtr = std::unique_ptr<QueryPipeline>;
@@ -97,4 +98,9 @@ private:
     std::vector<std::shared_ptr<Context>> interpreter_context;
 };
 
+void debugExplainStep(
+    const IQueryPlanStep & step,
+    QueryPlanStepFormatSettings & settings,
+    const QueryPlan::ExplainPlanOptions & options);
+
 }
diff --git a/src/Processors/QueryPlan/TotalsHavingStep.cpp b/src/Processors/QueryPlan/TotalsHavingStep.cpp
index 9947875e679..310f657b39d 100644
--- a/src/Processors/QueryPlan/TotalsHavingStep.cpp
+++ b/src/Processors/QueryPlan/TotalsHavingStep.cpp
@@ -74,7 +74,7 @@ static String totalsModeToString(TotalsMode totals_mode, double auto_include_thr
     __builtin_unreachable();
 }
 
-void TotalsHavingStep::describeActions(FormatSettings & settings) const
+void TotalsHavingStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Filter column: " << filter_column_name << '\n';
diff --git a/src/Processors/QueryPlan/TotalsHavingStep.h b/src/Processors/QueryPlan/TotalsHavingStep.h
index 7c1638013e5..88fd9bbb597 100644
--- a/src/Processors/QueryPlan/TotalsHavingStep.h
+++ b/src/Processors/QueryPlan/TotalsHavingStep.h
@@ -26,7 +26,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     bool overflow_row;
diff --git a/src/Processors/QueryPlan/UnionStep.cpp b/src/Processors/QueryPlan/UnionStep.cpp
index 630ff53f47d..413b52c69c2 100644
--- a/src/Processors/QueryPlan/UnionStep.cpp
+++ b/src/Processors/QueryPlan/UnionStep.cpp
@@ -36,7 +36,7 @@ QueryPipelinePtr UnionStep::updatePipeline(QueryPipelines pipelines)
     return pipeline;
 }
 
-void UnionStep::describePipeline(FormatSettings & settings) const
+void UnionStep::describePipeline(QueryPlanStepFormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/UnionStep.h b/src/Processors/QueryPlan/UnionStep.h
index e2e1f2c9efa..aeac7fa54ff 100644
--- a/src/Processors/QueryPlan/UnionStep.h
+++ b/src/Processors/QueryPlan/UnionStep.h
@@ -15,7 +15,7 @@ public:
 
     QueryPipelinePtr updatePipeline(QueryPipelines pipelines) override;
 
-    void describePipeline(FormatSettings & settings) const override;
+    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     Block header;
diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
index 82c589b8b20..d2d24985937 100644
--- a/src/Processors/QueryPlan/WindowStep.cpp
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -71,7 +71,7 @@ void WindowStep::transformPipeline(QueryPipeline & pipeline)
         "WindowStep transform for '" + window_description.window_name + "'");
 }
 
-void WindowStep::describeActions(FormatSettings & settings) const
+void WindowStep::describeActions(QueryPlanStepFormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Window: (";
diff --git a/src/Processors/QueryPlan/WindowStep.h b/src/Processors/QueryPlan/WindowStep.h
index 069d02c655c..25194820a0d 100644
--- a/src/Processors/QueryPlan/WindowStep.h
+++ b/src/Processors/QueryPlan/WindowStep.h
@@ -22,7 +22,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(FormatSettings & settings) const override;
+    void describeActions(QueryPlanStepFormatSettings & settings) const override;
 
 private:
     WindowDescription window_description;
diff --git a/src/Processors/Transforms/WindowTransform.cpp b/src/Processors/Transforms/WindowTransform.cpp
index 6e8b0ea8e39..b200e306213 100644
--- a/src/Processors/Transforms/WindowTransform.cpp
+++ b/src/Processors/Transforms/WindowTransform.cpp
@@ -77,6 +77,11 @@ void WindowTransform::transform(Chunk & chunk)
         ws.argument_columns.clear();
         for (const auto column_index : ws.argument_column_indices)
         {
+            // Aggregate functions can't work with constant columns, so we have to
+            // materialize them like the Aggregator does.
+            columns[column_index]
+                = std::move(columns[column_index])->convertToFullColumnIfConst();
+
             ws.argument_columns.push_back(columns[column_index].get());
         }
 
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 6c78108d734..19509de8a41 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -80,8 +80,19 @@ select number, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
 
 -- should work in ORDER BY though
+-- doesn't work now
+-- select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
-select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+-- at least it works in ORDER BY if you wrap it in a subquery
+
+select * from (select count(*) over () c from numbers(3)) order by c;
+
+-- must work in WHERE if you wrap it in a subquery
+
+1
+2
+3
+select * from (select count(*) over () c from numbers(3)) where c > 0;
 
 -- this one doesn't work yet -- looks like the column names clash, and the
 -- window count() is overwritten with aggregate count()
@@ -91,16 +102,9 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 -- an explain test would also be helpful, but it's too immature now and I don't
 -- want to change reference all the time
 
-9	9
-6	8
-7	8
-8	8
-3	5
-4	5
-5	5
-0	2
-1	2
-2	2
+1
+2
+3
 select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 5) order by number) as m from numbers(31) order by number settings max_block_size = 2;
 
 -- two functions over the same window
@@ -140,6 +144,8 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 30	30	1
 select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 3) order by number desc) as m from numbers(7) order by number settings max_block_size = 2;
 
+-- check that we can work with constant columns
+
 0	2	3
 1	2	2
 2	2	1
@@ -147,3 +153,13 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 4	5	2
 5	5	1
 6	6	1
+select median(x) over (partition by x) from (select 1 x);
+
+-- an empty window definition is valid as well
+
+1
+select groupArray(number) over () from numbers(3);
+
+[0]
+[0,1]
+[0,1,2]
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index a28d435d3f8..4bd9f1b7693 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -25,7 +25,14 @@ select number, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
 
 -- should work in ORDER BY though
-select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+-- doesn't work now
+-- select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+
+-- at least it works in ORDER BY if you wrap it in a subquery
+select * from (select count(*) over () c from numbers(3)) order by c;
+
+-- must work in WHERE if you wrap it in a subquery
+select * from (select count(*) over () c from numbers(3)) where c > 0;
 
 -- this one doesn't work yet -- looks like the column names clash, and the
 -- window count() is overwritten with aggregate count()
@@ -40,3 +47,9 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 -- an explain test would also be helpful, but it's too immature now and I don't
 -- want to change reference all the time
 select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 3) order by number desc) as m from numbers(7) order by number settings max_block_size = 2;
+
+-- check that we can work with constant columns
+select median(x) over (partition by x) from (select 1 x);
+
+-- an empty window definition is valid as well
+select groupArray(number) over () from numbers(3);

From 23d77a1698e90c8cd5f3f873fd95c313f012d5c9 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 09:59:38 +0300
Subject: [PATCH 375/504] remove debug output

---
 src/Interpreters/ActionsDAG.cpp             |  3 ---
 src/Interpreters/ExpressionAnalyzer.cpp     | 12 ------------
 src/Interpreters/InterpreterSelectQuery.cpp |  3 ---
 src/Interpreters/TreeOptimizer.cpp          | 12 ------------
 src/Interpreters/TreeRewriter.cpp           |  2 --
 src/Processors/QueryPlan/QueryPlan.cpp      |  5 -----
 6 files changed, 37 deletions(-)

diff --git a/src/Interpreters/ActionsDAG.cpp b/src/Interpreters/ActionsDAG.cpp
index f9505cc4199..2fc78261f17 100644
--- a/src/Interpreters/ActionsDAG.cpp
+++ b/src/Interpreters/ActionsDAG.cpp
@@ -691,9 +691,6 @@ ActionsDAGPtr ActionsDAG::makeConvertingActions(
 
 ActionsDAGPtr ActionsDAG::merge(ActionsDAG && first, ActionsDAG && second)
 {
-    fmt::print(stderr, "actiondag merge first '{}'\n, second '{}\n",
-        first.dumpDAG(), second.dumpDAG());
-
     /// first: x (1), x (2), y ==> x (2), z, x (3)
     /// second: x (1), x (2), x (3) ==> x (3), x (2), x (1)
     /// merge: x (1), x (2), x (3), y =(first)=> x (3), y, x (2), z, x (4) =(second)=> y, z, x (4), x (2), x (3)
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index e271f78b699..37f2791d8bd 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1032,8 +1032,6 @@ bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain,
     if (!select_query->having())
         return false;
 
-    fmt::print(stderr, "appendHaving:\n'{}'\n", select_query->dumpTree());
-
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
     getRootActionsForHaving(select_query->having(), only_types, step.actions());
@@ -1443,12 +1441,8 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
 
         if (has_window)
         {
-            fmt::print(stderr, "chain before window args: '{}'\n", chain.dumpChain());
-
             query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
 
-            fmt::print(stderr, "after window args chain: '{}'\n", chain.dumpChain());
-
             auto select_required_columns = chain.getLastStep().required_output;
 
             for (const auto & x : chain.getLastActions()->getNamesAndTypesList())
@@ -1457,8 +1451,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
             }
 
             finalize_chain(chain);
-            //auto & window_step = chain.addStep();
-            //window_step.actions_dag.query_analyzer.columns_after_window);
 
             auto & step = chain.lastStep(query_analyzer.columns_after_window);
             step.required_output = select_required_columns;
@@ -1473,8 +1465,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
                     step.required_output.push_back(child->getColumnName());
                 }
             }
-
-            fmt::print(stderr, "chain after window columns: '{}'\n", chain.dumpChain());
         }
 
         selected_columns = chain.getLastStep().required_output;
@@ -1488,8 +1478,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         final_projection = query_analyzer.appendProjectResult(chain);
 
         finalize_chain(chain);
-
-        fmt::print(stderr, "final chain: '{}'\n", chain.dumpChain());
     }
 
 
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 9a6e6b7593b..38cc19a00d6 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -521,9 +521,6 @@ Block InterpreterSelectQuery::getSampleBlockImpl()
             filter_info,
             source_header);
 
-    fmt::print(stderr, "ExpressionAnalysisResult: '{}'\nat\n{}\n",
-        analysis_result.dump(), StackTrace().toString());
-
     if (options.to_stage == QueryProcessingStage::Enum::FetchColumns)
     {
         auto header = source_header;
diff --git a/src/Interpreters/TreeOptimizer.cpp b/src/Interpreters/TreeOptimizer.cpp
index c02f68496e2..cee19c632fa 100644
--- a/src/Interpreters/TreeOptimizer.cpp
+++ b/src/Interpreters/TreeOptimizer.cpp
@@ -594,21 +594,9 @@ void TreeOptimizer::apply(ASTPtr & query, Aliases & aliases, const NameSet & sou
     if (settings.optimize_arithmetic_operations_in_aggregate_functions)
         optimizeAggregationFunctions(query);
 
-    const std::string before = select_query->dumpTree();
-    fmt::print(stderr, "before predicate pushdown:\n'{}'\n",
-        select_query->dumpTree());
     /// Push the predicate expression down to the subqueries.
     rewrite_subqueries = PredicateExpressionsOptimizer(context, tables_with_columns, settings).optimize(*select_query);
 
-    fmt::print(stderr, "after predicate pushdown:\n'{}'\n",
-        select_query->dumpTree());
-
-    if (before != select_query->dumpTree())
-    {
-        fmt::print(stderr, "predicate pushdown made changes at\n'{}'\n",
-            StackTrace().toString());
-    }
-
     /// GROUP BY injective function elimination.
     optimizeGroupBy(select_query, source_columns_set, context);
 
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 9dfd15a0e77..0dee5a8648d 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -443,8 +443,6 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
 
 std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSelectQuery & select_query)
 {
-    fmt::print(stderr, "AST at getWindowFunctions:\n'{}'\n", query->dumpTree());
-
     /// There can not be window functions inside the WHERE and PREWHERE.
     if (select_query.where())
         assertNoWindows(select_query.where(), "in WHERE");
diff --git a/src/Processors/QueryPlan/QueryPlan.cpp b/src/Processors/QueryPlan/QueryPlan.cpp
index ee7c5220b4c..a2fc8c689ff 100644
--- a/src/Processors/QueryPlan/QueryPlan.cpp
+++ b/src/Processors/QueryPlan/QueryPlan.cpp
@@ -498,11 +498,6 @@ static bool tryMergeExpressions(QueryPlan::Node * parent_node, QueryPlan::Node *
     auto & parent = parent_node->step;
     auto & child = child_node->step;
 
-    fmt::print(stderr,
-        "try merge expressions: parent '{}'\n, child '{}'\n at {}\n",
-        debugExplainStep(*parent), debugExplainStep(*child),
-        StackTrace().toString());
-
     /// TODO: FilterStep
     auto * parent_expr = typeid_cast<ExpressionStep *>(parent.get());
     auto * child_expr = typeid_cast<ExpressionStep *>(child.get());

From 54448ad4ce8ef69d7cee1bd8f3d0f37e761b060f Mon Sep 17 00:00:00 2001
From: vdimir <vdimirc@gmail.com>
Date: Thu, 24 Dec 2020 07:38:10 +0000
Subject: [PATCH 376/504] Remove unnecessary IPAddressDictionary::has overload

---
 src/Dictionaries/IPAddressDictionary.cpp | 104 ++++++-----------------
 src/Dictionaries/IPAddressDictionary.h   |   3 -
 2 files changed, 24 insertions(+), 83 deletions(-)

diff --git a/src/Dictionaries/IPAddressDictionary.cpp b/src/Dictionaries/IPAddressDictionary.cpp
index 53eb0c32e16..79661fd2a18 100644
--- a/src/Dictionaries/IPAddressDictionary.cpp
+++ b/src/Dictionaries/IPAddressDictionary.cpp
@@ -401,57 +401,32 @@ void IPAddressDictionary::has(const Columns & key_columns, const DataTypes & key
 {
     validateKeyTypes(key_types);
 
-    const auto & attribute = attributes.front();
-
-    switch (attribute.type)
+    const auto first_column = key_columns.front();
+    const auto rows = first_column->size();
+    if (first_column->isNumeric())
     {
-        case AttributeUnderlyingType::utUInt8:
-            has<UInt8>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utUInt16:
-            has<UInt16>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utUInt32:
-            has<UInt32>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utUInt64:
-            has<UInt64>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utUInt128:
-            has<UInt128>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utInt8:
-            has<Int8>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utInt16:
-            has<Int16>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utInt32:
-            has<Int32>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utInt64:
-            has<Int64>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utFloat32:
-            has<Float32>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utFloat64:
-            has<Float64>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utString:
-            has<StringRef>(attribute, key_columns, out);
-            break;
-
-        case AttributeUnderlyingType::utDecimal32:
-            has<Decimal32>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utDecimal64:
-            has<Decimal64>(attribute, key_columns, out);
-            break;
-        case AttributeUnderlyingType::utDecimal128:
-            has<Decimal128>(attribute, key_columns, out);
-            break;
+        uint8_t addrv6_buf[IPV6_BINARY_LENGTH];
+        for (const auto i : ext::range(0, rows))
+        {
+            auto addrv4 = UInt32(first_column->get64(i));
+            auto found = tryLookupIPv4(addrv4, addrv6_buf);
+            out[i] = (found != ipNotFound());
+        }
     }
+    else
+    {
+        for (const auto i : ext::range(0, rows))
+        {
+            auto addr = first_column->getDataAt(i);
+            if (unlikely(addr.size != IPV6_BINARY_LENGTH))
+                throw Exception("Expected key to be FixedString(16)", ErrorCodes::LOGICAL_ERROR);
+
+            auto found = tryLookupIPv6(reinterpret_cast<const uint8_t *>(addr.data));
+            out[i] = (found != ipNotFound());
+        }
+    }
+
+    query_count.fetch_add(rows, std::memory_order_relaxed);
 }
 
 void IPAddressDictionary::createAttributes()
@@ -960,37 +935,6 @@ const IPAddressDictionary::Attribute & IPAddressDictionary::getAttribute(const s
     return attributes[it->second];
 }
 
-template <typename T>
-void IPAddressDictionary::has(const Attribute &, const Columns & key_columns, PaddedPODArray<UInt8> & out) const
-{
-    const auto first_column = key_columns.front();
-    const auto rows = first_column->size();
-    if (first_column->isNumeric())
-    {
-        uint8_t addrv6_buf[IPV6_BINARY_LENGTH];
-        for (const auto i : ext::range(0, rows))
-        {
-            auto addrv4 = UInt32(first_column->get64(i));
-            auto found = tryLookupIPv4(addrv4, addrv6_buf);
-            out[i] = (found != ipNotFound());
-        }
-    }
-    else
-    {
-        for (const auto i : ext::range(0, rows))
-        {
-            auto addr = first_column->getDataAt(i);
-            if (unlikely(addr.size != IPV6_BINARY_LENGTH))
-                throw Exception("Expected key to be FixedString(16)", ErrorCodes::LOGICAL_ERROR);
-
-            auto found = tryLookupIPv6(reinterpret_cast<const uint8_t *>(addr.data));
-            out[i] = (found != ipNotFound());
-        }
-    }
-
-    query_count.fetch_add(rows, std::memory_order_relaxed);
-}
-
 Columns IPAddressDictionary::getKeyColumns() const
 {
     const auto * ipv4_col = std::get_if<IPv4Container>(&ip_column);
diff --git a/src/Dictionaries/IPAddressDictionary.h b/src/Dictionaries/IPAddressDictionary.h
index bf0a9755445..c63e6b01adf 100644
--- a/src/Dictionaries/IPAddressDictionary.h
+++ b/src/Dictionaries/IPAddressDictionary.h
@@ -228,9 +228,6 @@ private:
 
     const Attribute & getAttribute(const std::string & attribute_name) const;
 
-    template <typename T>
-    void has(const Attribute & attribute, const Columns & key_columns, PaddedPODArray<UInt8> & out) const;
-
     Columns getKeyColumns() const;
     RowIdxConstIter ipNotFound() const;
     RowIdxConstIter tryLookupIPv4(UInt32 addr, uint8_t * buf) const;

From be85056c716885607a2e89116d971e52f0c8055a Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Thu, 24 Dec 2020 11:25:47 +0300
Subject: [PATCH 377/504] added translation

---
 .../functions/other-functions.md              |  4 +-
 .../functions/other-functions.md              | 99 ++++++++++++++++++-
 2 files changed, 100 insertions(+), 3 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index b89b648bd8b..2dab5f45259 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1177,7 +1177,7 @@ Takes state of aggregate function. Returns result of aggregation (finalized stat
 **Syntax** 
 
 ``` sql
-finalizeAggregation(state);
+finalizeAggregation(state)
 ```
 
 **Parameters**
@@ -1246,7 +1246,7 @@ SELECT
     number,
     finalizeAggregation(one_row_sum_state) AS one_row_sum,
     runningAccumulate(one_row_sum_state) AS cumulative_sum
-FROM numbers(10)
+FROM numbers(10);
 ```
 
 Result:
diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index 91bf0f5b3a0..5a812a279c5 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1104,7 +1104,104 @@ SELECT formatReadableSize(filesystemCapacity()) AS "Capacity", toTypeName(filesy
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Принимает состояние агрегатной функции. Возвращает результат агрегирования.
+Принимает состояние агрегатной функции. Возвращает результат агрегирования. Полезно с комбинатором [-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state).
+
+**Синтаксис** 
+
+``` sql
+finalizeAggregation(state)
+```
+
+**Параметры**
+
+-   `state` — Состояние агрегатной функции. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction).
+
+**Возвращаемые значения**
+
+-   Значения, которые были агрегированы.
+
+Тип: тип, значения которого были агрегированы.
+
+**Примеры**
+
+Запрос:
+
+```sql
+SELECT finalizeAggregation(( SELECT countState(number) FROM numbers(10)));
+```
+
+Результат:
+
+```text
+┌─finalizeAggregation(_subquery16)─┐
+│                               10 │
+└──────────────────────────────────┘
+```
+
+Запрос:
+
+```sql
+SELECT finalizeAggregation(( SELECT sumState(number) FROM numbers(10)));
+```
+
+Результат:
+
+```text
+┌─finalizeAggregation(_subquery20)─┐
+│                               45 │
+└──────────────────────────────────┘
+```
+
+Обратите внимания, что значения NULL игнорируются. 
+
+Запрос:
+
+```sql
+SELECT finalizeAggregation(arrayReduce('anyState', [NULL, 2, 3]));
+```
+
+Результат:
+
+```text
+┌─finalizeAggregation(arrayReduce('anyState', [NULL, 2, 3]))─┐
+│                                                          2 │
+└────────────────────────────────────────────────────────────┘
+```
+
+Комбинированный пример:
+
+Запрос:
+
+```sql
+WITH initializeAggregation('sumState', number) AS one_row_sum_state
+SELECT
+    number,
+    finalizeAggregation(one_row_sum_state) AS one_row_sum,
+    runningAccumulate(one_row_sum_state) AS cumulative_sum
+FROM numbers(10);
+```
+
+Результат:
+
+```text
+┌─number─┬─one_row_sum─┬─cumulative_sum─┐
+│      0 │           0 │              0 │
+│      1 │           1 │              1 │
+│      2 │           2 │              3 │
+│      3 │           3 │              6 │
+│      4 │           4 │             10 │
+│      5 │           5 │             15 │
+│      6 │           6 │             21 │
+│      7 │           7 │             28 │
+│      8 │           8 │             36 │
+│      9 │           9 │             45 │
+└────────┴─────────────┴────────────────┘
+```
+
+**Смотрите также**
+
+-   [arrayReduce](../../sql-reference/functions/array-functions.md#arrayreduce)
+-   [initializeAggregation](../../sql-reference/aggregate-functions/reference/initializeAggregation.md)
 
 ## runningAccumulate {#runningaccumulate}
 

From f0f91fafafaa7e7359a65b14019881846409b8d4 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Thu, 24 Dec 2020 11:44:45 +0300
Subject: [PATCH 378/504] fixed a link

---
 docs/ru/sql-reference/functions/other-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index 5a812a279c5..e00c93fcff6 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1104,7 +1104,7 @@ SELECT formatReadableSize(filesystemCapacity()) AS "Capacity", toTypeName(filesy
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Принимает состояние агрегатной функции. Возвращает результат агрегирования. Полезно с комбинатором [-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state).
+Принимает состояние агрегатной функции. Возвращает результат агрегирования. Полезно с комбинатором [-State](../../sql-reference/aggregate-functions/combinators.md#state).
 
 **Синтаксис** 
 

From 912995cbaeedd487b72a7a4954e19637fc99dec1 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 11:49:55 +0300
Subject: [PATCH 379/504] some provision for aggregate fns as window fn args

(doesn't work yet)

also a perf test w/LIMIT BY
---
 src/Interpreters/ActionsVisitor.cpp           | 13 +++++++
 src/Interpreters/GetAggregatesVisitor.h       |  7 ++--
 src/Interpreters/TreeRewriter.cpp             | 28 +++++++++++++--
 tests/performance/window_functions.xml        | 36 +++++++++++++++++++
 .../01591_window_functions.reference          |  6 ++++
 .../0_stateless/01591_window_functions.sql    |  4 +++
 6 files changed, 89 insertions(+), 5 deletions(-)
 create mode 100644 tests/performance/window_functions.xml

diff --git a/src/Interpreters/ActionsVisitor.cpp b/src/Interpreters/ActionsVisitor.cpp
index c7bbc019518..bb47129f58c 100644
--- a/src/Interpreters/ActionsVisitor.cpp
+++ b/src/Interpreters/ActionsVisitor.cpp
@@ -749,6 +749,19 @@ void ActionsMatcher::visit(const ASTFunction & node, const ASTPtr & ast, Data &
             visit(node.window_order_by->clone(), data);
         }
 
+        // Also manually add columns for arguments of the window function itself.
+        // ActionVisitor is written in such a way that this method must itself
+        // descend into all needed function children. Window functions can't have
+        // any special functions as argument, so the code below that handles
+        // special arguments is not needed. This is analogous to the
+        // appendWindowFunctionsArguments() in SelectQueryExpressionAnalyzer and
+        // partially duplicates its code. Probably we can remove most of the
+        // logic from that function, but I don't yet have it all figured out...
+        for (const auto & arg : node.arguments->children)
+        {
+            visit(arg, data);
+        }
+
         // Don't need to do anything more for window functions here -- the
         // resulting column is added in ExpressionAnalyzer, similar to the
         // aggregate functions.
diff --git a/src/Interpreters/GetAggregatesVisitor.h b/src/Interpreters/GetAggregatesVisitor.h
index d416a5f240e..266187aaffb 100644
--- a/src/Interpreters/GetAggregatesVisitor.h
+++ b/src/Interpreters/GetAggregatesVisitor.h
@@ -33,11 +33,14 @@ public:
             return false;
         if (auto * func = node->as<ASTFunction>())
         {
-            if (isAggregateFunction(*func)
-                || func->is_window_function)
+            if (isAggregateFunction(*func))
             {
                 return false;
             }
+
+            // Window functions can contain aggregation results as arguments
+            // to the window functions, or columns of PARTITION BY or ORDER BY
+            // of the window.
         }
         return true;
     }
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 9816d1a3940..70d35fd40a8 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -29,6 +29,7 @@
 #include <DataTypes/DataTypeNullable.h>
 
 #include <IO/WriteHelpers.h>
+#include <IO/WriteBufferFromOStream.h>
 #include <Storages/IStorage.h>
 
 #include <AggregateFunctions/AggregateFunctionFactory.h>
@@ -445,6 +446,8 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
             for (auto & arg : node->arguments->children)
             {
                 assertNoAggregates(arg, "inside another aggregate function");
+                // We also can't have window functions inside aggregate functions,
+                // because the window functions are calculated later.
                 assertNoWindows(arg, "inside an aggregate function");
             }
         }
@@ -454,7 +457,9 @@ std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQu
 
 std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSelectQuery & select_query)
 {
-    /// There can not be window functions inside the WHERE and PREWHERE.
+    /// There can not be window functions inside the WHERE, PREWHERE and HAVING
+    if (select_query.having())
+        assertNoWindows(select_query.having(), "in HAVING");
     if (select_query.where())
         assertNoWindows(select_query.where(), "in WHERE");
     if (select_query.prewhere())
@@ -463,17 +468,34 @@ std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSel
     GetAggregatesVisitor::Data data;
     GetAggregatesVisitor(data).visit(query);
 
-    /// There can not be other window functions within the aggregate functions.
+    /// Window functions cannot be inside aggregates or other window functions.
+    /// Aggregate functions can be inside window functions because they are
+    /// calculated earlier.
     for (const ASTFunction * node : data.window_functions)
     {
         if (node->arguments)
         {
             for (auto & arg : node->arguments->children)
             {
-                assertNoAggregates(arg, "inside a window function");
                 assertNoWindows(arg, "inside another window function");
             }
         }
+
+        if (node->window_partition_by)
+        {
+            for (auto & arg : node->window_partition_by->children)
+            {
+                assertNoWindows(arg, "inside PARTITION BY of a window");
+            }
+        }
+
+        if (node->window_order_by)
+        {
+            for (auto & arg : node->window_order_by->children)
+            {
+                assertNoWindows(arg, "inside ORDER BY of a window");
+            }
+        }
     }
 
     return data.window_functions;
diff --git a/tests/performance/window_functions.xml b/tests/performance/window_functions.xml
new file mode 100644
index 00000000000..e4578a575a0
--- /dev/null
+++ b/tests/performance/window_functions.xml
@@ -0,0 +1,36 @@
+<test>
+    <preconditions>
+        <table_exists>hits_100m_single</table_exists>
+    </preconditions>
+
+    <!--
+        For some counters, find top 10 users by the numer of records.
+        First with LIMIT BY, next with window functions.
+    -->
+    <query><![CDATA[
+        select CounterID, UserID, count(*) c
+        from hits_100m_single
+        where CounterID < 10000
+        group by CounterID, UserID
+        order by c desc
+        limit 10 by CounterID
+        format Null
+    ]]></query>
+
+    <query><![CDATA[
+        select *
+        from (
+            select *,
+                count() over (partition by CounterID order by c desc) rank
+                from (
+                    select CounterID, UserID, count() c
+                    from hits_100m_single
+                    where CounterID < 10000
+                    group by CounterID, UserID
+                )
+        )
+        where rank <= 10
+        format Null
+    ]]></query>
+
+</test>
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 19509de8a41..704c5a98811 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -160,6 +160,12 @@ select median(x) over (partition by x) from (select 1 x);
 1
 select groupArray(number) over () from numbers(3);
 
+-- This one tests we properly process the window  function arguments.
+-- Seen errors like 'column `1` not found' from count(1).
+
 [0]
 [0,1]
 [0,1,2]
+select count(1) over (), max(number + 1) over () from numbers(3);
+
+1	3
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index 4bd9f1b7693..fd2831ea2ac 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -53,3 +53,7 @@ select median(x) over (partition by x) from (select 1 x);
 
 -- an empty window definition is valid as well
 select groupArray(number) over () from numbers(3);
+
+-- This one tests we properly process the window  function arguments.
+-- Seen errors like 'column `1` not found' from count(1).
+select count(1) over (), max(number + 1) over () from numbers(3);

From 2bb0f5772e7c6003eeee11de49244180c18e181f Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 12:06:12 +0300
Subject: [PATCH 380/504] proper signature

---
 src/Processors/QueryPlan/QueryPlan.h | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/src/Processors/QueryPlan/QueryPlan.h b/src/Processors/QueryPlan/QueryPlan.h
index 5ae991a328f..4c55718a062 100644
--- a/src/Processors/QueryPlan/QueryPlan.h
+++ b/src/Processors/QueryPlan/QueryPlan.h
@@ -98,9 +98,6 @@ private:
     std::vector<std::shared_ptr<Context>> interpreter_context;
 };
 
-void debugExplainStep(
-    const IQueryPlanStep & step,
-    QueryPlanStepFormatSettings & settings,
-    const QueryPlan::ExplainPlanOptions & options);
+std::string debugExplainStep(const IQueryPlanStep & step);
 
 }

From c6c2b3ba00bfbd85e4ac309c306b7619dc96f40a Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 12:09:35 +0300
Subject: [PATCH 381/504] Remove the forward declaration changes

---
 src/Processors/QueryPlan/AggregatingStep.cpp  |  4 +--
 src/Processors/QueryPlan/AggregatingStep.h    |  4 +--
 src/Processors/QueryPlan/ArrayJoinStep.cpp    |  2 +-
 src/Processors/QueryPlan/ArrayJoinStep.h      |  2 +-
 src/Processors/QueryPlan/CreatingSetsStep.cpp |  4 +--
 src/Processors/QueryPlan/CreatingSetsStep.h   |  4 +--
 src/Processors/QueryPlan/DistinctStep.cpp     |  2 +-
 src/Processors/QueryPlan/DistinctStep.h       |  2 +-
 src/Processors/QueryPlan/ExpressionStep.cpp   |  2 +-
 src/Processors/QueryPlan/ExpressionStep.h     |  2 +-
 src/Processors/QueryPlan/FillingStep.cpp      |  2 +-
 src/Processors/QueryPlan/FillingStep.h        |  2 +-
 src/Processors/QueryPlan/FilterStep.cpp       |  2 +-
 src/Processors/QueryPlan/FilterStep.h         |  2 +-
 .../QueryPlan/FinishSortingStep.cpp           |  2 +-
 src/Processors/QueryPlan/FinishSortingStep.h  |  2 +-
 src/Processors/QueryPlan/IQueryPlanStep.cpp   |  8 +++---
 src/Processors/QueryPlan/IQueryPlanStep.h     | 25 +++++++++----------
 src/Processors/QueryPlan/ISourceStep.cpp      |  2 +-
 src/Processors/QueryPlan/ISourceStep.h        |  2 +-
 .../QueryPlan/ITransformingStep.cpp           |  2 +-
 src/Processors/QueryPlan/ITransformingStep.h  |  2 +-
 src/Processors/QueryPlan/LimitByStep.cpp      |  2 +-
 src/Processors/QueryPlan/LimitByStep.h        |  2 +-
 src/Processors/QueryPlan/LimitStep.cpp        |  2 +-
 src/Processors/QueryPlan/LimitStep.h          |  2 +-
 src/Processors/QueryPlan/MergeSortingStep.cpp |  2 +-
 src/Processors/QueryPlan/MergeSortingStep.h   |  2 +-
 .../QueryPlan/MergingAggregatedStep.cpp       |  2 +-
 .../QueryPlan/MergingAggregatedStep.h         |  2 +-
 src/Processors/QueryPlan/MergingFinal.cpp     |  2 +-
 src/Processors/QueryPlan/MergingFinal.h       |  2 +-
 .../QueryPlan/MergingSortedStep.cpp           |  2 +-
 src/Processors/QueryPlan/MergingSortedStep.h  |  2 +-
 src/Processors/QueryPlan/OffsetStep.cpp       |  2 +-
 src/Processors/QueryPlan/OffsetStep.h         |  2 +-
 .../QueryPlan/PartialSortingStep.cpp          |  2 +-
 src/Processors/QueryPlan/PartialSortingStep.h |  2 +-
 src/Processors/QueryPlan/TotalsHavingStep.cpp |  2 +-
 src/Processors/QueryPlan/TotalsHavingStep.h   |  2 +-
 src/Processors/QueryPlan/UnionStep.cpp        |  2 +-
 src/Processors/QueryPlan/UnionStep.h          |  2 +-
 src/Processors/QueryPlan/WindowStep.cpp       |  2 +-
 src/Processors/QueryPlan/WindowStep.h         |  2 +-
 44 files changed, 61 insertions(+), 64 deletions(-)

diff --git a/src/Processors/QueryPlan/AggregatingStep.cpp b/src/Processors/QueryPlan/AggregatingStep.cpp
index 8b787859467..e8d4a262366 100644
--- a/src/Processors/QueryPlan/AggregatingStep.cpp
+++ b/src/Processors/QueryPlan/AggregatingStep.cpp
@@ -156,12 +156,12 @@ void AggregatingStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void AggregatingStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void AggregatingStep::describeActions(FormatSettings & settings) const
 {
     params.explain(settings.out, settings.offset);
 }
 
-void AggregatingStep::describePipeline(QueryPlanStepFormatSettings & settings) const
+void AggregatingStep::describePipeline(FormatSettings & settings) const
 {
     if (!aggregating.empty())
         IQueryPlanStep::describePipeline(aggregating, settings);
diff --git a/src/Processors/QueryPlan/AggregatingStep.h b/src/Processors/QueryPlan/AggregatingStep.h
index 3d12e79a662..853173895b3 100644
--- a/src/Processors/QueryPlan/AggregatingStep.h
+++ b/src/Processors/QueryPlan/AggregatingStep.h
@@ -29,8 +29,8 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings &) const override;
-    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings &) const override;
+    void describePipeline(FormatSettings & settings) const override;
 
 private:
     Aggregator::Params params;
diff --git a/src/Processors/QueryPlan/ArrayJoinStep.cpp b/src/Processors/QueryPlan/ArrayJoinStep.cpp
index 01cfd7c561a..b71d2176e52 100644
--- a/src/Processors/QueryPlan/ArrayJoinStep.cpp
+++ b/src/Processors/QueryPlan/ArrayJoinStep.cpp
@@ -69,7 +69,7 @@ void ArrayJoinStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void ArrayJoinStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void ArrayJoinStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     bool first = true;
diff --git a/src/Processors/QueryPlan/ArrayJoinStep.h b/src/Processors/QueryPlan/ArrayJoinStep.h
index 338ec7f0583..92c7e0a1304 100644
--- a/src/Processors/QueryPlan/ArrayJoinStep.h
+++ b/src/Processors/QueryPlan/ArrayJoinStep.h
@@ -15,7 +15,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     void updateInputStream(DataStream input_stream, Block result_header);
 
diff --git a/src/Processors/QueryPlan/CreatingSetsStep.cpp b/src/Processors/QueryPlan/CreatingSetsStep.cpp
index c6b77bc46c3..5868a7045f7 100644
--- a/src/Processors/QueryPlan/CreatingSetsStep.cpp
+++ b/src/Processors/QueryPlan/CreatingSetsStep.cpp
@@ -47,7 +47,7 @@ void CreatingSetStep::transformPipeline(QueryPipeline & pipeline)
     pipeline.addCreatingSetsTransform(getOutputStream().header, std::move(subquery_for_set), network_transfer_limits, context);
 }
 
-void CreatingSetStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void CreatingSetStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
 
@@ -102,7 +102,7 @@ QueryPipelinePtr CreatingSetsStep::updatePipeline(QueryPipelines pipelines)
     return main_pipeline;
 }
 
-void CreatingSetsStep::describePipeline(QueryPlanStepFormatSettings & settings) const
+void CreatingSetsStep::describePipeline(FormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/CreatingSetsStep.h b/src/Processors/QueryPlan/CreatingSetsStep.h
index f610eeb64d9..ec13ab2052e 100644
--- a/src/Processors/QueryPlan/CreatingSetsStep.h
+++ b/src/Processors/QueryPlan/CreatingSetsStep.h
@@ -22,7 +22,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     String description;
@@ -40,7 +40,7 @@ public:
 
     QueryPipelinePtr updatePipeline(QueryPipelines pipelines) override;
 
-    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
+    void describePipeline(FormatSettings & settings) const override;
 
 private:
     Processors processors;
diff --git a/src/Processors/QueryPlan/DistinctStep.cpp b/src/Processors/QueryPlan/DistinctStep.cpp
index ed2bae002ea..60966b08beb 100644
--- a/src/Processors/QueryPlan/DistinctStep.cpp
+++ b/src/Processors/QueryPlan/DistinctStep.cpp
@@ -79,7 +79,7 @@ void DistinctStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void DistinctStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void DistinctStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Columns: ";
diff --git a/src/Processors/QueryPlan/DistinctStep.h b/src/Processors/QueryPlan/DistinctStep.h
index 901e199ca39..4bfd73ce044 100644
--- a/src/Processors/QueryPlan/DistinctStep.h
+++ b/src/Processors/QueryPlan/DistinctStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     SizeLimits set_size_limits;
diff --git a/src/Processors/QueryPlan/ExpressionStep.cpp b/src/Processors/QueryPlan/ExpressionStep.cpp
index 028c9502c71..f7ca0a972dc 100644
--- a/src/Processors/QueryPlan/ExpressionStep.cpp
+++ b/src/Processors/QueryPlan/ExpressionStep.cpp
@@ -87,7 +87,7 @@ void ExpressionStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void ExpressionStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void ExpressionStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     bool first = true;
diff --git a/src/Processors/QueryPlan/ExpressionStep.h b/src/Processors/QueryPlan/ExpressionStep.h
index 2519765dbae..60f186688b0 100644
--- a/src/Processors/QueryPlan/ExpressionStep.h
+++ b/src/Processors/QueryPlan/ExpressionStep.h
@@ -26,7 +26,7 @@ public:
 
     void updateInputStream(DataStream input_stream, bool keep_header);
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     const ActionsDAGPtr & getExpression() const { return actions_dag; }
 
diff --git a/src/Processors/QueryPlan/FillingStep.cpp b/src/Processors/QueryPlan/FillingStep.cpp
index 18c71b89601..1a8fba97ee2 100644
--- a/src/Processors/QueryPlan/FillingStep.cpp
+++ b/src/Processors/QueryPlan/FillingStep.cpp
@@ -43,7 +43,7 @@ void FillingStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void FillingStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void FillingStep::describeActions(FormatSettings & settings) const
 {
     settings.out << String(settings.offset, ' ');
     dumpSortDescription(sort_description, input_streams.front().header, settings.out);
diff --git a/src/Processors/QueryPlan/FillingStep.h b/src/Processors/QueryPlan/FillingStep.h
index a7836e4de68..85736464a6c 100644
--- a/src/Processors/QueryPlan/FillingStep.h
+++ b/src/Processors/QueryPlan/FillingStep.h
@@ -15,7 +15,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     SortDescription sort_description;
diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp
index 98f111a1529..8fb405e685b 100644
--- a/src/Processors/QueryPlan/FilterStep.cpp
+++ b/src/Processors/QueryPlan/FilterStep.cpp
@@ -80,7 +80,7 @@ void FilterStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void FilterStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void FilterStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Filter column: " << filter_column_name << '\n';
diff --git a/src/Processors/QueryPlan/FilterStep.h b/src/Processors/QueryPlan/FilterStep.h
index 561fbf347da..72b02624faa 100644
--- a/src/Processors/QueryPlan/FilterStep.h
+++ b/src/Processors/QueryPlan/FilterStep.h
@@ -22,7 +22,7 @@ public:
 
     void updateInputStream(DataStream input_stream, bool keep_header);
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     const ActionsDAGPtr & getExpression() const { return actions_dag; }
     const String & getFilterColumnName() const { return filter_column_name; }
diff --git a/src/Processors/QueryPlan/FinishSortingStep.cpp b/src/Processors/QueryPlan/FinishSortingStep.cpp
index 7af8fad2980..d883bd0e0dd 100644
--- a/src/Processors/QueryPlan/FinishSortingStep.cpp
+++ b/src/Processors/QueryPlan/FinishSortingStep.cpp
@@ -85,7 +85,7 @@ void FinishSortingStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void FinishSortingStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void FinishSortingStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
 
diff --git a/src/Processors/QueryPlan/FinishSortingStep.h b/src/Processors/QueryPlan/FinishSortingStep.h
index 25dc0b78bda..4bb62037faa 100644
--- a/src/Processors/QueryPlan/FinishSortingStep.h
+++ b/src/Processors/QueryPlan/FinishSortingStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/IQueryPlanStep.cpp b/src/Processors/QueryPlan/IQueryPlanStep.cpp
index a864043ddcb..f06897e8488 100644
--- a/src/Processors/QueryPlan/IQueryPlanStep.cpp
+++ b/src/Processors/QueryPlan/IQueryPlanStep.cpp
@@ -18,8 +18,7 @@ const DataStream & IQueryPlanStep::getOutputStream() const
     return *output_stream;
 }
 
-static void doDescribeHeader(const Block & header, size_t count,
-    QueryPlanStepFormatSettings & settings)
+static void doDescribeHeader(const Block & header, size_t count, IQueryPlanStep::FormatSettings & settings)
 {
     String prefix(settings.offset, settings.indent_char);
     prefix += "Header";
@@ -53,8 +52,7 @@ static void doDescribeHeader(const Block & header, size_t count,
     }
 }
 
-static void doDescribeProcessor(const IProcessor & processor, size_t count,
-    QueryPlanStepFormatSettings & settings)
+static void doDescribeProcessor(const IProcessor & processor, size_t count, IQueryPlanStep::FormatSettings & settings)
 {
     settings.out << String(settings.offset, settings.indent_char) << processor.getName();
     if (count > 1)
@@ -91,7 +89,7 @@ static void doDescribeProcessor(const IProcessor & processor, size_t count,
     settings.offset += settings.indent;
 }
 
-void IQueryPlanStep::describePipeline(const Processors & processors, QueryPlanStepFormatSettings & settings)
+void IQueryPlanStep::describePipeline(const Processors & processors, FormatSettings & settings)
 {
     const IProcessor * prev = nullptr;
     size_t count = 0;
diff --git a/src/Processors/QueryPlan/IQueryPlanStep.h b/src/Processors/QueryPlan/IQueryPlanStep.h
index 1de1bbc438f..dfdbc9df442 100644
--- a/src/Processors/QueryPlan/IQueryPlanStep.h
+++ b/src/Processors/QueryPlan/IQueryPlanStep.h
@@ -61,16 +61,6 @@ public:
 
 using DataStreams = std::vector<DataStream>;
 
-// Not a nested class so that we can forward declare it.
-struct QueryPlanStepFormatSettings
-{
-    WriteBuffer & out;
-    size_t offset = 0;
-    const size_t indent = 2;
-    const char indent_char = ' ';
-    const bool write_header = false;
-};
-
 /// Single step of query plan.
 class IQueryPlanStep
 {
@@ -96,11 +86,20 @@ public:
     const std::string & getStepDescription() const { return step_description; }
     void setStepDescription(std::string description) { step_description = std::move(description); }
 
+    struct FormatSettings
+    {
+        WriteBuffer & out;
+        size_t offset = 0;
+        const size_t indent = 2;
+        const char indent_char = ' ';
+        const bool write_header = false;
+    };
+
     /// Get detailed description of step actions. This is shown in EXPLAIN query with options `actions = 1`.
-    virtual void describeActions(QueryPlanStepFormatSettings & /*settings*/) const {}
+    virtual void describeActions(FormatSettings & /*settings*/) const {}
 
     /// Get description of processors added in current step. Should be called after updatePipeline().
-    virtual void describePipeline(QueryPlanStepFormatSettings & /*settings*/) const {}
+    virtual void describePipeline(FormatSettings & /*settings*/) const {}
 
 protected:
     DataStreams input_streams;
@@ -109,7 +108,7 @@ protected:
     /// Text description about what current step does.
     std::string step_description;
 
-    static void describePipeline(const Processors & processors, QueryPlanStepFormatSettings & settings);
+    static void describePipeline(const Processors & processors, FormatSettings & settings);
 };
 
 using QueryPlanStepPtr = std::unique_ptr<IQueryPlanStep>;
diff --git a/src/Processors/QueryPlan/ISourceStep.cpp b/src/Processors/QueryPlan/ISourceStep.cpp
index 367be8e8ea5..1796f033896 100644
--- a/src/Processors/QueryPlan/ISourceStep.cpp
+++ b/src/Processors/QueryPlan/ISourceStep.cpp
@@ -19,7 +19,7 @@ QueryPipelinePtr ISourceStep::updatePipeline(QueryPipelines)
     return pipeline;
 }
 
-void ISourceStep::describePipeline(QueryPlanStepFormatSettings & settings) const
+void ISourceStep::describePipeline(FormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/ISourceStep.h b/src/Processors/QueryPlan/ISourceStep.h
index 050e39ac11b..fdb3dd566cb 100644
--- a/src/Processors/QueryPlan/ISourceStep.h
+++ b/src/Processors/QueryPlan/ISourceStep.h
@@ -14,7 +14,7 @@ public:
 
     virtual void initializePipeline(QueryPipeline & pipeline) = 0;
 
-    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
+    void describePipeline(FormatSettings & settings) const override;
 
 protected:
     /// We collect processors got after pipeline transformation.
diff --git a/src/Processors/QueryPlan/ITransformingStep.cpp b/src/Processors/QueryPlan/ITransformingStep.cpp
index 5430bafa27a..cb27bf38278 100644
--- a/src/Processors/QueryPlan/ITransformingStep.cpp
+++ b/src/Processors/QueryPlan/ITransformingStep.cpp
@@ -65,7 +65,7 @@ void ITransformingStep::updateDistinctColumns(const Block & res_header, NameSet
     }
 }
 
-void ITransformingStep::describePipeline(QueryPlanStepFormatSettings & settings) const
+void ITransformingStep::describePipeline(FormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/ITransformingStep.h b/src/Processors/QueryPlan/ITransformingStep.h
index b1bfebd4571..bd99478ec37 100644
--- a/src/Processors/QueryPlan/ITransformingStep.h
+++ b/src/Processors/QueryPlan/ITransformingStep.h
@@ -55,7 +55,7 @@ public:
     const TransformTraits & getTransformTraits() const { return transform_traits; }
     const DataStreamTraits & getDataStreamTraits() const { return data_stream_traits; }
 
-    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
+    void describePipeline(FormatSettings & settings) const override;
 
 protected:
     /// Clear distinct_columns if res_header doesn't contain all of them.
diff --git a/src/Processors/QueryPlan/LimitByStep.cpp b/src/Processors/QueryPlan/LimitByStep.cpp
index 15c64fcea44..9fcf5b60164 100644
--- a/src/Processors/QueryPlan/LimitByStep.cpp
+++ b/src/Processors/QueryPlan/LimitByStep.cpp
@@ -46,7 +46,7 @@ void LimitByStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void LimitByStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void LimitByStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
 
diff --git a/src/Processors/QueryPlan/LimitByStep.h b/src/Processors/QueryPlan/LimitByStep.h
index 01dcff5c912..9320735640c 100644
--- a/src/Processors/QueryPlan/LimitByStep.h
+++ b/src/Processors/QueryPlan/LimitByStep.h
@@ -16,7 +16,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     size_t group_length;
diff --git a/src/Processors/QueryPlan/LimitStep.cpp b/src/Processors/QueryPlan/LimitStep.cpp
index 38ec9265d4e..565d05956e5 100644
--- a/src/Processors/QueryPlan/LimitStep.cpp
+++ b/src/Processors/QueryPlan/LimitStep.cpp
@@ -50,7 +50,7 @@ void LimitStep::transformPipeline(QueryPipeline & pipeline)
     pipeline.addTransform(std::move(transform));
 }
 
-void LimitStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void LimitStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Limit " << limit << '\n';
diff --git a/src/Processors/QueryPlan/LimitStep.h b/src/Processors/QueryPlan/LimitStep.h
index a95bf28f896..eea186fea4a 100644
--- a/src/Processors/QueryPlan/LimitStep.h
+++ b/src/Processors/QueryPlan/LimitStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     size_t getLimitForSorting() const
     {
diff --git a/src/Processors/QueryPlan/MergeSortingStep.cpp b/src/Processors/QueryPlan/MergeSortingStep.cpp
index e2af023f598..b30286130b1 100644
--- a/src/Processors/QueryPlan/MergeSortingStep.cpp
+++ b/src/Processors/QueryPlan/MergeSortingStep.cpp
@@ -73,7 +73,7 @@ void MergeSortingStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void MergeSortingStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void MergeSortingStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/MergeSortingStep.h b/src/Processors/QueryPlan/MergeSortingStep.h
index c41c209dd55..a385a8a3e93 100644
--- a/src/Processors/QueryPlan/MergeSortingStep.h
+++ b/src/Processors/QueryPlan/MergeSortingStep.h
@@ -26,7 +26,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/MergingAggregatedStep.cpp b/src/Processors/QueryPlan/MergingAggregatedStep.cpp
index 2fddc88bbc1..85bbbeab59a 100644
--- a/src/Processors/QueryPlan/MergingAggregatedStep.cpp
+++ b/src/Processors/QueryPlan/MergingAggregatedStep.cpp
@@ -63,7 +63,7 @@ void MergingAggregatedStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void MergingAggregatedStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void MergingAggregatedStep::describeActions(FormatSettings & settings) const
 {
     return params->params.explain(settings.out, settings.offset);
 }
diff --git a/src/Processors/QueryPlan/MergingAggregatedStep.h b/src/Processors/QueryPlan/MergingAggregatedStep.h
index 2d9d902a294..5ffceb7f938 100644
--- a/src/Processors/QueryPlan/MergingAggregatedStep.h
+++ b/src/Processors/QueryPlan/MergingAggregatedStep.h
@@ -23,7 +23,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     AggregatingTransformParamsPtr params;
diff --git a/src/Processors/QueryPlan/MergingFinal.cpp b/src/Processors/QueryPlan/MergingFinal.cpp
index 46431e0c767..c1dec231f11 100644
--- a/src/Processors/QueryPlan/MergingFinal.cpp
+++ b/src/Processors/QueryPlan/MergingFinal.cpp
@@ -153,7 +153,7 @@ void MergingFinal::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void MergingFinal::describeActions(QueryPlanStepFormatSettings & settings) const
+void MergingFinal::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/MergingFinal.h b/src/Processors/QueryPlan/MergingFinal.h
index 207c191c9cb..c01f5c7f9a1 100644
--- a/src/Processors/QueryPlan/MergingFinal.h
+++ b/src/Processors/QueryPlan/MergingFinal.h
@@ -22,7 +22,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     size_t num_output_streams;
diff --git a/src/Processors/QueryPlan/MergingSortedStep.cpp b/src/Processors/QueryPlan/MergingSortedStep.cpp
index a1249ef412f..c59540b009f 100644
--- a/src/Processors/QueryPlan/MergingSortedStep.cpp
+++ b/src/Processors/QueryPlan/MergingSortedStep.cpp
@@ -62,7 +62,7 @@ void MergingSortedStep::transformPipeline(QueryPipeline & pipeline)
     }
 }
 
-void MergingSortedStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void MergingSortedStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/MergingSortedStep.h b/src/Processors/QueryPlan/MergingSortedStep.h
index 0c91d505f3d..483cfa5e8a7 100644
--- a/src/Processors/QueryPlan/MergingSortedStep.h
+++ b/src/Processors/QueryPlan/MergingSortedStep.h
@@ -21,7 +21,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/OffsetStep.cpp b/src/Processors/QueryPlan/OffsetStep.cpp
index 1a06f2ba05e..7ac3d3f2110 100644
--- a/src/Processors/QueryPlan/OffsetStep.cpp
+++ b/src/Processors/QueryPlan/OffsetStep.cpp
@@ -36,7 +36,7 @@ void OffsetStep::transformPipeline(QueryPipeline & pipeline)
     pipeline.addTransform(std::move(transform));
 }
 
-void OffsetStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void OffsetStep::describeActions(FormatSettings & settings) const
 {
     settings.out << String(settings.offset, ' ') << "Offset " << offset << '\n';
 }
diff --git a/src/Processors/QueryPlan/OffsetStep.h b/src/Processors/QueryPlan/OffsetStep.h
index 18ea58635af..17949371edf 100644
--- a/src/Processors/QueryPlan/OffsetStep.h
+++ b/src/Processors/QueryPlan/OffsetStep.h
@@ -15,7 +15,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     size_t offset;
diff --git a/src/Processors/QueryPlan/PartialSortingStep.cpp b/src/Processors/QueryPlan/PartialSortingStep.cpp
index 072217b4d0d..ce34eca9112 100644
--- a/src/Processors/QueryPlan/PartialSortingStep.cpp
+++ b/src/Processors/QueryPlan/PartialSortingStep.cpp
@@ -70,7 +70,7 @@ void PartialSortingStep::transformPipeline(QueryPipeline & pipeline)
     });
 }
 
-void PartialSortingStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void PartialSortingStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Sort description: ";
diff --git a/src/Processors/QueryPlan/PartialSortingStep.h b/src/Processors/QueryPlan/PartialSortingStep.h
index 6d9dbe2b9d8..172ef25c300 100644
--- a/src/Processors/QueryPlan/PartialSortingStep.h
+++ b/src/Processors/QueryPlan/PartialSortingStep.h
@@ -20,7 +20,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
     /// Add limit or change it to lower value.
     void updateLimit(size_t limit_);
diff --git a/src/Processors/QueryPlan/TotalsHavingStep.cpp b/src/Processors/QueryPlan/TotalsHavingStep.cpp
index 310f657b39d..9947875e679 100644
--- a/src/Processors/QueryPlan/TotalsHavingStep.cpp
+++ b/src/Processors/QueryPlan/TotalsHavingStep.cpp
@@ -74,7 +74,7 @@ static String totalsModeToString(TotalsMode totals_mode, double auto_include_thr
     __builtin_unreachable();
 }
 
-void TotalsHavingStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void TotalsHavingStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Filter column: " << filter_column_name << '\n';
diff --git a/src/Processors/QueryPlan/TotalsHavingStep.h b/src/Processors/QueryPlan/TotalsHavingStep.h
index 88fd9bbb597..7c1638013e5 100644
--- a/src/Processors/QueryPlan/TotalsHavingStep.h
+++ b/src/Processors/QueryPlan/TotalsHavingStep.h
@@ -26,7 +26,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     bool overflow_row;
diff --git a/src/Processors/QueryPlan/UnionStep.cpp b/src/Processors/QueryPlan/UnionStep.cpp
index 413b52c69c2..630ff53f47d 100644
--- a/src/Processors/QueryPlan/UnionStep.cpp
+++ b/src/Processors/QueryPlan/UnionStep.cpp
@@ -36,7 +36,7 @@ QueryPipelinePtr UnionStep::updatePipeline(QueryPipelines pipelines)
     return pipeline;
 }
 
-void UnionStep::describePipeline(QueryPlanStepFormatSettings & settings) const
+void UnionStep::describePipeline(FormatSettings & settings) const
 {
     IQueryPlanStep::describePipeline(processors, settings);
 }
diff --git a/src/Processors/QueryPlan/UnionStep.h b/src/Processors/QueryPlan/UnionStep.h
index aeac7fa54ff..e2e1f2c9efa 100644
--- a/src/Processors/QueryPlan/UnionStep.h
+++ b/src/Processors/QueryPlan/UnionStep.h
@@ -15,7 +15,7 @@ public:
 
     QueryPipelinePtr updatePipeline(QueryPipelines pipelines) override;
 
-    void describePipeline(QueryPlanStepFormatSettings & settings) const override;
+    void describePipeline(FormatSettings & settings) const override;
 
 private:
     Block header;
diff --git a/src/Processors/QueryPlan/WindowStep.cpp b/src/Processors/QueryPlan/WindowStep.cpp
index d2d24985937..82c589b8b20 100644
--- a/src/Processors/QueryPlan/WindowStep.cpp
+++ b/src/Processors/QueryPlan/WindowStep.cpp
@@ -71,7 +71,7 @@ void WindowStep::transformPipeline(QueryPipeline & pipeline)
         "WindowStep transform for '" + window_description.window_name + "'");
 }
 
-void WindowStep::describeActions(QueryPlanStepFormatSettings & settings) const
+void WindowStep::describeActions(FormatSettings & settings) const
 {
     String prefix(settings.offset, ' ');
     settings.out << prefix << "Window: (";
diff --git a/src/Processors/QueryPlan/WindowStep.h b/src/Processors/QueryPlan/WindowStep.h
index 25194820a0d..069d02c655c 100644
--- a/src/Processors/QueryPlan/WindowStep.h
+++ b/src/Processors/QueryPlan/WindowStep.h
@@ -22,7 +22,7 @@ public:
 
     void transformPipeline(QueryPipeline & pipeline) override;
 
-    void describeActions(QueryPlanStepFormatSettings & settings) const override;
+    void describeActions(FormatSettings & settings) const override;
 
 private:
     WindowDescription window_description;

From 61a35cdafd056e32879ab6389fbf33c9d89d58d9 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 12:11:37 +0300
Subject: [PATCH 382/504] final cleanup

---
 src/Processors/QueryPlan/QueryPlan.cpp | 10 +++++-----
 src/Processors/QueryPlan/QueryPlan.h   |  1 -
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/src/Processors/QueryPlan/QueryPlan.cpp b/src/Processors/QueryPlan/QueryPlan.cpp
index a2fc8c689ff..1b3ea16a213 100644
--- a/src/Processors/QueryPlan/QueryPlan.cpp
+++ b/src/Processors/QueryPlan/QueryPlan.cpp
@@ -204,7 +204,7 @@ void QueryPlan::addInterpreterContext(std::shared_ptr<Context> context)
 
 static void explainStep(
     const IQueryPlanStep & step,
-    QueryPlanStepFormatSettings & settings,
+    IQueryPlanStep::FormatSettings & settings,
     const QueryPlan::ExplainPlanOptions & options)
 {
     std::string prefix(settings.offset, ' ');
@@ -250,7 +250,7 @@ static void explainStep(
 std::string debugExplainStep(const IQueryPlanStep & step)
 {
     WriteBufferFromOwnString out;
-    QueryPlanStepFormatSettings settings{.out = out};
+    IQueryPlanStep::FormatSettings settings{.out = out};
     QueryPlan::ExplainPlanOptions options{.actions = true};
     explainStep(step, settings, options);
     return out.str();
@@ -260,7 +260,7 @@ void QueryPlan::explainPlan(WriteBuffer & buffer, const ExplainPlanOptions & opt
 {
     checkInitialized();
 
-    QueryPlanStepFormatSettings settings{.out = buffer, .write_header = options.header};
+    IQueryPlanStep::FormatSettings settings{.out = buffer, .write_header = options.header};
 
     struct Frame
     {
@@ -293,7 +293,7 @@ void QueryPlan::explainPlan(WriteBuffer & buffer, const ExplainPlanOptions & opt
     }
 }
 
-static void explainPipelineStep(IQueryPlanStep & step, QueryPlanStepFormatSettings & settings)
+static void explainPipelineStep(IQueryPlanStep & step, IQueryPlanStep::FormatSettings & settings)
 {
     settings.out << String(settings.offset, settings.indent_char) << "(" << step.getName() << ")\n";
     size_t current_offset = settings.offset;
@@ -306,7 +306,7 @@ void QueryPlan::explainPipeline(WriteBuffer & buffer, const ExplainPipelineOptio
 {
     checkInitialized();
 
-    QueryPlanStepFormatSettings settings{.out = buffer, .write_header = options.header};
+    IQueryPlanStep::FormatSettings settings{.out = buffer, .write_header = options.header};
 
     struct Frame
     {
diff --git a/src/Processors/QueryPlan/QueryPlan.h b/src/Processors/QueryPlan/QueryPlan.h
index 4c55718a062..9d2d7d93a36 100644
--- a/src/Processors/QueryPlan/QueryPlan.h
+++ b/src/Processors/QueryPlan/QueryPlan.h
@@ -13,7 +13,6 @@ class DataStream;
 
 class IQueryPlanStep;
 using QueryPlanStepPtr = std::unique_ptr<IQueryPlanStep>;
-struct QueryPlanStepFormatSettings;
 
 class QueryPipeline;
 using QueryPipelinePtr = std::unique_ptr<QueryPipeline>;

From d2a0551845b96566e454b753af63a901155f56c9 Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Thu, 24 Dec 2020 12:13:49 +0300
Subject: [PATCH 383/504] fixed arrayReduce

---
 docs/ru/sql-reference/functions/array-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/sql-reference/functions/array-functions.md b/docs/ru/sql-reference/functions/array-functions.md
index 16f21e3a80c..015d14b9de5 100644
--- a/docs/ru/sql-reference/functions/array-functions.md
+++ b/docs/ru/sql-reference/functions/array-functions.md
@@ -810,7 +810,7 @@ SELECT
 └──────────────┴───────────┘
 ```
 
-## arrayReduce (#arrayreduce}
+## arrayReduce {#arrayreduce}
 
 Применяет агрегатную функцию к элементам массива и возвращает ее результат. Имя агрегирующей функции передается как строка в одинарных кавычках `'max'`, `'sum'`. При использовании параметрических агрегатных функций, параметр указывается после имени функции в круглых скобках `'uniqUpTo(6)'`.
 

From 5e8cf89b9d6be6f58cfcda3b9cd1d0307e2eafa5 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 12:13:54 +0300
Subject: [PATCH 384/504] cleanup

---
 src/Interpreters/ExpressionAnalyzer.cpp | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 37f2791d8bd..0a380c66d23 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -981,6 +981,7 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
             getRootActionsNoMakeSet(f.function_node->clone(),
                 true /* no_subqueries */, step.actions());
 
+            // FIXME rewrite this comment
             // 1.2) result of window function: an empty INPUT.
             // It is an aggregate function, so it won't be added by getRootActions.
             // This is something of a hack. Other options:
@@ -999,12 +1000,6 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
             //     Expression pipelines. This is a "proper" way that would allow
             //     us to depend on window functions in other functions. But it's
             //     complicated so I avoid doing it for now.
-            ColumnWithTypeAndName col;
-            col.type = f.aggregate_function->getReturnType();
-            col.column = col.type->createColumn();
-            col.name = f.column_name;
-
-//            step.actions()->addInput(col);
             columns_after_window.push_back({f.column_name,
                 f.aggregate_function->getReturnType()});
 
@@ -1013,8 +1008,6 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
                 // 2.1) function arguments;
                 step.required_output.push_back(a->getColumnName());
             }
-//            // 2.2) function result;
-//            step.required_output.push_back(f.column_name);
         }
 
         // 2.3) PARTITION BY and ORDER BY columns.

From 71ab8c18fa25162105a0d0a62a1e549713e7f0ea Mon Sep 17 00:00:00 2001
From: spongedc <duchuangucas@live.com>
Date: Thu, 24 Dec 2020 17:16:47 +0800
Subject: [PATCH 385/504] support EXISTS DATABASE syntax

---
 src/Interpreters/InterpreterExistsQuery.cpp   |  6 +++++
 src/Interpreters/InterpreterFactory.cpp       |  4 ++++
 src/Parsers/ParserTablePropertiesQuery.cpp    | 24 ++++++++++++-------
 src/Parsers/TablePropertiesQueriesASTs.h      |  9 +++++++
 .../0_stateless/01048_exists_query.reference  |  2 ++
 .../0_stateless/01048_exists_query.sql        |  2 ++
 6 files changed, 39 insertions(+), 8 deletions(-)

diff --git a/src/Interpreters/InterpreterExistsQuery.cpp b/src/Interpreters/InterpreterExistsQuery.cpp
index 8539cb6816e..94a31db2e99 100644
--- a/src/Interpreters/InterpreterExistsQuery.cpp
+++ b/src/Interpreters/InterpreterExistsQuery.cpp
@@ -53,6 +53,12 @@ BlockInputStreamPtr InterpreterExistsQuery::executeImpl()
             result = DatabaseCatalog::instance().isTableExist({database, exists_query->table}, context);
         }
     }
+    else if ((exists_query = query_ptr->as<ASTExistsDatabaseQuery>()))
+    {
+        String database = context.resolveDatabase(exists_query->database);
+        context.checkAccess(AccessType::SHOW_DATABASES, database);
+        result = DatabaseCatalog::instance().isDatabaseExist(database);
+    }
     else if ((exists_query = query_ptr->as<ASTExistsDictionaryQuery>()))
     {
         if (exists_query->temporary)
diff --git a/src/Interpreters/InterpreterFactory.cpp b/src/Interpreters/InterpreterFactory.cpp
index 730f1db014c..7c2e82e76c1 100644
--- a/src/Interpreters/InterpreterFactory.cpp
+++ b/src/Interpreters/InterpreterFactory.cpp
@@ -148,6 +148,10 @@ std::unique_ptr<IInterpreter> InterpreterFactory::get(ASTPtr & query, Context &
     {
         return std::make_unique<InterpreterOptimizeQuery>(query, context);
     }
+    else if (query->as<ASTExistsDatabaseQuery>())
+    {
+        return std::make_unique<InterpreterExistsQuery>(query, context);
+    }
     else if (query->as<ASTExistsTableQuery>())
     {
         return std::make_unique<InterpreterExistsQuery>(query, context);
diff --git a/src/Parsers/ParserTablePropertiesQuery.cpp b/src/Parsers/ParserTablePropertiesQuery.cpp
index 2ee85a3330d..7073106bc13 100644
--- a/src/Parsers/ParserTablePropertiesQuery.cpp
+++ b/src/Parsers/ParserTablePropertiesQuery.cpp
@@ -34,15 +34,23 @@ bool ParserTablePropertiesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected &
     bool temporary = false;
     if (s_exists.ignore(pos, expected))
     {
-        if (s_temporary.ignore(pos, expected))
-            temporary = true;
-
-        if (s_table.checkWithoutMoving(pos, expected))
-            query = std::make_shared<ASTExistsTableQuery>();
-        else if (s_dictionary.checkWithoutMoving(pos, expected))
-            query = std::make_shared<ASTExistsDictionaryQuery>();
+        if (s_database.ignore(pos, expected))
+	    {
+            query = std::make_shared<ASTExistsDatabaseQuery>();
+	        parse_only_database_name = true;
+	    }
         else
-            query = std::make_shared<ASTExistsTableQuery>();
+        {
+            if (s_temporary.ignore(pos, expected))
+                temporary = true;
+
+            if (s_table.checkWithoutMoving(pos, expected))
+                query = std::make_shared<ASTExistsTableQuery>();
+            else if (s_dictionary.checkWithoutMoving(pos, expected))
+                query = std::make_shared<ASTExistsDictionaryQuery>();
+            else
+                query = std::make_shared<ASTExistsTableQuery>();
+        }
     }
     else if (s_show.ignore(pos, expected))
     {
diff --git a/src/Parsers/TablePropertiesQueriesASTs.h b/src/Parsers/TablePropertiesQueriesASTs.h
index 6a8e3b2ce83..3e0798dbcbd 100644
--- a/src/Parsers/TablePropertiesQueriesASTs.h
+++ b/src/Parsers/TablePropertiesQueriesASTs.h
@@ -7,6 +7,14 @@
 namespace DB
 {
 
+struct ASTExistsDatabaseQueryIDAndQueryNames
+{
+    static constexpr auto ID = "ExistsDatabaseQuery";
+    static constexpr auto Query = "EXISTS DATABASE";
+    /// No temporary databases are supported, just for parsing
+    static constexpr auto QueryTemporary = "";
+};
+
 struct ASTExistsTableQueryIDAndQueryNames
 {
     static constexpr auto ID = "ExistsTableQuery";
@@ -51,6 +59,7 @@ struct ASTDescribeQueryExistsQueryIDAndQueryNames
     static constexpr auto QueryTemporary = "DESCRIBE TEMPORARY TABLE";
 };
 
+using ASTExistsDatabaseQuery = ASTQueryWithTableAndOutputImpl<ASTExistsDatabaseQueryIDAndQueryNames>;
 using ASTExistsTableQuery = ASTQueryWithTableAndOutputImpl<ASTExistsTableQueryIDAndQueryNames>;
 using ASTExistsDictionaryQuery = ASTQueryWithTableAndOutputImpl<ASTExistsDictionaryQueryIDAndQueryNames>;
 using ASTShowCreateTableQuery = ASTQueryWithTableAndOutputImpl<ASTShowCreateTableQueryIDAndQueryNames>;
diff --git a/tests/queries/0_stateless/01048_exists_query.reference b/tests/queries/0_stateless/01048_exists_query.reference
index aa6264c9a0a..f1db7c70e71 100644
--- a/tests/queries/0_stateless/01048_exists_query.reference
+++ b/tests/queries/0_stateless/01048_exists_query.reference
@@ -2,6 +2,8 @@
 0
 0
 0
+1
+0
 0
 0
 1
diff --git a/tests/queries/0_stateless/01048_exists_query.sql b/tests/queries/0_stateless/01048_exists_query.sql
index 31b6d2af6c0..fca2c233c64 100644
--- a/tests/queries/0_stateless/01048_exists_query.sql
+++ b/tests/queries/0_stateless/01048_exists_query.sql
@@ -3,7 +3,9 @@ EXISTS TABLE db_01048.t_01048;
 EXISTS DICTIONARY db_01048.t_01048;
 
 DROP DATABASE IF EXISTS db_01048;
+EXISTS DATABASE db_01048;
 CREATE DATABASE db_01048;
+EXISTS DATABASE db_01048;
 
 DROP TABLE IF EXISTS db_01048.t_01048;
 EXISTS db_01048.t_01048;

From 34508f466f54a56cb4cd78c35809e55ad165fefd Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Thu, 24 Dec 2020 13:14:01 +0300
Subject: [PATCH 386/504] Update ParserTablePropertiesQuery.cpp

---
 src/Parsers/ParserTablePropertiesQuery.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Parsers/ParserTablePropertiesQuery.cpp b/src/Parsers/ParserTablePropertiesQuery.cpp
index 7073106bc13..d6060d0a419 100644
--- a/src/Parsers/ParserTablePropertiesQuery.cpp
+++ b/src/Parsers/ParserTablePropertiesQuery.cpp
@@ -35,9 +35,9 @@ bool ParserTablePropertiesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected &
     if (s_exists.ignore(pos, expected))
     {
         if (s_database.ignore(pos, expected))
-	    {
+        {
             query = std::make_shared<ASTExistsDatabaseQuery>();
-	        parse_only_database_name = true;
+            parse_only_database_name = true;
 	    }
         else
         {

From eb95a5a030f80fd5893b7916b7506521c3d219bf Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Thu, 24 Dec 2020 13:14:17 +0300
Subject: [PATCH 387/504] Update ParserTablePropertiesQuery.cpp

---
 src/Parsers/ParserTablePropertiesQuery.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Parsers/ParserTablePropertiesQuery.cpp b/src/Parsers/ParserTablePropertiesQuery.cpp
index d6060d0a419..dc080b2f13b 100644
--- a/src/Parsers/ParserTablePropertiesQuery.cpp
+++ b/src/Parsers/ParserTablePropertiesQuery.cpp
@@ -38,7 +38,7 @@ bool ParserTablePropertiesQuery::parseImpl(Pos & pos, ASTPtr & node, Expected &
         {
             query = std::make_shared<ASTExistsDatabaseQuery>();
             parse_only_database_name = true;
-	    }
+        }
         else
         {
             if (s_temporary.ignore(pos, expected))

From ce1e2b18a3065ceca54b6b78b62db8d527bb834e Mon Sep 17 00:00:00 2001
From: Yatsishin Ilya <2159081+qoega@users.noreply.github.com>
Date: Thu, 24 Dec 2020 13:11:07 +0300
Subject: [PATCH 388/504] Fix codespell warnings. Split style checks. Update
 style checks docker

---
 base/glibc-compatibility/musl/lgammal.c                |  2 +-
 base/glibc-compatibility/musl/powl.c                   |  2 +-
 base/pcg-random/pcg_random.hpp                         |  2 +-
 docker/test/style/Dockerfile                           |  5 ++++-
 .../AggregateFunctionGroupArrayInsertAt.h              |  2 +-
 src/Common/Config/ConfigProcessor.h                    |  2 +-
 src/Common/HashTable/HashTable.h                       |  2 +-
 src/Core/QueryProcessingStage.h                        |  2 +-
 src/Core/Types.h                                       |  2 +-
 src/DataTypes/DataTypeMap.cpp                          |  4 ++--
 src/DataTypes/IDataType.h                              |  2 +-
 src/Databases/DatabaseOnDisk.cpp                       |  4 ++--
 src/Dictionaries/CacheDictionary.inc.h                 |  2 +-
 src/Functions/FunctionsStringSearch.h                  |  2 +-
 src/Functions/abtesting.cpp                            |  2 +-
 src/Functions/array/arrayElement.cpp                   |  2 +-
 src/IO/WriteBufferFromVector.h                         |  2 +-
 src/Interpreters/InterpreterInsertQuery.cpp            |  2 +-
 src/Interpreters/InterpreterOptimizeQuery.cpp          |  2 +-
 src/Interpreters/TreeRewriter.cpp                      |  2 +-
 src/Parsers/New/AST/JoinExpr.cpp                       |  2 +-
 src/Parsers/New/AST/README.md                          |  2 +-
 src/Parsers/ParserOptimizeQuery.cpp                    |  2 +-
 src/Server/grpc_protos/clickhouse_grpc.proto           | 10 +++++-----
 src/Storages/MergeTree/MergeTreeRangeReader.cpp        |  4 ++--
 src/Storages/MergeTree/RPNBuilder.h                    |  2 +-
 .../tests/gtest_ReplicatedMergeTreeLogEntry.cpp        |  4 ++--
 .../RocksDB/EmbeddedRocksDBBlockInputStream.cpp        |  2 +-
 src/Storages/StorageInMemoryMetadata.h                 |  8 ++++----
 utils/check-style/check-typos                          |  8 ++++----
 utils/check-style/check-whitespaces                    |  9 +++++++++
 utils/check-style/codespell-ignore-words.list          |  9 ++++++++-
 32 files changed, 64 insertions(+), 45 deletions(-)
 create mode 100755 utils/check-style/check-whitespaces

diff --git a/base/glibc-compatibility/musl/lgammal.c b/base/glibc-compatibility/musl/lgammal.c
index 3b5d94c5051..19db11fba1a 100644
--- a/base/glibc-compatibility/musl/lgammal.c
+++ b/base/glibc-compatibility/musl/lgammal.c
@@ -38,7 +38,7 @@
  *                          = log(6.3*5.3) + lgamma(5.3)
  *                          = log(6.3*5.3*4.3*3.3*2.3) + lgamma(2.3)
  *   2. Polynomial approximation of lgamma around its
- *      minimun ymin=1.461632144968362245 to maintain monotonicity.
+ *      minimum ymin=1.461632144968362245 to maintain monotonicity.
  *      On [ymin-0.23, ymin+0.27] (i.e., [1.23164,1.73163]), use
  *              Let z = x-ymin;
  *              lgamma(x) = -1.214862905358496078218 + z^2*poly(z)
diff --git a/base/glibc-compatibility/musl/powl.c b/base/glibc-compatibility/musl/powl.c
index 508c65c9c7c..70cc3fd02ab 100644
--- a/base/glibc-compatibility/musl/powl.c
+++ b/base/glibc-compatibility/musl/powl.c
@@ -202,7 +202,7 @@ long double powl(long double x, long double y)
 	volatile long double z=0;
 	long double w=0, W=0, Wa=0, Wb=0, ya=0, yb=0, u=0;
 
-	/* make sure no invalid exception is raised by nan comparision */
+	/* make sure no invalid exception is raised by nan comparison */
 	if (isnan(x)) {
 		if (!isnan(y) && y == 0.0)
 			return 1.0;
diff --git a/base/pcg-random/pcg_random.hpp b/base/pcg-random/pcg_random.hpp
index f7778480c4c..abf83a60ee1 100644
--- a/base/pcg-random/pcg_random.hpp
+++ b/base/pcg-random/pcg_random.hpp
@@ -129,7 +129,7 @@ using namespace pcg_extras;
  *
  *      default_multiplier<uint32_t>::multiplier()
  *
- * gives you the default multipler for 32-bit integers.  We use the name
+ * gives you the default multiplier for 32-bit integers.  We use the name
  * of the constant and not a generic word like value to allow these classes
  * to be used as mixins.
  */
diff --git a/docker/test/style/Dockerfile b/docker/test/style/Dockerfile
index 1a4356caced..239a074969c 100644
--- a/docker/test/style/Dockerfile
+++ b/docker/test/style/Dockerfile
@@ -4,5 +4,8 @@ FROM ubuntu:20.04
 RUN apt-get update && env DEBIAN_FRONTEND=noninteractive apt-get install --yes shellcheck libxml2-utils git python3-pip && pip3 install codespell
 
 
-CMD cd /ClickHouse/utils/check-style && ./check-style -n | tee /test_output/style_output.txt && \
+CMD cd /ClickHouse/utils/check-style && \
+    ./check-style -n | tee /test_output/style_output.txt && \
+    ./check-typos | tee /test_output/typos_output.txt && \
+    ./check-whitespaces -n | tee /test_output/whitespaces_output.txt && \
     ./check-duplicate-includes.sh | tee /test_output/duplicate_output.txt
diff --git a/src/AggregateFunctions/AggregateFunctionGroupArrayInsertAt.h b/src/AggregateFunctions/AggregateFunctionGroupArrayInsertAt.h
index 6a5f2fbc899..a4800dd715e 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupArrayInsertAt.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupArrayInsertAt.h
@@ -107,7 +107,7 @@ public:
         /// TODO Do positions need to be 1-based for this function?
         size_t position = columns[1]->getUInt(row_num);
 
-        /// If position is larger than size to which array will be cutted - simply ignore value.
+        /// If position is larger than size to which array will be cut - simply ignore value.
         if (length_to_resize && position >= length_to_resize)
             return;
 
diff --git a/src/Common/Config/ConfigProcessor.h b/src/Common/Config/ConfigProcessor.h
index 5a942e73bdb..7a4102140d9 100644
--- a/src/Common/Config/ConfigProcessor.h
+++ b/src/Common/Config/ConfigProcessor.h
@@ -97,7 +97,7 @@ public:
     /// If preprocessed_dir is empty - calculate from loaded_config.path + /preprocessed_configs/
     void savePreprocessedConfig(const LoadedConfig & loaded_config, std::string preprocessed_dir);
 
-    /// Set path of main config.xml. It will be cutted from all configs placed to preprocessed_configs/
+    /// Set path of main config.xml. It will be cut from all configs placed to preprocessed_configs/
     static void setConfigPath(const std::string & config_path);
 
 public:
diff --git a/src/Common/HashTable/HashTable.h b/src/Common/HashTable/HashTable.h
index 2d5580bf709..15fa09490e6 100644
--- a/src/Common/HashTable/HashTable.h
+++ b/src/Common/HashTable/HashTable.h
@@ -1008,7 +1008,7 @@ public:
          * then deleting a erased_key_position will break search for it, so we need to move next_element
          * to erased_key_position. Now we have empty place at next_element, so we apply the identical
          * procedure for it.
-         * If an empty element is encoutered then means that there is no more next elements for which we can
+         * If an empty element is encountered then means that there is no more next elements for which we can
          * break the search so we can exit.
         */
 
diff --git a/src/Core/QueryProcessingStage.h b/src/Core/QueryProcessingStage.h
index 360cf70ebdd..01e7e12ab1e 100644
--- a/src/Core/QueryProcessingStage.h
+++ b/src/Core/QueryProcessingStage.h
@@ -44,7 +44,7 @@ namespace QueryProcessingStage
             : "Unknown stage";
     }
 
-    /// This methid is used for the program options,
+    /// This method is used for the program options,
     /// hence it accept under_score notation for stage:
     /// - complete
     /// - fetch_columns
diff --git a/src/Core/Types.h b/src/Core/Types.h
index 0d578e17517..cc8a4d7269a 100644
--- a/src/Core/Types.h
+++ b/src/Core/Types.h
@@ -185,7 +185,7 @@ using Decimal64 = Decimal<Int64>;
 using Decimal128 = Decimal<Int128>;
 using Decimal256 = Decimal<Int256>;
 
-// Distinguishable type to allow function resultion/deduction based on value type,
+// Distinguishable type to allow function resolution/deduction based on value type,
 // but also relatively easy to convert to/from Decimal64.
 class DateTime64 : public Decimal64
 {
diff --git a/src/DataTypes/DataTypeMap.cpp b/src/DataTypes/DataTypeMap.cpp
index 2047928721a..b207702d9b7 100644
--- a/src/DataTypes/DataTypeMap.cpp
+++ b/src/DataTypes/DataTypeMap.cpp
@@ -204,7 +204,7 @@ void DataTypeMap::serializeText(const IColumn & column, size_t row_num, WriteBuf
 
 void DataTypeMap::deserializeText(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    // need_safe_get_int_key is set for Interger to prevent to readIntTextUnsafe
+    // need_safe_get_int_key is set for Integer to prevent to readIntTextUnsafe
     bool need_safe_get_int_key = isInteger(key_type);
 
     deserializeTextImpl(column, istr, need_safe_get_int_key,
@@ -226,7 +226,7 @@ void DataTypeMap::serializeTextJSON(const IColumn & column, size_t row_num, Writ
 
 void DataTypeMap::deserializeTextJSON(IColumn & column, ReadBuffer & istr, const FormatSettings & settings) const
 {
-    // need_safe_get_int_key is set for Interger to prevent to readIntTextUnsafe
+    // need_safe_get_int_key is set for Integer to prevent to readIntTextUnsafe
     bool need_safe_get_int_key = isInteger(key_type);
 
     deserializeTextImpl(column, istr, need_safe_get_int_key,
diff --git a/src/DataTypes/IDataType.h b/src/DataTypes/IDataType.h
index 060572ee997..b51722ed96d 100644
--- a/src/DataTypes/IDataType.h
+++ b/src/DataTypes/IDataType.h
@@ -67,7 +67,7 @@ public:
       * If the data type require single stream (it's true for most of data types), the stream will have empty path.
       * Otherwise, the path can have components like "array elements", "array sizes", etc.
       *
-      * For multidimensional arrays, path can have arbiraty length.
+      * For multidimensional arrays, path can have arbitrary length.
       * As an example, for 2-dimensional arrays of numbers we have at least three streams:
       * - array sizes;                      (sizes of top level arrays)
       * - array elements / array sizes;     (sizes of second level (nested) arrays)
diff --git a/src/Databases/DatabaseOnDisk.cpp b/src/Databases/DatabaseOnDisk.cpp
index 2a8276ec6db..fb2d10a4fe7 100644
--- a/src/Databases/DatabaseOnDisk.cpp
+++ b/src/Databases/DatabaseOnDisk.cpp
@@ -229,7 +229,7 @@ void DatabaseOnDisk::removeDetachedPermanentlyFlag(const String & table_name, co
     }
     catch (Exception & e)
     {
-        e.addMessage("while trying to remove permanenty detached flag. Table {}.{} may still be marked as permanently detached, and will not be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
+        e.addMessage("while trying to remove permanently detached flag. Table {}.{} may still be marked as permanently detached, and will not be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
         throw;
     }
 }
@@ -264,7 +264,7 @@ void DatabaseOnDisk::detachTablePermanently(const String & table_name)
     }
     catch (Exception & e)
     {
-        e.addMessage("while trying to set permanenty detached flag. Table {}.{} may be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
+        e.addMessage("while trying to set permanently detached flag. Table {}.{} may be reattached during server restart.", backQuote(getDatabaseName()), backQuote(table_name));
         throw;
     }
 }
diff --git a/src/Dictionaries/CacheDictionary.inc.h b/src/Dictionaries/CacheDictionary.inc.h
index d724453f385..803b3b2566f 100644
--- a/src/Dictionaries/CacheDictionary.inc.h
+++ b/src/Dictionaries/CacheDictionary.inc.h
@@ -147,7 +147,7 @@ void CacheDictionary::getItemsNumberImpl(
     tryPushToUpdateQueueOrThrow(update_unit_ptr);
     waitForCurrentUpdateFinish(update_unit_ptr);
 
-    /// Add updated keys to asnwer.
+    /// Add updated keys to answer.
 
     const size_t attribute_index = getAttributeIndex(attribute.name);
 
diff --git a/src/Functions/FunctionsStringSearch.h b/src/Functions/FunctionsStringSearch.h
index 9e7d608aec6..75579b28fd3 100644
--- a/src/Functions/FunctionsStringSearch.h
+++ b/src/Functions/FunctionsStringSearch.h
@@ -29,7 +29,7 @@ namespace DB
   * multiMatchAnyIndex(haystack, [pattern_1, pattern_2, ..., pattern_n]) -- search by re2 regular expressions pattern_i; Returns index of any match or zero if none;
   * multiMatchAllIndices(haystack, [pattern_1, pattern_2, ..., pattern_n]) -- search by re2 regular expressions pattern_i; Returns an array of matched indices in any order;
   *
-  * countSubstrings(haystack, needle) -- count number of occurences of needle in haystack.
+  * countSubstrings(haystack, needle) -- count number of occurrences of needle in haystack.
   * countSubstringsCaseInsensitive(haystack, needle)
   *
   * Applies regexp re2 and pulls:
diff --git a/src/Functions/abtesting.cpp b/src/Functions/abtesting.cpp
index 1af929c8fb9..7bed2f4a2b9 100644
--- a/src/Functions/abtesting.cpp
+++ b/src/Functions/abtesting.cpp
@@ -236,7 +236,7 @@ public:
         if (const ColumnConst * col_higher_is_better = checkAndGetColumnConst<ColumnUInt8>(arguments[1].column.get()))
             higher_is_better = col_higher_is_better->getBool(0);
         else
-            throw Exception("Second argument for function " + getName() + " must be Constatnt boolean", ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+            throw Exception("Second argument for function " + getName() + " must be Constant boolean", ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
         if (const ColumnConst * col_const_arr = checkAndGetColumnConst<ColumnArray>(arguments[2].column.get()))
         {
diff --git a/src/Functions/array/arrayElement.cpp b/src/Functions/array/arrayElement.cpp
index 310b6b5b450..88166f04e0e 100644
--- a/src/Functions/array/arrayElement.cpp
+++ b/src/Functions/array/arrayElement.cpp
@@ -86,7 +86,7 @@ private:
     ColumnPtr executeTuple(const ColumnsWithTypeAndName & arguments, size_t input_rows_count) const;
 
     /** For a map the function finds the matched value for a key.
-     *  Currently implemented just as linear seach in array.
+     *  Currently implemented just as linear search in array.
      *  However, optimizations are possible.
      */
     ColumnPtr executeMap(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const;
diff --git a/src/IO/WriteBufferFromVector.h b/src/IO/WriteBufferFromVector.h
index ad0f8bd17a8..2a9810f3461 100644
--- a/src/IO/WriteBufferFromVector.h
+++ b/src/IO/WriteBufferFromVector.h
@@ -15,7 +15,7 @@ namespace ErrorCodes
 
 /** Writes data to existing std::vector or similar type. When not enough space, it doubles vector size.
   *
-  * In destructor, vector is cutted to the size of written data.
+  * In destructor, vector is cut to the size of written data.
   * You can call 'finalize' to resize earlier.
   *
   * The vector should live until this object is destroyed or until the 'finish' method is called.
diff --git a/src/Interpreters/InterpreterInsertQuery.cpp b/src/Interpreters/InterpreterInsertQuery.cpp
index 9ac936b5489..2a193dc7249 100644
--- a/src/Interpreters/InterpreterInsertQuery.cpp
+++ b/src/Interpreters/InterpreterInsertQuery.cpp
@@ -268,7 +268,7 @@ BlockIO InterpreterInsertQuery::execute()
                 const auto & selects = select_query.list_of_selects->children;
                 const auto & union_modes = select_query.list_of_modes;
 
-                /// ASTSelectWithUnionQuery is not normalized now, so it may pass some querys which can be Trivial select querys
+                /// ASTSelectWithUnionQuery is not normalized now, so it may pass some queries which can be Trivial select queries
                 is_trivial_insert_select
                     = std::all_of(
                           union_modes.begin(),
diff --git a/src/Interpreters/InterpreterOptimizeQuery.cpp b/src/Interpreters/InterpreterOptimizeQuery.cpp
index c382b86afd6..cda8471c32e 100644
--- a/src/Interpreters/InterpreterOptimizeQuery.cpp
+++ b/src/Interpreters/InterpreterOptimizeQuery.cpp
@@ -57,7 +57,7 @@ BlockIO InterpreterOptimizeQuery::execute()
         {
             // Deduplication is performed only for adjacent rows in a block,
             // and all rows in block are in the sorting key order within a single partition,
-            // hence deduplication always implicitly takes sorting keys and parition keys in account.
+            // hence deduplication always implicitly takes sorting keys and partition keys in account.
             // So we just explicitly state that limitation in order to avoid confusion.
             if (std::find(column_names.begin(), column_names.end(), required_col) == column_names.end())
                 throw Exception(ErrorCodes::THERE_IS_NO_COLUMN,
diff --git a/src/Interpreters/TreeRewriter.cpp b/src/Interpreters/TreeRewriter.cpp
index 9816d1a3940..fc3ad79f217 100644
--- a/src/Interpreters/TreeRewriter.cpp
+++ b/src/Interpreters/TreeRewriter.cpp
@@ -132,7 +132,7 @@ struct CustomizeAggregateFunctionsSuffixData
     }
 };
 
-// Used to rewrite aggregate functions with -OrNull suffix in some cases, such as sumIfOrNull, we shoule rewrite to sumOrNullIf
+// Used to rewrite aggregate functions with -OrNull suffix in some cases, such as sumIfOrNull, we should rewrite to sumOrNullIf
 struct CustomizeAggregateFunctionsMoveSuffixData
 {
     using TypeToVisit = ASTFunction;
diff --git a/src/Parsers/New/AST/JoinExpr.cpp b/src/Parsers/New/AST/JoinExpr.cpp
index a5effedc242..acbeae86897 100644
--- a/src/Parsers/New/AST/JoinExpr.cpp
+++ b/src/Parsers/New/AST/JoinExpr.cpp
@@ -57,7 +57,7 @@ JoinExpr::JoinExpr(JoinExpr::ExprType type, JoinExpr::JoinOpType op, JoinExpr::J
 
 ASTPtr JoinExpr::convertToOld() const
 {
-    /** The sole convertable chain of Join's may look like:
+    /** The sole convertible chain of Join's may look like:
      *
      *      … FROM table1 JOIN table2 ON SMTH JOIN table3 ON SMTH JOIN …
      *
diff --git a/src/Parsers/New/AST/README.md b/src/Parsers/New/AST/README.md
index 51e587f296c..4216a8dcfdc 100644
--- a/src/Parsers/New/AST/README.md
+++ b/src/Parsers/New/AST/README.md
@@ -13,7 +13,7 @@ Basic principles in code
 **predetermined order** and with **predetermined type**: some elements may be `nullptr` to preserve positions of other elements.
 - The order may be defined as a position in vector from the start, the last element, and some pattern of variable number of elements
 in between. It's convenient to define `enum ChildIndex : Uint8 {…}` with index numbers for each class.
-- If there is more than one variable pack of elements or the order can't be determenistic, then wrap elements into the lists and store the
+- If there is more than one variable pack of elements or the order can't be deterministic, then wrap elements into the lists and store the
 multi-level structure (see `ColumnExpr::ExprType::FUNCTION` for example).
 - Don't do multi-level structure just for nothing or to mimic the parse tree: the less is depth the better.
 - The whole grammar separates expressions for databases, tables and columns. That way we already assess the semantics on the parser level.
diff --git a/src/Parsers/ParserOptimizeQuery.cpp b/src/Parsers/ParserOptimizeQuery.cpp
index 629b88adb88..441cec1465e 100644
--- a/src/Parsers/ParserOptimizeQuery.cpp
+++ b/src/Parsers/ParserOptimizeQuery.cpp
@@ -14,7 +14,7 @@ bool ParserOptimizeQueryColumnsSpecification::parseImpl(Pos & pos, ASTPtr & node
 {
     // Do not allow APPLY and REPLACE transformers.
     // Since we use Columns Transformers only to get list of columns,
-    // ad we can't actuall modify content of the columns for deduplication.
+    // we can't actually modify content of the columns for deduplication.
     const auto allowed_transformers = ParserColumnsTransformers::ColumnTransformers{ParserColumnsTransformers::ColumnTransformer::EXCEPT};
 
     return ParserColumnsMatcher(allowed_transformers).parse(pos, node, expected)
diff --git a/src/Server/grpc_protos/clickhouse_grpc.proto b/src/Server/grpc_protos/clickhouse_grpc.proto
index c80a8bf1d40..5d16f9e8a79 100644
--- a/src/Server/grpc_protos/clickhouse_grpc.proto
+++ b/src/Server/grpc_protos/clickhouse_grpc.proto
@@ -21,11 +21,11 @@ message NameAndType {
    string type = 2;
 }
 
-// Desribes an external table - a table which will exists only while a query is executing.
+// Describes an external table - a table which will exists only while a query is executing.
 message ExternalTable {
    // Name of the table. If omitted, "_data" is used.
    string name = 1;
-   
+
    // Columns of the table. Types are required, names can be omitted. If the names are omitted, "_1", "_2", ... is used.
    repeated NameAndType columns = 2;
 
@@ -33,7 +33,7 @@ message ExternalTable {
    // If a method with streaming input (i.e. ExecuteQueryWithStreamInput() or ExecuteQueryWithStreamIO()) is used,
    // then data for insertion to the same external table can be splitted between multiple QueryInfos.
    string data = 3;
-   
+
    // Format of the data to insert to the external table.
    string format = 4;
 
@@ -57,7 +57,7 @@ message QueryInfo {
 
    // Delimiter for input_data, inserted between input_data from adjacent QueryInfos.
    string input_data_delimiter = 6;
-   
+
    // Default output format. If not specified, 'TabSeparated' is used.
    string output_format = 7;
 
@@ -71,7 +71,7 @@ message QueryInfo {
    string session_id = 12;
    bool session_check = 13;
    uint32 session_timeout = 14;
-   
+
    // Set `cancel` to true to stop executing the query.
    bool cancel = 15;
 
diff --git a/src/Storages/MergeTree/MergeTreeRangeReader.cpp b/src/Storages/MergeTree/MergeTreeRangeReader.cpp
index e22c6bb3d0c..c13146bd35c 100644
--- a/src/Storages/MergeTree/MergeTreeRangeReader.cpp
+++ b/src/Storages/MergeTree/MergeTreeRangeReader.cpp
@@ -637,7 +637,7 @@ MergeTreeRangeReader::ReadResult MergeTreeRangeReader::read(size_t max_rows, Mar
 
         if (!columns.empty())
         {
-            /// If some columns absent in part, then evaulate default values
+            /// If some columns absent in part, then evaluate default values
             if (should_evaluate_missing_defaults)
             {
                 auto block = prev_reader->sample_block.cloneWithColumns(read_result.columns);
@@ -681,7 +681,7 @@ MergeTreeRangeReader::ReadResult MergeTreeRangeReader::read(size_t max_rows, Mar
             merge_tree_reader->fillMissingColumns(read_result.columns, should_evaluate_missing_defaults,
                                                   read_result.num_rows);
 
-            /// If some columns absent in part, then evaulate default values
+            /// If some columns absent in part, then evaluate default values
             if (should_evaluate_missing_defaults)
                 merge_tree_reader->evaluateMissingDefaults({}, read_result.columns);
 
diff --git a/src/Storages/MergeTree/RPNBuilder.h b/src/Storages/MergeTree/RPNBuilder.h
index ed25252642b..292a120d28a 100644
--- a/src/Storages/MergeTree/RPNBuilder.h
+++ b/src/Storages/MergeTree/RPNBuilder.h
@@ -31,7 +31,7 @@ public:
           */
         block_with_constants = KeyCondition::getBlockWithConstants(query_info.query, query_info.syntax_analyzer_result, context);
 
-        /// Trasform WHERE section to Reverse Polish notation
+        /// Transform WHERE section to Reverse Polish notation
         const ASTSelectQuery & select = typeid_cast<const ASTSelectQuery &>(*query_info.query);
         if (select.where())
         {
diff --git a/src/Storages/MergeTree/tests/gtest_ReplicatedMergeTreeLogEntry.cpp b/src/Storages/MergeTree/tests/gtest_ReplicatedMergeTreeLogEntry.cpp
index 98e70bd6147..03f36d866bb 100644
--- a/src/Storages/MergeTree/tests/gtest_ReplicatedMergeTreeLogEntry.cpp
+++ b/src/Storages/MergeTree/tests/gtest_ReplicatedMergeTreeLogEntry.cpp
@@ -213,7 +213,7 @@ TEST_P(ReplicatedMergeTreeLogEntryDataTest, transcode)
 
 // Enabling this warning would ruin test brievity without adding anything else in return,
 // since most of the fields have default constructors or be will be zero-initialized as by standard,
-// so values are predicatable and stable accross runs.
+// so values are predicatable and stable across runs.
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wmissing-field-initializers"
 
@@ -301,7 +301,7 @@ INSTANTIATE_TEST_SUITE_P(Merge, ReplicatedMergeTreeLogEntryDataTest,
 
 // This is just an example of how to set all fields. Can't be used as is since depending on type,
 // only some fields are serialized/deserialized, and even if everything works perfectly,
-// some fileds in deserialized object would be unset (hence differ from expected).
+// some fields in deserialized object would be unset (hence differ from expected).
 // INSTANTIATE_TEST_SUITE_P(Full, ReplicatedMergeTreeLogEntryDataTest,
 //         ::testing::ValuesIn(std::initializer_list<ReplicatedMergeTreeLogEntryData>{
 //     {
diff --git a/src/Storages/RocksDB/EmbeddedRocksDBBlockInputStream.cpp b/src/Storages/RocksDB/EmbeddedRocksDBBlockInputStream.cpp
index 49b91aa7477..35c41cabd8b 100644
--- a/src/Storages/RocksDB/EmbeddedRocksDBBlockInputStream.cpp
+++ b/src/Storages/RocksDB/EmbeddedRocksDBBlockInputStream.cpp
@@ -56,7 +56,7 @@ Block EmbeddedRocksDBBlockInputStream::readImpl()
     finished = !iterator->Valid();
     if (!iterator->status().ok())
     {
-        throw Exception("Engine " + getName() + " got error while seeking key value datas: " + iterator->status().ToString(),
+        throw Exception("Engine " + getName() + " got error while seeking key value data: " + iterator->status().ToString(),
             ErrorCodes::ROCKSDB_ERROR);
     }
     return sample_block.cloneWithColumns(std::move(columns));
diff --git a/src/Storages/StorageInMemoryMetadata.h b/src/Storages/StorageInMemoryMetadata.h
index 3656edf71f4..4c5edf31efe 100644
--- a/src/Storages/StorageInMemoryMetadata.h
+++ b/src/Storages/StorageInMemoryMetadata.h
@@ -61,13 +61,13 @@ struct StorageInMemoryMetadata
     /// Sets constraints
     void setConstraints(ConstraintsDescription constraints_);
 
-    /// Set partition key for storage (methods bellow, are just wrappers for this struct).
+    /// Set partition key for storage (methods below, are just wrappers for this struct).
     void setPartitionKey(const KeyDescription & partition_key_);
-    /// Set sorting key for storage (methods bellow, are just wrappers for this struct).
+    /// Set sorting key for storage (methods below, are just wrappers for this struct).
     void setSortingKey(const KeyDescription & sorting_key_);
-    /// Set primary key for storage (methods bellow, are just wrappers for this struct).
+    /// Set primary key for storage (methods below, are just wrappers for this struct).
     void setPrimaryKey(const KeyDescription & primary_key_);
-    /// Set sampling key for storage (methods bellow, are just wrappers for this struct).
+    /// Set sampling key for storage (methods below, are just wrappers for this struct).
     void setSamplingKey(const KeyDescription & sampling_key_);
 
     /// Set common table TTLs
diff --git a/utils/check-style/check-typos b/utils/check-style/check-typos
index 14deb039059..72c19bc1140 100755
--- a/utils/check-style/check-typos
+++ b/utils/check-style/check-typos
@@ -1,14 +1,14 @@
 #!/usr/bin/env bash
+set +x
 
 # Check for typos in code.
 
 ROOT_PATH=$(git rev-parse --show-toplevel)
-CURDIR=$(dirname "${BASH_SOURCE[0]}")
 
 codespell \
-    --skip '*generated*,*gperf*,*.bin,*.mrk*,*.idx,checksums.txt,*.dat,*.pyc,*.kate-swp' \
-    --ignore-words "${CURDIR}/codespell-ignore-words.list" \
-    --exclude-file "${CURDIR}/codespell-ignore-lines.list" \
+    --skip '*generated*,*gperf*,*.bin,*.mrk*,*.idx,checksums.txt,*.dat,*.pyc,*.kate-swp,*obfuscateQueries.cpp' \
+    --ignore-words "${ROOT_PATH}/utils/check-style/codespell-ignore-words.list" \
+    --exclude-file "${ROOT_PATH}/utils/check-style/codespell-ignore-lines.list" \
     --quiet-level 2 \
     "$ROOT_PATH"/{src,base,programs,utils} \
     $@ | grep -P '.' \
diff --git a/utils/check-style/check-whitespaces b/utils/check-style/check-whitespaces
new file mode 100755
index 00000000000..1aa3fd8cf6d
--- /dev/null
+++ b/utils/check-style/check-whitespaces
@@ -0,0 +1,9 @@
+#!/usr/bin/env bash
+
+ROOT_PATH=$(git rev-parse --show-toplevel)
+EXCLUDE_DIRS='build/|integration/|widechar_width/|glibc-compatibility/|memcpy/|consistent-hashing/|Parsers/New'
+
+# Double whitespaces
+find $ROOT_PATH/{src,base,programs,utils} -name '*.h' -or -name '*.cpp' 2>/dev/null |
+    grep -vP $EXCLUDE_DIRS |
+    while read i; do $ROOT_PATH/utils/check-style/double-whitespaces.pl < $i || echo -e "^ File $i contains double whitespaces\n"; done
\ No newline at end of file
diff --git a/utils/check-style/codespell-ignore-words.list b/utils/check-style/codespell-ignore-words.list
index 4433e9917e3..680e7344ad1 100644
--- a/utils/check-style/codespell-ignore-words.list
+++ b/utils/check-style/codespell-ignore-words.list
@@ -1,7 +1,14 @@
-thenn
+wthenn
 fpr
 creat
 parsering
 nd
 ect
 pullrequest
+pullrequests
+ans
+splitted
+thenn
+ths
+offsett
+numer

From f36e8f13c03a1b447771f1937acb4d1743dd294a Mon Sep 17 00:00:00 2001
From: Yatsishin Ilya <2159081+qoega@users.noreply.github.com>
Date: Thu, 24 Dec 2020 13:48:10 +0300
Subject: [PATCH 389/504] more fixes after rebase

---
 src/Common/Exception.h                        |  2 +-
 src/Functions/ExtractString.h                 |  6 +++---
 src/Functions/FunctionsStringHash.cpp         | 12 ++++++------
 src/Server/grpc_protos/clickhouse_grpc.proto  |  2 +-
 utils/check-style/check-typos                 |  1 -
 utils/check-style/codespell-ignore-words.list |  4 +---
 6 files changed, 12 insertions(+), 15 deletions(-)

diff --git a/src/Common/Exception.h b/src/Common/Exception.h
index c26f3f41ab4..a6190c7ca24 100644
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@@ -98,7 +98,7 @@ private:
 
 
 /// Special class of exceptions, used mostly in ParallelParsingInputFormat for
-/// more convinient calculation of problem line number.
+/// more convenient calculation of problem line number.
 class ParsingException : public Exception
 {
 public:
diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index c4251f8c4a6..8313f583025 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -61,7 +61,7 @@ struct ExtractStringImpl
     // read a ASCII word
     static ALWAYS_INLINE inline size_t readOneASCIIWord(PaddedPODArray<UInt8> & word_buf, const char *& pos, const char * end)
     {
-        // jump seperators
+        // jump separators
         while (pos < end && !isAlphaNumericASCII(*pos))
             ++pos;
 
@@ -93,7 +93,7 @@ struct ExtractStringImpl
     // read one UTF8 word from pos to word
     static ALWAYS_INLINE inline size_t readOneUTF8Word(PaddedPODArray<UInt32> & word_buf, const char *& pos, const char * end)
     {
-        // jump UTF8 seperator
+        // jump UTF8 separator
         while (pos < end && isUTF8Sep(*pos))
             ++pos;
         word_buf.clear();
@@ -112,7 +112,7 @@ private:
         ((cont[Offset + I] = std::tolower(cont[Offset + I])), ...);
     }
 
-    // we use ASCII non-alphanum character as UTF8 seperator
+    // we use ASCII non-alphanum character as UTF8 separator
     static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNumericASCII(c); }
 
     // read one UTF8 character and return it
diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index d57be67ef7f..7062a47fa22 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -151,7 +151,7 @@ template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensit
 struct SimhashImpl
 {
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    // we made an assumption that the size of one word cann't exceed 128, which may not true
+    // we made an assumption that the size of one word can't exceed 128, which may not true
     // if some word's size exceed 128, it would be cut up to several word
     static constexpr size_t max_string_size = 1u << 15;
     static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
@@ -203,7 +203,7 @@ struct SimhashImpl
         return res_bit.to_ullong();
     }
 
-    // Simhash word shingle calculate funtion: String -> UInt64
+    // Simhash word shingle calculate function: String -> UInt64
     // this function extracting n word shingle from input string, and maintain a 64-dimensions vector as well
     // for each word shingle, calculate a 64 bit hash value, and update the vector according the hash value
     // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
@@ -211,8 +211,8 @@ struct SimhashImpl
     // word shingle hash value calculate:
     // 1. at the first, extracts N word shingles and calculate N hash values, store into an array, use this N hash values
     // to calculate the first word shingle hash value
-    // 2. next, we extrac one word each time, and calculate a new hash value of the new word,then use the latest N hash
-    // values to caculate the next word shingle hash value
+    // 2. next, we extract one word each time, and calculate a new hash value of the new word,then use the latest N hash
+    // values to calculate the next word shingle hash value
     static ALWAYS_INLINE inline UInt64 wordShinglesCalculateHashValue(
         const char * data,
         size_t size,
@@ -254,12 +254,12 @@ struct SimhashImpl
             // for example, N = 5, array |a0|a1|a2|a3|a4|, now , a0 is the oldest location,
             // so we need to store new word hash into location of a0, then ,this array become
             // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
-            // word hash value into locaion of a1, then array become |a5|a6|a2|a3|a4|
+            // word hash value into location of a1, then array become |a5|a6|a2|a3|a4|
             nword_hashes[offset] = Hash::hashSum(word_buf.data(), word_buf.size());
             offset = (offset + 1) % N;
             // according to the word hash storation way, in order to not lose the word shingle's
             // sequence information, when calculation word shingle hash value, we need provide the offset
-            // inforation, which is the offset of the first word's hash value of the word shingle
+            // information, which is the offset of the first word's hash value of the word shingle
             hash_value = hash_functor(nword_hashes, N, offset);
             std::bitset<64> bits(hash_value);
             for (size_t i = 0; i < 64; ++i)
diff --git a/src/Server/grpc_protos/clickhouse_grpc.proto b/src/Server/grpc_protos/clickhouse_grpc.proto
index 5d16f9e8a79..2f25973297c 100644
--- a/src/Server/grpc_protos/clickhouse_grpc.proto
+++ b/src/Server/grpc_protos/clickhouse_grpc.proto
@@ -31,7 +31,7 @@ message ExternalTable {
 
    // Data to insert to the external table.
    // If a method with streaming input (i.e. ExecuteQueryWithStreamInput() or ExecuteQueryWithStreamIO()) is used,
-   // then data for insertion to the same external table can be splitted between multiple QueryInfos.
+   // then data for insertion to the same external table can be split between multiple QueryInfos.
    string data = 3;
 
    // Format of the data to insert to the external table.
diff --git a/utils/check-style/check-typos b/utils/check-style/check-typos
index 72c19bc1140..b5858968841 100755
--- a/utils/check-style/check-typos
+++ b/utils/check-style/check-typos
@@ -1,5 +1,4 @@
 #!/usr/bin/env bash
-set +x
 
 # Check for typos in code.
 
diff --git a/utils/check-style/codespell-ignore-words.list b/utils/check-style/codespell-ignore-words.list
index 680e7344ad1..200b55d112d 100644
--- a/utils/check-style/codespell-ignore-words.list
+++ b/utils/check-style/codespell-ignore-words.list
@@ -1,4 +1,4 @@
-wthenn
+thenn
 fpr
 creat
 parsering
@@ -6,8 +6,6 @@ nd
 ect
 pullrequest
 pullrequests
-ans
-splitted
 thenn
 ths
 offsett

From 8c72fa063df28b25c04acdf24405f7347a04fd6c Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Thu, 24 Dec 2020 13:59:58 +0300
Subject: [PATCH 390/504] boop the CI

---
 src/Interpreters/ExpressionAnalyzer.cpp | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 0a380c66d23..4af2a17e164 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -1426,6 +1426,12 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
 
         has_order_by = query.orderBy() != nullptr;
+        // FIXME selected columns was set here. Should split order by and select
+        // and insert window functions in between.
+        // Will fix:
+        // 1) window function in DISTINCT
+        // 2) totally broken DISTINCT
+        // 3) window functions in order by
         before_order_and_select = query_analyzer.appendOrderBy(
                 chain,
                 only_types || (need_aggregate ? !second_stage : !first_stage),

From 3e9d177813b017e02e0e96bc8ead0a0469a3aa63 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 16:06:32 +0300
Subject: [PATCH 391/504] Slightly simplier cmake

---
 CMakeLists.txt             | 6 +++---
 base/common/CMakeLists.txt | 2 --
 contrib/CMakeLists.txt     | 5 ++++-
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 4ff492bda8c..52cb64fd763 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -252,15 +252,15 @@ endif ()
 option(WITH_COVERAGE "Profile the resulting binary/binaries" OFF)
 
 if (WITH_COVERAGE AND COMPILER_CLANG)
-    set(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
+    set(COMPILER_FLAGS "${COMPILER_FLAGS} -fprofile-instr-generate -fcoverage-mapping")
     # If we want to disable coverage for specific translation units
     set(WITHOUT_COVERAGE "-fno-profile-instr-generate -fno-coverage-mapping")
 endif()
 
 if (WITH_COVERAGE AND COMPILER_GCC)
+    set(COMPILER_FLAGS "${COMPILER_FLAGS} -fprofile-arcs -ftest-coverage")
     set(COVERAGE_OPTION "-lgcov")
-    set(COMPILER_COVERAGE_OPTIONS -fprofile-instr-generate -fcoverage-mapping)
-    set(WITH_COVERAGE "-fprofile-arcs -ftest-coverage")
+    set(WITHOUT_COVERAGE "-fno-profile-arcs -fno-test-coverage")
 endif()
 
 set(COMPILER_FLAGS "${COMPILER_FLAGS}")
diff --git a/base/common/CMakeLists.txt b/base/common/CMakeLists.txt
index 9a37f38ab14..cea52b443dd 100644
--- a/base/common/CMakeLists.txt
+++ b/base/common/CMakeLists.txt
@@ -37,8 +37,6 @@ add_library (common ${SRCS})
 
 if (WITH_COVERAGE)
     target_compile_definitions(common PUBLIC WITH_COVERAGE=1)
-    target_compile_options(common PUBLIC ${COMPILER_COVERAGE_OPTIONS})
-    target_link_options(common PUBLIC ${COMPILER_COVERAGE_OPTIONS})
 else ()
     target_compile_definitions(common PUBLIC WITH_COVERAGE=0)
 endif ()
diff --git a/contrib/CMakeLists.txt b/contrib/CMakeLists.txt
index a8cd1f07c1a..338d2e9f20c 100644
--- a/contrib/CMakeLists.txt
+++ b/contrib/CMakeLists.txt
@@ -15,7 +15,10 @@ set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -w")
 set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -w")
 
 if (WITH_COVERAGE)
-  add_compile_options(-O3 -DNDEBUG -finline)
+  set (WITHOUT_COVERAGE_LIST ${WITHOUT_COVERAGE})
+  separate_arguments(WITHOUT_COVERAGE_LIST)
+  # disable coverage for contib files and build with optimisations
+  add_compile_options(-O3 -DNDEBUG -finline-functions -finline-hint-functions ${WITHOUT_COVERAGE_LIST})
 endif()
 
 if (SANITIZE STREQUAL "undefined")

From 4c4f72786e1d174a1bdccfd014fb00577d854fce Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 24 Dec 2020 16:45:43 +0300
Subject: [PATCH 392/504] replaced name shadowing in gcc

---
 src/Functions/FunctionBinaryArithmetic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index d3639556ab6..a8ab5d1735e 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1361,8 +1361,8 @@ public:
 
     bool hasInformationAboutMonotonicity() const override
     {
-        const std::string_view name = Name::name;
-        return (name == "minus" || name == "plus" || name == "divide" || name == "intDiv");
+        const std::string_view name_view = Name::name;
+        return (name_view == "minus" || name_view == "plus" || name_view == "divide" || name_view == "intDiv");
     }
 
     Monotonicity getMonotonicityForRange(const IDataType &, const Field & left_point, const Field & right_point) const override

From fe80f98186763e01f8e8b122991e5e4665cf8c0c Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 24 Dec 2020 16:46:55 +0300
Subject: [PATCH 393/504] last fix add

---
 src/Functions/FunctionBinaryArithmetic.h | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index a8ab5d1735e..0e25f38e96d 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -1375,9 +1375,9 @@ public:
         if (applyVisitor(FieldVisitorAccurateEquals(), left_point, right_point))
             return {true, true, false};
 
-        const std::string_view name = Name::name;
+        const std::string_view name_view = Name::name;
 
-        if (name == "minus" || name == "plus")
+        if (name_view == "minus" || name_view == "plus")
         {
             // const +|- variable
             if (left.column && isColumnConst(*left.column))
@@ -1395,7 +1395,8 @@ public:
                 };
                 transform(left_point);
                 transform(right_point);
-                if (name == "plus")
+
+                if (name_view == "plus")
                 {
                     // Check if there is an overflow
                     if (applyVisitor(FieldVisitorAccurateLess(), left_point, right_point)
@@ -1437,7 +1438,7 @@ public:
                     return {false, true, false};
             }
         }
-        if (name == "divide" || name == "intDiv")
+        if (name_view == "divide" || name_view == "intDiv")
         {
             // const / variable
             if (left.column && isColumnConst(*left.column))

From 63dc8c546201a0182fed6de96b4e689c3c05df94 Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@jakalletti-build.sas.yp-c.yandex.net>
Date: Thu, 24 Dec 2020 17:24:51 +0300
Subject: [PATCH 394/504] better

---
 .../test_cluster_copier/test_trivial.py       | 35 ++++++++++++-------
 1 file changed, 23 insertions(+), 12 deletions(-)

diff --git a/tests/integration/test_cluster_copier/test_trivial.py b/tests/integration/test_cluster_copier/test_trivial.py
index ba6cefcec38..71b76915cfa 100644
--- a/tests/integration/test_cluster_copier/test_trivial.py
+++ b/tests/integration/test_cluster_copier/test_trivial.py
@@ -5,6 +5,7 @@ from contextlib import contextmanager
 
 import docker
 import pytest
+import random
 
 CURRENT_TEST_DIR = os.path.dirname(os.path.abspath(__file__))
 sys.path.insert(0, os.path.dirname(CURRENT_TEST_DIR))
@@ -85,9 +86,14 @@ def execute_task(task, cmd_options):
     zk = cluster.get_kazoo_client('zoo1')
     print("Use ZooKeeper server: {}:{}".format(zk.hosts[0][0], zk.hosts[0][1]))
 
+    try:
+        zk.delete("/clickhouse-copier", recursive=True)
+    except kazoo.exceptions.NoNodeError:
+        print("No node /clickhouse-copier. It is Ok in first test.")
+
     zk_task_path = task.zk_task_path
     zk.ensure_path(zk_task_path)
-    zk.create(zk_task_path + "/description", task.copier_task_config)
+    zk.create(zk_task_path + "/description", task.copier_task_config.encode())
 
     # Run cluster-copier processes on each node
     docker_api = docker.from_env().api
@@ -99,23 +105,28 @@ def execute_task(task, cmd_options):
            '--base-dir', '/var/log/clickhouse-server/copier']
     cmd += cmd_options
 
-    print(cmd)
+    copiers = list(cluster.instances.keys())
 
-    for instance_name, instance in cluster.instances.items():
+    for instance_name in copiers:
+        instance = cluster.instances[instance_name]
         container = instance.get_docker_handle()
+        instance.copy_file_to_container(os.path.join(CURRENT_TEST_DIR, "configs/config-copier.xml"),
+                                        "/etc/clickhouse-server/config-copier.xml")
+        print("Copied copier config to {}".format(instance.name))
         exec_id = docker_api.exec_create(container.id, cmd, stderr=True)
-        docker_api.exec_start(exec_id, detach=True)
-
+        output = docker_api.exec_start(exec_id).decode('utf8')
+        print(output)
         copiers_exec_ids.append(exec_id)
         print("Copier for {} ({}) has started".format(instance.name, instance.ip_address))
 
     # Wait for copiers stopping and check their return codes
-    for exec_id, instance in zip(copiers_exec_ids, iter(cluster.instances.values())):
+    for exec_id, instance_name in zip(copiers_exec_ids, copiers):
+        instance = cluster.instances[instance_name]
         while True:
             res = docker_api.exec_inspect(exec_id)
             if not res['Running']:
                 break
-            time.sleep(1)
+            time.sleep(0.5)
 
         assert res['ExitCode'] == 0, "Instance: {} ({}). Info: {}".format(instance.name, instance.ip_address, repr(res))
 
@@ -152,10 +163,10 @@ def test_trivial_copy(started_cluster, use_sample_offset):
 )
 def test_trivial_copy_with_copy_fault(started_cluster, use_sample_offset):
     if use_sample_offset:
-        execute_task(TaskTrivial(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY),
                                                     '--experimental-use-sample-offset', '1'])
     else:
-        execute_task(TaskTrivial(started_cluster), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--copy-fault-probability', str(COPYING_FAIL_PROBABILITY)])
 
 
 @pytest.mark.parametrize(
@@ -167,14 +178,14 @@ def test_trivial_copy_with_copy_fault(started_cluster, use_sample_offset):
 )
 def test_trivial_copy_with_move_fault(started_cluster, use_sample_offset):
     if use_sample_offset:
-        execute_task(TaskTrivial(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY),
                                                     '--experimental-use-sample-offset', '1'])
     else:
-        execute_task(TaskTrivial(started_cluster), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY)])
+        execute_task(TaskTrivial(started_cluster, use_sample_offset), ['--move-fault-probability', str(MOVING_FAIL_PROBABILITY)])
 
 
 if __name__ == '__main__':
     with contextmanager(started_cluster)() as cluster:
         for name, instance in list(cluster.instances.items()):
             print(name, instance.ip_address)
-        input("Cluster created, press any key to destroy...")
\ No newline at end of file
+        input("Cluster created, press any key to destroy...")

From ccb2fcc16124654350b2f8e765d9a263723b3152 Mon Sep 17 00:00:00 2001
From: Andy Yang <yangzhaohui168@gmail.com>
Date: Mon, 7 Dec 2020 17:35:15 +0800
Subject: [PATCH 395/504] Support bitmap64 feature

---
 contrib/croaring-cmake/CMakeLists.txt         |   1 +
 .../AggregateFunctionGroupBitmapData.h        | 600 +++++++++++++++---
 .../00829_bitmap64_function.reference         |  14 +
 .../0_stateless/00829_bitmap64_function.sql   |  73 +++
 4 files changed, 592 insertions(+), 96 deletions(-)
 create mode 100644 tests/queries/0_stateless/00829_bitmap64_function.reference
 create mode 100644 tests/queries/0_stateless/00829_bitmap64_function.sql

diff --git a/contrib/croaring-cmake/CMakeLists.txt b/contrib/croaring-cmake/CMakeLists.txt
index 3189795347b..8a8ca62e051 100644
--- a/contrib/croaring-cmake/CMakeLists.txt
+++ b/contrib/croaring-cmake/CMakeLists.txt
@@ -23,3 +23,4 @@ add_library(roaring ${SRCS})
 
 target_include_directories(roaring PRIVATE ${LIBRARY_DIR}/include/roaring)
 target_include_directories(roaring SYSTEM BEFORE PUBLIC ${LIBRARY_DIR}/include)
+target_include_directories(roaring SYSTEM BEFORE PUBLIC ${LIBRARY_DIR}/cpp)
diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
index d80e5e81f19..5e93fbb60d8 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
@@ -12,6 +12,7 @@
 // TODO: find out what it is. On github, they have proper interface headers like
 // this one: https://github.com/RoaringBitmap/CRoaring/blob/master/include/roaring/roaring.h
 #include <roaring/roaring.h>
+#include <roaring64map.hh>
 
 namespace DB
 {
@@ -26,25 +27,37 @@ class RoaringBitmapWithSmallSet : private boost::noncopyable
 private:
     using Small = SmallSet<T, small_set_size>;
     using ValueBuffer = std::vector<T>;
+    bool bUInt64 = sizeof(T) >= 8;
+
     Small small;
     roaring_bitmap_t * rb = nullptr;
+    std::shared_ptr<Roaring64Map> rb64 = nullptr;
 
     void toLarge()
     {
-        rb = roaring_bitmap_create();
+        if ( isUInt64() )
+            rb64 = std::make_shared<Roaring64Map>();
+        else
+            rb = roaring_bitmap_create();
 
-        for (const auto & x : small)
-            roaring_bitmap_add(rb, x.getValue());
+        for (const auto & x : small){
+            if ( isUInt64() )
+                rb64->add( static_cast<UInt64>(x.getValue()) );
+            else
+                roaring_bitmap_add(rb, x.getValue());
+        }
     }
 
 public:
-    bool isLarge() const { return rb != nullptr; }
+    bool isLarge() const { return rb != nullptr || rb64 != nullptr; }
 
-    bool isSmall() const { return rb == nullptr; }
+    bool isSmall() const { return rb == nullptr &&  rb64 == nullptr; }
+
+    bool isUInt64() const { return bUInt64; }
 
     ~RoaringBitmapWithSmallSet()
     {
-        if (isLarge())
+        if (isLarge() && !isUInt64() )
             roaring_bitmap_free(rb);
     }
 
@@ -59,19 +72,35 @@ public:
                 else
                 {
                     toLarge();
-                    roaring_bitmap_add(rb, value);
+                    if ( isUInt64() )
+                        rb64->add( static_cast<UInt64>(value) );
+                    else
+                        roaring_bitmap_add(rb, value);
                 }
             }
         }
         else
-            roaring_bitmap_add(rb, value);
+        {
+            if ( isUInt64() )
+                rb64->add( static_cast<UInt64>(value) );
+            else
+                roaring_bitmap_add(rb, value);
+        }
     }
 
     UInt64 size() const
     {
-        return isSmall()
-            ? small.size()
-            : roaring_bitmap_get_cardinality(rb);
+        if (isSmall())
+        {
+            return small.size();
+        }
+        else
+        {
+            if ( isUInt64() )
+                return rb64->cardinality();
+            else
+                return roaring_bitmap_get_cardinality(rb);
+        }
     }
 
     void merge(const RoaringBitmapWithSmallSet & r1)
@@ -81,7 +110,10 @@ public:
             if (isSmall())
                 toLarge();
 
-            roaring_bitmap_or_inplace(rb, r1.rb);
+            if ( isUInt64() )
+                *rb64 |= *r1.rb64;
+            else
+                roaring_bitmap_or_inplace(rb, r1.rb);
         }
         else
         {
@@ -92,39 +124,63 @@ public:
 
     void read(DB::ReadBuffer & in)
     {
-        bool is_large;
-        readBinary(is_large, in);
+        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
+        UInt8 containerType = 0;
+        readBinary(containerType, in);
 
-        if (is_large)
+        if( 0 == containerType )
         {
+            small.read(in);
+        } else if( 1 == containerType ) {
             std::string s;
             readStringBinary(s,in);
             rb = roaring_bitmap_portable_deserialize(s.c_str());
-            for (const auto & x : small) // merge from small
-                roaring_bitmap_add(rb, x.getValue());
+            // It has been persisted in the bitmap and does not need to merge from small
+//            for (const auto & x : small) // merge from small
+//                roaring_bitmap_add(rb, x.getValue());
+        } else {
+            std::string s;
+            readStringBinary(s,in);
+            rb64 = std::make_shared<Roaring64Map>( Roaring64Map::read(s.c_str()) );
         }
-        else
-            small.read(in);
     }
 
     void write(DB::WriteBuffer & out) const
     {
-        writeBinary(isLarge(), out);
+        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
+        UInt8 containerType = 0;
+        if (isLarge())
+        {
+            if ( isUInt64() )
+                containerType = 2;
+            else
+                containerType = 1;
+        }
+
+        writeBinary(containerType, out);
 
         if (isLarge())
         {
-            uint32_t expectedsize = roaring_bitmap_portable_size_in_bytes(rb);
-            std::string s(expectedsize,0);
-            roaring_bitmap_portable_serialize(rb, const_cast<char*>(s.data()));
-            writeStringBinary(s,out);
+            if ( isUInt64() ){
+                uint32_t expectedsize = rb64->getSizeInBytes();
+                std::string s(expectedsize,0);
+                rb64->write(const_cast<char*>(s.data()));
+                writeStringBinary(s,out);
+            } else {
+                uint32_t expectedsize = roaring_bitmap_portable_size_in_bytes(rb);
+                std::string s(expectedsize,0);
+                roaring_bitmap_portable_serialize(rb, const_cast<char*>(s.data()));
+                writeStringBinary(s,out);
+            }
         }
         else
             small.write(out);
     }
 
-
     roaring_bitmap_t * getRb() const { return rb; }
 
+    std::shared_ptr<Roaring64Map> getRb64() const { return rb64; }
+
     Small & getSmall() const { return small; }
 
     /**
@@ -138,6 +194,17 @@ public:
         return smallRb;
     }
 
+    /**
+     * Get a new Roaring64Map from elements of small
+     */
+    std::shared_ptr<Roaring64Map> getNewRb64FromSmall() const
+    {
+        std::shared_ptr<Roaring64Map> smallRb64 = std::make_shared<Roaring64Map>();
+        for (const auto & x : small)
+            smallRb64->add( static_cast<UInt64>(x.getValue()) );
+        return smallRb64;
+    }
+
     /**
      * Computes the intersection between two bitmaps
      */
@@ -162,8 +229,18 @@ public:
         else if (isSmall() && r1.isLarge())
         {
             for (const auto & x : small)
-                if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                    buffer.push_back(x.getValue());
+            {
+                if ( isUInt64() )
+                {
+                    if (rb64->contains(static_cast<UInt64>(x.getValue())) )
+                        buffer.push_back(x.getValue());
+                }
+                else
+                {
+                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
+                        buffer.push_back(x.getValue());
+                }
+            }
 
             // Clear out the original values
             small.clear();
@@ -175,10 +252,18 @@ public:
         }
         else
         {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            roaring_bitmap_and_inplace(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
+            if ( isUInt64() )
+            {
+                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+                *rb64 &= *newRb64;
+            }
+            else
+            {
+                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+                roaring_bitmap_and_inplace(rb, rb1);
+                if (r1.isSmall())
+                    roaring_bitmap_free(rb1);
+            }
         }
     }
 
@@ -194,10 +279,16 @@ public:
     {
         if (isSmall())
             toLarge();
-        roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-        roaring_bitmap_xor_inplace(rb, rb1);
-        if (r1.isSmall())
-            roaring_bitmap_free(rb1);
+
+        if ( isUInt64() ){
+            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+            *rb64 ^= *newRb64;
+        } else {
+            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+            roaring_bitmap_xor_inplace(rb, rb1);
+            if (r1.isSmall())
+                roaring_bitmap_free(rb1);
+        }
     }
 
     /**
@@ -224,8 +315,18 @@ public:
         else if (isSmall() && r1.isLarge())
         {
             for (const auto & x : small)
-                if (!roaring_bitmap_contains(r1.rb, x.getValue()))
-                    buffer.push_back(x.getValue());
+            {
+                if ( isUInt64() )
+                {
+                    if ( !rb64->contains(static_cast<UInt64>(x.getValue())))
+                        buffer.push_back(x.getValue());
+                }
+                else
+                {
+                    if (!roaring_bitmap_contains(r1.rb, x.getValue()))
+                        buffer.push_back(x.getValue());
+                }
+            }
 
             // Clear out the original values
             small.clear();
@@ -237,10 +338,18 @@ public:
         }
         else
         {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            roaring_bitmap_andnot_inplace(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
+            if ( isUInt64() )
+            {
+                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+                *rb64 -= *newRb64;
+            }
+            else
+            {
+                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+                roaring_bitmap_andnot_inplace(rb, rb1);
+                if (r1.isSmall())
+                    roaring_bitmap_free(rb1);
+            }
         }
     }
 
@@ -259,15 +368,31 @@ public:
         else if (isSmall() && r1.isLarge())
         {
             for (const auto & x : small)
-                if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                    ++retSize;
+            {
+                if ( isUInt64() )
+                {
+                    if ( rb64->contains(static_cast<UInt64>(x.getValue())))
+                        ++retSize;
+                } else {
+                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
+                        ++retSize;
+                }
+            }
         }
         else
         {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            retSize = roaring_bitmap_and_cardinality(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
+            if ( isUInt64() )
+            {
+                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+                retSize = ( *rb64 & *newRb64 ).cardinality();
+            }
+            else
+            {
+                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+                retSize = roaring_bitmap_and_cardinality(rb, rb1);
+                if (r1.isSmall())
+                    roaring_bitmap_free(rb1);
+            }
         }
         return retSize;
     }
@@ -311,10 +436,21 @@ public:
     {
         if (isSmall())
             toLarge();
-        roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-        UInt8 is_true = roaring_bitmap_equals(rb, rb1);
-        if (r1.isSmall())
-            roaring_bitmap_free(rb1);
+
+        UInt8 is_true = 0;
+        if ( isUInt64() )
+        {
+            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+            is_true = *rb64 == *newRb64;
+        }
+        else
+        {
+            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+            is_true = roaring_bitmap_equals(rb, rb1);
+            if (r1.isSmall())
+                roaring_bitmap_free(rb1);
+        }
+
         return is_true;
     }
 
@@ -335,18 +471,49 @@ public:
             else
             {
                 for (const auto & x : small)
-                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                        return 1;
+                {
+                    if ( isUInt64() )
+                    {
+                        if( r1.rb64->contains( static_cast<UInt64>(x.getValue()) ) )
+                            return 1;
+                    }
+                    else
+                    {
+                        if (roaring_bitmap_contains(r1.rb, x.getValue()))
+                            return 1;
+                    }
+                }
             }
         }
         else if (r1.isSmall())
         {
             for (const auto & x : r1.small)
-                if (roaring_bitmap_contains(rb, x.getValue()))
-                    return 1;
+            {
+                if ( isUInt64() )
+                {
+                    if( rb64->contains( static_cast<UInt64>(x.getValue()) ) )
+                        return 1;
+                }
+                else
+                {
+                    if (roaring_bitmap_contains(rb, x.getValue()))
+                        return 1;
+                }
+            }
+        }
+        else
+        {
+            if ( isUInt64() )
+            {
+                if( ( *rb64 & *r1.rb64 ).cardinality() > 0 )
+                    return 1;
+            }
+            else
+            {
+                if (roaring_bitmap_intersect(rb, r1.rb))
+                    return 1;
+            }
         }
-        else if (roaring_bitmap_intersect(rb, r1.rb))
-            return 1;
 
         return 0;
     }
@@ -379,28 +546,72 @@ public:
                 // r1_size + number of not found elements, if this sum becomes
                 // greater then r1 is not a subset.
                 for (const auto & x : small)
-                    if (!roaring_bitmap_contains(r1.rb, x.getValue()) && ++r1_size > small.size())
-                        return 0;
+                {
+                    if ( isUInt64() )
+                    {
+                        if (!r1.rb64->contains( static_cast<UInt64>(x.getValue()) ) && ++r1_size > small.size())
+                            return 0;
+                    }
+                    else
+                    {
+                        if (!roaring_bitmap_contains(r1.rb, x.getValue()) && ++r1_size > small.size())
+                            return 0;
+                    }
+                }
             }
         }
         else if (r1.isSmall())
         {
             for (const auto & x : r1.small)
-                if (!roaring_bitmap_contains(rb, x.getValue()))
-                    return 0;
+            {
+                if ( isUInt64() )
+                {
+                    if ( !rb64->contains( static_cast<UInt64>(x.getValue())) )
+                        return 0;
+                }
+                else
+                {
+                    if (!roaring_bitmap_contains(rb, x.getValue()))
+                        return 0;
+                }
+            }
+        }
+        else
+        {
+            if ( isUInt64() )
+            {
+                if (!r1.rb64->isSubset(*rb64))
+                    return 0;
+            }
+            else
+            {
+                if (!roaring_bitmap_is_subset(r1.rb, rb))
+                    return 0;
+            }
         }
-        else if (!roaring_bitmap_is_subset(r1.rb, rb))
-            return 0;
-
         return 1;
     }
 
     /**
      * Check whether this bitmap contains the argument.
      */
-    UInt8 rb_contains(const UInt32 x) const
+    UInt8 rb_contains(const UInt64 x) const
     {
-        return isSmall() ? small.find(x) != small.end() : roaring_bitmap_contains(rb, x);
+        if (isSmall())
+        {
+            return small.find(x) != small.end();
+        }
+        else
+        {
+            if ( isUInt64() )
+            {
+                return rb64->contains(x);
+            }
+            else
+            {
+                return roaring_bitmap_contains(rb, x);
+            }
+        }
     }
 
     /**
@@ -410,7 +621,11 @@ public:
     {
         if (isSmall())
             toLarge();
-        roaring_bitmap_remove(rb, offsetid);
+
+        if ( isUInt64() )
+            rb64->remove( offsetid );
+        else
+            roaring_bitmap_remove(rb, offsetid);
     }
 
     /**
@@ -423,7 +638,10 @@ public:
     {
         if (isSmall())
             toLarge();
-        roaring_bitmap_flip_inplace(rb, offsetstart, offsetend);
+        if ( isUInt64() )
+            rb64->flip( offsetstart, offsetend );
+        else
+            roaring_bitmap_flip_inplace(rb, offsetstart, offsetend);
     }
 
     /**
@@ -433,7 +651,11 @@ public:
     {
         if (isSmall())
             toLarge();
-        return roaring_bitmap_rank(rb, offsetid);
+
+        if ( isUInt64() )
+            return rb64->rank( offsetid );
+        else
+            return roaring_bitmap_rank(rb, offsetid);
     }
 
     /**
@@ -453,13 +675,23 @@ public:
         }
         else
         {
-            roaring_uint32_iterator_t iterator;
-            roaring_init_iterator(rb, &iterator);
-            while (iterator.has_value)
+            if ( isUInt64() )
             {
-                res_data.emplace_back(iterator.current_value);
-                roaring_advance_uint32_iterator(&iterator);
-                count++;
+                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++) {
+                    res_data.emplace_back( *iterator );
+                    count++;
+                }
+            }
+            else
+            {
+                roaring_uint32_iterator_t iterator;
+                roaring_init_iterator(rb, &iterator);
+                while (iterator.has_value)
+                {
+                    res_data.emplace_back(iterator.current_value);
+                    roaring_advance_uint32_iterator(&iterator);
+                    count++;
+                }
             }
         }
         return count;
@@ -468,7 +700,7 @@ public:
     /**
      * Return new set with specified range (not include the range_end)
      */
-    UInt64 rb_range(UInt32 range_start, UInt32 range_end, RoaringBitmapWithSmallSet & r1) const
+    UInt64 rb_range(UInt64 range_start, UInt64 range_end, RoaringBitmapWithSmallSet & r1) const
     {
         UInt64 count = 0;
         if (range_start >= range_end)
@@ -487,14 +719,33 @@ public:
         }
         else
         {
-            roaring_uint32_iterator_t iterator;
-            roaring_init_iterator(rb, &iterator);
-            roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
-            while (iterator.has_value && UInt32(iterator.current_value) < range_end)
+            if ( isUInt64() )
             {
-                r1.add(iterator.current_value);
-                roaring_advance_uint32_iterator(&iterator);
-                ++count;
+                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
+                {
+                    if( *iterator < range_start )
+                        continue;
+
+                    if( *iterator < range_end)
+                    {
+                        r1.add( *iterator );
+                        ++count;
+                    }
+                    else
+                        break;
+                }
+            }
+            else
+            {
+                roaring_uint32_iterator_t iterator;
+                roaring_init_iterator(rb, &iterator);
+                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
+                while (iterator.has_value && UInt32(iterator.current_value) < range_end)
+                {
+                    r1.add(iterator.current_value);
+                    roaring_advance_uint32_iterator(&iterator);
+                    ++count;
+                }
             }
         }
         return count;
@@ -503,7 +754,7 @@ public:
     /**
      * Return new set of the smallest `limit` values in set which is no less than `range_start`.
      */
-    UInt64 rb_limit(UInt32 range_start, UInt32 limit, RoaringBitmapWithSmallSet & r1) const
+    UInt64 rb_limit(UInt64 range_start, UInt64 limit, RoaringBitmapWithSmallSet & r1) const
     {
         UInt64 count = 0;
         if (isSmall())
@@ -526,14 +777,33 @@ public:
         }
         else
         {
-            roaring_uint32_iterator_t iterator;
-            roaring_init_iterator(rb, &iterator);
-            roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
-            while (UInt32(count) < limit && iterator.has_value)
+            if ( isUInt64() )
             {
-                r1.add(iterator.current_value);
-                roaring_advance_uint32_iterator(&iterator);
-                ++count;
+                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
+                {
+                    if( *iterator < range_start )
+                        continue;
+
+                    if( count < limit)
+                    {
+                        r1.add( *iterator );
+                        ++count;
+                    }
+                    else
+                        break;
+                }
+            }
+            else
+            {
+                roaring_uint32_iterator_t iterator;
+                roaring_init_iterator(rb, &iterator);
+                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
+                while (UInt32(count) < limit && iterator.has_value)
+                {
+                    r1.add(iterator.current_value);
+                    roaring_advance_uint32_iterator(&iterator);
+                    ++count;
+                }
             }
         }
         return count;
@@ -555,7 +825,10 @@ public:
         }
         else
         {
-            min_val = UInt64(roaring_bitmap_minimum(rb));
+            if ( isUInt64() )
+                min_val = rb64->minimum();
+            else
+                min_val = UInt64(roaring_bitmap_minimum(rb));
         }
         return min_val;
     }
@@ -576,7 +849,10 @@ public:
         }
         else
         {
-            max_val = UInt64(roaring_bitmap_maximum(rb));
+            if ( isUInt64() )
+                max_val = rb64->maximum();
+            else
+                max_val = UInt64(roaring_bitmap_maximum(rb));
         }
         return max_val;
     }
@@ -592,12 +868,144 @@ public:
         {
             if (from_vals[i] == to_vals[i])
                 continue;
-            bool changed = roaring_bitmap_remove_checked(rb, from_vals[i]);
-            if (changed)
-                roaring_bitmap_add(rb, to_vals[i]);
+
+            if ( isUInt64() )
+            {
+                bool changed = rb64->removeChecked( from_vals[i] );
+                if (changed)
+                    rb64->add(to_vals[i]);
+            }
+            else
+            {
+                bool changed = roaring_bitmap_remove_checked(rb, from_vals[i]);
+                if (changed)
+                    roaring_bitmap_add(rb, to_vals[i]);
+            }
         }
     }
 
+private:
+    /// To read and write the DB Buffer directly, migrate code from CRoaring
+    void db_roaring_bitmap_add_many(DB::ReadBuffer & db_buf, roaring_bitmap_t * r, size_t n_args)
+    {
+        void * container = nullptr; // hold value of last container touched
+        uint8_t typecode = 0; // typecode of last container touched
+        uint32_t prev = 0; // previous valued inserted
+        size_t i = 0; // index of value
+        int containerindex = 0;
+        if (n_args == 0)
+            return;
+        uint32_t val;
+        readBinary(val, db_buf);
+        container = containerptr_roaring_bitmap_add(r, val, &typecode, &containerindex);
+        prev = val;
+        ++i;
+        for (; i < n_args; ++i)
+        {
+            readBinary(val, db_buf);
+            if (((prev ^ val) >> 16) == 0)
+            { // no need to seek the container, it is at hand
+                // because we already have the container at hand, we can do the
+                // insertion
+                // automatically, bypassing the roaring_bitmap_add call
+                uint8_t newtypecode = typecode;
+                void * container2 = container_add(container, val & 0xFFFF, typecode, &newtypecode);
+                // rare instance when we need to
+                if (container2 != container)
+                {
+                    // change the container type
+                    container_free(container, typecode);
+                    ra_set_container_at_index(&r->high_low_container, containerindex, container2, newtypecode);
+                    typecode = newtypecode;
+                    container = container2;
+                }
+            }
+            else
+            {
+                container = containerptr_roaring_bitmap_add(r, val, &typecode, &containerindex);
+            }
+            prev = val;
+        }
+    }
+
+    void db_ra_to_uint32_array(DB::WriteBuffer & db_buf, roaring_array_t * ra) const
+    {
+        size_t ctr = 0;
+        for (Int32 i = 0; i < ra->size; ++i)
+        {
+            Int32 num_added = db_container_to_uint32_array(db_buf, ra->containers[i], ra->typecodes[i], (static_cast<UInt32>(ra->keys[i])) << 16);
+            ctr += num_added;
+        }
+    }
+
+    UInt32 db_container_to_uint32_array(DB::WriteBuffer & db_buf, const void * container, uint8_t typecode, UInt32 base) const
+    {
+        container = container_unwrap_shared(container, &typecode);
+        switch (typecode)
+        {
+            case BITSET_CONTAINER_TYPE_CODE:
+                return db_bitset_container_to_uint32_array(db_buf, static_cast<const bitset_container_t *>(container), base);
+            case ARRAY_CONTAINER_TYPE_CODE:
+                return db_array_container_to_uint32_array(db_buf, static_cast<const array_container_t *>(container), base);
+            case RUN_CONTAINER_TYPE_CODE:
+                return db_run_container_to_uint32_array(db_buf, static_cast<const run_container_t *>(container), base);
+        }
+        return 0;
+    }
+
+    UInt32 db_bitset_container_to_uint32_array(DB::WriteBuffer & db_buf, const bitset_container_t * cont, UInt32 base) const
+    {
+        return static_cast<UInt32>(db_bitset_extract_setbits(db_buf, cont->array, BITSET_CONTAINER_SIZE_IN_WORDS, base));
+    }
+
+    size_t db_bitset_extract_setbits(DB::WriteBuffer & db_buf, UInt64 * bitset, size_t length, UInt32 base) const
+    {
+        UInt32 outpos = 0;
+        for (size_t i = 0; i < length; ++i)
+        {
+            UInt64 w = bitset[i];
+            while (w != 0)
+            {
+                UInt64 t = w & (~w + 1); // on x64, should compile to BLSI (careful: the Intel compiler seems to fail)
+                UInt32 r = __builtin_ctzll(w); // on x64, should compile to TZCNT
+                UInt32 val = r + base;
+                writePODBinary(val, db_buf);
+                outpos++;
+                w ^= t;
+            }
+            base += 64;
+        }
+        return outpos;
+    }
+
+    int db_array_container_to_uint32_array(DB::WriteBuffer & db_buf, const array_container_t * cont, UInt32 base) const
+    {
+        UInt32 outpos = 0;
+        for (Int32 i = 0; i < cont->cardinality; ++i)
+        {
+            const UInt32 val = base + cont->array[i];
+            writePODBinary(val, db_buf);
+            outpos++;
+        }
+        return outpos;
+    }
+
+    int db_run_container_to_uint32_array(DB::WriteBuffer & db_buf, const run_container_t * cont, UInt32 base) const
+    {
+        UInt32 outpos = 0;
+        for (Int32 i = 0; i < cont->n_runs; ++i)
+        {
+            UInt32 run_start = base + cont->runs[i].value;
+            UInt16 le = cont->runs[i].length;
+            for (Int32 j = 0; j <= le; ++j)
+            {
+                UInt32 val = run_start + j;
+                writePODBinary(val, db_buf);
+                outpos++;
+            }
+        }
+        return outpos;
+    }
 };
 
 template <typename T>
diff --git a/tests/queries/0_stateless/00829_bitmap64_function.reference b/tests/queries/0_stateless/00829_bitmap64_function.reference
new file mode 100644
index 00000000000..20f5e19d780
--- /dev/null
+++ b/tests/queries/0_stateless/00829_bitmap64_function.reference
@@ -0,0 +1,14 @@
+2019-01-01	100
+2019-01-02	110
+2019-01-03	210
+210
+2019-01-01	100	[4294967296,4294967297,4294967298,4294967299,4294967300,4294967301,4294967302,4294967303,4294967304,4294967305,4294967306,4294967307,4294967308,4294967309,4294967310,4294967311,4294967312,4294967313,4294967314,4294967315,4294967316,4294967317,4294967318,4294967319,4294967320,4294967321,4294967322,4294967323,4294967324,4294967325,4294967326,4294967327,4294967328,4294967329,4294967330,4294967331,4294967332,4294967333,4294967334,4294967335,4294967336,4294967337,4294967338,4294967339,4294967340,4294967341,4294967342,4294967343,4294967344,4294967345,4294967346,4294967347,4294967348,4294967349,4294967350,4294967351,4294967352,4294967353,4294967354,4294967355,4294967356,4294967357,4294967358,4294967359,4294967360,4294967361,4294967362,4294967363,4294967364,4294967365,4294967366,4294967367,4294967368,4294967369,4294967370,4294967371,4294967372,4294967373,4294967374,4294967375,4294967376,4294967377,4294967378,4294967379,4294967380,4294967381,4294967382,4294967383,4294967384,4294967385,4294967386,4294967387,4294967388,4294967389,4294967390,4294967391,4294967392,4294967393,4294967394,4294967395]
+2019-01-02	110	[4294967385,4294967386,4294967387,4294967388,4294967389,4294967390,4294967391,4294967392,4294967393,4294967394,4294967395,4294967396,4294967397,4294967398,4294967399,4294967400,4294967401,4294967402,4294967403,4294967404,4294967405,4294967406,4294967407,4294967408,4294967409,4294967410,4294967411,4294967412,4294967413,4294967414,4294967415,4294967416,4294967417,4294967418,4294967419,4294967420,4294967421,4294967422,4294967423,4294967424,4294967425,4294967426,4294967427,4294967428,4294967429,4294967430,4294967431,4294967432,4294967433,4294967434,4294967435,4294967436,4294967437,4294967438,4294967439,4294967440,4294967441,4294967442,4294967443,4294967444,4294967445,4294967446,4294967447,4294967448,4294967449,4294967450,4294967451,4294967452,4294967453,4294967454,4294967455,4294967456,4294967457,4294967458,4294967459,4294967460,4294967461,4294967462,4294967463,4294967464,4294967465,4294967466,4294967467,4294967468,4294967469,4294967470,4294967471,4294967472,4294967473,4294967474,4294967475,4294967476,4294967477,4294967478,4294967479,4294967480,4294967481,4294967482,4294967483,4294967484,4294967485,4294967486,4294967487,4294967488,4294967489,4294967490,4294967491,4294967492,4294967493,4294967494]
+2019-01-03	210	[4294967296,4294967297,4294967298,4294967299,4294967300,4294967301,4294967302,4294967303,4294967304,4294967305,4294967306,4294967307,4294967308,4294967309,4294967310,4294967311,4294967312,4294967313,4294967314,4294967315,4294967316,4294967317,4294967318,4294967319,4294967320,4294967321,4294967322,4294967323,4294967324,4294967325,4294967326,4294967327,4294967328,4294967329,4294967330,4294967331,4294967332,4294967333,4294967334,4294967335,4294967336,4294967337,4294967338,4294967339,4294967340,4294967341,4294967342,4294967343,4294967344,4294967345,4294967346,4294967347,4294967348,4294967349,4294967350,4294967351,4294967352,4294967353,4294967354,4294967355,4294967356,4294967357,4294967358,4294967359,4294967360,4294967361,4294967362,4294967363,4294967364,4294967365,4294967366,4294967367,4294967368,4294967369,4294967370,4294967371,4294967372,4294967373,4294967374,4294967375,4294967376,4294967377,4294967378,4294967379,4294967380,4294967381,4294967382,4294967383,4294967384,4294967385,4294967386,4294967387,4294967388,4294967389,4294967390,4294967391,4294967392,4294967393,4294967394,4294967395,4294967396,4294967397,4294967398,4294967399,4294967400,4294967401,4294967402,4294967403,4294967404,4294967405,4294967406,4294967407,4294967408,4294967409,4294967410,4294967411,4294967412,4294967413,4294967414,4294967415,4294967416,4294967417,4294967418,4294967419,4294967420,4294967421,4294967422,4294967423,4294967424,4294967425,4294967426,4294967427,4294967428,4294967429,4294967430,4294967431,4294967432,4294967433,4294967434,4294967435,4294967436,4294967437,4294967438,4294967439,4294967440,4294967441,4294967442,4294967443,4294967444,4294967445,4294967446,4294967447,4294967448,4294967449,4294967450,4294967451,4294967452,4294967453,4294967454,4294967455,4294967456,4294967457,4294967458,4294967459,4294967460,4294967461,4294967462,4294967463,4294967464,4294967465,4294967466,4294967467,4294967468,4294967469,4294967470,4294967471,4294967472,4294967473,4294967474,4294967475,4294967476,4294967477,4294967478,4294967479,4294967480,4294967481,4294967482,4294967483,4294967484,4294967485,4294967486,4294967487,4294967488,4294967489,4294967490,4294967491,4294967492,4294967493,4294967494,4294967495,4294967496,4294967497,4294967498,4294967499,4294967500,4294967501,4294967502,4294967503,4294967504,4294967505]
+110	100	199	11	99	188
+110	100	199	11	99	188
+211
+211
+209
+[4294967296,4294967297,4294967298]
+[4294967296,4294967297,4294967298]
diff --git a/tests/queries/0_stateless/00829_bitmap64_function.sql b/tests/queries/0_stateless/00829_bitmap64_function.sql
new file mode 100644
index 00000000000..9d33dcefff9
--- /dev/null
+++ b/tests/queries/0_stateless/00829_bitmap64_function.sql
@@ -0,0 +1,73 @@
+DROP TABLE IF EXISTS bitmap_test;
+CREATE TABLE bitmap_test(pickup_date Date, city_id UInt32, uid UInt64)ENGINE = Memory;
+INSERT INTO bitmap_test SELECT '2019-01-01', 1, 4294967295 + number FROM numbers(1,100);
+INSERT INTO bitmap_test SELECT '2019-01-02', 1, 4294967295 + number FROM numbers(90,110);
+INSERT INTO bitmap_test SELECT '2019-01-03', 2, 4294967295 + number FROM numbers(1,210);
+
+
+DROP TABLE IF EXISTS bitmap_state_test;
+CREATE TABLE bitmap_state_test
+(
+	pickup_date Date,
+	city_id UInt32,
+    uv AggregateFunction( groupBitmap, UInt64 )
+)
+ENGINE = AggregatingMergeTree() PARTITION BY toYYYYMM(pickup_date) ORDER BY (pickup_date, city_id);
+
+
+INSERT INTO bitmap_state_test SELECT
+    pickup_date,
+    city_id,
+    groupBitmapState(uid) AS uv
+FROM bitmap_test
+GROUP BY pickup_date, city_id;
+
+SELECT pickup_date, groupBitmapMerge(uv) AS users from bitmap_state_test group by pickup_date;
+
+SELECT groupBitmap( uid ) AS user_num FROM bitmap_test;
+
+SELECT pickup_date, groupBitmap( uid ) AS user_num, bitmapToArray(groupBitmapState( uid )) AS users FROM bitmap_test GROUP BY pickup_date;
+
+SELECT
+    bitmapCardinality(day_today) AS today_users,
+    bitmapCardinality(day_before) AS before_users,
+    bitmapOrCardinality(day_today, day_before) AS all_users,
+    bitmapAndCardinality(day_today, day_before) AS old_users,
+    bitmapAndnotCardinality(day_today, day_before) AS new_users,
+    bitmapXorCardinality(day_today, day_before) AS diff_users
+FROM
+(
+ SELECT city_id, groupBitmapState( uid ) AS day_today FROM bitmap_test WHERE pickup_date = '2019-01-02' GROUP BY city_id
+) js1
+ALL LEFT JOIN
+(
+ SELECT city_id, groupBitmapState( uid ) AS day_before FROM bitmap_test WHERE pickup_date = '2019-01-01' GROUP BY city_id
+) js2
+USING city_id;
+
+SELECT
+    bitmapCardinality(day_today) AS today_users,
+    bitmapCardinality(day_before) AS before_users,
+    bitmapCardinality(bitmapOr(day_today, day_before))ll_users,
+    bitmapCardinality(bitmapAnd(day_today, day_before)) AS old_users,
+    bitmapCardinality(bitmapAndnot(day_today, day_before)) AS new_users,
+    bitmapCardinality(bitmapXor(day_today, day_before)) AS diff_users
+FROM
+(
+ SELECT city_id, groupBitmapState( uid ) AS day_today FROM bitmap_test WHERE pickup_date = '2019-01-02' GROUP BY city_id
+) js1
+ALL LEFT JOIN
+(
+ SELECT city_id, groupBitmapState( uid ) AS day_before FROM bitmap_test WHERE pickup_date = '2019-01-01' GROUP BY city_id
+) js2
+USING city_id;
+
+SELECT count(*) FROM bitmap_test WHERE bitmapHasAny((SELECT groupBitmapState(uid) FROM bitmap_test WHERE pickup_date = '2019-01-01'), bitmapBuild([uid]));
+
+SELECT count(*) FROM bitmap_test WHERE bitmapHasAny(bitmapBuild([uid]), (SELECT groupBitmapState(uid) FROM bitmap_test WHERE pickup_date = '2019-01-01'));
+
+SELECT count(*) FROM bitmap_test WHERE 0 = bitmapHasAny((SELECT groupBitmapState(uid) FROM bitmap_test WHERE pickup_date = '2019-01-01'), bitmapBuild([uid]));
+
+SELECT bitmapToArray(bitmapAnd(groupBitmapState(uid), bitmapBuild(CAST([4294967296, 4294967297, 4294967298], 'Array(UInt64)')))) FROM bitmap_test GROUP BY city_id;
+
+

From cc75c33064c0a380def32b44e134985f73ce0993 Mon Sep 17 00:00:00 2001
From: Andy Yang <yangzhaohui168@gmail.com>
Date: Mon, 7 Dec 2020 23:55:04 +0800
Subject: [PATCH 396/504] modify code style

---
 .../AggregateFunctionGroupBitmapData.h        | 1935 ++++++++---------
 1 file changed, 915 insertions(+), 1020 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
index 5e93fbb60d8..000edd64a0f 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
@@ -1,1020 +1,915 @@
-#pragma once
-
-#include <algorithm>
-#include <IO/ReadHelpers.h>
-#include <IO/WriteHelpers.h>
-#include <boost/noncopyable.hpp>
-#include <Common/HashTable/SmallTable.h>
-#include <Common/PODArray.h>
-
-// Include this header last, because it is an auto-generated dump of questionable
-// garbage that breaks the build (e.g. it changes _POSIX_C_SOURCE).
-// TODO: find out what it is. On github, they have proper interface headers like
-// this one: https://github.com/RoaringBitmap/CRoaring/blob/master/include/roaring/roaring.h
-#include <roaring/roaring.h>
-#include <roaring64map.hh>
-
-namespace DB
-{
-/**
-  * For a small number of values - an array of fixed size "on the stack".
-  * For large, roaring_bitmap_t is allocated.
-  * For a description of the roaring_bitmap_t, see: https://github.com/RoaringBitmap/CRoaring
-  */
-template <typename T, UInt8 small_set_size>
-class RoaringBitmapWithSmallSet : private boost::noncopyable
-{
-private:
-    using Small = SmallSet<T, small_set_size>;
-    using ValueBuffer = std::vector<T>;
-    bool bUInt64 = sizeof(T) >= 8;
-
-    Small small;
-    roaring_bitmap_t * rb = nullptr;
-    std::shared_ptr<Roaring64Map> rb64 = nullptr;
-
-    void toLarge()
-    {
-        if ( isUInt64() )
-            rb64 = std::make_shared<Roaring64Map>();
-        else
-            rb = roaring_bitmap_create();
-
-        for (const auto & x : small){
-            if ( isUInt64() )
-                rb64->add( static_cast<UInt64>(x.getValue()) );
-            else
-                roaring_bitmap_add(rb, x.getValue());
-        }
-    }
-
-public:
-    bool isLarge() const { return rb != nullptr || rb64 != nullptr; }
-
-    bool isSmall() const { return rb == nullptr &&  rb64 == nullptr; }
-
-    bool isUInt64() const { return bUInt64; }
-
-    ~RoaringBitmapWithSmallSet()
-    {
-        if (isLarge() && !isUInt64() )
-            roaring_bitmap_free(rb);
-    }
-
-    void add(T value)
-    {
-        if (isSmall())
-        {
-            if (small.find(value) == small.end())
-            {
-                if (!small.full())
-                    small.insert(value);
-                else
-                {
-                    toLarge();
-                    if ( isUInt64() )
-                        rb64->add( static_cast<UInt64>(value) );
-                    else
-                        roaring_bitmap_add(rb, value);
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-                rb64->add( static_cast<UInt64>(value) );
-            else
-                roaring_bitmap_add(rb, value);
-        }
-    }
-
-    UInt64 size() const
-    {
-        if (isSmall())
-        {
-            return small.size();
-        }
-        else
-        {
-            if ( isUInt64() )
-                return rb64->cardinality();
-            else
-                return roaring_bitmap_get_cardinality(rb);
-        }
-    }
-
-    void merge(const RoaringBitmapWithSmallSet & r1)
-    {
-        if (r1.isLarge())
-        {
-            if (isSmall())
-                toLarge();
-
-            if ( isUInt64() )
-                *rb64 |= *r1.rb64;
-            else
-                roaring_bitmap_or_inplace(rb, r1.rb);
-        }
-        else
-        {
-            for (const auto & x : r1.small)
-                add(x.getValue());
-        }
-    }
-
-    void read(DB::ReadBuffer & in)
-    {
-        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
-        UInt8 containerType = 0;
-        readBinary(containerType, in);
-
-        if( 0 == containerType )
-        {
-            small.read(in);
-        } else if( 1 == containerType ) {
-            std::string s;
-            readStringBinary(s,in);
-            rb = roaring_bitmap_portable_deserialize(s.c_str());
-            // It has been persisted in the bitmap and does not need to merge from small
-//            for (const auto & x : small) // merge from small
-//                roaring_bitmap_add(rb, x.getValue());
-        } else {
-            std::string s;
-            readStringBinary(s,in);
-            rb64 = std::make_shared<Roaring64Map>( Roaring64Map::read(s.c_str()) );
-        }
-    }
-
-    void write(DB::WriteBuffer & out) const
-    {
-        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
-        UInt8 containerType = 0;
-        if (isLarge())
-        {
-            if ( isUInt64() )
-                containerType = 2;
-            else
-                containerType = 1;
-        }
-
-        writeBinary(containerType, out);
-
-        if (isLarge())
-        {
-            if ( isUInt64() ){
-                uint32_t expectedsize = rb64->getSizeInBytes();
-                std::string s(expectedsize,0);
-                rb64->write(const_cast<char*>(s.data()));
-                writeStringBinary(s,out);
-            } else {
-                uint32_t expectedsize = roaring_bitmap_portable_size_in_bytes(rb);
-                std::string s(expectedsize,0);
-                roaring_bitmap_portable_serialize(rb, const_cast<char*>(s.data()));
-                writeStringBinary(s,out);
-            }
-        }
-        else
-            small.write(out);
-    }
-
-    roaring_bitmap_t * getRb() const { return rb; }
-
-    std::shared_ptr<Roaring64Map> getRb64() const { return rb64; }
-
-    Small & getSmall() const { return small; }
-
-    /**
-     * Get a new roaring_bitmap_t from elements of small
-     */
-    roaring_bitmap_t * getNewRbFromSmall() const
-    {
-        roaring_bitmap_t * smallRb = roaring_bitmap_create();
-        for (const auto & x : small)
-            roaring_bitmap_add(smallRb, x.getValue());
-        return smallRb;
-    }
-
-    /**
-     * Get a new Roaring64Map from elements of small
-     */
-    std::shared_ptr<Roaring64Map> getNewRb64FromSmall() const
-    {
-        std::shared_ptr<Roaring64Map> smallRb64 = std::make_shared<Roaring64Map>();
-        for (const auto & x : small)
-            smallRb64->add( static_cast<UInt64>(x.getValue()) );
-        return smallRb64;
-    }
-
-    /**
-     * Computes the intersection between two bitmaps
-     */
-    void rb_and(const RoaringBitmapWithSmallSet & r1)
-    {
-        ValueBuffer buffer;
-        if (isSmall() && r1.isSmall())
-        {
-            // intersect
-            for (const auto & x : small)
-                if (r1.small.find(x.getValue()) != r1.small.end())
-                    buffer.push_back(x.getValue());
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else if (isSmall() && r1.isLarge())
-        {
-            for (const auto & x : small)
-            {
-                if ( isUInt64() )
-                {
-                    if (rb64->contains(static_cast<UInt64>(x.getValue())) )
-                        buffer.push_back(x.getValue());
-                }
-                else
-                {
-                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                        buffer.push_back(x.getValue());
-                }
-            }
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-                *rb64 &= *newRb64;
-            }
-            else
-            {
-                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-                roaring_bitmap_and_inplace(rb, rb1);
-                if (r1.isSmall())
-                    roaring_bitmap_free(rb1);
-            }
-        }
-    }
-
-    /**
-     * Computes the union between two bitmaps.
-     */
-    void rb_or(const RoaringBitmapWithSmallSet & r1) { merge(r1); }
-
-    /**
-     * Computes the symmetric difference (xor) between two bitmaps.
-     */
-    void rb_xor(const RoaringBitmapWithSmallSet & r1)
-    {
-        if (isSmall())
-            toLarge();
-
-        if ( isUInt64() ){
-            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-            *rb64 ^= *newRb64;
-        } else {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            roaring_bitmap_xor_inplace(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
-        }
-    }
-
-    /**
-     * Computes the difference (andnot) between two bitmaps
-     */
-    void rb_andnot(const RoaringBitmapWithSmallSet & r1)
-    {
-        ValueBuffer buffer;
-        if (isSmall() && r1.isSmall())
-        {
-            // subtract
-            for (const auto & x : small)
-                if (r1.small.find(x.getValue()) == r1.small.end())
-                    buffer.push_back(x.getValue());
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else if (isSmall() && r1.isLarge())
-        {
-            for (const auto & x : small)
-            {
-                if ( isUInt64() )
-                {
-                    if ( !rb64->contains(static_cast<UInt64>(x.getValue())))
-                        buffer.push_back(x.getValue());
-                }
-                else
-                {
-                    if (!roaring_bitmap_contains(r1.rb, x.getValue()))
-                        buffer.push_back(x.getValue());
-                }
-            }
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-                *rb64 -= *newRb64;
-            }
-            else
-            {
-                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-                roaring_bitmap_andnot_inplace(rb, rb1);
-                if (r1.isSmall())
-                    roaring_bitmap_free(rb1);
-            }
-        }
-    }
-
-    /**
-     * Computes the cardinality of the intersection between two bitmaps.
-     */
-    UInt64 rb_and_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 retSize = 0;
-        if (isSmall() && r1.isSmall())
-        {
-            for (const auto & x : small)
-                if (r1.small.find(x.getValue()) != r1.small.end())
-                    ++retSize;
-        }
-        else if (isSmall() && r1.isLarge())
-        {
-            for (const auto & x : small)
-            {
-                if ( isUInt64() )
-                {
-                    if ( rb64->contains(static_cast<UInt64>(x.getValue())))
-                        ++retSize;
-                } else {
-                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                        ++retSize;
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-                retSize = ( *rb64 & *newRb64 ).cardinality();
-            }
-            else
-            {
-                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-                retSize = roaring_bitmap_and_cardinality(rb, rb1);
-                if (r1.isSmall())
-                    roaring_bitmap_free(rb1);
-            }
-        }
-        return retSize;
-    }
-
-    /**
-     * Computes the cardinality of the union between two bitmaps.
-    */
-    UInt64 rb_or_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 c1 = size();
-        UInt64 c2 = r1.size();
-        UInt64 inter = rb_and_cardinality(r1);
-        return c1 + c2 - inter;
-    }
-
-    /**
-     * Computes the cardinality of the symmetric difference (andnot) between two bitmaps.
-    */
-    UInt64 rb_xor_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 c1 = size();
-        UInt64 c2 = r1.size();
-        UInt64 inter = rb_and_cardinality(r1);
-        return c1 + c2 - 2 * inter;
-    }
-
-    /**
-     * Computes the cardinality of the difference (andnot) between two bitmaps.
-     */
-    UInt64 rb_andnot_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 c1 = size();
-        UInt64 inter = rb_and_cardinality(r1);
-        return c1 - inter;
-    }
-
-    /**
-     * Return 1 if the two bitmaps contain the same elements.
-     */
-    UInt8 rb_equals(const RoaringBitmapWithSmallSet & r1)
-    {
-        if (isSmall())
-            toLarge();
-
-        UInt8 is_true = 0;
-        if ( isUInt64() )
-        {
-            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-            is_true = *rb64 == *newRb64;
-        }
-        else
-        {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            is_true = roaring_bitmap_equals(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
-        }
-
-        return is_true;
-    }
-
-    /**
-     * Check whether two bitmaps intersect.
-     * Intersection with an empty set is always 0 (consistent with hasAny).
-     */
-    UInt8 rb_intersect(const RoaringBitmapWithSmallSet & r1) const
-    {
-        if (isSmall())
-        {
-            if (r1.isSmall())
-            {
-                for (const auto & x : r1.small)
-                    if (small.find(x.getValue()) != small.end())
-                        return 1;
-            }
-            else
-            {
-                for (const auto & x : small)
-                {
-                    if ( isUInt64() )
-                    {
-                        if( r1.rb64->contains( static_cast<UInt64>(x.getValue()) ) )
-                            return 1;
-                    }
-                    else
-                    {
-                        if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                            return 1;
-                    }
-                }
-            }
-        }
-        else if (r1.isSmall())
-        {
-            for (const auto & x : r1.small)
-            {
-                if ( isUInt64() )
-                {
-                    if( rb64->contains( static_cast<UInt64>(x.getValue()) ) )
-                        return 1;
-                }
-                else
-                {
-                    if (roaring_bitmap_contains(rb, x.getValue()))
-                        return 1;
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                if( ( *rb64 & *r1.rb64 ).cardinality() > 0 )
-                    return 1;
-            }
-            else
-            {
-                if (roaring_bitmap_intersect(rb, r1.rb))
-                    return 1;
-            }
-        }
-
-        return 0;
-    }
-
-    /**
-     * Check whether the argument is the subset of this set.
-     * Empty set is a subset of any other set (consistent with hasAll).
-     */
-    UInt8 rb_is_subset(const RoaringBitmapWithSmallSet & r1) const
-    {
-        if (isSmall())
-        {
-            if (r1.isSmall())
-            {
-                for (const auto & x : r1.small)
-                    if (small.find(x.getValue()) == small.end())
-                        return 0;
-            }
-            else
-            {
-                UInt64 r1_size = r1.size();
-
-                if (r1_size > small.size())
-                    return 0; // A bigger set can not be a subset of ours.
-
-                // This is a rare case with a small number of elements on
-                // both sides: r1 was promoted to large for some reason and
-                // it is still not larger than our small set.
-                // If r1 is our subset then our size must be equal to
-                // r1_size + number of not found elements, if this sum becomes
-                // greater then r1 is not a subset.
-                for (const auto & x : small)
-                {
-                    if ( isUInt64() )
-                    {
-                        if (!r1.rb64->contains( static_cast<UInt64>(x.getValue()) ) && ++r1_size > small.size())
-                            return 0;
-                    }
-                    else
-                    {
-                        if (!roaring_bitmap_contains(r1.rb, x.getValue()) && ++r1_size > small.size())
-                            return 0;
-                    }
-                }
-            }
-        }
-        else if (r1.isSmall())
-        {
-            for (const auto & x : r1.small)
-            {
-                if ( isUInt64() )
-                {
-                    if ( !rb64->contains( static_cast<UInt64>(x.getValue())) )
-                        return 0;
-                }
-                else
-                {
-                    if (!roaring_bitmap_contains(rb, x.getValue()))
-                        return 0;
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                if (!r1.rb64->isSubset(*rb64))
-                    return 0;
-            }
-            else
-            {
-                if (!roaring_bitmap_is_subset(r1.rb, rb))
-                    return 0;
-            }
-        }
-        return 1;
-    }
-
-    /**
-     * Check whether this bitmap contains the argument.
-     */
-    UInt8 rb_contains(const UInt64 x) const
-    {
-        if (isSmall())
-        {
-            return small.find(x) != small.end();
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                return rb64->contains(x);
-            }
-            else
-            {
-                return roaring_bitmap_contains(rb, x);
-            }
-        }
-    }
-
-    /**
-     * Remove value
-     */
-    void rb_remove(UInt64 offsetid)
-    {
-        if (isSmall())
-            toLarge();
-
-        if ( isUInt64() )
-            rb64->remove( offsetid );
-        else
-            roaring_bitmap_remove(rb, offsetid);
-    }
-
-    /**
-     * compute (in place) the negation of the roaring bitmap within a specified
-     * interval: [range_start, range_end). The number of negated values is
-     * range_end - range_start.
-     * Areas outside the range are passed through unchanged.
-     */
-    void rb_flip(UInt64 offsetstart, UInt64 offsetend)
-    {
-        if (isSmall())
-            toLarge();
-        if ( isUInt64() )
-            rb64->flip( offsetstart, offsetend );
-        else
-            roaring_bitmap_flip_inplace(rb, offsetstart, offsetend);
-    }
-
-    /**
-     * returns the number of integers that are smaller or equal to offsetid.
-     */
-    UInt64 rb_rank(UInt64 offsetid)
-    {
-        if (isSmall())
-            toLarge();
-
-        if ( isUInt64() )
-            return rb64->rank( offsetid );
-        else
-            return roaring_bitmap_rank(rb, offsetid);
-    }
-
-    /**
-     * Convert elements to integer array, return number of elements
-     */
-    template <typename Element>
-    UInt64 rb_to_array(PaddedPODArray<Element> & res_data) const
-    {
-        UInt64 count = 0;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                res_data.emplace_back(x.getValue());
-                count++;
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++) {
-                    res_data.emplace_back( *iterator );
-                    count++;
-                }
-            }
-            else
-            {
-                roaring_uint32_iterator_t iterator;
-                roaring_init_iterator(rb, &iterator);
-                while (iterator.has_value)
-                {
-                    res_data.emplace_back(iterator.current_value);
-                    roaring_advance_uint32_iterator(&iterator);
-                    count++;
-                }
-            }
-        }
-        return count;
-    }
-
-    /**
-     * Return new set with specified range (not include the range_end)
-     */
-    UInt64 rb_range(UInt64 range_start, UInt64 range_end, RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 count = 0;
-        if (range_start >= range_end)
-            return count;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt32(val) >= range_start && UInt32(val) < range_end)
-                {
-                    r1.add(val);
-                    ++count;
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
-                {
-                    if( *iterator < range_start )
-                        continue;
-
-                    if( *iterator < range_end)
-                    {
-                        r1.add( *iterator );
-                        ++count;
-                    }
-                    else
-                        break;
-                }
-            }
-            else
-            {
-                roaring_uint32_iterator_t iterator;
-                roaring_init_iterator(rb, &iterator);
-                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
-                while (iterator.has_value && UInt32(iterator.current_value) < range_end)
-                {
-                    r1.add(iterator.current_value);
-                    roaring_advance_uint32_iterator(&iterator);
-                    ++count;
-                }
-            }
-        }
-        return count;
-    }
-
-    /**
-     * Return new set of the smallest `limit` values in set which is no less than `range_start`.
-     */
-    UInt64 rb_limit(UInt64 range_start, UInt64 limit, RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 count = 0;
-        if (isSmall())
-        {
-            std::vector<T> answer;
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt32(val) >= range_start)
-                {
-                    answer.push_back(val);
-                }
-            }
-            sort(answer.begin(), answer.end());
-            if (limit > answer.size())
-                limit = answer.size();
-            for (size_t i = 0; i < limit; ++i)
-                r1.add(answer[i]);
-            count = UInt64(limit);
-        }
-        else
-        {
-            if ( isUInt64() )
-            {
-                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
-                {
-                    if( *iterator < range_start )
-                        continue;
-
-                    if( count < limit)
-                    {
-                        r1.add( *iterator );
-                        ++count;
-                    }
-                    else
-                        break;
-                }
-            }
-            else
-            {
-                roaring_uint32_iterator_t iterator;
-                roaring_init_iterator(rb, &iterator);
-                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
-                while (UInt32(count) < limit && iterator.has_value)
-                {
-                    r1.add(iterator.current_value);
-                    roaring_advance_uint32_iterator(&iterator);
-                    ++count;
-                }
-            }
-        }
-        return count;
-    }
-
-    UInt64 rb_min() const
-    {
-        UInt64 min_val = UINT32_MAX;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt64(val) < min_val)
-                {
-                    min_val = UInt64(val);
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-                min_val = rb64->minimum();
-            else
-                min_val = UInt64(roaring_bitmap_minimum(rb));
-        }
-        return min_val;
-    }
-
-    UInt64 rb_max() const
-    {
-        UInt64 max_val = 0;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt64(val) > max_val)
-                {
-                    max_val = UInt64(val);
-                }
-            }
-        }
-        else
-        {
-            if ( isUInt64() )
-                max_val = rb64->maximum();
-            else
-                max_val = UInt64(roaring_bitmap_maximum(rb));
-        }
-        return max_val;
-    }
-
-    /**
-     * Replace value
-     */
-    void rb_replace(const UInt32 * from_vals, const UInt32 * to_vals, size_t num)
-    {
-        if (isSmall())
-            toLarge();
-        for (size_t i = 0; i < num; ++i)
-        {
-            if (from_vals[i] == to_vals[i])
-                continue;
-
-            if ( isUInt64() )
-            {
-                bool changed = rb64->removeChecked( from_vals[i] );
-                if (changed)
-                    rb64->add(to_vals[i]);
-            }
-            else
-            {
-                bool changed = roaring_bitmap_remove_checked(rb, from_vals[i]);
-                if (changed)
-                    roaring_bitmap_add(rb, to_vals[i]);
-            }
-        }
-    }
-
-private:
-    /// To read and write the DB Buffer directly, migrate code from CRoaring
-    void db_roaring_bitmap_add_many(DB::ReadBuffer & db_buf, roaring_bitmap_t * r, size_t n_args)
-    {
-        void * container = nullptr; // hold value of last container touched
-        uint8_t typecode = 0; // typecode of last container touched
-        uint32_t prev = 0; // previous valued inserted
-        size_t i = 0; // index of value
-        int containerindex = 0;
-        if (n_args == 0)
-            return;
-        uint32_t val;
-        readBinary(val, db_buf);
-        container = containerptr_roaring_bitmap_add(r, val, &typecode, &containerindex);
-        prev = val;
-        ++i;
-        for (; i < n_args; ++i)
-        {
-            readBinary(val, db_buf);
-            if (((prev ^ val) >> 16) == 0)
-            { // no need to seek the container, it is at hand
-                // because we already have the container at hand, we can do the
-                // insertion
-                // automatically, bypassing the roaring_bitmap_add call
-                uint8_t newtypecode = typecode;
-                void * container2 = container_add(container, val & 0xFFFF, typecode, &newtypecode);
-                // rare instance when we need to
-                if (container2 != container)
-                {
-                    // change the container type
-                    container_free(container, typecode);
-                    ra_set_container_at_index(&r->high_low_container, containerindex, container2, newtypecode);
-                    typecode = newtypecode;
-                    container = container2;
-                }
-            }
-            else
-            {
-                container = containerptr_roaring_bitmap_add(r, val, &typecode, &containerindex);
-            }
-            prev = val;
-        }
-    }
-
-    void db_ra_to_uint32_array(DB::WriteBuffer & db_buf, roaring_array_t * ra) const
-    {
-        size_t ctr = 0;
-        for (Int32 i = 0; i < ra->size; ++i)
-        {
-            Int32 num_added = db_container_to_uint32_array(db_buf, ra->containers[i], ra->typecodes[i], (static_cast<UInt32>(ra->keys[i])) << 16);
-            ctr += num_added;
-        }
-    }
-
-    UInt32 db_container_to_uint32_array(DB::WriteBuffer & db_buf, const void * container, uint8_t typecode, UInt32 base) const
-    {
-        container = container_unwrap_shared(container, &typecode);
-        switch (typecode)
-        {
-            case BITSET_CONTAINER_TYPE_CODE:
-                return db_bitset_container_to_uint32_array(db_buf, static_cast<const bitset_container_t *>(container), base);
-            case ARRAY_CONTAINER_TYPE_CODE:
-                return db_array_container_to_uint32_array(db_buf, static_cast<const array_container_t *>(container), base);
-            case RUN_CONTAINER_TYPE_CODE:
-                return db_run_container_to_uint32_array(db_buf, static_cast<const run_container_t *>(container), base);
-        }
-        return 0;
-    }
-
-    UInt32 db_bitset_container_to_uint32_array(DB::WriteBuffer & db_buf, const bitset_container_t * cont, UInt32 base) const
-    {
-        return static_cast<UInt32>(db_bitset_extract_setbits(db_buf, cont->array, BITSET_CONTAINER_SIZE_IN_WORDS, base));
-    }
-
-    size_t db_bitset_extract_setbits(DB::WriteBuffer & db_buf, UInt64 * bitset, size_t length, UInt32 base) const
-    {
-        UInt32 outpos = 0;
-        for (size_t i = 0; i < length; ++i)
-        {
-            UInt64 w = bitset[i];
-            while (w != 0)
-            {
-                UInt64 t = w & (~w + 1); // on x64, should compile to BLSI (careful: the Intel compiler seems to fail)
-                UInt32 r = __builtin_ctzll(w); // on x64, should compile to TZCNT
-                UInt32 val = r + base;
-                writePODBinary(val, db_buf);
-                outpos++;
-                w ^= t;
-            }
-            base += 64;
-        }
-        return outpos;
-    }
-
-    int db_array_container_to_uint32_array(DB::WriteBuffer & db_buf, const array_container_t * cont, UInt32 base) const
-    {
-        UInt32 outpos = 0;
-        for (Int32 i = 0; i < cont->cardinality; ++i)
-        {
-            const UInt32 val = base + cont->array[i];
-            writePODBinary(val, db_buf);
-            outpos++;
-        }
-        return outpos;
-    }
-
-    int db_run_container_to_uint32_array(DB::WriteBuffer & db_buf, const run_container_t * cont, UInt32 base) const
-    {
-        UInt32 outpos = 0;
-        for (Int32 i = 0; i < cont->n_runs; ++i)
-        {
-            UInt32 run_start = base + cont->runs[i].value;
-            UInt16 le = cont->runs[i].length;
-            for (Int32 j = 0; j <= le; ++j)
-            {
-                UInt32 val = run_start + j;
-                writePODBinary(val, db_buf);
-                outpos++;
-            }
-        }
-        return outpos;
-    }
-};
-
-template <typename T>
-struct AggregateFunctionGroupBitmapData
-{
-    bool doneFirst = false;
-    RoaringBitmapWithSmallSet<T, 32> rbs;
-    static const char * name() { return "groupBitmap"; }
-};
-
-
-}
+#pragma once
+
+#include <algorithm>
+#include <IO/ReadHelpers.h>
+#include <IO/WriteHelpers.h>
+#include <boost/noncopyable.hpp>
+#include <Common/HashTable/SmallTable.h>
+#include <Common/PODArray.h>
+
+// Include this header last, because it is an auto-generated dump of questionable
+// garbage that breaks the build (e.g. it changes _POSIX_C_SOURCE).
+// TODO: find out what it is. On github, they have proper interface headers like
+// this one: https://github.com/RoaringBitmap/CRoaring/blob/master/include/roaring/roaring.h
+#include <roaring/roaring.h>
+#include "roaring.hh"
+#include <roaring64map.hh>
+
+namespace DB
+{
+/**
+  * For a small number of values - an array of fixed size "on the stack".
+  * For large, roaring_bitmap_t is allocated.
+  * For a description of the roaring_bitmap_t, see: https://github.com/RoaringBitmap/CRoaring
+  */
+template <typename T, UInt8 small_set_size>
+class RoaringBitmapWithSmallSet : private boost::noncopyable
+{
+private:
+    using Small = SmallSet<T, small_set_size>;
+    using ValueBuffer = std::vector<T>;
+    bool bUInt64 = sizeof(T) >= 8;
+
+    Small small;
+    roaring_bitmap_t * rb = nullptr;
+    std::shared_ptr<Roaring64Map> rb64 = nullptr;
+
+    void toLarge()
+    {
+        if (isUInt64())
+            rb64 = std::make_shared<Roaring64Map>();
+        else
+            rb = roaring_bitmap_create();
+
+        for (const auto & x : small)
+        {
+            if (isUInt64())
+                rb64->add(static_cast<UInt64>(x.getValue()));
+            else
+                roaring_bitmap_add(rb, x.getValue());
+        }
+    }
+
+public:
+    bool isLarge() const { return rb != nullptr || rb64 != nullptr; }
+
+    bool isSmall() const { return rb == nullptr && rb64 == nullptr; }
+
+    bool isUInt64() const { return bUInt64; }
+
+    ~RoaringBitmapWithSmallSet()
+    {
+        if (isLarge() && !isUInt64())
+            roaring_bitmap_free(rb);
+    }
+
+    void add(T value)
+    {
+        if (isSmall())
+        {
+            if (small.find(value) == small.end())
+            {
+                if (!small.full())
+                    small.insert(value);
+                else
+                {
+                    toLarge();
+                    if (isUInt64())
+                        rb64->add(static_cast<UInt64>(value));
+                    else
+                        roaring_bitmap_add(rb, value);
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+                rb64->add(static_cast<UInt64>(value));
+            else
+                roaring_bitmap_add(rb, value);
+        }
+    }
+
+    UInt64 size() const
+    {
+        if (isSmall())
+        {
+            return small.size();
+        }
+        else
+        {
+            if (isUInt64())
+                return rb64->cardinality();
+            else
+                return roaring_bitmap_get_cardinality(rb);
+        }
+    }
+
+    void merge(const RoaringBitmapWithSmallSet & r1)
+    {
+        if (r1.isLarge())
+        {
+            if (isSmall())
+                toLarge();
+
+            if (isUInt64())
+                *rb64 |= *r1.rb64;
+            else
+                roaring_bitmap_or_inplace(rb, r1.rb);
+        }
+        else
+        {
+            for (const auto & x : r1.small)
+                add(x.getValue());
+        }
+    }
+
+    void read(DB::ReadBuffer & in)
+    {
+        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
+        UInt8 containerType = 0;
+        readBinary(containerType, in);
+
+        if (0 == containerType)
+        {
+            small.read(in);
+        }
+        else if (1 == containerType)
+        {
+            std::string s;
+            readStringBinary(s, in);
+            rb = roaring_bitmap_portable_deserialize(s.c_str());
+            // It has been persisted in the bitmap and does not need to merge from small
+            //            for (const auto & x : small) // merge from small
+            //                roaring_bitmap_add(rb, x.getValue());
+        }
+        else
+        {
+            std::string s;
+            readStringBinary(s, in);
+            rb64 = std::make_shared<Roaring64Map>(Roaring64Map::read(s.c_str()));
+        }
+    }
+
+    void write(DB::WriteBuffer & out) const
+    {
+        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
+        UInt8 containerType = 0;
+        if (isLarge())
+        {
+            if (isUInt64())
+                containerType = 2;
+            else
+                containerType = 1;
+        }
+
+        writeBinary(containerType, out);
+
+        if (isLarge())
+        {
+            if (isUInt64())
+            {
+                uint32_t expectedsize = rb64->getSizeInBytes();
+                std::string s(expectedsize, 0);
+                rb64->write(const_cast<char *>(s.data()));
+                writeStringBinary(s, out);
+            }
+            else
+            {
+                uint32_t expectedsize = roaring_bitmap_portable_size_in_bytes(rb);
+                std::string s(expectedsize, 0);
+                roaring_bitmap_portable_serialize(rb, const_cast<char *>(s.data()));
+                writeStringBinary(s, out);
+            }
+        }
+        else
+            small.write(out);
+    }
+
+    roaring_bitmap_t * getRb() const { return rb; }
+
+    std::shared_ptr<Roaring64Map> getRb64() const { return rb64; }
+
+    Small & getSmall() const { return small; }
+
+    /**
+     * Get a new roaring_bitmap_t from elements of small
+     */
+    roaring_bitmap_t * getNewRbFromSmall() const
+    {
+        roaring_bitmap_t * smallRb = roaring_bitmap_create();
+        for (const auto & x : small)
+            roaring_bitmap_add(smallRb, x.getValue());
+        return smallRb;
+    }
+
+    /**
+     * Get a new Roaring64Map from elements of small
+     */
+    std::shared_ptr<Roaring64Map> getNewRb64FromSmall() const
+    {
+        std::shared_ptr<Roaring64Map> smallRb64 = std::make_shared<Roaring64Map>();
+        for (const auto & x : small)
+            smallRb64->add(static_cast<UInt64>(x.getValue()));
+        return smallRb64;
+    }
+
+    /**
+     * Computes the intersection between two bitmaps
+     */
+    void rb_and(const RoaringBitmapWithSmallSet & r1)
+    {
+        ValueBuffer buffer;
+        if (isSmall() && r1.isSmall())
+        {
+            // intersect
+            for (const auto & x : small)
+                if (r1.small.find(x.getValue()) != r1.small.end())
+                    buffer.push_back(x.getValue());
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else if (isSmall() && r1.isLarge())
+        {
+            for (const auto & x : small)
+            {
+                if (isUInt64())
+                {
+                    if (rb64->contains(static_cast<UInt64>(x.getValue())))
+                        buffer.push_back(x.getValue());
+                }
+                else
+                {
+                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
+                        buffer.push_back(x.getValue());
+                }
+            }
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+                *rb64 &= *newRb64;
+            }
+            else
+            {
+                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+                roaring_bitmap_and_inplace(rb, rb1);
+                if (r1.isSmall())
+                    roaring_bitmap_free(rb1);
+            }
+        }
+    }
+
+    /**
+     * Computes the union between two bitmaps.
+     */
+    void rb_or(const RoaringBitmapWithSmallSet & r1) { merge(r1); }
+
+    /**
+     * Computes the symmetric difference (xor) between two bitmaps.
+     */
+    void rb_xor(const RoaringBitmapWithSmallSet & r1)
+    {
+        if (isSmall())
+            toLarge();
+
+        if (isUInt64())
+        {
+            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+            *rb64 ^= *newRb64;
+        }
+        else
+        {
+            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+            roaring_bitmap_xor_inplace(rb, rb1);
+            if (r1.isSmall())
+                roaring_bitmap_free(rb1);
+        }
+    }
+
+    /**
+     * Computes the difference (andnot) between two bitmaps
+     */
+    void rb_andnot(const RoaringBitmapWithSmallSet & r1)
+    {
+        ValueBuffer buffer;
+        if (isSmall() && r1.isSmall())
+        {
+            // subtract
+            for (const auto & x : small)
+                if (r1.small.find(x.getValue()) == r1.small.end())
+                    buffer.push_back(x.getValue());
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else if (isSmall() && r1.isLarge())
+        {
+            for (const auto & x : small)
+            {
+                if (isUInt64())
+                {
+                    if (!rb64->contains(static_cast<UInt64>(x.getValue())))
+                        buffer.push_back(x.getValue());
+                }
+                else
+                {
+                    if (!roaring_bitmap_contains(r1.rb, x.getValue()))
+                        buffer.push_back(x.getValue());
+                }
+            }
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+                *rb64 -= *newRb64;
+            }
+            else
+            {
+                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+                roaring_bitmap_andnot_inplace(rb, rb1);
+                if (r1.isSmall())
+                    roaring_bitmap_free(rb1);
+            }
+        }
+    }
+
+    /**
+     * Computes the cardinality of the intersection between two bitmaps.
+     */
+    UInt64 rb_and_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 retSize = 0;
+        if (isSmall() && r1.isSmall())
+        {
+            for (const auto & x : small)
+                if (r1.small.find(x.getValue()) != r1.small.end())
+                    ++retSize;
+        }
+        else if (isSmall() && r1.isLarge())
+        {
+            for (const auto & x : small)
+            {
+                if (isUInt64())
+                {
+                    if (rb64->contains(static_cast<UInt64>(x.getValue())))
+                        ++retSize;
+                }
+                else
+                {
+                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
+                        ++retSize;
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+                retSize = (*rb64 & *newRb64).cardinality();
+            }
+            else
+            {
+                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+                retSize = roaring_bitmap_and_cardinality(rb, rb1);
+                if (r1.isSmall())
+                    roaring_bitmap_free(rb1);
+            }
+        }
+        return retSize;
+    }
+
+    /**
+     * Computes the cardinality of the union between two bitmaps.
+    */
+    UInt64 rb_or_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 c1 = size();
+        UInt64 c2 = r1.size();
+        UInt64 inter = rb_and_cardinality(r1);
+        return c1 + c2 - inter;
+    }
+
+    /**
+     * Computes the cardinality of the symmetric difference (andnot) between two bitmaps.
+    */
+    UInt64 rb_xor_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 c1 = size();
+        UInt64 c2 = r1.size();
+        UInt64 inter = rb_and_cardinality(r1);
+        return c1 + c2 - 2 * inter;
+    }
+
+    /**
+     * Computes the cardinality of the difference (andnot) between two bitmaps.
+     */
+    UInt64 rb_andnot_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 c1 = size();
+        UInt64 inter = rb_and_cardinality(r1);
+        return c1 - inter;
+    }
+
+    /**
+     * Return 1 if the two bitmaps contain the same elements.
+     */
+    UInt8 rb_equals(const RoaringBitmapWithSmallSet & r1)
+    {
+        if (isSmall())
+            toLarge();
+
+        UInt8 is_true = 0;
+        if (isUInt64())
+        {
+            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
+            is_true = *rb64 == *newRb64;
+        }
+        else
+        {
+            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
+            is_true = roaring_bitmap_equals(rb, rb1);
+            if (r1.isSmall())
+                roaring_bitmap_free(rb1);
+        }
+
+        return is_true;
+    }
+
+    /**
+     * Check whether two bitmaps intersect.
+     * Intersection with an empty set is always 0 (consistent with hasAny).
+     */
+    UInt8 rb_intersect(const RoaringBitmapWithSmallSet & r1) const
+    {
+        if (isSmall())
+        {
+            if (r1.isSmall())
+            {
+                for (const auto & x : r1.small)
+                    if (small.find(x.getValue()) != small.end())
+                        return 1;
+            }
+            else
+            {
+                for (const auto & x : small)
+                {
+                    if (isUInt64())
+                    {
+                        if (r1.rb64->contains(static_cast<UInt64>(x.getValue())))
+                            return 1;
+                    }
+                    else
+                    {
+                        if (roaring_bitmap_contains(r1.rb, x.getValue()))
+                            return 1;
+                    }
+                }
+            }
+        }
+        else if (r1.isSmall())
+        {
+            for (const auto & x : r1.small)
+            {
+                if (isUInt64())
+                {
+                    if (rb64->contains(static_cast<UInt64>(x.getValue())))
+                        return 1;
+                }
+                else
+                {
+                    if (roaring_bitmap_contains(rb, x.getValue()))
+                        return 1;
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                if ((*rb64 & *r1.rb64).cardinality() > 0)
+                    return 1;
+            }
+            else
+            {
+                if (roaring_bitmap_intersect(rb, r1.rb))
+                    return 1;
+            }
+        }
+
+        return 0;
+    }
+
+    /**
+     * Check whether the argument is the subset of this set.
+     * Empty set is a subset of any other set (consistent with hasAll).
+     */
+    UInt8 rb_is_subset(const RoaringBitmapWithSmallSet & r1) const
+    {
+        if (isSmall())
+        {
+            if (r1.isSmall())
+            {
+                for (const auto & x : r1.small)
+                    if (small.find(x.getValue()) == small.end())
+                        return 0;
+            }
+            else
+            {
+                UInt64 r1_size = r1.size();
+
+                if (r1_size > small.size())
+                    return 0; // A bigger set can not be a subset of ours.
+
+                // This is a rare case with a small number of elements on
+                // both sides: r1 was promoted to large for some reason and
+                // it is still not larger than our small set.
+                // If r1 is our subset then our size must be equal to
+                // r1_size + number of not found elements, if this sum becomes
+                // greater then r1 is not a subset.
+                for (const auto & x : small)
+                {
+                    if (isUInt64())
+                    {
+                        if (!r1.rb64->contains(static_cast<UInt64>(x.getValue())) && ++r1_size > small.size())
+                            return 0;
+                    }
+                    else
+                    {
+                        if (!roaring_bitmap_contains(r1.rb, x.getValue()) && ++r1_size > small.size())
+                            return 0;
+                    }
+                }
+            }
+        }
+        else if (r1.isSmall())
+        {
+            for (const auto & x : r1.small)
+            {
+                if (isUInt64())
+                {
+                    if (!rb64->contains(static_cast<UInt64>(x.getValue())))
+                        return 0;
+                }
+                else
+                {
+                    if (!roaring_bitmap_contains(rb, x.getValue()))
+                        return 0;
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                if (!r1.rb64->isSubset(*rb64))
+                    return 0;
+            }
+            else
+            {
+                if (!roaring_bitmap_is_subset(r1.rb, rb))
+                    return 0;
+            }
+        }
+        return 1;
+    }
+
+    /**
+     * Check whether this bitmap contains the argument.
+     */
+    UInt8 rb_contains(const UInt64 x) const
+    {
+        if (isSmall())
+        {
+            return small.find(x) != small.end();
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                return rb64->contains(x);
+            }
+            else
+            {
+                return roaring_bitmap_contains(rb, x);
+            }
+        }
+    }
+
+    /**
+     * Remove value
+     */
+    void rb_remove(UInt64 offsetid)
+    {
+        if (isSmall())
+            toLarge();
+
+        if (isUInt64())
+            rb64->remove(offsetid);
+        else
+            roaring_bitmap_remove(rb, offsetid);
+    }
+
+    /**
+     * compute (in place) the negation of the roaring bitmap within a specified
+     * interval: [range_start, range_end). The number of negated values is
+     * range_end - range_start.
+     * Areas outside the range are passed through unchanged.
+     */
+    void rb_flip(UInt64 offsetstart, UInt64 offsetend)
+    {
+        if (isSmall())
+            toLarge();
+        if (isUInt64())
+            rb64->flip(offsetstart, offsetend);
+        else
+            roaring_bitmap_flip_inplace(rb, offsetstart, offsetend);
+    }
+
+    /**
+     * returns the number of integers that are smaller or equal to offsetid.
+     */
+    UInt64 rb_rank(UInt64 offsetid)
+    {
+        if (isSmall())
+            toLarge();
+
+        if (isUInt64())
+            return rb64->rank(offsetid);
+        else
+            return roaring_bitmap_rank(rb, offsetid);
+    }
+
+    /**
+     * Convert elements to integer array, return number of elements
+     */
+    template <typename Element>
+    UInt64 rb_to_array(PaddedPODArray<Element> & res_data) const
+    {
+        UInt64 count = 0;
+        if (isSmall())
+        {
+            for (const auto & x : small)
+            {
+                res_data.emplace_back(x.getValue());
+                count++;
+            }
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
+                {
+                    res_data.emplace_back(*iterator);
+                    count++;
+                }
+            }
+            else
+            {
+                roaring_uint32_iterator_t iterator;
+                roaring_init_iterator(rb, &iterator);
+                while (iterator.has_value)
+                {
+                    res_data.emplace_back(iterator.current_value);
+                    roaring_advance_uint32_iterator(&iterator);
+                    count++;
+                }
+            }
+        }
+        return count;
+    }
+
+    /**
+     * Return new set with specified range (not include the range_end)
+     */
+    UInt64 rb_range(UInt64 range_start, UInt64 range_end, RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 count = 0;
+        if (range_start >= range_end)
+            return count;
+        if (isSmall())
+        {
+            for (const auto & x : small)
+            {
+                T val = x.getValue();
+                if (UInt32(val) >= range_start && UInt32(val) < range_end)
+                {
+                    r1.add(val);
+                    ++count;
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
+                {
+                    if (*iterator < range_start)
+                        continue;
+
+                    if (*iterator < range_end)
+                    {
+                        r1.add(*iterator);
+                        ++count;
+                    }
+                    else
+                        break;
+                }
+            }
+            else
+            {
+                roaring_uint32_iterator_t iterator;
+                roaring_init_iterator(rb, &iterator);
+                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
+                while (iterator.has_value && UInt32(iterator.current_value) < range_end)
+                {
+                    r1.add(iterator.current_value);
+                    roaring_advance_uint32_iterator(&iterator);
+                    ++count;
+                }
+            }
+        }
+        return count;
+    }
+
+    /**
+     * Return new set of the smallest `limit` values in set which is no less than `range_start`.
+     */
+    UInt64 rb_limit(UInt64 range_start, UInt64 limit, RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 count = 0;
+        if (isSmall())
+        {
+            std::vector<T> answer;
+            for (const auto & x : small)
+            {
+                T val = x.getValue();
+                if (UInt32(val) >= range_start)
+                {
+                    answer.push_back(val);
+                }
+            }
+            sort(answer.begin(), answer.end());
+            if (limit > answer.size())
+                limit = answer.size();
+            for (size_t i = 0; i < limit; ++i)
+                r1.add(answer[i]);
+            count = UInt64(limit);
+        }
+        else
+        {
+            if (isUInt64())
+            {
+                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
+                {
+                    if (*iterator < range_start)
+                        continue;
+
+                    if (count < limit)
+                    {
+                        r1.add(*iterator);
+                        ++count;
+                    }
+                    else
+                        break;
+                }
+            }
+            else
+            {
+                roaring_uint32_iterator_t iterator;
+                roaring_init_iterator(rb, &iterator);
+                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
+                while (UInt32(count) < limit && iterator.has_value)
+                {
+                    r1.add(iterator.current_value);
+                    roaring_advance_uint32_iterator(&iterator);
+                    ++count;
+                }
+            }
+        }
+        return count;
+    }
+
+    UInt64 rb_min() const
+    {
+        UInt64 min_val = UINT32_MAX;
+        if (isSmall())
+        {
+            for (const auto & x : small)
+            {
+                T val = x.getValue();
+                if (UInt64(val) < min_val)
+                {
+                    min_val = UInt64(val);
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+                min_val = rb64->minimum();
+            else
+                min_val = UInt64(roaring_bitmap_minimum(rb));
+        }
+        return min_val;
+    }
+
+    UInt64 rb_max() const
+    {
+        UInt64 max_val = 0;
+        if (isSmall())
+        {
+            for (const auto & x : small)
+            {
+                T val = x.getValue();
+                if (UInt64(val) > max_val)
+                {
+                    max_val = UInt64(val);
+                }
+            }
+        }
+        else
+        {
+            if (isUInt64())
+                max_val = rb64->maximum();
+            else
+                max_val = UInt64(roaring_bitmap_maximum(rb));
+        }
+        return max_val;
+    }
+
+    /**
+     * Replace value
+     */
+    void rb_replace(const UInt32 * from_vals, const UInt32 * to_vals, size_t num)
+    {
+        if (isSmall())
+            toLarge();
+        for (size_t i = 0; i < num; ++i)
+        {
+            if (from_vals[i] == to_vals[i])
+                continue;
+
+            if (isUInt64())
+            {
+                bool changed = rb64->removeChecked(from_vals[i]);
+                if (changed)
+                    rb64->add(to_vals[i]);
+            }
+            else
+            {
+                bool changed = roaring_bitmap_remove_checked(rb, from_vals[i]);
+                if (changed)
+                    roaring_bitmap_add(rb, to_vals[i]);
+            }
+        }
+    }
+
+private:
+ 
+};
+
+template <typename T>
+struct AggregateFunctionGroupBitmapData
+{
+    bool doneFirst = false;
+    RoaringBitmapWithSmallSet<T, 32> rbs;
+    static const char * name() { return "groupBitmap"; }
+};
+
+
+}

From c19ce7dad30c0223e214a0c97589a2adcc83e0c8 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Wed, 9 Dec 2020 12:44:25 +0800
Subject: [PATCH 397/504] rewrite with better code

---
 .../AggregateFunctionGroupBitmap.cpp          |   79 +-
 .../AggregateFunctionGroupBitmap.h            |   79 +-
 .../AggregateFunctionGroupBitmapData.h        | 1550 +++++++----------
 3 files changed, 704 insertions(+), 1004 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmap.cpp b/src/AggregateFunctions/AggregateFunctionGroupBitmap.cpp
index 25df0cb27bd..bf1d0af73ff 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmap.cpp
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmap.cpp
@@ -1,11 +1,10 @@
 #include <AggregateFunctions/AggregateFunctionFactory.h>
-#include <AggregateFunctions/Helpers.h>
 #include <AggregateFunctions/FactoryHelpers.h>
+#include <AggregateFunctions/Helpers.h>
 #include <DataTypes/DataTypeAggregateFunction.h>
 #include "registerAggregateFunctions.h"
 
-// TODO include this last because of a broken roaring header. See the comment
-// inside.
+// TODO include this last because of a broken roaring header. See the comment inside.
 #include <AggregateFunctions/AggregateFunctionGroupBitmap.h>
 
 namespace DB
@@ -17,46 +16,54 @@ namespace ErrorCodes
 
 namespace
 {
+    template <template <typename> class Data>
+    AggregateFunctionPtr createAggregateFunctionBitmap(const std::string & name, const DataTypes & argument_types, const Array & parameters)
+    {
+        assertNoParameters(name, parameters);
+        assertUnary(name, argument_types);
 
-template <template <typename> class Data>
-AggregateFunctionPtr createAggregateFunctionBitmap(const std::string & name, const DataTypes & argument_types, const Array & parameters)
-{
-    assertNoParameters(name, parameters);
-    assertUnary(name, argument_types);
+        if (!argument_types[0]->canBeUsedInBitOperations())
+            throw Exception(
+                "The type " + argument_types[0]->getName() + " of argument for aggregate function " + name
+                    + " is illegal, because it cannot be used in Bitmap operations",
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
-    if (!argument_types[0]->canBeUsedInBitOperations())
-        throw Exception("The type " + argument_types[0]->getName() + " of argument for aggregate function " + name
-            + " is illegal, because it cannot be used in Bitmap operations",
-            ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        AggregateFunctionPtr res(createWithUnsignedIntegerType<AggregateFunctionBitmap, Data>(*argument_types[0], argument_types[0]));
 
-    AggregateFunctionPtr res(createWithUnsignedIntegerType<AggregateFunctionBitmap, Data>(*argument_types[0], argument_types[0]));
+        if (!res)
+            throw Exception(
+                "Illegal type " + argument_types[0]->getName() + " of argument for aggregate function " + name,
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
-    if (!res)
-        throw Exception("Illegal type " + argument_types[0]->getName() + " of argument for aggregate function " + name, ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+        return res;
+    }
 
-    return res;
-}
+    // Additional aggregate functions to manipulate bitmaps.
+    template <template <typename, typename> class AggregateFunctionTemplate>
+    AggregateFunctionPtr
+    createAggregateFunctionBitmapL2(const std::string & name, const DataTypes & argument_types, const Array & parameters)
+    {
+        assertNoParameters(name, parameters);
+        assertUnary(name, argument_types);
+        DataTypePtr argument_type_ptr = argument_types[0];
+        WhichDataType which(*argument_type_ptr);
+        if (which.idx != TypeIndex::AggregateFunction)
+            throw Exception(
+                "Illegal type " + argument_types[0]->getName() + " of argument for aggregate function " + name,
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
-template <template <typename, typename> class AggregateFunctionTemplate>
-AggregateFunctionPtr createAggregateFunctionBitmapL2(const std::string & name, const DataTypes & argument_types, const Array & parameters)
-{
-    assertNoParameters(name, parameters);
-    assertUnary(name, argument_types);
-    DataTypePtr argument_type_ptr = argument_types[0];
-    WhichDataType which(*argument_type_ptr);
-    if (which.idx != TypeIndex::AggregateFunction)
-        throw Exception("Illegal type " + argument_types[0]->getName() + " of argument for aggregate function " + name, ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-
-    const DataTypeAggregateFunction& datatype_aggfunc = dynamic_cast<const DataTypeAggregateFunction&>(*argument_type_ptr);
-    AggregateFunctionPtr aggfunc = datatype_aggfunc.getFunction();
-    argument_type_ptr = aggfunc->getArgumentTypes()[0];
-    AggregateFunctionPtr res(createWithUnsignedIntegerType<AggregateFunctionTemplate, AggregateFunctionGroupBitmapData>(*argument_type_ptr, argument_type_ptr));
-    if (!res)
-        throw Exception("Illegal type " + argument_types[0]->getName() + " of argument for aggregate function " + name, ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
-
-    return res;
-}
+        const DataTypeAggregateFunction & datatype_aggfunc = dynamic_cast<const DataTypeAggregateFunction &>(*argument_type_ptr);
+        AggregateFunctionPtr aggfunc = datatype_aggfunc.getFunction();
+        argument_type_ptr = aggfunc->getArgumentTypes()[0];
+        AggregateFunctionPtr res(createWithUnsignedIntegerType<AggregateFunctionTemplate, AggregateFunctionGroupBitmapData>(
+            *argument_type_ptr, argument_type_ptr));
+        if (!res)
+            throw Exception(
+                "Illegal type " + argument_types[0]->getName() + " of argument for aggregate function " + name,
+                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
 
+        return res;
+    }
 }
 
 void registerAggregateFunctionsBitmap(AggregateFunctionFactory & factory)
diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmap.h b/src/AggregateFunctions/AggregateFunctionGroupBitmap.h
index e974cb753d3..315132cf05c 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmap.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmap.h
@@ -1,31 +1,26 @@
 #pragma once
 
-#include <Columns/ColumnVector.h>
-#include <Common/assert_cast.h>
 #include <AggregateFunctions/IAggregateFunction.h>
-#include <DataTypes/DataTypesNumber.h>
 #include <Columns/ColumnAggregateFunction.h>
+#include <Columns/ColumnVector.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Common/assert_cast.h>
 
 // TODO include this last because of a broken roaring header. See the comment inside.
 #include <AggregateFunctions/AggregateFunctionGroupBitmapData.h>
 
 namespace DB
 {
-
 /// Counts bitmap operation on numbers.
 template <typename T, typename Data>
 class AggregateFunctionBitmap final : public IAggregateFunctionDataHelper<Data, AggregateFunctionBitmap<T, Data>>
 {
 public:
-    AggregateFunctionBitmap(const DataTypePtr & type)
-            : IAggregateFunctionDataHelper<Data, AggregateFunctionBitmap<T, Data>>({type}, {}) {}
+    AggregateFunctionBitmap(const DataTypePtr & type) : IAggregateFunctionDataHelper<Data, AggregateFunctionBitmap<T, Data>>({type}, {}) { }
 
     String getName() const override { return Data::name(); }
 
-    DataTypePtr getReturnType() const override
-    {
-        return std::make_shared<DataTypeNumber<T>>();
-    }
+    DataTypePtr getReturnType() const override { return std::make_shared<DataTypeNumber<T>>(); }
 
     void add(AggregateDataPtr place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
@@ -37,15 +32,9 @@ public:
         this->data(place).rbs.merge(this->data(rhs).rbs);
     }
 
-    void serialize(ConstAggregateDataPtr place, WriteBuffer & buf) const override
-    {
-        this->data(place).rbs.write(buf);
-    }
+    void serialize(ConstAggregateDataPtr place, WriteBuffer & buf) const override { this->data(place).rbs.write(buf); }
 
-    void deserialize(AggregateDataPtr place, ReadBuffer & buf, Arena *) const override
-    {
-        this->data(place).rbs.read(buf);
-    }
+    void deserialize(AggregateDataPtr place, ReadBuffer & buf, Arena *) const override { this->data(place).rbs.read(buf); }
 
     void insertResultInto(AggregateDataPtr place, IColumn & to, Arena *) const override
     {
@@ -59,23 +48,22 @@ class AggregateFunctionBitmapL2 final : public IAggregateFunctionDataHelper<Data
 {
 public:
     AggregateFunctionBitmapL2(const DataTypePtr & type)
-            : IAggregateFunctionDataHelper<Data, AggregateFunctionBitmapL2<T, Data, Policy>>({type}, {}){}
+        : IAggregateFunctionDataHelper<Data, AggregateFunctionBitmapL2<T, Data, Policy>>({type}, {})
+    {
+    }
 
     String getName() const override { return Data::name(); }
 
-    DataTypePtr getReturnType() const override
-    {
-        return std::make_shared<DataTypeNumber<T>>();
-    }
+    DataTypePtr getReturnType() const override { return std::make_shared<DataTypeNumber<T>>(); }
 
     void add(AggregateDataPtr place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         Data & data_lhs = this->data(place);
         const Data & data_rhs = this->data(assert_cast<const ColumnAggregateFunction &>(*columns[0]).getData()[row_num]);
-        if (!data_lhs.doneFirst)
+        if (!data_lhs.init)
         {
-            data_lhs.doneFirst = true;
-            data_lhs.rbs.rb_or(data_rhs.rbs);
+            data_lhs.init = true;
+            data_lhs.rbs.merge(data_rhs.rbs);
         }
         else
         {
@@ -88,13 +76,13 @@ public:
         Data & data_lhs = this->data(place);
         const Data & data_rhs = this->data(rhs);
 
-        if (!data_rhs.doneFirst)
+        if (!data_rhs.init)
             return;
 
-        if (!data_lhs.doneFirst)
+        if (!data_lhs.init)
         {
-            data_lhs.doneFirst = true;
-            data_lhs.rbs.rb_or(data_rhs.rbs);
+            data_lhs.init = true;
+            data_lhs.rbs.merge(data_rhs.rbs);
         }
         else
         {
@@ -102,15 +90,9 @@ public:
         }
     }
 
-    void serialize(ConstAggregateDataPtr place, WriteBuffer & buf) const override
-    {
-        this->data(place).rbs.write(buf);
-    }
+    void serialize(ConstAggregateDataPtr place, WriteBuffer & buf) const override { this->data(place).rbs.write(buf); }
 
-    void deserialize(AggregateDataPtr place, ReadBuffer & buf, Arena *) const override
-    {
-        this->data(place).rbs.read(buf);
-    }
+    void deserialize(AggregateDataPtr place, ReadBuffer & buf, Arena *) const override { this->data(place).rbs.read(buf); }
 
     void insertResultInto(AggregateDataPtr place, IColumn & to, Arena *) const override
     {
@@ -122,39 +104,30 @@ template <typename Data>
 class BitmapAndPolicy
 {
 public:
-    static void apply(Data& lhs, const Data& rhs)
-    {
-        lhs.rbs.rb_and(rhs.rbs);
-    }
+    static void apply(Data & lhs, const Data & rhs) { lhs.rbs.rb_and(rhs.rbs); }
 };
 
 template <typename Data>
 class BitmapOrPolicy
 {
 public:
-    static void apply(Data& lhs, const Data& rhs)
-    {
-        lhs.rbs.rb_or(rhs.rbs);
-    }
+    static void apply(Data & lhs, const Data & rhs) { lhs.rbs.rb_or(rhs.rbs); }
 };
 
 template <typename Data>
 class BitmapXorPolicy
 {
 public:
-    static void apply(Data& lhs, const Data& rhs)
-    {
-        lhs.rbs.rb_xor(rhs.rbs);
-    }
+    static void apply(Data & lhs, const Data & rhs) { lhs.rbs.rb_xor(rhs.rbs); }
 };
 
 template <typename T, typename Data>
-using AggregateFunctionBitmapL2And = AggregateFunctionBitmapL2<T, Data, BitmapAndPolicy<Data> >;
+using AggregateFunctionBitmapL2And = AggregateFunctionBitmapL2<T, Data, BitmapAndPolicy<Data>>;
 
 template <typename T, typename Data>
-using AggregateFunctionBitmapL2Or = AggregateFunctionBitmapL2<T, Data, BitmapOrPolicy<Data> >;
+using AggregateFunctionBitmapL2Or = AggregateFunctionBitmapL2<T, Data, BitmapOrPolicy<Data>>;
 
 template <typename T, typename Data>
-using AggregateFunctionBitmapL2Xor = AggregateFunctionBitmapL2<T, Data, BitmapXorPolicy<Data> >;
+using AggregateFunctionBitmapL2Xor = AggregateFunctionBitmapL2<T, Data, BitmapXorPolicy<Data>>;
 
 }
diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
index 000edd64a0f..b26e47f0b63 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
@@ -1,915 +1,635 @@
-#pragma once
-
-#include <algorithm>
-#include <IO/ReadHelpers.h>
-#include <IO/WriteHelpers.h>
-#include <boost/noncopyable.hpp>
-#include <Common/HashTable/SmallTable.h>
-#include <Common/PODArray.h>
-
-// Include this header last, because it is an auto-generated dump of questionable
-// garbage that breaks the build (e.g. it changes _POSIX_C_SOURCE).
-// TODO: find out what it is. On github, they have proper interface headers like
-// this one: https://github.com/RoaringBitmap/CRoaring/blob/master/include/roaring/roaring.h
-#include <roaring/roaring.h>
-#include "roaring.hh"
-#include <roaring64map.hh>
-
-namespace DB
-{
-/**
-  * For a small number of values - an array of fixed size "on the stack".
-  * For large, roaring_bitmap_t is allocated.
-  * For a description of the roaring_bitmap_t, see: https://github.com/RoaringBitmap/CRoaring
-  */
-template <typename T, UInt8 small_set_size>
-class RoaringBitmapWithSmallSet : private boost::noncopyable
-{
-private:
-    using Small = SmallSet<T, small_set_size>;
-    using ValueBuffer = std::vector<T>;
-    bool bUInt64 = sizeof(T) >= 8;
-
-    Small small;
-    roaring_bitmap_t * rb = nullptr;
-    std::shared_ptr<Roaring64Map> rb64 = nullptr;
-
-    void toLarge()
-    {
-        if (isUInt64())
-            rb64 = std::make_shared<Roaring64Map>();
-        else
-            rb = roaring_bitmap_create();
-
-        for (const auto & x : small)
-        {
-            if (isUInt64())
-                rb64->add(static_cast<UInt64>(x.getValue()));
-            else
-                roaring_bitmap_add(rb, x.getValue());
-        }
-    }
-
-public:
-    bool isLarge() const { return rb != nullptr || rb64 != nullptr; }
-
-    bool isSmall() const { return rb == nullptr && rb64 == nullptr; }
-
-    bool isUInt64() const { return bUInt64; }
-
-    ~RoaringBitmapWithSmallSet()
-    {
-        if (isLarge() && !isUInt64())
-            roaring_bitmap_free(rb);
-    }
-
-    void add(T value)
-    {
-        if (isSmall())
-        {
-            if (small.find(value) == small.end())
-            {
-                if (!small.full())
-                    small.insert(value);
-                else
-                {
-                    toLarge();
-                    if (isUInt64())
-                        rb64->add(static_cast<UInt64>(value));
-                    else
-                        roaring_bitmap_add(rb, value);
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-                rb64->add(static_cast<UInt64>(value));
-            else
-                roaring_bitmap_add(rb, value);
-        }
-    }
-
-    UInt64 size() const
-    {
-        if (isSmall())
-        {
-            return small.size();
-        }
-        else
-        {
-            if (isUInt64())
-                return rb64->cardinality();
-            else
-                return roaring_bitmap_get_cardinality(rb);
-        }
-    }
-
-    void merge(const RoaringBitmapWithSmallSet & r1)
-    {
-        if (r1.isLarge())
-        {
-            if (isSmall())
-                toLarge();
-
-            if (isUInt64())
-                *rb64 |= *r1.rb64;
-            else
-                roaring_bitmap_or_inplace(rb, r1.rb);
-        }
-        else
-        {
-            for (const auto & x : r1.small)
-                add(x.getValue());
-        }
-    }
-
-    void read(DB::ReadBuffer & in)
-    {
-        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
-        UInt8 containerType = 0;
-        readBinary(containerType, in);
-
-        if (0 == containerType)
-        {
-            small.read(in);
-        }
-        else if (1 == containerType)
-        {
-            std::string s;
-            readStringBinary(s, in);
-            rb = roaring_bitmap_portable_deserialize(s.c_str());
-            // It has been persisted in the bitmap and does not need to merge from small
-            //            for (const auto & x : small) // merge from small
-            //                roaring_bitmap_add(rb, x.getValue());
-        }
-        else
-        {
-            std::string s;
-            readStringBinary(s, in);
-            rb64 = std::make_shared<Roaring64Map>(Roaring64Map::read(s.c_str()));
-        }
-    }
-
-    void write(DB::WriteBuffer & out) const
-    {
-        // Container type:  0: SmallSet,  1: RoaringBitmap32,  2: RoaringBitmap64
-        UInt8 containerType = 0;
-        if (isLarge())
-        {
-            if (isUInt64())
-                containerType = 2;
-            else
-                containerType = 1;
-        }
-
-        writeBinary(containerType, out);
-
-        if (isLarge())
-        {
-            if (isUInt64())
-            {
-                uint32_t expectedsize = rb64->getSizeInBytes();
-                std::string s(expectedsize, 0);
-                rb64->write(const_cast<char *>(s.data()));
-                writeStringBinary(s, out);
-            }
-            else
-            {
-                uint32_t expectedsize = roaring_bitmap_portable_size_in_bytes(rb);
-                std::string s(expectedsize, 0);
-                roaring_bitmap_portable_serialize(rb, const_cast<char *>(s.data()));
-                writeStringBinary(s, out);
-            }
-        }
-        else
-            small.write(out);
-    }
-
-    roaring_bitmap_t * getRb() const { return rb; }
-
-    std::shared_ptr<Roaring64Map> getRb64() const { return rb64; }
-
-    Small & getSmall() const { return small; }
-
-    /**
-     * Get a new roaring_bitmap_t from elements of small
-     */
-    roaring_bitmap_t * getNewRbFromSmall() const
-    {
-        roaring_bitmap_t * smallRb = roaring_bitmap_create();
-        for (const auto & x : small)
-            roaring_bitmap_add(smallRb, x.getValue());
-        return smallRb;
-    }
-
-    /**
-     * Get a new Roaring64Map from elements of small
-     */
-    std::shared_ptr<Roaring64Map> getNewRb64FromSmall() const
-    {
-        std::shared_ptr<Roaring64Map> smallRb64 = std::make_shared<Roaring64Map>();
-        for (const auto & x : small)
-            smallRb64->add(static_cast<UInt64>(x.getValue()));
-        return smallRb64;
-    }
-
-    /**
-     * Computes the intersection between two bitmaps
-     */
-    void rb_and(const RoaringBitmapWithSmallSet & r1)
-    {
-        ValueBuffer buffer;
-        if (isSmall() && r1.isSmall())
-        {
-            // intersect
-            for (const auto & x : small)
-                if (r1.small.find(x.getValue()) != r1.small.end())
-                    buffer.push_back(x.getValue());
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else if (isSmall() && r1.isLarge())
-        {
-            for (const auto & x : small)
-            {
-                if (isUInt64())
-                {
-                    if (rb64->contains(static_cast<UInt64>(x.getValue())))
-                        buffer.push_back(x.getValue());
-                }
-                else
-                {
-                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                        buffer.push_back(x.getValue());
-                }
-            }
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-                *rb64 &= *newRb64;
-            }
-            else
-            {
-                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-                roaring_bitmap_and_inplace(rb, rb1);
-                if (r1.isSmall())
-                    roaring_bitmap_free(rb1);
-            }
-        }
-    }
-
-    /**
-     * Computes the union between two bitmaps.
-     */
-    void rb_or(const RoaringBitmapWithSmallSet & r1) { merge(r1); }
-
-    /**
-     * Computes the symmetric difference (xor) between two bitmaps.
-     */
-    void rb_xor(const RoaringBitmapWithSmallSet & r1)
-    {
-        if (isSmall())
-            toLarge();
-
-        if (isUInt64())
-        {
-            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-            *rb64 ^= *newRb64;
-        }
-        else
-        {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            roaring_bitmap_xor_inplace(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
-        }
-    }
-
-    /**
-     * Computes the difference (andnot) between two bitmaps
-     */
-    void rb_andnot(const RoaringBitmapWithSmallSet & r1)
-    {
-        ValueBuffer buffer;
-        if (isSmall() && r1.isSmall())
-        {
-            // subtract
-            for (const auto & x : small)
-                if (r1.small.find(x.getValue()) == r1.small.end())
-                    buffer.push_back(x.getValue());
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else if (isSmall() && r1.isLarge())
-        {
-            for (const auto & x : small)
-            {
-                if (isUInt64())
-                {
-                    if (!rb64->contains(static_cast<UInt64>(x.getValue())))
-                        buffer.push_back(x.getValue());
-                }
-                else
-                {
-                    if (!roaring_bitmap_contains(r1.rb, x.getValue()))
-                        buffer.push_back(x.getValue());
-                }
-            }
-
-            // Clear out the original values
-            small.clear();
-
-            for (const auto & value : buffer)
-                small.insert(value);
-
-            buffer.clear();
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-                *rb64 -= *newRb64;
-            }
-            else
-            {
-                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-                roaring_bitmap_andnot_inplace(rb, rb1);
-                if (r1.isSmall())
-                    roaring_bitmap_free(rb1);
-            }
-        }
-    }
-
-    /**
-     * Computes the cardinality of the intersection between two bitmaps.
-     */
-    UInt64 rb_and_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 retSize = 0;
-        if (isSmall() && r1.isSmall())
-        {
-            for (const auto & x : small)
-                if (r1.small.find(x.getValue()) != r1.small.end())
-                    ++retSize;
-        }
-        else if (isSmall() && r1.isLarge())
-        {
-            for (const auto & x : small)
-            {
-                if (isUInt64())
-                {
-                    if (rb64->contains(static_cast<UInt64>(x.getValue())))
-                        ++retSize;
-                }
-                else
-                {
-                    if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                        ++retSize;
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-                retSize = (*rb64 & *newRb64).cardinality();
-            }
-            else
-            {
-                roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-                retSize = roaring_bitmap_and_cardinality(rb, rb1);
-                if (r1.isSmall())
-                    roaring_bitmap_free(rb1);
-            }
-        }
-        return retSize;
-    }
-
-    /**
-     * Computes the cardinality of the union between two bitmaps.
-    */
-    UInt64 rb_or_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 c1 = size();
-        UInt64 c2 = r1.size();
-        UInt64 inter = rb_and_cardinality(r1);
-        return c1 + c2 - inter;
-    }
-
-    /**
-     * Computes the cardinality of the symmetric difference (andnot) between two bitmaps.
-    */
-    UInt64 rb_xor_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 c1 = size();
-        UInt64 c2 = r1.size();
-        UInt64 inter = rb_and_cardinality(r1);
-        return c1 + c2 - 2 * inter;
-    }
-
-    /**
-     * Computes the cardinality of the difference (andnot) between two bitmaps.
-     */
-    UInt64 rb_andnot_cardinality(const RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 c1 = size();
-        UInt64 inter = rb_and_cardinality(r1);
-        return c1 - inter;
-    }
-
-    /**
-     * Return 1 if the two bitmaps contain the same elements.
-     */
-    UInt8 rb_equals(const RoaringBitmapWithSmallSet & r1)
-    {
-        if (isSmall())
-            toLarge();
-
-        UInt8 is_true = 0;
-        if (isUInt64())
-        {
-            std::shared_ptr<Roaring64Map> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.getRb64();
-            is_true = *rb64 == *newRb64;
-        }
-        else
-        {
-            roaring_bitmap_t * rb1 = r1.isSmall() ? r1.getNewRbFromSmall() : r1.getRb();
-            is_true = roaring_bitmap_equals(rb, rb1);
-            if (r1.isSmall())
-                roaring_bitmap_free(rb1);
-        }
-
-        return is_true;
-    }
-
-    /**
-     * Check whether two bitmaps intersect.
-     * Intersection with an empty set is always 0 (consistent with hasAny).
-     */
-    UInt8 rb_intersect(const RoaringBitmapWithSmallSet & r1) const
-    {
-        if (isSmall())
-        {
-            if (r1.isSmall())
-            {
-                for (const auto & x : r1.small)
-                    if (small.find(x.getValue()) != small.end())
-                        return 1;
-            }
-            else
-            {
-                for (const auto & x : small)
-                {
-                    if (isUInt64())
-                    {
-                        if (r1.rb64->contains(static_cast<UInt64>(x.getValue())))
-                            return 1;
-                    }
-                    else
-                    {
-                        if (roaring_bitmap_contains(r1.rb, x.getValue()))
-                            return 1;
-                    }
-                }
-            }
-        }
-        else if (r1.isSmall())
-        {
-            for (const auto & x : r1.small)
-            {
-                if (isUInt64())
-                {
-                    if (rb64->contains(static_cast<UInt64>(x.getValue())))
-                        return 1;
-                }
-                else
-                {
-                    if (roaring_bitmap_contains(rb, x.getValue()))
-                        return 1;
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                if ((*rb64 & *r1.rb64).cardinality() > 0)
-                    return 1;
-            }
-            else
-            {
-                if (roaring_bitmap_intersect(rb, r1.rb))
-                    return 1;
-            }
-        }
-
-        return 0;
-    }
-
-    /**
-     * Check whether the argument is the subset of this set.
-     * Empty set is a subset of any other set (consistent with hasAll).
-     */
-    UInt8 rb_is_subset(const RoaringBitmapWithSmallSet & r1) const
-    {
-        if (isSmall())
-        {
-            if (r1.isSmall())
-            {
-                for (const auto & x : r1.small)
-                    if (small.find(x.getValue()) == small.end())
-                        return 0;
-            }
-            else
-            {
-                UInt64 r1_size = r1.size();
-
-                if (r1_size > small.size())
-                    return 0; // A bigger set can not be a subset of ours.
-
-                // This is a rare case with a small number of elements on
-                // both sides: r1 was promoted to large for some reason and
-                // it is still not larger than our small set.
-                // If r1 is our subset then our size must be equal to
-                // r1_size + number of not found elements, if this sum becomes
-                // greater then r1 is not a subset.
-                for (const auto & x : small)
-                {
-                    if (isUInt64())
-                    {
-                        if (!r1.rb64->contains(static_cast<UInt64>(x.getValue())) && ++r1_size > small.size())
-                            return 0;
-                    }
-                    else
-                    {
-                        if (!roaring_bitmap_contains(r1.rb, x.getValue()) && ++r1_size > small.size())
-                            return 0;
-                    }
-                }
-            }
-        }
-        else if (r1.isSmall())
-        {
-            for (const auto & x : r1.small)
-            {
-                if (isUInt64())
-                {
-                    if (!rb64->contains(static_cast<UInt64>(x.getValue())))
-                        return 0;
-                }
-                else
-                {
-                    if (!roaring_bitmap_contains(rb, x.getValue()))
-                        return 0;
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                if (!r1.rb64->isSubset(*rb64))
-                    return 0;
-            }
-            else
-            {
-                if (!roaring_bitmap_is_subset(r1.rb, rb))
-                    return 0;
-            }
-        }
-        return 1;
-    }
-
-    /**
-     * Check whether this bitmap contains the argument.
-     */
-    UInt8 rb_contains(const UInt64 x) const
-    {
-        if (isSmall())
-        {
-            return small.find(x) != small.end();
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                return rb64->contains(x);
-            }
-            else
-            {
-                return roaring_bitmap_contains(rb, x);
-            }
-        }
-    }
-
-    /**
-     * Remove value
-     */
-    void rb_remove(UInt64 offsetid)
-    {
-        if (isSmall())
-            toLarge();
-
-        if (isUInt64())
-            rb64->remove(offsetid);
-        else
-            roaring_bitmap_remove(rb, offsetid);
-    }
-
-    /**
-     * compute (in place) the negation of the roaring bitmap within a specified
-     * interval: [range_start, range_end). The number of negated values is
-     * range_end - range_start.
-     * Areas outside the range are passed through unchanged.
-     */
-    void rb_flip(UInt64 offsetstart, UInt64 offsetend)
-    {
-        if (isSmall())
-            toLarge();
-        if (isUInt64())
-            rb64->flip(offsetstart, offsetend);
-        else
-            roaring_bitmap_flip_inplace(rb, offsetstart, offsetend);
-    }
-
-    /**
-     * returns the number of integers that are smaller or equal to offsetid.
-     */
-    UInt64 rb_rank(UInt64 offsetid)
-    {
-        if (isSmall())
-            toLarge();
-
-        if (isUInt64())
-            return rb64->rank(offsetid);
-        else
-            return roaring_bitmap_rank(rb, offsetid);
-    }
-
-    /**
-     * Convert elements to integer array, return number of elements
-     */
-    template <typename Element>
-    UInt64 rb_to_array(PaddedPODArray<Element> & res_data) const
-    {
-        UInt64 count = 0;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                res_data.emplace_back(x.getValue());
-                count++;
-            }
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
-                {
-                    res_data.emplace_back(*iterator);
-                    count++;
-                }
-            }
-            else
-            {
-                roaring_uint32_iterator_t iterator;
-                roaring_init_iterator(rb, &iterator);
-                while (iterator.has_value)
-                {
-                    res_data.emplace_back(iterator.current_value);
-                    roaring_advance_uint32_iterator(&iterator);
-                    count++;
-                }
-            }
-        }
-        return count;
-    }
-
-    /**
-     * Return new set with specified range (not include the range_end)
-     */
-    UInt64 rb_range(UInt64 range_start, UInt64 range_end, RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 count = 0;
-        if (range_start >= range_end)
-            return count;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt32(val) >= range_start && UInt32(val) < range_end)
-                {
-                    r1.add(val);
-                    ++count;
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
-                {
-                    if (*iterator < range_start)
-                        continue;
-
-                    if (*iterator < range_end)
-                    {
-                        r1.add(*iterator);
-                        ++count;
-                    }
-                    else
-                        break;
-                }
-            }
-            else
-            {
-                roaring_uint32_iterator_t iterator;
-                roaring_init_iterator(rb, &iterator);
-                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
-                while (iterator.has_value && UInt32(iterator.current_value) < range_end)
-                {
-                    r1.add(iterator.current_value);
-                    roaring_advance_uint32_iterator(&iterator);
-                    ++count;
-                }
-            }
-        }
-        return count;
-    }
-
-    /**
-     * Return new set of the smallest `limit` values in set which is no less than `range_start`.
-     */
-    UInt64 rb_limit(UInt64 range_start, UInt64 limit, RoaringBitmapWithSmallSet & r1) const
-    {
-        UInt64 count = 0;
-        if (isSmall())
-        {
-            std::vector<T> answer;
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt32(val) >= range_start)
-                {
-                    answer.push_back(val);
-                }
-            }
-            sort(answer.begin(), answer.end());
-            if (limit > answer.size())
-                limit = answer.size();
-            for (size_t i = 0; i < limit; ++i)
-                r1.add(answer[i]);
-            count = UInt64(limit);
-        }
-        else
-        {
-            if (isUInt64())
-            {
-                for (Roaring64Map::const_iterator iterator = rb64->begin(); iterator != rb64->end(); iterator++)
-                {
-                    if (*iterator < range_start)
-                        continue;
-
-                    if (count < limit)
-                    {
-                        r1.add(*iterator);
-                        ++count;
-                    }
-                    else
-                        break;
-                }
-            }
-            else
-            {
-                roaring_uint32_iterator_t iterator;
-                roaring_init_iterator(rb, &iterator);
-                roaring_move_uint32_iterator_equalorlarger(&iterator, range_start);
-                while (UInt32(count) < limit && iterator.has_value)
-                {
-                    r1.add(iterator.current_value);
-                    roaring_advance_uint32_iterator(&iterator);
-                    ++count;
-                }
-            }
-        }
-        return count;
-    }
-
-    UInt64 rb_min() const
-    {
-        UInt64 min_val = UINT32_MAX;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt64(val) < min_val)
-                {
-                    min_val = UInt64(val);
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-                min_val = rb64->minimum();
-            else
-                min_val = UInt64(roaring_bitmap_minimum(rb));
-        }
-        return min_val;
-    }
-
-    UInt64 rb_max() const
-    {
-        UInt64 max_val = 0;
-        if (isSmall())
-        {
-            for (const auto & x : small)
-            {
-                T val = x.getValue();
-                if (UInt64(val) > max_val)
-                {
-                    max_val = UInt64(val);
-                }
-            }
-        }
-        else
-        {
-            if (isUInt64())
-                max_val = rb64->maximum();
-            else
-                max_val = UInt64(roaring_bitmap_maximum(rb));
-        }
-        return max_val;
-    }
-
-    /**
-     * Replace value
-     */
-    void rb_replace(const UInt32 * from_vals, const UInt32 * to_vals, size_t num)
-    {
-        if (isSmall())
-            toLarge();
-        for (size_t i = 0; i < num; ++i)
-        {
-            if (from_vals[i] == to_vals[i])
-                continue;
-
-            if (isUInt64())
-            {
-                bool changed = rb64->removeChecked(from_vals[i]);
-                if (changed)
-                    rb64->add(to_vals[i]);
-            }
-            else
-            {
-                bool changed = roaring_bitmap_remove_checked(rb, from_vals[i]);
-                if (changed)
-                    roaring_bitmap_add(rb, to_vals[i]);
-            }
-        }
-    }
-
-private:
- 
-};
-
-template <typename T>
-struct AggregateFunctionGroupBitmapData
-{
-    bool doneFirst = false;
-    RoaringBitmapWithSmallSet<T, 32> rbs;
-    static const char * name() { return "groupBitmap"; }
-};
-
-
-}
+#pragma once
+
+#include <algorithm>
+#include <IO/ReadHelpers.h>
+#include <IO/WriteHelpers.h>
+#include <boost/noncopyable.hpp>
+#include <Common/HashTable/SmallTable.h>
+#include <Common/PODArray.h>
+
+// Include this header last, because it is an auto-generated dump of questionable
+// garbage that breaks the build (e.g. it changes _POSIX_C_SOURCE).
+// TODO: find out what it is. On github, they have proper interface headers like
+// this one: https://github.com/RoaringBitmap/CRoaring/blob/master/include/roaring/roaring.h
+#include <roaring.hh>
+#include <roaring64map.hh>
+
+namespace DB
+{
+/**
+  * For a small number of values - an array of fixed size "on the stack".
+  * For large, roaring bitmap is allocated.
+  * For a description of the roaring_bitmap_t, see: https://github.com/RoaringBitmap/CRoaring
+  */
+template <typename T, UInt8 small_set_size>
+class RoaringBitmapWithSmallSet : private boost::noncopyable
+{
+private:
+    using Small = SmallSet<T, small_set_size>;
+    using ValueBuffer = std::vector<T>;
+    Small small;
+    using RoaringBitmap = std::conditional_t<sizeof(T) >= 8, Roaring64Map, Roaring>;
+    using Value = std::conditional_t<sizeof(T) >= 8, UInt64, UInt32>;
+    std::shared_ptr<RoaringBitmap> rb64 = nullptr;
+
+    void toLarge()
+    {
+        rb64 = std::make_shared<RoaringBitmap>();
+        for (const auto & x : small)
+            rb64->add(static_cast<Value>(x.getValue()));
+        small.clear();
+    }
+
+public:
+    bool isLarge() const { return rb64 != nullptr; }
+
+    bool isSmall() const { return rb64 == nullptr; }
+
+    void add(T value)
+    {
+        if (isSmall())
+        {
+            if (small.find(value) == small.end())
+            {
+                if (!small.full())
+                    small.insert(value);
+                else
+                {
+                    toLarge();
+                    rb64->add(static_cast<Value>(value));
+                }
+            }
+        }
+        else
+        {
+            rb64->add(static_cast<Value>(value));
+        }
+    }
+
+    UInt64 size() const
+    {
+        if (isSmall())
+            return small.size();
+        else
+            return rb64->cardinality();
+    }
+
+    void merge(const RoaringBitmapWithSmallSet & r1)
+    {
+        if (r1.isLarge())
+        {
+            if (isSmall())
+                toLarge();
+
+            *rb64 |= *r1.rb64;
+        }
+        else
+        {
+            for (const auto & x : r1.small)
+                add(x.getValue());
+        }
+    }
+
+    void read(DB::ReadBuffer & in)
+    {
+        // Container type:  0: SmallSet,  1: RoaringBitmap
+        UInt8 containerType = 0;
+        readBinary(containerType, in);
+
+        if (0 == containerType)
+        {
+            small.read(in);
+        }
+        else if (1 == containerType)
+        {
+            size_t size;
+            readVarUInt(size, in);
+            std::unique_ptr<char[]> buf(new char[size]);
+            in.readStrict(buf.get(), size);
+            rb64 = std::make_shared<RoaringBitmap>(RoaringBitmap::read(buf.get()));
+        }
+    }
+
+    void write(DB::WriteBuffer & out) const
+    {
+        // Container type:  0: SmallSet,  1: RoaringBitmap
+        UInt8 containerType = isLarge() ? 1 : 0;
+        writeBinary(containerType, out);
+        if (0 == containerType)
+        {
+            small.write(out);
+        }
+        else if (1 == containerType)
+        {
+            auto size = rb64->getSizeInBytes();
+            writeVarUInt(size, out);
+            std::unique_ptr<char[]> buf(new char[size]);
+            rb64->write(buf.get());
+            out.write(buf.get(), size);
+        }
+    }
+
+    /**
+     * Get a new RoaringBitmap from elements of small
+     */
+    std::shared_ptr<RoaringBitmap> getNewRb64FromSmall() const
+    {
+        std::shared_ptr<RoaringBitmap> smallRb64 = std::make_shared<RoaringBitmap>();
+        for (const auto & x : small)
+            smallRb64->add(static_cast<Value>(x.getValue()));
+        return smallRb64;
+    }
+
+    /**
+     * Computes the intersection between two bitmaps
+     */
+    void rb_and(const RoaringBitmapWithSmallSet & r1)
+    {
+        ValueBuffer buffer;
+        if (isSmall() && r1.isSmall())
+        {
+            // intersect
+            for (const auto & x : small)
+                if (r1.small.find(x.getValue()) != r1.small.end())
+                    buffer.push_back(x.getValue());
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else if (isSmall() && r1.isLarge())
+        {
+            for (const auto & x : small)
+            {
+                if (rb64->contains(static_cast<Value>(x.getValue())))
+                    buffer.push_back(x.getValue());
+            }
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else
+        {
+            std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
+            *rb64 &= *newRb64;
+        }
+    }
+
+    /**
+     * Computes the union between two bitmaps.
+     */
+    void rb_or(const RoaringBitmapWithSmallSet & r1) { merge(r1); }
+
+    /**
+     * Computes the symmetric difference (xor) between two bitmaps.
+     */
+    void rb_xor(const RoaringBitmapWithSmallSet & r1)
+    {
+        if (isSmall())
+            toLarge();
+
+        std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
+        *rb64 ^= *newRb64;
+    }
+
+    /**
+     * Computes the difference (andnot) between two bitmaps
+     */
+    void rb_andnot(const RoaringBitmapWithSmallSet & r1)
+    {
+        ValueBuffer buffer;
+        if (isSmall() && r1.isSmall())
+        {
+            // subtract
+            for (const auto & x : small)
+                if (r1.small.find(x.getValue()) == r1.small.end())
+                    buffer.push_back(x.getValue());
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else if (isSmall() && r1.isLarge())
+        {
+            for (const auto & x : small)
+            {
+                if (!rb64->contains(static_cast<Value>(x.getValue())))
+                    buffer.push_back(x.getValue());
+            }
+
+            // Clear out the original values
+            small.clear();
+
+            for (const auto & value : buffer)
+                small.insert(value);
+
+            buffer.clear();
+        }
+        else
+        {
+            std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
+            *rb64 -= *newRb64;
+        }
+    }
+
+    /**
+     * Computes the cardinality of the intersection between two bitmaps.
+     */
+    UInt64 rb_and_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 ret = 0;
+        if (isSmall() && r1.isSmall())
+        {
+            for (const auto & x : small)
+                if (r1.small.find(x.getValue()) != r1.small.end())
+                    ++ret;
+        }
+        else if (isSmall() && r1.isLarge())
+        {
+            for (const auto & x : small)
+            {
+                if (rb64->contains(static_cast<Value>(x.getValue())))
+                    ++ret;
+            }
+        }
+        else
+        {
+            std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
+            ret = (*rb64 & *newRb64).cardinality();
+        }
+        return ret;
+    }
+
+    /**
+     * Computes the cardinality of the union between two bitmaps.
+    */
+    UInt64 rb_or_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 c1 = size();
+        UInt64 c2 = r1.size();
+        UInt64 inter = rb_and_cardinality(r1);
+        return c1 + c2 - inter;
+    }
+
+    /**
+     * Computes the cardinality of the symmetric difference (andnot) between two bitmaps.
+    */
+    UInt64 rb_xor_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 c1 = size();
+        UInt64 c2 = r1.size();
+        UInt64 inter = rb_and_cardinality(r1);
+        return c1 + c2 - 2 * inter;
+    }
+
+    /**
+     * Computes the cardinality of the difference (andnot) between two bitmaps.
+     */
+    UInt64 rb_andnot_cardinality(const RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 c1 = size();
+        UInt64 inter = rb_and_cardinality(r1);
+        return c1 - inter;
+    }
+
+    /**
+     * Return 1 if the two bitmaps contain the same elements.
+     */
+    UInt8 rb_equals(const RoaringBitmapWithSmallSet & r1)
+    {
+        if (isSmall())
+            toLarge();
+
+        std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
+        return *rb64 == *newRb64;
+    }
+
+    /**
+     * Check whether two bitmaps intersect.
+     * Intersection with an empty set is always 0 (consistent with hasAny).
+     */
+    UInt8 rb_intersect(const RoaringBitmapWithSmallSet & r1) const
+    {
+        if (isSmall())
+        {
+            if (r1.isSmall())
+            {
+                for (const auto & x : r1.small)
+                    if (small.find(x.getValue()) != small.end())
+                        return 1;
+            }
+            else
+            {
+                for (const auto & x : small)
+                {
+                    if (r1.rb64->contains(static_cast<Value>(x.getValue())))
+                        return 1;
+                }
+            }
+        }
+        else if (r1.isSmall())
+        {
+            for (const auto & x : r1.small)
+            {
+                if (rb64->contains(static_cast<Value>(x.getValue())))
+                    return 1;
+            }
+        }
+        else
+        {
+            if ((*rb64 & *r1.rb64).cardinality() > 0)
+                return 1;
+        }
+
+        return 0;
+    }
+
+    /**
+     * Check whether the argument is the subset of this set.
+     * Empty set is a subset of any other set (consistent with hasAll).
+     */
+    UInt8 rb_is_subset(const RoaringBitmapWithSmallSet & r1) const
+    {
+        if (isSmall())
+        {
+            if (r1.isSmall())
+            {
+                for (const auto & x : r1.small)
+                    if (small.find(x.getValue()) == small.end())
+                        return 0;
+            }
+            else
+            {
+                UInt64 r1_size = r1.size();
+
+                // A bigger set can not be a subset of ours.
+                if (r1_size > small.size())
+                    return 0;
+
+                // This is a rare case with a small number of elements on
+                // both sides: r1 was promoted to large for some reason and
+                // it is still not larger than our small set.
+                // If r1 is our subset then our size must be equal to
+                // r1_size + number of not found elements, if this sum becomes
+                // greater then r1 is not a subset.
+                for (const auto & x : small)
+                {
+                    if (!r1.rb64->contains(static_cast<Value>(x.getValue())) && ++r1_size > small.size())
+                        return 0;
+                }
+            }
+        }
+        else if (r1.isSmall())
+        {
+            for (const auto & x : r1.small)
+            {
+                if (!rb64->contains(static_cast<Value>(x.getValue())))
+                    return 0;
+            }
+        }
+        else
+        {
+            if (!r1.rb64->isSubset(*rb64))
+                return 0;
+        }
+        return 1;
+    }
+
+    /**
+     * Check whether this bitmap contains the argument.
+     */
+    UInt8 rb_contains(UInt64 x) const
+    {
+        if (isSmall())
+            return small.find(x) != small.end();
+        else
+            return rb64->contains(x);
+    }
+
+    /**
+     * Remove value
+     */
+    void rb_remove(UInt64 x)
+    {
+        if (isSmall())
+            toLarge();
+
+        rb64->remove(x);
+    }
+
+    /**
+     * compute (in place) the negation of the roaring bitmap within a specified
+     * interval: [range_start, range_end). The number of negated values is
+     * range_end - range_start.
+     * Areas outside the range are passed through unchanged.
+     */
+    void rb_flip(UInt64 begin, UInt64 end)
+    {
+        if (isSmall())
+            toLarge();
+
+        rb64->flip(begin, end);
+    }
+
+    /**
+     * returns the number of integers that are smaller or equal to offsetid.
+     */
+    UInt64 rb_rank(UInt64 x)
+    {
+        if (isSmall())
+            toLarge();
+
+        return rb64->rank(x);
+    }
+
+    /**
+     * Convert elements to integer array, return number of elements
+     */
+    template <typename Element>
+    UInt64 rb_to_array(PaddedPODArray<Element> & res) const
+    {
+        UInt64 count = 0;
+        if (isSmall())
+        {
+            for (const auto & x : small)
+            {
+                res.emplace_back(x.getValue());
+                ++count;
+            }
+        }
+        else
+        {
+            for (auto it = rb64->begin(); it != rb64->end(); ++it)
+            {
+                res.emplace_back(*it);
+                ++count;
+            }
+        }
+        return count;
+    }
+
+    /**
+     * Return new set with specified range (not include the range_end)
+     */
+    UInt64 rb_range(UInt64 range_start, UInt64 range_end, RoaringBitmapWithSmallSet & r1) const
+    {
+        UInt64 count = 0;
+        if (range_start >= range_end)
+            return count;
+        if (isSmall())
+        {
+            for (const auto & x : small)
+            {
+                T val = x.getValue();
+                if (UInt32(val) >= range_start && UInt32(val) < range_end)
+                {
+                    r1.add(val);
+                    ++count;
+                }
+            }
+        }
+        else
+        {
+            for (auto it = rb64->begin(); it != rb64->end(); ++it)
+            {
+                if (*it < range_start)
+                    continue;
+
+                if (*it < range_end)
+                {
+                    r1.add(*it);
+                    ++count;
+                }
+                else
+                    break;
+            }
+        }
+        return count;
+    }
+
+    /**
+     * Return new set of the smallest `limit` values in set which is no less than `range_start`.
+     */
+    UInt64 rb_limit(UInt64 range_start, UInt64 limit, RoaringBitmapWithSmallSet & r1) const
+    {
+        if (limit == 0)
+            return 0;
+
+        if (isSmall())
+        {
+            std::vector<T> answer;
+            for (const auto & x : small)
+            {
+                T val = x.getValue();
+                if (UInt32(val) >= range_start)
+                {
+                    answer.push_back(val);
+                }
+            }
+            if (limit < answer.size())
+            {
+                std::nth_element(answer.begin(), answer.begin() + limit, answer.end());
+                answer.resize(limit);
+            }
+
+            for (const auto & elem : answer)
+                r1.add(elem);
+            return answer.size();
+        }
+        else
+        {
+            UInt64 count = 0;
+            for (auto it = rb64->begin(); it != rb64->end(); ++it)
+            {
+                if (*it < range_start)
+                    continue;
+
+                if (count < limit)
+                {
+                    r1.add(*it);
+                    ++count;
+                }
+                else
+                    break;
+            }
+            return count;
+        }
+    }
+
+    UInt64 rb_min() const
+    {
+        if (isSmall())
+        {
+            auto min_val = std::numeric_limits<std::make_unsigned_t<T>>::max();
+            for (const auto & x : small)
+            {
+                auto val = x.getValue();
+                if (val < min_val)
+                    min_val = val;
+            }
+            return min_val;
+        }
+        else
+            return rb64->minimum();
+    }
+
+    UInt64 rb_max() const
+    {
+        if (isSmall())
+        {
+            auto max_val = std::numeric_limits<std::make_unsigned_t<T>>::min();
+            for (const auto & x : small)
+            {
+                auto val = x.getValue();
+                if (val < max_val)
+                    max_val = val;
+            }
+            return max_val;
+        }
+        else
+            return rb64->maximum();
+    }
+
+    /**
+     * Replace value
+     */
+    void rb_replace(const UInt32 * from_vals, const UInt32 * to_vals, size_t num)
+    {
+        if (isSmall())
+            toLarge();
+
+        for (size_t i = 0; i < num; ++i)
+        {
+            if (from_vals[i] == to_vals[i])
+                continue;
+            bool changed = rb64->removeChecked(from_vals[i]);
+            if (changed)
+                rb64->add(to_vals[i]);
+        }
+    }
+};
+
+template <typename T>
+struct AggregateFunctionGroupBitmapData
+{
+    // If false, all bitmap operations will be treated as merge to initialize the state
+    bool init = false;
+    RoaringBitmapWithSmallSet<T, 32> rbs;
+    static const char * name() { return "groupBitmap"; }
+};
+
+
+}

From e42f83ae7255c5fedc36f2790b19d4520f964329 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Wed, 9 Dec 2020 15:17:58 +0800
Subject: [PATCH 398/504] small fix

---
 .../AggregateFunctionGroupBitmapData.h        | 123 +++++++++---------
 1 file changed, 63 insertions(+), 60 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
index b26e47f0b63..f98c8772752 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
@@ -16,6 +16,13 @@
 
 namespace DB
 {
+
+enum BitmapKind
+{
+    Small = 0,
+    Bitmap = 1
+};
+
 /**
   * For a small number of values - an array of fixed size "on the stack".
   * For large, roaring bitmap is allocated.
@@ -25,25 +32,24 @@ template <typename T, UInt8 small_set_size>
 class RoaringBitmapWithSmallSet : private boost::noncopyable
 {
 private:
-    using Small = SmallSet<T, small_set_size>;
+    SmallSet<T, small_set_size> small;
     using ValueBuffer = std::vector<T>;
-    Small small;
-    using RoaringBitmap = std::conditional_t<sizeof(T) >= 8, Roaring64Map, Roaring>;
+    using RoaringBitmap = std::conditional_t<sizeof(T) >= 8, roaring::Roaring64Map, roaring::Roaring>;
     using Value = std::conditional_t<sizeof(T) >= 8, UInt64, UInt32>;
-    std::shared_ptr<RoaringBitmap> rb64 = nullptr;
+    std::shared_ptr<RoaringBitmap> rb = nullptr;
 
     void toLarge()
     {
-        rb64 = std::make_shared<RoaringBitmap>();
+        rb = std::make_shared<RoaringBitmap>();
         for (const auto & x : small)
-            rb64->add(static_cast<Value>(x.getValue()));
+            rb->add(static_cast<Value>(x.getValue()));
         small.clear();
     }
 
 public:
-    bool isLarge() const { return rb64 != nullptr; }
+    bool isLarge() const { return rb != nullptr; }
 
-    bool isSmall() const { return rb64 == nullptr; }
+    bool isSmall() const { return rb == nullptr; }
 
     void add(T value)
     {
@@ -56,13 +62,13 @@ public:
                 else
                 {
                     toLarge();
-                    rb64->add(static_cast<Value>(value));
+                    rb->add(static_cast<Value>(value));
                 }
             }
         }
         else
         {
-            rb64->add(static_cast<Value>(value));
+            rb->add(static_cast<Value>(value));
         }
     }
 
@@ -71,7 +77,7 @@ public:
         if (isSmall())
             return small.size();
         else
-            return rb64->cardinality();
+            return rb->cardinality();
     }
 
     void merge(const RoaringBitmapWithSmallSet & r1)
@@ -81,7 +87,7 @@ public:
             if (isSmall())
                 toLarge();
 
-            *rb64 |= *r1.rb64;
+            *rb |= *r1.rb;
         }
         else
         {
@@ -92,39 +98,36 @@ public:
 
     void read(DB::ReadBuffer & in)
     {
-        // Container type:  0: SmallSet,  1: RoaringBitmap
-        UInt8 containerType = 0;
-        readBinary(containerType, in);
-
-        if (0 == containerType)
+        UInt8 kind;
+        readBinary(kind, in);
+        if (BitmapKind::Small == kind)
         {
             small.read(in);
         }
-        else if (1 == containerType)
+        else if (BitmapKind::Bitmap == kind)
         {
             size_t size;
             readVarUInt(size, in);
             std::unique_ptr<char[]> buf(new char[size]);
             in.readStrict(buf.get(), size);
-            rb64 = std::make_shared<RoaringBitmap>(RoaringBitmap::read(buf.get()));
+            rb = std::make_shared<RoaringBitmap>(RoaringBitmap::read(buf.get()));
         }
     }
 
     void write(DB::WriteBuffer & out) const
     {
-        // Container type:  0: SmallSet,  1: RoaringBitmap
-        UInt8 containerType = isLarge() ? 1 : 0;
-        writeBinary(containerType, out);
-        if (0 == containerType)
+        UInt8 kind = isLarge() ? BitmapKind::Bitmap : BitmapKind::Small;
+        writeBinary(kind, out);
+        if (BitmapKind::Small == kind)
         {
             small.write(out);
         }
-        else if (1 == containerType)
+        else if (BitmapKind::Bitmap == kind)
         {
-            auto size = rb64->getSizeInBytes();
+            auto size = rb->getSizeInBytes();
             writeVarUInt(size, out);
             std::unique_ptr<char[]> buf(new char[size]);
-            rb64->write(buf.get());
+            rb->write(buf.get());
             out.write(buf.get(), size);
         }
     }
@@ -132,12 +135,12 @@ public:
     /**
      * Get a new RoaringBitmap from elements of small
      */
-    std::shared_ptr<RoaringBitmap> getNewRb64FromSmall() const
+    std::shared_ptr<RoaringBitmap> getNewRoaringBitmapFromSmall() const
     {
-        std::shared_ptr<RoaringBitmap> smallRb64 = std::make_shared<RoaringBitmap>();
+        std::shared_ptr<RoaringBitmap> ret = std::make_shared<RoaringBitmap>();
         for (const auto & x : small)
-            smallRb64->add(static_cast<Value>(x.getValue()));
-        return smallRb64;
+            ret->add(static_cast<Value>(x.getValue()));
+        return ret;
     }
 
     /**
@@ -165,7 +168,7 @@ public:
         {
             for (const auto & x : small)
             {
-                if (rb64->contains(static_cast<Value>(x.getValue())))
+                if (rb->contains(static_cast<Value>(x.getValue())))
                     buffer.push_back(x.getValue());
             }
 
@@ -179,8 +182,8 @@ public:
         }
         else
         {
-            std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
-            *rb64 &= *newRb64;
+            std::shared_ptr<RoaringBitmap> new_rb = r1.isSmall() ? r1.getNewRoaringBitmapFromSmall() : r1.rb;
+            *rb &= *new_rb;
         }
     }
 
@@ -197,8 +200,8 @@ public:
         if (isSmall())
             toLarge();
 
-        std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
-        *rb64 ^= *newRb64;
+        std::shared_ptr<RoaringBitmap> new_rb = r1.isSmall() ? r1.getNewRoaringBitmapFromSmall() : r1.rb;
+        *rb ^= *new_rb;
     }
 
     /**
@@ -226,7 +229,7 @@ public:
         {
             for (const auto & x : small)
             {
-                if (!rb64->contains(static_cast<Value>(x.getValue())))
+                if (!rb->contains(static_cast<Value>(x.getValue())))
                     buffer.push_back(x.getValue());
             }
 
@@ -240,8 +243,8 @@ public:
         }
         else
         {
-            std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
-            *rb64 -= *newRb64;
+            std::shared_ptr<RoaringBitmap> new_rb = r1.isSmall() ? r1.getNewRoaringBitmapFromSmall() : r1.rb;
+            *rb -= *new_rb;
         }
     }
 
@@ -261,14 +264,14 @@ public:
         {
             for (const auto & x : small)
             {
-                if (rb64->contains(static_cast<Value>(x.getValue())))
+                if (rb->contains(static_cast<Value>(x.getValue())))
                     ++ret;
             }
         }
         else
         {
-            std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
-            ret = (*rb64 & *newRb64).cardinality();
+            std::shared_ptr<RoaringBitmap> new_rb = r1.isSmall() ? r1.getNewRoaringBitmapFromSmall() : r1.rb;
+            ret = (*rb & *new_rb).cardinality();
         }
         return ret;
     }
@@ -313,8 +316,8 @@ public:
         if (isSmall())
             toLarge();
 
-        std::shared_ptr<RoaringBitmap> newRb64 = r1.isSmall() ? r1.getNewRb64FromSmall() : r1.rb64;
-        return *rb64 == *newRb64;
+        std::shared_ptr<RoaringBitmap> new_rb = r1.isSmall() ? r1.getNewRoaringBitmapFromSmall() : r1.rb;
+        return *rb == *new_rb;
     }
 
     /**
@@ -335,7 +338,7 @@ public:
             {
                 for (const auto & x : small)
                 {
-                    if (r1.rb64->contains(static_cast<Value>(x.getValue())))
+                    if (r1.rb->contains(static_cast<Value>(x.getValue())))
                         return 1;
                 }
             }
@@ -344,13 +347,13 @@ public:
         {
             for (const auto & x : r1.small)
             {
-                if (rb64->contains(static_cast<Value>(x.getValue())))
+                if (rb->contains(static_cast<Value>(x.getValue())))
                     return 1;
             }
         }
         else
         {
-            if ((*rb64 & *r1.rb64).cardinality() > 0)
+            if ((*rb & *r1.rb).cardinality() > 0)
                 return 1;
         }
 
@@ -387,7 +390,7 @@ public:
                 // greater then r1 is not a subset.
                 for (const auto & x : small)
                 {
-                    if (!r1.rb64->contains(static_cast<Value>(x.getValue())) && ++r1_size > small.size())
+                    if (!r1.rb->contains(static_cast<Value>(x.getValue())) && ++r1_size > small.size())
                         return 0;
                 }
             }
@@ -396,13 +399,13 @@ public:
         {
             for (const auto & x : r1.small)
             {
-                if (!rb64->contains(static_cast<Value>(x.getValue())))
+                if (!rb->contains(static_cast<Value>(x.getValue())))
                     return 0;
             }
         }
         else
         {
-            if (!r1.rb64->isSubset(*rb64))
+            if (!r1.rb->isSubset(*rb))
                 return 0;
         }
         return 1;
@@ -416,7 +419,7 @@ public:
         if (isSmall())
             return small.find(x) != small.end();
         else
-            return rb64->contains(x);
+            return rb->contains(x);
     }
 
     /**
@@ -427,7 +430,7 @@ public:
         if (isSmall())
             toLarge();
 
-        rb64->remove(x);
+        rb->remove(x);
     }
 
     /**
@@ -441,7 +444,7 @@ public:
         if (isSmall())
             toLarge();
 
-        rb64->flip(begin, end);
+        rb->flip(begin, end);
     }
 
     /**
@@ -452,7 +455,7 @@ public:
         if (isSmall())
             toLarge();
 
-        return rb64->rank(x);
+        return rb->rank(x);
     }
 
     /**
@@ -472,7 +475,7 @@ public:
         }
         else
         {
-            for (auto it = rb64->begin(); it != rb64->end(); ++it)
+            for (auto it = rb->begin(); it != rb->end(); ++it)
             {
                 res.emplace_back(*it);
                 ++count;
@@ -503,7 +506,7 @@ public:
         }
         else
         {
-            for (auto it = rb64->begin(); it != rb64->end(); ++it)
+            for (auto it = rb->begin(); it != rb->end(); ++it)
             {
                 if (*it < range_start)
                     continue;
@@ -552,7 +555,7 @@ public:
         else
         {
             UInt64 count = 0;
-            for (auto it = rb64->begin(); it != rb64->end(); ++it)
+            for (auto it = rb->begin(); it != rb->end(); ++it)
             {
                 if (*it < range_start)
                     continue;
@@ -583,7 +586,7 @@ public:
             return min_val;
         }
         else
-            return rb64->minimum();
+            return rb->minimum();
     }
 
     UInt64 rb_max() const
@@ -600,7 +603,7 @@ public:
             return max_val;
         }
         else
-            return rb64->maximum();
+            return rb->maximum();
     }
 
     /**
@@ -615,9 +618,9 @@ public:
         {
             if (from_vals[i] == to_vals[i])
                 continue;
-            bool changed = rb64->removeChecked(from_vals[i]);
+            bool changed = rb->removeChecked(from_vals[i]);
             if (changed)
-                rb64->add(to_vals[i]);
+                rb->add(to_vals[i]);
         }
     }
 };

From ccc426cd10c3215fd8e65e49a94e95071e061682 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Thu, 10 Dec 2020 00:56:31 +0800
Subject: [PATCH 399/504] update croaring

---
 contrib/croaring | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/croaring b/contrib/croaring
index 5f20740ec0d..d8402939b5c 160000
--- a/contrib/croaring
+++ b/contrib/croaring
@@ -1 +1 @@
-Subproject commit 5f20740ec0de5e153e8f4cb2ab91814e8b291a14
+Subproject commit d8402939b5c9fc134fd4fcf058fe0f7006d2b129

From 6648b89843dcc1dcce4445ecd6d65e7e0c2bf0d9 Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Thu, 10 Dec 2020 11:30:44 +0800
Subject: [PATCH 400/504] fix test

---
 src/AggregateFunctions/AggregateFunctionGroupBitmapData.h | 6 +++++-
 tests/queries/0_stateless/00829_bitmap_function.reference | 6 +++---
 2 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
index f98c8772752..734a63fd755 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
@@ -576,6 +576,8 @@ public:
     {
         if (isSmall())
         {
+            if (small.size())
+                return 0;
             auto min_val = std::numeric_limits<std::make_unsigned_t<T>>::max();
             for (const auto & x : small)
             {
@@ -593,11 +595,13 @@ public:
     {
         if (isSmall())
         {
+            if (small.size())
+                return 0;
             auto max_val = std::numeric_limits<std::make_unsigned_t<T>>::min();
             for (const auto & x : small)
             {
                 auto val = x.getValue();
-                if (val < max_val)
+                if (val > max_val)
                     max_val = val;
             }
             return max_val;
diff --git a/tests/queries/0_stateless/00829_bitmap_function.reference b/tests/queries/0_stateless/00829_bitmap_function.reference
index e03a7300db0..02633af568d 100644
--- a/tests/queries/0_stateless/00829_bitmap_function.reference
+++ b/tests/queries/0_stateless/00829_bitmap_function.reference
@@ -91,9 +91,9 @@ tag4	[0,1,2,3,4,5,6,7,8,9]	[5,999,2]	[2,888,20]	[0,1,3,4,6,7,8,9,20]
 [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,100,200,500]
 [30,31,32,33,100,200,500]
 [100,200,500]
-4294967295
-4294967295
-4294967295
+0
+0
+0
 1
 0
 0

From a797ac6d802c5597734c610455f3dc2d45500b9f Mon Sep 17 00:00:00 2001
From: Amos Bird <amosbird@gmail.com>
Date: Fri, 11 Dec 2020 00:31:22 +0800
Subject: [PATCH 401/504] fix test again...

---
 src/AggregateFunctions/AggregateFunctionGroupBitmapData.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
index 734a63fd755..5bca64095fa 100644
--- a/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupBitmapData.h
@@ -576,7 +576,7 @@ public:
     {
         if (isSmall())
         {
-            if (small.size())
+            if (small.empty())
                 return 0;
             auto min_val = std::numeric_limits<std::make_unsigned_t<T>>::max();
             for (const auto & x : small)
@@ -595,7 +595,7 @@ public:
     {
         if (isSmall())
         {
-            if (small.size())
+            if (small.empty())
                 return 0;
             auto max_val = std::numeric_limits<std::make_unsigned_t<T>>::min();
             for (const auto & x : small)

From cf4337ca41ab33a06c6167273cf51ccff893053e Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 18:47:38 +0300
Subject: [PATCH 402/504] Some intermediate stage

---
 .gitmodules                        |  3 +++
 contrib/NuRaft                     |  1 +
 contrib/boost-cmake/CMakeLists.txt | 18 +++++++++++++++++-
 src/Common/tests/CMakeLists.txt    |  3 +++
 4 files changed, 24 insertions(+), 1 deletion(-)
 create mode 160000 contrib/NuRaft

diff --git a/.gitmodules b/.gitmodules
index 03cd3ef94ae..63173402993 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -212,3 +212,6 @@
 [submodule "contrib/boringssl"]
 	path = contrib/boringssl
 	url = https://github.com/ClickHouse-Extras/boringssl.git
+[submodule "contrib/NuRaft"]
+	path = contrib/NuRaft
+	url = https://github.com/eBay/NuRaft.git
diff --git a/contrib/NuRaft b/contrib/NuRaft
new file mode 160000
index 00000000000..410bd149da8
--- /dev/null
+++ b/contrib/NuRaft
@@ -0,0 +1 @@
+Subproject commit 410bd149da84cdde60b4436b02b738749f4e87e1
diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index f7c1ce22e90..54d0ebe1d6c 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -12,10 +12,12 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         program_options
         regex
         context
+        coroutine
     )
 
     if(Boost_INCLUDE_DIR AND Boost_FILESYSTEM_LIBRARY AND Boost_FILESYSTEM_LIBRARY AND
-       Boost_PROGRAM_OPTIONS_LIBRARY AND Boost_REGEX_LIBRARY AND Boost_SYSTEM_LIBRARY AND Boost_CONTEXT_LIBRARY)
+        Boost_PROGRAM_OPTIONS_LIBRARY AND Boost_REGEX_LIBRARY AND Boost_SYSTEM_LIBRARY AND Boost_CONTEXT_LIBRARY AND
+        Boost_COROUTINE_LIBRARY)
 
         set(EXTERNAL_BOOST_FOUND 1)
 
@@ -29,6 +31,7 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         add_library (_boost_regex INTERFACE)
         add_library (_boost_system INTERFACE)
         add_library (_boost_context INTERFACE)
+        add_library (_boost_coroutine INTERFACE)
 
         target_link_libraries (_boost_filesystem INTERFACE ${Boost_FILESYSTEM_LIBRARY})
         target_link_libraries (_boost_iostreams INTERFACE ${Boost_IOSTREAMS_LIBRARY})
@@ -36,6 +39,7 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         target_link_libraries (_boost_regex INTERFACE ${Boost_REGEX_LIBRARY})
         target_link_libraries (_boost_system INTERFACE ${Boost_SYSTEM_LIBRARY})
         target_link_libraries (_boost_context INTERFACE ${Boost_CONTEXT_LIBRARY})
+        target_link_libraries (_boost_coroutine INTERFACE ${Boost_COROUTINE_LIBRARY})
 
         add_library (boost::filesystem ALIAS _boost_filesystem)
         add_library (boost::iostreams ALIAS _boost_iostreams)
@@ -43,6 +47,7 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         add_library (boost::regex ALIAS _boost_regex)
         add_library (boost::system ALIAS _boost_system)
         add_library (boost::context ALIAS _boost_context)
+        add_library (boost::coroutine ALIAS _boost_coroutine)
     else()
         set(EXTERNAL_BOOST_FOUND 0)
         message (${RECONFIGURE_MESSAGE_LEVEL} "Can't find system boost")
@@ -199,4 +204,15 @@ if (NOT EXTERNAL_BOOST_FOUND)
     add_library (_boost_context ${SRCS_CONTEXT})
     add_library (boost::context ALIAS _boost_context)
     target_include_directories (_boost_context PRIVATE ${LIBRARY_DIR})
+
+    # coroutine
+
+    set (SRCS_COROUTINE
+      ${LIBRARY_DIR}/libs/coroutine/detail/coroutine_context.cpp
+      ${LIBRARY_DIR}/libs/coroutine/exceptions.cpp
+      ${LIBRARY_DIR}/libs/coroutine/posix/stack_traits.cpp
+    )
+    add_library (_boost_coroutine ${SRCS_COROUTINE})
+    add_library (boost::coroutine ALIAS _boost_coroutine)
+    target_include_directories (_boost_coroutine PRIVATE ${LIBRARY_DIR})
 endif ()
diff --git a/src/Common/tests/CMakeLists.txt b/src/Common/tests/CMakeLists.txt
index cb36e2b97d2..2181763fb99 100644
--- a/src/Common/tests/CMakeLists.txt
+++ b/src/Common/tests/CMakeLists.txt
@@ -79,3 +79,6 @@ target_link_libraries (average PRIVATE clickhouse_common_io)
 
 add_executable (shell_command_inout shell_command_inout.cpp)
 target_link_libraries (shell_command_inout PRIVATE clickhouse_common_io)
+
+add_executable(asio_test asio_test.cpp)
+target_link_libraries (asio_test PRIVATE clickhouse_common_io)

From e7ffc135247c9fdf6d89f7c2d679081f8e28617c Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 24 Dec 2020 18:49:05 +0300
Subject: [PATCH 403/504] Renames.

---
 src/Functions/FunctionsStringHash.cpp | 226 ++++++++++++--------------
 1 file changed, 106 insertions(+), 120 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 7062a47fa22..2f1f900bda9 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -3,8 +3,6 @@
 #include <Functions/ExtractString.h>
 #include <Functions/FunctionFactory.h>
 #include <Functions/FunctionsHashing.h>
-#include <Common/HashTable/ClearableHashMap.h>
-#include <Common/HashTable/Hash.h>
 #include <Common/PODArray.h>
 
 #include <Core/Defines.h>
@@ -141,22 +139,19 @@ struct Hash
     }
 };
 
-// Simhash String -> UInt64
+// SimHash String -> UInt64
 // N: the length of ngram or words shingles
 // CodePoint: UInt8(ASCII) or UInt32(UTF8)
 // UTF8: means ASCII or UTF8, these two parameters CodePoint and UTF8 can only be (UInt8, false) or (UInt32, true)
 // Ngram: means ngram(true) or words shingles(false)
 // CaseInsensitive: means should we consider about letter case or not
 template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
-struct SimhashImpl
+struct SimHashImpl
 {
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    // we made an assumption that the size of one word can't exceed 128, which may not true
-    // if some word's size exceed 128, it would be cut up to several word
-    static constexpr size_t max_string_size = 1u << 15;
     static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
 
-    // Simhash ngram calculate function: String ->UInt64
+    // SimHash ngram calculate function: String ->UInt64
     // this function extracting ngram from input string, and maintain a 64-dimensions vector
     // for each ngram, calculate a 64 bit hash value, and update the vector according the hash value
     // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
@@ -203,7 +198,7 @@ struct SimhashImpl
         return res_bit.to_ullong();
     }
 
-    // Simhash word shingle calculate function: String -> UInt64
+    // SimHash word shingle calculate function: String -> UInt64
     // this function extracting n word shingle from input string, and maintain a 64-dimensions vector as well
     // for each word shingle, calculate a 64 bit hash value, and update the vector according the hash value
     // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
@@ -283,25 +278,21 @@ struct SimhashImpl
         {
             const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
             const size_t data_size = offsets[i] - offsets[i - 1] - 1;
-            if (data_size <= max_string_size)
+
+            if constexpr (Ngram)
             {
-                if constexpr (Ngram)
-                {
-                    if constexpr (!UTF8)
-                        res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
-                    else
-                        res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
-                }
+                if constexpr (!UTF8)
+                    res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
                 else
-                {
-                    if constexpr (!UTF8)
-                        res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
-                    else
-                        res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
-                }
+                    res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
             }
             else
-                res[i] = -1ull;
+            {
+                if constexpr (!UTF8)
+                    res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
+                else
+                    res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+            }
         }
     }
 };
@@ -333,7 +324,7 @@ private:
 };
 
 
-// Minhash: String -> Tuple(UInt64, UInt64)
+// MinHash: String -> Tuple(UInt64, UInt64)
 // for each string, we extract ngram or word shingle,
 // for each ngram or word shingle, calculate a hash value,
 // then we take the K minimum hash values to calculate a hashsum,
@@ -347,17 +338,16 @@ private:
 // Ngram: means ngram(true) or words shingles(false)
 // CaseInsensitive: means should we consider about letter case or not
 template <size_t N, size_t K, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
-struct MinhashImpl
+struct MinHashImpl
 {
     using Less = std::less<size_t>;
     using Greater = std::greater<size_t>;
     using MaxHeap = FixedHeap<std::less<size_t>, K, -1ULL>;
     using MinHeap = FixedHeap<std::greater<size_t>, K, 0>;
     using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    static constexpr size_t max_string_size = 1u << 15;
     static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
 
-    // Minhash ngram calculate function, String -> Tuple(UInt64, UInt64)
+    // MinHash ngram calculate function, String -> Tuple(UInt64, UInt64)
     // we extract ngram from input string, and calculate a hash value for each ngram
     // then we take the K minimum hash values to calculate a hashsum,
     // and take the K maximum hash values to calculate another hashsum,
@@ -397,7 +387,7 @@ struct MinhashImpl
         return std::make_tuple(res1, res2);
     }
 
-    // Minhash word shingle hash value calculate function: String ->Tuple(UInt64, UInt64)
+    // MinHash word shingle hash value calculate function: String ->Tuple(UInt64, UInt64)
     // for each word shingle, we calculate a hash value, but in fact, we just maintain the
     // K minimum and K maximum hash value
     static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> wordShinglesCalculateHashValue(
@@ -416,7 +406,7 @@ struct MinhashImpl
         // word buffer to store one word
         PaddedPODArray<CodePoint> word_buf;
         // how word shingle hash value calculation and word hash storation is same as we
-        // have descripted in Simhash wordShinglesCalculateHashValue function
+        // have descripted in SimHash wordShinglesCalculateHashValue function
         for (size_t i = 0; i < N && start < end; ++i)
         {
             read_one_word(word_buf, start, end);
@@ -456,171 +446,167 @@ struct MinhashImpl
         {
             const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
             const size_t data_size = offsets[i] - offsets[i - 1] - 1;
-            if (data_size <= max_string_size)
+
+            if constexpr (Ngram)
             {
-                if constexpr (Ngram)
-                {
-                    if constexpr (!UTF8)
-                        std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
-                    else
-                        std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
-                }
+                if constexpr (!UTF8)
+                    std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
                 else
-                {
-                    if constexpr (!UTF8)
-                        std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
-                    else
-                        std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
-                }
+                    std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
             }
             else
-                std::tie(res1[i], res2[i]) = std::make_tuple(-1ull, -1ull);
+            {
+                if constexpr (!UTF8)
+                    std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
+                else
+                    std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+            }
         }
     }
 };
 
-struct NameNgramSimhash
+struct NameNgramSimHash
 {
-    static constexpr auto name = "ngramSimhash";
+    static constexpr auto name = "ngramSimHash";
 };
 
-struct NameNgramSimhashCaseInsensitive
+struct NameNgramSimHashCaseInsensitive
 {
-    static constexpr auto name = "ngramSimhashCaseInsensitive";
+    static constexpr auto name = "ngramSimHashCaseInsensitive";
 };
 
-struct NameNgramSimhashUTF8
+struct NameNgramSimHashUTF8
 {
-    static constexpr auto name = "ngramSimhashUTF8";
+    static constexpr auto name = "ngramSimHashUTF8";
 };
 
-struct NameNgramSimhashCaseInsensitiveUTF8
+struct NameNgramSimHashCaseInsensitiveUTF8
 {
-    static constexpr auto name = "ngramSimhashCaseInsensitiveUTF8";
+    static constexpr auto name = "ngramSimHashCaseInsensitiveUTF8";
 };
 
-struct NameWordShingleSimhash
+struct NameWordShingleSimHash
 {
-    static constexpr auto name = "wordShingleSimhash";
+    static constexpr auto name = "wordShingleSimHash";
 };
 
-struct NameWordShingleSimhashCaseInsensitive
+struct NameWordShingleSimHashCaseInsensitive
 {
-    static constexpr auto name = "wordShingleSimhashCaseInsensitive";
+    static constexpr auto name = "wordShingleSimHashCaseInsensitive";
 };
 
-struct NameWordShingleSimhashUTF8
+struct NameWordShingleSimHashUTF8
 {
-    static constexpr auto name = "wordShingleSimhashUTF8";
+    static constexpr auto name = "wordShingleSimHashUTF8";
 };
 
-struct NameWordShingleSimhashCaseInsensitiveUTF8
+struct NameWordShingleSimHashCaseInsensitiveUTF8
 {
-    static constexpr auto name = "wordShingleSimhashCaseInsensitiveUTF8";
+    static constexpr auto name = "wordShingleSimHashCaseInsensitiveUTF8";
 };
 
-struct NameNgramMinhash
+struct NameNgramMinHash
 {
-    static constexpr auto name = "ngramMinhash";
+    static constexpr auto name = "ngramMinHash";
 };
 
-struct NameNgramMinhashCaseInsensitive
+struct NameNgramMinHashCaseInsensitive
 {
-    static constexpr auto name = "ngramMinhashCaseInsensitive";
+    static constexpr auto name = "ngramMinHashCaseInsensitive";
 };
 
-struct NameNgramMinhashUTF8
+struct NameNgramMinHashUTF8
 {
-    static constexpr auto name = "ngramMinhashUTF8";
+    static constexpr auto name = "ngramMinHashUTF8";
 };
 
-struct NameNgramMinhashCaseInsensitiveUTF8
+struct NameNgramMinHashCaseInsensitiveUTF8
 {
-    static constexpr auto name = "ngramMinhashCaseInsensitiveUTF8";
+    static constexpr auto name = "ngramMinHashCaseInsensitiveUTF8";
 };
 
-struct NameWordShingleMinhash
+struct NameWordShingleMinHash
 {
-    static constexpr auto name = "wordShingleMinhash";
+    static constexpr auto name = "wordShingleMinHash";
 };
 
-struct NameWordShingleMinhashCaseInsensitive
+struct NameWordShingleMinHashCaseInsensitive
 {
-    static constexpr auto name = "wordShingleMinhashCaseInsensitive";
+    static constexpr auto name = "wordShingleMinHashCaseInsensitive";
 };
 
-struct NameWordShingleMinhashUTF8
+struct NameWordShingleMinHashUTF8
 {
-    static constexpr auto name = "wordShingleMinhashUTF8";
+    static constexpr auto name = "wordShingleMinHashUTF8";
 };
 
-struct NameWordShingleMinhashCaseInsensitiveUTF8
+struct NameWordShingleMinHashCaseInsensitiveUTF8
 {
-    static constexpr auto name = "wordShingleMinhashCaseInsensitiveUTF8";
+    static constexpr auto name = "wordShingleMinHashCaseInsensitiveUTF8";
 };
 
-// Simhash
-using FunctionNgramSimhash = FunctionsStringHash<SimhashImpl<4, UInt8, false, true, false>, NameNgramSimhash, true>;
+// SimHash
+using FunctionNgramSimHash = FunctionsStringHash<SimHashImpl<4, UInt8, false, true, false>, NameNgramSimHash, true>;
 
-using FunctionNgramSimhashCaseInsensitive
-    = FunctionsStringHash<SimhashImpl<4, UInt8, false, true, true>, NameNgramSimhashCaseInsensitive, true>;
+using FunctionNgramSimHashCaseInsensitive
+    = FunctionsStringHash<SimHashImpl<4, UInt8, false, true, true>, NameNgramSimHashCaseInsensitive, true>;
 
-using FunctionNgramSimhashUTF8 = FunctionsStringHash<SimhashImpl<3, UInt32, true, true, false>, NameNgramSimhashUTF8, true>;
+using FunctionNgramSimHashUTF8 = FunctionsStringHash<SimHashImpl<3, UInt32, true, true, false>, NameNgramSimHashUTF8, true>;
 
-using FunctionNgramSimhashCaseInsensitiveUTF8
-    = FunctionsStringHash<SimhashImpl<3, UInt32, true, true, true>, NameNgramSimhashCaseInsensitiveUTF8, true>;
+using FunctionNgramSimHashCaseInsensitiveUTF8
+    = FunctionsStringHash<SimHashImpl<3, UInt32, true, true, true>, NameNgramSimHashCaseInsensitiveUTF8, true>;
 
-using FunctionWordShingleSimhash = FunctionsStringHash<SimhashImpl<3, UInt8, false, false, false>, NameWordShingleSimhash, true>;
+using FunctionWordShingleSimHash = FunctionsStringHash<SimHashImpl<3, UInt8, false, false, false>, NameWordShingleSimHash, true>;
 
-using FunctionWordShingleSimhashCaseInsensitive
-    = FunctionsStringHash<SimhashImpl<3, UInt8, false, false, true>, NameWordShingleSimhashCaseInsensitive, true>;
+using FunctionWordShingleSimHashCaseInsensitive
+    = FunctionsStringHash<SimHashImpl<3, UInt8, false, false, true>, NameWordShingleSimHashCaseInsensitive, true>;
 
-using FunctionWordShingleSimhashUTF8 = FunctionsStringHash<SimhashImpl<3, UInt32, true, false, false>, NameWordShingleSimhashUTF8, true>;
+using FunctionWordShingleSimHashUTF8 = FunctionsStringHash<SimHashImpl<3, UInt32, true, false, false>, NameWordShingleSimHashUTF8, true>;
 
-using FunctionWordShingleSimhashCaseInsensitiveUTF8
-    = FunctionsStringHash<SimhashImpl<3, UInt32, true, false, true>, NameWordShingleSimhashCaseInsensitiveUTF8, true>;
+using FunctionWordShingleSimHashCaseInsensitiveUTF8
+    = FunctionsStringHash<SimHashImpl<3, UInt32, true, false, true>, NameWordShingleSimHashCaseInsensitiveUTF8, true>;
 
-// Minhash
-using FunctionNgramMinhash = FunctionsStringHash<MinhashImpl<4, 6, UInt8, false, true, false>, NameNgramMinhash, false>;
+// MinHash
+using FunctionNgramMinHash = FunctionsStringHash<MinHashImpl<4, 6, UInt8, false, true, false>, NameNgramMinHash, false>;
 
-using FunctionNgramMinhashCaseInsensitive
-    = FunctionsStringHash<MinhashImpl<4, 6, UInt8, false, true, true>, NameNgramMinhashCaseInsensitive, false>;
+using FunctionNgramMinHashCaseInsensitive
+    = FunctionsStringHash<MinHashImpl<4, 6, UInt8, false, true, true>, NameNgramMinHashCaseInsensitive, false>;
 
-using FunctionNgramMinhashUTF8 = FunctionsStringHash<MinhashImpl<4, 6, UInt32, true, true, false>, NameNgramMinhashUTF8, false>;
+using FunctionNgramMinHashUTF8 = FunctionsStringHash<MinHashImpl<4, 6, UInt32, true, true, false>, NameNgramMinHashUTF8, false>;
 
-using FunctionNgramMinhashCaseInsensitiveUTF8
-    = FunctionsStringHash<MinhashImpl<4, 6, UInt32, true, true, true>, NameNgramMinhashCaseInsensitiveUTF8, false>;
+using FunctionNgramMinHashCaseInsensitiveUTF8
+    = FunctionsStringHash<MinHashImpl<4, 6, UInt32, true, true, true>, NameNgramMinHashCaseInsensitiveUTF8, false>;
 
-using FunctionWordShingleMinhash = FunctionsStringHash<MinhashImpl<3, 6, UInt8, false, false, false>, NameWordShingleMinhash, false>;
+using FunctionWordShingleMinHash = FunctionsStringHash<MinHashImpl<3, 6, UInt8, false, false, false>, NameWordShingleMinHash, false>;
 
-using FunctionWordShingleMinhashCaseInsensitive
-    = FunctionsStringHash<MinhashImpl<3, 6, UInt8, false, false, true>, NameWordShingleMinhashCaseInsensitive, false>;
+using FunctionWordShingleMinHashCaseInsensitive
+    = FunctionsStringHash<MinHashImpl<3, 6, UInt8, false, false, true>, NameWordShingleMinHashCaseInsensitive, false>;
 
-using FunctionWordShingleMinhashUTF8
-    = FunctionsStringHash<MinhashImpl<3, 6, UInt32, true, false, false>, NameWordShingleMinhashUTF8, false>;
+using FunctionWordShingleMinHashUTF8
+    = FunctionsStringHash<MinHashImpl<3, 6, UInt32, true, false, false>, NameWordShingleMinHashUTF8, false>;
 
-using FunctionWordShingleMinhashCaseInsensitiveUTF8
-    = FunctionsStringHash<MinhashImpl<3, 6, UInt32, true, false, true>, NameWordShingleMinhashCaseInsensitiveUTF8, false>;
+using FunctionWordShingleMinHashCaseInsensitiveUTF8
+    = FunctionsStringHash<MinHashImpl<3, 6, UInt32, true, false, true>, NameWordShingleMinHashCaseInsensitiveUTF8, false>;
 
 void registerFunctionsStringHash(FunctionFactory & factory)
 {
-    factory.registerFunction<FunctionNgramSimhash>();
-    factory.registerFunction<FunctionNgramSimhashCaseInsensitive>();
-    factory.registerFunction<FunctionNgramSimhashUTF8>();
-    factory.registerFunction<FunctionNgramSimhashCaseInsensitiveUTF8>();
-    factory.registerFunction<FunctionWordShingleSimhash>();
-    factory.registerFunction<FunctionWordShingleSimhashCaseInsensitive>();
-    factory.registerFunction<FunctionWordShingleSimhashUTF8>();
-    factory.registerFunction<FunctionWordShingleSimhashCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionNgramSimHash>();
+    factory.registerFunction<FunctionNgramSimHashCaseInsensitive>();
+    factory.registerFunction<FunctionNgramSimHashUTF8>();
+    factory.registerFunction<FunctionNgramSimHashCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionWordShingleSimHash>();
+    factory.registerFunction<FunctionWordShingleSimHashCaseInsensitive>();
+    factory.registerFunction<FunctionWordShingleSimHashUTF8>();
+    factory.registerFunction<FunctionWordShingleSimHashCaseInsensitiveUTF8>();
 
-    factory.registerFunction<FunctionNgramMinhash>();
-    factory.registerFunction<FunctionNgramMinhashCaseInsensitive>();
-    factory.registerFunction<FunctionNgramMinhashUTF8>();
-    factory.registerFunction<FunctionNgramMinhashCaseInsensitiveUTF8>();
-    factory.registerFunction<FunctionWordShingleMinhash>();
-    factory.registerFunction<FunctionWordShingleMinhashCaseInsensitive>();
-    factory.registerFunction<FunctionWordShingleMinhashUTF8>();
-    factory.registerFunction<FunctionWordShingleMinhashCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionNgramMinHash>();
+    factory.registerFunction<FunctionNgramMinHashCaseInsensitive>();
+    factory.registerFunction<FunctionNgramMinHashUTF8>();
+    factory.registerFunction<FunctionNgramMinHashCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionWordShingleMinHash>();
+    factory.registerFunction<FunctionWordShingleMinHashCaseInsensitive>();
+    factory.registerFunction<FunctionWordShingleMinHashUTF8>();
+    factory.registerFunction<FunctionWordShingleMinHashCaseInsensitiveUTF8>();
 }
 }
 

From 25d6f2776b917f45b02e5beb5c6f2e177861ae8b Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 18:51:12 +0300
Subject: [PATCH 404/504] Add ability to collect coverage from integration
 tests

---
 tests/integration/helpers/cluster.py | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/tests/integration/helpers/cluster.py b/tests/integration/helpers/cluster.py
index a65a420cd5b..d8ef6d89d44 100644
--- a/tests/integration/helpers/cluster.py
+++ b/tests/integration/helpers/cluster.py
@@ -195,6 +195,12 @@ class ClickHouseCluster:
 
         if tag is None:
             tag = self.docker_base_tag
+        if not env_variables:
+            env_variables = {}
+
+        # Code coverage files will be placed in database directory
+        # (affect only WITH_COVERAGE=1 build)
+        env_variables['LLVM_PROFILE_FILE'] = '/var/lib/clickhouse/server_%h_%p_%m.profraw'
 
         instance = ClickHouseInstance(
             cluster=self,
@@ -221,7 +227,7 @@ class ClickHouseCluster:
             clickhouse_path_dir=clickhouse_path_dir,
             with_odbc_drivers=with_odbc_drivers,
             hostname=hostname,
-            env_variables=env_variables or {},
+            env_variables=env_variables,
             image=image,
             tag=tag,
             stay_alive=stay_alive,
@@ -759,9 +765,11 @@ class ClickHouseCluster:
 
         if kill:
             try:
-                subprocess_check_call(self.base_cmd + ['kill'])
+                subprocess_check_call(self.base_cmd + ['stop', '--timeout', '20'])
             except Exception as e:
-                print("Kill command failed durung shutdown. {}".format(repr(e)))
+                print("Kill command failed during shutdown. {}".format(repr(e)))
+                print("Trying to kill forcefully")
+                subprocess_check_call(self.base_cmd + ['kill'])
 
         try:
             subprocess_check_call(self.base_cmd + ['down', '--volumes', '--remove-orphans'])

From 303c4f98cc8ccd91127ccf0b7bbc66dee43673cb Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 18:51:12 +0300
Subject: [PATCH 405/504] Add ability to collect coverage from integration
 tests

---
 tests/integration/helpers/cluster.py | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/tests/integration/helpers/cluster.py b/tests/integration/helpers/cluster.py
index a65a420cd5b..d8ef6d89d44 100644
--- a/tests/integration/helpers/cluster.py
+++ b/tests/integration/helpers/cluster.py
@@ -195,6 +195,12 @@ class ClickHouseCluster:
 
         if tag is None:
             tag = self.docker_base_tag
+        if not env_variables:
+            env_variables = {}
+
+        # Code coverage files will be placed in database directory
+        # (affect only WITH_COVERAGE=1 build)
+        env_variables['LLVM_PROFILE_FILE'] = '/var/lib/clickhouse/server_%h_%p_%m.profraw'
 
         instance = ClickHouseInstance(
             cluster=self,
@@ -221,7 +227,7 @@ class ClickHouseCluster:
             clickhouse_path_dir=clickhouse_path_dir,
             with_odbc_drivers=with_odbc_drivers,
             hostname=hostname,
-            env_variables=env_variables or {},
+            env_variables=env_variables,
             image=image,
             tag=tag,
             stay_alive=stay_alive,
@@ -759,9 +765,11 @@ class ClickHouseCluster:
 
         if kill:
             try:
-                subprocess_check_call(self.base_cmd + ['kill'])
+                subprocess_check_call(self.base_cmd + ['stop', '--timeout', '20'])
             except Exception as e:
-                print("Kill command failed durung shutdown. {}".format(repr(e)))
+                print("Kill command failed during shutdown. {}".format(repr(e)))
+                print("Trying to kill forcefully")
+                subprocess_check_call(self.base_cmd + ['kill'])
 
         try:
             subprocess_check_call(self.base_cmd + ['down', '--volumes', '--remove-orphans'])

From 8e5279cf6961ac62f1b2e8dadcf356a21e1f6eda Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Thu, 24 Dec 2020 19:14:09 +0300
Subject: [PATCH 406/504] fix gcc explicit namespace bug

---
 src/Functions/intDiv.cpp | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/Functions/intDiv.cpp b/src/Functions/intDiv.cpp
index 25c21f8439c..54e7f9ed5e1 100644
--- a/src/Functions/intDiv.cpp
+++ b/src/Functions/intDiv.cpp
@@ -88,6 +88,9 @@ struct DivideIntegralByConstantImpl
 
 }
 
+namespace impl_
+{
+
 template <> struct BinaryOperationImpl<UInt64, UInt8, DivideIntegralImpl<UInt64, UInt8>> : DivideIntegralByConstantImpl<UInt64, UInt8> {};
 template <> struct BinaryOperationImpl<UInt64, UInt16, DivideIntegralImpl<UInt64, UInt16>> : DivideIntegralByConstantImpl<UInt64, UInt16> {};
 template <> struct BinaryOperationImpl<UInt64, UInt32, DivideIntegralImpl<UInt64, UInt32>> : DivideIntegralByConstantImpl<UInt64, UInt32> {};
@@ -108,6 +111,7 @@ template <> struct BinaryOperationImpl<Int32, Int16, DivideIntegralImpl<Int32, I
 template <> struct BinaryOperationImpl<Int32, Int32, DivideIntegralImpl<Int32, Int32>> : DivideIntegralByConstantImpl<Int32, Int32> {};
 template <> struct BinaryOperationImpl<Int32, Int64, DivideIntegralImpl<Int32, Int64>> : DivideIntegralByConstantImpl<Int32, Int64> {};
 
+}
 
 struct NameIntDiv { static constexpr auto name = "intDiv"; };
 using FunctionIntDiv = BinaryArithmeticOverloadResolver<DivideIntegralImpl, NameIntDiv, false>;

From 5d94b7eec03a0a327f21c1249d083b859073559f Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 19:30:53 +0300
Subject: [PATCH 407/504] Remove non working code

---
 .../MergeTree/MergeTreeDataMergerMutator.cpp         |  3 +--
 src/Storages/MergeTree/MergeTreeIOSettings.h         |  3 ++-
 src/Storages/MergeTree/MergedBlockOutputStream.cpp   | 12 ------------
 src/Storages/MergeTree/MergedBlockOutputStream.h     |  1 -
 4 files changed, 3 insertions(+), 16 deletions(-)

diff --git a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
index 2365ef141b6..7fa4b3b4620 100644
--- a/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
+++ b/src/Storages/MergeTree/MergeTreeDataMergerMutator.cpp
@@ -682,7 +682,6 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataMergerMutator::mergePartsToTempor
     if (disk->exists(new_part_tmp_path))
         throw Exception("Directory " + fullPath(disk, new_part_tmp_path) + " already exists", ErrorCodes::DIRECTORY_ALREADY_EXISTS);
 
-    MergeTreeData::DataPart::ColumnToSize merged_column_to_size;
 
     Names all_column_names = metadata_snapshot->getColumns().getNamesOfPhysical();
     NamesAndTypesList storage_columns = metadata_snapshot->getColumns().getAllPhysical();
@@ -764,6 +763,7 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataMergerMutator::mergePartsToTempor
         rows_sources_uncompressed_write_buf = tmp_disk->writeFile(rows_sources_file_path);
         rows_sources_write_buf = std::make_unique<CompressedWriteBuffer>(*rows_sources_uncompressed_write_buf);
 
+        MergeTreeData::DataPart::ColumnToSize merged_column_to_size;
         for (const MergeTreeData::DataPartPtr & part : parts)
             part->accumulateColumnSizes(merged_column_to_size);
 
@@ -918,7 +918,6 @@ MergeTreeData::MutableDataPartPtr MergeTreeDataMergerMutator::mergePartsToTempor
         merging_columns,
         index_factory.getMany(metadata_snapshot->getSecondaryIndices()),
         compression_codec,
-        merged_column_to_size,
         data_settings->min_merge_bytes_to_use_direct_io,
         blocks_are_granules_size};
 
diff --git a/src/Storages/MergeTree/MergeTreeIOSettings.h b/src/Storages/MergeTree/MergeTreeIOSettings.h
index 7d8dda0c3ca..185211af84b 100644
--- a/src/Storages/MergeTree/MergeTreeIOSettings.h
+++ b/src/Storages/MergeTree/MergeTreeIOSettings.h
@@ -46,7 +46,8 @@ struct MergeTreeWriterSettings
     bool rewrite_primary_key;
     bool blocks_are_granules_size;
 
-    /// true if we write temporary files during alter.
+    /// Used for AIO threshold comparsion
+    /// FIXME currently doesn't work because WriteBufferAIO contain obscure bug(s)
     size_t estimated_size = 0;
 };
 
diff --git a/src/Storages/MergeTree/MergedBlockOutputStream.cpp b/src/Storages/MergeTree/MergedBlockOutputStream.cpp
index e45aaac6169..feaf46194d1 100644
--- a/src/Storages/MergeTree/MergedBlockOutputStream.cpp
+++ b/src/Storages/MergeTree/MergedBlockOutputStream.cpp
@@ -27,7 +27,6 @@ MergedBlockOutputStream::MergedBlockOutputStream(
         columns_list_,
         skip_indices,
         default_codec_,
-        {},
         data_part->storage.global_context.getSettings().min_bytes_to_use_direct_io,
         blocks_are_granules_size)
 {
@@ -39,7 +38,6 @@ MergedBlockOutputStream::MergedBlockOutputStream(
     const NamesAndTypesList & columns_list_,
     const MergeTreeIndices & skip_indices,
     CompressionCodecPtr default_codec_,
-    const MergeTreeData::DataPart::ColumnToSize & merged_column_to_size,
     size_t aio_threshold,
     bool blocks_are_granules_size)
     : IMergedBlockOutputStream(data_part, metadata_snapshot_)
@@ -54,16 +52,6 @@ MergedBlockOutputStream::MergedBlockOutputStream(
         /* rewrite_primary_key = */ true,
         blocks_are_granules_size);
 
-    if (aio_threshold > 0 && !merged_column_to_size.empty())
-    {
-        for (const auto & column : columns_list)
-        {
-            auto size_it = merged_column_to_size.find(column.name);
-            if (size_it != merged_column_to_size.end())
-                writer_settings.estimated_size += size_it->second;
-        }
-    }
-
     if (!part_path.empty())
         volume->getDisk()->createDirectories(part_path);
 
diff --git a/src/Storages/MergeTree/MergedBlockOutputStream.h b/src/Storages/MergeTree/MergedBlockOutputStream.h
index e12d937239d..3db0e45d207 100644
--- a/src/Storages/MergeTree/MergedBlockOutputStream.h
+++ b/src/Storages/MergeTree/MergedBlockOutputStream.h
@@ -27,7 +27,6 @@ public:
         const NamesAndTypesList & columns_list_,
         const MergeTreeIndices & skip_indices,
         CompressionCodecPtr default_codec_,
-        const MergeTreeData::DataPart::ColumnToSize & merged_column_to_size,
         size_t aio_threshold,
         bool blocks_are_granules_size = false);
 

From 64c4ade5b0e0c9a13d6c2475431daf431fc97554 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 19:34:53 +0300
Subject: [PATCH 408/504] Remove accident chage

---
 tests/integration/helpers/cluster.py | 14 +++-----------
 1 file changed, 3 insertions(+), 11 deletions(-)

diff --git a/tests/integration/helpers/cluster.py b/tests/integration/helpers/cluster.py
index d8ef6d89d44..a65a420cd5b 100644
--- a/tests/integration/helpers/cluster.py
+++ b/tests/integration/helpers/cluster.py
@@ -195,12 +195,6 @@ class ClickHouseCluster:
 
         if tag is None:
             tag = self.docker_base_tag
-        if not env_variables:
-            env_variables = {}
-
-        # Code coverage files will be placed in database directory
-        # (affect only WITH_COVERAGE=1 build)
-        env_variables['LLVM_PROFILE_FILE'] = '/var/lib/clickhouse/server_%h_%p_%m.profraw'
 
         instance = ClickHouseInstance(
             cluster=self,
@@ -227,7 +221,7 @@ class ClickHouseCluster:
             clickhouse_path_dir=clickhouse_path_dir,
             with_odbc_drivers=with_odbc_drivers,
             hostname=hostname,
-            env_variables=env_variables,
+            env_variables=env_variables or {},
             image=image,
             tag=tag,
             stay_alive=stay_alive,
@@ -765,11 +759,9 @@ class ClickHouseCluster:
 
         if kill:
             try:
-                subprocess_check_call(self.base_cmd + ['stop', '--timeout', '20'])
-            except Exception as e:
-                print("Kill command failed during shutdown. {}".format(repr(e)))
-                print("Trying to kill forcefully")
                 subprocess_check_call(self.base_cmd + ['kill'])
+            except Exception as e:
+                print("Kill command failed durung shutdown. {}".format(repr(e)))
 
         try:
             subprocess_check_call(self.base_cmd + ['down', '--volumes', '--remove-orphans'])

From 06819c187bd24a9c437fe3370f37405307a97952 Mon Sep 17 00:00:00 2001
From: Dmitriy <sevirov@yandex-team.ru>
Date: Thu, 24 Dec 2020 20:06:11 +0300
Subject: [PATCH 409/504] Update countMatches function
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Поставил ';' в конце запросов.
---
 docs/en/sql-reference/functions/string-search-functions.md | 6 +++---
 docs/ru/sql-reference/functions/string-search-functions.md | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/docs/en/sql-reference/functions/string-search-functions.md b/docs/en/sql-reference/functions/string-search-functions.md
index ef7e378ad11..96a2e74e1b7 100644
--- a/docs/en/sql-reference/functions/string-search-functions.md
+++ b/docs/en/sql-reference/functions/string-search-functions.md
@@ -618,7 +618,7 @@ Type: [UInt64](../../sql-reference/data-types/int-uint.md).
 Query:
 
 ``` sql
-SELECT countMatches('foobar.com', 'o+')
+SELECT countMatches('foobar.com', 'o+');
 ```
 
 Result:
@@ -632,13 +632,13 @@ Result:
 Query:
 
 ``` sql
-SELECT countMatches('aaaa', 'aa')
+SELECT countMatches('aaaa', 'aa');
 ```
 
 Result:
 
 ``` text
-┌─countMatches('aaaa', 'aa')─┐
+┌─countMatches('aaaa', 'aa')────┐
 │                             2 │
 └───────────────────────────────┘
 ```
diff --git a/docs/ru/sql-reference/functions/string-search-functions.md b/docs/ru/sql-reference/functions/string-search-functions.md
index 8bc2bb12382..996937c0d83 100644
--- a/docs/ru/sql-reference/functions/string-search-functions.md
+++ b/docs/ru/sql-reference/functions/string-search-functions.md
@@ -548,7 +548,7 @@ countMatches(haystack, pattern)
 Запрос:
 
 ``` sql
-SELECT countMatches('foobar.com', 'o+')
+SELECT countMatches('foobar.com', 'o+');
 ```
 
 Результат:
@@ -562,13 +562,13 @@ SELECT countMatches('foobar.com', 'o+')
 Запрос:
 
 ``` sql
-SELECT countMatches('aaaa', 'aa')
+SELECT countMatches('aaaa', 'aa');
 ```
 
 Результат:
 
 ``` text
-┌─countMatches('aaaa', 'aa')─┐
+┌─countMatches('aaaa', 'aa')────┐
 │                             2 │
 └───────────────────────────────┘
 ```

From 1ddea6d7eed5cadce087cf19e51413440571a472 Mon Sep 17 00:00:00 2001
From: Denis Glazachev <traceon@users.noreply.github.com>
Date: Thu, 24 Dec 2020 22:08:20 +0400
Subject: [PATCH 410/504] Allow AppleClang builds

---
 cmake/tools.cmake | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/cmake/tools.cmake b/cmake/tools.cmake
index 4e9258a5cb7..cc4046d2469 100644
--- a/cmake/tools.cmake
+++ b/cmake/tools.cmake
@@ -15,10 +15,6 @@ if (COMPILER_GCC)
 elseif (COMPILER_CLANG)
     # Require minimum version of clang/apple-clang
     if (CMAKE_CXX_COMPILER_ID MATCHES "AppleClang")
-        # If you are developer you can figure out what exact versions of AppleClang are Ok,
-        # remove the following line and commit changes below.
-        message (FATAL_ERROR "AppleClang is not supported, you should install clang from brew.")
-
         # AppleClang 10.0.1 (Xcode 10.2) corresponds to LLVM/Clang upstream version 7.0.0
         # AppleClang 11.0.0 (Xcode 11.0) corresponds to LLVM/Clang upstream version 8.0.0
         set (XCODE_MINIMUM_VERSION 10.2)
@@ -89,4 +85,4 @@ if (ARCH_PPC64LE)
     if (COMPILER_CLANG OR (COMPILER_GCC AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 8))
         message(FATAL_ERROR "Only gcc-8 or higher is supported for powerpc architecture")
     endif ()
-endif ()
\ No newline at end of file
+endif ()

From e8f4a19a100eb3321172cb9bc8a72fb99c1f943f Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Thu, 24 Dec 2020 21:57:31 +0300
Subject: [PATCH 411/504] Feractoring StringHash

---
 src/Functions/ExtractString.h | 68 ++++++++++-------------------------
 1 file changed, 18 insertions(+), 50 deletions(-)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index 8313f583025..7406a813143 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -15,47 +15,33 @@
 
 namespace DB
 {
+
+namespace ErrorCodes
+{
+    extern const int LOGICAL_ERROR;
+}
+
 // used by FunctionsStringSimilarity and FunctionsStringHash
 // includes extracting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
-template <size_t N, bool CaseInsensitive>
+template <bool CaseInsensitive>
 struct ExtractStringImpl
 {
     /// Padding form ColumnsString. It is a number of bytes we can always read starting from pos if pos < end.
     static constexpr size_t default_padding = 16;
 
+    const size_t shingle_size;
+    const size_t tail_size;
+
     /// Functions are read `default_padding - (N - 1)` bytes into the buffer. Window of size N is used.
     /// Read copies `N - 1` last bytes from buffer into beginning, and then reads new bytes.
-    static constexpr size_t buffer_size = default_padding + N - 1;
+    const size_t buffer_size = default_padding + tail_size;
 
-    // the length of code_points = buffer_size
-    // pos: the current beginning location that we want to copy data
-    // end: the end location of the string
-    static ALWAYS_INLINE size_t readASCIICodePoints(UInt8 * code_points, const char *& pos, const char * end)
+    explicit ExtractStringImpl(size_t shingle_size_)
+            : shingle_size(shingle_size_)
+            , tail_size(shingle_size > default_padding ? shingle_size : roundUpToPowerOfTwoOrZero(shingle_size - 1))
     {
-        /// Offset before which we copy some data.
-        constexpr size_t padding_offset = default_padding - N + 1;
-        /// We have an array like this for ASCII (N == 4, other cases are similar)
-        /// |a0|a1|a2|a3|a4|a5|a6|a7|a8|a9|a10|a11|a12|a13|a14|a15|a16|a17|a18|
-        /// And we copy                                ^^^^^^^^^^^^^^^ these bytes to the start
-        /// Actually it is enough to copy 3 bytes, but memcpy for 4 bytes translates into 1 instruction
-        memcpy(code_points, code_points + padding_offset, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt8));
-        /// Now we have an array
-        /// |a13|a14|a15|a16|a4|a5|a6|a7|a8|a9|a10|a11|a12|a13|a14|a15|a16|a17|a18|
-        ///              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-        /// Doing unaligned read of 16 bytes and copy them like above
-        /// 16 is also chosen to do two `movups`.
-        /// Such copying allow us to have 3 codepoints from the previous read to produce the 4-grams with them.
-        memcpy(code_points + (N - 1), pos, default_padding * sizeof(UInt8));
-
-        if constexpr (CaseInsensitive)
-        {
-            /// We really need template lambdas with C++20 to do it inline
-            unrollLowering<N - 1>(code_points, std::make_index_sequence<padding_offset>());
-        }
-        pos += padding_offset;
-        if (pos > end)
-            return default_padding - (pos - end);
-        return default_padding;
+        if (shingle_size == 0)
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Shingle size can't be zero");
     }
 
     // read a ASCII word
@@ -73,23 +59,12 @@ struct ExtractStringImpl
         word_buf.assign(word_start, pos);
         if (CaseInsensitive)
         {
-            std::transform(word_buf.begin(), word_buf.end(), word_buf.begin(), [](UInt8 c) { return std::tolower(c); });
+            for (auto & symbol : word_buf)
+                symbol = toLowerIfAlphaASCII(symbol);
         }
         return word_buf.size();
     }
 
-    static ALWAYS_INLINE inline size_t readUTF8CodePoints(UInt32 * code_points, const char *& pos, const char * end)
-    {
-        memcpy(code_points, code_points + default_padding - N + 1, roundUpToPowerOfTwoOrZero(N - 1) * sizeof(UInt32));
-
-        size_t num = N - 1;
-        while (num < default_padding && pos < end)
-        {
-            code_points[num++] = readOneUTF8Code(pos, end);
-        }
-        return num;
-    }
-
     // read one UTF8 word from pos to word
     static ALWAYS_INLINE inline size_t readOneUTF8Word(PaddedPODArray<UInt32> & word_buf, const char *& pos, const char * end)
     {
@@ -105,13 +80,6 @@ struct ExtractStringImpl
         return word_buf.size();
     }
 
-private:
-    template <size_t Offset, typename Container, size_t... I>
-    static ALWAYS_INLINE inline void unrollLowering(Container & cont, const std::index_sequence<I...> &)
-    {
-        ((cont[Offset + I] = std::tolower(cont[Offset + I])), ...);
-    }
-
     // we use ASCII non-alphanum character as UTF8 separator
     static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNumericASCII(c); }
 

From 1e3bd37380a1331727fa93443d20835c5cbc2a6f Mon Sep 17 00:00:00 2001
From: vdimir <vdimirc@gmail.com>
Date: Thu, 24 Dec 2020 07:31:09 +0000
Subject: [PATCH 412/504] Add option access_to_key_from_attributes to ip
 dictionary

---
 src/Dictionaries/IPAddressDictionary.cpp      | 84 +++++++++++++++----
 src/Dictionaries/IPAddressDictionary.h        |  9 +-
 .../0_stateless/01018_ip_dictionary.reference | 15 ++++
 .../0_stateless/01018_ip_dictionary.sql       | 21 ++++-
 4 files changed, 108 insertions(+), 21 deletions(-)

diff --git a/src/Dictionaries/IPAddressDictionary.cpp b/src/Dictionaries/IPAddressDictionary.cpp
index e04b9111ff1..a69e038c1e4 100644
--- a/src/Dictionaries/IPAddressDictionary.cpp
+++ b/src/Dictionaries/IPAddressDictionary.cpp
@@ -107,6 +107,18 @@ static std::pair<Poco::Net::IPAddress, UInt8> parseIPFromString(const std::strin
     }
 }
 
+static size_t formatIPWithPrefix(const unsigned char * src, UInt8 prefix_len, bool isv4, char * dst)
+{
+    char * ptr = dst;
+    if (isv4)
+        formatIPv4(src, ptr);
+    else
+        formatIPv6(src, ptr);
+    *(ptr - 1) = '/';
+    ptr = itoa(prefix_len, ptr);
+    return ptr - dst;
+}
+
 static void validateKeyTypes(const DataTypes & key_types)
 {
     if (key_types.empty() || key_types.size() > 2)
@@ -231,14 +243,21 @@ IPAddressDictionary::IPAddressDictionary(
     const DictionaryStructure & dict_struct_,
     DictionarySourcePtr source_ptr_,
     const DictionaryLifetime dict_lifetime_,
-    bool require_nonempty_)
+    bool require_nonempty_,
+    bool access_to_key_from_attributes_)
     : IDictionaryBase(dict_id_)
     , dict_struct(dict_struct_)
     , source_ptr{std::move(source_ptr_)}
     , dict_lifetime(dict_lifetime_)
     , require_nonempty(require_nonempty_)
+    , access_to_key_from_attributes(access_to_key_from_attributes_)
     , logger(&Poco::Logger::get("IPAddressDictionary"))
 {
+    if (access_to_key_from_attributes)
+    {
+        dict_struct.attributes.emplace_back(dict_struct.key->front());
+    }
+
     createAttributes();
 
     loadData();
@@ -453,8 +472,6 @@ void IPAddressDictionary::loadData()
     auto stream = source_ptr->loadAll();
     stream->readPrefix();
 
-    const auto attributes_size = attributes.size();
-
     std::vector<IPRecord> ip_records;
 
     bool has_ipv6 = false;
@@ -465,14 +482,19 @@ void IPAddressDictionary::loadData()
         element_count += rows;
 
         const ColumnPtr key_column_ptr = block.safeGetByPosition(0).column;
-        const auto attribute_column_ptrs = ext::map<Columns>(ext::range(0, attributes_size), [&](const size_t attribute_idx)
+
+        size_t attributes_size = dict_struct.attributes.size();
+        if (access_to_key_from_attributes)
         {
-            return block.safeGetByPosition(attribute_idx + 1).column;
-        });
+            /// last attribute contains key and will be filled in code below
+            attributes_size--;
+        }
+        const auto attribute_column_ptrs = ext::map<Columns>(ext::range(0, attributes_size),
+            [&](const size_t attribute_idx) { return block.safeGetByPosition(attribute_idx + 1).column; });
 
         for (const auto row : ext::range(0, rows))
         {
-            for (const auto attribute_idx : ext::range(0, attributes_size))
+            for (const auto attribute_idx : ext::range(0, attribute_column_ptrs.size()))
             {
                 const auto & attribute_column = *attribute_column_ptrs[attribute_idx];
                 auto & attribute = attributes[attribute_idx];
@@ -490,6 +512,33 @@ void IPAddressDictionary::loadData()
 
     stream->readSuffix();
 
+    if (access_to_key_from_attributes)
+    {
+        /// We format key attribute values here instead of filling with data from key_column
+        /// because string representation can be normalized if bits beyond mask are set.
+        /// Also all IPv4 will be displayed as mapped IPv6 if threre are any IPv6.
+        /// It's consistent with representation in table created with `ENGINE = Dictionary` from this dictionary.
+        char str_buffer[48];
+        if (has_ipv6)
+        {
+            uint8_t ip_buffer[IPV6_BINARY_LENGTH];
+            for (const auto & record : ip_records)
+            {
+                size_t str_len = formatIPWithPrefix(record.asIPv6Binary(ip_buffer), record.prefixIPv6(), false, str_buffer);
+                setAttributeValue(attributes.back(), String(str_buffer, str_len));
+            }
+        }
+        else
+        {
+            for (const auto & record : ip_records)
+            {
+                UInt32 addr = IPv4AsUInt32(record.addr.addr());
+                size_t str_len = formatIPWithPrefix(reinterpret_cast<const unsigned char *>(&addr), record.prefix, true, str_buffer);
+                setAttributeValue(attributes.back(), String(str_buffer, str_len));
+            }
+        }
+    }
+
     row_idx.reserve(ip_records.size());
     mask_column.reserve(ip_records.size());
 
@@ -679,7 +728,7 @@ void IPAddressDictionary::calculateBytesAllocated()
 template <typename T>
 void IPAddressDictionary::createAttributeImpl(Attribute & attribute, const Field & null_value)
 {
-    attribute.null_values = T(null_value.get<NearestFieldType<T>>());
+    attribute.null_values = null_value.isNull() ? T{} : T(null_value.get<NearestFieldType<T>>());
     attribute.maps.emplace<ContainerType<T>>();
 }
 
@@ -735,7 +784,8 @@ IPAddressDictionary::Attribute IPAddressDictionary::createAttributeWithType(cons
 
         case AttributeUnderlyingType::utString:
         {
-            attr.null_values = null_value.get<String>();
+
+            attr.null_values = null_value.isNull() ? String() : null_value.get<String>();
             attr.maps.emplace<ContainerType<StringRef>>();
             attr.string_arena = std::make_unique<Arena>();
             break;
@@ -979,14 +1029,12 @@ static auto keyViewGetter()
         for (size_t row : ext::range(0, key_ip_column.size()))
         {
             UInt8 mask = key_mask_column.getElement(row);
-            char * ptr = buffer;
+            size_t str_len;
             if constexpr (IsIPv4)
-                formatIPv4(reinterpret_cast<const unsigned char *>(&key_ip_column.getElement(row)), ptr);
+                str_len = formatIPWithPrefix(reinterpret_cast<const unsigned char *>(&key_ip_column.getElement(row)), mask, true, buffer);
             else
-                formatIPv6(reinterpret_cast<const unsigned char *>(key_ip_column.getDataAt(row).data), ptr);
-            *(ptr - 1) = '/';
-            ptr = itoa(mask, ptr);
-            column->insertData(buffer, ptr - buffer);
+                str_len = formatIPWithPrefix(reinterpret_cast<const unsigned char *>(key_ip_column.getDataAt(row).data), mask, false, buffer);
+            column->insertData(buffer, str_len);
         }
         return ColumnsWithTypeAndName{
             ColumnWithTypeAndName(std::move(column), std::make_shared<DataTypeString>(), dict_attributes.front().name)};
@@ -1120,8 +1168,12 @@ void registerDictionaryTrie(DictionaryFactory & factory)
         const auto dict_id = StorageID::fromDictionaryConfig(config, config_prefix);
         const DictionaryLifetime dict_lifetime{config, config_prefix + ".lifetime"};
         const bool require_nonempty = config.getBool(config_prefix + ".require_nonempty", false);
+
+        const auto & layout_prefix = config_prefix + ".layout.ip_trie";
+        const bool access_to_key_from_attributes = config.getBool(layout_prefix + ".access_to_key_from_attributes", false);
         // This is specialised dictionary for storing IPv4 and IPv6 prefixes.
-        return std::make_unique<IPAddressDictionary>(dict_id, dict_struct, std::move(source_ptr), dict_lifetime, require_nonempty);
+        return std::make_unique<IPAddressDictionary>(dict_id, dict_struct, std::move(source_ptr), dict_lifetime,
+                                                     require_nonempty, access_to_key_from_attributes);
     };
     factory.registerLayout("ip_trie", create_layout, true);
 }
diff --git a/src/Dictionaries/IPAddressDictionary.h b/src/Dictionaries/IPAddressDictionary.h
index c63e6b01adf..2009141ebcc 100644
--- a/src/Dictionaries/IPAddressDictionary.h
+++ b/src/Dictionaries/IPAddressDictionary.h
@@ -27,7 +27,8 @@ public:
         const DictionaryStructure & dict_struct_,
         DictionarySourcePtr source_ptr_,
         const DictionaryLifetime dict_lifetime_,
-        bool require_nonempty_);
+        bool require_nonempty_,
+        bool access_to_key_from_attributes_);
 
     std::string getKeyDescription() const { return key_description; }
 
@@ -45,7 +46,8 @@ public:
 
     std::shared_ptr<const IExternalLoadable> clone() const override
     {
-        return std::make_shared<IPAddressDictionary>(getDictionaryID(), dict_struct, source_ptr->clone(), dict_lifetime, require_nonempty);
+        return std::make_shared<IPAddressDictionary>(getDictionaryID(), dict_struct, source_ptr->clone(), dict_lifetime,
+                                                     require_nonempty, access_to_key_from_attributes);
     }
 
     const IDictionarySource * getSource() const override { return source_ptr.get(); }
@@ -238,10 +240,11 @@ private:
 
     static const uint8_t * getIPv6FromOffset(const IPv6Container & ipv6_col, size_t i);
 
-    const DictionaryStructure dict_struct;
+    DictionaryStructure dict_struct;
     const DictionarySourcePtr source_ptr;
     const DictionaryLifetime dict_lifetime;
     const bool require_nonempty;
+    const bool access_to_key_from_attributes;
     const std::string key_description{dict_struct.getKeyDescription()};
 
     /// Contains sorted IP subnetworks. If some addresses equals, subnet with lower mask is placed first.
diff --git a/tests/queries/0_stateless/01018_ip_dictionary.reference b/tests/queries/0_stateless/01018_ip_dictionary.reference
index becd2951be7..f11628001dd 100644
--- a/tests/queries/0_stateless/01018_ip_dictionary.reference
+++ b/tests/queries/0_stateless/01018_ip_dictionary.reference
@@ -123,6 +123,16 @@
 1
 1
 1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
 ***ipv6 trie dict***
 1
 1
@@ -273,6 +283,11 @@
 1
 1
 1
+1
+1
+1
+1
+1
 ***ipv6 trie dict mask***
 1
 1
diff --git a/tests/queries/0_stateless/01018_ip_dictionary.sql b/tests/queries/0_stateless/01018_ip_dictionary.sql
index 4cb56b77180..5df1afcd559 100644
--- a/tests/queries/0_stateless/01018_ip_dictionary.sql
+++ b/tests/queries/0_stateless/01018_ip_dictionary.sql
@@ -207,9 +207,20 @@ INSERT INTO database_for_dict.table_ipv4_trie VALUES ('127.255.255.255/32', 21);
 CREATE DICTIONARY database_for_dict.dict_ipv4_trie ( prefix String, val UInt32 )
 PRIMARY KEY prefix
 SOURCE(CLICKHOUSE(host 'localhost' port 9000 user 'default' db 'database_for_dict' table 'table_ipv4_trie'))
-LAYOUT(IP_TRIE())
+LAYOUT(IP_TRIE(ACCESS_TO_KEY_FROM_ATTRIBUTES 1))
 LIFETIME(MIN 10 MAX 100);
 
+SELECT '127.0.0.0/24' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.0.0.0')));
+SELECT '127.0.0.1/32' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.0.0.1')));
+SELECT '127.0.0.0/24' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.0.0.127')));
+SELECT '127.0.0.0/16' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.0.255.127')));
+SELECT '127.255.0.0/16' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.255.127.127')));
+SELECT '127.255.128.0/24' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.255.128.9')));
+SELECT '127.255.128.0/24' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.255.128.127')));
+SELECT '127.255.128.10/32' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.255.128.10')));
+SELECT '127.255.128.128/25' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.255.128.255')));
+SELECT '127.255.255.128/32' == dictGetString('database_for_dict.dict_ipv4_trie', 'prefix', tuple(IPv4StringToNum('127.255.255.128')));
+
 SELECT 3 == dictGetUInt32('database_for_dict.dict_ipv4_trie', 'val', tuple(IPv4StringToNum('127.0.0.0')));
 SELECT 4 == dictGetUInt32('database_for_dict.dict_ipv4_trie', 'val', tuple(IPv4StringToNum('127.0.0.1')));
 SELECT 3 == dictGetUInt32('database_for_dict.dict_ipv4_trie', 'val', tuple(IPv4StringToNum('127.0.0.127')));
@@ -274,7 +285,7 @@ CREATE DICTIONARY database_for_dict.dict_ip_trie
 )
 PRIMARY KEY prefix
 SOURCE(CLICKHOUSE(host 'localhost' port 9000 user 'default' db 'database_for_dict' table 'table_ip_trie'))
-LAYOUT(IP_TRIE())
+LAYOUT(IP_TRIE(ACCESS_TO_KEY_FROM_ATTRIBUTES 1))
 LIFETIME(MIN 10 MAX 100);
 
 SELECT 'US' == dictGetString('database_for_dict.dict_ip_trie', 'val', tuple(IPv6StringToNum('2620:0:870::')));
@@ -294,6 +305,12 @@ SELECT 'JA' == dictGetString('database_for_dict.dict_ip_trie', 'val', tuple(IPv4
 SELECT 1 == dictHas('database_for_dict.dict_ip_trie', tuple(IPv4StringToNum('127.0.0.1')));
 SELECT 1 == dictHas('database_for_dict.dict_ip_trie', tuple(IPv6StringToNum('::ffff:127.0.0.1')));
 
+SELECT '2620:0:870::/48' == dictGetString('database_for_dict.dict_ip_trie', 'prefix', tuple(IPv6StringToNum('2620:0:870::')));
+SELECT '2a02:6b8:1::/48' == dictGetString('database_for_dict.dict_ip_trie', 'prefix', tuple(IPv6StringToNum('2a02:6b8:1::1')));
+SELECT '2001:db8::/32' == dictGetString('database_for_dict.dict_ip_trie', 'prefix', tuple(IPv6StringToNum('2001:db8::1')));
+SELECT '::ffff:101.79.55.22/128' == dictGetString('database_for_dict.dict_ip_trie', 'prefix', tuple(IPv6StringToNum('::ffff:654f:3716')));
+SELECT '::ffff:101.79.55.22/128' == dictGetString('database_for_dict.dict_ip_trie', 'prefix', tuple(IPv6StringToNum('::ffff:101.79.55.22')));
+
 SELECT '0' == dictGetString('database_for_dict.dict_ip_trie', 'val', tuple(IPv6StringToNum('::0')));
 SELECT '1' == dictGetString('database_for_dict.dict_ip_trie', 'val', tuple(IPv6StringToNum('8000::')));
 SELECT '2' == dictGetString('database_for_dict.dict_ip_trie', 'val', tuple(IPv6StringToNum('c000::')));

From fcfe7f2c5d6481b44bc45a019fb4224cb4c47253 Mon Sep 17 00:00:00 2001
From: vdimir <vdimirc@gmail.com>
Date: Thu, 24 Dec 2020 15:46:45 +0000
Subject: [PATCH 413/504] Add option access_to_key_from_attributes to
 documentation for ip_trie

---
 .../external-dicts-dict-layout.md                      | 10 +++++++++-
 .../external-dicts-dict-layout.md                      |  8 +++++++-
 .../external-dicts-dict-layout.md                      |  6 ++++++
 .../external-dicts-dict-layout.md                      | 10 +++++++++-
 .../external-dicts-dict-layout.md                      |  6 ++++++
 5 files changed, 37 insertions(+), 3 deletions(-)

diff --git a/docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md b/docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
index a49ddf62eb9..05c418b1f15 100644
--- a/docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
+++ b/docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
@@ -406,6 +406,14 @@ Example:
             <null_value>??</null_value>
     </attribute>
     ...
+</structure>
+<layout>
+    <ip_trie>
+        <!-- Key attribute `prefix` can be retrieved via dictGetString. -->
+        <!-- This option increases memory usage. -->
+        <access_to_key_from_attributes>true</access_to_key_from_attributes>
+    </ip_trie>
+</layout>
 ```
 
 or
@@ -435,6 +443,6 @@ dictGetString('prefix', 'asn', tuple(IPv6StringToNum('2001:db8::1')))
 
 Other types are not supported yet. The function returns the attribute for the prefix that corresponds to this IP address. If there are overlapping prefixes, the most specific one is returned.
 
-Data is stored in a `trie`. It must completely fit into RAM.
+Data must completely fit into RAM.
 
 [Original article](https://clickhouse.tech/docs/en/query_language/dicts/external_dicts_dict_layout/) <!--hide-->
diff --git a/docs/fr/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md b/docs/fr/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
index 3bdc8d37d60..2569329fefd 100644
--- a/docs/fr/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
+++ b/docs/fr/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
@@ -367,6 +367,12 @@ Exemple:
             <null_value>??</null_value>
     </attribute>
     ...
+</structure>
+<layout>
+    <ip_trie>
+        <access_to_key_from_attributes>true</access_to_key_from_attributes>
+    </ip_trie>
+</layout>
 ```
 
 ou
@@ -396,6 +402,6 @@ dictGetString('prefix', 'asn', tuple(IPv6StringToNum('2001:db8::1')))
 
 Les autres types ne sont pas encore pris en charge. La fonction renvoie l'attribut du préfixe correspondant à cette adresse IP. S'il y a chevauchement des préfixes, le plus spécifique est retourné.
 
-Les données sont stockées dans une `trie`. Il doit complètement s'intégrer dans la RAM.
+Les données doit complètement s'intégrer dans la RAM.
 
 [Article Original](https://clickhouse.tech/docs/en/query_language/dicts/external_dicts_dict_layout/) <!--hide-->
diff --git a/docs/ja/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md b/docs/ja/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
index 167e0f1f5d1..ac632b99332 100644
--- a/docs/ja/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
+++ b/docs/ja/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
@@ -362,6 +362,12 @@ LAYOUT(DIRECT())
             <null_value>??</null_value>
     </attribute>
     ...
+</structure>
+<layout>
+    <ip_trie>
+        <access_to_key_from_attributes>true</access_to_key_from_attributes>
+    </ip_trie>
+</layout>
 ```
 
 または
diff --git a/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md b/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
index e97b1e421a4..fc4a3ac7285 100644
--- a/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
+++ b/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
@@ -404,6 +404,14 @@ LAYOUT(DIRECT())
             <null_value>??</null_value>
     </attribute>
     ...
+</structure>
+<layout>
+    <ip_trie>
+        <!-- Ключевой аттрибут `prefix` будет доступен через dictGetString -->
+        <!-- Эта опция увеличивает потреблямую память -->
+        <access_to_key_from_attributes>true</access_to_key_from_attributes>
+    </ip_trie>
+</layout>
 ```
 
 или
@@ -433,6 +441,6 @@ dictGetString('prefix', 'asn', tuple(IPv6StringToNum('2001:db8::1')))
 
 Никакие другие типы не поддерживаются. Функция возвращает атрибут для префикса, соответствующего данному IP-адресу. Если есть перекрывающиеся префиксы, возвращается наиболее специфический.
 
-Данные хранятся в побитовом дереве (`trie`), он должен полностью помещаться в оперативной памяти.
+Данные должны полностью помещаться в оперативной памяти.
 
 [Оригинальная статья](https://clickhouse.tech/docs/ru/query_language/dicts/external_dicts_dict_layout/) <!--hide-->
diff --git a/docs/zh/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md b/docs/zh/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
index 176859d3633..190fd74d705 100644
--- a/docs/zh/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
+++ b/docs/zh/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-layout.md
@@ -362,6 +362,12 @@ LAYOUT(DIRECT())
             <null_value>??</null_value>
     </attribute>
     ...
+</structure>
+<layout>
+    <ip_trie>
+        <access_to_key_from_attributes>true</access_to_key_from_attributes>
+    </ip_trie>
+</layout>
 ```
 
 或

From a16535687e088cf71727bc263a561d70f7022bb8 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Thu, 24 Dec 2020 22:52:39 +0300
Subject: [PATCH 414/504] Fix typo for parallel link jobs condition

---
 cmake/limit_jobs.cmake | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/cmake/limit_jobs.cmake b/cmake/limit_jobs.cmake
index 241fa509477..701f34f8dbd 100644
--- a/cmake/limit_jobs.cmake
+++ b/cmake/limit_jobs.cmake
@@ -44,7 +44,7 @@ if (ENABLE_THINLTO AND PARALLEL_LINK_JOBS GREATER 2)
     set (PARALLEL_LINK_JOBS 2)
 endif()
 
-if (PARALLEL_LINK_JOBS AND (NOT NUMBER_OF_LOGICAL_CORES OR PARALLEL_COMPILE_JOBS LESS NUMBER_OF_LOGICAL_CORES))
+if (PARALLEL_LINK_JOBS AND (NOT NUMBER_OF_LOGICAL_CORES OR PARALLEL_LINK_JOBS LESS NUMBER_OF_LOGICAL_CORES))
     set(CMAKE_JOB_POOL_LINK link_job_pool${CMAKE_CURRENT_SOURCE_DIR})
     string (REGEX REPLACE "[^a-zA-Z0-9]+" "_" CMAKE_JOB_POOL_LINK ${CMAKE_JOB_POOL_LINK})
     set_property(GLOBAL APPEND PROPERTY JOB_POOLS ${CMAKE_JOB_POOL_LINK}=${PARALLEL_LINK_JOBS})

From 4c69612fa7c24b01bda3a81edfce186114d20ad9 Mon Sep 17 00:00:00 2001
From: Azat Khuzhin <a3at.mail@gmail.com>
Date: Thu, 24 Dec 2020 22:51:34 +0300
Subject: [PATCH 415/504] Limit parallel link jobs only if ThinLTO *is* enabled

Before it checks only ENABLE_THINLTO flag, which enables ThinLTO only
for RelWithDebInfo.
---
 cmake/limit_jobs.cmake | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/cmake/limit_jobs.cmake b/cmake/limit_jobs.cmake
index 701f34f8dbd..b85260e6c76 100644
--- a/cmake/limit_jobs.cmake
+++ b/cmake/limit_jobs.cmake
@@ -36,10 +36,12 @@ if (NOT PARALLEL_LINK_JOBS AND AVAILABLE_PHYSICAL_MEMORY AND MAX_LINKER_MEMORY)
 endif ()
 
 # ThinLTO provides its own parallel linking
+# (which is enabled only for RELWITHDEBINFO)
+#
 # But use 2 parallel jobs, since:
 # - this is what llvm does
 # - and I've verfied that lld-11 does not use all available CPU time (in peak) while linking one binary
-if (ENABLE_THINLTO AND PARALLEL_LINK_JOBS GREATER 2)
+if (CMAKE_BUILD_TYPE_UC STREQUAL "RELWITHDEBINFO" AND ENABLE_THINLTO AND PARALLEL_LINK_JOBS GREATER 2)
     message(STATUS "ThinLTO provides its own parallel linking - limiting parallel link jobs to 2.")
     set (PARALLEL_LINK_JOBS 2)
 endif()

From 3c104035b1385a6f563782b55b7f7298f227295c Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 23:10:47 +0300
Subject: [PATCH 416/504] Add NuRaft library

---
 CMakeLists.txt                      |  1 +
 cmake/find/nuraft.cmake             | 19 ++++++++++++++
 contrib/CMakeLists.txt              |  4 +++
 contrib/boost                       |  2 +-
 contrib/nuraft-cmake/CMakeLists.txt | 40 +++++++++++++++++++++++++++++
 src/CMakeLists.txt                  |  3 +++
 6 files changed, 68 insertions(+), 1 deletion(-)
 create mode 100644 cmake/find/nuraft.cmake
 create mode 100644 contrib/nuraft-cmake/CMakeLists.txt

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 52cb64fd763..2e42ae3b6bd 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -468,6 +468,7 @@ include (cmake/find/rapidjson.cmake)
 include (cmake/find/fastops.cmake)
 include (cmake/find/odbc.cmake)
 include (cmake/find/rocksdb.cmake)
+include (cmake/find/nuraft.cmake)
 
 
 if(NOT USE_INTERNAL_PARQUET_LIBRARY)
diff --git a/cmake/find/nuraft.cmake b/cmake/find/nuraft.cmake
new file mode 100644
index 00000000000..92bf1e200ac
--- /dev/null
+++ b/cmake/find/nuraft.cmake
@@ -0,0 +1,19 @@
+option(ENABLE_NURAFT "Enable NuRaft" ${ENABLE_LIBRARIES})
+
+if (NOT ENABLE_NURAFT)
+    return()
+endif()
+
+if (NOT EXISTS "${ClickHouse_SOURCE_DIR}/contrib/NuRaft/CMakeLists.txt")
+    message (WARNING "submodule contrib/NuRaft is missing. to fix try run: \n git submodule update --init --recursive")
+    message (${RECONFIGURE_MESSAGE_LEVEL} "Can't find internal NuRaft library")
+    set (USE_NURAFT 0)
+    return()
+endif ()
+
+set (USE_NURAFT 1)
+set (NURAFT_LIBRARY nuraft)
+
+set (NURAFT_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/NuRaft/include")
+
+message (STATUS "Using NuRaft=${USE_NURAFT}: ${NURAFT_INCLUDE_DIR} : ${NURAFT_LIBRARY}")
diff --git a/contrib/CMakeLists.txt b/contrib/CMakeLists.txt
index 9fccee89c1d..671f7e3871b 100644
--- a/contrib/CMakeLists.txt
+++ b/contrib/CMakeLists.txt
@@ -300,5 +300,9 @@ if (USE_INTERNAL_ROCKSDB_LIBRARY)
     add_subdirectory(rocksdb-cmake)
 endif()
 
+if (USE_NURAFT)
+  add_subdirectory(nuraft-cmake)
+endif()
+
 add_subdirectory(fast_float)
 
diff --git a/contrib/boost b/contrib/boost
index 0b98b443aa7..809d0b72f42 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit 0b98b443aa7bb77d65efd7b23b3b8c8a0ab5f1f3
+Subproject commit 809d0b72f422d63977c066566d3aee1feea022ba
diff --git a/contrib/nuraft-cmake/CMakeLists.txt b/contrib/nuraft-cmake/CMakeLists.txt
new file mode 100644
index 00000000000..4a40daaaff8
--- /dev/null
+++ b/contrib/nuraft-cmake/CMakeLists.txt
@@ -0,0 +1,40 @@
+set (LIBRARY_DIR ${ClickHouse_SOURCE_DIR}/contrib/NuRaft)
+
+set (SRCS
+  ${LIBRARY_DIR}/src/handle_priority.cxx
+  ${LIBRARY_DIR}/src/buffer_serializer.cxx
+  ${LIBRARY_DIR}/src/peer.cxx
+  ${LIBRARY_DIR}/src/global_mgr.cxx
+  ${LIBRARY_DIR}/src/buffer.cxx
+  ${LIBRARY_DIR}/src/asio_service.cxx
+  ${LIBRARY_DIR}/src/handle_client_request.cxx
+  ${LIBRARY_DIR}/src/raft_server.cxx
+  ${LIBRARY_DIR}/src/snapshot.cxx
+  ${LIBRARY_DIR}/src/handle_commit.cxx
+  ${LIBRARY_DIR}/src/error_code.cxx
+  ${LIBRARY_DIR}/src/crc32.cxx
+  ${LIBRARY_DIR}/src/handle_snapshot_sync.cxx
+  ${LIBRARY_DIR}/src/stat_mgr.cxx
+  ${LIBRARY_DIR}/src/handle_join_leave.cxx
+  ${LIBRARY_DIR}/src/handle_user_cmd.cxx
+  ${LIBRARY_DIR}/src/handle_custom_notification.cxx
+  ${LIBRARY_DIR}/src/handle_vote.cxx
+  ${LIBRARY_DIR}/src/launcher.cxx
+  ${LIBRARY_DIR}/src/srv_config.cxx
+  ${LIBRARY_DIR}/src/snapshot_sync_req.cxx
+  ${LIBRARY_DIR}/src/handle_timeout.cxx
+  ${LIBRARY_DIR}/src/handle_append_entries.cxx
+  ${LIBRARY_DIR}/src/cluster_config.cxx
+)
+
+
+add_library(nuraft ${SRCS})
+
+target_compile_definitions(nuraft PRIVATE USE_BOOST_ASIO=1)
+
+target_include_directories (nuraft SYSTEM PRIVATE ${LIBRARY_DIR}/include/libnuraft)
+# for some reason include "asio.h" directly without "boost/" prefix.
+target_include_directories (nuraft SYSTEM PRIVATE ${ClickHouse_SOURCE_DIR}/contrib/boost/boost)
+target_link_libraries (nuraft PRIVATE boost::headers_only boost::coroutine)
+
+target_include_directories (nuraft SYSTEM PUBLIC ${LIBRARY_DIR}/include)
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 5f655f1f466..4e04f5607df 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -307,6 +307,9 @@ if (USE_KRB5)
     dbms_target_link_libraries(PRIVATE ${KRB5_LIBRARY})
 endif()
 
+if (USE_NURAFT)
+    dbms_target_link_libraries(PRIVATE ${NURAFT_LIBRARY})
+endif()
 
 if(RE2_INCLUDE_DIR)
     target_include_directories(clickhouse_common_io SYSTEM BEFORE PUBLIC ${RE2_INCLUDE_DIR})

From 4b1fdded4552c336e39c5443253425c0030eb62d Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Thu, 24 Dec 2020 23:12:50 +0300
Subject: [PATCH 417/504] Remove non existing binary

---
 src/Common/tests/CMakeLists.txt | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/src/Common/tests/CMakeLists.txt b/src/Common/tests/CMakeLists.txt
index 2181763fb99..cb36e2b97d2 100644
--- a/src/Common/tests/CMakeLists.txt
+++ b/src/Common/tests/CMakeLists.txt
@@ -79,6 +79,3 @@ target_link_libraries (average PRIVATE clickhouse_common_io)
 
 add_executable (shell_command_inout shell_command_inout.cpp)
 target_link_libraries (shell_command_inout PRIVATE clickhouse_common_io)
-
-add_executable(asio_test asio_test.cpp)
-target_link_libraries (asio_test PRIVATE clickhouse_common_io)

From a671f1359531bf08e86c74c082e7641cda5ee17a Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Fri, 25 Dec 2020 04:20:09 +0300
Subject: [PATCH 418/504] Fix flaky test
 01584_distributed_buffer_cannot_find_column

---
 src/Storages/StorageBuffer.cpp                              | 4 ++--
 src/Storages/StorageBuffer.h                                | 2 +-
 .../01584_distributed_buffer_cannot_find_column.sql         | 6 +++++-
 3 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/src/Storages/StorageBuffer.cpp b/src/Storages/StorageBuffer.cpp
index 7656da38a6c..070d2b676d7 100644
--- a/src/Storages/StorageBuffer.cpp
+++ b/src/Storages/StorageBuffer.cpp
@@ -580,7 +580,7 @@ void StorageBuffer::startup()
         LOG_WARNING(log, "Storage {} is run with readonly settings, it will not be able to insert data. Set appropriate system_profile to fix this.", getName());
     }
 
-    flush_handle = bg_pool.createTask(log->name() + "/Bg", [this]{ flushBack(); });
+    flush_handle = bg_pool.createTask(log->name() + "/Bg", [this]{ backgroundFlush(); });
     flush_handle->activateAndSchedule();
 }
 
@@ -829,7 +829,7 @@ void StorageBuffer::writeBlockToDestination(const Block & block, StoragePtr tabl
 }
 
 
-void StorageBuffer::flushBack()
+void StorageBuffer::backgroundFlush()
 {
     try
     {
diff --git a/src/Storages/StorageBuffer.h b/src/Storages/StorageBuffer.h
index 38596531a1a..ed8a405298b 100644
--- a/src/Storages/StorageBuffer.h
+++ b/src/Storages/StorageBuffer.h
@@ -157,7 +157,7 @@ private:
     /// `table` argument is passed, as it is sometimes evaluated beforehand. It must match the `destination`.
     void writeBlockToDestination(const Block & block, StoragePtr table);
 
-    void flushBack();
+    void backgroundFlush();
     void reschedule();
 
     BackgroundSchedulePool & bg_pool;
diff --git a/tests/queries/0_stateless/01584_distributed_buffer_cannot_find_column.sql b/tests/queries/0_stateless/01584_distributed_buffer_cannot_find_column.sql
index 3ba6a1c9528..4d58e51d690 100644
--- a/tests/queries/0_stateless/01584_distributed_buffer_cannot_find_column.sql
+++ b/tests/queries/0_stateless/01584_distributed_buffer_cannot_find_column.sql
@@ -8,11 +8,15 @@ CREATE TABLE realtimebuff(amount Int64,transID String,userID String,appID String
 
 insert into realtimebuff (amount,transID,userID,appID,appName,transType,orderSource,nau,fau,transactionType,supplier,fMerchant,bankConnCode,reqDate) values (100, '200312000295032','200223000028708','14', 'Data','1', '20','1', '0','123','abc', '1234a','ZPVBIDV', 1598256583);
 
+-- Data is written to the buffer table but has not been written to the Distributed table
 select sum(amount) = 100 from realtimebuff;
 
 OPTIMIZE TABLE realtimebuff;
-select sum(amount) IN (100, 200) from realtimebuff;
+-- Data has been flushed from Buffer table to the Distributed table and can possibly being sent to 0, 1 or 2 shards.
+-- Both shards reside on localhost in the same table.
+select sum(amount) IN (0, 100, 200) from realtimebuff;
 
+-- Data has been sent to all shards.
 SYSTEM FLUSH DISTRIBUTED realtimedistributed;
 select sum(amount) = 200 from realtimebuff;
 

From 05689a460bd3ceb4704165f336c05498d2537826 Mon Sep 17 00:00:00 2001
From: UnamedRus <UnamedRus@users.noreply.github.com>
Date: Fri, 25 Dec 2020 06:01:16 +0300
Subject: [PATCH 419/504] Docs for params support of col, tbl, db names.

---
 docs/ru/interfaces/cli.md | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/docs/ru/interfaces/cli.md b/docs/ru/interfaces/cli.md
index 662097c8a85..8556182af49 100644
--- a/docs/ru/interfaces/cli.md
+++ b/docs/ru/interfaces/cli.md
@@ -93,12 +93,13 @@ clickhouse-client --param_parName="[1, 2]"  -q "SELECT * FROM table WHERE a = {p
 ```
 
 -   `name` — идентификатор подстановки. В консольном клиенте его следует использовать как часть имени параметра `--param_<name> = value`.
--   `data type` — [тип данных](../sql-reference/data-types/index.md) значения. Например, структура данных `(integer, ('string', integer))` может иметь тип данных `Tuple(UInt8, Tuple(String, UInt8))` ([целочисленный](../sql-reference/data-types/int-uint.md) тип может быть и другим).
+-   `data type` — [тип данных](../sql-reference/data-types/index.md) значения. Например, структура данных `(integer, ('string', integer))` может иметь тип данных `Tuple(UInt8, Tuple(String, UInt8))` ([целочисленный](../sql-reference/data-types/int-uint.md) тип может быть и другим). В качестве параметра можно передать название столбца, таблицы и базы данных, в этом случае используется `Identifier` как тип данных.
 
 #### Пример {#primer}
 
 ``` bash
 $ clickhouse-client --param_tuple_in_tuple="(10, ('dt', 10))" -q "SELECT * FROM table WHERE val = {tuple_in_tuple:Tuple(UInt8, Tuple(String, UInt8))}"
+$ clickhouse-client --param_tbl="numbers" --param_db="system" --param_col="number" --query "SELECT {col:Identifier} FROM {db:Identifier}.{tbl:Identifier} LIMIT 10"
 ```
 
 ## Конфигурирование {#interfaces_cli_configuration}

From 8de33481b52ea99eb5a5562cf711b04fe995f659 Mon Sep 17 00:00:00 2001
From: UnamedRus <UnamedRus@users.noreply.github.com>
Date: Fri, 25 Dec 2020 06:10:08 +0300
Subject: [PATCH 420/504] Eng. Params docs for col, tbl, db names.

---
 docs/en/interfaces/cli.md | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/docs/en/interfaces/cli.md b/docs/en/interfaces/cli.md
index fe9a2f3ec51..c0c9b5668c2 100644
--- a/docs/en/interfaces/cli.md
+++ b/docs/en/interfaces/cli.md
@@ -87,12 +87,13 @@ Format a query as usual, then place the values that you want to pass from the ap
 ```
 
 -   `name` — Placeholder identifier. In the console client it should be used in app parameters as `--param_<name> = value`.
--   `data type` — [Data type](../sql-reference/data-types/index.md) of the app parameter value. For example, a data structure like `(integer, ('string', integer))` can have the `Tuple(UInt8, Tuple(String, UInt8))` data type (you can also use another [integer](../sql-reference/data-types/int-uint.md) types).
+-   `data type` — [Data type](../sql-reference/data-types/index.md) of the app parameter value. For example, a data structure like `(integer, ('string', integer))` can have the `Tuple(UInt8, Tuple(String, UInt8))` data type (you can also use another [integer](../sql-reference/data-types/int-uint.md) types). It's also possible to pass table, database, column names as parameter, in that case you would need to use `Identifier` as data type.
 
 #### Example {#example}
 
 ``` bash
 $ clickhouse-client --param_tuple_in_tuple="(10, ('dt', 10))" -q "SELECT * FROM table WHERE val = {tuple_in_tuple:Tuple(UInt8, Tuple(String, UInt8))}"
+$ clickhouse-client --param_tbl="numbers" --param_db="system" --param_col="number" --query "SELECT {col:Identifier} FROM {db:Identifier}.{tbl:Identifier} LIMIT 10"
 ```
 
 ## Configuring {#interfaces_cli_configuration}

From 5bd025a180520f461e791486537561996e27d0b9 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 25 Dec 2020 06:13:30 +0300
Subject: [PATCH 421/504] fix DISTINCT and ORDER BY

---
 src/Interpreters/ExpressionAnalyzer.cpp       | 65 +++++++++++--------
 src/Interpreters/ExpressionAnalyzer.h         |  5 +-
 src/Interpreters/InterpreterSelectQuery.cpp   | 20 ++++--
 .../01591_window_functions.reference          | 26 ++++++--
 .../0_stateless/01591_window_functions.sql    | 11 ++--
 5 files changed, 87 insertions(+), 40 deletions(-)

diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index 4af2a17e164..c10a8313bee 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -970,7 +970,9 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
     ExpressionActionsChain::Step & step = chain.lastStep(aggregated_columns);
 
     // 1) Add actions for window functions and their arguments;
-    // 2) Mark the columns that are really required.
+    // 2) Mark the columns that are really required. We have to mark them as
+    //    required because we finish the expression chain before processing the
+    //    window functions.
     for (const auto & [_, w] : window_descriptions)
     {
         for (const auto & f : w.window_functions)
@@ -1425,49 +1427,47 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         /// If there is aggregation, we execute expressions in SELECT and ORDER BY on the initiating server, otherwise on the source servers.
         query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
 
-        has_order_by = query.orderBy() != nullptr;
-        // FIXME selected columns was set here. Should split order by and select
-        // and insert window functions in between.
-        // Will fix:
-        // 1) window function in DISTINCT
-        // 2) totally broken DISTINCT
-        // 3) window functions in order by
-        before_order_and_select = query_analyzer.appendOrderBy(
-                chain,
-                only_types || (need_aggregate ? !second_stage : !first_stage),
-                optimize_read_in_order,
-                order_by_elements_actions);
-
+        // Window functions are processed in a separate expression chain after
+        // the main SELECT, similar to what we do for aggregate functions.
         if (has_window)
         {
             query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
 
-            auto select_required_columns = chain.getLastStep().required_output;
 
             for (const auto & x : chain.getLastActions()->getNamesAndTypesList())
             {
                 query_analyzer.columns_after_window.push_back(x);
             }
 
+            before_window = chain.getLastActions();
             finalize_chain(chain);
 
             auto & step = chain.lastStep(query_analyzer.columns_after_window);
-            step.required_output = select_required_columns;
+
+            // The output of this expression chain is the result of
+            // SELECT (before "final projection" i.e. renaming the columns), so
+            // we have to mark the expressions that are required in the output,
+            // again. We did it for the previous expression chain ("select w/o
+            // window functions") earlier, in appendSelect(). But that chain also
+            // produced the expressions required to calculate window functions.
+            // They are not needed in the final SELECT result. Knowing the correct
+            // list of columns is important when we apply SELECT DISTINCT later.
             const auto * select_query = query_analyzer.getSelectQuery();
             for (const auto & child : select_query->select()->children)
             {
-                if (const auto * function
-                        = typeid_cast<const ASTFunction *>(child.get());
-                    function && function->is_window_function)
-                {
-                    // See its counterpart in appendSelect()
-                    step.required_output.push_back(child->getColumnName());
-                }
+                step.required_output.push_back(child->getColumnName());
             }
         }
 
         selected_columns = chain.getLastStep().required_output;
 
+        has_order_by = query.orderBy() != nullptr;
+        before_order_by = query_analyzer.appendOrderBy(
+                chain,
+                only_types || (need_aggregate ? !second_stage : !first_stage),
+                optimize_read_in_order,
+                order_by_elements_actions);
+
         if (query_analyzer.appendLimitBy(chain, only_types || !second_stage))
         {
             before_limit_by = chain.getLastActions();
@@ -1479,7 +1479,6 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         finalize_chain(chain);
     }
 
-
     /// Before executing WHERE and HAVING, remove the extra columns from the block (mostly the aggregation keys).
     removeExtraColumns();
 
@@ -1610,9 +1609,9 @@ std::string ExpressionAnalysisResult::dump() const
         ss << "before_window " << before_window->dumpDAG() << "\n";
     }
 
-    if (before_order_and_select)
+    if (before_order_by)
     {
-        ss << "before_order_and_select " << before_order_and_select->dumpDAG() << "\n";
+        ss << "before_order_by " << before_order_by->dumpDAG() << "\n";
     }
 
     if (before_limit_by)
@@ -1625,6 +1624,20 @@ std::string ExpressionAnalysisResult::dump() const
         ss << "final_projection " << final_projection->dumpDAG() << "\n";
     }
 
+    if (!selected_columns.empty())
+    {
+        ss << "selected_columns ";
+        for (size_t i = 0; i < selected_columns.size(); i++)
+        {
+            if (i > 0)
+            {
+                ss << ", ";
+            }
+            ss << backQuote(selected_columns[i]);
+        }
+        ss << "\n";
+    }
+
     return ss.str();
 }
 
diff --git a/src/Interpreters/ExpressionAnalyzer.h b/src/Interpreters/ExpressionAnalyzer.h
index 1095689a775..ea43efa6036 100644
--- a/src/Interpreters/ExpressionAnalyzer.h
+++ b/src/Interpreters/ExpressionAnalyzer.h
@@ -204,11 +204,12 @@ struct ExpressionAnalysisResult
     ActionsDAGPtr before_aggregation;
     ActionsDAGPtr before_having;
     ActionsDAGPtr before_window;
-    ActionsDAGPtr before_order_and_select;
+    ActionsDAGPtr before_order_by;
     ActionsDAGPtr before_limit_by;
     ActionsDAGPtr final_projection;
 
-    /// Columns from the SELECT list, before renaming them to aliases.
+    /// Columns from the SELECT list, before renaming them to aliases. Used to
+    /// perform SELECT DISTINCT.
     Names selected_columns;
 
     /// Columns will be removed after prewhere actions execution.
diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 38cc19a00d6..eb7b78c35b4 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -538,7 +538,10 @@ Block InterpreterSelectQuery::getSampleBlockImpl()
     if (options.to_stage == QueryProcessingStage::Enum::WithMergeableState)
     {
         if (!analysis_result.need_aggregate)
-            return analysis_result.before_order_and_select->getResultColumns();
+        {
+            // What's the difference with selected_columns?
+            return analysis_result.before_order_by->getResultColumns();
+        }
 
         Block header = analysis_result.before_aggregation->getResultColumns();
 
@@ -564,7 +567,8 @@ Block InterpreterSelectQuery::getSampleBlockImpl()
 
     if (options.to_stage == QueryProcessingStage::Enum::WithMergeableStateAfterAggregation)
     {
-        return analysis_result.before_order_and_select->getResultColumns();
+        // What's the difference with selected_columns?
+        return analysis_result.before_order_by->getResultColumns();
     }
 
     return analysis_result.final_projection->getResultColumns();
@@ -958,8 +962,9 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
             }
             else
             {
-                executeExpression(query_plan, expressions.before_order_and_select, "Before ORDER BY and SELECT");
+                executeExpression(query_plan, expressions.before_window, "Before window functions");
                 executeWindow(query_plan);
+                executeExpression(query_plan, expressions.before_order_by, "Before ORDER BY");
                 executeDistinct(query_plan, true, expressions.selected_columns, true);
             }
 
@@ -1005,8 +1010,10 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
                 else if (expressions.hasHaving())
                     executeHaving(query_plan, expressions.before_having);
 
-                executeExpression(query_plan, expressions.before_order_and_select, "Before ORDER BY and SELECT");
+                executeExpression(query_plan, expressions.before_window,
+                    "Before window functions");
                 executeWindow(query_plan);
+                executeExpression(query_plan, expressions.before_order_by, "Before ORDER BY");
                 executeDistinct(query_plan, true, expressions.selected_columns, true);
 
             }
@@ -1745,6 +1752,11 @@ void InterpreterSelectQuery::executeRollupOrCube(QueryPlan & query_plan, Modific
 
 void InterpreterSelectQuery::executeExpression(QueryPlan & query_plan, const ActionsDAGPtr & expression, const std::string & description)
 {
+    if (!expression)
+    {
+        return;
+    }
+
     auto expression_step = std::make_unique<ExpressionStep>(query_plan.getCurrentDataStream(), expression);
 
     expression_step->setStepDescription(description);
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 704c5a98811..e56c6bdeb3a 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -79,12 +79,22 @@ select number, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 9	9
 select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
 
--- should work in ORDER BY though
--- doesn't work now
--- select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+-- should work in ORDER BY
 
--- at least it works in ORDER BY if you wrap it in a subquery
+select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
+-- also works in ORDER BY if you wrap it in a subquery
+
+9	9
+6	8
+7	8
+8	8
+3	5
+4	5
+5	5
+0	2
+1	2
+2	2
 select * from (select count(*) over () c from numbers(3)) order by c;
 
 -- must work in WHERE if you wrap it in a subquery
@@ -168,4 +178,12 @@ select groupArray(number) over () from numbers(3);
 [0,1,2]
 select count(1) over (), max(number + 1) over () from numbers(3);
 
+-- Should work in DISTINCT
+
 1	3
+select distinct sum(0) over () from numbers(2);
+
+0
+select distinct any(number) over () from numbers(2);
+
+0
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index fd2831ea2ac..8b82b8f95fc 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -24,11 +24,10 @@ select number, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 -- last stage of select, after all other functions.
 select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
 
--- should work in ORDER BY though
--- doesn't work now
--- select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
+-- should work in ORDER BY
+select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
--- at least it works in ORDER BY if you wrap it in a subquery
+-- also works in ORDER BY if you wrap it in a subquery
 select * from (select count(*) over () c from numbers(3)) order by c;
 
 -- must work in WHERE if you wrap it in a subquery
@@ -57,3 +56,7 @@ select groupArray(number) over () from numbers(3);
 -- This one tests we properly process the window  function arguments.
 -- Seen errors like 'column `1` not found' from count(1).
 select count(1) over (), max(number + 1) over () from numbers(3);
+
+-- Should work in DISTINCT
+select distinct sum(0) over () from numbers(2);
+select distinct any(number) over () from numbers(2);

From a38787553ce363c41df430d1642ac1ba922261c0 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 25 Dec 2020 06:15:36 +0300
Subject: [PATCH 422/504] perf test fix

---
 tests/performance/window_functions.xml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/tests/performance/window_functions.xml b/tests/performance/window_functions.xml
index e4578a575a0..dcc873b5029 100644
--- a/tests/performance/window_functions.xml
+++ b/tests/performance/window_functions.xml
@@ -3,6 +3,10 @@
         <table_exists>hits_100m_single</table_exists>
     </preconditions>
 
+    <settings>
+        <allow_experimental_window_functions>1</allow_experimental_window_functions>
+    </settings>
+
     <!--
         For some counters, find top 10 users by the numer of records.
         First with LIMIT BY, next with window functions.

From 998bd17877e887d0e4eb57c90c7a0ea1d0d8c250 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Fri, 25 Dec 2020 07:59:17 +0300
Subject: [PATCH 423/504] fix column name clash w/agg fns

also fix some test references influenced by the explain changes
---
 src/Interpreters/QueryNormalizer.cpp          |  30 ++--
 src/Interpreters/QueryNormalizer.h            |   2 +-
 src/Parsers/ASTFunction.cpp                   |  10 ++
 .../01508_explain_header.reference            |   2 +-
 ..._explain_select_with_union_query.reference | 144 +++++++++---------
 ...monotonous_functions_in_order_by.reference |   6 +-
 .../01591_window_functions.reference          |  28 +++-
 .../0_stateless/01591_window_functions.sql    |  16 +-
 8 files changed, 138 insertions(+), 100 deletions(-)

diff --git a/src/Interpreters/QueryNormalizer.cpp b/src/Interpreters/QueryNormalizer.cpp
index 25496c2b613..213e354a13f 100644
--- a/src/Interpreters/QueryNormalizer.cpp
+++ b/src/Interpreters/QueryNormalizer.cpp
@@ -148,9 +148,9 @@ void QueryNormalizer::visit(ASTSelectQuery & select, const ASTPtr &, Data & data
 /// Don't go into select query. It processes children itself.
 /// Do not go to the left argument of lambda expressions, so as not to replace the formal parameters
 ///  on aliases in expressions of the form 123 AS x, arrayMap(x -> 1, [2]).
-void QueryNormalizer::visitChildren(const ASTPtr & node, Data & data)
+void QueryNormalizer::visitChildren(ASTPtr & node, Data & data)
 {
-    if (const auto * func_node = node->as<ASTFunction>())
+    if (auto * func_node = node->as<ASTFunction>())
     {
         if (func_node->tryGetQueryArgument())
         {
@@ -159,21 +159,25 @@ void QueryNormalizer::visitChildren(const ASTPtr & node, Data & data)
             /// Don't go into query argument.
             return;
         }
-        /// We skip the first argument. We also assume that the lambda function can not have parameters.
-        size_t first_pos = 0;
-        if (func_node->name == "lambda")
-            first_pos = 1;
 
-        if (func_node->arguments)
+        // Process all function node children (arguments + window definition for
+        // window functions).
+        // We have to skip the first argument if it's the "lambda" function,
+        // because it contains formal parameters like x->x + 1;
+        const IAST * argument_to_skip = nullptr;
+        if (func_node->name == "lambda"
+            && func_node->arguments
+            && !func_node->arguments->children.empty())
         {
-            auto & func_children = func_node->arguments->children;
+            argument_to_skip = func_node->arguments->children[0].get();
+        }
 
-            for (size_t i = first_pos; i < func_children.size(); ++i)
+        for (auto & child : func_node->children)
+        {
+            if (child.get() != argument_to_skip
+                && needVisitChild(child))
             {
-                auto & child = func_children[i];
-
-                if (needVisitChild(child))
-                    visit(child, data);
+                visit(child, data);
             }
         }
     }
diff --git a/src/Interpreters/QueryNormalizer.h b/src/Interpreters/QueryNormalizer.h
index 5bd0064c002..951628d0738 100644
--- a/src/Interpreters/QueryNormalizer.h
+++ b/src/Interpreters/QueryNormalizer.h
@@ -69,7 +69,7 @@ private:
     static void visit(ASTTablesInSelectQueryElement &, const ASTPtr &, Data &);
     static void visit(ASTSelectQuery &, const ASTPtr &, Data &);
 
-    static void visitChildren(const ASTPtr &, Data & data);
+    static void visitChildren(ASTPtr & node, Data & data);
 };
 
 }
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index d5d03a540c9..980c395c7aa 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -39,6 +39,16 @@ void ASTFunction::appendColumnNameImpl(WriteBuffer & ostr) const
             (*it)->appendColumnName(ostr);
         }
     writeChar(')', ostr);
+
+    if (is_window_function)
+    {
+        writeCString(" OVER (", ostr);
+        FormatSettings settings{ostr, true /* one_line */};
+        FormatState state;
+        FormatStateStacked frame;
+        appendWindowDescription(settings, state, frame);
+        writeCString(")", ostr);
+    }
 }
 
 /** Get the text that identifies this element. */
diff --git a/tests/queries/0_stateless/01508_explain_header.reference b/tests/queries/0_stateless/01508_explain_header.reference
index 8cab5be1687..5f9e8cfed84 100644
--- a/tests/queries/0_stateless/01508_explain_header.reference
+++ b/tests/queries/0_stateless/01508_explain_header.reference
@@ -1,4 +1,4 @@
-Expression (Projection + Before ORDER BY and SELECT)
+Expression (Projection + Before ORDER BY)
 Header: x UInt8
   SettingQuotaAndLimits (Set limits and quota after reading from storage)
   Header: dummy UInt8
diff --git a/tests/queries/0_stateless/01556_explain_select_with_union_query.reference b/tests/queries/0_stateless/01556_explain_select_with_union_query.reference
index 442754ec344..e4aac5bda16 100644
--- a/tests/queries/0_stateless/01556_explain_select_with_union_query.reference
+++ b/tests/queries/0_stateless/01556_explain_select_with_union_query.reference
@@ -1,252 +1,252 @@
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
   Distinct
     Union
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
   Distinct
     Union
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Distinct
   Union
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
-    Expression (Projection + Before ORDER BY and SELECT)
+    Expression (Projection + Before ORDER BY)
       SettingQuotaAndLimits (Set limits and quota after reading from storage)
         ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
   Distinct
     Union
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
-      Expression (Projection + Before ORDER BY and SELECT)
+      Expression (Projection + Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
 Union
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
-  Expression (Projection + Before ORDER BY and SELECT)
+  Expression (Projection + Before ORDER BY)
     SettingQuotaAndLimits (Set limits and quota after reading from storage)
       ReadFromStorage (SystemOne)
diff --git a/tests/queries/0_stateless/01562_optimize_monotonous_functions_in_order_by.reference b/tests/queries/0_stateless/01562_optimize_monotonous_functions_in_order_by.reference
index 203efdab299..feca2cae5ea 100644
--- a/tests/queries/0_stateless/01562_optimize_monotonous_functions_in_order_by.reference
+++ b/tests/queries/0_stateless/01562_optimize_monotonous_functions_in_order_by.reference
@@ -9,7 +9,7 @@ Expression (Projection)
     MergingSorted (Merge sorted streams for ORDER BY)
       MergeSorting (Merge sorted blocks for ORDER BY)
         PartialSorting (Sort each block for ORDER BY)
-          Expression (Before ORDER BY and SELECT)
+          Expression (Before ORDER BY)
             SettingQuotaAndLimits (Set limits and quota after reading from storage)
               ReadFromStorage (MergeTree)
 SELECT
@@ -21,7 +21,7 @@ LIMIT 10
 Expression (Projection)
   Limit (preliminary LIMIT)
     FinishSorting
-      Expression (Before ORDER BY and SELECT)
+      Expression (Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (MergeTree  with order)
 SELECT
@@ -35,7 +35,7 @@ LIMIT 10
 Expression (Projection)
   Limit (preliminary LIMIT)
     FinishSorting
-      Expression (Before ORDER BY and SELECT)
+      Expression (Before ORDER BY)
         SettingQuotaAndLimits (Set limits and quota after reading from storage)
           ReadFromStorage (MergeTree  with order)
 SELECT
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index e56c6bdeb3a..6eb94856645 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -79,8 +79,15 @@ select number, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 9	9
 select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
 
+-- must work in WHERE if you wrap it in a subquery
+
+select * from (select count(*) over () c from numbers(3)) where c > 0;
+
 -- should work in ORDER BY
 
+1
+2
+3
 select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
 -- also works in ORDER BY if you wrap it in a subquery
@@ -97,24 +104,35 @@ select number, max(number) over (partition by intDiv(number, 3) order by number
 2	2
 select * from (select count(*) over () c from numbers(3)) order by c;
 
--- must work in WHERE if you wrap it in a subquery
+-- Example with window function only in ORDER BY. Here we make a rank of all
+-- numbers sorted descending, and then sort by this rank descending, and must get
+-- the ascending order.
 
 1
 2
 3
-select * from (select count(*) over () c from numbers(3)) where c > 0;
+select * from (select * from numbers(5) order by rand()) order by count() over (order by number desc) desc;
 
--- this one doesn't work yet -- looks like the column names clash, and the
--- window count() is overwritten with aggregate count()
--- select number, count(), count() over (partition by intDiv(number, 3)) from numbers(10) group by number order by count() desc;
+-- Aggregate functions as window function arguments. This query is semantically
+-- the same as the above one, only we replace `number` with
+-- `any(number) group by number` and so on.
+
+0
+1
+2
+3
+4
+select * from (select * from numbers(5) order by rand()) group by number order by sum(any(number)) over (order by min(number) desc) desc;
 
 -- different windows
 -- an explain test would also be helpful, but it's too immature now and I don't
 -- want to change reference all the time
 
 1
+0
 2
 3
+4
 select number, max(number) over (partition by intDiv(number, 3) order by number desc), count(number) over (partition by intDiv(number, 5) order by number) as m from numbers(31) order by number settings max_block_size = 2;
 
 -- two functions over the same window
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index 8b82b8f95fc..5039925aa42 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -24,18 +24,24 @@ select number, quantileExact(number) over (partition by intDiv(number, 3)) q fro
 -- last stage of select, after all other functions.
 select q * 10, quantileExact(number) over (partition by intDiv(number, 3)) q from numbers(10); -- { serverError 47 }
 
+-- must work in WHERE if you wrap it in a subquery
+select * from (select count(*) over () c from numbers(3)) where c > 0;
+
 -- should work in ORDER BY
 select number, max(number) over (partition by intDiv(number, 3) order by number desc) m from numbers(10) order by m desc, number;
 
 -- also works in ORDER BY if you wrap it in a subquery
 select * from (select count(*) over () c from numbers(3)) order by c;
 
--- must work in WHERE if you wrap it in a subquery
-select * from (select count(*) over () c from numbers(3)) where c > 0;
+-- Example with window function only in ORDER BY. Here we make a rank of all
+-- numbers sorted descending, and then sort by this rank descending, and must get
+-- the ascending order.
+select * from (select * from numbers(5) order by rand()) order by count() over (order by number desc) desc;
 
--- this one doesn't work yet -- looks like the column names clash, and the
--- window count() is overwritten with aggregate count()
--- select number, count(), count() over (partition by intDiv(number, 3)) from numbers(10) group by number order by count() desc;
+-- Aggregate functions as window function arguments. This query is semantically
+-- the same as the above one, only we replace `number` with
+-- `any(number) group by number` and so on.
+select * from (select * from numbers(5) order by rand()) group by number order by sum(any(number)) over (order by min(number) desc) desc;
 
 -- different windows
 -- an explain test would also be helpful, but it's too immature now and I don't

From 266d38813920a60968cd59c2cfead6b6f35b3a88 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 11:06:10 +0300
Subject: [PATCH 424/504] Better formatting

---
 contrib/nuraft-cmake/CMakeLists.txt | 52 ++++++++++++++---------------
 1 file changed, 26 insertions(+), 26 deletions(-)

diff --git a/contrib/nuraft-cmake/CMakeLists.txt b/contrib/nuraft-cmake/CMakeLists.txt
index 4a40daaaff8..fbb51626292 100644
--- a/contrib/nuraft-cmake/CMakeLists.txt
+++ b/contrib/nuraft-cmake/CMakeLists.txt
@@ -1,30 +1,30 @@
-set (LIBRARY_DIR ${ClickHouse_SOURCE_DIR}/contrib/NuRaft)
+set(LIBRARY_DIR ${ClickHouse_SOURCE_DIR}/contrib/NuRaft)
 
-set (SRCS
-  ${LIBRARY_DIR}/src/handle_priority.cxx
-  ${LIBRARY_DIR}/src/buffer_serializer.cxx
-  ${LIBRARY_DIR}/src/peer.cxx
-  ${LIBRARY_DIR}/src/global_mgr.cxx
-  ${LIBRARY_DIR}/src/buffer.cxx
-  ${LIBRARY_DIR}/src/asio_service.cxx
-  ${LIBRARY_DIR}/src/handle_client_request.cxx
-  ${LIBRARY_DIR}/src/raft_server.cxx
-  ${LIBRARY_DIR}/src/snapshot.cxx
-  ${LIBRARY_DIR}/src/handle_commit.cxx
-  ${LIBRARY_DIR}/src/error_code.cxx
-  ${LIBRARY_DIR}/src/crc32.cxx
-  ${LIBRARY_DIR}/src/handle_snapshot_sync.cxx
-  ${LIBRARY_DIR}/src/stat_mgr.cxx
-  ${LIBRARY_DIR}/src/handle_join_leave.cxx
-  ${LIBRARY_DIR}/src/handle_user_cmd.cxx
-  ${LIBRARY_DIR}/src/handle_custom_notification.cxx
-  ${LIBRARY_DIR}/src/handle_vote.cxx
-  ${LIBRARY_DIR}/src/launcher.cxx
-  ${LIBRARY_DIR}/src/srv_config.cxx
-  ${LIBRARY_DIR}/src/snapshot_sync_req.cxx
-  ${LIBRARY_DIR}/src/handle_timeout.cxx
-  ${LIBRARY_DIR}/src/handle_append_entries.cxx
-  ${LIBRARY_DIR}/src/cluster_config.cxx
+set(SRCS
+    ${LIBRARY_DIR}/src/handle_priority.cxx
+    ${LIBRARY_DIR}/src/buffer_serializer.cxx
+    ${LIBRARY_DIR}/src/peer.cxx
+    ${LIBRARY_DIR}/src/global_mgr.cxx
+    ${LIBRARY_DIR}/src/buffer.cxx
+    ${LIBRARY_DIR}/src/asio_service.cxx
+    ${LIBRARY_DIR}/src/handle_client_request.cxx
+    ${LIBRARY_DIR}/src/raft_server.cxx
+    ${LIBRARY_DIR}/src/snapshot.cxx
+    ${LIBRARY_DIR}/src/handle_commit.cxx
+    ${LIBRARY_DIR}/src/error_code.cxx
+    ${LIBRARY_DIR}/src/crc32.cxx
+    ${LIBRARY_DIR}/src/handle_snapshot_sync.cxx
+    ${LIBRARY_DIR}/src/stat_mgr.cxx
+    ${LIBRARY_DIR}/src/handle_join_leave.cxx
+    ${LIBRARY_DIR}/src/handle_user_cmd.cxx
+    ${LIBRARY_DIR}/src/handle_custom_notification.cxx
+    ${LIBRARY_DIR}/src/handle_vote.cxx
+    ${LIBRARY_DIR}/src/launcher.cxx
+    ${LIBRARY_DIR}/src/srv_config.cxx
+    ${LIBRARY_DIR}/src/snapshot_sync_req.cxx
+    ${LIBRARY_DIR}/src/handle_timeout.cxx
+    ${LIBRARY_DIR}/src/handle_append_entries.cxx
+    ${LIBRARY_DIR}/src/cluster_config.cxx
 )
 
 

From 236d8185e9d34a3da2a4689dd5c259d355c2bd71 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 11:18:26 +0300
Subject: [PATCH 425/504] Update boost

---
 contrib/boost | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/boost b/contrib/boost
index 809d0b72f42..729dc66b11a 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit 809d0b72f422d63977c066566d3aee1feea022ba
+Subproject commit 729dc66b11a006df6c3b32294635cc2572fea594

From 8d1b6f67b6886731edaf1a13187510d6be06c63b Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 12:28:08 +0300
Subject: [PATCH 426/504] Trying asio standalone

---
 contrib/boost-cmake/CMakeLists.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 54d0ebe1d6c..7c07e97ec08 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -24,6 +24,7 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         add_library (_boost_headers_only INTERFACE)
         add_library (boost::headers_only ALIAS _boost_headers_only)
         target_include_directories (_boost_headers_only SYSTEM BEFORE INTERFACE ${Boost_INCLUDE_DIR})
+        target_compile_definitions (_boost_headers_only PUBLIC BOOST_ASIO_STANDALONE)
 
         add_library (_boost_filesystem INTERFACE)
         add_library (_boost_iostreams INTERFACE)

From ca0a3de581a3e48a64998e9dad728bdc1cb153cd Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 25 Dec 2020 13:35:16 +0300
Subject: [PATCH 427/504] fixed typos

---
 docs/en/sql-reference/functions/other-functions.md | 2 +-
 docs/ru/sql-reference/functions/other-functions.md | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 2dab5f45259..46c422bd336 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1220,7 +1220,7 @@ Result:
 └──────────────────────────────────┘
 ```
 
-Note that NULL values are ignored. 
+Note that `NULL` values are ignored. 
 
 Query:
 
diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index e00c93fcff6..f61a41a6909 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1114,13 +1114,13 @@ finalizeAggregation(state)
 
 **Параметры**
 
--   `state` — Состояние агрегатной функции. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction).
+-   `state` — состояние агрегатной функции. [AggregateFunction](../../sql-reference/data-types/aggregatefunction.md#data-type-aggregatefunction).
 
 **Возвращаемые значения**
 
 -   Значения, которые были агрегированы.
 
-Тип: тип, значения которого были агрегированы.
+Тип: соответствует типу агрегируемых значений.
 
 **Примеры**
 
@@ -1152,7 +1152,7 @@ SELECT finalizeAggregation(( SELECT sumState(number) FROM numbers(10)));
 └──────────────────────────────────┘
 ```
 
-Обратите внимания, что значения NULL игнорируются. 
+Обратите внимание, что значения `NULL` игнорируются. 
 
 Запрос:
 

From dc77b06d30578dd8964c87aacbc7ed73330bec55 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 14:04:39 +0300
Subject: [PATCH 428/504] Fix build

---
 contrib/boost-cmake/CMakeLists.txt  | 1 -
 contrib/nuraft-cmake/CMakeLists.txt | 7 ++++++-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 7c07e97ec08..54d0ebe1d6c 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -24,7 +24,6 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         add_library (_boost_headers_only INTERFACE)
         add_library (boost::headers_only ALIAS _boost_headers_only)
         target_include_directories (_boost_headers_only SYSTEM BEFORE INTERFACE ${Boost_INCLUDE_DIR})
-        target_compile_definitions (_boost_headers_only PUBLIC BOOST_ASIO_STANDALONE)
 
         add_library (_boost_filesystem INTERFACE)
         add_library (_boost_iostreams INTERFACE)
diff --git a/contrib/nuraft-cmake/CMakeLists.txt b/contrib/nuraft-cmake/CMakeLists.txt
index fbb51626292..e5bb7f7d11b 100644
--- a/contrib/nuraft-cmake/CMakeLists.txt
+++ b/contrib/nuraft-cmake/CMakeLists.txt
@@ -30,11 +30,16 @@ set(SRCS
 
 add_library(nuraft ${SRCS})
 
-target_compile_definitions(nuraft PRIVATE USE_BOOST_ASIO=1)
+target_compile_definitions(nuraft PRIVATE USE_BOOST_ASIO=1 BOOST_ASIO_STANDALONE=1)
 
 target_include_directories (nuraft SYSTEM PRIVATE ${LIBRARY_DIR}/include/libnuraft)
 # for some reason include "asio.h" directly without "boost/" prefix.
 target_include_directories (nuraft SYSTEM PRIVATE ${ClickHouse_SOURCE_DIR}/contrib/boost/boost)
+
 target_link_libraries (nuraft PRIVATE boost::headers_only boost::coroutine)
 
+if(OPENSSL_SSL_LIBRARY AND OPENSSL_CRYPTO_LIBRARY)
+    target_link_libraries (nuraft PRIVATE ${OPENSSL_SSL_LIBRARY} ${OPENSSL_CRYPTO_LIBRARY})
+endif()
+
 target_include_directories (nuraft SYSTEM PUBLIC ${LIBRARY_DIR}/include)

From 04d55dc495315fd453608099c09aa7b896cc47fb Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 14:22:48 +0300
Subject: [PATCH 429/504] Feractoring StringHash

---
 src/Functions/ExtractString.h         |  85 ++-------
 src/Functions/FunctionsStringHash.cpp | 259 +++++++++++++++++++++-----
 2 files changed, 222 insertions(+), 122 deletions(-)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index 7406a813143..c4bb2e0fdf7 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -23,112 +23,51 @@ namespace ErrorCodes
 
 // used by FunctionsStringSimilarity and FunctionsStringHash
 // includes extracting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
-template <bool CaseInsensitive>
 struct ExtractStringImpl
 {
-    /// Padding form ColumnsString. It is a number of bytes we can always read starting from pos if pos < end.
-    static constexpr size_t default_padding = 16;
-
-    const size_t shingle_size;
-    const size_t tail_size;
-
-    /// Functions are read `default_padding - (N - 1)` bytes into the buffer. Window of size N is used.
-    /// Read copies `N - 1` last bytes from buffer into beginning, and then reads new bytes.
-    const size_t buffer_size = default_padding + tail_size;
-
-    explicit ExtractStringImpl(size_t shingle_size_)
-            : shingle_size(shingle_size_)
-            , tail_size(shingle_size > default_padding ? shingle_size : roundUpToPowerOfTwoOrZero(shingle_size - 1))
-    {
-        if (shingle_size == 0)
-            throw Exception(ErrorCodes::LOGICAL_ERROR, "Shingle size can't be zero");
-    }
-
     // read a ASCII word
-    static ALWAYS_INLINE inline size_t readOneASCIIWord(PaddedPODArray<UInt8> & word_buf, const char *& pos, const char * end)
+    static ALWAYS_INLINE inline const UInt8 * readOneASCIIWord(const UInt8 *& pos, const UInt8 * end)
     {
         // jump separators
         while (pos < end && !isAlphaNumericASCII(*pos))
             ++pos;
 
         // word start from here
-        const char * word_start = pos;
+        const UInt8 * word_start = pos;
         while (pos < end && isAlphaNumericASCII(*pos))
             ++pos;
 
-        word_buf.assign(word_start, pos);
-        if (CaseInsensitive)
-        {
-            for (auto & symbol : word_buf)
-                symbol = toLowerIfAlphaASCII(symbol);
-        }
-        return word_buf.size();
+        return word_start;
     }
 
     // read one UTF8 word from pos to word
-    static ALWAYS_INLINE inline size_t readOneUTF8Word(PaddedPODArray<UInt32> & word_buf, const char *& pos, const char * end)
+    static ALWAYS_INLINE inline const UInt8 * readOneUTF8Word(const UInt8 *& pos, const UInt8 * end)
     {
         // jump UTF8 separator
         while (pos < end && isUTF8Sep(*pos))
             ++pos;
-        word_buf.clear();
+
         // UTF8 word's character number
+        const UInt8 * word_start = pos;
+
         while (pos < end && !isUTF8Sep(*pos))
-        {
-            word_buf.push_back(readOneUTF8Code(pos, end));
-        }
-        return word_buf.size();
+            readOneUTF8Code(pos, end);
+
+        return word_start;
     }
 
     // we use ASCII non-alphanum character as UTF8 separator
     static ALWAYS_INLINE inline bool isUTF8Sep(const UInt8 c) { return c < 128 && !isAlphaNumericASCII(c); }
 
-    // read one UTF8 character and return it
-    static ALWAYS_INLINE inline UInt32 readOneUTF8Code(const char *& pos, const char * end)
+    // read one UTF8 character
+    static ALWAYS_INLINE inline void readOneUTF8Code(const UInt8 *& pos, const UInt8 * end)
     {
         size_t length = UTF8::seqLength(*pos);
 
         if (pos + length > end)
             length = end - pos;
-        UInt32 res;
-        switch (length)
-        {
-            case 1:
-                res = 0;
-                memcpy(&res, pos, 1);
-                break;
-            case 2:
-                res = 0;
-                memcpy(&res, pos, 2);
-                break;
-            case 3:
-                res = 0;
-                memcpy(&res, pos, 3);
-                break;
-            default:
-                memcpy(&res, pos, 4);
-        }
 
-        if constexpr (CaseInsensitive)
-        {
-            switch (length)
-            {
-                case 4:
-                    res &= ~(1u << (5 + 3 * CHAR_BIT));
-                    [[fallthrough]];
-                case 3:
-                    res &= ~(1u << (5 + 2 * CHAR_BIT));
-                    [[fallthrough]];
-                case 2:
-                    res &= ~(1u);
-                    res &= ~(1u << (5 + CHAR_BIT));
-                    [[fallthrough]];
-                default:
-                    res &= ~(1u << 5);
-            }
-        }
         pos += length;
-        return res;
     }
 };
 }
diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 2f1f900bda9..206cdc6429c 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -46,6 +46,17 @@ struct Hash
 #endif
     }
 
+    static UInt64 crc32u16(UInt64 crc [[maybe_unused]], UInt16 val [[maybe_unused]])
+    {
+#ifdef __SSE4_2__
+        return _mm_crc32_u16(crc, val);
+#elif defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)
+        return __crc32ch(crc, val);
+#else
+        throw Exception("String hash is not implemented without sse4.2 support", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+    }
+
     static UInt64 crc32u8(UInt64 crc [[maybe_unused]], UInt8 val [[maybe_unused]])
     {
 #ifdef __SSE4_2__
@@ -57,18 +68,70 @@ struct Hash
 #endif
     }
 
-    static ALWAYS_INLINE inline UInt64 ngramASCIIHash(const UInt8 * code_points)
+    template <bool CaseInsensitive>
+    static ALWAYS_INLINE inline UInt64 shingleHash(UInt64 crc, const UInt8 * start, size_t size)
     {
-        return crc32u64(-1ULL, unalignedLoad<UInt32>(code_points));
+        if (size & 1)
+        {
+            UInt8 x = *start;
+
+            if constexpr (CaseInsensitive)
+                x |= 0x20u; /// see toLowerIfAlphaASCII from StringUtils.h
+
+            crc = crc32u8(crc, x);
+            --size;
+            ++start;
+        }
+
+        if (size & 2)
+        {
+            UInt16 x = unalignedLoad<UInt16>(start);
+
+            if constexpr (CaseInsensitive)
+                x |= 0x2020u;
+
+            crc = crc32u16(crc, x);
+            size -= 2;
+            start += 2;
+        }
+
+        if (size & 4)
+        {
+            UInt32 x = unalignedLoad<UInt32>(start);
+
+            if constexpr (CaseInsensitive)
+                x |= 0x20202020u;
+
+            crc = crc32u32(crc, x);
+            size -= 4;
+            start += 4;
+        }
+
+        while (size)
+        {
+            UInt64 x = unalignedLoad<UInt64>(start);
+
+            if constexpr (CaseInsensitive)
+                x |= 0x2020202020202020u;
+
+            crc = crc32u64(crc, x);
+            size -= 8;
+            start += 8;
+        }
+
+        return crc;
     }
 
-    static ALWAYS_INLINE inline UInt64 ngramUTF8Hash(const UInt32 * code_points)
+    template <bool CaseInsensitive>
+    static ALWAYS_INLINE inline UInt64 shingleHash(const std::vector<StringRef> & shingle, size_t offset = 0)
     {
         UInt64 crc = -1ULL;
-        crc = crc32u64(crc, code_points[0]);
-        crc = crc32u64(crc, code_points[1]);
-        crc = crc32u64(crc, code_points[2]);
-        return crc;
+
+        for (size_t i = offset; i < shingle.size(); ++i)
+            crc = shingleHash<CaseInsensitive>(crc, shingle[i].data, shingle[i].size);
+
+        for (size_t i = 0; i < offset; ++i)
+            crc = shingleHash<CaseInsensitive>(crc, shingle[i].data, shingle[i].size);
     }
 
     static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
@@ -148,54 +211,82 @@ struct Hash
 template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimHashImpl
 {
-    using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
+    //using StrOp = ExtractStringImpl<N, CaseInsensitive>;
+    //static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
 
-    // SimHash ngram calculate function: String ->UInt64
+    /// Update fingerprint according to hash_value bits.
+    static ALWAYS_INLINE inline void updateFingerVector(Int64 * finger_vec, UInt64 hash_value)
+    {
+        for (size_t i = 0; i < 64; ++i)
+            finger_vec[i] += (hash_value & (1ULL << i)) ? 1 : -1;
+    }
+
+    /// Return a 64 bit value according to finger_vec.
+    static ALWAYS_INLINE inline UInt64 getSimHash(const Int64 * finger_vec)
+    {
+        UInt64 res = 0;
+
+        for (size_t i = 0; i < 64; ++i)
+            if (finger_vec[i] > 0)
+                res |= (1ULL << i);
+
+        return res;
+    }
+
+    // SimHash ngram calculate function: String -> UInt64
     // this function extracting ngram from input string, and maintain a 64-dimensions vector
     // for each ngram, calculate a 64 bit hash value, and update the vector according the hash value
     // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
-    static ALWAYS_INLINE inline UInt64 ngramCalculateHashValue(
-        const char * data,
-        size_t size,
-        size_t (*read_code_points)(CodePoint *, const char *&, const char *),
-        UInt64 (*hash_functor)(const CodePoint *))
+
+    static ALWAYS_INLINE inline UInt64 ngramHash(const UInt8 * data, size_t size, size_t shingle_size)
     {
-        const char * start = data;
-        const char * end = data + size;
-        // fingerprint vector, all dimensions initialized to zero at the first
+        if (size < shingle_size)
+            return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+
         Int64 finger_vec[64] = {};
-        CodePoint cp[simultaneously_codepoints_num] = {};
+        const UInt8 * end = data + size;
 
-        size_t found = read_code_points(cp, start, end);
-        size_t iter = N - 1;
-
-        do
+        for (const UInt8 * pos = data; pos + shingle_size <= end; ++pos)
         {
-            for (; iter + N <= found; ++iter)
-            {
-                // for each ngram, we can calculate an 64 bit hash
-                // then update finger_vec according to this hash value
-                // if the i'th bit is 1, finger_vec[i] plus 1, otherwise minus 1
-                UInt64 hash_value = hash_functor(cp + iter);
-                std::bitset<64> bits(hash_value);
-                for (size_t i = 0; i < 64; ++i)
-                {
-                    finger_vec[i] += ((bits.test(i)) ? 1 : -1);
-                }
-            }
-            iter = 0;
-        } while (start < end && (found = read_code_points(cp, start, end)));
-
-        // finally, we return a 64 bit value according to finger_vec
-        // if finger_vec[i] > 0, the i'th bit of the value is 1, otherwise 0
-        std::bitset<64> res_bit(0u);
-        for (size_t i = 0; i < 64; ++i)
-        {
-            if (finger_vec[i] > 0)
-                res_bit.set(i);
+            UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, pos, shingle_size);
+            updateFingerVector(finger_vec, hash_value);
         }
-        return res_bit.to_ullong();
+
+        return getSimHash(finger_vec);
+    }
+
+    static ALWAYS_INLINE inline UInt64 ngramHashUTF8(
+        size_t shingle_size,
+        const UInt8 * data,
+        size_t size)
+    {
+        const UInt8 * start = data;
+        const UInt8 * end = data + size;
+
+        const UInt8 * word_start = start;
+        const UInt8 * word_end = start;
+
+        for (size_t i = 0; i < shingle_size; ++i)
+        {
+            if (word_end >= end)
+                return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+
+            ExtractStringImpl::readOneUTF8Code(word_end, end);
+        }
+
+        Int64 finger_vec[64] = {};
+
+        while (word_end < end)
+        {
+            ExtractStringImpl::readOneUTF8Code(word_start, word_end);
+            ExtractStringImpl::readOneUTF8Code(word_end, end);
+
+            size_t length = word_end - word_start;
+            UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, word_start, length);
+            updateFingerVector(finger_vec, hash_value);
+        }
+
+        return getSimHash(finger_vec);
     }
 
     // SimHash word shingle calculate function: String -> UInt64
@@ -208,11 +299,81 @@ struct SimHashImpl
     // to calculate the first word shingle hash value
     // 2. next, we extract one word each time, and calculate a new hash value of the new word,then use the latest N hash
     // values to calculate the next word shingle hash value
+
+    static ALWAYS_INLINE inline UInt64 wordShingleHash(
+        const UInt8 * data,
+        size_t size,
+        size_t shingle_size)
+    {
+        const UInt8 * start = data;
+        const UInt8 * end = data + size;
+
+        // A 64 bit vector initialized to zero.
+        Int64 finger_vec[64] = {};
+        // An array to store N words.
+        std::vector<StringRef> words;
+        words.reserve(shingle_size);
+
+        // get first word shingle
+        while (start < end && words.size() < shingle_size)
+        {
+            const UInt8 * word_start = nullptr;
+
+            if constexpr (UTF8)
+                word_start = ExtractStringImpl::readOneUTF8Word(start, end);
+            else
+                word_start = ExtractStringImpl::readOneASCIIWord(start, end);
+
+            size_t length = start - word_start;
+
+            if (length)
+                words.emplace_back(word_start, length);
+        }
+
+        UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);
+        updateFingerVector(finger_vec, hash_value);
+
+        size_t offset = 0;
+        while (start < end)
+        {
+            const UInt8 * word_start = nullptr;
+
+            if constexpr (UTF8)
+                word_start = ExtractStringImpl::readOneUTF8Word(start, end);
+            else
+                word_start = ExtractStringImpl::readOneASCIIWord(start, end);
+
+            size_t length = start - word_start;
+
+            if (length == 0)
+                continue;
+
+            // we need to store the new word hash value to the oldest location.
+            // for example, N = 5, array |a0|a1|a2|a3|a4|, now , a0 is the oldest location,
+            // so we need to store new word hash into location of a0, then ,this array become
+            // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
+            // word hash value into location of a1, then array become |a5|a6|a2|a3|a4|
+            words[offset] = StringRef(word_start, length);
+            ++offset;
+            if (offset >= shingle_size)
+                offset = 0;
+
+            // according to the word hash storation way, in order to not lose the word shingle's
+            // sequence information, when calculation word shingle hash value, we need provide the offset
+            // information, which is the offset of the first word's hash value of the word shingle
+            hash_value = Hash::shingleHash<CaseInsensitive>(words, offset);
+            updateFingerVector(finger_vec, hash_value);
+        }
+
+        return getSimHash(finger_vec);
+    }
+
     static ALWAYS_INLINE inline UInt64 wordShinglesCalculateHashValue(
         const char * data,
         size_t size,
-        size_t (*read_one_word)(PaddedPODArray<CodePoint> &, const char *&, const char *),
-        UInt64 (*hash_functor)(const UInt64 *, size_t, size_t))
+        size_t shingle_size,
+        size_t heap_size,
+        size_t max_word_length)
     {
         const char * start = data;
         const char * end = data + size;
@@ -220,7 +381,7 @@ struct SimHashImpl
         // Also, a 64 bit vector initialized to zero
         Int64 finger_vec[64] = {};
         // a array to store N word hash values
-        UInt64 nword_hashes[N] = {};
+        std::vector<UInt64> word_hashes(shingle_size, 0);
         // word buffer to store one word
         PaddedPODArray<CodePoint> word_buf;
         // get first word shingle

From b4cac6a8a481657ad4e0169fbcf6b9607435e87e Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 15:36:49 +0300
Subject: [PATCH 430/504] Fix unbundled, splitted and disable freebsd build

---
 cmake/find/nuraft.cmake              | 12 ++++++++----
 contrib/boost-cmake/CMakeLists.txt   |  1 +
 docker/packager/unbundled/Dockerfile |  1 +
 3 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/cmake/find/nuraft.cmake b/cmake/find/nuraft.cmake
index 92bf1e200ac..8f67a09c5f6 100644
--- a/cmake/find/nuraft.cmake
+++ b/cmake/find/nuraft.cmake
@@ -11,9 +11,13 @@ if (NOT EXISTS "${ClickHouse_SOURCE_DIR}/contrib/NuRaft/CMakeLists.txt")
     return()
 endif ()
 
-set (USE_NURAFT 1)
-set (NURAFT_LIBRARY nuraft)
+if (NOT OS_FREEBSD)
+    set (USE_NURAFT 1)
+    set (NURAFT_LIBRARY nuraft)
 
-set (NURAFT_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/NuRaft/include")
+    set (NURAFT_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/NuRaft/include")
 
-message (STATUS "Using NuRaft=${USE_NURAFT}: ${NURAFT_INCLUDE_DIR} : ${NURAFT_LIBRARY}")
+    message (STATUS "Using NuRaft=${USE_NURAFT}: ${NURAFT_INCLUDE_DIR} : ${NURAFT_LIBRARY}")
+else()
+    message (${RECONFIGURE_MESSAGE_LEVEL} "Using internal NuRaft library on FreeBSD is not supported")
+endif()
diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 54d0ebe1d6c..616a19b3248 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -215,4 +215,5 @@ if (NOT EXTERNAL_BOOST_FOUND)
     add_library (_boost_coroutine ${SRCS_COROUTINE})
     add_library (boost::coroutine ALIAS _boost_coroutine)
     target_include_directories (_boost_coroutine PRIVATE ${LIBRARY_DIR})
+    target_link_libraries(_boost_coroutine PRIVATE _boost_context)
 endif ()
diff --git a/docker/packager/unbundled/Dockerfile b/docker/packager/unbundled/Dockerfile
index c6ebe95d44a..f640c595f14 100644
--- a/docker/packager/unbundled/Dockerfile
+++ b/docker/packager/unbundled/Dockerfile
@@ -22,6 +22,7 @@ RUN apt-get update \
         libboost-iostreams-dev \
         libboost-regex-dev \
         libboost-context-dev \
+        libboost-coroutine-dev \
         zlib1g-dev \
         liblz4-dev \
         libdouble-conversion-dev \

From ff80c1f3ce32cadbffc0c75780b50ebce071f452 Mon Sep 17 00:00:00 2001
From: Ivan <5627721+abyss7@users.noreply.github.com>
Date: Fri, 25 Dec 2020 15:59:02 +0300
Subject: [PATCH 431/504] Move --antlr arg outside of parser group

---
 tests/clickhouse-test | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/clickhouse-test b/tests/clickhouse-test
index 94ad1a34de8..3e4429449a7 100755
--- a/tests/clickhouse-test
+++ b/tests/clickhouse-test
@@ -886,7 +886,7 @@ if __name__ == '__main__':
     group=parser.add_mutually_exclusive_group(required=False)
     group.add_argument('--shard', action='store_true', default=None, dest='shard', help='Run sharding related tests (required to clickhouse-server listen 127.0.0.2 127.0.0.3)')
     group.add_argument('--no-shard', action='store_false', default=None, dest='shard', help='Do not run shard related tests')
-    group.add_argument('--antlr', action='store_true', default=False, dest='antlr', help='Use new ANTLR parser in tests')
+    parser.add_argument('--antlr', action='store_true', default=False, dest='antlr', help='Use new ANTLR parser in tests')
 
     args = parser.parse_args()
 

From 8ef23dde577fb62976cd540df5a7c40f8af486cb Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 16:06:54 +0300
Subject: [PATCH 432/504] Better asio

---
 contrib/boost-cmake/CMakeLists.txt | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/contrib/boost-cmake/CMakeLists.txt b/contrib/boost-cmake/CMakeLists.txt
index 616a19b3248..b9298f59f2b 100644
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@@ -81,6 +81,10 @@ if (NOT EXTERNAL_BOOST_FOUND)
     add_library (boost::headers_only ALIAS _boost_headers_only)
     target_include_directories (_boost_headers_only SYSTEM BEFORE INTERFACE ${LIBRARY_DIR})
 
+    # asio
+
+    target_compile_definitions (_boost_headers_only INTERFACE BOOST_ASIO_STANDALONE=1)
+
     # iostreams
 
     set (SRCS_IOSTREAMS

From a50615a22bfd432f25d89266bb095b34c4aa884e Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 16:38:04 +0300
Subject: [PATCH 433/504] More correct error code

---
 src/Common/ErrorCodes.cpp                   | 1 +
 src/Storages/StorageReplicatedMergeTree.cpp | 3 ++-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/Common/ErrorCodes.cpp b/src/Common/ErrorCodes.cpp
index f9908e4e612..8b14e17a138 100644
--- a/src/Common/ErrorCodes.cpp
+++ b/src/Common/ErrorCodes.cpp
@@ -530,6 +530,7 @@
     M(561, ZSTD_DECODER_FAILED) \
     M(562, TLD_LIST_NOT_FOUND) \
     M(563, CANNOT_READ_MAP_FROM_TEXT) \
+    M(564, INTERSERVER_SCHEME_DOESNT_MATCH) \
     \
     M(999, KEEPER_EXCEPTION) \
     M(1000, POCO_EXCEPTION) \
diff --git a/src/Storages/StorageReplicatedMergeTree.cpp b/src/Storages/StorageReplicatedMergeTree.cpp
index fe15d5dad29..a87b47a0eb7 100644
--- a/src/Storages/StorageReplicatedMergeTree.cpp
+++ b/src/Storages/StorageReplicatedMergeTree.cpp
@@ -122,6 +122,7 @@ namespace ErrorCodes
     extern const int ILLEGAL_TYPE_OF_ARGUMENT;
     extern const int UNKNOWN_POLICY;
     extern const int NO_SUCH_DATA_PART;
+    extern const int INTERSERVER_SCHEME_DOESNT_MATCH;
 }
 
 namespace ActionLocks
@@ -3509,7 +3510,7 @@ bool StorageReplicatedMergeTree::fetchPart(const String & part_name, const Stora
             if (interserver_scheme != address.scheme)
                 throw Exception("Interserver schemes are different: '" + interserver_scheme
                     + "' != '" + address.scheme + "', can't fetch part from " + address.host,
-                    ErrorCodes::LOGICAL_ERROR);
+                    ErrorCodes::INTERSERVER_SCHEME_DOESNT_MATCH);
 
             return fetcher.fetchPart(
                 metadata_snapshot, part_name, source_replica_path,

From 14e2236a469f7d5350051fbf71a45f842fcdc2c2 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 16:43:01 +0300
Subject: [PATCH 434/504] Fix freebsd one more time

---
 cmake/find/nuraft.cmake | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/cmake/find/nuraft.cmake b/cmake/find/nuraft.cmake
index 8f67a09c5f6..d31fe9c1de8 100644
--- a/cmake/find/nuraft.cmake
+++ b/cmake/find/nuraft.cmake
@@ -19,5 +19,6 @@ if (NOT OS_FREEBSD)
 
     message (STATUS "Using NuRaft=${USE_NURAFT}: ${NURAFT_INCLUDE_DIR} : ${NURAFT_LIBRARY}")
 else()
-    message (${RECONFIGURE_MESSAGE_LEVEL} "Using internal NuRaft library on FreeBSD is not supported")
+    set (USE_NURAFT 0)
+    message (STATUS "Using internal NuRaft library on FreeBSD is not supported")
 endif()

From cdc1f068d2c2a2399be6950ff857431e9ce838eb Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@jakalletti-build.sas.yp-c.yandex.net>
Date: Fri, 25 Dec 2020 17:13:26 +0300
Subject: [PATCH 435/504] skip arcadia

---
 tests/queries/0_stateless/arcadia_skip_list.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tests/queries/0_stateless/arcadia_skip_list.txt b/tests/queries/0_stateless/arcadia_skip_list.txt
index c297698641b..c1f7c14f58a 100644
--- a/tests/queries/0_stateless/arcadia_skip_list.txt
+++ b/tests/queries/0_stateless/arcadia_skip_list.txt
@@ -19,6 +19,7 @@
 00800_low_cardinality_distributed_insert
 00821_distributed_storage_with_join_on
 00829_bitmap_function
+00829_bitmap64_function
 00850_global_join_dups
 00858_issue_4756
 00877_memory_limit_for_new_delete

From 1a8dcc2b279870c128b45f0ad388f18b3508518e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 17:18:41 +0300
Subject: [PATCH 436/504] Feractoring StringHash

---
 src/Functions/FunctionsStringHash.cpp | 580 +++++++++++++-------------
 1 file changed, 290 insertions(+), 290 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 206cdc6429c..7be5b93d69e 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -7,13 +7,13 @@
 
 #include <Core/Defines.h>
 
-#include <bitset>
 #include <functional>
-#include <memory>
 #include <tuple>
 #include <vector>
 #include <common/unaligned.h>
 
+#include <city.h>
+
 namespace DB
 {
 
@@ -133,87 +133,15 @@ struct Hash
         for (size_t i = 0; i < offset; ++i)
             crc = shingleHash<CaseInsensitive>(crc, shingle[i].data, shingle[i].size);
     }
-
-    static ALWAYS_INLINE inline UInt64 wordShinglesHash(const UInt64 * hashes, size_t size, size_t offset)
-    {
-        UInt64 crc1 = -1ULL;
-        UInt64 crc2 = -1ULL;
-
-        for (size_t i = offset; i < size; i += 2)
-            crc1 = crc32u64(crc1, hashes[i]);
-        for (size_t i = offset + 1; i < size; i += 2)
-            crc2 = crc32u64(crc2, hashes[i]);
-
-        if ((size - offset) & 1)
-        {
-            for (size_t i = 0; i < offset; i += 2)
-                crc2 = crc32u64(crc2, hashes[i]);
-            for (size_t i = 1; i < offset; i += 2)
-                crc1 = crc32u64(crc1, hashes[i]);
-        }
-        else
-        {
-            for (size_t i = 0; i < offset; i += 2)
-                crc1 = crc32u64(crc1, hashes[i]);
-            for (size_t i = 1; i < offset; i += 2)
-                crc2 = crc32u64(crc2, hashes[i]);
-        }
-
-        return crc1 | (crc2 << 32u);
-    }
-
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt8 * hashes [[maybe_unused]], size_t K [[maybe_unused]])
-    {
-        UInt64 crc1 = -1ULL;
-        UInt64 crc2 = -1ULL;
-
-        for (size_t i = 0; i < K; i += 2)
-            crc1 = crc32u8(crc1, hashes[i]);
-        for (size_t i = 1; i < K; i += 2)
-            crc2 = crc32u8(crc2, hashes[i]);
-
-        return crc1 | (crc2 << 32u);
-    }
-
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt32 * hashes [[maybe_unused]], size_t K [[maybe_unused]])
-    {
-        UInt64 crc1 = -1ULL;
-        UInt64 crc2 = -1ULL;
-
-        for (size_t i = 0; i < K; i += 2)
-            crc1 = crc32u32(crc1, hashes[i]);
-        for (size_t i = 1; i < K; i += 2)
-            crc2 = crc32u32(crc2, hashes[i]);
-
-        return crc1 | (crc2 << 32u);
-    }
-
-    static ALWAYS_INLINE inline UInt64 hashSum(const UInt64 * hashes, size_t K)
-    {
-        UInt64 crc1 = -1ULL;
-        UInt64 crc2 = -1ULL;
-
-        for (size_t i = 0; i < K; i += 2)
-            crc1 = crc32u64(crc1, hashes[i]);
-        for (size_t i = 1; i < K; i += 2)
-            crc2 = crc32u64(crc2, hashes[i]);
-
-        return crc1 | (crc2 << 32u);
-    }
 };
 
 // SimHash String -> UInt64
-// N: the length of ngram or words shingles
-// CodePoint: UInt8(ASCII) or UInt32(UTF8)
 // UTF8: means ASCII or UTF8, these two parameters CodePoint and UTF8 can only be (UInt8, false) or (UInt32, true)
 // Ngram: means ngram(true) or words shingles(false)
 // CaseInsensitive: means should we consider about letter case or not
-template <size_t N, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
+template <bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimHashImpl
 {
-    //using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    //static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
-
     /// Update fingerprint according to hash_value bits.
     static ALWAYS_INLINE inline void updateFingerVector(Int64 * finger_vec, UInt64 hash_value)
     {
@@ -238,7 +166,7 @@ struct SimHashImpl
     // for each ngram, calculate a 64 bit hash value, and update the vector according the hash value
     // finally return a 64 bit value(UInt64), i'th bit is 1 means vector[i] > 0, otherwise, vector[i] < 0
 
-    static ALWAYS_INLINE inline UInt64 ngramHash(const UInt8 * data, size_t size, size_t shingle_size)
+    static ALWAYS_INLINE inline UInt64 ngramHashASCII(const UInt8 * data, size_t size, size_t shingle_size)
     {
         if (size < shingle_size)
             return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
@@ -255,10 +183,7 @@ struct SimHashImpl
         return getSimHash(finger_vec);
     }
 
-    static ALWAYS_INLINE inline UInt64 ngramHashUTF8(
-        size_t shingle_size,
-        const UInt8 * data,
-        size_t size)
+    static ALWAYS_INLINE inline UInt64 ngramHashUTF8(const UInt8 * data, size_t size, size_t shingle_size)
     {
         const UInt8 * start = data;
         const UInt8 * end = data + size;
@@ -300,10 +225,7 @@ struct SimHashImpl
     // 2. next, we extract one word each time, and calculate a new hash value of the new word,then use the latest N hash
     // values to calculate the next word shingle hash value
 
-    static ALWAYS_INLINE inline UInt64 wordShingleHash(
-        const UInt8 * data,
-        size_t size,
-        size_t shingle_size)
+    static ALWAYS_INLINE inline UInt64 wordShingleHash(const UInt8 * data, size_t size, size_t shingle_size)
     {
         const UInt8 * start = data;
         const UInt8 * end = data + size;
@@ -336,7 +258,7 @@ struct SimHashImpl
         size_t offset = 0;
         while (start < end)
         {
-            const UInt8 * word_start = nullptr;
+            const UInt8 * word_start;
 
             if constexpr (UTF8)
                 word_start = ExtractStringImpl::readOneUTF8Word(start, end);
@@ -368,123 +290,28 @@ struct SimHashImpl
         return getSimHash(finger_vec);
     }
 
-    static ALWAYS_INLINE inline UInt64 wordShinglesCalculateHashValue(
-        const char * data,
-        size_t size,
-        size_t shingle_size,
-        size_t heap_size,
-        size_t max_word_length)
-    {
-        const char * start = data;
-        const char * end = data + size;
-
-        // Also, a 64 bit vector initialized to zero
-        Int64 finger_vec[64] = {};
-        // a array to store N word hash values
-        std::vector<UInt64> word_hashes(shingle_size, 0);
-        // word buffer to store one word
-        PaddedPODArray<CodePoint> word_buf;
-        // get first word shingle
-        for (size_t i = 0; i < N && start < end; ++i)
-        {
-            read_one_word(word_buf, start, end);
-            if (!word_buf.empty())
-            {
-                // for each word, calculate a hash value and stored into the array
-                nword_hashes[i++] = Hash::hashSum(word_buf.data(), word_buf.size());
-            }
-        }
-
-        // calculate the first word shingle hash value
-        UInt64 hash_value = hash_functor(nword_hashes, N, 0);
-        std::bitset<64> first_bits(hash_value);
-        for (size_t i = 0; i < 64; ++i)
-        {
-            finger_vec[i] += ((first_bits.test(i)) ? 1 : -1);
-        }
-
-        size_t offset = 0;
-        while (start < end && read_one_word(word_buf, start, end))
-        {
-            // we need to store the new word hash value to the oldest location.
-            // for example, N = 5, array |a0|a1|a2|a3|a4|, now , a0 is the oldest location,
-            // so we need to store new word hash into location of a0, then ,this array become
-            // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
-            // word hash value into location of a1, then array become |a5|a6|a2|a3|a4|
-            nword_hashes[offset] = Hash::hashSum(word_buf.data(), word_buf.size());
-            offset = (offset + 1) % N;
-            // according to the word hash storation way, in order to not lose the word shingle's
-            // sequence information, when calculation word shingle hash value, we need provide the offset
-            // information, which is the offset of the first word's hash value of the word shingle
-            hash_value = hash_functor(nword_hashes, N, offset);
-            std::bitset<64> bits(hash_value);
-            for (size_t i = 0; i < 64; ++i)
-            {
-                finger_vec[i] += ((bits.test(i)) ? 1 : -1);
-            }
-        }
-
-        std::bitset<64> res_bit(0u);
-        for (size_t i = 0; i < 64; ++i)
-        {
-            if (finger_vec[i] > 0)
-                res_bit.set(i);
-        }
-        return res_bit.to_ullong();
-    }
-
-    static void apply(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, PaddedPODArray<UInt64> & res)
+    static void apply(const ColumnString::Chars & data, const ColumnString::Offsets & offsets, size_t shingle_size, PaddedPODArray<UInt64> & res)
     {
         for (size_t i = 0; i < offsets.size(); ++i)
         {
-            const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
+            const UInt8 * one_data = &data[offsets[i - 1]];
             const size_t data_size = offsets[i] - offsets[i - 1] - 1;
 
             if constexpr (Ngram)
             {
                 if constexpr (!UTF8)
-                    res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
+                    res[i] = ngramHashASCII(one_data, data_size, shingle_size);
                 else
-                    res[i] = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
+                    res[i] = ngramHashUTF8(one_data, data_size, shingle_size);
             }
             else
             {
-                if constexpr (!UTF8)
-                    res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
-                else
-                    res[i] = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+                res[i] = wordShingleHash(one_data, data_size, shingle_size);
             }
         }
     }
 };
 
-template <typename F, size_t K, size_t v>
-class FixedHeap
-{
-public:
-    FixedHeap() = delete;
-
-    explicit FixedHeap(F f_) : f(f_), data_t(std::make_shared<std::vector<UInt64>>(K, v))
-    {
-        std::make_heap(data_t->begin(), data_t->end(), f);
-    }
-
-    void insertAndReplace(UInt64 new_v)
-    {
-        data_t->push_back(new_v);
-        std::push_heap(data_t->begin(), data_t->end(), f);
-        std::pop_heap(data_t->begin(), data_t->end(), f);
-        data_t->pop_back();
-    }
-
-    const UInt64 * data() { return data_t->data(); }
-
-private:
-    F f;
-    std::shared_ptr<std::vector<UInt64>> data_t;
-};
-
-
 // MinHash: String -> Tuple(UInt64, UInt64)
 // for each string, we extract ngram or word shingle,
 // for each ngram or word shingle, calculate a hash value,
@@ -492,136 +319,237 @@ private:
 // and take the K maximum hash values to calculate another hashsum,
 // return this two hashsum: Tuple(hashsum1, hashsum2)
 //
-// N: the length of ngram or words shingles
-// K: the number of minimum hashes and maximum hashes that we keep
-// CodePoint: UInt8(ASCII) or UInt32(UTF8)
 // UTF8: means ASCII or UTF8, these two parameters CodePoint and UTF8 can only be (UInt8, false) or (UInt32, true)
 // Ngram: means ngram(true) or words shingles(false)
 // CaseInsensitive: means should we consider about letter case or not
-template <size_t N, size_t K, typename CodePoint, bool UTF8, bool Ngram, bool CaseInsensitive>
+template <bool UTF8, bool Ngram, bool CaseInsensitive>
 struct MinHashImpl
 {
-    using Less = std::less<size_t>;
-    using Greater = std::greater<size_t>;
-    using MaxHeap = FixedHeap<std::less<size_t>, K, -1ULL>;
-    using MinHeap = FixedHeap<std::greater<size_t>, K, 0>;
-    using StrOp = ExtractStringImpl<N, CaseInsensitive>;
-    static constexpr size_t simultaneously_codepoints_num = StrOp::buffer_size;
-
-    // MinHash ngram calculate function, String -> Tuple(UInt64, UInt64)
-    // we extract ngram from input string, and calculate a hash value for each ngram
-    // then we take the K minimum hash values to calculate a hashsum,
-    // and take the K maximum hash values to calculate another hashsum,
-    // return this two hashsum: Tuple(hashsum1, hashsum2)
-    static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> ngramCalculateHashValue(
-        const char * data,
-        size_t size,
-        size_t (*read_code_points)(CodePoint *, const char *&, const char *),
-        UInt64 (*hash_functor)(const CodePoint *))
+    template<typename Comp>
+    struct Heap
     {
-        const char * start = data;
-        const char * end = data + size;
-        // we just maintain the K minimu and K maximum hash values
-        MaxHeap k_minimum_hashes(Less{});
-        MinHeap k_maximum_hashes(Greater{});
-        CodePoint cp[simultaneously_codepoints_num] = {};
-
-        size_t found = read_code_points(cp, start, end);
-        size_t iter = N - 1;
-
-        do
+        void update(UInt64 hash, StringRef ref, size_t limit)
         {
-            for (; iter + N <= found; ++iter)
-            {
-                auto new_hash = hash_functor(cp + iter);
-                // insert the new hash value into array used to store K minimum value
-                // and K maximum value
-                k_minimum_hashes.insertAndReplace(new_hash);
-                k_maximum_hashes.insertAndReplace(new_hash);
-            }
-            iter = 0;
-        } while (start < end && (found = read_code_points(cp, start, end)));
+            if (values.count(hash))
+                return;
 
-        // calculate hashsum of the K minimum hash values and K maximum hash values
-        UInt64 res1 = Hash::hashSum(k_minimum_hashes.data(), K);
-        UInt64 res2 = Hash::hashSum(k_maximum_hashes.data(), K);
-        return std::make_tuple(res1, res2);
+            values[hash] = ref;
+
+            if (values.size() > limit)
+                values.erase(values.begin());
+        }
+
+        UInt64 getHash()
+        {
+            if (values.empty())
+                return 0;
+
+            auto it = values.begin();
+            UInt64 res = CityHash_v1_0_2::CityHash64(it->first, sizeof(UInt64));
+
+            for (; it != values.end(); ++it)
+                res = CityHash_v1_0_2::Hash128to64(CityHash_v1_0_2::uint128(res, it->first));
+
+            return res;
+        }
+
+        void fill(ColumnTuple & strings)
+        {
+            auto it = values.begin();
+            for (size_t i = 0; i < strings.size(); ++i)
+            {
+                auto & col_string = static_cast<ColumnString &>(strings.getColumn(i));
+                if (it != values.end())
+                {
+                    col_string.insertData(it->second.data, it->second.size);
+                    ++it;
+                }
+                else
+                    col_string.insertDefault();
+            }
+        }
+
+        std::map<UInt64, StringRef, Comp> values;
+    };
+
+    using MaxHeap = Heap<std::less<size_t>>;
+    using MinHeap = Heap<std::greater<size_t>>;
+
+    static ALWAYS_INLINE inline UInt64 ngramHashASCII(
+        MinHeap & min_heap,
+        MaxHeap & max_heap,
+        const UInt8 * data,
+        size_t size,
+        size_t shingle_size,
+        size_t heap_size)
+    {
+        if (size < shingle_size)
+            return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+
+        const UInt8 * end = data + size;
+
+        for (const UInt8 * pos = data; pos + shingle_size <= end; ++pos)
+        {
+            UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, pos, shingle_size);
+
+            // insert the new hash value into array used to store K minimum value
+            // and K maximum value
+            min_heap.update(hash_value, StringRef{pos, shingle_size}, heap_size);
+            max_heap.update(hash_value, StringRef{pos, shingle_size}, heap_size);
+        }
+    }
+
+    static ALWAYS_INLINE inline UInt64 ngramHashUTF8(
+        MinHeap & min_heap,
+        MaxHeap & max_heap,
+        const UInt8 * data,
+        size_t size,
+        size_t shingle_size,
+        size_t heap_size)
+    {
+        const UInt8 * start = data;
+        const UInt8 * end = data + size;
+
+        const UInt8 * word_start = start;
+        const UInt8 * word_end = start;
+
+        for (size_t i = 0; i < shingle_size; ++i)
+        {
+            if (word_end >= end)
+                return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+
+            ExtractStringImpl::readOneUTF8Code(word_end, end);
+        }
+
+        while (word_end < end)
+        {
+            ExtractStringImpl::readOneUTF8Code(word_start, word_end);
+            ExtractStringImpl::readOneUTF8Code(word_end, end);
+
+            size_t length = word_end - word_start;
+            UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, word_start, length);
+
+            min_heap.update(hash_value, StringRef{word_start, length}, heap_size);
+            max_heap.update(hash_value, StringRef{word_start, length}, heap_size);
+        }
     }
 
     // MinHash word shingle hash value calculate function: String ->Tuple(UInt64, UInt64)
     // for each word shingle, we calculate a hash value, but in fact, we just maintain the
     // K minimum and K maximum hash value
-    static ALWAYS_INLINE inline std::tuple<UInt64, UInt64> wordShinglesCalculateHashValue(
-        const char * data,
+    static ALWAYS_INLINE inline UInt64 wordShingleHash(
+        MinHeap & min_heap,
+        MaxHeap & max_heap,
+        const UInt8 * data,
         size_t size,
-        size_t (*read_one_word)(PaddedPODArray<CodePoint> &, const char *&, const char *),
-        UInt64 (*hash_functor)(const UInt64 *, size_t, size_t))
+        size_t shingle_size,
+        size_t heap_size)
     {
-        const char * start = data;
-        const char * end = start + size;
-        // also we just store the K minimu and K maximum hash values
-        MaxHeap k_minimum_hashes(Less{});
-        MinHeap k_maximum_hashes(Greater{});
-        // array to store n word hashes
-        UInt64 nword_hashes[N] = {};
-        // word buffer to store one word
-        PaddedPODArray<CodePoint> word_buf;
-        // how word shingle hash value calculation and word hash storation is same as we
-        // have descripted in SimHash wordShinglesCalculateHashValue function
-        for (size_t i = 0; i < N && start < end; ++i)
+        const UInt8 * start = data;
+        const UInt8 * end = data + size;
+
+        // An array to store N words.
+        std::vector<StringRef> words;
+        words.reserve(shingle_size);
+
+        // get first word shingle
+        while (start < end && words.size() < shingle_size)
         {
-            read_one_word(word_buf, start, end);
-            if (!word_buf.empty())
-            {
-                nword_hashes[i++] = Hash::hashSum(word_buf.data(), word_buf.size());
-            }
+            const UInt8 * word_start = nullptr;
+
+            if constexpr (UTF8)
+                word_start = ExtractStringImpl::readOneUTF8Word(start, end);
+            else
+                word_start = ExtractStringImpl::readOneASCIIWord(start, end);
+
+            size_t length = start - word_start;
+
+            if (length)
+                words.emplace_back(word_start, length);
         }
 
-        auto new_hash = hash_functor(nword_hashes, N, 0);
-        k_minimum_hashes.insertAndReplace(new_hash);
-        k_maximum_hashes.insertAndReplace(new_hash);
+        UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);
+        {
+            const char * shingle_start = words.front().data;
+            const char * shingle_end = words.back().data + words.back().size;
+            StringRef ref(shingle_start, shingle_end - shingle_start);
+            min_heap.update(hash_value, ref, heap_size);
+            max_heap.update(hash_value, ref, heap_size);
+        }
 
         size_t offset = 0;
-        while (start < end && read_one_word(word_buf, start, end))
+        while (start < end)
         {
-            nword_hashes[offset] = Hash::hashSum(word_buf.data(), word_buf.size());
-            offset = (offset + 1) % N;
-            new_hash = hash_functor(nword_hashes, N, offset);
-            k_minimum_hashes.insertAndReplace(new_hash);
-            k_maximum_hashes.insertAndReplace(new_hash);
-        }
+            const UInt8 * word_start;
 
-        // calculate hashsum
-        UInt64 res1 = Hash::hashSum(k_minimum_hashes.data(), K);
-        UInt64 res2 = Hash::hashSum(k_maximum_hashes.data(), K);
-        return std::make_tuple(res1, res2);
+            if constexpr (UTF8)
+                word_start = ExtractStringImpl::readOneUTF8Word(start, end);
+            else
+                word_start = ExtractStringImpl::readOneASCIIWord(start, end);
+
+            size_t length = start - word_start;
+
+            if (length == 0)
+                continue;
+
+            words[offset] = StringRef(word_start, length);
+            const char * shingle_end = words[offset].data + length;
+
+            ++offset;
+            if (offset >= shingle_size)
+                offset = 0;
+
+            const char * shingle_start = words[offset].data;
+
+            hash_value = Hash::shingleHash<CaseInsensitive>(words, offset);
+            StringRef ref(shingle_start, shingle_end - shingle_start);
+            min_heap.update(hash_value, ref, heap_size);
+            max_heap.update(hash_value, ref, heap_size);
+        }
     }
 
     static void apply(
         const ColumnString::Chars & data,
         const ColumnString::Offsets & offsets,
-        PaddedPODArray<UInt64> & res1,
-        PaddedPODArray<UInt64> & res2)
+        size_t shingle_size,
+        size_t heap_size,
+        PaddedPODArray<UInt64> * res1,
+        PaddedPODArray<UInt64> * res2,
+        ColumnTuple * res1_strings,
+        ColumnTuple * res2_strings)
     {
+        MinHeap min_heap;
+        MaxHeap max_heap;
+
         for (size_t i = 0; i < offsets.size(); ++i)
         {
             const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
             const size_t data_size = offsets[i] - offsets[i - 1] - 1;
 
+            min_heap.values.clear();
+            max_heap.values.clear();
+
             if constexpr (Ngram)
             {
                 if constexpr (!UTF8)
-                    std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readASCIICodePoints, Hash::ngramASCIIHash);
+                    ngramHashASCII(min_heap, max_heap, one_data, data_size, shingle_size, heap_size);
                 else
-                    std::tie(res1[i], res2[i]) = ngramCalculateHashValue(one_data, data_size, StrOp::readUTF8CodePoints, Hash::ngramUTF8Hash);
+                    ngramHashUTF8(min_heap, max_heap, one_data, data_size, shingle_size, heap_size);
             }
             else
             {
-                if constexpr (!UTF8)
-                    std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneASCIIWord, Hash::wordShinglesHash);
-                else
-                    std::tie(res1[i], res2[i]) = wordShinglesCalculateHashValue(one_data, data_size, StrOp::readOneUTF8Word, Hash::wordShinglesHash);
+                wordShingleHash(min_heap, max_heap, one_data, data_size, shingle_size, heap_size);
             }
+
+            if (res1)
+                (*res1)[i] = min_heap.getHash();
+            if (res2)
+                (*res2)[i] = max_heap.getHash();
+
+            if (res1_strings)
+                min_heap.fill(*res1_strings);
+            if (res2_strings)
+                max_heap.fill(*res2_strings);
         }
     }
 };
@@ -706,48 +634,111 @@ struct NameWordShingleMinHashCaseInsensitiveUTF8
     static constexpr auto name = "wordShingleMinHashCaseInsensitiveUTF8";
 };
 
+struct NameNgramMinHashArg
+{
+    static constexpr auto name = "ngramMinHashArg";
+};
+
+struct NameNgramMinHashArgCaseInsensitive
+{
+    static constexpr auto name = "ngramMinHashArgCaseInsensitive";
+};
+
+struct NameNgramMinHashArgUTF8
+{
+    static constexpr auto name = "ngramMinHashArgUTF8";
+};
+
+struct NameNgramMinHashArgCaseInsensitiveUTF8
+{
+    static constexpr auto name = "ngramMinHashArgCaseInsensitiveUTF8";
+};
+
+struct NameWordShingleMinHashArg
+{
+    static constexpr auto name = "wordShingleMinHashArg";
+};
+
+struct NameWordShingleMinHashArgCaseInsensitive
+{
+    static constexpr auto name = "wordShingleMinHashArgCaseInsensitive";
+};
+
+struct NameWordShingleMinHashArgUTF8
+{
+    static constexpr auto name = "wordShingleMinHashArgUTF8";
+};
+
+struct NameWordShingleMinHashArgCaseInsensitiveUTF8
+{
+    static constexpr auto name = "wordShingleMinHashArgCaseInsensitiveUTF8";
+};
+
 // SimHash
-using FunctionNgramSimHash = FunctionsStringHash<SimHashImpl<4, UInt8, false, true, false>, NameNgramSimHash, true>;
+using FunctionNgramSimHash = FunctionsStringHash<SimHashImpl<false, true, false>, NameNgramSimHash, true>;
 
 using FunctionNgramSimHashCaseInsensitive
-    = FunctionsStringHash<SimHashImpl<4, UInt8, false, true, true>, NameNgramSimHashCaseInsensitive, true>;
+    = FunctionsStringHash<SimHashImpl<false, true, true>, NameNgramSimHashCaseInsensitive, true>;
 
-using FunctionNgramSimHashUTF8 = FunctionsStringHash<SimHashImpl<3, UInt32, true, true, false>, NameNgramSimHashUTF8, true>;
+using FunctionNgramSimHashUTF8 = FunctionsStringHash<SimHashImpl<true, true, false>, NameNgramSimHashUTF8, true>;
 
 using FunctionNgramSimHashCaseInsensitiveUTF8
-    = FunctionsStringHash<SimHashImpl<3, UInt32, true, true, true>, NameNgramSimHashCaseInsensitiveUTF8, true>;
+    = FunctionsStringHash<SimHashImpl<true, true, true>, NameNgramSimHashCaseInsensitiveUTF8, true>;
 
-using FunctionWordShingleSimHash = FunctionsStringHash<SimHashImpl<3, UInt8, false, false, false>, NameWordShingleSimHash, true>;
+using FunctionWordShingleSimHash = FunctionsStringHash<SimHashImpl<false, false, false>, NameWordShingleSimHash, true>;
 
 using FunctionWordShingleSimHashCaseInsensitive
-    = FunctionsStringHash<SimHashImpl<3, UInt8, false, false, true>, NameWordShingleSimHashCaseInsensitive, true>;
+    = FunctionsStringHash<SimHashImpl<false, false, true>, NameWordShingleSimHashCaseInsensitive, true>;
 
-using FunctionWordShingleSimHashUTF8 = FunctionsStringHash<SimHashImpl<3, UInt32, true, false, false>, NameWordShingleSimHashUTF8, true>;
+using FunctionWordShingleSimHashUTF8 = FunctionsStringHash<SimHashImpl<true, false, false>, NameWordShingleSimHashUTF8, true>;
 
 using FunctionWordShingleSimHashCaseInsensitiveUTF8
-    = FunctionsStringHash<SimHashImpl<3, UInt32, true, false, true>, NameWordShingleSimHashCaseInsensitiveUTF8, true>;
+    = FunctionsStringHash<SimHashImpl<true, false, true>, NameWordShingleSimHashCaseInsensitiveUTF8, true>;
 
 // MinHash
-using FunctionNgramMinHash = FunctionsStringHash<MinHashImpl<4, 6, UInt8, false, true, false>, NameNgramMinHash, false>;
+using FunctionNgramMinHash = FunctionsStringHash<MinHashImpl<false, true, false>, NameNgramMinHash, false>;
 
 using FunctionNgramMinHashCaseInsensitive
-    = FunctionsStringHash<MinHashImpl<4, 6, UInt8, false, true, true>, NameNgramMinHashCaseInsensitive, false>;
+    = FunctionsStringHash<MinHashImpl<false, true, true>, NameNgramMinHashCaseInsensitive, false>;
 
-using FunctionNgramMinHashUTF8 = FunctionsStringHash<MinHashImpl<4, 6, UInt32, true, true, false>, NameNgramMinHashUTF8, false>;
+using FunctionNgramMinHashUTF8 = FunctionsStringHash<MinHashImpl<true, true, false>, NameNgramMinHashUTF8, false>;
 
 using FunctionNgramMinHashCaseInsensitiveUTF8
-    = FunctionsStringHash<MinHashImpl<4, 6, UInt32, true, true, true>, NameNgramMinHashCaseInsensitiveUTF8, false>;
+    = FunctionsStringHash<MinHashImpl<true, true, true>, NameNgramMinHashCaseInsensitiveUTF8, false>;
 
-using FunctionWordShingleMinHash = FunctionsStringHash<MinHashImpl<3, 6, UInt8, false, false, false>, NameWordShingleMinHash, false>;
+using FunctionWordShingleMinHash = FunctionsStringHash<MinHashImpl<false, false, false>, NameWordShingleMinHash, false>;
 
 using FunctionWordShingleMinHashCaseInsensitive
-    = FunctionsStringHash<MinHashImpl<3, 6, UInt8, false, false, true>, NameWordShingleMinHashCaseInsensitive, false>;
+    = FunctionsStringHash<MinHashImpl<false, false, true>, NameWordShingleMinHashCaseInsensitive, false>;
 
 using FunctionWordShingleMinHashUTF8
-    = FunctionsStringHash<MinHashImpl<3, 6, UInt32, true, false, false>, NameWordShingleMinHashUTF8, false>;
+    = FunctionsStringHash<MinHashImpl<true, false, false>, NameWordShingleMinHashUTF8, false>;
 
 using FunctionWordShingleMinHashCaseInsensitiveUTF8
-    = FunctionsStringHash<MinHashImpl<3, 6, UInt32, true, false, true>, NameWordShingleMinHashCaseInsensitiveUTF8, false>;
+    = FunctionsStringHash<MinHashImpl<true, false, true>, NameWordShingleMinHashCaseInsensitiveUTF8, false>;
+
+// MinHasArg
+
+using FunctionNgramMinHashArg = FunctionsStringHash<MinHashImpl<false, true, false>, NameNgramMinHashArg, false>;
+
+using FunctionNgramMinHashArgCaseInsensitive
+= FunctionsStringHash<MinHashImpl<false, true, true>, NameNgramMinHashArgCaseInsensitive, false>;
+
+using FunctionNgramMinHashArgUTF8 = FunctionsStringHash<MinHashImpl<true, true, false>, NameNgramMinHashArgUTF8, false>;
+
+using FunctionNgramMinHashArgCaseInsensitiveUTF8
+= FunctionsStringHash<MinHashImpl<true, true, true>, NameNgramMinHashArgCaseInsensitiveUTF8, false>;
+
+using FunctionWordShingleMinHashArg = FunctionsStringHash<MinHashImpl<false, false, false>, NameWordShingleMinHashArg, false>;
+
+using FunctionWordShingleMinHashArgCaseInsensitive
+= FunctionsStringHash<MinHashImpl<false, false, true>, NameWordShingleMinHashArgCaseInsensitive, false>;
+
+using FunctionWordShingleMinHashArgUTF8
+= FunctionsStringHash<MinHashImpl<true, false, false>, NameWordShingleMinHashArgUTF8, false>;
+
+using FunctionWordShingleMinHashArgCaseInsensitiveUTF8
+= FunctionsStringHash<MinHashImpl<true, false, true>, NameWordShingleMinHashArgCaseInsensitiveUTF8, false>;
 
 void registerFunctionsStringHash(FunctionFactory & factory)
 {
@@ -768,6 +759,15 @@ void registerFunctionsStringHash(FunctionFactory & factory)
     factory.registerFunction<FunctionWordShingleMinHashCaseInsensitive>();
     factory.registerFunction<FunctionWordShingleMinHashUTF8>();
     factory.registerFunction<FunctionWordShingleMinHashCaseInsensitiveUTF8>();
+
+    factory.registerFunction<FunctionNgramMinHashArg>();
+    factory.registerFunction<FunctionNgramMinHashArgCaseInsensitive>();
+    factory.registerFunction<FunctionNgramMinHashArgUTF8>();
+    factory.registerFunction<FunctionNgramMinHashArgCaseInsensitiveUTF8>();
+    factory.registerFunction<FunctionWordShingleMinHashArg>();
+    factory.registerFunction<FunctionWordShingleMinHashArgCaseInsensitive>();
+    factory.registerFunction<FunctionWordShingleMinHashArgUTF8>();
+    factory.registerFunction<FunctionWordShingleMinHashArgCaseInsensitiveUTF8>();
 }
 }
 

From 94f4a15cd0b74ad50160fbcc5cca56d98010dbd7 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 18:19:30 +0300
Subject: [PATCH 437/504] Feractoring StringHash

---
 src/Functions/FunctionsStringHash.cpp |  16 ++--
 src/Functions/FunctionsStringHash.h   | 121 ++++++++++++++++++++++++--
 2 files changed, 120 insertions(+), 17 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 7be5b93d69e..b36cc331dcb 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -719,26 +719,26 @@ using FunctionWordShingleMinHashCaseInsensitiveUTF8
 
 // MinHasArg
 
-using FunctionNgramMinHashArg = FunctionsStringHash<MinHashImpl<false, true, false>, NameNgramMinHashArg, false>;
+using FunctionNgramMinHashArg = FunctionsStringHash<MinHashImpl<false, true, false>, NameNgramMinHashArg, false, true>;
 
 using FunctionNgramMinHashArgCaseInsensitive
-= FunctionsStringHash<MinHashImpl<false, true, true>, NameNgramMinHashArgCaseInsensitive, false>;
+= FunctionsStringHash<MinHashImpl<false, true, true>, NameNgramMinHashArgCaseInsensitive, false, true>;
 
-using FunctionNgramMinHashArgUTF8 = FunctionsStringHash<MinHashImpl<true, true, false>, NameNgramMinHashArgUTF8, false>;
+using FunctionNgramMinHashArgUTF8 = FunctionsStringHash<MinHashImpl<true, true, false>, NameNgramMinHashArgUTF8, false, true>;
 
 using FunctionNgramMinHashArgCaseInsensitiveUTF8
-= FunctionsStringHash<MinHashImpl<true, true, true>, NameNgramMinHashArgCaseInsensitiveUTF8, false>;
+= FunctionsStringHash<MinHashImpl<true, true, true>, NameNgramMinHashArgCaseInsensitiveUTF8, false, true>;
 
-using FunctionWordShingleMinHashArg = FunctionsStringHash<MinHashImpl<false, false, false>, NameWordShingleMinHashArg, false>;
+using FunctionWordShingleMinHashArg = FunctionsStringHash<MinHashImpl<false, false, false>, NameWordShingleMinHashArg, false, true>;
 
 using FunctionWordShingleMinHashArgCaseInsensitive
-= FunctionsStringHash<MinHashImpl<false, false, true>, NameWordShingleMinHashArgCaseInsensitive, false>;
+= FunctionsStringHash<MinHashImpl<false, false, true>, NameWordShingleMinHashArgCaseInsensitive, false, true>;
 
 using FunctionWordShingleMinHashArgUTF8
-= FunctionsStringHash<MinHashImpl<true, false, false>, NameWordShingleMinHashArgUTF8, false>;
+= FunctionsStringHash<MinHashImpl<true, false, false>, NameWordShingleMinHashArgUTF8, false, true>;
 
 using FunctionWordShingleMinHashArgCaseInsensitiveUTF8
-= FunctionsStringHash<MinHashImpl<true, false, true>, NameWordShingleMinHashArgCaseInsensitiveUTF8, false>;
+= FunctionsStringHash<MinHashImpl<true, false, true>, NameWordShingleMinHashArgCaseInsensitiveUTF8, false, true>;
 
 void registerFunctionsStringHash(FunctionFactory & factory)
 {
diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index 979f2bd8a9d..2fe5418d94a 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -7,6 +7,7 @@
 #include <DataTypes/DataTypeFactory.h>
 #include <DataTypes/DataTypeTuple.h>
 #include <DataTypes/DataTypesNumber.h>
+#include <DataTypes/DataTypeString.h>
 #include <Functions/FunctionHelpers.h>
 #include <Functions/IFunctionImpl.h>
 
@@ -15,33 +16,107 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+    extern const int TOO_MANY_ARGUMENTS_FOR_FUNCTION;
+    extern const int TOO_FEW_ARGUMENTS_FOR_FUNCTION;
 }
 
 // FunctionStringHash
 // Simhash: String -> UInt64
 // Minhash: String -> (UInt64, UInt64)
-template <typename Impl, typename Name, bool is_simhash>
+template <typename Impl, typename Name, bool is_simhash, bool is_arg = false>
 class FunctionsStringHash : public IFunction
 {
 public:
     static constexpr auto name = Name::name;
+    static constexpr size_t default_shingle_size = 3;
+    static constexpr size_t default_num_hashes = 6;
 
     static FunctionPtr create(const Context &) { return std::make_shared<FunctionsStringHash>(); }
 
     String getName() const override { return name; }
 
-    size_t getNumberOfArguments() const override { return 1; }
+    size_t getNumberOfArguments() const override { return 0; }
+    bool isVariadic() const override { return true; }
 
-    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    ColumnNumbers getArgumentsThatAreAlwaysConstant() const override
     {
-        if (!isString(arguments[0]))
+        if constexpr (is_simhash)
+            return {1};
+        else
+            return {1, 2};
+    }
+
+    DataTypePtr getReturnTypeImpl(const ColumnsWithTypeAndName & arguments) const override
+    {
+        if (arguments.empty())
+            throw Exception(ErrorCodes::TOO_FEW_ARGUMENTS_FOR_FUNCTION, "Function {} expect at least one argument", getName());
+
+        if (!isString(arguments[0].type))
             throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
-                            "Function {} expect single String argument, got {}", getName(), arguments[0]->getName());
+                            "First argument of function {} must be String, got {}", getName(), arguments[0].type->getName());
+
+        size_t shingle_size = default_shingle_size;
+
+        if (arguments.size() > 1)
+        {
+            if (!isUnsignedInteger(arguments[1].type))
+                throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                                "Second argument (shingle size) of function {} must be unsigned integer, got {}",
+                                getName(), arguments[1].type->getName());
+
+            if (!arguments[1].column)
+                throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                                "Second argument (shingle size) of function {} must be constant", getName());
+
+            shingle_size = arguments[1].column->getUInt(0);
+        }
+
+        size_t num_hashes = default_num_hashes;
+
+        if (arguments.size() > 2)
+        {
+            if constexpr (is_simhash)
+                throw Exception(ErrorCodes::TOO_MANY_ARGUMENTS_FOR_FUNCTION,
+                                "Function {} expect no more then two arguments (text, shingle size), got {}",
+                                getName(), arguments.size());
+
+            if (!isUnsignedInteger(arguments[2].type))
+                throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                                "Third argument (num hashes) of function {} must be unsigned integer, got {}",
+                                getName(), arguments[2].type->getName());
+
+            if (!arguments[2].column)
+                throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                                "Third argument (num hashes) of function {} must be constant", getName());
+
+            num_hashes = arguments[2].column->getUInt(0);
+        }
+
+        if (arguments.size() > 3)
+        {
+            throw Exception(ErrorCodes::TOO_MANY_ARGUMENTS_FOR_FUNCTION,
+                            "Function {} expect no more then three arguments (text, shingle size, num hashes), got {}",
+                            getName(), arguments.size());
+        }
+
+        if (shingle_size == 0)
+            throw Exception("Second argument (shingle size) of function {} cannot be zero", getName());
+
+        if (num_hashes == 0)
+            throw Exception("Third argument (num hashes) of function {} cannot be zero", getName());
 
         auto type = std::make_shared<DataTypeUInt64>();
         if constexpr (is_simhash)
             return type;
 
+        if constexpr (is_arg)
+        {
+            DataTypePtr string_type = std::make_shared<DataTypeString>();
+            DataTypes types(num_hashes, string_type);
+            auto tuple_type = std::make_shared<DataTypeTuple>(types);
+            return std::make_shared<DataTypeTuple>(DataTypes{tuple_type, tuple_type});
+        }
+
         return std::make_shared<DataTypeTuple>(DataTypes{type, type});
     }
 
@@ -51,19 +126,47 @@ public:
     {
         const ColumnPtr & column = arguments[0].column;
 
+        size_t shingle_size = default_shingle_size;
+        size_t num_hashes = default_num_hashes;
+
+        if (arguments.size() > 1)
+            shingle_size = arguments[1].column->getUInt(0);
+
+        if (arguments.size() > 2)
+            num_hashes = arguments[2].column->getUInt(0);
+
         if constexpr (is_simhash)
         {
-            // non const string, const case is handled by useDefaultImplementationForConstants.
             auto col_res = ColumnVector<UInt64>::create();
             auto & vec_res = col_res->getData();
             vec_res.resize(column->size());
             const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
-            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_res);
+            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), shingle_size, vec_res);
             return col_res;
         }
+        else if constexpr (is_arg) // Min hash arg
+        {
+            MutableColumns min_columns(num_hashes);
+            MutableColumns max_columns(num_hashes);
+            for (size_t i = 0; i < num_hashes; ++i)
+            {
+                min_columns[i] = ColumnString::create();
+                max_columns[i] = ColumnString::create();
+            }
+
+            auto min_tuple = ColumnTuple::create(std::move(min_columns));
+            auto max_tuple = ColumnTuple::create(std::move(max_columns));
+
+            const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
+            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), shingle_size, num_hashes, nullptr, nullptr, &min_tuple, &max_tuple);
+
+            MutableColumns tuple_columns;
+            tuple_columns.emplace_back(std::move(min_tuple));
+            tuple_columns.emplace_back(std::move(max_tuple));
+            return ColumnTuple::create(std::move(tuple_columns));
+        }
         else // Min hash
         {
-            // non const string
             auto col_h1 = ColumnVector<UInt64>::create();
             auto col_h2 = ColumnVector<UInt64>::create();
             auto & vec_h1 = col_h1->getData();
@@ -71,7 +174,7 @@ public:
             vec_h1.resize(column->size());
             vec_h2.resize(column->size());
             const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
-            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), vec_h1, vec_h2);
+            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), shingle_size, num_hashes, &vec_h1, &vec_h2, nullptr, nullptr);
             MutableColumns tuple_columns;
             tuple_columns.emplace_back(std::move(col_h1));
             tuple_columns.emplace_back(std::move(col_h2));

From 4af0d73446a8845bf7349c332910562eb0876123 Mon Sep 17 00:00:00 2001
From: robot-clickhouse <robot-clickhouse@yandex-team.ru>
Date: Fri, 25 Dec 2020 18:38:21 +0300
Subject: [PATCH 438/504] Update version_date.tsv after release 20.8.11.17

---
 utils/list-versions/version_date.tsv | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/list-versions/version_date.tsv b/utils/list-versions/version_date.tsv
index e7e4b78c99b..002dc5965d3 100644
--- a/utils/list-versions/version_date.tsv
+++ b/utils/list-versions/version_date.tsv
@@ -18,6 +18,7 @@ v20.9.5.5-stable	2020-11-13
 v20.9.4.76-stable	2020-10-29
 v20.9.3.45-stable	2020-10-09
 v20.9.2.20-stable	2020-09-22
+v20.8.11.17-lts	2020-12-25
 v20.8.10.13-lts	2020-12-24
 v20.8.9.6-lts	2020-12-10
 v20.8.8.2-lts	2020-12-07

From 2c1d4439449dccb5ad45b1fb89e1d02e6e459e1b Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 18:47:17 +0300
Subject: [PATCH 439/504] Control alter sequential execution more correctly

---
 .../MergeTree/ReplicatedMergeTreeAltersSequence.cpp      | 9 ++++++---
 .../MergeTree/ReplicatedMergeTreeAltersSequence.h        | 5 ++---
 src/Storages/MergeTree/ReplicatedMergeTreeQueue.cpp      | 2 +-
 3 files changed, 9 insertions(+), 7 deletions(-)

diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp b/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp
index 81f80566990..34f5e06852c 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp
@@ -23,11 +23,13 @@ void ReplicatedMergeTreeAltersSequence::addMutationForAlter(int alter_version, s
 }
 
 void ReplicatedMergeTreeAltersSequence::addMetadataAlter(
-    int alter_version, bool have_mutation, std::lock_guard<std::mutex> & /*state_lock*/)
+    int alter_version, std::lock_guard<std::mutex> & /*state_lock*/)
 {
+    /// Data alter (mutation) always added before. See ReplicatedMergeTreeQueue::pullLogsToQueue.
+    /// So mutation alredy added to this sequence or doesn't exist.
     if (!queue_state.count(alter_version))
-        queue_state.emplace(alter_version, AlterState{.metadata_finished=false, .data_finished=!have_mutation});
-    else /// Data alter can be added before.
+        queue_state.emplace(alter_version, AlterState{.metadata_finished=false, .data_finished=true});
+    else
         queue_state[alter_version].metadata_finished = false;
 }
 
@@ -43,6 +45,7 @@ void ReplicatedMergeTreeAltersSequence::finishMetadataAlter(int alter_version, s
         queue_state.erase(alter_version);
     else
         queue_state[alter_version].metadata_finished = true;
+
 }
 
 void ReplicatedMergeTreeAltersSequence::finishDataAlter(int alter_version, std::lock_guard<std::mutex> & /*state_lock*/)
diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.h b/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.h
index b675a85581e..e5d3dd0a737 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.h
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.h
@@ -38,9 +38,8 @@ public:
     /// Add mutation for alter (alter data stage).
     void addMutationForAlter(int alter_version, std::lock_guard<std::mutex> & /*state_lock*/);
 
-    /// Add metadata for alter (alter metadata stage). If have_mutation=true, than we expect, that
-    /// corresponding mutation will be added.
-    void addMetadataAlter(int alter_version, bool have_mutation, std::lock_guard<std::mutex> & /*state_lock*/);
+    /// Add metadata for alter (alter metadata stage).
+    void addMetadataAlter(int alter_version, std::lock_guard<std::mutex> & /*state_lock*/);
 
     /// Finish metadata alter. If corresponding data alter finished, than we can remove
     /// alter from sequence.
diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeQueue.cpp b/src/Storages/MergeTree/ReplicatedMergeTreeQueue.cpp
index 9719df9726f..d87c6b75fbf 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeQueue.cpp
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeQueue.cpp
@@ -158,7 +158,7 @@ void ReplicatedMergeTreeQueue::insertUnlocked(
     if (entry->type == LogEntry::ALTER_METADATA)
     {
         LOG_TRACE(log, "Adding alter metadata version {} to the queue", entry->alter_version);
-        alter_sequence.addMetadataAlter(entry->alter_version, entry->have_mutation, state_lock);
+        alter_sequence.addMetadataAlter(entry->alter_version, state_lock);
     }
 }
 

From 0950c16fb23592fabd90b7663fa8af70d9fbeb49 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 18:49:29 +0300
Subject: [PATCH 440/504] Fix unexpected change

---
 src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp | 1 -
 1 file changed, 1 deletion(-)

diff --git a/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp b/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp
index 34f5e06852c..c2619317776 100644
--- a/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp
+++ b/src/Storages/MergeTree/ReplicatedMergeTreeAltersSequence.cpp
@@ -45,7 +45,6 @@ void ReplicatedMergeTreeAltersSequence::finishMetadataAlter(int alter_version, s
         queue_state.erase(alter_version);
     else
         queue_state[alter_version].metadata_finished = true;
-
 }
 
 void ReplicatedMergeTreeAltersSequence::finishDataAlter(int alter_version, std::lock_guard<std::mutex> & /*state_lock*/)

From b900ddd1c30a1d3e660e5f07eb999a58aa4524ca Mon Sep 17 00:00:00 2001
From: George <gyuton@yandex-team.ru>
Date: Fri, 25 Dec 2020 18:50:38 +0300
Subject: [PATCH 441/504] fixed description

---
 docs/en/sql-reference/functions/other-functions.md | 2 +-
 docs/ru/sql-reference/functions/other-functions.md | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/sql-reference/functions/other-functions.md b/docs/en/sql-reference/functions/other-functions.md
index 46c422bd336..982b7d215e7 100644
--- a/docs/en/sql-reference/functions/other-functions.md
+++ b/docs/en/sql-reference/functions/other-functions.md
@@ -1172,7 +1172,7 @@ Result:
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Takes state of aggregate function. Returns result of aggregation (finalized state). Useful with [-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state) combinator.
+Takes state of aggregate function. Returns result of aggregation (or finalized state when using[-State](../../sql-reference/aggregate-functions/combinators.md#agg-functions-combinator-state) combinator).
 
 **Syntax** 
 
diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index f61a41a6909..289c721bcce 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1104,7 +1104,7 @@ SELECT formatReadableSize(filesystemCapacity()) AS "Capacity", toTypeName(filesy
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Принимает состояние агрегатной функции. Возвращает результат агрегирования. Полезно с комбинатором [-State](../../sql-reference/aggregate-functions/combinators.md#state).
+Принимает состояние агрегатной функции. Возвращает результат агрегирования (или завершенное состояние при использовании комбинатора [-State](../../sql-reference/aggregate-functions/combinators.md#state)).
 
 **Синтаксис** 
 

From 2239f9b0c67cb152e0d3a06c574b53dfbc530c34 Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@jakalletti-build.sas.yp-c.yandex.net>
Date: Fri, 25 Dec 2020 18:54:38 +0300
Subject: [PATCH 442/504] done

---
 src/Functions/bar.cpp                                    | 9 +++++++--
 .../0_stateless/01621_bar_nan_arguments.reference        | 0
 tests/queries/0_stateless/01621_bar_nan_arguments.sql    | 2 ++
 3 files changed, 9 insertions(+), 2 deletions(-)
 create mode 100644 tests/queries/0_stateless/01621_bar_nan_arguments.reference
 create mode 100644 tests/queries/0_stateless/01621_bar_nan_arguments.sql

diff --git a/src/Functions/bar.cpp b/src/Functions/bar.cpp
index 12b22bae80c..84d5aae190e 100644
--- a/src/Functions/bar.cpp
+++ b/src/Functions/bar.cpp
@@ -7,6 +7,8 @@
 #include <Common/FieldVisitors.h>
 #include <IO/WriteHelpers.h>
 
+#include <iostream>
+
 
 namespace DB
 {
@@ -75,11 +77,14 @@ public:
         /// The maximum width of the bar in characters, by default.
         Float64 max_width = arguments.size() == 4 ? extractConstant<Float64>(arguments, 3, "Fourth") : 80;
 
+        if (isNaN(max_width))
+            throw Exception("Argument 'max_width' must not be NaN", ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+
         if (max_width < 1)
-            throw Exception("Max_width argument must be >= 1.", ErrorCodes::ARGUMENT_OUT_OF_BOUND);
+            throw Exception("Argument 'max_width' must be >= 1", ErrorCodes::ARGUMENT_OUT_OF_BOUND);
 
         if (max_width > 1000)
-            throw Exception("Too large max_width.", ErrorCodes::ARGUMENT_OUT_OF_BOUND);
+            throw Exception("Argument 'max_width' must be <= 1000", ErrorCodes::ARGUMENT_OUT_OF_BOUND);
 
         const auto & src = *arguments[0].column;
 
diff --git a/tests/queries/0_stateless/01621_bar_nan_arguments.reference b/tests/queries/0_stateless/01621_bar_nan_arguments.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01621_bar_nan_arguments.sql b/tests/queries/0_stateless/01621_bar_nan_arguments.sql
new file mode 100644
index 00000000000..c28cb53d7ce
--- /dev/null
+++ b/tests/queries/0_stateless/01621_bar_nan_arguments.sql
@@ -0,0 +1,2 @@
+SELECT bar((greatCircleAngle(65537, 2, 1, 1) - 1) * 65535, 1048576, 1048577, nan); -- { serverError 43 }
+select bar(1,1,1,nan); -- { serverError 43 }

From 4da43881324fedf6edbdb4e05d7b1eee6083d538 Mon Sep 17 00:00:00 2001
From: Nikita Mikhailov <jakalletti@jakalletti-build.sas.yp-c.yandex.net>
Date: Fri, 25 Dec 2020 18:59:31 +0300
Subject: [PATCH 443/504] better

---
 src/Functions/bar.cpp | 2 --
 1 file changed, 2 deletions(-)

diff --git a/src/Functions/bar.cpp b/src/Functions/bar.cpp
index 84d5aae190e..2eddf23af66 100644
--- a/src/Functions/bar.cpp
+++ b/src/Functions/bar.cpp
@@ -7,8 +7,6 @@
 #include <Common/FieldVisitors.h>
 #include <IO/WriteHelpers.h>
 
-#include <iostream>
-
 
 namespace DB
 {

From 4a87118f02a1ee455bf8bf3138cb5fc9c5abcfa4 Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Fri, 25 Dec 2020 19:04:38 +0300
Subject: [PATCH 444/504] Update other-functions.md

---
 docs/ru/sql-reference/functions/other-functions.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/sql-reference/functions/other-functions.md b/docs/ru/sql-reference/functions/other-functions.md
index 289c721bcce..dfbddbfb808 100644
--- a/docs/ru/sql-reference/functions/other-functions.md
+++ b/docs/ru/sql-reference/functions/other-functions.md
@@ -1104,7 +1104,7 @@ SELECT formatReadableSize(filesystemCapacity()) AS "Capacity", toTypeName(filesy
 
 ## finalizeAggregation {#function-finalizeaggregation}
 
-Принимает состояние агрегатной функции. Возвращает результат агрегирования (или завершенное состояние при использовании комбинатора [-State](../../sql-reference/aggregate-functions/combinators.md#state)).
+Принимает состояние агрегатной функции. Возвращает результат агрегирования (или конечное состояние при использовании комбинатора [-State](../../sql-reference/aggregate-functions/combinators.md#state)).
 
 **Синтаксис** 
 

From 281d87a007c3d118046e329d65b4804b2ba53e44 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 19:04:55 +0300
Subject: [PATCH 445/504] Fixing build.

---
 src/Functions/FunctionsStringHash.cpp | 40 +++++++++++++++------------
 src/Functions/FunctionsStringHash.h   |  4 +--
 2 files changed, 25 insertions(+), 19 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index b36cc331dcb..1cb1f2c0017 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -22,6 +22,12 @@ namespace ErrorCodes
     extern const int NOT_IMPLEMENTED;
 }
 
+struct BytesRef
+{
+    const UInt8 * data;
+    size_t size;
+};
+
 struct Hash
 {
     static UInt64 crc32u64(UInt64 crc [[maybe_unused]], UInt64 val [[maybe_unused]])
@@ -123,7 +129,7 @@ struct Hash
     }
 
     template <bool CaseInsensitive>
-    static ALWAYS_INLINE inline UInt64 shingleHash(const std::vector<StringRef> & shingle, size_t offset = 0)
+    static ALWAYS_INLINE inline UInt64 shingleHash(const std::vector<BytesRef> & shingle, size_t offset = 0)
     {
         UInt64 crc = -1ULL;
 
@@ -233,7 +239,7 @@ struct SimHashImpl
         // A 64 bit vector initialized to zero.
         Int64 finger_vec[64] = {};
         // An array to store N words.
-        std::vector<StringRef> words;
+        std::vector<BytesRef> words;
         words.reserve(shingle_size);
 
         // get first word shingle
@@ -275,7 +281,7 @@ struct SimHashImpl
             // so we need to store new word hash into location of a0, then ,this array become
             // |a5|a1|a2|a3|a4|, next time, a1 become the oldest location, we need to store new
             // word hash value into location of a1, then array become |a5|a6|a2|a3|a4|
-            words[offset] = StringRef(word_start, length);
+            words[offset] = BytesRef{word_start, length};
             ++offset;
             if (offset >= shingle_size)
                 offset = 0;
@@ -328,7 +334,7 @@ struct MinHashImpl
     template<typename Comp>
     struct Heap
     {
-        void update(UInt64 hash, StringRef ref, size_t limit)
+        void update(UInt64 hash, BytesRef ref, size_t limit)
         {
             if (values.count(hash))
                 return;
@@ -369,7 +375,7 @@ struct MinHashImpl
             }
         }
 
-        std::map<UInt64, StringRef, Comp> values;
+        std::map<UInt64, BytesRef, Comp> values;
     };
 
     using MaxHeap = Heap<std::less<size_t>>;
@@ -394,8 +400,8 @@ struct MinHashImpl
 
             // insert the new hash value into array used to store K minimum value
             // and K maximum value
-            min_heap.update(hash_value, StringRef{pos, shingle_size}, heap_size);
-            max_heap.update(hash_value, StringRef{pos, shingle_size}, heap_size);
+            min_heap.update(hash_value, BytesRef{pos, shingle_size}, heap_size);
+            max_heap.update(hash_value, BytesRef{pos, shingle_size}, heap_size);
         }
     }
 
@@ -429,8 +435,8 @@ struct MinHashImpl
             size_t length = word_end - word_start;
             UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, word_start, length);
 
-            min_heap.update(hash_value, StringRef{word_start, length}, heap_size);
-            max_heap.update(hash_value, StringRef{word_start, length}, heap_size);
+            min_heap.update(hash_value, BytesRef{word_start, length}, heap_size);
+            max_heap.update(hash_value, BytesRef{word_start, length}, heap_size);
         }
     }
 
@@ -449,7 +455,7 @@ struct MinHashImpl
         const UInt8 * end = data + size;
 
         // An array to store N words.
-        std::vector<StringRef> words;
+        std::vector<BytesRef> words;
         words.reserve(shingle_size);
 
         // get first word shingle
@@ -470,9 +476,9 @@ struct MinHashImpl
 
         UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);
         {
-            const char * shingle_start = words.front().data;
-            const char * shingle_end = words.back().data + words.back().size;
-            StringRef ref(shingle_start, shingle_end - shingle_start);
+            const UInt8 * shingle_start = words.front().data;
+            const UInt8 * shingle_end = words.back().data + words.back().size;
+            BytesRef ref{shingle_start, static_cast<size_t>(shingle_end - shingle_start)};
             min_heap.update(hash_value, ref, heap_size);
             max_heap.update(hash_value, ref, heap_size);
         }
@@ -492,17 +498,17 @@ struct MinHashImpl
             if (length == 0)
                 continue;
 
-            words[offset] = StringRef(word_start, length);
-            const char * shingle_end = words[offset].data + length;
+            words[offset] = BytesRef{word_start, length};
+            const UInt8 * shingle_end = words[offset].data + length;
 
             ++offset;
             if (offset >= shingle_size)
                 offset = 0;
 
-            const char * shingle_start = words[offset].data;
+            const UInt8 * shingle_start = words[offset].data;
 
             hash_value = Hash::shingleHash<CaseInsensitive>(words, offset);
-            StringRef ref(shingle_start, shingle_end - shingle_start);
+            BytesRef ref{shingle_start, static_cast<size_t>(shingle_end - shingle_start)};
             min_heap.update(hash_value, ref, heap_size);
             max_heap.update(hash_value, ref, heap_size);
         }
diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index 2fe5418d94a..d89a1d67855 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -100,10 +100,10 @@ public:
         }
 
         if (shingle_size == 0)
-            throw Exception("Second argument (shingle size) of function {} cannot be zero", getName());
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Second argument (shingle size) of function {} cannot be zero", getName());
 
         if (num_hashes == 0)
-            throw Exception("Third argument (num hashes) of function {} cannot be zero", getName());
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Third argument (num hashes) of function {} cannot be zero", getName());
 
         auto type = std::make_shared<DataTypeUInt64>();
         if constexpr (is_simhash)

From 7e7998f1213150fe1dbb078394b25ba38f2030df Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 19:11:05 +0300
Subject: [PATCH 446/504] Fixing build.

---
 src/Functions/FunctionsStringHash.cpp | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 1cb1f2c0017..8418e195b58 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -255,7 +255,7 @@ struct SimHashImpl
             size_t length = start - word_start;
 
             if (length)
-                words.emplace_back(word_start, length);
+                words.emplace_back(BytesRef{word_start, length});
         }
 
         UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);
@@ -350,10 +350,8 @@ struct MinHashImpl
             if (values.empty())
                 return 0;
 
-            auto it = values.begin();
-            UInt64 res = CityHash_v1_0_2::CityHash64(it->first, sizeof(UInt64));
-
-            for (; it != values.end(); ++it)
+            UInt64 res = 0;
+            for (auto it = values.begin(); it != values.end(); ++it)
                 res = CityHash_v1_0_2::Hash128to64(CityHash_v1_0_2::uint128(res, it->first));
 
             return res;
@@ -529,7 +527,7 @@ struct MinHashImpl
 
         for (size_t i = 0; i < offsets.size(); ++i)
         {
-            const char * one_data = reinterpret_cast<const char *>(&data[offsets[i - 1]]);
+            const UInt8 * one_data = &data[offsets[i - 1]];
             const size_t data_size = offsets[i] - offsets[i - 1] - 1;
 
             min_heap.values.clear();

From 42911e24385315a9f5af000b37d1e07d7e38e2e7 Mon Sep 17 00:00:00 2001
From: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date: Fri, 25 Dec 2020 19:28:10 +0300
Subject: [PATCH 447/504] implement CREATE OR REPLACE TABLE

---
 src/Databases/DatabaseOnDisk.cpp              |   3 +
 src/Interpreters/InterpreterCreateQuery.cpp   | 116 +++++++++++++++---
 src/Interpreters/InterpreterCreateQuery.h     |   1 +
 src/Parsers/ASTCreateQuery.cpp                |  19 ++-
 src/Parsers/ASTCreateQuery.h                  |   5 +-
 src/Parsers/ParserCreateQuery.cpp             |  25 ++--
 .../01185_create_or_replace_table.reference   |   8 ++
 .../01185_create_or_replace_table.sql         |  23 ++++
 8 files changed, 167 insertions(+), 33 deletions(-)
 create mode 100644 tests/queries/0_stateless/01185_create_or_replace_table.reference
 create mode 100644 tests/queries/0_stateless/01185_create_or_replace_table.sql

diff --git a/src/Databases/DatabaseOnDisk.cpp b/src/Databases/DatabaseOnDisk.cpp
index fb2d10a4fe7..717a56fb58e 100644
--- a/src/Databases/DatabaseOnDisk.cpp
+++ b/src/Databases/DatabaseOnDisk.cpp
@@ -103,12 +103,15 @@ String getObjectDefinitionFromCreateQuery(const ASTPtr & query)
         create->attach = true;
 
     /// We remove everything that is not needed for ATTACH from the query.
+    assert(!create->temporary);
     create->database.clear();
     create->as_database.clear();
     create->as_table.clear();
     create->if_not_exists = false;
     create->is_populate = false;
     create->replace_view = false;
+    create->replace_table = false;
+    create->create_or_replace = false;
 
     /// For views it is necessary to save the SELECT query itself, for the rest - on the contrary
     if (!create->is_view && !create->is_materialized_view && !create->is_live_view)
diff --git a/src/Interpreters/InterpreterCreateQuery.cpp b/src/Interpreters/InterpreterCreateQuery.cpp
index 2396313f980..3d5b80c7638 100644
--- a/src/Interpreters/InterpreterCreateQuery.cpp
+++ b/src/Interpreters/InterpreterCreateQuery.cpp
@@ -733,6 +733,36 @@ void InterpreterCreateQuery::assertOrSetUUID(ASTCreateQuery & create, const Data
         /// Ignore UUID if it's ON CLUSTER query
         create.uuid = UUIDHelpers::Nil;
     }
+
+    if (create.replace_table)
+    {
+        if (database->getUUID() == UUIDHelpers::Nil)
+            throw Exception(ErrorCodes::INCORRECT_QUERY,
+                            "{} query is supported only for Atomic databases",
+                            create.create_or_replace ? "CREATE OR REPLACE TABLE" : "REPLACE TABLE");
+
+        UUID uuid_of_table_to_replace;
+        if (create.create_or_replace)
+        {
+            uuid_of_table_to_replace = context.tryResolveStorageID(StorageID(create.database, create.table)).uuid;
+            if (uuid_of_table_to_replace == UUIDHelpers::Nil)
+            {
+                /// Convert to usual CREATE
+                create.replace_table = false;
+                assert(!database->isTableExist(create.table, context));
+            }
+            else
+                create.table = "_tmp_replace_" + toString(uuid_of_table_to_replace);
+        }
+        else
+        {
+            uuid_of_table_to_replace = context.resolveStorageID(StorageID(create.database, create.table)).uuid;
+            if (uuid_of_table_to_replace == UUIDHelpers::Nil)
+                throw Exception(ErrorCodes::UNKNOWN_TABLE, "Table {}.{} doesn't exist",
+                                backQuoteIfNeed(create.database), backQuoteIfNeed(create.table));
+            create.table = "_tmp_replace_" + toString(uuid_of_table_to_replace);
+        }
+    }
 }
 
 
@@ -756,9 +786,9 @@ BlockIO InterpreterCreateQuery::createTable(ASTCreateQuery & create)
         auto query = database->getCreateTableQuery(create.table, context);
         create = query->as<ASTCreateQuery &>(); // Copy the saved create query, but use ATTACH instead of CREATE
         if (create.is_dictionary)
-            throw Exception(
-                "Cannot ATTACH TABLE " + backQuoteIfNeed(database_name) + "." + backQuoteIfNeed(create.table) + ", it is a Dictionary",
-                ErrorCodes::INCORRECT_QUERY);
+            throw Exception(ErrorCodes::INCORRECT_QUERY,
+                            "Cannot ATTACH TABLE {}.{}, it is a Dictionary",
+                            backQuoteIfNeed(database_name), backQuoteIfNeed(create.table));
         create.attach = true;
         create.attach_short_syntax = true;
         create.if_not_exists = if_not_exists;
@@ -804,6 +834,9 @@ BlockIO InterpreterCreateQuery::createTable(ASTCreateQuery & create)
     /// Set and retrieve list of columns, indices and constraints. Set table engine if needed. Rewrite query in canonical way.
     TableProperties properties = setProperties(create);
 
+    if (create.replace_table)
+        return doCreateOrReplaceTable(create, properties);
+
     /// Actually creates table
     bool created = doCreateTable(create, properties);
     if (!created)   /// Table already exists
@@ -820,20 +853,19 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create,
     String data_path;
     DatabasePtr database;
 
-    const String table_name = create.table;
     bool need_add_to_database = !create.temporary;
     if (need_add_to_database)
     {
         /** If the request specifies IF NOT EXISTS, we allow concurrent CREATE queries (which do nothing).
           * If table doesn't exist, one thread is creating table, while others wait in DDLGuard.
           */
-        guard = DatabaseCatalog::instance().getDDLGuard(create.database, table_name);
+        guard = DatabaseCatalog::instance().getDDLGuard(create.database, create.table);
 
         database = DatabaseCatalog::instance().getDatabase(create.database);
         assertOrSetUUID(create, database);
 
         /// Table can be created before or it can be created concurrently in another thread, while we were waiting in DDLGuard.
-        if (database->isTableExist(table_name, context))
+        if (database->isTableExist(create.table, context))
         {
             /// TODO Check structure of table
             if (create.if_not_exists)
@@ -843,27 +875,27 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create,
                 /// when executing CREATE OR REPLACE VIEW, drop current existing view
                 auto drop_ast = std::make_shared<ASTDropQuery>();
                 drop_ast->database = create.database;
-                drop_ast->table = table_name;
+                drop_ast->table = create.table;
                 drop_ast->no_ddl_lock = true;
 
                 InterpreterDropQuery interpreter(drop_ast, context);
                 interpreter.execute();
             }
             else
-                throw Exception("Table " + create.database + "." + table_name + " already exists.", ErrorCodes::TABLE_ALREADY_EXISTS);
+                throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Table {}.{} already exists.", backQuoteIfNeed(create.database), backQuoteIfNeed(create.table));
         }
 
         data_path = database->getTableDataPath(create);
         if (!create.attach && !data_path.empty() && fs::exists(fs::path{context.getPath()} / data_path))
-            throw Exception("Directory for table data " + data_path + " already exists", ErrorCodes::TABLE_ALREADY_EXISTS);
+            throw Exception(ErrorCodes::TABLE_ALREADY_EXISTS, "Directory for table data {} already exists", String(data_path));
     }
     else
     {
-        if (create.if_not_exists && context.tryResolveStorageID({"", table_name}, Context::ResolveExternal))
+        if (create.if_not_exists && context.tryResolveStorageID({"", create.table}, Context::ResolveExternal))
             return false;
 
         auto temporary_table = TemporaryTableHolder(context, properties.columns, properties.constraints, query_ptr);
-        context.getSessionContext().addExternalTable(table_name, std::move(temporary_table));
+        context.getSessionContext().addExternalTable(create.table, std::move(temporary_table));
         return true;
     }
 
@@ -903,7 +935,7 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create,
                         "ATTACH ... FROM ... query is not supported for {} table engine, "
                         "because such tables do not store any data on disk. Use CREATE instead.", res->getName());
 
-    database->createTable(context, table_name, res, query_ptr);
+    database->createTable(context, create.table, res, query_ptr);
 
     /// Move table data to the proper place. Wo do not move data earlier to avoid situations
     /// when data directory moved, but table has not been created due to some error.
@@ -927,6 +959,50 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create,
     return true;
 }
 
+
+BlockIO InterpreterCreateQuery::doCreateOrReplaceTable(ASTCreateQuery & create,
+                                                       const InterpreterCreateQuery::TableProperties & properties)
+{
+    auto ast_drop = std::make_shared<ASTDropQuery>();
+    String table_to_replace_name = create.table;
+    bool created = false;
+    bool replaced = false;
+
+    try
+    {
+        [[maybe_unused]] bool done = doCreateTable(create, properties);
+        assert(done);
+        ast_drop->table = create.table;
+        ast_drop->database = create.database;
+        ast_drop->kind = ASTDropQuery::Drop;
+        created = true;
+        if (!create.replace_table)
+            return fillTableIfNeeded(create);
+
+        auto ast_rename = std::make_shared<ASTRenameQuery>();
+        ASTRenameQuery::Element elem
+        {
+            ASTRenameQuery::Table{create.database, create.table},
+            ASTRenameQuery::Table{create.database, table_to_replace_name}
+        };
+        ast_rename->elements.push_back(std::move(elem));
+        ast_rename->exchange = true;
+        InterpreterRenameQuery(ast_rename, context).execute();
+        replaced = true;
+
+        InterpreterDropQuery(ast_drop, context).execute();
+
+        create.table = table_to_replace_name;
+        return fillTableIfNeeded(create);
+    }
+    catch (...)
+    {
+        if (created && create.replace_table && !replaced)
+            InterpreterDropQuery(ast_drop, context).execute();
+        throw;
+    }
+}
+
 BlockIO InterpreterCreateQuery::fillTableIfNeeded(const ASTCreateQuery & create)
 {
     /// If the query is a CREATE SELECT, insert the data into the table.
@@ -1079,22 +1155,22 @@ AccessRightsElements InterpreterCreateQuery::getRequiredAccess() const
     }
     else if (create.is_view || create.is_materialized_view || create.is_live_view)
     {
-        if (create.temporary)
-            required_access.emplace_back(AccessType::CREATE_TEMPORARY_TABLE);
+        assert(!create.temporary);
+        if (create.replace_view)
+            required_access.emplace_back(AccessType::DROP_VIEW | AccessType::CREATE_VIEW, create.database, create.table);
         else
-        {
-            if (create.replace_view)
-                required_access.emplace_back(AccessType::DROP_VIEW | AccessType::CREATE_VIEW, create.database, create.table);
-            else
-                required_access.emplace_back(AccessType::CREATE_VIEW, create.database, create.table);
-        }
+            required_access.emplace_back(AccessType::CREATE_VIEW, create.database, create.table);
     }
     else
     {
         if (create.temporary)
             required_access.emplace_back(AccessType::CREATE_TEMPORARY_TABLE);
         else
+        {
+            if (create.replace_table)
+                required_access.emplace_back(AccessType::DROP_TABLE, create.database, create.table);
             required_access.emplace_back(AccessType::CREATE_TABLE, create.database, create.table);
+        }
     }
 
     if (create.to_table_id)
diff --git a/src/Interpreters/InterpreterCreateQuery.h b/src/Interpreters/InterpreterCreateQuery.h
index 347d84cef0b..c109b0b7760 100644
--- a/src/Interpreters/InterpreterCreateQuery.h
+++ b/src/Interpreters/InterpreterCreateQuery.h
@@ -81,6 +81,7 @@ private:
 
     /// Create IStorage and add it to database. If table already exists and IF NOT EXISTS specified, do nothing and return false.
     bool doCreateTable(ASTCreateQuery & create, const TableProperties & properties);
+    BlockIO doCreateOrReplaceTable(ASTCreateQuery & create, const InterpreterCreateQuery::TableProperties & properties);
     /// Inserts data in created table if it's CREATE ... SELECT
     BlockIO fillTableIfNeeded(const ASTCreateQuery & create);
 
diff --git a/src/Parsers/ASTCreateQuery.cpp b/src/Parsers/ASTCreateQuery.cpp
index 03db54c6957..1ef2f85585c 100644
--- a/src/Parsers/ASTCreateQuery.cpp
+++ b/src/Parsers/ASTCreateQuery.cpp
@@ -230,19 +230,28 @@ void ASTCreateQuery::formatQueryImpl(const FormatSettings & settings, FormatStat
 
     if (!is_dictionary)
     {
-        std::string what = "TABLE";
+        String action = "CREATE";
+        if (attach)
+            action = "ATTACH";
+        else if (replace_view)
+            action = "CREATE OR REPLACE";
+        else if (replace_table && create_or_replace)
+            action = "CREATE OR REPLACE";
+        else if (replace_table)
+            action = "REPLACE";
+
+        String what = "TABLE";
         if (is_view)
             what = "VIEW";
-        if (is_materialized_view)
+        else if (is_materialized_view)
             what = "MATERIALIZED VIEW";
-        if (is_live_view)
+        else if (is_live_view)
             what = "LIVE VIEW";
 
         settings.ostr
             << (settings.hilite ? hilite_keyword : "")
-                << (attach ? "ATTACH " : "CREATE ")
+                << action << " "
                 << (temporary ? "TEMPORARY " : "")
-                << (replace_view ? "OR REPLACE " : "")
                 << what << " "
                 << (if_not_exists ? "IF NOT EXISTS " : "")
             << (settings.hilite ? hilite_none : "")
diff --git a/src/Parsers/ASTCreateQuery.h b/src/Parsers/ASTCreateQuery.h
index 7b2deb99698..4f0d9777f04 100644
--- a/src/Parsers/ASTCreateQuery.h
+++ b/src/Parsers/ASTCreateQuery.h
@@ -64,7 +64,7 @@ public:
     bool replace_view{false}; /// CREATE OR REPLACE VIEW
     ASTColumns * columns_list = nullptr;
     ASTExpressionList * tables = nullptr;
-    //FIXME
+
     StorageID to_table_id = StorageID::createEmpty();   /// For CREATE MATERIALIZED VIEW mv TO table.
     ASTStorage * storage = nullptr;
     String as_database;
@@ -81,6 +81,9 @@ public:
 
     std::optional<String> attach_from_path = std::nullopt;
 
+    bool replace_table{false};
+    bool create_or_replace{false};
+
     /** Get the text that identifies this element. */
     String getID(char delim) const override { return (attach ? "AttachQuery" : "CreateQuery") + (delim + database) + delim + table; }
 
diff --git a/src/Parsers/ParserCreateQuery.cpp b/src/Parsers/ParserCreateQuery.cpp
index 5fe6a4eee80..c77ddc5f180 100644
--- a/src/Parsers/ParserCreateQuery.cpp
+++ b/src/Parsers/ParserCreateQuery.cpp
@@ -355,6 +355,8 @@ bool ParserCreateTableQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
 {
     ParserKeyword s_create("CREATE");
     ParserKeyword s_attach("ATTACH");
+    ParserKeyword s_replace("REPLACE");
+    ParserKeyword s_or_replace("OR REPLACE");
     ParserKeyword s_temporary("TEMPORARY");
     ParserKeyword s_table("TABLE");
     ParserKeyword s_if_not_exists("IF NOT EXISTS");
@@ -383,25 +385,32 @@ bool ParserCreateTableQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
 
     String cluster_str;
     bool attach = false;
+    bool replace = false;
+    bool or_replace = false;
     bool if_not_exists = false;
     bool is_temporary = false;
 
-    if (!s_create.ignore(pos, expected))
+    if (s_create.ignore(pos, expected))
     {
-        if (s_attach.ignore(pos, expected))
-            attach = true;
-        else
-            return false;
+        if (s_or_replace.ignore(pos, expected))
+            replace = or_replace = true;
     }
+    else if (s_attach.ignore(pos, expected))
+        attach = true;
+    else if (s_replace.ignore(pos, expected))
+        replace = true;
+    else
+        return false;
 
-    if (s_temporary.ignore(pos, expected))
+
+    if (!replace && !or_replace && s_temporary.ignore(pos, expected))
     {
         is_temporary = true;
     }
     if (!s_table.ignore(pos, expected))
         return false;
 
-    if (s_if_not_exists.ignore(pos, expected))
+    if (!replace && !or_replace && s_if_not_exists.ignore(pos, expected))
         if_not_exists = true;
 
     if (!table_name_p.parse(pos, table, expected))
@@ -505,6 +514,8 @@ bool ParserCreateTableQuery::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
         query->as_table_function = as_table_function;
 
     query->attach = attach;
+    query->replace_table = replace;
+    query->create_or_replace = or_replace;
     query->if_not_exists = if_not_exists;
     query->temporary = is_temporary;
 
diff --git a/tests/queries/0_stateless/01185_create_or_replace_table.reference b/tests/queries/0_stateless/01185_create_or_replace_table.reference
new file mode 100644
index 00000000000..84df5f0f5b5
--- /dev/null
+++ b/tests/queries/0_stateless/01185_create_or_replace_table.reference
@@ -0,0 +1,8 @@
+t1
+CREATE TABLE test_01185.t1\n(\n    `n` UInt64,\n    `s` String\n)\nENGINE = MergeTree\nORDER BY n\nSETTINGS index_granularity = 8192
+t1
+CREATE TABLE test_01185.t1\n(\n    `n` UInt64,\n    `s` Nullable(String)\n)\nENGINE = MergeTree\nORDER BY n\nSETTINGS index_granularity = 8192
+2	\N
+t1
+CREATE TABLE test_01185.t1\n(\n    `n` UInt64\n)\nENGINE = MergeTree\nORDER BY n\nSETTINGS index_granularity = 8192
+3
diff --git a/tests/queries/0_stateless/01185_create_or_replace_table.sql b/tests/queries/0_stateless/01185_create_or_replace_table.sql
new file mode 100644
index 00000000000..fe408cc7ac6
--- /dev/null
+++ b/tests/queries/0_stateless/01185_create_or_replace_table.sql
@@ -0,0 +1,23 @@
+drop database if exists test_01185;
+create database test_01185 engine=Atomic;
+
+replace table test_01185.t1 (n UInt64, s String) engine=MergeTree order by n; -- { serverError 60 }
+show tables from test_01185;
+create or replace table test_01185.t1 (n UInt64, s String) engine=MergeTree order by n;
+show tables from test_01185;
+show create table test_01185.t1;
+
+insert into test_01185.t1 values (1, 'test');
+create or replace table test_01185.t1 (n UInt64, s Nullable(String)) engine=MergeTree order by n;
+insert into test_01185.t1 values (2, null);
+show tables from test_01185;
+show create table test_01185.t1;
+select * from test_01185.t1;
+
+replace table test_01185.t1 (n UInt64) engine=MergeTree order by n;
+insert into test_01185.t1 values (3);
+show tables from test_01185;
+show create table test_01185.t1;
+select * from test_01185.t1;
+
+drop database test_01185;

From e335ba29fccbcff3888c337fd64fd46b3efae979 Mon Sep 17 00:00:00 2001
From: myrrc <me@myrrec.space>
Date: Fri, 25 Dec 2020 19:32:59 +0300
Subject: [PATCH 448/504] fixing build on gcc

---
 src/Functions/intDiv.cpp | 2 --
 src/Functions/modulo.cpp | 4 +++-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/Functions/intDiv.cpp b/src/Functions/intDiv.cpp
index 54e7f9ed5e1..e13c46e9cd1 100644
--- a/src/Functions/intDiv.cpp
+++ b/src/Functions/intDiv.cpp
@@ -90,7 +90,6 @@ struct DivideIntegralByConstantImpl
 
 namespace impl_
 {
-
 template <> struct BinaryOperationImpl<UInt64, UInt8, DivideIntegralImpl<UInt64, UInt8>> : DivideIntegralByConstantImpl<UInt64, UInt8> {};
 template <> struct BinaryOperationImpl<UInt64, UInt16, DivideIntegralImpl<UInt64, UInt16>> : DivideIntegralByConstantImpl<UInt64, UInt16> {};
 template <> struct BinaryOperationImpl<UInt64, UInt32, DivideIntegralImpl<UInt64, UInt32>> : DivideIntegralByConstantImpl<UInt64, UInt32> {};
@@ -110,7 +109,6 @@ template <> struct BinaryOperationImpl<Int32, Int8, DivideIntegralImpl<Int32, In
 template <> struct BinaryOperationImpl<Int32, Int16, DivideIntegralImpl<Int32, Int16>> : DivideIntegralByConstantImpl<Int32, Int16> {};
 template <> struct BinaryOperationImpl<Int32, Int32, DivideIntegralImpl<Int32, Int32>> : DivideIntegralByConstantImpl<Int32, Int32> {};
 template <> struct BinaryOperationImpl<Int32, Int64, DivideIntegralImpl<Int32, Int64>> : DivideIntegralByConstantImpl<Int32, Int64> {};
-
 }
 
 struct NameIntDiv { static constexpr auto name = "intDiv"; };
diff --git a/src/Functions/modulo.cpp b/src/Functions/modulo.cpp
index 1458d20c131..388e6ab9fb9 100644
--- a/src/Functions/modulo.cpp
+++ b/src/Functions/modulo.cpp
@@ -79,6 +79,8 @@ struct ModuloByConstantImpl
   * Can be expanded to all possible combinations, but more code is needed.
   */
 
+namespace impl_
+{
 template <> struct BinaryOperationImpl<UInt64, UInt8, ModuloImpl<UInt64, UInt8>> : ModuloByConstantImpl<UInt64, UInt8> {};
 template <> struct BinaryOperationImpl<UInt64, UInt16, ModuloImpl<UInt64, UInt16>> : ModuloByConstantImpl<UInt64, UInt16> {};
 template <> struct BinaryOperationImpl<UInt64, UInt32, ModuloImpl<UInt64, UInt32>> : ModuloByConstantImpl<UInt64, UInt32> {};
@@ -98,7 +100,7 @@ template <> struct BinaryOperationImpl<Int32, Int8, ModuloImpl<Int32, Int8>> : M
 template <> struct BinaryOperationImpl<Int32, Int16, ModuloImpl<Int32, Int16>> : ModuloByConstantImpl<Int32, Int16> {};
 template <> struct BinaryOperationImpl<Int32, Int32, ModuloImpl<Int32, Int32>> : ModuloByConstantImpl<Int32, Int32> {};
 template <> struct BinaryOperationImpl<Int32, Int64, ModuloImpl<Int32, Int64>> : ModuloByConstantImpl<Int32, Int64> {};
-
+}
 
 struct NameModulo { static constexpr auto name = "modulo"; };
 using FunctionModulo = BinaryArithmeticOverloadResolver<ModuloImpl, NameModulo, false>;

From ba5d2617ffb39634e21e0d6615281d74378a05d1 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 19:49:55 +0300
Subject: [PATCH 449/504] Fixing build.

---
 src/Functions/FunctionsStringHash.cpp | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 8418e195b58..ac60d4e3139 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -138,6 +138,8 @@ struct Hash
 
         for (size_t i = 0; i < offset; ++i)
             crc = shingleHash<CaseInsensitive>(crc, shingle[i].data, shingle[i].size);
+
+        return crc;
     }
 };
 
@@ -365,7 +367,7 @@ struct MinHashImpl
                 auto & col_string = static_cast<ColumnString &>(strings.getColumn(i));
                 if (it != values.end())
                 {
-                    col_string.insertData(it->second.data, it->second.size);
+                    col_string.insertData(static_cast<const char *>(it->second.data), it->second.size);
                     ++it;
                 }
                 else

From d204ad7ef1289643016ffae363a6234cc82ddcfc Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 19:54:13 +0300
Subject: [PATCH 450/504] Fix usage of concurrent bounded queue

---
 src/Common/ZooKeeper/TestKeeperStorage.cpp | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/src/Common/ZooKeeper/TestKeeperStorage.cpp b/src/Common/ZooKeeper/TestKeeperStorage.cpp
index 6b6ae93bbb9..daadba6519e 100644
--- a/src/Common/ZooKeeper/TestKeeperStorage.cpp
+++ b/src/Common/ZooKeeper/TestKeeperStorage.cpp
@@ -740,11 +740,13 @@ void TestKeeperStorage::putRequest(const Coordination::ZooKeeperRequestPtr & req
     request_info.session_id = session_id;
     request_info.response_callback = callback;
 
-    /// Put close requests without timeouts
-    auto timeout = request->getOpNum() == Coordination::OpNum::Close ? 0 : operation_timeout.totalMilliseconds();
     std::lock_guard lock(push_request_mutex);
-    if (!requests_queue.tryPush(std::move(request_info), timeout))
+    /// Put close requests without timeouts
+    if (request->getOpNum() == Coordination::OpNum::Close)
+        requests_queue.push(std::move(request_info));
+    else if (!requests_queue.tryPush(std::move(request_info), operation_timeout.totalMilliseconds()))
         throw Exception("Cannot push request to queue within operation timeout", ErrorCodes::TIMEOUT_EXCEEDED);
+
 }
 
 void TestKeeperStorage::putRequest(const Coordination::ZooKeeperRequestPtr & request, int64_t session_id, ResponseCallback callback, ResponseCallback watch_callback)
@@ -758,10 +760,11 @@ void TestKeeperStorage::putRequest(const Coordination::ZooKeeperRequestPtr & req
     if (request->has_watch)
         request_info.watch_callback = watch_callback;
 
-    /// Put close requests without timeouts
-    auto timeout = request->getOpNum() == Coordination::OpNum::Close ? 0 : operation_timeout.totalMilliseconds();
     std::lock_guard lock(push_request_mutex);
-    if (!requests_queue.tryPush(std::move(request_info), timeout))
+    /// Put close requests without timeouts
+    if (request->getOpNum() == Coordination::OpNum::Close)
+        requests_queue.push(std::move(request_info));
+    else if (!requests_queue.tryPush(std::move(request_info), operation_timeout.totalMilliseconds()))
         throw Exception("Cannot push request to queue within operation timeout", ErrorCodes::TIMEOUT_EXCEEDED);
 }
 

From 309b0a860a1f8c30bd6f987eb1880baa89ccf6e0 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 19:55:15 +0300
Subject: [PATCH 451/504] Fixing build.

---
 src/Functions/FunctionsStringHash.cpp | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index ac60d4e3139..a6791ce90c0 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -381,7 +381,7 @@ struct MinHashImpl
     using MaxHeap = Heap<std::less<size_t>>;
     using MinHeap = Heap<std::greater<size_t>>;
 
-    static ALWAYS_INLINE inline UInt64 ngramHashASCII(
+    static ALWAYS_INLINE inline void ngramHashASCII(
         MinHeap & min_heap,
         MaxHeap & max_heap,
         const UInt8 * data,
@@ -390,7 +390,12 @@ struct MinHashImpl
         size_t heap_size)
     {
         if (size < shingle_size)
-            return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+        {
+            UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+            min_heap.update(hash_value, BytesRef{data, size}, heap_size);
+            max_heap.update(hash_value, BytesRef{data, size}, heap_size);
+            return;
+        }
 
         const UInt8 * end = data + size;
 
@@ -405,7 +410,7 @@ struct MinHashImpl
         }
     }
 
-    static ALWAYS_INLINE inline UInt64 ngramHashUTF8(
+    static ALWAYS_INLINE inline void ngramHashUTF8(
         MinHeap & min_heap,
         MaxHeap & max_heap,
         const UInt8 * data,
@@ -422,7 +427,12 @@ struct MinHashImpl
         for (size_t i = 0; i < shingle_size; ++i)
         {
             if (word_end >= end)
-                return Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+            {
+                auto hash_value = Hash::shingleHash<CaseInsensitive>(-1ULL, data, size);
+                min_heap.update(hash_value, BytesRef{data, size}, heap_size);
+                max_heap.update(hash_value, BytesRef{data, size}, heap_size);
+                return;
+            }
 
             ExtractStringImpl::readOneUTF8Code(word_end, end);
         }
@@ -443,7 +453,7 @@ struct MinHashImpl
     // MinHash word shingle hash value calculate function: String ->Tuple(UInt64, UInt64)
     // for each word shingle, we calculate a hash value, but in fact, we just maintain the
     // K minimum and K maximum hash value
-    static ALWAYS_INLINE inline UInt64 wordShingleHash(
+    static ALWAYS_INLINE inline void wordShingleHash(
         MinHeap & min_heap,
         MaxHeap & max_heap,
         const UInt8 * data,

From 2c06a65e1cd59a058f9ee292c08d8c273781babc Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 19:56:51 +0300
Subject: [PATCH 452/504] Fixing build.

---
 src/Functions/FunctionsStringHash.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index a6791ce90c0..79ea9c16158 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -367,7 +367,7 @@ struct MinHashImpl
                 auto & col_string = static_cast<ColumnString &>(strings.getColumn(i));
                 if (it != values.end())
                 {
-                    col_string.insertData(static_cast<const char *>(it->second.data), it->second.size);
+                    col_string.insertData(reinterpret_cast<const char *>(it->second.data), it->second.size);
                     ++it;
                 }
                 else

From b53442e52175e62cf857666cd0f1be12e95443d5 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 20:00:42 +0300
Subject: [PATCH 453/504] Fixing build.

---
 src/Functions/FunctionsStringHash.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 79ea9c16158..1cbb0523165 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -481,7 +481,7 @@ struct MinHashImpl
             size_t length = start - word_start;
 
             if (length)
-                words.emplace_back(word_start, length);
+                words.emplace_back(BytesRef{word_start, length});
         }
 
         UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);

From 63de06fa31a24e8ef01265dc5d244fd2b866679e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 20:02:19 +0300
Subject: [PATCH 454/504] Fixing build.

---
 src/Functions/FunctionsStringHash.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionsStringHash.h b/src/Functions/FunctionsStringHash.h
index d89a1d67855..2bcb7fa1013 100644
--- a/src/Functions/FunctionsStringHash.h
+++ b/src/Functions/FunctionsStringHash.h
@@ -158,7 +158,7 @@ public:
             auto max_tuple = ColumnTuple::create(std::move(max_columns));
 
             const ColumnString * col_str_vector = checkAndGetColumn<ColumnString>(&*column);
-            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), shingle_size, num_hashes, nullptr, nullptr, &min_tuple, &max_tuple);
+            Impl::apply(col_str_vector->getChars(), col_str_vector->getOffsets(), shingle_size, num_hashes, nullptr, nullptr, min_tuple.get(), max_tuple.get());
 
             MutableColumns tuple_columns;
             tuple_columns.emplace_back(std::move(min_tuple));

From 721bb60edd055c432e8606472f5e63232b1de56e Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Fri, 25 Dec 2020 20:12:47 +0300
Subject: [PATCH 455/504] fix cast to map from tuple of arrays with unequal
 sizes

---
 src/Functions/FunctionsConversion.h           | 19 +++++++++----------
 .../0_stateless/01550_create_map_type.sql     |  1 +
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/src/Functions/FunctionsConversion.h b/src/Functions/FunctionsConversion.h
index 84a41a9c415..6669a369a9d 100644
--- a/src/Functions/FunctionsConversion.h
+++ b/src/Functions/FunctionsConversion.h
@@ -2473,24 +2473,23 @@ private:
             const auto * col = arguments.front().column.get();
             const auto & column_tuple = assert_cast<const ColumnTuple &>(*col);
 
-            if (column_tuple.getColumn(0).size() != column_tuple.getColumn(1).size())
-                throw Exception(ErrorCodes::TYPE_MISMATCH,
-                    "CAST AS Map can only be performed from tuple of arrays with equal sizes."
-                    " Size of keys: {}. Size of values: {}", column_tuple.getColumn(0).size(), column_tuple.getColumn(1).size());
-
-            ColumnPtr offsets;
+            Columns offsets(2);
             Columns converted_columns(2);
             for (size_t i = 0; i < 2; ++i)
             {
                 const auto & column_array = assert_cast<const ColumnArray &>(column_tuple.getColumn(i));
                 ColumnsWithTypeAndName element = {{column_array.getDataPtr(), from_kv_types[i], ""}};
                 converted_columns[i] = element_wrappers[i](element, to_kv_types[i], nullable_source, input_rows_count);
-
-                if (!offsets)
-                    offsets = column_array.getOffsetsPtr();
+                offsets[i] = column_array.getOffsetsPtr();
             }
 
-            return ColumnMap::create(converted_columns[0], converted_columns[1], offsets);
+            const auto & keys_offsets = assert_cast<const ColumnArray::ColumnOffsets &>(*offsets[0]).getData();
+            const auto & values_offsets = assert_cast<const ColumnArray::ColumnOffsets &>(*offsets[1]).getData();
+            if (keys_offsets != values_offsets)
+                throw Exception(ErrorCodes::TYPE_MISMATCH,
+                    "CAST AS Map can only be performed from tuple of arrays with equal sizes.");
+
+            return ColumnMap::create(converted_columns[0], converted_columns[1], offsets[0]);
         };
     }
 
diff --git a/tests/queries/0_stateless/01550_create_map_type.sql b/tests/queries/0_stateless/01550_create_map_type.sql
index 7a08886a5d8..695a40c42b4 100644
--- a/tests/queries/0_stateless/01550_create_map_type.sql
+++ b/tests/queries/0_stateless/01550_create_map_type.sql
@@ -63,3 +63,4 @@ SELECT sum(m['1']), sum(m['7']), sum(m['100']) FROM table_map;
 
 DROP TABLE IF EXISTS table_map;
 
+SELECT CAST(([2, 1, 1023], ['', '']), 'Map(UInt8, String)') AS map, map[10] -- { serverError 53}

From feadf02e20445a63274d7d05ac4acb4bf83a2978 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 20:14:10 +0300
Subject: [PATCH 456/504] Update test.

---
 .../0_stateless/01016_simhash_minhash.sql     | 132 +++++++++---------
 1 file changed, 66 insertions(+), 66 deletions(-)

diff --git a/tests/queries/0_stateless/01016_simhash_minhash.sql b/tests/queries/0_stateless/01016_simhash_minhash.sql
index 61b9ac14259..e27c851f9bb 100644
--- a/tests/queries/0_stateless/01016_simhash_minhash.sql
+++ b/tests/queries/0_stateless/01016_simhash_minhash.sql
@@ -1,22 +1,22 @@
-SELECT ngramSimhash('');
-SELECT ngramSimhash('what a cute cat.');
-SELECT ngramSimhashCaseInsensitive('what a cute cat.');
-SELECT ngramSimhashUTF8('what a cute cat.');
-SELECT ngramSimhashCaseInsensitiveUTF8('what a cute cat.');
-SELECT wordShingleSimhash('what a cute cat.');
-SELECT wordShingleSimhashCaseInsensitive('what a cute cat.');
-SELECT wordShingleSimhashUTF8('what a cute cat.');
-SELECT wordShingleSimhashCaseInsensitiveUTF8('what a cute cat.');
+SELECT ngramSimHash('');
+SELECT ngramSimHash('what a cute cat.');
+SELECT ngramSimHashCaseInsensitive('what a cute cat.');
+SELECT ngramSimHashUTF8('what a cute cat.');
+SELECT ngramSimHashCaseInsensitiveUTF8('what a cute cat.');
+SELECT wordShingleSimHash('what a cute cat.');
+SELECT wordShingleSimHashCaseInsensitive('what a cute cat.');
+SELECT wordShingleSimHashUTF8('what a cute cat.');
+SELECT wordShingleSimHashCaseInsensitiveUTF8('what a cute cat.');
 
-SELECT ngramMinhash('');
-SELECT ngramMinhash('what a cute cat.');
-SELECT ngramMinhashCaseInsensitive('what a cute cat.');
-SELECT ngramMinhashUTF8('what a cute cat.');
-SELECT ngramMinhashCaseInsensitiveUTF8('what a cute cat.');
-SELECT wordShingleMinhash('what a cute cat.');
-SELECT wordShingleMinhashCaseInsensitive('what a cute cat.');
-SELECT wordShingleMinhashUTF8('what a cute cat.');
-SELECT wordShingleMinhashCaseInsensitiveUTF8('what a cute cat.');
+SELECT ngramMinHash('');
+SELECT ngramMinHash('what a cute cat.');
+SELECT ngramMinHashCaseInsensitive('what a cute cat.');
+SELECT ngramMinHashUTF8('what a cute cat.');
+SELECT ngramMinHashCaseInsensitiveUTF8('what a cute cat.');
+SELECT wordShingleMinHash('what a cute cat.');
+SELECT wordShingleMinHashCaseInsensitive('what a cute cat.');
+SELECT wordShingleMinHashUTF8('what a cute cat.');
+SELECT wordShingleMinHashCaseInsensitiveUTF8('what a cute cat.');
 
 DROP TABLE IF EXISTS defaults;
 CREATE TABLE defaults
@@ -26,23 +26,23 @@ CREATE TABLE defaults
 
 INSERT INTO defaults values ('It is the latest occurrence of the Southeast European haze, the issue that occurs in constant intensity during every wet season. It has mainly been caused by forest fires resulting from illegal slash-and-burn clearing performed on behalf of the palm oil industry in Kazakhstan, principally on the islands, which then spread quickly in the dry season.') ('It is the latest occurrence of the Southeast Asian haze, the issue that occurs in constant intensity during every wet season. It has mainly been caused by forest fires resulting from illegal slash-and-burn clearing performed on behalf of the palm oil industry in Kazakhstan, principally on the islands, which then spread quickly in the dry season.');
 
-SELECT ngramSimhash(s) FROM defaults;
-SELECT ngramSimhashCaseInsensitive(s) FROM defaults;
-SELECT ngramSimhashUTF8(s) FROM defaults;
-SELECT ngramSimhashCaseInsensitiveUTF8(s) FROM defaults;
-SELECT wordShingleSimhash(s) FROM defaults;
-SELECT wordShingleSimhashCaseInsensitive(s) FROM defaults;
-SELECT wordShingleSimhashUTF8(s) FROM defaults;
-SELECT wordShingleSimhashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT ngramSimHash(s) FROM defaults;
+SELECT ngramSimHashCaseInsensitive(s) FROM defaults;
+SELECT ngramSimHashUTF8(s) FROM defaults;
+SELECT ngramSimHashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT wordShingleSimHash(s) FROM defaults;
+SELECT wordShingleSimHashCaseInsensitive(s) FROM defaults;
+SELECT wordShingleSimHashUTF8(s) FROM defaults;
+SELECT wordShingleSimHashCaseInsensitiveUTF8(s) FROM defaults;
 
-SELECT ngramMinhash(s) FROM defaults;
-SELECT ngramMinhashCaseInsensitive(s) FROM defaults;
-SELECT ngramMinhashUTF8(s) FROM defaults;
-SELECT ngramMinhashCaseInsensitiveUTF8(s) FROM defaults;
-SELECT wordShingleMinhash(s) FROM defaults;
-SELECT wordShingleMinhashCaseInsensitive(s) FROM defaults;
-SELECT wordShingleMinhashUTF8(s) FROM defaults;
-SELECT wordShingleMinhashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT ngramMinHash(s) FROM defaults;
+SELECT ngramMinHashCaseInsensitive(s) FROM defaults;
+SELECT ngramMinHashUTF8(s) FROM defaults;
+SELECT ngramMinHashCaseInsensitiveUTF8(s) FROM defaults;
+SELECT wordShingleMinHash(s) FROM defaults;
+SELECT wordShingleMinHashCaseInsensitive(s) FROM defaults;
+SELECT wordShingleMinHashUTF8(s) FROM defaults;
+SELECT wordShingleMinHashCaseInsensitiveUTF8(s) FROM defaults;
 
 TRUNCATE TABLE defaults;
 INSERT INTO defaults SELECT arrayJoin(splitByString('\n\n',
@@ -74,38 +74,38 @@ ClickHouse is simple and works out of the box. It simplifies all processing of y
 SELECT 'uniqExact', uniqExact(s) FROM defaults;
 
 
-SELECT 'ngramSimhash';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhash(s) as h FROM defaults GROUP BY h;
-SELECT 'ngramSimhashCaseInsensitive';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
-SELECT 'ngramSimhashUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhashUTF8(s) as h FROM defaults GROUP BY h;
-SELECT 'ngramSimhashCaseInsensitiveUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleSimhash';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhash(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleSimhashCaseInsensitive';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleSimhashUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhashUTF8(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleSimhashCaseInsensitiveUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimHash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimHash(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimHashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimHashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimHashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimHashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramSimHashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimHash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHash(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimHashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimHashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleSimHashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
 
-SELECT 'ngramMinhash';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhash(s) as h FROM defaults GROUP BY h;
-SELECT 'ngramMinhashCaseInsensitive';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
-SELECT 'ngramMinhashUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhashUTF8(s) as h FROM defaults GROUP BY h;
-SELECT 'ngramMinhashCaseInsensitiveUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleMinhash';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhash(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleMinhashCaseInsensitive';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhashCaseInsensitive(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleMinhashUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhashUTF8(s) as h FROM defaults GROUP BY h;
-SELECT 'wordShingleMinhashCaseInsensitiveUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinhashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinHash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHash(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinHashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinHashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'ngramMinHashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinHash';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHash(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinHashCaseInsensitive';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinHashUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT 'wordShingleMinHashCaseInsensitiveUTF8';
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
 
 DROP TABLE defaults;

From f8854fed5c1051bb590e2dcc093b38bed84e31df Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 20:30:44 +0300
Subject: [PATCH 457/504] Update test.

---
 src/Functions/FunctionsStringHash.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 1cbb0523165..432a81a9793 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -362,7 +362,7 @@ struct MinHashImpl
         void fill(ColumnTuple & strings)
         {
             auto it = values.begin();
-            for (size_t i = 0; i < strings.size(); ++i)
+            for (size_t i = 0; i < strings.tupleSize(); ++i)
             {
                 auto & col_string = static_cast<ColumnString &>(strings.getColumn(i));
                 if (it != values.end())

From 2836936079739a40c40ba8d87e0172ff60434503 Mon Sep 17 00:00:00 2001
From: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date: Fri, 25 Dec 2020 20:43:03 +0300
Subject: [PATCH 458/504] fix

---
 src/Interpreters/InterpreterCreateQuery.cpp | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/Interpreters/InterpreterCreateQuery.cpp b/src/Interpreters/InterpreterCreateQuery.cpp
index 3d5b80c7638..9c9726f39d6 100644
--- a/src/Interpreters/InterpreterCreateQuery.cpp
+++ b/src/Interpreters/InterpreterCreateQuery.cpp
@@ -81,6 +81,7 @@ namespace ErrorCodes
     extern const int LOGICAL_ERROR;
     extern const int PATH_ACCESS_DENIED;
     extern const int NOT_IMPLEMENTED;
+    extern const int UNKNOWN_TABLE;
 }
 
 namespace fs = std::filesystem;
@@ -894,8 +895,9 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create,
         if (create.if_not_exists && context.tryResolveStorageID({"", create.table}, Context::ResolveExternal))
             return false;
 
+        String temporary_table_name = create.table;
         auto temporary_table = TemporaryTableHolder(context, properties.columns, properties.constraints, query_ptr);
-        context.getSessionContext().addExternalTable(create.table, std::move(temporary_table));
+        context.getSessionContext().addExternalTable(temporary_table_name, std::move(temporary_table));
         return true;
     }
 

From 31191d839656559fa9c75ade1522a59583b2e6cc Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 20:58:44 +0300
Subject: [PATCH 459/504] Add min word size.

---
 src/Functions/ExtractString.h         |  4 ++--
 src/Functions/FunctionsStringHash.cpp | 16 ++++++++++------
 2 files changed, 12 insertions(+), 8 deletions(-)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index c4bb2e0fdf7..94ee01ae609 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -29,12 +29,12 @@ struct ExtractStringImpl
     static ALWAYS_INLINE inline const UInt8 * readOneASCIIWord(const UInt8 *& pos, const UInt8 * end)
     {
         // jump separators
-        while (pos < end && !isAlphaNumericASCII(*pos))
+        while (pos < end && isUTF8Sep(*pos))
             ++pos;
 
         // word start from here
         const UInt8 * word_start = pos;
-        while (pos < end && isAlphaNumericASCII(*pos))
+        while (pos < end && !isUTF8Sep(*pos))
             ++pos;
 
         return word_start;
diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 432a81a9793..eba985bd1e8 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -150,6 +150,8 @@ struct Hash
 template <bool UTF8, bool Ngram, bool CaseInsensitive>
 struct SimHashImpl
 {
+    static constexpr size_t min_word_size = 4;
+
     /// Update fingerprint according to hash_value bits.
     static ALWAYS_INLINE inline void updateFingerVector(Int64 * finger_vec, UInt64 hash_value)
     {
@@ -247,7 +249,7 @@ struct SimHashImpl
         // get first word shingle
         while (start < end && words.size() < shingle_size)
         {
-            const UInt8 * word_start = nullptr;
+            const UInt8 * word_start;
 
             if constexpr (UTF8)
                 word_start = ExtractStringImpl::readOneUTF8Word(start, end);
@@ -256,7 +258,7 @@ struct SimHashImpl
 
             size_t length = start - word_start;
 
-            if (length)
+            if (length >= min_word_size)
                 words.emplace_back(BytesRef{word_start, length});
         }
 
@@ -275,7 +277,7 @@ struct SimHashImpl
 
             size_t length = start - word_start;
 
-            if (length == 0)
+            if (length < min_word_size)
                 continue;
 
             // we need to store the new word hash value to the oldest location.
@@ -333,6 +335,8 @@ struct SimHashImpl
 template <bool UTF8, bool Ngram, bool CaseInsensitive>
 struct MinHashImpl
 {
+    static constexpr size_t min_word_size = 3;
+
     template<typename Comp>
     struct Heap
     {
@@ -471,7 +475,7 @@ struct MinHashImpl
         // get first word shingle
         while (start < end && words.size() < shingle_size)
         {
-            const UInt8 * word_start = nullptr;
+            const UInt8 * word_start;
 
             if constexpr (UTF8)
                 word_start = ExtractStringImpl::readOneUTF8Word(start, end);
@@ -480,7 +484,7 @@ struct MinHashImpl
 
             size_t length = start - word_start;
 
-            if (length)
+            if (length >= min_word_size)
                 words.emplace_back(BytesRef{word_start, length});
         }
 
@@ -505,7 +509,7 @@ struct MinHashImpl
 
             size_t length = start - word_start;
 
-            if (length == 0)
+            if (length < min_word_size)
                 continue;
 
             words[offset] = BytesRef{word_start, length};

From f1d433be4c81981af429379d4408e5dd381623be Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 21:03:08 +0300
Subject: [PATCH 460/504] Add min word size.

---
 src/Functions/FunctionsStringHash.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index eba985bd1e8..9c0b9f79448 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -335,7 +335,7 @@ struct SimHashImpl
 template <bool UTF8, bool Ngram, bool CaseInsensitive>
 struct MinHashImpl
 {
-    static constexpr size_t min_word_size = 3;
+    static constexpr size_t min_word_size = 4;
 
     template<typename Comp>
     struct Heap

From 2f9c5cf7455a3ec6ba562862b7c8657a4a86ca9a Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 21:07:03 +0300
Subject: [PATCH 461/504] Update test.

---
 .../0_stateless/01016_simhash_minhash.sql        | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/tests/queries/0_stateless/01016_simhash_minhash.sql b/tests/queries/0_stateless/01016_simhash_minhash.sql
index e27c851f9bb..225bdd2bdc6 100644
--- a/tests/queries/0_stateless/01016_simhash_minhash.sql
+++ b/tests/queries/0_stateless/01016_simhash_minhash.sql
@@ -83,13 +83,13 @@ SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimHashUTF
 SELECT 'ngramSimHashCaseInsensitiveUTF8';
 SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramSimHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleSimHash';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHash(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHash(s, 2) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleSimHashCaseInsensitive';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashCaseInsensitive(s, 2) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleSimHashUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashUTF8(s, 2) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleSimHashCaseInsensitiveUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleSimHashCaseInsensitiveUTF8(s, 2) as h FROM defaults GROUP BY h;
 
 SELECT 'ngramMinHash';
 SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHash(s) as h FROM defaults GROUP BY h;
@@ -100,12 +100,12 @@ SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHashUTF
 SELECT 'ngramMinHashCaseInsensitiveUTF8';
 SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), ngramMinHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleMinHash';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHash(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHash(s, 2, 3) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleMinHashCaseInsensitive';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashCaseInsensitive(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashCaseInsensitive(s, 2, 3) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleMinHashUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashUTF8(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashUTF8(s, 2, 3) as h FROM defaults GROUP BY h;
 SELECT 'wordShingleMinHashCaseInsensitiveUTF8';
-SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashCaseInsensitiveUTF8(s) as h FROM defaults GROUP BY h;
+SELECT arrayStringConcat(groupArray(s), '\n:::::::\n'), count(), wordShingleMinHashCaseInsensitiveUTF8(s, 2, 3) as h FROM defaults GROUP BY h;
 
 DROP TABLE defaults;

From 2cfbe056a04a23e8a506eff757e0a6cb2f4f4baf Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Fri, 25 Dec 2020 21:08:45 +0300
Subject: [PATCH 462/504] Update test.

---
 .../01016_simhash_minhash.reference           | 283 +++++++++---------
 1 file changed, 140 insertions(+), 143 deletions(-)

diff --git a/tests/queries/0_stateless/01016_simhash_minhash.reference b/tests/queries/0_stateless/01016_simhash_minhash.reference
index edd5afc1af7..3a668e6dcdb 100644
--- a/tests/queries/0_stateless/01016_simhash_minhash.reference
+++ b/tests/queries/0_stateless/01016_simhash_minhash.reference
@@ -1,144 +1,141 @@
-0
-2718169299
-2718169299
-1315333491
-1099965843
-5746351769509927967
-5746351769509927967
-8347269581771603092
-6041373934059725027
-(17178276249054052155,8864230932371215121)
-(14133097226001036899,7985237721476952807)
-(14133097226001036899,7985237721476952807)
-(4661257206578284012,15229878657590021759)
-(3087743741749030713,11631667950302077749)
-(11923981719512934676,1193672187225825732)
-(11923981719512934676,1193672187225825732)
-(17970606678134635272,3825545538448404526)
-(9422952829151664974,568010773615758889)
-2548869326
-2548869326
-401385678
-401385710
-2652202579
-2652235347
-2984455347
-2984488115
-12804820948382413807
-12804820948919350245
-11651601468065149391
-11651600368014488527
-18377198011227067677
-18233505035951822655
-5501050600367972694
-5501050600367972692
-(8590465925632898311,12699049311112305995)
-(8590465925632898311,15828587343885202011)
-(8590465925632898311,15824051019631343049)
-(8590465925632898311,12699049311222825283)
-(217966158370437743,14452995911556652133)
-(217966158370437743,14452995911556652133)
-(2170210914777151141,5341809779339553313)
-(12469866236432988845,5341809779339553313)
-(12271157076799061825,5514511977572226426)
-(11639913962681153226,2767634094725305612)
-(12271157075024394466,17994666970078080114)
-(12271157077109587702,13572452308677868240)
-(6252006845407214340,13538761942960976531)
-(13795977174459370328,6392395597500134035)
-(16118993428517222971,13602445809406467)
-(16118993428517222971,13602445809406467)
+18446744073709551615
+130877626
+130877626
+2414681787
+2414681787
+3795742796
+3795742796
+3795742796
+3795742796
+(10693559443859979498,10693559443859979498)
+(12862934800683464900,12912608544812513109)
+(12862934800683464900,12912608544812513109)
+(5701637312405877447,12912608544812513109)
+(5701637312405877447,12912608544812513109)
+(17357047205102710216,17357047205102710216)
+(17357047205102710216,17357047205102710216)
+(17357047205102710216,17357047205102710216)
+(17357047205102710216,17357047205102710216)
+3562273581
+3579050789
+3562257197
+3562258213
+3579050797
+3579050757
+3562258221
+3562258181
+3004171816
+2584740395
+437257770
+2651981610
+3004171816
+2584740395
+437257770
+2651981610
+(17614245890954671019,12771214424940442770)
+(17614245890954671019,12771214424940442770)
+(7128473921279637957,12771214424940442770)
+(7128473921279637957,12771214424940442770)
+(17614245890954671019,12771214424940442770)
+(17614245890954671019,12771214424940442770)
+(7128473921279637957,12771214424940442770)
+(7128473921279637957,12771214424940442770)
+(14260447771268573594,5578182242585518316)
+(14260447771268573594,16377939020851853906)
+(4363920713808688881,5013693163726625177)
+(14260447771268573594,3863279269132177973)
+(14260447771268573594,5578182242585518316)
+(14260447771268573594,16377939020851853906)
+(4363920713808688881,5013693163726625177)
+(14260447771268573594,3863279269132177973)
 uniqExact	6
-ngramSimhash
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	938403918
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	904817231
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	904849486
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	938469966
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	938404430
-ngramSimhashCaseInsensitive
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	938453071
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	938453599
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	938404430
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	636382047
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	938388046
-ngramSimhashUTF8
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2400625214
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2669060670
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	2671174174
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2669060798
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	2635506238
-ngramSimhashCaseInsensitiveUTF8
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2984307934
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	2967514366
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2715855070
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2967529694
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2984290526
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2984306910
-wordShingleSimhash
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384813566025024242
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2393820766427040734
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	2421405261516400471
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384883934767174398
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384813567165864670
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2384813567098766070
-wordShingleSimhashCaseInsensitive
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	11635224793909957342
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617192803208139478
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617192803208151794
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617192803208151766
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	3006891407629799254
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	11617263171950236406
-wordShingleSimhashUTF8
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	9097818277104946605
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9084246141658271116
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9084247241171471628
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	9088752215857929613
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9093255814816009484
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	9084247481822285196
-wordShingleSimhashCaseInsensitiveUTF8
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14788772559981154978
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	14497164445320454820
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14500537785782895266
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14787646625647636642
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	14500016612976573090
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	14787956717160870888
-ngramMinhash
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(15568933215262012353,16287411738807860353)
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(9473589826959436958,14264235017873782379)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(261441656340606110,13387826928927239258)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(261441656340606110,3305790294064680121)
-ngramMinhashCaseInsensitive
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(15568933215262012353,16287411738807860353)
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(9473589826959436958,14264235017873782379)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(3051755284325985438,3305790294064680121)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(3051755284325985438,13387826928927239258)
-ngramMinhashUTF8
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	4	(309830857064065611,7476109060377919216)
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(309830856946430871,7521913981442105351)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(309830857559697399,7476109060377919216)
-ngramMinhashCaseInsensitiveUTF8
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(13010809262502929096,2266175201446733829)
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	5	(16827851229372179144,976408052548769549)
-wordShingleMinhash
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(14343822344862533053,11776483993821900250)
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(18417749332128868312,11776483993821900250)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(18417749329907528200,14156831980621923226)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(4600092690178227586,11776483993821900250)
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(12998011837685887081,1565093152297016105)
-wordShingleMinhashCaseInsensitive
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(12998011837880940480,1565093152297016105)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(1100751419997894255,15225006848401474458)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(1100751419777226283,12993805708561478711)
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(1260401089202135898,12993805709529540523)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(1638964264353944555,12993805708561478711)
-wordShingleMinhashUTF8
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(742280067319112377,14237963017046410351)
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(7237654052534217600,14400297883226437452)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(742280067319112377,17574811665615962276)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(3458625375707825328,17574811665615962276)
-wordShingleMinhashCaseInsensitiveUTF8
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7032848390598450936,5104668712725998486)
-ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(15582670464629505464,13034678298246801511)
-ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(9935434838523508980,7648038926638343017)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	(7032848390598450936,16870743692447971238)
-ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7302041809563941951,6856814412450461959)
+ngramSimHash
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	3906262823
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2857686823
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	676648743
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	1012193063
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	3092567843
+ngramSimHashCaseInsensitive
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2891240999
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	3092567591
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	2824132391
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	3908359975
+ngramSimHashUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	3159676711
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	2	676648743
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	1012193063
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	2924795687
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	3897874215
+ngramSimHashCaseInsensitiveUTF8
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	3906262823
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	3092567591
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	2824132391
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	2891241255
+wordShingleSimHash
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	857724390
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	404215270
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	991679910
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	425963587
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	404215014
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	563598566
+wordShingleSimHashCaseInsensitive
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	959182215
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	429118950
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	421737795
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	964941252
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	965465540
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	420713958
+wordShingleSimHashUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	857724390
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	404215270
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	991679910
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	425963587
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	404215014
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	563598566
+wordShingleSimHashCaseInsensitiveUTF8
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	959182215
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	429118950
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	421737795
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	964941252
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	965465540
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	420713958
+ngramMinHash
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(6021986790841777095,17443426065825246292)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(13225377334870249827,17443426065825246292)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(4388091710993602029,17613327300639166679)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7962672159337006560,17443426065825246292)
+ngramMinHashCaseInsensitive
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(6021986790841777095,8535005350590298790)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(13225377334870249827,8535005350590298790)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(4388091710993602029,17613327300639166679)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7962672159337006560,8535005350590298790)
+ngramMinHashUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(6021986790841777095,17443426065825246292)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(13225377334870249827,17443426065825246292)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(4388091710993602029,17613327300639166679)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7962672159337006560,17443426065825246292)
+ngramMinHashCaseInsensitiveUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(6021986790841777095,8535005350590298790)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(13225377334870249827,8535005350590298790)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(4388091710993602029,17613327300639166679)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(7962672159337006560,8535005350590298790)
+wordShingleMinHash
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(18148981179837829400,14581416672396321264)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(16224204290372720939,13975393268888698430)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(5044918525503962090,12338022931991160906)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(18148981179837829400,6048943706095721476)
+wordShingleMinHashCaseInsensitive
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(15504011608613565061,6048943706095721476)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(16224204290372720939,13975393268888698430)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(5044918525503962090,3381836163833256482)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(15504011608613565061,14581416672396321264)
+wordShingleMinHashUTF8
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(18148981179837829400,14581416672396321264)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(16224204290372720939,13975393268888698430)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(5044918525503962090,12338022931991160906)
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(18148981179837829400,6048943706095721476)
+wordShingleMinHashCaseInsensitiveUTF8
+ClickHouse makes full use of all available hardware to process every request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (only used columns after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid single points of failure. Downtime for one site or the entire data center will not affect the system\'s read and write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they immediately become available for building reports. The SQL dialect allows you to express the desired result without resorting to any non-standard APIs that can be found in some alternative systems.	1	(15504011608613565061,6048943706095721476)
+ClickHouse uses all available hardware to its full potential to process each query as fast as possible. Peak processing performance for a single query stands at more than 2 terabytes per second (after decompression, only used columns). In distributed setup reads are automatically balanced among healthy replicas to avoid increasing latency.\nClickHouse supports multi-master asynchronous replication and can be deployed across multiple datacenters. All nodes are equal, which allows avoiding having single points of failure. Downtime of a single node or the whole datacenter wont affect the systems availability for both reads and writes.\nClickHouse is simple and works out-of-the-box. It streamlines all your data processing: ingest all your structured data into the system and it becomes instantly available for building reports. SQL dialect allows expressing the desired result without involving any custom non-standard API that could be found in some alternative systems.	1	(16224204290372720939,13975393268888698430)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all of your structured data into the system, and it is immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all your structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.\n:::::::\nClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (using columns after decompression only). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the read / write availability of the system.\nClickHouse is simple and works out of the box. It simplifies all processing of your data: it loads all structured data into the system and immediately becomes available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	3	(5044918525503962090,3381836163833256482)
+ClickHouse makes full use of all available hardware to process each request as quickly as possible. Peak performance for a single query is over 2 terabytes per second (used columns only after unpacking). In a distributed setup, reads are automatically balanced across healthy replicas to avoid increased latency.\nClickHouse supports asynchronous multi-master replication and can be deployed across multiple data centers. All nodes are equal to avoid a single point of failure. Downtime for one site or the entire data center will not affect the system\'s read / write availability.\nClickHouse is simple and works out of the box. It simplifies all the processing of your data: it loads all your structured data into the system, and they are immediately available for building reports. The SQL dialect allows you to express the desired result without resorting to any of the non-standard APIs found in some alternative systems.	1	(15504011608613565061,14581416672396321264)

From 34e298006d515a7a832de133b8715725fecef39b Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 22:24:26 +0300
Subject: [PATCH 463/504] Update to master boost

---
 contrib/boost | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/boost b/contrib/boost
index 729dc66b11a..8e259cd2a6b 160000
--- a/contrib/boost
+++ b/contrib/boost
@@ -1 +1 @@
-Subproject commit 729dc66b11a006df6c3b32294635cc2572fea594
+Subproject commit 8e259cd2a6b60d75dd17e73432f11bb7b9351bb1

From 4b51a401ec52425e6c47c30c875c69e06b614153 Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Fri, 25 Dec 2020 23:36:44 +0300
Subject: [PATCH 464/504] Update cctz to 2020e

---
 contrib/cctz | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/contrib/cctz b/contrib/cctz
index 260ba195ef6..c0f1bcb97fd 160000
--- a/contrib/cctz
+++ b/contrib/cctz
@@ -1 +1 @@
-Subproject commit 260ba195ef6c489968bae8c88c62a67cdac5ff9d
+Subproject commit c0f1bcb97fd2782f7c3f972fadd5aad5affac4b8

From 762e757557fae651adc1839744fbe061a99ca0ae Mon Sep 17 00:00:00 2001
From: UnamedRus <UnamedRus@users.noreply.github.com>
Date: Sat, 26 Dec 2020 01:46:47 +0300
Subject: [PATCH 465/504] Update docs/ru/interfaces/cli.md

Co-authored-by: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
---
 docs/ru/interfaces/cli.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/interfaces/cli.md b/docs/ru/interfaces/cli.md
index 8556182af49..03d669c5def 100644
--- a/docs/ru/interfaces/cli.md
+++ b/docs/ru/interfaces/cli.md
@@ -93,7 +93,7 @@ clickhouse-client --param_parName="[1, 2]"  -q "SELECT * FROM table WHERE a = {p
 ```
 
 -   `name` — идентификатор подстановки. В консольном клиенте его следует использовать как часть имени параметра `--param_<name> = value`.
--   `data type` — [тип данных](../sql-reference/data-types/index.md) значения. Например, структура данных `(integer, ('string', integer))` может иметь тип данных `Tuple(UInt8, Tuple(String, UInt8))` ([целочисленный](../sql-reference/data-types/int-uint.md) тип может быть и другим). В качестве параметра можно передать название столбца, таблицы и базы данных, в этом случае используется `Identifier` как тип данных.
+-   `data type` — [тип данных](../sql-reference/data-types/index.md) значения. Например, структура данных `(integer, ('string', integer))` может иметь тип данных `Tuple(UInt8, Tuple(String, UInt8))` ([целочисленный](../sql-reference/data-types/int-uint.md) тип может быть и другим). В качестве параметра можно передать название столбца, таблицы и базы данных, в этом случае используется тип данных`Identifier` .
 
 #### Пример {#primer}
 

From a087f23af6acf1f7f68b31c61ec7e5055e2c027b Mon Sep 17 00:00:00 2001
From: UnamedRus <UnamedRus@users.noreply.github.com>
Date: Sat, 26 Dec 2020 01:46:56 +0300
Subject: [PATCH 466/504] Update docs/en/interfaces/cli.md

Co-authored-by: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
---
 docs/en/interfaces/cli.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/en/interfaces/cli.md b/docs/en/interfaces/cli.md
index c0c9b5668c2..3474c22f48e 100644
--- a/docs/en/interfaces/cli.md
+++ b/docs/en/interfaces/cli.md
@@ -87,7 +87,7 @@ Format a query as usual, then place the values that you want to pass from the ap
 ```
 
 -   `name` — Placeholder identifier. In the console client it should be used in app parameters as `--param_<name> = value`.
--   `data type` — [Data type](../sql-reference/data-types/index.md) of the app parameter value. For example, a data structure like `(integer, ('string', integer))` can have the `Tuple(UInt8, Tuple(String, UInt8))` data type (you can also use another [integer](../sql-reference/data-types/int-uint.md) types). It's also possible to pass table, database, column names as parameter, in that case you would need to use `Identifier` as data type.
+-   `data type` — [Data type](../sql-reference/data-types/index.md) of the app parameter value. For example, a data structure like `(integer, ('string', integer))` can have the `Tuple(UInt8, Tuple(String, UInt8))` data type (you can also use another [integer](../sql-reference/data-types/int-uint.md) types). It's also possible to pass table, database, column names as a parameter, in that case you would need to use `Identifier` as a data type.
 
 #### Example {#example}
 

From 51b76f73671071c2065c88b3f193de2a942c9b43 Mon Sep 17 00:00:00 2001
From: Anna <annaconda@bk.ru>
Date: Sat, 26 Dec 2020 15:43:09 +0300
Subject: [PATCH 467/504] add desc for NULL modifiers

---
 docs/en/operations/settings/settings.md          | 11 +++++++++++
 docs/en/sql-reference/statements/create/table.md |  8 ++++++++
 2 files changed, 19 insertions(+)

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 7f0a61ccd42..4528f7010bf 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2470,4 +2470,15 @@ Possible values:
 
 Default value: `0`.
 
+## data_type_default_nullable {#data_type_default_nullable}
+
+Allows data types without explicit modifiers in table definition will be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable).
+
+Possible values:
+
+- 1 — `Nullable`-type expressions are allowed in a table definition.
+- 0 — `Nullable`-type expressions are not allowed in a table definition.
+
+Default value: `0`.
+
 [Original article](https://clickhouse.tech/docs/en/operations/settings/settings/) <!-- hide -->
diff --git a/docs/en/sql-reference/statements/create/table.md b/docs/en/sql-reference/statements/create/table.md
index e9952fc76fd..55e1ba66edc 100644
--- a/docs/en/sql-reference/statements/create/table.md
+++ b/docs/en/sql-reference/statements/create/table.md
@@ -99,6 +99,14 @@ If you add a new column to a table but later change its default expression, the
 
 It is not possible to set default values for elements in nested data structures.
 
+## NULL Or NOT NULL Modifiers
+
+`NULL` and `NOT NULL` modifiers after data type in table definition allow or do not allow it to be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable). 
+
+If the type is not `Nullable` and if `NULL` is specified, it will be treated as `Nullable`; if `NOT NULL` is specified, then no. For example, `INT NULL` is the same as `Nullable(INT)`. If the type is `Nullable` and `NULL` or `NOT NULL` modifiers are specified, the exception will be thrown.
+
+See also [data_type_default_nullable](../../operations/settings/settings.md#data_type_default_nullable) setting.
+
 ## Primary Key {#primary-key}
 
 You can define a [primary key](../../../engines/table-engines/mergetree-family/mergetree.md#primary-keys-and-indexes-in-queries) when creating a table. Primary key can be specified in two ways: 

From 99b50dbba3c452cfa93bc5b77b915241ab452c76 Mon Sep 17 00:00:00 2001
From: romanzhukov <romanzhukov@yandex-team.ru>
Date: Sat, 26 Dec 2020 16:10:37 +0300
Subject: [PATCH 468/504] DOCSUP-5043: Fix PR comments.

---
 .../operations/server-configuration-parameters/settings.md  | 6 +++---
 docs/ru/commercial/cloud.md                                 | 5 ++---
 .../operations/server-configuration-parameters/settings.md  | 4 ++--
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/docs/en/operations/server-configuration-parameters/settings.md b/docs/en/operations/server-configuration-parameters/settings.md
index 1877e474555..4095e84627a 100644
--- a/docs/en/operations/server-configuration-parameters/settings.md
+++ b/docs/en/operations/server-configuration-parameters/settings.md
@@ -80,7 +80,7 @@ List of prefixes for [custom settings](../../operations/settings/index.md#custom
 
 -   [Custom settings](../../operations/settings/index.md#custom_settings)
 
-## core_dump
+## core_dump {#server_configuration_parameters-core_dump}
 
 Configures soft limit for core dump file size.
 
@@ -88,7 +88,7 @@ Possible values:
 
 -   Positive integer.
 
-Default value: `1073741824`.
+Default value: `1073741824` (1 GB).
 
 !!! info "Note"
     Hard limit is configured via system tools
@@ -440,7 +440,7 @@ Limits total RAM usage by the ClickHouse server.
 Possible values:
 
 -   Positive integer.
--   0 (auto).
+-   0 — Auto.
 
 Default value: `0`.
 
diff --git a/docs/ru/commercial/cloud.md b/docs/ru/commercial/cloud.md
index 8d02a6332a7..885aac97312 100644
--- a/docs/ru/commercial/cloud.md
+++ b/docs/ru/commercial/cloud.md
@@ -21,13 +21,12 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 ## Altinity.Cloud {#altinity.cloud}
 
 [Altinity.Cloud](https://altinity.com/cloud-database/) — это полностью управляемый ClickHouse-as-a-Service для публичного облака Amazon.
+
 -   Быстрое развертывание кластеров ClickHouse на ресурсах Amazon
--   Easy scale-out/scale-in as well as vertical scaling of nodes
 -   Легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов
 -   Изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC
 -   Настраиваемые типы и объемы хранилищ
--   Cross-AZ scaling for performance and high availability
--   Cross-AZ масштабирование для повышения производительности и высокой доступности
+-   Cross-AZ масштабирование для повышения производительности и обеспечения высокой доступности
 -   Встроенный мониторинг и редактор SQL-запросов
 
 {## [Оригинальная статья](https://clickhouse.tech/docs/ru/commercial/cloud/) ##}
diff --git a/docs/ru/operations/server-configuration-parameters/settings.md b/docs/ru/operations/server-configuration-parameters/settings.md
index affe09c7703..b75f0371223 100644
--- a/docs/ru/operations/server-configuration-parameters/settings.md
+++ b/docs/ru/operations/server-configuration-parameters/settings.md
@@ -80,7 +80,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 
 -   [Пользовательские настройки](../../operations/settings/index.md#custom_settings)
 
-## core_dump
+## core_dump {#server_configuration_parameters-core_dump}
 
 Задает мягкое ограничение для размера файла дампа ядра.
 
@@ -88,7 +88,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 
 -   Положительное целое число.
 
-Значение по умолчанию: `1073741824`.
+Значение по умолчанию: `1073741824` (1 ГБ).
 
 !!! info "Примечание"
     Жесткое ограничение настраивается с помощью системных инструментов.

From 46791d49080af400167fe78e45152eaf72813590 Mon Sep 17 00:00:00 2001
From: vdimir <vdimirc@gmail.com>
Date: Sat, 26 Dec 2020 19:16:57 +0000
Subject: [PATCH 469/504] Do not throw logical error from IPAddressDictionary
 ctor

---
 src/Dictionaries/IPAddressDictionary.cpp | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/src/Dictionaries/IPAddressDictionary.cpp b/src/Dictionaries/IPAddressDictionary.cpp
index e04b9111ff1..6c8a7f15c62 100644
--- a/src/Dictionaries/IPAddressDictionary.cpp
+++ b/src/Dictionaries/IPAddressDictionary.cpp
@@ -21,10 +21,12 @@ namespace DB
 {
 namespace ErrorCodes
 {
+    extern const int BAD_ARGUMENTS;
+    extern const int CANNOT_PARSE_INPUT_ASSERTION_FAILED;
+    extern const int CANNOT_PARSE_NUMBER;
+    extern const int DICTIONARY_IS_EMPTY;
     extern const int LOGICAL_ERROR;
     extern const int TYPE_MISMATCH;
-    extern const int BAD_ARGUMENTS;
-    extern const int DICTIONARY_IS_EMPTY;
 }
 
 namespace
@@ -89,9 +91,9 @@ static std::pair<Poco::Net::IPAddress, UInt8> parseIPFromString(const std::strin
             const auto * addr_str_end = addr_str.data() + addr_str.size();
             auto [p, ec] = std::from_chars(addr_str.data() + pos + 1, addr_str_end, prefix);
             if (p != addr_str_end)
-                throw DB::Exception("extra characters at the end", ErrorCodes::LOGICAL_ERROR);
+                throw DB::Exception("extra characters at the end", ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
             if (ec != std::errc())
-                throw DB::Exception("mask is not a valid number", ErrorCodes::LOGICAL_ERROR);
+                throw DB::Exception("mask is not a valid number", ErrorCodes::CANNOT_PARSE_NUMBER);
 
             addr = addr & Poco::Net::IPAddress(prefix, addr.family());
             return {addr, prefix};
@@ -103,7 +105,7 @@ static std::pair<Poco::Net::IPAddress, UInt8> parseIPFromString(const std::strin
     catch (Poco::Exception & ex)
     {
         throw DB::Exception("can't parse address \"" + std::string(addr_str) + "\": " + ex.what(),
-            ErrorCodes::LOGICAL_ERROR);
+            ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
     }
 }
 

From 5c459e74c941c83097c6e33e5c5d1f280024a53c Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Sun, 27 Dec 2020 11:43:27 +0300
Subject: [PATCH 470/504] Update IPAddressDictionary.cpp

---
 src/Dictionaries/IPAddressDictionary.cpp | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/Dictionaries/IPAddressDictionary.cpp b/src/Dictionaries/IPAddressDictionary.cpp
index 6c8a7f15c62..2ed39f2784d 100644
--- a/src/Dictionaries/IPAddressDictionary.cpp
+++ b/src/Dictionaries/IPAddressDictionary.cpp
@@ -91,9 +91,9 @@ static std::pair<Poco::Net::IPAddress, UInt8> parseIPFromString(const std::strin
             const auto * addr_str_end = addr_str.data() + addr_str.size();
             auto [p, ec] = std::from_chars(addr_str.data() + pos + 1, addr_str_end, prefix);
             if (p != addr_str_end)
-                throw DB::Exception("extra characters at the end", ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
+                throw DB::Exception("Extra characters at the end of IP address", ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
             if (ec != std::errc())
-                throw DB::Exception("mask is not a valid number", ErrorCodes::CANNOT_PARSE_NUMBER);
+                throw DB::Exception("Mask for IP address is not a valid number", ErrorCodes::CANNOT_PARSE_NUMBER);
 
             addr = addr & Poco::Net::IPAddress(prefix, addr.family());
             return {addr, prefix};
@@ -104,7 +104,7 @@ static std::pair<Poco::Net::IPAddress, UInt8> parseIPFromString(const std::strin
     }
     catch (Poco::Exception & ex)
     {
-        throw DB::Exception("can't parse address \"" + std::string(addr_str) + "\": " + ex.what(),
+        throw DB::Exception("Can't parse address \"" + std::string(addr_str) + "\": " + ex.what(),
             ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED);
     }
 }

From be4832de6dc2fcbf0c9910022bcdbf4c2a96d0ad Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 12:54:59 +0300
Subject: [PATCH 471/504] Try fix test.

---
 src/Functions/FunctionsStringHash.cpp | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/src/Functions/FunctionsStringHash.cpp b/src/Functions/FunctionsStringHash.cpp
index 9c0b9f79448..e389e2f7f98 100644
--- a/src/Functions/FunctionsStringHash.cpp
+++ b/src/Functions/FunctionsStringHash.cpp
@@ -262,6 +262,9 @@ struct SimHashImpl
                 words.emplace_back(BytesRef{word_start, length});
         }
 
+        if (words.empty())
+            return 0;
+
         UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);
         updateFingerVector(finger_vec, hash_value);
 
@@ -488,6 +491,9 @@ struct MinHashImpl
                 words.emplace_back(BytesRef{word_start, length});
         }
 
+        if (words.empty())
+            return;
+
         UInt64 hash_value = Hash::shingleHash<CaseInsensitive>(words);
         {
             const UInt8 * shingle_start = words.front().data;

From d6f501f497f728ba5563ebc38d79396a535e4b45 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 13:30:00 +0300
Subject: [PATCH 472/504] Try fix Pipeline stuck

---
 src/Processors/QueryPlan/AddingDelayedSourceStep.cpp | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/src/Processors/QueryPlan/AddingDelayedSourceStep.cpp b/src/Processors/QueryPlan/AddingDelayedSourceStep.cpp
index 7247dc55ce2..99db95059a0 100644
--- a/src/Processors/QueryPlan/AddingDelayedSourceStep.cpp
+++ b/src/Processors/QueryPlan/AddingDelayedSourceStep.cpp
@@ -32,6 +32,11 @@ void AddingDelayedSourceStep::transformPipeline(QueryPipeline & pipeline)
 {
     source->setQueryPlanStep(this);
     pipeline.addDelayedStream(source);
+
+    /// Now, after adding delayed stream, it has implicit dependency on other port.
+    /// Here we add resize processor to remove this dependency.
+    /// Otherwise, if we add MergeSorting + MergingSorted transform to pipeline, we could get `Pipeline stuck`
+    pipeline.resize(pipeline.getNumStreams(), true);
 }
 
 }

From a85cfa81db00469e2985f8fc6886d95ac70c6e1f Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:02:21 +0300
Subject: [PATCH 473/504] Try fix Pipeline stuck

---
 src/Processors/DelayedPortsProcessor.cpp | 27 ++++++++++++++++++------
 src/Processors/DelayedPortsProcessor.h   |  2 +-
 src/Processors/QueryPipeline.cpp         |  2 +-
 3 files changed, 22 insertions(+), 9 deletions(-)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index 5a6297413e1..a634f348771 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -3,8 +3,10 @@
 namespace DB
 {
 
-DelayedPortsProcessor::DelayedPortsProcessor(const Block & header, size_t num_ports, const PortNumbers & delayed_ports)
-    : IProcessor(InputPorts(num_ports, header), OutputPorts(num_ports, header))
+DelayedPortsProcessor::DelayedPortsProcessor(
+    const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_delayed_ports_empty)
+    : IProcessor(InputPorts(num_ports, header),
+                 OutputPorts(num_ports - (assert_delayed_ports_empty ? delayed_ports.size() : 0), header))
     , num_delayed(delayed_ports.size())
 {
     port_pairs.resize(num_ports);
@@ -14,9 +16,13 @@ DelayedPortsProcessor::DelayedPortsProcessor(const Block & header, size_t num_po
     for (size_t i = 0; i < num_ports; ++i)
     {
         port_pairs[i].input_port = &*input_it;
-        port_pairs[i].output_port = &*output_it;
         ++input_it;
-        ++output_it;
+
+        if (output_it != outputs.end())
+        {
+            port_pairs[i].output_port = &*output_it;
+            ++output_it;
+        }
     }
 
     for (const auto & delayed : delayed_ports)
@@ -34,7 +40,7 @@ bool DelayedPortsProcessor::processPair(PortsPair & pair)
         }
     };
 
-    if (pair.output_port->isFinished())
+    if (pair.output_port && pair.output_port->isFinished())
     {
         pair.input_port->close();
         finish();
@@ -43,17 +49,24 @@ bool DelayedPortsProcessor::processPair(PortsPair & pair)
 
     if (pair.input_port->isFinished())
     {
-        pair.output_port->finish();
+        if (pair.output_port)
+            pair.output_port->finish();
         finish();
         return false;
     }
 
-    if (!pair.output_port->canPush())
+    if (pair.output_port && !pair.output_port->canPush())
         return false;
 
     pair.input_port->setNeeded();
     if (pair.input_port->hasData())
+    {
+        if (!pair.output_port)
+            throw Exception(ErrorCodes::LOGICAL_ERROR,
+                            "Input port for DelayedPortsProcessor is assumed to have no data, but it has one");
+
         pair.output_port->pushData(pair.input_port->pullData());
+    }
 
     return true;
 }
diff --git a/src/Processors/DelayedPortsProcessor.h b/src/Processors/DelayedPortsProcessor.h
index 44dd632f8a8..77409332d8e 100644
--- a/src/Processors/DelayedPortsProcessor.h
+++ b/src/Processors/DelayedPortsProcessor.h
@@ -11,7 +11,7 @@ namespace DB
 class DelayedPortsProcessor : public IProcessor
 {
 public:
-    DelayedPortsProcessor(const Block & header, size_t num_ports, const PortNumbers & delayed_ports);
+    DelayedPortsProcessor(const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_delayed_ports_empty = false);
 
     String getName() const override { return "DelayedPorts"; }
 
diff --git a/src/Processors/QueryPipeline.cpp b/src/Processors/QueryPipeline.cpp
index ffcab2e7a8d..ff086fd5b11 100644
--- a/src/Processors/QueryPipeline.cpp
+++ b/src/Processors/QueryPipeline.cpp
@@ -298,7 +298,7 @@ void QueryPipeline::addPipelineBefore(QueryPipeline pipeline)
     pipes.emplace_back(QueryPipeline::getPipe(std::move(pipeline)));
     pipe = Pipe::unitePipes(std::move(pipes), collected_processors);
 
-    auto processor = std::make_shared<DelayedPortsProcessor>(getHeader(), pipe.numOutputPorts(), delayed_streams);
+    auto processor = std::make_shared<DelayedPortsProcessor>(getHeader(), pipe.numOutputPorts(), delayed_streams, true);
     addTransform(std::move(processor));
 }
 

From b19146065f4f9ac424e1909046edf67f45feb6c8 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:06:14 +0300
Subject: [PATCH 474/504] Try fix Pipeline stuck

---
 src/Processors/DelayedPortsProcessor.cpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index a634f348771..d6a773d2390 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -11,6 +11,9 @@ DelayedPortsProcessor::DelayedPortsProcessor(
 {
     port_pairs.resize(num_ports);
 
+    for (const auto & delayed : delayed_ports)
+        port_pairs[delayed].is_delayed = true;
+
     auto input_it = inputs.begin();
     auto output_it = outputs.begin();
     for (size_t i = 0; i < num_ports; ++i)
@@ -18,15 +21,12 @@ DelayedPortsProcessor::DelayedPortsProcessor(
         port_pairs[i].input_port = &*input_it;
         ++input_it;
 
-        if (output_it != outputs.end())
+        if (!port_pairs[i].is_delayed || !assert_delayed_ports_empty)
         {
             port_pairs[i].output_port = &*output_it;
             ++output_it;
         }
     }
-
-    for (const auto & delayed : delayed_ports)
-        port_pairs[delayed].is_delayed = true;
 }
 
 bool DelayedPortsProcessor::processPair(PortsPair & pair)

From 30bbba6ed5d462840c3a33be5e6e4197161b7f29 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:13:17 +0300
Subject: [PATCH 475/504] Try fix Pipeline stuck

---
 src/Processors/DelayedPortsProcessor.cpp | 7 +++++--
 src/Processors/DelayedPortsProcessor.h   | 2 ++
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index d6a773d2390..a2dc833b039 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -10,6 +10,7 @@ DelayedPortsProcessor::DelayedPortsProcessor(
     , num_delayed(delayed_ports.size())
 {
     port_pairs.resize(num_ports);
+    output_to_pair.reserve(outputs.size());
 
     for (const auto & delayed : delayed_ports)
         port_pairs[delayed].is_delayed = true;
@@ -24,6 +25,7 @@ DelayedPortsProcessor::DelayedPortsProcessor(
         if (!port_pairs[i].is_delayed || !assert_delayed_ports_empty)
         {
             port_pairs[i].output_port = &*output_it;
+            output_to_pair.push_back(i);
             ++output_it;
         }
     }
@@ -78,8 +80,9 @@ IProcessor::Status DelayedPortsProcessor::prepare(const PortNumbers & updated_in
 
     for (const auto & output_number : updated_outputs)
     {
-        if (!skip_delayed || !port_pairs[output_number].is_delayed)
-            need_data = processPair(port_pairs[output_number]) || need_data;
+        auto pair_num = output_to_pair[output_number];
+        if (!skip_delayed || !port_pairs[pair_num].is_delayed)
+            need_data = processPair(port_pairs[pair_num]) || need_data;
     }
 
     for (const auto & input_number : updated_inputs)
diff --git a/src/Processors/DelayedPortsProcessor.h b/src/Processors/DelayedPortsProcessor.h
index 77409332d8e..1bcce20b5fa 100644
--- a/src/Processors/DelayedPortsProcessor.h
+++ b/src/Processors/DelayedPortsProcessor.h
@@ -31,6 +31,8 @@ private:
     size_t num_delayed;
     size_t num_finished = 0;
 
+    std::vector<size_t> output_to_pair;
+
     bool processPair(PortsPair & pair);
 };
 

From 7064a8500b681f516f5d7de8ae29539ada987440 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:16:40 +0300
Subject: [PATCH 476/504] Try fix Pipeline stuck

---
 src/Processors/DelayedPortsProcessor.cpp | 6 +++---
 src/Processors/DelayedPortsProcessor.h   | 2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index a2dc833b039..7689dd85245 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -4,9 +4,9 @@ namespace DB
 {
 
 DelayedPortsProcessor::DelayedPortsProcessor(
-    const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_delayed_ports_empty)
+    const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_main_ports_empty)
     : IProcessor(InputPorts(num_ports, header),
-                 OutputPorts(num_ports - (assert_delayed_ports_empty ? delayed_ports.size() : 0), header))
+                 OutputPorts((assert_main_ports_empty ? delayed_ports.size() : num_ports), header))
     , num_delayed(delayed_ports.size())
 {
     port_pairs.resize(num_ports);
@@ -22,7 +22,7 @@ DelayedPortsProcessor::DelayedPortsProcessor(
         port_pairs[i].input_port = &*input_it;
         ++input_it;
 
-        if (!port_pairs[i].is_delayed || !assert_delayed_ports_empty)
+        if (port_pairs[i].is_delayed || !assert_main_ports_empty)
         {
             port_pairs[i].output_port = &*output_it;
             output_to_pair.push_back(i);
diff --git a/src/Processors/DelayedPortsProcessor.h b/src/Processors/DelayedPortsProcessor.h
index 1bcce20b5fa..cb04e84b1ad 100644
--- a/src/Processors/DelayedPortsProcessor.h
+++ b/src/Processors/DelayedPortsProcessor.h
@@ -11,7 +11,7 @@ namespace DB
 class DelayedPortsProcessor : public IProcessor
 {
 public:
-    DelayedPortsProcessor(const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_delayed_ports_empty = false);
+    DelayedPortsProcessor(const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_main_ports_empty = false);
 
     String getName() const override { return "DelayedPorts"; }
 

From bc277892f4e7bc386adaaa41db47bd774ac6da38 Mon Sep 17 00:00:00 2001
From: Anna <annaconda@bk.ru>
Date: Sun, 27 Dec 2020 14:21:58 +0300
Subject: [PATCH 477/504] add desc for encodeXMLComponent

---
 .../functions/string-functions.md             | 32 +++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/docs/en/sql-reference/functions/string-functions.md b/docs/en/sql-reference/functions/string-functions.md
index a846a01f11f..21e6f224b67 100644
--- a/docs/en/sql-reference/functions/string-functions.md
+++ b/docs/en/sql-reference/functions/string-functions.md
@@ -558,4 +558,36 @@ Result:
 └─────┘
 ```
 
+## encodeXMLComponent {#encode-xml-component}
+
+Escape characters to place string into XML text node or attribute.
+
+**Syntax** 
+
+``` sql
+encodeXMLComponent(x)
+```
+
+**Parameters** 
+
+-   `x` — Sequence of characters. [String](../../sql-reference/data-types/string.md).
+
+**Returned value(s)**
+
+-   Sequence of characters.
+
+Type: [String](../../sql-reference/data-types/string.md).
+
+**Example**
+
+Query:
+
+``` sql
+```
+
+Result:
+
+``` text
+```
+
 [Original article](https://clickhouse.tech/docs/en/query_language/functions/string_functions/) <!--hide-->

From da6bfc2cfbc87da0e7c3c71554eec4acb1e4b96e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:22:46 +0300
Subject: [PATCH 478/504] Try fix Pipeline stuck

---
 src/Processors/DelayedPortsProcessor.cpp | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index 7689dd85245..9215409af47 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -78,6 +78,11 @@ IProcessor::Status DelayedPortsProcessor::prepare(const PortNumbers & updated_in
     bool skip_delayed = (num_finished + num_delayed) < port_pairs.size();
     bool need_data = false;
 
+    if (!updated_outputs.empty())
+        for (const auto & pair : port_pairs)
+            if (!pair.output_port)
+                pair.input_port->setNeeded();
+
     for (const auto & output_number : updated_outputs)
     {
         auto pair_num = output_to_pair[output_number];

From 1ebab1dc487bd0985c170efb82f29f87b8ba0b76 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:25:43 +0300
Subject: [PATCH 479/504] Try fix Pipeline stuck

---
 src/Processors/DelayedPortsProcessor.cpp | 7 ++++++-
 src/Processors/DelayedPortsProcessor.h   | 1 +
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index 9215409af47..ce21c5a018c 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -78,11 +78,16 @@ IProcessor::Status DelayedPortsProcessor::prepare(const PortNumbers & updated_in
     bool skip_delayed = (num_finished + num_delayed) < port_pairs.size();
     bool need_data = false;
 
-    if (!updated_outputs.empty())
+    if (!are_inputs_initialized && !updated_outputs.empty())
+    {
+        /// Activate inputs with no output.
         for (const auto & pair : port_pairs)
             if (!pair.output_port)
                 pair.input_port->setNeeded();
 
+        are_inputs_initialized = true;
+    }
+
     for (const auto & output_number : updated_outputs)
     {
         auto pair_num = output_to_pair[output_number];
diff --git a/src/Processors/DelayedPortsProcessor.h b/src/Processors/DelayedPortsProcessor.h
index cb04e84b1ad..3e44c945bd4 100644
--- a/src/Processors/DelayedPortsProcessor.h
+++ b/src/Processors/DelayedPortsProcessor.h
@@ -32,6 +32,7 @@ private:
     size_t num_finished = 0;
 
     std::vector<size_t> output_to_pair;
+    bool are_inputs_initialized = false;
 
     bool processPair(PortsPair & pair);
 };

From 7f1eb506ab70317b3aa701843198668f94ff4695 Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 14:28:13 +0300
Subject: [PATCH 480/504] Added test.

---
 .../0_stateless/01621_sort_after_join_pipeline_stuck.reference   | 0
 .../queries/0_stateless/01621_sort_after_join_pipeline_stuck.sql | 1 +
 2 files changed, 1 insertion(+)
 create mode 100644 tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.reference
 create mode 100644 tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.sql

diff --git a/tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.reference b/tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.reference
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.sql b/tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.sql
new file mode 100644
index 00000000000..ce9ae65e71e
--- /dev/null
+++ b/tests/queries/0_stateless/01621_sort_after_join_pipeline_stuck.sql
@@ -0,0 +1 @@
+SELECT k FROM (SELECT NULL, nullIf(number, 3) AS k, '1048575', (65536, -9223372036854775808), toString(number) AS a FROM system.numbers LIMIT 1048577) AS js1 ANY RIGHT JOIN (SELECT 1.000100016593933, nullIf(number, NULL) AS k, toString(number) AS b FROM system.numbers LIMIT 2, 255) AS js2 USING (k) ORDER BY 257 ASC NULLS LAST FORMAT Null;

From 436df8fe3ac79638fcd2badcb7f851285354e3cf Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Sun, 27 Dec 2020 14:36:08 +0300
Subject: [PATCH 481/504] Check fuzzer tests

---
 .../queries/0_stateless/01593_concurrent_alter_mutations_kill.sh | 1 +
 .../01593_concurrent_alter_mutations_kill_many_replicas.sh       | 1 +
 2 files changed, 2 insertions(+)

diff --git a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh
index dd8eb9d731e..fe243a7e929 100755
--- a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh
+++ b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh
@@ -24,6 +24,7 @@ function alter_thread
 function kill_mutation_thread
 {
     while true; do
+        # find any mutation and kill it
         mutation_id=$($CLICKHOUSE_CLIENT --query "SELECT mutation_id FROM system.mutations WHERE is_done=0 and database='${CLICKHOUSE_DATABASE}' and table='concurrent_mutate_kill' LIMIT 1")
         if [ ! -z "$mutation_id" ]; then
             $CLICKHOUSE_CLIENT --query "KILL MUTATION WHERE mutation_id='$mutation_id'" 1> /dev/null
diff --git a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh
index cfbb183263d..e9513557722 100755
--- a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh
+++ b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh
@@ -36,6 +36,7 @@ function alter_thread
 function kill_mutation_thread
 {
     while true; do
+        # find any mutation and kill it
         mutation_id=$($CLICKHOUSE_CLIENT --query "SELECT mutation_id FROM system.mutations WHERE is_done = 0 and table like 'concurrent_kill_%' and database='${CLICKHOUSE_DATABASE}' LIMIT 1")
         if [ ! -z "$mutation_id" ]; then
             $CLICKHOUSE_CLIENT --query "KILL MUTATION WHERE mutation_id='$mutation_id'" 1> /dev/null

From 45b8ddd7ab1e7cedfc0a2721f874e930a12808fd Mon Sep 17 00:00:00 2001
From: Anna <annaconda@bk.ru>
Date: Sun, 27 Dec 2020 14:47:15 +0300
Subject: [PATCH 482/504] fixed

---
 docs/en/operations/settings/settings.md       |  6 +++---
 .../sql-reference/statements/create/table.md  | 20 +++++++++----------
 2 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 4528f7010bf..7a2c8a487c1 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2472,12 +2472,12 @@ Default value: `0`.
 
 ## data_type_default_nullable {#data_type_default_nullable}
 
-Allows data types without explicit modifiers in table definition will be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable).
+Allows data types without explicit modifiers in column definition will be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable).
 
 Possible values:
 
-- 1 — `Nullable`-type expressions are allowed in a table definition.
-- 0 — `Nullable`-type expressions are not allowed in a table definition.
+- 1 — The data type in a column definition is `Nullable`.
+- 0 — The data type in a column definition is not `Nullable`.
 
 Default value: `0`.
 
diff --git a/docs/en/sql-reference/statements/create/table.md b/docs/en/sql-reference/statements/create/table.md
index 55e1ba66edc..b1a5fdb19b5 100644
--- a/docs/en/sql-reference/statements/create/table.md
+++ b/docs/en/sql-reference/statements/create/table.md
@@ -16,8 +16,8 @@ By default, tables are created only on the current server. Distributed DDL queri
 ``` sql
 CREATE TABLE [IF NOT EXISTS] [db.]table_name [ON CLUSTER cluster]
 (
-    name1 [type1] [DEFAULT|MATERIALIZED|ALIAS expr1] [compression_codec] [TTL expr1],
-    name2 [type2] [DEFAULT|MATERIALIZED|ALIAS expr2] [compression_codec] [TTL expr2],
+    name1 [type1] [NULL|NOT NULL] [DEFAULT|MATERIALIZED|ALIAS expr1] [compression_codec] [TTL expr1],
+    name2 [type2] [NULL|NOT NULL] [DEFAULT|MATERIALIZED|ALIAS expr2] [compression_codec] [TTL expr2],
     ...
 ) ENGINE = engine
 ```
@@ -57,6 +57,14 @@ In all cases, if `IF NOT EXISTS` is specified, the query won’t return an error
 
 There can be other clauses after the `ENGINE` clause in the query. See detailed documentation on how to create tables in the descriptions of [table engines](../../../engines/table-engines/index.md#table_engines).
 
+## NULL Or NOT NULL Modifiers {#null-modifiers}
+
+`NULL` and `NOT NULL` modifiers after data type in column definition allow or do not allow it to be [Nullable](../../../sql-reference/data-types/nullable.md#data_type-nullable). 
+
+If the type is not `Nullable` and if `NULL` is specified, it will be treated as `Nullable`; if `NOT NULL` is specified, then no. For example, `INT NULL` is the same as `Nullable(INT)`. If the type is `Nullable` and `NULL` or `NOT NULL` modifiers are specified, the exception will be thrown.
+
+See also [data_type_default_nullable](../../../operations/settings/settings.md#data_type_default_nullable) setting.
+
 ## Default Values {#create-default-values}
 
 The column description can specify an expression for a default value, in one of the following ways: `DEFAULT expr`, `MATERIALIZED expr`, `ALIAS expr`.
@@ -99,14 +107,6 @@ If you add a new column to a table but later change its default expression, the
 
 It is not possible to set default values for elements in nested data structures.
 
-## NULL Or NOT NULL Modifiers
-
-`NULL` and `NOT NULL` modifiers after data type in table definition allow or do not allow it to be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable). 
-
-If the type is not `Nullable` and if `NULL` is specified, it will be treated as `Nullable`; if `NOT NULL` is specified, then no. For example, `INT NULL` is the same as `Nullable(INT)`. If the type is `Nullable` and `NULL` or `NOT NULL` modifiers are specified, the exception will be thrown.
-
-See also [data_type_default_nullable](../../operations/settings/settings.md#data_type_default_nullable) setting.
-
 ## Primary Key {#primary-key}
 
 You can define a [primary key](../../../engines/table-engines/mergetree-family/mergetree.md#primary-keys-and-indexes-in-queries) when creating a table. Primary key can be specified in two ways: 

From 9c62adcb4b7422d277b6d7d9da268cbb421f561a Mon Sep 17 00:00:00 2001
From: Anna <annaconda@bk.ru>
Date: Sun, 27 Dec 2020 15:26:31 +0300
Subject: [PATCH 483/504] fixed

---
 docs/en/operations/settings/settings.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/docs/en/operations/settings/settings.md b/docs/en/operations/settings/settings.md
index 7a2c8a487c1..81c8787d442 100644
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@@ -2472,12 +2472,12 @@ Default value: `0`.
 
 ## data_type_default_nullable {#data_type_default_nullable}
 
-Allows data types without explicit modifiers in column definition will be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable).
+Allows data types without explicit modifiers [NULL or NOT NULL](../../sql-reference/statements/create/table.md#null-modifiers) in column definition will be [Nullable](../../sql-reference/data-types/nullable.md#data_type-nullable).
 
 Possible values:
 
-- 1 — The data type in a column definition is `Nullable`.
-- 0 — The data type in a column definition is not `Nullable`.
+- 1 — The data types in column definitions are set to `Nullable` by default.
+- 0 — The data types in column definitions are set to not `Nullable` by default.
 
 Default value: `0`.
 

From 2e69757062806a7683eccf76592db50cb471259a Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 16:53:42 +0300
Subject: [PATCH 484/504] Fix style

---
 src/Processors/DelayedPortsProcessor.cpp | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/src/Processors/DelayedPortsProcessor.cpp b/src/Processors/DelayedPortsProcessor.cpp
index ce21c5a018c..cb181a4e4ac 100644
--- a/src/Processors/DelayedPortsProcessor.cpp
+++ b/src/Processors/DelayedPortsProcessor.cpp
@@ -3,6 +3,11 @@
 namespace DB
 {
 
+namespace ErrorCodes
+{
+    extern const int LOGICAL_ERROR;
+}
+
 DelayedPortsProcessor::DelayedPortsProcessor(
     const Block & header, size_t num_ports, const PortNumbers & delayed_ports, bool assert_main_ports_empty)
     : IProcessor(InputPorts(num_ports, header),

From ca6ddfafc49f3b20908234da31047c8652a3570e Mon Sep 17 00:00:00 2001
From: Nikolai Kochetov <nik-kochetov@yandex-team.ru>
Date: Sun, 27 Dec 2020 16:55:55 +0300
Subject: [PATCH 485/504] Fix style

---
 src/Functions/ExtractString.h | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/src/Functions/ExtractString.h b/src/Functions/ExtractString.h
index 94ee01ae609..8b88a5a7c37 100644
--- a/src/Functions/ExtractString.h
+++ b/src/Functions/ExtractString.h
@@ -16,11 +16,6 @@
 namespace DB
 {
 
-namespace ErrorCodes
-{
-    extern const int LOGICAL_ERROR;
-}
-
 // used by FunctionsStringSimilarity and FunctionsStringHash
 // includes extracting ASCII ngram, UTF8 ngram, ASCII word and UTF8 word
 struct ExtractStringImpl

From 5337d4e2dc8f7b808b519e3f6bd3cb76bcf05d49 Mon Sep 17 00:00:00 2001
From: awesomeleo <awesomeleo@163.com>
Date: Sun, 27 Dec 2020 23:49:34 +0800
Subject: [PATCH 486/504] fix 18450, DB::Exception: Unsupported type
 FixedString

---
 src/Core/ExternalResultDescription.cpp | 3 +++
 src/Core/ExternalResultDescription.h   | 3 ++-
 src/Formats/MySQLBlockInputStream.cpp  | 4 ++++
 3 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/Core/ExternalResultDescription.cpp b/src/Core/ExternalResultDescription.cpp
index 7165d73b7d0..4be80a352c7 100644
--- a/src/Core/ExternalResultDescription.cpp
+++ b/src/Core/ExternalResultDescription.cpp
@@ -4,6 +4,7 @@
 #include <DataTypes/DataTypeDateTime64.h>
 #include <DataTypes/DataTypeNullable.h>
 #include <DataTypes/DataTypeString.h>
+#include <DataTypes/DataTypeFixedString.h>
 #include <DataTypes/DataTypeUUID.h>
 #include <DataTypes/DataTypesDecimal.h>
 #include <DataTypes/DataTypesNumber.h>
@@ -76,6 +77,8 @@ void ExternalResultDescription::init(const Block & sample_block_)
             types.emplace_back(ValueType::vtDecimal128, is_nullable);
         else if (typeid_cast<const DataTypeDecimal<Decimal256> *>(type))
             types.emplace_back(ValueType::vtDecimal256, is_nullable);
+        else if (typeid_cast<const DataTypeFixedString *>(type))
+            types.emplace_back(ValueType::vtFixedString, is_nullable);
         else
             throw Exception{"Unsupported type " + type->getName(), ErrorCodes::UNKNOWN_TYPE};
     }
diff --git a/src/Core/ExternalResultDescription.h b/src/Core/ExternalResultDescription.h
index f8ba2a6bba2..cc3b26ad841 100644
--- a/src/Core/ExternalResultDescription.h
+++ b/src/Core/ExternalResultDescription.h
@@ -30,7 +30,8 @@ struct ExternalResultDescription
         vtDecimal32,
         vtDecimal64,
         vtDecimal128,
-        vtDecimal256
+        vtDecimal256,
+        vtFixedString
     };
 
     Block sample_block;
diff --git a/src/Formats/MySQLBlockInputStream.cpp b/src/Formats/MySQLBlockInputStream.cpp
index 2ff8e8e5fb2..026f688a67f 100644
--- a/src/Formats/MySQLBlockInputStream.cpp
+++ b/src/Formats/MySQLBlockInputStream.cpp
@@ -8,6 +8,7 @@
 #    include <Columns/ColumnString.h>
 #    include <Columns/ColumnsNumber.h>
 #    include <Columns/ColumnDecimal.h>
+#    include <Columns/ColumnFixedString.h>
 #    include <DataTypes/IDataType.h>
 #    include <DataTypes/DataTypeNullable.h>
 #    include <IO/ReadHelpers.h>
@@ -110,6 +111,9 @@ namespace
                 data_type.deserializeAsWholeText(column, buffer, FormatSettings{});
                 break;
             }
+            case ValueType::vtFixedString:
+                assert_cast<ColumnFixedString &>(column).insertData(value.data(), value.size());
+                break;
         }
     }
 

From b086a78f814297806c6de3a0a23dd15f43a91e2f Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Sun, 27 Dec 2020 22:35:38 +0300
Subject: [PATCH 487/504] Update FunctionBinaryArithmetic.h

---
 src/Functions/FunctionBinaryArithmetic.h | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h
index 0e25f38e96d..848de344585 100644
--- a/src/Functions/FunctionBinaryArithmetic.h
+++ b/src/Functions/FunctionBinaryArithmetic.h
@@ -767,7 +767,7 @@ class FunctionBinaryArithmetic : public IFunction
 
     template <class LeftDataType, class RightDataType, class ResultDataType>
     ColumnPtr executeNumericWithDecimal(
-        const auto& left, const auto& right,
+        const auto & left, const auto & right,
         const ColumnConst * const col_left_const, const ColumnConst * const col_right_const,
         const auto * const col_left, const auto * const col_right,
         size_t col_left_size) const
@@ -789,7 +789,8 @@ class FunctionBinaryArithmetic : public IFunction
 
         typename ColVecResult::MutablePtr col_res = nullptr;
 
-        const ResultDataType type = [&] {
+        const ResultDataType type = [&]
+        {
             if constexpr (left_is_decimal && IsFloatingPoint<RightDataType>)
                 return RightDataType();
             else if constexpr (right_is_decimal && IsFloatingPoint<LeftDataType>)
@@ -798,7 +799,8 @@ class FunctionBinaryArithmetic : public IFunction
                 return decimalResultType<is_multiply, is_division>(left, right);
         }();
 
-        const ResultType scale_a = [&] {
+        const ResultType scale_a = [&]
+        {
             if constexpr (IsDataTypeDecimal<RightDataType> && is_division)
                 return right.getScaleMultiplier(); // the division impl uses only the scale_a
             else if constexpr (result_is_decimal)
@@ -825,7 +827,8 @@ class FunctionBinaryArithmetic : public IFunction
                 return 1; // the default value which won't cause any re-scale
         }();
 
-        const ResultType scale_b = [&] {
+        const ResultType scale_b = [&]
+        {
             if constexpr (result_is_decimal)
             {
                 if constexpr (is_multiply)
@@ -1000,9 +1003,9 @@ public:
 
                         const TimezoneMixin * tz = nullptr;
                         if constexpr (std::is_same_v<RightDataType, DataTypeDateTime>)
-                                tz = &right;
+                            tz = &right;
                         if constexpr (std::is_same_v<LeftDataType, DataTypeDateTime>)
-                                tz = &left;
+                            tz = &left;
                         type_res = std::make_shared<ResultDataType>(*tz);
                     }
                     else
@@ -1139,12 +1142,14 @@ public:
             const ColVecT1 * const col_right = checkAndGetColumn<ColVecT1>(col_right_raw);
 
             if constexpr (IsDataTypeDecimal<LeftDataType> || IsDataTypeDecimal<RightDataType>)
+            {
                 return executeNumericWithDecimal<LeftDataType, RightDataType, ResultDataType>(
                     left, right,
                     col_left_const, col_right_const,
                     col_left, col_right,
                     col_left_size);
-            else //can't avoid else and another indentation level, otherwise the compiler would try to instantiate
+            }
+            else // can't avoid else and another indentation level, otherwise the compiler would try to instantiate
                  // ColVecResult for Decimals which would lead to a compile error.
             {
                 using OpImpl = BinaryOperationImpl<T0, T1, Op<T0, T1>, ResultType>;

From 8c368688a847d7e87f6d34108872abe5d2e08edd Mon Sep 17 00:00:00 2001
From: Anna <annaconda@bk.ru>
Date: Sun, 27 Dec 2020 23:49:05 +0300
Subject: [PATCH 488/504] update

---
 .../en/sql-reference/functions/string-functions.md | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/docs/en/sql-reference/functions/string-functions.md b/docs/en/sql-reference/functions/string-functions.md
index 21e6f224b67..ddb27756df4 100644
--- a/docs/en/sql-reference/functions/string-functions.md
+++ b/docs/en/sql-reference/functions/string-functions.md
@@ -562,6 +562,8 @@ Result:
 
 Escape characters to place string into XML text node or attribute.
 
+The following five XML predefined entities will be replaced: `<`, `&`, `>`, `"`, `\'`.
+
 **Syntax** 
 
 ``` sql
@@ -570,11 +572,11 @@ encodeXMLComponent(x)
 
 **Parameters** 
 
--   `x` — Sequence of characters. [String](../../sql-reference/data-types/string.md).
+-   `x` — The sequence of characters. [String](../../sql-reference/data-types/string.md).
 
 **Returned value(s)**
 
--   Sequence of characters.
+-   The sequence of characters with escape characters.
 
 Type: [String](../../sql-reference/data-types/string.md).
 
@@ -583,11 +585,19 @@ Type: [String](../../sql-reference/data-types/string.md).
 Query:
 
 ``` sql
+SELECT encodeXMLComponent('Hello, "world"!');
+SELECT encodeXMLComponent('<123>');
+SELECT encodeXMLComponent('&clickhouse');
+SELECT encodeXMLComponent('\'foo\'');
 ```
 
 Result:
 
 ``` text
+Hello, &quot;world&quot;!
+&lt;123&gt;
+&amp;clickhouse
+&apos;foo&apos;
 ```
 
 [Original article](https://clickhouse.tech/docs/en/query_language/functions/string_functions/) <!--hide-->

From 2905f70cce59728bb91f415aeda7d6f4506c80b6 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Mon, 28 Dec 2020 12:56:38 +0300
Subject: [PATCH 489/504] fix aliases in partition by/order by

---
 programs/client/QueryFuzzer.cpp               |  4 +-
 src/Interpreters/ActionsVisitor.cpp           |  6 +--
 src/Interpreters/ExpressionAnalyzer.cpp       | 40 +++++++------------
 src/Interpreters/QueryNormalizer.cpp          | 40 +++++++++++--------
 src/Interpreters/QueryNormalizer.h            |  2 +-
 src/Parsers/ASTFunction.cpp                   |  9 +++--
 src/Parsers/ASTFunction.h                     | 22 ++++++++--
 src/Parsers/ExpressionElementParsers.cpp      |  9 +++--
 .../01591_window_functions.reference          | 16 +++++++-
 .../0_stateless/01591_window_functions.sql    |  6 ++-
 10 files changed, 93 insertions(+), 61 deletions(-)

diff --git a/programs/client/QueryFuzzer.cpp b/programs/client/QueryFuzzer.cpp
index 53ede4a3d92..fe0b6a975ce 100644
--- a/programs/client/QueryFuzzer.cpp
+++ b/programs/client/QueryFuzzer.cpp
@@ -405,8 +405,8 @@ void QueryFuzzer::fuzz(ASTPtr & ast)
 
         if (fn->is_window_function)
         {
-            fuzzColumnLikeExpressionList(fn->window_partition_by);
-            fuzzOrderByList(fn->window_order_by);
+            fuzzColumnLikeExpressionList(fn->window_partition_by.get());
+            fuzzOrderByList(fn->window_order_by.get());
         }
 
         fuzz(fn->children);
diff --git a/src/Interpreters/ActionsVisitor.cpp b/src/Interpreters/ActionsVisitor.cpp
index bb47129f58c..00600bebf07 100644
--- a/src/Interpreters/ActionsVisitor.cpp
+++ b/src/Interpreters/ActionsVisitor.cpp
@@ -738,15 +738,13 @@ void ActionsMatcher::visit(const ASTFunction & node, const ASTPtr & ast, Data &
     if (node.is_window_function)
     {
         // Also add columns from PARTITION BY and ORDER BY of window functions.
-        // Requiring a constant reference to a shared pointer to non-const AST
-        // doesn't really look sane, but the visitor does indeed require it.
         if (node.window_partition_by)
         {
-            visit(node.window_partition_by->clone(), data);
+            visit(node.window_partition_by, data);
         }
         if (node.window_order_by)
         {
-            visit(node.window_order_by->clone(), data);
+            visit(node.window_order_by, data);
         }
 
         // Also manually add columns for arguments of the window function itself.
diff --git a/src/Interpreters/ExpressionAnalyzer.cpp b/src/Interpreters/ExpressionAnalyzer.cpp
index c10a8313bee..a80b7799a98 100644
--- a/src/Interpreters/ExpressionAnalyzer.cpp
+++ b/src/Interpreters/ExpressionAnalyzer.cpp
@@ -983,36 +983,14 @@ void SelectQueryExpressionAnalyzer::appendWindowFunctionsArguments(
             getRootActionsNoMakeSet(f.function_node->clone(),
                 true /* no_subqueries */, step.actions());
 
-            // FIXME rewrite this comment
-            // 1.2) result of window function: an empty INPUT.
-            // It is an aggregate function, so it won't be added by getRootActions.
-            // This is something of a hack. Other options:
-            //  a] do it like aggregate function -- break the chain of actions
-            //     and manually add window functions to the starting list of
-            //     input columns. Logically this is similar to what we're doing
-            //     now, but would require to split the window function processing
-            //     into a full-fledged step after plain functions. This would be
-            //     somewhat cumbersome. With INPUT hack we can avoid a separate
-            //     step and pretend that window functions are almost "normal"
-            //     select functions. The limitation of both these ways is that
-            //     we can't reference window functions in other SELECT
-            //     expressions.
-            //  b] add a WINDOW action type, then sort, then split the chain on
-            //     each WINDOW action and insert the Window pipeline between the
-            //     Expression pipelines. This is a "proper" way that would allow
-            //     us to depend on window functions in other functions. But it's
-            //     complicated so I avoid doing it for now.
-            columns_after_window.push_back({f.column_name,
-                f.aggregate_function->getReturnType()});
-
+            // 2.1) function arguments;
             for (const auto & a : f.function_node->arguments->children)
             {
-                // 2.1) function arguments;
                 step.required_output.push_back(a->getColumnName());
             }
         }
 
-        // 2.3) PARTITION BY and ORDER BY columns.
+        // 2.1) PARTITION BY and ORDER BY columns.
         for (const auto & c : w.full_sort_description)
         {
             step.required_output.push_back(c.column_name);
@@ -1433,11 +1411,23 @@ ExpressionAnalysisResult::ExpressionAnalysisResult(
         {
             query_analyzer.appendWindowFunctionsArguments(chain, only_types || !first_stage);
 
-
+            // Build a list of output columns of the window step.
+            // 1) We need the columns that are the output of ExpressionActions.
             for (const auto & x : chain.getLastActions()->getNamesAndTypesList())
             {
                 query_analyzer.columns_after_window.push_back(x);
             }
+            // 2) We also have to manually add the output of the window function
+            // to the list of the output columns of the window step, because the
+            // window functions are not in the ExpressionActions.
+            for (const auto & [_, w] : query_analyzer.window_descriptions)
+            {
+                for (const auto & f : w.window_functions)
+                {
+                    query_analyzer.columns_after_window.push_back(
+                        {f.column_name, f.aggregate_function->getReturnType()});
+                }
+            }
 
             before_window = chain.getLastActions();
             finalize_chain(chain);
diff --git a/src/Interpreters/QueryNormalizer.cpp b/src/Interpreters/QueryNormalizer.cpp
index 213e354a13f..848681953e4 100644
--- a/src/Interpreters/QueryNormalizer.cpp
+++ b/src/Interpreters/QueryNormalizer.cpp
@@ -148,7 +148,7 @@ void QueryNormalizer::visit(ASTSelectQuery & select, const ASTPtr &, Data & data
 /// Don't go into select query. It processes children itself.
 /// Do not go to the left argument of lambda expressions, so as not to replace the formal parameters
 ///  on aliases in expressions of the form 123 AS x, arrayMap(x -> 1, [2]).
-void QueryNormalizer::visitChildren(ASTPtr & node, Data & data)
+void QueryNormalizer::visitChildren(IAST * node, Data & data)
 {
     if (auto * func_node = node->as<ASTFunction>())
     {
@@ -159,27 +159,33 @@ void QueryNormalizer::visitChildren(ASTPtr & node, Data & data)
             /// Don't go into query argument.
             return;
         }
+        /// We skip the first argument. We also assume that the lambda function can not have parameters.
+        size_t first_pos = 0;
+        if (func_node->name == "lambda")
+            first_pos = 1;
 
-        // Process all function node children (arguments + window definition for
-        // window functions).
-        // We have to skip the first argument if it's the "lambda" function,
-        // because it contains formal parameters like x->x + 1;
-        const IAST * argument_to_skip = nullptr;
-        if (func_node->name == "lambda"
-            && func_node->arguments
-            && !func_node->arguments->children.empty())
+        if (func_node->arguments)
         {
-            argument_to_skip = func_node->arguments->children[0].get();
-        }
+            auto & func_children = func_node->arguments->children;
 
-        for (auto & child : func_node->children)
-        {
-            if (child.get() != argument_to_skip
-                && needVisitChild(child))
+            for (size_t i = first_pos; i < func_children.size(); ++i)
             {
-                visit(child, data);
+                auto & child = func_children[i];
+
+                if (needVisitChild(child))
+                    visit(child, data);
             }
         }
+        
+        if (func_node->window_partition_by)
+        {
+            visitChildren(func_node->window_partition_by.get(), data);
+        }
+        
+        if (func_node->window_order_by)
+        {
+            visitChildren(func_node->window_order_by.get(), data);
+        }
     }
     else if (!node->as<ASTSelectQuery>())
     {
@@ -225,7 +231,7 @@ void QueryNormalizer::visit(ASTPtr & ast, Data & data)
     if (ast.get() != initial_ast.get())
         visit(ast, data);
     else
-        visitChildren(ast, data);
+        visitChildren(ast.get(), data);
 
     current_asts.erase(initial_ast.get());
     current_asts.erase(ast.get());
diff --git a/src/Interpreters/QueryNormalizer.h b/src/Interpreters/QueryNormalizer.h
index 951628d0738..e481f76ca8e 100644
--- a/src/Interpreters/QueryNormalizer.h
+++ b/src/Interpreters/QueryNormalizer.h
@@ -69,7 +69,7 @@ private:
     static void visit(ASTTablesInSelectQueryElement &, const ASTPtr &, Data &);
     static void visit(ASTSelectQuery &, const ASTPtr &, Data &);
 
-    static void visitChildren(ASTPtr & node, Data & data);
+    static void visitChildren(IAST * node, Data & data);
 };
 
 }
diff --git a/src/Parsers/ASTFunction.cpp b/src/Parsers/ASTFunction.cpp
index 980c395c7aa..d9159117b2d 100644
--- a/src/Parsers/ASTFunction.cpp
+++ b/src/Parsers/ASTFunction.cpp
@@ -67,17 +67,20 @@ ASTPtr ASTFunction::clone() const
 
     if (window_name)
     {
-        res->set(res->window_name, window_name->clone());
+        res->window_name = window_name->clone();
+        res->children.push_back(res->window_name);
     }
 
     if (window_partition_by)
     {
-        res->set(res->window_partition_by, window_partition_by->clone());
+        res->window_partition_by = window_partition_by->clone();
+        res->children.push_back(res->window_partition_by);
     }
 
     if (window_order_by)
     {
-        res->set(res->window_order_by, window_order_by->clone());
+        res->window_order_by = window_order_by->clone();
+        res->children.push_back(res->window_order_by);
     }
 
     return res;
diff --git a/src/Parsers/ASTFunction.h b/src/Parsers/ASTFunction.h
index 38e5f3f095c..4c20309fcb9 100644
--- a/src/Parsers/ASTFunction.h
+++ b/src/Parsers/ASTFunction.h
@@ -21,9 +21,25 @@ public:
     ASTPtr parameters;
 
     bool is_window_function = false;
-    ASTIdentifier * window_name;
-    ASTExpressionList * window_partition_by;
-    ASTExpressionList * window_order_by;
+
+    // We have to make these fields ASTPtr because this is what the visitors
+    // expect. Some of them take const ASTPtr & (makes no sense), and some
+    // take ASTPtr & and modify it. I don't understand how the latter is
+    // compatible with also having an owning `children` array -- apparently it
+    // leads to some dangling children that are not referenced by the fields of
+    // the AST class itself. Some older code hints at the idea of having
+    // ownership in `children` only, and making the class fields to be raw
+    // pointers of proper type (see e.g. IAST::set), but this is not compatible
+    // with the visitor interface.
+
+    // ASTIdentifier
+    ASTPtr window_name;
+
+    // ASTExpressionList
+    ASTPtr window_partition_by;
+
+    // ASTExpressionList of
+    ASTPtr window_order_by;
 
     /// do not print empty parentheses if there are no args - compatibility with new AST for data types and engine names.
     bool no_empty_args = false;
diff --git a/src/Parsers/ExpressionElementParsers.cpp b/src/Parsers/ExpressionElementParsers.cpp
index 726e28005e3..7c82c4aca1e 100644
--- a/src/Parsers/ExpressionElementParsers.cpp
+++ b/src/Parsers/ExpressionElementParsers.cpp
@@ -419,7 +419,8 @@ bool ParserWindowDefinition::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
         ParserIdentifier window_name_parser;
         if (window_name_parser.parse(pos, window_name_ast, expected))
         {
-            function->set(function->window_name, window_name_ast);
+            function->children.push_back(window_name_ast);
+            function->window_name = window_name_ast;
             return true;
         }
         else
@@ -442,7 +443,8 @@ bool ParserWindowDefinition::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
         ASTPtr partition_by_ast;
         if (columns_partition_by.parse(pos, partition_by_ast, expected))
         {
-            function->set(function->window_partition_by, partition_by_ast);
+            function->children.push_back(partition_by_ast);
+            function->window_partition_by = partition_by_ast;
         }
         else
         {
@@ -455,7 +457,8 @@ bool ParserWindowDefinition::parseImpl(Pos & pos, ASTPtr & node, Expected & expe
         ASTPtr order_by_ast;
         if (columns_order_by.parse(pos, order_by_ast, expected))
         {
-            function->set(function->window_order_by, order_by_ast);
+            function->children.push_back(order_by_ast);
+            function->window_order_by = order_by_ast;
         }
         else
         {
diff --git a/tests/queries/0_stateless/01591_window_functions.reference b/tests/queries/0_stateless/01591_window_functions.reference
index 6eb94856645..ce56860ed8b 100644
--- a/tests/queries/0_stateless/01591_window_functions.reference
+++ b/tests/queries/0_stateless/01591_window_functions.reference
@@ -122,14 +122,14 @@ select * from (select * from numbers(5) order by rand()) order by count() over (
 2
 3
 4
-select * from (select * from numbers(5) order by rand()) group by number order by sum(any(number)) over (order by min(number) desc) desc;
+select * from (select * from numbers(5) order by rand()) group by number order by sum(any(number + 1)) over (order by min(number) desc) desc;
 
 -- different windows
 -- an explain test would also be helpful, but it's too immature now and I don't
 -- want to change reference all the time
 
-1
 0
+1
 2
 3
 4
@@ -204,4 +204,16 @@ select distinct sum(0) over () from numbers(2);
 0
 select distinct any(number) over () from numbers(2);
 
+-- Various kinds of aliases are properly substituted into various parts of window
+-- function definition.
+
 0
+with number + 1 as x select intDiv(number, 3) as y, sum(x + y) over (partition by y order by x) from numbers(7);
+
+0	1
+0	3
+0	6
+1	5
+1	11
+1	18
+2	9
diff --git a/tests/queries/0_stateless/01591_window_functions.sql b/tests/queries/0_stateless/01591_window_functions.sql
index 5039925aa42..082a6652a65 100644
--- a/tests/queries/0_stateless/01591_window_functions.sql
+++ b/tests/queries/0_stateless/01591_window_functions.sql
@@ -41,7 +41,7 @@ select * from (select * from numbers(5) order by rand()) order by count() over (
 -- Aggregate functions as window function arguments. This query is semantically
 -- the same as the above one, only we replace `number` with
 -- `any(number) group by number` and so on.
-select * from (select * from numbers(5) order by rand()) group by number order by sum(any(number)) over (order by min(number) desc) desc;
+select * from (select * from numbers(5) order by rand()) group by number order by sum(any(number + 1)) over (order by min(number) desc) desc;
 
 -- different windows
 -- an explain test would also be helpful, but it's too immature now and I don't
@@ -66,3 +66,7 @@ select count(1) over (), max(number + 1) over () from numbers(3);
 -- Should work in DISTINCT
 select distinct sum(0) over () from numbers(2);
 select distinct any(number) over () from numbers(2);
+
+-- Various kinds of aliases are properly substituted into various parts of window
+-- function definition.
+with number + 1 as x select intDiv(number, 3) as y, sum(x + y) over (partition by y order by x) from numbers(7);

From 4adbdf364e64a20001925528dd1f4aeddbec577a Mon Sep 17 00:00:00 2001
From: hexiaoting <hewenting_ict@163.com>
Date: Thu, 17 Dec 2020 17:33:47 +0800
Subject: [PATCH 490/504] Fix bug for orderby withfill with limit clause

---
 src/Interpreters/InterpreterSelectQuery.cpp       | 15 ++++++++++++++-
 .../01614_with_fill_with_limit.reference          |  4 ++++
 .../0_stateless/01614_with_fill_with_limit.sql    | 15 +++++++++++++++
 3 files changed, 33 insertions(+), 1 deletion(-)
 create mode 100644 tests/queries/0_stateless/01614_with_fill_with_limit.reference
 create mode 100644 tests/queries/0_stateless/01614_with_fill_with_limit.sql

diff --git a/src/Interpreters/InterpreterSelectQuery.cpp b/src/Interpreters/InterpreterSelectQuery.cpp
index 38cc19a00d6..5e895bf732b 100644
--- a/src/Interpreters/InterpreterSelectQuery.cpp
+++ b/src/Interpreters/InterpreterSelectQuery.cpp
@@ -1029,10 +1029,23 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInpu
             /** Optimization - if there are several sources and there is LIMIT, then first apply the preliminary LIMIT,
               * limiting the number of rows in each up to `offset + limit`.
               */
+            bool has_withfill = false;
+            if (query.orderBy())
+            {
+                SortDescription order_descr = getSortDescription(query, *context);
+                for (auto & desc : order_descr)
+                    if (desc.with_fill)
+                    {
+                        has_withfill = true;
+                        break;
+                    }
+            }
+
             bool has_prelimit = false;
             if (!to_aggregation_stage &&
                 query.limitLength() && !query.limit_with_ties && !hasWithTotalsInAnySubqueryInFromClause(query) &&
-                !query.arrayJoinExpressionList() && !query.distinct && !expressions.hasLimitBy() && !settings.extremes)
+                !query.arrayJoinExpressionList() && !query.distinct && !expressions.hasLimitBy() && !settings.extremes &&
+                !has_withfill)
             {
                 executePreLimit(query_plan, false);
                 has_prelimit = true;
diff --git a/tests/queries/0_stateless/01614_with_fill_with_limit.reference b/tests/queries/0_stateless/01614_with_fill_with_limit.reference
new file mode 100644
index 00000000000..451e076552e
--- /dev/null
+++ b/tests/queries/0_stateless/01614_with_fill_with_limit.reference
@@ -0,0 +1,4 @@
+1	original
+2	
+1	original
+2	
diff --git a/tests/queries/0_stateless/01614_with_fill_with_limit.sql b/tests/queries/0_stateless/01614_with_fill_with_limit.sql
new file mode 100644
index 00000000000..119117af287
--- /dev/null
+++ b/tests/queries/0_stateless/01614_with_fill_with_limit.sql
@@ -0,0 +1,15 @@
+SELECT
+    toFloat32(number % 10) AS n,
+    'original' AS source
+FROM numbers(10)
+WHERE (number % 3) = 1
+ORDER BY n ASC WITH FILL STEP 1
+LIMIT 2;
+
+SELECT
+    toFloat32(number % 10) AS n,
+    'original' AS source
+FROM numbers(10)
+WHERE (number % 3) = 1
+ORDER BY n ASC WITH FILL STEP 1
+LIMIT 2 WITH TIES;

From 1c52fdb265ffbbd8387c10fd26e444c75451f634 Mon Sep 17 00:00:00 2001
From: Alexander Kuzmenkov <akuzm@yandex-team.ru>
Date: Mon, 28 Dec 2020 13:08:38 +0300
Subject: [PATCH 491/504] cleanup

---
 src/Interpreters/QueryNormalizer.cpp   |  4 ++--
 tests/performance/window_functions.xml | 20 +++++++++-----------
 2 files changed, 11 insertions(+), 13 deletions(-)

diff --git a/src/Interpreters/QueryNormalizer.cpp b/src/Interpreters/QueryNormalizer.cpp
index 848681953e4..38b54de2130 100644
--- a/src/Interpreters/QueryNormalizer.cpp
+++ b/src/Interpreters/QueryNormalizer.cpp
@@ -176,12 +176,12 @@ void QueryNormalizer::visitChildren(IAST * node, Data & data)
                     visit(child, data);
             }
         }
-        
+
         if (func_node->window_partition_by)
         {
             visitChildren(func_node->window_partition_by.get(), data);
         }
-        
+
         if (func_node->window_order_by)
         {
             visitChildren(func_node->window_order_by.get(), data);
diff --git a/tests/performance/window_functions.xml b/tests/performance/window_functions.xml
index dcc873b5029..f42345d0696 100644
--- a/tests/performance/window_functions.xml
+++ b/tests/performance/window_functions.xml
@@ -12,11 +12,11 @@
         First with LIMIT BY, next with window functions.
     -->
     <query><![CDATA[
-        select CounterID, UserID, count(*) c
+        select CounterID, UserID, count(*) user_hits
         from hits_100m_single
         where CounterID < 10000
         group by CounterID, UserID
-        order by c desc
+        order by user_hits desc
         limit 10 by CounterID
         format Null
     ]]></query>
@@ -24,16 +24,14 @@
     <query><![CDATA[
         select *
         from (
-            select *,
-                count() over (partition by CounterID order by c desc) rank
-                from (
-                    select CounterID, UserID, count() c
-                    from hits_100m_single
-                    where CounterID < 10000
-                    group by CounterID, UserID
-                )
+            select CounterID, UserID, count(*) user_hits,
+                count() over (partition by CounterID order by user_hits desc)
+                    user_rank
+            from hits_100m_single
+            where CounterID < 10000
+            group by CounterID, UserID
         )
-        where rank <= 10
+        where user_rank <= 10
         format Null
     ]]></query>
 

From 3d1273460bc9fdf7b92c5bbb12f0caa30fa830d5 Mon Sep 17 00:00:00 2001
From: robot-clickhouse <robot-clickhouse@yandex-team.ru>
Date: Mon, 28 Dec 2020 13:27:43 +0300
Subject: [PATCH 492/504] Update version_date.tsv after release 20.12.5.14

---
 utils/list-versions/version_date.tsv | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/list-versions/version_date.tsv b/utils/list-versions/version_date.tsv
index 002dc5965d3..1279da09905 100644
--- a/utils/list-versions/version_date.tsv
+++ b/utils/list-versions/version_date.tsv
@@ -1,3 +1,4 @@
+v20.12.5.14-stable	2020-12-28
 v20.12.4.5-stable	2020-12-24
 v20.12.3.3-stable	2020-12-09
 v20.12.2.1-stable	2020-12-09

From 67798c8811b63cd78e728499f17ea2e7fdb6d3b3 Mon Sep 17 00:00:00 2001
From: Roman Bug <rrrrrroman@gmail.com>
Date: Mon, 28 Dec 2020 13:39:51 +0300
Subject: [PATCH 493/504] Update docs/ru/commercial/cloud.md

Co-authored-by: Anna <42538400+adevyatova@users.noreply.github.com>
---
 docs/ru/commercial/cloud.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/commercial/cloud.md b/docs/ru/commercial/cloud.md
index 885aac97312..024297da7d8 100644
--- a/docs/ru/commercial/cloud.md
+++ b/docs/ru/commercial/cloud.md
@@ -22,7 +22,7 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 
 [Altinity.Cloud](https://altinity.com/cloud-database/) — это полностью управляемый ClickHouse-as-a-Service для публичного облака Amazon.
 
--   Быстрое развертывание кластеров ClickHouse на ресурсах Amazon
+-   быстрое развертывание кластеров ClickHouse на ресурсах Amazon.
 -   Легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов
 -   Изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC
 -   Настраиваемые типы и объемы хранилищ

From 5943a251b1f59cdd9f2d52209478f7f852d07254 Mon Sep 17 00:00:00 2001
From: Roman Bug <rrrrrroman@gmail.com>
Date: Mon, 28 Dec 2020 13:39:57 +0300
Subject: [PATCH 494/504] Update docs/ru/commercial/cloud.md

Co-authored-by: Anna <42538400+adevyatova@users.noreply.github.com>
---
 docs/ru/commercial/cloud.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/commercial/cloud.md b/docs/ru/commercial/cloud.md
index 024297da7d8..8fca835450c 100644
--- a/docs/ru/commercial/cloud.md
+++ b/docs/ru/commercial/cloud.md
@@ -23,7 +23,7 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 [Altinity.Cloud](https://altinity.com/cloud-database/) — это полностью управляемый ClickHouse-as-a-Service для публичного облака Amazon.
 
 -   быстрое развертывание кластеров ClickHouse на ресурсах Amazon.
--   Легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов
+-   легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов.
 -   Изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC
 -   Настраиваемые типы и объемы хранилищ
 -   Cross-AZ масштабирование для повышения производительности и обеспечения высокой доступности

From 281c767f30b029d5af19de80207cd8aec1d48a65 Mon Sep 17 00:00:00 2001
From: Roman Bug <rrrrrroman@gmail.com>
Date: Mon, 28 Dec 2020 13:40:05 +0300
Subject: [PATCH 495/504] Update docs/ru/commercial/cloud.md

Co-authored-by: Anna <42538400+adevyatova@users.noreply.github.com>
---
 docs/ru/commercial/cloud.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/commercial/cloud.md b/docs/ru/commercial/cloud.md
index 8fca835450c..b132aa58067 100644
--- a/docs/ru/commercial/cloud.md
+++ b/docs/ru/commercial/cloud.md
@@ -24,7 +24,7 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 
 -   быстрое развертывание кластеров ClickHouse на ресурсах Amazon.
 -   легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов.
--   Изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC
+-   изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC.
 -   Настраиваемые типы и объемы хранилищ
 -   Cross-AZ масштабирование для повышения производительности и обеспечения высокой доступности
 -   Встроенный мониторинг и редактор SQL-запросов

From 2c146fe2cd255469426bc96db2aeaf9cde881597 Mon Sep 17 00:00:00 2001
From: romanzhukov <romanzhukov@yandex-team.ru>
Date: Mon, 28 Dec 2020 14:24:24 +0300
Subject: [PATCH 496/504] DOCSUP-5054: Fix PR comments and update list items.

---
 docs/en/commercial/cloud.md                      |  1 +
 docs/ru/commercial/cloud.md                      | 16 ++++++++--------
 .../server-configuration-parameters/settings.md  |  2 +-
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/docs/en/commercial/cloud.md b/docs/en/commercial/cloud.md
index 03c8b7645a3..9dd08124f64 100644
--- a/docs/en/commercial/cloud.md
+++ b/docs/en/commercial/cloud.md
@@ -21,6 +21,7 @@ toc_title: Cloud
 ## Altinity.Cloud {#altinity.cloud}
 
 [Altinity.Cloud](https://altinity.com/cloud-database/) is a fully managed ClickHouse-as-a-Service for the Amazon public cloud.
+
 -   Fast deployment of ClickHouse clusters on Amazon resources
 -   Easy scale-out/scale-in as well as vertical scaling of nodes
 -   Isolated per-tenant VPCs with public endpoint or VPC peering
diff --git a/docs/ru/commercial/cloud.md b/docs/ru/commercial/cloud.md
index b132aa58067..4f57592b4c7 100644
--- a/docs/ru/commercial/cloud.md
+++ b/docs/ru/commercial/cloud.md
@@ -12,11 +12,11 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 
 [Yandex Managed Service for ClickHouse](https://cloud.yandex.ru/services/managed-clickhouse?utm_source=referrals&utm_medium=clickhouseofficialsite&utm_campaign=link3) предоставляет следующие ключевые возможности:
 
--   Полностью управляемый сервис ZooKeeper для [репликации ClickHouse](../engines/table-engines/mergetree-family/replication.md)
--   Выбор типа хранилища
--   Реплики в разных зонах доступности
--   Шифрование и изоляция
--   Автоматизированное техническое обслуживание
+-   полностью управляемый сервис ZooKeeper для [репликации ClickHouse](../engines/table-engines/mergetree-family/replication.md)
+-   выбор типа хранилища
+-   реплики в разных зонах доступности
+-   шифрование и изоляция
+-   автоматизированное техническое обслуживание
 
 ## Altinity.Cloud {#altinity.cloud}
 
@@ -25,8 +25,8 @@ toc_title: "\u041f\u043e\u0441\u0442\u0430\u0432\u0449\u0438\u043a\u0438\u0020\u
 -   быстрое развертывание кластеров ClickHouse на ресурсах Amazon.
 -   легкое горизонтальное масштабирование также, как и вертикальное масштабирование узлов.
 -   изолированные виртуальные сети для каждого клиента с общедоступным эндпоинтом или пирингом VPC.
--   Настраиваемые типы и объемы хранилищ
--   Cross-AZ масштабирование для повышения производительности и обеспечения высокой доступности
--   Встроенный мониторинг и редактор SQL-запросов
+-   настраиваемые типы и объемы хранилищ
+-   cross-az масштабирование для повышения производительности и обеспечения высокой доступности
+-   встроенный мониторинг и редактор SQL-запросов
 
 {## [Оригинальная статья](https://clickhouse.tech/docs/ru/commercial/cloud/) ##}
diff --git a/docs/ru/operations/server-configuration-parameters/settings.md b/docs/ru/operations/server-configuration-parameters/settings.md
index b75f0371223..d1911437baf 100644
--- a/docs/ru/operations/server-configuration-parameters/settings.md
+++ b/docs/ru/operations/server-configuration-parameters/settings.md
@@ -86,7 +86,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 
 Возможные значения:
 
--   Положительное целое число.
+-   положительное целое число.
 
 Значение по умолчанию: `1073741824` (1 ГБ).
 

From 50afe274614a86c400c2f50ea7f9ff5284b480ec Mon Sep 17 00:00:00 2001
From: alesapin <alesapin@gmail.com>
Date: Mon, 28 Dec 2020 14:46:22 +0300
Subject: [PATCH 497/504] More stable test

---
 .../01593_concurrent_alter_mutations_kill.sh   | 18 +++++++++++++++++-
 ...rrent_alter_mutations_kill_many_replicas.sh | 17 ++++++++++++++++-
 2 files changed, 33 insertions(+), 2 deletions(-)

diff --git a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh
index fe243a7e929..a521d90b768 100755
--- a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh
+++ b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill.sh
@@ -45,7 +45,23 @@ timeout $TIMEOUT bash -c kill_mutation_thread 2> /dev/null &
 wait
 
 $CLICKHOUSE_CLIENT --query "SYSTEM SYNC REPLICA concurrent_mutate_kill"
-$CLICKHOUSE_CLIENT --query "ALTER TABLE concurrent_mutate_kill MODIFY COLUMN value Int64 SETTINGS replication_alter_partitions_sync=2"
+
+# with timeout alter query can be not finished yet, so to execute new alter
+# we use retries
+counter=0
+while true; do
+    if $CLICKHOUSE_CLIENT --query "ALTER TABLE concurrent_mutate_kill MODIFY COLUMN value Int64 SETTINGS replication_alter_partitions_sync=2" 2> /dev/null ; then
+        break
+    fi
+
+    if [ "$counter" -gt 120 ]
+    then
+        break
+    fi
+    sleep 0.5
+    counter=$(($counter + 1))
+done
+
 $CLICKHOUSE_CLIENT --query "SHOW CREATE TABLE concurrent_mutate_kill"
 $CLICKHOUSE_CLIENT --query "OPTIMIZE TABLE concurrent_mutate_kill FINAL"
 $CLICKHOUSE_CLIENT --query "SELECT sum(value) FROM concurrent_mutate_kill"
diff --git a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh
index e9513557722..f5f751d6408 100755
--- a/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh
+++ b/tests/queries/0_stateless/01593_concurrent_alter_mutations_kill_many_replicas.sh
@@ -59,7 +59,22 @@ for i in $(seq $REPLICAS); do
     $CLICKHOUSE_CLIENT --query "SYSTEM SYNC REPLICA concurrent_kill_$i"
 done
 
-$CLICKHOUSE_CLIENT --query "ALTER TABLE concurrent_kill_$i MODIFY COLUMN value Int64 SETTINGS replication_alter_partitions_sync=2"
+# with timeout alter query can be not finished yet, so to execute new alter
+# we use retries
+counter=0
+while true; do
+    if $CLICKHOUSE_CLIENT --query "ALTER TABLE concurrent_kill_1 MODIFY COLUMN value Int64 SETTINGS replication_alter_partitions_sync=2" 2> /dev/null ; then
+        break
+    fi
+
+    if [ "$counter" -gt 120 ]
+    then
+        break
+    fi
+    sleep 0.5
+    counter=$(($counter + 1))
+done
+
 
 metadata_version=$($CLICKHOUSE_CLIENT --query "SELECT value FROM system.zookeeper WHERE path = '/clickhouse/tables/test_01593_concurrent_kill/replicas/$i/' and name = 'metadata_version'")
 for i in $(seq $REPLICAS); do

From f337be7ca2f02ed14959367e02f1e20ef98c581d Mon Sep 17 00:00:00 2001
From: Alexey Milovidov <milovidov@yandex-team.ru>
Date: Mon, 28 Dec 2020 16:26:22 +0300
Subject: [PATCH 498/504] Remove unneeded example

---
 src/Interpreters/tests/CMakeLists.txt |  3 -
 src/Interpreters/tests/context.cpp    | 90 ---------------------------
 2 files changed, 93 deletions(-)
 delete mode 100644 src/Interpreters/tests/context.cpp

diff --git a/src/Interpreters/tests/CMakeLists.txt b/src/Interpreters/tests/CMakeLists.txt
index 2c8440299fb..1bc9d7fbacb 100644
--- a/src/Interpreters/tests/CMakeLists.txt
+++ b/src/Interpreters/tests/CMakeLists.txt
@@ -32,9 +32,6 @@ target_link_libraries (string_hash_map_aggregation PRIVATE dbms)
 add_executable (string_hash_set string_hash_set.cpp)
 target_link_libraries (string_hash_set PRIVATE dbms)
 
-add_executable (context context.cpp)
-target_link_libraries (context PRIVATE dbms)
-
 add_executable (two_level_hash_map two_level_hash_map.cpp)
 target_include_directories (two_level_hash_map SYSTEM BEFORE PRIVATE ${SPARSEHASH_INCLUDE_DIR})
 target_link_libraries (two_level_hash_map PRIVATE dbms)
diff --git a/src/Interpreters/tests/context.cpp b/src/Interpreters/tests/context.cpp
deleted file mode 100644
index 9b908e26248..00000000000
--- a/src/Interpreters/tests/context.cpp
+++ /dev/null
@@ -1,90 +0,0 @@
-#include <iostream>
-/// #define BOOST_USE_UCONTEXT
-#include <Common/Fiber.h>
-// #include <boost/context/pooled_fixedsize_stack.hpp>
-// #include <boost/context/segmented_stack.hpp>
-#include <Common/Exception.h>
-#include <Common/FiberStack.h>
-
-void __attribute__((__noinline__)) foo(std::exception_ptr exception)
-{
-    if (exception)
-        std::rethrow_exception(exception);
-}
-
-void __attribute__((__noinline__)) bar(int a)
-{
-    std::cout << StackTrace().toString() << std::endl;
-
-    if (a > 0)
-        throw DB::Exception(0, "hello");
-}
-
-void __attribute__((__noinline__)) gar(int a)
-{
-    char buf[1024];
-    buf[1023] = a & 255;
-    if (a > 2)
-        return gar(a - 1);
-    else
-        bar(a);
-}
-
-int main(int, char **)
-try {
-    namespace ctx=boost::context;
-    int a;
-    std::exception_ptr exception;
-    // ctx::protected_fixedsize allocator
-    // ctx::pooled_fixedsize_stack(1024 * 64 + 2 * 2 * 1024 * 1024 * 16, 1)
-    ctx::fiber source{std::allocator_arg_t(), FiberStack(), [&](ctx::fiber&& sink)
-    {
-        a=0;
-        int b=1;
-        for (size_t i = 0; i < 9; ++i)
-        {
-            sink=std::move(sink).resume();
-            int next=a+b;
-            a=b;
-            b=next;
-        }
-        try
-        {
-            gar(1024);
-        }
-        catch (...)
-        {
-            std::cout << "Saving exception\n";
-            exception = std::current_exception();
-        }
-        return std::move(sink);
-    }};
-
-    for (int j=0;j<10;++j)
-    {
-        try
-        {
-            source=std::move(source).resume();
-        }
-        catch (DB::Exception & e)
-        {
-            std::cout << "Caught exception in resume " << e.getStackTraceString() << std::endl;
-        }
-        std::cout << a << " ";
-    }
-
-    std::cout << std::endl;
-
-    try
-    {
-        foo(exception);
-    }
-    catch (const DB::Exception & e)
-    {
-        std::cout << e.getStackTraceString() << std::endl;
-    }
-}
-catch (...)
-{
-    std::cerr << "Uncaught exception\n";
-}

From 819bd8059b9dcc3804b69f3c5a9c35953b5c55b8 Mon Sep 17 00:00:00 2001
From: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
Date: Mon, 28 Dec 2020 16:57:47 +0300
Subject: [PATCH 499/504] Update
 docs/ru/operations/server-configuration-parameters/settings.md

---
 docs/ru/operations/server-configuration-parameters/settings.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/operations/server-configuration-parameters/settings.md b/docs/ru/operations/server-configuration-parameters/settings.md
index d1911437baf..902dab0310d 100644
--- a/docs/ru/operations/server-configuration-parameters/settings.md
+++ b/docs/ru/operations/server-configuration-parameters/settings.md
@@ -82,7 +82,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 
 ## core_dump {#server_configuration_parameters-core_dump}
 
-Задает мягкое ограничение для размера файла дампа ядра.
+Задает мягкое ограничение для размера файла дампа памяти.
 
 Возможные значения:
 

From 35d62b73a7f766b7e9447e5e9a32e32baf2cf907 Mon Sep 17 00:00:00 2001
From: Ivan <5627721+abyss7@users.noreply.github.com>
Date: Mon, 28 Dec 2020 18:47:44 +0300
Subject: [PATCH 500/504] Check for CLICKHOUSE_CLIENT_OPT env before setting it
 (#18574)

---
 tests/clickhouse-test | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/tests/clickhouse-test b/tests/clickhouse-test
index 3e4429449a7..da7d055f250 100755
--- a/tests/clickhouse-test
+++ b/tests/clickhouse-test
@@ -873,6 +873,7 @@ if __name__ == '__main__':
     parser.add_argument('--use-skip-list', action='store_true', default=False, help="Use skip list to skip tests if found")
     parser.add_argument('--db-engine', help='Database engine name')
 
+    parser.add_argument('--antlr', action='store_true', default=False, dest='antlr', help='Use new ANTLR parser in tests')
     parser.add_argument('--no-stateless', action='store_true', help='Disable all stateless tests')
     parser.add_argument('--no-stateful', action='store_true', help='Disable all stateful tests')
     parser.add_argument('--skip', nargs='+', help="Skip these tests")
@@ -886,7 +887,6 @@ if __name__ == '__main__':
     group=parser.add_mutually_exclusive_group(required=False)
     group.add_argument('--shard', action='store_true', default=None, dest='shard', help='Run sharding related tests (required to clickhouse-server listen 127.0.0.2 127.0.0.3)')
     group.add_argument('--no-shard', action='store_false', default=None, dest='shard', help='Do not run shard related tests')
-    parser.add_argument('--antlr', action='store_true', default=False, dest='antlr', help='Use new ANTLR parser in tests')
 
     args = parser.parse_args()
 
@@ -967,7 +967,10 @@ if __name__ == '__main__':
         os.environ['CLICKHOUSE_URL_PARAMS'] += get_additional_client_options_url(args)
 
     if args.antlr:
-        os.environ['CLICKHOUSE_CLIENT_OPT'] += ' --use_antlr_parser=1'
+        if 'CLICKHOUSE_CLIENT_OPT' in os.environ:
+           os.environ['CLICKHOUSE_CLIENT_OPT'] += ' --use_antlr_parser=1'
+        else:
+           os.environ['CLICKHOUSE_CLIENT_OPT'] = '--use_antlr_parser=1'
 
     if args.extract_from_config is None:
         if os.access(args.binary + '-extract-from-config', os.X_OK):

From ecc03394724bf4c2bf2cab8f34fb68cacf3d392d Mon Sep 17 00:00:00 2001
From: Alexander Tokmakov <avtokmakov@yandex-team.ru>
Date: Mon, 28 Dec 2020 23:21:57 +0300
Subject: [PATCH 501/504] add test

---
 .../materialize_with_ddl.py                   | 29 +++++++++----------
 tests/integration/test_storage_mysql/test.py  |  7 +++++
 2 files changed, 21 insertions(+), 15 deletions(-)

diff --git a/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py b/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
index 47e325d0e96..c04194c8ebb 100644
--- a/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
+++ b/tests/integration/test_materialize_mysql_database/materialize_with_ddl.py
@@ -48,16 +48,15 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
                      "/* Need ClickHouse support read mysql decimal unsigned_decimal DECIMAL(19, 10) UNSIGNED, _decimal DECIMAL(19, 10), */"
                      "unsigned_float FLOAT UNSIGNED, _float FLOAT, "
                      "unsigned_double DOUBLE UNSIGNED, _double DOUBLE, "
-                     "_varchar VARCHAR(10), _char CHAR(10), "
+                     "_varchar VARCHAR(10), _char CHAR(10), binary_col BINARY(8), "
                      "/* Need ClickHouse support Enum('a', 'b', 'v') _enum ENUM('a', 'b', 'c'), */"
                      "_date Date, _datetime DateTime, _timestamp TIMESTAMP, _bool BOOLEAN) ENGINE = InnoDB;")
 
     # it already has some data
     mysql_node.query("""
-        INSERT INTO test_database.test_table_1 VALUES(1, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 
+        INSERT INTO test_database.test_table_1 VALUES(1, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 'binary',
         '2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', true);
         """)
-
     clickhouse_node.query(
         "CREATE DATABASE test_database ENGINE = MaterializeMySQL('{}:3306', 'test_database', 'root', 'clickhouse')".format(
             service_name))
@@ -65,51 +64,51 @@ def dml_with_materialize_mysql_database(clickhouse_node, mysql_node, service_nam
     assert "test_database" in clickhouse_node.query("SHOW DATABASES")
 
     check_query(clickhouse_node, "SELECT * FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
-                "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
+                "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\tbinary\\0\\0\t2020-01-01\t"
                 "2020-01-01 00:00:00\t2020-01-01 00:00:00\t1\n")
 
     mysql_node.query("""
-        INSERT INTO test_database.test_table_1 VALUES(2, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 
+        INSERT INTO test_database.test_table_1 VALUES(2, 1, -1, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 3.2, -3.2, 3.4, -3.4, 'varchar', 'char', 'binary',
         '2020-01-01', '2020-01-01 00:00:00', '2020-01-01 00:00:00', false);
         """)
 
     check_query(clickhouse_node, "SELECT * FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
-                "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
+                "1\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\tbinary\\0\\0\t2020-01-01\t"
                 "2020-01-01 00:00:00\t2020-01-01 00:00:00\t1\n2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
-                "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t2020-01-01 00:00:00\t0\n")
+                "varchar\tchar\tbinary\\0\\0\t2020-01-01\t2020-01-01 00:00:00\t2020-01-01 00:00:00\t0\n")
 
     mysql_node.query("UPDATE test_database.test_table_1 SET unsigned_tiny_int = 2 WHERE `key` = 1")
 
     check_query(clickhouse_node, """
         SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,
          small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, 
-         unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, 
+         unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, binary_col,
          _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ 
          _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV
         """,
-        "1\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
+        "1\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\tbinary\\0\\0\t2020-01-01\t"
         "2020-01-01 00:00:00\t1\n2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
-        "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t0\n")
+        "varchar\tchar\tbinary\\0\\0\t2020-01-01\t2020-01-01 00:00:00\t0\n")
 
     # update primary key
     mysql_node.query("UPDATE test_database.test_table_1 SET `key` = 3 WHERE `unsigned_tiny_int` = 2")
 
     check_query(clickhouse_node, "SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,"
                                  " small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, "
-                                 " unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, "
+                                 " unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, binary_col, "
                                  " _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ "
                                  " _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
                 "2\t1\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\t"
-                "varchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t0\n3\t2\t-1\t2\t-2\t3\t-3\t"
-                "4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t2020-01-01 00:00:00\t1\n")
+                "varchar\tchar\tbinary\\0\\0\t2020-01-01\t2020-01-01 00:00:00\t0\n3\t2\t-1\t2\t-2\t3\t-3\t"
+                "4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\tbinary\\0\\0\t2020-01-01\t2020-01-01 00:00:00\t1\n")
 
     mysql_node.query('DELETE FROM test_database.test_table_1 WHERE `key` = 2')
     check_query(clickhouse_node, "SELECT key, unsigned_tiny_int, tiny_int, unsigned_small_int,"
                                  " small_int, unsigned_medium_int, medium_int, unsigned_int, _int, unsigned_integer, _integer, "
-                                 " unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, "
+                                 " unsigned_bigint, _bigint, unsigned_float, _float, unsigned_double, _double, _varchar, _char, binary_col, "
                                  " _date, _datetime, /* exclude it, because ON UPDATE CURRENT_TIMESTAMP _timestamp, */ "
                                  " _bool FROM test_database.test_table_1 ORDER BY key FORMAT TSV",
-                "3\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\t2020-01-01\t"
+                "3\t2\t-1\t2\t-2\t3\t-3\t4\t-4\t5\t-5\t6\t-6\t3.2\t-3.2\t3.4\t-3.4\tvarchar\tchar\tbinary\\0\\0\t2020-01-01\t"
                 "2020-01-01 00:00:00\t1\n")
 
     mysql_node.query('DELETE FROM test_database.test_table_1 WHERE `unsigned_tiny_int` = 2')
diff --git a/tests/integration/test_storage_mysql/test.py b/tests/integration/test_storage_mysql/test.py
index 87033381e2c..7b23e20e200 100644
--- a/tests/integration/test_storage_mysql/test.py
+++ b/tests/integration/test_storage_mysql/test.py
@@ -148,6 +148,13 @@ def test_table_function(started_cluster):
     assert node1.query("SELECT sum(`money`) FROM {}".format(table_function)).rstrip() == '60000'
     conn.close()
 
+def test_binary_type(started_cluster):
+    conn = get_mysql_conn()
+    with conn.cursor() as cursor:
+        cursor.execute("CREATE TABLE clickhouse.binary_type (id INT PRIMARY KEY, data BINARY(16) NOT NULL)")
+    table_function = "mysql('mysql1:3306', 'clickhouse', '{}', 'root', 'clickhouse')".format('binary_type')
+    node1.query("INSERT INTO {} VALUES (42, 'clickhouse')".format('TABLE FUNCTION ' + table_function))
+    assert node1.query("SELECT * FROM {}".format(table_function)) == '42\tclickhouse\\0\\0\\0\\0\\0\\0\n'
 
 def test_enum_type(started_cluster):
     table_name = 'test_enum_type'

From 58b1e47e0d57ad4782856b50676c6803c272f7fa Mon Sep 17 00:00:00 2001
From: Anton Popov <pad11rus@gmail.com>
Date: Tue, 29 Dec 2020 16:27:47 +0300
Subject: [PATCH 502/504] Update string-functions.md

---
 docs/en/sql-reference/functions/string-functions.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/en/sql-reference/functions/string-functions.md b/docs/en/sql-reference/functions/string-functions.md
index ddb27756df4..83f2705693a 100644
--- a/docs/en/sql-reference/functions/string-functions.md
+++ b/docs/en/sql-reference/functions/string-functions.md
@@ -560,9 +560,9 @@ Result:
 
 ## encodeXMLComponent {#encode-xml-component}
 
-Escape characters to place string into XML text node or attribute.
+Escapes characters to place string into XML text node or attribute.
 
-The following five XML predefined entities will be replaced: `<`, `&`, `>`, `"`, `\'`.
+The following five XML predefined entities will be replaced: `<`, `&`, `>`, `"`, `'`.
 
 **Syntax** 
 

From 206660631cc757eae77ca4b96736681fd6854128 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Tue, 29 Dec 2020 17:50:25 +0300
Subject: [PATCH 503/504] Update datetime64.md

---
 docs/ru/sql-reference/data-types/datetime64.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/ru/sql-reference/data-types/datetime64.md b/docs/ru/sql-reference/data-types/datetime64.md
index b5e185c9132..6576bf9dc0d 100644
--- a/docs/ru/sql-reference/data-types/datetime64.md
+++ b/docs/ru/sql-reference/data-types/datetime64.md
@@ -96,7 +96,7 @@ FROM dt
 -   [Функции для работы с датой и временем](../../sql-reference/functions/date-time-functions.md)
 -   [Функции для работы с массивами](../../sql-reference/functions/array-functions.md)
 -   [Настройка `date_time_input_format`](../../operations/settings/settings.md#settings-date_time_input_format)
--   [Настройка `date_time_output_format`](../../operations/settings/settings.md#settings-date_time_output_format)
+-   [Настройка `date_time_output_format`](../../operations/settings/settings.md)
 -   [Конфигурационный параметр сервера `timezone`](../../operations/server-configuration-parameters/settings.md#server_configuration_parameters-timezone)
 -   [Операторы для работы с датой и временем](../../sql-reference/operators/index.md#operators-datetime)
 -   [Тип данных `Date`](date.md)

From 310c8138a69109f3e46e608df536081c58bace43 Mon Sep 17 00:00:00 2001
From: alexey-milovidov <milovidov@yandex-team.ru>
Date: Tue, 29 Dec 2020 17:56:55 +0300
Subject: [PATCH 504/504] Update formats.md

---
 docs/zh/interfaces/formats.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/zh/interfaces/formats.md b/docs/zh/interfaces/formats.md
index 808bff42335..b37ef559aa7 100644
--- a/docs/zh/interfaces/formats.md
+++ b/docs/zh/interfaces/formats.md
@@ -511,7 +511,7 @@ ClickHouse支持[NULL](../sql-reference/syntax.md), 在JSON输出中显示为`nu
 
 在这种格式中，一个JSON对象被解释为一个值。如果输入有几个JSON对象(逗号分隔)，它们将被解释为独立的行。
 
-这种格式只能对具有单个字段类型的表进行解析[String](../sql-reference/data-types/string.md)。其余的列必须设置为[DEFAULT](../sql-reference/statements/create/table.md#default)或[MATERIALIZED](../sql-reference/statements/create/table.md#materialized)，或者忽略。一旦将整个JSON对象收集为字符串，就可以使用[JSON函数](../sql-reference/functions/json-functions.md)运行它。
+这种格式只能对具有单个字段类型的表进行解析[String](../sql-reference/data-types/string.md)。其余的列必须设置为[DEFAULT](../sql-reference/statements/create.md)或[MATERIALIZED](../sql-reference/statements/create.md)，或者忽略。一旦将整个JSON对象收集为字符串，就可以使用[JSON函数](../sql-reference/functions/json-functions.md)运行它。
 
 **示例**