Fixed translation errors, part 1 [#CLICKHOUSE-3].

dbms/src/Common/Allocator.cpp

@@ -22,12 +22,12 @@ namespace ErrorCodes
 }
 
 
-/** Many modern allocators (for example, tcmalloc) do not know how to do a mremap for realloc,
+/** Many modern allocators (for example, tcmalloc) do not do a mremap for realloc,
   *  even in case of large enough chunks of memory.
   * Although this allows you to increase performance and reduce memory consumption during realloc.
-  * To fix this, do the mremap yourself if the chunk of memory is large enough.
+  * To fix this, we do mremap manually if the chunk of memory is large enough.
   * The threshold (64 MB) is chosen quite large, since changing the address space is
-  *  rather slow, especially in the case of a large number of threads.
+  *  very slow, especially in the case of a large number of threads.
   * We expect that the set of operations mmap/something to do/mremap can only be performed about 1000 times per second.
   *
   * PS. This is also required, because tcmalloc can not allocate a chunk of memory greater than 16 GB.

dbms/src/Common/ArenaWithFreeLists.h

@@ -13,7 +13,7 @@ namespace DB
   * For this, the requested size is rounded up to the power of two
   *  (or up to 8, if less, or using memory allocation outside Arena if the size is greater than 65536).
   * When freeing memory, for each size (14 options in all: 8, 16 ... 65536),
-  *  a one-link list of free blocks is kept track.
+  *  a single-linked list of free blocks is kept track.
   * When allocating, we take the head of the list of free blocks,
   *  or, if the list is empty - allocate a new block using Arena.
   */

dbms/src/Common/AutoArray.h

@@ -26,7 +26,7 @@ namespace DB
   * Designed for situations in which many arrays of the same small size are created,
   *  but the size is not known at compile time.
   * Also gives a significant advantage in cases where it is important that `sizeof` is minimal.
-  * For example, if arrays are put in an open-addressing hash table with inplace storage of values ​​(like HashMap)
+  * For example, if arrays are put in an open-addressing hash table with inplace storage of values (like HashMap)
   *
   * In this case, compared to std::vector:
   * - for arrays of 1 element size - an advantage of about 2 times;
@@ -82,7 +82,7 @@ public:
         init(size_, dont_init_elems);
     }
 
-    /** Preposition.
+    /** Move operations.
       */
     AutoArray(AutoArray && src)
     {

dbms/src/Common/CompactArray.h

@@ -18,7 +18,7 @@ namespace ErrorCodes
 /** Compact array for data storage, size `content_width`, in bits, of which is
   * less than one byte. Instead of storing each value in a separate
   * bytes, which leads to a waste of 37.5% of the space for content_width = 5, CompactArray stores
-  * adjacent `content_width`-bit values ​​in the byte array, that is actually CompactArray
+  * adjacent `content_width`-bit values in the byte array, that is actually CompactArray
   * simulates an array of `content_width`-bit values.
   */
 template <typename BucketIndex, UInt8 content_width, size_t bucket_count>

dbms/src/Common/Exception.h

@@ -74,7 +74,7 @@ void throwFromErrno(const std::string & s, int code = 0, int the_errno = errno);
 
 
 /** Try to write an exception to the log (and forget about it).
-  * Can be used in destructors in the catch block (...).
+  * Can be used in destructors in the catch-all block.
   */
 void tryLogCurrentException(const char * log_name, const std::string & start_of_message = "");
 void tryLogCurrentException(Poco::Logger * logger, const std::string & start_of_message = "");

dbms/src/Common/HashTable/ClearableHashSet.h

@@ -7,7 +7,7 @@
 /** A hash table that allows you to clear the table in O(1).
   * Even simpler than HashSet: Key and Mapped must be POD-types.
   *
-  * Instead of this class, you could just use the couple <version, key> in the HashSet as the key
+  * Instead of this class, you could just use the pair (version, key) in the HashSet as the key
   * but then the table would accumulate all the keys that it ever stored, and it was unreasonably growing.
   * This class goes a step further and considers the keys with the old version empty in the hash table.
   */

dbms/src/Common/HashTable/Hash.h

@@ -4,10 +4,17 @@
 
 
 /** Hash functions that are better than the trivial function std::hash.
-  * (when aggregated by the visitor ID, the performance increase is more than 5 times)
+  *
+  * Example: when aggregated by the visitor ID, the performance increase is more than 5 times.
+  * This is because of following reasons:
+  * - in Yandex, visitor identifier is an integer that has timestamp with seconds resolution in lower bits;
+  * - in typical implementation of standard library, hash function for integers is trivial and just use lower bits;
+  * - traffic is non-uniformly distributed across a day;
+  * - we are using open-addressing linear probing hash tables that are most critical to hash function quality,
+  *   and trivial hash function gives disasterous results.
   */
 
-/** Taken from MurmurHash.
+/** Taken from MurmurHash. This is Murmur finalizer.
   * Faster than intHash32 when inserting into the hash table UInt64 -> UInt64, where the key is the visitor ID.
   */
 inline DB::UInt64 intHash64(DB::UInt64 x)
@@ -22,20 +29,21 @@ inline DB::UInt64 intHash64(DB::UInt64 x)
 }
 
 /** CRC32C is not very high-quality as a hash function,
-  *  according to avalanche and bit independence tests, as well as a small number of bits,
+  *  according to avalanche and bit independence tests (see SMHasher software), as well as a small number of bits,
   *  but can behave well when used in hash tables,
   *  due to high speed (latency 3 + 1 clock cycle, throughput 1 clock cycle).
   * Works only with SSE 4.2 support.
-  * Used asm instead of intrinsics, so you do not have to build the entire project with -msse4.
   */
+#if __SSE4_2__
+#include <nmmintrin.h>
+#endif
+
 inline DB::UInt64 intHashCRC32(DB::UInt64 x)
 {
-#if defined(__x86_64__)
+#if __SSE4_2__
-    DB::UInt64 crc = -1ULL;
+    return _mm_crc32_u64(-1ULL, x);
-    asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x));
-    return crc;
 #else
-    /// On other platforms we do not need CRC32. NOTE This can be confusing.
+    /// On other platforms we do not have CRC32. NOTE This can be confusing.
     return intHash64(x);
 #endif
 }
@@ -128,7 +136,7 @@ struct TrivialHash
 };
 
 
-/** A relatively good non-cryptic hash function from UInt64 to UInt32.
+/** A relatively good non-cryptographic hash function from UInt64 to UInt32.
   * But worse (both in quality and speed) than just cutting intHash64.
   * Taken from here: http://www.concentric.net/~ttwang/tech/inthash.htm
   *
@@ -136,9 +144,14 @@ struct TrivialHash
   * This change did not affect the smhasher test results.
   *
   * It is recommended to use different salt for different tasks.
-  * That was the case that in the database values ​​were sorted by hash (for low-quality pseudo-random spread),
+  * That was the case that in the database values were sorted by hash (for low-quality pseudo-random spread),
   *  and in another place, in the aggregate function, the same hash was used in the hash table,
   *  as a result, this aggregate function was monstrously slowed due to collisions.
+  *
+  * NOTE Salting is far from perfect, because it commutes with first steps of calculation.
+  *
+  * NOTE As mentioned, this function is slower than intHash64.
+  * But occasionaly, it is faster, when written in a loop and loop is vectorized.
   */
 template <DB::UInt64 salt>
 inline DB::UInt32 intHash32(DB::UInt64 key)

dbms/src/Common/HashTable/HashMap.h

@@ -66,7 +66,7 @@ struct HashMapCell
     /// Do I 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 removed.
+    /// Whether the cell was deleted.
     bool isDeleted() const { return false; }
 
     void setMapped(const value_type & value_) { value.second = value_.second; }

dbms/src/Common/HashTable/HashTable.h

@@ -46,9 +46,9 @@ namespace ErrorCodes
 
 /** The state of the hash table that affects the properties of its cells.
   * Used as a template parameter.
-  * For example, there is an implementation of an instantly cleared hash table - ClearableHashMap.
+  * For example, there is an implementation of an instantly clearable hash table - ClearableHashMap.
   * For it, each cell holds the version number, and in the hash table itself is the current version.
-  *  When cleaning, the current version simply increases; All cells with a mismatching version are considered empty.
+  *  When clearing, the current version simply increases; All cells with a mismatching version are considered empty.
   *  Another example: for an approximate calculation of the number of unique visitors, there is a hash table for UniquesHashSet.
   *  It has the concept of "degree". At each overflow, cells with keys that do not divide by the corresponding power of the two are deleted.
   */
@@ -77,8 +77,8 @@ void set(T & x) { x = 0; }
 };
 
 
-/** Compile-time cell interface of the hash table.
+/** Compile-time interface for cell of the hash table.
-  * Different cells are used to implement different hash tables.
+  * Different cell types are used to implement different hash tables.
   * The cell must contain a key.
   * It can also contain a value and arbitrary additional data
   *  (example: the stored hash value; version number for ClearableHashMap).
@@ -95,7 +95,7 @@ struct HashTableCell
 
     /// Create a cell with the given key / key and value.
     HashTableCell(const Key & key_, const State & state) : key(key_) {}
-///    HashTableCell(const value_type & value_, const State & state) : key(value_) {}
+/// HashTableCell(const value_type & value_, const State & state) : key(value_) {}
 
     /// Get what the value_type of the container will be.
     value_type & getValue()             { return key; }
@@ -126,13 +126,13 @@ struct HashTableCell
     /// Set the key value to zero.
     void setZero() { ZeroTraits::set(key); }
 
-    /// Do I need to store the zero key separately (that is, can a zero key be inserted into the hash table).
+    /// 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 displayed value, if any (for HashMap), to the corresponding `value`.
+    /// Set the mapped value, if any (for HashMap), to the corresponding `value`.
     void setMapped(const value_type & value) {}
 
     /// Serialization, in binary and text form.
@@ -145,7 +145,7 @@ struct HashTableCell
 };
 
 
-/** Determines the size of the hash table, and when and how many times it should be resized.
+/** Determines the size of the hash table, and when and how much it should be resized.
   */
 template <size_t initial_size_degree = 8>
 struct HashTableGrower
@@ -195,7 +195,7 @@ struct HashTableGrower
 /** When used as a Grower, it turns a hash table into something like a lookup table.
   * It remains non-optimal - the cells store the keys.
   * Also, the compiler can not completely remove the code of passing through the collision resolution chain, although it is not needed.
-  * TODO Make a full lookup table.
+  * TODO Make a proper lookup table.
   */
 template <size_t key_bits>
 struct HashTableFixedGrower
@@ -212,7 +212,7 @@ struct HashTableFixedGrower
 };
 
 
-/** If you want to store the null key separately - a place to store it. */
+/** If you want to store the zero key separately - a place to store it. */
 template <bool need_zero_value_storage, typename Cell>
 struct ZeroValueStorage;
 
@@ -272,7 +272,7 @@ protected:
     using cell_type = Cell;
 
     size_t m_size = 0;        /// Amount of elements
-    Cell * buf;               /// A piece of memory for all elements except the element with key 0.
+    Cell * buf;               /// A piece of memory for all elements except the element with zero key.
     Grower grower;
 
 #ifdef DBMS_HASH_MAP_COUNT_COLLISIONS
@@ -334,7 +334,7 @@ protected:
 
         /** In case of exception for the object to remain in the correct state,
           *  changing the variable `grower` (which determines the buffer size of the hash table)
-          *  postpone for a moment after a real buffer change.
+          *  is postponed for a moment after a real buffer change.
           * The temporary variable `new_grower` is used to determine the new size.
           */
         Grower new_grower = grower;
@@ -410,7 +410,7 @@ protected:
         memcpy(&buf[place_value], &x, sizeof(x));
         x.setZero();
 
-        /// Then the elements that previously were in conflict with this can move to the old place.
+        /// Then the elements that previously were in collision with this can move to the old place.
     }
 
 
@@ -638,7 +638,7 @@ protected:
     }
 
 
-    /// Only for non-zero keys. Find the right place, insert the key there, if it does not already exist, return the iterator to the cell.
+    /// Only for non-zero keys. Find the right place, insert the key there, if it does not already exist. Set iterator to the cell in output parameter.
     void ALWAYS_INLINE emplaceNonZero(Key x, iterator & it, bool & inserted, size_t hash_value)
     {
         size_t place_value = findCell(x, hash_value, grower.place(hash_value));
@@ -664,8 +664,8 @@ protected:
             }
             catch (...)
             {
-                /** If you do not do it, then there will be problems.
+                /** If we have not resized successfully, then there will be problems.
-                  * After all, there remains a key, but uninitialized mapped-value,
+                  * There remains a key, but uninitialized mapped-value,
                   *  which, perhaps, can not even be called a destructor.
                   */
                 --m_size;
@@ -698,7 +698,7 @@ public:
       * return the iterator to a position that can be used for `placement new` of value,
       * as well as the flag - whether a new key was inserted.
       *
-      * You are required to make `placement new` of value ​​if you inserted a new key,
+      * You have to make `placement new` of value if you inserted a new key,
       * since when destroying a hash table, it will call the destructor!
       *
       * Example usage:

dbms/src/Common/HashTable/SmallTable.h

@@ -215,7 +215,7 @@ public:
       * return the iterator to a position that can be used for `placement new` of value,
       * as well as the flag - whether a new key was inserted.
       *
-      * You have to make `placement new` of value ​​if you inserted a new key,
+      * You have to make `placement new` of value if you inserted a new key,
       * since when destroying a hash table, a destructor will be called for it!
       *
       * Example usage:

dbms/src/Common/HashTable/TwoLevelHashTable.h

@@ -224,7 +224,7 @@ public:
       * return the iterator to a position that can be used for `placement new` value,
       * as well as the flag - whether a new key was inserted.
       *
-      * You have to make `placement new` values ​​if you inserted a new key,
+      * You have to make `placement new` values if you inserted a new key,
       * since when destroying a hash table, the destructor will be invoked for it!
       *
       * Example usage:

dbms/src/Common/HyperLogLogBiasEstimator.h

@@ -22,7 +22,7 @@ public:
         return false;
     }
 
-    /// Maximum number of unique values ​​to which the correction should apply
+    /// Maximum number of unique values to which the correction should apply
     /// from the LinearCounting algorithm.
     static double getThreshold()
     {
@@ -83,7 +83,7 @@ private:
 };
 
 /** Trivial case of HyperLogLogBiasEstimator: used if we do not want to fix
-  * error. This has meaning for small values ​​of the accuracy parameter, for example 5 or 12.
+  * error. This has meaning for small values of the accuracy parameter, for example 5 or 12.
   * Then the corrections from the original version of the HyperLogLog algorithm are applied.
   * See "HyperLogLog: The analysis of a near-optimal cardinality estimation algorithm"
   * (P. Flajolet et al., AOFA '07: Proceedings of the 2007 International Conference on Analysis

dbms/src/Common/SimpleCache.h

@@ -8,7 +8,7 @@
 
 /** The simplest cache for a free function.
   * You can also pass a static class method or lambda without capturing.
-  * The size is unlimited. Values ​​are not obsolete.
+  * The size is unlimited. Values are not obsolete.
   * To synchronize, use mutex.
   * Suitable only for the simplest cases.
   *

dbms/src/Common/UInt128.h

@@ -4,6 +4,10 @@
 #include <IO/ReadHelpers.h>
 #include <IO/WriteHelpers.h>
 
+#if __SSE4_2__
+#include <nmmintrin.h>
+#endif
+
 
 namespace DB
 {
@@ -42,15 +46,15 @@ struct UInt128Hash
     }
 };
 
-#if defined(__x86_64__)
+#if __SSE4_2__
 
 struct UInt128HashCRC32
 {
     size_t operator()(UInt128 x) const
     {
         UInt64 crc = -1ULL;
-        asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x.first));
+        crc = _mm_crc32_u64(crc, x.first);
-        asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x.second));
+        crc = _mm_crc32_u64(crc, x.second);
         return crc;
     }
 };
@@ -122,17 +126,17 @@ struct UInt256Hash
     }
 };
 
-#if defined(__x86_64__)
+#if __SSE4_2__
 
 struct UInt256HashCRC32
 {
     size_t operator()(UInt256 x) const
     {
         UInt64 crc = -1ULL;
-        asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x.a));
+        crc = _mm_crc32_u64(crc, x.a);
-        asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x.b));
+        crc = _mm_crc32_u64(crc, x.b);
-        asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x.c));
+        crc = _mm_crc32_u64(crc, x.c);
-        asm("crc32q %[x], %[crc]\n" : [crc] "+r" (crc) : [x] "rm" (x.d));
+        crc = _mm_crc32_u64(crc, x.d);
         return crc;
     }
 };

dbms/src/Common/Volnitsky.h

@@ -19,7 +19,7 @@
   * - when initializing, fill in an open-addressing linear probing hash table of the form
   *    hash from the bigram of needle -> the position of this bigram in needle + 1.
   *    (one is added only to distinguish zero offset from an empty cell)
-  * - the keys are not stored in the hash table, only the values ​​are stored;
+  * - the keys are not stored in the hash table, only the values are stored;
   * - bigrams can be inserted several times if they occur in the needle several times;
   * - when searching, take from haystack bigram, which should correspond to the last bigram of needle (comparing from the end);
   * - look for it in the hash table, if found - get the offset from the hash table and compare the string bytewise;

dbms/src/Core/BlockInfo.h

@@ -15,7 +15,7 @@ struct BlockInfo
 {
     /** is_overflows:
       * After running GROUP BY ... WITH TOTALS with the max_rows_to_group_by and group_by_overflow_mode = 'any' settings,
-      *  a row is inserted in the separate block with aggregated values ​​that have not passed max_rows_to_group_by.
+      *  a row is inserted in the separate block with aggregated values that have not passed max_rows_to_group_by.
       * If it is such a block, then is_overflows is set to true for it.
       */
 
@@ -36,10 +36,10 @@ struct BlockInfo
 
 #undef DECLARE_FIELD
 
-    /// Write the values ​​in binary form. NOTE: You could use protobuf, but it would be overkill for this case.
+    /// Write the values in binary form. NOTE: You could use protobuf, but it would be overkill for this case.
     void write(WriteBuffer & out) const;
 
-    /// Read the values ​​in binary form.
+    /// Read the values in binary form.
     void read(ReadBuffer & in);
 };
 

dbms/src/Core/Defines.h

@@ -27,7 +27,7 @@
 #define DEFAULT_MAX_COMPRESS_BLOCK_SIZE                         1048576
 
 /** Which blocks by default read the data (by number of rows).
-  * Smaller values ​​give better cache locality, less consumption of RAM, but more overhead to process the query.
+  * Smaller values give better cache locality, less consumption of RAM, but more overhead to process the query.
   */
 #define DEFAULT_BLOCK_SIZE                                         65536
 

dbms/src/Core/StringRef.h

@@ -171,17 +171,7 @@ struct StringRefHash64
 
 #if __SSE4_2__
 
-#ifdef __SSE4_1__
 #include <smmintrin.h>
-#else
-
-inline UInt64 _mm_crc32_u64(UInt64 crc, UInt64 value)
-{
-    asm("crc32q %[value], %[crc]\n" : [crc] "+r" (crc) : [value] "rm" (value));
-    return crc;
-}
-
-#endif
 
 /// Parts are taken from CityHash.
 

dbms/src/Core/Types.h

@@ -9,7 +9,7 @@
 namespace DB
 {
 
-/** Data types for representing values ​​from a database in RAM.
+/** Data types for representing values from a database in RAM.
   */
 
 STRONG_TYPEDEF(char, Null);

dbms/src/DataTypes/DataTypeAggregateFunction.h

@@ -41,7 +41,7 @@ public:
 
     DataTypePtr clone() const override { return std::make_shared<DataTypeAggregateFunction>(function, argument_types, parameters); }
 
-    /// NOTE These two functions for serializing single values ​​are incompatible with the functions below.
+    /// NOTE These two functions for serializing single values are incompatible with the functions below.
     void serializeBinary(const Field & field, WriteBuffer & ostr) const override;
     void deserializeBinary(Field & field, ReadBuffer & istr) const override;
 

dbms/src/DataTypes/DataTypeSet.h

@@ -6,7 +6,7 @@
 namespace DB
 {
 
-/** The data type corresponding to the set of values ​​in the IN section.
+/** The data type corresponding to the set of values in the IN section.
   * Used only as an intermediate option when evaluating expressions.
   */
 class DataTypeSet final : public IDataTypeDummy

dbms/src/Databases/DatabaseCloud.h

@@ -26,7 +26,7 @@ namespace DB
   *
   * cloud_path                   - the path to the "cloud"; There may be several different independent clouds
         /table_definitions       - set of unique table definitions so you do not write them many times for a large number of tables
-            /hash128 -> sql      - mapping: hash from table definition (identifier) ​​-> table definition itself as CREATE query
+            /hash128 -> sql      - mapping: hash from table definition (identifier) -> table definition itself as CREATE query
         /tables                  - list of tables
             /database_name       - name of the database
                 /name_hash_mod -> compressed_table_list

dbms/src/Storages/MergeTree/MergeTreeDataSelectExecutor.h

@@ -28,7 +28,7 @@ public:
         QueryProcessingStage::Enum & processed_stage,
         size_t max_block_size,
         unsigned threads,
-        size_t * inout_part_index,    /// If not nullptr, from this counter values are taken ​​for the virtual column _part_index.
+        size_t * inout_part_index,    /// If not nullptr, from this counter values are taken for the virtual column _part_index.
         Int64 max_block_number_to_read) const;
 
 private:

dbms/src/Storages/MergeTree/MergedBlockOutputStream.h

@@ -136,7 +136,7 @@ public:
 private:
     void init();
 
-    /** If `permutation` is given, it rearranges the values ​​in the columns when writing.
+    /** If `permutation` is given, it rearranges the values in the columns when writing.
       * This is necessary to not keep the whole block in the RAM to sort it.
       */
     void writeImpl(const Block & block, const IColumn::Permutation * permutation);

dbms/src/Storages/MergeTree/ReplicatedMergeTreeLogEntry.h

@@ -63,7 +63,7 @@ struct ReplicatedMergeTreeLogEntryData
     /// The name of resulting part.
     /// For DROP_RANGE, the name of a non-existent part. You need to remove all the parts covered by it.
     String new_part_name;
-    String block_id;    /// For parts of level zero, the block identifier for deduplication (node ​​name in /blocks /).
+    String block_id;    /// For parts of level zero, the block identifier for deduplication (node name in /blocks /).
 
     Strings parts_to_merge;
     bool deduplicate = false; /// Do deduplicate on merge

dbms/src/Storages/MergeTree/ReplicatedMergeTreeRestartingThread.h

@@ -14,7 +14,7 @@ class StorageReplicatedMergeTree;
 
 
 /** Initializes ZK session.
-  * Exposes ephemeral nodes. It sets the node values ​​that are required for replica detection.
+  * Exposes ephemeral nodes. It sets the node values that are required for replica detection.
   * Starts participation in the leader selection. Starts all background threads.
   * Then monitors whether the session has expired. And if it expired, it will reinitialize it.
   */