Merge branch 'master' into improve_access_type

.github/workflows/master.yml

@@ -149,7 +149,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH"
   SplitBuildSmokeTest:
     needs: [BuilderDebSplitted]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, style-checker]
     steps:
       - name: Set envs
@@ -316,7 +315,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinRelease:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -362,7 +360,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinGCC:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -636,7 +633,6 @@ jobs:
 ##########################################################################################
   BuilderDebSplitted:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -682,7 +678,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinTidy:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -728,7 +723,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinDarwin:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -774,7 +768,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinAarch64:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -820,7 +813,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinFreeBSD:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -866,7 +858,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinDarwinAarch64:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs
@@ -912,7 +903,6 @@ jobs:
           sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
   BuilderBinPPC64:
     needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
     runs-on: [self-hosted, builder]
     steps:
       - name: Set envs

docker/test/stateful/run.sh

@@ -115,7 +115,7 @@ function run_tests()
     fi
 
     set +e
-    clickhouse-test --testname --shard --zookeeper --check-zookeeper-session --no-stateless --hung-check --print-time \
+    clickhouse-test -j 2 --testname --shard --zookeeper --check-zookeeper-session --no-stateless --hung-check --print-time \
         --skip 00168_parallel_processing_on_replicas "${ADDITIONAL_OPTIONS[@]}" \
         "$SKIP_TESTS_OPTION" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee test_output/test_result.txt
 

docs/en/sql-reference/aggregate-functions/reference/grouparraysorted.md

@@ -0,0 +1,48 @@
+---
+toc_priority: 108
+---
+
+# groupArraySorted {#groupArraySorted}
+
+Returns an array with the first N items in ascending order.
+
+``` sql
+groupArraySorted(N)(column)
+```
+
+**Arguments**
+
+-   `N` – The number of elements to return.
+
+If the parameter is omitted, default value 10 is used.
+
+**Arguments**
+
+-   `column` – The value.
+-   `expr` — Optional. The field or expresion to sort by. If not set values are sorted by themselves.
+
+**Example**
+
+Gets the first 10 numbers:
+
+``` sql
+SELECT groupArraySorted(10)(number) FROM numbers(100)
+```
+
+``` text
+┌─groupArraySorted(10)(number)─┐
+│ [0,1,2,3,4,5,6,7,8,9]        │
+└──────────────────────────────┘
+```
+
+Or the last 10:
+
+``` sql
+SELECT groupArraySorted(10)(number, -number) FROM numbers(100)
+```
+
+``` text
+┌─groupArraySorted(10)(number, negate(number))─┐
+│ [99,98,97,96,95,94,93,92,91,90]              │
+└──────────────────────────────────────────────┘
+```

docs/en/sql-reference/aggregate-functions/reference/index.md

@@ -35,6 +35,7 @@ ClickHouse-specific aggregate functions:
 -   [groupArrayInsertAt](../../../sql-reference/aggregate-functions/reference/grouparrayinsertat.md)
 -   [groupArrayMovingAvg](../../../sql-reference/aggregate-functions/reference/grouparraymovingavg.md)
 -   [groupArrayMovingSum](../../../sql-reference/aggregate-functions/reference/grouparraymovingsum.md)
+-   [groupArraySorted](../../../sql-reference/aggregate-functions/reference/grouparraysorted.md)
 -   [groupBitAnd](../../../sql-reference/aggregate-functions/reference/groupbitand.md)
 -   [groupBitOr](../../../sql-reference/aggregate-functions/reference/groupbitor.md)
 -   [groupBitXor](../../../sql-reference/aggregate-functions/reference/groupbitxor.md)

src/AggregateFunctions/AggregateFunctionGroupArraySorted.cpp

@@ -0,0 +1,147 @@
+#include <AggregateFunctions/AggregateFunctionFactory.h>
+#include <AggregateFunctions/AggregateFunctionGroupArraySorted.h>
+#include <AggregateFunctions/FactoryHelpers.h>
+#include <AggregateFunctions/Helpers.h>
+#include <DataTypes/DataTypeDate.h>
+#include <DataTypes/DataTypeDateTime.h>
+#include <DataTypes/DataTypeString.h>
+#include <Common/FieldVisitorConvertToNumber.h>
+
+
+static inline constexpr UInt64 GROUP_SORTED_ARRAY_MAX_SIZE = 0xFFFFFF;
+static inline constexpr UInt64 GROUP_SORTED_ARRAY_DEFAULT_THRESHOLD = 10;
+
+
+namespace DB
+{
+struct Settings;
+
+namespace ErrorCodes
+{
+    extern const int ARGUMENT_OUT_OF_BOUND;
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+}
+
+
+namespace
+{
+    template <typename T, bool expr_sorted, typename TColumnB, bool is_plain_b>
+    class AggregateFunctionGroupArraySortedNumeric : public AggregateFunctionGroupArraySorted<T, false, expr_sorted, TColumnB, is_plain_b>
+    {
+        using AggregateFunctionGroupArraySorted<T, false, expr_sorted, TColumnB, is_plain_b>::AggregateFunctionGroupArraySorted;
+    };
+
+    template <typename T, bool expr_sorted, typename TColumnB, bool is_plain_b>
+    class AggregateFunctionGroupArraySortedFieldType
+        : public AggregateFunctionGroupArraySorted<typename T::FieldType, false, expr_sorted, TColumnB, is_plain_b>
+    {
+        using AggregateFunctionGroupArraySorted<typename T::FieldType, false, expr_sorted, TColumnB, is_plain_b>::
+            AggregateFunctionGroupArraySorted;
+        DataTypePtr getReturnType() const override { return std::make_shared<DataTypeArray>(std::make_shared<T>()); }
+    };
+
+    template <template <typename, bool, typename, bool> class AggregateFunctionTemplate, typename TColumnA, bool expr_sorted, typename TColumnB, bool is_plain_b, typename... TArgs>
+    AggregateFunctionPtr
+    createAggregateFunctionGroupArraySortedTypedFinal(TArgs && ... args)
+    {
+        return AggregateFunctionPtr(new AggregateFunctionTemplate<TColumnA, expr_sorted, TColumnB, is_plain_b>(std::forward<TArgs>(args)...));
+    }
+
+    template <bool expr_sorted = false, typename TColumnB = UInt64, bool is_plain_b = false>
+    AggregateFunctionPtr
+    createAggregateFunctionGroupArraySortedTyped(const DataTypes & argument_types, const Array & params, UInt64 threshold)
+    {
+#define DISPATCH(A, C, B) \
+    if (which.idx == TypeIndex::A) \
+        return createAggregateFunctionGroupArraySortedTypedFinal<C, B, expr_sorted, TColumnB, is_plain_b>(threshold, argument_types, params);
+#define DISPATCH_NUMERIC(A) DISPATCH(A, AggregateFunctionGroupArraySortedNumeric, A)
+        WhichDataType which(argument_types[0]);
+        FOR_NUMERIC_TYPES(DISPATCH_NUMERIC)
+        DISPATCH(Enum8, AggregateFunctionGroupArraySortedNumeric, Int8)
+        DISPATCH(Enum16, AggregateFunctionGroupArraySortedNumeric, Int16)
+        DISPATCH(Date, AggregateFunctionGroupArraySortedFieldType, DataTypeDate)
+        DISPATCH(DateTime, AggregateFunctionGroupArraySortedFieldType, DataTypeDateTime)
+#undef DISPATCH
+#undef DISPATCH_NUMERIC
+
+        if (argument_types[0]->isValueUnambiguouslyRepresentedInContiguousMemoryRegion())
+        {
+            return AggregateFunctionPtr(new AggregateFunctionGroupArraySorted<StringRef, true, expr_sorted, TColumnB, is_plain_b>(
+                threshold, argument_types, params));
+        }
+        else
+        {
+            return AggregateFunctionPtr(new AggregateFunctionGroupArraySorted<StringRef, false, expr_sorted, TColumnB, is_plain_b>(
+                threshold, argument_types, params));
+        }
+    }
+
+
+    AggregateFunctionPtr createAggregateFunctionGroupArraySorted(
+        const std::string & name, const DataTypes & argument_types, const Array & params, const Settings *)
+    {
+        UInt64 threshold = GROUP_SORTED_ARRAY_DEFAULT_THRESHOLD;
+
+        if (params.size() == 1)
+        {
+            UInt64 k = applyVisitor(FieldVisitorConvertToNumber<UInt64>(), params[0]);
+
+            if (k > GROUP_SORTED_ARRAY_MAX_SIZE)
+                throw Exception(
+                    "Too large parameter(s) for aggregate function " + name + ". Maximum: " + toString(GROUP_SORTED_ARRAY_MAX_SIZE),
+                    ErrorCodes::ARGUMENT_OUT_OF_BOUND);
+
+            if (k == 0)
+                throw Exception("Parameter 0 is illegal for aggregate function " + name, ErrorCodes::ARGUMENT_OUT_OF_BOUND);
+
+            threshold = k;
+        }
+        else if (!params.empty())
+        {
+            throw Exception("Aggregate function " + name + " only supports 1 parameter.", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+        }
+
+        if (argument_types.size() == 2)
+        {
+            if (isNumber(argument_types[1]))
+            {
+#define DISPATCH2(A, B) \
+    if (which.idx == TypeIndex::A) \
+        return createAggregateFunctionGroupArraySortedTyped<true, B>(argument_types, params, threshold);
+#define DISPATCH(A) DISPATCH2(A, A)
+                WhichDataType which(argument_types[1]);
+                FOR_NUMERIC_TYPES(DISPATCH)
+                DISPATCH2(Enum8, Int8)
+                DISPATCH2(Enum16, Int16)
+#undef DISPATCH
+#undef DISPATCH2
+                throw Exception("Invalid parameter type.", ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+            }
+            else if (argument_types[1]->isValueUnambiguouslyRepresentedInContiguousMemoryRegion())
+            {
+                return createAggregateFunctionGroupArraySortedTyped<true, StringRef, true>(argument_types, params, threshold);
+            }
+            else
+            {
+                return createAggregateFunctionGroupArraySortedTyped<true, StringRef, false>(argument_types, params, threshold);
+            }
+        }
+        else if (argument_types.size() == 1)
+        {
+            return createAggregateFunctionGroupArraySortedTyped<>(argument_types, params, threshold);
+        }
+        else
+        {
+            throw Exception(
+                "Aggregate function " + name + " requires one or two parameters.", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+        }
+    }
+}
+
+void registerAggregateFunctionGroupArraySorted(AggregateFunctionFactory & factory)
+{
+    AggregateFunctionProperties properties = {.returns_default_when_only_null = false, .is_order_dependent = true};
+    factory.registerFunction("groupArraySorted", {createAggregateFunctionGroupArraySorted, properties});
+}
+}

src/AggregateFunctions/AggregateFunctionGroupArraySorted.h

@@ -0,0 +1,310 @@
+#pragma once
+
+#include <Columns/ColumnArray.h>
+#include <DataTypes/DataTypeArray.h>
+
+#include <AggregateFunctions/AggregateFunctionGroupArraySortedData.h>
+#include <AggregateFunctions/IAggregateFunction.h>
+
+namespace DB
+{
+template <typename TColumn, bool is_plain>
+inline TColumn readItem(const IColumn * column, Arena * arena, size_t row)
+{
+    if constexpr (std::is_same_v<TColumn, StringRef>)
+    {
+        if constexpr (is_plain)
+        {
+            StringRef str = column->getDataAt(row);
+            auto ptr = arena->alloc(str.size);
+            std::copy(str.data, str.data + str.size, ptr);
+            return StringRef(ptr, str.size);
+        }
+        else
+        {
+            const char * begin = nullptr;
+            return column->serializeValueIntoArena(row, *arena, begin);
+        }
+    }
+    else
+    {
+        if constexpr (std::is_same_v<TColumn, UInt64>)
+            return column->getUInt(row);
+        else
+            return column->getInt(row);
+    }
+}
+
+template <typename TColumn, typename TFilter = void>
+size_t
+getFirstNElements_low_threshold(const TColumn * data, int num_elements, int threshold, size_t * results, const TFilter * filter = nullptr)
+{
+    for (int i = 0; i < threshold; i++)
+    {
+        results[i] = 0;
+    }
+
+    threshold = std::min(num_elements, threshold);
+    int current_max = 0;
+    int cur;
+    int z;
+    for (int i = 0; i < num_elements; i++)
+    {
+        if constexpr (!std::is_same_v<TFilter, void>)
+        {
+            if (filter[i] == 0)
+                continue;
+        }
+
+        //Starting from the highest values and we look for the immediately lower than the given one
+        for (cur = current_max; cur > 0; cur--)
+        {
+            if (data[i] > data[results[cur - 1]])
+                break;
+        }
+
+        if (cur < threshold)
+        {
+            //Move all the higher values 1 position to the right
+            for (z = std::min(threshold - 1, current_max); z > cur; z--)
+                results[z] = results[z - 1];
+
+            if (current_max < threshold)
+                ++current_max;
+
+            //insert element into the given position
+            results[cur] = i;
+        }
+    }
+
+    return current_max;
+}
+
+template <typename T>
+struct SortableItem
+{
+    T a;
+    size_t b;
+    bool operator<(const SortableItem & other) const { return (this->a < other.a); }
+};
+
+template <typename TColumn, typename TFilter = void>
+size_t getFirstNElements_high_threshold(
+    const TColumn * data, size_t num_elements, size_t threshold, size_t * results, const TFilter * filter = nullptr)
+{
+    std::vector<SortableItem<TColumn>> dataIndexed(num_elements);
+    size_t num_elements_filtered = 0;
+
+    for (size_t i = 0; i < num_elements; i++)
+    {
+        if constexpr (!std::is_same_v<TFilter, void>)
+        {
+            if (filter[i] == 0)
+                continue;
+        }
+
+        dataIndexed.data()[num_elements_filtered].a = data[i];
+        dataIndexed.data()[num_elements_filtered].b = i;
+        num_elements_filtered++;
+    }
+
+    threshold = std::min(num_elements_filtered, threshold);
+
+    std::nth_element(dataIndexed.data(), dataIndexed.data() + threshold, dataIndexed.data() + num_elements_filtered);
+    std::sort(dataIndexed.data(), dataIndexed.data() + threshold);
+
+    for (size_t i = 0; i < threshold; i++)
+    {
+        results[i] = dataIndexed[i].b;
+    }
+
+    return threshold;
+}
+
+static const size_t THRESHOLD_MAX_CUSTOM_FUNCTION = 1000;
+
+template <typename TColumn>
+size_t getFirstNElements(const TColumn * data, size_t num_elements, size_t threshold, size_t * results, const UInt8 * filter = nullptr)
+{
+    if (threshold < THRESHOLD_MAX_CUSTOM_FUNCTION)
+    {
+        if (filter != nullptr)
+            return getFirstNElements_low_threshold(data, num_elements, threshold, results, filter);
+        else
+            return getFirstNElements_low_threshold(data, num_elements, threshold, results);
+    }
+    else
+    {
+        if (filter != nullptr)
+            return getFirstNElements_high_threshold(data, num_elements, threshold, results, filter);
+        else
+            return getFirstNElements_high_threshold(data, num_elements, threshold, results);
+    }
+}
+
+template <typename TColumnA, bool is_plain_a, bool use_column_b, typename TColumnB, bool is_plain_b>
+class AggregateFunctionGroupArraySorted : public IAggregateFunctionDataHelper<
+                                              AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>,
+                                              AggregateFunctionGroupArraySorted<TColumnA, is_plain_a, use_column_b, TColumnB, is_plain_b>>
+{
+protected:
+    using State = AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>;
+    using Base = IAggregateFunctionDataHelper<
+        AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>,
+        AggregateFunctionGroupArraySorted>;
+
+    UInt64 threshold;
+    DataTypePtr & input_data_type;
+    mutable std::mutex mutex;
+
+    static void deserializeAndInsert(StringRef str, IColumn & data_to);
+
+public:
+    AggregateFunctionGroupArraySorted(UInt64 threshold_, const DataTypes & argument_types_, const Array & params)
+        : IAggregateFunctionDataHelper<
+            AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>,
+            AggregateFunctionGroupArraySorted>(argument_types_, params)
+        , threshold(threshold_)
+        , input_data_type(this->argument_types[0])
+    {
+    }
+
+    void create(AggregateDataPtr place) const override
+    {
+        Base::create(place);
+        this->data(place).threshold = threshold;
+    }
+
+    String getName() const override { return "groupArraySorted"; }
+
+    DataTypePtr getReturnType() const override { return std::make_shared<DataTypeArray>(input_data_type); }
+
+    bool allocatesMemoryInArena() const override
+    {
+        if constexpr (std::is_same_v<TColumnA, StringRef>)
+            return true;
+        else
+            return false;
+    }
+
+    void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena * arena) const override
+    {
+        State & data = this->data(place);
+        if constexpr (use_column_b)
+        {
+            data.add(
+                readItem<TColumnA, is_plain_a>(columns[0], arena, row_num), readItem<TColumnB, is_plain_b>(columns[1], arena, row_num));
+        }
+        else
+        {
+            data.add(readItem<TColumnA, is_plain_a>(columns[0], arena, row_num));
+        }
+    }
+
+    template <typename TColumn, bool is_plain, typename TFunc>
+    void
+    forFirstRows(size_t batch_size, const IColumn ** columns, size_t data_column, Arena * arena, ssize_t if_argument_pos, TFunc func) const
+    {
+        const TColumn * values = nullptr;
+        std::unique_ptr<std::vector<TColumn>> values_vector;
+        std::vector<size_t> best_rows(threshold);
+
+        if constexpr (std::is_same_v<TColumn, StringRef>)
+        {
+            values_vector.reset(new std::vector<TColumn>(batch_size));
+            for (size_t i = 0; i < batch_size; i++)
+                (*values_vector)[i] = readItem<TColumn, is_plain>(columns[data_column], arena, i);
+            values = (*values_vector).data();
+        }
+        else
+        {
+            const auto & column = assert_cast<const ColumnVector<TColumn> &>(*columns[data_column]);
+            values = column.getData().data();
+        }
+
+        const UInt8 * filter = nullptr;
+        StringRef refFilter;
+
+        if (if_argument_pos >= 0)
+        {
+            refFilter = columns[if_argument_pos]->getRawData();
+            filter = reinterpret_cast<const UInt8 *>(refFilter.data);
+        }
+
+        size_t num_elements = getFirstNElements(values, batch_size, threshold, best_rows.data(), filter);
+        for (size_t i = 0; i < num_elements; i++)
+        {
+            func(best_rows[i], values);
+        }
+    }
+
+    void addBatchSinglePlace(
+        size_t batch_size, AggregateDataPtr place, const IColumn ** columns, Arena * arena, ssize_t if_argument_pos) const override
+    {
+        State & data = this->data(place);
+
+        if constexpr (use_column_b)
+        {
+            forFirstRows<TColumnB, is_plain_b>(
+                batch_size, columns, 1, arena, if_argument_pos, [columns, &arena, &data](size_t row, const TColumnB * values)
+                {
+                    data.add(readItem<TColumnA, is_plain_a>(columns[0], arena, row), values[row]);
+                });
+        }
+        else
+        {
+            forFirstRows<TColumnA, is_plain_a>(
+                batch_size, columns, 0, arena, if_argument_pos, [&data](size_t row, const TColumnA * values)
+                {
+                    data.add(values[row]);
+                });
+        }
+    }
+
+    void merge(AggregateDataPtr __restrict place, ConstAggregateDataPtr rhs, Arena *) const override
+    {
+        this->data(place).merge(this->data(rhs));
+    }
+
+    void serialize(ConstAggregateDataPtr __restrict place, WriteBuffer & buf, std::optional<size_t> /* version */) const override
+    {
+        this->data(place).serialize(buf);
+    }
+
+    void
+    deserialize(AggregateDataPtr __restrict place, ReadBuffer & buf, std::optional<size_t> /* version  */, Arena * arena) const override
+    {
+        this->data(place).deserialize(buf, arena);
+    }
+
+    void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena * /*arena*/) const override
+    {
+        ColumnArray & arr_to = assert_cast<ColumnArray &>(to);
+        ColumnArray::Offsets & offsets_to = arr_to.getOffsets();
+
+        auto & values = this->data(place).values;
+        offsets_to.push_back(offsets_to.back() + values.size());
+
+        IColumn & data_to = arr_to.getData();
+        for (auto value : values)
+        {
+            if constexpr (std::is_same_v<TColumnA, StringRef>)
+            {
+                auto str = State::itemValue(value);
+                if constexpr (is_plain_a)
+                {
+                    data_to.insertData(str.data, str.size);
+                }
+                else
+                {
+                    data_to.deserializeAndInsertFromArena(str.data);
+                }
+            }
+            else
+            {
+                data_to.insert(State::itemValue(value));
+            }
+        }
+    }
+};
+}

src/AggregateFunctions/AggregateFunctionGroupArraySortedData.h

@@ -0,0 +1,162 @@
+#pragma once
+
+#include <IO/ReadBuffer.h>
+#include <IO/ReadHelpers.h>
+#include <IO/VarInt.h>
+#include <IO/WriteBuffer.h>
+#include <IO/WriteHelpers.h>
+
+
+static inline constexpr UInt64 GROUP_SORTED_DEFAULT_THRESHOLD = 0xFFFFFF;
+
+namespace DB
+{
+template <typename T>
+static void writeOneItem(WriteBuffer & buf, T item)
+{
+    if constexpr (std::numeric_limits<T>::is_signed)
+    {
+        writeVarInt(item, buf);
+    }
+    else
+    {
+        writeVarUInt(item, buf);
+    }
+}
+
+static void writeOneItem(WriteBuffer & buf, const StringRef & item)
+{
+    writeBinary(item, buf);
+}
+
+template <typename T>
+static void readOneItem(ReadBuffer & buf, Arena * /*arena*/, T & item)
+{
+    if constexpr (std::numeric_limits<T>::is_signed)
+    {
+        DB::Int64 val;
+        readVarT(val, buf);
+        item = val;
+    }
+    else
+    {
+        DB::UInt64 val;
+        readVarT(val, buf);
+        item = val;
+    }
+}
+
+static void readOneItem(ReadBuffer & buf, Arena * arena, StringRef & item)
+{
+    item = readStringBinaryInto(*arena, buf);
+}
+
+template <typename Storage>
+struct AggregateFunctionGroupArraySortedDataBase
+{
+    typedef typename Storage::value_type ValueType;
+    AggregateFunctionGroupArraySortedDataBase(UInt64 threshold_ = GROUP_SORTED_DEFAULT_THRESHOLD) : threshold(threshold_) { }
+
+    virtual ~AggregateFunctionGroupArraySortedDataBase() { }
+    inline void narrowDown()
+    {
+        while (values.size() > threshold)
+            values.erase(--values.end());
+    }
+
+    void merge(const AggregateFunctionGroupArraySortedDataBase & other)
+    {
+        values.merge(Storage(other.values));
+        narrowDown();
+    }
+
+    void serialize(WriteBuffer & buf) const
+    {
+        writeOneItem(buf, UInt64(values.size()));
+        for (auto value : values)
+        {
+            serializeItem(buf, value);
+        }
+    }
+
+    virtual void serializeItem(WriteBuffer & buf, ValueType & val) const = 0;
+    virtual ValueType deserializeItem(ReadBuffer & buf, Arena * arena) const = 0;
+
+    void deserialize(ReadBuffer & buf, Arena * arena)
+    {
+        values.clear();
+        UInt64 length;
+        readOneItem(buf, nullptr, length);
+
+        while (length--)
+        {
+            values.insert(deserializeItem(buf, arena));
+        }
+
+        narrowDown();
+    }
+
+    UInt64 threshold;
+    Storage values;
+};
+
+template <typename T, bool expr_sorted, typename TIndex>
+struct AggregateFunctionGroupArraySortedData
+{
+};
+
+template <typename T, typename TIndex>
+struct AggregateFunctionGroupArraySortedData<T, true, TIndex> : public AggregateFunctionGroupArraySortedDataBase<std::multimap<TIndex, T>>
+{
+    using Base = AggregateFunctionGroupArraySortedDataBase<std::multimap<TIndex, T>>;
+    using Base::Base;
+
+    void add(T item, TIndex weight)
+    {
+        Base::values.insert({weight, item});
+        Base::narrowDown();
+    }
+
+    void serializeItem(WriteBuffer & buf, typename Base::ValueType & value) const override
+    {
+        writeOneItem(buf, value.first);
+        writeOneItem(buf, value.second);
+    }
+
+    virtual typename Base::ValueType deserializeItem(ReadBuffer & buf, Arena * arena) const override
+    {
+        TIndex first;
+        T second;
+        readOneItem(buf, arena, first);
+        readOneItem(buf, arena, second);
+
+        return {first, second};
+    }
+
+    static T itemValue(typename Base::ValueType & value) { return value.second; }
+};
+
+template <typename T, typename TIndex>
+struct AggregateFunctionGroupArraySortedData<T, false, TIndex> : public AggregateFunctionGroupArraySortedDataBase<std::multiset<T>>
+{
+    using Base = AggregateFunctionGroupArraySortedDataBase<std::multiset<T>>;
+    using Base::Base;
+
+    void add(T item)
+    {
+        Base::values.insert(item);
+        Base::narrowDown();
+    }
+
+    void serializeItem(WriteBuffer & buf, typename Base::ValueType & value) const override { writeOneItem(buf, value); }
+
+    typename Base::ValueType deserializeItem(ReadBuffer & buf, Arena * arena) const override
+    {
+        T value;
+        readOneItem(buf, arena, value);
+        return value;
+    }
+
+    static T itemValue(typename Base::ValueType & value) { return value; }
+};
+}

src/AggregateFunctions/registerAggregateFunctions.cpp

@@ -59,6 +59,7 @@ void registerAggregateFunctionNothing(AggregateFunctionFactory &);
 void registerAggregateFunctionExponentialMovingAverage(AggregateFunctionFactory &);
 void registerAggregateFunctionSparkbar(AggregateFunctionFactory &);
 void registerAggregateFunctionIntervalLengthSum(AggregateFunctionFactory &);
+void registerAggregateFunctionGroupArraySorted(AggregateFunctionFactory & factory);
 
 class AggregateFunctionCombinatorFactory;
 void registerAggregateFunctionCombinatorIf(AggregateFunctionCombinatorFactory &);
@@ -130,6 +131,7 @@ void registerAggregateFunctions()
         registerAggregateFunctionIntervalLengthSum(factory);
         registerAggregateFunctionExponentialMovingAverage(factory);
         registerAggregateFunctionSparkbar(factory);
+        registerAggregateFunctionGroupArraySorted(factory);
 
         registerWindowFunctions(factory);
     }

src/Common/Exception.h

@@ -35,10 +35,10 @@ public:
     {}
 
     // Format message with fmt::format, like the logging functions.
-    template <typename ...Args>
-    Exception(int code, const std::string & fmt, Args&&... args)
-        : Exception(fmt::format(fmt::runtime(fmt), std::forward<Args>(args)...), code)
-    {}
+    template <typename... Args>
+    Exception(int code, fmt::format_string<Args...> fmt, Args &&... args) : Exception(fmt::format(fmt, std::forward<Args>(args)...), code)
+    {
+    }
 
     struct CreateFromPocoTag {};
     struct CreateFromSTDTag {};
@@ -52,10 +52,10 @@ public:
     const char * what() const throw() override { return message().data(); }
 
     /// Add something to the existing message.
-    template <typename ...Args>
-    void addMessage(const std::string& format, Args&&... args)
+    template <typename... Args>
+    void addMessage(fmt::format_string<Args...> format, Args &&... args)
     {
-        extendedMessage(fmt::format(fmt::runtime(format), std::forward<Args>(args)...));
+        extendedMessage(fmt::format(format, std::forward<Args>(args)...));
     }
 
     void addMessage(const std::string& message)
@@ -117,10 +117,10 @@ public:
     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)
-        : Exception(fmt::format(fmt::runtime(fmt), std::forward<Args>(args)...), code)
-    {}
+    template <typename... Args>
+    ParsingException(int code, fmt::format_string<Args...> fmt, Args &&... args) : Exception(code, fmt, std::forward<Args>(args)...)
+    {
+    }
 
 
     std::string displayText() const

src/Common/HashTable/FixedHashTable.h

@@ -67,6 +67,9 @@ struct FixedHashTableCalculatedSize
 {
     size_t getSize(const Cell * buf, const typename Cell::State & state, size_t num_cells) const
     {
+        if (!buf)
+            return 0;
+
         size_t res = 0;
         for (const Cell * end = buf + num_cells; buf != end; ++buf)
             if (!buf->isZero(state))
@@ -76,6 +79,9 @@ struct FixedHashTableCalculatedSize
 
     bool isEmpty(const Cell * buf, const typename Cell::State & state, size_t num_cells) const
     {
+        if (!buf)
+            return true;
+
         for (const Cell * end = buf + num_cells; buf != end; ++buf)
             if (!buf->isZero(state))
                 return false;

src/Common/HashTable/TwoLevelHashTable.h

@@ -94,6 +94,12 @@ public:
 
     TwoLevelHashTable() = default;
 
+    explicit TwoLevelHashTable(size_t size_hint)
+    {
+        for (auto & impl : impls)
+            impl.reserve(size_hint / NUM_BUCKETS);
+    }
+
     /// Copy the data from another (normal) hash table. It should have the same hash function.
     template <typename Source>
     explicit TwoLevelHashTable(const Source & src)

src/Common/ProfileEvents.cpp

@@ -285,6 +285,9 @@
     \
     M(MainConfigLoads, "Number of times the main configuration was reloaded.") \
     \
+    M(AggregationPreallocatedElementsInHashTables, "How many elements were preallocated in hash tables for aggregation.") \
+    M(AggregationHashTablesInitializedAsTwoLevel, "How many hash tables were inited as two-level for aggregation.") \
+    \
     M(MergeTreeMetadataCacheGet, "Number of rocksdb reads(used for merge tree metadata cache)") \
     M(MergeTreeMetadataCachePut, "Number of rocksdb puts(used for merge tree metadata cache)") \
     M(MergeTreeMetadataCacheDelete, "Number of rocksdb deletes(used for merge tree metadata cache)") \

src/Common/RangeGenerator.h

@@ -0,0 +1,46 @@
+#pragma once
+
+#include <optional>
+#include <cmath>
+
+namespace DB
+{
+
+class RangeGenerator
+{
+public:
+    explicit RangeGenerator(size_t total_size_, size_t range_step_, size_t range_start = 0)
+        : from(range_start), range_step(range_step_), total_size(total_size_)
+    {
+    }
+
+    size_t totalRanges() const { return static_cast<size_t>(round(static_cast<float>(total_size - from) / range_step)); }
+
+    using Range = std::pair<size_t, size_t>;
+
+    // return upper exclusive range of values, i.e. [from_range, to_range>
+    std::optional<Range> nextRange()
+    {
+        if (from >= total_size)
+        {
+            return std::nullopt;
+        }
+
+        auto to = from + range_step;
+        if (to >= total_size)
+        {
+            to = total_size;
+        }
+
+        Range range{from, to};
+        from = to;
+        return range;
+    }
+
+private:
+    size_t from;
+    size_t range_step;
+    size_t total_size;
+};
+
+}

src/Common/format.h

@@ -0,0 +1,178 @@
+#pragma once
+
+#include <base/types.h>
+#include <Common/Exception.h>
+#include <Common/PODArray.h>
+#include <Common/StringUtils/StringUtils.h>
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int BAD_ARGUMENTS;
+}
+
+namespace Format
+{
+    using IndexPositions = PODArrayWithStackMemory<UInt64, 64>;
+
+    static inline void parseNumber(const String & description, UInt64 l, UInt64 r, UInt64 & res, UInt64 argument_number)
+    {
+        res = 0;
+        for (UInt64 pos = l; pos < r; ++pos)
+        {
+            if (!isNumericASCII(description[pos]))
+                throw Exception("Not a number in curly braces at position " + std::to_string(pos), ErrorCodes::BAD_ARGUMENTS);
+            res = res * 10 + description[pos] - '0';
+            if (res >= argument_number)
+                throw Exception(
+                    "Too big number for arguments, must be at most " + std::to_string(argument_number - 1), ErrorCodes::BAD_ARGUMENTS);
+        }
+    }
+
+    static inline void init(
+        const String & pattern,
+        size_t argument_number,
+        const std::vector<std::optional<String>> & constant_strings,
+        IndexPositions & index_positions,
+        std::vector<String> & substrings)
+    {
+        /// Is current position after open curly brace.
+        bool is_open_curly = false;
+        /// The position of last open token.
+        size_t last_open = -1;
+
+        /// Is formatting in a plain {} token.
+        std::optional<bool> is_plain_numbering;
+        UInt64 index_if_plain = 0;
+
+        /// Left position of adding substrings, just to the closed brace position or the start of the string.
+        /// Invariant --- the start of substring is in this position.
+        size_t start_pos = 0;
+
+        /// A flag to decide whether we should glue the constant strings.
+        bool glue_to_next = false;
+
+        /// Handling double braces (escaping).
+        auto double_brace_removal = [](String & str)
+        {
+            size_t i = 0;
+            bool should_delete = true;
+            str.erase(
+                std::remove_if(
+                    str.begin(),
+                    str.end(),
+                    [&i, &should_delete, &str](char)
+                    {
+                        bool is_double_brace = (str[i] == '{' && str[i + 1] == '{') || (str[i] == '}' && str[i + 1] == '}');
+                        ++i;
+                        if (is_double_brace && should_delete)
+                        {
+                            should_delete = false;
+                            return true;
+                        }
+                        should_delete = true;
+                        return false;
+                    }),
+                str.end());
+        };
+
+        index_positions.emplace_back();
+
+        for (size_t i = 0; i < pattern.size(); ++i)
+        {
+            if (pattern[i] == '{')
+            {
+                /// Escaping handling
+                /// It is safe to access because of null termination
+                if (pattern[i + 1] == '{')
+                {
+                    ++i;
+                    continue;
+                }
+
+                if (is_open_curly)
+                    throw Exception("Two open curly braces without close one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
+
+                String to_add = String(pattern.data() + start_pos, i - start_pos);
+                double_brace_removal(to_add);
+                if (!glue_to_next)
+                    substrings.emplace_back(to_add);
+                else
+                    substrings.back() += to_add;
+
+                glue_to_next = false;
+
+                is_open_curly = true;
+                last_open = i + 1;
+            }
+            else if (pattern[i] == '}')
+            {
+                if (pattern[i + 1] == '}')
+                {
+                    ++i;
+                    continue;
+                }
+
+                if (!is_open_curly)
+                    throw Exception("Closed curly brace without open one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
+
+                is_open_curly = false;
+
+                if (last_open == i)
+                {
+                    if (is_plain_numbering && !*is_plain_numbering)
+                        throw Exception(
+                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
+                    is_plain_numbering = true;
+                    if (index_if_plain >= argument_number)
+                        throw Exception("Argument is too big for formatting", ErrorCodes::BAD_ARGUMENTS);
+                    index_positions.back() = index_if_plain++;
+                }
+                else
+                {
+                    if (is_plain_numbering && *is_plain_numbering)
+                        throw Exception(
+                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
+                    is_plain_numbering = false;
+
+                    UInt64 arg;
+                    parseNumber(pattern, last_open, i, arg, argument_number);
+
+                    if (arg >= argument_number)
+                        throw Exception(
+                            "Argument is too big for formatting. Note that indexing starts from zero", ErrorCodes::BAD_ARGUMENTS);
+
+                    index_positions.back() = arg;
+                }
+
+                if (!constant_strings.empty() && constant_strings[index_positions.back()])
+                {
+                    /// The next string should be glued to last `A {} C`.format('B') -> `A B C`.
+                    glue_to_next = true;
+                    substrings.back() += *constant_strings[index_positions.back()];
+                }
+                else
+                    index_positions.emplace_back(); /// Otherwise we commit arg number and proceed.
+
+                start_pos = i + 1;
+            }
+        }
+
+        if (is_open_curly)
+            throw Exception("Last open curly brace is not closed", ErrorCodes::BAD_ARGUMENTS);
+
+        String to_add = String(pattern.data() + start_pos, pattern.size() - start_pos);
+        double_brace_removal(to_add);
+
+        if (!glue_to_next)
+            substrings.emplace_back(to_add);
+        else
+            substrings.back() += to_add;
+
+        index_positions.pop_back();
+    }
+}
+
+}

src/Core/Settings.h

@@ -500,6 +500,10 @@ class IColumn;
     M(Bool, optimize_rewrite_sum_if_to_count_if, true, "Rewrite sumIf() and sum(if()) function countIf() function when logically equivalent", 0) \
     M(UInt64, insert_shard_id, 0, "If non zero, when insert into a distributed table, the data will be inserted into the shard `insert_shard_id` synchronously. Possible values range from 1 to `shards_number` of corresponding distributed table", 0) \
     \
+    M(Bool, collect_hash_table_stats_during_aggregation, true, "Enable collecting hash table statistics to optimize memory allocation", 0) \
+    M(UInt64, max_entries_for_hash_table_stats, 10'000, "How many entries hash table statistics collected during aggregation is allowed to have", 0) \
+    M(UInt64, max_size_to_preallocate_for_aggregation, 10'000'000, "For how many elements it is allowed to preallocate space in all hash tables in total before aggregation", 0) \
+    \
     /** Experimental feature for moving data between shards. */ \
     \
     M(Bool, allow_experimental_query_deduplication, false, "Experimental data deduplication for SELECT queries based on part UUIDs", 0) \

src/Databases/DatabaseOnDisk.cpp

@@ -6,6 +6,7 @@
 #include <IO/WriteHelpers.h>
 #include <Interpreters/Context.h>
 #include <Interpreters/InterpreterCreateQuery.h>
+#include <Interpreters/ApplyWithSubqueryVisitor.h>
 #include <Parsers/ASTCreateQuery.h>
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ParserCreateQuery.h>
@@ -55,6 +56,9 @@ std::pair<String, StoragePtr> createTableFromAST(
     ast_create_query.attach = true;
     ast_create_query.setDatabase(database_name);
 
+    if (ast_create_query.select && ast_create_query.isView())
+        ApplyWithSubqueryVisitor().visit(*ast_create_query.select);
+
     if (ast_create_query.as_table_function)
     {
         const auto & factory = TableFunctionFactory::instance();

src/Databases/DatabaseReplicatedWorker.cpp

@@ -179,8 +179,12 @@ String DatabaseReplicatedDDLWorker::tryEnqueueAndExecuteEntry(DDLLogEntry & entr
 
     if (!task->was_executed)
     {
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Entry {} was executed, but was not committed: code {}: {}",
-                        task->execution_status.code, task->execution_status.message);
+        throw Exception(
+            ErrorCodes::LOGICAL_ERROR,
+            "Entry {} was executed, but was not committed: code {}: {}",
+            task->entry_name,
+            task->execution_status.code,
+            task->execution_status.message);
     }
 
     try_node->setAlreadyRemoved();

src/Dictionaries/XDBCDictionarySource.cpp

@@ -50,7 +50,7 @@ namespace
         {
             if (!qualified_name.database.empty())
                 throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
-                    "Dictionary source of type {} specifies a schema but schema is not supported by {}-driver",
+                    "Dictionary source specifies a schema but schema is not supported by {}-driver",
                     bridge_.getName());
         }
 

src/Disks/IO/CachedReadBufferFromRemoteFS.cpp

@@ -392,8 +392,13 @@ void CachedReadBufferFromRemoteFS::predownload(FileSegmentPtr & file_segment)
                 if (bytes_to_predownload)
                     throw Exception(
                         ErrorCodes::LOGICAL_ERROR,
-                        "Failed to predownload remaining {} bytes. Current file segment: {}, current download offset: {}, expected: {}, eof: {}",
-                        file_segment->range().toString(), file_segment->getDownloadOffset(), file_offset_of_buffer_end, implementation_buffer->eof());
+                        "Failed to predownload remaining {} bytes. Current file segment: {}, current download offset: {}, expected: {}, "
+                        "eof: {}",
+                        bytes_to_predownload,
+                        file_segment->range().toString(),
+                        file_segment->getDownloadOffset(),
+                        file_offset_of_buffer_end,
+                        implementation_buffer->eof());
 
                 auto result = implementation_buffer->hasPendingData();
 

src/Disks/IO/ReadBufferFromRemoteFSGather.cpp

@@ -44,7 +44,7 @@ SeekableReadBufferPtr ReadBufferFromS3Gather::createImplementationBuffer(const S
     {
         return std::make_unique<ReadBufferFromS3>(
             client_ptr, bucket, fs::path(metadata.remote_fs_root_path) / path, max_single_read_retries,
-            settings, /* use_external_buffer */true, read_until_position, /* restricted_seek */true);
+            settings, /* use_external_buffer */true, /* offset */ 0, read_until_position, /* restricted_seek */true);
     };
 
     if (with_cache)

src/Formats/FormatSchemaInfo.cpp

@@ -85,9 +85,12 @@ FormatSchemaInfo::FormatSchemaInfo(const String & format_schema, const String &
     else if (path.has_parent_path() && !fs::weakly_canonical(default_schema_directory_path / path).string().starts_with(fs::weakly_canonical(default_schema_directory_path).string()))
     {
         if (is_server)
-            throw Exception(ErrorCodes::BAD_ARGUMENTS,
-                            "Path in the 'format_schema' setting shouldn't go outside the 'format_schema_path' directory: {} ({} not in {})",
-                            path.string());
+            throw Exception(
+                ErrorCodes::BAD_ARGUMENTS,
+                "Path in the 'format_schema' setting shouldn't go outside the 'format_schema_path' directory: {} ({} not in {})",
+                default_schema_directory(),
+                path.string(),
+                default_schema_directory());
         path = default_schema_directory_path / path;
         schema_path = path.filename();
         schema_directory = path.parent_path() / "";

src/Functions/FunctionsConversion.h

@@ -887,7 +887,7 @@ struct ConvertImplGenericToString
         const IColumn & col_from = *col_with_type_and_name.column;
 
         size_t size = col_from.size();
-        auto col_to = result_type->createColumn();
+        auto col_to = removeNullable(result_type)->createColumn();
 
         {
             ColumnStringHelpers::WriteHelper write_helper(

src/Functions/FunctionsStringArray.h

@@ -259,7 +259,7 @@ public:
             throw Exception(
                 ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
                 "Function '{}' needs at least 2 arguments, at most 3 arguments; passed {}.",
-                arguments.size());
+                name, arguments.size());
 
         if (!isString(arguments[0]))
             throw Exception("Illegal type " + arguments[0]->getName() + " of first argument of function " + getName() + ". Must be String.",

src/Functions/IFunction.cpp

@@ -181,9 +181,12 @@ ColumnPtr IExecutableFunction::defaultImplementationForNulls(
         // Default implementation for nulls returns null result for null arguments,
         // so the result type must be nullable.
         if (!result_type->isNullable())
-            throw Exception(ErrorCodes::LOGICAL_ERROR,
-                            "Function {} with Null argument and default implementation for Nulls "
-                            "is expected to return Nullable result, got {}", result_type->getName());
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR,
+                "Function {} with Null argument and default implementation for Nulls "
+                "is expected to return Nullable result, got {}",
+                getName(),
+                result_type->getName());
 
         return result_type->createColumnConstWithDefaultValue(input_rows_count);
     }

src/Functions/castOrDefault.cpp

@@ -231,7 +231,7 @@ private:
             {
                 throw Exception(ErrorCodes::BAD_ARGUMENTS,
                     "Function {} decimal scale should have native UInt type. Actual {}",
-                    scale_argument.type->getName());
+                    getName(), scale_argument.type->getName());
             }
 
             scale = arguments[additional_argument_index].column->getUInt(0);

src/Functions/concat.cpp

@@ -52,23 +52,21 @@ public:
     {
         if (arguments.size() < 2)
             throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at least 2.",
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
-
-        if (arguments.size() > FormatImpl::argument_threshold)
-            throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at most " + std::to_string(FormatImpl::argument_threshold),
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
+                "Number of arguments for function {} doesn't match: passed {}, should be at least 2",
+                getName(),
+                arguments.size());
 
         for (const auto arg_idx : collections::range(0, arguments.size()))
         {
             const auto * arg = arguments[arg_idx].get();
             if (!isStringOrFixedString(arg))
-                throw Exception{"Illegal type " + arg->getName() + " of argument " + std::to_string(arg_idx + 1) + " of function "
-                                    + getName(),
-                                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT};
+                throw Exception(
+                    ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                    "Illegal type {} of argument {} of function {}",
+                    arg->getName(),
+                    arg_idx + 1,
+                    getName());
         }
 
         return std::make_shared<DataTypeString>();
@@ -125,7 +123,7 @@ private:
         std::vector<const ColumnString::Chars *> data(num_arguments);
         std::vector<const ColumnString::Offsets *> offsets(num_arguments);
         std::vector<size_t> fixed_string_sizes(num_arguments);
-        std::vector<String> constant_strings(num_arguments);
+        std::vector<std::optional<String>> constant_strings(num_arguments);
         bool has_column_string = false;
         bool has_column_fixed_string = false;
         for (size_t i = 0; i < num_arguments; ++i)

src/Functions/dateName.cpp

@@ -112,7 +112,7 @@ public:
             || (res = executeType<DataTypeDateTime64>(arguments, result_type))))
             throw Exception(
                 ErrorCodes::ILLEGAL_COLUMN,
-                "Illegal column {} of function {], must be Date or DateTime.",
+                "Illegal column {} of function {}, must be Date or DateTime.",
                 arguments[1].column->getName(),
                 getName());
 

src/Functions/formatString.cpp

@@ -45,25 +45,23 @@ public:
 
     DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
     {
-        if (arguments.empty())
+        if (arguments.size() < 2)
             throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at least 1",
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
-
-        if (arguments.size() > FormatImpl::argument_threshold)
-            throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at most " + std::to_string(FormatImpl::argument_threshold),
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
+                "Number of arguments for function {} doesn't match: passed {}, should be at least 2",
+                getName(),
+                arguments.size());
 
         for (const auto arg_idx : collections::range(0, arguments.size()))
         {
             const auto * arg = arguments[arg_idx].get();
             if (!isStringOrFixedString(arg))
                 throw Exception(
-                    "Illegal type " + arg->getName() + " of argument " + std::to_string(arg_idx + 1) + " of function " + getName(),
-                    ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+                    ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                    "Illegal type {} of argument {} of function {}",
+                    arg->getName(),
+                    arg_idx + 1,
+                    getName());
         }
 
         return std::make_shared<DataTypeString>();
@@ -84,7 +82,7 @@ public:
         std::vector<const ColumnString::Chars *> data(arguments.size() - 1);
         std::vector<const ColumnString::Offsets *> offsets(arguments.size() - 1);
         std::vector<size_t> fixed_string_sizes(arguments.size() - 1);
-        std::vector<String> constant_strings(arguments.size() - 1);
+        std::vector<std::optional<String>> constant_strings(arguments.size() - 1);
 
         bool has_column_string = false;
         bool has_column_fixed_string = false;

src/Functions/formatString.h

@@ -4,8 +4,10 @@
 #include <base/types.h>
 #include <Common/Exception.h>
 #include <Common/StringUtils/StringUtils.h>
+#include <Common/format.h>
 #include <Common/memcpySmall.h>
 
+
 #include <algorithm>
 #include <optional>
 #include <string>
@@ -15,15 +17,9 @@
 
 namespace DB
 {
-namespace ErrorCodes
-{
-    extern const int BAD_ARGUMENTS;
-}
 
 struct FormatImpl
 {
-    static constexpr size_t small_argument_threshold = 1024;
-    static constexpr size_t argument_threshold = std::numeric_limits<UInt32>::max();
     static constexpr size_t right_padding = 15;
 
     template <typename... Args>
@@ -39,165 +35,10 @@ struct FormatImpl
             format<false, false>(std::forward<Args>(args)...);
     }
 
-    static void parseNumber(const String & description, UInt64 l, UInt64 r, UInt64 & res)
-    {
-        res = 0;
-        for (UInt64 pos = l; pos < r; ++pos)
-        {
-            if (!isNumericASCII(description[pos]))
-                throw Exception("Not a number in curly braces at position " + std::to_string(pos), ErrorCodes::BAD_ARGUMENTS);
-            res = res * 10 + description[pos] - '0';
-            if (res >= argument_threshold)
-                throw Exception(
-                    "Too big number for arguments, must be at most " + std::to_string(argument_threshold), ErrorCodes::BAD_ARGUMENTS);
-        }
-    }
-
-    static inline void init(
-        const String & pattern,
-        const std::vector<const ColumnString::Chars *> & data,
-        size_t argument_number,
-        const std::vector<String> & constant_strings,
-        UInt64 * index_positions_ptr,
-        std::vector<String> & substrings)
-    {
-        /// Is current position after open curly brace.
-        bool is_open_curly = false;
-        /// The position of last open token.
-        size_t last_open = -1;
-
-        /// Is formatting in a plain {} token.
-        std::optional<bool> is_plain_numbering;
-        UInt64 index_if_plain = 0;
-
-        /// Left position of adding substrings, just to the closed brace position or the start of the string.
-        /// Invariant --- the start of substring is in this position.
-        size_t start_pos = 0;
-
-        /// A flag to decide whether we should glue the constant strings.
-        bool glue_to_next = false;
-
-        /// Handling double braces (escaping).
-        auto double_brace_removal = [](String & str)
-        {
-            size_t i = 0;
-            bool should_delete = true;
-            str.erase(
-                std::remove_if(
-                    str.begin(),
-                    str.end(),
-                    [&i, &should_delete, &str](char)
-                    {
-                        bool is_double_brace = (str[i] == '{' && str[i + 1] == '{') || (str[i] == '}' && str[i + 1] == '}');
-                        ++i;
-                        if (is_double_brace && should_delete)
-                        {
-                            should_delete = false;
-                            return true;
-                        }
-                        should_delete = true;
-                        return false;
-                    }),
-                str.end());
-        };
-
-        for (size_t i = 0; i < pattern.size(); ++i)
-        {
-            if (pattern[i] == '{')
-            {
-                /// Escaping handling
-                /// It is safe to access because of null termination
-                if (pattern[i + 1] == '{')
-                {
-                    ++i;
-                    continue;
-                }
-
-                if (is_open_curly)
-                    throw Exception("Two open curly braces without close one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
-
-                String to_add = String(pattern.data() + start_pos, i - start_pos);
-                double_brace_removal(to_add);
-                if (!glue_to_next)
-                    substrings.emplace_back(to_add);
-                else
-                    substrings.back() += to_add;
-
-                glue_to_next = false;
-
-                is_open_curly = true;
-                last_open = i + 1;
-            }
-            else if (pattern[i] == '}')
-            {
-                if (pattern[i + 1] == '}')
-                {
-                    ++i;
-                    continue;
-                }
-
-                if (!is_open_curly)
-                    throw Exception("Closed curly brace without open one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
-
-                is_open_curly = false;
-
-                if (last_open == i)
-                {
-                    if (is_plain_numbering && !*is_plain_numbering)
-                        throw Exception(
-                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
-                    is_plain_numbering = true;
-                    if (index_if_plain >= argument_number)
-                        throw Exception("Argument is too big for formatting", ErrorCodes::BAD_ARGUMENTS);
-                    *index_positions_ptr = index_if_plain++;
-                }
-                else
-                {
-                    if (is_plain_numbering && *is_plain_numbering)
-                        throw Exception(
-                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
-                    is_plain_numbering = false;
-
-                    UInt64 arg;
-                    parseNumber(pattern, last_open, i, arg);
-
-                    if (arg >= argument_number)
-                        throw Exception(
-                            "Argument is too big for formatting. Note that indexing starts from zero", ErrorCodes::BAD_ARGUMENTS);
-
-                    *index_positions_ptr = arg;
-                }
-
-                /// Constant string.
-                if (!data[*index_positions_ptr])
-                {
-                    /// The next string should be glued to last `A {} C`.format('B') -> `A B C`.
-                    glue_to_next = true;
-                    substrings.back() += constant_strings[*index_positions_ptr];
-                }
-                else
-                    ++index_positions_ptr; /// Otherwise we commit arg number and proceed.
-
-                start_pos = i + 1;
-            }
-        }
-
-        if (is_open_curly)
-            throw Exception("Last open curly brace is not closed", ErrorCodes::BAD_ARGUMENTS);
-
-        String to_add = String(pattern.data() + start_pos, pattern.size() - start_pos);
-        double_brace_removal(to_add);
-
-        if (!glue_to_next)
-            substrings.emplace_back(to_add);
-        else
-            substrings.back() += to_add;
-    }
-
     /// data for ColumnString and ColumnFixed. Nullptr means no data, it is const string.
     /// offsets for ColumnString, nullptr is an indicator that there is a fixed string rather than ColumnString.
     /// fixed_string_N for savings N to fixed strings.
-    /// constant_strings for constant strings. If data[i] is nullptr, than it is constant string.
+    /// constant_strings for constant strings. If data[i] is nullptr, it is constant string.
     /// res_data is result_data, res_offsets is offset result.
     /// input_rows_count is the number of rows processed.
     /// Precondition: data.size() == offsets.size() == fixed_string_N.size() == constant_strings.size().
@@ -207,29 +48,22 @@ struct FormatImpl
         const std::vector<const ColumnString::Chars *> & data,
         const std::vector<const ColumnString::Offsets *> & offsets,
         [[maybe_unused]] /* Because sometimes !has_column_fixed_string */ const std::vector<size_t> & fixed_string_N,
-        const std::vector<String> & constant_strings,
+        const std::vector<std::optional<String>> & constant_strings,
         ColumnString::Chars & res_data,
         ColumnString::Offsets & res_offsets,
         size_t input_rows_count)
     {
         const size_t argument_number = offsets.size();
 
-        UInt64 small_index_positions_buffer[small_argument_threshold];
-        /// The subsequent indexes of strings we should use. e.g `Hello world {1} {3} {1} {0}` this array will be filled with [1, 3, 1, 0, ... (garbage)] but without constant string indices.
-        UInt64 * index_positions = small_index_positions_buffer;
-
-        std::unique_ptr<UInt64[]> big_index_positions_buffer;
-        if (argument_number > small_argument_threshold)
-        {
-            big_index_positions_buffer.reset(new UInt64[argument_number]);
-            index_positions = big_index_positions_buffer.get();
-        }
+        /// The subsequent indexes of strings we should use. e.g `Hello world {1} {3} {1} {0}` this
+        /// array will be filled with [1, 3, 1, 0] but without constant string indices.
+        Format::IndexPositions index_positions;
 
         /// Vector of substrings of pattern that will be copied to the answer, not string view because of escaping and iterators invalidation.
         /// These are exactly what is between {} tokens, for `Hello {} world {}` we will have [`Hello `, ` world `, ``].
         std::vector<String> substrings;
 
-        init(pattern, data, argument_number, constant_strings, index_positions, substrings);
+        Format::init(pattern, argument_number, constant_strings, index_positions, substrings);
 
         UInt64 final_size = 0;
 
@@ -271,7 +105,7 @@ struct FormatImpl
                 for (size_t j = 1; j < substrings.size(); ++j)
                 {
                     UInt64 arg = index_positions[j - 1];
-                    auto offset_ptr = offsets[arg];
+                    const auto * offset_ptr = offsets[arg];
                     UInt64 arg_offset = 0;
                     UInt64 size = 0;
 

src/IO/ParallelReadBuffer.cpp

@@ -237,7 +237,7 @@ void ParallelReadBuffer::readerThreadFunction(ReadWorkerPtr read_worker)
         while (!emergency_stop && !read_worker->cancel)
         {
             if (!read_worker->reader->next())
-                throw Exception("Failed to read all the data from the reader", ErrorCodes::LOGICAL_ERROR);
+                throw Exception(ErrorCodes::LOGICAL_ERROR, "Failed to read all the data from the reader, missing {} bytes", read_worker->bytes_left);
 
             if (emergency_stop || read_worker->cancel)
                 break;

src/IO/ParallelReadBuffer.h

@@ -82,8 +82,8 @@ public:
         std::unique_ptr<ReadBufferFactory> reader_factory_,
         ThreadPool * pool,
         size_t max_working_readers,
-        WorkerSetup worker_setup = {},
-        WorkerCleanup worker_cleanup = {});
+        WorkerSetup worker_setup = [](ThreadStatus &){},
+        WorkerCleanup worker_cleanup = [](ThreadStatus &){});
 
     ~ParallelReadBuffer() override { finishAndWait(); }
 

src/IO/ReadBufferFromS3.cpp

@@ -1,4 +1,5 @@
 #include <Common/config.h>
+#include "IO/S3Common.h"
 
 #if USE_AWS_S3
 
@@ -42,6 +43,7 @@ ReadBufferFromS3::ReadBufferFromS3(
     UInt64 max_single_read_retries_,
     const ReadSettings & settings_,
     bool use_external_buffer_,
+    size_t offset_,
     size_t read_until_position_,
     bool restricted_seek_)
     : SeekableReadBufferWithSize(nullptr, 0)
@@ -49,9 +51,10 @@ ReadBufferFromS3::ReadBufferFromS3(
     , bucket(bucket_)
     , key(key_)
     , max_single_read_retries(max_single_read_retries_)
+    , offset(offset_)
+    , read_until_position(read_until_position_)
     , read_settings(settings_)
     , use_external_buffer(use_external_buffer_)
-    , read_until_position(read_until_position_)
     , restricted_seek(restricted_seek_)
 {
 }
@@ -210,13 +213,14 @@ std::optional<size_t> ReadBufferFromS3::getTotalSize()
     if (file_size)
         return file_size;
 
-    Aws::S3::Model::HeadObjectRequest request;
-    request.SetBucket(bucket);
-    request.SetKey(key);
+    auto object_size = S3::getObjectSize(client_ptr, bucket, key, false);
 
-    auto outcome = client_ptr->HeadObject(request);
-    auto head_result = outcome.GetResultWithOwnership();
-    file_size = head_result.GetContentLength();
+    if (!object_size)
+    {
+        return std::nullopt;
+    }
+
+    file_size = object_size;
     return file_size;
 }
 
@@ -234,6 +238,11 @@ void ReadBufferFromS3::setReadUntilPosition(size_t position)
     }
 }
 
+SeekableReadBuffer::Range ReadBufferFromS3::getRemainingReadRange() const
+{
+    return Range{.left = static_cast<size_t>(offset), .right = read_until_position ? std::optional{read_until_position - 1} : std::nullopt};
+}
+
 std::unique_ptr<ReadBuffer> ReadBufferFromS3::initialize()
 {
     Aws::S3::Model::GetObjectRequest req;
@@ -272,6 +281,36 @@ std::unique_ptr<ReadBuffer> ReadBufferFromS3::initialize()
         throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
 }
 
+SeekableReadBufferPtr ReadBufferS3Factory::getReader()
+{
+    const auto next_range = range_generator.nextRange();
+    if (!next_range)
+    {
+        return nullptr;
+    }
+
+    auto reader = std::make_shared<ReadBufferFromS3>(
+        client_ptr,
+        bucket,
+        key,
+        s3_max_single_read_retries,
+        read_settings,
+        false /*use_external_buffer*/,
+        next_range->first,
+        next_range->second);
+    return reader;
+}
+
+off_t ReadBufferS3Factory::seek(off_t off, [[maybe_unused]] int whence)
+{
+    range_generator = RangeGenerator{object_size, range_step, static_cast<size_t>(off)};
+    return off;
+}
+
+std::optional<size_t> ReadBufferS3Factory::getTotalSize()
+{
+    return object_size;
+}
 }
 
 #endif

src/IO/ReadBufferFromS3.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <Common/RangeGenerator.h>
 #include <Common/config.h>
 
 #if USE_AWS_S3
@@ -7,6 +8,7 @@
 #include <memory>
 
 #include <IO/HTTPCommon.h>
+#include <IO/ParallelReadBuffer.h>
 #include <IO/ReadBuffer.h>
 #include <IO/ReadSettings.h>
 #include <IO/SeekableReadBuffer.h>
@@ -30,7 +32,9 @@ private:
     String bucket;
     String key;
     UInt64 max_single_read_retries;
+
     off_t offset = 0;
+    off_t read_until_position = 0;
 
     Aws::S3::Model::GetObjectResult read_result;
     std::unique_ptr<ReadBuffer> impl;
@@ -45,6 +49,7 @@ public:
         UInt64 max_single_read_retries_,
         const ReadSettings & settings_,
         bool use_external_buffer = false,
+        size_t offset_ = 0,
         size_t read_until_position_ = 0,
         bool restricted_seek_ = false);
 
@@ -58,7 +63,7 @@ public:
 
     void setReadUntilPosition(size_t position) override;
 
-    Range getRemainingReadRange() const override { return Range{ .left = static_cast<size_t>(offset), .right = read_until_position }; }
+    Range getRemainingReadRange() const override;
 
     size_t getFileOffsetOfBufferEnd() const override { return offset; }
 
@@ -69,13 +74,55 @@ private:
 
     bool use_external_buffer;
 
-    off_t read_until_position = 0;
-
     /// There is different seek policy for disk seek and for non-disk seek
     /// (non-disk seek is applied for seekable input formats: orc, arrow, parquet).
     bool restricted_seek;
 };
 
+/// Creates separate ReadBufferFromS3 for sequence of ranges of particular object
+class ReadBufferS3Factory : public ParallelReadBuffer::ReadBufferFactory
+{
+public:
+    explicit ReadBufferS3Factory(
+        std::shared_ptr<Aws::S3::S3Client> client_ptr_,
+        const String & bucket_,
+        const String & key_,
+        size_t range_step_,
+        size_t object_size_,
+        UInt64 s3_max_single_read_retries_,
+        const ReadSettings & read_settings_)
+        : client_ptr(client_ptr_)
+        , bucket(bucket_)
+        , key(key_)
+        , read_settings(read_settings_)
+        , range_generator(object_size_, range_step_)
+        , range_step(range_step_)
+        , object_size(object_size_)
+        , s3_max_single_read_retries(s3_max_single_read_retries_)
+    {
+        assert(range_step > 0);
+        assert(range_step < object_size);
+    }
+
+    SeekableReadBufferPtr getReader() override;
+
+    off_t seek(off_t off, [[maybe_unused]] int whence) override;
+
+    std::optional<size_t> getTotalSize() override;
+
+private:
+    std::shared_ptr<Aws::S3::S3Client> client_ptr;
+    const String bucket;
+    const String key;
+    ReadSettings read_settings;
+
+    RangeGenerator range_generator;
+    size_t range_step;
+    size_t object_size;
+
+    UInt64 s3_max_single_read_retries;
+};
+
 }
 
 #endif

src/IO/ReadWriteBufferFromHTTP.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <functional>
+#include <Common/RangeGenerator.h>
 #include <IO/ConnectionTimeouts.h>
 #include <IO/HTTPCommon.h>
 #include <IO/ParallelReadBuffer.h>
@@ -635,43 +636,6 @@ public:
     void buildNewSession(const Poco::URI & uri) override { session = makeHTTPSession(uri, timeouts); }
 };
 
-class RangeGenerator
-{
-public:
-    explicit RangeGenerator(size_t total_size_, size_t range_step_, size_t range_start = 0)
-        : from(range_start), range_step(range_step_), total_size(total_size_)
-    {
-    }
-
-    size_t totalRanges() const { return static_cast<size_t>(round(static_cast<float>(total_size - from) / range_step)); }
-
-    using Range = std::pair<size_t, size_t>;
-
-    // return upper exclusive range of values, i.e. [from_range, to_range>
-    std::optional<Range> nextRange()
-    {
-        if (from >= total_size)
-        {
-            return std::nullopt;
-        }
-
-        auto to = from + range_step;
-        if (to >= total_size)
-        {
-            to = total_size;
-        }
-
-        Range range{from, to};
-        from = to;
-        return range;
-    }
-
-private:
-    size_t from;
-    size_t range_step;
-    size_t total_size;
-};
-
 class ReadWriteBufferFromHTTP : public detail::ReadWriteBufferFromHTTPBase<std::shared_ptr<UpdatableSession>>
 {
     using Parent = detail::ReadWriteBufferFromHTTPBase<std::shared_ptr<UpdatableSession>>;

src/IO/S3Common.cpp

@@ -24,6 +24,7 @@
 #    include <aws/core/utils/UUID.h>
 #    include <aws/core/http/HttpClientFactory.h>
 #    include <aws/s3/S3Client.h>
+#    include <aws/s3/model/HeadObjectRequest.h>  // Y_IGNORE
 
 #    include <IO/S3/PocoHTTPClientFactory.h>
 #    include <IO/S3/PocoHTTPClient.h>
@@ -682,6 +683,7 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int BAD_ARGUMENTS;
+    extern const int S3_ERROR;
 }
 
 namespace S3
@@ -839,6 +841,26 @@ namespace S3
             throw Exception(ErrorCodes::BAD_ARGUMENTS, "Bucket name length is out of bounds in virtual hosted style S3 URI:     {}{}",
                             quoteString(bucket), !uri.empty() ? " (" + uri.toString() + ")" : "");
     }
+
+    size_t getObjectSize(std::shared_ptr<Aws::S3::S3Client> client_ptr, const String & bucket, const String & key, bool throw_on_error)
+    {
+        Aws::S3::Model::HeadObjectRequest req;
+        req.SetBucket(bucket);
+        req.SetKey(key);
+
+        Aws::S3::Model::HeadObjectOutcome outcome = client_ptr->HeadObject(req);
+
+        if (outcome.IsSuccess())
+        {
+            auto read_result = outcome.GetResultWithOwnership();
+            return static_cast<size_t>(read_result.GetContentLength());
+        }
+        else if (throw_on_error)
+        {
+            throw DB::Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
+        }
+        return 0;
+    }
 }
 
 }

src/IO/S3Common.h

@@ -75,6 +75,8 @@ struct URI
     static void validateBucket(const String & bucket, const Poco::URI & uri);
 };
 
+size_t getObjectSize(std::shared_ptr<Aws::S3::S3Client> client_ptr, const String & bucket, const String & key, bool throw_on_error = true);
+
 }
 
 #endif

src/Interpreters/Aggregator.cpp

@@ -1,4 +1,6 @@
+#include <algorithm>
 #include <future>
+#include <numeric>
 #include <Poco/Util/Application.h>
 
 #include <base/sort.h>
@@ -15,6 +17,7 @@
 #include <IO/WriteBufferFromFile.h>
 #include <Compression/CompressedWriteBuffer.h>
 #include <Interpreters/Aggregator.h>
+#include <Common/LRUCache.h>
 #include <Common/MemoryTracker.h>
 #include <Common/CurrentThread.h>
 #include <Common/typeid_cast.h>
@@ -27,12 +30,236 @@
 #include <Interpreters/JIT/CompiledExpressionCache.h>
 #include <Core/ProtocolDefines.h>
 
+#include <Parsers/ASTSelectQuery.h>
 
 namespace ProfileEvents
 {
-    extern const Event ExternalAggregationWritePart;
-    extern const Event ExternalAggregationCompressedBytes;
-    extern const Event ExternalAggregationUncompressedBytes;
+extern const Event ExternalAggregationWritePart;
+extern const Event ExternalAggregationCompressedBytes;
+extern const Event ExternalAggregationUncompressedBytes;
+extern const Event AggregationPreallocatedElementsInHashTables;
+extern const Event AggregationHashTablesInitializedAsTwoLevel;
+}
+
+namespace
+{
+/** Collects observed HashMap-s sizes to avoid redundant intermediate resizes.
+  */
+class HashTablesStatistics
+{
+public:
+    struct Entry
+    {
+        size_t sum_of_sizes; // used to determine if it's better to convert aggregation to two-level from the beginning
+        size_t median_size; // roughly the size we're going to preallocate on each thread
+    };
+
+    using Cache = DB::LRUCache<UInt64, Entry>;
+    using CachePtr = std::shared_ptr<Cache>;
+    using Params = DB::Aggregator::Params::StatsCollectingParams;
+
+    /// Collection and use of the statistics should be enabled.
+    std::optional<Entry> getSizeHint(const Params & params)
+    {
+        if (!params.isCollectionAndUseEnabled())
+            throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Collection and use of the statistics should be enabled.");
+
+        std::lock_guard lock(mutex);
+        const auto cache = getHashTableStatsCache(params, lock);
+        if (const auto hint = cache->get(params.key))
+        {
+            LOG_DEBUG(
+                &Poco::Logger::get("Aggregator"),
+                "An entry for key={} found in cache: sum_of_sizes={}, median_size={}",
+                params.key,
+                hint->sum_of_sizes,
+                hint->median_size);
+            return *hint;
+        }
+        return std::nullopt;
+    }
+
+    /// Collection and use of the statistics should be enabled.
+    void update(size_t sum_of_sizes, size_t median_size, const Params & params)
+    {
+        if (!params.isCollectionAndUseEnabled())
+            throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Collection and use of the statistics should be enabled.");
+
+        std::lock_guard lock(mutex);
+        const auto cache = getHashTableStatsCache(params, lock);
+        const auto hint = cache->get(params.key);
+        // We'll maintain the maximum among all the observed values until the next prediction turns out to be too wrong.
+        if (!hint || sum_of_sizes < hint->sum_of_sizes / 2 || hint->sum_of_sizes < sum_of_sizes || median_size < hint->median_size / 2
+            || hint->median_size < median_size)
+        {
+            LOG_DEBUG(
+                &Poco::Logger::get("Aggregator"),
+                "Statistics updated for key={}: new sum_of_sizes={}, median_size={}",
+                params.key,
+                sum_of_sizes,
+                median_size);
+            cache->set(params.key, std::make_shared<Entry>(Entry{.sum_of_sizes = sum_of_sizes, .median_size = median_size}));
+        }
+    }
+
+    std::optional<DB::HashTablesCacheStatistics> getCacheStats() const
+    {
+        std::lock_guard lock(mutex);
+        if (hash_table_stats)
+        {
+            size_t hits = 0, misses = 0;
+            hash_table_stats->getStats(hits, misses);
+            return DB::HashTablesCacheStatistics{.entries = hash_table_stats->count(), .hits = hits, .misses = misses};
+        }
+        return std::nullopt;
+    }
+
+    static size_t calculateCacheKey(const DB::ASTPtr & select_query)
+    {
+        if (!select_query)
+            throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Query ptr cannot be null");
+
+        const auto & select = select_query->as<DB::ASTSelectQuery &>();
+
+        // It may happen in some corner cases like `select 1 as num group by num`.
+        if (!select.tables())
+            return 0;
+
+        SipHash hash;
+        hash.update(select.tables()->getTreeHash());
+        if (const auto where = select.where())
+            hash.update(where->getTreeHash());
+        if (const auto group_by = select.groupBy())
+            hash.update(group_by->getTreeHash());
+        return hash.get64();
+    }
+
+private:
+    CachePtr getHashTableStatsCache(const Params & params, const std::lock_guard<std::mutex> &)
+    {
+        if (!hash_table_stats || hash_table_stats->maxSize() != params.max_entries_for_hash_table_stats)
+            hash_table_stats = std::make_shared<Cache>(params.max_entries_for_hash_table_stats);
+        return hash_table_stats;
+    }
+
+    mutable std::mutex mutex;
+    CachePtr hash_table_stats;
+};
+
+HashTablesStatistics & getHashTablesStatistics()
+{
+    static HashTablesStatistics hash_tables_stats;
+    return hash_tables_stats;
+}
+
+bool worthConvertToTwoLevel(
+    size_t group_by_two_level_threshold, size_t result_size, size_t group_by_two_level_threshold_bytes, auto result_size_bytes)
+{
+    // params.group_by_two_level_threshold will be equal to 0 if we have only one thread to execute aggregation (refer to AggregatingStep::transformPipeline).
+    return (group_by_two_level_threshold && result_size >= group_by_two_level_threshold)
+        || (group_by_two_level_threshold_bytes && result_size_bytes >= static_cast<Int64>(group_by_two_level_threshold_bytes));
+}
+
+DB::AggregatedDataVariants::Type convertToTwoLevelTypeIfPossible(DB::AggregatedDataVariants::Type type)
+{
+    using Type = DB::AggregatedDataVariants::Type;
+    switch (type)
+    {
+#define M(NAME) \
+    case Type::NAME: \
+        return Type::NAME##_two_level;
+        APPLY_FOR_VARIANTS_CONVERTIBLE_TO_TWO_LEVEL(M)
+#undef M
+        default:
+            return type;
+    }
+    __builtin_unreachable();
+}
+
+void initDataVariantsWithSizeHint(
+    DB::AggregatedDataVariants & result, DB::AggregatedDataVariants::Type method_chosen, const DB::Aggregator::Params & params)
+{
+    const auto & stats_collecting_params = params.stats_collecting_params;
+    if (stats_collecting_params.isCollectionAndUseEnabled())
+    {
+        if (auto hint = getHashTablesStatistics().getSizeHint(stats_collecting_params))
+        {
+            const auto max_threads = params.group_by_two_level_threshold != 0 ? std::max(params.max_threads, 1ul) : 1;
+            const auto lower_limit = hint->sum_of_sizes / max_threads;
+            const auto upper_limit = stats_collecting_params.max_size_to_preallocate_for_aggregation / max_threads;
+            const auto adjusted = std::min(std::max(lower_limit, hint->median_size), upper_limit);
+            if (worthConvertToTwoLevel(
+                    params.group_by_two_level_threshold,
+                    hint->sum_of_sizes,
+                    /*group_by_two_level_threshold_bytes*/ 0,
+                    /*result_size_bytes*/ 0))
+                method_chosen = convertToTwoLevelTypeIfPossible(method_chosen);
+            result.init(method_chosen, adjusted);
+            ProfileEvents::increment(ProfileEvents::AggregationHashTablesInitializedAsTwoLevel, result.isTwoLevel());
+            return;
+        }
+    }
+    result.init(method_chosen);
+}
+
+/// Collection and use of the statistics should be enabled.
+void updateStatistics(const DB::ManyAggregatedDataVariants & data_variants, const DB::Aggregator::Params::StatsCollectingParams & params)
+{
+    if (!params.isCollectionAndUseEnabled())
+        throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Collection and use of the statistics should be enabled.");
+
+    std::vector<size_t> sizes(data_variants.size());
+    for (size_t i = 0; i < data_variants.size(); ++i)
+        sizes[i] = data_variants[i]->size();
+    const auto median_size = sizes.begin() + sizes.size() / 2; // not precisely though...
+    std::nth_element(sizes.begin(), median_size, sizes.end());
+    const auto sum_of_sizes = std::accumulate(sizes.begin(), sizes.end(), 0ull);
+    getHashTablesStatistics().update(sum_of_sizes, *median_size, params);
+}
+
+// The std::is_constructible trait isn't suitable here because some classes have template constructors with semantics different from providing size hints.
+// Also string hash table variants are not supported due to the fact that both local perf tests and tests in CI showed slowdowns for them.
+template <typename...>
+struct HasConstructorOfNumberOfElements : std::false_type
+{
+};
+
+template <typename... Ts>
+struct HasConstructorOfNumberOfElements<HashMapTable<Ts...>> : std::true_type
+{
+};
+
+template <typename Key, typename Cell, typename Hash, typename Grower, typename Allocator, template <typename...> typename ImplTable>
+struct HasConstructorOfNumberOfElements<TwoLevelHashMapTable<Key, Cell, Hash, Grower, Allocator, ImplTable>> : std::true_type
+{
+};
+
+template <typename... Ts>
+struct HasConstructorOfNumberOfElements<HashTable<Ts...>> : std::true_type
+{
+};
+
+template <typename... Ts>
+struct HasConstructorOfNumberOfElements<TwoLevelHashTable<Ts...>> : std::true_type
+{
+};
+
+template <template <typename> typename Method, typename Base>
+struct HasConstructorOfNumberOfElements<Method<Base>> : HasConstructorOfNumberOfElements<Base>
+{
+};
+
+template <typename Method>
+auto constructWithReserveIfPossible(size_t size_hint)
+{
+    if constexpr (HasConstructorOfNumberOfElements<typename Method::Data>::value)
+    {
+        ProfileEvents::increment(ProfileEvents::AggregationPreallocatedElementsInHashTables, size_hint);
+        return std::make_unique<Method>(size_hint);
+    }
+    else
+        return std::make_unique<Method>();
+}
 }
 
 namespace DB
@@ -64,6 +291,10 @@ AggregatedDataVariants::~AggregatedDataVariants()
     }
 }
 
+std::optional<HashTablesCacheStatistics> getHashTablesCacheStatistics()
+{
+    return getHashTablesStatistics().getCacheStats();
+}
 
 void AggregatedDataVariants::convertToTwoLevel()
 {
@@ -88,6 +319,47 @@ void AggregatedDataVariants::convertToTwoLevel()
     }
 }
 
+void AggregatedDataVariants::init(Type type_, std::optional<size_t> size_hint)
+{
+    switch (type_)
+    {
+        case Type::EMPTY:
+            break;
+        case Type::without_key:
+            break;
+
+#define M(NAME, IS_TWO_LEVEL) \
+    case Type::NAME: \
+        if (size_hint) \
+            (NAME) = constructWithReserveIfPossible<decltype(NAME)::element_type>(*size_hint); \
+        else \
+            (NAME) = std::make_unique<decltype(NAME)::element_type>(); \
+        break;
+            APPLY_FOR_AGGREGATED_VARIANTS(M)
+#undef M
+    }
+
+    type = type_;
+}
+
+Aggregator::Params::StatsCollectingParams::StatsCollectingParams() = default;
+
+Aggregator::Params::StatsCollectingParams::StatsCollectingParams(
+    const ASTPtr & select_query_,
+    bool collect_hash_table_stats_during_aggregation_,
+    size_t max_entries_for_hash_table_stats_,
+    size_t max_size_to_preallocate_for_aggregation_)
+    : key(collect_hash_table_stats_during_aggregation_ ? HashTablesStatistics::calculateCacheKey(select_query_) : 0)
+    , max_entries_for_hash_table_stats(max_entries_for_hash_table_stats_)
+    , max_size_to_preallocate_for_aggregation(max_size_to_preallocate_for_aggregation_)
+{
+}
+
+bool Aggregator::Params::StatsCollectingParams::isCollectionAndUseEnabled() const
+{
+    return key != 0;
+}
+
 Block Aggregator::getHeader(bool final) const
 {
     return params.getHeader(final);
@@ -237,8 +509,7 @@ public:
 
 #endif
 
-Aggregator::Aggregator(const Params & params_)
-    : params(params_)
+Aggregator::Aggregator(const Params & params_) : params(params_)
 {
     /// Use query-level memory tracker
     if (auto * memory_tracker_child = CurrentThread::getMemoryTracker())
@@ -292,7 +563,6 @@ Aggregator::Aggregator(const Params & params_)
 #if USE_EMBEDDED_COMPILER
     compileAggregateFunctionsIfNeeded();
 #endif
-
 }
 
 #if USE_EMBEDDED_COMPILER
@@ -958,7 +1228,7 @@ bool Aggregator::executeOnBlock(Columns columns, UInt64 num_rows, AggregatedData
     /// How to perform the aggregation?
     if (result.empty())
     {
-        result.init(method_chosen);
+        initDataVariantsWithSizeHint(result, method_chosen, params);
         result.keys_size = params.keys_size;
         result.key_sizes = key_sizes;
         LOG_TRACE(log, "Aggregation method: {}", result.getMethodName());
@@ -1038,9 +1308,8 @@ bool Aggregator::executeOnBlock(Columns columns, UInt64 num_rows, AggregatedData
     /// Here all the results in the sum are taken into account, from different threads.
     auto result_size_bytes = current_memory_usage - memory_usage_before_aggregation;
 
-    bool worth_convert_to_two_level
-        = (params.group_by_two_level_threshold && result_size >= params.group_by_two_level_threshold)
-        || (params.group_by_two_level_threshold_bytes && result_size_bytes >= static_cast<Int64>(params.group_by_two_level_threshold_bytes));
+    bool worth_convert_to_two_level = worthConvertToTwoLevel(
+        params.group_by_two_level_threshold, result_size, params.group_by_two_level_threshold_bytes, result_size_bytes);
 
     /** Converting to a two-level data structure.
       * It allows you to make, in the subsequent, an effective merge - either economical from memory or parallel.
@@ -1327,10 +1596,7 @@ void Aggregator::convertToBlockImpl(
 
 
 template <typename Mapped>
-inline void Aggregator::insertAggregatesIntoColumns(
-    Mapped & mapped,
-    MutableColumns & final_aggregate_columns,
-    Arena * arena) const
+inline void Aggregator::insertAggregatesIntoColumns(Mapped & mapped, MutableColumns & final_aggregate_columns, Arena * arena) const
 {
     /** Final values of aggregate functions are inserted to columns.
       * Then states of aggregate functions, that are not longer needed, are destroyed.
@@ -2179,6 +2445,9 @@ ManyAggregatedDataVariants Aggregator::prepareVariantsToMerge(ManyAggregatedData
 
     LOG_TRACE(log, "Merging aggregated data");
 
+    if (params.stats_collecting_params.isCollectionAndUseEnabled())
+        updateStatistics(data_variants, params.stats_collecting_params);
+
     ManyAggregatedDataVariants non_empty_data;
     non_empty_data.reserve(data_variants.size());
     for (auto & data : data_variants)
@@ -2388,9 +2657,8 @@ bool Aggregator::mergeOnBlock(Block block, AggregatedDataVariants & result, bool
     /// Here all the results in the sum are taken into account, from different threads.
     auto result_size_bytes = current_memory_usage - memory_usage_before_aggregation;
 
-    bool worth_convert_to_two_level
-        = (params.group_by_two_level_threshold && result_size >= params.group_by_two_level_threshold)
-        || (params.group_by_two_level_threshold_bytes && result_size_bytes >= static_cast<Int64>(params.group_by_two_level_threshold_bytes));
+    bool worth_convert_to_two_level = worthConvertToTwoLevel(
+        params.group_by_two_level_threshold, result_size, params.group_by_two_level_threshold_bytes, result_size_bytes);
 
     /** Converting to a two-level data structure.
       * It allows you to make, in the subsequent, an effective merge - either economical from memory or parallel.

src/Interpreters/Aggregator.h

@@ -34,6 +34,7 @@
 #include <Columns/ColumnNullable.h>
 #include <Columns/ColumnLowCardinality.h>
 
+#include <Parsers/IAST_fwd.h>
 
 namespace DB
 {
@@ -129,6 +130,7 @@ private:
 template <typename Base>
 struct AggregationDataWithNullKeyTwoLevel : public Base
 {
+    using Base::Base;
     using Base::impls;
 
     AggregationDataWithNullKeyTwoLevel() = default;
@@ -183,6 +185,8 @@ struct AggregationMethodOneNumber
 
     AggregationMethodOneNumber() = default;
 
+    explicit AggregationMethodOneNumber(size_t size_hint) : data(size_hint) { }
+
     template <typename Other>
     explicit AggregationMethodOneNumber(const Other & other) : data(other.data)
     {
@@ -225,6 +229,8 @@ struct AggregationMethodString
     {
     }
 
+    explicit AggregationMethodString(size_t size_hint) : data(size_hint) { }
+
     using State = ColumnsHashing::HashMethodString<typename Data::value_type, Mapped>;
 
     static const bool low_cardinality_optimization = false;
@@ -250,6 +256,8 @@ struct AggregationMethodStringNoCache
 
     AggregationMethodStringNoCache() = default;
 
+    explicit AggregationMethodStringNoCache(size_t size_hint) : data(size_hint) { }
+
     template <typename Other>
     explicit AggregationMethodStringNoCache(const Other & other) : data(other.data)
     {
@@ -280,6 +288,8 @@ struct AggregationMethodFixedString
 
     AggregationMethodFixedString() = default;
 
+    explicit AggregationMethodFixedString(size_t size_hint) : data(size_hint) { }
+
     template <typename Other>
     explicit AggregationMethodFixedString(const Other & other) : data(other.data)
     {
@@ -309,6 +319,8 @@ struct AggregationMethodFixedStringNoCache
 
     AggregationMethodFixedStringNoCache() = default;
 
+    explicit AggregationMethodFixedStringNoCache(size_t size_hint) : data(size_hint) { }
+
     template <typename Other>
     explicit AggregationMethodFixedStringNoCache(const Other & other) : data(other.data)
     {
@@ -382,6 +394,8 @@ struct AggregationMethodKeysFixed
 
     AggregationMethodKeysFixed() = default;
 
+    explicit AggregationMethodKeysFixed(size_t size_hint) : data(size_hint) { }
+
     template <typename Other>
     explicit AggregationMethodKeysFixed(const Other & other) : data(other.data)
     {
@@ -473,6 +487,8 @@ struct AggregationMethodSerialized
 
     AggregationMethodSerialized() = default;
 
+    explicit AggregationMethodSerialized(size_t size_hint) : data(size_hint) { }
+
     template <typename Other>
     explicit AggregationMethodSerialized(const Other & other) : data(other.data)
     {
@@ -652,21 +668,7 @@ struct AggregatedDataVariants : private boost::noncopyable
 
     ~AggregatedDataVariants();
 
-    void init(Type type_)
-    {
-        switch (type_)
-        {
-            case Type::EMPTY:       break;
-            case Type::without_key: break;
-
-        #define M(NAME, IS_TWO_LEVEL) \
-            case Type::NAME: (NAME) = std::make_unique<decltype(NAME)::element_type>(); break;
-            APPLY_FOR_AGGREGATED_VARIANTS(M)
-        #undef M
-        }
-
-        type = type_;
-    }
+    void init(Type type_, std::optional<size_t> size_hint = std::nullopt);
 
     /// Number of rows (different keys).
     size_t size() const
@@ -929,29 +931,61 @@ public:
         bool compile_aggregate_expressions;
         size_t min_count_to_compile_aggregate_expression;
 
+        struct StatsCollectingParams
+        {
+            StatsCollectingParams();
+
+            StatsCollectingParams(
+                const ASTPtr & select_query_,
+                bool collect_hash_table_stats_during_aggregation_,
+                size_t max_entries_for_hash_table_stats_,
+                size_t max_size_to_preallocate_for_aggregation_);
+
+            bool isCollectionAndUseEnabled() const;
+
+            const UInt64 key = 0;
+            const size_t max_entries_for_hash_table_stats = 0;
+            const size_t max_size_to_preallocate_for_aggregation = 0;
+        };
+        StatsCollectingParams stats_collecting_params;
+
         Params(
             const Block & src_header_,
-            const ColumnNumbers & keys_, const AggregateDescriptions & aggregates_,
-            bool overflow_row_, size_t max_rows_to_group_by_, OverflowMode group_by_overflow_mode_,
-            size_t group_by_two_level_threshold_, size_t group_by_two_level_threshold_bytes_,
+            const ColumnNumbers & keys_,
+            const AggregateDescriptions & aggregates_,
+            bool overflow_row_,
+            size_t max_rows_to_group_by_,
+            OverflowMode group_by_overflow_mode_,
+            size_t group_by_two_level_threshold_,
+            size_t group_by_two_level_threshold_bytes_,
             size_t max_bytes_before_external_group_by_,
             bool empty_result_for_aggregation_by_empty_set_,
-            VolumePtr tmp_volume_, size_t max_threads_,
+            VolumePtr tmp_volume_,
+            size_t max_threads_,
             size_t min_free_disk_space_,
             bool compile_aggregate_expressions_,
             size_t min_count_to_compile_aggregate_expression_,
-            const Block & intermediate_header_ = {})
-            : src_header(src_header_),
-            intermediate_header(intermediate_header_),
-            keys(keys_), aggregates(aggregates_), keys_size(keys.size()), aggregates_size(aggregates.size()),
-            overflow_row(overflow_row_), max_rows_to_group_by(max_rows_to_group_by_), group_by_overflow_mode(group_by_overflow_mode_),
-            group_by_two_level_threshold(group_by_two_level_threshold_), group_by_two_level_threshold_bytes(group_by_two_level_threshold_bytes_),
-            max_bytes_before_external_group_by(max_bytes_before_external_group_by_),
-            empty_result_for_aggregation_by_empty_set(empty_result_for_aggregation_by_empty_set_),
-            tmp_volume(tmp_volume_), max_threads(max_threads_),
-            min_free_disk_space(min_free_disk_space_),
-            compile_aggregate_expressions(compile_aggregate_expressions_),
-            min_count_to_compile_aggregate_expression(min_count_to_compile_aggregate_expression_)
+            const Block & intermediate_header_ = {},
+            const StatsCollectingParams & stats_collecting_params_ = {})
+            : src_header(src_header_)
+            , intermediate_header(intermediate_header_)
+            , keys(keys_)
+            , aggregates(aggregates_)
+            , keys_size(keys.size())
+            , aggregates_size(aggregates.size())
+            , overflow_row(overflow_row_)
+            , max_rows_to_group_by(max_rows_to_group_by_)
+            , group_by_overflow_mode(group_by_overflow_mode_)
+            , group_by_two_level_threshold(group_by_two_level_threshold_)
+            , group_by_two_level_threshold_bytes(group_by_two_level_threshold_bytes_)
+            , max_bytes_before_external_group_by(max_bytes_before_external_group_by_)
+            , empty_result_for_aggregation_by_empty_set(empty_result_for_aggregation_by_empty_set_)
+            , tmp_volume(tmp_volume_)
+            , max_threads(max_threads_)
+            , min_free_disk_space(min_free_disk_space_)
+            , compile_aggregate_expressions(compile_aggregate_expressions_)
+            , min_count_to_compile_aggregate_expression(min_count_to_compile_aggregate_expression_)
+            , stats_collecting_params(stats_collecting_params_)
         {
         }
 
@@ -1350,4 +1384,13 @@ APPLY_FOR_AGGREGATED_VARIANTS(M)
 
 #undef M
 
+
+struct HashTablesCacheStatistics
+{
+    size_t entries = 0;
+    size_t hits = 0;
+    size_t misses = 0;
+};
+
+std::optional<HashTablesCacheStatistics> getHashTablesCacheStatistics();
 }

src/Interpreters/AsynchronousMetrics.cpp

@@ -1,3 +1,4 @@
+#include <Interpreters/Aggregator.h>
 #include <Interpreters/AsynchronousMetrics.h>
 #include <Interpreters/AsynchronousMetricLog.h>
 #include <Interpreters/JIT/CompiledExpressionCache.h>
@@ -630,6 +631,15 @@ void AsynchronousMetrics::update(std::chrono::system_clock::time_point update_ti
 
     new_values["Uptime"] = getContext()->getUptimeSeconds();
 
+    {
+        if (const auto stats = getHashTablesCacheStatistics())
+        {
+            new_values["HashTableStatsCacheEntries"] = stats->entries;
+            new_values["HashTableStatsCacheHits"] = stats->hits;
+            new_values["HashTableStatsCacheMisses"] = stats->misses;
+        }
+    }
+
     /// Process process memory usage according to OS
 #if defined(OS_LINUX) || defined(OS_FREEBSD)
     {

src/Interpreters/CatBoostModel.cpp

@@ -169,6 +169,7 @@ public:
         if (columns.size() != float_features_count + cat_features_count)
             throw Exception(ErrorCodes::BAD_ARGUMENTS,
                 "Number of columns is different with number of features: columns size {} float features size {} + cat features size {}",
+                columns.size(),
                 float_features_count,
                 cat_features_count);
 

src/Interpreters/DatabaseCatalog.cpp

@@ -233,7 +233,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         {
             assert(!db_and_table.first && !db_and_table.second);
             if (exception)
-                exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Table {} doesn't exist", table_id.getNameForLogs());
+                exception->emplace(fmt::format("Table {} doesn't exist", table_id.getNameForLogs()), ErrorCodes::UNKNOWN_TABLE);
             return {};
         }
 
@@ -263,7 +263,7 @@ 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(ErrorCodes::DATABASE_ACCESS_DENIED, "Direct access to `{}` database is not allowed", String(TEMPORARY_DATABASE));
+            exception->emplace(fmt::format("Direct access to `{}` database is not allowed", TEMPORARY_DATABASE), ErrorCodes::DATABASE_ACCESS_DENIED);
         return {};
     }
 
@@ -274,7 +274,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
         if (databases.end() == it)
         {
             if (exception)
-                exception->emplace(ErrorCodes::UNKNOWN_DATABASE, "Database {} doesn't exist", backQuoteIfNeed(table_id.getDatabaseName()));
+                exception->emplace(fmt::format("Database {} doesn't exist", backQuoteIfNeed(table_id.getDatabaseName())), ErrorCodes::UNKNOWN_DATABASE);
             return {};
         }
         database = it->second;
@@ -282,7 +282,7 @@ DatabaseAndTable DatabaseCatalog::getTableImpl(
 
     auto table = database->tryGetTable(table_id.table_name, context_);
     if (!table && exception)
-            exception->emplace(ErrorCodes::UNKNOWN_TABLE, "Table {} doesn't exist", table_id.getNameForLogs());
+            exception->emplace(fmt::format("Table {} doesn't exist", table_id.getNameForLogs()), ErrorCodes::UNKNOWN_TABLE);
     if (!table)
         database = nullptr;
 

src/Interpreters/InterpreterSelectQuery.cpp

@@ -2082,6 +2082,12 @@ void InterpreterSelectQuery::executeAggregation(QueryPlan & query_plan, const Ac
 
     const Settings & settings = context->getSettingsRef();
 
+    const auto stats_collecting_params = Aggregator::Params::StatsCollectingParams(
+        query_ptr,
+        settings.collect_hash_table_stats_during_aggregation,
+        settings.max_entries_for_hash_table_stats,
+        settings.max_size_to_preallocate_for_aggregation);
+
     Aggregator::Params params(
         header_before_aggregation,
         keys,
@@ -2099,7 +2105,9 @@ void InterpreterSelectQuery::executeAggregation(QueryPlan & query_plan, const Ac
         settings.max_threads,
         settings.min_free_disk_space_for_temporary_data,
         settings.compile_aggregate_expressions,
-        settings.min_count_to_compile_aggregate_expression);
+        settings.min_count_to_compile_aggregate_expression,
+        Block{},
+        stats_collecting_params);
 
     SortDescription group_by_sort_description;
 

src/Interpreters/executeDDLQueryOnCluster.cpp

@@ -320,12 +320,13 @@ Chunk DDLQueryStatusSource::generate()
             if (throw_on_timeout)
             {
                 if (!first_exception)
-                    first_exception = std::make_unique<Exception>(ErrorCodes::TIMEOUT_EXCEEDED, msg_format,
-                        node_path, timeout_seconds, num_unfinished_hosts, num_active_hosts);
+                    first_exception = std::make_unique<Exception>(
+                        fmt::format(msg_format, node_path, timeout_seconds, num_unfinished_hosts, num_active_hosts),
+                        ErrorCodes::TIMEOUT_EXCEEDED);
                 return {};
             }
 
-            LOG_INFO(log, fmt::runtime(msg_format), node_path, timeout_seconds, num_unfinished_hosts, num_active_hosts);
+            LOG_INFO(log, msg_format, node_path, timeout_seconds, num_unfinished_hosts, num_active_hosts);
 
             NameSet unfinished_hosts = waiting_hosts;
             for (const auto & host_id : finished_hosts)
@@ -358,9 +359,12 @@ Chunk DDLQueryStatusSource::generate()
             /// Paradoxically, this exception will be throw even in case of "never_throw" mode.
 
             if (!first_exception)
-                first_exception = std::make_unique<Exception>(ErrorCodes::UNFINISHED,
-                    "Cannot provide query execution status. The query's node {} has been deleted by the cleaner"
-                    " since it was finished (or its lifetime is expired)", node_path);
+                first_exception = std::make_unique<Exception>(
+                    fmt::format(
+                        "Cannot provide query execution status. The query's node {} has been deleted by the cleaner"
+                        " since it was finished (or its lifetime is expired)",
+                        node_path),
+                    ErrorCodes::UNFINISHED);
             return {};
         }
 
@@ -386,7 +390,8 @@ Chunk DDLQueryStatusSource::generate()
             if (status.code != 0 && !first_exception
                 && context->getSettingsRef().distributed_ddl_output_mode != DistributedDDLOutputMode::NEVER_THROW)
             {
-                first_exception = std::make_unique<Exception>(status.code, "There was an error on [{}:{}]: {}", host, port, status.message);
+                first_exception = std::make_unique<Exception>(
+                    fmt::format("There was an error on [{}:{}]: {}", host, port, status.message), status.code);
             }
 
             ++num_hosts_finished;

src/Processors/Formats/Impl/MsgPackRowInputFormat.cpp

@@ -359,7 +359,7 @@ bool MsgPackVisitor::visit_ext(const char * value, uint32_t size)
         return true;
     }
 
-    throw Exception(ErrorCodes::BAD_ARGUMENTS, "Unsupported MsgPack extension type: {%x}", type);
+    throw Exception(ErrorCodes::BAD_ARGUMENTS, "Unsupported MsgPack extension type: {:x}", type);
 }
 
 void MsgPackVisitor::parse_error(size_t, size_t) // NOLINT
@@ -498,7 +498,7 @@ DataTypePtr MsgPackSchemaReader::getDataType(const msgpack::object & object)
             msgpack::object_ext object_ext = object.via.ext;
             if (object_ext.type() == int8_t(MsgPackExtensionTypes::UUIDType))
                 return std::make_shared<DataTypeUUID>();
-            throw Exception(ErrorCodes::BAD_ARGUMENTS, "Msgpack extension type {%x} is not supported", object_ext.type());
+            throw Exception(ErrorCodes::BAD_ARGUMENTS, "Msgpack extension type {:x} is not supported", object_ext.type());
         }
     }
     __builtin_unreachable();

src/Storages/Kafka/ReadBufferFromKafkaConsumer.cpp

@@ -1,10 +1,13 @@
+// Needs to go first because its partial specialization of fmt::formatter
+// should be defined before any instantiation
+#include <fmt/ostream.h>
+
 #include <Storages/Kafka/ReadBufferFromKafkaConsumer.h>
 
 #include <base/logger_useful.h>
 
 #include <cppkafka/cppkafka.h>
 #include <boost/algorithm/string/join.hpp>
-#include <fmt/ostream.h>
 #include <algorithm>
 
 namespace DB

src/Storages/StorageS3.cpp

@@ -1,4 +1,6 @@
 #include <Common/config.h>
+#include "IO/ParallelReadBuffer.h"
+#include "IO/IOThreadPool.h"
 #include "Parsers/ASTCreateQuery.h"
 
 #if USE_AWS_S3
@@ -238,7 +240,8 @@ StorageS3Source::StorageS3Source(
     String compression_hint_,
     const std::shared_ptr<Aws::S3::S3Client> & client_,
     const String & bucket_,
-    std::shared_ptr<IteratorWrapper> file_iterator_)
+    std::shared_ptr<IteratorWrapper> file_iterator_,
+    const size_t download_thread_num_)
     : SourceWithProgress(getHeader(sample_block_, need_path, need_file))
     , WithContext(context_)
     , name(std::move(name_))
@@ -254,6 +257,7 @@ StorageS3Source::StorageS3Source(
     , with_file_column(need_file)
     , with_path_column(need_path)
     , file_iterator(file_iterator_)
+    , download_thread_num(download_thread_num_)
 {
     initialize();
 }
@@ -275,28 +279,79 @@ bool StorageS3Source::initialize()
 
     file_path = fs::path(bucket) / current_key;
 
-    read_buf = wrapReadBufferWithCompressionMethod(
-        std::make_unique<ReadBufferFromS3>(client, bucket, current_key, max_single_read_retries, getContext()->getReadSettings()),
-        chooseCompressionMethod(current_key, compression_hint));
+    read_buf = wrapReadBufferWithCompressionMethod(createS3ReadBuffer(current_key), chooseCompressionMethod(current_key, compression_hint));
+
     auto input_format = getContext()->getInputFormat(format, *read_buf, sample_block, max_block_size, format_settings);
     QueryPipelineBuilder builder;
     builder.init(Pipe(input_format));
 
     if (columns_desc.hasDefaults())
     {
-        builder.addSimpleTransform([&](const Block & header)
-        {
-            return std::make_shared<AddingDefaultsTransform>(header, columns_desc, *input_format, getContext());
-        });
+        builder.addSimpleTransform(
+            [&](const Block & header)
+            { return std::make_shared<AddingDefaultsTransform>(header, columns_desc, *input_format, getContext()); });
     }
 
     pipeline = std::make_unique<QueryPipeline>(QueryPipelineBuilder::getPipeline(std::move(builder)));
     reader = std::make_unique<PullingPipelineExecutor>(*pipeline);
 
-    initialized = false;
     return true;
 }
 
+std::unique_ptr<ReadBuffer> StorageS3Source::createS3ReadBuffer(const String & key)
+{
+    const size_t object_size = DB::S3::getObjectSize(client, bucket, key, false);
+
+    auto download_buffer_size = getContext()->getSettings().max_download_buffer_size;
+    const bool use_parallel_download = download_buffer_size > 0 && download_thread_num > 1;
+    const bool object_too_small = object_size < download_thread_num * download_buffer_size;
+    if (!use_parallel_download || object_too_small)
+    {
+        LOG_TRACE(log, "Downloading object of size {} from S3 in single thread", object_size);
+        return std::make_unique<ReadBufferFromS3>(client, bucket, key, max_single_read_retries, getContext()->getReadSettings());
+    }
+
+    assert(object_size > 0);
+
+    if (download_buffer_size < DBMS_DEFAULT_BUFFER_SIZE)
+    {
+        LOG_WARNING(log, "Downloading buffer {} bytes too small, set at least {} bytes", download_buffer_size, DBMS_DEFAULT_BUFFER_SIZE);
+        download_buffer_size = DBMS_DEFAULT_BUFFER_SIZE;
+    }
+
+    auto factory = std::make_unique<ReadBufferS3Factory>(
+        client, bucket, key, download_buffer_size, object_size, max_single_read_retries, getContext()->getReadSettings());
+    LOG_TRACE(
+        log, "Downloading from S3 in {} threads. Object size: {}, Range size: {}.", download_thread_num, object_size, download_buffer_size);
+
+    ThreadGroupStatusPtr running_group = CurrentThread::isInitialized() && CurrentThread::get().getThreadGroup()
+        ? CurrentThread::get().getThreadGroup()
+        : MainThreadStatus::getInstance().getThreadGroup();
+
+    ContextPtr query_context = CurrentThread::isInitialized() ? CurrentThread::get().getQueryContext() : nullptr;
+
+    auto worker_cleanup = [has_running_group = running_group == nullptr](ThreadStatus & thread_status)
+    {
+        if (has_running_group)
+            thread_status.detachQuery(false);
+    };
+
+    auto worker_setup = [query_context = std::move(query_context),
+                         running_group = std::move(running_group)](ThreadStatus & thread_status)
+    {
+        /// Save query context if any, because cache implementation needs it.
+        if (query_context)
+            thread_status.attachQueryContext(query_context);
+
+        /// To be able to pass ProfileEvents.
+        if (running_group)
+            thread_status.attachQuery(running_group);
+    };
+
+    return std::make_unique<ParallelReadBuffer>(
+        std::move(factory), &IOThreadPool::get(), download_thread_num, std::move(worker_setup), std::move(worker_cleanup));
+}
+
 String StorageS3Source::getName() const
 {
     return name;
@@ -670,6 +725,7 @@ Pipe StorageS3::read(
         block_for_format = storage_snapshot->metadata->getSampleBlock();
     }
 
+    const size_t max_download_threads = local_context->getSettingsRef().max_download_threads;
     for (size_t i = 0; i < num_streams; ++i)
     {
         pipes.emplace_back(std::make_shared<StorageS3Source>(
@@ -686,7 +742,8 @@ Pipe StorageS3::read(
             compression_method,
             client_auth.client,
             client_auth.uri.bucket,
-            iterator_wrapper));
+            iterator_wrapper,
+            max_download_threads));
     }
     auto pipe = Pipe::unitePipes(std::move(pipes));
 

src/Storages/StorageS3.h

@@ -74,7 +74,8 @@ public:
         String compression_hint_,
         const std::shared_ptr<Aws::S3::S3Client> & client_,
         const String & bucket,
-        std::shared_ptr<IteratorWrapper> file_iterator_);
+        std::shared_ptr<IteratorWrapper> file_iterator_,
+        size_t download_thread_num);
 
     String getName() const override;
 
@@ -101,13 +102,17 @@ private:
     std::unique_ptr<PullingPipelineExecutor> reader;
     /// onCancel and generate can be called concurrently
     std::mutex reader_mutex;
-    bool initialized = false;
     bool with_file_column = false;
     bool with_path_column = false;
     std::shared_ptr<IteratorWrapper> file_iterator;
+    size_t download_thread_num = 1;
+
+    Poco::Logger * log = &Poco::Logger::get("StorageS3Source");
 
     /// Recreate ReadBuffer and BlockInputStream for each file.
     bool initialize();
+
+    std::unique_ptr<ReadBuffer> createS3ReadBuffer(const String & key);
 };
 
 /**

tests/ci/build_check.py

@@ -54,6 +54,7 @@ def get_packager_cmd(
     build_version: str,
     image_version: str,
     ccache_path: str,
+    official: bool,
 ) -> str:
     package_type = build_config["package_type"]
     comp = build_config["compiler"]
@@ -83,6 +84,9 @@ def get_packager_cmd(
     if _can_export_binaries(build_config):
         cmd += " --with-binaries=tests"
 
+    if official:
+        cmd += " --official"
+
     return cmd
 
 
@@ -254,9 +258,11 @@ def main():
 
     logging.info("Got version from repo %s", version.string)
 
+    official_flag = pr_info.number == 0
     version_type = "testing"
     if "release" in pr_info.labels or "release-lts" in pr_info.labels:
         version_type = "stable"
+        official_flag = True
 
     update_version_local(REPO_COPY, version, version_type)
 
@@ -290,7 +296,9 @@ def main():
         version.string,
         image_version,
         ccache_path,
+        official=official_flag,
     )
+
     logging.info("Going to run packager with %s", packager_cmd)
 
     build_clickhouse_log = os.path.join(TEMP_PATH, "build_log")

tests/ci/build_report_check.py

@@ -233,7 +233,11 @@ if __name__ == "__main__":
     if ok_builds == 0 or some_builds_are_missing:
         summary_status = "error"
 
-    description = f"{ok_builds}/{total_builds} builds are OK"
+    addition = ""
+    if some_builds_are_missing:
+        addition = "(some builds are missing)"
+
+    description = f"{ok_builds}/{total_builds} builds are OK {addition}"
 
     print("::notice ::Report url: {}".format(url))
 
@@ -244,3 +248,6 @@ if __name__ == "__main__":
         state=summary_status,
         target_url=url,
     )
+
+    if summary_status == "error":
+        sys.exit(1)

tests/ci/cherry_pick_utils/parser.py

@@ -20,8 +20,6 @@ class Description:
 
     def __init__(self, pull_request):
         self.label_name = str()
-        self.legal = False
-
         self._parse(pull_request["bodyText"])
 
     def _parse(self, text):
@@ -39,12 +37,6 @@ class Description:
                 category = stripped
                 next_category = False
 
-            if (
-                stripped
-                == "I hereby agree to the terms of the CLA available at: https://yandex.ru/legal/cla/?lang=en"
-            ):
-                self.legal = True
-
             category_headers = (
                 "Category (leave one):",
                 "Changelog category (leave one):",

tests/ci/clickhouse_helper.py

@@ -15,7 +15,7 @@ class ClickHouseHelper:
         self.url = url
         self.auth = {
             "X-ClickHouse-User": get_parameter_from_ssm("clickhouse-test-stat-login"),
-            "X-ClickHouse-Key": get_parameter_from_ssm("clickhouse-test-stat-password")
+            "X-ClickHouse-Key": get_parameter_from_ssm("clickhouse-test-stat-password"),
         }
 
     @staticmethod

tests/ci/commit_status_helper.py

@@ -59,3 +59,17 @@ def post_commit_status_to_file(file_path, description, state, report_url):
     with open(file_path, "w", encoding="utf-8") as f:
         out = csv.writer(f, delimiter="\t")
         out.writerow([state, report_url, description])
+
+
+def remove_labels(gh, pr_info, labels_names):
+    repo = gh.get_repo(GITHUB_REPOSITORY)
+    pull_request = repo.get_pull(pr_info.number)
+    for label in labels_names:
+        pull_request.remove_from_labels(label)
+
+
+def post_labels(gh, pr_info, labels_names):
+    repo = gh.get_repo(GITHUB_REPOSITORY)
+    pull_request = repo.get_pull(pr_info.number)
+    for label in labels_names:
+        pull_request.add_to_labels(label)

tests/ci/compatibility_check.py

@@ -197,4 +197,8 @@ if __name__ == "__main__":
         report_url,
         CHECK_NAME,
     )
+
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+
+    if state == "error":
+        sys.exit(1)

tests/ci/docker_images_check.py

@@ -7,6 +7,7 @@ import platform
 import shutil
 import subprocess
 import time
+import sys
 from typing import Dict, List, Optional, Set, Tuple, Union
 
 from github import Github
@@ -461,6 +462,9 @@ def main():
     ch_helper = ClickHouseHelper()
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
 
+    if status == "error":
+        sys.exit(1)
+
 
 if __name__ == "__main__":
     main()

tests/ci/docs_check.py

@@ -114,4 +114,7 @@ if __name__ == "__main__":
         report_url,
         NAME,
     )
+
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+    if status == "error":
+        sys.exit(1)

tests/ci/fast_test_check.py

@@ -208,7 +208,7 @@ if __name__ == "__main__":
 
     # Refuse other checks to run if fast test failed
     if state != "success":
-        if "force-tests" in pr_info.labels:
+        if "force-tests" in pr_info.labels and state != "error":
             print("'force-tests' enabled, will report success")
         else:
             sys.exit(1)

tests/ci/integration_test_check.py

@@ -279,4 +279,8 @@ if __name__ == "__main__":
         report_url,
         check_name_with_group,
     )
+
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+
+    if state == "error":
+        sys.exit(1)

tests/ci/performance_comparison_check.py

@@ -217,3 +217,6 @@ if __name__ == "__main__":
     post_commit_status(
         gh, pr_info.sha, check_name_with_group, message, status, report_url
     )
+
+    if status == "error":
+        sys.exit(1)

tests/ci/pr_info.py

@@ -236,6 +236,15 @@ class PRInfo:
                 return True
         return False
 
+    def has_changes_in_submodules(self):
+        if self.changed_files is None or not self.changed_files:
+            return True
+
+        for f in self.changed_files:
+            if "contrib" in f:
+                return True
+        return False
+
     def can_skip_builds_and_use_version_from_master(self):
         # TODO: See a broken loop
         if "force tests" in self.labels:

tests/ci/run_check.py

@@ -8,7 +8,7 @@ from github import Github
 from env_helper import GITHUB_RUN_URL, GITHUB_REPOSITORY, GITHUB_SERVER_URL
 from pr_info import PRInfo
 from get_robot_token import get_best_robot_token
-from commit_status_helper import get_commit
+from commit_status_helper import get_commit, post_labels, remove_labels
 
 NAME = "Run Check (actions)"
 
@@ -22,6 +22,7 @@ OK_SKIP_LABELS = {"release", "pr-backport", "pr-cherrypick"}
 CAN_BE_TESTED_LABEL = "can be tested"
 DO_NOT_TEST_LABEL = "do not test"
 FORCE_TESTS_LABEL = "force tests"
+SUBMODULE_CHANGED_LABEL = "submodule changed"
 
 # Individual trusted contirbutors who are not in any trusted organization.
 # Can be changed in runtime: we will append users that we learned to be in
@@ -81,6 +82,25 @@ TRUSTED_CONTRIBUTORS = {
     ]
 }
 
+MAP_CATEGORY_TO_LABEL = {
+    "New Feature": "pr-feature",
+    "Bug Fix": "pr-bugfix",
+    "Bug Fix (user-visible misbehaviour in official stable or prestable release)": "pr-bugfix",
+    "Improvement": "pr-improvement",
+    "Performance Improvement": "pr-performance",
+    "Backward Incompatible Change": "pr-backward-incompatible",
+    "Build/Testing/Packaging Improvement": "pr-build",
+    "Build Improvement": "pr-build",
+    "Build/Testing Improvement": "pr-build",
+    "Build": "pr-build",
+    "Packaging Improvement": "pr-build",
+    "Not for changelog (changelog entry is not required)": "pr-not-for-changelog",
+    "Not for changelog": "pr-not-for-changelog",
+    "Documentation (changelog entry is not required)": "pr-documentation",
+    "Documentation": "pr-documentation",
+    # 'Other': doesn't match anything
+}
+
 
 def pr_is_by_trusted_user(pr_user_login, pr_user_orgs):
     if pr_user_login.lower() in TRUSTED_CONTRIBUTORS:
@@ -168,7 +188,7 @@ def check_pr_description(pr_info):
                     + second_category
                     + "'"
                 )
-                return result_status[:140]
+                return result_status[:140], category
 
         elif re.match(
             r"(?i)^[>*_ ]*(short\s*description|change\s*log\s*entry)", lines[i]
@@ -190,30 +210,57 @@ def check_pr_description(pr_info):
             i += 1
 
     if not category:
-        return "Changelog category is empty"
+        return "Changelog category is empty", category
 
     # 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)",
         category,
     ):
-        return ""
+        return "", category
 
     if not entry:
-        return f"Changelog entry required for category '{category}'"
+        return f"Changelog entry required for category '{category}'", category
 
-    return ""
+    return "", category
 
 
 if __name__ == "__main__":
     logging.basicConfig(level=logging.INFO)
 
-    pr_info = PRInfo(need_orgs=True, pr_event_from_api=True)
+    pr_info = PRInfo(need_orgs=True, pr_event_from_api=True, need_changed_files=True)
     can_run, description, labels_state = should_run_checks_for_pr(pr_info)
     gh = Github(get_best_robot_token())
     commit = get_commit(gh, pr_info.sha)
 
-    description_report = check_pr_description(pr_info)[:139]
+    description_report, category = check_pr_description(pr_info)
+    pr_labels_to_add = []
+    pr_labels_to_remove = []
+    if (
+        category in MAP_CATEGORY_TO_LABEL
+        and MAP_CATEGORY_TO_LABEL[category] not in pr_info.labels
+    ):
+        pr_labels_to_add.append(MAP_CATEGORY_TO_LABEL[category])
+
+    for label in pr_info.labels:
+        if (
+            label in MAP_CATEGORY_TO_LABEL.values()
+            and category in MAP_CATEGORY_TO_LABEL
+            and label != MAP_CATEGORY_TO_LABEL[category]
+        ):
+            pr_labels_to_remove.append(label)
+
+    if pr_info.has_changes_in_submodules():
+        pr_labels_to_add.append(SUBMODULE_CHANGED_LABEL)
+    elif SUBMODULE_CHANGED_LABEL in pr_info.labels:
+        pr_labels_to_remove.append(SUBMODULE_CHANGED_LABEL)
+
+    if pr_labels_to_add:
+        post_labels(gh, pr_info, pr_labels_to_add)
+
+    if pr_labels_to_remove:
+        remove_labels(gh, pr_info, pr_labels_to_remove)
+
     if description_report:
         print("::notice ::Cannot run, description does not match the template")
         logging.info(
@@ -225,7 +272,7 @@ if __name__ == "__main__":
         )
         commit.create_status(
             context=NAME,
-            description=description_report,
+            description=description_report[:139],
             state="failure",
             target_url=url,
         )

tests/ci/split_build_smoke_check.py

@@ -147,4 +147,8 @@ if __name__ == "__main__":
         report_url,
         CHECK_NAME,
     )
+
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+
+    if state == "error":
+        sys.exit(1)

tests/ci/stress_check.py

@@ -177,3 +177,6 @@ if __name__ == "__main__":
         check_name,
     )
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+
+    if state == "error":
+        sys.exit(1)

tests/ci/style_check.py

@@ -118,3 +118,6 @@ if __name__ == "__main__":
         NAME,
     )
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+
+    if state == "error":
+        sys.exit(1)

tests/ci/unit_tests_check.py

@@ -173,4 +173,8 @@ if __name__ == "__main__":
         report_url,
         check_name,
     )
+
     ch_helper.insert_events_into(db="default", table="checks", events=prepared_events)
+
+    if state == "error":
+        sys.exit(1)

tests/clickhouse-test

@@ -52,6 +52,7 @@ MESSAGES_TO_RETRY = [
     "DB::Exception: New table appeared in database being dropped or detached. Try again",
     "is already started to be removing by another replica right now",
     "DB::Exception: Cannot enqueue query",
+    "environment: line 1: wait_for: No record of process", # Something weird from bash internals, let's just retry
     "is executing longer than distributed_ddl_task_timeout" # FIXME
 ]
 
@@ -116,7 +117,7 @@ def clickhouse_execute_http(base_args, query, timeout=30, settings=None, default
 def clickhouse_execute(base_args, query, timeout=30, settings=None):
     return clickhouse_execute_http(base_args, query, timeout, settings).strip()
 
-def clickhouse_execute_json(base_args, query, timeout=30, settings=None):
+def clickhouse_execute_json(base_args, query, timeout=60, settings=None):
     data = clickhouse_execute_http(base_args, query, timeout, settings, 'JSONEachRow')
     if not data:
         return None

tests/integration/test_global_overcommit_tracker/test.py

@@ -30,7 +30,7 @@ def test_overcommited_is_killed():
 
     responses_A = list()
     responses_B = list()
-    for _ in range(100):
+    for _ in range(500):
         responses_A.append(node.get_query_request(TEST_QUERY_A, user="A"))
         responses_B.append(node.get_query_request(TEST_QUERY_B, user="B"))
 

tests/integration/test_materialized_view_restart_server/test.py

@@ -0,0 +1,25 @@
+import pytest
+from helpers.cluster import ClickHouseCluster
+
+cluster = ClickHouseCluster(__file__)
+node = cluster.add_instance("node", stay_alive=True)
+
+
+@pytest.fixture(scope="module")
+def start_cluster():
+    try:
+        cluster.start()
+        yield cluster
+    finally:
+        cluster.shutdown()
+
+
+def test_materialized_view_with_subquery(start_cluster):
+    node.query("create table test (x UInt32) engine=TinyLog()")
+    node.query(
+        "create materialized view mv engine = TinyLog() as with subquery as (select * from test) select * from subquery"
+    )
+    node.restart_clickhouse(kill=True)
+    node.query("insert into test select 1")
+    result = node.query("select * from mv")
+    assert int(result) == 1

tests/integration/test_storage_s3/test.py

@@ -517,7 +517,7 @@ def test_put_get_with_globs(started_cluster):
         # ("'minio','minio123',",True), Redirect with credentials not working with nginx.
     ],
 )
-def test_multipart_put(started_cluster, maybe_auth, positive):
+def test_multipart(started_cluster, maybe_auth, positive):
     # type: (ClickHouseCluster) -> None
 
     bucket = (
@@ -535,8 +535,9 @@ def test_multipart_put(started_cluster, maybe_auth, positive):
 
     one_line_length = 6  # 3 digits, 2 commas, 1 line separator.
 
+    total_rows = csv_size_bytes // one_line_length
     # Generate data having size more than one part
-    int_data = [[1, 2, 3] for i in range(csv_size_bytes // one_line_length)]
+    int_data = [[1, 2, 3] for i in range(total_rows)]
     csv_data = "".join(["{},{},{}\n".format(x, y, z) for x, y, z in int_data])
 
     assert len(csv_data) > min_part_size_bytes
@@ -573,6 +574,37 @@ def test_multipart_put(started_cluster, maybe_auth, positive):
 
         assert csv_data == get_s3_file_content(started_cluster, bucket, filename)
 
+    # select uploaded data from many threads
+    select_query = (
+        "select sum(column1), sum(column2), sum(column3) "
+        "from s3('http://{host}:{port}/{bucket}/{filename}', {auth}'CSV', '{table_format}')".format(
+            host=started_cluster.minio_redirect_host,
+            port=started_cluster.minio_redirect_port,
+            bucket=bucket,
+            filename=filename,
+            auth=maybe_auth,
+            table_format=table_format,
+        )
+    )
+    try:
+        select_result = run_query(
+            instance,
+            select_query,
+            settings={
+                "max_download_threads": random.randint(4, 16),
+                "max_download_buffer_size": 1024 * 1024,
+            },
+        )
+    except helpers.client.QueryRuntimeException:
+        if positive:
+            raise
+    else:
+        assert positive
+        assert (
+            select_result
+            == "\t".join(map(str, [total_rows, total_rows * 2, total_rows * 3])) + "\n"
+        )
+
 
 def test_remote_host_filter(started_cluster):
     instance = started_cluster.instances["restricted_dummy"]

tests/performance/group_array_sorted.xml

@@ -0,0 +1,27 @@
+<test>
+    <settings>
+        <max_threads>10</max_threads>
+    </settings>
+    <substitutions>
+        <substitution>
+            <name>items</name>
+            <values>
+                <value>1000</value>
+                <value>100000</value>
+                <value>10000000</value>
+            </values>
+        </substitution>
+    </substitutions>
+
+    <create_query>CREATE TABLE test ( `id` UInt64, `value` UInt64, `text` String ) ENGINE = Memory</create_query>
+    <fill_query>INSERT INTO test SELECT number as id, rand64() as value, toString(number) as text FROM numbers({items})</fill_query>
+    <query>SELECT groupArraySorted(10)(id, value) FROM test</query>
+    <query>SELECT groupArraySorted(10)(text, value) FROM test</query>
+    <query>SELECT groupArraySorted(10)((id, text), value) FROM test</query>
+    <query>SELECT groupArraySorted(10)(text) FROM test</query>
+    <query>SELECT groupArraySorted(10000)(id, value) FROM test</query>
+    <query>SELECT groupArraySorted(10000)(text, value) FROM test</query>
+    <query>SELECT groupArraySorted(10000)((id, text), value) FROM test</query>
+    <query>SELECT groupArraySorted(10000)(text) FROM test</query>
+    <drop_query>DROP TABLE IF EXISTS test</drop_query>
+</test>

tests/performance/hash_table_sizes_stats.xml

@@ -0,0 +1,29 @@
+<test>
+    <preconditions>
+        <table_exists>hits_10m_single</table_exists>
+        <table_exists>hits_100m_single</table_exists>
+    </preconditions>
+
+    <settings>
+        <max_size_to_preallocate_for_aggregation>1000000000</max_size_to_preallocate_for_aggregation>
+    </settings>
+
+    <query>SELECT number FROM numbers(5000000) GROUP BY number FORMAT Null</query>
+    <query>SELECT number FROM numbers(10000000) GROUP BY number FORMAT Null</query>
+    <query short="1">SELECT number FROM numbers_mt(500000) GROUP BY number FORMAT Null</query>
+    <query short="1">SELECT number FROM numbers_mt(1000000) GROUP BY number FORMAT Null</query>
+    <query>SELECT number FROM numbers_mt(10000000) GROUP BY number FORMAT Null</query>
+    <query>SELECT number FROM numbers_mt(50000000) GROUP BY number FORMAT Null</query>
+    <query>WITH number % 524289 AS k, toUInt64(k) AS k1, k1 + 1 AS k2 SELECT k1, k2, count() FROM numbers(100000000) GROUP BY k1, k2 FORMAT Null</query>
+    <query>SELECT number FROM numbers_mt(10000000) GROUP BY number FORMAT Null SETTINGS group_by_two_level_threshold = 1e12, group_by_two_level_threshold_bytes = 1e12</query>
+    <query>SELECT number FROM numbers_mt(50000000) GROUP BY number FORMAT Null SETTINGS group_by_two_level_threshold = 1e12, group_by_two_level_threshold_bytes = 1e12</query>
+
+    <query>SELECT WatchID FROM hits_10m_single GROUP BY WatchID FORMAT Null</query>
+    <query>SELECT WatchID FROM hits_100m_single GROUP BY WatchID FORMAT Null</query>
+    <query>SELECT ClientIP AS x, x - 1, x - 2, x - 3, count() AS c FROM hits_10m_single GROUP BY x, x - 1, x - 2, x - 3 ORDER BY c DESC LIMIT 10</query>
+    <query>SELECT ClientIP AS x, x - 1, x - 2, x - 3, count() AS c FROM hits_100m_single GROUP BY x, x - 1, x - 2, x - 3 ORDER BY c DESC LIMIT 10</query>
+    <query>SELECT WatchID, ClientIP, count() AS c, sum(Refresh), avg(ResolutionWidth) FROM hits_10m_single WHERE SearchPhrase != '' GROUP BY WatchID, ClientIP ORDER BY c DESC LIMIT 10</query>
+    <query>SELECT WatchID, ClientIP, count() AS c, sum(Refresh), avg(ResolutionWidth) FROM hits_100m_single WHERE SearchPhrase != '' GROUP BY WatchID, ClientIP ORDER BY c DESC LIMIT 10</query>
+    <query>SELECT min(MobilePhoneModel) FROM hits_10m_single WHERE MobilePhoneModel != '' GROUP BY intHash32(UserID) % 1000000 FORMAT Null</query>
+    <query>SELECT min(MobilePhoneModel) FROM hits_100m_single WHERE MobilePhoneModel != '' GROUP BY intHash32(UserID) % 1000000 FORMAT Null</query>
+</test>

tests/queries/0_stateless/00385_storage_file_and_clickhouse-local_app_long.sh

@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# Tags: long
+# Tags: long, no-random-settings
 
 set -e
 

tests/queries/0_stateless/00417_kill_query.sh

@@ -21,3 +21,5 @@ $CLICKHOUSE_CLIENT -q "KILL QUERY WHERE 0 ASYNC"
 $CLICKHOUSE_CLIENT -q "KILL QUERY WHERE 0 FORMAT TabSeparated"
 $CLICKHOUSE_CLIENT -q "KILL QUERY WHERE 0 SYNC FORMAT TabSeparated"
 $CLICKHOUSE_CLIENT -q "KILL QUERY WHERE 1 TEST" &>/dev/null
+
+clickhouse_test_wait_queries 60

tests/queries/0_stateless/00484_preferred_max_column_in_block_size_bytes.sql

@@ -1,3 +1,5 @@
+-- Tags: no-random-settings
+
 drop table if exists tab_00484;
 create table tab_00484 (date Date, x UInt64, s FixedString(128)) engine = MergeTree PARTITION BY date ORDER BY (date, x) SETTINGS min_bytes_for_wide_part = 0;
 insert into tab_00484 select today(), number, toFixedString('', 128) from system.numbers limit 8192;

tests/queries/0_stateless/02122_parallel_formatting.sh

@@ -18,8 +18,8 @@ for format in ${formats}; do
     diff $non_parallel_file $parallel_file
 
     echo $format-2
-    $CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals limit 190000 format $format" --extremes=1 --output_format_parallel_formatting=0 --output_format_pretty_max_rows=1000000 | grep -a -v "elapsed" > $non_parallel_file
-    $CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals limit 190000 format $format" --extremes=1 --output_format_parallel_formatting=1 --output_format_pretty_max_rows=1000000 | grep -a -v "elapsed" > $parallel_file
+    $CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals order by number limit 190000 format $format" --extremes=1 --output_format_parallel_formatting=0 --output_format_pretty_max_rows=1000000 | grep -a -v "elapsed" > $non_parallel_file
+    $CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals order by number limit 190000 format $format" --extremes=1 --output_format_parallel_formatting=1 --output_format_pretty_max_rows=1000000 | grep -a -v "elapsed" > $parallel_file
 
     diff $non_parallel_file $parallel_file
 done
@@ -33,15 +33,17 @@ $CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(nu
 diff $non_parallel_file $parallel_file
 
 echo "CustomSeparated-2"
-$CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals limit 190000 format CustomSeparated $CUSTOM_SETTINGS" --output_format_parallel_formatting=0 --extremes=1 > $non_parallel_file
-$CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals limit 190000 format CustomSeparated $CUSTOM_SETTINGS" --output_format_parallel_formatting=1 --extremes=1 > $parallel_file
+$CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals order by number limit 190000 format CustomSeparated $CUSTOM_SETTINGS" --output_format_parallel_formatting=0 --extremes=1 > $non_parallel_file
+$CLICKHOUSE_CLIENT -q "select number, number + 1, concat('string: ', toString(number)) from numbers(200000) group by number with totals order by number limit 190000 format CustomSeparated $CUSTOM_SETTINGS" --output_format_parallel_formatting=1 --extremes=1 > $parallel_file
 
 diff $non_parallel_file $parallel_file
 
-echo -ne '{prefix} \n${data}\n $$ suffix $$\n' > "$CUR_DIR"/02122_template_format_resultset.tmp
-echo -ne 'x:${x:Quoted}, y:${y:Quoted}, s:${s:Quoted}' > "$CUR_DIR"/02122_template_format_row.tmp
+resultset_path=$CUR_DIR/$CLICKHOUSE_TEST_UNIQUE_NAME"_template_format_resultset.tmp"
+echo -ne '{prefix} \n${data}\n $$ suffix $$\n' > $resultset_path
+row_path=$CUR_DIR/$CLICKHOUSE_TEST_UNIQUE_NAME"_template_format_row.tmp"
+echo -ne 'x:${x:Quoted}, y:${y:Quoted}, s:${s:Quoted}' > $row_path
 
-TEMPLATE_SETTINGS="SETTINGS format_template_resultset = '$CUR_DIR/02122_template_format_resultset.tmp', format_template_row = '$CUR_DIR/02122_template_format_row.tmp', format_template_rows_between_delimiter = ';\n'"
+TEMPLATE_SETTINGS="SETTINGS format_template_resultset = '$resultset_path', format_template_row = '$row_path', format_template_rows_between_delimiter = ';\n'"
 
 echo "Template-1"
 $CLICKHOUSE_CLIENT -q "select number as x, number + 1 as y, concat('string: ', toString(number)) as s from numbers(200000) format Template $TEMPLATE_SETTINGS" --output_format_parallel_formatting=0 > $non_parallel_file
@@ -49,14 +51,14 @@ $CLICKHOUSE_CLIENT -q "select number as x, number + 1 as y, concat('string: ', t
 
 diff $non_parallel_file $parallel_file
 
-echo -ne '{prefix} \n${data}\n $$ suffix $$\n${totals}\n${min}\n${max}\n${rows:Quoted}\n${rows_before_limit:Quoted}\n${rows_read:Quoted}\n${bytes_read:Quoted}\n' > "$CUR_DIR"/02122_template_format_resultset.tmp
+echo -ne '{prefix} \n${data}\n $$ suffix $$\n${totals}\n${min}\n${max}\n${rows:Quoted}\n${rows_before_limit:Quoted}\n${rows_read:Quoted}\n${bytes_read:Quoted}\n' > $resultset_path
 
 echo "Template-2"
-$CLICKHOUSE_CLIENT -q "select number as x, number + 1 as y, concat('string: ', toString(number)) as s from numbers(200000) group by number with totals limit 190000 format Template $TEMPLATE_SETTINGS" --output_format_parallel_formatting=0 --extremes=1 > $non_parallel_file
-$CLICKHOUSE_CLIENT -q "select number as x, number + 1 as y, concat('string: ', toString(number)) as s from numbers(200000) group by number with totals limit 190000 format Template $TEMPLATE_SETTINGS" --output_format_parallel_formatting=1 --extremes=1 > $parallel_file
+$CLICKHOUSE_CLIENT -q "select number as x, number + 1 as y, concat('string: ', toString(number)) as s from numbers(200000) group by number with totals order by number limit 190000 format Template $TEMPLATE_SETTINGS" --output_format_parallel_formatting=0 --extremes=1 > $non_parallel_file
+$CLICKHOUSE_CLIENT -q "select number as x, number + 1 as y, concat('string: ', toString(number)) as s from numbers(200000) group by number with totals order by number limit 190000 format Template $TEMPLATE_SETTINGS" --output_format_parallel_formatting=1 --extremes=1 > $parallel_file
 
 diff $non_parallel_file $parallel_file
 
 rm $non_parallel_file $parallel_file
-rm "$CUR_DIR"/02122_template_format_resultset.tmp "$CUR_DIR"/02122_template_format_row.tmp
+rm $resultset_path $row_path
 

tests/queries/0_stateless/02151_hash_table_sizes_stats.reference

@@ -0,0 +1,21 @@
+1
+--
+1
+--
+1
+--
+1
+--
+1
+1
+--
+1
+--
+1
+1
+--
+1
+--
+1
+1
+--

tests/queries/0_stateless/02151_hash_table_sizes_stats.sh

@@ -0,0 +1,90 @@
+#!/usr/bin/env bash
+# Tags: long
+
+# shellcheck disable=SC2154
+
+unset CLICKHOUSE_LOG_COMMENT
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+# shellcheck source=../shell_config.sh
+. "$CURDIR"/../shell_config.sh
+
+
+# tests rely on that all the rows are unique and max_threads divides table_size
+table_size=10000
+max_threads=5
+
+
+prepare_table() {
+  table_name="t_hash_table_sizes_stats_$RANDOM$RANDOM"
+  $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS $table_name;"
+  if [ -z "$1" ]; then
+    $CLICKHOUSE_CLIENT -q "CREATE TABLE $table_name(number UInt64) Engine=MergeTree() ORDER BY tuple();"
+  else
+    $CLICKHOUSE_CLIENT -q "CREATE TABLE $table_name(number UInt64) Engine=MergeTree() ORDER BY $1;"
+  fi
+  $CLICKHOUSE_CLIENT -q "SYSTEM STOP MERGES $table_name;"
+  for ((i = 1; i <= max_threads; i++)); do
+    cnt=$((table_size / max_threads))
+    from=$(((i - 1) * cnt))
+    $CLICKHOUSE_CLIENT -q "INSERT INTO $table_name SELECT * FROM numbers($from, $cnt);"
+  done
+}
+
+prepare_table_with_sorting_key() {
+  prepare_table "$1"
+}
+
+run_query() {
+  query_id="${CLICKHOUSE_DATABASE}_hash_table_sizes_stats_$RANDOM$RANDOM"
+  $CLICKHOUSE_CLIENT --query_id="$query_id" --multiquery -q "
+    SET max_block_size = $((table_size / 10));
+    SET merge_tree_min_rows_for_concurrent_read = 1;
+    SET max_untracked_memory = 0;
+    SET max_size_to_preallocate_for_aggregation = 1e12;
+    $query"
+}
+
+check_preallocated_elements() {
+  $CLICKHOUSE_CLIENT -q "SYSTEM FLUSH LOGS"
+  # rows may be distributed in any way including "everything goes to the one particular thread"
+  min=$1
+  if [ -z "$2" ]; then
+    max=$1
+  else
+    max=$2
+  fi
+  $CLICKHOUSE_CLIENT --param_query_id="$query_id" -q "
+    SELECT COUNT(*)
+      FROM system.query_log
+     WHERE event_date >= yesterday() AND query_id = {query_id:String} AND current_database = currentDatabase()
+           AND ProfileEvents['AggregationPreallocatedElementsInHashTables'] BETWEEN $min AND $max"
+}
+
+check_convertion_to_two_level() {
+  $CLICKHOUSE_CLIENT -q "SYSTEM FLUSH LOGS"
+  # rows may be distributed in any way including "everything goes to the one particular thread"
+  $CLICKHOUSE_CLIENT --param_query_id="$query_id" -q "
+    SELECT SUM(ProfileEvents['AggregationHashTablesInitializedAsTwoLevel']) BETWEEN 1 AND $max_threads
+      FROM system.query_log
+     WHERE event_date >= yesterday() AND query_id = {query_id:String} AND current_database = currentDatabase()"
+}
+
+print_border() {
+  echo "--"
+}
+
+
+# shellcheck source=./02151_hash_table_sizes_stats.testcases
+source "$CURDIR"/02151_hash_table_sizes_stats.testcases
+
+
+test_one_thread_simple_group_by
+test_one_thread_simple_group_by_with_limit
+test_one_thread_simple_group_by_with_join_and_subquery
+test_several_threads_simple_group_by_with_limit_single_level_ht
+test_several_threads_simple_group_by_with_limit_two_level_ht
+test_several_threads_simple_group_by_with_limit_and_rollup_single_level_ht
+test_several_threads_simple_group_by_with_limit_and_rollup_two_level_ht
+test_several_threads_simple_group_by_with_limit_and_cube_single_level_ht
+test_several_threads_simple_group_by_with_limit_and_cube_two_level_ht

tests/queries/0_stateless/02151_hash_table_sizes_stats.testcases

@@ -0,0 +1,195 @@
+test_one_thread_simple_group_by() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint --
+    SELECT number
+      FROM $table_name
+  GROUP BY number
+  SETTINGS max_threads = 1
+    FORMAT Null;"
+
+  run_query
+  run_query
+  check_preallocated_elements $expected_size_hint
+  print_border
+}
+
+test_one_thread_simple_group_by_with_limit() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+    SELECT number
+      FROM $table_name
+  GROUP BY number
+     LIMIT 5
+  SETTINGS max_threads = 1
+    FORMAT Null;"
+
+  run_query
+  run_query
+  check_preallocated_elements $expected_size_hint
+  print_border
+}
+
+test_one_thread_simple_group_by_with_join_and_subquery() {
+  expected_size_hint=$((table_size + table_size / 2))
+  prepare_table
+
+  query="
+  -- expected two size_hints for different keys: for the inner ($table_size) and the outer aggregation ($((table_size / 2)))
+    SELECT number
+      FROM $table_name AS t1
+      JOIN
+      (
+        SELECT number
+          FROM $table_name AS t2
+      GROUP BY number
+         LIMIT $((table_size / 2))
+      ) AS t3 USING(number)
+  GROUP BY number
+  SETTINGS max_threads = 1,
+           distributed_product_mode = 'local'
+    FORMAT Null;"
+
+  run_query
+  run_query
+  check_preallocated_elements $expected_size_hint
+  print_border
+}
+
+test_several_threads_simple_group_by_with_limit_single_level_ht() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+    SELECT number
+      FROM $table_name
+  GROUP BY number
+     LIMIT 5
+  SETTINGS max_threads = $max_threads,
+           group_by_two_level_threshold = $((expected_size_hint + 1)),
+           group_by_two_level_threshold_bytes = $((table_size * 1000))
+    FORMAT Null;"
+
+  run_query
+  run_query
+  check_preallocated_elements $((expected_size_hint / max_threads)) $((expected_size_hint * max_threads))
+  print_border
+}
+
+test_several_threads_simple_group_by_with_limit_two_level_ht() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+    SELECT number
+      FROM $table_name
+  GROUP BY number
+     LIMIT 5
+  SETTINGS max_threads = $max_threads,
+           group_by_two_level_threshold = $expected_size_hint,
+           group_by_two_level_threshold_bytes = $((table_size * 1000))
+    FORMAT Null;"
+
+  run_query
+  run_query
+  check_convertion_to_two_level
+  check_preallocated_elements $((expected_size_hint / max_threads)) $((expected_size_hint * max_threads))
+  print_border
+}
+
+test_several_threads_simple_group_by_with_limit_and_rollup_single_level_ht() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+       SELECT number
+         FROM $table_name
+     GROUP BY number
+  WITH ROLLUP
+        LIMIT 5
+     SETTINGS max_threads = $max_threads,
+              group_by_two_level_threshold = $((expected_size_hint + 1)),
+              group_by_two_level_threshold_bytes = $((table_size * 1000))
+       FORMAT Null;"
+
+  run_query
+  run_query
+  check_preallocated_elements $((expected_size_hint / max_threads)) $((expected_size_hint * max_threads))
+  print_border
+}
+
+test_several_threads_simple_group_by_with_limit_and_rollup_two_level_ht() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+       SELECT number
+         FROM $table_name
+     GROUP BY number
+  WITH ROLLUP
+        LIMIT 5
+     SETTINGS max_threads = $max_threads,
+              group_by_two_level_threshold = $expected_size_hint,
+              group_by_two_level_threshold_bytes = $((table_size * 1000))
+       FORMAT Null;"
+
+  run_query
+  run_query
+  check_convertion_to_two_level
+  check_preallocated_elements $((expected_size_hint / max_threads)) $((expected_size_hint * max_threads))
+  print_border
+}
+
+test_several_threads_simple_group_by_with_limit_and_cube_single_level_ht() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+     SELECT number
+       FROM $table_name
+   GROUP BY number
+  WITH CUBE
+      LIMIT 5
+   SETTINGS max_threads = $max_threads,
+            group_by_two_level_threshold = $((expected_size_hint + 1)),
+            group_by_two_level_threshold_bytes = $((table_size * 1000))
+     FORMAT Null;"
+
+  run_query
+  run_query
+  check_preallocated_elements $((expected_size_hint / max_threads)) $((expected_size_hint * max_threads))
+  print_border
+}
+
+test_several_threads_simple_group_by_with_limit_and_cube_two_level_ht() {
+  expected_size_hint=$table_size
+  prepare_table
+
+  query="
+  -- size_hint = $expected_size_hint despite the presence of limit --
+     SELECT number
+       FROM $table_name
+   GROUP BY number
+  WITH CUBE
+      LIMIT 5
+   SETTINGS max_threads = $max_threads,
+            group_by_two_level_threshold = $expected_size_hint,
+            group_by_two_level_threshold_bytes = $((table_size * 1000))
+     FORMAT Null;"
+
+  run_query
+  run_query
+  check_convertion_to_two_level
+  check_preallocated_elements $((expected_size_hint / max_threads)) $((expected_size_hint * max_threads))
+  print_border
+}

tests/queries/0_stateless/02151_hash_table_sizes_stats_distributed.reference

@@ -0,0 +1,33 @@
+1
+1
+--
+1
+1
+--
+1
+1
+--
+1
+1
+--
+1
+1
+1
+1
+--
+1
+1
+--
+1
+1
+1
+1
+--
+1
+1
+--
+1
+1
+1
+1
+--

tests/queries/0_stateless/02151_hash_table_sizes_stats_distributed.sh

@@ -0,0 +1,97 @@
+#!/usr/bin/env bash
+# Tags: long, distributed
+
+# These tests don't use `current_database = currentDatabase()` condition, because database name isn't propagated during remote queries.
+
+# shellcheck disable=SC2154
+
+unset CLICKHOUSE_LOG_COMMENT
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+# shellcheck source=../shell_config.sh
+. "$CURDIR"/../shell_config.sh
+
+
+# tests rely on that all the rows are unique and max_threads divides table_size
+table_size=10000
+max_threads=5
+
+
+prepare_table() {
+  table_name="t_hash_table_sizes_stats_$RANDOM$RANDOM"
+  $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS $table_name;"
+  if [ -z "$1" ]; then
+    $CLICKHOUSE_CLIENT -q "CREATE TABLE $table_name(number UInt64) Engine=MergeTree() ORDER BY tuple();"
+  else
+    $CLICKHOUSE_CLIENT -q "CREATE TABLE $table_name(number UInt64) Engine=MergeTree() ORDER BY $1;"
+  fi
+  $CLICKHOUSE_CLIENT -q "SYSTEM STOP MERGES $table_name;"
+  for ((i = 1; i <= max_threads; i++)); do
+    cnt=$((table_size / max_threads))
+    from=$(((i - 1) * cnt))
+    $CLICKHOUSE_CLIENT -q "INSERT INTO $table_name SELECT * FROM numbers($from, $cnt);"
+  done
+  $CLICKHOUSE_CLIENT -q "DROP TABLE IF EXISTS ${table_name}_d;"
+  $CLICKHOUSE_CLIENT -q "CREATE TABLE ${table_name}_d AS $table_name ENGINE = Distributed(test_cluster_two_shards, currentDatabase(), $table_name);"
+  table_name="${table_name}_d"
+}
+
+prepare_table_with_sorting_key() {
+  prepare_table "$1"
+}
+
+run_query() {
+  query_id="${CLICKHOUSE_DATABASE}_hash_table_sizes_stats_$RANDOM$RANDOM"
+  $CLICKHOUSE_CLIENT --query_id="$query_id" --multiquery -q "
+    SET max_block_size = $((table_size / 10));
+    SET merge_tree_min_rows_for_concurrent_read = 1;
+    SET max_untracked_memory = 0;
+    SET prefer_localhost_replica = 1;
+    $query"
+}
+
+check_preallocated_elements() {
+  $CLICKHOUSE_CLIENT -q "SYSTEM FLUSH LOGS"
+  # rows may be distributed in any way including "everything goes to the one particular thread"
+  min=$1
+  if [ -z "$2" ]; then
+    max=$1
+  else
+    max=$2
+  fi
+  $CLICKHOUSE_CLIENT --param_query_id="$query_id" -q "
+    SELECT COUNT(*)
+      FROM system.query_log
+     WHERE event_date >= yesterday() AND (query_id = {query_id:String} OR initial_query_id = {query_id:String})
+           AND ProfileEvents['AggregationPreallocatedElementsInHashTables'] BETWEEN $min AND $max
+  GROUP BY query_id"
+}
+
+check_convertion_to_two_level() {
+  $CLICKHOUSE_CLIENT -q "SYSTEM FLUSH LOGS"
+  # rows may be distributed in any way including "everything goes to the one particular thread"
+  $CLICKHOUSE_CLIENT --param_query_id="$query_id" -q "
+    SELECT SUM(ProfileEvents['AggregationHashTablesInitializedAsTwoLevel']) BETWEEN 1 AND $max_threads
+      FROM system.query_log
+     WHERE event_date >= yesterday() AND (query_id = {query_id:String} OR initial_query_id = {query_id:String})
+  GROUP BY query_id"
+}
+
+print_border() {
+  echo "--"
+}
+
+
+# shellcheck source=./02151_hash_table_sizes_stats.testcases
+source "$CURDIR"/02151_hash_table_sizes_stats.testcases
+
+
+test_one_thread_simple_group_by
+test_one_thread_simple_group_by_with_limit
+test_one_thread_simple_group_by_with_join_and_subquery
+test_several_threads_simple_group_by_with_limit_single_level_ht
+test_several_threads_simple_group_by_with_limit_two_level_ht
+test_several_threads_simple_group_by_with_limit_and_rollup_single_level_ht
+test_several_threads_simple_group_by_with_limit_and_rollup_two_level_ht
+test_several_threads_simple_group_by_with_limit_and_cube_single_level_ht
+test_several_threads_simple_group_by_with_limit_and_cube_two_level_ht

tests/queries/0_stateless/02152_http_external_tables_memory_tracking.sh

@@ -23,7 +23,7 @@ function run_and_check()
     echo "Checking $*"
 
     # Run query with external table (implicit StorageMemory user)
-    $CLICKHOUSE_CURL -sS -F "s=@$tmp_file;" "$CLICKHOUSE_URL&s_structure=key+Int&query=SELECT+count()+FROM+s&memory_profiler_sample_probability=1&query_id=$query_id&$*" -o /dev/null
+    $CLICKHOUSE_CURL -sS -F "s=@$tmp_file;" "$CLICKHOUSE_URL&s_structure=key+Int&query=SELECT+count()+FROM+s&memory_profiler_sample_probability=1&max_untracked_memory=0&query_id=$query_id&$*" -o /dev/null
 
     ${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}" --data-binary @- <<<'SYSTEM FLUSH LOGS'
 

tests/queries/0_stateless/02158_grouparraysorted.reference

@@ -0,0 +1,18 @@
+[0,1,2,3,4]
+[0,1,2,3,4,5,6,7,8,9]
+[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99]
+[999,998,997,996,995,994,993,992,991,990,989,988,987,986,985,984,983,982,981,980,979,978,977,976,975,974,973,972,971,970,969,968,967,966,965,964,963,962,961,960,959,958,957,956,955,954,953,952,951,950,949,948,947,946,945,944,943,942,941,940,939,938,937,936,935,934,933,932,931,930,929,928,927,926,925,924,923,922,921,920,919,918,917,916,915,914,913,912,911,910,909,908,907,906,905,904,903,902,901,900]
+['0','1','2','3','4']
+['0','1','2','3','4']
+['9','8','7','6','5']
+[(0,'0'),(1,'1'),(2,'2'),(3,'3'),(4,'4')]
+['0','1','10','11','12']
+['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','34','35','36','37','38','39','40','41','42','43','44','45','46','47','48','49']
+[0,0,1,1,2,2,3,3,4,4]
+pablo	[1,2]
+luis	[1,3]
+pablo	[1,2]
+luis	[1,3]
+[4,5,6,7,8]
+[10,11,12,13,14]
+['10','11','12','13','14']

tests/queries/0_stateless/02158_grouparraysorted.sql

@@ -0,0 +1,43 @@
+
+SELECT groupArraySorted(5)(number) from numbers(100);
+
+SELECT groupArraySorted(number, number) from numbers(100);
+
+SELECT groupArraySorted(100)(number, number) from numbers(1000);
+
+SELECT groupArraySorted(100)(number, -number) from numbers(1000);
+
+SELECT groupArraySorted(5)(str, number) FROM (SELECT toString(number) as str, number FROM numbers(10));
+
+SELECT groupArraySorted(5)(text) FROM (select toString(number) as text from numbers(10));
+
+SELECT groupArraySorted(5)(text, -number) FROM (select toString(number) as text, number from numbers(10));
+
+SELECT groupArraySorted(5)((number,text)) from (SELECT toString(number) as text, number FROM numbers(100));
+
+SELECT groupArraySorted(5)(text,text) from (SELECT toString(number) as text FROM numbers(100));
+
+SELECT groupArraySorted(50)(text,(number,text)) from (SELECT toString(number) as text, number FROM numbers(100));
+
+SELECT groupArraySorted(10)(toInt64(number/2)) FROM numbers(100);
+
+
+DROP TABLE IF EXISTS test;
+DROP VIEW IF EXISTS mv_test;
+CREATE TABLE test (`n` String, `h` Int64) ENGINE = MergeTree ORDER BY n;
+CREATE MATERIALIZED VIEW mv_test (`n` String, `h` AggregateFunction(groupArraySorted(2), Int64, Int64)) ENGINE = AggregatingMergeTree ORDER BY n AS SELECT n, groupArraySortedState(2)(h, h) as h FROM test GROUP BY n;
+INSERT INTO test VALUES ('pablo',1)('pablo', 2)('luis', 1)('luis', 3)('pablo', 5)('pablo',4)('pablo', 5)('luis', 6)('luis', 7)('pablo', 8)('pablo',9)('pablo',10)('luis',11)('luis',12)('pablo',13);
+SELECT n, groupArraySortedMerge(2)(h) from mv_test GROUP BY n;
+
+DROP TABLE IF EXISTS test;
+DROP VIEW IF EXISTS mv_test;
+CREATE TABLE test (`n` String, `h` Int64) ENGINE = MergeTree ORDER BY n;
+CREATE MATERIALIZED VIEW mv_test (`n` String, `h` AggregateFunction(groupArraySorted(2), Int64)) ENGINE = AggregatingMergeTree ORDER BY n AS SELECT n, groupArraySortedState(2)(h) as h FROM test GROUP BY n;
+INSERT INTO test VALUES ('pablo',1)('pablo', 2)('luis', 1)('luis', 3)('pablo', 5)('pablo',4)('pablo', 5)('luis', 6)('luis', 7)('pablo', 8)('pablo',9)('pablo',10)('luis',11)('luis',12)('pablo',13);
+SELECT n, groupArraySortedMerge(2)(h) from mv_test GROUP BY n;
+DROP TABLE test;
+DROP VIEW mv_test;
+
+SELECT groupArraySortedIf(5)(number, number, number>3) from numbers(100);
+SELECT groupArraySortedIf(5)(number, toString(number), number>3) from numbers(100);
+SELECT groupArraySortedIf(5)(toString(number), number>3) from numbers(100);

tests/queries/0_stateless/02167_format_from_file_extension.sh

@@ -5,6 +5,13 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 # shellcheck source=../shell_config.sh
 . "$CURDIR"/../shell_config.sh
 
+function cleanup()
+{
+    # this command expects an error message like 'Code: 107. DB::Exception: Received <...> nonexist.txt doesn't exist. (FILE_DOESNT_EXIST)'
+    user_files_path=$($CLICKHOUSE_CLIENT --query "select _path,_file from file('nonexist.txt', 'CSV', 'val1 char')" 2>&1 | grep -E '^Code: 107.*FILE_DOESNT_EXIST' | head -1 | awk '{gsub("/nonexist.txt","",$9); print $9}')
+    rm $user_files_path/test_02167.*
+}
+trap cleanup EXIT
 
 for format in TSV TabSeparated TSVWithNames TSVWithNamesAndTypes CSV Parquet ORC Arrow JSONEachRow JSONCompactEachRow CustomSeparatedWithNamesAndTypes
 do

tests/queries/0_stateless/02245_format_string_stack_overflow.reference

@@ -0,0 +1 @@
+00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000

tests/queries/0_stateless/02248_nullable_custom_types_to_string.reference

@@ -0,0 +1,6 @@
+true
+\N
+0.0.0.0
+\N
+::ffff:127.0.0.1
+\N

tests/queries/0_stateless/02248_nullable_custom_types_to_string.sql

@@ -0,0 +1,6 @@
+select toString(toNullable(true));
+select toString(CAST(NULL, 'Nullable(Bool)'));
+select toString(toNullable(toIPv4('0.0.0.0')));
+select toString(CAST(NULL, 'Nullable(IPv4)'));
+select toString(toNullable(toIPv6('::ffff:127.0.0.1')));
+select toString(CAST(NULL, 'Nullable(IPv6)'));

tests/queries/1_stateful/00140_rename.sql

@@ -1,4 +1,4 @@
--- Tags: no-replicated-database
+-- Tags: no-replicated-database, no-parallel
 -- Tag no-replicated-database: Does not support renaming of multiple tables in single query
 
 RENAME TABLE test.hits TO test.visits_tmp, test.visits TO test.hits, test.visits_tmp TO test.visits;

tests/queries/1_stateful/00152_insert_different_granularity.sql

@@ -1,4 +1,4 @@
--- Tags: no-tsan, no-replicated-database
+-- Tags: no-tsan, no-replicated-database, no-parallel
 -- Tag no-replicated-database: Fails due to additional replicas or shards
 
 DROP TABLE IF EXISTS fixed_granularity_table;

tests/queries/1_stateful/00170_s3_cache.reference

@@ -1,4 +1,5 @@
 -- { echo }
+
 SET max_memory_usage='20G';
 SELECT count() FROM test.hits_s3;
 8873898

tests/queries/1_stateful/00170_s3_cache.sql

@@ -1,4 +1,7 @@
+-- Tags: no-parallel
+
 -- { echo }
+
 SET max_memory_usage='20G';
 SELECT count() FROM test.hits_s3;
 SELECT count() FROM test.hits_s3 WHERE AdvEngineID != 0;

utils/github/__init__.py

@@ -1 +0,0 @@
-# -*- coding: utf-8 -*-

utils/github/backport.py

@@ -1,185 +0,0 @@
-# -*- coding: utf-8 -*-
-
-try:
-    from clickhouse.utils.github.cherrypick import CherryPick
-    from clickhouse.utils.github.query import Query as RemoteRepo
-    from clickhouse.utils.github.local import Repository as LocalRepo
-except:
-    from .cherrypick import CherryPick
-    from .query import Query as RemoteRepo
-    from .local import Repository as LocalRepo
-
-import argparse
-import logging
-import re
-import sys
-
-
-class Backport:
-    def __init__(self, token, owner, name, team):
-        self._gh = RemoteRepo(
-            token, owner=owner, name=name, team=team, max_page_size=30, min_page_size=7
-        )
-        self._token = token
-        self.default_branch_name = self._gh.default_branch
-        self.ssh_url = self._gh.ssh_url
-
-    def getPullRequests(self, from_commit):
-        return self._gh.get_pull_requests(from_commit)
-
-    def getBranchesWithRelease(self):
-        branches = set()
-        for pull_request in self._gh.find_pull_requests("release"):
-            branches.add(pull_request["headRefName"])
-        return branches
-
-    def execute(self, repo, upstream, until_commit, run_cherrypick):
-        repo = LocalRepo(repo, upstream, self.default_branch_name)
-        all_branches = repo.get_release_branches()  # [(branch_name, base_commit)]
-
-        release_branches = self.getBranchesWithRelease()
-
-        branches = []
-        # iterate over all branches to preserve their precedence.
-        for branch in all_branches:
-            if branch[0] in release_branches:
-                branches.append(branch)
-
-        if not branches:
-            logging.info("No release branches found!")
-            return
-
-        for branch in branches:
-            logging.info("Found release branch: %s", branch[0])
-
-        if not until_commit:
-            until_commit = branches[0][1]
-        pull_requests = self.getPullRequests(until_commit)
-
-        backport_map = {}
-
-        RE_MUST_BACKPORT = re.compile(r"^v(\d+\.\d+)-must-backport$")
-        RE_NO_BACKPORT = re.compile(r"^v(\d+\.\d+)-no-backport$")
-        RE_BACKPORTED = re.compile(r"^v(\d+\.\d+)-backported$")
-
-        # pull-requests are sorted by ancestry from the most recent.
-        for pr in pull_requests:
-            while repo.comparator(branches[-1][1]) >= repo.comparator(
-                pr["mergeCommit"]["oid"]
-            ):
-                logging.info(
-                    "PR #{} is already inside {}. Dropping this branch for further PRs".format(
-                        pr["number"], branches[-1][0]
-                    )
-                )
-                branches.pop()
-
-            logging.info("Processing PR #{}".format(pr["number"]))
-
-            assert len(branches)
-
-            branch_set = set([branch[0] for branch in branches])
-
-            # First pass. Find all must-backports
-            for label in pr["labels"]["nodes"]:
-                if label["name"] == "pr-must-backport":
-                    backport_map[pr["number"]] = branch_set.copy()
-                    continue
-                matched = RE_MUST_BACKPORT.match(label["name"])
-                if matched:
-                    if pr["number"] not in backport_map:
-                        backport_map[pr["number"]] = set()
-                    backport_map[pr["number"]].add(matched.group(1))
-
-            # Second pass. Find all no-backports
-            for label in pr["labels"]["nodes"]:
-                if label["name"] == "pr-no-backport" and pr["number"] in backport_map:
-                    del backport_map[pr["number"]]
-                    break
-                matched_no_backport = RE_NO_BACKPORT.match(label["name"])
-                matched_backported = RE_BACKPORTED.match(label["name"])
-                if (
-                    matched_no_backport
-                    and pr["number"] in backport_map
-                    and matched_no_backport.group(1) in backport_map[pr["number"]]
-                ):
-                    backport_map[pr["number"]].remove(matched_no_backport.group(1))
-                    logging.info(
-                        "\tskipping %s because of forced no-backport",
-                        matched_no_backport.group(1),
-                    )
-                elif (
-                    matched_backported
-                    and pr["number"] in backport_map
-                    and matched_backported.group(1) in backport_map[pr["number"]]
-                ):
-                    backport_map[pr["number"]].remove(matched_backported.group(1))
-                    logging.info(
-                        "\tskipping %s because it's already backported manually",
-                        matched_backported.group(1),
-                    )
-
-        for pr, branches in list(backport_map.items()):
-            logging.info("PR #%s needs to be backported to:", pr)
-            for branch in branches:
-                logging.info(
-                    "\t%s, and the status is: %s",
-                    branch,
-                    run_cherrypick(self._token, pr, branch),
-                )
-
-        # print API costs
-        logging.info("\nGitHub API total costs per query:")
-        for name, value in list(self._gh.api_costs.items()):
-            logging.info("%s : %s", name, value)
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        "--token", type=str, required=True, help="token for Github access"
-    )
-    parser.add_argument(
-        "--repo",
-        type=str,
-        required=True,
-        help="path to full repository",
-        metavar="PATH",
-    )
-    parser.add_argument(
-        "--til", type=str, help="check PRs from HEAD til this commit", metavar="COMMIT"
-    )
-    parser.add_argument(
-        "--dry-run",
-        action="store_true",
-        help="do not create or merge any PRs",
-        default=False,
-    )
-    parser.add_argument(
-        "--verbose",
-        "-v",
-        action="store_true",
-        help="more verbose output",
-        default=False,
-    )
-    parser.add_argument(
-        "--upstream",
-        "-u",
-        type=str,
-        help="remote name of upstream in repository",
-        default="origin",
-    )
-    args = parser.parse_args()
-
-    if args.verbose:
-        logging.basicConfig(
-            format="%(message)s", stream=sys.stdout, level=logging.DEBUG
-        )
-    else:
-        logging.basicConfig(format="%(message)s", stream=sys.stdout, level=logging.INFO)
-
-    cherrypick_run = lambda token, pr, branch: CherryPick(
-        token, "ClickHouse", "ClickHouse", "core", pr, branch
-    ).execute(args.repo, args.dry_run)
-    bp = Backport(args.token, "ClickHouse", "ClickHouse", "core")
-    bp.execute(args.repo, args.upstream, args.til, cherrypick_run)

utils/github/cherrypick.py

@@ -1,323 +0,0 @@
-# -*- coding: utf-8 -*-
-
-"""
-Backports changes from PR to release branch.
-Requires multiple separate runs as part of the implementation.
-
-First run should do the following:
-1. Merge release branch with a first parent of merge-commit of PR (using 'ours' strategy). (branch: backport/{branch}/{pr})
-2. Create temporary branch over merge-commit to use it for PR creation. (branch: cherrypick/{merge_commit})
-3. Create PR from temporary branch to backport branch (emulating cherry-pick).
-
-Second run checks PR from previous run to be merged or at least being mergeable. If it's not merged then try to merge it.
-
-Third run creates PR from backport branch (with merged previous PR) to release branch.
-"""
-
-try:
-    from clickhouse.utils.github.query import Query as RemoteRepo
-except:
-    from .query import Query as RemoteRepo
-
-import argparse
-from enum import Enum
-import logging
-import os
-import subprocess
-import sys
-
-
-class CherryPick:
-    class Status(Enum):
-        DISCARDED = "discarded"
-        NOT_INITIATED = "not started"
-        FIRST_MERGEABLE = "waiting for 1st stage"
-        FIRST_CONFLICTS = "conflicts on 1st stage"
-        SECOND_MERGEABLE = "waiting for 2nd stage"
-        SECOND_CONFLICTS = "conflicts on 2nd stage"
-        MERGED = "backported"
-
-    def _run(self, args):
-        out = subprocess.check_output(args).rstrip()
-        logging.debug(out)
-        return out
-
-    def __init__(self, token, owner, name, team, pr_number, target_branch):
-        self._gh = RemoteRepo(token, owner=owner, name=name, team=team)
-        self._pr = self._gh.get_pull_request(pr_number)
-
-        self.ssh_url = self._gh.ssh_url
-
-        # TODO: check if pull-request is merged.
-
-        self.merge_commit_oid = self._pr["mergeCommit"]["oid"]
-
-        self.target_branch = target_branch
-        self.backport_branch = "backport/{branch}/{pr}".format(
-            branch=target_branch, pr=pr_number
-        )
-        self.cherrypick_branch = "cherrypick/{branch}/{oid}".format(
-            branch=target_branch, oid=self.merge_commit_oid
-        )
-
-    def getCherryPickPullRequest(self):
-        return self._gh.find_pull_request(
-            base=self.backport_branch, head=self.cherrypick_branch
-        )
-
-    def createCherryPickPullRequest(self, repo_path):
-        DESCRIPTION = (
-            "This pull-request is a first step of an automated backporting.\n"
-            "It contains changes like after calling a local command `git cherry-pick`.\n"
-            "If you intend to continue backporting this changes, then resolve all conflicts if any.\n"
-            "Otherwise, if you do not want to backport them, then just close this pull-request.\n"
-            "\n"
-            "The check results does not matter at this step - you can safely ignore them.\n"
-            "Also this pull-request will be merged automatically as it reaches the mergeable state, but you always can merge it manually.\n"
-        )
-
-        # FIXME: replace with something better than os.system()
-        git_prefix = [
-            "git",
-            "-C",
-            repo_path,
-            "-c",
-            "user.email=robot-clickhouse@yandex-team.ru",
-            "-c",
-            "user.name=robot-clickhouse",
-        ]
-        base_commit_oid = self._pr["mergeCommit"]["parents"]["nodes"][0]["oid"]
-
-        # Create separate branch for backporting, and make it look like real cherry-pick.
-        self._run(git_prefix + ["checkout", "-f", self.target_branch])
-        self._run(git_prefix + ["checkout", "-B", self.backport_branch])
-        self._run(git_prefix + ["merge", "-s", "ours", "--no-edit", base_commit_oid])
-
-        # Create secondary branch to allow pull request with cherry-picked commit.
-        self._run(
-            git_prefix + ["branch", "-f", self.cherrypick_branch, self.merge_commit_oid]
-        )
-
-        self._run(
-            git_prefix
-            + [
-                "push",
-                "-f",
-                "origin",
-                "{branch}:{branch}".format(branch=self.backport_branch),
-            ]
-        )
-        self._run(
-            git_prefix
-            + [
-                "push",
-                "-f",
-                "origin",
-                "{branch}:{branch}".format(branch=self.cherrypick_branch),
-            ]
-        )
-
-        # Create pull-request like a local cherry-pick
-        pr = self._gh.create_pull_request(
-            source=self.cherrypick_branch,
-            target=self.backport_branch,
-            title="Cherry pick #{number} to {target}: {title}".format(
-                number=self._pr["number"],
-                target=self.target_branch,
-                title=self._pr["title"].replace('"', '\\"'),
-            ),
-            description="Original pull-request #{}\n\n{}".format(
-                self._pr["number"], DESCRIPTION
-            ),
-        )
-
-        # FIXME: use `team` to leave a single eligible assignee.
-        self._gh.add_assignee(pr, self._pr["author"])
-        self._gh.add_assignee(pr, self._pr["mergedBy"])
-
-        self._gh.set_label(pr, "do not test")
-        self._gh.set_label(pr, "pr-cherrypick")
-
-        return pr
-
-    def mergeCherryPickPullRequest(self, cherrypick_pr):
-        return self._gh.merge_pull_request(cherrypick_pr["id"])
-
-    def getBackportPullRequest(self):
-        return self._gh.find_pull_request(
-            base=self.target_branch, head=self.backport_branch
-        )
-
-    def createBackportPullRequest(self, cherrypick_pr, repo_path):
-        DESCRIPTION = (
-            "This pull-request is a last step of an automated backporting.\n"
-            "Treat it as a standard pull-request: look at the checks and resolve conflicts.\n"
-            "Merge it only if you intend to backport changes to the target branch, otherwise just close it.\n"
-        )
-
-        git_prefix = [
-            "git",
-            "-C",
-            repo_path,
-            "-c",
-            "user.email=robot-clickhouse@clickhouse.com",
-            "-c",
-            "user.name=robot-clickhouse",
-        ]
-
-        pr_title = "Backport #{number} to {target}: {title}".format(
-            number=self._pr["number"],
-            target=self.target_branch,
-            title=self._pr["title"].replace('"', '\\"'),
-        )
-
-        self._run(git_prefix + ["checkout", "-f", self.backport_branch])
-        self._run(git_prefix + ["pull", "--ff-only", "origin", self.backport_branch])
-        self._run(
-            git_prefix
-            + [
-                "reset",
-                "--soft",
-                self._run(
-                    git_prefix
-                    + [
-                        "merge-base",
-                        "origin/" + self.target_branch,
-                        self.backport_branch,
-                    ]
-                ),
-            ]
-        )
-        self._run(git_prefix + ["commit", "-a", "--allow-empty", "-m", pr_title])
-        self._run(
-            git_prefix
-            + [
-                "push",
-                "-f",
-                "origin",
-                "{branch}:{branch}".format(branch=self.backport_branch),
-            ]
-        )
-
-        pr = self._gh.create_pull_request(
-            source=self.backport_branch,
-            target=self.target_branch,
-            title=pr_title,
-            description="Original pull-request #{}\nCherry-pick pull-request #{}\n\n{}".format(
-                self._pr["number"], cherrypick_pr["number"], DESCRIPTION
-            ),
-        )
-
-        # FIXME: use `team` to leave a single eligible assignee.
-        self._gh.add_assignee(pr, self._pr["author"])
-        self._gh.add_assignee(pr, self._pr["mergedBy"])
-
-        self._gh.set_label(pr, "pr-backport")
-
-        return pr
-
-    def execute(self, repo_path, dry_run=False):
-        pr1 = self.getCherryPickPullRequest()
-        if not pr1:
-            if not dry_run:
-                pr1 = self.createCherryPickPullRequest(repo_path)
-                logging.debug(
-                    "Created PR with cherry-pick of %s to %s: %s",
-                    self._pr["number"],
-                    self.target_branch,
-                    pr1["url"],
-                )
-            else:
-                return CherryPick.Status.NOT_INITIATED
-        else:
-            logging.debug(
-                "Found PR with cherry-pick of %s to %s: %s",
-                self._pr["number"],
-                self.target_branch,
-                pr1["url"],
-            )
-
-        if not pr1["merged"] and pr1["mergeable"] == "MERGEABLE" and not pr1["closed"]:
-            if not dry_run:
-                pr1 = self.mergeCherryPickPullRequest(pr1)
-                logging.debug(
-                    "Merged PR with cherry-pick of %s to %s: %s",
-                    self._pr["number"],
-                    self.target_branch,
-                    pr1["url"],
-                )
-
-        if not pr1["merged"]:
-            logging.debug(
-                "Waiting for PR with cherry-pick of %s to %s: %s",
-                self._pr["number"],
-                self.target_branch,
-                pr1["url"],
-            )
-
-            if pr1["closed"]:
-                return CherryPick.Status.DISCARDED
-            elif pr1["mergeable"] == "CONFLICTING":
-                return CherryPick.Status.FIRST_CONFLICTS
-            else:
-                return CherryPick.Status.FIRST_MERGEABLE
-
-        pr2 = self.getBackportPullRequest()
-        if not pr2:
-            if not dry_run:
-                pr2 = self.createBackportPullRequest(pr1, repo_path)
-                logging.debug(
-                    "Created PR with backport of %s to %s: %s",
-                    self._pr["number"],
-                    self.target_branch,
-                    pr2["url"],
-                )
-            else:
-                return CherryPick.Status.FIRST_MERGEABLE
-        else:
-            logging.debug(
-                "Found PR with backport of %s to %s: %s",
-                self._pr["number"],
-                self.target_branch,
-                pr2["url"],
-            )
-
-        if pr2["merged"]:
-            return CherryPick.Status.MERGED
-        elif pr2["closed"]:
-            return CherryPick.Status.DISCARDED
-        elif pr2["mergeable"] == "CONFLICTING":
-            return CherryPick.Status.SECOND_CONFLICTS
-        else:
-            return CherryPick.Status.SECOND_MERGEABLE
-
-
-if __name__ == "__main__":
-    logging.basicConfig(format="%(message)s", stream=sys.stdout, level=logging.DEBUG)
-
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        "--token", "-t", type=str, required=True, help="token for Github access"
-    )
-    parser.add_argument("--pr", type=str, required=True, help="PR# to cherry-pick")
-    parser.add_argument(
-        "--branch",
-        "-b",
-        type=str,
-        required=True,
-        help="target branch name for cherry-pick",
-    )
-    parser.add_argument(
-        "--repo",
-        "-r",
-        type=str,
-        required=True,
-        help="path to full repository",
-        metavar="PATH",
-    )
-    args = parser.parse_args()
-
-    cp = CherryPick(
-        args.token, "ClickHouse", "ClickHouse", "core", args.pr, args.branch
-    )
-    cp.execute(args.repo)

utils/github/local.py

@@ -1,108 +0,0 @@
-# -*- coding: utf-8 -*-
-
-import functools
-import logging
-import os
-import re
-
-
-class RepositoryBase:
-    def __init__(self, repo_path):
-        import git
-
-        self._repo = git.Repo(repo_path, search_parent_directories=(not repo_path))
-
-        # comparator of commits
-        def cmp(x, y):
-            if str(x) == str(y):
-                return 0
-            if self._repo.is_ancestor(x, y):
-                return -1
-            else:
-                return 1
-
-        self.comparator = functools.cmp_to_key(cmp)
-
-    def get_head_commit(self):
-        return self._repo.commit(self._default)
-
-    def iterate(self, begin, end):
-        rev_range = "{}...{}".format(begin, end)
-        for commit in self._repo.iter_commits(rev_range, first_parent=True):
-            yield commit
-
-
-class Repository(RepositoryBase):
-    def __init__(self, repo_path, remote_name, default_branch_name):
-        super(Repository, self).__init__(repo_path)
-        self._remote = self._repo.remotes[remote_name]
-        self._remote.fetch()
-        self._default = self._remote.refs[default_branch_name]
-
-    def get_release_branches(self):
-        """
-        Returns sorted list of tuples:
-         * remote branch (git.refs.remote.RemoteReference),
-         * base commit (git.Commit),
-         * head (git.Commit)).
-        List is sorted by commits in ascending order.
-        """
-        release_branches = []
-
-        RE_RELEASE_BRANCH_REF = re.compile(r"^refs/remotes/.+/\d+\.\d+$")
-
-        for branch in [
-            r for r in self._remote.refs if RE_RELEASE_BRANCH_REF.match(r.path)
-        ]:
-            base = self._repo.merge_base(self._default, self._repo.commit(branch))
-            if not base:
-                logging.info(
-                    "Branch %s is not based on branch %s. Ignoring.",
-                    branch.path,
-                    self._default,
-                )
-            elif len(base) > 1:
-                logging.info(
-                    "Branch %s has more than one base commit. Ignoring.", branch.path
-                )
-            else:
-                release_branches.append((os.path.basename(branch.name), base[0]))
-
-        return sorted(release_branches, key=lambda x: self.comparator(x[1]))
-
-
-class BareRepository(RepositoryBase):
-    def __init__(self, repo_path, default_branch_name):
-        super(BareRepository, self).__init__(repo_path)
-        self._default = self._repo.branches[default_branch_name]
-
-    def get_release_branches(self):
-        """
-        Returns sorted list of tuples:
-         * branch (git.refs.head?),
-         * base commit (git.Commit),
-         * head (git.Commit)).
-        List is sorted by commits in ascending order.
-        """
-        release_branches = []
-
-        RE_RELEASE_BRANCH_REF = re.compile(r"^refs/heads/\d+\.\d+$")
-
-        for branch in [
-            r for r in self._repo.branches if RE_RELEASE_BRANCH_REF.match(r.path)
-        ]:
-            base = self._repo.merge_base(self._default, self._repo.commit(branch))
-            if not base:
-                logging.info(
-                    "Branch %s is not based on branch %s. Ignoring.",
-                    branch.path,
-                    self._default,
-                )
-            elif len(base) > 1:
-                logging.info(
-                    "Branch %s has more than one base commit. Ignoring.", branch.path
-                )
-            else:
-                release_branches.append((os.path.basename(branch.name), base[0]))
-
-        return sorted(release_branches, key=lambda x: self.comparator(x[1]))

utils/github/parser.py

@@ -1,64 +0,0 @@
-# -*- coding: utf-8 -*-
-
-
-class Description:
-    """Parsed description representation"""
-
-    MAP_CATEGORY_TO_LABEL = {
-        "New Feature": "pr-feature",
-        "Bug Fix": "pr-bugfix",
-        "Improvement": "pr-improvement",
-        "Performance Improvement": "pr-performance",
-        # 'Backward Incompatible Change': doesn't match anything
-        "Build/Testing/Packaging Improvement": "pr-build",
-        "Non-significant (changelog entry is not needed)": "pr-non-significant",
-        "Non-significant (changelog entry is not required)": "pr-non-significant",
-        "Non-significant": "pr-non-significant",
-        "Documentation (changelog entry is not required)": "pr-documentation",
-        # 'Other': doesn't match anything
-    }
-
-    def __init__(self, pull_request):
-        self.label_name = str()
-        self.legal = False
-
-        self._parse(pull_request["bodyText"])
-
-    def _parse(self, text):
-        lines = text.splitlines()
-        next_category = False
-        category = str()
-
-        for line in lines:
-            stripped = line.strip()
-
-            if not stripped:
-                continue
-
-            if next_category:
-                category = stripped
-                next_category = False
-
-            if (
-                stripped
-                == "I hereby agree to the terms of the CLA available at: https://yandex.ru/legal/cla/?lang=en"
-            ):
-                self.legal = True
-
-            category_headers = (
-                "Category (leave one):",
-                "Changelog category (leave one):",
-                "Changelog category:",
-                "Category:",
-            )
-
-            if stripped in category_headers:
-                next_category = True
-
-        if category in Description.MAP_CATEGORY_TO_LABEL:
-            self.label_name = Description.MAP_CATEGORY_TO_LABEL[category]
-        else:
-            if not category:
-                print("Cannot find category in pr description")
-            else:
-                print(("Unknown category: " + category))

utils/github/query.py

@@ -1,492 +0,0 @@
-# -*- coding: utf-8 -*-
-
-import requests
-
-
-class Query:
-    """
-    Implements queries to the Github API using GraphQL
-    """
-
-    _PULL_REQUEST = """
-        author {{
-            ... on User {{
-                id
-                login
-            }}
-        }}
-
-        baseRepository {{
-            nameWithOwner
-        }}
-
-        mergeCommit {{
-            oid
-            parents(first: {min_page_size}) {{
-                totalCount
-                nodes {{
-                    oid
-                }}
-            }}
-        }}
-
-        mergedBy {{
-            ... on User {{
-                id
-                login
-            }}
-        }}
-
-        baseRefName
-        closed
-        headRefName
-        id
-        mergeable
-        merged
-        number
-        title
-        url
-    """
-
-    def __init__(self, token, owner, name, team, max_page_size=100, min_page_size=10):
-        self._PULL_REQUEST = Query._PULL_REQUEST.format(min_page_size=min_page_size)
-
-        self._token = token
-        self._owner = owner
-        self._name = name
-        self._team = team
-
-        self._max_page_size = max_page_size
-        self._min_page_size = min_page_size
-
-        self.api_costs = {}
-
-        repo = self.get_repository()
-        self._id = repo["id"]
-        self.ssh_url = repo["sshUrl"]
-        self.default_branch = repo["defaultBranchRef"]["name"]
-
-        self.members = set(self.get_members())
-
-    def get_repository(self):
-        _QUERY = """
-            repository(owner: "{owner}" name: "{name}") {{
-                defaultBranchRef {{
-                    name
-                }}
-                id
-                sshUrl
-            }}
-        """
-
-        query = _QUERY.format(owner=self._owner, name=self._name)
-        return self._run(query)["repository"]
-
-    def get_members(self):
-        """Get all team members for organization
-
-        Returns:
-            members: a map of members' logins to ids
-        """
-
-        _QUERY = """
-            organization(login: "{organization}") {{
-                team(slug: "{team}") {{
-                    members(first: {max_page_size} {next}) {{
-                        pageInfo {{
-                            hasNextPage
-                            endCursor
-                        }}
-                        nodes {{
-                            id
-                            login
-                        }}
-                    }}
-                }}
-            }}
-        """
-
-        members = {}
-        not_end = True
-        query = _QUERY.format(
-            organization=self._owner,
-            team=self._team,
-            max_page_size=self._max_page_size,
-            next="",
-        )
-
-        while not_end:
-            result = self._run(query)["organization"]["team"]
-            if result is None:
-                break
-            result = result["members"]
-            not_end = result["pageInfo"]["hasNextPage"]
-            query = _QUERY.format(
-                organization=self._owner,
-                team=self._team,
-                max_page_size=self._max_page_size,
-                next='after: "{}"'.format(result["pageInfo"]["endCursor"]),
-            )
-
-            members += dict([(node["login"], node["id"]) for node in result["nodes"]])
-
-        return members
-
-    def get_pull_request(self, number):
-        _QUERY = """
-            repository(owner: "{owner}" name: "{name}") {{
-                pullRequest(number: {number}) {{
-                    {pull_request_data}
-                }}
-            }}
-        """
-
-        query = _QUERY.format(
-            owner=self._owner,
-            name=self._name,
-            number=number,
-            pull_request_data=self._PULL_REQUEST,
-            min_page_size=self._min_page_size,
-        )
-        return self._run(query)["repository"]["pullRequest"]
-
-    def find_pull_request(self, base, head):
-        _QUERY = """
-            repository(owner: "{owner}" name: "{name}") {{
-                pullRequests(first: {min_page_size} baseRefName: "{base}" headRefName: "{head}") {{
-                    nodes {{
-                        {pull_request_data}
-                    }}
-                    totalCount
-                }}
-            }}
-        """
-
-        query = _QUERY.format(
-            owner=self._owner,
-            name=self._name,
-            base=base,
-            head=head,
-            pull_request_data=self._PULL_REQUEST,
-            min_page_size=self._min_page_size,
-        )
-        result = self._run(query)["repository"]["pullRequests"]
-        if result["totalCount"] > 0:
-            return result["nodes"][0]
-        else:
-            return {}
-
-    def find_pull_requests(self, label_name):
-        """
-        Get all pull-requests filtered by label name
-        """
-        _QUERY = """
-            repository(owner: "{owner}" name: "{name}") {{
-                pullRequests(first: {min_page_size} labels: "{label_name}" states: OPEN) {{
-                    nodes {{
-                        {pull_request_data}
-                    }}
-                }}
-            }}
-        """
-
-        query = _QUERY.format(
-            owner=self._owner,
-            name=self._name,
-            label_name=label_name,
-            pull_request_data=self._PULL_REQUEST,
-            min_page_size=self._min_page_size,
-        )
-        return self._run(query)["repository"]["pullRequests"]["nodes"]
-
-    def get_pull_requests(self, before_commit):
-        """
-        Get all merged pull-requests from the HEAD of default branch to the last commit (excluding)
-        """
-
-        _QUERY = """
-            repository(owner: "{owner}" name: "{name}") {{
-                defaultBranchRef {{
-                    target {{
-                        ... on Commit {{
-                            history(first: {max_page_size} {next}) {{
-                                pageInfo {{
-                                    hasNextPage
-                                    endCursor
-                                }}
-                                nodes {{
-                                    oid
-                                    associatedPullRequests(first: {min_page_size}) {{
-                                        totalCount
-                                        nodes {{
-                                            ... on PullRequest {{
-                                                {pull_request_data}
-
-                                                labels(first: {min_page_size}) {{
-                                                    totalCount
-                                                    pageInfo {{
-                                                        hasNextPage
-                                                        endCursor
-                                                    }}
-                                                    nodes {{
-                                                        name
-                                                        color
-                                                    }}
-                                                }}
-                                            }}
-                                        }}
-                                    }}
-                                }}
-                            }}
-                        }}
-                    }}
-                }}
-            }}
-        """
-
-        pull_requests = []
-        not_end = True
-        query = _QUERY.format(
-            owner=self._owner,
-            name=self._name,
-            max_page_size=self._max_page_size,
-            min_page_size=self._min_page_size,
-            pull_request_data=self._PULL_REQUEST,
-            next="",
-        )
-
-        while not_end:
-            result = self._run(query)["repository"]["defaultBranchRef"]["target"][
-                "history"
-            ]
-            not_end = result["pageInfo"]["hasNextPage"]
-            query = _QUERY.format(
-                owner=self._owner,
-                name=self._name,
-                max_page_size=self._max_page_size,
-                min_page_size=self._min_page_size,
-                pull_request_data=self._PULL_REQUEST,
-                next='after: "{}"'.format(result["pageInfo"]["endCursor"]),
-            )
-
-            for commit in result["nodes"]:
-                # FIXME: maybe include `before_commit`?
-                if str(commit["oid"]) == str(before_commit):
-                    not_end = False
-                    break
-
-                # TODO: fetch all pull-requests that were merged in a single commit.
-                assert (
-                    commit["associatedPullRequests"]["totalCount"]
-                    <= self._min_page_size
-                )
-
-                for pull_request in commit["associatedPullRequests"]["nodes"]:
-                    if (
-                        pull_request["baseRepository"]["nameWithOwner"]
-                        == "{}/{}".format(self._owner, self._name)
-                        and pull_request["baseRefName"] == self.default_branch
-                        and pull_request["mergeCommit"]["oid"] == commit["oid"]
-                    ):
-                        pull_requests.append(pull_request)
-
-        return pull_requests
-
-    def create_pull_request(
-        self, source, target, title, description="", draft=False, can_modify=True
-    ):
-        _QUERY = """
-            createPullRequest(input: {{
-                baseRefName: "{target}",
-                headRefName: "{source}",
-                repositoryId: "{id}",
-                title: "{title}",
-                body: "{body}",
-                draft: {draft},
-                maintainerCanModify: {modify}
-            }}) {{
-                pullRequest {{
-                    {pull_request_data}
-                }}
-            }}
-        """
-
-        query = _QUERY.format(
-            target=target,
-            source=source,
-            id=self._id,
-            title=title,
-            body=description,
-            draft="true" if draft else "false",
-            modify="true" if can_modify else "false",
-            pull_request_data=self._PULL_REQUEST,
-        )
-        return self._run(query, is_mutation=True)["createPullRequest"]["pullRequest"]
-
-    def merge_pull_request(self, id):
-        _QUERY = """
-            mergePullRequest(input: {{
-                pullRequestId: "{id}"
-            }}) {{
-                pullRequest {{
-                    {pull_request_data}
-                }}
-            }}
-        """
-
-        query = _QUERY.format(id=id, pull_request_data=self._PULL_REQUEST)
-        return self._run(query, is_mutation=True)["mergePullRequest"]["pullRequest"]
-
-    # FIXME: figure out how to add more assignees at once
-    def add_assignee(self, pr, assignee):
-        _QUERY = """
-            addAssigneesToAssignable(input: {{
-                assignableId: "{id1}",
-                assigneeIds: "{id2}"
-            }}) {{
-                clientMutationId
-            }}
-        """
-
-        query = _QUERY.format(id1=pr["id"], id2=assignee["id"])
-        self._run(query, is_mutation=True)
-
-    def set_label(self, pull_request, label_name):
-        """
-        Set label by name to the pull request
-
-        Args:
-            pull_request: JSON object returned by `get_pull_requests()`
-            label_name (string): label name
-        """
-
-        _GET_LABEL = """
-            repository(owner: "{owner}" name: "{name}") {{
-                labels(first: {max_page_size} {next} query: "{label_name}") {{
-                    pageInfo {{
-                        hasNextPage
-                        endCursor
-                    }}
-                    nodes {{
-                        id
-                        name
-                        color
-                    }}
-                }}
-            }}
-        """
-
-        _SET_LABEL = """
-            addLabelsToLabelable(input: {{
-                labelableId: "{pr_id}",
-                labelIds: "{label_id}"
-            }}) {{
-                clientMutationId
-            }}
-        """
-
-        labels = []
-        not_end = True
-        query = _GET_LABEL.format(
-            owner=self._owner,
-            name=self._name,
-            label_name=label_name,
-            max_page_size=self._max_page_size,
-            next="",
-        )
-
-        while not_end:
-            result = self._run(query)["repository"]["labels"]
-            not_end = result["pageInfo"]["hasNextPage"]
-            query = _GET_LABEL.format(
-                owner=self._owner,
-                name=self._name,
-                label_name=label_name,
-                max_page_size=self._max_page_size,
-                next='after: "{}"'.format(result["pageInfo"]["endCursor"]),
-            )
-
-            labels += [label for label in result["nodes"]]
-
-        if not labels:
-            return
-
-        query = _SET_LABEL.format(pr_id=pull_request["id"], label_id=labels[0]["id"])
-        self._run(query, is_mutation=True)
-
-    def _run(self, query, is_mutation=False):
-        from requests.adapters import HTTPAdapter
-        from urllib3.util.retry import Retry
-
-        def requests_retry_session(
-            retries=3,
-            backoff_factor=0.3,
-            status_forcelist=(500, 502, 504),
-            session=None,
-        ):
-            session = session or requests.Session()
-            retry = Retry(
-                total=retries,
-                read=retries,
-                connect=retries,
-                backoff_factor=backoff_factor,
-                status_forcelist=status_forcelist,
-            )
-            adapter = HTTPAdapter(max_retries=retry)
-            session.mount("http://", adapter)
-            session.mount("https://", adapter)
-            return session
-
-        headers = {"Authorization": "bearer {}".format(self._token)}
-        if is_mutation:
-            query = """
-            mutation {{
-                {query}
-            }}
-            """.format(
-                query=query
-            )
-        else:
-            query = """
-            query {{
-                {query}
-                rateLimit {{
-                    cost
-                    remaining
-                }}
-            }}
-            """.format(
-                query=query
-            )
-
-        while True:
-            request = requests_retry_session().post(
-                "https://api.github.com/graphql", json={"query": query}, headers=headers
-            )
-            if request.status_code == 200:
-                result = request.json()
-                if "errors" in result:
-                    raise Exception(
-                        "Errors occurred: {}\nOriginal query: {}".format(
-                            result["errors"], query
-                        )
-                    )
-
-                if not is_mutation:
-                    import inspect
-
-                    caller = inspect.getouterframes(inspect.currentframe(), 2)[1][3]
-                    if caller not in list(self.api_costs.keys()):
-                        self.api_costs[caller] = 0
-                    self.api_costs[caller] += result["data"]["rateLimit"]["cost"]
-
-                return result["data"]
-            else:
-                import json
-
-                raise Exception(
-                    "Query failed with code {code}:\n{json}".format(
-                        code=request.status_code,
-                        json=json.dumps(request.json(), indent=4),
-                    )
-                )

utils/tests-visualizer/index.html

@@ -84,14 +84,14 @@ let render_data_query = `
     SELECT groupArray([d, n, fail]) FROM
     (
         SELECT n, check_start_time::Date - start_date AS d, max(test_status LIKE 'F%' OR test_status LIKE 'E%') AS fail
-        FROM "gh-data".checks
+        FROM "default".checks
 
         INNER JOIN
         (
             SELECT test_name, toUInt16(rowNumberInAllBlocks()) AS n FROM
             (
                 SELECT DISTINCT test_name
-                FROM "gh-data".checks
+                FROM "default".checks
                 WHERE match(test_name, '^\\d+_') AND check_name ILIKE '%stateless%' AND check_start_time > now() - INTERVAL 1 DAY
                 ORDER BY test_name
             )
@@ -112,7 +112,7 @@ let test_names_query = `
     SELECT test_name, toUInt16(rowNumberInAllBlocks()) AS n FROM
     (
         SELECT DISTINCT test_name
-        FROM "gh-data".checks
+        FROM "default".checks
         WHERE match(test_name, '^\\d+_') AND check_name ILIKE '%stateless%' AND check_start_time > now() - INTERVAL 1 DAY
         ORDER BY test_name
     ) FORMAT JSONCompact`;