From 77a5865a22d290033aec1894b2c79e688f713238 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sat, 13 Jan 2024 22:30:30 +0100 Subject: [PATCH 001/244] Adding FP16 --- base/base/BFloat16.h | 9 ++++++++ base/base/DecomposedFloat.h | 10 +++++++++ base/base/TypeLists.h | 2 +- base/base/TypeName.h | 1 + base/base/extended_types.h | 20 ++++++++++++++--- base/base/wide_integer.h | 2 +- base/base/wide_integer_impl.h | 8 ++++++- .../AggregateFunctionGroupArray.cpp | 4 ++-- .../AggregateFunctionGroupArrayMoving.cpp | 2 +- .../AggregateFunctionIntervalLengthSum.cpp | 4 ++-- .../AggregateFunctionSparkbar.cpp | 6 ++--- src/AggregateFunctions/AggregateFunctionSum.h | 6 ++--- .../AggregateFunctionUniqCombined.h | 2 +- src/AggregateFunctions/QuantileTDigest.h | 2 +- src/AggregateFunctions/ReservoirSampler.h | 2 +- .../ReservoirSamplerDeterministic.h | 2 +- src/Columns/ColumnArray.cpp | 4 ++++ src/Columns/ColumnNullable.cpp | 2 ++ src/Columns/ColumnVector.cpp | 13 ++++++----- src/Columns/ColumnVector.h | 1 + src/Columns/ColumnsCommon.cpp | 1 + src/Columns/ColumnsNumber.h | 1 + src/Columns/MaskOperations.cpp | 2 ++ src/Columns/tests/gtest_column_vector.cpp | 1 + src/Columns/tests/gtest_low_cardinality.cpp | 1 + src/Common/FieldVisitorConvertToNumber.h | 4 ++-- src/Common/HashTable/Hash.h | 1 + src/Common/HashTable/HashTable.h | 2 +- src/Common/NaNUtils.h | 6 ++--- src/Common/findExtreme.h | 2 +- src/Common/transformEndianness.h | 2 +- src/Core/AccurateComparison.h | 18 +++++++-------- src/Core/DecimalFunctions.h | 2 +- src/Core/Field.h | 1 + src/Core/SortCursor.h | 1 + src/Core/TypeId.h | 2 ++ src/Core/Types_fwd.h | 7 +----- src/Core/callOnTypeIndex.h | 3 +++ src/DataTypes/DataTypeNumberBase.cpp | 1 + src/DataTypes/DataTypeNumberBase.h | 1 + src/DataTypes/DataTypesDecimal.h | 5 +++-- src/DataTypes/DataTypesNumber.cpp | 1 + src/DataTypes/DataTypesNumber.h | 1 + src/DataTypes/IDataType.h | 5 ++++- src/DataTypes/NumberTraits.h | 22 +++++++++---------- .../Serializations/SerializationNumber.cpp | 1 + src/DataTypes/Utils.cpp | 7 ++++++ src/DataTypes/getLeastSupertype.cpp | 6 ++++- src/DataTypes/getMostSubtype.cpp | 6 ++++- src/Formats/ProtobufSerializer.cpp | 2 +- src/Functions/DivisionUtils.h | 16 +++++++------- src/Functions/FunctionMathUnary.h | 4 ++-- src/Functions/FunctionsConversion.h | 12 +++++----- src/Functions/FunctionsRound.h | 2 +- src/Functions/factorial.cpp | 2 +- src/Functions/minus.cpp | 4 ++-- src/Functions/moduloOrZero.cpp | 2 +- src/Functions/multiply.cpp | 4 ++-- src/Functions/plus.cpp | 4 ++-- src/Functions/sign.cpp | 2 +- src/IO/ReadHelpers.h | 4 +++- src/IO/WriteHelpers.h | 22 +++++++++++-------- src/Interpreters/RowRefs.cpp | 2 +- 63 files changed, 192 insertions(+), 105 deletions(-) create mode 100644 base/base/BFloat16.h diff --git a/base/base/BFloat16.h b/base/base/BFloat16.h new file mode 100644 index 00000000000..17c3ebe9ef3 --- /dev/null +++ b/base/base/BFloat16.h @@ -0,0 +1,9 @@ +#pragma once + +using BFloat16 = __bf16; + +namespace std +{ + inline constexpr bool isfinite(BFloat16) { return true; } + inline constexpr bool signbit(BFloat16) { return false; } +} diff --git a/base/base/DecomposedFloat.h b/base/base/DecomposedFloat.h index f152637b94e..fda7ee8d3f4 100644 --- a/base/base/DecomposedFloat.h +++ b/base/base/DecomposedFloat.h @@ -10,6 +10,15 @@ template struct FloatTraits; +template <> +struct FloatTraits<__bf16> +{ + using UInt = uint16_t; + static constexpr size_t bits = 16; + static constexpr size_t exponent_bits = 8; + static constexpr size_t mantissa_bits = bits - exponent_bits - 1; +}; + template <> struct FloatTraits { @@ -217,3 +226,4 @@ struct DecomposedFloat using DecomposedFloat64 = DecomposedFloat; using DecomposedFloat32 = DecomposedFloat; +using DecomposedFloat16 = DecomposedFloat<__bf16>; diff --git a/base/base/TypeLists.h b/base/base/TypeLists.h index 6c1283d054c..ce3111b1da3 100644 --- a/base/base/TypeLists.h +++ b/base/base/TypeLists.h @@ -9,7 +9,7 @@ namespace DB { using TypeListNativeInt = TypeList; -using TypeListFloat = TypeList; +using TypeListFloat = TypeList; using TypeListNativeNumber = TypeListConcat; using TypeListWideInt = TypeList; using TypeListInt = TypeListConcat; diff --git a/base/base/TypeName.h b/base/base/TypeName.h index 9005b5a2bf4..1f4b475d653 100644 --- a/base/base/TypeName.h +++ b/base/base/TypeName.h @@ -32,6 +32,7 @@ TN_MAP(Int32) TN_MAP(Int64) TN_MAP(Int128) TN_MAP(Int256) +TN_MAP(BFloat16) TN_MAP(Float32) TN_MAP(Float64) TN_MAP(String) diff --git a/base/base/extended_types.h b/base/base/extended_types.h index b58df45a97e..39665784141 100644 --- a/base/base/extended_types.h +++ b/base/base/extended_types.h @@ -4,6 +4,8 @@ #include #include +#include + using Int128 = wide::integer<128, signed>; using UInt128 = wide::integer<128, unsigned>; @@ -24,6 +26,7 @@ struct is_signed // NOLINT(readability-identifier-naming) template <> struct is_signed { static constexpr bool value = true; }; template <> struct is_signed { static constexpr bool value = true; }; +template <> struct is_signed { static constexpr bool value = true; }; template inline constexpr bool is_signed_v = is_signed::value; @@ -47,8 +50,6 @@ template concept is_integer = || std::is_same_v || std::is_same_v; -template concept is_floating_point = std::is_floating_point_v; - template struct is_arithmetic // NOLINT(readability-identifier-naming) { @@ -59,11 +60,24 @@ template <> struct is_arithmetic { static constexpr bool value = true; } template <> struct is_arithmetic { static constexpr bool value = true; }; template <> struct is_arithmetic { static constexpr bool value = true; }; template <> struct is_arithmetic { static constexpr bool value = true; }; - +template <> struct is_arithmetic { static constexpr bool value = true; }; template inline constexpr bool is_arithmetic_v = is_arithmetic::value; + +template +struct is_floating_point // NOLINT(readability-identifier-naming) +{ + static constexpr bool value = std::is_floating_point_v; +}; + +template <> struct is_floating_point { static constexpr bool value = true; }; + +template +inline constexpr bool is_floating_point_v = is_floating_point::value; + + template struct make_unsigned // NOLINT(readability-identifier-naming) { diff --git a/base/base/wide_integer.h b/base/base/wide_integer.h index ffd30460c03..877ef5bd137 100644 --- a/base/base/wide_integer.h +++ b/base/base/wide_integer.h @@ -117,6 +117,7 @@ public: constexpr operator long double() const noexcept; constexpr operator double() const noexcept; constexpr operator float() const noexcept; + constexpr operator __bf16() const noexcept; struct _impl; @@ -262,4 +263,3 @@ struct hash>; // NOLINTEND(*) #include "wide_integer_impl.h" - diff --git a/base/base/wide_integer_impl.h b/base/base/wide_integer_impl.h index c1fd7b69b7f..7b95164e44d 100644 --- a/base/base/wide_integer_impl.h +++ b/base/base/wide_integer_impl.h @@ -154,7 +154,7 @@ struct common_type, Arithmetic> static_assert(wide::ArithmeticConcept()); using type = std::conditional_t< - std::is_floating_point_v, + is_floating_point_v, Arithmetic, std::conditional_t< sizeof(Arithmetic) * 8 < Bits, @@ -1291,6 +1291,12 @@ constexpr integer::operator float() const noexcept return static_cast(static_cast(*this)); } +template +constexpr integer::operator __bf16() const noexcept +{ + return static_cast<__bf16>(static_cast(*this)); +} + // Unary operators template constexpr integer operator~(const integer & lhs) noexcept diff --git a/src/AggregateFunctions/AggregateFunctionGroupArray.cpp b/src/AggregateFunctions/AggregateFunctionGroupArray.cpp index 6c6397e35d5..bcefa6b93dc 100644 --- a/src/AggregateFunctions/AggregateFunctionGroupArray.cpp +++ b/src/AggregateFunctions/AggregateFunctionGroupArray.cpp @@ -74,7 +74,7 @@ template struct GroupArraySamplerData { /// For easy serialization. - static_assert(std::has_unique_object_representations_v || std::is_floating_point_v); + static_assert(std::has_unique_object_representations_v || is_floating_point_v); // Switch to ordinary Allocator after 4096 bytes to avoid fragmentation and trash in Arena using Allocator = MixedAlignedArenaAllocator; @@ -116,7 +116,7 @@ template struct GroupArrayNumericData { /// For easy serialization. - static_assert(std::has_unique_object_representations_v || std::is_floating_point_v); + static_assert(std::has_unique_object_representations_v || is_floating_point_v); // Switch to ordinary Allocator after 4096 bytes to avoid fragmentation and trash in Arena using Allocator = MixedAlignedArenaAllocator; diff --git a/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp b/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp index 026b8d1956f..ee6a82686c5 100644 --- a/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp +++ b/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp @@ -38,7 +38,7 @@ template struct MovingData { /// For easy serialization. - static_assert(std::has_unique_object_representations_v || std::is_floating_point_v); + static_assert(std::has_unique_object_representations_v || is_floating_point_v); using Accumulator = T; diff --git a/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp b/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp index eacd0596757..06156643aa0 100644 --- a/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp +++ b/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp @@ -187,7 +187,7 @@ public: static DataTypePtr createResultType() { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return std::make_shared(); return std::make_shared(); } @@ -227,7 +227,7 @@ public: void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena *) const override { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) assert_cast(to).getData().push_back(getIntervalLengthSum(this->data(place))); else assert_cast(to).getData().push_back(getIntervalLengthSum(this->data(place))); diff --git a/src/AggregateFunctions/AggregateFunctionSparkbar.cpp b/src/AggregateFunctions/AggregateFunctionSparkbar.cpp index b6e538520a8..f4214f3a133 100644 --- a/src/AggregateFunctions/AggregateFunctionSparkbar.cpp +++ b/src/AggregateFunctions/AggregateFunctionSparkbar.cpp @@ -50,7 +50,7 @@ struct AggregateFunctionSparkbarData auto [it, inserted] = points.insert({x, y}); if (!inserted) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { it->getMapped() += y; return it->getMapped(); @@ -197,7 +197,7 @@ private: Y res; bool has_overfllow = false; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) res = histogram[index] + point.getMapped(); else has_overfllow = common::addOverflow(histogram[index], point.getMapped(), res); @@ -246,7 +246,7 @@ private: } constexpr auto levels_num = static_cast(BAR_LEVELS - 1); - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { y = y / (y_max / levels_num) + 1; } diff --git a/src/AggregateFunctions/AggregateFunctionSum.h b/src/AggregateFunctions/AggregateFunctionSum.h index 5781ab69c6b..81df3244b38 100644 --- a/src/AggregateFunctions/AggregateFunctionSum.h +++ b/src/AggregateFunctions/AggregateFunctionSum.h @@ -69,7 +69,7 @@ struct AggregateFunctionSumData size_t count = end - start; const auto * end_ptr = ptr + count; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { /// Compiler cannot unroll this loop, do it manually. /// (at least for floats, most likely due to the lack of -fassociative-math) @@ -164,7 +164,7 @@ struct AggregateFunctionSumData return; } - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { /// For floating point we use a similar trick as above, except that now we reinterpret the floating point number as an unsigned /// integer of the same size and use a mask instead (0 to discard, 0xFF..FF to keep) @@ -277,7 +277,7 @@ struct AggregateFunctionSumData template struct AggregateFunctionSumKahanData { - static_assert(std::is_floating_point_v, + static_assert(is_floating_point_v, "It doesn't make sense to use Kahan Summation algorithm for non floating point types"); T sum{}; diff --git a/src/AggregateFunctions/AggregateFunctionUniqCombined.h b/src/AggregateFunctions/AggregateFunctionUniqCombined.h index 10774442610..19e2665f9af 100644 --- a/src/AggregateFunctions/AggregateFunctionUniqCombined.h +++ b/src/AggregateFunctions/AggregateFunctionUniqCombined.h @@ -114,7 +114,7 @@ public: /// Initially UInt128 was introduced only for UUID, and then the other big-integer types were added. hash = static_cast(sipHash64(value)); } - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { hash = static_cast(intHash64(bit_cast(value))); } diff --git a/src/AggregateFunctions/QuantileTDigest.h b/src/AggregateFunctions/QuantileTDigest.h index 979c3f2af15..1407b73e669 100644 --- a/src/AggregateFunctions/QuantileTDigest.h +++ b/src/AggregateFunctions/QuantileTDigest.h @@ -380,7 +380,7 @@ public: ResultType getImpl(Float64 level) { if (centroids.empty()) - return std::is_floating_point_v ? std::numeric_limits::quiet_NaN() : 0; + return is_floating_point_v ? std::numeric_limits::quiet_NaN() : 0; compress(); diff --git a/src/AggregateFunctions/ReservoirSampler.h b/src/AggregateFunctions/ReservoirSampler.h index 37fc05a2e4c..242540102b8 100644 --- a/src/AggregateFunctions/ReservoirSampler.h +++ b/src/AggregateFunctions/ReservoirSampler.h @@ -278,6 +278,6 @@ private: if (OnEmpty == ReservoirSamplerOnEmpty::THROW) throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Quantile of empty ReservoirSampler"); else - return NanLikeValueConstructor>::getValue(); + return NanLikeValueConstructor>::getValue(); } }; diff --git a/src/AggregateFunctions/ReservoirSamplerDeterministic.h b/src/AggregateFunctions/ReservoirSamplerDeterministic.h index daed0b98ca3..75af6638183 100644 --- a/src/AggregateFunctions/ReservoirSamplerDeterministic.h +++ b/src/AggregateFunctions/ReservoirSamplerDeterministic.h @@ -273,7 +273,7 @@ private: if (OnEmpty == ReservoirSamplerDeterministicOnEmpty::THROW) throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Quantile of empty ReservoirSamplerDeterministic"); else - return NanLikeValueConstructor>::getValue(); + return NanLikeValueConstructor>::getValue(); } }; diff --git a/src/Columns/ColumnArray.cpp b/src/Columns/ColumnArray.cpp index 1cb8188bce6..4aaaf01e5ea 100644 --- a/src/Columns/ColumnArray.cpp +++ b/src/Columns/ColumnArray.cpp @@ -574,6 +574,8 @@ ColumnPtr ColumnArray::filter(const Filter & filt, ssize_t result_size_hint) con return filterNumber(filt, result_size_hint); if (typeid_cast(data.get())) return filterNumber(filt, result_size_hint); + if (typeid_cast(data.get())) + return filterNumber(filt, result_size_hint); if (typeid_cast(data.get())) return filterNumber(filt, result_size_hint); if (typeid_cast(data.get())) @@ -993,6 +995,8 @@ ColumnPtr ColumnArray::replicate(const Offsets & replicate_offsets) const return replicateNumber(replicate_offsets); if (typeid_cast(data.get())) return replicateNumber(replicate_offsets); + if (typeid_cast(data.get())) + return replicateNumber(replicate_offsets); if (typeid_cast(data.get())) return replicateNumber(replicate_offsets); if (typeid_cast(data.get())) diff --git a/src/Columns/ColumnNullable.cpp b/src/Columns/ColumnNullable.cpp index 4ee6bb3d586..3513ac06dcd 100644 --- a/src/Columns/ColumnNullable.cpp +++ b/src/Columns/ColumnNullable.cpp @@ -171,6 +171,8 @@ StringRef ColumnNullable::serializeValueIntoArena(size_t n, Arena & arena, char return static_cast(nested_column.get())->serializeValueIntoArena(n, arena, begin, &arr[n]); case TypeIndex::Int256: return static_cast(nested_column.get())->serializeValueIntoArena(n, arena, begin, &arr[n]); + case TypeIndex::BFloat16: + return static_cast(nested_column.get())->serializeValueIntoArena(n, arena, begin, &arr[n]); case TypeIndex::Float32: return static_cast(nested_column.get())->serializeValueIntoArena(n, arena, begin, &arr[n]); case TypeIndex::Float64: diff --git a/src/Columns/ColumnVector.cpp b/src/Columns/ColumnVector.cpp index b1cf449dfde..bad84e7147c 100644 --- a/src/Columns/ColumnVector.cpp +++ b/src/Columns/ColumnVector.cpp @@ -141,7 +141,7 @@ struct ColumnVector::less_stable if (unlikely(parent.data[lhs] == parent.data[rhs])) return lhs < rhs; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { if (unlikely(std::isnan(parent.data[lhs]) && std::isnan(parent.data[rhs]))) { @@ -173,7 +173,7 @@ struct ColumnVector::greater_stable if (unlikely(parent.data[lhs] == parent.data[rhs])) return lhs < rhs; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { if (unlikely(std::isnan(parent.data[lhs]) && std::isnan(parent.data[rhs]))) { @@ -259,7 +259,7 @@ void ColumnVector::getPermutation(IColumn::PermutationSortDirection direction bool sort_is_stable = stability == IColumn::PermutationSortStability::Stable; /// TODO: LSD RadixSort is currently not stable if direction is descending, or value is floating point - bool use_radix_sort = (sort_is_stable && ascending && !std::is_floating_point_v) || !sort_is_stable; + bool use_radix_sort = (sort_is_stable && ascending && !is_floating_point_v) || !sort_is_stable; /// Thresholds on size. Lower threshold is arbitrary. Upper threshold is chosen by the type for histogram counters. if (data_size >= 256 && data_size <= std::numeric_limits::max() && use_radix_sort) @@ -286,7 +286,7 @@ void ColumnVector::getPermutation(IColumn::PermutationSortDirection direction /// Radix sort treats all NaNs to be greater than all numbers. /// If the user needs the opposite, we must move them accordingly. - if (std::is_floating_point_v && nan_direction_hint < 0) + if (is_floating_point_v && nan_direction_hint < 0) { size_t nans_to_move = 0; @@ -333,7 +333,7 @@ void ColumnVector::updatePermutation(IColumn::PermutationSortDirection direct if constexpr (is_arithmetic_v && !is_big_int_v) { /// TODO: LSD RadixSort is currently not stable if direction is descending, or value is floating point - bool use_radix_sort = (sort_is_stable && ascending && !std::is_floating_point_v) || !sort_is_stable; + bool use_radix_sort = (sort_is_stable && ascending && !is_floating_point_v) || !sort_is_stable; size_t size = end - begin; /// Thresholds on size. Lower threshold is arbitrary. Upper threshold is chosen by the type for histogram counters. @@ -356,7 +356,7 @@ void ColumnVector::updatePermutation(IColumn::PermutationSortDirection direct /// Radix sort treats all NaNs to be greater than all numbers. /// If the user needs the opposite, we must move them accordingly. - if (std::is_floating_point_v && nan_direction_hint < 0) + if (is_floating_point_v && nan_direction_hint < 0) { size_t nans_to_move = 0; @@ -970,6 +970,7 @@ template class ColumnVector; template class ColumnVector; template class ColumnVector; template class ColumnVector; +template class ColumnVector; template class ColumnVector; template class ColumnVector; template class ColumnVector; diff --git a/src/Columns/ColumnVector.h b/src/Columns/ColumnVector.h index fab2d5f06aa..c976fac3bab 100644 --- a/src/Columns/ColumnVector.h +++ b/src/Columns/ColumnVector.h @@ -570,6 +570,7 @@ extern template class ColumnVector; extern template class ColumnVector; extern template class ColumnVector; extern template class ColumnVector; +extern template class ColumnVector; extern template class ColumnVector; extern template class ColumnVector; extern template class ColumnVector; diff --git a/src/Columns/ColumnsCommon.cpp b/src/Columns/ColumnsCommon.cpp index 4ac84e10750..444f5fae87a 100644 --- a/src/Columns/ColumnsCommon.cpp +++ b/src/Columns/ColumnsCommon.cpp @@ -328,6 +328,7 @@ INSTANTIATE(Int32) INSTANTIATE(Int64) INSTANTIATE(Int128) INSTANTIATE(Int256) +INSTANTIATE(BFloat16) INSTANTIATE(Float32) INSTANTIATE(Float64) INSTANTIATE(Decimal32) diff --git a/src/Columns/ColumnsNumber.h b/src/Columns/ColumnsNumber.h index ae7eddb0b22..2dce2269079 100644 --- a/src/Columns/ColumnsNumber.h +++ b/src/Columns/ColumnsNumber.h @@ -23,6 +23,7 @@ using ColumnInt64 = ColumnVector; using ColumnInt128 = ColumnVector; using ColumnInt256 = ColumnVector; +using ColumnBFloat16 = ColumnVector; using ColumnFloat32 = ColumnVector; using ColumnFloat64 = ColumnVector; diff --git a/src/Columns/MaskOperations.cpp b/src/Columns/MaskOperations.cpp index b84268356a7..ca4ca263811 100644 --- a/src/Columns/MaskOperations.cpp +++ b/src/Columns/MaskOperations.cpp @@ -63,6 +63,7 @@ INSTANTIATE(Int32) INSTANTIATE(Int64) INSTANTIATE(Int128) INSTANTIATE(Int256) +INSTANTIATE(BFloat16) INSTANTIATE(Float32) INSTANTIATE(Float64) INSTANTIATE(Decimal32) @@ -225,6 +226,7 @@ MaskInfo extractMaskImpl( || extractMaskNumeric(mask, column, null_value, null_bytemap, nulls, mask_info) || extractMaskNumeric(mask, column, null_value, null_bytemap, nulls, mask_info) || extractMaskNumeric(mask, column, null_value, null_bytemap, nulls, mask_info) + || extractMaskNumeric(mask, column, null_value, null_bytemap, nulls, mask_info) || extractMaskNumeric(mask, column, null_value, null_bytemap, nulls, mask_info) || extractMaskNumeric(mask, column, null_value, null_bytemap, nulls, mask_info))) throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Cannot convert column {} to mask.", column->getName()); diff --git a/src/Columns/tests/gtest_column_vector.cpp b/src/Columns/tests/gtest_column_vector.cpp index b71d4a095ab..3a084a89079 100644 --- a/src/Columns/tests/gtest_column_vector.cpp +++ b/src/Columns/tests/gtest_column_vector.cpp @@ -93,6 +93,7 @@ TEST(ColumnVector, Filter) testFilter(); testFilter(); testFilter(); + testFilter(); testFilter(); testFilter(); testFilter(); diff --git a/src/Columns/tests/gtest_low_cardinality.cpp b/src/Columns/tests/gtest_low_cardinality.cpp index 5e01279b7df..965c0d219b9 100644 --- a/src/Columns/tests/gtest_low_cardinality.cpp +++ b/src/Columns/tests/gtest_low_cardinality.cpp @@ -45,6 +45,7 @@ TEST(ColumnLowCardinality, Insert) testLowCardinalityNumberInsert(std::make_shared()); testLowCardinalityNumberInsert(std::make_shared()); + testLowCardinalityNumberInsert(std::make_shared()); testLowCardinalityNumberInsert(std::make_shared()); testLowCardinalityNumberInsert(std::make_shared()); } diff --git a/src/Common/FieldVisitorConvertToNumber.h b/src/Common/FieldVisitorConvertToNumber.h index bf8c8c8638e..38144650b97 100644 --- a/src/Common/FieldVisitorConvertToNumber.h +++ b/src/Common/FieldVisitorConvertToNumber.h @@ -58,7 +58,7 @@ public: T operator() (const Float64 & x) const { - if constexpr (!std::is_floating_point_v) + if constexpr (!is_floating_point_v) { if (!isFinite(x)) { @@ -88,7 +88,7 @@ public: template T operator() (const DecimalField & x) const { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return x.getValue().template convertTo() / x.getScaleMultiplier().template convertTo(); else return (x.getValue() / x.getScaleMultiplier()). template convertTo(); diff --git a/src/Common/HashTable/Hash.h b/src/Common/HashTable/Hash.h index fb6afcde133..b4bc6af1cef 100644 --- a/src/Common/HashTable/Hash.h +++ b/src/Common/HashTable/Hash.h @@ -322,6 +322,7 @@ DEFINE_HASH(Int32) DEFINE_HASH(Int64) DEFINE_HASH(Int128) DEFINE_HASH(Int256) +DEFINE_HASH(BFloat16) DEFINE_HASH(Float32) DEFINE_HASH(Float64) DEFINE_HASH(DB::UUID) diff --git a/src/Common/HashTable/HashTable.h b/src/Common/HashTable/HashTable.h index f23c4ca15dd..e4d5d3868c8 100644 --- a/src/Common/HashTable/HashTable.h +++ b/src/Common/HashTable/HashTable.h @@ -91,7 +91,7 @@ inline bool bitEquals(T && a, T && b) { using RealT = std::decay_t; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return 0 == memcmp(&a, &b, sizeof(RealT)); /// Note that memcmp with constant size is compiler builtin. else return a == b; diff --git a/src/Common/NaNUtils.h b/src/Common/NaNUtils.h index 1c5a619e919..6363e3e61a2 100644 --- a/src/Common/NaNUtils.h +++ b/src/Common/NaNUtils.h @@ -9,7 +9,7 @@ template inline bool isNaN(T x) { /// To be sure, that this function is zero-cost for non-floating point types. - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return std::isnan(x); else return false; @@ -19,7 +19,7 @@ inline bool isNaN(T x) template inline bool isFinite(T x) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return std::isfinite(x); else return true; @@ -29,7 +29,7 @@ inline bool isFinite(T x) template T NaNOrZero() { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return std::numeric_limits::quiet_NaN(); else return {}; diff --git a/src/Common/findExtreme.h b/src/Common/findExtreme.h index b38c24697c0..611af023d33 100644 --- a/src/Common/findExtreme.h +++ b/src/Common/findExtreme.h @@ -11,7 +11,7 @@ namespace DB { template -concept is_any_native_number = (is_any_of); +concept is_any_native_number = (is_any_of); template std::optional findExtremeMin(const T * __restrict ptr, size_t start, size_t end); diff --git a/src/Common/transformEndianness.h b/src/Common/transformEndianness.h index 1657305acda..2a0c45efe38 100644 --- a/src/Common/transformEndianness.h +++ b/src/Common/transformEndianness.h @@ -38,7 +38,7 @@ inline void transformEndianness(T & x) } template -requires std::is_floating_point_v +requires is_floating_point_v inline void transformEndianness(T & value) { if constexpr (ToEndian != FromEndian) diff --git a/src/Core/AccurateComparison.h b/src/Core/AccurateComparison.h index a201c136e3a..82d06876fe3 100644 --- a/src/Core/AccurateComparison.h +++ b/src/Core/AccurateComparison.h @@ -25,7 +25,7 @@ bool lessOp(A a, B b) return a < b; /// float vs float - if constexpr (std::is_floating_point_v && std::is_floating_point_v) + if constexpr (is_floating_point_v && is_floating_point_v) return a < b; /// anything vs NaN @@ -49,7 +49,7 @@ bool lessOp(A a, B b) } /// int vs float - if constexpr (is_integer && std::is_floating_point_v) + if constexpr (is_integer && is_floating_point_v) { if constexpr (sizeof(A) <= 4) return static_cast(a) < static_cast(b); @@ -57,7 +57,7 @@ bool lessOp(A a, B b) return DecomposedFloat(b).greater(a); } - if constexpr (std::is_floating_point_v && is_integer) + if constexpr (is_floating_point_v && is_integer) { if constexpr (sizeof(B) <= 4) return static_cast(a) < static_cast(b); @@ -65,8 +65,8 @@ bool lessOp(A a, B b) return DecomposedFloat(a).less(b); } - static_assert(is_integer || std::is_floating_point_v); - static_assert(is_integer || std::is_floating_point_v); + static_assert(is_integer || is_floating_point_v); + static_assert(is_integer || is_floating_point_v); UNREACHABLE(); } @@ -101,7 +101,7 @@ bool equalsOp(A a, B b) return a == b; /// float vs float - if constexpr (std::is_floating_point_v && std::is_floating_point_v) + if constexpr (is_floating_point_v && is_floating_point_v) return a == b; /// anything vs NaN @@ -125,7 +125,7 @@ bool equalsOp(A a, B b) } /// int vs float - if constexpr (is_integer && std::is_floating_point_v) + if constexpr (is_integer && is_floating_point_v) { if constexpr (sizeof(A) <= 4) return static_cast(a) == static_cast(b); @@ -133,7 +133,7 @@ bool equalsOp(A a, B b) return DecomposedFloat(b).equals(a); } - if constexpr (std::is_floating_point_v && is_integer) + if constexpr (is_floating_point_v && is_integer) { if constexpr (sizeof(B) <= 4) return static_cast(a) == static_cast(b); @@ -163,7 +163,7 @@ inline bool NO_SANITIZE_UNDEFINED convertNumeric(From value, To & result) return true; } - if constexpr (std::is_floating_point_v && std::is_floating_point_v) + if constexpr (is_floating_point_v && is_floating_point_v) { /// Note that NaNs doesn't compare equal to anything, but they are still in range of any Float type. if (isNaN(value)) diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h index 8dad00c3a1e..c5bc4ad70f6 100644 --- a/src/Core/DecimalFunctions.h +++ b/src/Core/DecimalFunctions.h @@ -310,7 +310,7 @@ ReturnType convertToImpl(const DecimalType & decimal, UInt32 scale, To & result) using DecimalNativeType = typename DecimalType::NativeType; static constexpr bool throw_exception = std::is_void_v; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { result = static_cast(decimal.value) / static_cast(scaleMultiplier(scale)); } diff --git a/src/Core/Field.h b/src/Core/Field.h index 6afa98ed9c0..be70eb1ea07 100644 --- a/src/Core/Field.h +++ b/src/Core/Field.h @@ -251,6 +251,7 @@ template <> struct NearestFieldTypeImpl> { using Type = template <> struct NearestFieldTypeImpl> { using Type = DecimalField; }; template <> struct NearestFieldTypeImpl> { using Type = DecimalField; }; template <> struct NearestFieldTypeImpl> { using Type = DecimalField; }; +template <> struct NearestFieldTypeImpl { using Type = Float64; }; template <> struct NearestFieldTypeImpl { using Type = Float64; }; template <> struct NearestFieldTypeImpl { using Type = Float64; }; template <> struct NearestFieldTypeImpl { using Type = String; }; diff --git a/src/Core/SortCursor.h b/src/Core/SortCursor.h index 3c412fa1f17..a9dc90a8fa1 100644 --- a/src/Core/SortCursor.h +++ b/src/Core/SortCursor.h @@ -687,6 +687,7 @@ private: SortingQueueImpl>, strategy>, SortingQueueImpl>, strategy>, + SortingQueueImpl>, strategy>, SortingQueueImpl>, strategy>, SortingQueueImpl>, strategy>, diff --git a/src/Core/TypeId.h b/src/Core/TypeId.h index 9c634d2321c..73fa7da37e2 100644 --- a/src/Core/TypeId.h +++ b/src/Core/TypeId.h @@ -21,6 +21,7 @@ enum class TypeIndex Int64, Int128, Int256, + BFloat16, Float32, Float64, Date, @@ -91,6 +92,7 @@ TYPEID_MAP(Int32) TYPEID_MAP(Int64) TYPEID_MAP(Int128) TYPEID_MAP(Int256) +TYPEID_MAP(BFloat16) TYPEID_MAP(Float32) TYPEID_MAP(Float64) TYPEID_MAP(UUID) diff --git a/src/Core/Types_fwd.h b/src/Core/Types_fwd.h index a59e4b6eab8..2dffc910f9b 100644 --- a/src/Core/Types_fwd.h +++ b/src/Core/Types_fwd.h @@ -21,6 +21,7 @@ using Int128 = wide::integer<128, signed>; using UInt128 = wide::integer<128, unsigned>; using Int256 = wide::integer<256, signed>; using UInt256 = wide::integer<256, unsigned>; +using BFloat16 = __bf16; namespace DB { @@ -28,16 +29,10 @@ namespace DB using UUID = StrongTypedef; struct IPv4; - struct IPv6; struct Null; -using UInt128 = ::UInt128; -using UInt256 = ::UInt256; -using Int128 = ::Int128; -using Int256 = ::Int256; - enum class TypeIndex; /// Not a data type in database, defined just for convenience. diff --git a/src/Core/callOnTypeIndex.h b/src/Core/callOnTypeIndex.h index f5f67df563b..68aba2263c7 100644 --- a/src/Core/callOnTypeIndex.h +++ b/src/Core/callOnTypeIndex.h @@ -62,6 +62,7 @@ static bool callOnBasicType(TypeIndex number, F && f) { switch (number) { + case TypeIndex::BFloat16: return f(TypePair()); case TypeIndex::Float32: return f(TypePair()); case TypeIndex::Float64: return f(TypePair()); default: @@ -132,6 +133,7 @@ static inline bool callOnBasicTypes(TypeIndex type_num1, TypeIndex type_num2, F { switch (type_num1) { + case TypeIndex::BFloat16: return callOnBasicType(type_num2, std::forward(f)); case TypeIndex::Float32: return callOnBasicType(type_num2, std::forward(f)); case TypeIndex::Float64: return callOnBasicType(type_num2, std::forward(f)); default: @@ -189,6 +191,7 @@ static bool callOnIndexAndDataType(TypeIndex number, F && f, ExtraArgs && ... ar case TypeIndex::Int128: return f(TypePair, T>(), std::forward(args)...); case TypeIndex::Int256: return f(TypePair, T>(), std::forward(args)...); + case TypeIndex::BFloat16: return f(TypePair, T>(), std::forward(args)...); case TypeIndex::Float32: return f(TypePair, T>(), std::forward(args)...); case TypeIndex::Float64: return f(TypePair, T>(), std::forward(args)...); diff --git a/src/DataTypes/DataTypeNumberBase.cpp b/src/DataTypes/DataTypeNumberBase.cpp index be448fe1491..636d557f4d0 100644 --- a/src/DataTypes/DataTypeNumberBase.cpp +++ b/src/DataTypes/DataTypeNumberBase.cpp @@ -42,6 +42,7 @@ template class DataTypeNumberBase; template class DataTypeNumberBase; template class DataTypeNumberBase; template class DataTypeNumberBase; +template class DataTypeNumberBase; template class DataTypeNumberBase; template class DataTypeNumberBase; diff --git a/src/DataTypes/DataTypeNumberBase.h b/src/DataTypes/DataTypeNumberBase.h index 3a5b11c5124..11b9427a14d 100644 --- a/src/DataTypes/DataTypeNumberBase.h +++ b/src/DataTypes/DataTypeNumberBase.h @@ -68,6 +68,7 @@ extern template class DataTypeNumberBase; extern template class DataTypeNumberBase; extern template class DataTypeNumberBase; extern template class DataTypeNumberBase; +extern template class DataTypeNumberBase; extern template class DataTypeNumberBase; extern template class DataTypeNumberBase; diff --git a/src/DataTypes/DataTypesDecimal.h b/src/DataTypes/DataTypesDecimal.h index e2b433cbe2f..12d061b11e5 100644 --- a/src/DataTypes/DataTypesDecimal.h +++ b/src/DataTypes/DataTypesDecimal.h @@ -3,6 +3,7 @@ #include #include #include +#include #include #include #include @@ -209,9 +210,9 @@ inline ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & static constexpr bool throw_exception = std::is_same_v; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { - if (!std::isfinite(value)) + if (!isFinite(value)) { if constexpr (throw_exception) throw Exception(ErrorCodes::DECIMAL_OVERFLOW, "{} convert overflow. Cannot convert infinity or NaN to decimal", ToDataType::family_name); diff --git a/src/DataTypes/DataTypesNumber.cpp b/src/DataTypes/DataTypesNumber.cpp index 1c0c418411b..81c64df9711 100644 --- a/src/DataTypes/DataTypesNumber.cpp +++ b/src/DataTypes/DataTypesNumber.cpp @@ -54,6 +54,7 @@ void registerDataTypeNumbers(DataTypeFactory & factory) factory.registerDataType("Int32", createNumericDataType); factory.registerDataType("Int64", createNumericDataType); + factory.registerDataType("BFloat16", createNumericDataType); factory.registerDataType("Float32", createNumericDataType); factory.registerDataType("Float64", createNumericDataType); diff --git a/src/DataTypes/DataTypesNumber.h b/src/DataTypes/DataTypesNumber.h index 0c1f88a7925..1fe95f58e99 100644 --- a/src/DataTypes/DataTypesNumber.h +++ b/src/DataTypes/DataTypesNumber.h @@ -63,6 +63,7 @@ using DataTypeInt8 = DataTypeNumber; using DataTypeInt16 = DataTypeNumber; using DataTypeInt32 = DataTypeNumber; using DataTypeInt64 = DataTypeNumber; +using DataTypeBFloat16 = DataTypeNumber; using DataTypeFloat32 = DataTypeNumber; using DataTypeFloat64 = DataTypeNumber; diff --git a/src/DataTypes/IDataType.h b/src/DataTypes/IDataType.h index eabf066bc3d..ac71a61683a 100644 --- a/src/DataTypes/IDataType.h +++ b/src/DataTypes/IDataType.h @@ -372,9 +372,10 @@ struct WhichDataType constexpr bool isDecimal256() const { return idx == TypeIndex::Decimal256; } constexpr bool isDecimal() const { return isDecimal32() || isDecimal64() || isDecimal128() || isDecimal256(); } + constexpr bool isBFloat16() const { return idx == TypeIndex::BFloat16; } constexpr bool isFloat32() const { return idx == TypeIndex::Float32; } constexpr bool isFloat64() const { return idx == TypeIndex::Float64; } - constexpr bool isFloat() const { return isFloat32() || isFloat64(); } + constexpr bool isFloat() const { return isBFloat16() || isFloat32() || isFloat64(); } constexpr bool isNativeNumber() const { return isNativeInteger() || isFloat(); } constexpr bool isNumber() const { return isInteger() || isFloat() || isDecimal(); } @@ -558,6 +559,7 @@ template inline constexpr bool IsDataTypeEnum> = tr M(Int16) \ M(Int32) \ M(Int64) \ + M(BFloat16) \ M(Float32) \ M(Float64) @@ -574,6 +576,7 @@ template inline constexpr bool IsDataTypeEnum> = tr M(Int64) \ M(Int128) \ M(Int256) \ + M(BFloat16) \ M(Float32) \ M(Float64) } diff --git a/src/DataTypes/NumberTraits.h b/src/DataTypes/NumberTraits.h index cf283d3358c..35a6238c71a 100644 --- a/src/DataTypes/NumberTraits.h +++ b/src/DataTypes/NumberTraits.h @@ -74,7 +74,7 @@ template struct ResultOfAdditionMultiplication { using Type = typename Construct< is_signed_v || is_signed_v, - std::is_floating_point_v || std::is_floating_point_v, + is_floating_point_v || is_floating_point_v, nextSize(max(sizeof(A), sizeof(B)))>::Type; }; @@ -82,7 +82,7 @@ template struct ResultOfSubtraction { using Type = typename Construct< true, - std::is_floating_point_v || std::is_floating_point_v, + is_floating_point_v || is_floating_point_v, nextSize(max(sizeof(A), sizeof(B)))>::Type; }; @@ -113,7 +113,7 @@ template struct ResultOfModulo /// Example: toInt32(-199) % toUInt8(200) will return -199 that does not fit in Int8, only in Int16. static constexpr size_t size_of_result = result_is_signed ? nextSize(sizeof(B)) : sizeof(B); using Type0 = typename Construct::Type; - using Type = std::conditional_t || std::is_floating_point_v, Float64, Type0>; + using Type = std::conditional_t || is_floating_point_v, Float64, Type0>; }; template struct ResultOfPositiveModulo @@ -121,21 +121,21 @@ template struct ResultOfPositiveModulo /// function positive_modulo always return non-negative number. static constexpr size_t size_of_result = sizeof(B); using Type0 = typename Construct::Type; - using Type = std::conditional_t || std::is_floating_point_v, Float64, Type0>; + using Type = std::conditional_t || is_floating_point_v, Float64, Type0>; }; template struct ResultOfModuloLegacy { using Type0 = typename Construct || is_signed_v, false, sizeof(B)>::Type; - using Type = std::conditional_t || std::is_floating_point_v, Float64, Type0>; + using Type = std::conditional_t || is_floating_point_v, Float64, Type0>; }; template struct ResultOfNegate { using Type = typename Construct< true, - std::is_floating_point_v, + is_floating_point_v, is_signed_v ? sizeof(A) : nextSize(sizeof(A))>::Type; }; @@ -143,7 +143,7 @@ template struct ResultOfAbs { using Type = typename Construct< false, - std::is_floating_point_v, + is_floating_point_v, sizeof(A)>::Type; }; @@ -154,7 +154,7 @@ template struct ResultOfBit using Type = typename Construct< is_signed_v || is_signed_v, false, - std::is_floating_point_v || std::is_floating_point_v ? 8 : max(sizeof(A), sizeof(B))>::Type; + is_floating_point_v || is_floating_point_v ? 8 : max(sizeof(A), sizeof(B))>::Type; }; template struct ResultOfBitNot @@ -180,7 +180,7 @@ template struct ResultOfBitNot template struct ResultOfIf { - static constexpr bool has_float = std::is_floating_point_v || std::is_floating_point_v; + static constexpr bool has_float = is_floating_point_v || is_floating_point_v; static constexpr bool has_integer = is_integer || is_integer; static constexpr bool has_signed = is_signed_v || is_signed_v; static constexpr bool has_unsigned = !is_signed_v || !is_signed_v; @@ -189,7 +189,7 @@ struct ResultOfIf static constexpr size_t max_size_of_unsigned_integer = max(is_signed_v ? 0 : sizeof(A), is_signed_v ? 0 : sizeof(B)); static constexpr size_t max_size_of_signed_integer = max(is_signed_v ? sizeof(A) : 0, is_signed_v ? sizeof(B) : 0); static constexpr size_t max_size_of_integer = max(is_integer ? sizeof(A) : 0, is_integer ? sizeof(B) : 0); - static constexpr size_t max_size_of_float = max(std::is_floating_point_v ? sizeof(A) : 0, std::is_floating_point_v ? sizeof(B) : 0); + static constexpr size_t max_size_of_float = max(is_floating_point_v ? sizeof(A) : 0, is_floating_point_v ? sizeof(B) : 0); using ConstructedType = typename Construct= max_size_of_float) @@ -211,7 +211,7 @@ template struct ToInteger using Type = typename Construct< is_signed_v, false, - std::is_floating_point_v ? 8 : sizeof(A)>::Type; + is_floating_point_v ? 8 : sizeof(A)>::Type; }; diff --git a/src/DataTypes/Serializations/SerializationNumber.cpp b/src/DataTypes/Serializations/SerializationNumber.cpp index b6c7e4618b8..805253fccee 100644 --- a/src/DataTypes/Serializations/SerializationNumber.cpp +++ b/src/DataTypes/Serializations/SerializationNumber.cpp @@ -176,6 +176,7 @@ template class SerializationNumber; template class SerializationNumber; template class SerializationNumber; template class SerializationNumber; +template class SerializationNumber; template class SerializationNumber; template class SerializationNumber; diff --git a/src/DataTypes/Utils.cpp b/src/DataTypes/Utils.cpp index e58331a8bcb..d1e314e77dc 100644 --- a/src/DataTypes/Utils.cpp +++ b/src/DataTypes/Utils.cpp @@ -54,6 +54,13 @@ bool canBeSafelyCasted(const DataTypePtr & from_type, const DataTypePtr & to_typ return false; } + case TypeIndex::BFloat16: + { + if (to_which_type.isFloat32() || to_which_type.isFloat64() || to_which_type.isString()) + return true; + + return false; + } case TypeIndex::Float32: { if (to_which_type.isFloat64() || to_which_type.isString()) diff --git a/src/DataTypes/getLeastSupertype.cpp b/src/DataTypes/getLeastSupertype.cpp index e5bdb4b267f..0ed075563e2 100644 --- a/src/DataTypes/getLeastSupertype.cpp +++ b/src/DataTypes/getLeastSupertype.cpp @@ -108,6 +108,8 @@ DataTypePtr getNumericType(const TypeIndexSet & types) maximize(max_bits_of_signed_integer, 128); else if (type == TypeIndex::Int256) maximize(max_bits_of_signed_integer, 256); + else if (type == TypeIndex::BFloat16) + maximize(max_mantissa_bits_of_floating, 8); else if (type == TypeIndex::Float32) maximize(max_mantissa_bits_of_floating, 24); else if (type == TypeIndex::Float64) @@ -144,7 +146,9 @@ DataTypePtr getNumericType(const TypeIndexSet & types) if (max_mantissa_bits_of_floating) { size_t min_mantissa_bits = std::max(min_bit_width_of_integer, max_mantissa_bits_of_floating); - if (min_mantissa_bits <= 24) + if (min_mantissa_bits <= 8) + return std::make_shared(); + else if (min_mantissa_bits <= 24) return std::make_shared(); else if (min_mantissa_bits <= 53) return std::make_shared(); diff --git a/src/DataTypes/getMostSubtype.cpp b/src/DataTypes/getMostSubtype.cpp index 33b5735456e..d0ea716f2ff 100644 --- a/src/DataTypes/getMostSubtype.cpp +++ b/src/DataTypes/getMostSubtype.cpp @@ -297,6 +297,8 @@ DataTypePtr getMostSubtype(const DataTypes & types, bool throw_if_result_is_noth minimize(min_bits_of_signed_integer, 128); else if (typeid_cast(type.get())) minimize(min_bits_of_signed_integer, 256); + else if (typeid_cast(type.get())) + minimize(min_mantissa_bits_of_floating, 8); else if (typeid_cast(type.get())) minimize(min_mantissa_bits_of_floating, 24); else if (typeid_cast(type.get())) @@ -313,7 +315,9 @@ DataTypePtr getMostSubtype(const DataTypes & types, bool throw_if_result_is_noth /// If the result must be floating. if (!min_bits_of_signed_integer && !min_bits_of_unsigned_integer) { - if (min_mantissa_bits_of_floating <= 24) + if (min_mantissa_bits_of_floating <= 8) + return std::make_shared(); + else if (min_mantissa_bits_of_floating <= 24) return std::make_shared(); else if (min_mantissa_bits_of_floating <= 53) return std::make_shared(); diff --git a/src/Formats/ProtobufSerializer.cpp b/src/Formats/ProtobufSerializer.cpp index dd37c25719c..872991709af 100644 --- a/src/Formats/ProtobufSerializer.cpp +++ b/src/Formats/ProtobufSerializer.cpp @@ -540,7 +540,7 @@ namespace case FieldTypeId::TYPE_ENUM: { - if (std::is_floating_point_v) + if (is_floating_point_v) incompatibleColumnType(TypeName); write_function = [this](NumberType value) diff --git a/src/Functions/DivisionUtils.h b/src/Functions/DivisionUtils.h index ff07309e248..2508bd2b62b 100644 --- a/src/Functions/DivisionUtils.h +++ b/src/Functions/DivisionUtils.h @@ -47,9 +47,9 @@ inline auto checkedDivision(A a, B b) { throwIfDivisionLeadsToFPE(a, b); - if constexpr (is_big_int_v && std::is_floating_point_v) + if constexpr (is_big_int_v && is_floating_point_v) return static_cast(a) / b; - else if constexpr (is_big_int_v && std::is_floating_point_v) + else if constexpr (is_big_int_v && is_floating_point_v) return a / static_cast(b); else if constexpr (is_big_int_v && is_big_int_v) return static_cast(a / b); @@ -86,17 +86,17 @@ struct DivideIntegralImpl { /// Comparisons are not strict to avoid rounding issues when operand is implicitly casted to float. - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) if (isNaN(a) || a >= std::numeric_limits::max() || a <= std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) if (isNaN(b) || b >= std::numeric_limits::max() || b <= std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); auto res = checkedDivision(CastA(a), CastB(b)); - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) if (isNaN(res) || res >= static_cast(std::numeric_limits::max()) || res <= std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division, because it will produce infinite or too large number"); @@ -122,18 +122,18 @@ struct ModuloImpl template static inline Result apply(A a, B b) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { /// This computation is similar to `fmod` but the latter is not inlined and has 40 times worse performance. return static_cast(a) - trunc(static_cast(a) / static_cast(b)) * static_cast(b); } else { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) if (isNaN(a) || a > std::numeric_limits::max() || a < std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) if (isNaN(b) || b > std::numeric_limits::max() || b < std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); diff --git a/src/Functions/FunctionMathUnary.h b/src/Functions/FunctionMathUnary.h index 9f400932356..8395855a564 100644 --- a/src/Functions/FunctionMathUnary.h +++ b/src/Functions/FunctionMathUnary.h @@ -66,7 +66,7 @@ private: /// Process all data as a whole and use FastOps implementation /// If the argument is integer, convert to Float64 beforehand - if constexpr (!std::is_floating_point_v) + if constexpr (!is_floating_point_v) { PODArray tmp_vec(size); for (size_t i = 0; i < size; ++i) @@ -150,7 +150,7 @@ private: { using Types = std::decay_t; using Type = typename Types::RightType; - using ReturnType = std::conditional_t, Float64, Type>; + using ReturnType = std::conditional_t, Float64, Type>; using ColVecType = ColumnVectorOrDecimal; const auto col_vec = checkAndGetColumn(col.column.get()); diff --git a/src/Functions/FunctionsConversion.h b/src/Functions/FunctionsConversion.h index eed75788fcd..fe4b14f5053 100644 --- a/src/Functions/FunctionsConversion.h +++ b/src/Functions/FunctionsConversion.h @@ -291,7 +291,7 @@ struct ConvertImpl else { /// If From Data is Nan or Inf and we convert to integer type, throw exception - if constexpr (std::is_floating_point_v && !std::is_floating_point_v) + if constexpr (is_floating_point_v && !is_floating_point_v) { if (!isFinite(vec_from[i])) { @@ -1314,7 +1314,7 @@ inline void convertFromTime(DataTypeDateTime::FieldType & x, t template void parseImpl(typename DataType::FieldType & x, ReadBuffer & rb, const DateLUTImpl *, bool precise_float_parsing) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { if (precise_float_parsing) readFloatTextPrecise(x, rb); @@ -1378,7 +1378,7 @@ inline void parseImpl(DataTypeIPv6::FieldType & x, ReadBuffer & rb template bool tryParseImpl(typename DataType::FieldType & x, ReadBuffer & rb, const DateLUTImpl *, bool precise_float_parsing) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { if (precise_float_parsing) return tryReadFloatTextPrecise(x, rb); @@ -2350,9 +2350,9 @@ private: using RightT = typename RightDataType::FieldType; static constexpr bool bad_left = - is_decimal || std::is_floating_point_v || is_big_int_v || is_signed_v; + is_decimal || is_floating_point_v || is_big_int_v || is_signed_v; static constexpr bool bad_right = - is_decimal || std::is_floating_point_v || is_big_int_v || is_signed_v; + is_decimal || is_floating_point_v || is_big_int_v || is_signed_v; /// Disallow int vs UUID conversion (but support int vs UInt128 conversion) if constexpr ((bad_left && std::is_same_v) || @@ -2678,7 +2678,7 @@ struct ToNumberMonotonicity /// Float cases. /// When converting to Float, the conversion is always monotonic. - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) return { .is_monotonic = true, .is_always_monotonic = true }; const auto * low_cardinality = typeid_cast(&type); diff --git a/src/Functions/FunctionsRound.h b/src/Functions/FunctionsRound.h index 3d1028c6d35..d775d616eb2 100644 --- a/src/Functions/FunctionsRound.h +++ b/src/Functions/FunctionsRound.h @@ -461,7 +461,7 @@ template - using FunctionRoundingImpl = std::conditional_t, + using FunctionRoundingImpl = std::conditional_t, FloatRoundingImpl, IntegerRoundingImpl>; diff --git a/src/Functions/factorial.cpp b/src/Functions/factorial.cpp index b814e8198e6..be545e398cd 100644 --- a/src/Functions/factorial.cpp +++ b/src/Functions/factorial.cpp @@ -21,7 +21,7 @@ struct FactorialImpl static inline NO_SANITIZE_UNDEFINED ResultType apply(A a) { - if constexpr (std::is_floating_point_v || is_over_big_int) + if constexpr (is_floating_point_v || is_over_big_int) throw Exception( ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Illegal type of argument of function factorial, should not be floating point or big int"); diff --git a/src/Functions/minus.cpp b/src/Functions/minus.cpp index 04877a42b18..109e5894f5e 100644 --- a/src/Functions/minus.cpp +++ b/src/Functions/minus.cpp @@ -17,8 +17,8 @@ struct MinusImpl { if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) - static_cast(static_cast(b)); } diff --git a/src/Functions/moduloOrZero.cpp b/src/Functions/moduloOrZero.cpp index 3551ae74c5f..bfd786940ce 100644 --- a/src/Functions/moduloOrZero.cpp +++ b/src/Functions/moduloOrZero.cpp @@ -17,7 +17,7 @@ struct ModuloOrZeroImpl template static inline Result apply(A a, B b) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { /// This computation is similar to `fmod` but the latter is not inlined and has 40 times worse performance. return ResultType(a) - trunc(ResultType(a) / ResultType(b)) * ResultType(b); diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp index 4dc8cd10f31..ef51fe6061e 100644 --- a/src/Functions/multiply.cpp +++ b/src/Functions/multiply.cpp @@ -18,8 +18,8 @@ struct MultiplyImpl { if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) * static_cast(static_cast(b)); } diff --git a/src/Functions/plus.cpp b/src/Functions/plus.cpp index cd9cf6cec5c..ea79fb4702a 100644 --- a/src/Functions/plus.cpp +++ b/src/Functions/plus.cpp @@ -19,8 +19,8 @@ struct PlusImpl /// Next everywhere, static_cast - so that there is no wrong result in expressions of the form Int64 c = UInt32(a) * Int32(-1). if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) + static_cast(static_cast(b)); } diff --git a/src/Functions/sign.cpp b/src/Functions/sign.cpp index 6c849760eed..59a307e43bb 100644 --- a/src/Functions/sign.cpp +++ b/src/Functions/sign.cpp @@ -13,7 +13,7 @@ struct SignImpl static inline NO_SANITIZE_UNDEFINED ResultType apply(A a) { - if constexpr (is_decimal || std::is_floating_point_v) + if constexpr (is_decimal || is_floating_point_v) return a < A(0) ? -1 : a == A(0) ? 0 : 1; else if constexpr (is_signed_v) return a < 0 ? -1 : a == 0 ? 0 : 1; diff --git a/src/IO/ReadHelpers.h b/src/IO/ReadHelpers.h index 85584d63ee8..6068f49f5bf 100644 --- a/src/IO/ReadHelpers.h +++ b/src/IO/ReadHelpers.h @@ -1316,7 +1316,9 @@ inline bool tryReadText(UUID & x, ReadBuffer & buf) { return tryReadUUIDText(x, inline bool tryReadText(IPv4 & x, ReadBuffer & buf) { return tryReadIPv4Text(x, buf); } inline bool tryReadText(IPv6 & x, ReadBuffer & buf) { return tryReadIPv6Text(x, buf); } -inline void readText(is_floating_point auto & x, ReadBuffer & buf) { readFloatText(x, buf); } +template +requires is_floating_point_v +inline void readText(T & x, ReadBuffer & buf) { readFloatText(x, buf); } inline void readText(String & x, ReadBuffer & buf) { readEscapedString(x, buf); } diff --git a/src/IO/WriteHelpers.h b/src/IO/WriteHelpers.h index b4f8b476b11..c6a86b05f4d 100644 --- a/src/IO/WriteHelpers.h +++ b/src/IO/WriteHelpers.h @@ -153,6 +153,7 @@ inline void writeBoolText(bool x, WriteBuffer & buf) template +requires is_floating_point_v inline size_t writeFloatTextFastPath(T x, char * buffer) { Int64 result = 0; @@ -169,10 +170,13 @@ inline size_t writeFloatTextFastPath(T x, char * buffer) } else { - if (DecomposedFloat32(x).isIntegerInRepresentableRange()) - result = itoa(Int32(x), buffer) - buffer; + /// This will support 16-bit floats as well. + float f32 = x; + + if (DecomposedFloat32(f32).isIntegerInRepresentableRange()) + result = itoa(Int32(f32), buffer) - buffer; else - result = jkj::dragonbox::to_chars_n(x, buffer) - buffer; + result = jkj::dragonbox::to_chars_n(f32, buffer) - buffer; } if (result <= 0) @@ -181,10 +185,9 @@ inline size_t writeFloatTextFastPath(T x, char * buffer) } template +requires is_floating_point_v inline void writeFloatText(T x, WriteBuffer & buf) { - static_assert(std::is_same_v || std::is_same_v, "Argument for writeFloatText must be float or double"); - using Converter = DoubleConverter; if (likely(buf.available() >= Converter::MAX_REPRESENTATION_LENGTH)) { @@ -530,7 +533,7 @@ void writeJSONNumber(T x, WriteBuffer & ostr, const FormatSettings & settings) bool is_finite = isFinite(x); const bool need_quote = (is_integer && (sizeof(T) >= 8) && settings.json.quote_64bit_integers) - || (settings.json.quote_denormals && !is_finite) || (is_floating_point && (sizeof(T) >= 8) && settings.json.quote_64bit_floats); + || (settings.json.quote_denormals && !is_finite) || (is_floating_point_v && (sizeof(T) >= 8) && settings.json.quote_64bit_floats); if (need_quote) writeChar('"', ostr); @@ -541,7 +544,7 @@ void writeJSONNumber(T x, WriteBuffer & ostr, const FormatSettings & settings) writeCString("null", ostr); else { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point_v) { if (std::signbit(x)) { @@ -800,7 +803,6 @@ inline void writeXMLStringForTextElement(std::string_view s, WriteBuffer & buf) } /// @brief Serialize `uuid` into an array of characters in big-endian byte order. -/// @param uuid UUID to serialize. /// @return Array of characters in big-endian byte order. std::array formatUUID(const UUID & uuid); @@ -1065,7 +1067,9 @@ inline void writeText(is_integer auto x, WriteBuffer & buf) writeIntText(x, buf); } -inline void writeText(is_floating_point auto x, WriteBuffer & buf) { writeFloatText(x, buf); } +template +requires is_floating_point_v +inline void writeText(T x, WriteBuffer & buf) { writeFloatText(x, buf); } inline void writeText(is_enum auto x, WriteBuffer & buf) { writeText(magic_enum::enum_name(x), buf); } diff --git a/src/Interpreters/RowRefs.cpp b/src/Interpreters/RowRefs.cpp index 4335cde47f9..61caacd8346 100644 --- a/src/Interpreters/RowRefs.cpp +++ b/src/Interpreters/RowRefs.cpp @@ -181,7 +181,7 @@ private: if (!sorted.load(std::memory_order_relaxed)) { - if constexpr (std::is_arithmetic_v && !std::is_floating_point_v) + if constexpr (std::is_arithmetic_v && !is_floating_point_v) { if (likely(entries.size() > 256)) { From 2750f8ca1d9b7336fbfec6fc0e73a8fbf17eadee Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 2 Jun 2024 02:27:48 +0200 Subject: [PATCH 002/244] Whitespace --- src/Storages/StorageSet.cpp | 1 - 1 file changed, 1 deletion(-) diff --git a/src/Storages/StorageSet.cpp b/src/Storages/StorageSet.cpp index 205a90423bf..a8c8e81e23d 100644 --- a/src/Storages/StorageSet.cpp +++ b/src/Storages/StorageSet.cpp @@ -130,7 +130,6 @@ StorageSetOrJoinBase::StorageSetOrJoinBase( storage_metadata.setComment(comment); setInMemoryMetadata(storage_metadata); - if (relative_path_.empty()) throw Exception(ErrorCodes::INCORRECT_FILE_NAME, "Join and Set storages require data path"); From 6e08f415c49afeac27ce08f97cde365dbf5940a2 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 2 Jun 2024 04:26:14 +0200 Subject: [PATCH 003/244] Preparation --- base/base/DecomposedFloat.h | 9 ++++++++ base/base/EnumReflection.h | 2 +- base/base/extended_types.h | 14 ++++-------- base/base/wide_integer_impl.h | 2 +- .../AggregateFunctionGroupArray.cpp | 4 ++-- .../AggregateFunctionGroupArrayMoving.cpp | 2 +- .../AggregateFunctionIntervalLengthSum.cpp | 4 ++-- .../AggregateFunctionSparkbar.cpp | 6 ++--- src/AggregateFunctions/AggregateFunctionSum.h | 6 ++--- src/AggregateFunctions/QuantileTDigest.h | 2 +- src/AggregateFunctions/ReservoirSampler.h | 2 +- .../ReservoirSamplerDeterministic.h | 2 +- src/Columns/ColumnVector.cpp | 16 +++++++------- src/Common/FieldVisitorConvertToNumber.h | 4 ++-- src/Common/HashTable/HashTable.h | 2 +- src/Common/NaNUtils.h | 14 ++++++------ src/Common/findExtreme.cpp | 2 +- src/Common/transformEndianness.h | 2 +- src/Core/AccurateComparison.h | 18 +++++++-------- src/Core/DecimalFunctions.h | 2 +- src/DataTypes/DataTypesDecimal.cpp | 5 +++-- src/DataTypes/NumberTraits.h | 22 +++++++++---------- src/Formats/ProtobufSerializer.cpp | 2 +- src/Functions/DivisionUtils.h | 16 +++++++------- src/Functions/FunctionMathUnary.h | 4 ++-- src/Functions/FunctionsConversion.cpp | 12 +++++----- src/Functions/FunctionsJSON.h | 4 ++-- src/Functions/FunctionsRound.h | 2 +- src/Functions/FunctionsVisitParam.h | 2 +- src/Functions/abs.cpp | 2 +- src/Functions/array/arrayAggregation.cpp | 2 +- src/Functions/factorial.cpp | 2 +- src/Functions/if.cpp | 16 +++++++------- src/Functions/minus.cpp | 4 ++-- src/Functions/moduloOrZero.cpp | 2 +- src/Functions/multiply.cpp | 4 ++-- src/Functions/plus.cpp | 4 ++-- src/Functions/sign.cpp | 2 +- src/IO/ReadHelpers.h | 2 +- src/IO/WriteHelpers.h | 10 ++++----- src/Interpreters/RowRefs.cpp | 2 +- 41 files changed, 120 insertions(+), 116 deletions(-) diff --git a/base/base/DecomposedFloat.h b/base/base/DecomposedFloat.h index 0997c39db16..b5bc3f08357 100644 --- a/base/base/DecomposedFloat.h +++ b/base/base/DecomposedFloat.h @@ -96,6 +96,15 @@ struct DecomposedFloat && ((mantissa() & ((1ULL << (Traits::mantissa_bits - normalizedExponent())) - 1)) == 0)); } + bool isFinite() const + { + return exponent() != ((1ull << Traits::exponent_bits) - 1); + } + + bool isNaN() const + { + return !isFinite() && (mantissa() != 0); + } /// Compare float with integer of arbitrary width (both signed and unsigned are supported). Assuming two's complement arithmetic. /// This function is generic, big integers (128, 256 bit) are supported as well. diff --git a/base/base/EnumReflection.h b/base/base/EnumReflection.h index 4a9de4d17a3..963c7e3f1b9 100644 --- a/base/base/EnumReflection.h +++ b/base/base/EnumReflection.h @@ -4,7 +4,7 @@ #include -template concept is_enum = std::is_enum_v; +template concept is_enum = std::is_enum_v; namespace detail { diff --git a/base/base/extended_types.h b/base/base/extended_types.h index de654152649..7ddf7de7e22 100644 --- a/base/base/extended_types.h +++ b/base/base/extended_types.h @@ -43,7 +43,7 @@ template <> struct is_unsigned { static constexpr bool value = true; }; template inline constexpr bool is_unsigned_v = is_unsigned::value; -template concept is_integer = +template concept is_integer = std::is_integral_v || std::is_same_v || std::is_same_v @@ -65,16 +65,10 @@ template <> struct is_arithmetic { static constexpr bool value = true; template inline constexpr bool is_arithmetic_v = is_arithmetic::value; -template -struct is_floating_point // NOLINT(readability-identifier-naming) -{ - static constexpr bool value = std::is_floating_point_v; -}; +template concept is_floating_point = + std::is_floating_point_v + || std::is_same_v; -template <> struct is_floating_point { static constexpr bool value = true; }; - -template -inline constexpr bool is_floating_point_v = is_floating_point::value; #define FOR_EACH_ARITHMETIC_TYPE(M) \ M(DataTypeDate) \ diff --git a/base/base/wide_integer_impl.h b/base/base/wide_integer_impl.h index c950fd27fa3..d0bbd7df9d4 100644 --- a/base/base/wide_integer_impl.h +++ b/base/base/wide_integer_impl.h @@ -154,7 +154,7 @@ struct common_type, Arithmetic> static_assert(wide::ArithmeticConcept()); using type = std::conditional_t< - is_floating_point_v, + std::is_floating_point_v || std::is_same_v, Arithmetic, std::conditional_t< sizeof(Arithmetic) * 8 < Bits, diff --git a/src/AggregateFunctions/AggregateFunctionGroupArray.cpp b/src/AggregateFunctions/AggregateFunctionGroupArray.cpp index 0b478fe3c04..3a0bbb001c3 100644 --- a/src/AggregateFunctions/AggregateFunctionGroupArray.cpp +++ b/src/AggregateFunctions/AggregateFunctionGroupArray.cpp @@ -73,7 +73,7 @@ template struct GroupArraySamplerData { /// For easy serialization. - static_assert(std::has_unique_object_representations_v || is_floating_point_v); + static_assert(std::has_unique_object_representations_v || is_floating_point); // Switch to ordinary Allocator after 4096 bytes to avoid fragmentation and trash in Arena using Allocator = MixedAlignedArenaAllocator; @@ -115,7 +115,7 @@ template struct GroupArrayNumericData { /// For easy serialization. - static_assert(std::has_unique_object_representations_v || is_floating_point_v); + static_assert(std::has_unique_object_representations_v || is_floating_point); // Switch to ordinary Allocator after 4096 bytes to avoid fragmentation and trash in Arena using Allocator = MixedAlignedArenaAllocator; diff --git a/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp b/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp index ee6a82686c5..a9a09d7abd5 100644 --- a/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp +++ b/src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp @@ -38,7 +38,7 @@ template struct MovingData { /// For easy serialization. - static_assert(std::has_unique_object_representations_v || is_floating_point_v); + static_assert(std::has_unique_object_representations_v || is_floating_point); using Accumulator = T; diff --git a/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp b/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp index 06156643aa0..e5404add820 100644 --- a/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp +++ b/src/AggregateFunctions/AggregateFunctionIntervalLengthSum.cpp @@ -187,7 +187,7 @@ public: static DataTypePtr createResultType() { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) return std::make_shared(); return std::make_shared(); } @@ -227,7 +227,7 @@ public: void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena *) const override { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) assert_cast(to).getData().push_back(getIntervalLengthSum(this->data(place))); else assert_cast(to).getData().push_back(getIntervalLengthSum(this->data(place))); diff --git a/src/AggregateFunctions/AggregateFunctionSparkbar.cpp b/src/AggregateFunctions/AggregateFunctionSparkbar.cpp index 5b6fc3b315c..33412d50b21 100644 --- a/src/AggregateFunctions/AggregateFunctionSparkbar.cpp +++ b/src/AggregateFunctions/AggregateFunctionSparkbar.cpp @@ -50,7 +50,7 @@ struct AggregateFunctionSparkbarData auto [it, inserted] = points.insert({x, y}); if (!inserted) { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { it->getMapped() += y; return it->getMapped(); @@ -197,7 +197,7 @@ private: Y res; bool has_overfllow = false; - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) res = histogram[index] + point.getMapped(); else has_overfllow = common::addOverflow(histogram[index], point.getMapped(), res); @@ -246,7 +246,7 @@ private: } constexpr auto levels_num = static_cast(BAR_LEVELS - 1); - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { y = y / (y_max / levels_num) + 1; } diff --git a/src/AggregateFunctions/AggregateFunctionSum.h b/src/AggregateFunctions/AggregateFunctionSum.h index c663c632280..d0d600be70b 100644 --- a/src/AggregateFunctions/AggregateFunctionSum.h +++ b/src/AggregateFunctions/AggregateFunctionSum.h @@ -69,7 +69,7 @@ struct AggregateFunctionSumData size_t count = end - start; const auto * end_ptr = ptr + count; - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { /// Compiler cannot unroll this loop, do it manually. /// (at least for floats, most likely due to the lack of -fassociative-math) @@ -193,7 +193,7 @@ struct AggregateFunctionSumData Impl::add(sum, local_sum); return; } - else if constexpr (is_floating_point_v) + else if constexpr (is_floating_point) { /// For floating point we use a similar trick as above, except that now we reinterpret the floating point number as an unsigned /// integer of the same size and use a mask instead (0 to discard, 0xFF..FF to keep) @@ -306,7 +306,7 @@ struct AggregateFunctionSumData template struct AggregateFunctionSumKahanData { - static_assert(is_floating_point_v, + static_assert(is_floating_point, "It doesn't make sense to use Kahan Summation algorithm for non floating point types"); T sum{}; diff --git a/src/AggregateFunctions/QuantileTDigest.h b/src/AggregateFunctions/QuantileTDigest.h index 408e500e941..a693c57e6d8 100644 --- a/src/AggregateFunctions/QuantileTDigest.h +++ b/src/AggregateFunctions/QuantileTDigest.h @@ -379,7 +379,7 @@ public: ResultType getImpl(Float64 level) { if (centroids.empty()) - return is_floating_point_v ? std::numeric_limits::quiet_NaN() : 0; + return is_floating_point ? std::numeric_limits::quiet_NaN() : 0; compress(); diff --git a/src/AggregateFunctions/ReservoirSampler.h b/src/AggregateFunctions/ReservoirSampler.h index 182a49af2ca..c21e76614c1 100644 --- a/src/AggregateFunctions/ReservoirSampler.h +++ b/src/AggregateFunctions/ReservoirSampler.h @@ -278,6 +278,6 @@ private: if (OnEmpty == ReservoirSamplerOnEmpty::THROW) throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Quantile of empty ReservoirSampler"); else - return NanLikeValueConstructor>::getValue(); + return NanLikeValueConstructor>::getValue(); } }; diff --git a/src/AggregateFunctions/ReservoirSamplerDeterministic.h b/src/AggregateFunctions/ReservoirSamplerDeterministic.h index c9afcb21549..7fe5d23f4e4 100644 --- a/src/AggregateFunctions/ReservoirSamplerDeterministic.h +++ b/src/AggregateFunctions/ReservoirSamplerDeterministic.h @@ -272,7 +272,7 @@ private: if (OnEmpty == ReservoirSamplerDeterministicOnEmpty::THROW) throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Quantile of empty ReservoirSamplerDeterministic"); else - return NanLikeValueConstructor>::getValue(); + return NanLikeValueConstructor>::getValue(); } }; diff --git a/src/Columns/ColumnVector.cpp b/src/Columns/ColumnVector.cpp index 19849b8a1c6..2b137231faa 100644 --- a/src/Columns/ColumnVector.cpp +++ b/src/Columns/ColumnVector.cpp @@ -118,7 +118,7 @@ struct ColumnVector::less_stable if (unlikely(parent.data[lhs] == parent.data[rhs])) return lhs < rhs; - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { if (unlikely(std::isnan(parent.data[lhs]) && std::isnan(parent.data[rhs]))) { @@ -150,7 +150,7 @@ struct ColumnVector::greater_stable if (unlikely(parent.data[lhs] == parent.data[rhs])) return lhs < rhs; - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { if (unlikely(std::isnan(parent.data[lhs]) && std::isnan(parent.data[rhs]))) { @@ -224,9 +224,9 @@ void ColumnVector::getPermutation(IColumn::PermutationSortDirection direction iota(res.data(), data_size, IColumn::Permutation::value_type(0)); - if constexpr (has_find_extreme_implementation && !std::is_floating_point_v) + if constexpr (has_find_extreme_implementation && !is_floating_point) { - /// Disabled for:floating point + /// Disabled for floating point: /// * floating point: We don't deal with nan_direction_hint /// * stability::Stable: We might return any value, not the first if ((limit == 1) && (stability == IColumn::PermutationSortStability::Unstable)) @@ -256,7 +256,7 @@ void ColumnVector::getPermutation(IColumn::PermutationSortDirection direction bool sort_is_stable = stability == IColumn::PermutationSortStability::Stable; /// TODO: LSD RadixSort is currently not stable if direction is descending, or value is floating point - bool use_radix_sort = (sort_is_stable && ascending && !is_floating_point_v) || !sort_is_stable; + bool use_radix_sort = (sort_is_stable && ascending && !is_floating_point) || !sort_is_stable; /// Thresholds on size. Lower threshold is arbitrary. Upper threshold is chosen by the type for histogram counters. if (data_size >= 256 && data_size <= std::numeric_limits::max() && use_radix_sort) @@ -283,7 +283,7 @@ void ColumnVector::getPermutation(IColumn::PermutationSortDirection direction /// Radix sort treats all NaNs to be greater than all numbers. /// If the user needs the opposite, we must move them accordingly. - if (is_floating_point_v && nan_direction_hint < 0) + if (is_floating_point && nan_direction_hint < 0) { size_t nans_to_move = 0; @@ -330,7 +330,7 @@ void ColumnVector::updatePermutation(IColumn::PermutationSortDirection direct if constexpr (is_arithmetic_v && !is_big_int_v) { /// TODO: LSD RadixSort is currently not stable if direction is descending, or value is floating point - bool use_radix_sort = (sort_is_stable && ascending && !is_floating_point_v) || !sort_is_stable; + bool use_radix_sort = (sort_is_stable && ascending && !is_floating_point) || !sort_is_stable; size_t size = end - begin; /// Thresholds on size. Lower threshold is arbitrary. Upper threshold is chosen by the type for histogram counters. @@ -353,7 +353,7 @@ void ColumnVector::updatePermutation(IColumn::PermutationSortDirection direct /// Radix sort treats all NaNs to be greater than all numbers. /// If the user needs the opposite, we must move them accordingly. - if (is_floating_point_v && nan_direction_hint < 0) + if (is_floating_point && nan_direction_hint < 0) { size_t nans_to_move = 0; diff --git a/src/Common/FieldVisitorConvertToNumber.h b/src/Common/FieldVisitorConvertToNumber.h index 646caadce35..ebd084df54d 100644 --- a/src/Common/FieldVisitorConvertToNumber.h +++ b/src/Common/FieldVisitorConvertToNumber.h @@ -58,7 +58,7 @@ public: T operator() (const Float64 & x) const { - if constexpr (!is_floating_point_v) + if constexpr (!is_floating_point) { if (!isFinite(x)) { @@ -88,7 +88,7 @@ public: template T operator() (const DecimalField & x) const { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) return x.getValue().template convertTo() / x.getScaleMultiplier().template convertTo(); else return (x.getValue() / x.getScaleMultiplier()).template convertTo(); diff --git a/src/Common/HashTable/HashTable.h b/src/Common/HashTable/HashTable.h index fd8832a56a3..8237c81461f 100644 --- a/src/Common/HashTable/HashTable.h +++ b/src/Common/HashTable/HashTable.h @@ -91,7 +91,7 @@ inline bool bitEquals(T && a, T && b) { using RealT = std::decay_t; - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) /// Note that memcmp with constant size is compiler builtin. return 0 == memcmp(&a, &b, sizeof(RealT)); /// NOLINT else diff --git a/src/Common/NaNUtils.h b/src/Common/NaNUtils.h index 0e885541599..3e4af902104 100644 --- a/src/Common/NaNUtils.h +++ b/src/Common/NaNUtils.h @@ -3,24 +3,24 @@ #include #include #include +#include template inline bool isNaN(T x) { /// To be sure, that this function is zero-cost for non-floating point types. - if constexpr (is_floating_point_v) - return std::isnan(x); + if constexpr (is_floating_point) + return DecomposedFloat(x).isNaN(); else return false; } - template inline bool isFinite(T x) { - if constexpr (is_floating_point_v) - return std::isfinite(x); + if constexpr (is_floating_point) + return DecomposedFloat(x).isFinite(); else return true; } @@ -28,7 +28,7 @@ inline bool isFinite(T x) template bool canConvertTo(Float64 x) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) return true; if (!isFinite(x)) return false; @@ -41,7 +41,7 @@ bool canConvertTo(Float64 x) template T NaNOrZero() { - if constexpr (is_floating_point_v) + if constexpr (std::is_floating_point_v) return std::numeric_limits::quiet_NaN(); else return {}; diff --git a/src/Common/findExtreme.cpp b/src/Common/findExtreme.cpp index ce3bbb86d7c..a29750b848a 100644 --- a/src/Common/findExtreme.cpp +++ b/src/Common/findExtreme.cpp @@ -47,7 +47,7 @@ MULTITARGET_FUNCTION_AVX2_SSE42( /// Unroll the loop manually for floating point, since the compiler doesn't do it without fastmath /// as it might change the return value - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { constexpr size_t unroll_block = 512 / sizeof(T); /// Chosen via benchmarks with AVX2 so YMMV size_t unrolled_end = i + (((count - i) / unroll_block) * unroll_block); diff --git a/src/Common/transformEndianness.h b/src/Common/transformEndianness.h index 2a0c45efe38..e6e04ec75af 100644 --- a/src/Common/transformEndianness.h +++ b/src/Common/transformEndianness.h @@ -38,7 +38,7 @@ inline void transformEndianness(T & x) } template -requires is_floating_point_v +requires is_floating_point inline void transformEndianness(T & value) { if constexpr (ToEndian != FromEndian) diff --git a/src/Core/AccurateComparison.h b/src/Core/AccurateComparison.h index c1e93b8055a..87ff14e40e7 100644 --- a/src/Core/AccurateComparison.h +++ b/src/Core/AccurateComparison.h @@ -25,7 +25,7 @@ bool lessOp(A a, B b) return a < b; /// float vs float - if constexpr (is_floating_point_v && is_floating_point_v) + if constexpr (is_floating_point && is_floating_point) return a < b; /// anything vs NaN @@ -49,7 +49,7 @@ bool lessOp(A a, B b) } /// int vs float - if constexpr (is_integer && is_floating_point_v) + if constexpr (is_integer && is_floating_point) { if constexpr (sizeof(A) <= 4) return static_cast(a) < static_cast(b); @@ -57,7 +57,7 @@ bool lessOp(A a, B b) return DecomposedFloat(b).greater(a); } - if constexpr (is_floating_point_v && is_integer) + if constexpr (is_floating_point && is_integer) { if constexpr (sizeof(B) <= 4) return static_cast(a) < static_cast(b); @@ -65,8 +65,8 @@ bool lessOp(A a, B b) return DecomposedFloat(a).less(b); } - static_assert(is_integer || is_floating_point_v); - static_assert(is_integer || is_floating_point_v); + static_assert(is_integer || is_floating_point); + static_assert(is_integer || is_floating_point); UNREACHABLE(); } @@ -101,7 +101,7 @@ bool equalsOp(A a, B b) return a == b; /// float vs float - if constexpr (is_floating_point_v && is_floating_point_v) + if constexpr (is_floating_point && is_floating_point) return a == b; /// anything vs NaN @@ -125,7 +125,7 @@ bool equalsOp(A a, B b) } /// int vs float - if constexpr (is_integer && is_floating_point_v) + if constexpr (is_integer && is_floating_point) { if constexpr (sizeof(A) <= 4) return static_cast(a) == static_cast(b); @@ -133,7 +133,7 @@ bool equalsOp(A a, B b) return DecomposedFloat(b).equals(a); } - if constexpr (is_floating_point_v && is_integer) + if constexpr (is_floating_point && is_integer) { if constexpr (sizeof(B) <= 4) return static_cast(a) == static_cast(b); @@ -163,7 +163,7 @@ inline bool NO_SANITIZE_UNDEFINED convertNumeric(From value, To & result) return true; } - if constexpr (is_floating_point_v && is_floating_point_v) + if constexpr (is_floating_point && is_floating_point) { /// Note that NaNs doesn't compare equal to anything, but they are still in range of any Float type. if (isNaN(value)) diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h index c5bc4ad70f6..435cef61145 100644 --- a/src/Core/DecimalFunctions.h +++ b/src/Core/DecimalFunctions.h @@ -310,7 +310,7 @@ ReturnType convertToImpl(const DecimalType & decimal, UInt32 scale, To & result) using DecimalNativeType = typename DecimalType::NativeType; static constexpr bool throw_exception = std::is_void_v; - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { result = static_cast(decimal.value) / static_cast(scaleMultiplier(scale)); } diff --git a/src/DataTypes/DataTypesDecimal.cpp b/src/DataTypes/DataTypesDecimal.cpp index 77a7a3e7237..d87eff97675 100644 --- a/src/DataTypes/DataTypesDecimal.cpp +++ b/src/DataTypes/DataTypesDecimal.cpp @@ -2,6 +2,7 @@ #include #include +#include #include #include #include @@ -269,9 +270,9 @@ ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & value, static constexpr bool throw_exception = std::is_same_v; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { - if (!std::isfinite(value)) + if (!isFinite(value)) { if constexpr (throw_exception) throw Exception(ErrorCodes::DECIMAL_OVERFLOW, "{} convert overflow. Cannot convert infinity or NaN to decimal", ToDataType::family_name); diff --git a/src/DataTypes/NumberTraits.h b/src/DataTypes/NumberTraits.h index ad1e9eaa67b..ee0d9812097 100644 --- a/src/DataTypes/NumberTraits.h +++ b/src/DataTypes/NumberTraits.h @@ -74,7 +74,7 @@ template struct ResultOfAdditionMultiplication { using Type = typename Construct< is_signed_v || is_signed_v, - is_floating_point_v || is_floating_point_v, + is_floating_point || is_floating_point, nextSize(max(sizeof(A), sizeof(B)))>::Type; }; @@ -82,7 +82,7 @@ template struct ResultOfSubtraction { using Type = typename Construct< true, - is_floating_point_v || is_floating_point_v, + is_floating_point || is_floating_point, nextSize(max(sizeof(A), sizeof(B)))>::Type; }; @@ -113,7 +113,7 @@ template struct ResultOfModulo /// Example: toInt32(-199) % toUInt8(200) will return -199 that does not fit in Int8, only in Int16. static constexpr size_t size_of_result = result_is_signed ? nextSize(sizeof(B)) : sizeof(B); using Type0 = typename Construct::Type; - using Type = std::conditional_t || is_floating_point_v, Float64, Type0>; + using Type = std::conditional_t || is_floating_point, Float64, Type0>; }; template struct ResultOfPositiveModulo @@ -121,21 +121,21 @@ template struct ResultOfPositiveModulo /// function positive_modulo always return non-negative number. static constexpr size_t size_of_result = sizeof(B); using Type0 = typename Construct::Type; - using Type = std::conditional_t || is_floating_point_v, Float64, Type0>; + using Type = std::conditional_t || is_floating_point, Float64, Type0>; }; template struct ResultOfModuloLegacy { using Type0 = typename Construct || is_signed_v, false, sizeof(B)>::Type; - using Type = std::conditional_t || is_floating_point_v, Float64, Type0>; + using Type = std::conditional_t || is_floating_point, Float64, Type0>; }; template struct ResultOfNegate { using Type = typename Construct< true, - is_floating_point_v, + is_floating_point, is_signed_v ? sizeof(A) : nextSize(sizeof(A))>::Type; }; @@ -143,7 +143,7 @@ template struct ResultOfAbs { using Type = typename Construct< false, - is_floating_point_v, + is_floating_point, sizeof(A)>::Type; }; @@ -154,7 +154,7 @@ template struct ResultOfBit using Type = typename Construct< is_signed_v || is_signed_v, false, - is_floating_point_v || is_floating_point_v ? 8 : max(sizeof(A), sizeof(B))>::Type; + is_floating_point || is_floating_point ? 8 : max(sizeof(A), sizeof(B))>::Type; }; template struct ResultOfBitNot @@ -180,7 +180,7 @@ template struct ResultOfBitNot template struct ResultOfIf { - static constexpr bool has_float = is_floating_point_v || is_floating_point_v; + static constexpr bool has_float = is_floating_point || is_floating_point; static constexpr bool has_integer = is_integer || is_integer; static constexpr bool has_signed = is_signed_v || is_signed_v; static constexpr bool has_unsigned = !is_signed_v || !is_signed_v; @@ -189,7 +189,7 @@ struct ResultOfIf static constexpr size_t max_size_of_unsigned_integer = max(is_signed_v ? 0 : sizeof(A), is_signed_v ? 0 : sizeof(B)); static constexpr size_t max_size_of_signed_integer = max(is_signed_v ? sizeof(A) : 0, is_signed_v ? sizeof(B) : 0); static constexpr size_t max_size_of_integer = max(is_integer ? sizeof(A) : 0, is_integer ? sizeof(B) : 0); - static constexpr size_t max_size_of_float = max(is_floating_point_v ? sizeof(A) : 0, is_floating_point_v ? sizeof(B) : 0); + static constexpr size_t max_size_of_float = max(is_floating_point ? sizeof(A) : 0, is_floating_point ? sizeof(B) : 0); using ConstructedType = typename Construct= max_size_of_float) @@ -211,7 +211,7 @@ template struct ToInteger using Type = typename Construct< is_signed_v, false, - is_floating_point_v ? 8 : sizeof(A)>::Type; + is_floating_point ? 8 : sizeof(A)>::Type; }; diff --git a/src/Formats/ProtobufSerializer.cpp b/src/Formats/ProtobufSerializer.cpp index 7f03bdeb45d..86b11f45b72 100644 --- a/src/Formats/ProtobufSerializer.cpp +++ b/src/Formats/ProtobufSerializer.cpp @@ -541,7 +541,7 @@ namespace case FieldTypeId::TYPE_ENUM: { - if (is_floating_point_v) + if (is_floating_point) incompatibleColumnType(TypeName); write_function = [this](NumberType value) diff --git a/src/Functions/DivisionUtils.h b/src/Functions/DivisionUtils.h index 1a241c7171a..e8f5da342f8 100644 --- a/src/Functions/DivisionUtils.h +++ b/src/Functions/DivisionUtils.h @@ -47,9 +47,9 @@ inline auto checkedDivision(A a, B b) { throwIfDivisionLeadsToFPE(a, b); - if constexpr (is_big_int_v && is_floating_point_v) + if constexpr (is_big_int_v && is_floating_point) return static_cast(a) / b; - else if constexpr (is_big_int_v && is_floating_point_v) + else if constexpr (is_big_int_v && is_floating_point) return a / static_cast(b); else if constexpr (is_big_int_v && is_big_int_v) return static_cast(a / b); @@ -86,17 +86,17 @@ struct DivideIntegralImpl { /// Comparisons are not strict to avoid rounding issues when operand is implicitly casted to float. - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) if (isNaN(a) || a >= std::numeric_limits::max() || a <= std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) if (isNaN(b) || b >= std::numeric_limits::max() || b <= std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); auto res = checkedDivision(CastA(a), CastB(b)); - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) if (isNaN(res) || res >= static_cast(std::numeric_limits::max()) || res <= std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division, because it will produce infinite or too large number"); @@ -122,18 +122,18 @@ struct ModuloImpl template static Result apply(A a, B b) { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { /// This computation is similar to `fmod` but the latter is not inlined and has 40 times worse performance. return static_cast(a) - trunc(static_cast(a) / static_cast(b)) * static_cast(b); } else { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) if (isNaN(a) || a > std::numeric_limits::max() || a < std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) if (isNaN(b) || b > std::numeric_limits::max() || b < std::numeric_limits::lowest()) throw Exception(ErrorCodes::ILLEGAL_DIVISION, "Cannot perform integer division on infinite or too large floating point numbers"); diff --git a/src/Functions/FunctionMathUnary.h b/src/Functions/FunctionMathUnary.h index 8395855a564..2cbd9b2e03c 100644 --- a/src/Functions/FunctionMathUnary.h +++ b/src/Functions/FunctionMathUnary.h @@ -66,7 +66,7 @@ private: /// Process all data as a whole and use FastOps implementation /// If the argument is integer, convert to Float64 beforehand - if constexpr (!is_floating_point_v) + if constexpr (!is_floating_point) { PODArray tmp_vec(size); for (size_t i = 0; i < size; ++i) @@ -150,7 +150,7 @@ private: { using Types = std::decay_t; using Type = typename Types::RightType; - using ReturnType = std::conditional_t, Float64, Type>; + using ReturnType = std::conditional_t, Float64, Type>; using ColVecType = ColumnVectorOrDecimal; const auto col_vec = checkAndGetColumn(col.column.get()); diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp index 44d0b750af9..8512ea5726f 100644 --- a/src/Functions/FunctionsConversion.cpp +++ b/src/Functions/FunctionsConversion.cpp @@ -638,7 +638,7 @@ inline void convertFromTime(DataTypeDateTime::FieldType & x, t template void parseImpl(typename DataType::FieldType & x, ReadBuffer & rb, const DateLUTImpl *, bool precise_float_parsing) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { if (precise_float_parsing) readFloatTextPrecise(x, rb); @@ -702,7 +702,7 @@ inline void parseImpl(DataTypeIPv6::FieldType & x, ReadBuffer & rb template bool tryParseImpl(typename DataType::FieldType & x, ReadBuffer & rb, const DateLUTImpl *, bool precise_float_parsing) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { if (precise_float_parsing) return tryReadFloatTextPrecise(x, rb); @@ -1767,7 +1767,7 @@ struct ConvertImpl else { /// If From Data is Nan or Inf and we convert to integer type, throw exception - if constexpr (std::is_floating_point_v && !std::is_floating_point_v) + if constexpr (is_floating_point && !is_floating_point) { if (!isFinite(vec_from[i])) { @@ -2253,9 +2253,9 @@ private: using RightT = typename RightDataType::FieldType; static constexpr bool bad_left = - is_decimal || std::is_floating_point_v || is_big_int_v || is_signed_v; + is_decimal || is_floating_point || is_big_int_v || is_signed_v; static constexpr bool bad_right = - is_decimal || std::is_floating_point_v || is_big_int_v || is_signed_v; + is_decimal || is_floating_point || is_big_int_v || is_signed_v; /// Disallow int vs UUID conversion (but support int vs UInt128 conversion) if constexpr ((bad_left && std::is_same_v) || @@ -2578,7 +2578,7 @@ struct ToNumberMonotonicity /// Float cases. /// When converting to Float, the conversion is always monotonic. - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) return { .is_monotonic = true, .is_always_monotonic = true }; const auto * low_cardinality = typeid_cast(&type); diff --git a/src/Functions/FunctionsJSON.h b/src/Functions/FunctionsJSON.h index 8a2ad457d34..65c1a6fb2d2 100644 --- a/src/Functions/FunctionsJSON.h +++ b/src/Functions/FunctionsJSON.h @@ -741,7 +741,7 @@ public: switch (element.type()) { case ElementType::DOUBLE: - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { /// We permit inaccurate conversion of double to float. /// Example: double 0.1 from JSON is not representable in float. @@ -769,7 +769,7 @@ public: case ElementType::STRING: { auto rb = ReadBufferFromMemory{element.getString()}; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { if (!tryReadFloatText(value, rb) || !rb.eof()) return false; diff --git a/src/Functions/FunctionsRound.h b/src/Functions/FunctionsRound.h index ab62deed45d..46fbe70458d 100644 --- a/src/Functions/FunctionsRound.h +++ b/src/Functions/FunctionsRound.h @@ -453,7 +453,7 @@ template - using FunctionRoundingImpl = std::conditional_t, + using FunctionRoundingImpl = std::conditional_t, FloatRoundingImpl, IntegerRoundingImpl>; diff --git a/src/Functions/FunctionsVisitParam.h b/src/Functions/FunctionsVisitParam.h index 5e13fbbad5c..fd59ea3a9c1 100644 --- a/src/Functions/FunctionsVisitParam.h +++ b/src/Functions/FunctionsVisitParam.h @@ -57,7 +57,7 @@ struct ExtractNumericType ResultType x = 0; if (!in.eof()) { - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) tryReadFloatText(x, in); else tryReadIntText(x, in); diff --git a/src/Functions/abs.cpp b/src/Functions/abs.cpp index 9ac2363f765..3a618686b30 100644 --- a/src/Functions/abs.cpp +++ b/src/Functions/abs.cpp @@ -22,7 +22,7 @@ struct AbsImpl return a < 0 ? static_cast(~a) + 1 : static_cast(a); else if constexpr (is_integer && is_unsigned_v) return static_cast(a); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) return static_cast(std::abs(a)); } diff --git a/src/Functions/array/arrayAggregation.cpp b/src/Functions/array/arrayAggregation.cpp index 03aa5fb9086..9c17e1095c5 100644 --- a/src/Functions/array/arrayAggregation.cpp +++ b/src/Functions/array/arrayAggregation.cpp @@ -85,7 +85,7 @@ struct ArrayAggregateResultImpl std::conditional_t, Decimal128, std::conditional_t, Decimal256, std::conditional_t, Decimal128, - std::conditional_t, Float64, + std::conditional_t, Float64, std::conditional_t, Int64, UInt64>>>>>>>>>>>; }; diff --git a/src/Functions/factorial.cpp b/src/Functions/factorial.cpp index 3b46d9e867f..32bdc84b954 100644 --- a/src/Functions/factorial.cpp +++ b/src/Functions/factorial.cpp @@ -21,7 +21,7 @@ struct FactorialImpl static NO_SANITIZE_UNDEFINED ResultType apply(A a) { - if constexpr (is_floating_point_v || is_over_big_int) + if constexpr (is_floating_point || is_over_big_int) throw Exception( ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Illegal type of argument of function factorial, should not be floating point or big int"); diff --git a/src/Functions/if.cpp b/src/Functions/if.cpp index 7a6d37d810d..dded3d46652 100644 --- a/src/Functions/if.cpp +++ b/src/Functions/if.cpp @@ -87,7 +87,7 @@ inline void fillVectorVector(const ArrayCond & cond, const ArrayA & a, const Arr { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a[a_index]) + (!cond[i]) * static_cast(b[b_index]); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a[a_index], b[b_index], res[i]) } @@ -105,7 +105,7 @@ inline void fillVectorVector(const ArrayCond & cond, const ArrayA & a, const Arr { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a[a_index]) + (!cond[i]) * static_cast(b[i]); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a[a_index], b[i], res[i]) } @@ -122,7 +122,7 @@ inline void fillVectorVector(const ArrayCond & cond, const ArrayA & a, const Arr { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a[i]) + (!cond[i]) * static_cast(b[b_index]); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a[i], b[b_index], res[i]) } @@ -138,7 +138,7 @@ inline void fillVectorVector(const ArrayCond & cond, const ArrayA & a, const Arr { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a[i]) + (!cond[i]) * static_cast(b[i]); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a[i], b[i], res[i]) } @@ -162,7 +162,7 @@ inline void fillVectorConstant(const ArrayCond & cond, const ArrayA & a, B b, Ar { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a[a_index]) + (!cond[i]) * static_cast(b); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a[a_index], b, res[i]) } @@ -178,7 +178,7 @@ inline void fillVectorConstant(const ArrayCond & cond, const ArrayA & a, B b, Ar { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a[i]) + (!cond[i]) * static_cast(b); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a[i], b, res[i]) } @@ -200,7 +200,7 @@ inline void fillConstantVector(const ArrayCond & cond, A a, const ArrayB & b, Ar { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a) + (!cond[i]) * static_cast(b[b_index]); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a, b[b_index], res[i]) } @@ -216,7 +216,7 @@ inline void fillConstantVector(const ArrayCond & cond, A a, const ArrayB & b, Ar { if constexpr (is_native_int_or_decimal_v) res[i] = !!cond[i] * static_cast(a) + (!cond[i]) * static_cast(b[i]); - else if constexpr (std::is_floating_point_v) + else if constexpr (is_floating_point) { BRANCHFREE_IF_FLOAT(ResultType, cond[i], a, b[i], res[i]) } diff --git a/src/Functions/minus.cpp b/src/Functions/minus.cpp index 4d86442ad7e..cf318db805b 100644 --- a/src/Functions/minus.cpp +++ b/src/Functions/minus.cpp @@ -17,8 +17,8 @@ struct MinusImpl { if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) - static_cast(static_cast(b)); } diff --git a/src/Functions/moduloOrZero.cpp b/src/Functions/moduloOrZero.cpp index d233e4e4ce2..5a4d1539345 100644 --- a/src/Functions/moduloOrZero.cpp +++ b/src/Functions/moduloOrZero.cpp @@ -17,7 +17,7 @@ struct ModuloOrZeroImpl template static Result apply(A a, B b) { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { /// This computation is similar to `fmod` but the latter is not inlined and has 40 times worse performance. return ResultType(a) - trunc(ResultType(a) / ResultType(b)) * ResultType(b); diff --git a/src/Functions/multiply.cpp b/src/Functions/multiply.cpp index 559143a43b4..740ab81d0d9 100644 --- a/src/Functions/multiply.cpp +++ b/src/Functions/multiply.cpp @@ -18,8 +18,8 @@ struct MultiplyImpl { if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) * static_cast(static_cast(b)); } diff --git a/src/Functions/plus.cpp b/src/Functions/plus.cpp index 00136e50c5b..26921713f78 100644 --- a/src/Functions/plus.cpp +++ b/src/Functions/plus.cpp @@ -19,8 +19,8 @@ struct PlusImpl /// Next everywhere, static_cast - so that there is no wrong result in expressions of the form Int64 c = UInt32(a) * Int32(-1). if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) + static_cast(static_cast(b)); } diff --git a/src/Functions/sign.cpp b/src/Functions/sign.cpp index 16f0efd2201..a6396a58c0c 100644 --- a/src/Functions/sign.cpp +++ b/src/Functions/sign.cpp @@ -13,7 +13,7 @@ struct SignImpl static NO_SANITIZE_UNDEFINED ResultType apply(A a) { - if constexpr (is_decimal || is_floating_point_v) + if constexpr (is_decimal || is_floating_point) return a < A(0) ? -1 : a == A(0) ? 0 : 1; else if constexpr (is_signed_v) return a < 0 ? -1 : a == 0 ? 0 : 1; diff --git a/src/IO/ReadHelpers.h b/src/IO/ReadHelpers.h index 6dda5a9b089..f1fcbb07af5 100644 --- a/src/IO/ReadHelpers.h +++ b/src/IO/ReadHelpers.h @@ -1382,7 +1382,7 @@ inline bool tryReadText(IPv4 & x, ReadBuffer & buf) { return tryReadIPv4Text(x, inline bool tryReadText(IPv6 & x, ReadBuffer & buf) { return tryReadIPv6Text(x, buf); } template -requires is_floating_point_v +requires is_floating_point inline void readText(T & x, ReadBuffer & buf) { readFloatText(x, buf); } inline void readText(String & x, ReadBuffer & buf) { readEscapedString(x, buf); } diff --git a/src/IO/WriteHelpers.h b/src/IO/WriteHelpers.h index cdeabfcf352..a4eefeaffe2 100644 --- a/src/IO/WriteHelpers.h +++ b/src/IO/WriteHelpers.h @@ -150,7 +150,7 @@ inline void writeBoolText(bool x, WriteBuffer & buf) template -requires is_floating_point_v +requires is_floating_point inline size_t writeFloatTextFastPath(T x, char * buffer) { Int64 result = 0; @@ -182,7 +182,7 @@ inline size_t writeFloatTextFastPath(T x, char * buffer) } template -requires is_floating_point_v +requires is_floating_point inline void writeFloatText(T x, WriteBuffer & buf) { using Converter = DoubleConverter; @@ -530,7 +530,7 @@ void writeJSONNumber(T x, WriteBuffer & ostr, const FormatSettings & settings) bool is_finite = isFinite(x); const bool need_quote = (is_integer && (sizeof(T) >= 8) && settings.json.quote_64bit_integers) - || (settings.json.quote_denormals && !is_finite) || (is_floating_point_v && (sizeof(T) >= 8) && settings.json.quote_64bit_floats); + || (settings.json.quote_denormals && !is_finite) || (is_floating_point && (sizeof(T) >= 8) && settings.json.quote_64bit_floats); if (need_quote) writeChar('"', ostr); @@ -541,7 +541,7 @@ void writeJSONNumber(T x, WriteBuffer & ostr, const FormatSettings & settings) writeCString("null", ostr); else { - if constexpr (is_floating_point_v) + if constexpr (is_floating_point) { if (std::signbit(x)) { @@ -1065,7 +1065,7 @@ inline void writeText(is_integer auto x, WriteBuffer & buf) } template -requires is_floating_point_v +requires is_floating_point inline void writeText(T x, WriteBuffer & buf) { writeFloatText(x, buf); } inline void writeText(is_enum auto x, WriteBuffer & buf) { writeText(magic_enum::enum_name(x), buf); } diff --git a/src/Interpreters/RowRefs.cpp b/src/Interpreters/RowRefs.cpp index 9785ba46dab..c5ffbb96d6f 100644 --- a/src/Interpreters/RowRefs.cpp +++ b/src/Interpreters/RowRefs.cpp @@ -183,7 +183,7 @@ private: if (sorted.load(std::memory_order_relaxed)) return; - if constexpr (std::is_arithmetic_v && !std::is_floating_point_v) + if constexpr (std::is_arithmetic_v && !std::is_floating_point) { if (likely(entries.size() > 256)) { From bf2a8f6a7f6eb8073b60468058f8259cf4a4f341 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 2 Jun 2024 20:43:02 +0200 Subject: [PATCH 004/244] Preparation --- base/base/BFloat16.h | 17 +- src/AggregateFunctions/AggregateFunctionSum.h | 13 +- .../AggregateFunctionUniq.h | 2 +- src/Core/DecimalFunctions.h | 10 +- src/Core/iostream_debug_helpers.cpp | 149 ------------------ src/Core/iostream_debug_helpers.h | 49 ------ src/DataTypes/DataTypesDecimal.cpp | 15 +- src/Dictionaries/RangeHashedDictionary.h | 3 +- src/Functions/FunctionsRound.h | 3 +- src/Functions/array/mapPopulateSeries.cpp | 32 ++-- src/Functions/exp.cpp | 9 +- src/Functions/log.cpp | 9 +- src/Functions/minus.cpp | 4 +- src/Functions/sigmoid.cpp | 10 +- src/Functions/tanh.cpp | 9 +- src/IO/WriteHelpers.h | 14 +- src/Interpreters/RowRefs.cpp | 2 +- src/Parsers/iostream_debug_helpers.cpp | 35 ---- src/Parsers/iostream_debug_helpers.h | 17 -- 19 files changed, 110 insertions(+), 292 deletions(-) delete mode 100644 src/Core/iostream_debug_helpers.cpp delete mode 100644 src/Core/iostream_debug_helpers.h delete mode 100644 src/Parsers/iostream_debug_helpers.cpp delete mode 100644 src/Parsers/iostream_debug_helpers.h diff --git a/base/base/BFloat16.h b/base/base/BFloat16.h index 17c3ebe9ef3..99eab5c67cb 100644 --- a/base/base/BFloat16.h +++ b/base/base/BFloat16.h @@ -1,9 +1,22 @@ #pragma once +#include + + using BFloat16 = __bf16; namespace std { - inline constexpr bool isfinite(BFloat16) { return true; } - inline constexpr bool signbit(BFloat16) { return false; } + inline constexpr bool isfinite(BFloat16 x) { return (bit_cast(x) & 0b0111111110000000) != 0b0111111110000000; } + inline constexpr bool signbit(BFloat16 x) { return bit_cast(x) & 0b1000000000000000; } +} + +inline Float32 BFloat16ToFloat32(BFloat16 x) +{ + return bit_cast(static_cast(bit_cast(x)) << 16); +} + +inline BFloat16 Float32ToBFloat16(Float32 x) +{ + return bit_cast(std::bit_cast(x) >> 16); } diff --git a/src/AggregateFunctions/AggregateFunctionSum.h b/src/AggregateFunctions/AggregateFunctionSum.h index d0d600be70b..f6c51241a5c 100644 --- a/src/AggregateFunctions/AggregateFunctionSum.h +++ b/src/AggregateFunctions/AggregateFunctionSum.h @@ -193,12 +193,11 @@ struct AggregateFunctionSumData Impl::add(sum, local_sum); return; } - else if constexpr (is_floating_point) + else if constexpr (is_floating_point && (sizeof(Value) == 4 || sizeof(Value) == 8)) { - /// For floating point we use a similar trick as above, except that now we reinterpret the floating point number as an unsigned + /// For floating point we use a similar trick as above, except that now we reinterpret the floating point number as an unsigned /// integer of the same size and use a mask instead (0 to discard, 0xFF..FF to keep) - static_assert(sizeof(Value) == 4 || sizeof(Value) == 8); - using equivalent_integer = typename std::conditional_t; + using EquivalentInteger = typename std::conditional_t; constexpr size_t unroll_count = 128 / sizeof(T); T partial_sums[unroll_count]{}; @@ -209,11 +208,11 @@ struct AggregateFunctionSumData { for (size_t i = 0; i < unroll_count; ++i) { - equivalent_integer value; - std::memcpy(&value, &ptr[i], sizeof(Value)); + EquivalentInteger value; + memcpy(&value, &ptr[i], sizeof(Value)); value &= (!condition_map[i] != add_if_zero) - 1; Value d; - std::memcpy(&d, &value, sizeof(Value)); + memcpy(&d, &value, sizeof(Value)); Impl::add(partial_sums[i], d); } ptr += unroll_count; diff --git a/src/AggregateFunctions/AggregateFunctionUniq.h b/src/AggregateFunctions/AggregateFunctionUniq.h index cef23f766c7..cd2d3c1eb18 100644 --- a/src/AggregateFunctions/AggregateFunctionUniq.h +++ b/src/AggregateFunctions/AggregateFunctionUniq.h @@ -257,7 +257,7 @@ template struct AggregateFunctionUniqTraits { static UInt64 hash(T x) { - if constexpr (std::is_same_v || std::is_same_v) + if constexpr (is_floating_point) { return bit_cast(x); } diff --git a/src/Core/DecimalFunctions.h b/src/Core/DecimalFunctions.h index 435cef61145..abd660a8a7f 100644 --- a/src/Core/DecimalFunctions.h +++ b/src/Core/DecimalFunctions.h @@ -17,6 +17,7 @@ class DataTypeNumber; namespace ErrorCodes { + extern const int NOT_IMPLEMENTED; extern const int DECIMAL_OVERFLOW; extern const int ARGUMENT_OUT_OF_BOUND; } @@ -310,7 +311,14 @@ ReturnType convertToImpl(const DecimalType & decimal, UInt32 scale, To & result) using DecimalNativeType = typename DecimalType::NativeType; static constexpr bool throw_exception = std::is_void_v; - if constexpr (is_floating_point) + if constexpr (std::is_same_v) + { + if constexpr (throw_exception) + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Conversion from Decimal to BFloat16 is not implemented"); + else + return ReturnType(false); + } + else if constexpr (is_floating_point) { result = static_cast(decimal.value) / static_cast(scaleMultiplier(scale)); } diff --git a/src/Core/iostream_debug_helpers.cpp b/src/Core/iostream_debug_helpers.cpp deleted file mode 100644 index 38e61ac4fca..00000000000 --- a/src/Core/iostream_debug_helpers.cpp +++ /dev/null @@ -1,149 +0,0 @@ -#include "iostream_debug_helpers.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - - -namespace DB -{ - -template <> -std::ostream & operator<< (std::ostream & stream, const Field & what) -{ - stream << applyVisitor(FieldVisitorDump(), what); - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const NameAndTypePair & what) -{ - stream << "NameAndTypePair(name = " << what.name << ", type = " << what.type << ")"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const IDataType & what) -{ - stream << "IDataType(name = " << what.getName() << ", default = " << what.getDefault() << ")"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const IStorage & what) -{ - auto table_id = what.getStorageID(); - stream << "IStorage(name = " << what.getName() << ", tableName = " << table_id.table_name << ") {" - << what.getInMemoryMetadataPtr()->getColumns().getAllPhysical().toString() << "}"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const TableLockHolder &) -{ - stream << "TableStructureReadLock()"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const IFunctionOverloadResolver & what) -{ - stream << "IFunction(name = " << what.getName() << ", variadic = " << what.isVariadic() << ", args = " << what.getNumberOfArguments() - << ")"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const Block & what) -{ - stream << "Block(" - << "num_columns = " << what.columns() << "){" << what.dumpStructure() << "}"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const ColumnWithTypeAndName & what) -{ - stream << "ColumnWithTypeAndName(name = " << what.name << ", type = " << *what.type << ", column = "; - return dumpValue(stream, what.column) << ")"; -} - -std::ostream & operator<<(std::ostream & stream, const IColumn & what) -{ - stream << "IColumn(" << what.dumpStructure() << ")"; - stream << "{"; - for (size_t i = 0; i < what.size(); ++i) - { - if (i) - stream << ", "; - stream << applyVisitor(FieldVisitorDump(), what[i]); - } - stream << "}"; - - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const Packet & what) -{ - stream << "Packet(" - << "type = " << what.type; - // types description: Core/Protocol.h - if (what.exception) - stream << "exception = " << what.exception.get(); - // TODO: profile_info - stream << ") {" << what.block << "}"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const ExpressionActions & what) -{ - stream << "ExpressionActions(" << what.dumpActions() << ")"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const TreeRewriterResult & what) -{ - stream << "SyntaxAnalyzerResult{"; - stream << "storage=" << what.storage << "; "; - if (!what.source_columns.empty()) - { - stream << "source_columns="; - dumpValue(stream, what.source_columns); - stream << "; "; - } - if (!what.aliases.empty()) - { - stream << "aliases="; - dumpValue(stream, what.aliases); - stream << "; "; - } - if (!what.array_join_result_to_source.empty()) - { - stream << "array_join_result_to_source="; - dumpValue(stream, what.array_join_result_to_source); - stream << "; "; - } - if (!what.array_join_alias_to_name.empty()) - { - stream << "array_join_alias_to_name="; - dumpValue(stream, what.array_join_alias_to_name); - stream << "; "; - } - if (!what.array_join_name_to_alias.empty()) - { - stream << "array_join_name_to_alias="; - dumpValue(stream, what.array_join_name_to_alias); - stream << "; "; - } - stream << "rewrite_subqueries=" << what.rewrite_subqueries << "; "; - stream << "}"; - - return stream; -} - -} diff --git a/src/Core/iostream_debug_helpers.h b/src/Core/iostream_debug_helpers.h deleted file mode 100644 index e40bf74583e..00000000000 --- a/src/Core/iostream_debug_helpers.h +++ /dev/null @@ -1,49 +0,0 @@ -#pragma once -#include - -namespace DB -{ - -// Use template to disable implicit casting for certain overloaded types such as Field, which leads -// to overload resolution ambiguity. -class Field; -template -requires std::is_same_v -std::ostream & operator<<(std::ostream & stream, const T & what); - -struct NameAndTypePair; -std::ostream & operator<<(std::ostream & stream, const NameAndTypePair & what); - -class IDataType; -std::ostream & operator<<(std::ostream & stream, const IDataType & what); - -class IStorage; -std::ostream & operator<<(std::ostream & stream, const IStorage & what); - -class IFunctionOverloadResolver; -std::ostream & operator<<(std::ostream & stream, const IFunctionOverloadResolver & what); - -class IFunctionBase; -std::ostream & operator<<(std::ostream & stream, const IFunctionBase & what); - -class Block; -std::ostream & operator<<(std::ostream & stream, const Block & what); - -struct ColumnWithTypeAndName; -std::ostream & operator<<(std::ostream & stream, const ColumnWithTypeAndName & what); - -class IColumn; -std::ostream & operator<<(std::ostream & stream, const IColumn & what); - -struct Packet; -std::ostream & operator<<(std::ostream & stream, const Packet & what); - -class ExpressionActions; -std::ostream & operator<<(std::ostream & stream, const ExpressionActions & what); - -struct TreeRewriterResult; -std::ostream & operator<<(std::ostream & stream, const TreeRewriterResult & what); -} - -/// some operator<< should be declared before operator<<(... std::shared_ptr<>) -#include diff --git a/src/DataTypes/DataTypesDecimal.cpp b/src/DataTypes/DataTypesDecimal.cpp index d87eff97675..e0304e46b05 100644 --- a/src/DataTypes/DataTypesDecimal.cpp +++ b/src/DataTypes/DataTypesDecimal.cpp @@ -20,6 +20,7 @@ namespace ErrorCodes extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH; extern const int ILLEGAL_TYPE_OF_ARGUMENT; extern const int DECIMAL_OVERFLOW; + extern const int NOT_IMPLEMENTED; } @@ -262,15 +263,19 @@ FOR_EACH_ARITHMETIC_TYPE(INVOKE); template requires (is_arithmetic_v && IsDataTypeDecimal) -ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType & result) +ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & /*value*/, UInt32 /*scale*/, typename ToDataType::FieldType & /*result*/) { - using FromFieldType = typename FromDataType::FieldType; +/* using FromFieldType = typename FromDataType::FieldType; using ToFieldType = typename ToDataType::FieldType; using ToNativeType = typename ToFieldType::NativeType; static constexpr bool throw_exception = std::is_same_v; - if constexpr (is_floating_point) + if constexpr (std::is_same_v) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Conversion from BFloat16 to Decimal is not implemented"); + } + else if constexpr (is_floating_point) { if (!isFinite(value)) { @@ -302,7 +307,9 @@ ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & value, return ReturnType(convertDecimalsImpl, ToDataType, ReturnType>(static_cast(value), 0, scale, result)); else return ReturnType(convertDecimalsImpl, ToDataType, ReturnType>(static_cast(value), 0, scale, result)); - } + }*/ + + return ReturnType(); } #define DISPATCH(FROM_DATA_TYPE, TO_DATA_TYPE) \ diff --git a/src/Dictionaries/RangeHashedDictionary.h b/src/Dictionaries/RangeHashedDictionary.h index bf004dbe32b..4950e7c8ee6 100644 --- a/src/Dictionaries/RangeHashedDictionary.h +++ b/src/Dictionaries/RangeHashedDictionary.h @@ -298,7 +298,8 @@ namespace impl using Types = std::decay_t; using DataType = typename Types::LeftType; - if constexpr (IsDataTypeDecimalOrNumber || IsDataTypeDateOrDateTime || IsDataTypeEnum) + if constexpr ((IsDataTypeDecimalOrNumber || IsDataTypeDateOrDateTime || IsDataTypeEnum) + && !std::is_same_v) { using ColumnType = typename DataType::ColumnType; func(TypePair()); diff --git a/src/Functions/FunctionsRound.h b/src/Functions/FunctionsRound.h index 46fbe70458d..7eea0d74975 100644 --- a/src/Functions/FunctionsRound.h +++ b/src/Functions/FunctionsRound.h @@ -579,7 +579,8 @@ public: using Types = std::decay_t; using DataType = typename Types::LeftType; - if constexpr (IsDataTypeNumber || IsDataTypeDecimal) + if constexpr ((IsDataTypeNumber || IsDataTypeDecimal) + && !std::is_same_v) { using FieldType = typename DataType::FieldType; res = Dispatcher::apply(column.column.get(), scale_arg); diff --git a/src/Functions/array/mapPopulateSeries.cpp b/src/Functions/array/mapPopulateSeries.cpp index 0db71ab2cf8..759696147c3 100644 --- a/src/Functions/array/mapPopulateSeries.cpp +++ b/src/Functions/array/mapPopulateSeries.cpp @@ -453,23 +453,29 @@ private: using ValueType = typename Types::RightType; static constexpr bool key_and_value_are_numbers = IsDataTypeNumber && IsDataTypeNumber; - static constexpr bool key_is_float = std::is_same_v || std::is_same_v; - if constexpr (key_and_value_are_numbers && !key_is_float) + if constexpr (key_and_value_are_numbers) { - using KeyFieldType = typename KeyType::FieldType; - using ValueFieldType = typename ValueType::FieldType; + if constexpr (is_floating_point) + { + return false; + } + else + { + using KeyFieldType = typename KeyType::FieldType; + using ValueFieldType = typename ValueType::FieldType; - executeImplTyped( - input.key_column, - input.value_column, - input.offsets_column, - input.max_key_column, - std::move(result_columns.result_key_column), - std::move(result_columns.result_value_column), - std::move(result_columns.result_offset_column)); + executeImplTyped( + input.key_column, + input.value_column, + input.offsets_column, + input.max_key_column, + std::move(result_columns.result_key_column), + std::move(result_columns.result_value_column), + std::move(result_columns.result_offset_column)); - return true; + return true; + } } return false; diff --git a/src/Functions/exp.cpp b/src/Functions/exp.cpp index d352cda7460..9b8207afe30 100644 --- a/src/Functions/exp.cpp +++ b/src/Functions/exp.cpp @@ -21,7 +21,14 @@ namespace template static void execute(const T * src, size_t size, T * dst) { - NFastOps::Exp(src, size, dst); + if constexpr (std::is_same_v) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Function `{}` is not implemented for BFloat16", name); + } + else + { + NFastOps::Exp(src, size, dst); + } } }; } diff --git a/src/Functions/log.cpp b/src/Functions/log.cpp index 9096b8c6f22..d5e10c90c83 100644 --- a/src/Functions/log.cpp +++ b/src/Functions/log.cpp @@ -20,7 +20,14 @@ struct LogName { static constexpr auto name = "log"; }; template static void execute(const T * src, size_t size, T * dst) { - NFastOps::Log(src, size, dst); + if constexpr (std::is_same_v) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Function `{}` is not implemented for BFloat16", name); + } + else + { + NFastOps::Log(src, size, dst); + } } }; diff --git a/src/Functions/minus.cpp b/src/Functions/minus.cpp index cf318db805b..a372e8d5d78 100644 --- a/src/Functions/minus.cpp +++ b/src/Functions/minus.cpp @@ -17,8 +17,8 @@ struct MinusImpl { if constexpr (is_big_int_v || is_big_int_v) { - using CastA = std::conditional_t, B, A>; - using CastB = std::conditional_t, A, B>; + using CastA = std::conditional_t, B, A>; + using CastB = std::conditional_t, A, B>; return static_cast(static_cast(a)) - static_cast(static_cast(b)); } diff --git a/src/Functions/sigmoid.cpp b/src/Functions/sigmoid.cpp index d121bdc7389..1179329845d 100644 --- a/src/Functions/sigmoid.cpp +++ b/src/Functions/sigmoid.cpp @@ -21,7 +21,14 @@ namespace template static void execute(const T * src, size_t size, T * dst) { - NFastOps::Sigmoid<>(src, size, dst); + if constexpr (std::is_same_v) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Function `{}` is not implemented for BFloat16", name); + } + else + { + NFastOps::Sigmoid<>(src, size, dst); + } } }; } @@ -47,4 +54,3 @@ REGISTER_FUNCTION(Sigmoid) } } - diff --git a/src/Functions/tanh.cpp b/src/Functions/tanh.cpp index bdefa5263d7..49788b31970 100644 --- a/src/Functions/tanh.cpp +++ b/src/Functions/tanh.cpp @@ -19,7 +19,14 @@ struct TanhName { static constexpr auto name = "tanh"; }; template static void execute(const T * src, size_t size, T * dst) { - NFastOps::Tanh<>(src, size, dst); + if constexpr (std::is_same_v) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Function `{}` is not implemented for BFloat16", name); + } + else + { + NFastOps::Tanh<>(src, size, dst); + } } }; diff --git a/src/IO/WriteHelpers.h b/src/IO/WriteHelpers.h index a4eefeaffe2..d2e2868b245 100644 --- a/src/IO/WriteHelpers.h +++ b/src/IO/WriteHelpers.h @@ -155,7 +155,7 @@ inline size_t writeFloatTextFastPath(T x, char * buffer) { Int64 result = 0; - if constexpr (std::is_same_v) + if constexpr (std::is_same_v) { /// The library Ryu has low performance on integers. /// This workaround improves performance 6..10 times. @@ -165,10 +165,16 @@ inline size_t writeFloatTextFastPath(T x, char * buffer) else result = jkj::dragonbox::to_chars_n(x, buffer) - buffer; } - else + else if constexpr (std::is_same_v) { - /// This will support 16-bit floats as well. - float f32 = x; + if (DecomposedFloat32(x).isIntegerInRepresentableRange()) + result = itoa(Int32(x), buffer) - buffer; + else + result = jkj::dragonbox::to_chars_n(x, buffer) - buffer; + } + else if constexpr (std::is_same_v) + { + Float32 f32 = BFloat16ToFloat32(x); if (DecomposedFloat32(f32).isIntegerInRepresentableRange()) result = itoa(Int32(f32), buffer) - buffer; diff --git a/src/Interpreters/RowRefs.cpp b/src/Interpreters/RowRefs.cpp index c5ffbb96d6f..a0fad8840e6 100644 --- a/src/Interpreters/RowRefs.cpp +++ b/src/Interpreters/RowRefs.cpp @@ -183,7 +183,7 @@ private: if (sorted.load(std::memory_order_relaxed)) return; - if constexpr (std::is_arithmetic_v && !std::is_floating_point) + if constexpr (std::is_arithmetic_v && !is_floating_point) { if (likely(entries.size() > 256)) { diff --git a/src/Parsers/iostream_debug_helpers.cpp b/src/Parsers/iostream_debug_helpers.cpp deleted file mode 100644 index b74d337b22d..00000000000 --- a/src/Parsers/iostream_debug_helpers.cpp +++ /dev/null @@ -1,35 +0,0 @@ -#include "iostream_debug_helpers.h" -#include -#include -#include -#include -#include -#include - -namespace DB -{ - -std::ostream & operator<<(std::ostream & stream, const Token & what) -{ - stream << "Token (type="<< static_cast(what.type) <<"){"<< std::string{what.begin, what.end} << "}"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const Expected & what) -{ - stream << "Expected {variants="; - dumpValue(stream, what.variants) - << "; max_parsed_pos=" << what.max_parsed_pos << "}"; - return stream; -} - -std::ostream & operator<<(std::ostream & stream, const IAST & what) -{ - WriteBufferFromOStream buf(stream, 4096); - buf << "IAST{"; - what.dumpTree(buf); - buf << "}"; - return stream; -} - -} diff --git a/src/Parsers/iostream_debug_helpers.h b/src/Parsers/iostream_debug_helpers.h deleted file mode 100644 index 39f52ebcbc2..00000000000 --- a/src/Parsers/iostream_debug_helpers.h +++ /dev/null @@ -1,17 +0,0 @@ -#pragma once -#include - -namespace DB -{ -struct Token; -std::ostream & operator<<(std::ostream & stream, const Token & what); - -struct Expected; -std::ostream & operator<<(std::ostream & stream, const Expected & what); - -class IAST; -std::ostream & operator<<(std::ostream & stream, const IAST & what); - -} - -#include From 83b79fce832ef4bd3fa230ae8a66d23935724985 Mon Sep 17 00:00:00 2001 From: qhsong Date: Wed, 8 May 2024 16:18:13 +0800 Subject: [PATCH 005/244] Add repeatable uniq ID for processor and step --- src/Common/ThreadStatus.cpp | 10 +++++++ src/Common/ThreadStatus.h | 9 +++++++ src/Interpreters/Context.h | 2 ++ src/Interpreters/ProcessorsProfileLog.cpp | 4 +++ src/Interpreters/ProcessorsProfileLog.h | 4 ++- src/Interpreters/executeQuery.cpp | 2 ++ .../Executors/ExecutionThreadContext.cpp | 2 +- src/Processors/IProcessor.h | 26 ++++++++++++++++--- src/Processors/QueryPlan/IQueryPlanStep.h | 12 ++++++++- src/Processors/QueryPlan/QueryPlan.cpp | 1 + src/QueryPipeline/QueryPipelineBuilder.cpp | 4 +-- src/QueryPipeline/printPipeline.cpp | 4 +-- src/QueryPipeline/printPipeline.h | 2 +- .../01786_explain_merge_tree.reference | 2 ++ 14 files changed, 73 insertions(+), 11 deletions(-) diff --git a/src/Common/ThreadStatus.cpp b/src/Common/ThreadStatus.cpp index ad96018a17e..74b5475da77 100644 --- a/src/Common/ThreadStatus.cpp +++ b/src/Common/ThreadStatus.cpp @@ -221,6 +221,16 @@ bool ThreadStatus::isQueryCanceled() const return false; } +size_t ThreadStatus::incrStepIndex() +{ + return ++(*local_data.step_count); +} + +size_t ThreadStatus::incrProcessorIndex() +{ + return ++(*local_data.processor_count); +} + ThreadStatus::~ThreadStatus() { flushUntrackedMemory(); diff --git a/src/Common/ThreadStatus.h b/src/Common/ThreadStatus.h index 0c02ab8fdb0..97b45c01e54 100644 --- a/src/Common/ThreadStatus.h +++ b/src/Common/ThreadStatus.h @@ -10,6 +10,7 @@ #include +#include #include #include #include @@ -90,6 +91,11 @@ public: String query_for_logs; UInt64 normalized_query_hash = 0; + //QueryPlan can not build parallel, but processor may build parallel in expand() function. + //so we use atomic_size_t for processor_count + std::shared_ptr step_count = std::make_shared(0); + std::shared_ptr processor_count = std::make_shared(0); + QueryIsCanceledPredicate query_is_canceled_predicate = {}; }; @@ -309,6 +315,9 @@ public: void initGlobalProfiler(UInt64 global_profiler_real_time_period, UInt64 global_profiler_cpu_time_period); + size_t incrStepIndex(); + size_t incrProcessorIndex(); + private: void applyGlobalSettings(); void applyQuerySettings(); diff --git a/src/Interpreters/Context.h b/src/Interpreters/Context.h index d1ff5b4c2b2..692d71a3384 100644 --- a/src/Interpreters/Context.h +++ b/src/Interpreters/Context.h @@ -1336,7 +1336,9 @@ private: std::shared_ptr getClustersImpl(std::lock_guard & lock) const; /// Throttling + public: + ThrottlerPtr getReplicatedFetchesThrottler() const; ThrottlerPtr getReplicatedSendsThrottler() const; diff --git a/src/Interpreters/ProcessorsProfileLog.cpp b/src/Interpreters/ProcessorsProfileLog.cpp index 7dec2a3163a..8e1cf278c63 100644 --- a/src/Interpreters/ProcessorsProfileLog.cpp +++ b/src/Interpreters/ProcessorsProfileLog.cpp @@ -42,6 +42,8 @@ ColumnsDescription ProcessorProfileLogElement::getColumnsDescription() {"input_bytes", std::make_shared(), "The number of bytes consumed by processor."}, {"output_rows", std::make_shared(), "The number of rows generated by processor."}, {"output_bytes", std::make_shared(), "The number of bytes generated by processor."}, + {"processor_uniq_id", std::make_shared(), "The uniq processor id in pipeline."}, + {"step_uniq_id", std::make_shared(), "The uniq step id in plan."}, }; } @@ -75,6 +77,8 @@ void ProcessorProfileLogElement::appendToBlock(MutableColumns & columns) const columns[i++]->insert(input_bytes); columns[i++]->insert(output_rows); columns[i++]->insert(output_bytes); + columns[i++]->insert(processor_uniq_id); + columns[i++]->insert(step_uniq_id); } diff --git a/src/Interpreters/ProcessorsProfileLog.h b/src/Interpreters/ProcessorsProfileLog.h index 8319d373f39..1b2abaa8ede 100644 --- a/src/Interpreters/ProcessorsProfileLog.h +++ b/src/Interpreters/ProcessorsProfileLog.h @@ -17,12 +17,14 @@ struct ProcessorProfileLogElement UInt64 id{}; std::vector parent_ids; - UInt64 plan_step{}; + UInt64 plan_step; UInt64 plan_group{}; String initial_query_id; String query_id; String processor_name; + String processor_uniq_id; + String step_uniq_id; /// Milliseconds spend in IProcessor::work() UInt32 elapsed_us{}; diff --git a/src/Interpreters/executeQuery.cpp b/src/Interpreters/executeQuery.cpp index 33a4cf2a74c..59573e912e4 100644 --- a/src/Interpreters/executeQuery.cpp +++ b/src/Interpreters/executeQuery.cpp @@ -460,6 +460,8 @@ void logQueryFinish( processor_elem.plan_step = reinterpret_cast(processor->getQueryPlanStep()); processor_elem.plan_group = processor->getQueryPlanStepGroup(); + processor_elem.processor_uniq_id = processor->getUniqID(); + processor_elem.step_uniq_id = processor->getStepUniqID(); processor_elem.processor_name = processor->getName(); diff --git a/src/Processors/Executors/ExecutionThreadContext.cpp b/src/Processors/Executors/ExecutionThreadContext.cpp index 05669725f9a..06b4b53c817 100644 --- a/src/Processors/Executors/ExecutionThreadContext.cpp +++ b/src/Processors/Executors/ExecutionThreadContext.cpp @@ -79,7 +79,7 @@ bool ExecutionThreadContext::executeTask() if (trace_processors) { - span = std::make_unique(node->processor->getName()); + span = std::make_unique(node->processor->getUniqID()); span->addAttribute("thread_number", thread_number); } std::optional execution_time_watch; diff --git a/src/Processors/IProcessor.h b/src/Processors/IProcessor.h index 56b4509fe00..b99ebeb5fa5 100644 --- a/src/Processors/IProcessor.h +++ b/src/Processors/IProcessor.h @@ -3,7 +3,9 @@ #include #include #include - +#include +#include +#include class EventCounter; @@ -121,7 +123,10 @@ protected: OutputPorts outputs; public: - IProcessor() = default; + IProcessor() + { + setProcessorIndex(); + } IProcessor(InputPorts inputs_, OutputPorts outputs_) : inputs(std::move(inputs_)), outputs(std::move(outputs_)) @@ -130,9 +135,16 @@ public: port.processor = this; for (auto & port : outputs) port.processor = this; + setProcessorIndex(); + } + + void setProcessorIndex() + { + processor_index = CurrentThread::get().incrProcessorIndex(); } virtual String getName() const = 0; + String getUniqID() const { return fmt::format("{}_{}", getName(), processor_index); } enum class Status { @@ -300,11 +312,16 @@ public: /// Step of QueryPlan from which processor was created. void setQueryPlanStep(IQueryPlanStep * step, size_t group = 0) { - query_plan_step = step; + if (step != nullptr) + { + query_plan_step = step; + step_uniq_id = step->getUniqID(); + } query_plan_step_group = group; } IQueryPlanStep * getQueryPlanStep() const { return query_plan_step; } + const String &getStepUniqID() const { return step_uniq_id; } size_t getQueryPlanStepGroup() const { return query_plan_step_group; } uint64_t getElapsedUs() const { return elapsed_us; } @@ -392,7 +409,10 @@ private: size_t stream_number = NO_STREAM; IQueryPlanStep * query_plan_step = nullptr; + String step_uniq_id; size_t query_plan_step_group = 0; + + size_t processor_index = 0; }; diff --git a/src/Processors/QueryPlan/IQueryPlanStep.h b/src/Processors/QueryPlan/IQueryPlanStep.h index ac5ea259d2e..ec5ac9ad4dc 100644 --- a/src/Processors/QueryPlan/IQueryPlanStep.h +++ b/src/Processors/QueryPlan/IQueryPlanStep.h @@ -2,6 +2,9 @@ #include #include #include +#include +#include +#include namespace DB { @@ -71,6 +74,10 @@ using QueryPlanRawPtrs = std::list; class IQueryPlanStep { public: + IQueryPlanStep() + { + step_index = CurrentThread::get().incrStepIndex(); + } virtual ~IQueryPlanStep() = default; virtual String getName() const = 0; @@ -138,7 +145,7 @@ public: } virtual bool canUpdateInputStream() const { return false; } - + String getUniqID() const { return fmt::format("{}_{}", getName(), step_index); } protected: virtual void updateOutputStream() { throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Not implemented"); } @@ -153,6 +160,9 @@ protected: Processors processors; static void describePipeline(const Processors & processors, FormatSettings & settings); + +private: + size_t step_index = 0; }; using QueryPlanStepPtr = std::unique_ptr; diff --git a/src/Processors/QueryPlan/QueryPlan.cpp b/src/Processors/QueryPlan/QueryPlan.cpp index 0fae7e8df4d..f651870453b 100644 --- a/src/Processors/QueryPlan/QueryPlan.cpp +++ b/src/Processors/QueryPlan/QueryPlan.cpp @@ -206,6 +206,7 @@ QueryPipelineBuilderPtr QueryPlan::buildQueryPipeline( static void explainStep(const IQueryPlanStep & step, JSONBuilder::JSONMap & map, const QueryPlan::ExplainPlanOptions & options) { map.add("Node Type", step.getName()); + map.add("Node Id", step.getUniqID()); if (options.description) { diff --git a/src/QueryPipeline/QueryPipelineBuilder.cpp b/src/QueryPipeline/QueryPipelineBuilder.cpp index 67a8fe5dcab..4b6f15905ce 100644 --- a/src/QueryPipeline/QueryPipelineBuilder.cpp +++ b/src/QueryPipeline/QueryPipelineBuilder.cpp @@ -400,10 +400,10 @@ std::unique_ptr QueryPipelineBuilder::joinPipelinesRightLe left->pipe.collected_processors = collected_processors; - /// Collect the NEW processors for the right pipeline. - QueryPipelineProcessorsCollector collector(*right); /// Remember the last step of the right pipeline. IQueryPlanStep * step = right->pipe.processors->back()->getQueryPlanStep(); + /// Collect the NEW processors for the right pipeline. + QueryPipelineProcessorsCollector collector(*right, step); /// In case joined subquery has totals, and we don't, add default chunk to totals. bool default_totals = false; diff --git a/src/QueryPipeline/printPipeline.cpp b/src/QueryPipeline/printPipeline.cpp index 40c88502ed0..1726d776921 100644 --- a/src/QueryPipeline/printPipeline.cpp +++ b/src/QueryPipeline/printPipeline.cpp @@ -113,7 +113,7 @@ void printPipelineCompact(const Processors & processors, WriteBuffer & out, bool if (item.first != nullptr) { out << " subgraph cluster_" << next_step << " {\n"; - out << " label =\"" << item.first->getName() << "\";\n"; + out << " label =\"" << item.first->getUniqID() << "\";\n"; out << " style=filled;\n"; out << " color=lightgrey;\n"; out << " node [style=filled,color=white];\n"; @@ -125,7 +125,7 @@ void printPipelineCompact(const Processors & processors, WriteBuffer & out, bool for (const auto & node : item.second) { const auto & processor = node->agents.front(); - out << " n" << node->id << " [label=\"" << processor->getName(); + out << " n" << node->id << " [label=\"" << processor->getUniqID(); if (node->agents.size() > 1) out << " × " << node->agents.size(); diff --git a/src/QueryPipeline/printPipeline.h b/src/QueryPipeline/printPipeline.h index 2bdbd8f7a07..e6799251851 100644 --- a/src/QueryPipeline/printPipeline.h +++ b/src/QueryPipeline/printPipeline.h @@ -30,7 +30,7 @@ void printPipeline(const Processors & processors, const Statuses & statuses, Wri for (const auto & processor : processors) { const auto & description = processor->getDescription(); - out << " n" << get_proc_id(*processor) << "[label=\"" << processor->getName() << (description.empty() ? "" : ":") << description; + out << " n" << get_proc_id(*processor) << "[label=\"" << processor->getUniqID() << (description.empty() ? "" : ":") << description; if (statuses_iter != statuses.end()) { diff --git a/tests/queries/0_stateless/01786_explain_merge_tree.reference b/tests/queries/0_stateless/01786_explain_merge_tree.reference index 3a015d32539..36ebbe1a1da 100644 --- a/tests/queries/0_stateless/01786_explain_merge_tree.reference +++ b/tests/queries/0_stateless/01786_explain_merge_tree.reference @@ -29,6 +29,7 @@ Granules: 2/3 ----------------- "Node Type": "ReadFromMergeTree", + "Node Id": "ReadFromMergeTree_0", "Description": "default.test_index", "Indexes": [ { @@ -126,6 +127,7 @@ Granules: 3/6 ----------------- "Node Type": "ReadFromMergeTree", + "Node Id": "ReadFromMergeTree_0", "Description": "default.test_index", "Indexes": [ { From 59bd7447fcb1db44bc77d93339b36dae684d5daf Mon Sep 17 00:00:00 2001 From: qhsong Date: Tue, 30 Jul 2024 10:23:51 +0800 Subject: [PATCH 006/244] Fix testcase --- src/Common/ThreadStatus.h | 2 +- src/Interpreters/Context.h | 2 -- src/Processors/IProcessor.h | 5 ++++- src/Processors/QueryPlan/IQueryPlanStep.h | 5 ++++- .../0_stateless/01786_explain_merge_tree.reference | 4 ++-- .../0_stateless/01823_explain_json.reference | 13 +++++++++++-- .../03213_distributed_analyzer.reference | 2 +- 7 files changed, 23 insertions(+), 10 deletions(-) diff --git a/src/Common/ThreadStatus.h b/src/Common/ThreadStatus.h index 97b45c01e54..fd384ad1603 100644 --- a/src/Common/ThreadStatus.h +++ b/src/Common/ThreadStatus.h @@ -93,7 +93,7 @@ public: //QueryPlan can not build parallel, but processor may build parallel in expand() function. //so we use atomic_size_t for processor_count - std::shared_ptr step_count = std::make_shared(0); + std::shared_ptr step_count = std::make_shared(0); std::shared_ptr processor_count = std::make_shared(0); QueryIsCanceledPredicate query_is_canceled_predicate = {}; diff --git a/src/Interpreters/Context.h b/src/Interpreters/Context.h index 1f331f0d094..cb553d07513 100644 --- a/src/Interpreters/Context.h +++ b/src/Interpreters/Context.h @@ -1404,9 +1404,7 @@ private: std::shared_ptr getClustersImpl(std::lock_guard & lock) const; /// Throttling - public: - ThrottlerPtr getReplicatedFetchesThrottler() const; ThrottlerPtr getReplicatedSendsThrottler() const; diff --git a/src/Processors/IProcessor.h b/src/Processors/IProcessor.h index 3933a79ab55..d426d5ef9ba 100644 --- a/src/Processors/IProcessor.h +++ b/src/Processors/IProcessor.h @@ -140,7 +140,10 @@ public: void setProcessorIndex() { - processor_index = CurrentThread::get().incrProcessorIndex(); + if (CurrentThread::isInitialized()) [[likely]] + { + processor_index = CurrentThread::get().incrProcessorIndex(); + } } virtual String getName() const = 0; diff --git a/src/Processors/QueryPlan/IQueryPlanStep.h b/src/Processors/QueryPlan/IQueryPlanStep.h index acd8857b9df..500e0812983 100644 --- a/src/Processors/QueryPlan/IQueryPlanStep.h +++ b/src/Processors/QueryPlan/IQueryPlanStep.h @@ -76,7 +76,10 @@ class IQueryPlanStep public: IQueryPlanStep() { - step_index = CurrentThread::get().incrStepIndex(); + if (CurrentThread::isInitialized()) [[likely]] + { + step_index = CurrentThread::get().incrStepIndex(); + } } virtual ~IQueryPlanStep() = default; diff --git a/tests/queries/0_stateless/01786_explain_merge_tree.reference b/tests/queries/0_stateless/01786_explain_merge_tree.reference index 36ebbe1a1da..75736669905 100644 --- a/tests/queries/0_stateless/01786_explain_merge_tree.reference +++ b/tests/queries/0_stateless/01786_explain_merge_tree.reference @@ -29,7 +29,7 @@ Granules: 2/3 ----------------- "Node Type": "ReadFromMergeTree", - "Node Id": "ReadFromMergeTree_0", + "Node Id": "ReadFromMergeTree_1", "Description": "default.test_index", "Indexes": [ { @@ -127,7 +127,7 @@ Granules: 3/6 ----------------- "Node Type": "ReadFromMergeTree", - "Node Id": "ReadFromMergeTree_0", + "Node Id": "ReadFromMergeTree_1", "Description": "default.test_index", "Indexes": [ { diff --git a/tests/queries/0_stateless/01823_explain_json.reference b/tests/queries/0_stateless/01823_explain_json.reference index 23fb34c2192..1aa5aa134e9 100644 --- a/tests/queries/0_stateless/01823_explain_json.reference +++ b/tests/queries/0_stateless/01823_explain_json.reference @@ -2,20 +2,25 @@ { "Plan": { "Node Type": "Union", + "Node Id": "Union_11", "Plans": [ { "Node Type": "Expression", + "Node Id": "Expression_14", "Plans": [ { - "Node Type": "ReadFromStorage" + "Node Type": "ReadFromStorage", + "Node Id": "ReadFromStorage_1" } ] }, { "Node Type": "Expression", + "Node Id": "Expression_17", "Plans": [ { - "Node Type": "ReadFromStorage" + "Node Type": "ReadFromStorage", + "Node Id": "ReadFromStorage_5" } ] } @@ -35,6 +40,7 @@ } -------- "Node Type": "Aggregating", + "Node Id": "Aggregating_4", "Header": [ { "Name": "__table1.number", @@ -73,13 +79,16 @@ ], -------- "Node Type": "ArrayJoin", + "Node Id": "ArrayJoin_5", "Left": false, "Columns": ["__table1.x", "__table1.y"], -------- "Node Type": "Distinct", + "Node Id": "Distinct_5", "Columns": ["intDiv(__table1.number, 2_UInt8)", "intDiv(__table1.number, 3_UInt8)"], -- "Node Type": "Distinct", + "Node Id": "Distinct_4", "Columns": ["intDiv(__table1.number, 2_UInt8)", "intDiv(__table1.number, 3_UInt8)"], -------- "Sort Description": [ diff --git a/tests/queries/0_stateless/03213_distributed_analyzer.reference b/tests/queries/0_stateless/03213_distributed_analyzer.reference index 9d63c0a7a5e..2456192ca9d 100644 --- a/tests/queries/0_stateless/03213_distributed_analyzer.reference +++ b/tests/queries/0_stateless/03213_distributed_analyzer.reference @@ -1 +1 @@ -['digraph','{',' rankdir="LR";',' { node [shape = rect]',' n1 [label="Remote"];',' }','}','digraph','{',' rankdir="LR";',' { node [shape = rect]',' n1 [label="Remote"];',' }','}'] +['digraph','{',' rankdir="LR";',' { node [shape = rect]',' n1 [label="Remote_4"];',' }','}','digraph','{',' rankdir="LR";',' { node [shape = rect]',' n1 [label="Remote_4"];',' }','}'] From d10b79020edda1f35f9f3637447cd57d90352bae Mon Sep 17 00:00:00 2001 From: vdimir Date: Mon, 14 Oct 2024 13:39:03 +0000 Subject: [PATCH 007/244] Refactoring TempDataOnDisk --- src/Common/CurrentMetrics.cpp | 1 + .../gtest_cascade_and_memory_write_buffer.cpp | 2 +- src/Interpreters/Aggregator.cpp | 12 +- src/Interpreters/Aggregator.h | 9 +- src/Interpreters/Context.cpp | 20 +- src/Interpreters/GraceHashJoin.cpp | 60 +- src/Interpreters/GraceHashJoin.h | 2 +- src/Interpreters/HashJoin/HashJoin.cpp | 29 +- src/Interpreters/HashJoin/HashJoin.h | 5 +- src/Interpreters/TableJoin.h | 7 +- src/Interpreters/TemporaryDataOnDisk.cpp | 647 ++++++++---------- src/Interpreters/TemporaryDataOnDisk.h | 233 ++++--- src/Interpreters/tests/gtest_filecache.cpp | 85 ++- .../Algorithms/CollapsingSortedAlgorithm.cpp | 6 +- .../Algorithms/CollapsingSortedAlgorithm.h | 6 +- .../Algorithms/MergingSortedAlgorithm.cpp | 3 +- .../Algorithms/MergingSortedAlgorithm.h | 6 +- .../Algorithms/ReplacingSortedAlgorithm.cpp | 6 +- .../Algorithms/ReplacingSortedAlgorithm.h | 6 +- .../VersionedCollapsingAlgorithm.cpp | 6 +- .../Algorithms/VersionedCollapsingAlgorithm.h | 6 +- .../Merges/CollapsingSortedTransform.h | 2 +- .../Merges/MergingSortedTransform.cpp | 2 +- .../Merges/MergingSortedTransform.h | 2 +- .../Merges/ReplacingSortedTransform.h | 4 +- .../Merges/VersionedCollapsingTransform.h | 4 +- .../QueryPlan/BuildQueryPipelineSettings.h | 1 - src/Processors/QueryPlan/SortingStep.cpp | 6 +- .../Transforms/AggregatingTransform.cpp | 27 +- .../Transforms/MergeSortingTransform.cpp | 44 +- .../Transforms/MergeSortingTransform.h | 5 +- src/QueryPipeline/QueryPipelineBuilder.h | 6 + src/QueryPipeline/QueryPlanResourceHolder.h | 2 + src/Server/HTTPHandler.cpp | 39 +- src/Storages/MergeTree/MergeTask.cpp | 124 +--- src/Storages/MergeTree/MergeTask.h | 5 +- .../MergeTree/MergeTreeSelectProcessor.cpp | 5 +- 37 files changed, 646 insertions(+), 789 deletions(-) diff --git a/src/Common/CurrentMetrics.cpp b/src/Common/CurrentMetrics.cpp index bd62e7e8aae..8d232e11df3 100644 --- a/src/Common/CurrentMetrics.cpp +++ b/src/Common/CurrentMetrics.cpp @@ -45,6 +45,7 @@ M(TemporaryFilesForSort, "Number of temporary files created for external sorting") \ M(TemporaryFilesForAggregation, "Number of temporary files created for external aggregation") \ M(TemporaryFilesForJoin, "Number of temporary files created for JOIN") \ + M(TemporaryFilesForMerge, "Number of temporary files for vertical merge") \ M(TemporaryFilesUnknown, "Number of temporary files created without known purpose") \ M(Read, "Number of read (read, pread, io_getevents, etc.) syscalls in fly") \ M(RemoteRead, "Number of read with remote reader in fly") \ diff --git a/src/Disks/tests/gtest_cascade_and_memory_write_buffer.cpp b/src/Disks/tests/gtest_cascade_and_memory_write_buffer.cpp index 23b783173c8..6fd7570c4eb 100644 --- a/src/Disks/tests/gtest_cascade_and_memory_write_buffer.cpp +++ b/src/Disks/tests/gtest_cascade_and_memory_write_buffer.cpp @@ -69,7 +69,7 @@ static void testCascadeBufferRedability( auto rbuf = wbuf_readable.tryGetReadBuffer(); ASSERT_FALSE(!rbuf); - concat.appendBuffer(wrapReadBufferPointer(std::move(rbuf))); + concat.appendBuffer(std::move(rbuf)); } std::string decoded_data; diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp index 0fe1c74ed17..e6fecc37cfa 100644 --- a/src/Interpreters/Aggregator.cpp +++ b/src/Interpreters/Aggregator.cpp @@ -335,7 +335,7 @@ Aggregator::Aggregator(const Block & header_, const Params & params_) : header(header_) , keys_positions(calculateKeysPositions(header, params_)) , params(params_) - , tmp_data(params.tmp_data_scope ? std::make_unique(params.tmp_data_scope, CurrentMetrics::TemporaryFilesForAggregation) : nullptr) + , tmp_data(params.tmp_data_scope ? params.tmp_data_scope->childScope(CurrentMetrics::TemporaryFilesForAggregation) : nullptr) , min_bytes_for_prefetch(getMinBytesForPrefetch()) { /// Use query-level memory tracker @@ -1519,10 +1519,10 @@ void Aggregator::writeToTemporaryFile(AggregatedDataVariants & data_variants, si Stopwatch watch; size_t rows = data_variants.size(); - auto & out_stream = tmp_data->createStream(getHeader(false), max_temp_file_size); + auto & out_stream = tmp_files.emplace_back(getHeader(false), tmp_data.get(), max_temp_file_size); ProfileEvents::increment(ProfileEvents::ExternalAggregationWritePart); - LOG_DEBUG(log, "Writing part of aggregation data into temporary file {}", out_stream.getPath()); + LOG_DEBUG(log, "Writing part of aggregation data into temporary file {}", out_stream.getHolder()->describeFilePath()); /// Flush only two-level data and possibly overflow data. @@ -1643,7 +1643,7 @@ template void Aggregator::writeToTemporaryFileImpl( AggregatedDataVariants & data_variants, Method & method, - TemporaryFileStream & out) const + TemporaryBlockStreamHolder & out) const { size_t max_temporary_block_size_rows = 0; size_t max_temporary_block_size_bytes = 0; @@ -1660,14 +1660,14 @@ void Aggregator::writeToTemporaryFileImpl( for (UInt32 bucket = 0; bucket < Method::Data::NUM_BUCKETS; ++bucket) { Block block = convertOneBucketToBlock(data_variants, method, data_variants.aggregates_pool, false, bucket); - out.write(block); + out->write(block); update_max_sizes(block); } if (params.overflow_row) { Block block = prepareBlockAndFillWithoutKey(data_variants, false, true); - out.write(block); + out->write(block); update_max_sizes(block); } diff --git a/src/Interpreters/Aggregator.h b/src/Interpreters/Aggregator.h index 4de0a640219..bc28d3dccb8 100644 --- a/src/Interpreters/Aggregator.h +++ b/src/Interpreters/Aggregator.h @@ -309,9 +309,9 @@ public: /// For external aggregation. void writeToTemporaryFile(AggregatedDataVariants & data_variants, size_t max_temp_file_size = 0) const; - bool hasTemporaryData() const { return tmp_data && !tmp_data->empty(); } + bool hasTemporaryData() const { return !tmp_files.empty(); } - const TemporaryDataOnDisk & getTemporaryData() const { return *tmp_data; } + std::vector & getTemporaryData() { return tmp_files; } /// Get data structure of the result. Block getHeader(bool final) const; @@ -355,7 +355,8 @@ private: LoggerPtr log = getLogger("Aggregator"); /// For external aggregation. - TemporaryDataOnDiskPtr tmp_data; + TemporaryDataOnDiskScopePtr tmp_data; + mutable std::vector tmp_files; size_t min_bytes_for_prefetch = 0; @@ -456,7 +457,7 @@ private: void writeToTemporaryFileImpl( AggregatedDataVariants & data_variants, Method & method, - TemporaryFileStream & out) const; + TemporaryBlockStreamHolder & out) const; /// Merge NULL key data from hash table `src` into `dst`. template diff --git a/src/Interpreters/Context.cpp b/src/Interpreters/Context.cpp index 85cde959b66..6ada12e63f9 100644 --- a/src/Interpreters/Context.cpp +++ b/src/Interpreters/Context.cpp @@ -353,6 +353,8 @@ struct ContextSharedPart : boost::noncopyable /// Child scopes for more fine-grained accounting are created per user/query/etc. /// Initialized once during server startup. TemporaryDataOnDiskScopePtr root_temp_data_on_disk TSA_GUARDED_BY(mutex); + /// TODO: remove, use only root_temp_data_on_disk + VolumePtr temporary_volume_legacy; mutable OnceFlag async_loader_initialized; mutable std::unique_ptr async_loader; /// Thread pool for asynchronous initialization of arbitrary DAG of `LoadJob`s (used for tables loading) @@ -783,10 +785,9 @@ struct ContextSharedPart : boost::noncopyable } /// Special volumes might also use disks that require shutdown. - auto & tmp_data = root_temp_data_on_disk; - if (tmp_data && tmp_data->getVolume()) + if (temporary_volume_legacy) { - auto & disks = tmp_data->getVolume()->getDisks(); + auto & disks = temporary_volume_legacy->getDisks(); for (auto & disk : disks) disk->shutdown(); } @@ -1166,8 +1167,8 @@ VolumePtr Context::getGlobalTemporaryVolume() const SharedLockGuard lock(shared->mutex); /// Calling this method we just bypass the `temp_data_on_disk` and write to the file on the volume directly. /// Volume is the same for `root_temp_data_on_disk` (always set) and `temp_data_on_disk` (if it's set). - if (shared->root_temp_data_on_disk) - return shared->root_temp_data_on_disk->getVolume(); + if (shared->temporary_volume_legacy) + return shared->temporary_volume_legacy; return nullptr; } @@ -1288,7 +1289,8 @@ void Context::setTemporaryStoragePath(const String & path, size_t max_size) TemporaryDataOnDiskSettings temporary_data_on_disk_settings; temporary_data_on_disk_settings.max_size_on_disk = max_size; - shared->root_temp_data_on_disk = std::make_shared(std::move(volume), std::move(temporary_data_on_disk_settings)); + shared->root_temp_data_on_disk = std::make_shared(volume, std::move(temporary_data_on_disk_settings)); + shared->temporary_volume_legacy = volume; } void Context::setTemporaryStoragePolicy(const String & policy_name, size_t max_size) @@ -1336,7 +1338,8 @@ void Context::setTemporaryStoragePolicy(const String & policy_name, size_t max_s TemporaryDataOnDiskSettings temporary_data_on_disk_settings; temporary_data_on_disk_settings.max_size_on_disk = max_size; - shared->root_temp_data_on_disk = std::make_shared(std::move(volume), std::move(temporary_data_on_disk_settings)); + shared->root_temp_data_on_disk = std::make_shared(volume, std::move(temporary_data_on_disk_settings)); + shared->temporary_volume_legacy = volume; } void Context::setTemporaryStorageInCache(const String & cache_disk_name, size_t max_size) @@ -1360,7 +1363,8 @@ void Context::setTemporaryStorageInCache(const String & cache_disk_name, size_t TemporaryDataOnDiskSettings temporary_data_on_disk_settings; temporary_data_on_disk_settings.max_size_on_disk = max_size; - shared->root_temp_data_on_disk = std::make_shared(std::move(volume), file_cache.get(), std::move(temporary_data_on_disk_settings)); + shared->root_temp_data_on_disk = std::make_shared(file_cache.get(), std::move(temporary_data_on_disk_settings)); + shared->temporary_volume_legacy = volume; } void Context::setFlagsPath(const String & path) diff --git a/src/Interpreters/GraceHashJoin.cpp b/src/Interpreters/GraceHashJoin.cpp index 978782c851f..a2010b7d94b 100644 --- a/src/Interpreters/GraceHashJoin.cpp +++ b/src/Interpreters/GraceHashJoin.cpp @@ -41,15 +41,15 @@ namespace class AccumulatedBlockReader { public: - AccumulatedBlockReader(TemporaryFileStream & reader_, + AccumulatedBlockReader(TemporaryBlockStreamReaderHolder reader_, std::mutex & mutex_, size_t result_block_size_ = 0) - : reader(reader_) + : reader(std::move(reader_)) , mutex(mutex_) , result_block_size(result_block_size_) { - if (!reader.isWriteFinished()) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Reading not finished file"); + if (!reader) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Reader is nullptr"); } Block read() @@ -63,7 +63,7 @@ namespace size_t rows_read = 0; do { - Block block = reader.read(); + Block block = reader->read(); rows_read += block.rows(); if (!block) { @@ -81,7 +81,7 @@ namespace } private: - TemporaryFileStream & reader; + TemporaryBlockStreamReaderHolder reader; std::mutex & mutex; const size_t result_block_size; @@ -124,12 +124,12 @@ class GraceHashJoin::FileBucket : boost::noncopyable public: using BucketLock = std::unique_lock; - explicit FileBucket(size_t bucket_index_, TemporaryFileStream & left_file_, TemporaryFileStream & right_file_, LoggerPtr log_) - : idx{bucket_index_} - , left_file{left_file_} - , right_file{right_file_} - , state{State::WRITING_BLOCKS} - , log{log_} + explicit FileBucket(size_t bucket_index_, TemporaryBlockStreamHolder left_file_, TemporaryBlockStreamHolder right_file_, LoggerPtr log_) + : idx(bucket_index_) + , left_file(std::move(left_file_)) + , right_file(std::move(right_file_)) + , state(State::WRITING_BLOCKS) + , log(log_) { } @@ -157,12 +157,6 @@ public: return addBlockImpl(block, right_file, lock); } - bool finished() const - { - std::unique_lock left_lock(left_file_mutex); - return left_file.isEof(); - } - bool empty() const { return is_empty.load(); } AccumulatedBlockReader startJoining() @@ -172,24 +166,21 @@ public: std::unique_lock left_lock(left_file_mutex); std::unique_lock right_lock(right_file_mutex); - left_file.finishWriting(); - right_file.finishWriting(); - state = State::JOINING_BLOCKS; } - return AccumulatedBlockReader(right_file, right_file_mutex); + return AccumulatedBlockReader(right_file.getReadStream(), right_file_mutex); } AccumulatedBlockReader getLeftTableReader() { ensureState(State::JOINING_BLOCKS); - return AccumulatedBlockReader(left_file, left_file_mutex); + return AccumulatedBlockReader(left_file.getReadStream(), left_file_mutex); } const size_t idx; private: - bool addBlockImpl(const Block & block, TemporaryFileStream & writer, std::unique_lock & lock) + bool addBlockImpl(const Block & block, TemporaryBlockStreamHolder & writer, std::unique_lock & lock) { ensureState(State::WRITING_BLOCKS); @@ -199,7 +190,7 @@ private: if (block.rows()) is_empty = false; - writer.write(block); + writer->write(block); return true; } @@ -217,8 +208,8 @@ private: throw Exception(ErrorCodes::LOGICAL_ERROR, "Invalid state transition, expected {}, got {}", expected, state.load()); } - TemporaryFileStream & left_file; - TemporaryFileStream & right_file; + TemporaryBlockStreamHolder left_file; + TemporaryBlockStreamHolder right_file; mutable std::mutex left_file_mutex; mutable std::mutex right_file_mutex; @@ -274,7 +265,7 @@ GraceHashJoin::GraceHashJoin( , max_num_buckets{context->getSettingsRef()[Setting::grace_hash_join_max_buckets]} , left_key_names(table_join->getOnlyClause().key_names_left) , right_key_names(table_join->getOnlyClause().key_names_right) - , tmp_data(std::make_unique(tmp_data_, CurrentMetrics::TemporaryFilesForJoin)) + , tmp_data(tmp_data_->childScope(CurrentMetrics::TemporaryFilesForJoin)) , hash_join(makeInMemoryJoin("grace0")) , hash_join_sample_block(hash_join->savedBlockSample()) { @@ -398,10 +389,10 @@ void GraceHashJoin::addBuckets(const size_t bucket_count) for (size_t i = 0; i < bucket_count; ++i) try { - auto & left_file = tmp_data->createStream(left_sample_block); - auto & right_file = tmp_data->createStream(prepareRightBlock(right_sample_block)); + TemporaryBlockStreamHolder left_file = TemporaryBlockStreamHolder(left_sample_block, tmp_data.get()); + TemporaryBlockStreamHolder right_file = TemporaryBlockStreamHolder(prepareRightBlock(right_sample_block), tmp_data.get()); - BucketPtr new_bucket = std::make_shared(current_size + i, left_file, right_file, log); + BucketPtr new_bucket = std::make_shared(current_size + i, std::move(left_file), std::move(right_file), log); tmp_buckets.emplace_back(std::move(new_bucket)); } catch (...) @@ -632,12 +623,9 @@ IBlocksStreamPtr GraceHashJoin::getDelayedBlocks() for (bucket_idx = bucket_idx + 1; bucket_idx < buckets.size(); ++bucket_idx) { current_bucket = buckets[bucket_idx].get(); - if (current_bucket->finished() || current_bucket->empty()) + if (current_bucket->empty()) { - LOG_TRACE(log, "Skipping {} {} bucket {}", - current_bucket->finished() ? "finished" : "", - current_bucket->empty() ? "empty" : "", - bucket_idx); + LOG_TRACE(log, "Skipping empty bucket {}", bucket_idx); continue; } diff --git a/src/Interpreters/GraceHashJoin.h b/src/Interpreters/GraceHashJoin.h index d31d6886af7..938c9b1facf 100644 --- a/src/Interpreters/GraceHashJoin.h +++ b/src/Interpreters/GraceHashJoin.h @@ -132,7 +132,7 @@ private: Names left_key_names; Names right_key_names; - TemporaryDataOnDiskPtr tmp_data; + TemporaryDataOnDiskScopePtr tmp_data; Buckets buckets; mutable SharedMutex rehash_mutex; diff --git a/src/Interpreters/HashJoin/HashJoin.cpp b/src/Interpreters/HashJoin/HashJoin.cpp index 3e7f3deea8b..af23b520abb 100644 --- a/src/Interpreters/HashJoin/HashJoin.cpp +++ b/src/Interpreters/HashJoin/HashJoin.cpp @@ -35,11 +35,6 @@ #include #include -namespace CurrentMetrics -{ - extern const Metric TemporaryFilesForJoin; -} - namespace DB { @@ -64,7 +59,7 @@ struct NotProcessedCrossJoin : public ExtraBlock { size_t left_position; size_t right_block; - std::unique_ptr reader; + TemporaryBlockStreamReaderHolder reader; }; @@ -106,10 +101,7 @@ HashJoin::HashJoin(std::shared_ptr table_join_, const Block & right_s , instance_id(instance_id_) , asof_inequality(table_join->getAsofInequality()) , data(std::make_shared()) - , tmp_data( - table_join_->getTempDataOnDisk() - ? std::make_unique(table_join_->getTempDataOnDisk(), CurrentMetrics::TemporaryFilesForJoin) - : nullptr) + , tmp_data(table_join_->getTempDataOnDisk()) , right_sample_block(right_sample_block_) , max_joined_block_rows(table_join->maxJoinedBlockRows()) , instance_log_id(!instance_id_.empty() ? "(" + instance_id_ + ") " : "") @@ -520,10 +512,9 @@ bool HashJoin::addBlockToJoin(const Block & source_block_, bool check_limits) && (tmp_stream || (max_bytes_in_join && getTotalByteCount() + block_to_save.allocatedBytes() >= max_bytes_in_join) || (max_rows_in_join && getTotalRowCount() + block_to_save.rows() >= max_rows_in_join))) { - if (tmp_stream == nullptr) - { - tmp_stream = &tmp_data->createStream(right_sample_block); - } + if (!tmp_stream) + tmp_stream = TemporaryBlockStreamHolder(right_sample_block, tmp_data.get()); + tmp_stream->write(block_to_save); return true; } @@ -730,7 +721,7 @@ void HashJoin::joinBlockImplCross(Block & block, ExtraBlockPtr & not_processed) { size_t start_left_row = 0; size_t start_right_block = 0; - std::unique_ptr reader = nullptr; + TemporaryBlockStreamReaderHolder reader; if (not_processed) { auto & continuation = static_cast(*not_processed); @@ -804,11 +795,9 @@ void HashJoin::joinBlockImplCross(Block & block, ExtraBlockPtr & not_processed) if (tmp_stream && rows_added <= max_joined_block_rows) { - if (reader == nullptr) - { - tmp_stream->finishWritingAsyncSafe(); - reader = tmp_stream->getReadStream(); - } + if (!reader) + reader = tmp_stream.getReadStream(); + while (auto block_right = reader->read()) { ++block_number; diff --git a/src/Interpreters/HashJoin/HashJoin.h b/src/Interpreters/HashJoin/HashJoin.h index 4c1ebbcdc66..0f50e110db9 100644 --- a/src/Interpreters/HashJoin/HashJoin.h +++ b/src/Interpreters/HashJoin/HashJoin.h @@ -423,8 +423,9 @@ private: std::vector key_sizes; /// Needed to do external cross join - TemporaryDataOnDiskPtr tmp_data; - TemporaryFileStream* tmp_stream{nullptr}; + TemporaryDataOnDiskScopePtr tmp_data; + TemporaryBlockStreamHolder tmp_stream; + mutable std::once_flag finish_writing; /// Block with columns from the right-side table. Block right_sample_block; diff --git a/src/Interpreters/TableJoin.h b/src/Interpreters/TableJoin.h index e1bae55a4ed..4ecbc9eb960 100644 --- a/src/Interpreters/TableJoin.h +++ b/src/Interpreters/TableJoin.h @@ -20,6 +20,11 @@ #include #include +namespace CurrentMetrics +{ + extern const Metric TemporaryFilesForJoin; +} + namespace DB { @@ -265,7 +270,7 @@ public: VolumePtr getGlobalTemporaryVolume() { return tmp_volume; } - TemporaryDataOnDiskScopePtr getTempDataOnDisk() { return tmp_data; } + TemporaryDataOnDiskScopePtr getTempDataOnDisk() { return tmp_data ? tmp_data->childScope(CurrentMetrics::TemporaryFilesForJoin) : nullptr; } ActionsDAG createJoinedBlockActions(ContextPtr context) const; diff --git a/src/Interpreters/TemporaryDataOnDisk.cpp b/src/Interpreters/TemporaryDataOnDisk.cpp index 81796678f24..c3b24fb783b 100644 --- a/src/Interpreters/TemporaryDataOnDisk.cpp +++ b/src/Interpreters/TemporaryDataOnDisk.cpp @@ -27,11 +27,266 @@ namespace DB namespace ErrorCodes { - extern const int TOO_MANY_ROWS_OR_BYTES; + extern const int INVALID_STATE; extern const int LOGICAL_ERROR; extern const int NOT_ENOUGH_SPACE; + extern const int TOO_MANY_ROWS_OR_BYTES; } +namespace +{ + +inline CompressionCodecPtr getCodec(const TemporaryDataOnDiskSettings & settings) +{ + if (settings.compression_codec.empty()) + return CompressionCodecFactory::instance().get("NONE"); + + return CompressionCodecFactory::instance().get(settings.compression_codec); +} + +} + +TemporaryFileHolder::TemporaryFileHolder() +{ + ProfileEvents::increment(ProfileEvents::ExternalProcessingFilesTotal); +} + + +class TemporaryFileInLocalCache : public TemporaryFileHolder +{ +public: + explicit TemporaryFileInLocalCache(FileCache & file_cache, size_t max_file_size = 0) + { + const auto key = FileSegment::Key::random(); + segment_holder = file_cache.set( + key, 0, std::max(10_MiB, max_file_size), + CreateFileSegmentSettings(FileSegmentKind::Ephemeral), FileCache::getCommonUser()); + + chassert(segment_holder->size() == 1); + segment_holder->front().getKeyMetadata()->createBaseDirectory(/* throw_if_failed */true); + } + + std::unique_ptr write() override + { + return std::make_unique(&segment_holder->front()); + } + + std::unique_ptr read(size_t buffer_size) const override + { + return std::make_unique(segment_holder->front().getPath(), /* buf_size = */ buffer_size); + } + + String describeFilePath() const override + { + return fmt::format("fscache://{}", segment_holder->front().getPath()); + } + +private: + FileSegmentsHolderPtr segment_holder; +}; + +class TemporaryFileOnLocalDisk : public TemporaryFileHolder +{ +public: + explicit TemporaryFileOnLocalDisk(VolumePtr volume, size_t max_file_size = 0) + : path_to_file("tmp" + toString(UUIDHelpers::generateV4())) + { + if (max_file_size > 0) + { + auto reservation = volume->reserve(max_file_size); + if (!reservation) + throw Exception(ErrorCodes::NOT_ENOUGH_SPACE, "Not enough space on temporary disk"); + disk = reservation->getDisk(); + } + else + { + disk = volume->getDisk(); + } + chassert(disk); + } + + std::unique_ptr write() override + { + return disk->writeFile(path_to_file); + } + + std::unique_ptr read(size_t buffer_size) const override + { + ReadSettings settings; + settings.local_fs_buffer_size = buffer_size; + settings.remote_fs_buffer_size = buffer_size; + settings.prefetch_buffer_size = buffer_size; + + return disk->readFile(path_to_file, settings); + } + + String describeFilePath() const override + { + return fmt::format("disk({})://{}/{}", disk->getName(), disk->getPath(), path_to_file); + } + + ~TemporaryFileOnLocalDisk() override + try + { + if (disk->exists(path_to_file)) + disk->removeRecursive(path_to_file); + else + LOG_WARNING(getLogger("TemporaryFileOnLocalDisk"), "Temporary path '{}' does not exist in '{}' on disk {}", path_to_file, disk->getPath(), disk->getName()); + } + catch (...) + { + tryLogCurrentException(__PRETTY_FUNCTION__); + } + +private: + DiskPtr disk; + String path_to_file; +}; + +TemporaryFileProvider createTemporaryFileProvider(VolumePtr volume) +{ + if (!volume) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Volume is not initialized"); + return [volume](size_t max_size) -> std::unique_ptr + { + return std::make_unique(volume, max_size); + }; +} + +TemporaryFileProvider createTemporaryFileProvider(FileCache * file_cache) +{ + if (!file_cache || !file_cache->isInitialized()) + throw Exception(ErrorCodes::LOGICAL_ERROR, "File cache is not initialized"); + return [file_cache](size_t max_size) -> std::unique_ptr + { + return std::make_unique(*file_cache, max_size); + }; +} + +TemporaryDataOnDiskScopePtr TemporaryDataOnDiskScope::childScope(CurrentMetrics::Metric current_metric) +{ + TemporaryDataOnDiskSettings child_settings = settings; + child_settings.current_metric = current_metric; + return std::make_shared(shared_from_this(), child_settings); +} + +TemporaryDataReadBuffer::TemporaryDataReadBuffer(std::unique_ptr in_) + : ReadBuffer(nullptr, 0) + , compressed_buf(std::move(in_)) +{ + BufferBase::set(compressed_buf->buffer().begin(), compressed_buf->buffer().size(), compressed_buf->offset()); +} + +bool TemporaryDataReadBuffer::nextImpl() +{ + compressed_buf->position() = position(); + if (!compressed_buf->next()) + { + set(compressed_buf->position(), 0); + return false; + } + BufferBase::set(compressed_buf->buffer().begin(), compressed_buf->buffer().size(), compressed_buf->offset()); + return true; +} + +TemporaryDataBuffer::TemporaryDataBuffer(TemporaryDataOnDiskScope * parent_, size_t max_file_size) + : WriteBuffer(nullptr, 0) + , parent(parent_) + , file_holder(parent->file_provider(max_file_size == 0 ? parent->getSettings().max_size_on_disk : max_file_size)) + , out_compressed_buf(file_holder->write(), getCodec(parent->getSettings())) +{ + WriteBuffer::set(out_compressed_buf->buffer().begin(), out_compressed_buf->buffer().size()); +} + +void TemporaryDataBuffer::nextImpl() +{ + if (!out_compressed_buf) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file buffer writing has been finished"); + + out_compressed_buf->position() = position(); + out_compressed_buf->next(); + BufferBase::set(out_compressed_buf->buffer().begin(), out_compressed_buf->buffer().size(), out_compressed_buf->offset()); + updateAllocAndCheck(); +} + +String TemporaryDataBuffer::describeFilePath() const +{ + return file_holder->describeFilePath(); +} + +TemporaryDataBuffer::~TemporaryDataBuffer() +{ + if (out_compressed_buf) + // read() nor finishWriting() was called + cancel(); +} + +void TemporaryDataBuffer::cancelImpl() noexcept +{ + if (out_compressed_buf) + { + /// CompressedWriteBuffer doesn't call cancel/finalize for wrapped buffer + out_compressed_buf->cancel(); + out_compressed_buf.getHolder()->cancel(); + out_compressed_buf.reset(); + } +} + +void TemporaryDataBuffer::finalizeImpl() +{ + if (!out_compressed_buf) + return; + + /// CompressedWriteBuffer doesn't call cancel/finalize for wrapped buffer + out_compressed_buf->finalize(); + out_compressed_buf.getHolder()->finalize(); + + updateAllocAndCheck(); + out_compressed_buf.reset(); +} + +TemporaryDataBuffer::Stat TemporaryDataBuffer::finishWriting() +{ + /// TemporaryDataBuffer::read can be called from multiple threads + std::call_once(write_finished, [this] + { + if (canceled) + throw Exception(ErrorCodes::INVALID_STATE, "Writing to temporary file buffer was not successful"); + next(); + finalize(); + }); + return stat; +} + +std::unique_ptr TemporaryDataBuffer::read() +{ + finishWriting(); + + /// Keep buffer size less that file size, to avoid memory overhead for large amounts of small files + size_t buffer_size = std::min(stat.compressed_size, DBMS_DEFAULT_BUFFER_SIZE); + return std::make_unique(file_holder->read(buffer_size)); +} + +void TemporaryDataBuffer::updateAllocAndCheck() +{ + if (!out_compressed_buf) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file buffer writing has been finished"); + + size_t new_compressed_size = out_compressed_buf->getCompressedBytes(); + size_t new_uncompressed_size = out_compressed_buf->getUncompressedBytes(); + + if (unlikely(new_compressed_size < stat.compressed_size || new_uncompressed_size < stat.uncompressed_size)) + { + throw Exception(ErrorCodes::LOGICAL_ERROR, + "Temporary file {} size decreased after write: compressed: {} -> {}, uncompressed: {} -> {}", + file_holder ? file_holder->describeFilePath() : "NULL", + new_compressed_size, stat.compressed_size, new_uncompressed_size, stat.uncompressed_size); + } + + parent->deltaAllocAndCheck(new_compressed_size - stat.compressed_size, new_uncompressed_size - stat.uncompressed_size); + stat.compressed_size = new_compressed_size; + stat.uncompressed_size = new_uncompressed_size; +} void TemporaryDataOnDiskScope::deltaAllocAndCheck(ssize_t compressed_delta, ssize_t uncompressed_delta) { @@ -54,391 +309,25 @@ void TemporaryDataOnDiskScope::deltaAllocAndCheck(ssize_t compressed_delta, ssiz stat.uncompressed_size += uncompressed_delta; } -TemporaryDataOnDisk::TemporaryDataOnDisk(TemporaryDataOnDiskScopePtr parent_) - : TemporaryDataOnDiskScope(parent_, parent_->getSettings()) +TemporaryBlockStreamHolder::TemporaryBlockStreamHolder(const Block & header_, TemporaryDataOnDiskScope * parent_, size_t max_file_size) + : WrapperGuard(std::make_unique(parent_, max_file_size), DBMS_TCP_PROTOCOL_VERSION, header_) + , header(header_) {} -TemporaryDataOnDisk::TemporaryDataOnDisk(TemporaryDataOnDiskScopePtr parent_, CurrentMetrics::Metric metric_scope) - : TemporaryDataOnDiskScope(parent_, parent_->getSettings()) - , current_metric_scope(metric_scope) -{} - -std::unique_ptr TemporaryDataOnDisk::createRawStream(size_t max_file_size) +TemporaryDataBuffer::Stat TemporaryBlockStreamHolder::finishWriting() const { - if (file_cache && file_cache->isInitialized()) - { - auto holder = createCacheFile(max_file_size); - return std::make_unique(std::move(holder)); - } - if (volume) - { - auto tmp_file = createRegularFile(max_file_size); - return std::make_unique(std::move(tmp_file)); - } + if (!holder) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary block stream is not initialized"); - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryDataOnDiskScope has no cache and no volume"); + impl->flush(); + return holder->finishWriting(); } -TemporaryFileStream & TemporaryDataOnDisk::createStream(const Block & header, size_t max_file_size) +TemporaryBlockStreamReaderHolder TemporaryBlockStreamHolder::getReadStream() const { - if (file_cache && file_cache->isInitialized()) - { - auto holder = createCacheFile(max_file_size); - - std::lock_guard lock(mutex); - TemporaryFileStreamPtr & tmp_stream = streams.emplace_back(std::make_unique(std::move(holder), header, this)); - return *tmp_stream; - } - if (volume) - { - auto tmp_file = createRegularFile(max_file_size); - std::lock_guard lock(mutex); - TemporaryFileStreamPtr & tmp_stream - = streams.emplace_back(std::make_unique(std::move(tmp_file), header, this)); - return *tmp_stream; - } - - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryDataOnDiskScope has no cache and no volume"); -} - -FileSegmentsHolderPtr TemporaryDataOnDisk::createCacheFile(size_t max_file_size) -{ - if (!file_cache) - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryDataOnDiskScope has no cache"); - - ProfileEvents::increment(ProfileEvents::ExternalProcessingFilesTotal); - - const auto key = FileSegment::Key::random(); - auto holder = file_cache->set( - key, 0, std::max(10_MiB, max_file_size), - CreateFileSegmentSettings(FileSegmentKind::Ephemeral), FileCache::getCommonUser()); - - chassert(holder->size() == 1); - holder->back().getKeyMetadata()->createBaseDirectory(/* throw_if_failed */true); - - return holder; -} - -TemporaryFileOnDiskHolder TemporaryDataOnDisk::createRegularFile(size_t max_file_size) -{ - if (!volume) - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryDataOnDiskScope has no volume"); - - DiskPtr disk; - if (max_file_size > 0) - { - auto reservation = volume->reserve(max_file_size); - if (!reservation) - throw Exception(ErrorCodes::NOT_ENOUGH_SPACE, "Not enough space on temporary disk"); - disk = reservation->getDisk(); - } - else - { - disk = volume->getDisk(); - } - /// We do not increment ProfileEvents::ExternalProcessingFilesTotal here because it is incremented in TemporaryFileOnDisk constructor. - return std::make_unique(disk, current_metric_scope); -} - -std::vector TemporaryDataOnDisk::getStreams() const -{ - std::vector res; - std::lock_guard lock(mutex); - res.reserve(streams.size()); - for (const auto & stream : streams) - res.push_back(stream.get()); - return res; -} - -bool TemporaryDataOnDisk::empty() const -{ - std::lock_guard lock(mutex); - return streams.empty(); -} - -static inline CompressionCodecPtr getCodec(const TemporaryDataOnDiskSettings & settings) -{ - if (settings.compression_codec.empty()) - return CompressionCodecFactory::instance().get("NONE"); - - return CompressionCodecFactory::instance().get(settings.compression_codec); -} - -struct TemporaryFileStream::OutputWriter -{ - OutputWriter(std::unique_ptr out_buf_, const Block & header_, const TemporaryDataOnDiskSettings & settings) - : out_buf(std::move(out_buf_)) - , out_compressed_buf(*out_buf, getCodec(settings)) - , out_writer(out_compressed_buf, DBMS_TCP_PROTOCOL_VERSION, header_) - { - } - - size_t write(const Block & block) - { - if (finalized) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot write to finalized stream"); - size_t written_bytes = out_writer.write(block); - num_rows += block.rows(); - return written_bytes; - } - - void flush() - { - if (finalized) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot flush finalized stream"); - - out_compressed_buf.next(); - out_buf->next(); - out_writer.flush(); - } - - void finalize() - { - if (finalized) - return; - - /// if we called finalize() explicitly, and got an exception, - /// we don't want to get it again in the destructor, so set finalized flag first - finalized = true; - - out_writer.flush(); - out_compressed_buf.finalize(); - out_buf->finalize(); - } - - ~OutputWriter() - { - try - { - finalize(); - } - catch (...) - { - tryLogCurrentException(__PRETTY_FUNCTION__); - } - } - - std::unique_ptr out_buf; - CompressedWriteBuffer out_compressed_buf; - NativeWriter out_writer; - - std::atomic_size_t num_rows = 0; - - bool finalized = false; -}; - -TemporaryFileStream::Reader::Reader(const String & path_, const Block & header_, size_t size_) - : path(path_) - , size(size_ ? std::min(size_, DBMS_DEFAULT_BUFFER_SIZE) : DBMS_DEFAULT_BUFFER_SIZE) - , header(header_) -{ - LOG_TEST(getLogger("TemporaryFileStream"), "Reading {} from {}", header_.dumpStructure(), path); -} - -TemporaryFileStream::Reader::Reader(const String & path_, size_t size_) - : path(path_) - , size(size_ ? std::min(size_, DBMS_DEFAULT_BUFFER_SIZE) : DBMS_DEFAULT_BUFFER_SIZE) -{ - LOG_TEST(getLogger("TemporaryFileStream"), "Reading from {}", path); -} - -Block TemporaryFileStream::Reader::read() -{ - if (!in_reader) - { - if (fs::exists(path)) - in_file_buf = std::make_unique(path, size); - else - in_file_buf = std::make_unique(); - - in_compressed_buf = std::make_unique(*in_file_buf); - if (header.has_value()) - in_reader = std::make_unique(*in_compressed_buf, header.value(), DBMS_TCP_PROTOCOL_VERSION); - else - in_reader = std::make_unique(*in_compressed_buf, DBMS_TCP_PROTOCOL_VERSION); - } - return in_reader->read(); -} - -TemporaryFileStream::TemporaryFileStream(TemporaryFileOnDiskHolder file_, const Block & header_, TemporaryDataOnDisk * parent_) - : parent(parent_) - , header(header_) - , file(std::move(file_)) - , out_writer(std::make_unique(std::make_unique(file->getAbsolutePath()), header, parent->settings)) -{ - LOG_TEST(getLogger("TemporaryFileStream"), "Writing to temporary file {}", file->getAbsolutePath()); -} - -TemporaryFileStream::TemporaryFileStream(FileSegmentsHolderPtr segments_, const Block & header_, TemporaryDataOnDisk * parent_) - : parent(parent_) - , header(header_) - , segment_holder(std::move(segments_)) -{ - if (segment_holder->size() != 1) - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryFileStream can be created only from single segment"); - auto out_buf = std::make_unique(&segment_holder->front()); - - LOG_TEST(getLogger("TemporaryFileStream"), "Writing to temporary file {}", out_buf->getFileName()); - out_writer = std::make_unique(std::move(out_buf), header, parent_->settings); -} - -size_t TemporaryFileStream::write(const Block & block) -{ - if (!out_writer) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Writing has been finished"); - - updateAllocAndCheck(); - size_t bytes_written = out_writer->write(block); - return bytes_written; -} - -void TemporaryFileStream::flush() -{ - if (!out_writer) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Writing has been finished"); - - out_writer->flush(); -} - -TemporaryFileStream::Stat TemporaryFileStream::finishWriting() -{ - if (isWriteFinished()) - return stat; - - if (out_writer) - { - out_writer->finalize(); - /// The amount of written data can be changed after finalization, some buffers can be flushed - /// Need to update the stat - updateAllocAndCheck(); - out_writer.reset(); - - /// reader will be created at the first read call, not to consume memory before it is needed - } - return stat; -} - -TemporaryFileStream::Stat TemporaryFileStream::finishWritingAsyncSafe() -{ - std::call_once(finish_writing, [this]{ finishWriting(); }); - return stat; -} - -bool TemporaryFileStream::isWriteFinished() const -{ - assert(in_reader == nullptr || out_writer == nullptr); - return out_writer == nullptr; -} - -Block TemporaryFileStream::read() -{ - if (!isWriteFinished()) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Writing has been not finished"); - - if (isEof()) - return {}; - - if (!in_reader) - { - in_reader = std::make_unique(getPath(), header, getSize()); - } - - Block block = in_reader->read(); - if (!block) - { - /// finalize earlier to release resources, do not wait for the destructor - this->release(); - } - return block; -} - -std::unique_ptr TemporaryFileStream::getReadStream() -{ - if (!isWriteFinished()) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Writing has been not finished"); - - if (isEof()) - return nullptr; - - return std::make_unique(getPath(), header, getSize()); -} - -void TemporaryFileStream::updateAllocAndCheck() -{ - assert(out_writer); - size_t new_compressed_size = out_writer->out_compressed_buf.getCompressedBytes(); - size_t new_uncompressed_size = out_writer->out_compressed_buf.getUncompressedBytes(); - - if (unlikely(new_compressed_size < stat.compressed_size || new_uncompressed_size < stat.uncompressed_size)) - { - throw Exception(ErrorCodes::LOGICAL_ERROR, - "Temporary file {} size decreased after write: compressed: {} -> {}, uncompressed: {} -> {}", - getPath(), new_compressed_size, stat.compressed_size, new_uncompressed_size, stat.uncompressed_size); - } - - parent->deltaAllocAndCheck(new_compressed_size - stat.compressed_size, new_uncompressed_size - stat.uncompressed_size); - stat.compressed_size = new_compressed_size; - stat.uncompressed_size = new_uncompressed_size; - stat.num_rows = out_writer->num_rows; -} - -bool TemporaryFileStream::isEof() const -{ - return file == nullptr && !segment_holder; -} - -void TemporaryFileStream::release() -{ - if (in_reader) - in_reader.reset(); - - if (out_writer) - { - out_writer->finalize(); - out_writer.reset(); - } - - if (file) - { - file.reset(); - parent->deltaAllocAndCheck(-stat.compressed_size, -stat.uncompressed_size); - } - - if (segment_holder) - segment_holder.reset(); -} - -String TemporaryFileStream::getPath() const -{ - if (file) - return file->getAbsolutePath(); - if (segment_holder && !segment_holder->empty()) - return segment_holder->front().getPath(); - - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryFileStream has no file"); -} - -size_t TemporaryFileStream::getSize() const -{ - if (file) - return file->getDisk()->getFileSize(file->getRelativePath()); - if (segment_holder && !segment_holder->empty()) - return segment_holder->front().getReservedSize(); - - throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryFileStream has no file"); -} - -TemporaryFileStream::~TemporaryFileStream() -{ - try - { - release(); - } - catch (...) - { - tryLogCurrentException(__PRETTY_FUNCTION__); - assert(false); /// deltaAllocAndCheck with negative can't throw exception - } + if (!holder) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary block stream is not initialized"); + return TemporaryBlockStreamReaderHolder(holder->read(), header, DBMS_TCP_PROTOCOL_VERSION); } } diff --git a/src/Interpreters/TemporaryDataOnDisk.h b/src/Interpreters/TemporaryDataOnDisk.h index d541c93e031..f8d14b00ac5 100644 --- a/src/Interpreters/TemporaryDataOnDisk.h +++ b/src/Interpreters/TemporaryDataOnDisk.h @@ -4,15 +4,21 @@ #include #include -#include +#include #include -#include -#include +#include + #include #include -#include -#include +#include +#include + +#include + +#include + +class FileCacheTest_TemporaryDataReadBufferSize_Test; namespace CurrentMetrics { @@ -25,11 +31,10 @@ namespace DB class TemporaryDataOnDiskScope; using TemporaryDataOnDiskScopePtr = std::shared_ptr; -class TemporaryDataOnDisk; -using TemporaryDataOnDiskPtr = std::unique_ptr; +class TemporaryDataBuffer; +using TemporaryDataBufferPtr = std::unique_ptr; -class TemporaryFileStream; -using TemporaryFileStreamPtr = std::unique_ptr; +class TemporaryFileHolder; class FileCache; @@ -40,15 +45,26 @@ struct TemporaryDataOnDiskSettings /// Compression codec for temporary data, if empty no compression will be used. LZ4 by default String compression_codec = "LZ4"; + + /// Read/Write internal buffer size + size_t buffer_size = DBMS_DEFAULT_BUFFER_SIZE; + + /// Metrics counter to increment when temporary file in current scope are created + CurrentMetrics::Metric current_metric = CurrentMetrics::TemporaryFilesUnknown; }; +/// Creates temporary files located on specified resource (disk, fs_cache, etc.) +using TemporaryFileProvider = std::function(size_t)>; +TemporaryFileProvider createTemporaryFileProvider(VolumePtr volume); +TemporaryFileProvider createTemporaryFileProvider(FileCache * file_cache); + /* * Used to account amount of temporary data written to disk. * If limit is set, throws exception if limit is exceeded. * Data can be nested, so parent scope accounts all data written by children. * Scopes are: global -> per-user -> per-query -> per-purpose (sorting, aggregation, etc). */ -class TemporaryDataOnDiskScope : boost::noncopyable +class TemporaryDataOnDiskScope : boost::noncopyable, public std::enable_shared_from_this { public: struct StatAtomic @@ -57,164 +73,155 @@ public: std::atomic uncompressed_size; }; - explicit TemporaryDataOnDiskScope(VolumePtr volume_, TemporaryDataOnDiskSettings settings_) - : volume(std::move(volume_)) + /// Root scope + template + TemporaryDataOnDiskScope(T && storage, TemporaryDataOnDiskSettings settings_) + : file_provider(createTemporaryFileProvider(std::forward(storage))) , settings(std::move(settings_)) {} - explicit TemporaryDataOnDiskScope(VolumePtr volume_, FileCache * file_cache_, TemporaryDataOnDiskSettings settings_) - : volume(std::move(volume_)) - , file_cache(file_cache_) - , settings(std::move(settings_)) - {} - explicit TemporaryDataOnDiskScope(TemporaryDataOnDiskScopePtr parent_, TemporaryDataOnDiskSettings settings_) + TemporaryDataOnDiskScope(TemporaryDataOnDiskScopePtr parent_, TemporaryDataOnDiskSettings settings_) : parent(std::move(parent_)) - , volume(parent->volume) - , file_cache(parent->file_cache) + , file_provider(parent->file_provider) , settings(std::move(settings_)) {} - /// TODO: remove - /// Refactor all code that uses volume directly to use TemporaryDataOnDisk. - VolumePtr getVolume() const { return volume; } + TemporaryDataOnDiskScopePtr childScope(CurrentMetrics::Metric current_metric); const TemporaryDataOnDiskSettings & getSettings() const { return settings; } - protected: + friend class TemporaryDataBuffer; + void deltaAllocAndCheck(ssize_t compressed_delta, ssize_t uncompressed_delta); TemporaryDataOnDiskScopePtr parent = nullptr; - VolumePtr volume = nullptr; - FileCache * file_cache = nullptr; + TemporaryFileProvider file_provider; StatAtomic stat; const TemporaryDataOnDiskSettings settings; }; -/* - * Holds the set of temporary files. - * New file stream is created with `createStream`. - * Streams are owned by this object and will be deleted when it is deleted. - * It's a leaf node in temporary data scope tree. - */ -class TemporaryDataOnDisk : private TemporaryDataOnDiskScope +/** Used to hold the wrapper and wrapped object together. + * This class provides a convenient way to manage the lifetime of both the wrapper and the wrapped object. + * The wrapper class (Impl) stores a reference to the wrapped object (Holder), and both objects are owned by this class. + * The lifetime of the wrapper and the wrapped object should be the same. + * This pattern is commonly used when the caller only needs to interact with the wrapper and doesn't need to be aware of the wrapped object. + * Examples: CompressedWriteBuffer and WriteBuffer, and NativeReader and ReadBuffer. + */ +template +class WrapperGuard { - friend class TemporaryFileStream; /// to allow it to call `deltaAllocAndCheck` to account data - public: - using TemporaryDataOnDiskScope::StatAtomic; + WrapperGuard() = default; - explicit TemporaryDataOnDisk(TemporaryDataOnDiskScopePtr parent_); + template + WrapperGuard(std::unique_ptr holder_, Args && ... args) + : holder(std::move(holder_)) + , impl(std::make_unique(*holder, std::forward(args)...)) + {} - explicit TemporaryDataOnDisk(TemporaryDataOnDiskScopePtr parent_, CurrentMetrics::Metric metric_scope); + Impl * operator->() { return impl.get(); } + const Impl * operator->() const { return impl.get(); } + Impl & operator*() { return *impl; } + const Impl & operator*() const { return *impl; } + operator bool() const { return impl != nullptr; } - /// If max_file_size > 0, then check that there's enough space on the disk and throw an exception in case of lack of free space - TemporaryFileStream & createStream(const Block & header, size_t max_file_size = 0); + const Holder * getHolder() const { return holder.get(); } + Holder * getHolder() { return holder.get(); } - /// Write raw data directly into buffer. - /// Differences from `createStream`: - /// 1) it doesn't account data in parent scope - /// 2) returned buffer owns resources (instead of TemporaryDataOnDisk itself) - /// If max_file_size > 0, then check that there's enough space on the disk and throw an exception in case of lack of free space - std::unique_ptr createRawStream(size_t max_file_size = 0); + void reset() + { + impl.reset(); + holder.reset(); + } - std::vector getStreams() const; - bool empty() const; - - const StatAtomic & getStat() const { return stat; } - -private: - FileSegmentsHolderPtr createCacheFile(size_t max_file_size); - TemporaryFileOnDiskHolder createRegularFile(size_t max_file_size); - - mutable std::mutex mutex; - std::vector streams TSA_GUARDED_BY(mutex); - - typename CurrentMetrics::Metric current_metric_scope = CurrentMetrics::TemporaryFilesUnknown; +protected: + std::unique_ptr holder; + std::unique_ptr impl; }; -/* - * Data can be written into this stream and then read. - * After finish writing, call `finishWriting` and then either call `read` or 'getReadStream'(only one of the two) to read the data. - * Account amount of data written to disk in parent scope. - */ -class TemporaryFileStream : boost::noncopyable +/// Owns temporary file and provides access to it. +/// On destruction, file is removed and all resources are freed. +/// Lifetime of read/write buffers should be less than lifetime of TemporaryFileHolder. +class TemporaryFileHolder { public: - struct Reader - { - Reader(const String & path, const Block & header_, size_t size = 0); + TemporaryFileHolder(); - explicit Reader(const String & path, size_t size = 0); + virtual std::unique_ptr write() = 0; + virtual std::unique_ptr read(size_t buffer_size) const = 0; - Block read(); + /// Get location for logging purposes + virtual String describeFilePath() const = 0; - const std::string path; - const size_t size; - const std::optional header; + virtual ~TemporaryFileHolder() = default; +}; - std::unique_ptr in_file_buf; - std::unique_ptr in_compressed_buf; - std::unique_ptr in_reader; - }; +class TemporaryDataReadBuffer : public ReadBuffer +{ +public: + explicit TemporaryDataReadBuffer(std::unique_ptr in_); + +private: + friend class ::FileCacheTest_TemporaryDataReadBufferSize_Test; + + bool nextImpl() override; + + WrapperGuard compressed_buf; +}; + +/// Writes data to buffer provided by file_holder, and accounts amount of written data in parent scope. +class TemporaryDataBuffer : public WriteBuffer +{ +public: struct Stat { - /// Statistics for file - /// Non-atomic because we don't allow to `read` or `write` into single file from multiple threads size_t compressed_size = 0; size_t uncompressed_size = 0; - size_t num_rows = 0; }; - TemporaryFileStream(TemporaryFileOnDiskHolder file_, const Block & header_, TemporaryDataOnDisk * parent_); - TemporaryFileStream(FileSegmentsHolderPtr segments_, const Block & header_, TemporaryDataOnDisk * parent_); - - size_t write(const Block & block); - void flush(); + explicit TemporaryDataBuffer(TemporaryDataOnDiskScope * parent_, size_t max_file_size = 0); + void nextImpl() override; + void finalizeImpl() override; + void cancelImpl() noexcept override; + std::unique_ptr read(); Stat finishWriting(); - Stat finishWritingAsyncSafe(); - bool isWriteFinished() const; - std::unique_ptr getReadStream(); + String describeFilePath() const; - Block read(); - - String getPath() const; - size_t getSize() const; - - Block getHeader() const { return header; } - - /// Read finished and file released - bool isEof() const; - - ~TemporaryFileStream(); + ~TemporaryDataBuffer() override; private: void updateAllocAndCheck(); - /// Release everything, close reader and writer, delete file - void release(); - - TemporaryDataOnDisk * parent; - - Block header; - - /// Data can be stored in file directly or in the cache - TemporaryFileOnDiskHolder file; - FileSegmentsHolderPtr segment_holder; + TemporaryDataOnDiskScope * parent; + std::unique_ptr file_holder; + WrapperGuard out_compressed_buf; + std::once_flag write_finished; Stat stat; +}; - std::once_flag finish_writing; +using TemporaryBlockStreamReaderHolder = WrapperGuard; - struct OutputWriter; - std::unique_ptr out_writer; +class TemporaryBlockStreamHolder : public WrapperGuard +{ +public: + TemporaryBlockStreamHolder() = default; - std::unique_ptr in_reader; + TemporaryBlockStreamHolder(const Block & header_, TemporaryDataOnDiskScope * parent_, size_t max_file_size = 0); + + TemporaryBlockStreamReaderHolder getReadStream() const; + + TemporaryDataBuffer::Stat finishWriting() const; + const Block & getHeader() const { return header; } + +private: + Block header; }; } diff --git a/src/Interpreters/tests/gtest_filecache.cpp b/src/Interpreters/tests/gtest_filecache.cpp index 007b31d9fdc..ae45443d4bd 100644 --- a/src/Interpreters/tests/gtest_filecache.cpp +++ b/src/Interpreters/tests/gtest_filecache.cpp @@ -934,7 +934,7 @@ static Block generateBlock(size_t size = 0) return block; } -static size_t readAllTemporaryData(TemporaryFileStream & stream) +static size_t readAllTemporaryData(NativeReader & stream) { Block block; size_t read_rows = 0; @@ -947,6 +947,7 @@ static size_t readAllTemporaryData(TemporaryFileStream & stream) } TEST_F(FileCacheTest, temporaryData) +try { ServerUUID::setRandomForUnitTests(); DB::FileCacheSettings settings; @@ -959,7 +960,7 @@ TEST_F(FileCacheTest, temporaryData) file_cache.initialize(); const auto user = FileCache::getCommonUser(); - auto tmp_data_scope = std::make_shared(nullptr, &file_cache, TemporaryDataOnDiskSettings{}); + auto tmp_data_scope = std::make_shared(&file_cache, TemporaryDataOnDiskSettings{}); auto some_data_holder = file_cache.getOrSet(FileCacheKey::fromPath("some_data"), 0, 5_KiB, 5_KiB, CreateFileSegmentSettings{}, 0, user); @@ -982,12 +983,17 @@ TEST_F(FileCacheTest, temporaryData) size_t size_used_with_temporary_data; size_t segments_used_with_temporary_data; + + { - auto tmp_data = std::make_unique(tmp_data_scope); + TemporaryBlockStreamHolder stream(generateBlock(), tmp_data_scope.get()); + ASSERT_TRUE(stream); + /// Do nothitng with stream, just create it and destroy. + } - auto & stream = tmp_data->createStream(generateBlock()); - - ASSERT_GT(stream.write(generateBlock(100)), 0); + { + TemporaryBlockStreamHolder stream(generateBlock(), tmp_data_scope.get()); + ASSERT_GT(stream->write(generateBlock(100)), 0); ASSERT_GT(file_cache.getUsedCacheSize(), 0); ASSERT_GT(file_cache.getFileSegmentsNum(), 0); @@ -995,22 +1001,22 @@ TEST_F(FileCacheTest, temporaryData) size_t used_size_before_attempt = file_cache.getUsedCacheSize(); /// data can't be evicted because it is still held by `some_data_holder` ASSERT_THROW({ - stream.write(generateBlock(2000)); - stream.flush(); + stream->write(generateBlock(2000)); + stream.finishWriting(); }, DB::Exception); + ASSERT_THROW(stream.finishWriting(), DB::Exception); + ASSERT_EQ(file_cache.getUsedCacheSize(), used_size_before_attempt); } { size_t before_used_size = file_cache.getUsedCacheSize(); - auto tmp_data = std::make_unique(tmp_data_scope); - - auto write_buf_stream = tmp_data->createRawStream(); + auto write_buf_stream = std::make_unique(tmp_data_scope.get()); write_buf_stream->write("1234567890", 10); write_buf_stream->write("abcde", 5); - auto read_buf = dynamic_cast(write_buf_stream.get())->tryGetReadBuffer(); + auto read_buf = write_buf_stream->read(); ASSERT_GT(file_cache.getUsedCacheSize(), before_used_size + 10); @@ -1023,22 +1029,22 @@ TEST_F(FileCacheTest, temporaryData) } { - auto tmp_data = std::make_unique(tmp_data_scope); - auto & stream = tmp_data->createStream(generateBlock()); + TemporaryBlockStreamHolder stream(generateBlock(), tmp_data_scope.get()); - ASSERT_GT(stream.write(generateBlock(100)), 0); + ASSERT_GT(stream->write(generateBlock(100)), 0); some_data_holder.reset(); - stream.write(generateBlock(2000)); + stream->write(generateBlock(2000)); - auto stat = stream.finishWriting(); + stream.finishWriting(); - ASSERT_TRUE(fs::exists(stream.getPath())); - ASSERT_GT(fs::file_size(stream.getPath()), 100); + String file_path = stream.getHolder()->describeFilePath().substr(strlen("fscache://")); - ASSERT_EQ(stat.num_rows, 2100); - ASSERT_EQ(readAllTemporaryData(stream), 2100); + ASSERT_TRUE(fs::exists(file_path)) << "File " << file_path << " should exist"; + ASSERT_GT(fs::file_size(file_path), 100) << "File " << file_path << " should be larger than 100 bytes"; + + ASSERT_EQ(readAllTemporaryData(*stream.getReadStream()), 2100); size_used_with_temporary_data = file_cache.getUsedCacheSize(); segments_used_with_temporary_data = file_cache.getFileSegmentsNum(); @@ -1054,6 +1060,11 @@ TEST_F(FileCacheTest, temporaryData) ASSERT_LE(file_cache.getUsedCacheSize(), size_used_before_temporary_data); ASSERT_LE(file_cache.getFileSegmentsNum(), segments_used_before_temporary_data); } +catch (...) +{ + std::cerr << getCurrentExceptionMessage(true) << std::endl; + throw; +} TEST_F(FileCacheTest, CachedReadBuffer) { @@ -1148,18 +1159,22 @@ TEST_F(FileCacheTest, TemporaryDataReadBufferSize) DB::FileCache file_cache("cache", settings); file_cache.initialize(); - auto tmp_data_scope = std::make_shared(/*volume=*/nullptr, &file_cache, /*settings=*/TemporaryDataOnDiskSettings{}); - - auto tmp_data = std::make_unique(tmp_data_scope); + auto tmp_data_scope = std::make_shared(&file_cache, TemporaryDataOnDiskSettings{}); auto block = generateBlock(/*size=*/3); - auto & stream = tmp_data->createStream(block); - stream.write(block); - stream.finishWriting(); + TemporaryBlockStreamHolder stream(block, tmp_data_scope.get()); - /// We allocate buffer of size min(getSize(), DBMS_DEFAULT_BUFFER_SIZE) + stream->write(block); + auto stat = stream.finishWriting(); + + /// We allocate buffer of size min(stat.compressed_size, DBMS_DEFAULT_BUFFER_SIZE) /// We do care about buffer size because realistic external group by could generate 10^5 temporary files - ASSERT_EQ(stream.getSize(), 62); + ASSERT_EQ(stat.compressed_size, 62); + + auto reader = stream.getReadStream(); + auto * read_buf = reader.getHolder(); + const auto & internal_buffer = static_cast(read_buf)->compressed_buf.getHolder()->internalBuffer(); + ASSERT_EQ(internal_buffer.size(), 62); } /// Temporary data stored on disk @@ -1170,16 +1185,14 @@ TEST_F(FileCacheTest, TemporaryDataReadBufferSize) disk = createDisk("temporary_data_read_buffer_size_test_dir"); VolumePtr volume = std::make_shared("volume", disk); - auto tmp_data_scope = std::make_shared(/*volume=*/volume, /*cache=*/nullptr, /*settings=*/TemporaryDataOnDiskSettings{}); - - auto tmp_data = std::make_unique(tmp_data_scope); + auto tmp_data_scope = std::make_shared(volume, TemporaryDataOnDiskSettings{}); auto block = generateBlock(/*size=*/3); - auto & stream = tmp_data->createStream(block); - stream.write(block); - stream.finishWriting(); + TemporaryBlockStreamHolder stream(block, tmp_data_scope.get()); + stream->write(block); + auto stat = stream.finishWriting(); - ASSERT_EQ(stream.getSize(), 62); + ASSERT_EQ(stat.compressed_size, 62); } } diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp index 07ee8f4ddef..1560e88ffef 100644 --- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp @@ -5,6 +5,7 @@ #include #include #include +#include #include @@ -29,17 +30,18 @@ CollapsingSortedAlgorithm::CollapsingSortedAlgorithm( size_t max_block_size_rows_, size_t max_block_size_bytes_, LoggerPtr log_, - WriteBuffer * out_row_sources_buf_, + std::shared_ptr temp_data_buffer_, bool use_average_block_sizes) : IMergingAlgorithmWithSharedChunks( header_, num_inputs, std::move(description_), - out_row_sources_buf_, + temp_data_buffer_.get(), max_row_refs, std::make_unique(use_average_block_sizes, max_block_size_rows_, max_block_size_bytes_)) , sign_column_number(header_.getPositionByName(sign_column)) , only_positive_sign(only_positive_sign_) + , temp_data_buffer(temp_data_buffer_) , log(log_) { } diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h index 99fd95d82d9..b7bb9914cf8 100644 --- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h +++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h @@ -11,6 +11,8 @@ namespace Poco namespace DB { +class TemporaryDataBuffer; + /** Merges several sorted inputs to one. * For each group of consecutive identical values of the primary key (the columns by which the data is sorted), * keeps no more than one row with the value of the column `sign_column = -1` ("negative row") @@ -35,7 +37,7 @@ public: size_t max_block_size_rows_, size_t max_block_size_bytes_, LoggerPtr log_, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr temp_data_buffer_ = nullptr, bool use_average_block_sizes = false); const char * getName() const override { return "CollapsingSortedAlgorithm"; } @@ -62,6 +64,8 @@ private: PODArray current_row_sources; /// Sources of rows with the current primary key size_t count_incorrect_data = 0; /// To prevent too many error messages from writing to the log. + std::shared_ptr temp_data_buffer = nullptr; + LoggerPtr log; void reportIncorrectData(); diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp index 3a9cf7ee141..d4e4ba6aa5f 100644 --- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp @@ -3,6 +3,7 @@ #include #include #include +#include namespace DB { @@ -15,7 +16,7 @@ MergingSortedAlgorithm::MergingSortedAlgorithm( size_t max_block_size_bytes_, SortingQueueStrategy sorting_queue_strategy_, UInt64 limit_, - WriteBuffer * out_row_sources_buf_, + std::shared_ptr out_row_sources_buf_, bool use_average_block_sizes) : header(std::move(header_)) , merged_data(use_average_block_sizes, max_block_size_, max_block_size_bytes_) diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h index c889668a38e..fc300e41026 100644 --- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h +++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h @@ -9,6 +9,8 @@ namespace DB { +class TemporaryDataBuffer; + /// Merges several sorted inputs into one sorted output. class MergingSortedAlgorithm final : public IMergingAlgorithm { @@ -21,7 +23,7 @@ public: size_t max_block_size_bytes_, SortingQueueStrategy sorting_queue_strategy_, UInt64 limit_ = 0, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false); void addInput(); @@ -45,7 +47,7 @@ private: /// Used in Vertical merge algorithm to gather non-PK/non-index columns (on next step) /// If it is not nullptr then it should be populated during execution - WriteBuffer * out_row_sources_buf = nullptr; + std::shared_ptr out_row_sources_buf = nullptr; /// Chunks currently being merged. Inputs current_inputs; diff --git a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp index cd347d371d9..a3a33080f52 100644 --- a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp @@ -5,6 +5,7 @@ #include #include #include +#include namespace DB { @@ -37,12 +38,13 @@ ReplacingSortedAlgorithm::ReplacingSortedAlgorithm( const String & version_column, size_t max_block_size_rows, size_t max_block_size_bytes, - WriteBuffer * out_row_sources_buf_, + std::shared_ptr temp_data_buffer_, bool use_average_block_sizes, bool cleanup_, bool enable_vertical_final_) - : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), out_row_sources_buf_, max_row_refs, std::make_unique(use_average_block_sizes, max_block_size_rows, max_block_size_bytes)) + : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), temp_data_buffer_.get(), max_row_refs, std::make_unique(use_average_block_sizes, max_block_size_rows, max_block_size_bytes)) , cleanup(cleanup_), enable_vertical_final(enable_vertical_final_) + , temp_data_buffer(temp_data_buffer_) { if (!is_deleted_column.empty()) is_deleted_column_number = header_.getPositionByName(is_deleted_column); diff --git a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h index 2f23f2a5c4d..d3b9837a253 100644 --- a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h +++ b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h @@ -24,6 +24,8 @@ struct ChunkSelectFinalIndices : public ChunkInfoCloneable temp_data_buffer_ = nullptr, bool use_average_block_sizes = false, bool cleanup = false, bool enable_vertical_final_ = false); @@ -59,6 +61,8 @@ private: RowRef selected_row; /// Last row with maximum version for current primary key, may extend lifetime of chunk in input source size_t max_pos = 0; /// The position (into current_row_sources) of the row with the highest version. + std::shared_ptr temp_data_buffer = nullptr; + /// Sources of rows with the current primary key. PODArray current_row_sources; diff --git a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp index 9f124c6ba18..1ceb1f46234 100644 --- a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp @@ -1,6 +1,7 @@ #include #include #include +#include namespace DB { @@ -14,12 +15,13 @@ VersionedCollapsingAlgorithm::VersionedCollapsingAlgorithm( const String & sign_column_, size_t max_block_size_rows_, size_t max_block_size_bytes_, - WriteBuffer * out_row_sources_buf_, + std::shared_ptr temp_data_buffer_, bool use_average_block_sizes) - : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), out_row_sources_buf_, MAX_ROWS_IN_MULTIVERSION_QUEUE, std::make_unique(use_average_block_sizes, max_block_size_rows_, max_block_size_bytes_)) + : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), temp_data_buffer_.get(), MAX_ROWS_IN_MULTIVERSION_QUEUE, std::make_unique(use_average_block_sizes, max_block_size_rows_, max_block_size_bytes_)) /// -1 for +1 in FixedSizeDequeWithGaps's internal buffer. 3 is a reasonable minimum size to collapse anything. , max_rows_in_queue(std::min(std::max(3, max_block_size_rows_), MAX_ROWS_IN_MULTIVERSION_QUEUE) - 1) , current_keys(max_rows_in_queue) + , temp_data_buffer(temp_data_buffer_) { sign_column_number = header_.getPositionByName(sign_column_); } diff --git a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h index e6d20ddac75..6f877459147 100644 --- a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h +++ b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h @@ -8,6 +8,8 @@ namespace DB { +class TemporaryDataBuffer; + /** Merges several sorted inputs to one. * For each group of consecutive identical values of the sorting key * (the columns by which the data is sorted, including specially specified version column), @@ -22,7 +24,7 @@ public: SortDescription description_, const String & sign_column_, size_t max_block_size_rows, size_t max_block_size_bytes, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr temp_data_buffer_ = nullptr, bool use_average_block_sizes = false); const char * getName() const override { return "VersionedCollapsingAlgorithm"; } @@ -37,6 +39,8 @@ private: FixedSizeDequeWithGaps current_keys; Int8 sign_in_queue = 0; + std::shared_ptr temp_data_buffer = nullptr; + std::queue current_row_sources; /// Sources of rows with the current primary key void insertGap(size_t gap_size); diff --git a/src/Processors/Merges/CollapsingSortedTransform.h b/src/Processors/Merges/CollapsingSortedTransform.h index 99fb700abf1..9b09c802783 100644 --- a/src/Processors/Merges/CollapsingSortedTransform.h +++ b/src/Processors/Merges/CollapsingSortedTransform.h @@ -23,7 +23,7 @@ public: bool only_positive_sign, size_t max_block_size_rows, size_t max_block_size_bytes, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false) : IMergingTransform( num_inputs, header, header, /*have_all_inputs_=*/ true, /*limit_hint_=*/ 0, /*always_read_till_end_=*/ false, diff --git a/src/Processors/Merges/MergingSortedTransform.cpp b/src/Processors/Merges/MergingSortedTransform.cpp index d2895a2a2e9..13330dcff6d 100644 --- a/src/Processors/Merges/MergingSortedTransform.cpp +++ b/src/Processors/Merges/MergingSortedTransform.cpp @@ -20,7 +20,7 @@ MergingSortedTransform::MergingSortedTransform( SortingQueueStrategy sorting_queue_strategy, UInt64 limit_, bool always_read_till_end_, - WriteBuffer * out_row_sources_buf_, + std::shared_ptr out_row_sources_buf_, bool use_average_block_sizes, bool have_all_inputs_) : IMergingTransform( diff --git a/src/Processors/Merges/MergingSortedTransform.h b/src/Processors/Merges/MergingSortedTransform.h index 6e52450efa7..fb8e5ce74e3 100644 --- a/src/Processors/Merges/MergingSortedTransform.h +++ b/src/Processors/Merges/MergingSortedTransform.h @@ -20,7 +20,7 @@ public: SortingQueueStrategy sorting_queue_strategy, UInt64 limit_ = 0, bool always_read_till_end_ = false, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false, bool have_all_inputs_ = true); diff --git a/src/Processors/Merges/ReplacingSortedTransform.h b/src/Processors/Merges/ReplacingSortedTransform.h index dc262aab9ee..a9d9f4fb619 100644 --- a/src/Processors/Merges/ReplacingSortedTransform.h +++ b/src/Processors/Merges/ReplacingSortedTransform.h @@ -21,7 +21,7 @@ public: const String & is_deleted_column, const String & version_column, size_t max_block_size_rows, size_t max_block_size_bytes, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr temp_data_buffer_ = nullptr, bool use_average_block_sizes = false, bool cleanup = false, bool enable_vertical_final = false) @@ -34,7 +34,7 @@ public: version_column, max_block_size_rows, max_block_size_bytes, - out_row_sources_buf_, + temp_data_buffer_, use_average_block_sizes, cleanup, enable_vertical_final) diff --git a/src/Processors/Merges/VersionedCollapsingTransform.h b/src/Processors/Merges/VersionedCollapsingTransform.h index 32b5d7bf343..0bdccd4795d 100644 --- a/src/Processors/Merges/VersionedCollapsingTransform.h +++ b/src/Processors/Merges/VersionedCollapsingTransform.h @@ -21,7 +21,7 @@ public: SortDescription description_, const String & sign_column_, size_t max_block_size_rows, size_t max_block_size_bytes, - WriteBuffer * out_row_sources_buf_ = nullptr, + std::shared_ptr temp_data_buffer_ = nullptr, bool use_average_block_sizes = false) : IMergingTransform( num_inputs, header, header, /*have_all_inputs_=*/ true, /*limit_hint_=*/ 0, /*always_read_till_end_=*/ false, @@ -31,7 +31,7 @@ public: sign_column_, max_block_size_rows, max_block_size_bytes, - out_row_sources_buf_, + temp_data_buffer_, use_average_block_sizes) { } diff --git a/src/Processors/QueryPlan/BuildQueryPipelineSettings.h b/src/Processors/QueryPlan/BuildQueryPipelineSettings.h index d99f9a7d1f1..1c03a4d74cd 100644 --- a/src/Processors/QueryPlan/BuildQueryPipelineSettings.h +++ b/src/Processors/QueryPlan/BuildQueryPipelineSettings.h @@ -20,7 +20,6 @@ struct BuildQueryPipelineSettings ExpressionActionsSettings actions_settings; QueryStatusPtr process_list_element; ProgressCallback progress_callback = nullptr; - TemporaryFileLookupPtr temporary_file_lookup; const ExpressionActionsSettings & getActionsSettings() const { return actions_settings; } static BuildQueryPipelineSettings fromContext(ContextPtr from); diff --git a/src/Processors/QueryPlan/SortingStep.cpp b/src/Processors/QueryPlan/SortingStep.cpp index 5ad2f1f62d5..4fde246f764 100644 --- a/src/Processors/QueryPlan/SortingStep.cpp +++ b/src/Processors/QueryPlan/SortingStep.cpp @@ -280,9 +280,9 @@ void SortingStep::mergeSorting( if (increase_sort_description_compile_attempts) increase_sort_description_compile_attempts = false; - auto tmp_data_on_disk = sort_settings.tmp_data - ? std::make_unique(sort_settings.tmp_data, CurrentMetrics::TemporaryFilesForSort) - : std::unique_ptr(); + TemporaryDataOnDiskScopePtr tmp_data_on_disk = nullptr; + if (sort_settings.tmp_data) + tmp_data_on_disk = sort_settings.tmp_data->childScope(CurrentMetrics::TemporaryFilesForSort); return std::make_shared( header, diff --git a/src/Processors/Transforms/AggregatingTransform.cpp b/src/Processors/Transforms/AggregatingTransform.cpp index 68f23898018..2c54788b995 100644 --- a/src/Processors/Transforms/AggregatingTransform.cpp +++ b/src/Processors/Transforms/AggregatingTransform.cpp @@ -54,9 +54,9 @@ namespace class SourceFromNativeStream : public ISource { public: - explicit SourceFromNativeStream(TemporaryFileStream * tmp_stream_) - : ISource(tmp_stream_->getHeader()) - , tmp_stream(tmp_stream_) + explicit SourceFromNativeStream(const Block & header, TemporaryBlockStreamReaderHolder tmp_stream_) + : ISource(header) + , tmp_stream(std::move(tmp_stream_)) {} String getName() const override { return "SourceFromNativeStream"; } @@ -69,7 +69,7 @@ namespace auto block = tmp_stream->read(); if (!block) { - tmp_stream = nullptr; + tmp_stream.reset(); return {}; } return convertToChunk(block); @@ -78,7 +78,7 @@ namespace std::optional getReadProgress() override { return std::nullopt; } private: - TemporaryFileStream * tmp_stream; + TemporaryBlockStreamReaderHolder tmp_stream; }; } @@ -811,15 +811,18 @@ void AggregatingTransform::initGenerate() Pipes pipes; /// Merge external data from all aggregators used in query. - for (const auto & aggregator : *params->aggregator_list_ptr) + for (auto & aggregator : *params->aggregator_list_ptr) { - const auto & tmp_data = aggregator.getTemporaryData(); - for (auto * tmp_stream : tmp_data.getStreams()) - pipes.emplace_back(Pipe(std::make_unique(tmp_stream))); + auto & tmp_data = aggregator.getTemporaryData(); + num_streams += tmp_data.size(); - num_streams += tmp_data.getStreams().size(); - compressed_size += tmp_data.getStat().compressed_size; - uncompressed_size += tmp_data.getStat().uncompressed_size; + for (auto & tmp_stream : tmp_data) + { + auto stat = tmp_stream.finishWriting(); + compressed_size += stat.compressed_size; + uncompressed_size += stat.uncompressed_size; + pipes.emplace_back(Pipe(std::make_unique(tmp_stream.getHeader(), tmp_stream.getReadStream()))); + } } LOG_DEBUG( diff --git a/src/Processors/Transforms/MergeSortingTransform.cpp b/src/Processors/Transforms/MergeSortingTransform.cpp index c45192e7118..ba157dabffb 100644 --- a/src/Processors/Transforms/MergeSortingTransform.cpp +++ b/src/Processors/Transforms/MergeSortingTransform.cpp @@ -27,15 +27,20 @@ namespace ProfileEvents namespace DB { +namespace ErrorCodes +{ + extern const int LOGICAL_ERROR; +} + class BufferingToFileTransform : public IAccumulatingTransform { public: - BufferingToFileTransform(const Block & header, TemporaryFileStream & tmp_stream_, LoggerPtr log_) + BufferingToFileTransform(const Block & header, TemporaryBlockStreamHolder tmp_stream_, LoggerPtr log_) : IAccumulatingTransform(header, header) - , tmp_stream(tmp_stream_) + , tmp_stream(std::move(tmp_stream_)) , log(log_) { - LOG_INFO(log, "Sorting and writing part of data into temporary file {}", tmp_stream.getPath()); + LOG_INFO(log, "Sorting and writing part of data into temporary file {}", tmp_stream.getHolder()->describeFilePath()); ProfileEvents::increment(ProfileEvents::ExternalSortWritePart); } @@ -44,14 +49,15 @@ public: void consume(Chunk chunk) override { Block block = getInputPort().getHeader().cloneWithColumns(chunk.detachColumns()); - tmp_stream.write(block); + tmp_stream->write(block); } Chunk generate() override { - if (!tmp_stream.isWriteFinished()) + if (!tmp_read_stream) { auto stat = tmp_stream.finishWriting(); + tmp_read_stream = tmp_stream.getReadStream(); ProfileEvents::increment(ProfileEvents::ExternalProcessingCompressedBytesTotal, stat.compressed_size); ProfileEvents::increment(ProfileEvents::ExternalProcessingUncompressedBytesTotal, stat.uncompressed_size); @@ -59,10 +65,11 @@ public: ProfileEvents::increment(ProfileEvents::ExternalSortUncompressedBytes, stat.uncompressed_size); LOG_INFO(log, "Done writing part of data into temporary file {}, compressed {}, uncompressed {} ", - tmp_stream.getPath(), ReadableSize(static_cast(stat.compressed_size)), ReadableSize(static_cast(stat.uncompressed_size))); + tmp_stream.getHolder()->describeFilePath(), + ReadableSize(static_cast(stat.compressed_size)), ReadableSize(static_cast(stat.uncompressed_size))); } - Block block = tmp_stream.read(); + Block block = tmp_read_stream->read(); if (!block) return {}; @@ -71,7 +78,8 @@ public: } private: - TemporaryFileStream & tmp_stream; + TemporaryBlockStreamHolder tmp_stream; + TemporaryBlockStreamReaderHolder tmp_read_stream; LoggerPtr log; }; @@ -86,7 +94,7 @@ MergeSortingTransform::MergeSortingTransform( size_t max_bytes_before_remerge_, double remerge_lowered_memory_bytes_ratio_, size_t max_bytes_before_external_sort_, - TemporaryDataOnDiskPtr tmp_data_, + TemporaryDataOnDiskScopePtr tmp_data_, size_t min_free_disk_space_) : SortingTransform(header, description_, max_merged_block_size_, limit_, increase_sort_description_compile_attempts) , max_bytes_before_remerge(max_bytes_before_remerge_) @@ -168,9 +176,13 @@ void MergeSortingTransform::consume(Chunk chunk) */ if (max_bytes_before_external_sort && sum_bytes_in_blocks > max_bytes_before_external_sort) { + if (!tmp_data) + throw Exception(ErrorCodes::LOGICAL_ERROR, "TemporaryDataOnDisk is not set for MergeSortingTransform"); + temporary_files_num++; + /// If there's less free disk space than reserve_size, an exception will be thrown size_t reserve_size = sum_bytes_in_blocks + min_free_disk_space; - auto & tmp_stream = tmp_data->createStream(header_without_constants, reserve_size); + TemporaryBlockStreamHolder tmp_stream(header_without_constants, tmp_data.get(), reserve_size); size_t max_merged_block_size = this->max_merged_block_size; if (max_block_bytes > 0 && sum_rows_in_blocks > 0 && sum_bytes_in_blocks > 0) { @@ -179,7 +191,7 @@ void MergeSortingTransform::consume(Chunk chunk) max_merged_block_size = std::max(std::min(max_merged_block_size, max_block_bytes / avg_row_bytes), 128UL); } merge_sorter = std::make_unique(header_without_constants, std::move(chunks), description, max_merged_block_size, limit); - auto current_processor = std::make_shared(header_without_constants, tmp_stream, log); + auto current_processor = std::make_shared(header_without_constants, std::move(tmp_stream), log); processors.emplace_back(current_processor); @@ -221,14 +233,14 @@ void MergeSortingTransform::generate() { if (!generated_prefix) { - size_t num_tmp_files = tmp_data ? tmp_data->getStreams().size() : 0; - if (num_tmp_files == 0) - merge_sorter - = std::make_unique(header_without_constants, std::move(chunks), description, max_merged_block_size, limit); + if (temporary_files_num == 0) + { + merge_sorter = std::make_unique(header_without_constants, std::move(chunks), description, max_merged_block_size, limit); + } else { ProfileEvents::increment(ProfileEvents::ExternalSortMerge); - LOG_INFO(log, "There are {} temporary sorted parts to merge", num_tmp_files); + LOG_INFO(log, "There are {} temporary sorted parts to merge", temporary_files_num); processors.emplace_back(std::make_shared( header_without_constants, std::move(chunks), description, max_merged_block_size, limit)); diff --git a/src/Processors/Transforms/MergeSortingTransform.h b/src/Processors/Transforms/MergeSortingTransform.h index a39dd66caa0..f7cb63d518b 100644 --- a/src/Processors/Transforms/MergeSortingTransform.h +++ b/src/Processors/Transforms/MergeSortingTransform.h @@ -29,7 +29,7 @@ public: size_t max_bytes_before_remerge_, double remerge_lowered_memory_bytes_ratio_, size_t max_bytes_before_external_sort_, - TemporaryDataOnDiskPtr tmp_data_, + TemporaryDataOnDiskScopePtr tmp_data_, size_t min_free_disk_space_); String getName() const override { return "MergeSortingTransform"; } @@ -45,7 +45,8 @@ private: size_t max_bytes_before_remerge; double remerge_lowered_memory_bytes_ratio; size_t max_bytes_before_external_sort; - TemporaryDataOnDiskPtr tmp_data; + TemporaryDataOnDiskScopePtr tmp_data; + size_t temporary_files_num = 0; size_t min_free_disk_space; size_t max_block_bytes; diff --git a/src/QueryPipeline/QueryPipelineBuilder.h b/src/QueryPipeline/QueryPipelineBuilder.h index a9e5b1535c0..1e274a97a08 100644 --- a/src/QueryPipeline/QueryPipelineBuilder.h +++ b/src/QueryPipeline/QueryPipelineBuilder.h @@ -197,6 +197,12 @@ public: void setQueryIdHolder(std::shared_ptr query_id_holder) { resources.query_id_holders.emplace_back(std::move(query_id_holder)); } void addContext(ContextPtr context) { resources.interpreter_context.emplace_back(std::move(context)); } + template + void addResource(Resource resource, std::vector QueryPlanResourceHolder::*field) + { + (resources.*field).push_back(std::move(resource)); + } + /// Convert query pipeline to pipe. static Pipe getPipe(QueryPipelineBuilder pipeline, QueryPlanResourceHolder & resources); static QueryPipeline getPipeline(QueryPipelineBuilder builder); diff --git a/src/QueryPipeline/QueryPlanResourceHolder.h b/src/QueryPipeline/QueryPlanResourceHolder.h index 10f7f39ab09..ee2ecc25cd5 100644 --- a/src/QueryPipeline/QueryPlanResourceHolder.h +++ b/src/QueryPipeline/QueryPlanResourceHolder.h @@ -13,6 +13,7 @@ class QueryPlan; class Context; struct QueryIdHolder; +class TemporaryDataBuffer; struct QueryPlanResourceHolder { @@ -33,6 +34,7 @@ struct QueryPlanResourceHolder std::vector storage_holders; std::vector table_locks; std::vector> query_id_holders; + std::vector> rows_sources_temporary_file; }; } diff --git a/src/Server/HTTPHandler.cpp b/src/Server/HTTPHandler.cpp index 8a9ae05b355..52b56860543 100644 --- a/src/Server/HTTPHandler.cpp +++ b/src/Server/HTTPHandler.cpp @@ -162,15 +162,16 @@ void HTTPHandler::pushDelayedResults(Output & used_output) for (auto & write_buf : write_buffers) { - if (!write_buf) - continue; - - IReadableWriteBuffer * write_buf_concrete = dynamic_cast(write_buf.get()); - if (write_buf_concrete) + if (auto * write_buf_concrete = dynamic_cast(write_buf.get())) { - ReadBufferPtr reread_buf = write_buf_concrete->tryGetReadBuffer(); - if (reread_buf) - read_buffers.emplace_back(wrapReadBufferPointer(reread_buf)); + if (auto reread_buf = write_buf_concrete->read()) + read_buffers.emplace_back(std::move(reread_buf)); + } + + if (auto * write_buf_concrete = dynamic_cast(write_buf.get())) + { + if (auto reread_buf = write_buf_concrete->tryGetReadBuffer()) + read_buffers.emplace_back(std::move(reread_buf)); } } @@ -312,21 +313,19 @@ void HTTPHandler::processQuery( if (buffer_size_memory > 0 || buffer_until_eof) { - CascadeWriteBuffer::WriteBufferPtrs cascade_buffer1; - CascadeWriteBuffer::WriteBufferConstructors cascade_buffer2; + CascadeWriteBuffer::WriteBufferPtrs cascade_buffers; + CascadeWriteBuffer::WriteBufferConstructors cascade_buffers_lazy; if (buffer_size_memory > 0) - cascade_buffer1.emplace_back(std::make_shared(buffer_size_memory)); + cascade_buffers.emplace_back(std::make_shared(buffer_size_memory)); if (buffer_until_eof) { - auto tmp_data = std::make_shared(server.context()->getTempDataOnDisk()); - - auto create_tmp_disk_buffer = [tmp_data] (const WriteBufferPtr &) -> WriteBufferPtr { - return tmp_data->createRawStream(); - }; - - cascade_buffer2.emplace_back(std::move(create_tmp_disk_buffer)); + auto tmp_data = server.context()->getTempDataOnDisk(); + cascade_buffers_lazy.emplace_back([tmp_data](const WriteBufferPtr &) -> WriteBufferPtr + { + return std::make_unique(tmp_data.get()); + }); } else { @@ -342,10 +341,10 @@ void HTTPHandler::processQuery( return next_buffer; }; - cascade_buffer2.emplace_back(push_memory_buffer_and_continue); + cascade_buffers_lazy.emplace_back(push_memory_buffer_and_continue); } - used_output.out_delayed_and_compressed_holder = std::make_unique(std::move(cascade_buffer1), std::move(cascade_buffer2)); + used_output.out_delayed_and_compressed_holder = std::make_unique(std::move(cascade_buffers), std::move(cascade_buffers_lazy)); used_output.out_maybe_delayed_and_compressed = used_output.out_delayed_and_compressed_holder.get(); } else diff --git a/src/Storages/MergeTree/MergeTask.cpp b/src/Storages/MergeTree/MergeTask.cpp index 74d6d60ba1b..5c9d4ea61a2 100644 --- a/src/Storages/MergeTree/MergeTask.cpp +++ b/src/Storages/MergeTree/MergeTask.cpp @@ -118,68 +118,6 @@ static ColumnsStatistics getStatisticsForColumns( return all_statistics; } -/// Manages the "rows_sources" temporary file that is used during vertical merge. -class RowsSourcesTemporaryFile : public ITemporaryFileLookup -{ -public: - /// A logical name of the temporary file under which it will be known to the plan steps that use it. - static constexpr auto FILE_ID = "rows_sources"; - - explicit RowsSourcesTemporaryFile(TemporaryDataOnDiskScopePtr temporary_data_on_disk_) - : tmp_disk(std::make_unique(temporary_data_on_disk_)) - , uncompressed_write_buffer(tmp_disk->createRawStream()) - , tmp_file_name_on_disk(uncompressed_write_buffer->getFileName()) - { - } - - WriteBuffer & getTemporaryFileForWriting(const String & name) override - { - if (name != FILE_ID) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected temporary file name requested: {}", name); - - if (write_buffer) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file was already requested for writing, there musto be only one writer"); - - write_buffer = (std::make_unique(*uncompressed_write_buffer)); - return *write_buffer; - } - - std::unique_ptr getTemporaryFileForReading(const String & name) override - { - if (name != FILE_ID) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected temporary file name requested: {}", name); - - if (!finalized) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file is not finalized yet"); - - /// tmp_disk might not create real file if no data was written to it. - if (final_size == 0) - return std::make_unique(); - - /// Reopen the file for each read so that multiple reads can be performed in parallel and there is no need to seek to the beginning. - auto raw_file_read_buffer = std::make_unique(tmp_file_name_on_disk); - return std::make_unique(std::move(raw_file_read_buffer)); - } - - /// Returns written data size in bytes - size_t finalizeWriting() - { - write_buffer->finalize(); - uncompressed_write_buffer->finalize(); - finalized = true; - final_size = write_buffer->count(); - return final_size; - } - -private: - std::unique_ptr tmp_disk; - std::unique_ptr uncompressed_write_buffer; - std::unique_ptr write_buffer; - const String tmp_file_name_on_disk; - bool finalized = false; - size_t final_size = 0; -}; - static void addMissedColumnsToSerializationInfos( size_t num_rows_in_parts, const Names & part_columns, @@ -480,7 +418,7 @@ bool MergeTask::ExecuteAndFinalizeHorizontalPart::prepare() const } case MergeAlgorithm::Vertical: { - ctx->rows_sources_temporary_file = std::make_shared(global_ctx->context->getTempDataOnDisk()); + ctx->rows_sources_temporary_file = std::make_unique(global_ctx->context->getTempDataOnDisk().get()); std::map local_merged_column_to_size; for (const auto & part : global_ctx->future_part->parts) @@ -854,22 +792,11 @@ bool MergeTask::VerticalMergeStage::prepareVerticalMergeForAllColumns() const if (global_ctx->chosen_merge_algorithm != MergeAlgorithm::Vertical) return false; - size_t sum_input_rows_exact = global_ctx->merge_list_element_ptr->rows_read; - size_t input_rows_filtered = *global_ctx->input_rows_filtered; global_ctx->merge_list_element_ptr->columns_written = global_ctx->merging_columns.size(); global_ctx->merge_list_element_ptr->progress.store(ctx->column_sizes->keyColumnsWeight(), std::memory_order_relaxed); /// Ensure data has written to disk. - size_t rows_sources_count = ctx->rows_sources_temporary_file->finalizeWriting(); - /// In special case, when there is only one source part, and no rows were skipped, we may have - /// skipped writing rows_sources file. Otherwise rows_sources_count must be equal to the total - /// number of input rows. - if ((rows_sources_count > 0 || global_ctx->future_part->parts.size() > 1) && sum_input_rows_exact != rows_sources_count + input_rows_filtered) - throw Exception( - ErrorCodes::LOGICAL_ERROR, - "Number of rows in source parts ({}) excluding filtered rows ({}) differs from number " - "of bytes written to rows_sources file ({}). It is a bug.", - sum_input_rows_exact, input_rows_filtered, rows_sources_count); + ctx->rows_sources_temporary_file->finishWriting(); ctx->it_name_and_type = global_ctx->gathering_columns.cbegin(); @@ -901,12 +828,12 @@ class ColumnGathererStep : public ITransformingStep public: ColumnGathererStep( const Header & input_header_, - const String & rows_sources_temporary_file_name_, + std::unique_ptr rows_sources_read_buf_, UInt64 merge_block_size_rows_, UInt64 merge_block_size_bytes_, bool is_result_sparse_) : ITransformingStep(input_header_, input_header_, getTraits()) - , rows_sources_temporary_file_name(rows_sources_temporary_file_name_) + , rows_sources_read_buf(std::move(rows_sources_read_buf_)) , merge_block_size_rows(merge_block_size_rows_) , merge_block_size_bytes(merge_block_size_bytes_) , is_result_sparse(is_result_sparse_) @@ -914,15 +841,13 @@ public: String getName() const override { return "ColumnGatherer"; } - void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & pipeline_settings) override + void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & /* pipeline_settings */) override { - const auto &header = pipeline.getHeader(); + const auto & header = pipeline.getHeader(); const auto input_streams_count = pipeline.getNumStreams(); - if (!pipeline_settings.temporary_file_lookup) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file lookup is not set in pipeline settings for vertical merge"); - - auto rows_sources_read_buf = pipeline_settings.temporary_file_lookup->getTemporaryFileForReading(rows_sources_temporary_file_name); + if (!rows_sources_read_buf) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary data buffer for rows sources is not set"); auto transform = std::make_unique( header, @@ -957,7 +882,7 @@ private: } MergeTreeData::MergingParams merging_params{}; - const String rows_sources_temporary_file_name; + std::unique_ptr rows_sources_read_buf; const UInt64 merge_block_size_rows; const UInt64 merge_block_size_bytes; const bool is_result_sparse; @@ -1008,7 +933,7 @@ MergeTask::VerticalMergeRuntimeContext::PreparedColumnPipeline MergeTask::Vertic const auto data_settings = global_ctx->data->getSettings(); auto merge_step = std::make_unique( merge_column_query_plan.getCurrentHeader(), - RowsSourcesTemporaryFile::FILE_ID, + ctx->rows_sources_temporary_file->read(), (*data_settings)[MergeTreeSetting::merge_max_block_size], (*data_settings)[MergeTreeSetting::merge_max_block_size_bytes], is_result_sparse); @@ -1037,9 +962,9 @@ MergeTask::VerticalMergeRuntimeContext::PreparedColumnPipeline MergeTask::Vertic } auto pipeline_settings = BuildQueryPipelineSettings::fromContext(global_ctx->context); - pipeline_settings.temporary_file_lookup = ctx->rows_sources_temporary_file; auto optimization_settings = QueryPlanOptimizationSettings::fromContext(global_ctx->context); auto builder = merge_column_query_plan.buildQueryPipeline(optimization_settings, pipeline_settings); + builder->addResource>(ctx->rows_sources_temporary_file, &QueryPlanResourceHolder::rows_sources_temporary_file); return {QueryPipelineBuilder::getPipeline(std::move(*builder)), std::move(indexes_to_recalc)}; } @@ -1401,7 +1326,7 @@ public: const SortDescription & sort_description_, const Names partition_key_columns_, const MergeTreeData::MergingParams & merging_params_, - const String & rows_sources_temporary_file_name_, + std::shared_ptr rows_sources_temporary_file_, UInt64 merge_block_size_rows_, UInt64 merge_block_size_bytes_, bool blocks_are_granules_size_, @@ -1411,7 +1336,7 @@ public: , sort_description(sort_description_) , partition_key_columns(partition_key_columns_) , merging_params(merging_params_) - , rows_sources_temporary_file_name(rows_sources_temporary_file_name_) + , rows_sources_temporary_file(rows_sources_temporary_file_) , merge_block_size_rows(merge_block_size_rows_) , merge_block_size_bytes(merge_block_size_bytes_) , blocks_are_granules_size(blocks_are_granules_size_) @@ -1421,7 +1346,7 @@ public: String getName() const override { return "MergeParts"; } - void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & pipeline_settings) override + void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & /* pipeline_settings */) override { /// The order of the streams is important: when the key is matched, the elements go in the order of the source stream number. /// In the merged part, the lines with the same key must be in the ascending order of the identifier of original part, @@ -1431,14 +1356,6 @@ public: const auto &header = pipeline.getHeader(); const auto input_streams_count = pipeline.getNumStreams(); - WriteBuffer * rows_sources_write_buf = nullptr; - if (!rows_sources_temporary_file_name.empty()) - { - if (!pipeline_settings.temporary_file_lookup) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file lookup is not set in pipeline settings for vertical merge"); - rows_sources_write_buf = &pipeline_settings.temporary_file_lookup->getTemporaryFileForWriting(rows_sources_temporary_file_name); - } - switch (merging_params.mode) { case MergeTreeData::MergingParams::Ordinary: @@ -1451,14 +1368,14 @@ public: SortingQueueStrategy::Default, /* limit_= */0, /* always_read_till_end_= */false, - rows_sources_write_buf, + rows_sources_temporary_file, blocks_are_granules_size); break; case MergeTreeData::MergingParams::Collapsing: merged_transform = std::make_shared( header, input_streams_count, sort_description, merging_params.sign_column, false, - merge_block_size_rows, merge_block_size_bytes, rows_sources_write_buf, blocks_are_granules_size); + merge_block_size_rows, merge_block_size_bytes, rows_sources_temporary_file, blocks_are_granules_size); break; case MergeTreeData::MergingParams::Summing: @@ -1473,7 +1390,7 @@ public: case MergeTreeData::MergingParams::Replacing: merged_transform = std::make_shared( header, input_streams_count, sort_description, merging_params.is_deleted_column, merging_params.version_column, - merge_block_size_rows, merge_block_size_bytes, rows_sources_write_buf, blocks_are_granules_size, + merge_block_size_rows, merge_block_size_bytes, rows_sources_temporary_file, blocks_are_granules_size, cleanup); break; @@ -1486,7 +1403,7 @@ public: case MergeTreeData::MergingParams::VersionedCollapsing: merged_transform = std::make_shared( header, input_streams_count, sort_description, merging_params.sign_column, - merge_block_size_rows, merge_block_size_bytes, rows_sources_write_buf, blocks_are_granules_size); + merge_block_size_rows, merge_block_size_bytes, rows_sources_temporary_file, blocks_are_granules_size); break; } @@ -1528,7 +1445,7 @@ private: const SortDescription sort_description; const Names partition_key_columns; const MergeTreeData::MergingParams merging_params{}; - const String rows_sources_temporary_file_name; + std::shared_ptr rows_sources_temporary_file; const UInt64 merge_block_size_rows; const UInt64 merge_block_size_bytes; const bool blocks_are_granules_size; @@ -1697,7 +1614,7 @@ void MergeTask::ExecuteAndFinalizeHorizontalPart::createMergedStream() const sort_description, partition_key_columns, global_ctx->merging_params, - (is_vertical_merge ? RowsSourcesTemporaryFile::FILE_ID : ""), /// rows_sources temporaty file is used only for vertical merge + (is_vertical_merge ? ctx->rows_sources_temporary_file : nullptr), /// rows_sources temporaty file is used only for vertical merge (*data_settings)[MergeTreeSetting::merge_max_block_size], (*data_settings)[MergeTreeSetting::merge_max_block_size_bytes], ctx->blocks_are_granules_size, @@ -1762,7 +1679,6 @@ void MergeTask::ExecuteAndFinalizeHorizontalPart::createMergedStream() const { auto pipeline_settings = BuildQueryPipelineSettings::fromContext(global_ctx->context); - pipeline_settings.temporary_file_lookup = ctx->rows_sources_temporary_file; auto optimization_settings = QueryPlanOptimizationSettings::fromContext(global_ctx->context); auto builder = merge_parts_query_plan.buildQueryPipeline(optimization_settings, pipeline_settings); diff --git a/src/Storages/MergeTree/MergeTask.h b/src/Storages/MergeTree/MergeTask.h index 5a4fb1ec0b8..a3d72127627 100644 --- a/src/Storages/MergeTree/MergeTask.h +++ b/src/Storages/MergeTree/MergeTask.h @@ -41,7 +41,6 @@ namespace DB class MergeTask; using MergeTaskPtr = std::shared_ptr; -class RowsSourcesTemporaryFile; /** * Overview of the merge algorithm @@ -235,7 +234,7 @@ private: bool force_ttl{false}; CompressionCodecPtr compression_codec{nullptr}; size_t sum_input_rows_upper_bound{0}; - std::shared_ptr rows_sources_temporary_file; + std::shared_ptr rows_sources_temporary_file; std::optional column_sizes{}; /// For projections to rebuild @@ -314,7 +313,7 @@ private: struct VerticalMergeRuntimeContext : public IStageRuntimeContext { /// Begin dependencies from previous stage - std::shared_ptr rows_sources_temporary_file; + std::shared_ptr rows_sources_temporary_file; std::optional column_sizes; CompressionCodecPtr compression_codec; std::list::const_iterator it_name_and_type; diff --git a/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp b/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp index 76bcf41d6d8..a53d4213cbd 100644 --- a/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp +++ b/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp @@ -113,10 +113,11 @@ MergeTreeSelectProcessor::MergeTreeSelectProcessor( } if (!prewhere_actions.steps.empty()) - LOG_TRACE(log, "PREWHERE condition was split into {} steps: {}", prewhere_actions.steps.size(), prewhere_actions.dumpConditions()); + LOG_TRACE(log, "PREWHERE condition was split into {} steps", prewhere_actions.steps.size()); if (prewhere_info) - LOG_TEST(log, "Original PREWHERE DAG:\n{}\nPREWHERE actions:\n{}", + LOG_TEST(log, "Original PREWHERE DAG:{}\n{}\nPREWHERE actions:\n{}", + prewhere_actions.dumpConditions(), prewhere_info->prewhere_actions.dumpDAG(), (!prewhere_actions.steps.empty() ? prewhere_actions.dump() : std::string(""))); } From db2aab199db6e542c5a87c30466c358a2207c30a Mon Sep 17 00:00:00 2001 From: vdimir Date: Wed, 16 Oct 2024 11:57:11 +0000 Subject: [PATCH 008/244] log --- src/Interpreters/TemporaryDataOnDisk.cpp | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/src/Interpreters/TemporaryDataOnDisk.cpp b/src/Interpreters/TemporaryDataOnDisk.cpp index c3b24fb783b..6cc49fe83c8 100644 --- a/src/Interpreters/TemporaryDataOnDisk.cpp +++ b/src/Interpreters/TemporaryDataOnDisk.cpp @@ -58,6 +58,7 @@ public: explicit TemporaryFileInLocalCache(FileCache & file_cache, size_t max_file_size = 0) { const auto key = FileSegment::Key::random(); + LOG_TRACE(getLogger("TemporaryFileOnLocalDisk"), "Creating temporary file in cache with key {}", key); segment_holder = file_cache.set( key, 0, std::max(10_MiB, max_file_size), CreateFileSegmentSettings(FileSegmentKind::Ephemeral), FileCache::getCommonUser()); @@ -91,6 +92,7 @@ public: explicit TemporaryFileOnLocalDisk(VolumePtr volume, size_t max_file_size = 0) : path_to_file("tmp" + toString(UUIDHelpers::generateV4())) { + LOG_TRACE(getLogger("TemporaryFileOnLocalDisk"), "Creating temporary file '{}'", path_to_file); if (max_file_size > 0) { auto reservation = volume->reserve(max_file_size); @@ -129,9 +131,14 @@ public: try { if (disk->exists(path_to_file)) + { + LOG_TRACE(getLogger("TemporaryFileOnLocalDisk"), "Removing temporary file '{}'", path_to_file); disk->removeRecursive(path_to_file); + } else + { LOG_WARNING(getLogger("TemporaryFileOnLocalDisk"), "Temporary path '{}' does not exist in '{}' on disk {}", path_to_file, disk->getPath(), disk->getName()); + } } catch (...) { From b09d3c5479edcecaa041df6e6de7a45a2d407aa8 Mon Sep 17 00:00:00 2001 From: vdimir Date: Wed, 16 Oct 2024 12:01:21 +0000 Subject: [PATCH 009/244] fix --- src/Interpreters/Context.cpp | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/src/Interpreters/Context.cpp b/src/Interpreters/Context.cpp index 6ada12e63f9..f0e29dcdc41 100644 --- a/src/Interpreters/Context.cpp +++ b/src/Interpreters/Context.cpp @@ -1256,6 +1256,10 @@ try /// We skip directories (for example, 'http_buffers' - it's used for buffering of the results) and all other file types. } } + else + { + fs::create_directories(path); + } } catch (...) { From f238530cc5d222f62611214a9434138d79aabefd Mon Sep 17 00:00:00 2001 From: vdimir Date: Thu, 17 Oct 2024 15:10:39 +0000 Subject: [PATCH 010/244] w --- src/Interpreters/Aggregator.cpp | 5 ++++ src/Interpreters/Aggregator.h | 2 +- src/Interpreters/GraceHashJoin.cpp | 4 +-- src/Interpreters/HashJoin/HashJoin.cpp | 15 +++++------ src/Interpreters/HashJoin/HashJoin.h | 2 +- src/Interpreters/TemporaryDataOnDisk.h | 25 ++++++++++--------- .../Transforms/MergeSortingTransform.cpp | 4 +-- src/Storages/MergeTree/MergeTask.cpp | 2 +- 8 files changed, 33 insertions(+), 26 deletions(-) diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp index e6fecc37cfa..cdc819d3a32 100644 --- a/src/Interpreters/Aggregator.cpp +++ b/src/Interpreters/Aggregator.cpp @@ -1639,6 +1639,11 @@ Block Aggregator::convertOneBucketToBlock(AggregatedDataVariants & variants, Are return block; } +std::vector & Aggregator::getTemporaryData() +{ + return tmp_files; +} + template void Aggregator::writeToTemporaryFileImpl( AggregatedDataVariants & data_variants, diff --git a/src/Interpreters/Aggregator.h b/src/Interpreters/Aggregator.h index bc28d3dccb8..3ac5ca30ed4 100644 --- a/src/Interpreters/Aggregator.h +++ b/src/Interpreters/Aggregator.h @@ -311,7 +311,7 @@ public: bool hasTemporaryData() const { return !tmp_files.empty(); } - std::vector & getTemporaryData() { return tmp_files; } + std::vector & getTemporaryData(); /// Get data structure of the result. Block getHeader(bool final) const; diff --git a/src/Interpreters/GraceHashJoin.cpp b/src/Interpreters/GraceHashJoin.cpp index a2010b7d94b..3fb83c3ce47 100644 --- a/src/Interpreters/GraceHashJoin.cpp +++ b/src/Interpreters/GraceHashJoin.cpp @@ -389,8 +389,8 @@ void GraceHashJoin::addBuckets(const size_t bucket_count) for (size_t i = 0; i < bucket_count; ++i) try { - TemporaryBlockStreamHolder left_file = TemporaryBlockStreamHolder(left_sample_block, tmp_data.get()); - TemporaryBlockStreamHolder right_file = TemporaryBlockStreamHolder(prepareRightBlock(right_sample_block), tmp_data.get()); + TemporaryBlockStreamHolder left_file(left_sample_block, tmp_data.get()); + TemporaryBlockStreamHolder right_file(prepareRightBlock(right_sample_block), tmp_data.get()); BucketPtr new_bucket = std::make_shared(current_size + i, std::move(left_file), std::move(right_file), log); tmp_buckets.emplace_back(std::move(new_bucket)); diff --git a/src/Interpreters/HashJoin/HashJoin.cpp b/src/Interpreters/HashJoin/HashJoin.cpp index af23b520abb..a2c9f94a6ae 100644 --- a/src/Interpreters/HashJoin/HashJoin.cpp +++ b/src/Interpreters/HashJoin/HashJoin.cpp @@ -59,7 +59,7 @@ struct NotProcessedCrossJoin : public ExtraBlock { size_t left_position; size_t right_block; - TemporaryBlockStreamReaderHolder reader; + std::optional reader; }; @@ -513,9 +513,9 @@ bool HashJoin::addBlockToJoin(const Block & source_block_, bool check_limits) || (max_rows_in_join && getTotalRowCount() + block_to_save.rows() >= max_rows_in_join))) { if (!tmp_stream) - tmp_stream = TemporaryBlockStreamHolder(right_sample_block, tmp_data.get()); + tmp_stream.emplace(right_sample_block, tmp_data.get()); - tmp_stream->write(block_to_save); + tmp_stream.value()->write(block_to_save); return true; } @@ -721,13 +721,14 @@ void HashJoin::joinBlockImplCross(Block & block, ExtraBlockPtr & not_processed) { size_t start_left_row = 0; size_t start_right_block = 0; - TemporaryBlockStreamReaderHolder reader; + std::optional reader; if (not_processed) { auto & continuation = static_cast(*not_processed); start_left_row = continuation.left_position; start_right_block = continuation.right_block; - reader = std::move(continuation.reader); + if (continuation.reader) + reader = std::move(*continuation.reader); not_processed.reset(); } @@ -796,9 +797,9 @@ void HashJoin::joinBlockImplCross(Block & block, ExtraBlockPtr & not_processed) if (tmp_stream && rows_added <= max_joined_block_rows) { if (!reader) - reader = tmp_stream.getReadStream(); + reader = tmp_stream->getReadStream(); - while (auto block_right = reader->read()) + while (auto block_right = reader.value()->read()) { ++block_number; process_right_block(block_right); diff --git a/src/Interpreters/HashJoin/HashJoin.h b/src/Interpreters/HashJoin/HashJoin.h index 0f50e110db9..8572c5df096 100644 --- a/src/Interpreters/HashJoin/HashJoin.h +++ b/src/Interpreters/HashJoin/HashJoin.h @@ -424,7 +424,7 @@ private: /// Needed to do external cross join TemporaryDataOnDiskScopePtr tmp_data; - TemporaryBlockStreamHolder tmp_stream; + std::optional tmp_stream; mutable std::once_flag finish_writing; /// Block with columns from the right-side table. diff --git a/src/Interpreters/TemporaryDataOnDisk.h b/src/Interpreters/TemporaryDataOnDisk.h index f8d14b00ac5..86fa9e57e81 100644 --- a/src/Interpreters/TemporaryDataOnDisk.h +++ b/src/Interpreters/TemporaryDataOnDisk.h @@ -114,18 +114,19 @@ template class WrapperGuard { public: - WrapperGuard() = default; - template WrapperGuard(std::unique_ptr holder_, Args && ... args) : holder(std::move(holder_)) , impl(std::make_unique(*holder, std::forward(args)...)) - {} + { + chassert(holder); + chassert(impl); + } - Impl * operator->() { return impl.get(); } - const Impl * operator->() const { return impl.get(); } - Impl & operator*() { return *impl; } - const Impl & operator*() const { return *impl; } + Impl * operator->() { chassert(impl); chassert(holder); return impl.get(); } + const Impl * operator->() const { chassert(impl); chassert(holder); return impl.get(); } + Impl & operator*() { chassert(impl); chassert(holder); return *impl; } + const Impl & operator*() const { chassert(impl); chassert(holder); return *impl; } operator bool() const { return impl != nullptr; } const Holder * getHolder() const { return holder.get(); } @@ -153,13 +154,13 @@ public: virtual std::unique_ptr write() = 0; virtual std::unique_ptr read(size_t buffer_size) const = 0; - /// Get location for logging purposes + /// Get location for logging virtual String describeFilePath() const = 0; virtual ~TemporaryFileHolder() = default; }; - +/// Reads raw data from temporary file class TemporaryDataReadBuffer : public ReadBuffer { public: @@ -173,7 +174,7 @@ private: WrapperGuard compressed_buf; }; -/// Writes data to buffer provided by file_holder, and accounts amount of written data in parent scope. +/// Writes raw data to buffer provided by file_holder, and accounts amount of written data in parent scope. class TemporaryDataBuffer : public WriteBuffer { public: @@ -206,13 +207,13 @@ private: Stat stat; }; + +/// High level interfaces for reading and writing temporary data by blocks. using TemporaryBlockStreamReaderHolder = WrapperGuard; class TemporaryBlockStreamHolder : public WrapperGuard { public: - TemporaryBlockStreamHolder() = default; - TemporaryBlockStreamHolder(const Block & header_, TemporaryDataOnDiskScope * parent_, size_t max_file_size = 0); TemporaryBlockStreamReaderHolder getReadStream() const; diff --git a/src/Processors/Transforms/MergeSortingTransform.cpp b/src/Processors/Transforms/MergeSortingTransform.cpp index ba157dabffb..d3299ea651f 100644 --- a/src/Processors/Transforms/MergeSortingTransform.cpp +++ b/src/Processors/Transforms/MergeSortingTransform.cpp @@ -69,7 +69,7 @@ public: ReadableSize(static_cast(stat.compressed_size)), ReadableSize(static_cast(stat.uncompressed_size))); } - Block block = tmp_read_stream->read(); + Block block = tmp_read_stream.value()->read(); if (!block) return {}; @@ -79,7 +79,7 @@ public: private: TemporaryBlockStreamHolder tmp_stream; - TemporaryBlockStreamReaderHolder tmp_read_stream; + std::optional tmp_read_stream; LoggerPtr log; }; diff --git a/src/Storages/MergeTree/MergeTask.cpp b/src/Storages/MergeTree/MergeTask.cpp index 5c9d4ea61a2..1009458574e 100644 --- a/src/Storages/MergeTree/MergeTask.cpp +++ b/src/Storages/MergeTree/MergeTask.cpp @@ -964,7 +964,7 @@ MergeTask::VerticalMergeRuntimeContext::PreparedColumnPipeline MergeTask::Vertic auto pipeline_settings = BuildQueryPipelineSettings::fromContext(global_ctx->context); auto optimization_settings = QueryPlanOptimizationSettings::fromContext(global_ctx->context); auto builder = merge_column_query_plan.buildQueryPipeline(optimization_settings, pipeline_settings); - builder->addResource>(ctx->rows_sources_temporary_file, &QueryPlanResourceHolder::rows_sources_temporary_file); + builder->addResource(ctx->rows_sources_temporary_file, &QueryPlanResourceHolder::rows_sources_temporary_file); return {QueryPipelineBuilder::getPipeline(std::move(*builder)), std::move(indexes_to_recalc)}; } From 017d9557c5c6d41c671d55c21fb2e8810d231dd3 Mon Sep 17 00:00:00 2001 From: vdimir Date: Thu, 17 Oct 2024 15:33:33 +0000 Subject: [PATCH 011/244] f --- src/Interpreters/Aggregator.cpp | 12 +++++++++++- src/Interpreters/Aggregator.h | 7 ++++--- 2 files changed, 15 insertions(+), 4 deletions(-) diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp index cdc819d3a32..bb9e22e5a1b 100644 --- a/src/Interpreters/Aggregator.cpp +++ b/src/Interpreters/Aggregator.cpp @@ -1519,7 +1519,10 @@ void Aggregator::writeToTemporaryFile(AggregatedDataVariants & data_variants, si Stopwatch watch; size_t rows = data_variants.size(); + std::unique_lock lk(tmp_files_mutex); auto & out_stream = tmp_files.emplace_back(getHeader(false), tmp_data.get(), max_temp_file_size); + lk.unlock(); + ProfileEvents::increment(ProfileEvents::ExternalAggregationWritePart); LOG_DEBUG(log, "Writing part of aggregation data into temporary file {}", out_stream.getHolder()->describeFilePath()); @@ -1639,11 +1642,18 @@ Block Aggregator::convertOneBucketToBlock(AggregatedDataVariants & variants, Are return block; } -std::vector & Aggregator::getTemporaryData() +std::list & Aggregator::getTemporaryData() { return tmp_files; } +bool Aggregator::hasTemporaryData() const +{ + std::lock_guard lk(tmp_files_mutex); + return !tmp_files.empty(); +} + + template void Aggregator::writeToTemporaryFileImpl( AggregatedDataVariants & data_variants, diff --git a/src/Interpreters/Aggregator.h b/src/Interpreters/Aggregator.h index 3ac5ca30ed4..451583946eb 100644 --- a/src/Interpreters/Aggregator.h +++ b/src/Interpreters/Aggregator.h @@ -309,9 +309,9 @@ public: /// For external aggregation. void writeToTemporaryFile(AggregatedDataVariants & data_variants, size_t max_temp_file_size = 0) const; - bool hasTemporaryData() const { return !tmp_files.empty(); } + bool hasTemporaryData() const; - std::vector & getTemporaryData(); + std::list & getTemporaryData(); /// Get data structure of the result. Block getHeader(bool final) const; @@ -356,7 +356,8 @@ private: /// For external aggregation. TemporaryDataOnDiskScopePtr tmp_data; - mutable std::vector tmp_files; + mutable std::mutex tmp_files_mutex; + mutable std::list tmp_files; size_t min_bytes_for_prefetch = 0; From a5b9083f2c2f03345f1b14630d9bae8c25996697 Mon Sep 17 00:00:00 2001 From: vdimir Date: Fri, 18 Oct 2024 14:40:47 +0000 Subject: [PATCH 012/244] f --- src/Interpreters/TemporaryDataOnDisk.cpp | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/src/Interpreters/TemporaryDataOnDisk.cpp b/src/Interpreters/TemporaryDataOnDisk.cpp index 6cc49fe83c8..c0c9d0a80c5 100644 --- a/src/Interpreters/TemporaryDataOnDisk.cpp +++ b/src/Interpreters/TemporaryDataOnDisk.cpp @@ -9,13 +9,14 @@ #include #include #include +#include #include #include #include #include #include -#include "Common/Exception.h" +#include namespace ProfileEvents { @@ -130,7 +131,7 @@ public: ~TemporaryFileOnLocalDisk() override try { - if (disk->exists(path_to_file)) + if (disk->existsFile(path_to_file)) { LOG_TRACE(getLogger("TemporaryFileOnLocalDisk"), "Removing temporary file '{}'", path_to_file); disk->removeRecursive(path_to_file); From e8fdacdeced2a1bab0600524db90fc6cb29aaaf2 Mon Sep 17 00:00:00 2001 From: vdimir Date: Mon, 21 Oct 2024 16:50:16 +0000 Subject: [PATCH 013/244] fix --- src/Interpreters/Cache/Metadata.cpp | 1 + src/Interpreters/TemporaryDataOnDisk.cpp | 5 ++++- 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/src/Interpreters/Cache/Metadata.cpp b/src/Interpreters/Cache/Metadata.cpp index 2ee985b1c31..6a2cca33a13 100644 --- a/src/Interpreters/Cache/Metadata.cpp +++ b/src/Interpreters/Cache/Metadata.cpp @@ -470,6 +470,7 @@ private: void CacheMetadata::cleanupThreadFunc() { + LOG_DEBUG(log, "Cleanup thread started"); while (true) { Key key; diff --git a/src/Interpreters/TemporaryDataOnDisk.cpp b/src/Interpreters/TemporaryDataOnDisk.cpp index c0c9d0a80c5..ea29afbe1fa 100644 --- a/src/Interpreters/TemporaryDataOnDisk.cpp +++ b/src/Interpreters/TemporaryDataOnDisk.cpp @@ -59,7 +59,7 @@ public: explicit TemporaryFileInLocalCache(FileCache & file_cache, size_t max_file_size = 0) { const auto key = FileSegment::Key::random(); - LOG_TRACE(getLogger("TemporaryFileOnLocalDisk"), "Creating temporary file in cache with key {}", key); + LOG_TRACE(getLogger("TemporaryFileInLocalCache"), "Creating temporary file in cache with key {}", key); segment_holder = file_cache.set( key, 0, std::max(10_MiB, max_file_size), CreateFileSegmentSettings(FileSegmentKind::Ephemeral), FileCache::getCommonUser()); @@ -270,6 +270,9 @@ std::unique_ptr TemporaryDataBuffer::read() { finishWriting(); + if (stat.compressed_size == 0 && stat.uncompressed_size == 0) + return std::make_unique(std::make_unique()); + /// Keep buffer size less that file size, to avoid memory overhead for large amounts of small files size_t buffer_size = std::min(stat.compressed_size, DBMS_DEFAULT_BUFFER_SIZE); return std::make_unique(file_holder->read(buffer_size)); From 881f1a94ae72433a1b1c49ee76877a3af66b1527 Mon Sep 17 00:00:00 2001 From: vdimir Date: Mon, 21 Oct 2024 16:52:18 +0000 Subject: [PATCH 014/244] fix tidy --- src/Interpreters/TemporaryDataOnDisk.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/Interpreters/TemporaryDataOnDisk.h b/src/Interpreters/TemporaryDataOnDisk.h index 86fa9e57e81..eab3571dd07 100644 --- a/src/Interpreters/TemporaryDataOnDisk.h +++ b/src/Interpreters/TemporaryDataOnDisk.h @@ -115,7 +115,7 @@ class WrapperGuard { public: template - WrapperGuard(std::unique_ptr holder_, Args && ... args) + explicit WrapperGuard(std::unique_ptr holder_, Args && ... args) : holder(std::move(holder_)) , impl(std::make_unique(*holder, std::forward(args)...)) { @@ -127,7 +127,7 @@ public: const Impl * operator->() const { chassert(impl); chassert(holder); return impl.get(); } Impl & operator*() { chassert(impl); chassert(holder); return *impl; } const Impl & operator*() const { chassert(impl); chassert(holder); return *impl; } - operator bool() const { return impl != nullptr; } + operator bool() const { return impl != nullptr; } /// NOLINT const Holder * getHolder() const { return holder.get(); } Holder * getHolder() { return holder.get(); } From 0e702fc56d55900aaf9ce18696f18cd4855d9d17 Mon Sep 17 00:00:00 2001 From: vdimir Date: Mon, 21 Oct 2024 16:57:32 +0000 Subject: [PATCH 015/244] upd tests/integration/test_tmp_policy/test.py --- tests/integration/test_tmp_policy/test.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tests/integration/test_tmp_policy/test.py b/tests/integration/test_tmp_policy/test.py index 870a70b127a..097f93660b2 100644 --- a/tests/integration/test_tmp_policy/test.py +++ b/tests/integration/test_tmp_policy/test.py @@ -35,8 +35,8 @@ def test_disk_selection(start_cluster): node.query(query, settings=settings) assert node.contains_in_log( - "Writing part of aggregation data into temporary file /disk1/" + "Writing part of aggregation data into temporary file.*/disk1/" ) assert node.contains_in_log( - "Writing part of aggregation data into temporary file /disk2/" + "Writing part of aggregation data into temporary file.*/disk2/" ) From eccf5dd15e91663adb0c54e045f6c87e789656b3 Mon Sep 17 00:00:00 2001 From: vdimir Date: Tue, 22 Oct 2024 13:19:42 +0000 Subject: [PATCH 016/244] better error message --- src/Interpreters/TemporaryDataOnDisk.cpp | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/src/Interpreters/TemporaryDataOnDisk.cpp b/src/Interpreters/TemporaryDataOnDisk.cpp index ea29afbe1fa..60bfd379a72 100644 --- a/src/Interpreters/TemporaryDataOnDisk.cpp +++ b/src/Interpreters/TemporaryDataOnDisk.cpp @@ -15,6 +15,8 @@ #include #include +#include +#include #include #include @@ -98,7 +100,24 @@ public: { auto reservation = volume->reserve(max_file_size); if (!reservation) - throw Exception(ErrorCodes::NOT_ENOUGH_SPACE, "Not enough space on temporary disk"); + { + auto disks = volume->getDisks(); + Strings disks_info; + for (const auto & d : disks) + { + auto to_double = [](auto x) { return static_cast(x); }; + disks_info.push_back(fmt::format("{}: available: {} unreserved: {}, total: {}, keeping: {}", + d->getName(), + ReadableSize(d->getAvailableSpace().transform(to_double).value_or(NaNOrZero())), + ReadableSize(d->getUnreservedSpace().transform(to_double).value_or(NaNOrZero())), + ReadableSize(d->getTotalSpace().transform(to_double).value_or(NaNOrZero())), + ReadableSize(d->getKeepingFreeSpace()))); + } + + throw Exception(ErrorCodes::NOT_ENOUGH_SPACE, + "Not enough space on temporary disk, cannot reserve {} bytes on [{}]", + max_file_size, fmt::join(disks_info, ", ")); + } disk = reservation->getDisk(); } else From 349af95cd1e2c998391bec9710bacb0458175835 Mon Sep 17 00:00:00 2001 From: vdimir Date: Fri, 25 Oct 2024 10:45:32 +0000 Subject: [PATCH 017/244] fix data race --- src/Interpreters/Cache/Metadata.cpp | 1 - 1 file changed, 1 deletion(-) diff --git a/src/Interpreters/Cache/Metadata.cpp b/src/Interpreters/Cache/Metadata.cpp index 6a2cca33a13..2ee985b1c31 100644 --- a/src/Interpreters/Cache/Metadata.cpp +++ b/src/Interpreters/Cache/Metadata.cpp @@ -470,7 +470,6 @@ private: void CacheMetadata::cleanupThreadFunc() { - LOG_DEBUG(log, "Cleanup thread started"); while (true) { Key key; From 084f878fb19995763e3db825752dc61c9d768b43 Mon Sep 17 00:00:00 2001 From: vdimir Date: Fri, 25 Oct 2024 11:10:33 +0000 Subject: [PATCH 018/244] log --- src/Interpreters/Aggregator.cpp | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp index bb9e22e5a1b..2dd6513d498 100644 --- a/src/Interpreters/Aggregator.cpp +++ b/src/Interpreters/Aggregator.cpp @@ -1504,6 +1504,7 @@ bool Aggregator::executeOnBlock(Columns columns, && worth_convert_to_two_level) { size_t size = current_memory_usage + params.min_free_disk_space; + LOG_DEBUG(&Poco::Logger::get("XXXX"), "{}:{}: {} {}", __FILE__, __LINE__, current_memory_usage, params.min_free_disk_space); writeToTemporaryFile(result, size); } @@ -1520,6 +1521,7 @@ void Aggregator::writeToTemporaryFile(AggregatedDataVariants & data_variants, si size_t rows = data_variants.size(); std::unique_lock lk(tmp_files_mutex); + LOG_DEBUG(&Poco::Logger::get("XXXX"), "{}:{}: max_temp_file_size {}", __FILE__, __LINE__, max_temp_file_size); auto & out_stream = tmp_files.emplace_back(getHeader(false), tmp_data.get(), max_temp_file_size); lk.unlock(); @@ -2932,6 +2934,7 @@ bool Aggregator::mergeOnBlock(Block block, AggregatedDataVariants & result, bool && worth_convert_to_two_level) { size_t size = current_memory_usage + params.min_free_disk_space; + LOG_DEBUG(&Poco::Logger::get("XXXX"), "{}:{}: {} {}", __FILE__, __LINE__, current_memory_usage, params.min_free_disk_space); writeToTemporaryFile(result, size); } From 54b93953847699f1f9d14939bd1e0067d933dbba Mon Sep 17 00:00:00 2001 From: vdimir Date: Fri, 25 Oct 2024 11:11:19 +0000 Subject: [PATCH 019/244] fix typo --- src/Storages/MergeTree/MergeTask.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Storages/MergeTree/MergeTask.cpp b/src/Storages/MergeTree/MergeTask.cpp index 1009458574e..d781cef9f17 100644 --- a/src/Storages/MergeTree/MergeTask.cpp +++ b/src/Storages/MergeTree/MergeTask.cpp @@ -1614,7 +1614,7 @@ void MergeTask::ExecuteAndFinalizeHorizontalPart::createMergedStream() const sort_description, partition_key_columns, global_ctx->merging_params, - (is_vertical_merge ? ctx->rows_sources_temporary_file : nullptr), /// rows_sources temporaty file is used only for vertical merge + (is_vertical_merge ? ctx->rows_sources_temporary_file : nullptr), /// rows_sources' temporary file is used only for vertical merge (*data_settings)[MergeTreeSetting::merge_max_block_size], (*data_settings)[MergeTreeSetting::merge_max_block_size_bytes], ctx->blocks_are_granules_size, From c952d9d8153ce59458fdb69a208b361c7454cab1 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 25 Oct 2024 20:55:51 +0000 Subject: [PATCH 020/244] Right JOIN with parallel replicas --- .../ClusterProxy/executeQuery.cpp | 4 +- src/Planner/PlannerJoinTree.cpp | 2 + src/Planner/findParallelReplicasQuery.cpp | 52 ++++++++++++---- src/Storages/buildQueryTreeForShard.cpp | 62 +++++++++++++------ 4 files changed, 86 insertions(+), 34 deletions(-) diff --git a/src/Interpreters/ClusterProxy/executeQuery.cpp b/src/Interpreters/ClusterProxy/executeQuery.cpp index e88fdeb0379..4b1f3094be3 100644 --- a/src/Interpreters/ClusterProxy/executeQuery.cpp +++ b/src/Interpreters/ClusterProxy/executeQuery.cpp @@ -477,8 +477,8 @@ void executeQueryWithParallelReplicas( QueryPlanStepPtr analyzed_read_from_merge_tree) { auto logger = getLogger("executeQueryWithParallelReplicas"); - LOG_DEBUG(logger, "Executing read from {}, header {}, query ({}), stage {} with parallel replicas", - storage_id.getNameForLogs(), header.dumpStructure(), query_ast->formatForLogging(), processed_stage); + LOG_DEBUG(logger, "Executing read from {}, header {}, query ({}), stage {} with parallel replicas\n{}", + storage_id.getNameForLogs(), header.dumpStructure(), query_ast->formatForLogging(), processed_stage, StackTrace().toString()); const auto & settings = context->getSettingsRef(); diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 39c1352c9cf..7889a358d95 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -665,6 +665,8 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres bool is_single_table_expression, bool wrap_read_columns_in_subquery) { + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "table_expression:\n{}", table_expression->dumpTree()); + const auto & query_context = planner_context->getQueryContext(); const auto & settings = query_context->getSettingsRef(); diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index b97a9a36381..891e5034f44 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -100,14 +100,19 @@ std::stack getSupportingParallelReplicasQuery(const IQueryTre auto join_kind = join_node.getKind(); auto join_strictness = join_node.getStrictness(); - bool can_parallelize_join = - join_kind == JoinKind::Left - || (join_kind == JoinKind::Inner && join_strictness == JoinStrictness::All); - - if (!can_parallelize_join) + if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner && join_strictness == JoinStrictness::All)) + { + query_tree_node = join_node.getLeftTableExpression().get(); + } + else if (join_kind == JoinKind::Right) + { + query_tree_node = join_node.getRightTableExpression().get(); + } + else + { return {}; + } - query_tree_node = join_node.getLeftTableExpression().get(); break; } default: @@ -310,13 +315,15 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr static const TableNode * findTableForParallelReplicas(const IQueryTreeNode * query_tree_node) { - std::stack right_join_nodes; - while (query_tree_node || !right_join_nodes.empty()) + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "\n{}", StackTrace().toString()); + + std::stack join_nodes; + while (query_tree_node || !join_nodes.empty()) { if (!query_tree_node) { - query_tree_node = right_join_nodes.top(); - right_join_nodes.pop(); + query_tree_node = join_nodes.top(); + join_nodes.pop(); } auto join_tree_node_type = query_tree_node->getNodeType(); @@ -365,8 +372,23 @@ static const TableNode * findTableForParallelReplicas(const IQueryTreeNode * que case QueryTreeNodeType::JOIN: { const auto & join_node = query_tree_node->as(); - query_tree_node = join_node.getLeftTableExpression().get(); - right_join_nodes.push(join_node.getRightTableExpression().get()); + const auto join_kind = join_node.getKind(); + const auto join_strictness = join_node.getStrictness(); + + if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner and join_strictness == JoinStrictness::All)) + { + query_tree_node = join_node.getLeftTableExpression().get(); + join_nodes.push(join_node.getRightTableExpression().get()); + } + else if (join_kind == JoinKind::Right) + { + query_tree_node = join_node.getRightTableExpression().get(); + join_nodes.push(join_node.getLeftTableExpression().get()); + } + else + { + return nullptr; + } break; } default: @@ -400,7 +422,9 @@ const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tr if (!context->canUseParallelReplicasOnFollower()) return nullptr; - return findTableForParallelReplicas(query_tree_node.get()); + const auto * res = findTableForParallelReplicas(query_tree_node.get()); + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Table found {}", res->getStorageID().getFullTableName()); + return res; } JoinTreeQueryPlan buildQueryPlanForParallelReplicas( @@ -408,6 +432,8 @@ JoinTreeQueryPlan buildQueryPlanForParallelReplicas( const PlannerContextPtr & planner_context, std::shared_ptr storage_limits) { + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "\n{}", StackTrace().toString()); + auto processed_stage = QueryProcessingStage::WithMergeableState; auto context = planner_context->getQueryContext(); diff --git a/src/Storages/buildQueryTreeForShard.cpp b/src/Storages/buildQueryTreeForShard.cpp index bbf32c68d19..df9bfd049fb 100644 --- a/src/Storages/buildQueryTreeForShard.cpp +++ b/src/Storages/buildQueryTreeForShard.cpp @@ -314,6 +314,35 @@ TableNodePtr executeSubqueryNode(const QueryTreeNodePtr & subquery_node, return temporary_table_expression_node; } +QueryTreeNodePtr getSubqueryFromTableExpression( + const QueryTreeNodePtr & join_table_expression, + const std::unordered_map & column_source_to_columns, + const ContextPtr & context) +{ + auto join_table_expression_node_type = join_table_expression->getNodeType(); + QueryTreeNodePtr subquery_node; + + if (join_table_expression_node_type == QueryTreeNodeType::QUERY || join_table_expression_node_type == QueryTreeNodeType::UNION) + { + subquery_node = join_table_expression; + } + else if ( + join_table_expression_node_type == QueryTreeNodeType::TABLE || join_table_expression_node_type == QueryTreeNodeType::TABLE_FUNCTION) + { + const auto & columns = column_source_to_columns.at(join_table_expression).columns; + subquery_node = buildSubqueryToReadColumnsFromTableExpression(columns, join_table_expression, context); + } + else + { + throw Exception( + ErrorCodes::LOGICAL_ERROR, + "Expected JOIN right table expression to be table, table function, query or union node. Actual {}", + join_table_expression->formatASTForErrorMessage()); + } + + return subquery_node; +} + } QueryTreeNodePtr buildQueryTreeForShard(const PlannerContextPtr & planner_context, QueryTreeNodePtr query_tree_to_modify) @@ -335,37 +364,32 @@ QueryTreeNodePtr buildQueryTreeForShard(const PlannerContextPtr & planner_contex { if (auto * join_node = global_in_or_join_node.query_node->as()) { - auto join_right_table_expression = join_node->getRightTableExpression(); - auto join_right_table_expression_node_type = join_right_table_expression->getNodeType(); - - QueryTreeNodePtr subquery_node; - - if (join_right_table_expression_node_type == QueryTreeNodeType::QUERY || - join_right_table_expression_node_type == QueryTreeNodeType::UNION) + QueryTreeNodePtr join_table_expression; + const auto join_kind = join_node->getKind(); + const auto join_strictness = join_node->getStrictness(); + if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner && join_strictness == JoinStrictness::All)) { - subquery_node = join_right_table_expression; + join_table_expression = join_node->getRightTableExpression(); } - else if (join_right_table_expression_node_type == QueryTreeNodeType::TABLE || - join_right_table_expression_node_type == QueryTreeNodeType::TABLE_FUNCTION) + else if (join_kind == JoinKind::Right) { - const auto & columns = column_source_to_columns.at(join_right_table_expression).columns; - subquery_node = buildSubqueryToReadColumnsFromTableExpression(columns, - join_right_table_expression, - planner_context->getQueryContext()); + join_table_expression = join_node->getLeftTableExpression(); } else { - throw Exception(ErrorCodes::LOGICAL_ERROR, - "Expected JOIN right table expression to be table, table function, query or union node. Actual {}", - join_right_table_expression->formatASTForErrorMessage()); + throw Exception( + ErrorCodes::LOGICAL_ERROR, "Unexpected join kind: {}", join_kind); } + auto subquery_node + = getSubqueryFromTableExpression(join_table_expression, column_source_to_columns, planner_context->getQueryContext()); + auto temporary_table_expression_node = executeSubqueryNode(subquery_node, planner_context->getMutableQueryContext(), global_in_or_join_node.subquery_depth); - temporary_table_expression_node->setAlias(join_right_table_expression->getAlias()); + temporary_table_expression_node->setAlias(join_table_expression->getAlias()); - replacement_map.emplace(join_right_table_expression.get(), std::move(temporary_table_expression_node)); + replacement_map.emplace(join_table_expression.get(), std::move(temporary_table_expression_node)); continue; } if (auto * in_function_node = global_in_or_join_node.query_node->as()) From aeffae571c91909a7196bec6e952026932c43cc6 Mon Sep 17 00:00:00 2001 From: Max Kainov Date: Sat, 26 Oct 2024 19:35:33 +0200 Subject: [PATCH 021/244] CI: Functional Tests with praktika --- .github/workflows/pr.yaml | 287 ++++++++++++++++++++ .github/workflows/pull_request.yml | 212 --------------- ci/jobs/build_clickhouse.py | 12 +- ci/jobs/check_style.py | 2 +- ci/jobs/fast_test.py | 2 +- ci/jobs/functional_stateless_tests.py | 48 ++++ ci/jobs/scripts/functional_tests_results.py | 3 +- ci/praktika/__main__.py | 23 +- ci/praktika/_environment.py | 3 +- ci/praktika/_settings.py | 3 +- ci/praktika/hook_cache.py | 6 +- ci/praktika/json.html | 50 +++- ci/praktika/result.py | 2 +- ci/praktika/runner.py | 56 ++-- ci/praktika/yaml_generator.py | 6 +- ci/settings/definitions.py | 1 + ci/settings/settings.py | 2 + ci/workflows/pull_request.py | 16 +- 18 files changed, 477 insertions(+), 257 deletions(-) create mode 100644 .github/workflows/pr.yaml delete mode 100644 .github/workflows/pull_request.yml create mode 100644 ci/jobs/functional_stateless_tests.py diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml new file mode 100644 index 00000000000..34c794f6088 --- /dev/null +++ b/.github/workflows/pr.yaml @@ -0,0 +1,287 @@ +# generated by praktika + +name: PR + +on: + pull_request: + branches: ['master'] + +# Cancel the previous wf run in PRs. +concurrency: + group: ${{ github.workflow }}-${{ github.ref }} + cancel-in-progress: true + +env: + # Force the stdout and stderr streams to be unbuffered + PYTHONUNBUFFERED: 1 + GH_TOKEN: ${{ github.token }} + +# Allow updating GH commit statuses and PR comments to post an actual job reports link +permissions: write-all + +jobs: + + config_workflow: + runs-on: [ci_services] + needs: [] + name: "Config Workflow" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Config Workflow''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Config Workflow''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + docker_builds: + runs-on: [ci_services_ebs] + needs: [config_workflow] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'RG9ja2VyIEJ1aWxkcw==') }} + name: "Docker Builds" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Docker Builds''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Docker Builds''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + style_check: + runs-on: [ci_services] + needs: [config_workflow, docker_builds] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3R5bGUgQ2hlY2s=') }} + name: "Style Check" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Style Check''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Style Check''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + fast_test: + runs-on: [builder] + needs: [config_workflow, docker_builds] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'RmFzdCB0ZXN0') }} + name: "Fast test" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Fast test''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Fast test''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + build_amd64_debug: + runs-on: [builder] + needs: [config_workflow, docker_builds] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'QnVpbGQgYW1kNjQgZGVidWc=') }} + name: "Build amd64 debug" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Build amd64 debug''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Build amd64 debug''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + stateless_tests_amd_debug: + runs-on: [builder] + needs: [config_workflow, docker_builds, build_amd64_debug] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKQ==') }} + name: "Stateless tests (amd, debug)" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Stateless tests (amd, debug)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Stateless tests (amd, debug)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + finish_workflow: + runs-on: [ci_services] + needs: [config_workflow, docker_builds, style_check, fast_test, build_amd64_debug, stateless_tests_amd_debug] + if: ${{ !cancelled() }} + name: "Finish Workflow" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Finish Workflow''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Finish Workflow''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi diff --git a/.github/workflows/pull_request.yml b/.github/workflows/pull_request.yml deleted file mode 100644 index e4eb44b2774..00000000000 --- a/.github/workflows/pull_request.yml +++ /dev/null @@ -1,212 +0,0 @@ -# yamllint disable rule:comments-indentation -name: PullRequestCI - -env: - # Force the stdout and stderr streams to be unbuffered - PYTHONUNBUFFERED: 1 - -on: # yamllint disable-line rule:truthy - pull_request: - types: - - synchronize - - reopened - - opened - branches: - - master - -# Cancel the previous wf run in PRs. -concurrency: - group: ${{ github.workflow }}-${{ github.ref }} - cancel-in-progress: true - -jobs: - RunConfig: - runs-on: [self-hosted, style-checker-aarch64] - outputs: - data: ${{ steps.runconfig.outputs.CI_DATA }} - steps: - - name: Check out repository code - uses: ClickHouse/checkout@v1 - with: - clear-repository: true # to ensure correct digests - fetch-depth: 0 # to get a version - filter: tree:0 - - name: Debug Info - uses: ./.github/actions/debug - - name: Set pending Sync status - run: | - python3 "$GITHUB_WORKSPACE/tests/ci/ci.py" --set-pending-status - - name: Labels check - run: | - cd "$GITHUB_WORKSPACE/tests/ci" - python3 run_check.py - - name: Python unit tests - run: | - cd "$GITHUB_WORKSPACE/tests/ci" - echo "Testing the main ci directory" - python3 -m unittest discover -s . -p 'test_*.py' - - name: PrepareRunConfig - id: runconfig - run: | - python3 "$GITHUB_WORKSPACE/tests/ci/ci.py" --configure --outfile ${{ runner.temp }}/ci_run_data.json - - echo "::group::CI configuration" - python3 -m json.tool ${{ runner.temp }}/ci_run_data.json - echo "::endgroup::" - - { - echo 'CI_DATA<> "$GITHUB_OUTPUT" - - name: Re-create GH statuses for skipped jobs if any - run: | - python3 "$GITHUB_WORKSPACE/tests/ci/ci.py" --infile ${{ runner.temp }}/ci_run_data.json --update-gh-statuses - BuildDockers: - needs: [RunConfig] - if: ${{ !failure() && !cancelled() && toJson(fromJson(needs.RunConfig.outputs.data).docker_data.missing_multi) != '[]' }} - uses: ./.github/workflows/docker_test_images.yml - with: - data: ${{ needs.RunConfig.outputs.data }} - StyleCheck: - needs: [RunConfig, BuildDockers] - if: ${{ !failure() && !cancelled() && contains(fromJson(needs.RunConfig.outputs.data).jobs_data.jobs_to_do, 'Style check')}} - uses: ./.github/workflows/reusable_test.yml - with: - test_name: Style check - runner_type: style-checker-aarch64 - run_command: | - python3 style_check.py - data: ${{ needs.RunConfig.outputs.data }} - secrets: - secret_envs: | - ROBOT_CLICKHOUSE_SSH_KEY< "$WORKFLOW_RESULT_FILE" << 'EOF' - ${{ toJson(needs) }} - EOF - python3 merge_pr.py --set-ci-status - - name: Check Workflow results - uses: ./.github/actions/check_workflow - with: - needs: ${{ toJson(needs) }} - - ################################# Stage Final ################################# - # - FinishCheck: - if: ${{ !failure() && !cancelled() }} - needs: [RunConfig, BuildDockers, StyleCheck, FastTest, Builds_1, Builds_2, Builds_Report, Tests_1, Tests_2_ww, Tests_2] - runs-on: [self-hosted, style-checker-aarch64] - steps: - - name: Check out repository code - uses: ClickHouse/checkout@v1 - with: - filter: tree:0 - - name: Finish label - run: | - cd "$GITHUB_WORKSPACE/tests/ci" - python3 finish_check.py --wf-status ${{ contains(needs.*.result, 'failure') && 'failure' || 'success' }} - -############################################################################################# -###################################### JEPSEN TESTS ######################################### -############################################################################################# - # This is special test NOT INCLUDED in FinishCheck - # When it's skipped, all dependent tasks will be skipped too. - # DO NOT add it there - Jepsen: - # we need concurrency as the job uses dedicated instances in the cloud - concurrency: - group: jepsen - if: ${{ !failure() && !cancelled() && contains(fromJson(needs.RunConfig.outputs.data).jobs_data.jobs_to_do, 'ClickHouse Keeper Jepsen') }} - needs: [RunConfig, Builds_1] - uses: ./.github/workflows/reusable_test.yml - with: - test_name: ClickHouse Keeper Jepsen - runner_type: style-checker-aarch64 - data: ${{ needs.RunConfig.outputs.data }} diff --git a/ci/jobs/build_clickhouse.py b/ci/jobs/build_clickhouse.py index 21ed8091608..cfa358b4059 100644 --- a/ci/jobs/build_clickhouse.py +++ b/ci/jobs/build_clickhouse.py @@ -13,8 +13,14 @@ class JobStages(metaclass=MetaClasses.WithIter): def parse_args(): parser = argparse.ArgumentParser(description="ClickHouse Build Job") - parser.add_argument("BUILD_TYPE", help="Type: ") - parser.add_argument("--param", help="Optional custom job start stage", default=None) + parser.add_argument( + "BUILD_TYPE", help="Type: __" + ) + parser.add_argument( + "--param", + help="Optional user-defined job start stage (for local run)", + default=None, + ) return parser.parse_args() @@ -95,7 +101,7 @@ def main(): Shell.check(f"ls -l {build_dir}/programs/") res = results[-1].is_ok() - Result.create_from(results=results, stopwatch=stop_watch).finish_job_accordingly() + Result.create_from(results=results, stopwatch=stop_watch).complete_job() if __name__ == "__main__": diff --git a/ci/jobs/check_style.py b/ci/jobs/check_style.py index f9cdc76302d..d4b81abc92c 100644 --- a/ci/jobs/check_style.py +++ b/ci/jobs/check_style.py @@ -379,4 +379,4 @@ if __name__ == "__main__": ) ) - Result.create_from(results=results, stopwatch=stop_watch).finish_job_accordingly() + Result.create_from(results=results, stopwatch=stop_watch).complete_job() diff --git a/ci/jobs/fast_test.py b/ci/jobs/fast_test.py index 1dcd65b6ed2..dc5e1c975a6 100644 --- a/ci/jobs/fast_test.py +++ b/ci/jobs/fast_test.py @@ -330,7 +330,7 @@ def main(): CH.terminate() - Result.create_from(results=results, stopwatch=stop_watch).finish_job_accordingly() + Result.create_from(results=results, stopwatch=stop_watch).complete_job() if __name__ == "__main__": diff --git a/ci/jobs/functional_stateless_tests.py b/ci/jobs/functional_stateless_tests.py new file mode 100644 index 00000000000..dfdd5821a19 --- /dev/null +++ b/ci/jobs/functional_stateless_tests.py @@ -0,0 +1,48 @@ +import argparse + +from praktika.result import Result +from praktika.settings import Settings +from praktika.utils import MetaClasses, Shell, Utils + + +class JobStages(metaclass=MetaClasses.WithIter): + CHECKOUT_SUBMODULES = "checkout" + CMAKE = "cmake" + BUILD = "build" + + +def parse_args(): + parser = argparse.ArgumentParser(description="ClickHouse Build Job") + parser.add_argument("BUILD_TYPE", help="Type: ") + parser.add_argument("--param", help="Optional custom job start stage", default=None) + return parser.parse_args() + + +def main(): + + args = parse_args() + + stop_watch = Utils.Stopwatch() + + stages = list(JobStages) + stage = args.param or JobStages.CHECKOUT_SUBMODULES + if stage: + assert stage in JobStages, f"--param must be one of [{list(JobStages)}]" + print(f"Job will start from stage [{stage}]") + while stage in stages: + stages.pop(0) + stages.insert(0, stage) + + res = True + results = [] + + if res and JobStages.CHECKOUT_SUBMODULES in stages: + info = Shell.get_output(f"ls -l {Settings.INPUT_DIR}") + results.append(Result(name="TEST", status=Result.Status.SUCCESS, info=info)) + res = results[-1].is_ok() + + Result.create_from(results=results, stopwatch=stop_watch).complete_job() + + +if __name__ == "__main__": + main() diff --git a/ci/jobs/scripts/functional_tests_results.py b/ci/jobs/scripts/functional_tests_results.py index 5ac9d6b985d..aba3e4f7f5b 100755 --- a/ci/jobs/scripts/functional_tests_results.py +++ b/ci/jobs/scripts/functional_tests_results.py @@ -1,7 +1,6 @@ import dataclasses from typing import List -from praktika.environment import Environment from praktika.result import Result OK_SIGN = "[ OK " @@ -250,7 +249,7 @@ class FTResultsProcessor: # test_results.sort(key=test_result_comparator) return Result.create_from( - name=Environment.JOB_NAME, + name="Tests", results=test_results, status=state, files=[self.tests_output_file], diff --git a/ci/praktika/__main__.py b/ci/praktika/__main__.py index 7f472ecd9ae..fbb9f92909a 100644 --- a/ci/praktika/__main__.py +++ b/ci/praktika/__main__.py @@ -37,6 +37,24 @@ def create_parser(): type=str, default=None, ) + run_parser.add_argument( + "--pr", + help="PR number. Optional parameter for local run. Set if you want an required artifact to be uploaded from CI run in that PR", + type=int, + default=None, + ) + run_parser.add_argument( + "--sha", + help="Commit sha. Optional parameter for local run. Set if you want an required artifact to be uploaded from CI run on that sha, head sha will be used if not set", + type=str, + default=None, + ) + run_parser.add_argument( + "--branch", + help="Commit sha. Optional parameter for local run. Set if you want an required artifact to be uploaded from CI run on that branch, main branch name will be used if not set", + type=str, + default=None, + ) run_parser.add_argument( "--ci", help="When not set - dummy env will be generated, for local test", @@ -85,9 +103,12 @@ if __name__ == "__main__": workflow=workflow, job=job, docker=args.docker, - dummy_env=not args.ci, + local_run=not args.ci, no_docker=args.no_docker, param=args.param, + pr=args.pr, + branch=args.branch, + sha=args.sha, ) else: parser.print_help() diff --git a/ci/praktika/_environment.py b/ci/praktika/_environment.py index ce9c6f5b486..4ac8ad319f9 100644 --- a/ci/praktika/_environment.py +++ b/ci/praktika/_environment.py @@ -159,7 +159,8 @@ class _Environment(MetaClasses.Serializable): @classmethod def get_s3_prefix_static(cls, pr_number, branch, sha, latest=False): prefix = "" - if pr_number > 0: + assert sha or latest + if pr_number and pr_number > 0: prefix += f"{pr_number}" else: prefix += f"{branch}" diff --git a/ci/praktika/_settings.py b/ci/praktika/_settings.py index 3052d8ef877..1777257f484 100644 --- a/ci/praktika/_settings.py +++ b/ci/praktika/_settings.py @@ -1,5 +1,4 @@ import dataclasses -from pathlib import Path from typing import Dict, Iterable, List, Optional @@ -8,6 +7,7 @@ class _Settings: ###################################### # Pipeline generation settings # ###################################### + MAIN_BRANCH = "main" CI_PATH = "./ci" WORKFLOW_PATH_PREFIX: str = "./.github/workflows" WORKFLOWS_DIRECTORY: str = f"{CI_PATH}/workflows" @@ -111,6 +111,7 @@ _USER_DEFINED_SETTINGS = [ "CI_DB_INSERT_TIMEOUT_SEC", "SECRET_GH_APP_PEM_KEY", "SECRET_GH_APP_ID", + "MAIN_BRANCH", ] diff --git a/ci/praktika/hook_cache.py b/ci/praktika/hook_cache.py index b1b5c654f20..5cfedec0144 100644 --- a/ci/praktika/hook_cache.py +++ b/ci/praktika/hook_cache.py @@ -8,11 +8,9 @@ from praktika.utils import Utils class CacheRunnerHooks: @classmethod - def configure(cls, _workflow): - workflow_config = RunConfig.from_fs(_workflow.name) + def configure(cls, workflow): + workflow_config = RunConfig.from_fs(workflow.name) cache = Cache() - assert _Environment.get().WORKFLOW_NAME - workflow = _get_workflows(name=_Environment.get().WORKFLOW_NAME)[0] print(f"Workflow Configure, workflow [{workflow.name}]") assert ( workflow.enable_cache diff --git a/ci/praktika/json.html b/ci/praktika/json.html index 2f8c3e45d0b..af03ed702f8 100644 --- a/ci/praktika/json.html +++ b/ci/praktika/json.html @@ -89,6 +89,17 @@ letter-spacing: -0.5px; } + .dropdown-value { + width: 100px; + font-weight: normal; + font-family: inherit; + background-color: transparent; + color: inherit; + /*border: none;*/ + /*outline: none;*/ + /*cursor: pointer;*/ + } + #result-container { background-color: var(--tile-background); margin-left: calc(var(--status-width) + 20px); @@ -282,6 +293,12 @@ } } + function updateUrlParameter(paramName, paramValue) { + const url = new URL(window.location.href); + url.searchParams.set(paramName, paramValue); + window.location.href = url.toString(); + } + // Attach the toggle function to the click event of the icon document.getElementById('theme-toggle').addEventListener('click', toggleTheme); @@ -291,14 +308,14 @@ const monthNames = ["Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"]; const month = monthNames[date.getMonth()]; - const year = date.getFullYear(); + //const year = date.getFullYear(); const hours = String(date.getHours()).padStart(2, '0'); const minutes = String(date.getMinutes()).padStart(2, '0'); const seconds = String(date.getSeconds()).padStart(2, '0'); //const milliseconds = String(date.getMilliseconds()).padStart(2, '0'); return showDate - ? `${day}-${month}-${year} ${hours}:${minutes}:${seconds}` + ? `${day}'${month} ${hours}:${minutes}:${seconds}` : `${hours}:${minutes}:${seconds}`; } @@ -346,7 +363,7 @@ return 'status-other'; } - function addKeyValueToStatus(key, value) { + function addKeyValueToStatus(key, value, options = null) { const statusContainer = document.getElementById('status-container'); @@ -357,10 +374,25 @@ keyElement.className = 'json-key'; keyElement.textContent = key + ':'; - const valueElement = document.createElement('div'); - valueElement.className = 'json-value'; - valueElement.textContent = value; - + let valueElement + if (value) { + valueElement = document.createElement('div'); + valueElement.className = 'json-value'; + valueElement.textContent = value; + } else if (options) { + valueElement = document.createElement('select'); + valueElement.className = 'dropdown-value'; + valueElement.addEventListener('change', (event) => { + const selectedValue = event.target.value; + updateUrlParameter(key, selectedValue); + }); + options.forEach(optionValue => { + const option = document.createElement('option'); + option.value = optionValue; + option.textContent = optionValue; + valueElement.appendChild(option); + }); + } keyValuePair.appendChild(keyElement) keyValuePair.appendChild(valueElement) statusContainer.appendChild(keyValuePair); @@ -487,7 +519,7 @@ const columnSymbols = { name: '📂', - status: '✔️', + status: '⏯️', start_time: '🕒', duration: '⏳', info: 'ℹ️', @@ -726,7 +758,7 @@ } else { console.error("TODO") } - addKeyValueToStatus("sha", sha); + addKeyValueToStatus("sha", null, [sha, 'lala']); if (nameParams[1]) { addKeyValueToStatus("job", nameParams[1]); } diff --git a/ci/praktika/result.py b/ci/praktika/result.py index 3d3c986d5f9..2ba8309ad60 100644 --- a/ci/praktika/result.py +++ b/ci/praktika/result.py @@ -318,7 +318,7 @@ class Result(MetaClasses.Serializable): files=[log_file] if log_file else None, ) - def finish_job_accordingly(self): + def complete_job(self): self.dump() if not self.is_ok(): print("ERROR: Job Failed") diff --git a/ci/praktika/runner.py b/ci/praktika/runner.py index 797a799a74d..823c7e0f36d 100644 --- a/ci/praktika/runner.py +++ b/ci/praktika/runner.py @@ -19,7 +19,7 @@ from praktika.utils import Shell, TeePopen, Utils class Runner: @staticmethod - def generate_dummy_environment(workflow, job): + def generate_local_run_environment(workflow, job, pr=None, branch=None, sha=None): print("WARNING: Generate dummy env for local test") Shell.check( f"mkdir -p {Settings.TEMP_DIR} {Settings.INPUT_DIR} {Settings.OUTPUT_DIR}" @@ -28,9 +28,9 @@ class Runner: WORKFLOW_NAME=workflow.name, JOB_NAME=job.name, REPOSITORY="", - BRANCH="", - SHA="", - PR_NUMBER=-1, + BRANCH=branch or Settings.MAIN_BRANCH if not pr else "", + SHA=sha or Shell.get_output("git rev-parse HEAD"), + PR_NUMBER=pr or -1, EVENT_TYPE="", JOB_OUTPUT_STREAM="", EVENT_FILE_PATH="", @@ -86,7 +86,7 @@ class Runner: return 0 - def _pre_run(self, workflow, job): + def _pre_run(self, workflow, job, local_run=False): env = _Environment.get() result = Result( @@ -96,9 +96,10 @@ class Runner: ) result.dump() - if workflow.enable_report and job.name != Settings.CI_CONFIG_JOB_NAME: - print("Update Job and Workflow Report") - HtmlRunnerHooks.pre_run(workflow, job) + if not local_run: + if workflow.enable_report and job.name != Settings.CI_CONFIG_JOB_NAME: + print("Update Job and Workflow Report") + HtmlRunnerHooks.pre_run(workflow, job) print("Download required artifacts") required_artifacts = [] @@ -133,11 +134,17 @@ class Runner: env.dump() if job.run_in_docker and not no_docker: - # TODO: add support for any image, including not from ci config (e.g. ubuntu:latest) - docker_tag = RunConfig.from_fs(workflow.name).digest_dockers[ - job.run_in_docker - ] - docker = docker or f"{job.run_in_docker}:{docker_tag}" + if ":" in job.run_in_docker: + docker_name, docker_tag = job.run_in_docker.split(":") + print( + f"WARNING: Job [{job.name}] use custom docker image with a tag - praktika won't control docker version" + ) + else: + docker_name, docker_tag = ( + job.run_in_docker, + RunConfig.from_fs(workflow.name).digest_dockers[job.run_in_docker], + ) + docker = docker or f"{docker_name}:{docker_tag}" cmd = f"docker run --rm --user \"$(id -u):$(id -g)\" -e PYTHONPATH='{Settings.DOCKER_WD}:{Settings.DOCKER_WD}/ci' --volume ./:{Settings.DOCKER_WD} --volume {Settings.TEMP_DIR}:{Settings.TEMP_DIR} --workdir={Settings.DOCKER_WD} {docker} {job.command}" else: cmd = job.command @@ -285,14 +292,23 @@ class Runner: return True def run( - self, workflow, job, docker="", dummy_env=False, no_docker=False, param=None + self, + workflow, + job, + docker="", + local_run=False, + no_docker=False, + param=None, + pr=None, + sha=None, + branch=None, ): res = True setup_env_code = -10 prerun_code = -10 run_code = -10 - if res and not dummy_env: + if res and not local_run: print( f"\n\n=== Setup env script [{job.name}], workflow [{workflow.name}] ===" ) @@ -309,13 +325,15 @@ class Runner: traceback.print_exc() print(f"=== Setup env finished ===\n\n") else: - self.generate_dummy_environment(workflow, job) + self.generate_local_run_environment( + workflow, job, pr=pr, branch=branch, sha=sha + ) - if res and not dummy_env: + if res: res = False print(f"=== Pre run script [{job.name}], workflow [{workflow.name}] ===") try: - prerun_code = self._pre_run(workflow, job) + prerun_code = self._pre_run(workflow, job, local_run=local_run) res = prerun_code == 0 if not res: print(f"ERROR: Pre-run failed with exit code [{prerun_code}]") @@ -339,7 +357,7 @@ class Runner: traceback.print_exc() print(f"=== Run scrip finished ===\n\n") - if not dummy_env: + if not local_run: print(f"=== Post run script [{job.name}], workflow [{workflow.name}] ===") self._post_run(workflow, job, setup_env_code, prerun_code, run_code) print(f"=== Post run scrip finished ===") diff --git a/ci/praktika/yaml_generator.py b/ci/praktika/yaml_generator.py index 00c469fec0c..fb918b4ddba 100644 --- a/ci/praktika/yaml_generator.py +++ b/ci/praktika/yaml_generator.py @@ -102,7 +102,11 @@ jobs: run: | . /tmp/praktika_setup_env.sh set -o pipefail - {PYTHON} -m praktika run --job '''{JOB_NAME}''' --workflow "{WORKFLOW_NAME}" --ci |& tee {RUN_LOG} + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''{JOB_NAME}''' --workflow "{WORKFLOW_NAME}" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''{JOB_NAME}''' --workflow "{WORKFLOW_NAME}" --ci |& tee /tmp/praktika/praktika_run.log + fi {UPLOADS_GITHUB}\ """ diff --git a/ci/settings/definitions.py b/ci/settings/definitions.py index 176e865e6f3..c67bdee015b 100644 --- a/ci/settings/definitions.py +++ b/ci/settings/definitions.py @@ -231,3 +231,4 @@ class JobNames: STYLE_CHECK = "Style Check" FAST_TEST = "Fast test" BUILD_AMD_DEBUG = "Build amd64 debug" + STATELESS_TESTS = "Stateless tests (amd, debug)" diff --git a/ci/settings/settings.py b/ci/settings/settings.py index 8d5e7bc3c87..0f3b1efcee0 100644 --- a/ci/settings/settings.py +++ b/ci/settings/settings.py @@ -4,6 +4,8 @@ from ci.settings.definitions import ( RunnerLabels, ) +MAIN_BRANCH = "master" + S3_ARTIFACT_PATH = f"{S3_BUCKET_NAME}/artifacts" CI_CONFIG_RUNS_ON = [RunnerLabels.CI_SERVICES] DOCKER_BUILD_RUNS_ON = [RunnerLabels.CI_SERVICES_EBS] diff --git a/ci/workflows/pull_request.py b/ci/workflows/pull_request.py index 74129177efb..c7715b40fca 100644 --- a/ci/workflows/pull_request.py +++ b/ci/workflows/pull_request.py @@ -59,6 +59,19 @@ job_build_amd_debug = Job.Config( provides=[ArtifactNames.ch_debug_binary], ) +stateless_tests_job = Job.Config( + name=JobNames.STATELESS_TESTS, + runs_on=[RunnerLabels.BUILDER], + command="python3 ./ci/jobs/functional_stateless_tests.py amd_debug", + run_in_docker="clickhouse/fasttest:latest", + digest_config=Job.CacheDigestConfig( + include_paths=[ + "./ci/jobs/functional_stateless_tests.py", + ], + ), + requires=[ArtifactNames.ch_debug_binary], +) + workflow = Workflow.Config( name="PR", event=Workflow.Event.PULL_REQUEST, @@ -67,6 +80,7 @@ workflow = Workflow.Config( style_check_job, fast_test_job, job_build_amd_debug, + stateless_tests_job, ], artifacts=[ Artifact.Config( @@ -91,4 +105,4 @@ if __name__ == "__main__": # local job test inside praktika environment from praktika.runner import Runner - Runner().run(workflow, fast_test_job, docker="fasttest", dummy_env=True) + Runner().run(workflow, fast_test_job, docker="fasttest", local_run=True) From 2c3363e40e1856f7b5ce8eb23c301ee2ee403f36 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=D0=9A=D0=B8=D1=80=D0=B8=D0=BB=D0=BB=20=D0=93=D0=B0=D1=80?= =?UTF-8?q?=D0=B1=D0=B0=D1=80?= Date: Mon, 28 Oct 2024 19:00:37 +0300 Subject: [PATCH 022/244] Hard limit on replicated tables, dicts, views --- src/Common/CurrentMetrics.cpp | 1 + src/Core/ServerSettings.cpp | 3 + src/Databases/DatabasesCommon.cpp | 6 +- src/Interpreters/InterpreterCreateQuery.cpp | 63 ++++++++++++++++--- src/Interpreters/InterpreterCreateQuery.h | 2 + src/Storages/Utils.cpp | 16 +++-- src/Storages/Utils.h | 2 +- .../test_table_db_num_limit/config/config.xml | 12 ++++ .../test_table_db_num_limit/test.py | 43 ++++++++++--- 9 files changed, 121 insertions(+), 27 deletions(-) diff --git a/src/Common/CurrentMetrics.cpp b/src/Common/CurrentMetrics.cpp index e9d5e07c914..542838813de 100644 --- a/src/Common/CurrentMetrics.cpp +++ b/src/Common/CurrentMetrics.cpp @@ -242,6 +242,7 @@ M(PartsActive, "Active data part, used by current and upcoming SELECTs.") \ M(AttachedDatabase, "Active databases.") \ M(AttachedTable, "Active tables.") \ + M(AttachedReplicatedTable, "Active replicated tables.") \ M(AttachedView, "Active views.") \ M(AttachedDictionary, "Active dictionaries.") \ M(PartsOutdated, "Not active data part, but could be used by only current SELECTs, could be deleted after SELECTs finishes.") \ diff --git a/src/Core/ServerSettings.cpp b/src/Core/ServerSettings.cpp index 8c0864e78b7..2240b45a49f 100644 --- a/src/Core/ServerSettings.cpp +++ b/src/Core/ServerSettings.cpp @@ -128,7 +128,10 @@ namespace DB M(UInt64, max_database_num_to_warn, 1000lu, "If the number of databases is greater than this value, the server will create a warning that will displayed to user.", 0) \ M(UInt64, max_part_num_to_warn, 100000lu, "If the number of parts is greater than this value, the server will create a warning that will displayed to user.", 0) \ M(UInt64, max_table_num_to_throw, 0lu, "If number of tables is greater than this value, server will throw an exception. 0 means no limitation. View, remote tables, dictionary, system tables are not counted. Only count table in Atomic/Ordinary/Replicated/Lazy database engine.", 0) \ + M(UInt64, max_replicated_table_num_to_throw, 0lu, "If number of replicated tables is greater than this value, server will throw an exception. 0 means no limitation. Only count table in Atomic/Ordinary/Replicated/Lazy database engine.", 0) \ M(UInt64, max_database_num_to_throw, 0lu, "If number of databases is greater than this value, server will throw an exception. 0 means no limitation.", 0) \ + M(UInt64, max_dictionary_num_to_throw, 0lu, "If number of dictionaries is greater than this value, server will throw an exception. 0 means no limitation. Only count table in Atomic/Ordinary/Replicated/Lazy database engine.", 0) \ + M(UInt64, max_view_num_to_throw, 0lu, "If number of views is greater than this value, server will throw an exception. 0 means no limitation. Only count table in Atomic/Ordinary/Replicated/Lazy database engine.", 0) \ M(UInt64, max_authentication_methods_per_user, 100, "The maximum number of authentication methods a user can be created with or altered. Changing this setting does not affect existing users. Zero means unlimited", 0) \ M(UInt64, concurrent_threads_soft_limit_num, 0, "Sets how many concurrent thread can be allocated before applying CPU pressure. Zero means unlimited.", 0) \ M(UInt64, concurrent_threads_soft_limit_ratio_to_cores, 0, "Same as concurrent_threads_soft_limit_num, but with ratio to cores.", 0) \ diff --git a/src/Databases/DatabasesCommon.cpp b/src/Databases/DatabasesCommon.cpp index d26ec9d6eec..23d199cd160 100644 --- a/src/Databases/DatabasesCommon.cpp +++ b/src/Databases/DatabasesCommon.cpp @@ -382,7 +382,8 @@ StoragePtr DatabaseWithOwnTablesBase::detachTableUnlocked(const String & table_n if (!table_storage->isSystemStorage() && !DatabaseCatalog::isPredefinedDatabase(database_name)) { LOG_TEST(log, "Counting detached table {} to database {}", table_name, database_name); - CurrentMetrics::sub(getAttachedCounterForStorage(table_storage)); + for (auto metric : getAttachedCountersForStorage(table_storage)) + CurrentMetrics::sub(metric); } auto table_id = table_storage->getStorageID(); @@ -430,7 +431,8 @@ void DatabaseWithOwnTablesBase::attachTableUnlocked(const String & table_name, c if (!table->isSystemStorage() && !DatabaseCatalog::isPredefinedDatabase(database_name)) { LOG_TEST(log, "Counting attached table {} to database {}", table_name, database_name); - CurrentMetrics::add(getAttachedCounterForStorage(table)); + for (auto metric : getAttachedCountersForStorage(table)) + CurrentMetrics::add(metric); } } diff --git a/src/Interpreters/InterpreterCreateQuery.cpp b/src/Interpreters/InterpreterCreateQuery.cpp index 6057afefd02..f8e85733911 100644 --- a/src/Interpreters/InterpreterCreateQuery.cpp +++ b/src/Interpreters/InterpreterCreateQuery.cpp @@ -98,6 +98,9 @@ namespace CurrentMetrics { extern const Metric AttachedTable; + extern const Metric AttachedReplicatedTable; + extern const Metric AttachedDictionary; + extern const Metric AttachedView; } namespace DB @@ -146,7 +149,10 @@ namespace ServerSetting { extern const ServerSettingsBool ignore_empty_sql_security_in_create_view_query; extern const ServerSettingsUInt64 max_database_num_to_throw; + extern const ServerSettingsUInt64 max_dictionary_num_to_throw; extern const ServerSettingsUInt64 max_table_num_to_throw; + extern const ServerSettingsUInt64 max_replicated_table_num_to_throw; + extern const ServerSettingsUInt64 max_view_num_to_throw; } namespace ErrorCodes @@ -1914,16 +1920,8 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create, } } - UInt64 table_num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_table_num_to_throw]; - if (table_num_limit > 0 && !internal) - { - UInt64 table_count = CurrentMetrics::get(CurrentMetrics::AttachedTable); - if (table_count >= table_num_limit) - throw Exception(ErrorCodes::TOO_MANY_TABLES, - "Too many tables. " - "The limit (server configuration parameter `max_table_num_to_throw`) is set to {}, the current number of tables is {}", - table_num_limit, table_count); - } + if (!internal) + throwIfTooManyEntities(create, res); database->createTable(getContext(), create.getTable(), res, query_ptr); @@ -1950,6 +1948,51 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create, } +void InterpreterCreateQuery::throwIfTooManyEntities(ASTCreateQuery & create, StoragePtr storage) const +{ + if (auto * replicated_storage = typeid_cast(storage.get())) + { + UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_replicated_table_num_to_throw]; + UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedReplicatedTable); + if (attached_count >= num_limit) + throw Exception(ErrorCodes::TOO_MANY_TABLES, + "Too many replicated tables. " + "The limit (server configuration parameter `max_replicated_table_num_to_throw`) is set to {}, the current number is {}", + num_limit, attached_count); + } + else if (create.is_dictionary) + { + UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_dictionary_num_to_throw]; + UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedDictionary); + if (attached_count >= num_limit) + throw Exception(ErrorCodes::TOO_MANY_TABLES, + "Too many dictionaries. " + "The limit (server configuration parameter `max_dictionary_num_to_throw`) is set to {}, the current number is {}", + num_limit, attached_count); + } + else if (create.isView()) + { + UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_view_num_to_throw]; + UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedView); + if (attached_count >= num_limit) + throw Exception(ErrorCodes::TOO_MANY_TABLES, + "Too many views. " + "The limit (server configuration parameter `max_view_num_to_throw`) is set to {}, the current number is {}", + num_limit, attached_count); + } + else + { + UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_table_num_to_throw]; + UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedTable); + if (attached_count >= num_limit) + throw Exception(ErrorCodes::TOO_MANY_TABLES, + "Too many tables. " + "The limit (server configuration parameter `max_table_num_to_throw`) is set to {}, the current number is {}", + num_limit, attached_count); + } +} + + BlockIO InterpreterCreateQuery::doCreateOrReplaceTable(ASTCreateQuery & create, const InterpreterCreateQuery::TableProperties & properties, LoadingStrictnessLevel mode) { diff --git a/src/Interpreters/InterpreterCreateQuery.h b/src/Interpreters/InterpreterCreateQuery.h index cb7af25383e..24cf308951c 100644 --- a/src/Interpreters/InterpreterCreateQuery.h +++ b/src/Interpreters/InterpreterCreateQuery.h @@ -122,6 +122,8 @@ private: BlockIO executeQueryOnCluster(ASTCreateQuery & create); + void throwIfTooManyEntities(ASTCreateQuery & create, StoragePtr storage) const; + ASTPtr query_ptr; /// Skip safety threshold when loading tables. diff --git a/src/Storages/Utils.cpp b/src/Storages/Utils.cpp index bd03a96c7cc..72aeb0d158d 100644 --- a/src/Storages/Utils.cpp +++ b/src/Storages/Utils.cpp @@ -1,10 +1,13 @@ +#include #include #include +#include namespace CurrentMetrics { extern const Metric AttachedTable; + extern const Metric AttachedReplicatedTable; extern const Metric AttachedView; extern const Metric AttachedDictionary; } @@ -12,17 +15,20 @@ namespace CurrentMetrics namespace DB { - CurrentMetrics::Metric getAttachedCounterForStorage(const StoragePtr & storage) + std::vector getAttachedCountersForStorage(const StoragePtr & storage) { if (storage->isView()) { - return CurrentMetrics::AttachedView; + return {CurrentMetrics::AttachedView}; } if (storage->isDictionary()) { - return CurrentMetrics::AttachedDictionary; + return {CurrentMetrics::AttachedDictionary}; } - - return CurrentMetrics::AttachedTable; + if (auto * replicated_storage = typeid_cast(storage.get())) + { + return {CurrentMetrics::AttachedTable, CurrentMetrics::AttachedReplicatedTable}; + } + return {CurrentMetrics::AttachedTable}; } } diff --git a/src/Storages/Utils.h b/src/Storages/Utils.h index c86c2a4c341..eb302178485 100644 --- a/src/Storages/Utils.h +++ b/src/Storages/Utils.h @@ -6,5 +6,5 @@ namespace DB { - CurrentMetrics::Metric getAttachedCounterForStorage(const StoragePtr & storage); + std::vector getAttachedCountersForStorage(const StoragePtr & storage); } diff --git a/tests/integration/test_table_db_num_limit/config/config.xml b/tests/integration/test_table_db_num_limit/config/config.xml index 9a573b158fe..a4246c79694 100644 --- a/tests/integration/test_table_db_num_limit/config/config.xml +++ b/tests/integration/test_table_db_num_limit/config/config.xml @@ -1,5 +1,17 @@ + + + + + node1 + 9000 + + + + + 10 + 5 10 diff --git a/tests/integration/test_table_db_num_limit/test.py b/tests/integration/test_table_db_num_limit/test.py index b3aff6ddca2..ce981ffca3c 100644 --- a/tests/integration/test_table_db_num_limit/test.py +++ b/tests/integration/test_table_db_num_limit/test.py @@ -1,11 +1,14 @@ import pytest -from helpers.client import QueryRuntimeException from helpers.cluster import ClickHouseCluster cluster = ClickHouseCluster(__file__) -node = cluster.add_instance("node", main_configs=["config/config.xml"]) +node = cluster.add_instance( + "node1", + with_zookeeper=True, + main_configs=["config/config.xml"], +) @pytest.fixture(scope="module") @@ -24,10 +27,9 @@ def test_table_db_limit(started_cluster): for i in range(9): node.query("create database db{}".format(i)) - with pytest.raises(QueryRuntimeException) as exp_info: - node.query("create database db_exp".format(i)) - - assert "TOO_MANY_DATABASES" in str(exp_info) + assert "TOO_MANY_DATABASES" in node.query_and_get_error( + "create database db_exp".format(i) + ) for i in range(10): node.query("create table t{} (a Int32) Engine = Log".format(i)) @@ -35,13 +37,36 @@ def test_table_db_limit(started_cluster): # This checks that system tables are not accounted in the number of tables. node.query("system flush logs") + # Regular tables for i in range(10): node.query("drop table t{}".format(i)) for i in range(10): node.query("create table t{} (a Int32) Engine = Log".format(i)) - with pytest.raises(QueryRuntimeException) as exp_info: - node.query("create table default.tx (a Int32) Engine = Log") + assert "TOO_MANY_TABLES" in node.query_and_get_error( + "create table default.tx (a Int32) Engine = Log" + ) - assert "TOO_MANY_TABLES" in str(exp_info) + # Replicated tables + for i in range(10): + node.query("drop table t{}".format(i)) + + for i in range(5): + node.query( + "create table t{} (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/t{}', 'r1') order by a".format( + i, i + ) + ) + + assert "Too many replicated tables" in node.query_and_get_error( + "create table tx (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/tx', 'r1') order by a" + ) + + # Checks that replicated tables are also counted as regular tables + for i in range(5, 10): + node.query("create table t{} (a Int32) Engine = Log".format(i)) + + assert "TOO_MANY_TABLES" in node.query_and_get_error( + "create table tx (a Int32) Engine = Log" + ) From b03a296542de52c3cb2b6f309a4bc496e4a70454 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Mon, 28 Oct 2024 23:25:38 +0000 Subject: [PATCH 023/244] Fix right join - disabling PR lead to dup result --- src/Planner/PlannerJoinTree.cpp | 24 +++++++++++++++++------- 1 file changed, 17 insertions(+), 7 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 7889a358d95..834e572b167 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -665,11 +665,15 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres bool is_single_table_expression, bool wrap_read_columns_in_subquery) { - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "table_expression:\n{}", table_expression->dumpTree()); - const auto & query_context = planner_context->getQueryContext(); const auto & settings = query_context->getSettingsRef(); + LOG_DEBUG( + getLogger(__PRETTY_FUNCTION__), + "pr_enabled={} table_expression:\n{}", + settings[Setting::allow_experimental_parallel_reading_from_replicas].toString(), + table_expression->dumpTree()); + auto & table_expression_data = planner_context->getTableExpressionDataOrThrow(table_expression); QueryProcessingStage::Enum from_stage = QueryProcessingStage::Enum::FetchColumns; @@ -914,11 +918,11 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres /// It is just a safety check needed until we have a proper sending plan to replicas. /// If we have a non-trivial storage like View it might create its own Planner inside read(), run findTableForParallelReplicas() /// and find some other table that might be used for reading with parallel replicas. It will lead to errors. - const bool other_table_already_chosen_for_reading_with_parallel_replicas - = planner_context->getGlobalPlannerContext()->parallel_replicas_table - && !table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas; - if (other_table_already_chosen_for_reading_with_parallel_replicas) - planner_context->getMutableQueryContext()->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); + // const bool other_table_already_chosen_for_reading_with_parallel_replicas + // = planner_context->getGlobalPlannerContext()->parallel_replicas_table + // && !table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas; + // if (other_table_already_chosen_for_reading_with_parallel_replicas) + // planner_context->getMutableQueryContext()->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); storage->read( query_plan, @@ -930,6 +934,8 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres max_block_size, max_streams); + LOG_DEBUG(getLogger("dumpQueryPlan"), "\n{}", dumpQueryPlan(query_plan)); + auto parallel_replicas_enabled_for_storage = [](const StoragePtr & table, const Settings & query_settings) { if (!table->isMergeTree()) @@ -1249,6 +1255,8 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_ const ColumnIdentifierSet & outer_scope_columns, PlannerContextPtr & planner_context) { + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Join expression: {}", join_table_expression->dumpTree()); + auto & join_node = join_table_expression->as(); if (left_join_tree_query_plan.from_stage != QueryProcessingStage::FetchColumns) throw Exception(ErrorCodes::UNSUPPORTED_METHOD, @@ -1921,6 +1929,8 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, "Expected 1 query plan for JOIN TREE. Actual {}", query_plans_stack.size()); + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "JOIN query plan:\n{}", dumpQueryPlan(query_plans_stack.back().query_plan)); + return std::move(query_plans_stack.back()); } From dc976c48d284fa79ad05fe234130ed3794522511 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Mon, 28 Oct 2024 23:36:57 +0000 Subject: [PATCH 024/244] Test --- .../03254_pr_join_on_dups.reference | 273 ++++++++++++++++++ .../0_stateless/03254_pr_join_on_dups.sql | 81 ++++++ 2 files changed, 354 insertions(+) create mode 100644 tests/queries/0_stateless/03254_pr_join_on_dups.reference create mode 100644 tests/queries/0_stateless/03254_pr_join_on_dups.sql diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.reference b/tests/queries/0_stateless/03254_pr_join_on_dups.reference new file mode 100644 index 00000000000..58602bafb5d --- /dev/null +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.reference @@ -0,0 +1,273 @@ +inner +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +9 l9 \N 9 r9 nr9 +inner subs +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +9 l9 \N 9 r9 nr9 +inner expr +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +9 l9 \N 9 r9 nr9 +left +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 +left subs +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 +left expr +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 +right +0 \N 6 r7 nr7 +0 \N 7 r8 nr8 +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +9 l9 \N 9 r9 nr9 +right subs +0 \N 6 r7 nr7 +0 \N 7 r8 nr8 +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +9 l9 \N 9 r9 nr9 +full +0 \N 6 r7 nr7 +0 \N 7 r8 nr8 +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 +full subs +0 \N 6 r7 nr7 +0 \N 7 r8 nr8 +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 +self inner +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l2 2 2 l3 3 +2 l3 3 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 4 l5 \N +4 l5 \N 4 l6 \N +4 l6 \N 4 l5 \N +4 l6 \N 4 l6 \N +5 l7 \N 5 l7 \N +8 l8 \N 8 l8 \N +9 l9 \N 9 l9 \N +self inner nullable +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +self inner nullable vs not nullable +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l2 2 +3 l4 4 2 l3 3 +4 l5 \N 3 l4 4 +4 l6 \N 3 l4 4 +self inner nullable vs not nullable 2 +4 r6 nr6 4 r6 nr6 +6 r7 nr7 6 r7 nr7 +7 r8 nr8 7 r8 nr8 +9 r9 nr9 9 r9 nr9 +self left +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l2 2 2 l3 3 +2 l3 3 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 4 l5 \N +4 l5 \N 4 l6 \N +4 l6 \N 4 l5 \N +4 l6 \N 4 l6 \N +5 l7 \N 5 l7 \N +8 l8 \N 8 l8 \N +9 l9 \N 9 l9 \N +self left nullable +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 0 \N +4 l6 \N 0 \N +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 0 \N +self left nullable vs not nullable +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l2 2 +3 l4 4 2 l3 3 +4 l5 \N 3 l4 4 +4 l6 \N 3 l4 4 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 0 \N +self left nullable vs not nullable 2 +1 r1 \N 0 \N +1 r2 \N 0 \N +2 r3 \N 0 \N +3 r4 \N 0 \N +3 r5 \N 0 \N +4 r6 nr6 4 r6 nr6 +6 r7 nr7 6 r7 nr7 +7 r8 nr8 7 r8 nr8 +9 r9 nr9 9 r9 nr9 +self right +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l2 2 2 l3 3 +2 l3 3 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 4 l5 \N +4 l5 \N 4 l6 \N +4 l6 \N 4 l5 \N +4 l6 \N 4 l6 \N +5 l7 \N 5 l7 \N +8 l8 \N 8 l8 \N +9 l9 \N 9 l9 \N +self right nullable +0 \N 4 l5 \N +0 \N 4 l6 \N +0 \N 5 l7 \N +0 \N 8 l8 \N +0 \N 9 l9 \N +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +self right nullable vs not nullable +0 \N 4 l5 \N +0 \N 4 l6 \N +0 \N 5 l7 \N +0 \N 8 l8 \N +0 \N 9 l9 \N +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l2 2 +3 l4 4 2 l3 3 +4 l5 \N 3 l4 4 +4 l6 \N 3 l4 4 +self full +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l2 2 2 l3 3 +2 l3 3 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 4 l5 \N +4 l5 \N 4 l6 \N +4 l6 \N 4 l5 \N +4 l6 \N 4 l6 \N +5 l7 \N 5 l7 \N +8 l8 \N 8 l8 \N +9 l9 \N 9 l9 \N +self full nullable +0 \N 4 l5 \N +0 \N 4 l6 \N +0 \N 5 l7 \N +0 \N 8 l8 \N +0 \N 9 l9 \N +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 0 \N +4 l6 \N 0 \N +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 0 \N +self full nullable vs not nullable +0 \N 4 l5 \N +0 \N 4 l6 \N +0 \N 5 l7 \N +0 \N 8 l8 \N +0 \N 9 l9 \N +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l2 2 +3 l4 4 2 l3 3 +4 l5 \N 3 l4 4 +4 l6 \N 3 l4 4 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 0 \N diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.sql b/tests/queries/0_stateless/03254_pr_join_on_dups.sql new file mode 100644 index 00000000000..71695c0d486 --- /dev/null +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.sql @@ -0,0 +1,81 @@ +drop table if exists X sync; +drop table if exists Y sync; + +create table X (id Int32, x_a String, x_b Nullable(Int32)) engine ReplicatedMergeTree('/clickhouse/{database}/X', '1') order by tuple(); +create table Y (id Int32, y_a String, y_b Nullable(String)) engine ReplicatedMergeTree('/clickhouse/{database}/Y', '1') order by tuple(); + +insert into X (id, x_a, x_b) values (1, 'l1', 1), (2, 'l2', 2), (2, 'l3', 3), (3, 'l4', 4); +insert into X (id, x_a) values (4, 'l5'), (4, 'l6'), (5, 'l7'), (8, 'l8'), (9, 'l9'); +insert into Y (id, y_a) values (1, 'r1'), (1, 'r2'), (2, 'r3'), (3, 'r4'), (3, 'r5'); +insert into Y (id, y_a, y_b) values (4, 'r6', 'nr6'), (6, 'r7', 'nr7'), (7, 'r8', 'nr8'), (9, 'r9', 'nr9'); + +set enable_parallel_replicas = 2, max_parallel_replicas = 3, parallel_replicas_for_non_replicated_merge_tree = 1, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost', parallel_replicas_local_plan=1; + +select 'inner'; +select X.*, Y.* from X inner join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; +select 'inner subs'; +select s.*, j.* from (select * from X) as s inner join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; +select 'inner expr'; +select X.*, Y.* from X inner join Y on (X.id + 1) = (Y.id + 1) order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; + +select 'left'; +select X.*, Y.* from X left join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; +select 'left subs'; +select s.*, j.* from (select * from X) as s left join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; +select 'left expr'; +select X.*, Y.* from X left join Y on (X.id + 1) = (Y.id + 1) order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; + +select 'right'; +select X.*, Y.* from X right join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; +select 'right subs'; +select s.*, j.* from (select * from X) as s right join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; +--select 'right expr'; +--select X.*, Y.* from X right join Y on (X.id + 1) = (Y.id + 1) order by id; + +select 'full'; +select X.*, Y.* from X full join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; +select 'full subs'; +select s.*, j.* from (select * from X) as s full join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; +--select 'full expr'; +--select X.*, Y.* from X full join Y on (X.id + 1) = (Y.id + 1) order by id; + +select 'self inner'; +select X.*, s.* from X inner join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self inner nullable'; +select X.*, s.* from X inner join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self inner nullable vs not nullable'; +select X.*, s.* from X inner join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +-- TODO: s.y_b == '' instead of NULL +select 'self inner nullable vs not nullable 2'; +select Y.*, s.* from Y inner join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by Y.id, Y.y_a, Y.y_b, s.id, s.y_a, s.y_b; + +select 'self left'; +select X.*, s.* from X left join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self left nullable'; +select X.*, s.* from X left join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self left nullable vs not nullable'; +select X.*, s.* from X left join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +-- TODO: s.y_b == '' instead of NULL +select 'self left nullable vs not nullable 2'; +select Y.*, s.* from Y left join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by Y.id, Y.y_a, Y.y_b, s.id, s.y_a, s.y_b; + +select 'self right'; +select X.*, s.* from X right join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self right nullable'; +select X.*, s.* from X right join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self right nullable vs not nullable'; +select X.*, s.* from X right join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +--select 'self right nullable vs not nullable 2'; +--select Y.*, s.* from Y right join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by id; + +select 'self full'; +select X.*, s.* from X full join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self full nullable'; +select X.*, s.* from X full join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self full nullable vs not nullable'; +select X.*, s.* from X full join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +--select 'self full nullable vs not nullable 2'; +--select Y.*, s.* from Y full join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by id; + +-- drop table X; +-- drop table Y; From aaba95ca8ceac01fcc22416a400d52c8a169cafd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=D0=9A=D0=B8=D1=80=D0=B8=D0=BB=D0=BB=20=D0=93=D0=B0=D1=80?= =?UTF-8?q?=D0=B1=D0=B0=D1=80?= Date: Tue, 29 Oct 2024 11:41:37 +0300 Subject: [PATCH 025/244] Simplify and fix limit check --- src/Interpreters/InterpreterCreateQuery.cpp | 51 ++++++--------------- 1 file changed, 15 insertions(+), 36 deletions(-) diff --git a/src/Interpreters/InterpreterCreateQuery.cpp b/src/Interpreters/InterpreterCreateQuery.cpp index f8e85733911..3a6e7bc1653 100644 --- a/src/Interpreters/InterpreterCreateQuery.cpp +++ b/src/Interpreters/InterpreterCreateQuery.cpp @@ -1950,46 +1950,25 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create, void InterpreterCreateQuery::throwIfTooManyEntities(ASTCreateQuery & create, StoragePtr storage) const { + auto check_and_throw = [&](auto setting, CurrentMetrics::Metric metric, String setting_name, String entity_name) + { + UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[setting]; + UInt64 attached_count = CurrentMetrics::get(metric); + if (num_limit > 0 && attached_count >= num_limit) + throw Exception(ErrorCodes::TOO_MANY_TABLES, + "Too many {}. " + "The limit (server configuration parameter `{}`) is set to {}, the current number is {}", + entity_name, setting_name, num_limit, attached_count); + }; + if (auto * replicated_storage = typeid_cast(storage.get())) - { - UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_replicated_table_num_to_throw]; - UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedReplicatedTable); - if (attached_count >= num_limit) - throw Exception(ErrorCodes::TOO_MANY_TABLES, - "Too many replicated tables. " - "The limit (server configuration parameter `max_replicated_table_num_to_throw`) is set to {}, the current number is {}", - num_limit, attached_count); - } + check_and_throw(ServerSetting::max_replicated_table_num_to_throw, CurrentMetrics::AttachedReplicatedTable, "max_replicated_table_num_to_throw", "replicated tables"); else if (create.is_dictionary) - { - UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_dictionary_num_to_throw]; - UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedDictionary); - if (attached_count >= num_limit) - throw Exception(ErrorCodes::TOO_MANY_TABLES, - "Too many dictionaries. " - "The limit (server configuration parameter `max_dictionary_num_to_throw`) is set to {}, the current number is {}", - num_limit, attached_count); - } + check_and_throw(ServerSetting::max_dictionary_num_to_throw, CurrentMetrics::AttachedDictionary, "max_dictionary_num_to_throw", "dictionaries"); else if (create.isView()) - { - UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_view_num_to_throw]; - UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedView); - if (attached_count >= num_limit) - throw Exception(ErrorCodes::TOO_MANY_TABLES, - "Too many views. " - "The limit (server configuration parameter `max_view_num_to_throw`) is set to {}, the current number is {}", - num_limit, attached_count); - } + check_and_throw(ServerSetting::max_view_num_to_throw, CurrentMetrics::AttachedView, "max_view_num_to_throw", "views"); else - { - UInt64 num_limit = getContext()->getGlobalContext()->getServerSettings()[ServerSetting::max_table_num_to_throw]; - UInt64 attached_count = CurrentMetrics::get(CurrentMetrics::AttachedTable); - if (attached_count >= num_limit) - throw Exception(ErrorCodes::TOO_MANY_TABLES, - "Too many tables. " - "The limit (server configuration parameter `max_table_num_to_throw`) is set to {}, the current number is {}", - num_limit, attached_count); - } + check_and_throw(ServerSetting::max_table_num_to_throw, CurrentMetrics::AttachedTable, "max_table_num_to_throw", "tables"); } From 19c95b2f0e52bd3794d160605e24c59abc5101b1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=D0=9A=D0=B8=D1=80=D0=B8=D0=BB=D0=BB=20=D0=93=D0=B0=D1=80?= =?UTF-8?q?=D0=B1=D0=B0=D1=80?= Date: Tue, 29 Oct 2024 11:44:50 +0300 Subject: [PATCH 026/244] Test dictionaries --- .../test_table_db_num_limit/config/config.xml | 1 + tests/integration/test_table_db_num_limit/test.py | 12 ++++++++++++ 2 files changed, 13 insertions(+) diff --git a/tests/integration/test_table_db_num_limit/config/config.xml b/tests/integration/test_table_db_num_limit/config/config.xml index a4246c79694..bfe50325d3f 100644 --- a/tests/integration/test_table_db_num_limit/config/config.xml +++ b/tests/integration/test_table_db_num_limit/config/config.xml @@ -10,6 +10,7 @@ + 10 10 5 10 diff --git a/tests/integration/test_table_db_num_limit/test.py b/tests/integration/test_table_db_num_limit/test.py index ce981ffca3c..bcfa60e48cd 100644 --- a/tests/integration/test_table_db_num_limit/test.py +++ b/tests/integration/test_table_db_num_limit/test.py @@ -48,6 +48,18 @@ def test_table_db_limit(started_cluster): "create table default.tx (a Int32) Engine = Log" ) + # Dictionaries + for i in range(10): + node.query( + "create dictionary d{} (a Int32) primary key a source(null()) layout(flat()) lifetime(1000)".format( + i + ) + ) + + assert "TOO_MANY_TABLES" in node.query_and_get_error( + "create dictionary dx (a Int32) primary key a source(null()) layout(flat()) lifetime(1000)" + ) + # Replicated tables for i in range(10): node.query("drop table t{}".format(i)) From e2c2e67c7b4915da6302a516826573cf1ccee701 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Tue, 29 Oct 2024 10:02:24 +0000 Subject: [PATCH 027/244] Fix --- src/Planner/findParallelReplicasQuery.cpp | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 891e5034f44..58a7f48ee2b 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -423,7 +423,10 @@ const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tr return nullptr; const auto * res = findTableForParallelReplicas(query_tree_node.get()); - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Table found {}", res->getStorageID().getFullTableName()); + if (res) + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Table found {}", res->getStorageID().getFullTableName()); + else + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not table found"); return res; } From c7fce84729435f98222d0e02ba035cdd6085a0df Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Tue, 29 Oct 2024 12:17:46 +0000 Subject: [PATCH 028/244] Cleanup --- src/Interpreters/ClusterProxy/executeQuery.cpp | 4 ++-- src/Planner/Planner.cpp | 2 +- src/Planner/PlannerJoinTree.cpp | 16 +++++----------- src/Planner/findParallelReplicasQuery.cpp | 4 ++-- src/Planner/findQueryForParallelReplicas.h | 4 ++-- .../03173_parallel_replicas_join_bug.sh | 3 +++ 6 files changed, 15 insertions(+), 18 deletions(-) diff --git a/src/Interpreters/ClusterProxy/executeQuery.cpp b/src/Interpreters/ClusterProxy/executeQuery.cpp index 4b1f3094be3..e88fdeb0379 100644 --- a/src/Interpreters/ClusterProxy/executeQuery.cpp +++ b/src/Interpreters/ClusterProxy/executeQuery.cpp @@ -477,8 +477,8 @@ void executeQueryWithParallelReplicas( QueryPlanStepPtr analyzed_read_from_merge_tree) { auto logger = getLogger("executeQueryWithParallelReplicas"); - LOG_DEBUG(logger, "Executing read from {}, header {}, query ({}), stage {} with parallel replicas\n{}", - storage_id.getNameForLogs(), header.dumpStructure(), query_ast->formatForLogging(), processed_stage, StackTrace().toString()); + LOG_DEBUG(logger, "Executing read from {}, header {}, query ({}), stage {} with parallel replicas", + storage_id.getNameForLogs(), header.dumpStructure(), query_ast->formatForLogging(), processed_stage); const auto & settings = context->getSettingsRef(); diff --git a/src/Planner/Planner.cpp b/src/Planner/Planner.cpp index 8d3c75fdabb..17277dfe8cd 100644 --- a/src/Planner/Planner.cpp +++ b/src/Planner/Planner.cpp @@ -274,7 +274,7 @@ FiltersForTableExpressionMap collectFiltersForAnalysis(const QueryTreeNodePtr & return res; } -FiltersForTableExpressionMap collectFiltersForAnalysis(const QueryTreeNodePtr & query_tree_node, SelectQueryOptions & select_query_options) +FiltersForTableExpressionMap collectFiltersForAnalysis(const QueryTreeNodePtr & query_tree_node, const SelectQueryOptions & select_query_options) { if (select_query_options.only_analyze) return {}; diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 834e572b167..5c08cc27aff 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -918,11 +918,11 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres /// It is just a safety check needed until we have a proper sending plan to replicas. /// If we have a non-trivial storage like View it might create its own Planner inside read(), run findTableForParallelReplicas() /// and find some other table that might be used for reading with parallel replicas. It will lead to errors. - // const bool other_table_already_chosen_for_reading_with_parallel_replicas - // = planner_context->getGlobalPlannerContext()->parallel_replicas_table - // && !table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas; - // if (other_table_already_chosen_for_reading_with_parallel_replicas) - // planner_context->getMutableQueryContext()->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); + const bool other_table_already_chosen_for_reading_with_parallel_replicas + = planner_context->getGlobalPlannerContext()->parallel_replicas_table + && !table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas; + if (other_table_already_chosen_for_reading_with_parallel_replicas) + planner_context->getMutableQueryContext()->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); storage->read( query_plan, @@ -934,8 +934,6 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres max_block_size, max_streams); - LOG_DEBUG(getLogger("dumpQueryPlan"), "\n{}", dumpQueryPlan(query_plan)); - auto parallel_replicas_enabled_for_storage = [](const StoragePtr & table, const Settings & query_settings) { if (!table->isMergeTree()) @@ -1255,8 +1253,6 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_ const ColumnIdentifierSet & outer_scope_columns, PlannerContextPtr & planner_context) { - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Join expression: {}", join_table_expression->dumpTree()); - auto & join_node = join_table_expression->as(); if (left_join_tree_query_plan.from_stage != QueryProcessingStage::FetchColumns) throw Exception(ErrorCodes::UNSUPPORTED_METHOD, @@ -1929,8 +1925,6 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, "Expected 1 query plan for JOIN TREE. Actual {}", query_plans_stack.size()); - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "JOIN query plan:\n{}", dumpQueryPlan(query_plans_stack.back().query_plan)); - return std::move(query_plans_stack.back()); } diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 58a7f48ee2b..d92500e82fc 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -250,7 +250,7 @@ const QueryNode * findQueryForParallelReplicas( return res; } -const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tree_node, SelectQueryOptions & select_query_options) +const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tree_node, const SelectQueryOptions & select_query_options) { if (select_query_options.only_analyze) return nullptr; @@ -404,7 +404,7 @@ static const TableNode * findTableForParallelReplicas(const IQueryTreeNode * que return nullptr; } -const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tree_node, SelectQueryOptions & select_query_options) +const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tree_node, const SelectQueryOptions & select_query_options) { if (select_query_options.only_analyze) return nullptr; diff --git a/src/Planner/findQueryForParallelReplicas.h b/src/Planner/findQueryForParallelReplicas.h index cdce4ad0b47..83aa11c8c64 100644 --- a/src/Planner/findQueryForParallelReplicas.h +++ b/src/Planner/findQueryForParallelReplicas.h @@ -15,10 +15,10 @@ struct SelectQueryOptions; /// Find a query which can be executed with parallel replicas up to WithMergableStage. /// Returned query will always contain some (>1) subqueries, possibly with joins. -const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tree_node, SelectQueryOptions & select_query_options); +const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tree_node, const SelectQueryOptions & select_query_options); /// Find a table from which we should read on follower replica. It's the left-most table within all JOINs and UNIONs. -const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tree_node, SelectQueryOptions & select_query_options); +const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tree_node, const SelectQueryOptions & select_query_options); struct JoinTreeQueryPlan; diff --git a/tests/queries/0_stateless/03173_parallel_replicas_join_bug.sh b/tests/queries/0_stateless/03173_parallel_replicas_join_bug.sh index 289a49c72f4..1ee3d729cb4 100755 --- a/tests/queries/0_stateless/03173_parallel_replicas_join_bug.sh +++ b/tests/queries/0_stateless/03173_parallel_replicas_join_bug.sh @@ -6,12 +6,15 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) $CLICKHOUSE_CLIENT -q " + DROP TABLE IF EXISTS ids; CREATE TABLE ids (id UUID, whatever String) Engine=MergeTree ORDER BY tuple(); INSERT INTO ids VALUES ('a1451105-722e-4fe7-bfaa-65ad2ae249c2', 'whatever'); + DROP TABLE IF EXISTS data; CREATE TABLE data (id UUID, event_time DateTime, status String) Engine=MergeTree ORDER BY tuple(); INSERT INTO data VALUES ('a1451105-722e-4fe7-bfaa-65ad2ae249c2', '2000-01-01', 'CREATED'); + DROP TABLE IF EXISTS data2; CREATE TABLE data2 (id UUID, event_time DateTime, status String) Engine=MergeTree ORDER BY tuple(); INSERT INTO data2 VALUES ('a1451105-722e-4fe7-bfaa-65ad2ae249c2', '2000-01-02', 'CREATED'); " From 772209e6c0bd0a124d6605a6fe6ef873df8ec161 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=D0=9A=D0=B8=D1=80=D0=B8=D0=BB=D0=BB=20=D0=93=D0=B0=D1=80?= =?UTF-8?q?=D0=B1=D0=B0=D1=80?= Date: Tue, 29 Oct 2024 16:23:21 +0300 Subject: [PATCH 029/244] Test other replica and cleanup --- .../test_table_db_num_limit/config/config.xml | 5 ++- .../config/config1.xml | 4 ++ .../config/config2.xml | 4 ++ .../test_table_db_num_limit/test.py | 40 +++++++++++++++++-- 4 files changed, 48 insertions(+), 5 deletions(-) create mode 100644 tests/integration/test_table_db_num_limit/config/config1.xml create mode 100644 tests/integration/test_table_db_num_limit/config/config2.xml diff --git a/tests/integration/test_table_db_num_limit/config/config.xml b/tests/integration/test_table_db_num_limit/config/config.xml index bfe50325d3f..88438d51b94 100644 --- a/tests/integration/test_table_db_num_limit/config/config.xml +++ b/tests/integration/test_table_db_num_limit/config/config.xml @@ -6,13 +6,16 @@ node1 9000 + + node2 + 9000 + 10 10 - 5 10 diff --git a/tests/integration/test_table_db_num_limit/config/config1.xml b/tests/integration/test_table_db_num_limit/config/config1.xml new file mode 100644 index 00000000000..73b695f3cd6 --- /dev/null +++ b/tests/integration/test_table_db_num_limit/config/config1.xml @@ -0,0 +1,4 @@ + + 5 + + diff --git a/tests/integration/test_table_db_num_limit/config/config2.xml b/tests/integration/test_table_db_num_limit/config/config2.xml new file mode 100644 index 00000000000..e46ca03d70f --- /dev/null +++ b/tests/integration/test_table_db_num_limit/config/config2.xml @@ -0,0 +1,4 @@ + + 3 + + diff --git a/tests/integration/test_table_db_num_limit/test.py b/tests/integration/test_table_db_num_limit/test.py index bcfa60e48cd..53a644a262c 100644 --- a/tests/integration/test_table_db_num_limit/test.py +++ b/tests/integration/test_table_db_num_limit/test.py @@ -7,7 +7,15 @@ cluster = ClickHouseCluster(__file__) node = cluster.add_instance( "node1", with_zookeeper=True, - main_configs=["config/config.xml"], + macros={"replica": "r1"}, + main_configs=["config/config.xml", "config/config1.xml"], +) + +node2 = cluster.add_instance( + "node2", + with_zookeeper=True, + macros={"replica": "r2"}, + main_configs=["config/config.xml", "config/config2.xml"], ) @@ -64,15 +72,27 @@ def test_table_db_limit(started_cluster): for i in range(10): node.query("drop table t{}".format(i)) - for i in range(5): + for i in range(3): node.query( - "create table t{} (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/t{}', 'r1') order by a".format( + "create table t{} on cluster 'cluster' (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/t{}', '{{replica}}') order by a".format( + i, i + ) + ) + + # Test limit on other replica + assert "Too many replicated tables" in node2.query_and_get_error( + "create table tx (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/tx', '{replica}') order by a" + ) + + for i in range(3, 5): + node.query( + "create table t{} (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/t{}', '{{replica}}') order by a".format( i, i ) ) assert "Too many replicated tables" in node.query_and_get_error( - "create table tx (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/tx', 'r1') order by a" + "create table tx (a Int32) Engine = ReplicatedMergeTree('/clickhouse/tables/tx', '{replica}') order by a" ) # Checks that replicated tables are also counted as regular tables @@ -82,3 +102,15 @@ def test_table_db_limit(started_cluster): assert "TOO_MANY_TABLES" in node.query_and_get_error( "create table tx (a Int32) Engine = Log" ) + + # Cleanup + for i in range(10): + node.query("drop table t{} sync".format(i)) + for i in range(3): + node2.query("drop table t{} sync".format(i)) + node.query("system drop replica 'r1' from ZKPATH '/clickhouse/tables/tx'") + node.query("system drop replica 'r2' from ZKPATH '/clickhouse/tables/tx'") + for i in range(9): + node.query("drop database db{}".format(i)) + for i in range(10): + node.query("drop dictionary d{}".format(i)) From 04f68594dcf3dccc5eaecd542e00073af39777d9 Mon Sep 17 00:00:00 2001 From: Amos Bird Date: Tue, 29 Oct 2024 21:36:43 +0800 Subject: [PATCH 030/244] Print method in clickhouse-compressor --stat. --- programs/compressor/Compressor.cpp | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/programs/compressor/Compressor.cpp b/programs/compressor/Compressor.cpp index 819f16cfd64..fc07a0adc66 100644 --- a/programs/compressor/Compressor.cpp +++ b/programs/compressor/Compressor.cpp @@ -33,12 +33,12 @@ namespace DB namespace { -/// Outputs sizes of uncompressed and compressed blocks for compressed file. +/// Outputs method, sizes of uncompressed and compressed blocks for compressed file. void checkAndWriteHeader(DB::ReadBuffer & in, DB::WriteBuffer & out) { while (!in.eof()) { - in.ignore(16); /// checksum + in.ignore(16); /// checksum char header[COMPRESSED_BLOCK_HEADER_SIZE]; in.readStrict(header, COMPRESSED_BLOCK_HEADER_SIZE); @@ -50,6 +50,13 @@ void checkAndWriteHeader(DB::ReadBuffer & in, DB::WriteBuffer & out) UInt32 size_decompressed = unalignedLoad(&header[5]); + auto method_byte = static_cast(header[0]); + auto method = magic_enum::enum_cast(method_byte); + if (method) + DB::writeText(magic_enum::enum_name(*method), out); + else + DB::writeText(fmt::format("UNKNOWN({})", method_byte), out); + DB::writeChar('\t', out); DB::writeText(size_decompressed, out); DB::writeChar('\t', out); DB::writeText(size_compressed, out); From 0fda9bf238d261269b2dd7f47c79898ceaf931cb Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Tue, 29 Oct 2024 14:38:57 +0000 Subject: [PATCH 031/244] Fix 03080_incorrect_join_with_merge.sql --- src/Storages/buildQueryTreeForShard.cpp | 4 ++-- .../queries/0_stateless/03080_incorrect_join_with_merge.sql | 5 +++++ 2 files changed, 7 insertions(+), 2 deletions(-) diff --git a/src/Storages/buildQueryTreeForShard.cpp b/src/Storages/buildQueryTreeForShard.cpp index df9bfd049fb..8d8af134a05 100644 --- a/src/Storages/buildQueryTreeForShard.cpp +++ b/src/Storages/buildQueryTreeForShard.cpp @@ -366,8 +366,8 @@ QueryTreeNodePtr buildQueryTreeForShard(const PlannerContextPtr & planner_contex { QueryTreeNodePtr join_table_expression; const auto join_kind = join_node->getKind(); - const auto join_strictness = join_node->getStrictness(); - if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner && join_strictness == JoinStrictness::All)) + // const auto join_strictness = join_node->getStrictness(); + if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner /* && join_strictness == JoinStrictness::All*/)) { join_table_expression = join_node->getRightTableExpression(); } diff --git a/tests/queries/0_stateless/03080_incorrect_join_with_merge.sql b/tests/queries/0_stateless/03080_incorrect_join_with_merge.sql index a34c71a44e2..a743c5bdffb 100644 --- a/tests/queries/0_stateless/03080_incorrect_join_with_merge.sql +++ b/tests/queries/0_stateless/03080_incorrect_join_with_merge.sql @@ -2,6 +2,7 @@ SET enable_analyzer=1; SET distributed_foreground_insert=1; +DROP TABLE IF EXISTS first_table_lr SYNC; CREATE TABLE first_table_lr ( id String, @@ -11,6 +12,7 @@ ENGINE = ReplicatedMergeTree('/clickhouse/tables/{database}/test_03080/alter', ' ORDER BY id; +DROP TABLE IF EXISTS first_table; CREATE TABLE first_table ( id String, @@ -19,6 +21,7 @@ CREATE TABLE first_table ENGINE = Distributed('test_shard_localhost', currentDatabase(), 'first_table_lr'); +DROP TABLE IF EXISTS second_table_lr; CREATE TABLE second_table_lr ( id String, @@ -26,6 +29,7 @@ CREATE TABLE second_table_lr ) ENGINE = MergeTree() ORDER BY id; +DROP TABLE IF EXISTS second_table; CREATE TABLE second_table ( id String, @@ -36,6 +40,7 @@ ENGINE = Distributed('test_shard_localhost', currentDatabase(), 'second_table_lr INSERT INTO first_table VALUES ('1', '2'), ('3', '4'); INSERT INTO second_table VALUES ('1', '2'), ('3', '4'); +DROP TABLE IF EXISTS two_tables; CREATE TABLE two_tables ( id String, From 66f750ea6f12c08f99c7fecea700d8c7f1eaeeb7 Mon Sep 17 00:00:00 2001 From: vdimir Date: Tue, 29 Oct 2024 15:12:03 +0000 Subject: [PATCH 032/244] remove debug logs --- src/Interpreters/Aggregator.cpp | 3 --- src/Parsers/CreateQueryUUIDs.cpp | 2 +- 2 files changed, 1 insertion(+), 4 deletions(-) diff --git a/src/Interpreters/Aggregator.cpp b/src/Interpreters/Aggregator.cpp index 2dd6513d498..bb9e22e5a1b 100644 --- a/src/Interpreters/Aggregator.cpp +++ b/src/Interpreters/Aggregator.cpp @@ -1504,7 +1504,6 @@ bool Aggregator::executeOnBlock(Columns columns, && worth_convert_to_two_level) { size_t size = current_memory_usage + params.min_free_disk_space; - LOG_DEBUG(&Poco::Logger::get("XXXX"), "{}:{}: {} {}", __FILE__, __LINE__, current_memory_usage, params.min_free_disk_space); writeToTemporaryFile(result, size); } @@ -1521,7 +1520,6 @@ void Aggregator::writeToTemporaryFile(AggregatedDataVariants & data_variants, si size_t rows = data_variants.size(); std::unique_lock lk(tmp_files_mutex); - LOG_DEBUG(&Poco::Logger::get("XXXX"), "{}:{}: max_temp_file_size {}", __FILE__, __LINE__, max_temp_file_size); auto & out_stream = tmp_files.emplace_back(getHeader(false), tmp_data.get(), max_temp_file_size); lk.unlock(); @@ -2934,7 +2932,6 @@ bool Aggregator::mergeOnBlock(Block block, AggregatedDataVariants & result, bool && worth_convert_to_two_level) { size_t size = current_memory_usage + params.min_free_disk_space; - LOG_DEBUG(&Poco::Logger::get("XXXX"), "{}:{}: {} {}", __FILE__, __LINE__, current_memory_usage, params.min_free_disk_space); writeToTemporaryFile(result, size); } diff --git a/src/Parsers/CreateQueryUUIDs.cpp b/src/Parsers/CreateQueryUUIDs.cpp index c788cc7a025..14cf5761a11 100644 --- a/src/Parsers/CreateQueryUUIDs.cpp +++ b/src/Parsers/CreateQueryUUIDs.cpp @@ -31,7 +31,7 @@ CreateQueryUUIDs::CreateQueryUUIDs(const ASTCreateQuery & query, bool generate_r /// If we generate random UUIDs for already existing tables then those UUIDs will not be correct making those inner target table inaccessible. /// Thus it's not safe for example to replace /// "ATTACH MATERIALIZED VIEW mv AS SELECT a FROM b" with - /// "ATTACH MATERIALIZED VIEW mv TO INNER UUID "XXXX" AS SELECT a FROM b" + /// "ATTACH MATERIALIZED VIEW mv TO INNER UUID "248372b7-02c4-4c88-a5e1-282a83cc572a" AS SELECT a FROM b" /// This replacement is safe only for CREATE queries when inner target tables don't exist yet. if (!query.attach) { From bebef8d0d96e27c9823419b3a7f669d62c6a6a56 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Tue, 29 Oct 2024 23:58:39 +0000 Subject: [PATCH 033/244] Fix right joins again --- src/Planner/PlannerJoinTree.cpp | 38 +++++++++++++++++++++++---------- 1 file changed, 27 insertions(+), 11 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 5c08cc27aff..0007dc9d158 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -665,7 +665,7 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres bool is_single_table_expression, bool wrap_read_columns_in_subquery) { - const auto & query_context = planner_context->getQueryContext(); + auto query_context = planner_context->getQueryContext(); const auto & settings = query_context->getSettingsRef(); LOG_DEBUG( @@ -922,17 +922,33 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres = planner_context->getGlobalPlannerContext()->parallel_replicas_table && !table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas; if (other_table_already_chosen_for_reading_with_parallel_replicas) - planner_context->getMutableQueryContext()->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); + { + chassert(query_context->canUseParallelReplicasOnFollower()); - storage->read( - query_plan, - columns_names, - storage_snapshot, - table_expression_query_info, - query_context, - from_stage, - max_block_size, - max_streams); + auto mutable_context = Context::createCopy(query_context); + mutable_context->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); + storage->read( + query_plan, + columns_names, + storage_snapshot, + table_expression_query_info, + mutable_context, + from_stage, + max_block_size, + max_streams); + } + else + { + storage->read( + query_plan, + columns_names, + storage_snapshot, + table_expression_query_info, + query_context, + from_stage, + max_block_size, + max_streams); + } auto parallel_replicas_enabled_for_storage = [](const StoragePtr & table, const Settings & query_settings) { From c5d6acf5e3ff24122518abb992e78c73954f8703 Mon Sep 17 00:00:00 2001 From: Amos Bird Date: Wed, 30 Oct 2024 09:00:18 +0800 Subject: [PATCH 034/244] Fix --- programs/compressor/Compressor.cpp | 3 +++ 1 file changed, 3 insertions(+) diff --git a/programs/compressor/Compressor.cpp b/programs/compressor/Compressor.cpp index fc07a0adc66..69936912d49 100644 --- a/programs/compressor/Compressor.cpp +++ b/programs/compressor/Compressor.cpp @@ -1,3 +1,6 @@ +/// For magic_enum to properly get enum name of DB::CompressionMethodByte +#define MAGIC_ENUM_RANGE_MAX 256 + #include #include #include From 10ee24d9a0c749624b86840f78b2bd3bbdb221d9 Mon Sep 17 00:00:00 2001 From: Amos Bird Date: Wed, 30 Oct 2024 09:41:18 +0800 Subject: [PATCH 035/244] Fix multiple codecs and add test --- programs/compressor/Compressor.cpp | 26 +++++----------- .../getCompressionCodecForFile.cpp | 31 ++++++++++++++----- src/Compression/getCompressionCodecForFile.h | 4 +++ .../03260_compressor_stat.reference | 1 + .../0_stateless/03260_compressor_stat.sh | 13 ++++++++ 5 files changed, 49 insertions(+), 26 deletions(-) create mode 100644 tests/queries/0_stateless/03260_compressor_stat.reference create mode 100755 tests/queries/0_stateless/03260_compressor_stat.sh diff --git a/programs/compressor/Compressor.cpp b/programs/compressor/Compressor.cpp index 69936912d49..7bb434d40a8 100644 --- a/programs/compressor/Compressor.cpp +++ b/programs/compressor/Compressor.cpp @@ -1,6 +1,3 @@ -/// For magic_enum to properly get enum name of DB::CompressionMethodByte -#define MAGIC_ENUM_RANGE_MAX 256 - #include #include #include @@ -14,9 +11,12 @@ #include #include #include +#include +#include #include #include #include +#include #include #include #include @@ -41,31 +41,19 @@ void checkAndWriteHeader(DB::ReadBuffer & in, DB::WriteBuffer & out) { while (!in.eof()) { - in.ignore(16); /// checksum - - char header[COMPRESSED_BLOCK_HEADER_SIZE]; - in.readStrict(header, COMPRESSED_BLOCK_HEADER_SIZE); - - UInt32 size_compressed = unalignedLoad(&header[1]); + UInt32 size_compressed; + UInt32 size_decompressed; + auto codec = DB::getCompressionCodecForFile(in, size_compressed, size_decompressed, true /* skip_to_next_block */); if (size_compressed > DBMS_MAX_COMPRESSED_SIZE) throw DB::Exception(DB::ErrorCodes::TOO_LARGE_SIZE_COMPRESSED, "Too large size_compressed. Most likely corrupted data."); - UInt32 size_decompressed = unalignedLoad(&header[5]); - - auto method_byte = static_cast(header[0]); - auto method = magic_enum::enum_cast(method_byte); - if (method) - DB::writeText(magic_enum::enum_name(*method), out); - else - DB::writeText(fmt::format("UNKNOWN({})", method_byte), out); + DB::writeText(queryToString(codec->getFullCodecDesc()), out); DB::writeChar('\t', out); DB::writeText(size_decompressed, out); DB::writeChar('\t', out); DB::writeText(size_compressed, out); DB::writeChar('\n', out); - - in.ignore(size_compressed - COMPRESSED_BLOCK_HEADER_SIZE); } } diff --git a/src/Compression/getCompressionCodecForFile.cpp b/src/Compression/getCompressionCodecForFile.cpp index 027ee0ac57a..b04e4b6371a 100644 --- a/src/Compression/getCompressionCodecForFile.cpp +++ b/src/Compression/getCompressionCodecForFile.cpp @@ -10,33 +10,50 @@ namespace DB { - using Checksum = CityHash_v1_0_2::uint128; -CompressionCodecPtr getCompressionCodecForFile(const IDataPartStorage & data_part_storage, const String & relative_path) +CompressionCodecPtr +getCompressionCodecForFile(ReadBuffer & read_buffer, UInt32 & size_compressed, UInt32 & size_decompressed, bool skip_to_next_block) { - auto read_buffer = data_part_storage.readFile(relative_path, {}, std::nullopt, std::nullopt); - read_buffer->ignore(sizeof(Checksum)); + read_buffer.ignore(sizeof(Checksum)); UInt8 header_size = ICompressionCodec::getHeaderSize(); + size_t starting_bytes = read_buffer.count(); PODArray compressed_buffer; compressed_buffer.resize(header_size); - read_buffer->readStrict(compressed_buffer.data(), header_size); + read_buffer.readStrict(compressed_buffer.data(), header_size); uint8_t method = ICompressionCodec::readMethod(compressed_buffer.data()); + size_compressed = unalignedLoad(&compressed_buffer[1]); + size_decompressed = unalignedLoad(&compressed_buffer[5]); if (method == static_cast(CompressionMethodByte::Multiple)) { compressed_buffer.resize(1); - read_buffer->readStrict(compressed_buffer.data(), 1); + read_buffer.readStrict(compressed_buffer.data(), 1); compressed_buffer.resize(1 + compressed_buffer[0]); - read_buffer->readStrict(compressed_buffer.data() + 1, compressed_buffer[0]); + read_buffer.readStrict(compressed_buffer.data() + 1, compressed_buffer[0]); auto codecs_bytes = CompressionCodecMultiple::getCodecsBytesFromData(compressed_buffer.data()); Codecs codecs; for (auto byte : codecs_bytes) codecs.push_back(CompressionCodecFactory::instance().get(byte)); + if (skip_to_next_block) + read_buffer.ignore(size_compressed - (read_buffer.count() - starting_bytes)); + return std::make_shared(codecs); } + + if (skip_to_next_block) + read_buffer.ignore(size_compressed - (read_buffer.count() - starting_bytes)); + return CompressionCodecFactory::instance().get(method); } +CompressionCodecPtr getCompressionCodecForFile(const IDataPartStorage & data_part_storage, const String & relative_path) +{ + auto read_buffer = data_part_storage.readFile(relative_path, {}, std::nullopt, std::nullopt); + UInt32 size_compressed; + UInt32 size_decompressed; + return getCompressionCodecForFile(*read_buffer, size_compressed, size_decompressed, false); +} + } diff --git a/src/Compression/getCompressionCodecForFile.h b/src/Compression/getCompressionCodecForFile.h index b6f22750e4d..535befa37e1 100644 --- a/src/Compression/getCompressionCodecForFile.h +++ b/src/Compression/getCompressionCodecForFile.h @@ -13,4 +13,8 @@ namespace DB /// from metadata. CompressionCodecPtr getCompressionCodecForFile(const IDataPartStorage & data_part_storage, const String & relative_path); +/// Same as above which is used by clickhouse-compressor to print compression statistics of each data block. +CompressionCodecPtr +getCompressionCodecForFile(ReadBuffer & read_buffer, UInt32 & size_compressed, UInt32 & size_decompressed, bool skip_to_next_block); + } diff --git a/tests/queries/0_stateless/03260_compressor_stat.reference b/tests/queries/0_stateless/03260_compressor_stat.reference new file mode 100644 index 00000000000..ba84b26cc48 --- /dev/null +++ b/tests/queries/0_stateless/03260_compressor_stat.reference @@ -0,0 +1 @@ +CODEC(Delta(1), LZ4) 14 48 diff --git a/tests/queries/0_stateless/03260_compressor_stat.sh b/tests/queries/0_stateless/03260_compressor_stat.sh new file mode 100755 index 00000000000..6efa7b6ee0a --- /dev/null +++ b/tests/queries/0_stateless/03260_compressor_stat.sh @@ -0,0 +1,13 @@ +#!/usr/bin/env bash + +CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) +# shellcheck source=../shell_config.sh +. "$CURDIR"/../shell_config.sh + +echo "Hello, World!" > 03260_test_data + +$CLICKHOUSE_COMPRESSOR --codec 'Delta' --codec 'LZ4' --input '03260_test_data' --output '03260_test_out' + +$CLICKHOUSE_COMPRESSOR --stat '03260_test_out' + +rm -f 03260_test_data 03260_test_out From bd9cfaecea93dc3b6d469f3898fcde9506ae5f9b Mon Sep 17 00:00:00 2001 From: Amos Bird Date: Wed, 30 Oct 2024 14:35:06 +0800 Subject: [PATCH 036/244] No need to create tmp files --- tests/queries/0_stateless/03260_compressor_stat.sh | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) diff --git a/tests/queries/0_stateless/03260_compressor_stat.sh b/tests/queries/0_stateless/03260_compressor_stat.sh index 6efa7b6ee0a..8a03541763c 100755 --- a/tests/queries/0_stateless/03260_compressor_stat.sh +++ b/tests/queries/0_stateless/03260_compressor_stat.sh @@ -4,10 +4,4 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) # shellcheck source=../shell_config.sh . "$CURDIR"/../shell_config.sh -echo "Hello, World!" > 03260_test_data - -$CLICKHOUSE_COMPRESSOR --codec 'Delta' --codec 'LZ4' --input '03260_test_data' --output '03260_test_out' - -$CLICKHOUSE_COMPRESSOR --stat '03260_test_out' - -rm -f 03260_test_data 03260_test_out +echo "Hello, World!" | $CLICKHOUSE_COMPRESSOR --codec 'Delta' --codec 'LZ4' | $CLICKHOUSE_COMPRESSOR --stat From 6004cb8ff4fc0b751f9cd0821a4d9214cfd63e3e Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Wed, 30 Oct 2024 11:08:21 +0000 Subject: [PATCH 037/244] Remove current_table_chosen_for_reading_with_parallel_replicas flag --- src/Planner/PlannerJoinTree.cpp | 8 +++----- src/Storages/SelectQueryInfo.h | 2 -- src/Storages/StorageMergeTree.cpp | 4 +--- src/Storages/StorageReplicatedMergeTree.cpp | 5 +---- 4 files changed, 5 insertions(+), 14 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 0007dc9d158..5e29c1a6a81 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -702,8 +702,6 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres table_expression_query_info.table_expression = table_expression; if (const auto & filter_actions = table_expression_data.getFilterActions()) table_expression_query_info.filter_actions_dag = std::make_shared(filter_actions->clone()); - table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas - = table_node == planner_context->getGlobalPlannerContext()->parallel_replicas_table; size_t max_streams = settings[Setting::max_threads]; size_t max_threads_execute_query = settings[Setting::max_threads]; @@ -918,10 +916,10 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres /// It is just a safety check needed until we have a proper sending plan to replicas. /// If we have a non-trivial storage like View it might create its own Planner inside read(), run findTableForParallelReplicas() /// and find some other table that might be used for reading with parallel replicas. It will lead to errors. - const bool other_table_already_chosen_for_reading_with_parallel_replicas + const bool other_table_chosen_for_reading_with_parallel_replicas = planner_context->getGlobalPlannerContext()->parallel_replicas_table - && !table_expression_query_info.current_table_chosen_for_reading_with_parallel_replicas; - if (other_table_already_chosen_for_reading_with_parallel_replicas) + && table_node != planner_context->getGlobalPlannerContext()->parallel_replicas_table; + if (other_table_chosen_for_reading_with_parallel_replicas) { chassert(query_context->canUseParallelReplicasOnFollower()); diff --git a/src/Storages/SelectQueryInfo.h b/src/Storages/SelectQueryInfo.h index 7ad6a733c6f..f67274f227a 100644 --- a/src/Storages/SelectQueryInfo.h +++ b/src/Storages/SelectQueryInfo.h @@ -162,8 +162,6 @@ struct SelectQueryInfo /// It's guaranteed to be present in JOIN TREE of `query_tree` QueryTreeNodePtr table_expression; - bool current_table_chosen_for_reading_with_parallel_replicas = false; - /// Table expression modifiers for storage std::optional table_expression_modifiers; diff --git a/src/Storages/StorageMergeTree.cpp b/src/Storages/StorageMergeTree.cpp index 40cd6e01dba..55f79a54f2e 100644 --- a/src/Storages/StorageMergeTree.cpp +++ b/src/Storages/StorageMergeTree.cpp @@ -276,9 +276,7 @@ void StorageMergeTree::read( } const bool enable_parallel_reading = local_context->canUseParallelReplicasOnFollower() - && local_context->getSettingsRef()[Setting::parallel_replicas_for_non_replicated_merge_tree] - && (!local_context->getSettingsRef()[Setting::allow_experimental_analyzer] - || query_info.current_table_chosen_for_reading_with_parallel_replicas); + && local_context->getSettingsRef()[Setting::parallel_replicas_for_non_replicated_merge_tree]; if (auto plan = reader.read( column_names, diff --git a/src/Storages/StorageReplicatedMergeTree.cpp b/src/Storages/StorageReplicatedMergeTree.cpp index fc3245eafcf..3f1d2bc6a1c 100644 --- a/src/Storages/StorageReplicatedMergeTree.cpp +++ b/src/Storages/StorageReplicatedMergeTree.cpp @@ -5640,10 +5640,7 @@ void StorageReplicatedMergeTree::readLocalImpl( const size_t max_block_size, const size_t num_streams) { - const bool enable_parallel_reading = local_context->canUseParallelReplicasOnFollower() - && (!local_context->getSettingsRef()[Setting::allow_experimental_analyzer] - || query_info.current_table_chosen_for_reading_with_parallel_replicas); - + const bool enable_parallel_reading = local_context->canUseParallelReplicasOnFollower(); auto plan = reader.read( column_names, storage_snapshot, query_info, local_context, max_block_size, num_streams, From b724f2c33141fb0348742d6b48c4b58763450ff7 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Wed, 30 Oct 2024 12:24:56 +0000 Subject: [PATCH 038/244] Fix FULL joins --- src/Planner/PlannerJoinTree.cpp | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 5e29c1a6a81..ac05f893cd2 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -916,13 +916,11 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres /// It is just a safety check needed until we have a proper sending plan to replicas. /// If we have a non-trivial storage like View it might create its own Planner inside read(), run findTableForParallelReplicas() /// and find some other table that might be used for reading with parallel replicas. It will lead to errors. - const bool other_table_chosen_for_reading_with_parallel_replicas - = planner_context->getGlobalPlannerContext()->parallel_replicas_table + const bool no_tables_or_another_table_chosen_for_reading_with_parallel_replicas_mode + = query_context->canUseParallelReplicasOnFollower() && table_node != planner_context->getGlobalPlannerContext()->parallel_replicas_table; - if (other_table_chosen_for_reading_with_parallel_replicas) + if (no_tables_or_another_table_chosen_for_reading_with_parallel_replicas_mode) { - chassert(query_context->canUseParallelReplicasOnFollower()); - auto mutable_context = Context::createCopy(query_context); mutable_context->setSetting("allow_experimental_parallel_reading_from_replicas", Field(0)); storage->read( @@ -984,7 +982,10 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres query_plan = std::move(query_plan_parallel_replicas); } } - else if (ClusterProxy::canUseParallelReplicasOnInitiator(query_context)) + else if ( + ClusterProxy::canUseParallelReplicasOnInitiator(query_context) + && planner_context->getGlobalPlannerContext()->parallel_replicas_node + && planner_context->getGlobalPlannerContext()->parallel_replicas_node == query_node) { // (1) find read step QueryPlan::Node * node = query_plan.getRootNode(); From 8245e3d7ef5530d55763700b0c1aeae1697dd26c Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Wed, 30 Oct 2024 21:23:29 +0000 Subject: [PATCH 039/244] Fix --- src/Planner/PlannerJoinTree.cpp | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index ac05f893cd2..481cb9b8649 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -982,10 +982,7 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres query_plan = std::move(query_plan_parallel_replicas); } } - else if ( - ClusterProxy::canUseParallelReplicasOnInitiator(query_context) - && planner_context->getGlobalPlannerContext()->parallel_replicas_node - && planner_context->getGlobalPlannerContext()->parallel_replicas_node == query_node) + else if (ClusterProxy::canUseParallelReplicasOnInitiator(query_context)) { // (1) find read step QueryPlan::Node * node = query_plan.getRootNode(); From b229fb1664c8ed5b2c19ff569bb94c51e2f8cbec Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Thu, 31 Oct 2024 12:04:24 +0000 Subject: [PATCH 040/244] Check if the mutation query is valid. --- src/Interpreters/MutationsInterpreter.cpp | 3 +++ .../03256_invalid_mutation_query.reference | 0 .../03256_invalid_mutation_query.sql | 19 +++++++++++++++++++ 3 files changed, 22 insertions(+) create mode 100644 tests/queries/0_stateless/03256_invalid_mutation_query.reference create mode 100644 tests/queries/0_stateless/03256_invalid_mutation_query.sql diff --git a/src/Interpreters/MutationsInterpreter.cpp b/src/Interpreters/MutationsInterpreter.cpp index 0f25d5ac21c..da99b217341 100644 --- a/src/Interpreters/MutationsInterpreter.cpp +++ b/src/Interpreters/MutationsInterpreter.cpp @@ -1386,6 +1386,9 @@ void MutationsInterpreter::validate() } } + // Make sure the mutations query is valid + prepareQueryAffectedQueryTree(commands, source.getStorage(), context); + QueryPlan plan; initQueryPlan(stages.front(), plan); diff --git a/tests/queries/0_stateless/03256_invalid_mutation_query.reference b/tests/queries/0_stateless/03256_invalid_mutation_query.reference new file mode 100644 index 00000000000..e69de29bb2d diff --git a/tests/queries/0_stateless/03256_invalid_mutation_query.sql b/tests/queries/0_stateless/03256_invalid_mutation_query.sql new file mode 100644 index 00000000000..010f96414d4 --- /dev/null +++ b/tests/queries/0_stateless/03256_invalid_mutation_query.sql @@ -0,0 +1,19 @@ +DROP TABLE IF EXISTS t; +DROP TABLE IF EXISTS t2; + +CREATE TABLE t (x int) ENGINE = MergeTree() ORDER BY (); + +DELETE FROM t WHERE y in (SELECT y FROM t); -- { serverError 47 } +DELETE FROM t WHERE x in (SELECT y FROM t); -- { serverError 47 } +DELETE FROM t WHERE x IN (SELECT * FROM t2); -- { serverError 60 } +ALTER TABLE t DELETE WHERE x in (SELECT y FROM t); -- { serverError 47 } +ALTER TABLE t UPDATE x = 1 WHERE x IN (SELECT y FROM t); -- { serverError 47 } + +ALTER TABLE t ADD COLUMN y int; +DELETE FROM t WHERE y in (SELECT y FROM t); + +CREATE TABLE t2 (x int) ENGINE = MergeTree() ORDER BY (); +DELETE FROM t WHERE x IN (SELECT * FROM t2); + +DROP TABLE t; +DROP TABLE t2; From 0808d7f0fb96e9f6c6536b9033cf2f7499cbb383 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Thu, 31 Oct 2024 12:26:46 +0000 Subject: [PATCH 041/244] Fix FULL JOINs again --- src/Planner/Planner.cpp | 3 ++- src/Planner/Planner.h | 1 + src/Planner/PlannerJoinTree.cpp | 15 ++++++++++++--- src/Planner/findParallelReplicasQuery.cpp | 8 +------- 4 files changed, 16 insertions(+), 11 deletions(-) diff --git a/src/Planner/Planner.cpp b/src/Planner/Planner.cpp index 17277dfe8cd..260462652fc 100644 --- a/src/Planner/Planner.cpp +++ b/src/Planner/Planner.cpp @@ -1263,6 +1263,7 @@ Planner::Planner(const QueryTreeNodePtr & query_tree_, findQueryForParallelReplicas(query_tree, select_query_options), findTableForParallelReplicas(query_tree, select_query_options), collectFiltersForAnalysis(query_tree, select_query_options)))) + , root_planner(true) { } @@ -1537,7 +1538,7 @@ void Planner::buildPlanForQueryNode() JoinTreeQueryPlan join_tree_query_plan; if (planner_context->getMutableQueryContext()->canUseTaskBasedParallelReplicas() - && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &query_node) + && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &query_node && !root_planner) { join_tree_query_plan = buildQueryPlanForParallelReplicas(query_node, planner_context, select_query_info.storage_limits); } diff --git a/src/Planner/Planner.h b/src/Planner/Planner.h index ae78f05cbd4..bf11c9ef9cd 100644 --- a/src/Planner/Planner.h +++ b/src/Planner/Planner.h @@ -82,6 +82,7 @@ private: StorageLimitsList storage_limits; std::set used_row_policies; QueryNodeToPlanStepMapping query_node_to_plan_step_mapping; + bool root_planner = false; }; } diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 481cb9b8649..160d7f07d5b 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -659,6 +659,7 @@ std::unique_ptr createComputeAliasColumnsStep( } JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expression, + const QueryNode & parent_query_node, const SelectQueryInfo & select_query_info, const SelectQueryOptions & select_query_options, PlannerContextPtr & planner_context, @@ -982,7 +983,10 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres query_plan = std::move(query_plan_parallel_replicas); } } - else if (ClusterProxy::canUseParallelReplicasOnInitiator(query_context)) + else if ( + ClusterProxy::canUseParallelReplicasOnInitiator(query_context) + && planner_context->getGlobalPlannerContext()->parallel_replicas_node + && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &parent_query_node) { // (1) find read step QueryPlan::Node * node = query_plan.getRootNode(); @@ -1815,6 +1819,7 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, const ColumnIdentifierSet & outer_scope_columns, PlannerContextPtr & planner_context) { + const QueryNode & parent_query_node = query_node->as(); auto table_expressions_stack = buildTableExpressionsStack(query_node->as().getJoinTree()); size_t table_expressions_stack_size = table_expressions_stack.size(); bool is_single_table_expression = table_expressions_stack_size == 1; @@ -1850,7 +1855,9 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, * Examples: Distributed, LiveView, Merge storages. */ auto left_table_expression = table_expressions_stack.front(); - auto left_table_expression_query_plan = buildQueryPlanForTableExpression(left_table_expression, + auto left_table_expression_query_plan = buildQueryPlanForTableExpression( + left_table_expression, + parent_query_node, select_query_info, select_query_options, planner_context, @@ -1923,7 +1930,9 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, * table expression in subquery. */ bool is_remote = planner_context->getTableExpressionDataOrThrow(table_expression).isRemote(); - query_plans_stack.push_back(buildQueryPlanForTableExpression(table_expression, + query_plans_stack.push_back(buildQueryPlanForTableExpression( + table_expression, + parent_query_node, select_query_info, select_query_options, planner_context, diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index d92500e82fc..63c0ce8eb68 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -101,17 +101,11 @@ std::stack getSupportingParallelReplicasQuery(const IQueryTre auto join_strictness = join_node.getStrictness(); if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner && join_strictness == JoinStrictness::All)) - { query_tree_node = join_node.getLeftTableExpression().get(); - } else if (join_kind == JoinKind::Right) - { query_tree_node = join_node.getRightTableExpression().get(); - } else - { return {}; - } break; } @@ -275,7 +269,7 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr /// We don't have any subquery and storage can process parallel replicas by itself. if (stack.top() == query_tree_node.get()) - return nullptr; + return query_node; /// This is needed to avoid infinite recursion. auto mutable_context = Context::createCopy(context); From 9015454b37627712eac4eae5126378ae68d8e98c Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 1 Nov 2024 11:06:21 +0000 Subject: [PATCH 042/244] Add setting --- src/Core/Settings.cpp | 5 +++++ src/Core/SettingsChangesHistory.cpp | 1 + src/Interpreters/MutationsInterpreter.cpp | 6 ++++-- 3 files changed, 10 insertions(+), 2 deletions(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index 6c269e22c35..17e2e1cc599 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -3640,6 +3640,11 @@ Given that, for example, dictionaries, can be out of sync across nodes, mutation ``` +)", 0) \ + DECLARE(Bool, validate_mutation_query, true, R"( +Validate mutation queries before accepting them. Mutations are executed in the background, and running an invalid query will cause mutations to get stuck, requiring manual intervention. + +Only change this setting if you encounter a backward-incompatible bug. )", 0) \ DECLARE(Seconds, lock_acquire_timeout, DBMS_DEFAULT_LOCK_ACQUIRE_TIMEOUT_SEC, R"( Defines how many seconds a locking request waits before failing. diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index 3fe3e960dc6..613b9e2281a 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -64,6 +64,7 @@ static std::initializer_listgetSettingsRef()[Setting::validate_mutation_query]) + // Make sure the mutation query is valid + prepareQueryAffectedQueryTree(commands, source.getStorage(), context); QueryPlan plan; From 7691b7dd4435d1df5cd43cdf9169277aa9e81996 Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 1 Nov 2024 11:06:49 +0000 Subject: [PATCH 043/244] Fix test --- .../integration/test_failed_mutations/test.py | 32 +++++++------------ .../03256_invalid_mutation_query.sql | 2 ++ 2 files changed, 14 insertions(+), 20 deletions(-) diff --git a/tests/integration/test_failed_mutations/test.py b/tests/integration/test_failed_mutations/test.py index 5a2bf874da2..8d2ee46e748 100644 --- a/tests/integration/test_failed_mutations/test.py +++ b/tests/integration/test_failed_mutations/test.py @@ -27,6 +27,9 @@ REPLICATED_POSTPONE_MUTATION_LOG = ( POSTPONE_MUTATION_LOG = ( "According to exponential backoff policy, do not perform mutations for the part" ) +FAILING_MUTATION_QUERY = ( + "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT throwIf(1))" +) all_nodes = [node_with_backoff, node_no_backoff] @@ -83,17 +86,13 @@ def test_exponential_backoff_with_merge_tree(started_cluster, node, found_in_log assert not node.contains_in_log(POSTPONE_MUTATION_LOG) # Executing incorrect mutation. - node.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM notexist_table) SETTINGS allow_nondeterministic_mutations=1" - ) + node.query(FAILING_MUTATION_QUERY) check_logs() node.query("KILL MUTATION WHERE table='test_mutations'") # Check that after kill new parts mutations are postponing. - node.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM notexist_table) SETTINGS allow_nondeterministic_mutations=1" - ) + node.query(FAILING_MUTATION_QUERY) check_logs() @@ -101,9 +100,7 @@ def test_exponential_backoff_with_merge_tree(started_cluster, node, found_in_log def test_exponential_backoff_with_replicated_tree(started_cluster): prepare_cluster(True) - node_with_backoff.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM notexist_table) SETTINGS allow_nondeterministic_mutations=1" - ) + node_with_backoff.query(FAILING_MUTATION_QUERY) assert node_with_backoff.wait_for_log_line(REPLICATED_POSTPONE_MUTATION_LOG) assert not node_no_backoff.contains_in_log(REPLICATED_POSTPONE_MUTATION_LOG) @@ -114,7 +111,7 @@ def test_exponential_backoff_create_dependent_table(started_cluster): # Executing incorrect mutation. node_with_backoff.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS allow_nondeterministic_mutations=1" + "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS validate_mutation_query = 0" ) # Creating dependent table for mutation. @@ -148,9 +145,7 @@ def test_exponential_backoff_setting_override(started_cluster): node.query("INSERT INTO test_mutations SELECT * FROM system.numbers LIMIT 10") # Executing incorrect mutation. - node.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS allow_nondeterministic_mutations=1" - ) + node.query(FAILING_MUTATION_QUERY) assert not node.contains_in_log(POSTPONE_MUTATION_LOG) @@ -166,9 +161,7 @@ def test_backoff_clickhouse_restart(started_cluster, replicated_table): node = node_with_backoff # Executing incorrect mutation. - node.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS allow_nondeterministic_mutations=1" - ) + node.query(FAILING_MUTATION_QUERY) assert node.wait_for_log_line( REPLICATED_POSTPONE_MUTATION_LOG if replicated_table else POSTPONE_MUTATION_LOG ) @@ -193,11 +186,10 @@ def test_no_backoff_after_killing_mutation(started_cluster, replicated_table): node = node_with_backoff # Executing incorrect mutation. - node.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS allow_nondeterministic_mutations=1" - ) + node.query(FAILING_MUTATION_QUERY) + # Executing correct mutation. - node.query("ALTER TABLE test_mutations DELETE WHERE x=1") + node.query("ALTER TABLE test_mutations DELETE WHERE x=1") assert node.wait_for_log_line( REPLICATED_POSTPONE_MUTATION_LOG if replicated_table else POSTPONE_MUTATION_LOG ) diff --git a/tests/queries/0_stateless/03256_invalid_mutation_query.sql b/tests/queries/0_stateless/03256_invalid_mutation_query.sql index 010f96414d4..2c554cabb9e 100644 --- a/tests/queries/0_stateless/03256_invalid_mutation_query.sql +++ b/tests/queries/0_stateless/03256_invalid_mutation_query.sql @@ -9,6 +9,8 @@ DELETE FROM t WHERE x IN (SELECT * FROM t2); -- { serverError 60 } ALTER TABLE t DELETE WHERE x in (SELECT y FROM t); -- { serverError 47 } ALTER TABLE t UPDATE x = 1 WHERE x IN (SELECT y FROM t); -- { serverError 47 } +DELETE FROM t WHERE x IN (SELECT foo FROM bar) SETTINGS validate_mutation_query = 0; + ALTER TABLE t ADD COLUMN y int; DELETE FROM t WHERE y in (SELECT y FROM t); From 752dfead2c5fc686b64d062b7f032196657295ff Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 1 Nov 2024 13:06:29 +0000 Subject: [PATCH 044/244] Only RIGHT JOINs test --- src/Planner/PlannerJoinTree.cpp | 17 +++-- src/Planner/findParallelReplicasQuery.cpp | 11 ++- .../03254_pr_join_on_dups.reference | 72 ------------------- .../0_stateless/03254_pr_join_on_dups.sql | 28 +------- 4 files changed, 26 insertions(+), 102 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index 160d7f07d5b..d79aa626d5e 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -659,7 +659,7 @@ std::unique_ptr createComputeAliasColumnsStep( } JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expression, - const QueryNode & parent_query_node, + [[maybe_unused]] const QueryNode & parent_query_node, const SelectQueryInfo & select_query_info, const SelectQueryOptions & select_query_options, PlannerContextPtr & planner_context, @@ -958,6 +958,14 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres return true; }; + LOG_DEBUG( + getLogger(__PRETTY_FUNCTION__), + "parallel_replicas_node={} parent_query_node={}", + UInt64(planner_context->getGlobalPlannerContext()->parallel_replicas_node), + UInt64(&parent_query_node)); + + // const JoinNode * table_join_node = parent_query_node.getJoinTree()->as(); + /// query_plan can be empty if there is nothing to read if (query_plan.isInitialized() && parallel_replicas_enabled_for_storage(storage, settings)) { @@ -984,9 +992,10 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres } } else if ( - ClusterProxy::canUseParallelReplicasOnInitiator(query_context) - && planner_context->getGlobalPlannerContext()->parallel_replicas_node - && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &parent_query_node) + ClusterProxy::canUseParallelReplicasOnInitiator(query_context)) + // && (!table_join_node + // || (table_join_node && planner_context->getGlobalPlannerContext()->parallel_replicas_node + // && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &parent_query_node))) { // (1) find read step QueryPlan::Node * node = query_plan.getRootNode(); diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 63c0ce8eb68..8d818daa575 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -265,11 +265,17 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr auto stack = getSupportingParallelReplicasQuery(query_tree_node.get()); /// Empty stack means that storage does not support parallel replicas. if (stack.empty()) + { + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found"); return nullptr; + } /// We don't have any subquery and storage can process parallel replicas by itself. if (stack.top() == query_tree_node.get()) + { + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query itself:\n{}", query_tree_node->dumpTree()); return query_node; + } /// This is needed to avoid infinite recursion. auto mutable_context = Context::createCopy(context); @@ -303,7 +309,10 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr new_stack.pop(); } } - + if (!res) + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found 2"); + else + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query:\n{}", query_tree_node->dumpTree()); return res; } diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.reference b/tests/queries/0_stateless/03254_pr_join_on_dups.reference index 58602bafb5d..95cb0d8cae2 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.reference +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.reference @@ -88,34 +88,6 @@ right subs 4 l5 \N 4 r6 nr6 4 l6 \N 4 r6 nr6 9 l9 \N 9 r9 nr9 -full -0 \N 6 r7 nr7 -0 \N 7 r8 nr8 -1 l1 1 1 r1 \N -1 l1 1 1 r2 \N -2 l2 2 2 r3 \N -2 l3 3 2 r3 \N -3 l4 4 3 r4 \N -3 l4 4 3 r5 \N -4 l5 \N 4 r6 nr6 -4 l6 \N 4 r6 nr6 -5 l7 \N 0 \N -8 l8 \N 0 \N -9 l9 \N 9 r9 nr9 -full subs -0 \N 6 r7 nr7 -0 \N 7 r8 nr8 -1 l1 1 1 r1 \N -1 l1 1 1 r2 \N -2 l2 2 2 r3 \N -2 l3 3 2 r3 \N -3 l4 4 3 r4 \N -3 l4 4 3 r5 \N -4 l5 \N 4 r6 nr6 -4 l6 \N 4 r6 nr6 -5 l7 \N 0 \N -8 l8 \N 0 \N -9 l9 \N 9 r9 nr9 self inner 1 l1 1 1 l1 1 2 l2 2 2 l2 2 @@ -227,47 +199,3 @@ self right nullable vs not nullable 3 l4 4 2 l3 3 4 l5 \N 3 l4 4 4 l6 \N 3 l4 4 -self full -1 l1 1 1 l1 1 -2 l2 2 2 l2 2 -2 l2 2 2 l3 3 -2 l3 3 2 l2 2 -2 l3 3 2 l3 3 -3 l4 4 3 l4 4 -4 l5 \N 4 l5 \N -4 l5 \N 4 l6 \N -4 l6 \N 4 l5 \N -4 l6 \N 4 l6 \N -5 l7 \N 5 l7 \N -8 l8 \N 8 l8 \N -9 l9 \N 9 l9 \N -self full nullable -0 \N 4 l5 \N -0 \N 4 l6 \N -0 \N 5 l7 \N -0 \N 8 l8 \N -0 \N 9 l9 \N -1 l1 1 1 l1 1 -2 l2 2 2 l2 2 -2 l3 3 2 l3 3 -3 l4 4 3 l4 4 -4 l5 \N 0 \N -4 l6 \N 0 \N -5 l7 \N 0 \N -8 l8 \N 0 \N -9 l9 \N 0 \N -self full nullable vs not nullable -0 \N 4 l5 \N -0 \N 4 l6 \N -0 \N 5 l7 \N -0 \N 8 l8 \N -0 \N 9 l9 \N -1 l1 1 1 l1 1 -2 l2 2 2 l2 2 -2 l3 3 2 l2 2 -3 l4 4 2 l3 3 -4 l5 \N 3 l4 4 -4 l6 \N 3 l4 4 -5 l7 \N 0 \N -8 l8 \N 0 \N -9 l9 \N 0 \N diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.sql b/tests/queries/0_stateless/03254_pr_join_on_dups.sql index 71695c0d486..22e94507c83 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.sql +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.sql @@ -9,7 +9,7 @@ insert into X (id, x_a) values (4, 'l5'), (4, 'l6'), (5, 'l7'), (8, 'l8'), insert into Y (id, y_a) values (1, 'r1'), (1, 'r2'), (2, 'r3'), (3, 'r4'), (3, 'r5'); insert into Y (id, y_a, y_b) values (4, 'r6', 'nr6'), (6, 'r7', 'nr7'), (7, 'r8', 'nr8'), (9, 'r9', 'nr9'); -set enable_parallel_replicas = 2, max_parallel_replicas = 3, parallel_replicas_for_non_replicated_merge_tree = 1, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost', parallel_replicas_local_plan=1; +set enable_parallel_replicas = 1, max_parallel_replicas = 3, parallel_replicas_for_non_replicated_merge_tree = 1, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost', parallel_replicas_local_plan=1; select 'inner'; select X.*, Y.* from X inner join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; @@ -29,15 +29,6 @@ select 'right'; select X.*, Y.* from X right join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; select 'right subs'; select s.*, j.* from (select * from X) as s right join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; ---select 'right expr'; ---select X.*, Y.* from X right join Y on (X.id + 1) = (Y.id + 1) order by id; - -select 'full'; -select X.*, Y.* from X full join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; -select 'full subs'; -select s.*, j.* from (select * from X) as s full join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; ---select 'full expr'; ---select X.*, Y.* from X full join Y on (X.id + 1) = (Y.id + 1) order by id; select 'self inner'; select X.*, s.* from X inner join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; @@ -45,7 +36,6 @@ select 'self inner nullable'; select X.*, s.* from X inner join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; select 'self inner nullable vs not nullable'; select X.*, s.* from X inner join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; --- TODO: s.y_b == '' instead of NULL select 'self inner nullable vs not nullable 2'; select Y.*, s.* from Y inner join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by Y.id, Y.y_a, Y.y_b, s.id, s.y_a, s.y_b; @@ -55,7 +45,6 @@ select 'self left nullable'; select X.*, s.* from X left join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; select 'self left nullable vs not nullable'; select X.*, s.* from X left join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; --- TODO: s.y_b == '' instead of NULL select 'self left nullable vs not nullable 2'; select Y.*, s.* from Y left join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by Y.id, Y.y_a, Y.y_b, s.id, s.y_a, s.y_b; @@ -65,17 +54,6 @@ select 'self right nullable'; select X.*, s.* from X right join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; select 'self right nullable vs not nullable'; select X.*, s.* from X right join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; ---select 'self right nullable vs not nullable 2'; ---select Y.*, s.* from Y right join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by id; -select 'self full'; -select X.*, s.* from X full join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; -select 'self full nullable'; -select X.*, s.* from X full join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; -select 'self full nullable vs not nullable'; -select X.*, s.* from X full join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; ---select 'self full nullable vs not nullable 2'; ---select Y.*, s.* from Y full join (select * from Y) as s on concat('n', Y.y_a) = s.y_b order by id; - --- drop table X; --- drop table Y; +drop table X sync; +drop table Y sync; From 31f761508875de1fdc678429b316e19556538eb4 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 1 Nov 2024 14:52:42 +0000 Subject: [PATCH 045/244] Fix --- src/Planner/Planner.cpp | 4 ++-- src/Planner/Planner.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/src/Planner/Planner.cpp b/src/Planner/Planner.cpp index 260462652fc..4b5a2b903c0 100644 --- a/src/Planner/Planner.cpp +++ b/src/Planner/Planner.cpp @@ -1263,7 +1263,7 @@ Planner::Planner(const QueryTreeNodePtr & query_tree_, findQueryForParallelReplicas(query_tree, select_query_options), findTableForParallelReplicas(query_tree, select_query_options), collectFiltersForAnalysis(query_tree, select_query_options)))) - , root_planner(true) + // , root_planner(true) { } @@ -1538,7 +1538,7 @@ void Planner::buildPlanForQueryNode() JoinTreeQueryPlan join_tree_query_plan; if (planner_context->getMutableQueryContext()->canUseTaskBasedParallelReplicas() - && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &query_node && !root_planner) + && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &query_node /* && !root_planner*/) { join_tree_query_plan = buildQueryPlanForParallelReplicas(query_node, planner_context, select_query_info.storage_limits); } diff --git a/src/Planner/Planner.h b/src/Planner/Planner.h index bf11c9ef9cd..8d771c343c3 100644 --- a/src/Planner/Planner.h +++ b/src/Planner/Planner.h @@ -82,7 +82,7 @@ private: StorageLimitsList storage_limits; std::set used_row_policies; QueryNodeToPlanStepMapping query_node_to_plan_step_mapping; - bool root_planner = false; + // bool root_planner = false; }; } From 8f86168c65ad74e6203c59620f4667d0083e3c9e Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 1 Nov 2024 14:53:06 +0000 Subject: [PATCH 046/244] Fix test --- tests/integration/test_failed_mutations/test.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tests/integration/test_failed_mutations/test.py b/tests/integration/test_failed_mutations/test.py index 8d2ee46e748..c7e571ae171 100644 --- a/tests/integration/test_failed_mutations/test.py +++ b/tests/integration/test_failed_mutations/test.py @@ -28,7 +28,7 @@ POSTPONE_MUTATION_LOG = ( "According to exponential backoff policy, do not perform mutations for the part" ) FAILING_MUTATION_QUERY = ( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT throwIf(1))" + "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT throwIf(1)) SETTINGS allow_nondeterministic_mutations = 1" ) all_nodes = [node_with_backoff, node_no_backoff] @@ -111,7 +111,7 @@ def test_exponential_backoff_create_dependent_table(started_cluster): # Executing incorrect mutation. node_with_backoff.query( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS validate_mutation_query = 0" + "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT x FROM dep_table) SETTINGS allow_nondeterministic_mutations = 1, validate_mutation_query = 0" ) # Creating dependent table for mutation. From ac0902b08820dcd64cb41ba6bd34e4957fe8eadf Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 1 Nov 2024 15:57:44 +0000 Subject: [PATCH 047/244] Fix --- src/Planner/findParallelReplicasQuery.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 8d818daa575..e89f06d6cc3 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -274,7 +274,8 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr if (stack.top() == query_tree_node.get()) { LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query itself:\n{}", query_tree_node->dumpTree()); - return query_node; + return nullptr; + // return query_node; } /// This is needed to avoid infinite recursion. From 67b773dcddc61c01d603b16ac59632e9a8cc4f26 Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 1 Nov 2024 16:01:17 +0000 Subject: [PATCH 048/244] Fix style --- tests/integration/test_failed_mutations/test.py | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/tests/integration/test_failed_mutations/test.py b/tests/integration/test_failed_mutations/test.py index c7e571ae171..24b67ff86e5 100644 --- a/tests/integration/test_failed_mutations/test.py +++ b/tests/integration/test_failed_mutations/test.py @@ -27,9 +27,7 @@ REPLICATED_POSTPONE_MUTATION_LOG = ( POSTPONE_MUTATION_LOG = ( "According to exponential backoff policy, do not perform mutations for the part" ) -FAILING_MUTATION_QUERY = ( - "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT throwIf(1)) SETTINGS allow_nondeterministic_mutations = 1" -) +FAILING_MUTATION_QUERY = "ALTER TABLE test_mutations DELETE WHERE x IN (SELECT throwIf(1)) SETTINGS allow_nondeterministic_mutations = 1" all_nodes = [node_with_backoff, node_no_backoff] From 38a3c6707525fba84c190e6a7e42f791b2da5659 Mon Sep 17 00:00:00 2001 From: vdimir Date: Fri, 1 Nov 2024 18:17:08 +0000 Subject: [PATCH 049/244] revert unnecessary changes --- .../Algorithms/CollapsingSortedAlgorithm.cpp | 6 +- .../Algorithms/CollapsingSortedAlgorithm.h | 6 +- .../Algorithms/MergingSortedAlgorithm.cpp | 3 +- .../Algorithms/MergingSortedAlgorithm.h | 6 +- .../Algorithms/ReplacingSortedAlgorithm.cpp | 6 +- .../Algorithms/ReplacingSortedAlgorithm.h | 6 +- .../VersionedCollapsingAlgorithm.cpp | 6 +- .../Algorithms/VersionedCollapsingAlgorithm.h | 6 +- .../Merges/CollapsingSortedTransform.h | 2 +- .../Merges/MergingSortedTransform.cpp | 2 +- .../Merges/MergingSortedTransform.h | 2 +- .../Merges/ReplacingSortedTransform.h | 4 +- .../Merges/VersionedCollapsingTransform.h | 4 +- .../QueryPlan/BuildQueryPipelineSettings.h | 1 + src/QueryPipeline/QueryPipelineBuilder.h | 6 - src/QueryPipeline/QueryPlanResourceHolder.h | 2 - src/Storages/MergeTree/MergeTask.cpp | 129 +++++++++++++++--- src/Storages/MergeTree/MergeTask.h | 5 +- .../MergeTree/MergeTreeSelectProcessor.cpp | 5 +- 19 files changed, 134 insertions(+), 73 deletions(-) diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp index 1560e88ffef..07ee8f4ddef 100644 --- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.cpp @@ -5,7 +5,6 @@ #include #include #include -#include #include @@ -30,18 +29,17 @@ CollapsingSortedAlgorithm::CollapsingSortedAlgorithm( size_t max_block_size_rows_, size_t max_block_size_bytes_, LoggerPtr log_, - std::shared_ptr temp_data_buffer_, + WriteBuffer * out_row_sources_buf_, bool use_average_block_sizes) : IMergingAlgorithmWithSharedChunks( header_, num_inputs, std::move(description_), - temp_data_buffer_.get(), + out_row_sources_buf_, max_row_refs, std::make_unique(use_average_block_sizes, max_block_size_rows_, max_block_size_bytes_)) , sign_column_number(header_.getPositionByName(sign_column)) , only_positive_sign(only_positive_sign_) - , temp_data_buffer(temp_data_buffer_) , log(log_) { } diff --git a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h index b7bb9914cf8..99fd95d82d9 100644 --- a/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h +++ b/src/Processors/Merges/Algorithms/CollapsingSortedAlgorithm.h @@ -11,8 +11,6 @@ namespace Poco namespace DB { -class TemporaryDataBuffer; - /** Merges several sorted inputs to one. * For each group of consecutive identical values of the primary key (the columns by which the data is sorted), * keeps no more than one row with the value of the column `sign_column = -1` ("negative row") @@ -37,7 +35,7 @@ public: size_t max_block_size_rows_, size_t max_block_size_bytes_, LoggerPtr log_, - std::shared_ptr temp_data_buffer_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false); const char * getName() const override { return "CollapsingSortedAlgorithm"; } @@ -64,8 +62,6 @@ private: PODArray current_row_sources; /// Sources of rows with the current primary key size_t count_incorrect_data = 0; /// To prevent too many error messages from writing to the log. - std::shared_ptr temp_data_buffer = nullptr; - LoggerPtr log; void reportIncorrectData(); diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp index d4e4ba6aa5f..3a9cf7ee141 100644 --- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.cpp @@ -3,7 +3,6 @@ #include #include #include -#include namespace DB { @@ -16,7 +15,7 @@ MergingSortedAlgorithm::MergingSortedAlgorithm( size_t max_block_size_bytes_, SortingQueueStrategy sorting_queue_strategy_, UInt64 limit_, - std::shared_ptr out_row_sources_buf_, + WriteBuffer * out_row_sources_buf_, bool use_average_block_sizes) : header(std::move(header_)) , merged_data(use_average_block_sizes, max_block_size_, max_block_size_bytes_) diff --git a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h index fc300e41026..c889668a38e 100644 --- a/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h +++ b/src/Processors/Merges/Algorithms/MergingSortedAlgorithm.h @@ -9,8 +9,6 @@ namespace DB { -class TemporaryDataBuffer; - /// Merges several sorted inputs into one sorted output. class MergingSortedAlgorithm final : public IMergingAlgorithm { @@ -23,7 +21,7 @@ public: size_t max_block_size_bytes_, SortingQueueStrategy sorting_queue_strategy_, UInt64 limit_ = 0, - std::shared_ptr out_row_sources_buf_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false); void addInput(); @@ -47,7 +45,7 @@ private: /// Used in Vertical merge algorithm to gather non-PK/non-index columns (on next step) /// If it is not nullptr then it should be populated during execution - std::shared_ptr out_row_sources_buf = nullptr; + WriteBuffer * out_row_sources_buf = nullptr; /// Chunks currently being merged. Inputs current_inputs; diff --git a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp index a3a33080f52..cd347d371d9 100644 --- a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.cpp @@ -5,7 +5,6 @@ #include #include #include -#include namespace DB { @@ -38,13 +37,12 @@ ReplacingSortedAlgorithm::ReplacingSortedAlgorithm( const String & version_column, size_t max_block_size_rows, size_t max_block_size_bytes, - std::shared_ptr temp_data_buffer_, + WriteBuffer * out_row_sources_buf_, bool use_average_block_sizes, bool cleanup_, bool enable_vertical_final_) - : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), temp_data_buffer_.get(), max_row_refs, std::make_unique(use_average_block_sizes, max_block_size_rows, max_block_size_bytes)) + : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), out_row_sources_buf_, max_row_refs, std::make_unique(use_average_block_sizes, max_block_size_rows, max_block_size_bytes)) , cleanup(cleanup_), enable_vertical_final(enable_vertical_final_) - , temp_data_buffer(temp_data_buffer_) { if (!is_deleted_column.empty()) is_deleted_column_number = header_.getPositionByName(is_deleted_column); diff --git a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h index d3b9837a253..2f23f2a5c4d 100644 --- a/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h +++ b/src/Processors/Merges/Algorithms/ReplacingSortedAlgorithm.h @@ -24,8 +24,6 @@ struct ChunkSelectFinalIndices : public ChunkInfoCloneable temp_data_buffer_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false, bool cleanup = false, bool enable_vertical_final_ = false); @@ -61,8 +59,6 @@ private: RowRef selected_row; /// Last row with maximum version for current primary key, may extend lifetime of chunk in input source size_t max_pos = 0; /// The position (into current_row_sources) of the row with the highest version. - std::shared_ptr temp_data_buffer = nullptr; - /// Sources of rows with the current primary key. PODArray current_row_sources; diff --git a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp index 1ceb1f46234..9f124c6ba18 100644 --- a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp +++ b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.cpp @@ -1,7 +1,6 @@ #include #include #include -#include namespace DB { @@ -15,13 +14,12 @@ VersionedCollapsingAlgorithm::VersionedCollapsingAlgorithm( const String & sign_column_, size_t max_block_size_rows_, size_t max_block_size_bytes_, - std::shared_ptr temp_data_buffer_, + WriteBuffer * out_row_sources_buf_, bool use_average_block_sizes) - : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), temp_data_buffer_.get(), MAX_ROWS_IN_MULTIVERSION_QUEUE, std::make_unique(use_average_block_sizes, max_block_size_rows_, max_block_size_bytes_)) + : IMergingAlgorithmWithSharedChunks(header_, num_inputs, std::move(description_), out_row_sources_buf_, MAX_ROWS_IN_MULTIVERSION_QUEUE, std::make_unique(use_average_block_sizes, max_block_size_rows_, max_block_size_bytes_)) /// -1 for +1 in FixedSizeDequeWithGaps's internal buffer. 3 is a reasonable minimum size to collapse anything. , max_rows_in_queue(std::min(std::max(3, max_block_size_rows_), MAX_ROWS_IN_MULTIVERSION_QUEUE) - 1) , current_keys(max_rows_in_queue) - , temp_data_buffer(temp_data_buffer_) { sign_column_number = header_.getPositionByName(sign_column_); } diff --git a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h index 6f877459147..e6d20ddac75 100644 --- a/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h +++ b/src/Processors/Merges/Algorithms/VersionedCollapsingAlgorithm.h @@ -8,8 +8,6 @@ namespace DB { -class TemporaryDataBuffer; - /** Merges several sorted inputs to one. * For each group of consecutive identical values of the sorting key * (the columns by which the data is sorted, including specially specified version column), @@ -24,7 +22,7 @@ public: SortDescription description_, const String & sign_column_, size_t max_block_size_rows, size_t max_block_size_bytes, - std::shared_ptr temp_data_buffer_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false); const char * getName() const override { return "VersionedCollapsingAlgorithm"; } @@ -39,8 +37,6 @@ private: FixedSizeDequeWithGaps current_keys; Int8 sign_in_queue = 0; - std::shared_ptr temp_data_buffer = nullptr; - std::queue current_row_sources; /// Sources of rows with the current primary key void insertGap(size_t gap_size); diff --git a/src/Processors/Merges/CollapsingSortedTransform.h b/src/Processors/Merges/CollapsingSortedTransform.h index 9b09c802783..99fb700abf1 100644 --- a/src/Processors/Merges/CollapsingSortedTransform.h +++ b/src/Processors/Merges/CollapsingSortedTransform.h @@ -23,7 +23,7 @@ public: bool only_positive_sign, size_t max_block_size_rows, size_t max_block_size_bytes, - std::shared_ptr out_row_sources_buf_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false) : IMergingTransform( num_inputs, header, header, /*have_all_inputs_=*/ true, /*limit_hint_=*/ 0, /*always_read_till_end_=*/ false, diff --git a/src/Processors/Merges/MergingSortedTransform.cpp b/src/Processors/Merges/MergingSortedTransform.cpp index 13330dcff6d..d2895a2a2e9 100644 --- a/src/Processors/Merges/MergingSortedTransform.cpp +++ b/src/Processors/Merges/MergingSortedTransform.cpp @@ -20,7 +20,7 @@ MergingSortedTransform::MergingSortedTransform( SortingQueueStrategy sorting_queue_strategy, UInt64 limit_, bool always_read_till_end_, - std::shared_ptr out_row_sources_buf_, + WriteBuffer * out_row_sources_buf_, bool use_average_block_sizes, bool have_all_inputs_) : IMergingTransform( diff --git a/src/Processors/Merges/MergingSortedTransform.h b/src/Processors/Merges/MergingSortedTransform.h index fb8e5ce74e3..6e52450efa7 100644 --- a/src/Processors/Merges/MergingSortedTransform.h +++ b/src/Processors/Merges/MergingSortedTransform.h @@ -20,7 +20,7 @@ public: SortingQueueStrategy sorting_queue_strategy, UInt64 limit_ = 0, bool always_read_till_end_ = false, - std::shared_ptr out_row_sources_buf_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false, bool have_all_inputs_ = true); diff --git a/src/Processors/Merges/ReplacingSortedTransform.h b/src/Processors/Merges/ReplacingSortedTransform.h index a9d9f4fb619..dc262aab9ee 100644 --- a/src/Processors/Merges/ReplacingSortedTransform.h +++ b/src/Processors/Merges/ReplacingSortedTransform.h @@ -21,7 +21,7 @@ public: const String & is_deleted_column, const String & version_column, size_t max_block_size_rows, size_t max_block_size_bytes, - std::shared_ptr temp_data_buffer_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false, bool cleanup = false, bool enable_vertical_final = false) @@ -34,7 +34,7 @@ public: version_column, max_block_size_rows, max_block_size_bytes, - temp_data_buffer_, + out_row_sources_buf_, use_average_block_sizes, cleanup, enable_vertical_final) diff --git a/src/Processors/Merges/VersionedCollapsingTransform.h b/src/Processors/Merges/VersionedCollapsingTransform.h index 0bdccd4795d..32b5d7bf343 100644 --- a/src/Processors/Merges/VersionedCollapsingTransform.h +++ b/src/Processors/Merges/VersionedCollapsingTransform.h @@ -21,7 +21,7 @@ public: SortDescription description_, const String & sign_column_, size_t max_block_size_rows, size_t max_block_size_bytes, - std::shared_ptr temp_data_buffer_ = nullptr, + WriteBuffer * out_row_sources_buf_ = nullptr, bool use_average_block_sizes = false) : IMergingTransform( num_inputs, header, header, /*have_all_inputs_=*/ true, /*limit_hint_=*/ 0, /*always_read_till_end_=*/ false, @@ -31,7 +31,7 @@ public: sign_column_, max_block_size_rows, max_block_size_bytes, - temp_data_buffer_, + out_row_sources_buf_, use_average_block_sizes) { } diff --git a/src/Processors/QueryPlan/BuildQueryPipelineSettings.h b/src/Processors/QueryPlan/BuildQueryPipelineSettings.h index 1c03a4d74cd..d99f9a7d1f1 100644 --- a/src/Processors/QueryPlan/BuildQueryPipelineSettings.h +++ b/src/Processors/QueryPlan/BuildQueryPipelineSettings.h @@ -20,6 +20,7 @@ struct BuildQueryPipelineSettings ExpressionActionsSettings actions_settings; QueryStatusPtr process_list_element; ProgressCallback progress_callback = nullptr; + TemporaryFileLookupPtr temporary_file_lookup; const ExpressionActionsSettings & getActionsSettings() const { return actions_settings; } static BuildQueryPipelineSettings fromContext(ContextPtr from); diff --git a/src/QueryPipeline/QueryPipelineBuilder.h b/src/QueryPipeline/QueryPipelineBuilder.h index 1e274a97a08..a9e5b1535c0 100644 --- a/src/QueryPipeline/QueryPipelineBuilder.h +++ b/src/QueryPipeline/QueryPipelineBuilder.h @@ -197,12 +197,6 @@ public: void setQueryIdHolder(std::shared_ptr query_id_holder) { resources.query_id_holders.emplace_back(std::move(query_id_holder)); } void addContext(ContextPtr context) { resources.interpreter_context.emplace_back(std::move(context)); } - template - void addResource(Resource resource, std::vector QueryPlanResourceHolder::*field) - { - (resources.*field).push_back(std::move(resource)); - } - /// Convert query pipeline to pipe. static Pipe getPipe(QueryPipelineBuilder pipeline, QueryPlanResourceHolder & resources); static QueryPipeline getPipeline(QueryPipelineBuilder builder); diff --git a/src/QueryPipeline/QueryPlanResourceHolder.h b/src/QueryPipeline/QueryPlanResourceHolder.h index ee2ecc25cd5..10f7f39ab09 100644 --- a/src/QueryPipeline/QueryPlanResourceHolder.h +++ b/src/QueryPipeline/QueryPlanResourceHolder.h @@ -13,7 +13,6 @@ class QueryPlan; class Context; struct QueryIdHolder; -class TemporaryDataBuffer; struct QueryPlanResourceHolder { @@ -34,7 +33,6 @@ struct QueryPlanResourceHolder std::vector storage_holders; std::vector table_locks; std::vector> query_id_holders; - std::vector> rows_sources_temporary_file; }; } diff --git a/src/Storages/MergeTree/MergeTask.cpp b/src/Storages/MergeTree/MergeTask.cpp index 9c2bd59e7cb..e73bc18557c 100644 --- a/src/Storages/MergeTree/MergeTask.cpp +++ b/src/Storages/MergeTree/MergeTask.cpp @@ -65,6 +65,11 @@ namespace ProfileEvents extern const Event MergeProjectionStageExecuteMilliseconds; } +namespace CurrentMetrics +{ + extern const Metric TemporaryFilesForMerge; +} + namespace DB { namespace Setting @@ -124,6 +129,66 @@ static ColumnsStatistics getStatisticsForColumns( return all_statistics; } + +/// Manages the "rows_sources" temporary file that is used during vertical merge. +class RowsSourcesTemporaryFile : public ITemporaryFileLookup +{ +public: + /// A logical name of the temporary file under which it will be known to the plan steps that use it. + static constexpr auto FILE_ID = "rows_sources"; + + explicit RowsSourcesTemporaryFile(TemporaryDataOnDiskScopePtr temporary_data_on_disk_) + : temporary_data_on_disk(temporary_data_on_disk_->childScope(CurrentMetrics::TemporaryFilesForMerge)) + { + } + + WriteBuffer & getTemporaryFileForWriting(const String & name) override + { + if (name != FILE_ID) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected temporary file name requested: {}", name); + + if (tmp_data_buffer) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file was already requested for writing, there musto be only one writer"); + + tmp_data_buffer = std::make_unique(temporary_data_on_disk.get()); + return *tmp_data_buffer; + } + + std::unique_ptr getTemporaryFileForReading(const String & name) override + { + if (name != FILE_ID) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected temporary file name requested: {}", name); + + if (!finalized) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file is not finalized yet"); + + /// tmp_disk might not create real file if no data was written to it. + if (final_size == 0) + return std::make_unique(); + + /// Reopen the file for each read so that multiple reads can be performed in parallel and there is no need to seek to the beginning. + return tmp_data_buffer->read(); + } + + /// Returns written data size in bytes + size_t finalizeWriting() + { + if (!tmp_data_buffer) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file was not requested for writing"); + + auto stat = tmp_data_buffer->finishWriting(); + finalized = true; + final_size = stat.uncompressed_size; + return final_size; + } + +private: + std::unique_ptr tmp_data_buffer; + TemporaryDataOnDiskScopePtr temporary_data_on_disk; + bool finalized = false; + size_t final_size = 0; +}; + static void addMissedColumnsToSerializationInfos( size_t num_rows_in_parts, const Names & part_columns, @@ -425,7 +490,7 @@ bool MergeTask::ExecuteAndFinalizeHorizontalPart::prepare() const } case MergeAlgorithm::Vertical: { - ctx->rows_sources_temporary_file = std::make_unique(global_ctx->context->getTempDataOnDisk().get()); + ctx->rows_sources_temporary_file = std::make_shared(global_ctx->context->getTempDataOnDisk()); std::map local_merged_column_to_size; for (const auto & part : global_ctx->future_part->parts) @@ -802,11 +867,24 @@ bool MergeTask::VerticalMergeStage::prepareVerticalMergeForAllColumns() const if (global_ctx->chosen_merge_algorithm != MergeAlgorithm::Vertical) return false; + size_t sum_input_rows_exact = global_ctx->merge_list_element_ptr->rows_read; + size_t input_rows_filtered = *global_ctx->input_rows_filtered; global_ctx->merge_list_element_ptr->columns_written = global_ctx->merging_columns.size(); global_ctx->merge_list_element_ptr->progress.store(ctx->column_sizes->keyColumnsWeight(), std::memory_order_relaxed); /// Ensure data has written to disk. - ctx->rows_sources_temporary_file->finishWriting(); + size_t rows_sources_count = ctx->rows_sources_temporary_file->finalizeWriting(); + /// In special case, when there is only one source part, and no rows were skipped, we may have + /// skipped writing rows_sources file. Otherwise rows_sources_count must be equal to the total + /// number of input rows. + /// Note that only one byte index is written for each row, so number of rows is equals to the number of bytes written. + if ((rows_sources_count > 0 || global_ctx->future_part->parts.size() > 1) && sum_input_rows_exact != rows_sources_count + input_rows_filtered) + throw Exception( + ErrorCodes::LOGICAL_ERROR, + "Number of rows in source parts ({}) excluding filtered rows ({}) differs from number " + "of bytes written to rows_sources file ({}). It is a bug.", + sum_input_rows_exact, input_rows_filtered, rows_sources_count); + ctx->it_name_and_type = global_ctx->gathering_columns.cbegin(); @@ -838,12 +916,12 @@ class ColumnGathererStep : public ITransformingStep public: ColumnGathererStep( const Header & input_header_, - std::unique_ptr rows_sources_read_buf_, + const String & rows_sources_temporary_file_name_, UInt64 merge_block_size_rows_, UInt64 merge_block_size_bytes_, bool is_result_sparse_) : ITransformingStep(input_header_, input_header_, getTraits()) - , rows_sources_read_buf(std::move(rows_sources_read_buf_)) + , rows_sources_temporary_file_name(rows_sources_temporary_file_name_) , merge_block_size_rows(merge_block_size_rows_) , merge_block_size_bytes(merge_block_size_bytes_) , is_result_sparse(is_result_sparse_) @@ -851,13 +929,15 @@ public: String getName() const override { return "ColumnGatherer"; } - void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & /* pipeline_settings */) override + void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & pipeline_settings) override { - const auto & header = pipeline.getHeader(); + const auto &header = pipeline.getHeader(); const auto input_streams_count = pipeline.getNumStreams(); - if (!rows_sources_read_buf) - throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary data buffer for rows sources is not set"); + if (!pipeline_settings.temporary_file_lookup) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file lookup is not set in pipeline settings for vertical merge"); + + auto rows_sources_read_buf = pipeline_settings.temporary_file_lookup->getTemporaryFileForReading(rows_sources_temporary_file_name); auto transform = std::make_unique( header, @@ -892,7 +972,7 @@ private: } MergeTreeData::MergingParams merging_params{}; - std::unique_ptr rows_sources_read_buf; + const String rows_sources_temporary_file_name; const UInt64 merge_block_size_rows; const UInt64 merge_block_size_bytes; const bool is_result_sparse; @@ -943,7 +1023,7 @@ MergeTask::VerticalMergeRuntimeContext::PreparedColumnPipeline MergeTask::Vertic const auto data_settings = global_ctx->data->getSettings(); auto merge_step = std::make_unique( merge_column_query_plan.getCurrentHeader(), - ctx->rows_sources_temporary_file->read(), + RowsSourcesTemporaryFile::FILE_ID, (*data_settings)[MergeTreeSetting::merge_max_block_size], (*data_settings)[MergeTreeSetting::merge_max_block_size_bytes], is_result_sparse); @@ -972,9 +1052,9 @@ MergeTask::VerticalMergeRuntimeContext::PreparedColumnPipeline MergeTask::Vertic } auto pipeline_settings = BuildQueryPipelineSettings::fromContext(global_ctx->context); + pipeline_settings.temporary_file_lookup = ctx->rows_sources_temporary_file; auto optimization_settings = QueryPlanOptimizationSettings::fromContext(global_ctx->context); auto builder = merge_column_query_plan.buildQueryPipeline(optimization_settings, pipeline_settings); - builder->addResource(ctx->rows_sources_temporary_file, &QueryPlanResourceHolder::rows_sources_temporary_file); return {QueryPipelineBuilder::getPipeline(std::move(*builder)), std::move(indexes_to_recalc)}; } @@ -1347,7 +1427,7 @@ public: const SortDescription & sort_description_, const Names partition_key_columns_, const MergeTreeData::MergingParams & merging_params_, - std::shared_ptr rows_sources_temporary_file_, + const String & rows_sources_temporary_file_name_, UInt64 merge_block_size_rows_, UInt64 merge_block_size_bytes_, bool blocks_are_granules_size_, @@ -1357,7 +1437,7 @@ public: , sort_description(sort_description_) , partition_key_columns(partition_key_columns_) , merging_params(merging_params_) - , rows_sources_temporary_file(rows_sources_temporary_file_) + , rows_sources_temporary_file_name(rows_sources_temporary_file_name_) , merge_block_size_rows(merge_block_size_rows_) , merge_block_size_bytes(merge_block_size_bytes_) , blocks_are_granules_size(blocks_are_granules_size_) @@ -1367,7 +1447,7 @@ public: String getName() const override { return "MergeParts"; } - void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & /* pipeline_settings */) override + void transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & pipeline_settings) override { /// The order of the streams is important: when the key is matched, the elements go in the order of the source stream number. /// In the merged part, the lines with the same key must be in the ascending order of the identifier of original part, @@ -1377,6 +1457,14 @@ public: const auto & header = pipeline.getHeader(); const auto input_streams_count = pipeline.getNumStreams(); + WriteBuffer * rows_sources_write_buf = nullptr; + if (!rows_sources_temporary_file_name.empty()) + { + if (!pipeline_settings.temporary_file_lookup) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Temporary file lookup is not set in pipeline settings for vertical merge"); + rows_sources_write_buf = &pipeline_settings.temporary_file_lookup->getTemporaryFileForWriting(rows_sources_temporary_file_name); + } + switch (merging_params.mode) { case MergeTreeData::MergingParams::Ordinary: @@ -1389,14 +1477,14 @@ public: SortingQueueStrategy::Default, /* limit_= */0, /* always_read_till_end_= */false, - rows_sources_temporary_file, + rows_sources_write_buf, blocks_are_granules_size); break; case MergeTreeData::MergingParams::Collapsing: merged_transform = std::make_shared( header, input_streams_count, sort_description, merging_params.sign_column, false, - merge_block_size_rows, merge_block_size_bytes, rows_sources_temporary_file, blocks_are_granules_size); + merge_block_size_rows, merge_block_size_bytes, rows_sources_write_buf, blocks_are_granules_size); break; case MergeTreeData::MergingParams::Summing: @@ -1411,7 +1499,7 @@ public: case MergeTreeData::MergingParams::Replacing: merged_transform = std::make_shared( header, input_streams_count, sort_description, merging_params.is_deleted_column, merging_params.version_column, - merge_block_size_rows, merge_block_size_bytes, rows_sources_temporary_file, blocks_are_granules_size, + merge_block_size_rows, merge_block_size_bytes, rows_sources_write_buf, blocks_are_granules_size, cleanup); break; @@ -1424,7 +1512,7 @@ public: case MergeTreeData::MergingParams::VersionedCollapsing: merged_transform = std::make_shared( header, input_streams_count, sort_description, merging_params.sign_column, - merge_block_size_rows, merge_block_size_bytes, rows_sources_temporary_file, blocks_are_granules_size); + merge_block_size_rows, merge_block_size_bytes, rows_sources_write_buf, blocks_are_granules_size); break; } @@ -1466,7 +1554,7 @@ private: const SortDescription sort_description; const Names partition_key_columns; const MergeTreeData::MergingParams merging_params{}; - std::shared_ptr rows_sources_temporary_file; + const String rows_sources_temporary_file_name; const UInt64 merge_block_size_rows; const UInt64 merge_block_size_bytes; const bool blocks_are_granules_size; @@ -1635,7 +1723,7 @@ void MergeTask::ExecuteAndFinalizeHorizontalPart::createMergedStream() const sort_description, partition_key_columns, global_ctx->merging_params, - (is_vertical_merge ? ctx->rows_sources_temporary_file : nullptr), /// rows_sources' temporary file is used only for vertical merge + (is_vertical_merge ? RowsSourcesTemporaryFile::FILE_ID : ""), /// rows_sources' temporary file is used only for vertical merge (*data_settings)[MergeTreeSetting::merge_max_block_size], (*data_settings)[MergeTreeSetting::merge_max_block_size_bytes], ctx->blocks_are_granules_size, @@ -1700,6 +1788,7 @@ void MergeTask::ExecuteAndFinalizeHorizontalPart::createMergedStream() const { auto pipeline_settings = BuildQueryPipelineSettings::fromContext(global_ctx->context); + pipeline_settings.temporary_file_lookup = ctx->rows_sources_temporary_file; auto optimization_settings = QueryPlanOptimizationSettings::fromContext(global_ctx->context); auto builder = merge_parts_query_plan.buildQueryPipeline(optimization_settings, pipeline_settings); diff --git a/src/Storages/MergeTree/MergeTask.h b/src/Storages/MergeTree/MergeTask.h index a6969e3aa48..53792165987 100644 --- a/src/Storages/MergeTree/MergeTask.h +++ b/src/Storages/MergeTree/MergeTask.h @@ -42,6 +42,7 @@ namespace DB class MergeTask; using MergeTaskPtr = std::shared_ptr; +class RowsSourcesTemporaryFile; /** * Overview of the merge algorithm @@ -243,7 +244,7 @@ private: bool force_ttl{false}; CompressionCodecPtr compression_codec{nullptr}; size_t sum_input_rows_upper_bound{0}; - std::shared_ptr rows_sources_temporary_file; + std::shared_ptr rows_sources_temporary_file; std::optional column_sizes{}; /// For projections to rebuild @@ -322,7 +323,7 @@ private: struct VerticalMergeRuntimeContext : public IStageRuntimeContext { /// Begin dependencies from previous stage - std::shared_ptr rows_sources_temporary_file; + std::shared_ptr rows_sources_temporary_file; std::optional column_sizes; CompressionCodecPtr compression_codec; std::list::const_iterator it_name_and_type; diff --git a/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp b/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp index 5e9674fb5d6..5efd33ce09a 100644 --- a/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp +++ b/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp @@ -111,11 +111,10 @@ MergeTreeSelectProcessor::MergeTreeSelectProcessor( } if (!prewhere_actions.steps.empty()) - LOG_TRACE(log, "PREWHERE condition was split into {} steps", prewhere_actions.steps.size()); + LOG_TRACE(log, "PREWHERE condition was split into {} steps: {}", prewhere_actions.steps.size(), prewhere_actions.dumpConditions()); if (prewhere_info) - LOG_TEST(log, "Original PREWHERE DAG:{}\n{}\nPREWHERE actions:\n{}", - prewhere_actions.dumpConditions(), + LOG_TEST(log, "Original PREWHERE DAG:\n{}\nPREWHERE actions:\n{}", prewhere_info->prewhere_actions.dumpDAG(), (!prewhere_actions.steps.empty() ? prewhere_actions.dump() : std::string(""))); } From 6d5c707d2cfc029528ba1a32ceb4cd313e198147 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 1 Nov 2024 21:32:07 +0000 Subject: [PATCH 050/244] Cleanup --- src/Planner/findParallelReplicasQuery.cpp | 14 +++++++------- .../02771_parallel_replicas_analyzer.sql | 4 ++-- 2 files changed, 9 insertions(+), 9 deletions(-) diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index e89f06d6cc3..5db67d7c793 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -97,8 +97,8 @@ std::stack getSupportingParallelReplicasQuery(const IQueryTre case QueryTreeNodeType::JOIN: { const auto & join_node = query_tree_node->as(); - auto join_kind = join_node.getKind(); - auto join_strictness = join_node.getStrictness(); + const auto join_kind = join_node.getKind(); + const auto join_strictness = join_node.getStrictness(); if (join_kind == JoinKind::Left || (join_kind == JoinKind::Inner && join_strictness == JoinStrictness::All)) query_tree_node = join_node.getLeftTableExpression().get(); @@ -266,7 +266,7 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr /// Empty stack means that storage does not support parallel replicas. if (stack.empty()) { - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found"); + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found"); return nullptr; } @@ -310,10 +310,10 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr new_stack.pop(); } } - if (!res) - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found 2"); - else - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query:\n{}", query_tree_node->dumpTree()); + // if (!res) + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found 2"); + // else + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query:\n{}", query_tree_node->dumpTree()); return res; } diff --git a/tests/queries/0_stateless/02771_parallel_replicas_analyzer.sql b/tests/queries/0_stateless/02771_parallel_replicas_analyzer.sql index 081077ba460..a2d26a8fc78 100644 --- a/tests/queries/0_stateless/02771_parallel_replicas_analyzer.sql +++ b/tests/queries/0_stateless/02771_parallel_replicas_analyzer.sql @@ -1,5 +1,5 @@ -- Tags: zookeeper -DROP TABLE IF EXISTS join_inner_table__fuzz_146_replicated; +DROP TABLE IF EXISTS join_inner_table__fuzz_146_replicated SYNC; CREATE TABLE join_inner_table__fuzz_146_replicated ( `id` UUID, @@ -52,4 +52,4 @@ WHERE GROUP BY is_initial_query, query ORDER BY is_initial_query DESC, c, query; -DROP TABLE join_inner_table__fuzz_146_replicated; +DROP TABLE join_inner_table__fuzz_146_replicated SYNC; From 1e3f08ab3e48d666cd5e3b02cfecf50915738377 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Sat, 2 Nov 2024 19:44:03 +0000 Subject: [PATCH 051/244] Only with analyzer --- tests/queries/0_stateless/03254_pr_join_on_dups.sql | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.sql b/tests/queries/0_stateless/03254_pr_join_on_dups.sql index 22e94507c83..5f2f209d0b0 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.sql +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.sql @@ -9,7 +9,7 @@ insert into X (id, x_a) values (4, 'l5'), (4, 'l6'), (5, 'l7'), (8, 'l8'), insert into Y (id, y_a) values (1, 'r1'), (1, 'r2'), (2, 'r3'), (3, 'r4'), (3, 'r5'); insert into Y (id, y_a, y_b) values (4, 'r6', 'nr6'), (6, 'r7', 'nr7'), (7, 'r8', 'nr8'), (9, 'r9', 'nr9'); -set enable_parallel_replicas = 1, max_parallel_replicas = 3, parallel_replicas_for_non_replicated_merge_tree = 1, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost', parallel_replicas_local_plan=1; +set enable_analyzer = 1, enable_parallel_replicas = 1, max_parallel_replicas = 3, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost'; select 'inner'; select X.*, Y.* from X inner join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; From 2530fd233f3c4d81ff7ad6f18ec0e3a73320c8d0 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Sat, 2 Nov 2024 21:36:02 +0000 Subject: [PATCH 052/244] Added 03261_pr_semi_anti_join --- .../03261_pr_semi_anti_join.reference | 16 +++++++++++ .../0_stateless/03261_pr_semi_anti_join.sql | 27 +++++++++++++++++++ 2 files changed, 43 insertions(+) create mode 100644 tests/queries/0_stateless/03261_pr_semi_anti_join.reference create mode 100644 tests/queries/0_stateless/03261_pr_semi_anti_join.sql diff --git a/tests/queries/0_stateless/03261_pr_semi_anti_join.reference b/tests/queries/0_stateless/03261_pr_semi_anti_join.reference new file mode 100644 index 00000000000..782147f1f6f --- /dev/null +++ b/tests/queries/0_stateless/03261_pr_semi_anti_join.reference @@ -0,0 +1,16 @@ +semi left +2 a3 2 b1 +2 a6 2 b1 +4 a5 4 b3 +semi right +2 a3 2 b1 +2 a3 2 b2 +4 a5 4 b3 +4 a5 4 b4 +4 a5 4 b5 +anti left +0 a1 0 +1 a2 1 +3 a4 3 +anti right +0 5 b6 diff --git a/tests/queries/0_stateless/03261_pr_semi_anti_join.sql b/tests/queries/0_stateless/03261_pr_semi_anti_join.sql new file mode 100644 index 00000000000..d2ea3725d6b --- /dev/null +++ b/tests/queries/0_stateless/03261_pr_semi_anti_join.sql @@ -0,0 +1,27 @@ +DROP TABLE IF EXISTS t1 SYNC; +DROP TABLE IF EXISTS t2 SYNC; +create table X (id Int32, x_a String, x_b Nullable(Int32)) engine ReplicatedMergeTree('/clickhouse/{database}/X', '1') order by tuple(); + +CREATE TABLE t1 (x UInt32, s String) engine ReplicatedMergeTree('/clickhouse/{database}/t1', '1') order by tuple(); +CREATE TABLE t2 (x UInt32, s String) engine ReplicatedMergeTree('/clickhouse/{database}/t2', '1') order by tuple(); + +INSERT INTO t1 (x, s) VALUES (0, 'a1'), (1, 'a2'), (2, 'a3'), (3, 'a4'), (4, 'a5'), (2, 'a6'); +INSERT INTO t2 (x, s) VALUES (2, 'b1'), (2, 'b2'), (4, 'b3'), (4, 'b4'), (4, 'b5'), (5, 'b6'); + +SET join_use_nulls = 0; +set enable_analyzer = 1, enable_parallel_replicas = 1, max_parallel_replicas = 3, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost'; + +SELECT 'semi left'; +SELECT t1.*, t2.* FROM t1 SEMI LEFT JOIN t2 USING(x) ORDER BY t1.x, t2.x, t1.s, t2.s; + +SELECT 'semi right'; +SELECT t1.*, t2.* FROM t1 SEMI RIGHT JOIN t2 USING(x) ORDER BY t1.x, t2.x, t1.s, t2.s; + +SELECT 'anti left'; +SELECT t1.*, t2.* FROM t1 ANTI LEFT JOIN t2 USING(x) ORDER BY t1.x, t2.x, t1.s, t2.s; + +SELECT 'anti right'; +SELECT t1.*, t2.* FROM t1 ANTI RIGHT JOIN t2 USING(x) ORDER BY t1.x, t2.x, t1.s, t2.s; + +DROP TABLE t1 SYNC; +DROP TABLE t2 SYNC; From c7f970405885d6dae54c9eb94201c662528ab965 Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Mon, 4 Nov 2024 09:45:26 +0000 Subject: [PATCH 053/244] Try fix integration test --- tests/integration/test_quorum_inserts/test.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/integration/test_quorum_inserts/test.py b/tests/integration/test_quorum_inserts/test.py index eefc4882e8e..66f96d61b3e 100644 --- a/tests/integration/test_quorum_inserts/test.py +++ b/tests/integration/test_quorum_inserts/test.py @@ -366,7 +366,7 @@ def test_insert_quorum_with_ttl(started_cluster): zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster") -def test_insert_quorum_with_keeper_loss_connection(): +def test_insert_quorum_with_keeper_loss_connection(started_cluster): zero.query( "DROP TABLE IF EXISTS test_insert_quorum_with_keeper_fail ON CLUSTER cluster" ) From 935a29485c60038b14e4e8c87c8e021fc05f7928 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Mon, 4 Nov 2024 14:32:54 +0000 Subject: [PATCH 054/244] Fix logs --- src/Planner/findParallelReplicasQuery.cpp | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 5db67d7c793..314a7f06137 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -273,7 +273,7 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr /// We don't have any subquery and storage can process parallel replicas by itself. if (stack.top() == query_tree_node.get()) { - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query itself:\n{}", query_tree_node->dumpTree()); + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query itself:\n{}", query_tree_node->dumpTree()); return nullptr; // return query_node; } @@ -427,10 +427,10 @@ const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tr return nullptr; const auto * res = findTableForParallelReplicas(query_tree_node.get()); - if (res) - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Table found {}", res->getStorageID().getFullTableName()); - else - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not table found"); + // if (res) + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Table found {}", res->getStorageID().getFullTableName()); + // else + // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not table found"); return res; } From 47b1b2c1584babf53eadf062a8421e1ce481580c Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Mon, 4 Nov 2024 14:51:43 +0000 Subject: [PATCH 055/244] Try fix integration test - second attempt --- tests/integration/test_quorum_inserts/test.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/integration/test_quorum_inserts/test.py b/tests/integration/test_quorum_inserts/test.py index 66f96d61b3e..f64864185c5 100644 --- a/tests/integration/test_quorum_inserts/test.py +++ b/tests/integration/test_quorum_inserts/test.py @@ -368,7 +368,7 @@ def test_insert_quorum_with_ttl(started_cluster): def test_insert_quorum_with_keeper_loss_connection(started_cluster): zero.query( - "DROP TABLE IF EXISTS test_insert_quorum_with_keeper_fail ON CLUSTER cluster" + "DROP TABLE IF EXISTS test_insert_quorum_with_keeper_loss ON CLUSTER cluster" ) create_query = ( "CREATE TABLE test_insert_quorum_with_keeper_loss" From 876158672c07361f54574c2eefabff5de9e0a48f Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Mon, 4 Nov 2024 17:53:48 +0000 Subject: [PATCH 056/244] Fix integration test: Sync all drop table calls --- tests/integration/test_quorum_inserts/test.py | 24 +++++++++---------- 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/tests/integration/test_quorum_inserts/test.py b/tests/integration/test_quorum_inserts/test.py index f64864185c5..350da822c80 100644 --- a/tests/integration/test_quorum_inserts/test.py +++ b/tests/integration/test_quorum_inserts/test.py @@ -46,7 +46,7 @@ def started_cluster(): def test_simple_add_replica(started_cluster): - zero.query("DROP TABLE IF EXISTS test_simple ON CLUSTER cluster") + zero.query("DROP TABLE IF EXISTS test_simple ON CLUSTER cluster SYNC") create_query = ( "CREATE TABLE test_simple " @@ -82,12 +82,12 @@ def test_simple_add_replica(started_cluster): assert "1\t2011-01-01\n" == first.query("SELECT * from test_simple") assert "1\t2011-01-01\n" == second.query("SELECT * from test_simple") - zero.query("DROP TABLE IF EXISTS test_simple ON CLUSTER cluster") + zero.query("DROP TABLE IF EXISTS test_simple ON CLUSTER cluster SYNC") def test_drop_replica_and_achieve_quorum(started_cluster): zero.query( - "DROP TABLE IF EXISTS test_drop_replica_and_achieve_quorum ON CLUSTER cluster" + "DROP TABLE IF EXISTS test_drop_replica_and_achieve_quorum ON CLUSTER cluster SYNC" ) create_query = ( @@ -156,7 +156,7 @@ def test_insert_quorum_with_drop_partition(started_cluster, add_new_data): if add_new_data else "test_quorum_insert_with_drop_partition" ) - zero.query(f"DROP TABLE IF EXISTS {table_name} ON CLUSTER cluster") + zero.query(f"DROP TABLE IF EXISTS {table_name} ON CLUSTER cluster SYNC") create_query = ( f"CREATE TABLE {table_name} ON CLUSTER cluster " @@ -208,7 +208,7 @@ def test_insert_quorum_with_drop_partition(started_cluster, add_new_data): assert TSV("") == TSV(zero.query(f"SELECT * FROM {table_name}")) assert TSV("") == TSV(second.query(f"SELECT * FROM {table_name}")) - zero.query(f"DROP TABLE IF EXISTS {table_name} ON CLUSTER cluster") + zero.query(f"DROP TABLE IF EXISTS {table_name} ON CLUSTER cluster SYNC") @pytest.mark.parametrize(("add_new_data"), [False, True]) @@ -224,8 +224,8 @@ def test_insert_quorum_with_move_partition(started_cluster, add_new_data): if add_new_data else "test_insert_quorum_with_move_partition_destination" ) - zero.query(f"DROP TABLE IF EXISTS {source_table_name} ON CLUSTER cluster") - zero.query(f"DROP TABLE IF EXISTS {destination_table_name} ON CLUSTER cluster") + zero.query(f"DROP TABLE IF EXISTS {source_table_name} ON CLUSTER cluster SYNC") + zero.query(f"DROP TABLE IF EXISTS {destination_table_name} ON CLUSTER cluster SYNC") create_source = ( f"CREATE TABLE {source_table_name} ON CLUSTER cluster " @@ -291,12 +291,12 @@ def test_insert_quorum_with_move_partition(started_cluster, add_new_data): assert TSV("") == TSV(zero.query(f"SELECT * FROM {source_table_name}")) assert TSV("") == TSV(second.query(f"SELECT * FROM {source_table_name}")) - zero.query(f"DROP TABLE IF EXISTS {source_table_name} ON CLUSTER cluster") - zero.query(f"DROP TABLE IF EXISTS {destination_table_name} ON CLUSTER cluster") + zero.query(f"DROP TABLE IF EXISTS {source_table_name} ON CLUSTER cluster SYNC") + zero.query(f"DROP TABLE IF EXISTS {destination_table_name} ON CLUSTER cluster SYNC") def test_insert_quorum_with_ttl(started_cluster): - zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster") + zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster SYNC") create_query = ( "CREATE TABLE test_insert_quorum_with_ttl " @@ -363,12 +363,12 @@ def test_insert_quorum_with_ttl(started_cluster): ) ) - zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster") + zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster SYNC") def test_insert_quorum_with_keeper_loss_connection(started_cluster): zero.query( - "DROP TABLE IF EXISTS test_insert_quorum_with_keeper_loss ON CLUSTER cluster" + "DROP TABLE IF EXISTS test_insert_quorum_with_keeper_loss ON CLUSTER cluster SYNC" ) create_query = ( "CREATE TABLE test_insert_quorum_with_keeper_loss" From 64fbc9eb8d328db7013525fd6bb34fe0939b7c68 Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Mon, 4 Nov 2024 18:06:08 +0000 Subject: [PATCH 057/244] Style --- tests/integration/test_quorum_inserts/test.py | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/tests/integration/test_quorum_inserts/test.py b/tests/integration/test_quorum_inserts/test.py index 350da822c80..0809d2c003f 100644 --- a/tests/integration/test_quorum_inserts/test.py +++ b/tests/integration/test_quorum_inserts/test.py @@ -296,7 +296,9 @@ def test_insert_quorum_with_move_partition(started_cluster, add_new_data): def test_insert_quorum_with_ttl(started_cluster): - zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster SYNC") + zero.query( + "DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster SYNC" + ) create_query = ( "CREATE TABLE test_insert_quorum_with_ttl " @@ -363,7 +365,9 @@ def test_insert_quorum_with_ttl(started_cluster): ) ) - zero.query("DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster SYNC") + zero.query( + "DROP TABLE IF EXISTS test_insert_quorum_with_ttl ON CLUSTER cluster SYNC" + ) def test_insert_quorum_with_keeper_loss_connection(started_cluster): From a6b55563c73ff10b42569d17ee24457ffff91e4e Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Mon, 4 Nov 2024 18:32:05 +0000 Subject: [PATCH 058/244] Fix FULL joins --- src/Planner/PlannerJoinTree.cpp | 37 +++++----- .../03254_pr_join_on_dups.reference | 72 +++++++++++++++++++ .../0_stateless/03254_pr_join_on_dups.sql | 12 ++++ 3 files changed, 103 insertions(+), 18 deletions(-) diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index d79aa626d5e..c2acbd661c8 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -659,7 +659,7 @@ std::unique_ptr createComputeAliasColumnsStep( } JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expression, - [[maybe_unused]] const QueryNode & parent_query_node, + const QueryTreeNodePtr & parent_join_tree, const SelectQueryInfo & select_query_info, const SelectQueryOptions & select_query_options, PlannerContextPtr & planner_context, @@ -958,17 +958,22 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres return true; }; - LOG_DEBUG( - getLogger(__PRETTY_FUNCTION__), - "parallel_replicas_node={} parent_query_node={}", - UInt64(planner_context->getGlobalPlannerContext()->parallel_replicas_node), - UInt64(&parent_query_node)); - - // const JoinNode * table_join_node = parent_query_node.getJoinTree()->as(); - /// query_plan can be empty if there is nothing to read if (query_plan.isInitialized() && parallel_replicas_enabled_for_storage(storage, settings)) { + const bool allow_parallel_replicas_for_table_expression = [](const QueryTreeNodePtr & join_tree_node) + { + const JoinNode * join_node = join_tree_node->as(); + if (!join_node) + return true; + + const auto join_kind = join_node->getKind(); + if (join_kind == JoinKind::Left || join_kind == JoinKind::Right || join_kind == JoinKind::Inner) + return true; + + return false; + }(parent_join_tree); + if (query_context->canUseParallelReplicasCustomKey() && query_context->getClientInfo().distributed_depth == 0) { if (auto cluster = query_context->getClusterForParallelReplicas(); @@ -991,11 +996,7 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres query_plan = std::move(query_plan_parallel_replicas); } } - else if ( - ClusterProxy::canUseParallelReplicasOnInitiator(query_context)) - // && (!table_join_node - // || (table_join_node && planner_context->getGlobalPlannerContext()->parallel_replicas_node - // && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &parent_query_node))) + else if (ClusterProxy::canUseParallelReplicasOnInitiator(query_context) && allow_parallel_replicas_for_table_expression) { // (1) find read step QueryPlan::Node * node = query_plan.getRootNode(); @@ -1828,8 +1829,8 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, const ColumnIdentifierSet & outer_scope_columns, PlannerContextPtr & planner_context) { - const QueryNode & parent_query_node = query_node->as(); - auto table_expressions_stack = buildTableExpressionsStack(query_node->as().getJoinTree()); + const QueryTreeNodePtr & join_tree_node = query_node->as().getJoinTree(); + auto table_expressions_stack = buildTableExpressionsStack(join_tree_node); size_t table_expressions_stack_size = table_expressions_stack.size(); bool is_single_table_expression = table_expressions_stack_size == 1; @@ -1866,7 +1867,7 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, auto left_table_expression = table_expressions_stack.front(); auto left_table_expression_query_plan = buildQueryPlanForTableExpression( left_table_expression, - parent_query_node, + join_tree_node, select_query_info, select_query_options, planner_context, @@ -1941,7 +1942,7 @@ JoinTreeQueryPlan buildJoinTreeQueryPlan(const QueryTreeNodePtr & query_node, bool is_remote = planner_context->getTableExpressionDataOrThrow(table_expression).isRemote(); query_plans_stack.push_back(buildQueryPlanForTableExpression( table_expression, - parent_query_node, + join_tree_node, select_query_info, select_query_options, planner_context, diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.reference b/tests/queries/0_stateless/03254_pr_join_on_dups.reference index 95cb0d8cae2..58602bafb5d 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.reference +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.reference @@ -88,6 +88,34 @@ right subs 4 l5 \N 4 r6 nr6 4 l6 \N 4 r6 nr6 9 l9 \N 9 r9 nr9 +full +0 \N 6 r7 nr7 +0 \N 7 r8 nr8 +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 +full subs +0 \N 6 r7 nr7 +0 \N 7 r8 nr8 +1 l1 1 1 r1 \N +1 l1 1 1 r2 \N +2 l2 2 2 r3 \N +2 l3 3 2 r3 \N +3 l4 4 3 r4 \N +3 l4 4 3 r5 \N +4 l5 \N 4 r6 nr6 +4 l6 \N 4 r6 nr6 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 9 r9 nr9 self inner 1 l1 1 1 l1 1 2 l2 2 2 l2 2 @@ -199,3 +227,47 @@ self right nullable vs not nullable 3 l4 4 2 l3 3 4 l5 \N 3 l4 4 4 l6 \N 3 l4 4 +self full +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l2 2 2 l3 3 +2 l3 3 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 4 l5 \N +4 l5 \N 4 l6 \N +4 l6 \N 4 l5 \N +4 l6 \N 4 l6 \N +5 l7 \N 5 l7 \N +8 l8 \N 8 l8 \N +9 l9 \N 9 l9 \N +self full nullable +0 \N 4 l5 \N +0 \N 4 l6 \N +0 \N 5 l7 \N +0 \N 8 l8 \N +0 \N 9 l9 \N +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l3 3 +3 l4 4 3 l4 4 +4 l5 \N 0 \N +4 l6 \N 0 \N +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 0 \N +self full nullable vs not nullable +0 \N 4 l5 \N +0 \N 4 l6 \N +0 \N 5 l7 \N +0 \N 8 l8 \N +0 \N 9 l9 \N +1 l1 1 1 l1 1 +2 l2 2 2 l2 2 +2 l3 3 2 l2 2 +3 l4 4 2 l3 3 +4 l5 \N 3 l4 4 +4 l6 \N 3 l4 4 +5 l7 \N 0 \N +8 l8 \N 0 \N +9 l9 \N 0 \N diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.sql b/tests/queries/0_stateless/03254_pr_join_on_dups.sql index 5f2f209d0b0..222f7693090 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.sql +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.sql @@ -30,6 +30,11 @@ select X.*, Y.* from X right join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, select 'right subs'; select s.*, j.* from (select * from X) as s right join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; +select 'full'; +select X.*, Y.* from X full join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; +select 'full subs'; +select s.*, j.* from (select * from X) as s full join (select * from Y) as j on s.id = j.id order by s.id, s.x_a, s.x_b, j.id, j.y_a, j.y_b; + select 'self inner'; select X.*, s.* from X inner join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; select 'self inner nullable'; @@ -55,5 +60,12 @@ select X.*, s.* from X right join (select * from X) as s on X.x_b = s.x_b order select 'self right nullable vs not nullable'; select X.*, s.* from X right join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self full'; +select X.*, s.* from X full join (select * from X) as s on X.id = s.id order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self full nullable'; +select X.*, s.* from X full join (select * from X) as s on X.x_b = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; +select 'self full nullable vs not nullable'; +select X.*, s.* from X full join (select * from X) as s on X.id = s.x_b order by X.id, X.x_a, X.x_b, s.id, s.x_a, s.x_b; + drop table X sync; drop table Y sync; From 8c5ab63345f385a75caa943f4b50169a13e3b470 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Mon, 4 Nov 2024 22:37:57 +0000 Subject: [PATCH 059/244] Cleanup --- src/Planner/Planner.cpp | 3 +-- src/Planner/Planner.h | 1 - src/Planner/PlannerJoinTree.cpp | 6 ------ src/Planner/findParallelReplicasQuery.cpp | 22 +--------------------- 4 files changed, 2 insertions(+), 30 deletions(-) diff --git a/src/Planner/Planner.cpp b/src/Planner/Planner.cpp index 4b5a2b903c0..17277dfe8cd 100644 --- a/src/Planner/Planner.cpp +++ b/src/Planner/Planner.cpp @@ -1263,7 +1263,6 @@ Planner::Planner(const QueryTreeNodePtr & query_tree_, findQueryForParallelReplicas(query_tree, select_query_options), findTableForParallelReplicas(query_tree, select_query_options), collectFiltersForAnalysis(query_tree, select_query_options)))) - // , root_planner(true) { } @@ -1538,7 +1537,7 @@ void Planner::buildPlanForQueryNode() JoinTreeQueryPlan join_tree_query_plan; if (planner_context->getMutableQueryContext()->canUseTaskBasedParallelReplicas() - && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &query_node /* && !root_planner*/) + && planner_context->getGlobalPlannerContext()->parallel_replicas_node == &query_node) { join_tree_query_plan = buildQueryPlanForParallelReplicas(query_node, planner_context, select_query_info.storage_limits); } diff --git a/src/Planner/Planner.h b/src/Planner/Planner.h index 8d771c343c3..ae78f05cbd4 100644 --- a/src/Planner/Planner.h +++ b/src/Planner/Planner.h @@ -82,7 +82,6 @@ private: StorageLimitsList storage_limits; std::set used_row_policies; QueryNodeToPlanStepMapping query_node_to_plan_step_mapping; - // bool root_planner = false; }; } diff --git a/src/Planner/PlannerJoinTree.cpp b/src/Planner/PlannerJoinTree.cpp index c2acbd661c8..c1b8f999f22 100644 --- a/src/Planner/PlannerJoinTree.cpp +++ b/src/Planner/PlannerJoinTree.cpp @@ -669,12 +669,6 @@ JoinTreeQueryPlan buildQueryPlanForTableExpression(QueryTreeNodePtr table_expres auto query_context = planner_context->getQueryContext(); const auto & settings = query_context->getSettingsRef(); - LOG_DEBUG( - getLogger(__PRETTY_FUNCTION__), - "pr_enabled={} table_expression:\n{}", - settings[Setting::allow_experimental_parallel_reading_from_replicas].toString(), - table_expression->dumpTree()); - auto & table_expression_data = planner_context->getTableExpressionDataOrThrow(table_expression); QueryProcessingStage::Enum from_stage = QueryProcessingStage::Enum::FetchColumns; diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 314a7f06137..bda96f0c31f 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -265,18 +265,11 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr auto stack = getSupportingParallelReplicasQuery(query_tree_node.get()); /// Empty stack means that storage does not support parallel replicas. if (stack.empty()) - { - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found"); return nullptr; - } /// We don't have any subquery and storage can process parallel replicas by itself. if (stack.top() == query_tree_node.get()) - { - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query itself:\n{}", query_tree_node->dumpTree()); return nullptr; - // return query_node; - } /// This is needed to avoid infinite recursion. auto mutable_context = Context::createCopy(context); @@ -310,17 +303,11 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr new_stack.pop(); } } - // if (!res) - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not found 2"); - // else - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Query:\n{}", query_tree_node->dumpTree()); return res; } static const TableNode * findTableForParallelReplicas(const IQueryTreeNode * query_tree_node) { - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "\n{}", StackTrace().toString()); - std::stack join_nodes; while (query_tree_node || !join_nodes.empty()) { @@ -426,12 +413,7 @@ const TableNode * findTableForParallelReplicas(const QueryTreeNodePtr & query_tr if (!context->canUseParallelReplicasOnFollower()) return nullptr; - const auto * res = findTableForParallelReplicas(query_tree_node.get()); - // if (res) - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Table found {}", res->getStorageID().getFullTableName()); - // else - // LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Not table found"); - return res; + return findTableForParallelReplicas(query_tree_node.get()); } JoinTreeQueryPlan buildQueryPlanForParallelReplicas( @@ -439,8 +421,6 @@ JoinTreeQueryPlan buildQueryPlanForParallelReplicas( const PlannerContextPtr & planner_context, std::shared_ptr storage_limits) { - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "\n{}", StackTrace().toString()); - auto processed_stage = QueryProcessingStage::WithMergeableState; auto context = planner_context->getQueryContext(); From e198b205092dcb0bec14b8a3a08763cc68a4a1b9 Mon Sep 17 00:00:00 2001 From: Max Kainov Date: Tue, 29 Oct 2024 21:09:03 +0000 Subject: [PATCH 060/244] CI: Stateless Tests with praktika --- .github/workflows/pr.yaml | 168 ++++++++++++++++++-- ci/__init__.py | 0 ci/docker/stateless-test/Dockerfile | 107 +++++++++++++ ci/docker/stateless-test/requirements.txt | 5 + ci/jobs/__init__.py | 0 ci/jobs/build_clickhouse.py | 65 ++++++-- ci/jobs/fast_test.py | 117 +------------- ci/jobs/functional_stateless_tests.py | 119 +++++++++++++- ci/jobs/scripts/__init__.py | 0 ci/jobs/scripts/clickhouse_proc.py | 144 +++++++++++++++++ ci/jobs/scripts/functional_tests_results.py | 3 + ci/praktika/_settings.py | 3 + ci/praktika/hook_html.py | 110 ++++++++++--- ci/praktika/job.py | 46 +++++- ci/praktika/json.html | 156 +++++++++++------- ci/praktika/mangle.py | 1 - ci/praktika/native_jobs.py | 5 +- ci/praktika/param.py | 8 + ci/praktika/result.py | 19 ++- ci/praktika/runner.py | 20 ++- ci/praktika/s3.py | 2 +- ci/praktika/workflow.py | 1 + ci/praktika/yaml_generator.py | 3 + ci/settings/definitions.py | 38 +++-- ci/workflows/pull_request.py | 53 ++++-- tests/clickhouse-test | 11 +- tests/config/config.d/ssl_certs.xml | 4 +- tests/config/install.sh | 24 ++- tests/docker_scripts/setup_minio.sh | 28 ++-- 29 files changed, 955 insertions(+), 305 deletions(-) create mode 100644 ci/__init__.py create mode 100644 ci/docker/stateless-test/Dockerfile create mode 100644 ci/docker/stateless-test/requirements.txt create mode 100644 ci/jobs/__init__.py create mode 100644 ci/jobs/scripts/__init__.py create mode 100644 ci/jobs/scripts/clickhouse_proc.py create mode 100644 ci/praktika/param.py diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml index 34c794f6088..0c3f74aeac8 100644 --- a/.github/workflows/pr.yaml +++ b/.github/workflows/pr.yaml @@ -30,6 +30,9 @@ jobs: steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | @@ -68,6 +71,9 @@ jobs: steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | @@ -106,6 +112,9 @@ jobs: steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | @@ -144,6 +153,9 @@ jobs: steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | @@ -172,16 +184,19 @@ jobs: python3 -m praktika run --job '''Fast test''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi - build_amd64_debug: + build_amd_debug: runs-on: [builder] needs: [config_workflow, docker_builds] - if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'QnVpbGQgYW1kNjQgZGVidWc=') }} - name: "Build amd64 debug" + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'QnVpbGQgKGFtZF9kZWJ1Zyk=') }} + name: "Build (amd_debug)" outputs: data: ${{ steps.run.outputs.DATA }} steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | @@ -205,21 +220,24 @@ jobs: . /tmp/praktika_setup_env.sh set -o pipefail if command -v ts &> /dev/null; then - python3 -m praktika run --job '''Build amd64 debug''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Build (amd_debug)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log else - python3 -m praktika run --job '''Build amd64 debug''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Build (amd_debug)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi - stateless_tests_amd_debug: + build_amd_release: runs-on: [builder] - needs: [config_workflow, docker_builds, build_amd64_debug] - if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKQ==') }} - name: "Stateless tests (amd, debug)" + needs: [config_workflow, docker_builds] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'QnVpbGQgKGFtZF9yZWxlYXNlKQ==') }} + name: "Build (amd_release)" outputs: data: ${{ steps.run.outputs.DATA }} steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | @@ -243,14 +261,137 @@ jobs: . /tmp/praktika_setup_env.sh set -o pipefail if command -v ts &> /dev/null; then - python3 -m praktika run --job '''Stateless tests (amd, debug)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Build (amd_release)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log else - python3 -m praktika run --job '''Stateless tests (amd, debug)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Build (amd_release)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + stateless_tests_amd_debug_parallel_1_2: + runs-on: [builder] + needs: [config_workflow, docker_builds, build_amd_debug] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKSAocGFyYWxsZWwgMS8yKQ==') }} + name: "Stateless tests (amd, debug) (parallel 1/2)" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 1/2)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 1/2)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + stateless_tests_amd_debug_parallel_2_2: + runs-on: [builder] + needs: [config_workflow, docker_builds, build_amd_debug] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKSAocGFyYWxsZWwgMi8yKQ==') }} + name: "Stateless tests (amd, debug) (parallel 2/2)" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 2/2)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 2/2)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + stateless_tests_amd_debug_non_parallel: + runs-on: [style-checker] + needs: [config_workflow, docker_builds, build_amd_debug] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKSAobm9uLXBhcmFsbGVsKQ==') }} + name: "Stateless tests (amd, debug) (non-parallel)" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Stateless tests (amd, debug) (non-parallel)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Stateless tests (amd, debug) (non-parallel)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi finish_workflow: runs-on: [ci_services] - needs: [config_workflow, docker_builds, style_check, fast_test, build_amd64_debug, stateless_tests_amd_debug] + needs: [config_workflow, docker_builds, style_check, fast_test, build_amd_debug, build_amd_release, stateless_tests_amd_debug_parallel_1_2, stateless_tests_amd_debug_parallel_2_2, stateless_tests_amd_debug_non_parallel] if: ${{ !cancelled() }} name: "Finish Workflow" outputs: @@ -258,6 +399,9 @@ jobs: steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{ github.event.pull_reguest.head.sha }} - name: Prepare env script run: | diff --git a/ci/__init__.py b/ci/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/ci/docker/stateless-test/Dockerfile b/ci/docker/stateless-test/Dockerfile new file mode 100644 index 00000000000..4abd8204f1d --- /dev/null +++ b/ci/docker/stateless-test/Dockerfile @@ -0,0 +1,107 @@ +# docker build -t clickhouse/stateless-test . +FROM ubuntu:22.04 + +# ARG for quick switch to a given ubuntu mirror +ARG apt_archive="http://archive.ubuntu.com" +RUN sed -i "s|http://archive.ubuntu.com|$apt_archive|g" /etc/apt/sources.list + +ARG odbc_driver_url="https://github.com/ClickHouse/clickhouse-odbc/releases/download/v1.1.6.20200320/clickhouse-odbc-1.1.6-Linux.tar.gz" + +# moreutils - provides ts fo FT +# expect, bzip2 - requried by FT +# bsdmainutils - provides hexdump for FT + +# golang version 1.13 on Ubuntu 20 is enough for tests +RUN apt-get update -y \ + && env DEBIAN_FRONTEND=noninteractive \ + apt-get install --yes --no-install-recommends \ + awscli \ + brotli \ + lz4 \ + expect \ + moreutils \ + bzip2 \ + bsdmainutils \ + golang \ + lsof \ + mysql-client=8.0* \ + ncdu \ + netcat-openbsd \ + nodejs \ + npm \ + odbcinst \ + openjdk-11-jre-headless \ + openssl \ + postgresql-client \ + python3 \ + python3-pip \ + qemu-user-static \ + sqlite3 \ + sudo \ + tree \ + unixodbc \ + rustc \ + cargo \ + zstd \ + file \ + jq \ + pv \ + zip \ + unzip \ + p7zip-full \ + curl \ + wget \ + && apt-get clean \ + && rm -rf /var/lib/apt/lists/* /var/cache/debconf /tmp/* + +ARG PROTOC_VERSION=25.1 +RUN curl -OL https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-linux-x86_64.zip \ + && unzip protoc-${PROTOC_VERSION}-linux-x86_64.zip -d /usr/local \ + && rm protoc-${PROTOC_VERSION}-linux-x86_64.zip + +COPY requirements.txt / +RUN pip3 install --no-cache-dir -r /requirements.txt + +RUN mkdir -p /tmp/clickhouse-odbc-tmp \ + && cd /tmp/clickhouse-odbc-tmp \ + && curl -L ${odbc_driver_url} | tar --strip-components=1 -xz clickhouse-odbc-1.1.6-Linux \ + && mkdir /usr/local/lib64 -p \ + && cp /tmp/clickhouse-odbc-tmp/lib64/*.so /usr/local/lib64/ \ + && odbcinst -i -d -f /tmp/clickhouse-odbc-tmp/share/doc/clickhouse-odbc/config/odbcinst.ini.sample \ + && odbcinst -i -s -l -f /tmp/clickhouse-odbc-tmp/share/doc/clickhouse-odbc/config/odbc.ini.sample \ + && sed -i 's"=libclickhouseodbc"=/usr/local/lib64/libclickhouseodbc"' /etc/odbcinst.ini \ + && rm -rf /tmp/clickhouse-odbc-tmp + +ENV TZ=Europe/Amsterdam +RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone + +ENV NUM_TRIES=1 + +# Unrelated to vars in setup_minio.sh, but should be the same there +# to have the same binaries for local running scenario +ARG MINIO_SERVER_VERSION=2024-08-03T04-33-23Z +ARG MINIO_CLIENT_VERSION=2024-07-31T15-58-33Z +ARG TARGETARCH + +# Download Minio-related binaries +RUN arch=${TARGETARCH:-amd64} \ + && curl -L "https://dl.min.io/server/minio/release/linux-${arch}/archive/minio.RELEASE.${MINIO_SERVER_VERSION}" -o /minio \ + && curl -L "https://dl.min.io/client/mc/release/linux-${arch}/archive/mc.RELEASE.${MINIO_CLIENT_VERSION}" -o /mc \ + && chmod +x /mc /minio + +ENV MINIO_ROOT_USER="clickhouse" +ENV MINIO_ROOT_PASSWORD="clickhouse" + +# for minio to work without root +RUN chmod 777 /home +ENV HOME="/home" +ENV TEMP_DIR="/tmp/praktika" +ENV PATH="/wd/tests:/tmp/praktika/input:$PATH" + +RUN curl -L --no-verbose -O 'https://archive.apache.org/dist/hadoop/common/hadoop-3.3.1/hadoop-3.3.1.tar.gz' \ + && tar -xvf hadoop-3.3.1.tar.gz \ + && rm -rf hadoop-3.3.1.tar.gz + + +RUN npm install -g azurite@3.30.0 \ + && npm install -g tslib && npm install -g node diff --git a/ci/docker/stateless-test/requirements.txt b/ci/docker/stateless-test/requirements.txt new file mode 100644 index 00000000000..d556d23485f --- /dev/null +++ b/ci/docker/stateless-test/requirements.txt @@ -0,0 +1,5 @@ +Jinja2==3.1.3 +numpy==1.26.4 +requests==2.32.3 +pandas==1.5.3 +scipy==1.12.0 diff --git a/ci/jobs/__init__.py b/ci/jobs/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/ci/jobs/build_clickhouse.py b/ci/jobs/build_clickhouse.py index cfa358b4059..3db88938f23 100644 --- a/ci/jobs/build_clickhouse.py +++ b/ci/jobs/build_clickhouse.py @@ -1,5 +1,6 @@ import argparse +from praktika.param import get_param from praktika.result import Result from praktika.settings import Settings from praktika.utils import MetaClasses, Shell, Utils @@ -14,7 +15,9 @@ class JobStages(metaclass=MetaClasses.WithIter): def parse_args(): parser = argparse.ArgumentParser(description="ClickHouse Build Job") parser.add_argument( - "BUILD_TYPE", help="Type: __" + "--build-type", + help="Type: __", + default=None, ) parser.add_argument( "--param", @@ -24,6 +27,18 @@ def parse_args(): return parser.parse_args() +CMAKE_CMD = """cmake --debug-trycompile -DCMAKE_VERBOSE_MAKEFILE=1 -LA \ +-DCMAKE_BUILD_TYPE={BUILD_TYPE} \ +-DSANITIZE={SANITIZER} \ +-DENABLE_CHECK_HEAVY_BUILDS=1 -DENABLE_CLICKHOUSE_SELF_EXTRACTING=1 -DENABLE_TESTS=0 \ +-DENABLE_UTILS=0 -DCMAKE_FIND_PACKAGE_NO_PACKAGE_REGISTRY=ON -DCMAKE_INSTALL_PREFIX=/usr \ +-DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_INSTALL_LOCALSTATEDIR=/var -DCMAKE_SKIP_INSTALL_ALL_DEPENDENCY=ON \ +{AUX_DEFS} \ +-DCMAKE_C_COMPILER=clang-18 -DCMAKE_CXX_COMPILER=clang++-18 \ +-DCOMPILER_CACHE={CACHE_TYPE} \ +-DENABLE_BUILD_PROFILING=1 {DIR}""" + + def main(): args = parse_args() @@ -42,20 +57,45 @@ def main(): cmake_build_type = "Release" sanitizer = "" - if "debug" in args.BUILD_TYPE.lower(): - print("Build type set: debug") - cmake_build_type = "Debug" + if args.build_type and get_param(): + assert ( + False + ), "Build type must provided via job parameter (CI case) or via --build-type input argument not both" - if "asan" in args.BUILD_TYPE.lower(): - print("Sanitizer set: address") - sanitizer = "address" + build_type = args.build_type or get_param() + assert ( + build_type + ), "build_type must be provided either as input argument or as a parameter of parametrized job in CI" + build_type = build_type.lower() # if Environment.is_local_run(): # build_cache_type = "disabled" # else: - build_cache_type = "sccache" + CACHE_TYPE = "sccache" + + if "debug" in build_type: + print("Build type set: debug") + BUILD_TYPE = "Debug" + AUX_DEFS = " -DSPLIT_DEBUG_SYMBOLS=ON -DBUILD_STANDALONE_KEEPER=1 " + elif "release" in build_type: + print("Build type set: release") + BUILD_TYPE = "None" + AUX_DEFS = " -DENABLE_TESTS=1 " + + if "asan" in build_type: + print("Sanitizer set: address") + SANITIZER = "address" + else: + SANITIZER = "" + + cmake_cmd = CMAKE_CMD.format( + BUILD_TYPE=BUILD_TYPE, + CACHE_TYPE=CACHE_TYPE, + SANITIZER=SANITIZER, + AUX_DEFS=AUX_DEFS, + DIR=Utils.cwd(), + ) - current_directory = Utils.cwd() build_dir = f"{Settings.TEMP_DIR}/build" res = True @@ -75,12 +115,7 @@ def main(): results.append( Result.create_from_command_execution( name="Cmake configuration", - command=f"cmake --debug-trycompile -DCMAKE_VERBOSE_MAKEFILE=1 -LA -DCMAKE_BUILD_TYPE={cmake_build_type} \ - -DSANITIZE={sanitizer} -DENABLE_CHECK_HEAVY_BUILDS=1 -DENABLE_CLICKHOUSE_SELF_EXTRACTING=1 -DENABLE_TESTS=0 \ - -DENABLE_UTILS=0 -DCMAKE_FIND_PACKAGE_NO_PACKAGE_REGISTRY=ON -DCMAKE_INSTALL_PREFIX=/usr \ - -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_INSTALL_LOCALSTATEDIR=/var -DCMAKE_SKIP_INSTALL_ALL_DEPENDENCY=ON \ - -DCMAKE_C_COMPILER=clang-18 -DCMAKE_CXX_COMPILER=clang++-18 -DCOMPILER_CACHE={build_cache_type} -DENABLE_TESTS=1 \ - -DENABLE_BUILD_PROFILING=1 {current_directory}", + command=cmake_cmd, workdir=build_dir, with_log=True, ) diff --git a/ci/jobs/fast_test.py b/ci/jobs/fast_test.py index dc5e1c975a6..cb7d925fead 100644 --- a/ci/jobs/fast_test.py +++ b/ci/jobs/fast_test.py @@ -1,120 +1,13 @@ import argparse -import threading -from pathlib import Path from praktika.result import Result from praktika.settings import Settings from praktika.utils import MetaClasses, Shell, Utils +from ci.jobs.scripts.clickhouse_proc import ClickHouseProc from ci.jobs.scripts.functional_tests_results import FTResultsProcessor -class ClickHouseProc: - def __init__(self): - self.ch_config_dir = f"{Settings.TEMP_DIR}/etc/clickhouse-server" - self.pid_file = f"{self.ch_config_dir}/clickhouse-server.pid" - self.config_file = f"{self.ch_config_dir}/config.xml" - self.user_files_path = f"{self.ch_config_dir}/user_files" - self.test_output_file = f"{Settings.OUTPUT_DIR}/test_result.txt" - self.command = f"clickhouse-server --config-file {self.config_file} --pid-file {self.pid_file} -- --path {self.ch_config_dir} --user_files_path {self.user_files_path} --top_level_domains_path {self.ch_config_dir}/top_level_domains --keeper_server.storage_path {self.ch_config_dir}/coordination" - self.proc = None - self.pid = 0 - nproc = int(Utils.cpu_count() / 2) - self.fast_test_command = f"clickhouse-test --hung-check --fast-tests-only --no-random-settings --no-random-merge-tree-settings --no-long --testname --shard --zookeeper --check-zookeeper-session --order random --print-time --report-logs-stats --jobs {nproc} -- '' | ts '%Y-%m-%d %H:%M:%S' \ - | tee -a \"{self.test_output_file}\"" - # TODO: store info in case of failure - self.info = "" - self.info_file = "" - - Utils.set_env("CLICKHOUSE_CONFIG_DIR", self.ch_config_dir) - Utils.set_env("CLICKHOUSE_CONFIG", self.config_file) - Utils.set_env("CLICKHOUSE_USER_FILES", self.user_files_path) - Utils.set_env("CLICKHOUSE_SCHEMA_FILES", f"{self.ch_config_dir}/format_schemas") - - def start(self): - print("Starting ClickHouse server") - Shell.check(f"rm {self.pid_file}") - - def run_clickhouse(): - self.proc = Shell.run_async( - self.command, verbose=True, suppress_output=True - ) - - thread = threading.Thread(target=run_clickhouse) - thread.daemon = True # Allow program to exit even if thread is still running - thread.start() - - # self.proc = Shell.run_async(self.command, verbose=True) - - started = False - try: - for _ in range(5): - pid = Shell.get_output(f"cat {self.pid_file}").strip() - if not pid: - Utils.sleep(1) - continue - started = True - print(f"Got pid from fs [{pid}]") - _ = int(pid) - break - except Exception: - pass - - if not started: - stdout = self.proc.stdout.read().strip() if self.proc.stdout else "" - stderr = self.proc.stderr.read().strip() if self.proc.stderr else "" - Utils.print_formatted_error("Failed to start ClickHouse", stdout, stderr) - return False - - print(f"ClickHouse server started successfully, pid [{pid}]") - return True - - def wait_ready(self): - res, out, err = 0, "", "" - attempts = 30 - delay = 2 - for attempt in range(attempts): - res, out, err = Shell.get_res_stdout_stderr( - 'clickhouse-client --query "select 1"', verbose=True - ) - if out.strip() == "1": - print("Server ready") - break - else: - print(f"Server not ready, wait") - Utils.sleep(delay) - else: - Utils.print_formatted_error( - f"Server not ready after [{attempts*delay}s]", out, err - ) - return False - return True - - def run_fast_test(self): - if Path(self.test_output_file).exists(): - Path(self.test_output_file).unlink() - exit_code = Shell.run(self.fast_test_command) - return exit_code == 0 - - def terminate(self): - print("Terminate ClickHouse process") - timeout = 10 - if self.proc: - Utils.terminate_process_group(self.proc.pid) - - self.proc.terminate() - try: - self.proc.wait(timeout=10) - print(f"Process {self.proc.pid} terminated gracefully.") - except Exception: - print( - f"Process {self.proc.pid} did not terminate in {timeout} seconds, killing it..." - ) - Utils.terminate_process_group(self.proc.pid, force=True) - self.proc.wait() # Wait for the process to be fully killed - print(f"Process {self.proc} was killed.") - - def clone_submodules(): submodules_to_update = [ "contrib/sysroot", @@ -240,7 +133,7 @@ def main(): Shell.check(f"rm -rf {build_dir} && mkdir -p {build_dir}") results.append( Result.create_from_command_execution( - name="Checkout Submodules for Minimal Build", + name="Checkout Submodules", command=clone_submodules, ) ) @@ -295,8 +188,8 @@ def main(): if res and JobStages.CONFIG in stages: commands = [ f"rm -rf {Settings.TEMP_DIR}/etc/ && mkdir -p {Settings.TEMP_DIR}/etc/clickhouse-client {Settings.TEMP_DIR}/etc/clickhouse-server", - f"cp {current_directory}/programs/server/config.xml {current_directory}/programs/server/users.xml {Settings.TEMP_DIR}/etc/clickhouse-server/", - f"{current_directory}/tests/config/install.sh {Settings.TEMP_DIR}/etc/clickhouse-server {Settings.TEMP_DIR}/etc/clickhouse-client", + f"cp ./programs/server/config.xml ./programs/server/users.xml {Settings.TEMP_DIR}/etc/clickhouse-server/", + f"./tests/config/install.sh {Settings.TEMP_DIR}/etc/clickhouse-server {Settings.TEMP_DIR}/etc/clickhouse-client --fast-test", # f"cp -a {current_directory}/programs/server/config.d/log_to_console.xml {Settings.TEMP_DIR}/etc/clickhouse-server/config.d/", f"rm -f {Settings.TEMP_DIR}/etc/clickhouse-server/config.d/secure_ports.xml", update_path_ch_config, @@ -310,7 +203,7 @@ def main(): ) res = results[-1].is_ok() - CH = ClickHouseProc() + CH = ClickHouseProc(fast_test=True) if res and JobStages.TEST in stages: stop_watch_ = Utils.Stopwatch() step_name = "Start ClickHouse Server" diff --git a/ci/jobs/functional_stateless_tests.py b/ci/jobs/functional_stateless_tests.py index dfdd5821a19..d77522ed73a 100644 --- a/ci/jobs/functional_stateless_tests.py +++ b/ci/jobs/functional_stateless_tests.py @@ -1,31 +1,78 @@ import argparse +import os +from pathlib import Path +from praktika.param import get_param from praktika.result import Result from praktika.settings import Settings from praktika.utils import MetaClasses, Shell, Utils +from ci.jobs.scripts.clickhouse_proc import ClickHouseProc +from ci.jobs.scripts.functional_tests_results import FTResultsProcessor +from ci.settings.definitions import azure_secret + class JobStages(metaclass=MetaClasses.WithIter): - CHECKOUT_SUBMODULES = "checkout" - CMAKE = "cmake" - BUILD = "build" + INSTALL_CLICKHOUSE = "install" + START = "start" + TEST = "test" def parse_args(): parser = argparse.ArgumentParser(description="ClickHouse Build Job") - parser.add_argument("BUILD_TYPE", help="Type: ") + parser.add_argument( + "BUILD_TYPE", help="Type: __" + ) parser.add_argument("--param", help="Optional custom job start stage", default=None) return parser.parse_args() +def run_stateless_test( + no_parallel: bool, no_sequiential: bool, batch_num: int, batch_total: int +): + assert not (no_parallel and no_sequiential) + test_output_file = f"{Settings.OUTPUT_DIR}/test_result.txt" + aux = "" + nproc = int(Utils.cpu_count() / 2) + if batch_num and batch_total: + aux = f"--run-by-hash-total {batch_total} --run-by-hash-num {batch_num-1}" + statless_test_command = f"clickhouse-test --testname --shard --zookeeper --check-zookeeper-session --hung-check --print-time \ + --no-drop-if-fail --capture-client-stacktrace --queries /repo/tests/queries --test-runs 1 --hung-check \ + {'--no-parallel' if no_parallel else ''} {'--no-sequential' if no_sequiential else ''} \ + --print-time --jobs {nproc} --report-coverage --report-logs-stats {aux} \ + --queries ./tests/queries -- '' | ts '%Y-%m-%d %H:%M:%S' \ + | tee -a \"{test_output_file}\"" + if Path(test_output_file).exists(): + Path(test_output_file).unlink() + Shell.run(statless_test_command, verbose=True) + + def main(): args = parse_args() + params = get_param().split(" ") + parallel_or_sequential = None + no_parallel = False + no_sequential = False + if params: + parallel_or_sequential = params[0] + if len(params) > 1: + batch_num, total_batches = map(int, params[1].split("/")) + else: + batch_num, total_batches = 0, 0 + if parallel_or_sequential: + no_parallel = parallel_or_sequential == "non-parallel" + no_sequential = parallel_or_sequential == "parallel" + + os.environ["AZURE_CONNECTION_STRING"] = Shell.get_output( + f"aws ssm get-parameter --region us-east-1 --name azure_connection_string --with-decryption --output text --query Parameter.Value", + verbose=True, + ) stop_watch = Utils.Stopwatch() stages = list(JobStages) - stage = args.param or JobStages.CHECKOUT_SUBMODULES + stage = args.param or JobStages.INSTALL_CLICKHOUSE if stage: assert stage in JobStages, f"--param must be one of [{list(JobStages)}]" print(f"Job will start from stage [{stage}]") @@ -36,9 +83,65 @@ def main(): res = True results = [] - if res and JobStages.CHECKOUT_SUBMODULES in stages: - info = Shell.get_output(f"ls -l {Settings.INPUT_DIR}") - results.append(Result(name="TEST", status=Result.Status.SUCCESS, info=info)) + Utils.add_to_PATH(f"{Settings.INPUT_DIR}:tests") + + if res and JobStages.INSTALL_CLICKHOUSE in stages: + commands = [ + f"chmod +x {Settings.INPUT_DIR}/clickhouse", + f"ln -sf {Settings.INPUT_DIR}/clickhouse {Settings.INPUT_DIR}/clickhouse-server", + f"ln -sf {Settings.INPUT_DIR}/clickhouse {Settings.INPUT_DIR}/clickhouse-client", + f"rm -rf {Settings.TEMP_DIR}/etc/ && mkdir -p {Settings.TEMP_DIR}/etc/clickhouse-client {Settings.TEMP_DIR}/etc/clickhouse-server", + f"cp programs/server/config.xml programs/server/users.xml {Settings.TEMP_DIR}/etc/clickhouse-server/", + f"./tests/config/install.sh {Settings.TEMP_DIR}/etc/clickhouse-server {Settings.TEMP_DIR}/etc/clickhouse-client --s3-storage", + # update_path_ch_config, + f"sed -i 's|>/var/|>{Settings.TEMP_DIR}/var/|g; s|>/etc/|>{Settings.TEMP_DIR}/etc/|g' {Settings.TEMP_DIR}/etc/clickhouse-server/config.xml", + f"sed -i 's|>/etc/|>{Settings.TEMP_DIR}/etc/|g' {Settings.TEMP_DIR}/etc/clickhouse-server/config.d/ssl_certs.xml", + f"clickhouse-server --version", + ] + results.append( + Result.create_from_command_execution( + name="Install ClickHouse", command=commands, with_log=True + ) + ) + res = results[-1].is_ok() + + CH = ClickHouseProc() + if res and JobStages.START in stages: + stop_watch_ = Utils.Stopwatch() + step_name = "Start ClickHouse Server" + print(step_name) + res = res and CH.start_minio() + res = res and CH.start() + res = res and CH.wait_ready() + results.append( + Result.create_from( + name=step_name, + status=res, + stopwatch=stop_watch_, + files=( + [ + "/tmp/praktika/var/log/clickhouse-server/clickhouse-server.log", + "/tmp/praktika/var/log/clickhouse-server/clickhouse-server.err.log", + ] + if not res + else [] + ), + ) + ) + res = results[-1].is_ok() + + if res and JobStages.TEST in stages: + stop_watch_ = Utils.Stopwatch() + step_name = "Tests" + print(step_name) + run_stateless_test( + no_parallel=no_parallel, + no_sequiential=no_sequential, + batch_num=batch_num, + batch_total=total_batches, + ) + results.append(FTResultsProcessor(wd=Settings.OUTPUT_DIR).run()) + results[-1].set_timing(stopwatch=stop_watch_) res = results[-1].is_ok() Result.create_from(results=results, stopwatch=stop_watch).complete_job() diff --git a/ci/jobs/scripts/__init__.py b/ci/jobs/scripts/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/ci/jobs/scripts/clickhouse_proc.py b/ci/jobs/scripts/clickhouse_proc.py new file mode 100644 index 00000000000..cc822eab693 --- /dev/null +++ b/ci/jobs/scripts/clickhouse_proc.py @@ -0,0 +1,144 @@ +import threading +import time +from pathlib import Path + +from praktika.settings import Settings +from praktika.utils import Shell, Utils + + +class ClickHouseProc: + BACKUPS_XML = """ + + + local + {CH_RUNTIME_DIR}/var/lib/clickhouse/disks/backups/ + + +""" + + def __init__(self, fast_test=False): + self.ch_config_dir = f"{Settings.TEMP_DIR}/etc/clickhouse-server" + self.pid_file = f"{self.ch_config_dir}/clickhouse-server.pid" + self.config_file = f"{self.ch_config_dir}/config.xml" + self.user_files_path = f"{self.ch_config_dir}/user_files" + self.test_output_file = f"{Settings.OUTPUT_DIR}/test_result.txt" + self.command = f"clickhouse-server --config-file {self.config_file} --pid-file {self.pid_file} -- --path {self.ch_config_dir} --user_files_path {self.user_files_path} --top_level_domains_path {self.ch_config_dir}/top_level_domains --keeper_server.storage_path {self.ch_config_dir}/coordination" + self.proc = None + self.pid = 0 + nproc = int(Utils.cpu_count() / 2) + self.fast_test_command = f"clickhouse-test --hung-check --fast-tests-only --no-random-settings --no-random-merge-tree-settings --no-long --testname --shard --zookeeper --check-zookeeper-session --order random --print-time --report-logs-stats --jobs {nproc} -- '' | ts '%Y-%m-%d %H:%M:%S' \ + | tee -a \"{self.test_output_file}\"" + # TODO: store info in case of failure + self.info = "" + self.info_file = "" + + self.minio_cmd = f"tests/docker_scripts/setup_minio.sh stateless 2>&1 > {Settings.OUTPUT_DIR}/minio.log" + + Utils.set_env("CLICKHOUSE_CONFIG_DIR", self.ch_config_dir) + Utils.set_env("CLICKHOUSE_CONFIG", self.config_file) + Utils.set_env("CLICKHOUSE_USER_FILES", self.user_files_path) + Utils.set_env("CLICKHOUSE_SCHEMA_FILES", f"{self.ch_config_dir}/format_schemas") + + if not fast_test: + with open(f"{self.ch_config_dir}/config.d/backups.xml", "w") as file: + file.write(self.BACKUPS_XML) + + self.minio_proc = None + + def start_minio(self): + print("Starting minio") + + def run_minio(): + self.minio_proc = Shell.run_async( + self.minio_cmd, verbose=True, suppress_output=True + ) + + thread = threading.Thread(target=run_minio) + thread.daemon = True # Allow program to exit even if thread is still running + thread.start() + time.sleep(5) + return thread.is_alive() + + def start(self): + print("Starting ClickHouse server") + Shell.check(f"rm {self.pid_file}") + + def run_clickhouse(): + self.proc = Shell.run_async( + self.command, verbose=True, suppress_output=False + ) + + thread = threading.Thread(target=run_clickhouse) + thread.daemon = True # Allow program to exit even if thread is still running + thread.start() + + started = False + try: + for _ in range(5): + pid = Shell.get_output(f"cat {self.pid_file}").strip() + if not pid: + Utils.sleep(1) + continue + started = True + print(f"Got pid from fs [{pid}]") + _ = int(pid) + break + except Exception: + pass + + if not started: + stdout = self.proc.stdout.read().strip() if self.proc.stdout else "" + stderr = self.proc.stderr.read().strip() if self.proc.stderr else "" + Utils.print_formatted_error("Failed to start ClickHouse", stdout, stderr) + return False + + print(f"ClickHouse server started successfully, pid [{pid}]") + return True + + def wait_ready(self): + res, out, err = 0, "", "" + attempts = 30 + delay = 2 + for attempt in range(attempts): + res, out, err = Shell.get_res_stdout_stderr( + 'clickhouse-client --query "select 1"', verbose=True + ) + if out.strip() == "1": + print("Server ready") + break + else: + print(f"Server not ready, wait") + Utils.sleep(delay) + else: + Utils.print_formatted_error( + f"Server not ready after [{attempts*delay}s]", out, err + ) + return False + return True + + def run_fast_test(self): + if Path(self.test_output_file).exists(): + Path(self.test_output_file).unlink() + exit_code = Shell.run(self.fast_test_command) + return exit_code == 0 + + def terminate(self): + print("Terminate ClickHouse process") + timeout = 10 + if self.proc: + Utils.terminate_process_group(self.proc.pid) + + self.proc.terminate() + try: + self.proc.wait(timeout=10) + print(f"Process {self.proc.pid} terminated gracefully.") + except Exception: + print( + f"Process {self.proc.pid} did not terminate in {timeout} seconds, killing it..." + ) + Utils.terminate_process_group(self.proc.pid, force=True) + self.proc.wait() # Wait for the process to be fully killed + print(f"Process {self.proc} was killed.") + + if self.minio_proc: + Utils.terminate_process_group(self.minio_proc.pid) diff --git a/ci/jobs/scripts/functional_tests_results.py b/ci/jobs/scripts/functional_tests_results.py index aba3e4f7f5b..06989fb0a44 100755 --- a/ci/jobs/scripts/functional_tests_results.py +++ b/ci/jobs/scripts/functional_tests_results.py @@ -232,6 +232,8 @@ class FTResultsProcessor: else: pass + info = f"Total: {s.total - s.skipped}, Failed: {s.failed}" + # TODO: !!! # def test_result_comparator(item): # # sort by status then by check name @@ -253,6 +255,7 @@ class FTResultsProcessor: results=test_results, status=state, files=[self.tests_output_file], + info=info, with_info_from_results=False, ) diff --git a/ci/praktika/_settings.py b/ci/praktika/_settings.py index 1777257f484..17da1519e37 100644 --- a/ci/praktika/_settings.py +++ b/ci/praktika/_settings.py @@ -80,6 +80,8 @@ class _Settings: CI_DB_TABLE_NAME = "" CI_DB_INSERT_TIMEOUT_SEC = 5 + DISABLE_MERGE_COMMIT = True + _USER_DEFINED_SETTINGS = [ "S3_ARTIFACT_PATH", @@ -112,6 +114,7 @@ _USER_DEFINED_SETTINGS = [ "SECRET_GH_APP_PEM_KEY", "SECRET_GH_APP_ID", "MAIN_BRANCH", + "DISABLE_MERGE_COMMIT", ] diff --git a/ci/praktika/hook_html.py b/ci/praktika/hook_html.py index f4bd4435511..cea84192d0d 100644 --- a/ci/praktika/hook_html.py +++ b/ci/praktika/hook_html.py @@ -11,50 +11,112 @@ from praktika.result import Result, ResultInfo from praktika.runtime import RunConfig from praktika.s3 import S3 from praktika.settings import Settings -from praktika.utils import Shell, Utils +from praktika.utils import Utils @dataclasses.dataclass class GitCommit: - date: str - message: str + # date: str + # message: str sha: str @staticmethod - def from_json(json_data: str) -> List["GitCommit"]: + def from_json(file) -> List["GitCommit"]: commits = [] + json_data = None try: - data = json.loads(json_data) - + with open(file, "r", encoding="utf-8") as f: + json_data = json.load(f) commits = [ GitCommit( - message=commit["messageHeadline"], - sha=commit["oid"], - date=commit["committedDate"], + # message=commit["messageHeadline"], + sha=commit["sha"], + # date=commit["committedDate"], ) - for commit in data.get("commits", []) + for commit in json_data ] except Exception as e: print( - f"ERROR: Failed to deserialize commit's data: [{json_data}], ex: [{e}]" + f"ERROR: Failed to deserialize commit's data [{json_data}], ex: [{e}]" ) return commits + @classmethod + def update_s3_data(cls): + env = _Environment.get() + sha = env.SHA + if not sha: + print("WARNING: Failed to retrieve commit sha") + return + commits = cls.pull_from_s3() + for commit in commits: + if sha == commit.sha: + print( + f"INFO: Sha already present in commits data [{sha}] - skip data update" + ) + return + commits.append(GitCommit(sha=sha)) + cls.push_to_s3(commits) + return + + @classmethod + def dump(cls, commits): + commits_ = [] + for commit in commits: + commits_.append(dataclasses.asdict(commit)) + with open(cls.file_name(), "w", encoding="utf8") as f: + json.dump(commits_, f) + + @classmethod + def pull_from_s3(cls): + local_path = Path(cls.file_name()) + file_name = local_path.name + env = _Environment.get() + s3_path = f"{Settings.HTML_S3_PATH}/{cls.get_s3_prefix(pr_number=env.PR_NUMBER, branch=env.BRANCH)}/{file_name}" + if not S3.copy_file_from_s3(s3_path=s3_path, local_path=local_path): + print(f"WARNING: failed to cp file [{s3_path}] from s3") + return [] + return cls.from_json(local_path) + + @classmethod + def push_to_s3(cls, commits): + print(f"INFO: push commits data to s3, commits num [{len(commits)}]") + cls.dump(commits) + local_path = Path(cls.file_name()) + file_name = local_path.name + env = _Environment.get() + s3_path = f"{Settings.HTML_S3_PATH}/{cls.get_s3_prefix(pr_number=env.PR_NUMBER, branch=env.BRANCH)}/{file_name}" + if not S3.copy_file_to_s3(s3_path=s3_path, local_path=local_path, text=True): + print(f"WARNING: failed to cp file [{local_path}] to s3") + + @classmethod + def get_s3_prefix(cls, pr_number, branch): + prefix = "" + assert pr_number or branch + if pr_number and pr_number > 0: + prefix += f"{pr_number}" + else: + prefix += f"{branch}" + return prefix + + @classmethod + def file_name(cls): + return f"{Settings.TEMP_DIR}/commits.json" + + # def _get_pr_commits(pr_number): + # res = [] + # if not pr_number: + # return res + # output = Shell.get_output(f"gh pr view {pr_number} --json commits") + # if output: + # res = GitCommit.from_json(output) + # return res + class HtmlRunnerHooks: @classmethod def configure(cls, _workflow): - - def _get_pr_commits(pr_number): - res = [] - if not pr_number: - return res - output = Shell.get_output(f"gh pr view {pr_number} --json commits") - if output: - res = GitCommit.from_json(output) - return res - # generate pending Results for all jobs in the workflow if _workflow.enable_cache: skip_jobs = RunConfig.from_fs(_workflow.name).cache_success @@ -106,11 +168,9 @@ class HtmlRunnerHooks: Utils.raise_with_error( "Failed to set both GH commit status and PR comment with Workflow Status, cannot proceed" ) - if env.PR_NUMBER: - commits = _get_pr_commits(env.PR_NUMBER) - # TODO: upload commits data to s3 to visualise it on a report page - print(commits) + # TODO: enable for branch, add commit number limiting + GitCommit.update_s3_data() @classmethod def pre_run(cls, _workflow, _job): diff --git a/ci/praktika/job.py b/ci/praktika/job.py index d0d4232cfa2..99eb08938b8 100644 --- a/ci/praktika/job.py +++ b/ci/praktika/job.py @@ -52,30 +52,57 @@ class Job: self, parameter: Optional[List[Any]] = None, runs_on: Optional[List[List[str]]] = None, + provides: Optional[List[List[str]]] = None, + requires: Optional[List[List[str]]] = None, timeout: Optional[List[int]] = None, ): assert ( parameter or runs_on ), "Either :parameter or :runs_on must be non empty list for parametrisation" + if runs_on: + assert isinstance(runs_on, list) and isinstance(runs_on[0], list) if not parameter: parameter = [None] * len(runs_on) if not runs_on: runs_on = [None] * len(parameter) if not timeout: timeout = [None] * len(parameter) + if not provides: + provides = [None] * len(parameter) + if not requires: + requires = [None] * len(parameter) assert ( - len(parameter) == len(runs_on) == len(timeout) - ), "Parametrization lists must be of the same size" + len(parameter) + == len(runs_on) + == len(timeout) + == len(provides) + == len(requires) + ), f"Parametrization lists must be of the same size [{len(parameter)}, {len(runs_on)}, {len(timeout)}, {len(provides)}, {len(requires)}]" res = [] - for parameter_, runs_on_, timeout_ in zip(parameter, runs_on, timeout): + for parameter_, runs_on_, timeout_, provides_, requires_ in zip( + parameter, runs_on, timeout, provides, requires + ): obj = copy.deepcopy(self) + assert ( + not obj.provides + ), "Job.Config.provides must be empty for parametrized jobs" if parameter_: obj.parameter = parameter_ if runs_on_: obj.runs_on = runs_on_ if timeout_: obj.timeout = timeout_ + if provides_: + assert ( + not obj.provides + ), "Job.Config.provides must be empty for parametrized jobs" + obj.provides = provides_ + if requires_: + assert ( + not obj.requires + ), "Job.Config.requires and parametrize(requires=...) are both set" + obj.requires = requires_ obj.name = obj.get_job_name_with_parameter() res.append(obj) return res @@ -84,13 +111,16 @@ class Job: name, parameter, runs_on = self.name, self.parameter, self.runs_on res = name name_params = [] - if isinstance(parameter, list) or isinstance(parameter, dict): - name_params.append(json.dumps(parameter)) - elif parameter is not None: - name_params.append(parameter) - if runs_on: + if parameter: + if isinstance(parameter, list) or isinstance(parameter, dict): + name_params.append(json.dumps(parameter)) + else: + name_params.append(parameter) + elif runs_on: assert isinstance(runs_on, list) name_params.append(json.dumps(runs_on)) + else: + assert False if name_params: name_params = [str(param) for param in name_params] res += f" ({', '.join(name_params)})" diff --git a/ci/praktika/json.html b/ci/praktika/json.html index af03ed702f8..f86a7b27ecb 100644 --- a/ci/praktika/json.html +++ b/ci/praktika/json.html @@ -200,10 +200,7 @@ } th.name-column, td.name-column { - max-width: 400px; /* Set the maximum width for the column */ - white-space: nowrap; /* Prevent text from wrapping */ - overflow: hidden; /* Hide the overflowed text */ - text-overflow: ellipsis; /* Show ellipsis (...) for overflowed text */ + min-width: 350px; } th.status-column, td.status-column { @@ -364,7 +361,6 @@ } function addKeyValueToStatus(key, value, options = null) { - const statusContainer = document.getElementById('status-container'); let keyValuePair = document.createElement('div'); @@ -374,27 +370,40 @@ keyElement.className = 'json-key'; keyElement.textContent = key + ':'; - let valueElement - if (value) { - valueElement = document.createElement('div'); - valueElement.className = 'json-value'; - valueElement.textContent = value; - } else if (options) { + let valueElement; + + if (options) { + // Create dropdown if options are provided valueElement = document.createElement('select'); valueElement.className = 'dropdown-value'; + + options.forEach(optionValue => { + const option = document.createElement('option'); + option.value = optionValue; + option.textContent = optionValue.slice(0, 10); + + // Set the initially selected option + if (optionValue === value) { + option.selected = true; + } + + valueElement.appendChild(option); + }); + + // Update the URL parameter when the selected value changes valueElement.addEventListener('change', (event) => { const selectedValue = event.target.value; updateUrlParameter(key, selectedValue); }); - options.forEach(optionValue => { - const option = document.createElement('option'); - option.value = optionValue; - option.textContent = optionValue; - valueElement.appendChild(option); - }); + } else { + // Create a simple text display if no options are provided + valueElement = document.createElement('div'); + valueElement.className = 'json-value'; + valueElement.textContent = value || 'N/A'; // Display 'N/A' if value is null } - keyValuePair.appendChild(keyElement) - keyValuePair.appendChild(valueElement) + + keyValuePair.appendChild(keyElement); + keyValuePair.appendChild(valueElement); statusContainer.appendChild(keyValuePair); } @@ -518,12 +527,12 @@ const columns = ['name', 'status', 'start_time', 'duration', 'info']; const columnSymbols = { - name: '📂', - status: '⏯️', + name: '🗂️', + status: '🧾', start_time: '🕒', duration: '⏳', - info: 'ℹ️', - files: '📄' + info: '📝', + files: '📎' }; function createResultsTable(results, nest_level) { @@ -532,16 +541,14 @@ const thead = document.createElement('thead'); const tbody = document.createElement('tbody'); - // Get the current URL parameters - const currentUrl = new URL(window.location.href); - // Create table headers based on the fixed columns const headerRow = document.createElement('tr'); columns.forEach(column => { const th = document.createElement('th'); - th.textContent = th.textContent = columnSymbols[column] || column; + th.textContent = columnSymbols[column] || column; th.style.cursor = 'pointer'; // Make headers clickable - th.addEventListener('click', () => sortTable(results, column, tbody, nest_level)); // Add click event to sort the table + th.setAttribute('data-sort-direction', 'asc'); // Default sort direction + th.addEventListener('click', () => sortTable(results, column, columnSymbols[column] || column, tbody, nest_level, columns)); // Add click event to sort the table headerRow.appendChild(th); }); thead.appendChild(headerRow); @@ -605,39 +612,33 @@ }); } - function sortTable(results, key, tbody, nest_level) { + function sortTable(results, column, key, tbody, nest_level, columns) { // Find the table header element for the given key - let th = null; - const tableHeaders = document.querySelectorAll('th'); // Select all table headers - tableHeaders.forEach(header => { - if (header.textContent.trim().toLowerCase() === key.toLowerCase()) { - th = header; - } - }); + const tableHeaders = document.querySelectorAll('th'); + let th = Array.from(tableHeaders).find(header => header.textContent === key); if (!th) { console.error(`No table header found for key: ${key}`); return; } - // Determine the current sort direction - let ascending = th.getAttribute('data-sort-direction') === 'asc' ? false : true; + const ascending = th.getAttribute('data-sort-direction') === 'asc'; + th.setAttribute('data-sort-direction', ascending ? 'desc' : 'asc'); - // Toggle the sort direction for the next click - th.setAttribute('data-sort-direction', ascending ? 'asc' : 'desc'); - - // Sort the results array by the given key results.sort((a, b) => { - if (a[key] < b[key]) return ascending ? -1 : 1; - if (a[key] > b[key]) return ascending ? 1 : -1; + if (a[column] < b[column]) return ascending ? -1 : 1; + if (a[column] > b[column]) return ascending ? 1 : -1; return 0; }); + // Clear the existing rows in tbody + tbody.innerHTML = ''; + // Re-populate the table with sorted data populateTableRows(tbody, results, columns, nest_level); } - function loadJSON(PR, sha, nameParams) { + function loadResultsJSON(PR, sha, nameParams) { const infoElement = document.getElementById('info-container'); let lastModifiedTime = null; const task = nameParams[0].toLowerCase(); @@ -753,22 +754,61 @@ } }); - if (PR) { - addKeyValueToStatus("PR", PR) - } else { - console.error("TODO") - } - addKeyValueToStatus("sha", null, [sha, 'lala']); - if (nameParams[1]) { - addKeyValueToStatus("job", nameParams[1]); - } - addKeyValueToStatus("workflow", nameParams[0]); + let path_commits_json = ''; + let commitsArray = []; - if (PR && sha && root_name) { - loadJSON(PR, sha, nameParams); + if (PR) { + addKeyValueToStatus("PR", PR); + const baseUrl = window.location.origin + window.location.pathname.replace('/json.html', ''); + path_commits_json = `${baseUrl}/${encodeURIComponent(PR)}/commits.json`; } else { - document.getElementById('title').textContent = 'Error: Missing required URL parameters: PR, sha, or name_0'; + // Placeholder for a different path when PR is missing + console.error("PR parameter is missing. Setting alternate commits path."); + path_commits_json = '/path/to/alternative/commits.json'; } + + function loadCommitsArray(path) { + return fetch(path, { cache: "no-cache" }) + .then(response => { + if (!response.ok) { + console.error(`HTTP error! status: ${response.status}`) + return []; + } + return response.json(); + }) + .then(data => { + if (Array.isArray(data) && data.every(item => typeof item === 'object' && item.hasOwnProperty('sha'))) { + return data.map(item => item.sha); + } else { + throw new Error('Invalid data format: expected array of objects with a "sha" key'); + } + }) + .catch(error => { + console.error('Error loading commits JSON:', error); + return []; // Return an empty array if an error occurs + }); + } + + loadCommitsArray(path_commits_json) + .then(data => { + commitsArray = data; + }) + .finally(() => { + // Proceed with the rest of the initialization + addKeyValueToStatus("sha", sha || "latest", commitsArray.concat(["latest"])); + + if (nameParams[1]) { + addKeyValueToStatus("job", nameParams[1]); + } + addKeyValueToStatus("workflow", nameParams[0]); + + // Check if all required parameters are present to load JSON + if (PR && sha && root_name) { + loadResultsJSON(PR, sha, nameParams); + } else { + document.getElementById('title').textContent = 'Error: Missing required URL parameters: PR, sha, or name_0'; + } + }); } window.onload = init; diff --git a/ci/praktika/mangle.py b/ci/praktika/mangle.py index 89fc52cf849..bca33f9e660 100644 --- a/ci/praktika/mangle.py +++ b/ci/praktika/mangle.py @@ -58,7 +58,6 @@ def _update_workflow_artifacts(workflow): artifact_job = {} for job in workflow.jobs: for artifact_name in job.provides: - assert artifact_name not in artifact_job artifact_job[artifact_name] = job.name for artifact in workflow.artifacts: artifact._provided_by = artifact_job[artifact.name] diff --git a/ci/praktika/native_jobs.py b/ci/praktika/native_jobs.py index f7fd4ca190b..16ffa9056e9 100644 --- a/ci/praktika/native_jobs.py +++ b/ci/praktika/native_jobs.py @@ -151,7 +151,7 @@ def _config_workflow(workflow: Workflow.Config, job_name): status = Result.Status.ERROR print("ERROR: ", info) else: - Shell.check(f"{Settings.PYTHON_INTERPRETER} -m praktika --generate") + assert Shell.check(f"{Settings.PYTHON_INTERPRETER} -m praktika yaml") exit_code, output, err = Shell.get_res_stdout_stderr( f"git diff-index HEAD -- {Settings.WORKFLOW_PATH_PREFIX}" ) @@ -250,6 +250,9 @@ def _config_workflow(workflow: Workflow.Config, job_name): info_lines.append(job_name + ": " + info) results.append(result_) + if workflow.enable_merge_commit: + assert False, "NOT implemented" + # config: if workflow.dockers: print("Calculate docker's digests") diff --git a/ci/praktika/param.py b/ci/praktika/param.py new file mode 100644 index 00000000000..f5727198e0d --- /dev/null +++ b/ci/praktika/param.py @@ -0,0 +1,8 @@ +from praktika._environment import _Environment + + +# TODO: find better place and/or right storage for parameter +def get_param(): + env = _Environment.get() + assert env.PARAMETER + return env.PARAMETER diff --git a/ci/praktika/result.py b/ci/praktika/result.py index 2ba8309ad60..f473cf3ed05 100644 --- a/ci/praktika/result.py +++ b/ci/praktika/result.py @@ -1,7 +1,6 @@ import dataclasses import datetime import sys -from collections.abc import Container from pathlib import Path from typing import Any, Dict, List, Optional @@ -68,8 +67,9 @@ class Result(MetaClasses.Serializable): if isinstance(status, bool): status = Result.Status.SUCCESS if status else Result.Status.FAILED if not results and not status: - print("ERROR: Either .results or .status must be provided") - raise + Utils.raise_with_error( + f"Either .results ({results}) or .status ({status}) must be provided" + ) if not name: name = _Environment.get().JOB_NAME if not name: @@ -78,10 +78,10 @@ class Result(MetaClasses.Serializable): result_status = status or Result.Status.SUCCESS infos = [] if info: - if isinstance(info, Container): - infos += info + if isinstance(info, str): + infos += [info] else: - infos.append(info) + infos += info if results and not status: for result in results: if result.status not in (Result.Status.SUCCESS, Result.Status.FAILED): @@ -112,7 +112,7 @@ class Result(MetaClasses.Serializable): return self.status not in (Result.Status.PENDING, Result.Status.RUNNING) def is_running(self): - return self.status not in (Result.Status.RUNNING,) + return self.status in (Result.Status.RUNNING,) def is_ok(self): return self.status in (Result.Status.SKIPPED, Result.Status.SUCCESS) @@ -180,6 +180,11 @@ class Result(MetaClasses.Serializable): ) return self + def set_timing(self, stopwatch: Utils.Stopwatch): + self.start_time = stopwatch.start_time + self.duration = stopwatch.duration + return self + def update_sub_result(self, result: "Result"): assert self.results, "BUG?" for i, result_ in enumerate(self.results): diff --git a/ci/praktika/runner.py b/ci/praktika/runner.py index 823c7e0f36d..5db1a89ce99 100644 --- a/ci/praktika/runner.py +++ b/ci/praktika/runner.py @@ -125,15 +125,24 @@ class Runner: return 0 def _run(self, workflow, job, docker="", no_docker=False, param=None): + # re-set envs for local run + env = _Environment.get() + env.JOB_NAME = job.name + env.PARAMETER = job.parameter + env.dump() + if param: if not isinstance(param, str): Utils.raise_with_error( f"Custom param for local tests must be of type str, got [{type(param)}]" ) - env = _Environment.get() - env.dump() if job.run_in_docker and not no_docker: + job.run_in_docker, docker_settings = ( + job.run_in_docker.split("+")[0], + job.run_in_docker.split("+")[1:], + ) + from_root = "root" in docker_settings if ":" in job.run_in_docker: docker_name, docker_tag = job.run_in_docker.split(":") print( @@ -145,7 +154,7 @@ class Runner: RunConfig.from_fs(workflow.name).digest_dockers[job.run_in_docker], ) docker = docker or f"{docker_name}:{docker_tag}" - cmd = f"docker run --rm --user \"$(id -u):$(id -g)\" -e PYTHONPATH='{Settings.DOCKER_WD}:{Settings.DOCKER_WD}/ci' --volume ./:{Settings.DOCKER_WD} --volume {Settings.TEMP_DIR}:{Settings.TEMP_DIR} --workdir={Settings.DOCKER_WD} {docker} {job.command}" + cmd = f"docker run --rm --name praktika {'--user $(id -u):$(id -g)' if not from_root else ''} -e PYTHONPATH='{Settings.DOCKER_WD}:{Settings.DOCKER_WD}/ci' --volume ./:{Settings.DOCKER_WD} --volume {Settings.TEMP_DIR}:{Settings.TEMP_DIR} --workdir={Settings.DOCKER_WD} {docker} {job.command}" else: cmd = job.command @@ -226,7 +235,8 @@ class Runner: print(info) result.set_info(info).set_status(Result.Status.ERROR).dump() - result.set_files(files=[Settings.RUN_LOG]) + if not result.is_ok(): + result.set_files(files=[Settings.RUN_LOG]) result.update_duration().dump() if result.info and result.status != Result.Status.SUCCESS: @@ -329,7 +339,7 @@ class Runner: workflow, job, pr=pr, branch=branch, sha=sha ) - if res: + if res and (not local_run or pr or sha or branch): res = False print(f"=== Pre run script [{job.name}], workflow [{workflow.name}] ===") try: diff --git a/ci/praktika/s3.py b/ci/praktika/s3.py index 8cfb70a9076..04a08622dcd 100644 --- a/ci/praktika/s3.py +++ b/ci/praktika/s3.py @@ -52,7 +52,7 @@ class S3: cmd += " --content-type text/plain" res = cls.run_command_with_retries(cmd) if not res: - raise + raise RuntimeError() bucket = s3_path.split("/")[0] endpoint = Settings.S3_BUCKET_TO_HTTP_ENDPOINT[bucket] assert endpoint diff --git a/ci/praktika/workflow.py b/ci/praktika/workflow.py index 41e8056f9ef..8c5ec12440f 100644 --- a/ci/praktika/workflow.py +++ b/ci/praktika/workflow.py @@ -31,6 +31,7 @@ class Workflow: enable_report: bool = False enable_merge_ready_status: bool = False enable_cidb: bool = False + enable_merge_commit: bool = False def is_event_pull_request(self): return self.event == Workflow.Event.PULL_REQUEST diff --git a/ci/praktika/yaml_generator.py b/ci/praktika/yaml_generator.py index fb918b4ddba..1422a835663 100644 --- a/ci/praktika/yaml_generator.py +++ b/ci/praktika/yaml_generator.py @@ -80,6 +80,9 @@ jobs: steps: - name: Checkout code uses: actions/checkout@v4 + with: + clear-repository: true + ref: ${{{{ github.event.pull_reguest.head.sha }}}} {JOB_ADDONS} - name: Prepare env script run: | diff --git a/ci/settings/definitions.py b/ci/settings/definitions.py index c67bdee015b..33173756924 100644 --- a/ci/settings/definitions.py +++ b/ci/settings/definitions.py @@ -8,23 +8,30 @@ class RunnerLabels: CI_SERVICES = "ci_services" CI_SERVICES_EBS = "ci_services_ebs" BUILDER = "builder" + STYLE_CHECKER = "style-checker" BASE_BRANCH = "master" +azure_secret = Secret.Config( + name="azure_connection_string", + type=Secret.Type.AWS_SSM_VAR, +) + SECRETS = [ Secret.Config( name="dockerhub_robot_password", type=Secret.Type.AWS_SSM_VAR, ), - Secret.Config( - name="woolenwolf_gh_app.clickhouse-app-id", - type=Secret.Type.AWS_SSM_SECRET, - ), - Secret.Config( - name="woolenwolf_gh_app.clickhouse-app-key", - type=Secret.Type.AWS_SSM_SECRET, - ), + azure_secret, + # Secret.Config( + # name="woolenwolf_gh_app.clickhouse-app-id", + # type=Secret.Type.AWS_SSM_SECRET, + # ), + # Secret.Config( + # name="woolenwolf_gh_app.clickhouse-app-key", + # type=Secret.Type.AWS_SSM_SECRET, + # ), ] DOCKERS = [ @@ -118,12 +125,12 @@ DOCKERS = [ # platforms=Docker.Platforms.arm_amd, # depends_on=["clickhouse/test-base"], # ), - # Docker.Config( - # name="clickhouse/stateless-test", - # path="./ci/docker/test/stateless", - # platforms=Docker.Platforms.arm_amd, - # depends_on=["clickhouse/test-base"], - # ), + Docker.Config( + name="clickhouse/stateless-test", + path="./ci/docker/stateless-test", + platforms=Docker.Platforms.arm_amd, + depends_on=[], + ), # Docker.Config( # name="clickhouse/stateful-test", # path="./ci/docker/test/stateful", @@ -230,5 +237,6 @@ DOCKERS = [ class JobNames: STYLE_CHECK = "Style Check" FAST_TEST = "Fast test" - BUILD_AMD_DEBUG = "Build amd64 debug" + BUILD = "Build" + BUILD_AMD_DEBUG = "Build (amd, debug)" STATELESS_TESTS = "Stateless tests (amd, debug)" diff --git a/ci/workflows/pull_request.py b/ci/workflows/pull_request.py index c7715b40fca..10dd77a0414 100644 --- a/ci/workflows/pull_request.py +++ b/ci/workflows/pull_request.py @@ -13,7 +13,8 @@ from ci.settings.definitions import ( class ArtifactNames: - ch_debug_binary = "clickhouse_debug_binary" + CH_AMD_DEBUG = "CH_AMD_DEBUG" + CH_AMD_RELEASE = "CH_AMD_RELEASE" style_check_job = Job.Config( @@ -37,10 +38,10 @@ fast_test_job = Job.Config( ), ) -job_build_amd_debug = Job.Config( - name=JobNames.BUILD_AMD_DEBUG, +amd_build_jobs = Job.Config( + name=JobNames.BUILD, runs_on=[RunnerLabels.BUILDER], - command="python3 ./ci/jobs/build_clickhouse.py amd_debug", + command="python3 ./ci/jobs/build_clickhouse.py", run_in_docker="clickhouse/fasttest", digest_config=Job.CacheDigestConfig( include_paths=[ @@ -56,20 +57,30 @@ job_build_amd_debug = Job.Config( "./tests/ci/version_helper.py", ], ), - provides=[ArtifactNames.ch_debug_binary], +).parametrize( + parameter=["amd_debug", "amd_release"], + provides=[[ArtifactNames.CH_AMD_DEBUG], [ArtifactNames.CH_AMD_RELEASE]], ) -stateless_tests_job = Job.Config( +statless_batch_num = 2 +stateless_tests_amd_debug_jobs = Job.Config( name=JobNames.STATELESS_TESTS, runs_on=[RunnerLabels.BUILDER], command="python3 ./ci/jobs/functional_stateless_tests.py amd_debug", - run_in_docker="clickhouse/fasttest:latest", + run_in_docker="clickhouse/stateless-test", digest_config=Job.CacheDigestConfig( include_paths=[ "./ci/jobs/functional_stateless_tests.py", ], ), - requires=[ArtifactNames.ch_debug_binary], + requires=[ArtifactNames.CH_AMD_DEBUG], +).parametrize( + parameter=[ + f"parallel {i+1}/{statless_batch_num}" for i in range(statless_batch_num) + ] + + ["non-parallel"], + runs_on=[[RunnerLabels.BUILDER] for _ in range(statless_batch_num)] + + [[RunnerLabels.STYLE_CHECKER]], ) workflow = Workflow.Config( @@ -79,15 +90,20 @@ workflow = Workflow.Config( jobs=[ style_check_job, fast_test_job, - job_build_amd_debug, - stateless_tests_job, + *amd_build_jobs, + *stateless_tests_amd_debug_jobs, ], artifacts=[ Artifact.Config( - name=ArtifactNames.ch_debug_binary, + name=ArtifactNames.CH_AMD_DEBUG, type=Artifact.Type.S3, path=f"{Settings.TEMP_DIR}/build/programs/clickhouse", - ) + ), + Artifact.Config( + name=ArtifactNames.CH_AMD_RELEASE, + type=Artifact.Type.S3, + path=f"{Settings.TEMP_DIR}/build/programs/clickhouse", + ), ], dockers=DOCKERS, secrets=SECRETS, @@ -101,8 +117,11 @@ WORKFLOWS = [ ] # type: List[Workflow.Config] -if __name__ == "__main__": - # local job test inside praktika environment - from praktika.runner import Runner - - Runner().run(workflow, fast_test_job, docker="fasttest", local_run=True) +# if __name__ == "__main__": +# # local job test inside praktika environment +# from praktika.runner import Runner +# from praktika.digest import Digest +# +# print(Digest().calc_job_digest(amd_debug_build_job)) +# +# Runner().run(workflow, fast_test_job, docker="fasttest", local_run=True) diff --git a/tests/clickhouse-test b/tests/clickhouse-test index 100a6358dcf..3396b10814a 100755 --- a/tests/clickhouse-test +++ b/tests/clickhouse-test @@ -2153,9 +2153,9 @@ class TestSuite: self.sequential_tests = [] self.parallel_tests = [] for test_name in self.all_tests: - if self.is_sequential_test(test_name): + if self.is_sequential_test(test_name) and not args.no_sequential: self.sequential_tests.append(test_name) - else: + elif not args.no_parallel: self.parallel_tests.append(test_name) def is_sequential_test(self, test_name): @@ -3290,7 +3290,10 @@ def parse_args(): help='Replace random database name with "default" in stderr', ) parser.add_argument( - "--parallel", default="1/1", help="One parallel test run number/total" + "--no-sequential", action="store_true", help="Not run no-parallel" + ) + parser.add_argument( + "--no-parallel", action="store_true", help="Run only no-parallel" ) parser.add_argument( "-j", "--jobs", default=1, nargs="?", type=int, help="Run all tests in parallel" @@ -3339,7 +3342,7 @@ def parse_args(): parser.add_argument( "--sequential", nargs="+", - help="Run these tests sequentially even if --parallel specified", + help="Run all tests sequentially", ) parser.add_argument( "--no-long", action="store_true", dest="no_long", help="Do not run long tests" diff --git a/tests/config/config.d/ssl_certs.xml b/tests/config/config.d/ssl_certs.xml index 26b679f39df..c20fef89e00 100644 --- a/tests/config/config.d/ssl_certs.xml +++ b/tests/config/config.d/ssl_certs.xml @@ -1,8 +1,8 @@ - /etc/clickhouse-server/server.crt - /etc/clickhouse-server/server.key + /tmp/praktika/etc/clickhouse-server/server.crt + /tmp/praktika/etc/clickhouse-server/server.key diff --git a/tests/config/install.sh b/tests/config/install.sh index be47298f6a4..cdae5741fce 100755 --- a/tests/config/install.sh +++ b/tests/config/install.sh @@ -9,6 +9,21 @@ DEST_SERVER_PATH="${1:-/etc/clickhouse-server}" DEST_CLIENT_PATH="${2:-/etc/clickhouse-client}" SRC_PATH="$( cd "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )" +shift # DEST_SERVER_PATH +shift # DEST_CLIENT_PATH + +FAST_TEST=0 +S3_STORAGE=0 + +while [[ "$#" -gt 0 ]]; do + case $1 in + --fast-test) FAST_TEST=1 ;; + --s3-storage) S3_STORAGE=1 ;; + *) echo "Unknown option: $1" ; exit 1 ;; + esac + shift +done + echo "Going to install test configs from $SRC_PATH into $DEST_SERVER_PATH" mkdir -p $DEST_SERVER_PATH/config.d/ @@ -72,9 +87,8 @@ ln -sf $SRC_PATH/config.d/serverwide_trace_collector.xml $DEST_SERVER_PATH/confi ln -sf $SRC_PATH/config.d/rocksdb.xml $DEST_SERVER_PATH/config.d/ # Not supported with fasttest. -if [ "${DEST_SERVER_PATH}" = "/etc/clickhouse-server" ] -then - ln -sf $SRC_PATH/config.d/legacy_geobase.xml $DEST_SERVER_PATH/config.d/ +if [ "${DEST_SERVER_PATH}" = "/etc/clickhouse-server" ] || [ "$FAST_TEST" != "1" ]; then + ln -sf "$SRC_PATH/config.d/legacy_geobase.xml" "$DEST_SERVER_PATH/config.d/" fi ln -sf $SRC_PATH/users.d/log_queries.xml $DEST_SERVER_PATH/users.d/ @@ -185,7 +199,7 @@ elif [[ "$USE_AZURE_STORAGE_FOR_MERGE_TREE" == "1" ]]; then ln -sf $SRC_PATH/config.d/azure_storage_policy_by_default.xml $DEST_SERVER_PATH/config.d/ fi -if [[ -n "$EXPORT_S3_STORAGE_POLICIES" ]]; then +if [[ "$EXPORT_S3_STORAGE_POLICIES" == "1" ]] || [[ "$S3_STORAGE" = "1" ]]; then ln -sf $SRC_PATH/config.d/azure_storage_conf.xml $DEST_SERVER_PATH/config.d/ ln -sf $SRC_PATH/config.d/storage_conf.xml $DEST_SERVER_PATH/config.d/ ln -sf $SRC_PATH/config.d/storage_conf_02944.xml $DEST_SERVER_PATH/config.d/ @@ -195,7 +209,7 @@ if [[ -n "$EXPORT_S3_STORAGE_POLICIES" ]]; then ln -sf $SRC_PATH/users.d/s3_cache_new.xml $DEST_SERVER_PATH/users.d/ fi -if [[ -n "$USE_DATABASE_REPLICATED" ]] && [[ "$USE_DATABASE_REPLICATED" -eq 1 ]]; then +if [[ "$USE_DATABASE_REPLICATED" == "1" ]]; then ln -sf $SRC_PATH/users.d/database_replicated.xml $DEST_SERVER_PATH/users.d/ ln -sf $SRC_PATH/config.d/database_replicated.xml $DEST_SERVER_PATH/config.d/ rm /etc/clickhouse-server/config.d/zookeeper.xml diff --git a/tests/docker_scripts/setup_minio.sh b/tests/docker_scripts/setup_minio.sh index 40e93e713a1..837c05a9c5d 100755 --- a/tests/docker_scripts/setup_minio.sh +++ b/tests/docker_scripts/setup_minio.sh @@ -5,6 +5,12 @@ set -euxf -o pipefail export MINIO_ROOT_USER=${MINIO_ROOT_USER:-clickhouse} export MINIO_ROOT_PASSWORD=${MINIO_ROOT_PASSWORD:-clickhouse} +if [ -d "$TEMP_DIR" ]; then + cd "$TEMP_DIR" + # add / for minio mc in docker + PATH="/:.:$PATH" +fi + usage() { echo $"Usage: $0 (default path: /usr/share/clickhouse-test)" exit 1 @@ -70,9 +76,10 @@ download_minio() { } start_minio() { + pwd mkdir -p ./minio_data - ./minio --version - ./minio server --address ":11111" ./minio_data & + minio --version + minio server --address ":11111" ./minio_data & wait_for_it lsof -i :11111 sleep 5 @@ -80,12 +87,14 @@ start_minio() { setup_minio() { local test_type=$1 - ./mc alias set clickminio http://localhost:11111 clickhouse clickhouse - ./mc admin user add clickminio test testtest - ./mc admin policy attach clickminio readwrite --user=test - ./mc mb --ignore-existing clickminio/test + echo "setup_minio(), test_type=$test_type" + mc alias set clickminio http://localhost:11111 clickhouse clickhouse + mc admin user add clickminio test testtest + mc admin policy attach clickminio readwrite --user=test ||: + mc mb --ignore-existing clickminio/test if [ "$test_type" = "stateless" ]; then - ./mc anonymous set public clickminio/test + echo "Create @test bucket in minio" + mc anonymous set public clickminio/test fi } @@ -95,12 +104,13 @@ upload_data() { local query_dir=$1 local test_path=$2 local data_path=${test_path}/queries/${query_dir}/data_minio + echo "upload_data() data_path=$data_path" # iterating over globs will cause redundant file variable to be # a path to a file, not a filename # shellcheck disable=SC2045 if [ -d "${data_path}" ]; then - ./mc cp --recursive "${data_path}"/ clickminio/test/ + mc cp --recursive "${data_path}"/ clickminio/test/ fi } @@ -138,7 +148,7 @@ wait_for_it() { main() { local query_dir query_dir=$(check_arg "$@") - if [ ! -f ./minio ]; then + if ! (minio --version && mc --version); then download_minio fi start_minio From 98ee0893318bcfd4e0d63b564f513b37579bd3c8 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Tue, 5 Nov 2024 17:31:47 +0000 Subject: [PATCH 061/244] Cleanup --- tests/queries/0_stateless/03261_pr_semi_anti_join.sql | 1 - 1 file changed, 1 deletion(-) diff --git a/tests/queries/0_stateless/03261_pr_semi_anti_join.sql b/tests/queries/0_stateless/03261_pr_semi_anti_join.sql index d2ea3725d6b..2d671756d6e 100644 --- a/tests/queries/0_stateless/03261_pr_semi_anti_join.sql +++ b/tests/queries/0_stateless/03261_pr_semi_anti_join.sql @@ -1,6 +1,5 @@ DROP TABLE IF EXISTS t1 SYNC; DROP TABLE IF EXISTS t2 SYNC; -create table X (id Int32, x_a String, x_b Nullable(Int32)) engine ReplicatedMergeTree('/clickhouse/{database}/X', '1') order by tuple(); CREATE TABLE t1 (x UInt32, s String) engine ReplicatedMergeTree('/clickhouse/{database}/t1', '1') order by tuple(); CREATE TABLE t2 (x UInt32, s String) engine ReplicatedMergeTree('/clickhouse/{database}/t2', '1') order by tuple(); From 6a8df5ea89724d7686f6c520bc436b7cb80294bd Mon Sep 17 00:00:00 2001 From: nauu Date: Wed, 6 Nov 2024 14:57:14 +0800 Subject: [PATCH 062/244] support the endpoint of oss accelerator --- src/IO/S3/URI.cpp | 12 ++++++++++-- src/IO/tests/gtest_s3_uri.cpp | 16 ++++++++++++++++ 2 files changed, 26 insertions(+), 2 deletions(-) diff --git a/src/IO/S3/URI.cpp b/src/IO/S3/URI.cpp index 7c6a21941eb..ad746ff3326 100644 --- a/src/IO/S3/URI.cpp +++ b/src/IO/S3/URI.cpp @@ -37,7 +37,7 @@ URI::URI(const std::string & uri_, bool allow_archive_path_syntax) /// Case when bucket name represented in domain name of S3 URL. /// E.g. (https://bucket-name.s3.region.amazonaws.com/key) /// https://docs.aws.amazon.com/AmazonS3/latest/dev/VirtualHosting.html#virtual-hosted-style-access - static const RE2 virtual_hosted_style_pattern(R"((.+)\.(s3express[\-a-z0-9]+|s3|cos|obs|oss|eos)([.\-][a-z0-9\-.:]+))"); + static const RE2 virtual_hosted_style_pattern(R"((.+)\.(s3express[\-a-z0-9]+|s3|cos|obs|oss-data-acc|oss|eos)([.\-][a-z0-9\-.:]+))"); /// Case when AWS Private Link Interface is being used /// E.g. (bucket.vpce-07a1cd78f1bd55c5f-j3a3vg6w.s3.us-east-1.vpce.amazonaws.com/bucket-name/key) @@ -115,7 +115,15 @@ URI::URI(const std::string & uri_, bool allow_archive_path_syntax) && re2::RE2::FullMatch(uri.getAuthority(), virtual_hosted_style_pattern, &bucket, &name, &endpoint_authority_from_uri)) { is_virtual_hosted_style = true; - endpoint = uri.getScheme() + "://" + name + endpoint_authority_from_uri; + if (name == "oss-data-acc") + { + bucket = bucket.substr(0, bucket.find(".")); + endpoint = uri.getScheme() + "://" + uri.getHost().substr(bucket.length() + 1); + } + else + { + endpoint = uri.getScheme() + "://" + name + endpoint_authority_from_uri; + } validateBucket(bucket, uri); if (!uri.getPath().empty()) diff --git a/src/IO/tests/gtest_s3_uri.cpp b/src/IO/tests/gtest_s3_uri.cpp index 8696fab0616..6167313b634 100644 --- a/src/IO/tests/gtest_s3_uri.cpp +++ b/src/IO/tests/gtest_s3_uri.cpp @@ -212,6 +212,22 @@ TEST(S3UriTest, validPatterns) ASSERT_EQ("", uri.version_id); ASSERT_EQ(true, uri.is_virtual_hosted_style); } + { + S3::URI uri("https://bucket-test1.oss-cn-beijing-internal.aliyuncs.com/ab-test"); + ASSERT_EQ("https://oss-cn-beijing-internal.aliyuncs.com", uri.endpoint); + ASSERT_EQ("bucket-test1", uri.bucket); + ASSERT_EQ("ab-test", uri.key); + ASSERT_EQ("", uri.version_id); + ASSERT_EQ(true, uri.is_virtual_hosted_style); + } + { + S3::URI uri("https://bucket-test.cn-beijing-internal.oss-data-acc.aliyuncs.com/ab-test"); + ASSERT_EQ("https://cn-beijing-internal.oss-data-acc.aliyuncs.com", uri.endpoint); + ASSERT_EQ("bucket-test", uri.bucket); + ASSERT_EQ("ab-test", uri.key); + ASSERT_EQ("", uri.version_id); + ASSERT_EQ(true, uri.is_virtual_hosted_style); + } } TEST(S3UriTest, versionIdChecks) From 699b9d40263078285a8fce6a031bc74ce72c16d3 Mon Sep 17 00:00:00 2001 From: kevinyhzou Date: Wed, 6 Nov 2024 20:20:44 +0800 Subject: [PATCH 063/244] fix comments --- src/Functions/parseDateTime.cpp | 277 ++++++++++++++++++++++---------- 1 file changed, 194 insertions(+), 83 deletions(-) diff --git a/src/Functions/parseDateTime.cpp b/src/Functions/parseDateTime.cpp index 65bc65fb45c..976be53a21e 100644 --- a/src/Functions/parseDateTime.cpp +++ b/src/Functions/parseDateTime.cpp @@ -57,8 +57,15 @@ namespace Null }; + enum class ReturnType: uint8_t + { + DateTime, + DateTime64 + }; + constexpr Int32 minYear = 1970; constexpr Int32 maxYear = 2106; + constexpr Int32 maxPrecisionOfDateTime64 = 6; const std::unordered_map> dayOfWeekMap{ {"mon", {"day", 1}}, @@ -570,8 +577,8 @@ namespace } }; - /// _FUNC_(str[, format, timezone]) - template + /// _FUNC_(str[scale, format, timezone]) + template class FunctionParseDateTimeImpl : public IFunction { public: @@ -602,79 +609,112 @@ namespace {"time", static_cast(&isString), nullptr, "String"} }; - FunctionArgumentDescriptors optional_args{ - {"format", static_cast(&isString), nullptr, "String"}, - {"timezone", static_cast(&isString), &isColumnConst, "const String"} - }; - + FunctionArgumentDescriptors optional_args; + if constexpr (return_type == ReturnType::DateTime64) + { + optional_args = { + {"precision or format", static_cast([](const IDataType & data_type) -> bool { + return isUInt(data_type) || isString(data_type); + }), nullptr, "Number or String"}, + {"format", static_cast(&isString), nullptr, "String"}, + {"timezone", static_cast(&isString), &isColumnConst, "const String"} + }; + } + else + optional_args = { + {"format", static_cast(&isString), nullptr, "String"}, + {"timezone", static_cast(&isString), &isColumnConst, "const String"} + }; validateFunctionArguments(*this, arguments, mandatory_args, optional_args); String time_zone_name = getTimeZone(arguments).getTimeZone(); - DataTypePtr date_type = nullptr; - if constexpr (parseDateTime64) + DataTypePtr data_type; + if constexpr (return_type == ReturnType::DateTime64) { - String format = getFormat(arguments); - std::vector instructions = parseFormat(format); - UInt32 scale = 0; - if (!instructions.empty()) + if (arguments.size() == 1) + return std::make_shared(0, time_zone_name); + else { - for (const auto & ins : instructions) + UInt32 precision = 0; + if (isUInt(arguments[1].type)) { - if (scale > 0) - break; - const String fragment = ins.getFragment(); + const auto * col_precision = checkAndGetColumnConst(arguments[1].column.get()); + if (col_precision) + precision = col_precision->getValue(); + else + throw Exception(ErrorCodes::BAD_ARGUMENTS, + "The input precision value may exceed the max value of `DateTime64`: {}.", + maxPrecisionOfDateTime64); + } + /// Construct the return type `DataTypDateTime64` with precision and time zone name. The precision value can be specified or be extracted + /// from the format string by computing how many 'S' characters are contained in the format's micorsceond fragment. + String format = getFormat(arguments, precision); + std::vector instructions = parseFormat(format); + for (const auto & instruction : instructions) + { + const String & fragment = instruction.getFragment(); + UInt32 val = 0; for (char ch : fragment) { if (ch != 'S') { - scale = 0; + val = 0; break; } else - scale++; + val++; } + /// If the precision is already specified by the second parameter, but it not equals the value that extract from the format string, + /// then we should throw an exception; If the precision is not specified, then we set its value as the extracted one. + if (val != 0 && precision != 0 && val != precision) + throw Exception(ErrorCodes::BAD_ARGUMENTS, + "The precision of input format string {} not equals the given precision value {}.", + format, + precision); + else if (precision == 0 && val != 0) + precision = val; } + if (precision > maxPrecisionOfDateTime64) + throw Exception(ErrorCodes::BAD_ARGUMENTS, + "The precision of the input format string {} exceed the max precision value {}.", + format, + maxPrecisionOfDateTime64); + data_type = std::make_shared(precision, time_zone_name); } - date_type = std::make_shared(scale, time_zone_name); } else - date_type = std::make_shared(time_zone_name); + data_type = std::make_shared(time_zone_name); if (error_handling == ErrorHandling::Null) - return std::make_shared(date_type); - return date_type; + return std::make_shared(data_type); + return data_type; } ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override { - ColumnUInt8::MutablePtr col_null_map; + DataTypePtr non_null_result_type; if constexpr (error_handling == ErrorHandling::Null) - col_null_map = ColumnUInt8::create(input_rows_count, 0); - if constexpr (parseDateTime64) + non_null_result_type = removeNullable(result_type); + else + non_null_result_type = result_type; + + if constexpr (return_type == ReturnType::DateTime64) { - const DataTypeDateTime64 * datatime64_type = checkAndGetDataType(removeNullable(result_type).get()); - auto col_res = ColumnDateTime64::create(input_rows_count, datatime64_type->getScale()); - PaddedPODArray & res_data = col_res->getData(); - executeImpl2(arguments, result_type, input_rows_count, res_data, col_null_map); - if constexpr (error_handling == ErrorHandling::Null) - return ColumnNullable::create(std::move(col_res), std::move(col_null_map)); - else - return col_res; + const auto * datatime64_type = checkAndGetDataType(non_null_result_type.get()); + MutableColumnPtr col_res = ColumnDateTime64::create(input_rows_count, datatime64_type->getScale()); + ColumnDateTime64 * col_datetime64 = assert_cast(col_res.get()); + return executeImpl2(arguments, result_type, input_rows_count, col_res, col_datetime64->getData()); } else { - auto col_res = ColumnDateTime::create(input_rows_count); - PaddedPODArray & res_data = col_res->getData(); - executeImpl2(arguments, result_type, input_rows_count, res_data, col_null_map); - if constexpr (error_handling == ErrorHandling::Null) - return ColumnNullable::create(std::move(col_res), std::move(col_null_map)); - else - return col_res; + MutableColumnPtr col_res = ColumnDateTime::create(input_rows_count); + ColumnDateTime * col_datetime = assert_cast(col_res.get()); + return executeImpl2(arguments, result_type, input_rows_count, col_res, col_datetime->getData()); } } template - void executeImpl2(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count, - PaddedPODArray & res_data, ColumnUInt8::MutablePtr & col_null_map) const + ColumnPtr executeImpl2(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count, + MutableColumnPtr & col_res, PaddedPODArray & res_data) const { const auto * col_str = checkAndGetColumn(arguments[0].column.get()); if (!col_str) @@ -683,8 +723,21 @@ namespace "Illegal column {} of first ('str') argument of function {}. Must be string.", arguments[0].column->getName(), getName()); + + ColumnUInt8::MutablePtr col_null_map; + if constexpr (error_handling == ErrorHandling::Null) + col_null_map = ColumnUInt8::create(input_rows_count, 0); - String format = getFormat(arguments); + Int64 multiplier = 0; + UInt32 precision = 0; + if constexpr (return_type == ReturnType::DateTime64) + { + const DataTypeDateTime64 * datatime64_type = checkAndGetDataType(removeNullable(result_type).get()); + precision = datatime64_type->getScale(); + multiplier = DecimalUtils::scaleMultiplier(precision); + } + + String format = getFormat(arguments, precision); const auto & time_zone = getTimeZone(arguments); std::vector instructions = parseFormat(format); @@ -733,8 +786,8 @@ namespace Int64OrError result = 0; - /// Ensure all input was consumed - if (!parseDateTime64 && cur < end) + /// Ensure all input was consumed when the return type is `DateTime`. + if (return_type == ReturnType::DateTime && cur < end) { result = tl::unexpected(ErrorCodeAndMessage( ErrorCodes::CANNOT_PARSE_DATETIME, @@ -747,12 +800,8 @@ namespace { if (result = datetime.buildDateTime(time_zone); result.has_value()) { - if constexpr (parseDateTime64) - { - const DataTypeDateTime64 * datatime64_type = checkAndGetDataType(removeNullable(result_type).get()); - Int64 multiplier = DecimalUtils::scaleMultiplier(datatime64_type->getScale()); + if constexpr (return_type == ReturnType::DateTime64) res_data[i] = static_cast(*result) * multiplier + datetime.microsecond; - } else res_data[i] = static_cast(*result); } @@ -777,6 +826,10 @@ namespace } } } + if constexpr (error_handling == ErrorHandling::Null) + return ColumnNullable::create(std::move(col_res), std::move(col_null_map)); + else + return std::move(col_res); } @@ -808,7 +861,7 @@ namespace explicit Instruction(const String & literal_) : literal(literal_), fragment("LITERAL") { } explicit Instruction(String && literal_) : literal(std::move(literal_)), fragment("LITERAL") { } - String getFragment() const { return fragment; } + const String & getFragment() const { return fragment; } /// For debug [[maybe_unused]] String toString() const @@ -1695,7 +1748,7 @@ namespace } [[nodiscard]] - static PosOrError jodaMicroSecondOfSecond(size_t repetitions, Pos cur, Pos end, const String & fragment, DateTime & date) + static PosOrError jodaMicrosecondOfSecond(size_t repetitions, Pos cur, Pos end, const String & fragment, DateTime & date) { Int32 microsecond; ASSIGN_RESULT_OR_RETURN_ERROR(cur, (readNumberWithVariableLength(cur, end, false, false, false, repetitions, std::max(repetitions, 2uz), fragment, microsecond))) @@ -1704,25 +1757,25 @@ namespace } [[nodiscard]] - static PosOrError jodaTimezoneId(size_t, Pos cur, Pos end, const String &, DateTime & date) + static PosOrError jodaTimezone(size_t, Pos cur, Pos end, const String &, DateTime & date) { - String dateTimeZone; + String read_time_zone; while (cur <= end) { - dateTimeZone += *cur; + read_time_zone += *cur; ++cur; } - const DateLUTImpl & date_time_zone = DateLUT::instance(dateTimeZone); + const DateLUTImpl & date_time_zone = DateLUT::instance(read_time_zone); const auto result = date.buildDateTime(date_time_zone); if (result.has_value()) { - const auto timezoneOffset = date_time_zone.timezoneOffset(*result); + const DateLUTImpl::Time timezone_offset = date_time_zone.timezoneOffset(*result); date.has_time_zone_offset = true; - date.time_zone_offset = timezoneOffset; + date.time_zone_offset = timezone_offset; return cur; } else - RETURN_ERROR(ErrorCodes::CANNOT_PARSE_DATETIME, "Unable to build date time from timezone {}", dateTimeZone) + RETURN_ERROR(ErrorCodes::CANNOT_PARSE_DATETIME, "Unable to build date time from timezone {}", read_time_zone) } [[nodiscard]] @@ -1745,8 +1798,22 @@ namespace Int32 hour; ASSIGN_RESULT_OR_RETURN_ERROR(cur, (readNumberWithVariableLength(cur, end, false, false, false, repetitions, std::max(repetitions, 2uz), fragment, hour))) + if (hour < 0 || hour > 23) + RETURN_ERROR( + ErrorCodes::CANNOT_PARSE_DATETIME, + "Unable to parse fragment {} from {} because of the hour of datetime not in range [0, 23]: {}", + fragment, + std::string_view(cur, end - cur), + std::string_view(cur, 1)) Int32 minute; ASSIGN_RESULT_OR_RETURN_ERROR(cur, (readNumberWithVariableLength(cur, end, false, false, false, repetitions, std::max(repetitions, 2uz), fragment, minute))) + if (minute < 0 || minute > 59) + RETURN_ERROR( + ErrorCodes::CANNOT_PARSE_DATETIME, + "Unable to parse fragment {} from {} because of the minute of datetime not in range [0, 59]: {}", + fragment, + std::string_view(cur, end - cur), + std::string_view(cur, 1)) date.has_time_zone_offset = true; date.time_zone_offset = sign * (hour * 3600 + minute * 60); return cur; @@ -2133,10 +2200,10 @@ namespace instructions.emplace_back(ACTION_ARGS_WITH_BIND(Instruction::jodaSecondOfMinute, repetitions)); break; case 'S': - instructions.emplace_back(ACTION_ARGS_WITH_BIND(Instruction::jodaMicroSecondOfSecond, repetitions)); + instructions.emplace_back(ACTION_ARGS_WITH_BIND(Instruction::jodaMicrosecondOfSecond, repetitions)); break; case 'z': - instructions.emplace_back(ACTION_ARGS_WITH_BIND(Instruction::jodaTimezoneId, repetitions)); + instructions.emplace_back(ACTION_ARGS_WITH_BIND(Instruction::jodaTimezone, repetitions)); break; case 'Z': instructions.emplace_back(ACTION_ARGS_WITH_BIND(Instruction::jodaTimezoneOffset, repetitions)); @@ -2156,26 +2223,45 @@ namespace } - String getFormat(const ColumnsWithTypeAndName & arguments) const + String getFormat(const ColumnsWithTypeAndName & arguments, UInt32 precision) const { - if (arguments.size() == 1) + size_t format_arg_index = 1; + if constexpr (return_type == ReturnType::DateTime64) { - if constexpr (parse_syntax == ParseSyntax::MySQL) - return "%Y-%m-%d %H:%i:%s"; + /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22.22.123', 3), then the format is treated + /// as default value `yyyy-MM-dd HH:mm:ss`. + /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 'yyyy-MM-dd HH:mm:ss.SSS')`, + /// then the second argument is the format. + /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS')`, + /// then the third argument is the format. + if (arguments.size() > 1 && isString(removeNullable(arguments[1].type))) + format_arg_index = 1; else - return "yyyy-MM-dd HH:mm:ss"; + format_arg_index = 2; + } + + if (arguments.size() <= format_arg_index) + { + String format; + if constexpr (parse_syntax == ParseSyntax::MySQL) + format = "%Y-%m-%d %H:%i:%s"; + else + format = "yyyy-MM-dd HH:mm:ss"; + if (precision > 0) + format += "." + String(precision, 'S'); + return format; } else { - if (!arguments[1].column || !isColumnConst(*arguments[1].column)) - throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Argument at index {} for function {} must be constant", 1, getName()); + if (!arguments[format_arg_index].column || !isColumnConst(*arguments[format_arg_index].column)) + throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Argument at index {} for function {} must be constant", format_arg_index, getName()); - const auto * col_format = checkAndGetColumnConst(arguments[1].column.get()); + const auto * col_format = checkAndGetColumnConst(arguments[format_arg_index].column.get()); if (!col_format) throw Exception( ErrorCodes::ILLEGAL_COLUMN, "Illegal column {} of second ('format') argument of function {}. Must be constant string.", - arguments[1].column->getName(), + arguments[format_arg_index].column->getName(), getName()); return col_format->getValue(); } @@ -2183,15 +2269,19 @@ namespace const DateLUTImpl & getTimeZone(const ColumnsWithTypeAndName & arguments) const { - if (arguments.size() < 3) + size_t timezone_arg_index = 2; + if constexpr (return_type == ReturnType::DateTime64) + timezone_arg_index = 3; + + if (arguments.size() <= timezone_arg_index) return DateLUT::instance(); - const auto * col = checkAndGetColumnConst(arguments[2].column.get()); + const auto * col = checkAndGetColumnConst(arguments[timezone_arg_index].column.get()); if (!col) throw Exception( ErrorCodes::ILLEGAL_COLUMN, "Illegal column {} of third ('timezone') argument of function {}. Must be constant String.", - arguments[2].column->getName(), + arguments[timezone_arg_index].column->getName(), getName()); String time_zone = col->getValue(); @@ -2229,6 +2319,21 @@ namespace static constexpr auto name = "parseDateTimeInJodaSyntaxOrNull"; }; + struct NameParseDateTime64 + { + static constexpr auto name = "parseDateTime64"; + }; + + struct NameParseDateTime64OrZero + { + static constexpr auto name = "parseDateTime64OrZero"; + }; + + struct NameParseDateTime64OrNull + { + static constexpr auto name = "parseDateTime64OrNull"; + }; + struct NameParseDateTime64InJodaSyntax { static constexpr auto name = "parseDateTime64InJodaSyntax"; @@ -2244,15 +2349,18 @@ namespace static constexpr auto name = "parseDateTime64InJodaSyntaxOrNull"; }; - using FunctionParseDateTime = FunctionParseDateTimeImpl; - using FunctionParseDateTimeOrZero = FunctionParseDateTimeImpl; - using FunctionParseDateTimeOrNull = FunctionParseDateTimeImpl; - using FunctionParseDateTimeInJodaSyntax = FunctionParseDateTimeImpl; - using FunctionParseDateTimeInJodaSyntaxOrZero = FunctionParseDateTimeImpl; - using FunctionParseDateTimeInJodaSyntaxOrNull = FunctionParseDateTimeImpl; - using FunctionParseDateTime64InJodaSyntax = FunctionParseDateTimeImpl; - using FunctionParseDateTime64InJodaSyntaxOrZero = FunctionParseDateTimeImpl; - using FunctionParseDateTime64InJodaSyntaxOrNull = FunctionParseDateTimeImpl; + using FunctionParseDateTime = FunctionParseDateTimeImpl; + using FunctionParseDateTimeOrZero = FunctionParseDateTimeImpl; + using FunctionParseDateTimeOrNull = FunctionParseDateTimeImpl; + using FunctionParseDateTime64 = FunctionParseDateTimeImpl; + using FunctionParseDateTime64OrZero = FunctionParseDateTimeImpl; + using FunctionParseDateTime64OrNull = FunctionParseDateTimeImpl; + using FunctionParseDateTimeInJodaSyntax = FunctionParseDateTimeImpl; + using FunctionParseDateTimeInJodaSyntaxOrZero = FunctionParseDateTimeImpl; + using FunctionParseDateTimeInJodaSyntaxOrNull = FunctionParseDateTimeImpl; + using FunctionParseDateTime64InJodaSyntax = FunctionParseDateTimeImpl; + using FunctionParseDateTime64InJodaSyntaxOrZero = FunctionParseDateTimeImpl; + using FunctionParseDateTime64InJodaSyntaxOrNull = FunctionParseDateTimeImpl; } REGISTER_FUNCTION(ParseDateTime) @@ -2262,6 +2370,9 @@ REGISTER_FUNCTION(ParseDateTime) factory.registerFunction(); factory.registerFunction(); factory.registerAlias("str_to_date", FunctionParseDateTimeOrNull::name, FunctionFactory::Case::Insensitive); + factory.registerFunction(); + factory.registerFunction(); + factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); From 338af374d88c134b39d75dd1f56f5630cd41fcc2 Mon Sep 17 00:00:00 2001 From: Sema Checherinda Date: Wed, 6 Nov 2024 09:52:25 +0100 Subject: [PATCH 064/244] remove the method remove in favor of the method removeIfExists --- .../AzureBlobStorage/AzureObjectStorage.cpp | 20 ++++++++--------- .../AzureBlobStorage/AzureObjectStorage.h | 4 ++-- .../Cached/CachedObjectStorage.cpp | 22 +++++++++---------- .../Cached/CachedObjectStorage.h | 4 ++-- .../DiskObjectStorageTransaction.cpp | 6 ++--- .../ObjectStorages/HDFS/HDFSObjectStorage.h | 4 ++-- src/Disks/ObjectStorages/IObjectStorage.h | 4 ++-- .../ObjectStorages/Local/LocalObjectStorage.h | 4 ++-- .../MetadataStorageFromPlainObjectStorage.cpp | 4 ++-- ...torageFromPlainObjectStorageOperations.cpp | 4 ++-- .../ObjectStorages/S3/S3ObjectStorage.cpp | 16 +++++++------- src/Disks/ObjectStorages/S3/S3ObjectStorage.h | 4 ++-- .../ObjectStorages/Web/WebObjectStorage.cpp | 16 +++++++------- .../ObjectStorages/Web/WebObjectStorage.h | 4 ++-- .../ObjectStorageQueueSource.cpp | 2 +- 15 files changed, 58 insertions(+), 60 deletions(-) diff --git a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp index 673c82806bd..959afa65672 100644 --- a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp +++ b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp @@ -278,17 +278,17 @@ void AzureObjectStorage::removeObjectImpl(const StoredObject & object, const Sha } /// Remove file. Throws exception if file doesn't exists or it's a directory. -void AzureObjectStorage::removeObject(const StoredObject & object) -{ - removeObjectImpl(object, client.get(), false); -} +// void AzureObjectStorage::removeObject(const StoredObject & object) +// { +// removeObjectImpl(object, client.get(), false); +// } -void AzureObjectStorage::removeObjects(const StoredObjects & objects) -{ - auto client_ptr = client.get(); - for (const auto & object : objects) - removeObjectImpl(object, client_ptr, false); -} +// void AzureObjectStorage::removeObjects(const StoredObjects & objects) +// { +// auto client_ptr = client.get(); +// for (const auto & object : objects) +// removeObjectImpl(object, client_ptr, false); +// } void AzureObjectStorage::removeObjectIfExists(const StoredObject & object) { diff --git a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h index 58225eccd90..433fe7a852e 100644 --- a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h +++ b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h @@ -60,9 +60,9 @@ public: const WriteSettings & write_settings = {}) override; /// Remove file. Throws exception if file doesn't exists or it's a directory. - void removeObject(const StoredObject & object) override; + //void removeObject(const StoredObject & object) override; - void removeObjects(const StoredObjects & objects) override; + //void removeObjects(const StoredObjects & objects) override; void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp index 163ff3a9c68..f2750e6814f 100644 --- a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp +++ b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp @@ -148,19 +148,19 @@ void CachedObjectStorage::removeCacheIfExists(const std::string & path_key_for_c cache->removeKeyIfExists(getCacheKey(path_key_for_cache), FileCache::getCommonUser().user_id); } -void CachedObjectStorage::removeObject(const StoredObject & object) -{ - removeCacheIfExists(object.remote_path); - object_storage->removeObject(object); -} +// void CachedObjectStorage::removeObject(const StoredObject & object) +// { +// removeCacheIfExists(object.remote_path); +// object_storage->removeObject(object); +// } -void CachedObjectStorage::removeObjects(const StoredObjects & objects) -{ - for (const auto & object : objects) - removeCacheIfExists(object.remote_path); +// void CachedObjectStorage::removeObjects(const StoredObjects & objects) +// { +// for (const auto & object : objects) +// removeCacheIfExists(object.remote_path); - object_storage->removeObjects(objects); -} +// object_storage->removeObjects(objects); +// } void CachedObjectStorage::removeObjectIfExists(const StoredObject & object) { diff --git a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h index b77baf21e40..7e10057e04c 100644 --- a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h +++ b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h @@ -45,9 +45,9 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - void removeObject(const StoredObject & object) override; + // void removeObject(const StoredObject & object) override; - void removeObjects(const StoredObjects & objects) override; + // void removeObjects(const StoredObjects & objects) override; void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Disks/ObjectStorages/DiskObjectStorageTransaction.cpp b/src/Disks/ObjectStorages/DiskObjectStorageTransaction.cpp index 64323fb6f3c..19de2bb78af 100644 --- a/src/Disks/ObjectStorages/DiskObjectStorageTransaction.cpp +++ b/src/Disks/ObjectStorages/DiskObjectStorageTransaction.cpp @@ -480,8 +480,7 @@ struct WriteFileObjectStorageOperation final : public IDiskObjectStorageOperatio void undo() override { - if (object_storage.exists(object)) - object_storage.removeObject(object); + object_storage.removeObjectIfExists(object); } void finalize() override @@ -543,8 +542,7 @@ struct CopyFileObjectStorageOperation final : public IDiskObjectStorageOperation void undo() override { - for (const auto & object : created_objects) - destination_object_storage.removeObject(object); + destination_object_storage.removeObjectsIfExist(created_objects); } void finalize() override diff --git a/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h b/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h index b53161beb76..317399b4753 100644 --- a/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h +++ b/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h @@ -78,9 +78,9 @@ public: const WriteSettings & write_settings = {}) override; /// Remove file. Throws exception if file doesn't exists or it's a directory. - void removeObject(const StoredObject & object) override; + void removeObject(const StoredObject & object); - void removeObjects(const StoredObjects & objects) override; + void removeObjects(const StoredObjects & objects); void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Disks/ObjectStorages/IObjectStorage.h b/src/Disks/ObjectStorages/IObjectStorage.h index 8dde96b8b16..adb36762539 100644 --- a/src/Disks/ObjectStorages/IObjectStorage.h +++ b/src/Disks/ObjectStorages/IObjectStorage.h @@ -161,11 +161,11 @@ public: virtual bool isRemote() const = 0; /// Remove object. Throws exception if object doesn't exists. - virtual void removeObject(const StoredObject & object) = 0; + // virtual void removeObject(const StoredObject & object) = 0; /// Remove multiple objects. Some object storages can do batch remove in a more /// optimal way. - virtual void removeObjects(const StoredObjects & objects) = 0; + // virtual void removeObjects(const StoredObjects & objects) = 0; /// Remove object on path if exists virtual void removeObjectIfExists(const StoredObject & object) = 0; diff --git a/src/Disks/ObjectStorages/Local/LocalObjectStorage.h b/src/Disks/ObjectStorages/Local/LocalObjectStorage.h index f1a0391a984..ffc151bda04 100644 --- a/src/Disks/ObjectStorages/Local/LocalObjectStorage.h +++ b/src/Disks/ObjectStorages/Local/LocalObjectStorage.h @@ -42,9 +42,9 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - void removeObject(const StoredObject & object) override; + void removeObject(const StoredObject & object); - void removeObjects(const StoredObjects & objects) override; + void removeObjects(const StoredObjects & objects); void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorage.cpp b/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorage.cpp index d56c5d9143c..27aa9304de7 100644 --- a/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorage.cpp +++ b/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorage.cpp @@ -203,7 +203,7 @@ void MetadataStorageFromPlainObjectStorageTransaction::unlinkFile(const std::str { auto object_key = metadata_storage.object_storage->generateObjectKeyForPath(path, std::nullopt /* key_prefix */); auto object = StoredObject(object_key.serialize()); - metadata_storage.object_storage->removeObject(object); + metadata_storage.object_storage->removeObjectIfExists(object); } void MetadataStorageFromPlainObjectStorageTransaction::removeDirectory(const std::string & path) @@ -211,7 +211,7 @@ void MetadataStorageFromPlainObjectStorageTransaction::removeDirectory(const std if (metadata_storage.object_storage->isWriteOnce()) { for (auto it = metadata_storage.iterateDirectory(path); it->isValid(); it->next()) - metadata_storage.object_storage->removeObject(StoredObject(it->path())); + metadata_storage.object_storage->removeObjectIfExists(StoredObject(it->path())); } else { diff --git a/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorageOperations.cpp b/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorageOperations.cpp index ea57d691908..62015631aa5 100644 --- a/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorageOperations.cpp +++ b/src/Disks/ObjectStorages/MetadataStorageFromPlainObjectStorageOperations.cpp @@ -107,7 +107,7 @@ void MetadataStorageFromPlainObjectStorageCreateDirectoryOperation::undo(std::un auto metric = object_storage->getMetadataStorageMetrics().directory_map_size; CurrentMetrics::sub(metric, 1); - object_storage->removeObject(StoredObject(metadata_object_key.serialize(), path / PREFIX_PATH_FILE_NAME)); + object_storage->removeObjectIfExists(StoredObject(metadata_object_key.serialize(), path / PREFIX_PATH_FILE_NAME)); } else if (write_created) object_storage->removeObjectIfExists(StoredObject(metadata_object_key.serialize(), path / PREFIX_PATH_FILE_NAME)); @@ -247,7 +247,7 @@ void MetadataStorageFromPlainObjectStorageRemoveDirectoryOperation::execute(std: auto metadata_object_key = createMetadataObjectKey(key_prefix, metadata_key_prefix); auto metadata_object = StoredObject(/*remote_path*/ metadata_object_key.serialize(), /*local_path*/ path / PREFIX_PATH_FILE_NAME); - object_storage->removeObject(metadata_object); + object_storage->removeObjectIfExists(metadata_object); { std::lock_guard lock(path_map.mutex); diff --git a/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp b/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp index 47ef97401f2..7ed118c6b07 100644 --- a/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp +++ b/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp @@ -326,20 +326,20 @@ void S3ObjectStorage::removeObjectsImpl(const StoredObjects & objects, bool if_e ProfileEvents::DiskS3DeleteObjects); } -void S3ObjectStorage::removeObject(const StoredObject & object) -{ - removeObjectImpl(object, false); -} +// void S3ObjectStorage::removeObject(const StoredObject & object) +// { +// removeObjectImpl(object, false); +// } void S3ObjectStorage::removeObjectIfExists(const StoredObject & object) { removeObjectImpl(object, true); } -void S3ObjectStorage::removeObjects(const StoredObjects & objects) -{ - removeObjectsImpl(objects, false); -} +// void S3ObjectStorage::removeObjects(const StoredObjects & objects) +// { +// removeObjectsImpl(objects, false); +// } void S3ObjectStorage::removeObjectsIfExist(const StoredObjects & objects) { diff --git a/src/Disks/ObjectStorages/S3/S3ObjectStorage.h b/src/Disks/ObjectStorages/S3/S3ObjectStorage.h index d6e84cf57ef..a2aeaf8a43c 100644 --- a/src/Disks/ObjectStorages/S3/S3ObjectStorage.h +++ b/src/Disks/ObjectStorages/S3/S3ObjectStorage.h @@ -102,11 +102,11 @@ public: ObjectStorageIteratorPtr iterate(const std::string & path_prefix, size_t max_keys) const override; /// Uses `DeleteObjectRequest`. - void removeObject(const StoredObject & object) override; + //void removeObject(const StoredObject & object) override; /// Uses `DeleteObjectsRequest` if it is allowed by `s3_capabilities`, otherwise `DeleteObjectRequest`. /// `DeleteObjectsRequest` is not supported on GCS, see https://issuetracker.google.com/issues/162653700 . - void removeObjects(const StoredObjects & objects) override; + //void removeObjects(const StoredObjects & objects) override; /// Uses `DeleteObjectRequest`. void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp b/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp index 871d3b506f6..1503d5819eb 100644 --- a/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp +++ b/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp @@ -254,15 +254,15 @@ std::unique_ptr WebObjectStorage::writeObject( /// NOLI throwNotAllowed(); } -void WebObjectStorage::removeObject(const StoredObject &) -{ - throwNotAllowed(); -} +// void WebObjectStorage::removeObject(const StoredObject &) +// { +// throwNotAllowed(); +// } -void WebObjectStorage::removeObjects(const StoredObjects &) -{ - throwNotAllowed(); -} +// void WebObjectStorage::removeObjects(const StoredObjects &) +// { +// throwNotAllowed(); +// } void WebObjectStorage::removeObjectIfExists(const StoredObject &) { diff --git a/src/Disks/ObjectStorages/Web/WebObjectStorage.h b/src/Disks/ObjectStorages/Web/WebObjectStorage.h index 573221b7e21..ae52cc20f9b 100644 --- a/src/Disks/ObjectStorages/Web/WebObjectStorage.h +++ b/src/Disks/ObjectStorages/Web/WebObjectStorage.h @@ -47,9 +47,9 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - void removeObject(const StoredObject & object) override; + // void removeObject(const StoredObject & object) override; - void removeObjects(const StoredObjects & objects) override; + // void removeObjects(const StoredObjects & objects) override; void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Storages/ObjectStorageQueue/ObjectStorageQueueSource.cpp b/src/Storages/ObjectStorageQueue/ObjectStorageQueueSource.cpp index ba1a97bc2fb..e702f07208a 100644 --- a/src/Storages/ObjectStorageQueue/ObjectStorageQueueSource.cpp +++ b/src/Storages/ObjectStorageQueue/ObjectStorageQueueSource.cpp @@ -659,7 +659,7 @@ void ObjectStorageQueueSource::applyActionAfterProcessing(const String & path) { if (files_metadata->getTableMetadata().after_processing == ObjectStorageQueueAction::DELETE) { - object_storage->removeObject(StoredObject(path)); + object_storage->removeObjectIfExists(StoredObject(path)); } } From 020b69647a65dd740cddfbf62730f37de14a4eb8 Mon Sep 17 00:00:00 2001 From: avogar Date: Wed, 6 Nov 2024 15:15:29 +0000 Subject: [PATCH 065/244] Fix counting column size in wide part for Dynamid and JSON types --- .../MergeTree/MergeTreeDataPartWide.cpp | 2 +- .../MergeTree/MergeTreeReaderWide.cpp | 2 +- ...umn_sizes_with_dynamic_structure.reference | 1 + ...62_column_sizes_with_dynamic_structure.sql | 22 +++++++++++++++++++ 4 files changed, 25 insertions(+), 2 deletions(-) create mode 100644 tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.reference create mode 100644 tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.sql diff --git a/src/Storages/MergeTree/MergeTreeDataPartWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWide.cpp index d6f213463f2..d8470ba8405 100644 --- a/src/Storages/MergeTree/MergeTreeDataPartWide.cpp +++ b/src/Storages/MergeTree/MergeTreeDataPartWide.cpp @@ -108,7 +108,7 @@ ColumnSize MergeTreeDataPartWide::getColumnSizeImpl( auto mrk_checksum = checksums.files.find(*stream_name + getMarksFileExtension()); if (mrk_checksum != checksums.files.end()) size.marks += mrk_checksum->second.file_size; - }); + }, column.type, getColumnSample(column)); return size; } diff --git a/src/Storages/MergeTree/MergeTreeReaderWide.cpp b/src/Storages/MergeTree/MergeTreeReaderWide.cpp index 77231d8d392..885bd1ded8c 100644 --- a/src/Storages/MergeTree/MergeTreeReaderWide.cpp +++ b/src/Storages/MergeTree/MergeTreeReaderWide.cpp @@ -172,7 +172,7 @@ size_t MergeTreeReaderWide::readRows( throw; } - if (column->empty()) + if (column->empty() && max_rows_to_read > 0) res_columns[pos] = nullptr; } diff --git a/tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.reference b/tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.reference new file mode 100644 index 00000000000..5cab16ed96d --- /dev/null +++ b/tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.reference @@ -0,0 +1 @@ +test 10.00 million 352.87 MiB 39.43 MiB 39.45 MiB diff --git a/tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.sql b/tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.sql new file mode 100644 index 00000000000..21e6515fc99 --- /dev/null +++ b/tests/queries/0_stateless/03262_column_sizes_with_dynamic_structure.sql @@ -0,0 +1,22 @@ +-- Tags: no-random-settings + +set allow_experimental_dynamic_type = 1; +set allow_experimental_json_type = 1; + +drop table if exists test; +create table test (d Dynamic, json JSON) engine=MergeTree order by tuple() settings min_rows_for_wide_part=0, min_bytes_for_wide_part=1; +insert into test select number, '{"a" : 42, "b" : "Hello, World"}' from numbers(10000000); + +SELECT + `table`, + formatReadableQuantity(sum(rows)) AS rows, + formatReadableSize(sum(data_uncompressed_bytes)) AS data_size_uncompressed, + formatReadableSize(sum(data_compressed_bytes)) AS data_size_compressed, + formatReadableSize(sum(bytes_on_disk)) AS total_size_on_disk +FROM system.parts +WHERE active AND (database = currentDatabase()) AND (`table` = 'test') +GROUP BY `table` +ORDER BY `table` ASC; + +drop table test; + From c1345d38c8e987838704a4ae7da6cb05af8257c2 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Wed, 6 Nov 2024 15:44:22 +0000 Subject: [PATCH 066/244] Fix flakiness in 03254_pr_join_on_dups --- src/Interpreters/IJoin.h | 1 - src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp | 1 - tests/queries/0_stateless/03254_pr_join_on_dups.sql | 1 + 3 files changed, 1 insertion(+), 2 deletions(-) diff --git a/src/Interpreters/IJoin.h b/src/Interpreters/IJoin.h index 8f648de2538..5a83137ca2a 100644 --- a/src/Interpreters/IJoin.h +++ b/src/Interpreters/IJoin.h @@ -1,7 +1,6 @@ #pragma once #include -#include #include #include diff --git a/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp b/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp index c0b31864eac..35d340b4bbf 100644 --- a/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp +++ b/src/Processors/QueryPlan/Optimizations/optimizeJoin.cpp @@ -16,7 +16,6 @@ #include #include -#include namespace DB::QueryPlanOptimizations { diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.sql b/tests/queries/0_stateless/03254_pr_join_on_dups.sql index 222f7693090..166910d496f 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.sql +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.sql @@ -10,6 +10,7 @@ insert into Y (id, y_a) values (1, 'r1'), (1, 'r2'), (2, 'r3'), (3, 'r4'), insert into Y (id, y_a, y_b) values (4, 'r6', 'nr6'), (6, 'r7', 'nr7'), (7, 'r8', 'nr8'), (9, 'r9', 'nr9'); set enable_analyzer = 1, enable_parallel_replicas = 1, max_parallel_replicas = 3, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost'; +set min_bytes_to_use_direct_io = 0; -- min_bytes_to_use_direct_io > 0 is broken and leads to unexpected results, https://github.com/ClickHouse/ClickHouse/issues/65690 select 'inner'; select X.*, Y.* from X inner join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; From df632b6f1e4d825644138c77c2ae4a25943a7fe8 Mon Sep 17 00:00:00 2001 From: Sema Checherinda Date: Wed, 6 Nov 2024 16:44:52 +0100 Subject: [PATCH 067/244] clean up --- .../AzureBlobStorage/AzureObjectStorage.cpp | 13 ------------- .../AzureBlobStorage/AzureObjectStorage.h | 5 ----- .../ObjectStorages/Cached/CachedObjectStorage.cpp | 14 -------------- .../ObjectStorages/Cached/CachedObjectStorage.h | 4 ---- src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp | 10 ---------- src/Disks/ObjectStorages/S3/S3ObjectStorage.h | 7 ------- src/Disks/ObjectStorages/Web/WebObjectStorage.cpp | 10 ---------- src/Disks/ObjectStorages/Web/WebObjectStorage.h | 4 ---- 8 files changed, 67 deletions(-) diff --git a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp index 959afa65672..b8386bcf967 100644 --- a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp +++ b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp @@ -277,19 +277,6 @@ void AzureObjectStorage::removeObjectImpl(const StoredObject & object, const Sha } } -/// Remove file. Throws exception if file doesn't exists or it's a directory. -// void AzureObjectStorage::removeObject(const StoredObject & object) -// { -// removeObjectImpl(object, client.get(), false); -// } - -// void AzureObjectStorage::removeObjects(const StoredObjects & objects) -// { -// auto client_ptr = client.get(); -// for (const auto & object : objects) -// removeObjectImpl(object, client_ptr, false); -// } - void AzureObjectStorage::removeObjectIfExists(const StoredObject & object) { removeObjectImpl(object, client.get(), true); diff --git a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h index 433fe7a852e..401493be367 100644 --- a/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h +++ b/src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.h @@ -59,11 +59,6 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - /// Remove file. Throws exception if file doesn't exists or it's a directory. - //void removeObject(const StoredObject & object) override; - - //void removeObjects(const StoredObjects & objects) override; - void removeObjectIfExists(const StoredObject & object) override; void removeObjectsIfExist(const StoredObjects & objects) override; diff --git a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp index f2750e6814f..779b8830fab 100644 --- a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp +++ b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.cpp @@ -148,20 +148,6 @@ void CachedObjectStorage::removeCacheIfExists(const std::string & path_key_for_c cache->removeKeyIfExists(getCacheKey(path_key_for_cache), FileCache::getCommonUser().user_id); } -// void CachedObjectStorage::removeObject(const StoredObject & object) -// { -// removeCacheIfExists(object.remote_path); -// object_storage->removeObject(object); -// } - -// void CachedObjectStorage::removeObjects(const StoredObjects & objects) -// { -// for (const auto & object : objects) -// removeCacheIfExists(object.remote_path); - -// object_storage->removeObjects(objects); -// } - void CachedObjectStorage::removeObjectIfExists(const StoredObject & object) { removeCacheIfExists(object.remote_path); diff --git a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h index 7e10057e04c..77aa635b89b 100644 --- a/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h +++ b/src/Disks/ObjectStorages/Cached/CachedObjectStorage.h @@ -45,10 +45,6 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - // void removeObject(const StoredObject & object) override; - - // void removeObjects(const StoredObjects & objects) override; - void removeObjectIfExists(const StoredObject & object) override; void removeObjectsIfExist(const StoredObjects & objects) override; diff --git a/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp b/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp index 7ed118c6b07..9fca3cad688 100644 --- a/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp +++ b/src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp @@ -326,21 +326,11 @@ void S3ObjectStorage::removeObjectsImpl(const StoredObjects & objects, bool if_e ProfileEvents::DiskS3DeleteObjects); } -// void S3ObjectStorage::removeObject(const StoredObject & object) -// { -// removeObjectImpl(object, false); -// } - void S3ObjectStorage::removeObjectIfExists(const StoredObject & object) { removeObjectImpl(object, true); } -// void S3ObjectStorage::removeObjects(const StoredObjects & objects) -// { -// removeObjectsImpl(objects, false); -// } - void S3ObjectStorage::removeObjectsIfExist(const StoredObjects & objects) { removeObjectsImpl(objects, true); diff --git a/src/Disks/ObjectStorages/S3/S3ObjectStorage.h b/src/Disks/ObjectStorages/S3/S3ObjectStorage.h index a2aeaf8a43c..4b9c968ede9 100644 --- a/src/Disks/ObjectStorages/S3/S3ObjectStorage.h +++ b/src/Disks/ObjectStorages/S3/S3ObjectStorage.h @@ -101,13 +101,6 @@ public: ObjectStorageIteratorPtr iterate(const std::string & path_prefix, size_t max_keys) const override; - /// Uses `DeleteObjectRequest`. - //void removeObject(const StoredObject & object) override; - - /// Uses `DeleteObjectsRequest` if it is allowed by `s3_capabilities`, otherwise `DeleteObjectRequest`. - /// `DeleteObjectsRequest` is not supported on GCS, see https://issuetracker.google.com/issues/162653700 . - //void removeObjects(const StoredObjects & objects) override; - /// Uses `DeleteObjectRequest`. void removeObjectIfExists(const StoredObject & object) override; diff --git a/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp b/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp index 1503d5819eb..35abc0ed0df 100644 --- a/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp +++ b/src/Disks/ObjectStorages/Web/WebObjectStorage.cpp @@ -254,16 +254,6 @@ std::unique_ptr WebObjectStorage::writeObject( /// NOLI throwNotAllowed(); } -// void WebObjectStorage::removeObject(const StoredObject &) -// { -// throwNotAllowed(); -// } - -// void WebObjectStorage::removeObjects(const StoredObjects &) -// { -// throwNotAllowed(); -// } - void WebObjectStorage::removeObjectIfExists(const StoredObject &) { throwNotAllowed(); diff --git a/src/Disks/ObjectStorages/Web/WebObjectStorage.h b/src/Disks/ObjectStorages/Web/WebObjectStorage.h index ae52cc20f9b..1e612bd359c 100644 --- a/src/Disks/ObjectStorages/Web/WebObjectStorage.h +++ b/src/Disks/ObjectStorages/Web/WebObjectStorage.h @@ -47,10 +47,6 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - // void removeObject(const StoredObject & object) override; - - // void removeObjects(const StoredObjects & objects) override; - void removeObjectIfExists(const StoredObject & object) override; void removeObjectsIfExist(const StoredObjects & objects) override; From 4ad8273e5f3d16f5a95220824223800b4a356e26 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Wed, 6 Nov 2024 17:31:24 +0000 Subject: [PATCH 068/244] Enable merge filters optimization. --- src/Core/Settings.cpp | 2 +- src/Core/SettingsChangesHistory.cpp | 1 + .../QueryPlanOptimizationSettings.h | 2 +- .../03262_filter_push_down_view.reference | 2 ++ .../03262_filter_push_down_view.sql | 36 +++++++++++++++++++ 5 files changed, 41 insertions(+), 2 deletions(-) create mode 100644 tests/queries/0_stateless/03262_filter_push_down_view.reference create mode 100644 tests/queries/0_stateless/03262_filter_push_down_view.sql diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index 081e07ca2ce..6f8047bbdf8 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -4554,7 +4554,7 @@ Possible values: - 0 - Disable - 1 - Enable )", 0) \ - DECLARE(Bool, query_plan_merge_filters, false, R"( + DECLARE(Bool, query_plan_merge_filters, true, R"( Allow to merge filters in the query plan )", 0) \ DECLARE(Bool, query_plan_filter_push_down, true, R"( diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index ed87fde8b7e..12350b6cdaf 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -74,6 +74,7 @@ static std::initializer_list Date: Tue, 5 Nov 2024 14:02:43 +0000 Subject: [PATCH 069/244] Reduce the general critical section for query_metric_log - Use a separate mutex for each query to reduce the contention period for queries_mutex. - Refactor to use std::mutex instead of std::recursive_mutex for queries_mutex. - In case we're running late to schedule the next task, schedule it immediately. - Fix LockGuard because unlocking twice is undefined behavior. --- base/base/defines.h | 1 + src/Common/LockGuard.h | 32 +++++- src/Interpreters/QueryMetricLog.cpp | 165 +++++++++++++++++++--------- src/Interpreters/QueryMetricLog.h | 43 ++++++-- 4 files changed, 179 insertions(+), 62 deletions(-) diff --git a/base/base/defines.h b/base/base/defines.h index 5685a6d9833..a0c3c0d1de5 100644 --- a/base/base/defines.h +++ b/base/base/defines.h @@ -145,6 +145,7 @@ #define TSA_TRY_ACQUIRE_SHARED(...) __attribute__((try_acquire_shared_capability(__VA_ARGS__))) /// function tries to acquire a shared capability and returns a boolean value indicating success or failure #define TSA_RELEASE_SHARED(...) __attribute__((release_shared_capability(__VA_ARGS__))) /// function releases the given shared capability #define TSA_SCOPED_LOCKABLE __attribute__((scoped_lockable)) /// object of a class has scoped lockable capability +#define TSA_RETURN_CAPABILITY(...) __attribute__((lock_returned(__VA_ARGS__))) /// to return capabilities in functions /// Macros for suppressing TSA warnings for specific reads/writes (instead of suppressing it for the whole function) /// They use a lambda function to apply function attribute to a single statement. This enable us to suppress warnings locally instead of diff --git a/src/Common/LockGuard.h b/src/Common/LockGuard.h index 8a98c5f553a..03c8a3e7617 100644 --- a/src/Common/LockGuard.h +++ b/src/Common/LockGuard.h @@ -1,23 +1,47 @@ #pragma once -#include #include +#include +#include namespace DB { +namespace ErrorCodes +{ + extern const int LOGICAL_ERROR; +}; + /** LockGuard provides RAII-style locking mechanism for a mutex. - ** It's intended to be used like std::unique_ptr but with TSA annotations + ** It's intended to be used like std::unique_lock but with TSA annotations */ template class TSA_SCOPED_LOCKABLE LockGuard { public: - explicit LockGuard(Mutex & mutex_) TSA_ACQUIRE(mutex_) : mutex(mutex_) { mutex.lock(); } - ~LockGuard() TSA_RELEASE() { mutex.unlock(); } + explicit LockGuard(Mutex & mutex_) TSA_ACQUIRE(mutex_) : mutex(mutex_) { lock(); } + ~LockGuard() TSA_RELEASE() { if (locked) unlock(); } + + void lock() TSA_ACQUIRE() + { + /// Don't allow recursive_mutex for now. + if (locked) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Can't lock twice the same mutex"); + mutex.lock(); + locked = true; + } + + void unlock() TSA_RELEASE() + { + if (!locked) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Can't unlock the mutex without locking it first"); + mutex.unlock(); + locked = false; + } private: Mutex & mutex; + bool locked = false; }; template typename TLockGuard, typename Mutex> diff --git a/src/Interpreters/QueryMetricLog.cpp b/src/Interpreters/QueryMetricLog.cpp index 5ab3fe590e0..e784c357b29 100644 --- a/src/Interpreters/QueryMetricLog.cpp +++ b/src/Interpreters/QueryMetricLog.cpp @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -16,7 +17,6 @@ #include #include -#include namespace DB @@ -24,6 +24,20 @@ namespace DB static auto logger = getLogger("QueryMetricLog"); +namespace ErrorCodes +{ + extern const int LOGICAL_ERROR; +}; + +String timePointToString(QueryMetricLog::TimePoint time) +{ + /// fmtlib supports subsecond formatting in 10.0.0. We're in 9.1.0, so we need to add the milliseconds ourselves. + auto seconds = std::chrono::time_point_cast(time); + auto microseconds = std::chrono::duration_cast(time - seconds).count(); + + return fmt::format("{:%Y.%m.%d %H:%M:%S}.{:06}", seconds, microseconds); +} + ColumnsDescription QueryMetricLogElement::getColumnsDescription() { ColumnsDescription result; @@ -87,36 +101,69 @@ void QueryMetricLog::shutdown() Base::shutdown(); } -void QueryMetricLog::startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds) +void QueryMetricLog::collectMetric(const ProcessList & process_list, String query_id) { - QueryMetricLogStatus status; - status.interval_milliseconds = interval_milliseconds; - status.next_collect_time = start_time + std::chrono::milliseconds(interval_milliseconds); + auto current_time = std::chrono::system_clock::now(); + const auto query_info = process_list.getQueryInfo(query_id, false, true, false); + if (!query_info) + { + LOG_TRACE(logger, "Query {} is not running anymore, so we couldn't get its QueryStatusInfo", query_id); + return; + } + + LockGuard global_lock(queries_mutex); + auto it = queries.find(query_id); + + /// The query might have finished while the scheduled task is running. + if (it == queries.end()) + { + global_lock.unlock(); + LOG_TRACE(logger, "Query {} not found in the list. Finished while this collecting task was running", query_id); + return; + } + + auto & query_status = it->second; + if (!query_status.mutex) + { + global_lock.unlock(); + LOG_TRACE(logger, "Query {} finished while this collecting task was running", query_id); + return; + } + + LockGuard query_lock(query_status.getMutex()); + global_lock.unlock(); + + auto elem = query_status.createLogMetricElement(query_id, *query_info, current_time); + if (elem) + add(std::move(elem.value())); +} + +/// We use TSA_NO_THREAD_SAFETY_ANALYSIS to prevent TSA complaining that we're modifying the query_status fields +/// without locking the mutex. Since we're building it from scratch, there's no harm in not holding it. +/// If we locked it to make TSA happy, TSAN build would falsely complain about +/// lock-order-inversion (potential deadlock) +/// which is not a real issue since QueryMetricLogStatus's mutex cannot be locked by anything else +/// until we add it to the queries map. +void QueryMetricLog::startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds) TSA_NO_THREAD_SAFETY_ANALYSIS +{ + QueryMetricLogStatus query_status; + query_status.interval_milliseconds = interval_milliseconds; + query_status.next_collect_time = start_time + std::chrono::milliseconds(interval_milliseconds); auto context = getContext(); const auto & process_list = context->getProcessList(); - status.task = context->getSchedulePool().createTask("QueryMetricLog", [this, &process_list, query_id] { - auto current_time = std::chrono::system_clock::now(); - const auto query_info = process_list.getQueryInfo(query_id, false, true, false); - if (!query_info) - { - LOG_TRACE(logger, "Query {} is not running anymore, so we couldn't get its QueryStatusInfo", query_id); - return; - } - - auto elem = createLogMetricElement(query_id, *query_info, current_time); - if (elem) - add(std::move(elem.value())); + query_status.task = context->getSchedulePool().createTask("QueryMetricLog", [this, &process_list, query_id] { + collectMetric(process_list, query_id); }); - std::lock_guard lock(queries_mutex); - status.task->scheduleAfter(interval_milliseconds); - queries.emplace(query_id, std::move(status)); + LockGuard global_lock(queries_mutex); + query_status.scheduleNext(query_id); + queries.emplace(query_id, std::move(query_status)); } void QueryMetricLog::finishQuery(const String & query_id, TimePoint finish_time, QueryStatusInfoPtr query_info) { - std::unique_lock lock(queries_mutex); + LockGuard global_lock(queries_mutex); auto it = queries.find(query_id); /// finishQuery may be called from logExceptionBeforeStart when the query has not even started @@ -124,9 +171,19 @@ void QueryMetricLog::finishQuery(const String & query_id, TimePoint finish_time, if (it == queries.end()) return; + auto & query_status = it->second; + decltype(query_status.mutex) query_mutex; + LockGuard query_lock(query_status.getMutex()); + + /// Move the query mutex here so that we hold it until the end, after removing the query from queries. + query_mutex = std::move(query_status.mutex); + query_status.mutex = {}; + + global_lock.unlock(); + if (query_info) { - auto elem = createLogMetricElement(query_id, *query_info, finish_time, false); + auto elem = query_status.createLogMetricElement(query_id, *query_info, finish_time, false); if (elem) add(std::move(elem.value())); } @@ -139,51 +196,62 @@ void QueryMetricLog::finishQuery(const String & query_id, TimePoint finish_time, /// that order. { /// Take ownership of the task so that we can destroy it in this scope after unlocking `queries_mutex`. - auto task = std::move(it->second.task); + auto task = std::move(query_status.task); /// Build an empty task for the old task to make sure it does not lock any mutex on its destruction. - it->second.task = {}; + query_status.task = {}; + query_lock.unlock(); + global_lock.lock(); queries.erase(query_id); /// Ensure `queries_mutex` is unlocked before calling task's destructor at the end of this /// scope which will lock `exec_mutex`. - lock.unlock(); + global_lock.unlock(); } } -std::optional QueryMetricLog::createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next) +void QueryMetricLogStatus::scheduleNext(String query_id) { - /// fmtlib supports subsecond formatting in 10.0.0. We're in 9.1.0, so we need to add the milliseconds ourselves. - auto seconds = std::chrono::time_point_cast(query_info_time); - auto microseconds = std::chrono::duration_cast(query_info_time - seconds).count(); - LOG_DEBUG(logger, "Collecting query_metric_log for query {} with QueryStatusInfo from {:%Y.%m.%d %H:%M:%S}.{:06}. Schedule next: {}", query_id, seconds, microseconds, schedule_next); - - std::unique_lock lock(queries_mutex); - auto query_status_it = queries.find(query_id); - - /// The query might have finished while the scheduled task is running. - if (query_status_it == queries.end()) + const auto now = std::chrono::system_clock::now(); + if (next_collect_time > now) { - lock.unlock(); - LOG_TRACE(logger, "Query {} finished already while this collecting task was running", query_id); - return {}; + const auto wait_time = std::chrono::duration_cast(next_collect_time - now).count(); + task->scheduleAfter(wait_time); } - - auto & query_status = query_status_it->second; - if (query_info_time <= query_status.last_collect_time) + else + { + LOG_TRACE(logger, "The next collecting task for query {} should have already run at {}. Scheduling it right now", + query_id, timePointToString(next_collect_time)); + task->schedule(); + } +} + +std::optional QueryMetricLogStatus::createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next) +{ + LOG_TRACE(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Schedule next: {}", + query_id, interval_milliseconds, timePointToString(query_info_time), schedule_next); + + if (query_info_time <= last_collect_time) { - lock.unlock(); LOG_TRACE(logger, "Query {} has a more recent metrics collected. Skipping this one", query_id); return {}; } - query_status.last_collect_time = query_info_time; + /// Leave some margin because task->scheduleAfter takes a value in milliseconds. + /// So, we can expect up to 1ms of drift since BackgroundSchedulePool will compare + /// time points in milliseconds. + static auto error_margin = std::chrono::milliseconds(1); + if (schedule_next && query_info_time + error_margin < next_collect_time) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Task to collect metric for query {} scheduled at {} but run at {}", + query_id, timePointToString(next_collect_time), timePointToString(query_info_time)); + + last_collect_time = query_info_time; QueryMetricLogElement elem; elem.event_time = timeInSeconds(query_info_time); elem.event_time_microseconds = timeInMicroseconds(query_info_time); - elem.query_id = query_status_it->first; + elem.query_id = query_id; elem.memory_usage = query_info.memory_usage > 0 ? query_info.memory_usage : 0; elem.peak_memory_usage = query_info.peak_memory_usage > 0 ? query_info.peak_memory_usage : 0; @@ -192,7 +260,7 @@ std::optional QueryMetricLog::createLogMetricElement(cons for (ProfileEvents::Event i = ProfileEvents::Event(0), end = ProfileEvents::end(); i < end; ++i) { const auto & new_value = (*(query_info.profile_counters))[i]; - auto & old_value = query_status.last_profile_events[i]; + auto & old_value = last_profile_events[i]; /// Profile event counters are supposed to be monotonic. However, at least the `NetworkReceiveBytes` can be inaccurate. /// So, since in the future the counter should always have a bigger value than in the past, we skip this event. @@ -214,9 +282,8 @@ std::optional QueryMetricLog::createLogMetricElement(cons if (schedule_next) { - query_status.next_collect_time += std::chrono::milliseconds(query_status.interval_milliseconds); - const auto wait_time = std::chrono::duration_cast(query_status.next_collect_time - std::chrono::system_clock::now()).count(); - query_status.task->scheduleAfter(wait_time); + next_collect_time += std::chrono::milliseconds(interval_milliseconds); + scheduleNext(query_id); } return elem; diff --git a/src/Interpreters/QueryMetricLog.h b/src/Interpreters/QueryMetricLog.h index 802cee7bf26..65764229b0a 100644 --- a/src/Interpreters/QueryMetricLog.h +++ b/src/Interpreters/QueryMetricLog.h @@ -1,5 +1,6 @@ #pragma once +#include #include #include #include @@ -11,11 +12,17 @@ #include #include +#include namespace DB { +namespace ErrorCodes +{ + extern const int LOGICAL_ERROR; +}; + /** QueryMetricLogElement is a log of query metric values measured at regular time interval. */ @@ -36,31 +43,49 @@ struct QueryMetricLogElement struct QueryMetricLogStatus { + using TimePoint = std::chrono::system_clock::time_point; + using Mutex = std::mutex; + UInt64 interval_milliseconds; - std::chrono::system_clock::time_point last_collect_time; - std::chrono::system_clock::time_point next_collect_time; - std::vector last_profile_events = std::vector(ProfileEvents::end()); - BackgroundSchedulePool::TaskHolder task; + std::chrono::system_clock::time_point last_collect_time TSA_GUARDED_BY(getMutex()); + std::chrono::system_clock::time_point next_collect_time TSA_GUARDED_BY(getMutex()); + std::vector last_profile_events TSA_GUARDED_BY(getMutex()) = std::vector(ProfileEvents::end()); + BackgroundSchedulePool::TaskHolder task TSA_GUARDED_BY(getMutex()); + + /// We need to be able to move it for the hash map, so we need to add an indirection here. + std::unique_ptr mutex = std::make_unique(); + + /// Return a reference to the mutex, used for Thread Sanitizer annotations. + Mutex & getMutex() const TSA_RETURN_CAPABILITY(mutex) + { + if (!mutex) + throw Exception(ErrorCodes::LOGICAL_ERROR, "Mutex cannot be NULL"); + return *mutex; + } + + void scheduleNext(String query_id) TSA_REQUIRES(getMutex()); + std::optional createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next = true) TSA_REQUIRES(getMutex()); }; class QueryMetricLog : public SystemLog { using SystemLog::SystemLog; - using TimePoint = std::chrono::system_clock::time_point; using Base = SystemLog; public: + using TimePoint = std::chrono::system_clock::time_point; + void shutdown() final; - // Both startQuery and finishQuery are called from the thread that executes the query + /// Both startQuery and finishQuery are called from the thread that executes the query. void startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds); void finishQuery(const String & query_id, TimePoint finish_time, QueryStatusInfoPtr query_info = nullptr); private: - std::optional createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next = true); + void collectMetric(const ProcessList & process_list, String query_id); - std::recursive_mutex queries_mutex; - std::unordered_map queries; + std::mutex queries_mutex; + std::unordered_map queries TSA_GUARDED_BY(queries_mutex); }; } From 1c74206bf2fddd2aad8f96699769e20dd122979a Mon Sep 17 00:00:00 2001 From: kevinyhzou Date: Thu, 7 Nov 2024 18:00:14 +0800 Subject: [PATCH 070/244] add parseDateTime64 functions --- src/Functions/parseDateTime.cpp | 145 ++++++++++++------ .../03252_parse_datetime64.reference | 17 ++ .../0_stateless/03252_parse_datetime64.sql | 32 ++++ ..._parse_datetime64_in_joda_syntax.reference | 32 ++-- .../03252_parse_datetime64_in_joda_syntax.sql | 60 ++++++-- 5 files changed, 214 insertions(+), 72 deletions(-) create mode 100644 tests/queries/0_stateless/03252_parse_datetime64.reference create mode 100644 tests/queries/0_stateless/03252_parse_datetime64.sql diff --git a/src/Functions/parseDateTime.cpp b/src/Functions/parseDateTime.cpp index 976be53a21e..9f7f78dcbe2 100644 --- a/src/Functions/parseDateTime.cpp +++ b/src/Functions/parseDateTime.cpp @@ -193,6 +193,7 @@ namespace Int32 minute = 0; /// range [0, 59] Int32 second = 0; /// range [0, 59] Int32 microsecond = 0; /// range [0, 999999] + UInt32 scale = 0; /// The microsecond scale of DateTime64. bool is_am = true; /// If is_hour_of_half_day = true and is_am = false (i.e. pm) then add 12 hours to the result DateTime bool hour_starts_at_1 = false; /// Whether the hour is clockhour @@ -221,6 +222,7 @@ namespace minute = 0; second = 0; microsecond = 0; + scale = 0; is_am = true; hour_starts_at_1 = false; @@ -599,7 +601,7 @@ namespace bool useDefaultImplementationForConstants() const override { return true; } bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo & /*arguments*/) const override { return false; } - ColumnNumbers getArgumentsThatAreAlwaysConstant() const override { return {1, 2}; } + ColumnNumbers getArgumentsThatAreAlwaysConstant() const override { return {1, 2, 3}; } bool isVariadic() const override { return true; } size_t getNumberOfArguments() const override { return 0; } @@ -613,9 +615,9 @@ namespace if constexpr (return_type == ReturnType::DateTime64) { optional_args = { - {"precision or format", static_cast([](const IDataType & data_type) -> bool { + {"scale/format", static_cast([](const IDataType & data_type) -> bool { return isUInt(data_type) || isString(data_type); - }), nullptr, "Number or String"}, + }), nullptr, "UInt or String"}, {"format", static_cast(&isString), nullptr, "String"}, {"timezone", static_cast(&isString), &isColumnConst, "const String"} }; @@ -631,24 +633,34 @@ namespace DataTypePtr data_type; if constexpr (return_type == ReturnType::DateTime64) { + UInt32 scale = 0; if (arguments.size() == 1) - return std::make_shared(0, time_zone_name); + { + /// In MySQL parse syntax, the scale of microseond is 6. + if constexpr (parse_syntax == ParseSyntax::MySQL) + scale = 6; + } else { - UInt32 precision = 0; if (isUInt(arguments[1].type)) { - const auto * col_precision = checkAndGetColumnConst(arguments[1].column.get()); - if (col_precision) - precision = col_precision->getValue(); + const auto * col_scale = checkAndGetColumnConst(arguments[1].column.get()); + if (col_scale) + scale = col_scale->getValue(); else throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The input precision value may exceed the max value of `DateTime64`: {}.", + "The input scale value may exceed the max value of `DateTime64`: {}.", maxPrecisionOfDateTime64); } - /// Construct the return type `DataTypDateTime64` with precision and time zone name. The precision value can be specified or be extracted - /// from the format string by computing how many 'S' characters are contained in the format's micorsceond fragment. - String format = getFormat(arguments, precision); + else + { + if constexpr (parse_syntax == ParseSyntax::MySQL) + scale = 6; + } + + /// Construct the return type `DataTypDateTime64` with scale and time zone name. The scale value can be specified or be extracted + /// from the format string by c how many 'S' characters are contained in the format's micorsceond fragment. + String format = getFormat(arguments, scale); std::vector instructions = parseFormat(format); for (const auto & instruction : instructions) { @@ -664,26 +676,27 @@ namespace else val++; } - /// If the precision is already specified by the second parameter, but it not equals the value that extract from the format string, - /// then we should throw an exception; If the precision is not specified, then we set its value as the extracted one. - if (val != 0 && precision != 0 && val != precision) + /// If the scale is already specified by the second argument, but it not equals the value that extract from the format string, + /// then we should throw an exception; If the scale is not specified, then we should set its value as the extracted one. + if (val != 0 && scale != 0 && val != scale) throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The precision of input format string {} not equals the given precision value {}.", + "The scale of input format string {} not equals the given scale value {}.", format, - precision); - else if (precision == 0 && val != 0) - precision = val; + scale); + else if (scale == 0 && val != 0) + scale = val; } - if (precision > maxPrecisionOfDateTime64) + if (scale > maxPrecisionOfDateTime64) throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The precision of the input format string {} exceed the max precision value {}.", + "The scale of the input format string {} exceed the max scale value {}.", format, maxPrecisionOfDateTime64); - data_type = std::make_shared(precision, time_zone_name); } + data_type = std::make_shared(scale, time_zone_name); } else data_type = std::make_shared(time_zone_name); + if (error_handling == ErrorHandling::Null) return std::make_shared(data_type); return data_type; @@ -729,15 +742,15 @@ namespace col_null_map = ColumnUInt8::create(input_rows_count, 0); Int64 multiplier = 0; - UInt32 precision = 0; + UInt32 scale = 0; if constexpr (return_type == ReturnType::DateTime64) { const DataTypeDateTime64 * datatime64_type = checkAndGetDataType(removeNullable(result_type).get()); - precision = datatime64_type->getScale(); - multiplier = DecimalUtils::scaleMultiplier(precision); + scale = datatime64_type->getScale(); + multiplier = DecimalUtils::scaleMultiplier(scale); } - String format = getFormat(arguments, precision); + const String format = getFormat(arguments, scale); const auto & time_zone = getTimeZone(arguments); std::vector instructions = parseFormat(format); @@ -746,6 +759,9 @@ namespace for (size_t i = 0; i < input_rows_count; ++i) { datetime.reset(); + if constexpr (return_type == ReturnType::DateTime64) + datetime.scale = scale; + StringRef str_ref = col_str->getDataAt(i); Pos cur = str_ref.data; Pos end = str_ref.data + str_ref.size; @@ -787,7 +803,7 @@ namespace Int64OrError result = 0; /// Ensure all input was consumed when the return type is `DateTime`. - if (return_type == ReturnType::DateTime && cur < end) + if (cur < end) { result = tl::unexpected(ErrorCodeAndMessage( ErrorCodes::CANNOT_PARSE_DATETIME, @@ -938,6 +954,28 @@ namespace return cur; } + template + [[nodiscard]] + static PosOrError readNumber6(Pos cur, Pos end, [[maybe_unused]] const String & fragment, T & res) + { + if constexpr (need_check_space == NeedCheckSpace::Yes) + RETURN_ERROR_IF_FAILED(checkSpace(cur, end, 6, "readNumber6 requires size >= 6", fragment)) + + res = (*cur - '0'); + ++cur; + res = res * 10 + (*cur - '0'); + ++cur; + res = res * 10 + (*cur - '0'); + ++cur; + res = res * 10 + (*cur - '0'); + ++cur; + res = res * 10 + (*cur - '0'); + ++cur; + res = res * 10 + (*cur - '0'); + ++cur; + return cur; + } + [[nodiscard]] static VoidOrError checkSpace(Pos cur, Pos end, size_t len, const String & msg, const String & fragment) { @@ -1358,13 +1396,18 @@ namespace } [[nodiscard]] - static PosOrError mysqlMicrosecond(Pos cur, Pos end, const String & fragment, DateTime & /*date*/) + static PosOrError mysqlMicrosecond(Pos cur, Pos end, const String & fragment, DateTime & date) { - RETURN_ERROR_IF_FAILED(checkSpace(cur, end, 6, "mysqlMicrosecond requires size >= 6", fragment)) - - for (size_t i = 0; i < 6; ++i) - ASSIGN_RESULT_OR_RETURN_ERROR(cur, (assertNumber(cur, end, fragment))) - + if (date.scale != 6) + RETURN_ERROR( + ErrorCodes::CANNOT_PARSE_DATETIME, + "Unable to parse fragment {} from {} because of the microsecond's scale {} is not 6", + fragment, + std::string_view(cur, end - cur), + std::to_string(date.scale)) + Int32 microsecond = 0; + ASSIGN_RESULT_OR_RETURN_ERROR(cur, (readNumber6(cur, end, fragment, microsecond))) + RETURN_ERROR_IF_FAILED(date.setMicrosecond(microsecond)) return cur; } @@ -1775,7 +1818,7 @@ namespace return cur; } else - RETURN_ERROR(ErrorCodes::CANNOT_PARSE_DATETIME, "Unable to build date time from timezone {}", read_time_zone) + RETURN_ERROR(ErrorCodes::CANNOT_PARSE_DATETIME, "Unable to parse date time from timezone {}", read_time_zone) } [[nodiscard]] @@ -2223,7 +2266,7 @@ namespace } - String getFormat(const ColumnsWithTypeAndName & arguments, UInt32 precision) const + String getFormat(const ColumnsWithTypeAndName & arguments, UInt32 scale) const { size_t format_arg_index = 1; if constexpr (return_type == ReturnType::DateTime64) @@ -2247,15 +2290,17 @@ namespace format = "%Y-%m-%d %H:%i:%s"; else format = "yyyy-MM-dd HH:mm:ss"; - if (precision > 0) - format += "." + String(precision, 'S'); + if (scale > 0) + { + if constexpr (parse_syntax == ParseSyntax::MySQL) + format += ".%f"; + else + format += "." + String(scale, 'S'); + } return format; } else { - if (!arguments[format_arg_index].column || !isColumnConst(*arguments[format_arg_index].column)) - throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Argument at index {} for function {} must be constant", format_arg_index, getName()); - const auto * col_format = checkAndGetColumnConst(arguments[format_arg_index].column.get()); if (!col_format) throw Exception( @@ -2269,18 +2314,24 @@ namespace const DateLUTImpl & getTimeZone(const ColumnsWithTypeAndName & arguments) const { - size_t timezone_arg_index = 2; - if constexpr (return_type == ReturnType::DateTime64) - timezone_arg_index = 3; - - if (arguments.size() <= timezone_arg_index) + if (arguments.size() < 3) return DateLUT::instance(); - + else if constexpr (return_type == ReturnType::DateTime64) + { + /// If the return type is DateTime64, and the second argument is UInt type for scale, then it has 2 reasonable situations: + /// the first like parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-07 17:27.30.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT+8') + /// the second like parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-07 17:27.30.123456', 6, '%Y-%m-%d %H:%i:%s.%f'). And for the + /// first one, we should return the last argument as its timezone, and for the second one, we should return the default time zone as + /// `DateLUT::instance()`. + if (isUInt(arguments[1].type) && arguments.size() < 4) + return DateLUT::instance(); + } + size_t timezone_arg_index = arguments.size() - 1; const auto * col = checkAndGetColumnConst(arguments[timezone_arg_index].column.get()); if (!col) throw Exception( ErrorCodes::ILLEGAL_COLUMN, - "Illegal column {} of third ('timezone') argument of function {}. Must be constant String.", + "Illegal column {} of ('timezone') argument of function {}. Must be constant String.", arguments[timezone_arg_index].column->getName(), getName()); diff --git a/tests/queries/0_stateless/03252_parse_datetime64.reference b/tests/queries/0_stateless/03252_parse_datetime64.reference new file mode 100644 index 00000000000..27dcef6bf68 --- /dev/null +++ b/tests/queries/0_stateless/03252_parse_datetime64.reference @@ -0,0 +1,17 @@ +2024-10-09 10:30:10.123456 +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +2024-10-09 10:30:10.123456 +1970-01-01 08:00:00.000000 +1970-01-01 08:00:00.000 +1970-01-01 08:00:00.000 +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +1970-01-01 08:00:00.000 +2024-10-09 10:30:10.123456 +\N +\N +\N +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +\N diff --git a/tests/queries/0_stateless/03252_parse_datetime64.sql b/tests/queries/0_stateless/03252_parse_datetime64.sql new file mode 100644 index 00000000000..d28b6e586f7 --- /dev/null +++ b/tests/queries/0_stateless/03252_parse_datetime64.sql @@ -0,0 +1,32 @@ +set session_timezone = 'Asia/Shanghai'; + +select parseDateTime64('2024-10-09 10:30:10.123456'); +select parseDateTime64('2024-10-09 10:30:10.123'); -- { serverError NOT_ENOUGH_SPACE } +select parseDateTime64('2024-10-09 10:30:10', 3); -- { serverError NOT_ENOUGH_SPACE } +select parseDateTime64('2024-10-09 10:30:10.', 3); -- { serverError CANNOT_PARSE_DATETIME } +select parseDateTime64('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} +select parseDateTime64('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64('2024-10-09 10:30:10.123456', 6), parseDateTime64('2024-10-09 10:30:10.123456', '%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'), parseDateTime64('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7'); +select parseDateTime64('2024-10-09 10:30:10.123', 3, '%Y-%m-%d %H:%i:%s.%f'); -- { serverError CANNOT_PARSE_DATETIME } +select parseDateTime64('2024-10-09 10:30:10.123', 6, '%Y-%m-%d %H:%i:%s.%f'); -- { serverError NOT_ENOUGH_SPACE } + +select parseDateTime64OrZero('2024-10-09 10:30:10.123456'); +select parseDateTime64OrZero('2024-10-09 10:30:10.123'); +select parseDateTime64OrZero('2024-10-09 10:30:10', 3); +select parseDateTime64OrZero('2024-10-09 10:30:10.', 3); +select parseDateTime64OrZero('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} +select parseDateTime64OrZero('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6), parseDateTime64OrZero('2024-10-09 10:30:10.123456', '%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'), parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7'); +select parseDateTime64OrZero('2024-10-09 10:30:10.123', 3, '%Y-%m-%d %H:%i:%s.%f'); + +select parseDateTime64OrNull('2024-10-09 10:30:10.123456'); +select parseDateTime64OrNull('2024-10-09 10:30:10.123'); +select parseDateTime64OrNull('2024-10-09 10:30:10', 3); +select parseDateTime64OrNull('2024-10-09 10:30:10.', 3); +select parseDateTime64OrNull('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} +select parseDateTime64OrNull('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6), parseDateTime64OrZero('2024-10-09 10:30:10.123456', '%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'), parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7');; +select parseDateTime64OrNull('2024-10-09 10:30:10.123', 3, '%Y-%m-%d %H:%i:%s.%f'); \ No newline at end of file diff --git a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference index 063b76b152c..0b4a28c4b38 100644 --- a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference +++ b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference @@ -1,14 +1,26 @@ -2024-10-09 10:30:10.123 -2024-10-09 10:30:10.123456 -2024-10-10 02:30:10.123456 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-10 02:42:10.123456 2024-10-10 01:30:10.123456 -2024-10-09 10:30:10.123 -2024-10-09 10:30:10.123456 -1970-01-01 08:00:00.000000000 -2024-10-10 02:30:10.123456 2024-10-10 01:30:10.123456 -2024-10-09 10:30:10.123 -2024-10-09 10:30:10.123456 +1970-01-01 08:00:00.000 +1970-01-01 08:00:00.000 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-10 02:42:10.123456 +1970-01-01 08:00:00.000000 +2024-10-10 01:30:10.123456 +2024-10-10 01:30:10.123456 +1970-01-01 08:00:00.000000 +\N +\N +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 +2024-10-10 02:42:10.123456 \N -2024-10-10 02:30:10.123456 2024-10-10 01:30:10.123456 +2024-10-10 01:30:10.123456 +\N diff --git a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql index 9ea854bc324..8482677e9c9 100644 --- a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql +++ b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql @@ -1,19 +1,49 @@ set session_timezone = 'Asia/Shanghai'; -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'); -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456789', 'yyyy-MM-dd HH:mm:ss.SSSSSSSSS'); -- { serverError CANNOT_PARSE_DATETIME } -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456-0800', 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456America/Los_Angeles', 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10', 3); -- { serverError NOT_ENOUGH_SPACE } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.', 3); -- { serverError CANNOT_PARSE_DATETIME } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 3), parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 6); +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSS'); +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSSS'); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456-0812', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456-08123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZZZ'); -- {serverError CANNOT_PARSE_DATETIME} +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456America/Los_Angeles', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456America/Los_Angeles', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSzzz'); +-- incorrect timezone offset and timezone +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456-8000', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); -- { serverError CANNOT_PARSE_DATETIME } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456ABCD', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); -- { serverError BAD_ARGUMENTS } -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'); -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456789', 'yyyy-MM-dd HH:mm:ss.SSSSSSSSS'); -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456-0800', 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456America/Los_Angeles', 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10', 3); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.', 3); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 3), parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 6); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSS'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSSS'); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456-0812', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456-08123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZZZ'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456America/Los_Angeles', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456America/Los_Angeles', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSzzz'); +-- incorrect timezone offset and timezone +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456-8000', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); +select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456ABCD', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); -- { serverError BAD_ARGUMENTS } -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'); -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456789', 'yyyy-MM-dd HH:mm:ss.SSSSSSSSS'); -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456-0800', 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456America/Los_Angeles', 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10', 3); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.', 3); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 3), parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 6); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSS'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSSS'); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456-0812', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456-08123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZZZ'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456America/Los_Angeles', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456America/Los_Angeles', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSzzz'); +-- incorrect timezone offset and timezone +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456-8000', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); +select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456ABCD', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); -- { serverError BAD_ARGUMENTS } \ No newline at end of file From a6b08187b31b3d3d6a5432bf8f39fe81ab5d81a7 Mon Sep 17 00:00:00 2001 From: kevinyhzou Date: Thu, 7 Nov 2024 20:03:44 +0800 Subject: [PATCH 071/244] checkstyle and doc --- .../functions/type-conversion-functions.md | 46 ++++++++++++++++++- src/Functions/parseDateTime.cpp | 27 +++++------ 2 files changed, 57 insertions(+), 16 deletions(-) diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md index 91bae2fe9da..c44d9ddb12b 100644 --- a/docs/en/sql-reference/functions/type-conversion-functions.md +++ b/docs/en/sql-reference/functions/type-conversion-functions.md @@ -6867,9 +6867,53 @@ Same as for [parseDateTimeInJodaSyntax](#parsedatetimeinjodasyntax) except that Same as for [parseDateTimeInJodaSyntax](#parsedatetimeinjodasyntax) except that it returns `NULL` when it encounters a date format that cannot be processed. +## parseDateTime64 + +Converts a [String](../data-types/string.md) to [DateTime64](../data-types/datetime64.md) according to a [MySQL format string](https://dev.mysql.com/doc/refman/8.0/en/date-and-time-functions.html#function_date-format). + +**Syntax** + +``` sql +parseDateTime64(str[, [scale, [format[, timezone]]]]) +``` + +**Arguments** + +- `str` — The String to be parsed +- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default is 6 if not specified. +- `format` — The format string. Optional. `%Y-%m-%d %H:%i:%s` if not specified. +- `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. + +**Returned value(s)** + +Returns [DateTime64](../data-types/datetime64.md) type values parsed from input string according to a MySQL style format string. + +## parseDateTime64OrZero +Same as for [parseDateTime64](#parsedatetime64) except that it returns zero date when it encounters a date format that cannot be processed. + +## parseDateTime64OrNull +Same as for [parseDateTime64](#parsedatetime64) except that it returns zero date when it encounters a date format that cannot be processed. + ## parseDateTime64InJodaSyntax -Similar to [parseDateTimeInJodaSyntax](#parsedatetimeinjodasyntax). Differently, it returns a value of type [DateTime64](../data-types/datetime64.md). +Converts a [String](../data-types/string.md) to [DateTime64](../data-types/datetime64.md) according to a [Joda format string](https://joda-time.sourceforge.net/apidocs/org/joda/time/format/DateTimeFormat.html). + +**Syntax** + +``` sql +parseDateTime64InJodaSyntax(str[, [scale, [format[, timezone]]]]) +``` + +**Arguments** + +- `str` — The String to be parsed +- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default is 6 if not specified. +- `format` — The format string. Optional. `yyyy-MM-dd HH:mm:ss` if not specified. +- `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. + +**Returned value(s)** + +Returns [DateTime64](../data-types/datetime64.md) type values parsed from input string according to a joda style format string. ## parseDateTime64InJodaSyntaxOrZero diff --git a/src/Functions/parseDateTime.cpp b/src/Functions/parseDateTime.cpp index 9f7f78dcbe2..5743278e104 100644 --- a/src/Functions/parseDateTime.cpp +++ b/src/Functions/parseDateTime.cpp @@ -613,15 +613,12 @@ namespace FunctionArgumentDescriptors optional_args; if constexpr (return_type == ReturnType::DateTime64) - { - optional_args = { - {"scale/format", static_cast([](const IDataType & data_type) -> bool { - return isUInt(data_type) || isString(data_type); - }), nullptr, "UInt or String"}, + optional_args = {{"scale/format", static_cast( + [](const IDataType & data_type) -> bool { return isUInt(data_type) || isString(data_type); } + ), nullptr, "UInt or String"}, {"format", static_cast(&isString), nullptr, "String"}, {"timezone", static_cast(&isString), &isColumnConst, "const String"} }; - } else optional_args = { {"format", static_cast(&isString), nullptr, "String"}, @@ -659,7 +656,7 @@ namespace } /// Construct the return type `DataTypDateTime64` with scale and time zone name. The scale value can be specified or be extracted - /// from the format string by c how many 'S' characters are contained in the format's micorsceond fragment. + /// from the format string by counting how many 'S' characters are contained in the format's micorsceond fragment. String format = getFormat(arguments, scale); std::vector instructions = parseFormat(format); for (const auto & instruction : instructions) @@ -676,7 +673,7 @@ namespace else val++; } - /// If the scale is already specified by the second argument, but it not equals the value that extract from the format string, + /// If the scale is already specified by the second argument, but it not equals the value that extract from the format string, /// then we should throw an exception; If the scale is not specified, then we should set its value as the extracted one. if (val != 0 && scale != 0 && val != scale) throw Exception(ErrorCodes::BAD_ARGUMENTS, @@ -687,7 +684,7 @@ namespace scale = val; } if (scale > maxPrecisionOfDateTime64) - throw Exception(ErrorCodes::BAD_ARGUMENTS, + throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale of the input format string {} exceed the max scale value {}.", format, maxPrecisionOfDateTime64); @@ -709,7 +706,7 @@ namespace non_null_result_type = removeNullable(result_type); else non_null_result_type = result_type; - + if constexpr (return_type == ReturnType::DateTime64) { const auto * datatime64_type = checkAndGetDataType(non_null_result_type.get()); @@ -726,7 +723,7 @@ namespace } template - ColumnPtr executeImpl2(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count, + ColumnPtr executeImpl2(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count, MutableColumnPtr & col_res, PaddedPODArray & res_data) const { const auto * col_str = checkAndGetColumn(arguments[0].column.get()); @@ -736,7 +733,7 @@ namespace "Illegal column {} of first ('str') argument of function {}. Must be string.", arguments[0].column->getName(), getName()); - + ColumnUInt8::MutablePtr col_null_map; if constexpr (error_handling == ErrorHandling::Null) col_null_map = ColumnUInt8::create(input_rows_count, 0); @@ -802,7 +799,7 @@ namespace Int64OrError result = 0; - /// Ensure all input was consumed when the return type is `DateTime`. + /// Ensure all input was consumed. if (cur < end) { result = tl::unexpected(ErrorCodeAndMessage( @@ -2273,7 +2270,7 @@ namespace { /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22.22.123', 3), then the format is treated /// as default value `yyyy-MM-dd HH:mm:ss`. - /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 'yyyy-MM-dd HH:mm:ss.SSS')`, + /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 'yyyy-MM-dd HH:mm:ss.SSS')`, /// then the second argument is the format. /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS')`, /// then the third argument is the format. @@ -2321,7 +2318,7 @@ namespace /// If the return type is DateTime64, and the second argument is UInt type for scale, then it has 2 reasonable situations: /// the first like parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-07 17:27.30.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT+8') /// the second like parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-07 17:27.30.123456', 6, '%Y-%m-%d %H:%i:%s.%f'). And for the - /// first one, we should return the last argument as its timezone, and for the second one, we should return the default time zone as + /// first one, we should return the last argument as its timezone, and for the second one, we should return the default time zone as /// `DateLUT::instance()`. if (isUInt(arguments[1].type) && arguments.size() < 4) return DateLUT::instance(); From 552b0fc8d0f106db1a85805ab883debe7e491e9c Mon Sep 17 00:00:00 2001 From: kssenii Date: Thu, 7 Nov 2024 13:11:33 +0100 Subject: [PATCH 072/244] Rename a setting --- src/Core/Settings.cpp | 3 ++- src/IO/ReadSettings.h | 2 +- src/Interpreters/Cache/QueryLimit.cpp | 2 +- src/Interpreters/Context.cpp | 4 ++-- 4 files changed, 6 insertions(+), 5 deletions(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index c2ffc2ddf0e..d9668849fd2 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -4852,7 +4852,7 @@ Allows to record the filesystem caching log for each query DECLARE(Bool, read_from_filesystem_cache_if_exists_otherwise_bypass_cache, false, R"( Allow to use the filesystem cache in passive mode - benefit from the existing cache entries, but don't put more entries into the cache. If you set this setting for heavy ad-hoc queries and leave it disabled for short real-time queries, this will allows to avoid cache threshing by too heavy queries and to improve the overall system efficiency. )", 0) \ - DECLARE(Bool, skip_download_if_exceeds_query_cache, true, R"( + DECLARE(Bool, filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit, true, R"( Skip download from remote filesystem if exceeds query cache size )", 0) \ DECLARE(UInt64, filesystem_cache_max_download_size, (128UL * 1024 * 1024 * 1024), R"( @@ -5887,6 +5887,7 @@ Experimental data deduplication for SELECT queries based on part UUIDs MAKE_OBSOLETE(M, Bool, use_mysql_types_in_show_columns, false) \ MAKE_OBSOLETE(M, Bool, s3queue_allow_experimental_sharded_mode, false) \ MAKE_OBSOLETE(M, LightweightMutationProjectionMode, lightweight_mutation_projection_mode, LightweightMutationProjectionMode::THROW) \ + MAKE_OBSOLETE(M, Bool, skip_download_if_exceeds_query_cache, true) \ /* moved to config.xml: see also src/Core/ServerSettings.h */ \ MAKE_DEPRECATED_BY_SERVER_CONFIG(M, UInt64, background_buffer_flush_schedule_pool_size, 16) \ MAKE_DEPRECATED_BY_SERVER_CONFIG(M, UInt64, background_pool_size, 16) \ diff --git a/src/IO/ReadSettings.h b/src/IO/ReadSettings.h index 6ed02212095..103ce7df54b 100644 --- a/src/IO/ReadSettings.h +++ b/src/IO/ReadSettings.h @@ -68,7 +68,7 @@ struct ReadSettings std::shared_ptr page_cache; size_t filesystem_cache_max_download_size = (128UL * 1024 * 1024 * 1024); - bool skip_download_if_exceeds_query_cache = true; + bool filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit = true; size_t remote_read_min_bytes_for_seek = DBMS_DEFAULT_BUFFER_SIZE; diff --git a/src/Interpreters/Cache/QueryLimit.cpp b/src/Interpreters/Cache/QueryLimit.cpp index b18d23a5b7f..a7c964022a5 100644 --- a/src/Interpreters/Cache/QueryLimit.cpp +++ b/src/Interpreters/Cache/QueryLimit.cpp @@ -53,7 +53,7 @@ FileCacheQueryLimit::QueryContextPtr FileCacheQueryLimit::getOrSetQueryContext( { it->second = std::make_shared( settings.filesystem_cache_max_download_size, - !settings.skip_download_if_exceeds_query_cache); + !settings.filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit); } return it->second; diff --git a/src/Interpreters/Context.cpp b/src/Interpreters/Context.cpp index c1fa2c8549a..7b7cdfa2104 100644 --- a/src/Interpreters/Context.cpp +++ b/src/Interpreters/Context.cpp @@ -236,7 +236,7 @@ namespace Setting extern const SettingsUInt64 remote_fs_read_backoff_max_tries; extern const SettingsUInt64 remote_read_min_bytes_for_seek; extern const SettingsBool throw_on_error_from_cache_on_write_operations; - extern const SettingsBool skip_download_if_exceeds_query_cache; + extern const SettingsBool filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit; extern const SettingsBool s3_allow_parallel_part_upload; extern const SettingsBool use_page_cache_for_disks_without_file_cache; extern const SettingsUInt64 use_structure_from_insertion_table_in_table_functions; @@ -5753,7 +5753,7 @@ ReadSettings Context::getReadSettings() const res.filesystem_cache_allow_background_download_during_fetch = settings_ref[Setting::filesystem_cache_enable_background_download_during_fetch]; res.filesystem_cache_max_download_size = settings_ref[Setting::filesystem_cache_max_download_size]; - res.skip_download_if_exceeds_query_cache = settings_ref[Setting::skip_download_if_exceeds_query_cache]; + res.filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit = settings_ref[Setting::filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit]; res.page_cache = getPageCache(); res.use_page_cache_for_disks_without_file_cache = settings_ref[Setting::use_page_cache_for_disks_without_file_cache]; From 3332bce1dc94e7fddccc1865df01eb029e5f7e52 Mon Sep 17 00:00:00 2001 From: kevinyhzou Date: Thu, 7 Nov 2024 20:38:44 +0800 Subject: [PATCH 073/244] fix doc and comments --- .../functions/type-conversion-functions.md | 6 +++--- src/Functions/parseDateTime.cpp | 16 ++++++++-------- 2 files changed, 11 insertions(+), 11 deletions(-) diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md index c44d9ddb12b..8043b21744a 100644 --- a/docs/en/sql-reference/functions/type-conversion-functions.md +++ b/docs/en/sql-reference/functions/type-conversion-functions.md @@ -6880,7 +6880,7 @@ parseDateTime64(str[, [scale, [format[, timezone]]]]) **Arguments** - `str` — The String to be parsed -- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default is 6 if not specified. +- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default 6 if not specified. - `format` — The format string. Optional. `%Y-%m-%d %H:%i:%s` if not specified. - `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. @@ -6892,7 +6892,7 @@ Returns [DateTime64](../data-types/datetime64.md) type values parsed from input Same as for [parseDateTime64](#parsedatetime64) except that it returns zero date when it encounters a date format that cannot be processed. ## parseDateTime64OrNull -Same as for [parseDateTime64](#parsedatetime64) except that it returns zero date when it encounters a date format that cannot be processed. +Same as for [parseDateTime64](#parsedatetime64) except that it returns `NULL` when it encounters a date format that cannot be processed. ## parseDateTime64InJodaSyntax @@ -6907,7 +6907,7 @@ parseDateTime64InJodaSyntax(str[, [scale, [format[, timezone]]]]) **Arguments** - `str` — The String to be parsed -- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default is 6 if not specified. +- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default 0 if not specified. - `format` — The format string. Optional. `yyyy-MM-dd HH:mm:ss` if not specified. - `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. diff --git a/src/Functions/parseDateTime.cpp b/src/Functions/parseDateTime.cpp index 5743278e104..7190c1ad6f8 100644 --- a/src/Functions/parseDateTime.cpp +++ b/src/Functions/parseDateTime.cpp @@ -65,7 +65,7 @@ namespace constexpr Int32 minYear = 1970; constexpr Int32 maxYear = 2106; - constexpr Int32 maxPrecisionOfDateTime64 = 6; + constexpr Int32 maxScaleOfDateTime64 = 6; const std::unordered_map> dayOfWeekMap{ {"mon", {"day", 1}}, @@ -193,7 +193,7 @@ namespace Int32 minute = 0; /// range [0, 59] Int32 second = 0; /// range [0, 59] Int32 microsecond = 0; /// range [0, 999999] - UInt32 scale = 0; /// The microsecond scale of DateTime64. + UInt32 scale = 0; /// The scale of DateTime64, range [0, 6]. bool is_am = true; /// If is_hour_of_half_day = true and is_am = false (i.e. pm) then add 12 hours to the result DateTime bool hour_starts_at_1 = false; /// Whether the hour is clockhour @@ -646,8 +646,8 @@ namespace scale = col_scale->getValue(); else throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The input scale value may exceed the max value of `DateTime64`: {}.", - maxPrecisionOfDateTime64); + "The input scale value may exceed the max scale value of `DateTime64`: {}.", + maxScaleOfDateTime64); } else { @@ -656,7 +656,7 @@ namespace } /// Construct the return type `DataTypDateTime64` with scale and time zone name. The scale value can be specified or be extracted - /// from the format string by counting how many 'S' characters are contained in the format's micorsceond fragment. + /// from the format string by counting how many 'S' characters are contained in the format's microsceond fragment. String format = getFormat(arguments, scale); std::vector instructions = parseFormat(format); for (const auto & instruction : instructions) @@ -683,11 +683,11 @@ namespace else if (scale == 0 && val != 0) scale = val; } - if (scale > maxPrecisionOfDateTime64) + if (scale > maxScaleOfDateTime64) throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale of the input format string {} exceed the max scale value {}.", format, - maxPrecisionOfDateTime64); + maxScaleOfDateTime64); } data_type = std::make_shared(scale, time_zone_name); } @@ -1398,7 +1398,7 @@ namespace if (date.scale != 6) RETURN_ERROR( ErrorCodes::CANNOT_PARSE_DATETIME, - "Unable to parse fragment {} from {} because of the microsecond's scale {} is not 6", + "Unable to parse fragment {} from {} because of the datetime scale {} is not 6", fragment, std::string_view(cur, end - cur), std::to_string(date.scale)) From ca23e5254c2cca5e6b3f4a9c7ccd65f70be42fc4 Mon Sep 17 00:00:00 2001 From: avogar Date: Thu, 7 Nov 2024 12:44:57 +0000 Subject: [PATCH 074/244] Fix for tmp parts --- src/Storages/MergeTree/IMergeTreeDataPart.cpp | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/src/Storages/MergeTree/IMergeTreeDataPart.cpp b/src/Storages/MergeTree/IMergeTreeDataPart.cpp index 20d7528d38a..fb934a77512 100644 --- a/src/Storages/MergeTree/IMergeTreeDataPart.cpp +++ b/src/Storages/MergeTree/IMergeTreeDataPart.cpp @@ -2501,7 +2501,7 @@ ColumnPtr IMergeTreeDataPart::getColumnSample(const NameAndTypePair & column) co { const size_t total_mark = getMarksCount(); /// If column doesn't have dynamic subcolumns or part has no data, just create column using it's type. - if (!column.type->hasDynamicSubcolumns() || !total_mark) + if (is_temp || !column.type->hasDynamicSubcolumns() || !total_mark) return column.type->createColumn(); /// Otherwise, read sample column with 0 rows from the part, so it will load dynamic structure. @@ -2510,22 +2510,24 @@ ColumnPtr IMergeTreeDataPart::getColumnSample(const NameAndTypePair & column) co StorageMetadataPtr metadata_ptr = storage.getInMemoryMetadataPtr(); StorageSnapshotPtr storage_snapshot_ptr = std::make_shared(storage, metadata_ptr); + MergeTreeReaderSettings settings; + settings.can_read_part_without_marks = true; MergeTreeReaderPtr reader = getReader( cols, storage_snapshot_ptr, - MarkRanges{MarkRange(0, 1)}, + MarkRanges{MarkRange(0, total_mark)}, /*virtual_fields=*/ {}, /*uncompressed_cache=*/{}, storage.getContext()->getMarkCache().get(), std::make_shared(), - MergeTreeReaderSettings{}, + settings, ValueSizeMap{}, ReadBufferFromFileBase::ProfileCallback{}); Columns result; result.resize(1); - reader->readRows(0, 1, false, 0, result); + reader->readRows(0, total_mark, false, 0, result); return result[0]; } From e5fc37bc7e6c707cd7ea14bb3c4888f94118a126 Mon Sep 17 00:00:00 2001 From: kssenii Date: Thu, 7 Nov 2024 17:27:51 +0100 Subject: [PATCH 075/244] Add alias --- src/Core/Settings.cpp | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index d9668849fd2..328f950da1d 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -4854,7 +4854,7 @@ Allow to use the filesystem cache in passive mode - benefit from the existing ca )", 0) \ DECLARE(Bool, filesystem_cache_skip_download_if_exceeds_per_query_cache_write_limit, true, R"( Skip download from remote filesystem if exceeds query cache size -)", 0) \ +)", 0) ALIAS(skip_download_if_exceeds_query_cache) \ DECLARE(UInt64, filesystem_cache_max_download_size, (128UL * 1024 * 1024 * 1024), R"( Max remote filesystem cache size that can be downloaded by a single query )", 0) \ @@ -5887,7 +5887,6 @@ Experimental data deduplication for SELECT queries based on part UUIDs MAKE_OBSOLETE(M, Bool, use_mysql_types_in_show_columns, false) \ MAKE_OBSOLETE(M, Bool, s3queue_allow_experimental_sharded_mode, false) \ MAKE_OBSOLETE(M, LightweightMutationProjectionMode, lightweight_mutation_projection_mode, LightweightMutationProjectionMode::THROW) \ - MAKE_OBSOLETE(M, Bool, skip_download_if_exceeds_query_cache, true) \ /* moved to config.xml: see also src/Core/ServerSettings.h */ \ MAKE_DEPRECATED_BY_SERVER_CONFIG(M, UInt64, background_buffer_flush_schedule_pool_size, 16) \ MAKE_DEPRECATED_BY_SERVER_CONFIG(M, UInt64, background_pool_size, 16) \ From 0ac6ce56bd08e25fc9c22022fec21f3346a753c5 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Thu, 7 Nov 2024 18:19:26 +0000 Subject: [PATCH 076/244] Trying to fix short-circuit for FilterStep. --- src/Processors/QueryPlan/FilterStep.cpp | 98 ++++++++++++++++++++++++- 1 file changed, 97 insertions(+), 1 deletion(-) diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp index 862e03d74f2..64c46332c34 100644 --- a/src/Processors/QueryPlan/FilterStep.cpp +++ b/src/Processors/QueryPlan/FilterStep.cpp @@ -5,6 +5,9 @@ #include #include #include +#include +#include +#include namespace DB { @@ -24,6 +27,78 @@ static ITransformingStep::Traits getTraits() }; } +static bool isTrivialSubtree(const ActionsDAG::Node * node) +{ + while (node->type == ActionsDAG::ActionType::ALIAS) + node = node->children.at(0); + + return node->type != ActionsDAG::ActionType::FUNCTION && node->type != ActionsDAG::ActionType::ARRAY_JOIN; +} + +struct ActionsAndName +{ + ActionsDAG dag; + std::string name; +}; + +static ActionsAndName splitSingleAndFilter(ActionsDAG & dag, const ActionsDAG::Node * filter_node) +{ + auto name = filter_node->result_name; + auto split_result = dag.split({filter_node}, true); + dag = std::move(split_result.second); + split_result.first.getOutputs().emplace(split_result.first.getOutputs().begin(), split_result.split_nodes_mapping[filter_node]); + return ActionsAndName{std::move(split_result.first), std::move(name)}; +} + +static std::optional trySplitSingleAndFilter(ActionsDAG & dag, const std::string & filter_name) +{ + const auto * filter = &dag.findInOutputs(filter_name); + while (filter->type == ActionsDAG::ActionType::ALIAS) + filter = filter->children.at(0); + + if (filter->type != ActionsDAG::ActionType::FUNCTION || filter->function_base->getName() != "and") + return {}; + + const ActionsDAG::Node * condition_to_split = nullptr; + std::stack nodes; + nodes.push(filter); + while (!nodes.empty()) + { + const auto * node = nodes.top(); + nodes.pop(); + + if (node->type == ActionsDAG::ActionType::FUNCTION && node->function_base->getName() == "and") + { + for (const auto * child : node->children | std::ranges::views::reverse) + nodes.push(child); + + continue; + } + + if (isTrivialSubtree(node)) + continue; + + /// Do not split subtree if it's the last non-trivial one. + /// So, split the first found condition only when there is a another one found. + if (condition_to_split) + return splitSingleAndFilter(dag, condition_to_split); + + condition_to_split = node; + } + + return {}; +} + +std::vector splitAndChainIntoMultipleFilters(ActionsDAG & dag, const std::string & filter_name) +{ + std::vector res; + + while (auto condition = trySplitSingleAndFilter(dag, filter_name)) + res.push_back(std::move(*condition)); + + return res; +} + FilterStep::FilterStep( const Header & input_header_, ActionsDAG actions_dag_, @@ -50,6 +125,17 @@ FilterStep::FilterStep( void FilterStep::transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & settings) { + auto and_atoms = splitAndChainIntoMultipleFilters(actions_dag, filter_column_name); + for (auto & and_atom : and_atoms) + { + auto expression = std::make_shared(std::move(and_atom.dag), settings.getActionsSettings()); + pipeline.addSimpleTransform([&](const Block & header, QueryPipelineBuilder::StreamType stream_type) + { + bool on_totals = stream_type == QueryPipelineBuilder::StreamType::Totals; + return std::make_shared(header, expression, and_atom.name, true, on_totals); + }); + } + auto expression = std::make_shared(std::move(actions_dag), settings.getActionsSettings()); pipeline.addSimpleTransform([&](const Block & header, QueryPipelineBuilder::StreamType stream_type) @@ -76,13 +162,23 @@ void FilterStep::transformPipeline(QueryPipelineBuilder & pipeline, const BuildQ void FilterStep::describeActions(FormatSettings & settings) const { String prefix(settings.offset, settings.indent_char); + + auto cloned_dag = actions_dag.clone(); + auto and_atoms = splitAndChainIntoMultipleFilters(cloned_dag, filter_column_name); + for (auto & and_atom : and_atoms) + { + auto expression = std::make_shared(std::move(and_atom.dag)); + settings.out << prefix << "AND column: " << and_atom.name; + expression->describeActions(settings.out, prefix); + } + settings.out << prefix << "Filter column: " << filter_column_name; if (remove_filter_column) settings.out << " (removed)"; settings.out << '\n'; - auto expression = std::make_shared(actions_dag.clone()); + auto expression = std::make_shared(std::move(cloned_dag)); expression->describeActions(settings.out, prefix); } From 1157028e35139ead00c7475045795bde966a2e40 Mon Sep 17 00:00:00 2001 From: avogar Date: Thu, 7 Nov 2024 19:53:30 +0000 Subject: [PATCH 077/244] Fix getting column sample for not finalized part --- src/Columns/ColumnVariant.cpp | 2 +- src/Storages/MergeTree/IMergeTreeDataPart.cpp | 11 ++++++----- src/Storages/MergeTree/IMergeTreeDataPart.h | 6 +++--- src/Storages/MergeTree/IMergeTreeDataPartWriter.h | 2 ++ src/Storages/MergeTree/MergeTreeDataPartCompact.cpp | 2 +- src/Storages/MergeTree/MergeTreeDataPartCompact.h | 2 +- src/Storages/MergeTree/MergeTreeDataPartWide.cpp | 8 ++++---- src/Storages/MergeTree/MergeTreeDataPartWide.h | 4 ++-- .../MergeTree/MergeTreeDataPartWriterOnDisk.h | 2 ++ src/Storages/MergeTree/MergedBlockOutputStream.cpp | 2 +- 10 files changed, 23 insertions(+), 18 deletions(-) diff --git a/src/Columns/ColumnVariant.cpp b/src/Columns/ColumnVariant.cpp index 54f0421fc4b..2fa59b8e33c 100644 --- a/src/Columns/ColumnVariant.cpp +++ b/src/Columns/ColumnVariant.cpp @@ -952,7 +952,7 @@ ColumnPtr ColumnVariant::permute(const Permutation & perm, size_t limit) const if (hasOnlyNulls()) { if (limit) - return cloneResized(limit); + return cloneResized(limit ? std::min(size(), limit) : size()); /// If no limit, we can just return current immutable column. return this->getPtr(); diff --git a/src/Storages/MergeTree/IMergeTreeDataPart.cpp b/src/Storages/MergeTree/IMergeTreeDataPart.cpp index b631d991e90..f73b52dbafd 100644 --- a/src/Storages/MergeTree/IMergeTreeDataPart.cpp +++ b/src/Storages/MergeTree/IMergeTreeDataPart.cpp @@ -2252,18 +2252,18 @@ void IMergeTreeDataPart::checkConsistencyWithProjections(bool require_part_metad proj_part->checkConsistency(require_part_metadata); } -void IMergeTreeDataPart::calculateColumnsAndSecondaryIndicesSizesOnDisk() +void IMergeTreeDataPart::calculateColumnsAndSecondaryIndicesSizesOnDisk(std::optional columns_sample) { - calculateColumnsSizesOnDisk(); + calculateColumnsSizesOnDisk(columns_sample); calculateSecondaryIndicesSizesOnDisk(); } -void IMergeTreeDataPart::calculateColumnsSizesOnDisk() +void IMergeTreeDataPart::calculateColumnsSizesOnDisk(std::optional columns_sample) { if (getColumns().empty() || checksums.empty()) throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot calculate columns sizes when columns or checksums are not initialized"); - calculateEachColumnSizes(columns_sizes, total_columns_size); + calculateEachColumnSizes(columns_sizes, total_columns_size, columns_sample); } void IMergeTreeDataPart::calculateSecondaryIndicesSizesOnDisk() @@ -2501,7 +2501,7 @@ ColumnPtr IMergeTreeDataPart::getColumnSample(const NameAndTypePair & column) co { const size_t total_mark = getMarksCount(); /// If column doesn't have dynamic subcolumns or part has no data, just create column using it's type. - if (is_temp || !column.type->hasDynamicSubcolumns() || !total_mark) + if (!column.type->hasDynamicSubcolumns() || !total_mark) return column.type->createColumn(); /// Otherwise, read sample column with 0 rows from the part, so it will load dynamic structure. @@ -2527,6 +2527,7 @@ ColumnPtr IMergeTreeDataPart::getColumnSample(const NameAndTypePair & column) co Columns result; result.resize(1); + LOG_DEBUG(getLogger("IMergeTreeDataPart"), "getColumnSample"); reader->readRows(0, total_mark, false, 0, result); return result[0]; } diff --git a/src/Storages/MergeTree/IMergeTreeDataPart.h b/src/Storages/MergeTree/IMergeTreeDataPart.h index b41a1d840e1..a7051a2491a 100644 --- a/src/Storages/MergeTree/IMergeTreeDataPart.h +++ b/src/Storages/MergeTree/IMergeTreeDataPart.h @@ -426,7 +426,7 @@ public: bool shallParticipateInMerges(const StoragePolicyPtr & storage_policy) const; /// Calculate column and secondary indices sizes on disk. - void calculateColumnsAndSecondaryIndicesSizesOnDisk(); + void calculateColumnsAndSecondaryIndicesSizesOnDisk(std::optional columns_sample = std::nullopt); std::optional getRelativePathForPrefix(const String & prefix, bool detached = false, bool broken = false) const; @@ -631,7 +631,7 @@ protected: /// Fill each_columns_size and total_size with sizes from columns files on /// disk using columns and checksums. - virtual void calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size) const = 0; + virtual void calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size, std::optional columns_sample) const = 0; std::optional getRelativePathForDetachedPart(const String & prefix, bool broken) const; @@ -713,7 +713,7 @@ private: void loadPartitionAndMinMaxIndex(); - void calculateColumnsSizesOnDisk(); + void calculateColumnsSizesOnDisk(std::optional columns_sample = std::nullopt); void calculateSecondaryIndicesSizesOnDisk(); diff --git a/src/Storages/MergeTree/IMergeTreeDataPartWriter.h b/src/Storages/MergeTree/IMergeTreeDataPartWriter.h index d1c76505d7c..8923f6a59ca 100644 --- a/src/Storages/MergeTree/IMergeTreeDataPartWriter.h +++ b/src/Storages/MergeTree/IMergeTreeDataPartWriter.h @@ -54,6 +54,8 @@ public: const MergeTreeIndexGranularity & getIndexGranularity() const { return index_granularity; } + virtual Block getColumnsSample() const = 0; + protected: SerializationPtr getSerialization(const String & column_name) const; diff --git a/src/Storages/MergeTree/MergeTreeDataPartCompact.cpp b/src/Storages/MergeTree/MergeTreeDataPartCompact.cpp index 14c2da82de1..8856f467b90 100644 --- a/src/Storages/MergeTree/MergeTreeDataPartCompact.cpp +++ b/src/Storages/MergeTree/MergeTreeDataPartCompact.cpp @@ -80,7 +80,7 @@ MergeTreeDataPartWriterPtr createMergeTreeDataPartCompactWriter( } -void MergeTreeDataPartCompact::calculateEachColumnSizes(ColumnSizeByName & /*each_columns_size*/, ColumnSize & total_size) const +void MergeTreeDataPartCompact::calculateEachColumnSizes(ColumnSizeByName & /*each_columns_size*/, ColumnSize & total_size, std::optional /*columns_sample*/) const { auto bin_checksum = checksums.files.find(DATA_FILE_NAME_WITH_EXTENSION); if (bin_checksum != checksums.files.end()) diff --git a/src/Storages/MergeTree/MergeTreeDataPartCompact.h b/src/Storages/MergeTree/MergeTreeDataPartCompact.h index 8e279571578..c394de0d7c1 100644 --- a/src/Storages/MergeTree/MergeTreeDataPartCompact.h +++ b/src/Storages/MergeTree/MergeTreeDataPartCompact.h @@ -70,7 +70,7 @@ private: void loadIndexGranularity() override; /// Compact parts don't support per column size, only total size - void calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size) const override; + void calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size, std::optional columns_sample) const override; }; } diff --git a/src/Storages/MergeTree/MergeTreeDataPartWide.cpp b/src/Storages/MergeTree/MergeTreeDataPartWide.cpp index b3b6a0dded6..39f96ba06ad 100644 --- a/src/Storages/MergeTree/MergeTreeDataPartWide.cpp +++ b/src/Storages/MergeTree/MergeTreeDataPartWide.cpp @@ -82,7 +82,7 @@ MergeTreeDataPartWriterPtr createMergeTreeDataPartWideWriter( /// Takes into account the fact that several columns can e.g. share their .size substreams. /// When calculating totals these should be counted only once. ColumnSize MergeTreeDataPartWide::getColumnSizeImpl( - const NameAndTypePair & column, std::unordered_set * processed_substreams) const + const NameAndTypePair & column, std::unordered_set * processed_substreams, std::optional columns_sample) const { ColumnSize size; if (checksums.empty()) @@ -108,7 +108,7 @@ ColumnSize MergeTreeDataPartWide::getColumnSizeImpl( auto mrk_checksum = checksums.files.find(*stream_name + getMarksFileExtension()); if (mrk_checksum != checksums.files.end()) size.marks += mrk_checksum->second.file_size; - }, column.type, getColumnSample(column)); + }, column.type, columns_sample && columns_sample->has(column.name) ? columns_sample->getByName(column.name).column : getColumnSample(column)); return size; } @@ -374,12 +374,12 @@ std::optional MergeTreeDataPartWide::getFileNameForColumn(const NameAndT return filename; } -void MergeTreeDataPartWide::calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size) const +void MergeTreeDataPartWide::calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size, std::optional columns_sample) const { std::unordered_set processed_substreams; for (const auto & column : columns) { - ColumnSize size = getColumnSizeImpl(column, &processed_substreams); + ColumnSize size = getColumnSizeImpl(column, &processed_substreams, columns_sample); each_columns_size[column.name] = size; total_size.add(size); diff --git a/src/Storages/MergeTree/MergeTreeDataPartWide.h b/src/Storages/MergeTree/MergeTreeDataPartWide.h index 022a5fb746c..a6d4897ed87 100644 --- a/src/Storages/MergeTree/MergeTreeDataPartWide.h +++ b/src/Storages/MergeTree/MergeTreeDataPartWide.h @@ -64,9 +64,9 @@ private: /// Loads marks index granularity into memory void loadIndexGranularity() override; - ColumnSize getColumnSizeImpl(const NameAndTypePair & column, std::unordered_set * processed_substreams) const; + ColumnSize getColumnSizeImpl(const NameAndTypePair & column, std::unordered_set * processed_substreams, std::optional columns_sample) const; - void calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size) const override; + void calculateEachColumnSizes(ColumnSizeByName & each_columns_size, ColumnSize & total_size, std::optional columns_sample) const override; }; diff --git a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h index 49d654c15e1..b22d58ba51e 100644 --- a/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h +++ b/src/Storages/MergeTree/MergeTreeDataPartWriterOnDisk.h @@ -123,6 +123,8 @@ public: written_offset_columns = written_offset_columns_; } + Block getColumnsSample() const override { return block_sample; } + protected: /// Count index_granularity for block and store in `index_granularity` size_t computeIndexGranularity(const Block & block) const; diff --git a/src/Storages/MergeTree/MergedBlockOutputStream.cpp b/src/Storages/MergeTree/MergedBlockOutputStream.cpp index 77c34aae30a..604b2fda20a 100644 --- a/src/Storages/MergeTree/MergedBlockOutputStream.cpp +++ b/src/Storages/MergeTree/MergedBlockOutputStream.cpp @@ -207,7 +207,7 @@ MergedBlockOutputStream::Finalizer MergedBlockOutputStream::finalizePartAsync( new_part->setBytesOnDisk(checksums.getTotalSizeOnDisk()); new_part->setBytesUncompressedOnDisk(checksums.getTotalSizeUncompressedOnDisk()); new_part->index_granularity = writer->getIndexGranularity(); - new_part->calculateColumnsAndSecondaryIndicesSizesOnDisk(); + new_part->calculateColumnsAndSecondaryIndicesSizesOnDisk(writer->getColumnsSample()); /// In mutation, existing_rows_count is already calculated in PartMergerWriter /// In merge situation, lightweight deleted rows was physically deleted, existing_rows_count equals rows_count From 0ff0c96b007108ab222a264e4a3bf8aa7cb7a18e Mon Sep 17 00:00:00 2001 From: avogar Date: Thu, 7 Nov 2024 20:01:40 +0000 Subject: [PATCH 078/244] Remove logging --- src/Storages/MergeTree/IMergeTreeDataPart.cpp | 1 - 1 file changed, 1 deletion(-) diff --git a/src/Storages/MergeTree/IMergeTreeDataPart.cpp b/src/Storages/MergeTree/IMergeTreeDataPart.cpp index f73b52dbafd..4e400fb1f94 100644 --- a/src/Storages/MergeTree/IMergeTreeDataPart.cpp +++ b/src/Storages/MergeTree/IMergeTreeDataPart.cpp @@ -2527,7 +2527,6 @@ ColumnPtr IMergeTreeDataPart::getColumnSample(const NameAndTypePair & column) co Columns result; result.resize(1); - LOG_DEBUG(getLogger("IMergeTreeDataPart"), "getColumnSample"); reader->readRows(0, total_mark, false, 0, result); return result[0]; } From a828e3e923ef06666d4582c34868750bbbee3e6a Mon Sep 17 00:00:00 2001 From: Max Kainov Date: Tue, 5 Nov 2024 12:59:14 +0000 Subject: [PATCH 079/244] test --- .github/workflows/pr.yaml | 108 ++++++++++++++++---------- ci/jobs/build_clickhouse.py | 30 +++---- ci/jobs/fast_test.py | 2 + ci/jobs/functional_stateless_tests.py | 92 +++++++++++++--------- ci/jobs/scripts/clickhouse_proc.py | 39 +++------- ci/praktika/_environment.py | 12 +-- ci/praktika/cidb.py | 2 +- ci/praktika/digest.py | 20 ++--- ci/praktika/environment.py | 3 - ci/praktika/hook_html.py | 20 ++--- ci/praktika/job.py | 1 + ci/praktika/json.html | 22 +++--- ci/praktika/mangle.py | 55 +++++++------ ci/praktika/native_jobs.py | 6 +- ci/praktika/param.py | 8 -- ci/praktika/result.py | 6 -- ci/praktika/runner.py | 7 +- ci/praktika/utils.py | 27 +++---- ci/praktika/validator.py | 89 ++++++++++----------- ci/praktika/yaml_generator.py | 11 +-- ci/settings/definitions.py | 5 +- ci/workflows/pull_request.py | 54 +++++++++---- tests/clickhouse-test | 30 +++---- tests/config/install.sh | 2 +- tests/docker_scripts/setup_minio.sh | 6 +- 25 files changed, 334 insertions(+), 323 deletions(-) delete mode 100644 ci/praktika/environment.py delete mode 100644 ci/praktika/param.py diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml index 0c3f74aeac8..51bb9b52d10 100644 --- a/.github/workflows/pr.yaml +++ b/.github/workflows/pr.yaml @@ -31,8 +31,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -72,8 +71,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -113,8 +111,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -154,8 +151,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -195,8 +191,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -236,8 +231,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -266,19 +260,18 @@ jobs: python3 -m praktika run --job '''Build (amd_release)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi - stateless_tests_amd_debug_parallel_1_2: + stateless_tests_amd_debugparallel: runs-on: [builder] needs: [config_workflow, docker_builds, build_amd_debug] - if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKSAocGFyYWxsZWwgMS8yKQ==') }} - name: "Stateless tests (amd, debug) (parallel 1/2)" + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWRfZGVidWcscGFyYWxsZWwp') }} + name: "Stateless tests (amd_debug,parallel)" outputs: data: ${{ steps.run.outputs.DATA }} steps: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -302,24 +295,63 @@ jobs: . /tmp/praktika_setup_env.sh set -o pipefail if command -v ts &> /dev/null; then - python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 1/2)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Stateless tests (amd_debug,parallel)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log else - python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 1/2)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Stateless tests (amd_debug,parallel)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi - stateless_tests_amd_debug_parallel_2_2: + stateless_tests_amd_debugnon_parallel: + runs-on: [func-tester] + needs: [config_workflow, docker_builds, build_amd_debug] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWRfZGVidWcsbm9uLXBhcmFsbGVsKQ==') }} + name: "Stateless tests (amd_debug,non-parallel)" + outputs: + data: ${{ steps.run.outputs.DATA }} + steps: + - name: Checkout code + uses: actions/checkout@v4 + with: + ref: ${{ github.head_ref }} + + - name: Prepare env script + run: | + cat > /tmp/praktika_setup_env.sh << 'ENV_SETUP_SCRIPT_EOF' + export PYTHONPATH=./ci:. + + cat > /tmp/praktika/workflow_config_pr.json << 'EOF' + ${{ needs.config_workflow.outputs.data }} + EOF + cat > /tmp/praktika/workflow_status.json << 'EOF' + ${{ toJson(needs) }} + EOF + ENV_SETUP_SCRIPT_EOF + + rm -rf /tmp/praktika/input /tmp/praktika/output /tmp/praktika + mkdir -p /tmp/praktika /tmp/praktika/input /tmp/praktika/output + + - name: Run + id: run + run: | + . /tmp/praktika_setup_env.sh + set -o pipefail + if command -v ts &> /dev/null; then + python3 -m praktika run --job '''Stateless tests (amd_debug,non-parallel)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + else + python3 -m praktika run --job '''Stateless tests (amd_debug,non-parallel)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + fi + + stateless_tests_amd_releaseparallel: runs-on: [builder] - needs: [config_workflow, docker_builds, build_amd_debug] - if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKSAocGFyYWxsZWwgMi8yKQ==') }} - name: "Stateless tests (amd, debug) (parallel 2/2)" + needs: [config_workflow, docker_builds, build_amd_release] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWRfcmVsZWFzZSxwYXJhbGxlbCk=') }} + name: "Stateless tests (amd_release,parallel)" outputs: data: ${{ steps.run.outputs.DATA }} steps: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -343,24 +375,23 @@ jobs: . /tmp/praktika_setup_env.sh set -o pipefail if command -v ts &> /dev/null; then - python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 2/2)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Stateless tests (amd_release,parallel)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log else - python3 -m praktika run --job '''Stateless tests (amd, debug) (parallel 2/2)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Stateless tests (amd_release,parallel)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi - stateless_tests_amd_debug_non_parallel: - runs-on: [style-checker] - needs: [config_workflow, docker_builds, build_amd_debug] - if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWQsIGRlYnVnKSAobm9uLXBhcmFsbGVsKQ==') }} - name: "Stateless tests (amd, debug) (non-parallel)" + stateless_tests_amd_releasenon_parallel: + runs-on: [func-tester] + needs: [config_workflow, docker_builds, build_amd_release] + if: ${{ !failure() && !cancelled() && !contains(fromJson(needs.config_workflow.outputs.data).cache_success_base64, 'U3RhdGVsZXNzIHRlc3RzIChhbWRfcmVsZWFzZSxub24tcGFyYWxsZWwp') }} + name: "Stateless tests (amd_release,non-parallel)" outputs: data: ${{ steps.run.outputs.DATA }} steps: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | @@ -384,14 +415,14 @@ jobs: . /tmp/praktika_setup_env.sh set -o pipefail if command -v ts &> /dev/null; then - python3 -m praktika run --job '''Stateless tests (amd, debug) (non-parallel)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Stateless tests (amd_release,non-parallel)''' --workflow "PR" --ci |& ts '[%Y-%m-%d %H:%M:%S]' | tee /tmp/praktika/praktika_run.log else - python3 -m praktika run --job '''Stateless tests (amd, debug) (non-parallel)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log + python3 -m praktika run --job '''Stateless tests (amd_release,non-parallel)''' --workflow "PR" --ci |& tee /tmp/praktika/praktika_run.log fi finish_workflow: runs-on: [ci_services] - needs: [config_workflow, docker_builds, style_check, fast_test, build_amd_debug, build_amd_release, stateless_tests_amd_debug_parallel_1_2, stateless_tests_amd_debug_parallel_2_2, stateless_tests_amd_debug_non_parallel] + needs: [config_workflow, docker_builds, style_check, fast_test, build_amd_debug, build_amd_release, stateless_tests_amd_debugparallel, stateless_tests_amd_debugnon_parallel, stateless_tests_amd_releaseparallel, stateless_tests_amd_releasenon_parallel] if: ${{ !cancelled() }} name: "Finish Workflow" outputs: @@ -400,8 +431,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{ github.event.pull_reguest.head.sha }} + ref: ${{ github.head_ref }} - name: Prepare env script run: | diff --git a/ci/jobs/build_clickhouse.py b/ci/jobs/build_clickhouse.py index 3db88938f23..1e6d2c648a7 100644 --- a/ci/jobs/build_clickhouse.py +++ b/ci/jobs/build_clickhouse.py @@ -1,6 +1,5 @@ import argparse -from praktika.param import get_param from praktika.result import Result from praktika.settings import Settings from praktika.utils import MetaClasses, Shell, Utils @@ -16,8 +15,7 @@ def parse_args(): parser = argparse.ArgumentParser(description="ClickHouse Build Job") parser.add_argument( "--build-type", - help="Type: __", - default=None, + help="Type: ,,", ) parser.add_argument( "--param", @@ -30,7 +28,7 @@ def parse_args(): CMAKE_CMD = """cmake --debug-trycompile -DCMAKE_VERBOSE_MAKEFILE=1 -LA \ -DCMAKE_BUILD_TYPE={BUILD_TYPE} \ -DSANITIZE={SANITIZER} \ --DENABLE_CHECK_HEAVY_BUILDS=1 -DENABLE_CLICKHOUSE_SELF_EXTRACTING=1 -DENABLE_TESTS=0 \ +-DENABLE_CHECK_HEAVY_BUILDS=1 -DENABLE_CLICKHOUSE_SELF_EXTRACTING=1 \ -DENABLE_UTILS=0 -DCMAKE_FIND_PACKAGE_NO_PACKAGE_REGISTRY=ON -DCMAKE_INSTALL_PREFIX=/usr \ -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_INSTALL_LOCALSTATEDIR=/var -DCMAKE_SKIP_INSTALL_ALL_DEPENDENCY=ON \ {AUX_DEFS} \ @@ -54,33 +52,26 @@ def main(): stages.pop(0) stages.insert(0, stage) - cmake_build_type = "Release" - sanitizer = "" - - if args.build_type and get_param(): - assert ( - False - ), "Build type must provided via job parameter (CI case) or via --build-type input argument not both" - - build_type = args.build_type or get_param() + build_type = args.build_type assert ( build_type ), "build_type must be provided either as input argument or as a parameter of parametrized job in CI" build_type = build_type.lower() - # if Environment.is_local_run(): - # build_cache_type = "disabled" - # else: CACHE_TYPE = "sccache" if "debug" in build_type: print("Build type set: debug") BUILD_TYPE = "Debug" - AUX_DEFS = " -DSPLIT_DEBUG_SYMBOLS=ON -DBUILD_STANDALONE_KEEPER=1 " + AUX_DEFS = ( + " -DENABLE_TESTS=1 -DSPLIT_DEBUG_SYMBOLS=ON -DBUILD_STANDALONE_KEEPER=1 " + ) elif "release" in build_type: print("Build type set: release") - BUILD_TYPE = "None" - AUX_DEFS = " -DENABLE_TESTS=1 " + BUILD_TYPE = "RelWithDebInfo" + AUX_DEFS = " -DENABLE_TESTS=0 " + else: + assert False if "asan" in build_type: print("Sanitizer set: address") @@ -136,6 +127,7 @@ def main(): Shell.check(f"ls -l {build_dir}/programs/") res = results[-1].is_ok() + Result.create_from(results=results, stopwatch=stop_watch).complete_job() diff --git a/ci/jobs/fast_test.py b/ci/jobs/fast_test.py index cb7d925fead..03a4c0cd496 100644 --- a/ci/jobs/fast_test.py +++ b/ci/jobs/fast_test.py @@ -215,11 +215,13 @@ def main(): ) if res and JobStages.TEST in stages: + stop_watch_ = Utils.Stopwatch() step_name = "Tests" print(step_name) res = res and CH.run_fast_test() if res: results.append(FTResultsProcessor(wd=Settings.OUTPUT_DIR).run()) + results[-1].set_timing(stopwatch=stop_watch_) CH.terminate() diff --git a/ci/jobs/functional_stateless_tests.py b/ci/jobs/functional_stateless_tests.py index d77522ed73a..0481086d80a 100644 --- a/ci/jobs/functional_stateless_tests.py +++ b/ci/jobs/functional_stateless_tests.py @@ -1,15 +1,13 @@ import argparse -import os +import time from pathlib import Path -from praktika.param import get_param from praktika.result import Result from praktika.settings import Settings from praktika.utils import MetaClasses, Shell, Utils from ci.jobs.scripts.clickhouse_proc import ClickHouseProc from ci.jobs.scripts.functional_tests_results import FTResultsProcessor -from ci.settings.definitions import azure_secret class JobStages(metaclass=MetaClasses.WithIter): @@ -21,9 +19,14 @@ class JobStages(metaclass=MetaClasses.WithIter): def parse_args(): parser = argparse.ArgumentParser(description="ClickHouse Build Job") parser.add_argument( - "BUILD_TYPE", help="Type: __" + "--ch-path", help="Path to clickhouse binary", default=f"{Settings.INPUT_DIR}" ) - parser.add_argument("--param", help="Optional custom job start stage", default=None) + parser.add_argument( + "--test-options", + help="Comma separated option(s): parallel|non-parallel|BATCH_NUM/BTATCH_TOT|..", + default="", + ) + parser.add_argument("--param", help="Optional job start stage", default=None) return parser.parse_args() @@ -50,28 +53,31 @@ def run_stateless_test( def main(): args = parse_args() - params = get_param().split(" ") - parallel_or_sequential = None - no_parallel = False - no_sequential = False - if params: - parallel_or_sequential = params[0] - if len(params) > 1: - batch_num, total_batches = map(int, params[1].split("/")) - else: - batch_num, total_batches = 0, 0 - if parallel_or_sequential: - no_parallel = parallel_or_sequential == "non-parallel" - no_sequential = parallel_or_sequential == "parallel" + test_options = args.test_options.split(",") + no_parallel = "non-parallel" in test_options + no_sequential = "parallel" in test_options + batch_num, total_batches = 0, 0 + for to in test_options: + if "/" in to: + batch_num, total_batches = map(int, to.split("/")) - os.environ["AZURE_CONNECTION_STRING"] = Shell.get_output( - f"aws ssm get-parameter --region us-east-1 --name azure_connection_string --with-decryption --output text --query Parameter.Value", - verbose=True, - ) + # os.environ["AZURE_CONNECTION_STRING"] = Shell.get_output( + # f"aws ssm get-parameter --region us-east-1 --name azure_connection_string --with-decryption --output text --query Parameter.Value", + # verbose=True, + # strict=True + # ) + + ch_path = args.ch_path + assert Path( + ch_path + "/clickhouse" + ).is_file(), f"clickhouse binary not found under [{ch_path}]" stop_watch = Utils.Stopwatch() stages = list(JobStages) + + logs_to_attach = [] + stage = args.param or JobStages.INSTALL_CLICKHOUSE if stage: assert stage in JobStages, f"--param must be one of [{list(JobStages)}]" @@ -83,19 +89,22 @@ def main(): res = True results = [] - Utils.add_to_PATH(f"{Settings.INPUT_DIR}:tests") + Utils.add_to_PATH(f"{ch_path}:tests") if res and JobStages.INSTALL_CLICKHOUSE in stages: commands = [ - f"chmod +x {Settings.INPUT_DIR}/clickhouse", - f"ln -sf {Settings.INPUT_DIR}/clickhouse {Settings.INPUT_DIR}/clickhouse-server", - f"ln -sf {Settings.INPUT_DIR}/clickhouse {Settings.INPUT_DIR}/clickhouse-client", + f"chmod +x {ch_path}/clickhouse", + f"ln -sf {ch_path}/clickhouse {ch_path}/clickhouse-server", + f"ln -sf {ch_path}/clickhouse {ch_path}/clickhouse-client", f"rm -rf {Settings.TEMP_DIR}/etc/ && mkdir -p {Settings.TEMP_DIR}/etc/clickhouse-client {Settings.TEMP_DIR}/etc/clickhouse-server", f"cp programs/server/config.xml programs/server/users.xml {Settings.TEMP_DIR}/etc/clickhouse-server/", f"./tests/config/install.sh {Settings.TEMP_DIR}/etc/clickhouse-server {Settings.TEMP_DIR}/etc/clickhouse-client --s3-storage", # update_path_ch_config, - f"sed -i 's|>/var/|>{Settings.TEMP_DIR}/var/|g; s|>/etc/|>{Settings.TEMP_DIR}/etc/|g' {Settings.TEMP_DIR}/etc/clickhouse-server/config.xml", - f"sed -i 's|>/etc/|>{Settings.TEMP_DIR}/etc/|g' {Settings.TEMP_DIR}/etc/clickhouse-server/config.d/ssl_certs.xml", + # f"sed -i 's|>/var/|>{Settings.TEMP_DIR}/var/|g; s|>/etc/|>{Settings.TEMP_DIR}/etc/|g' {Settings.TEMP_DIR}/etc/clickhouse-server/config.xml", + # f"sed -i 's|>/etc/|>{Settings.TEMP_DIR}/etc/|g' {Settings.TEMP_DIR}/etc/clickhouse-server/config.d/ssl_certs.xml", + f"for file in /tmp/praktika/etc/clickhouse-server/config.d/*.xml; do [ -f $file ] && echo Change config $file && sed -i 's|>/var/log|>{Settings.TEMP_DIR}/var/log|g; s|>/etc/|>{Settings.TEMP_DIR}/etc/|g' $(readlink -f $file); done", + f"for file in /tmp/praktika/etc/clickhouse-server/*.xml; do [ -f $file ] && echo Change config $file && sed -i 's|>/var/log|>{Settings.TEMP_DIR}/var/log|g; s|>/etc/|>{Settings.TEMP_DIR}/etc/|g' $(readlink -f $file); done", + f"for file in /tmp/praktika/etc/clickhouse-server/config.d/*.xml; do [ -f $file ] && echo Change config $file && sed -i 's|local_disk|{Settings.TEMP_DIR}/local_disk|g' $(readlink -f $file); done", f"clickhouse-server --version", ] results.append( @@ -110,22 +119,27 @@ def main(): stop_watch_ = Utils.Stopwatch() step_name = "Start ClickHouse Server" print(step_name) - res = res and CH.start_minio() + minio_log = "/tmp/praktika/output/minio.log" + res = res and CH.start_minio(log_file_path=minio_log) + logs_to_attach += [minio_log] + time.sleep(10) + Shell.check("ps -ef | grep minio", verbose=True) + res = res and Shell.check( + "aws s3 ls s3://test --endpoint-url http://localhost:11111/", verbose=True + ) res = res and CH.start() res = res and CH.wait_ready() + if res: + print("ch started") + logs_to_attach += [ + "/tmp/praktika/var/log/clickhouse-server/clickhouse-server.log", + "/tmp/praktika/var/log/clickhouse-server/clickhouse-server.err.log", + ] results.append( Result.create_from( name=step_name, status=res, stopwatch=stop_watch_, - files=( - [ - "/tmp/praktika/var/log/clickhouse-server/clickhouse-server.log", - "/tmp/praktika/var/log/clickhouse-server/clickhouse-server.err.log", - ] - if not res - else [] - ), ) ) res = results[-1].is_ok() @@ -144,7 +158,9 @@ def main(): results[-1].set_timing(stopwatch=stop_watch_) res = results[-1].is_ok() - Result.create_from(results=results, stopwatch=stop_watch).complete_job() + Result.create_from( + results=results, stopwatch=stop_watch, files=logs_to_attach if not res else [] + ).complete_job() if __name__ == "__main__": diff --git a/ci/jobs/scripts/clickhouse_proc.py b/ci/jobs/scripts/clickhouse_proc.py index cc822eab693..c43283e75e0 100644 --- a/ci/jobs/scripts/clickhouse_proc.py +++ b/ci/jobs/scripts/clickhouse_proc.py @@ -1,5 +1,4 @@ -import threading -import time +import subprocess from pathlib import Path from praktika.settings import Settings @@ -39,39 +38,25 @@ class ClickHouseProc: Utils.set_env("CLICKHOUSE_USER_FILES", self.user_files_path) Utils.set_env("CLICKHOUSE_SCHEMA_FILES", f"{self.ch_config_dir}/format_schemas") - if not fast_test: - with open(f"{self.ch_config_dir}/config.d/backups.xml", "w") as file: - file.write(self.BACKUPS_XML) + # if not fast_test: + # with open(f"{self.ch_config_dir}/config.d/backups.xml", "w") as file: + # file.write(self.BACKUPS_XML) self.minio_proc = None - def start_minio(self): - print("Starting minio") - - def run_minio(): - self.minio_proc = Shell.run_async( - self.minio_cmd, verbose=True, suppress_output=True + def start_minio(self, log_file_path): + command = ["tests/docker_scripts/setup_minio.sh", "stateless", "./tests"] + with open(log_file_path, "w") as log_file: + process = subprocess.Popen( + command, stdout=log_file, stderr=subprocess.STDOUT ) - - thread = threading.Thread(target=run_minio) - thread.daemon = True # Allow program to exit even if thread is still running - thread.start() - time.sleep(5) - return thread.is_alive() + print(f"Started setup_minio.sh asynchronously with PID {process.pid}") + return True def start(self): print("Starting ClickHouse server") Shell.check(f"rm {self.pid_file}") - - def run_clickhouse(): - self.proc = Shell.run_async( - self.command, verbose=True, suppress_output=False - ) - - thread = threading.Thread(target=run_clickhouse) - thread.daemon = True # Allow program to exit even if thread is still running - thread.start() - + self.proc = subprocess.Popen(self.command, stderr=subprocess.STDOUT, shell=True) started = False try: for _ in range(5): diff --git a/ci/praktika/_environment.py b/ci/praktika/_environment.py index 4ac8ad319f9..1c6b547ddde 100644 --- a/ci/praktika/_environment.py +++ b/ci/praktika/_environment.py @@ -30,7 +30,6 @@ class _Environment(MetaClasses.Serializable): INSTANCE_ID: str INSTANCE_LIFE_CYCLE: str LOCAL_RUN: bool = False - PARAMETER: Any = None REPORT_INFO: List[str] = dataclasses.field(default_factory=list) name = "environment" @@ -172,18 +171,15 @@ class _Environment(MetaClasses.Serializable): # TODO: find a better place for the function. This file should not import praktika.settings # as it's requires reading users config, that's why imports nested inside the function - def get_report_url(self): + def get_report_url(self, settings): import urllib - from praktika.settings import Settings - from praktika.utils import Utils - - path = Settings.HTML_S3_PATH - for bucket, endpoint in Settings.S3_BUCKET_TO_HTTP_ENDPOINT.items(): + path = settings.HTML_S3_PATH + for bucket, endpoint in settings.S3_BUCKET_TO_HTTP_ENDPOINT.items(): if bucket in path: path = path.replace(bucket, endpoint) break - REPORT_URL = f"https://{path}/{Path(Settings.HTML_PAGE_FILE).name}?PR={self.PR_NUMBER}&sha={self.SHA}&name_0={urllib.parse.quote(self.WORKFLOW_NAME, safe='')}&name_1={urllib.parse.quote(self.JOB_NAME, safe='')}" + REPORT_URL = f"https://{path}/{Path(settings.HTML_PAGE_FILE).name}?PR={self.PR_NUMBER}&sha={self.SHA}&name_0={urllib.parse.quote(self.WORKFLOW_NAME, safe='')}&name_1={urllib.parse.quote(self.JOB_NAME, safe='')}" return REPORT_URL def is_local_run(self): diff --git a/ci/praktika/cidb.py b/ci/praktika/cidb.py index 087845ec762..53088c102cd 100644 --- a/ci/praktika/cidb.py +++ b/ci/praktika/cidb.py @@ -52,7 +52,7 @@ class CIDB: check_status=result.status, check_duration_ms=int(result.duration * 1000), check_start_time=Utils.timestamp_to_str(result.start_time), - report_url=env.get_report_url(), + report_url=env.get_report_url(settings=Settings), pull_request_url=env.CHANGE_URL, base_ref=env.BASE_BRANCH, base_repo=env.REPOSITORY, diff --git a/ci/praktika/digest.py b/ci/praktika/digest.py index 93b62b13dc0..a1f2eecf9b6 100644 --- a/ci/praktika/digest.py +++ b/ci/praktika/digest.py @@ -31,6 +31,9 @@ class Digest: cache_key = self._hash_digest_config(config) if cache_key in self.digest_cache: + print( + f"calc digest for job [{job_config.name}]: hash_key [{cache_key}] - from cache" + ) return self.digest_cache[cache_key] included_files = Utils.traverse_paths( @@ -38,12 +41,9 @@ class Digest: job_config.digest_config.exclude_paths, sorted=True, ) - print( f"calc digest for job [{job_config.name}]: hash_key [{cache_key}], include [{len(included_files)}] files" ) - # Sort files to ensure consistent hash calculation - included_files.sort() # Calculate MD5 hash res = "" @@ -52,11 +52,11 @@ class Digest: print(f"NOTE: empty digest config [{config}] - return dummy digest") else: hash_md5 = hashlib.md5() - for file_path in included_files: - res = self._calc_file_digest(file_path, hash_md5) - assert res - self.digest_cache[cache_key] = res - return res + for i, file_path in enumerate(included_files): + hash_md5 = self._calc_file_digest(file_path, hash_md5) + digest = hash_md5.hexdigest()[: Settings.CACHE_DIGEST_LEN] + self.digest_cache[cache_key] = digest + return digest def calc_docker_digest( self, @@ -103,10 +103,10 @@ class Digest: print( f"WARNING: No valid file resolved by link {file_path} -> {resolved_path} - skipping digest calculation" ) - return hash_md5.hexdigest()[: Settings.CACHE_DIGEST_LEN] + return hash_md5 with open(resolved_path, "rb") as f: for chunk in iter(lambda: f.read(4096), b""): hash_md5.update(chunk) - return hash_md5.hexdigest()[: Settings.CACHE_DIGEST_LEN] + return hash_md5 diff --git a/ci/praktika/environment.py b/ci/praktika/environment.py deleted file mode 100644 index 8f53aa6230b..00000000000 --- a/ci/praktika/environment.py +++ /dev/null @@ -1,3 +0,0 @@ -from praktika._environment import _Environment - -Environment = _Environment.get() diff --git a/ci/praktika/hook_html.py b/ci/praktika/hook_html.py index cea84192d0d..ca2692d1b22 100644 --- a/ci/praktika/hook_html.py +++ b/ci/praktika/hook_html.py @@ -1,6 +1,5 @@ import dataclasses import json -import urllib.parse from pathlib import Path from typing import List @@ -132,17 +131,9 @@ class HtmlRunnerHooks: result = Result.generate_skipped(job.name) results.append(result) summary_result = Result.generate_pending(_workflow.name, results=results) - summary_result.aux_links.append(env.CHANGE_URL) - summary_result.aux_links.append(env.RUN_URL) + summary_result.links.append(env.CHANGE_URL) + summary_result.links.append(env.RUN_URL) summary_result.start_time = Utils.timestamp() - page_url = "/".join( - ["https:/", Settings.HTML_S3_PATH, str(Path(Settings.HTML_PAGE_FILE).name)] - ) - for bucket, endpoint in Settings.S3_BUCKET_TO_HTTP_ENDPOINT.items(): - page_url = page_url.replace(bucket, endpoint) - # TODO: add support for non-PRs (use branch?) - page_url += f"?PR={env.PR_NUMBER}&sha=latest&name_0={urllib.parse.quote(env.WORKFLOW_NAME, safe='')}" - summary_result.html_link = page_url # clean the previous latest results in PR if any if env.PR_NUMBER: @@ -152,13 +143,14 @@ class HtmlRunnerHooks: unlock=False, ) + page_url = env.get_report_url(settings=Settings) print(f"CI Status page url [{page_url}]") res1 = GH.post_commit_status( name=_workflow.name, status=Result.Status.PENDING, description="", - url=page_url, + url=env.get_report_url(settings=Settings), ) res2 = GH.post_pr_comment( comment_body=f"Workflow [[{_workflow.name}]({page_url})], commit [{_Environment.get().SHA[:8]}]", @@ -248,11 +240,11 @@ class HtmlRunnerHooks: ) if workflow_result.status != old_status: print( - f"Update GH commit status [{result.name}]: [{old_status} -> {workflow_result.status}], link [{workflow_result.html_link}]" + f"Update GH commit status [{result.name}]: [{old_status} -> {workflow_result.status}]" ) GH.post_commit_status( name=workflow_result.name, status=GH.convert_to_gh_status(workflow_result.status), description="", - url=workflow_result.html_link, + url=env.get_report_url(settings=Settings), ) diff --git a/ci/praktika/job.py b/ci/praktika/job.py index 99eb08938b8..595a86456e9 100644 --- a/ci/praktika/job.py +++ b/ci/praktika/job.py @@ -89,6 +89,7 @@ class Job: ), "Job.Config.provides must be empty for parametrized jobs" if parameter_: obj.parameter = parameter_ + obj.command = obj.command.format(PARAMETER=parameter_) if runs_on_: obj.runs_on = runs_on_ if timeout_: diff --git a/ci/praktika/json.html b/ci/praktika/json.html index f86a7b27ecb..4e15a67ba76 100644 --- a/ci/praktika/json.html +++ b/ci/praktika/json.html @@ -663,20 +663,20 @@ let targetData = navigatePath(data, nameParams); let nest_level = nameParams.length; + // Add footer links from top-level Result + if (Array.isArray(data.links) && data.links.length > 0) { + data.links.forEach(link => { + const a = document.createElement('a'); + a.href = link; + a.textContent = link.split('/').pop(); + a.target = '_blank'; + footerRight.appendChild(a); + }); + } + if (targetData) { infoElement.style.display = 'none'; - // Handle footer links if present - if (Array.isArray(data.aux_links) && data.aux_links.length > 0) { - data.aux_links.forEach(link => { - const a = document.createElement('a'); - a.href = link; - a.textContent = link.split('/').pop(); - a.target = '_blank'; - footerRight.appendChild(a); - }); - } - addStatusToStatus(targetData.status, targetData.start_time, targetData.duration) // Handle links diff --git a/ci/praktika/mangle.py b/ci/praktika/mangle.py index bca33f9e660..b16d52fbbbf 100644 --- a/ci/praktika/mangle.py +++ b/ci/praktika/mangle.py @@ -14,35 +14,34 @@ def _get_workflows(name=None, file=None): """ res = [] - with ContextManager.cd(): - directory = Path(_Settings.WORKFLOWS_DIRECTORY) - for py_file in directory.glob("*.py"): - if file and file not in str(py_file): - continue - module_name = py_file.name.removeprefix(".py") - spec = importlib.util.spec_from_file_location( - module_name, f"{_Settings.WORKFLOWS_DIRECTORY}/{module_name}" - ) - assert spec - foo = importlib.util.module_from_spec(spec) - assert spec.loader - spec.loader.exec_module(foo) - try: - for workflow in foo.WORKFLOWS: - if name: - if name == workflow.name: - print(f"Read workflow [{name}] config from [{module_name}]") - res = [workflow] - break - else: - continue + directory = Path(_Settings.WORKFLOWS_DIRECTORY) + for py_file in directory.glob("*.py"): + if file and file not in str(py_file): + continue + module_name = py_file.name.removeprefix(".py") + spec = importlib.util.spec_from_file_location( + module_name, f"{_Settings.WORKFLOWS_DIRECTORY}/{module_name}" + ) + assert spec + foo = importlib.util.module_from_spec(spec) + assert spec.loader + spec.loader.exec_module(foo) + try: + for workflow in foo.WORKFLOWS: + if name: + if name == workflow.name: + print(f"Read workflow [{name}] config from [{module_name}]") + res = [workflow] + break else: - res += foo.WORKFLOWS - print(f"Read workflow configs from [{module_name}]") - except Exception as e: - print( - f"WARNING: Failed to add WORKFLOWS config from [{module_name}], exception [{e}]" - ) + continue + else: + res += foo.WORKFLOWS + print(f"Read workflow configs from [{module_name}]") + except Exception as e: + print( + f"WARNING: Failed to add WORKFLOWS config from [{module_name}], exception [{e}]" + ) if not res: Utils.raise_with_error(f"Failed to find workflow [{name or file}]") diff --git a/ci/praktika/native_jobs.py b/ci/praktika/native_jobs.py index 16ffa9056e9..58af211988b 100644 --- a/ci/praktika/native_jobs.py +++ b/ci/praktika/native_jobs.py @@ -342,7 +342,7 @@ def _finish_workflow(workflow, job_name): f"NOTE: Result for [{result.name}] has not ok status [{result.status}]" ) ready_for_merge_status = Result.Status.FAILED - failed_results.append(result.name.split("(", maxsplit=1)[0]) # cut name + failed_results.append(result.name) if failed_results: ready_for_merge_description = f"failed: {', '.join(failed_results)}" @@ -362,9 +362,7 @@ def _finish_workflow(workflow, job_name): unlock=False, ) # no lock - no unlock - Result.from_fs(job_name).set_status(Result.Status.SUCCESS).set_info( - ready_for_merge_description - ) + Result.from_fs(job_name).set_status(Result.Status.SUCCESS) if __name__ == "__main__": diff --git a/ci/praktika/param.py b/ci/praktika/param.py deleted file mode 100644 index f5727198e0d..00000000000 --- a/ci/praktika/param.py +++ /dev/null @@ -1,8 +0,0 @@ -from praktika._environment import _Environment - - -# TODO: find better place and/or right storage for parameter -def get_param(): - env = _Environment.get() - assert env.PARAMETER - return env.PARAMETER diff --git a/ci/praktika/result.py b/ci/praktika/result.py index f473cf3ed05..842deacbcbd 100644 --- a/ci/praktika/result.py +++ b/ci/praktika/result.py @@ -26,10 +26,6 @@ class Result(MetaClasses.Serializable): files (List[str]): A list of file paths or names related to the result. links (List[str]): A list of URLs related to the result (e.g., links to reports or resources). info (str): Additional information about the result. Free-form text. - # TODO: rename - aux_links (List[str]): A list of auxiliary links that provide additional context for the result. - # TODO: remove - html_link (str): A direct link to an HTML representation of the result (e.g., a detailed report page). Inner Class: Status: Defines possible statuses for the task, such as "success", "failure", etc. @@ -51,8 +47,6 @@ class Result(MetaClasses.Serializable): files: List[str] = dataclasses.field(default_factory=list) links: List[str] = dataclasses.field(default_factory=list) info: str = "" - aux_links: List[str] = dataclasses.field(default_factory=list) - html_link: str = "" @staticmethod def create_from( diff --git a/ci/praktika/runner.py b/ci/praktika/runner.py index 5db1a89ce99..1ac8748d1c0 100644 --- a/ci/praktika/runner.py +++ b/ci/praktika/runner.py @@ -80,7 +80,6 @@ class Runner: print("Read GH Environment") env = _Environment.from_env() env.JOB_NAME = job.name - env.PARAMETER = job.parameter env.dump() print(env) @@ -128,7 +127,6 @@ class Runner: # re-set envs for local run env = _Environment.get() env.JOB_NAME = job.name - env.PARAMETER = job.parameter env.dump() if param: @@ -143,6 +141,7 @@ class Runner: job.run_in_docker.split("+")[1:], ) from_root = "root" in docker_settings + settings = [s for s in docker_settings if s.startswith("--")] if ":" in job.run_in_docker: docker_name, docker_tag = job.run_in_docker.split(":") print( @@ -154,9 +153,11 @@ class Runner: RunConfig.from_fs(workflow.name).digest_dockers[job.run_in_docker], ) docker = docker or f"{docker_name}:{docker_tag}" - cmd = f"docker run --rm --name praktika {'--user $(id -u):$(id -g)' if not from_root else ''} -e PYTHONPATH='{Settings.DOCKER_WD}:{Settings.DOCKER_WD}/ci' --volume ./:{Settings.DOCKER_WD} --volume {Settings.TEMP_DIR}:{Settings.TEMP_DIR} --workdir={Settings.DOCKER_WD} {docker} {job.command}" + cmd = f"docker run --rm --name praktika {'--user $(id -u):$(id -g)' if not from_root else ''} -e PYTHONPATH='{Settings.DOCKER_WD}:{Settings.DOCKER_WD}/ci' --volume ./:{Settings.DOCKER_WD} --volume {Settings.TEMP_DIR}:{Settings.TEMP_DIR} --workdir={Settings.DOCKER_WD} {' '.join(settings)} {docker} {job.command}" else: cmd = job.command + python_path = os.getenv("PYTHONPATH", ":") + os.environ["PYTHONPATH"] = f".:{python_path}" if param: print(f"Custom --param [{param}] will be passed to job's script") diff --git a/ci/praktika/utils.py b/ci/praktika/utils.py index b96c78e4fa7..62eb13b3e19 100644 --- a/ci/praktika/utils.py +++ b/ci/praktika/utils.py @@ -81,25 +81,26 @@ class MetaClasses: class ContextManager: @staticmethod @contextmanager - def cd(to: Optional[Union[Path, str]] = None) -> Iterator[None]: + def cd(to: Optional[Union[Path, str]]) -> Iterator[None]: """ changes current working directory to @path or `git root` if @path is None :param to: :return: """ - if not to: - try: - to = Shell.get_output_or_raise("git rev-parse --show-toplevel") - except: - pass - if not to: - if Path(_Settings.DOCKER_WD).is_dir(): - to = _Settings.DOCKER_WD - if not to: - assert False, "FIX IT" - assert to + # if not to: + # try: + # to = Shell.get_output_or_raise("git rev-parse --show-toplevel") + # except: + # pass + # if not to: + # if Path(_Settings.DOCKER_WD).is_dir(): + # to = _Settings.DOCKER_WD + # if not to: + # assert False, "FIX IT" + # assert to old_pwd = os.getcwd() - os.chdir(to) + if to: + os.chdir(to) try: yield finally: diff --git a/ci/praktika/validator.py b/ci/praktika/validator.py index 29edc0a27ed..d612881b819 100644 --- a/ci/praktika/validator.py +++ b/ci/praktika/validator.py @@ -119,61 +119,58 @@ class Validator: def validate_file_paths_in_run_command(cls, workflow: Workflow.Config) -> None: if not Settings.VALIDATE_FILE_PATHS: return - with ContextManager.cd(): - for job in workflow.jobs: - run_command = job.command - command_parts = run_command.split(" ") - for part in command_parts: - if ">" in part: - return - if "/" in part: - assert ( - Path(part).is_file() or Path(part).is_dir() - ), f"Apparently run command [{run_command}] for job [{job}] has invalid path [{part}]. Setting to disable check: VALIDATE_FILE_PATHS" + for job in workflow.jobs: + run_command = job.command + command_parts = run_command.split(" ") + for part in command_parts: + if ">" in part: + return + if "/" in part: + assert ( + Path(part).is_file() or Path(part).is_dir() + ), f"Apparently run command [{run_command}] for job [{job}] has invalid path [{part}]. Setting to disable check: VALIDATE_FILE_PATHS" @classmethod def validate_file_paths_in_digest_configs(cls, workflow: Workflow.Config) -> None: if not Settings.VALIDATE_FILE_PATHS: return - with ContextManager.cd(): - for job in workflow.jobs: - if not job.digest_config: - continue - for include_path in chain( - job.digest_config.include_paths, job.digest_config.exclude_paths - ): - if "*" in include_path: - assert glob.glob( - include_path, recursive=True - ), f"Apparently file glob [{include_path}] in job [{job.name}] digest_config [{job.digest_config}] invalid, workflow [{workflow.name}]. Setting to disable check: VALIDATE_FILE_PATHS" - else: - assert ( - Path(include_path).is_file() or Path(include_path).is_dir() - ), f"Apparently file path [{include_path}] in job [{job.name}] digest_config [{job.digest_config}] invalid, workflow [{workflow.name}]. Setting to disable check: VALIDATE_FILE_PATHS" + for job in workflow.jobs: + if not job.digest_config: + continue + for include_path in chain( + job.digest_config.include_paths, job.digest_config.exclude_paths + ): + if "*" in include_path: + assert glob.glob( + include_path, recursive=True + ), f"Apparently file glob [{include_path}] in job [{job.name}] digest_config [{job.digest_config}] invalid, workflow [{workflow.name}]. Setting to disable check: VALIDATE_FILE_PATHS" + else: + assert ( + Path(include_path).is_file() or Path(include_path).is_dir() + ), f"Apparently file path [{include_path}] in job [{job.name}] digest_config [{job.digest_config}] invalid, workflow [{workflow.name}]. Setting to disable check: VALIDATE_FILE_PATHS" @classmethod def validate_requirements_txt_files(cls, workflow: Workflow.Config) -> None: - with ContextManager.cd(): - for job in workflow.jobs: - if job.job_requirements: - if job.job_requirements.python_requirements_txt: - path = Path(job.job_requirements.python_requirements_txt) - message = f"File with py requirement [{path}] does not exist" - if job.name in ( - Settings.DOCKER_BUILD_JOB_NAME, - Settings.CI_CONFIG_JOB_NAME, - Settings.FINISH_WORKFLOW_JOB_NAME, - ): - message += '\n If all requirements already installed on your runners - add setting INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS""' - message += "\n If requirements needs to be installed - add requirements file (Settings.INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS):" - message += "\n echo jwt==1.3.1 > ./ci/requirements.txt" - message += ( - "\n echo requests==2.32.3 >> ./ci/requirements.txt" - ) - message += "\n echo https://clickhouse-builds.s3.amazonaws.com/packages/praktika-0.1-py3-none-any.whl >> ./ci/requirements.txt" - cls.evaluate_check( - path.is_file(), message, job.name, workflow.name + for job in workflow.jobs: + if job.job_requirements: + if job.job_requirements.python_requirements_txt: + path = Path(job.job_requirements.python_requirements_txt) + message = f"File with py requirement [{path}] does not exist" + if job.name in ( + Settings.DOCKER_BUILD_JOB_NAME, + Settings.CI_CONFIG_JOB_NAME, + Settings.FINISH_WORKFLOW_JOB_NAME, + ): + message += '\n If all requirements already installed on your runners - add setting INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS""' + message += "\n If requirements needs to be installed - add requirements file (Settings.INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS):" + message += "\n echo jwt==1.3.1 > ./ci/requirements.txt" + message += ( + "\n echo requests==2.32.3 >> ./ci/requirements.txt" ) + message += "\n echo https://clickhouse-builds.s3.amazonaws.com/packages/praktika-0.1-py3-none-any.whl >> ./ci/requirements.txt" + cls.evaluate_check( + path.is_file(), message, job.name, workflow.name + ) @classmethod def validate_dockers(cls, workflow: Workflow.Config): diff --git a/ci/praktika/yaml_generator.py b/ci/praktika/yaml_generator.py index 1422a835663..f56715755e8 100644 --- a/ci/praktika/yaml_generator.py +++ b/ci/praktika/yaml_generator.py @@ -81,8 +81,7 @@ jobs: - name: Checkout code uses: actions/checkout@v4 with: - clear-repository: true - ref: ${{{{ github.event.pull_reguest.head.sha }}}} + ref: ${{{{ github.head_ref }}}} {JOB_ADDONS} - name: Prepare env script run: | @@ -191,12 +190,10 @@ jobs: False ), f"Workflow event not yet supported [{workflow_config.event}]" - with ContextManager.cd(): - with open(self._get_workflow_file_name(workflow_config.name), "w") as f: - f.write(yaml_workflow_str) + with open(self._get_workflow_file_name(workflow_config.name), "w") as f: + f.write(yaml_workflow_str) - with ContextManager.cd(): - Shell.check("git add ./.github/workflows/*.yaml") + Shell.check("git add ./.github/workflows/*.yaml") class PullRequestPushYamlGen: diff --git a/ci/settings/definitions.py b/ci/settings/definitions.py index 33173756924..99fec8b5402 100644 --- a/ci/settings/definitions.py +++ b/ci/settings/definitions.py @@ -8,7 +8,7 @@ class RunnerLabels: CI_SERVICES = "ci_services" CI_SERVICES_EBS = "ci_services_ebs" BUILDER = "builder" - STYLE_CHECKER = "style-checker" + FUNC_TESTER_AMD = "func-tester" BASE_BRANCH = "master" @@ -238,5 +238,4 @@ class JobNames: STYLE_CHECK = "Style Check" FAST_TEST = "Fast test" BUILD = "Build" - BUILD_AMD_DEBUG = "Build (amd, debug)" - STATELESS_TESTS = "Stateless tests (amd, debug)" + STATELESS = "Stateless tests" diff --git a/ci/workflows/pull_request.py b/ci/workflows/pull_request.py index 10dd77a0414..0d505ae27c4 100644 --- a/ci/workflows/pull_request.py +++ b/ci/workflows/pull_request.py @@ -41,8 +41,9 @@ fast_test_job = Job.Config( amd_build_jobs = Job.Config( name=JobNames.BUILD, runs_on=[RunnerLabels.BUILDER], - command="python3 ./ci/jobs/build_clickhouse.py", + command="python3 ./ci/jobs/build_clickhouse.py --build-type {PARAMETER}", run_in_docker="clickhouse/fasttest", + timeout=3600 * 2, digest_config=Job.CacheDigestConfig( include_paths=[ "./src", @@ -55,6 +56,7 @@ amd_build_jobs = Job.Config( "./docker/packager/packager", "./rust", "./tests/ci/version_helper.py", + "./ci/jobs/build_clickhouse.py", ], ), ).parametrize( @@ -62,27 +64,53 @@ amd_build_jobs = Job.Config( provides=[[ArtifactNames.CH_AMD_DEBUG], [ArtifactNames.CH_AMD_RELEASE]], ) -statless_batch_num = 2 -stateless_tests_amd_debug_jobs = Job.Config( - name=JobNames.STATELESS_TESTS, +stateless_tests_jobs = Job.Config( + name=JobNames.STATELESS, runs_on=[RunnerLabels.BUILDER], - command="python3 ./ci/jobs/functional_stateless_tests.py amd_debug", - run_in_docker="clickhouse/stateless-test", + command="python3 ./ci/jobs/functional_stateless_tests.py --test-options {PARAMETER}", + run_in_docker="clickhouse/stateless-test+--security-opt seccomp=unconfined+--volume=/tmp/praktika:/var/lib/clickhouse", digest_config=Job.CacheDigestConfig( include_paths=[ "./ci/jobs/functional_stateless_tests.py", ], ), - requires=[ArtifactNames.CH_AMD_DEBUG], ).parametrize( parameter=[ - f"parallel {i+1}/{statless_batch_num}" for i in range(statless_batch_num) - ] - + ["non-parallel"], - runs_on=[[RunnerLabels.BUILDER] for _ in range(statless_batch_num)] - + [[RunnerLabels.STYLE_CHECKER]], + "amd_debug,parallel", + "amd_debug,non-parallel", + "amd_release,parallel", + "amd_release,non-parallel", + ], + runs_on=[ + [RunnerLabels.BUILDER], + [RunnerLabels.FUNC_TESTER_AMD], + [RunnerLabels.BUILDER], + [RunnerLabels.FUNC_TESTER_AMD], + ], + requires=[ + [ArtifactNames.CH_AMD_DEBUG], + [ArtifactNames.CH_AMD_DEBUG], + [ArtifactNames.CH_AMD_RELEASE], + [ArtifactNames.CH_AMD_RELEASE], + ], ) +# stateless_tests_amd_release_jobs = Job.Config( +# name=JobNames.STATELESS_AMD_RELEASE, +# runs_on=[RunnerLabels.BUILDER], +# command="python3 ./ci/jobs/functional_stateless_tests.py --test-options {PARAMETER}", +# run_in_docker="clickhouse/stateless-test+--security-opt seccomp=unconfined+--volume=/tmp/praktika:/var/lib/clickhouse", +# digest_config=Job.CacheDigestConfig( +# include_paths=[ +# "./ci/jobs/functional_stateless_tests.py", +# ], +# ), +# requires=[ArtifactNames.CH_AMD_RELEASE], +# ).parametrize( +# parameter=["parallel", "non-parallel"], +# runs_on=[[RunnerLabels.BUILDER], [RunnerLabels.FUNC_TESTER_AMD]], +# ) + workflow = Workflow.Config( name="PR", event=Workflow.Event.PULL_REQUEST, @@ -91,7 +119,7 @@ workflow = Workflow.Config( style_check_job, fast_test_job, *amd_build_jobs, - *stateless_tests_amd_debug_jobs, + *stateless_tests_jobs, ], artifacts=[ Artifact.Config( diff --git a/tests/clickhouse-test b/tests/clickhouse-test index 3396b10814a..a0ec080ed75 100755 --- a/tests/clickhouse-test +++ b/tests/clickhouse-test @@ -2619,14 +2619,14 @@ def run_tests_process(*args, **kwargs): def do_run_tests(jobs, test_suite: TestSuite): - if jobs > 1 and len(test_suite.parallel_tests) > 0: - print( - "Found", - len(test_suite.parallel_tests), - "parallel tests and", - len(test_suite.sequential_tests), - "sequential tests", - ) + print( + "Found", + len(test_suite.parallel_tests), + "parallel tests and", + len(test_suite.sequential_tests), + "sequential tests", + ) + if test_suite.parallel_tests: tests_n = len(test_suite.parallel_tests) jobs = min(jobs, tests_n) @@ -2639,6 +2639,7 @@ def do_run_tests(jobs, test_suite: TestSuite): # It makes it more difficult to detect real flaky tests, # because the distribution and the amount # of failures will be nearly the same for all tests from the group. + # TODO: add shuffle for sequential tests random.shuffle(test_suite.parallel_tests) batch_size = len(test_suite.parallel_tests) // jobs @@ -2684,6 +2685,7 @@ def do_run_tests(jobs, test_suite: TestSuite): if not p.is_alive(): processes.remove(p) + if test_suite.sequential_tests: run_tests_array( ( test_suite.sequential_tests, @@ -2693,17 +2695,7 @@ def do_run_tests(jobs, test_suite: TestSuite): ) ) - return len(test_suite.sequential_tests) + len(test_suite.parallel_tests) - num_tests = len(test_suite.all_tests) - run_tests_array( - ( - test_suite.all_tests, - num_tests, - test_suite, - False, - ) - ) - return num_tests + return len(test_suite.sequential_tests) + len(test_suite.parallel_tests) def is_test_from_dir(suite_dir, case): diff --git a/tests/config/install.sh b/tests/config/install.sh index cdae5741fce..9630977b9c1 100755 --- a/tests/config/install.sh +++ b/tests/config/install.sh @@ -200,7 +200,7 @@ elif [[ "$USE_AZURE_STORAGE_FOR_MERGE_TREE" == "1" ]]; then fi if [[ "$EXPORT_S3_STORAGE_POLICIES" == "1" ]] || [[ "$S3_STORAGE" = "1" ]]; then - ln -sf $SRC_PATH/config.d/azure_storage_conf.xml $DEST_SERVER_PATH/config.d/ + #ln -sf $SRC_PATH/config.d/azure_storage_conf.xml $DEST_SERVER_PATH/config.d/ ln -sf $SRC_PATH/config.d/storage_conf.xml $DEST_SERVER_PATH/config.d/ ln -sf $SRC_PATH/config.d/storage_conf_02944.xml $DEST_SERVER_PATH/config.d/ ln -sf $SRC_PATH/config.d/storage_conf_02963.xml $DEST_SERVER_PATH/config.d/ diff --git a/tests/docker_scripts/setup_minio.sh b/tests/docker_scripts/setup_minio.sh index 837c05a9c5d..88839c39674 100755 --- a/tests/docker_scripts/setup_minio.sh +++ b/tests/docker_scripts/setup_minio.sh @@ -4,8 +4,10 @@ set -euxf -o pipefail export MINIO_ROOT_USER=${MINIO_ROOT_USER:-clickhouse} export MINIO_ROOT_PASSWORD=${MINIO_ROOT_PASSWORD:-clickhouse} +TEST_DIR=${2:-/repo/tests/} if [ -d "$TEMP_DIR" ]; then + TEST_DIR=$(readlink -f $TEST_DIR) cd "$TEMP_DIR" # add / for minio mc in docker PATH="/:.:$PATH" @@ -79,7 +81,7 @@ start_minio() { pwd mkdir -p ./minio_data minio --version - minio server --address ":11111" ./minio_data & + nohup minio server --address ":11111" ./minio_data & wait_for_it lsof -i :11111 sleep 5 @@ -153,7 +155,7 @@ main() { fi start_minio setup_minio "$1" - upload_data "${query_dir}" "${2:-/repo/tests/}" + upload_data "${query_dir}" "$TEST_DIR" setup_aws_credentials } From bd875401115fb8116302f446c2dec27835b5e958 Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 8 Nov 2024 09:45:51 +0000 Subject: [PATCH 080/244] Update tests/queries/0_stateless/03256_invalid_mutation_query.sql MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: János Benjamin Antal --- .../0_stateless/03256_invalid_mutation_query.sql | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/tests/queries/0_stateless/03256_invalid_mutation_query.sql b/tests/queries/0_stateless/03256_invalid_mutation_query.sql index 2c554cabb9e..9b4e8f9a7ea 100644 --- a/tests/queries/0_stateless/03256_invalid_mutation_query.sql +++ b/tests/queries/0_stateless/03256_invalid_mutation_query.sql @@ -3,11 +3,11 @@ DROP TABLE IF EXISTS t2; CREATE TABLE t (x int) ENGINE = MergeTree() ORDER BY (); -DELETE FROM t WHERE y in (SELECT y FROM t); -- { serverError 47 } -DELETE FROM t WHERE x in (SELECT y FROM t); -- { serverError 47 } -DELETE FROM t WHERE x IN (SELECT * FROM t2); -- { serverError 60 } -ALTER TABLE t DELETE WHERE x in (SELECT y FROM t); -- { serverError 47 } -ALTER TABLE t UPDATE x = 1 WHERE x IN (SELECT y FROM t); -- { serverError 47 } +DELETE FROM t WHERE y in (SELECT x FROM t); -- { serverError UNKNOWN_IDENTIFIER } +DELETE FROM t WHERE x in (SELECT y FROM t); -- { serverError UNKNOWN_IDENTIFIER } +DELETE FROM t WHERE x IN (SELECT * FROM t2); -- { serverError UNKNOWN_TABLE } +ALTER TABLE t DELETE WHERE x in (SELECT y FROM t); -- { serverError UNKNOWN_IDENTIFIER } +ALTER TABLE t UPDATE x = 1 WHERE x IN (SELECT y FROM t); -- { serverError UNKNOWN_IDENTIFIER } DELETE FROM t WHERE x IN (SELECT foo FROM bar) SETTINGS validate_mutation_query = 0; From 2d70dd11d27837f2d73fa2b2496ac5d17c1c5a67 Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 8 Nov 2024 09:47:23 +0000 Subject: [PATCH 081/244] Make it work without the new analyzer --- src/Interpreters/MutationsInterpreter.cpp | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/src/Interpreters/MutationsInterpreter.cpp b/src/Interpreters/MutationsInterpreter.cpp index 589791ac871..a35353a6b2a 100644 --- a/src/Interpreters/MutationsInterpreter.cpp +++ b/src/Interpreters/MutationsInterpreter.cpp @@ -1387,9 +1387,17 @@ void MutationsInterpreter::validate() } } + // Make sure the mutation query is valid if (context->getSettingsRef()[Setting::validate_mutation_query]) - // Make sure the mutation query is valid - prepareQueryAffectedQueryTree(commands, source.getStorage(), context); + { + if (context->getSettingsRef()[Setting::allow_experimental_analyzer]) + prepareQueryAffectedQueryTree(commands, source.getStorage(), context); + else + { + ASTPtr select_query = prepareQueryAffectedAST(commands, source.getStorage(), context); + InterpreterSelectQuery(select_query, context, source.getStorage(), metadata_snapshot); + } + } QueryPlan plan; From dd1ca389dbc3b3f5c5f456bc0d070a972acca806 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Fri, 8 Nov 2024 10:45:13 +0000 Subject: [PATCH 082/244] Trying to cast filter column. --- src/Processors/QueryPlan/FilterStep.cpp | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp index 64c46332c34..7613aac618e 100644 --- a/src/Processors/QueryPlan/FilterStep.cpp +++ b/src/Processors/QueryPlan/FilterStep.cpp @@ -5,6 +5,8 @@ #include #include #include +#include +#include #include #include #include @@ -46,7 +48,19 @@ static ActionsAndName splitSingleAndFilter(ActionsDAG & dag, const ActionsDAG::N auto name = filter_node->result_name; auto split_result = dag.split({filter_node}, true); dag = std::move(split_result.second); - split_result.first.getOutputs().emplace(split_result.first.getOutputs().begin(), split_result.split_nodes_mapping[filter_node]); + + const auto * split_filter_node = split_result.split_nodes_mapping[filter_node]; + auto filter_type = removeLowCardinality(split_filter_node->result_type); + if (!filter_type->onlyNull() && !isUInt8(removeNullable(filter_type))) + { + DataTypePtr cast_type = std::make_shared(); + if (filter_type->isNullable()) + cast_type = std::make_shared(std::move(cast_type)); + + split_result.first.addCast(*split_filter_node, cast_type, {}); + } + + split_result.first.getOutputs().emplace(split_result.first.getOutputs().begin(), split_filter_node); return ActionsAndName{std::move(split_result.first), std::move(name)}; } @@ -168,7 +182,7 @@ void FilterStep::describeActions(FormatSettings & settings) const for (auto & and_atom : and_atoms) { auto expression = std::make_shared(std::move(and_atom.dag)); - settings.out << prefix << "AND column: " << and_atom.name; + settings.out << prefix << "AND column: " << and_atom.name << '\n'; expression->describeActions(settings.out, prefix); } From b370fefb3c8e5583904ab5fe6b21e4ebbb7de5ad Mon Sep 17 00:00:00 2001 From: Christoph Wurm Date: Fri, 8 Nov 2024 10:53:30 +0000 Subject: [PATCH 083/244] Fix test 03173_forbid_qualify --- tests/queries/0_stateless/03173_forbid_qualify.reference | 1 - tests/queries/0_stateless/03173_forbid_qualify.sql | 3 +-- 2 files changed, 1 insertion(+), 3 deletions(-) diff --git a/tests/queries/0_stateless/03173_forbid_qualify.reference b/tests/queries/0_stateless/03173_forbid_qualify.reference index c2f595d8c4b..648ff45ff18 100644 --- a/tests/queries/0_stateless/03173_forbid_qualify.reference +++ b/tests/queries/0_stateless/03173_forbid_qualify.reference @@ -1,3 +1,2 @@ 100 49 -100 diff --git a/tests/queries/0_stateless/03173_forbid_qualify.sql b/tests/queries/0_stateless/03173_forbid_qualify.sql index 0a41385c52f..04c65fdab9c 100644 --- a/tests/queries/0_stateless/03173_forbid_qualify.sql +++ b/tests/queries/0_stateless/03173_forbid_qualify.sql @@ -7,5 +7,4 @@ select count() from test_qualify; -- 100 select * from test_qualify qualify row_number() over (order by number) = 50 SETTINGS enable_analyzer = 1; -- 49 select * from test_qualify qualify row_number() over (order by number) = 50 SETTINGS enable_analyzer = 0; -- { serverError NOT_IMPLEMENTED } -delete from test_qualify where number in (select number from test_qualify qualify row_number() over (order by number) = 50); -- { serverError UNFINISHED } -select count() from test_qualify; -- 100 +delete from test_qualify where number in (select number from test_qualify qualify row_number() over (order by number) = 50); -- { serverError NOT_IMPLEMENTED } From 164e3c26677a209bc7990d326869e71eb3be3bef Mon Sep 17 00:00:00 2001 From: kssenii Date: Fri, 8 Nov 2024 11:54:43 +0100 Subject: [PATCH 084/244] Update settings changes history --- src/Core/SettingsChangesHistory.cpp | 1 + 1 file changed, 1 insertion(+) diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index 64964f294bd..efa47302343 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -74,6 +74,7 @@ static std::initializer_list Date: Fri, 8 Nov 2024 10:55:52 +0000 Subject: [PATCH 085/244] Better test fix --- tests/queries/0_stateless/03173_forbid_qualify.reference | 1 + tests/queries/0_stateless/03173_forbid_qualify.sql | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/tests/queries/0_stateless/03173_forbid_qualify.reference b/tests/queries/0_stateless/03173_forbid_qualify.reference index 648ff45ff18..c2f595d8c4b 100644 --- a/tests/queries/0_stateless/03173_forbid_qualify.reference +++ b/tests/queries/0_stateless/03173_forbid_qualify.reference @@ -1,2 +1,3 @@ 100 49 +100 diff --git a/tests/queries/0_stateless/03173_forbid_qualify.sql b/tests/queries/0_stateless/03173_forbid_qualify.sql index 04c65fdab9c..f7b05a1eb7e 100644 --- a/tests/queries/0_stateless/03173_forbid_qualify.sql +++ b/tests/queries/0_stateless/03173_forbid_qualify.sql @@ -7,4 +7,5 @@ select count() from test_qualify; -- 100 select * from test_qualify qualify row_number() over (order by number) = 50 SETTINGS enable_analyzer = 1; -- 49 select * from test_qualify qualify row_number() over (order by number) = 50 SETTINGS enable_analyzer = 0; -- { serverError NOT_IMPLEMENTED } -delete from test_qualify where number in (select number from test_qualify qualify row_number() over (order by number) = 50); -- { serverError NOT_IMPLEMENTED } +delete from test_qualify where number in (select number from test_qualify qualify row_number() over (order by number) = 50) SETTINGS validate_mutation_query = 0; -- { serverError UNFINISHED } +select count() from test_qualify; -- 100 From 96383d42b184df2e05e9d8aa5ee83dbce4105800 Mon Sep 17 00:00:00 2001 From: Pablo Marcos Date: Fri, 8 Nov 2024 11:38:43 +0000 Subject: [PATCH 086/244] Small refactor to ease debugging when something happens on the CI --- src/Interpreters/QueryMetricLog.cpp | 13 ++++++------- .../03203_system_query_metric_log.reference | 6 +++--- .../0_stateless/03203_system_query_metric_log.sh | 6 +++--- 3 files changed, 12 insertions(+), 13 deletions(-) diff --git a/src/Interpreters/QueryMetricLog.cpp b/src/Interpreters/QueryMetricLog.cpp index e784c357b29..4fbe4f9e1b5 100644 --- a/src/Interpreters/QueryMetricLog.cpp +++ b/src/Interpreters/QueryMetricLog.cpp @@ -148,7 +148,7 @@ void QueryMetricLog::startQuery(const String & query_id, TimePoint start_time, U { QueryMetricLogStatus query_status; query_status.interval_milliseconds = interval_milliseconds; - query_status.next_collect_time = start_time + std::chrono::milliseconds(interval_milliseconds); + query_status.next_collect_time = start_time; auto context = getContext(); const auto & process_list = context->getProcessList(); @@ -213,6 +213,7 @@ void QueryMetricLog::finishQuery(const String & query_id, TimePoint finish_time, void QueryMetricLogStatus::scheduleNext(String query_id) { + next_collect_time += std::chrono::milliseconds(interval_milliseconds); const auto now = std::chrono::system_clock::now(); if (next_collect_time > now) { @@ -229,8 +230,9 @@ void QueryMetricLogStatus::scheduleNext(String query_id) std::optional QueryMetricLogStatus::createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next) { - LOG_TRACE(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Schedule next: {}", - query_id, interval_milliseconds, timePointToString(query_info_time), schedule_next); + LOG_TRACE(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Next collection time: {}", + query_id, interval_milliseconds, timePointToString(query_info_time), + schedule_next ? timePointToString(next_collect_time + std::chrono::milliseconds(interval_milliseconds)) : "finished"); if (query_info_time <= last_collect_time) { @@ -276,15 +278,12 @@ std::optional QueryMetricLogStatus::createLogMetricElemen } else { - LOG_TRACE(logger, "Query {} has no profile counters", query_id); + LOG_WARNING(logger, "Query {} has no profile counters", query_id); elem.profile_events = std::vector(ProfileEvents::end()); } if (schedule_next) - { - next_collect_time += std::chrono::milliseconds(interval_milliseconds); scheduleNext(query_id); - } return elem; } diff --git a/tests/queries/0_stateless/03203_system_query_metric_log.reference b/tests/queries/0_stateless/03203_system_query_metric_log.reference index 940b0c4e178..fa8e27a7e90 100644 --- a/tests/queries/0_stateless/03203_system_query_metric_log.reference +++ b/tests/queries/0_stateless/03203_system_query_metric_log.reference @@ -23,8 +23,8 @@ --Interval 123: check that the SleepFunctionCalls, SleepFunctionMilliseconds and ProfileEvent_SleepFunctionElapsedMicroseconds are correct 1 --Check that a query_metric_log_interval=0 disables the collection -0 +1 -Check that a query which execution time is less than query_metric_log_interval is never collected -0 +1 --Check that there is a final event when queries finish -3 +1 diff --git a/tests/queries/0_stateless/03203_system_query_metric_log.sh b/tests/queries/0_stateless/03203_system_query_metric_log.sh index bf94be79d7c..abcd14c8e5d 100755 --- a/tests/queries/0_stateless/03203_system_query_metric_log.sh +++ b/tests/queries/0_stateless/03203_system_query_metric_log.sh @@ -84,17 +84,17 @@ check_log 123 # query_metric_log_interval=0 disables the collection altogether $CLICKHOUSE_CLIENT -m -q """ SELECT '--Check that a query_metric_log_interval=0 disables the collection'; - SELECT count() FROM system.query_metric_log WHERE event_date >= yesterday() AND query_id = '${query_prefix}_0' + SELECT count() == 0 FROM system.query_metric_log WHERE event_date >= yesterday() AND query_id = '${query_prefix}_0' """ # a quick query that takes less than query_metric_log_interval is never collected $CLICKHOUSE_CLIENT -m -q """ SELECT '-Check that a query which execution time is less than query_metric_log_interval is never collected'; - SELECT count() FROM system.query_metric_log WHERE event_date >= yesterday() AND query_id = '${query_prefix}_fast' + SELECT count() == 0 FROM system.query_metric_log WHERE event_date >= yesterday() AND query_id = '${query_prefix}_fast' """ # a query that takes more than query_metric_log_interval is collected including the final row $CLICKHOUSE_CLIENT -m -q """ SELECT '--Check that there is a final event when queries finish'; - SELECT count() FROM system.query_metric_log WHERE event_date >= yesterday() AND query_id = '${query_prefix}_1000' + SELECT count() > 2 FROM system.query_metric_log WHERE event_date >= yesterday() AND query_id = '${query_prefix}_1000' """ From 0dbc041d8bc49d2760fe85a8a76431395571dfb8 Mon Sep 17 00:00:00 2001 From: Pablo Marcos Date: Fri, 8 Nov 2024 12:00:34 +0000 Subject: [PATCH 087/244] Log when the query finishes for system.query_metric_log ASAP There are logs where we can see that after the query finishes, executeQuery takes up to 2 seconds to call finishQuery. --- src/Interpreters/executeQuery.cpp | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/src/Interpreters/executeQuery.cpp b/src/Interpreters/executeQuery.cpp index 9250c069283..4507126b7b3 100644 --- a/src/Interpreters/executeQuery.cpp +++ b/src/Interpreters/executeQuery.cpp @@ -506,6 +506,7 @@ void logQueryFinish( auto time_now = std::chrono::system_clock::now(); QueryStatusInfo info = process_list_elem->getInfo(true, settings[Setting::log_profile_events]); + logQueryMetricLogFinish(context, internal, elem.client_info.current_query_id, time_now, std::make_shared(info)); elem.type = QueryLogElementType::QUERY_FINISH; addStatusInfoToQueryLogElement(elem, info, query_ast, context); @@ -551,6 +552,7 @@ void logQueryFinish( if (auto query_log = context->getQueryLog()) query_log->add(elem); } + if (settings[Setting::log_processors_profiles]) { if (auto processors_profile_log = context->getProcessorsProfileLog()) @@ -598,8 +600,6 @@ void logQueryFinish( } } } - - logQueryMetricLogFinish(context, internal, elem.client_info.current_query_id, time_now, std::make_shared(info)); } if (query_span) @@ -669,6 +669,7 @@ void logQueryException( { elem.query_duration_ms = start_watch.elapsedMilliseconds(); } + logQueryMetricLogFinish(context, internal, elem.client_info.current_query_id, time_now, info); elem.query_cache_usage = QueryCache::Usage::None; @@ -698,8 +699,6 @@ void logQueryException( query_span->addAttribute("clickhouse.exception_code", elem.exception_code); query_span->finish(); } - - logQueryMetricLogFinish(context, internal, elem.client_info.current_query_id, time_now, info); } void logExceptionBeforeStart( @@ -753,6 +752,8 @@ void logExceptionBeforeStart( elem.client_info = context->getClientInfo(); + logQueryMetricLogFinish(context, false, elem.client_info.current_query_id, std::chrono::system_clock::now(), nullptr); + elem.log_comment = settings[Setting::log_comment]; if (elem.log_comment.size() > settings[Setting::max_query_size]) elem.log_comment.resize(settings[Setting::max_query_size]); @@ -797,8 +798,6 @@ void logExceptionBeforeStart( ProfileEvents::increment(ProfileEvents::FailedInsertQuery); } } - - logQueryMetricLogFinish(context, false, elem.client_info.current_query_id, std::chrono::system_clock::now(), nullptr); } void validateAnalyzerSettings(ASTPtr ast, bool context_value) From 6f74b3236bef52beed01aca5007dad13df7a5ae4 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Fri, 8 Nov 2024 12:22:57 +0000 Subject: [PATCH 088/244] Fix some tests. --- src/Core/SettingsChangesHistory.cpp | 1 - src/Processors/QueryPlan/FilterStep.cpp | 4 +- .../01655_plan_optimizations.reference | 5 +-- .../0_stateless/01655_plan_optimizations.sh | 4 +- .../02496_remove_redundant_sorting.reference | 13 +++---- ...rouping_sets_predicate_push_down.reference | 38 ++++++++----------- 6 files changed, 27 insertions(+), 38 deletions(-) diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index dedf8279533..8f01bacf254 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -75,7 +75,6 @@ static std::initializer_listresult_name; auto split_result = dag.split({filter_node}, true); dag = std::move(split_result.second); @@ -57,10 +56,11 @@ static ActionsAndName splitSingleAndFilter(ActionsDAG & dag, const ActionsDAG::N if (filter_type->isNullable()) cast_type = std::make_shared(std::move(cast_type)); - split_result.first.addCast(*split_filter_node, cast_type, {}); + split_filter_node = &split_result.first.addCast(*split_filter_node, cast_type, {}); } split_result.first.getOutputs().emplace(split_result.first.getOutputs().begin(), split_filter_node); + auto name = split_filter_node->result_name; return ActionsAndName{std::move(split_result.first), std::move(name)}; } diff --git a/tests/queries/0_stateless/01655_plan_optimizations.reference b/tests/queries/0_stateless/01655_plan_optimizations.reference index edf93b4b39f..7fc7556e85b 100644 --- a/tests/queries/0_stateless/01655_plan_optimizations.reference +++ b/tests/queries/0_stateless/01655_plan_optimizations.reference @@ -82,12 +82,12 @@ Filter column: notEquals(__table1.y, 0_UInt8) 9 10 > one condition of filter should be pushed down after aggregating, other two conditions are ANDed Filter column -FUNCTION and(minus(s, 8) :: 5, minus(s, 4) :: 2) -> and(notEquals(y, 0), minus(s, 8), minus(s, 4)) +FUNCTION and(minus(s, 8) :: 3, minus(s, 4) :: 5) -> and(notEquals(y, 0), minus(s, 8), minus(s, 4)) Aggregating Filter column: notEquals(y, 0) > (analyzer) one condition of filter should be pushed down after aggregating, other two conditions are ANDed Filter column -FUNCTION and(minus(__table1.s, 8_UInt8) :: 1, minus(__table1.s, 4_UInt8) :: 2) -> and(notEquals(__table1.y, 0_UInt8), minus(__table1.s, 8_UInt8), minus(__table1.s, 4_UInt8)) +FUNCTION and(minus(__table1.s, 8_UInt8) :: 3, minus(__table1.s, 4_UInt8) :: 5) -> and(notEquals(__table1.y, 0_UInt8), minus(__table1.s, 8_UInt8), minus(__table1.s, 4_UInt8)) Aggregating Filter column: notEquals(__table1.y, 0_UInt8) 0 1 @@ -163,7 +163,6 @@ Filter column: notEquals(__table1.y, 2_UInt8) > filter is pushed down before CreatingSets CreatingSets Filter -Filter 1 3 > one condition of filter is pushed down before LEFT JOIN diff --git a/tests/queries/0_stateless/01655_plan_optimizations.sh b/tests/queries/0_stateless/01655_plan_optimizations.sh index 42cdac8c01f..04ab9bbd11c 100755 --- a/tests/queries/0_stateless/01655_plan_optimizations.sh +++ b/tests/queries/0_stateless/01655_plan_optimizations.sh @@ -89,14 +89,14 @@ $CLICKHOUSE_CLIENT --enable_analyzer=0 --convert_query_to_cnf=0 -q " select sum(x) as s, y from (select number as x, number + 1 as y from numbers(10)) group by y ) where y != 0 and s - 8 and s - 4 settings enable_optimize_predicate_expression=0" | - grep -o "Aggregating\|Filter column\|Filter column: notEquals(y, 0)\|FUNCTION and(minus(s, 8) :: 5, minus(s, 4) :: 2) -> and(notEquals(y, 0), minus(s, 8), minus(s, 4))" + grep -o "Aggregating\|Filter column\|Filter column: notEquals(y, 0)\|FUNCTION and(minus(s, 8) :: 3, minus(s, 4) :: 5) -> and(notEquals(y, 0), minus(s, 8), minus(s, 4))" echo "> (analyzer) one condition of filter should be pushed down after aggregating, other two conditions are ANDed" $CLICKHOUSE_CLIENT --enable_analyzer=1 --convert_query_to_cnf=0 -q " explain actions = 1 select s, y from ( select sum(x) as s, y from (select number as x, number + 1 as y from numbers(10)) group by y ) where y != 0 and s - 8 and s - 4 settings enable_optimize_predicate_expression=0" | - grep -o "Aggregating\|Filter column\|Filter column: notEquals(__table1.y, 0_UInt8)\|FUNCTION and(minus(__table1.s, 8_UInt8) :: 1, minus(__table1.s, 4_UInt8) :: 2) -> and(notEquals(__table1.y, 0_UInt8), minus(__table1.s, 8_UInt8), minus(__table1.s, 4_UInt8))" + grep -o "Aggregating\|Filter column\|Filter column: notEquals(__table1.y, 0_UInt8)\|FUNCTION and(minus(__table1.s, 8_UInt8) :: 3, minus(__table1.s, 4_UInt8) :: 5) -> and(notEquals(__table1.y, 0_UInt8), minus(__table1.s, 8_UInt8), minus(__table1.s, 4_UInt8))" $CLICKHOUSE_CLIENT -q " select s, y from ( select sum(x) as s, y from (select number as x, number + 1 as y from numbers(10)) group by y diff --git a/tests/queries/0_stateless/02496_remove_redundant_sorting.reference b/tests/queries/0_stateless/02496_remove_redundant_sorting.reference index 7824fd8cba9..00db41e8ac5 100644 --- a/tests/queries/0_stateless/02496_remove_redundant_sorting.reference +++ b/tests/queries/0_stateless/02496_remove_redundant_sorting.reference @@ -332,13 +332,12 @@ SETTINGS optimize_aggregators_of_group_by_keys=0 -- avoid removing any() as it d Expression (Projection) Sorting (Sorting for ORDER BY) Expression (Before ORDER BY) - Filter ((WHERE + (Projection + Before ORDER BY))) - Filter (HAVING) - Aggregating - Expression ((Before GROUP BY + Projection)) - Sorting (Sorting for ORDER BY) - Expression ((Before ORDER BY + (Projection + Before ORDER BY))) - ReadFromSystemNumbers + Filter (((WHERE + (Projection + Before ORDER BY)) + HAVING)) + Aggregating + Expression ((Before GROUP BY + Projection)) + Sorting (Sorting for ORDER BY) + Expression ((Before ORDER BY + (Projection + Before ORDER BY))) + ReadFromSystemNumbers -- execute 1 2 diff --git a/tests/queries/0_stateless/02554_fix_grouping_sets_predicate_push_down.reference b/tests/queries/0_stateless/02554_fix_grouping_sets_predicate_push_down.reference index 9bb0c022752..a382e14ce03 100644 --- a/tests/queries/0_stateless/02554_fix_grouping_sets_predicate_push_down.reference +++ b/tests/queries/0_stateless/02554_fix_grouping_sets_predicate_push_down.reference @@ -28,21 +28,17 @@ WHERE type_1 = \'all\' (Expression) ExpressionTransform × 2 (Filter) - FilterTransform × 2 - (Filter) - FilterTransform × 2 - (Filter) - FilterTransform × 2 - (Aggregating) - ExpressionTransform × 2 - AggregatingTransform × 2 - Copy 1 → 2 - (Expression) - ExpressionTransform - (Expression) - ExpressionTransform - (ReadFromMergeTree) - MergeTreeSelect(pool: ReadPoolInOrder, algorithm: InOrder) 0 → 1 + FilterTransform × 6 + (Aggregating) + ExpressionTransform × 2 + AggregatingTransform × 2 + Copy 1 → 2 + (Expression) + ExpressionTransform + (Expression) + ExpressionTransform + (ReadFromMergeTree) + MergeTreeSelect(pool: ReadPoolInOrder, algorithm: InOrder) 0 → 1 (Expression) ExpressionTransform × 2 (Filter) @@ -68,14 +64,10 @@ ExpressionTransform × 2 ExpressionTransform × 2 AggregatingTransform × 2 Copy 1 → 2 - (Filter) - FilterTransform - (Filter) - FilterTransform - (Expression) - ExpressionTransform - (ReadFromMergeTree) - MergeTreeSelect(pool: ReadPoolInOrder, algorithm: InOrder) 0 → 1 + (Expression) + ExpressionTransform + (ReadFromMergeTree) + MergeTreeSelect(pool: ReadPoolInOrder, algorithm: InOrder) 0 → 1 (Expression) ExpressionTransform × 2 (Aggregating) From a8d07555d4d01e1f261dd5c6c6f003a5581c2339 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 8 Nov 2024 12:31:22 +0000 Subject: [PATCH 089/244] Update 02967_parallel_replicas_joins_and_analyzer EXPLAIN with RIGHT JOIN changed --- ...llel_replicas_joins_and_analyzer.reference | 99 +++++++------------ 1 file changed, 35 insertions(+), 64 deletions(-) diff --git a/tests/queries/0_stateless/02967_parallel_replicas_joins_and_analyzer.reference b/tests/queries/0_stateless/02967_parallel_replicas_joins_and_analyzer.reference index 93003b6cf6d..1269f792e76 100644 --- a/tests/queries/0_stateless/02967_parallel_replicas_joins_and_analyzer.reference +++ b/tests/queries/0_stateless/02967_parallel_replicas_joins_and_analyzer.reference @@ -266,24 +266,13 @@ Expression Join Expression Join - Union - Expression - Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas - Union - Expression - Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas - Union - Expression Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas + ReadFromMemoryStorage + Expression + Expression + ReadFromMergeTree + Expression + ReadFromMemoryStorage -- -- RIGHT JOIN in sub5: sub5 -> WithMergableStage with sub1 as (select x, y from tab1 where x != 2), @@ -317,27 +306,19 @@ select * from sub5 order by x SETTINGS enable_parallel_replicas = 2, max_paralle Expression Sorting Expression - Join - Union - Expression - Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas + Sorting Expression Join - Union - Expression + Expression + ReadFromMemoryStorage + Expression + Join Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas - Union - Expression + Expression + ReadFromMergeTree Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas + Expression + ReadFromMergeTree -- -- Subqueries for IN allowed with sub1 as (select x, y from tab1 where x in (select number from numbers(16) where number != 2)), @@ -722,28 +703,22 @@ sub4 as (select z, a from tab3 where z != 8), sub5 as (select x, l.y, y, z, rr.z, a from sub3 ll any left join sub4 rr on ll.z = rr.z) select * from sub5 SETTINGS enable_parallel_replicas = 2, max_parallel_replicas = 2, parallel_replicas_for_non_replicated_merge_tree = 1, prefer_localhost_replica = 1, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost', enable_analyzer=1; -Expression - Join - Expression - Join - Union +Union + Expression + Join + Expression + Join Expression Expression ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas - Union Expression Expression ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas - Union Expression Expression ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas + Expression + ReadFromRemoteParallelReplicas -- -- RIGHT JOIN in sub5: sub5 -> WithMergableStage with sub1 as (select x, y from tab1 where x != 2), @@ -776,28 +751,24 @@ sub5 as (select z, a, x, y, r.y, ll.z from sub4 rr any right join sub3 ll on ll. select * from sub5 order by x SETTINGS enable_parallel_replicas = 2, max_parallel_replicas = 2, parallel_replicas_for_non_replicated_merge_tree = 1, prefer_localhost_replica = 1, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost', enable_analyzer=1; Expression Sorting - Expression - Join - Union + Union + Expression + Sorting Expression - Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas - Expression - Join - Union + Join Expression Expression ReadFromMergeTree Expression - ReadFromRemoteParallelReplicas - Union - Expression - Expression - ReadFromMergeTree - Expression - ReadFromRemoteParallelReplicas + Join + Expression + Expression + ReadFromMergeTree + Expression + Expression + ReadFromMergeTree + Expression + ReadFromRemoteParallelReplicas -- -- Subqueries for IN allowed with sub1 as (select x, y from tab1 where x in (select number from numbers(16) where number != 2)), From 955f537bd5ef2f4a29717ac4999ce2af47b4c039 Mon Sep 17 00:00:00 2001 From: Pablo Marcos Date: Fri, 8 Nov 2024 12:28:06 +0000 Subject: [PATCH 090/244] Add a new setting query_metric_log_debug to avoid the noise --- src/Core/Settings.cpp | 5 +++++ src/Core/SettingsChangesHistory.cpp | 1 + src/Interpreters/QueryMetricLog.cpp | 18 +++++++++++------- src/Interpreters/QueryMetricLog.h | 3 ++- src/Interpreters/executeQuery.cpp | 3 ++- .../03203_system_query_metric_log.sh | 10 +++++----- 6 files changed, 26 insertions(+), 14 deletions(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index 081e07ca2ce..d07cd7352a1 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -2784,6 +2784,11 @@ If set to any negative value, it will take the value `collect_interval_milliseco To disable the collection of a single query, set `query_metric_log_interval` to 0. Default value: -1 + )", 0) \ + DECLARE(Bool, query_metric_log_debug, false, R"( +Turns on debugging traces for system.query_metric_log + +Default value: false )", 0) \ DECLARE(LogsLevel, send_logs_level, LogsLevel::fatal, R"( Send server text logs with specified minimum level to client. Valid values: 'trace', 'debug', 'information', 'warning', 'error', 'fatal', 'none' diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index ed87fde8b7e..a3e21aa670f 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -75,6 +75,7 @@ static std::initializer_listgetProcessList(); @@ -222,21 +223,24 @@ void QueryMetricLogStatus::scheduleNext(String query_id) } else { - LOG_TRACE(logger, "The next collecting task for query {} should have already run at {}. Scheduling it right now", - query_id, timePointToString(next_collect_time)); + if (debug_traces) + LOG_DEBUG(logger, "The next collecting task for query {} should have already run at {}. Scheduling it right now", + query_id, timePointToString(next_collect_time)); task->schedule(); } } std::optional QueryMetricLogStatus::createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next) { - LOG_TRACE(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Next collection time: {}", - query_id, interval_milliseconds, timePointToString(query_info_time), - schedule_next ? timePointToString(next_collect_time + std::chrono::milliseconds(interval_milliseconds)) : "finished"); + if (debug_traces) + LOG_DEBUG(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Next collection time: {}", + query_id, interval_milliseconds, timePointToString(query_info_time), + schedule_next ? timePointToString(next_collect_time + std::chrono::milliseconds(interval_milliseconds)) : "finished"); if (query_info_time <= last_collect_time) { - LOG_TRACE(logger, "Query {} has a more recent metrics collected. Skipping this one", query_id); + if (debug_traces) + LOG_DEBUG(logger, "Query {} has a more recent metrics collected. Skipping this one", query_id); return {}; } diff --git a/src/Interpreters/QueryMetricLog.h b/src/Interpreters/QueryMetricLog.h index 65764229b0a..5f301b2cd13 100644 --- a/src/Interpreters/QueryMetricLog.h +++ b/src/Interpreters/QueryMetricLog.h @@ -51,6 +51,7 @@ struct QueryMetricLogStatus std::chrono::system_clock::time_point next_collect_time TSA_GUARDED_BY(getMutex()); std::vector last_profile_events TSA_GUARDED_BY(getMutex()) = std::vector(ProfileEvents::end()); BackgroundSchedulePool::TaskHolder task TSA_GUARDED_BY(getMutex()); + bool debug_traces = false; /// We need to be able to move it for the hash map, so we need to add an indirection here. std::unique_ptr mutex = std::make_unique(); @@ -78,7 +79,7 @@ public: void shutdown() final; /// Both startQuery and finishQuery are called from the thread that executes the query. - void startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds); + void startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds, bool debug_traces = false); void finishQuery(const String & query_id, TimePoint finish_time, QueryStatusInfoPtr query_info = nullptr); private: diff --git a/src/Interpreters/executeQuery.cpp b/src/Interpreters/executeQuery.cpp index 4507126b7b3..794d3dab0e6 100644 --- a/src/Interpreters/executeQuery.cpp +++ b/src/Interpreters/executeQuery.cpp @@ -146,6 +146,7 @@ namespace Setting extern const SettingsQueryCacheSystemTableHandling query_cache_system_table_handling; extern const SettingsSeconds query_cache_ttl; extern const SettingsInt64 query_metric_log_interval; + extern const SettingsBool query_metric_log_debug; extern const SettingsOverflowMode read_overflow_mode; extern const SettingsOverflowMode read_overflow_mode_leaf; extern const SettingsOverflowMode result_overflow_mode; @@ -455,7 +456,7 @@ QueryLogElement logQueryStart( { auto interval_milliseconds = getQueryMetricLogInterval(context); if (interval_milliseconds > 0) - query_metric_log->startQuery(elem.client_info.current_query_id, query_start_time, interval_milliseconds); + query_metric_log->startQuery(elem.client_info.current_query_id, query_start_time, interval_milliseconds, settings[Setting::query_metric_log_debug]); } return elem; diff --git a/tests/queries/0_stateless/03203_system_query_metric_log.sh b/tests/queries/0_stateless/03203_system_query_metric_log.sh index abcd14c8e5d..4bc764b777c 100755 --- a/tests/queries/0_stateless/03203_system_query_metric_log.sh +++ b/tests/queries/0_stateless/03203_system_query_metric_log.sh @@ -6,11 +6,11 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) readonly query_prefix=$CLICKHOUSE_DATABASE -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_1000" -q "SELECT sleep(2.5) FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_400" -q "SELECT sleep(2.5) SETTINGS query_metric_log_interval=400 FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_123" -q "SELECT sleep(2.5) SETTINGS query_metric_log_interval=123 FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_0" -q "SELECT sleep(2.5) SETTINGS query_metric_log_interval=0 FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_fast" -q "SELECT sleep(0.1) FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_1000" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_400" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true, query_metric_log_interval=400 FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_123" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true, query_metric_log_interval=123 FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_0" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true, query_metric_log_interval=0 FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_fast" -q "SELECT sleep(0.1) SETTINGS query_metric_log_debug=true FORMAT Null" & wait From fdc18a6a28e6d44efbcd5a5e488e62478b411011 Mon Sep 17 00:00:00 2001 From: kssenii Date: Fri, 8 Nov 2024 15:01:38 +0100 Subject: [PATCH 091/244] Add a separate setting for background download max size --- src/Interpreters/Cache/FileCache.cpp | 17 ++++++++++++++- src/Interpreters/Cache/FileCache.h | 2 ++ src/Interpreters/Cache/FileCacheSettings.cpp | 3 +++ src/Interpreters/Cache/FileCacheSettings.h | 2 ++ src/Interpreters/Cache/FileCache_fwd.h | 1 + src/Interpreters/Cache/FileSegment.cpp | 5 ++++- src/Interpreters/Cache/Metadata.cpp | 22 +++++++++++++++++++- src/Interpreters/Cache/Metadata.h | 6 ++++++ 8 files changed, 55 insertions(+), 3 deletions(-) diff --git a/src/Interpreters/Cache/FileCache.cpp b/src/Interpreters/Cache/FileCache.cpp index 7de3f7af78d..bda91b31692 100644 --- a/src/Interpreters/Cache/FileCache.cpp +++ b/src/Interpreters/Cache/FileCache.cpp @@ -103,7 +103,11 @@ FileCache::FileCache(const std::string & cache_name, const FileCacheSettings & s , keep_current_elements_to_max_ratio(1 - settings.keep_free_space_elements_ratio) , keep_up_free_space_remove_batch(settings.keep_free_space_remove_batch) , log(getLogger("FileCache(" + cache_name + ")")) - , metadata(settings.base_path, settings.background_download_queue_size_limit, settings.background_download_threads, write_cache_per_user_directory) + , metadata(settings.base_path, + settings.background_download_queue_size_limit, + settings.background_download_threads, + settings.background_download_max_file_segment_size, + write_cache_per_user_directory) { if (settings.cache_policy == "LRU") { @@ -1600,6 +1604,17 @@ void FileCache::applySettingsIfPossible(const FileCacheSettings & new_settings, } } + if (new_settings.background_download_max_file_segment_size != actual_settings.background_download_max_file_segment_size) + { + metadata.setBackgroundDownloadMaxFileSegmentSize(new_settings.background_download_max_file_segment_size); + + LOG_INFO(log, "Changed background_download_max_file_segment_size from {} to {}", + actual_settings.background_download_max_file_segment_size, + new_settings.background_download_max_file_segment_size); + + actual_settings.background_download_max_file_segment_size = new_settings.background_download_max_file_segment_size; + } + if (new_settings.max_size != actual_settings.max_size || new_settings.max_elements != actual_settings.max_elements) { diff --git a/src/Interpreters/Cache/FileCache.h b/src/Interpreters/Cache/FileCache.h index 810ed481300..79966e60ad9 100644 --- a/src/Interpreters/Cache/FileCache.h +++ b/src/Interpreters/Cache/FileCache.h @@ -161,6 +161,8 @@ public: size_t getMaxFileSegmentSize() const { return max_file_segment_size; } + size_t getBackgroundDownloadMaxFileSegmentSize() const { return metadata.getBackgroundDownloadMaxFileSegmentSize(); } + bool tryReserve( FileSegment & file_segment, size_t size, diff --git a/src/Interpreters/Cache/FileCacheSettings.cpp b/src/Interpreters/Cache/FileCacheSettings.cpp index e162d6b7551..9cd0ee750de 100644 --- a/src/Interpreters/Cache/FileCacheSettings.cpp +++ b/src/Interpreters/Cache/FileCacheSettings.cpp @@ -62,6 +62,9 @@ void FileCacheSettings::loadImpl(FuncHas has, FuncGetUInt get_uint, FuncGetStrin if (has("background_download_queue_size_limit")) background_download_queue_size_limit = get_uint("background_download_queue_size_limit"); + if (has("background_download_max_file_segment_size")) + background_download_threads = get_uint("background_download_max_file_segment_size"); + if (has("load_metadata_threads")) load_metadata_threads = get_uint("load_metadata_threads"); diff --git a/src/Interpreters/Cache/FileCacheSettings.h b/src/Interpreters/Cache/FileCacheSettings.h index 72a2b6c3369..9cf72a2bdff 100644 --- a/src/Interpreters/Cache/FileCacheSettings.h +++ b/src/Interpreters/Cache/FileCacheSettings.h @@ -43,6 +43,8 @@ struct FileCacheSettings double keep_free_space_elements_ratio = FILECACHE_DEFAULT_FREE_SPACE_ELEMENTS_RATIO; size_t keep_free_space_remove_batch = FILECACHE_DEFAULT_FREE_SPACE_REMOVE_BATCH; + size_t background_download_max_file_segment_size = FILECACHE_DEFAULT_MAX_FILE_SEGMENT_SIZE_WITH_BACKGROUND_DOWLOAD; + void loadFromConfig(const Poco::Util::AbstractConfiguration & config, const std::string & config_prefix); void loadFromCollection(const NamedCollection & collection); diff --git a/src/Interpreters/Cache/FileCache_fwd.h b/src/Interpreters/Cache/FileCache_fwd.h index da75f30f0e8..3d461abd065 100644 --- a/src/Interpreters/Cache/FileCache_fwd.h +++ b/src/Interpreters/Cache/FileCache_fwd.h @@ -6,6 +6,7 @@ namespace DB static constexpr int FILECACHE_DEFAULT_MAX_FILE_SEGMENT_SIZE = 32 * 1024 * 1024; /// 32Mi static constexpr int FILECACHE_DEFAULT_FILE_SEGMENT_ALIGNMENT = 4 * 1024 * 1024; /// 4Mi +static constexpr int FILECACHE_DEFAULT_MAX_FILE_SEGMENT_SIZE_WITH_BACKGROUND_DOWLOAD = 4 * 1024 * 1024; /// 4Mi static constexpr int FILECACHE_DEFAULT_BACKGROUND_DOWNLOAD_THREADS = 5; static constexpr int FILECACHE_DEFAULT_BACKGROUND_DOWNLOAD_QUEUE_SIZE_LIMIT = 5000; static constexpr int FILECACHE_DEFAULT_LOAD_METADATA_THREADS = 16; diff --git a/src/Interpreters/Cache/FileSegment.cpp b/src/Interpreters/Cache/FileSegment.cpp index 541f0f5607a..2455461435b 100644 --- a/src/Interpreters/Cache/FileSegment.cpp +++ b/src/Interpreters/Cache/FileSegment.cpp @@ -708,7 +708,10 @@ void FileSegment::complete(bool allow_background_download) if (is_last_holder) { bool added_to_download_queue = false; - if (allow_background_download && background_download_enabled && remote_file_reader) + if (allow_background_download + && background_download_enabled + && remote_file_reader + && downloaded_size < cache->getBackgroundDownloadMaxFileSegmentSize()) { ProfileEvents::increment(ProfileEvents::FilesystemCacheBackgroundDownloadQueuePush); added_to_download_queue = locked_key->addToDownloadQueue(offset(), segment_lock); /// Finish download in background. diff --git a/src/Interpreters/Cache/Metadata.cpp b/src/Interpreters/Cache/Metadata.cpp index 231545212cd..257401d2ce6 100644 --- a/src/Interpreters/Cache/Metadata.cpp +++ b/src/Interpreters/Cache/Metadata.cpp @@ -168,6 +168,7 @@ CacheMetadata::CacheMetadata( const std::string & path_, size_t background_download_queue_size_limit_, size_t background_download_threads_, + size_t background_download_max_file_segment_size_, bool write_cache_per_user_directory_) : path(path_) , cleanup_queue(std::make_shared()) @@ -175,6 +176,7 @@ CacheMetadata::CacheMetadata( , write_cache_per_user_directory(write_cache_per_user_directory_) , log(getLogger("CacheMetadata")) , download_threads_num(background_download_threads_) + , download_max_file_segment_size(background_download_max_file_segment_size_) { } @@ -630,6 +632,9 @@ void CacheMetadata::downloadThreadFunc(const bool & stop_flag) auto & file_segment = holder->front(); + if (file_segment.getDownloadedSize() >= download_max_file_segment_size) + continue; + if (file_segment.getOrSetDownloader() != FileSegment::getCallerId()) continue; @@ -701,10 +706,25 @@ void CacheMetadata::downloadImpl(FileSegment & file_segment, std::optional(reader->getPosition())) reader->seek(offset, SEEK_SET); - while (!reader->eof()) + bool stop = false; + const size_t max_file_segment_size = download_max_file_segment_size.load(); + + while (!stop && !reader->eof()) { auto size = reader->available(); + const size_t downloaded_size = file_segment.getDownloadedSize(); + if (downloaded_size >= max_file_segment_size) + break; + + if (download_max_file_segment_size + size > max_file_segment_size) + { + /// Do not download more than download_max_file_segment_size + /// because we want to leave right boundary of file segment aligned. + size = max_file_segment_size - downloaded_size; + stop = true; + } + std::string failure_reason; if (!file_segment.reserve(size, reserve_space_lock_wait_timeout_milliseconds, failure_reason)) { diff --git a/src/Interpreters/Cache/Metadata.h b/src/Interpreters/Cache/Metadata.h index 0e85ead3265..526b82c9a68 100644 --- a/src/Interpreters/Cache/Metadata.h +++ b/src/Interpreters/Cache/Metadata.h @@ -165,6 +165,7 @@ public: const std::string & path_, size_t background_download_queue_size_limit_, size_t background_download_threads_, + size_t background_download_max_file_segment_size_, bool write_cache_per_user_directory_); void startup(); @@ -210,6 +211,10 @@ public: bool setBackgroundDownloadThreads(size_t threads_num); size_t getBackgroundDownloadThreads() const { return download_threads.size(); } + + void setBackgroundDownloadMaxFileSegmentSize(size_t max_file_segment_size) { download_max_file_segment_size = max_file_segment_size; } + size_t getBackgroundDownloadMaxFileSegmentSize() const { return download_max_file_segment_size; } + bool setBackgroundDownloadQueueSizeLimit(size_t size); bool isBackgroundDownloadEnabled(); @@ -241,6 +246,7 @@ private: }; std::atomic download_threads_num; + std::atomic download_max_file_segment_size; std::vector> download_threads; std::unique_ptr cleanup_thread; From 1561a0115fa740c746ccb054552de3ad751e12ae Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 8 Nov 2024 14:30:02 +0000 Subject: [PATCH 092/244] Fix test, set min_bytes_to_use_direct_io expicitly --- tests/queries/0_stateless/03254_pr_join_on_dups.sql | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tests/queries/0_stateless/03254_pr_join_on_dups.sql b/tests/queries/0_stateless/03254_pr_join_on_dups.sql index 166910d496f..aca4fc6b6c3 100644 --- a/tests/queries/0_stateless/03254_pr_join_on_dups.sql +++ b/tests/queries/0_stateless/03254_pr_join_on_dups.sql @@ -1,6 +1,8 @@ drop table if exists X sync; drop table if exists Y sync; +set min_bytes_to_use_direct_io = 0; -- min_bytes_to_use_direct_io > 0 is broken and leads to unexpected results, https://github.com/ClickHouse/ClickHouse/issues/65690 + create table X (id Int32, x_a String, x_b Nullable(Int32)) engine ReplicatedMergeTree('/clickhouse/{database}/X', '1') order by tuple(); create table Y (id Int32, y_a String, y_b Nullable(String)) engine ReplicatedMergeTree('/clickhouse/{database}/Y', '1') order by tuple(); @@ -10,7 +12,6 @@ insert into Y (id, y_a) values (1, 'r1'), (1, 'r2'), (2, 'r3'), (3, 'r4'), insert into Y (id, y_a, y_b) values (4, 'r6', 'nr6'), (6, 'r7', 'nr7'), (7, 'r8', 'nr8'), (9, 'r9', 'nr9'); set enable_analyzer = 1, enable_parallel_replicas = 1, max_parallel_replicas = 3, cluster_for_parallel_replicas = 'test_cluster_one_shard_three_replicas_localhost'; -set min_bytes_to_use_direct_io = 0; -- min_bytes_to_use_direct_io > 0 is broken and leads to unexpected results, https://github.com/ClickHouse/ClickHouse/issues/65690 select 'inner'; select X.*, Y.* from X inner join Y on X.id = Y.id order by X.id, X.x_a, X.x_b, Y.id, Y.y_a, Y.y_b; From 00e1c5cf0b3e5ddcfee91e8ba1a9ba0e4607354a Mon Sep 17 00:00:00 2001 From: Kseniia Sumarokova <54203879+kssenii@users.noreply.github.com> Date: Fri, 8 Nov 2024 16:30:54 +0100 Subject: [PATCH 093/244] Update FileCacheSettings.cpp --- src/Interpreters/Cache/FileCacheSettings.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Interpreters/Cache/FileCacheSettings.cpp b/src/Interpreters/Cache/FileCacheSettings.cpp index 9cd0ee750de..8f0c5206211 100644 --- a/src/Interpreters/Cache/FileCacheSettings.cpp +++ b/src/Interpreters/Cache/FileCacheSettings.cpp @@ -63,7 +63,7 @@ void FileCacheSettings::loadImpl(FuncHas has, FuncGetUInt get_uint, FuncGetStrin background_download_queue_size_limit = get_uint("background_download_queue_size_limit"); if (has("background_download_max_file_segment_size")) - background_download_threads = get_uint("background_download_max_file_segment_size"); + background_download_max_file_segment_size = get_uint("background_download_max_file_segment_size"); if (has("load_metadata_threads")) load_metadata_threads = get_uint("load_metadata_threads"); From e18ff6e56bda4916064f3135c6b2e59436606235 Mon Sep 17 00:00:00 2001 From: Kseniia Sumarokova <54203879+kssenii@users.noreply.github.com> Date: Fri, 8 Nov 2024 16:31:58 +0100 Subject: [PATCH 094/244] Update Metadata.cpp --- src/Interpreters/Cache/Metadata.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Interpreters/Cache/Metadata.cpp b/src/Interpreters/Cache/Metadata.cpp index 257401d2ce6..2ef8f76aca0 100644 --- a/src/Interpreters/Cache/Metadata.cpp +++ b/src/Interpreters/Cache/Metadata.cpp @@ -717,7 +717,7 @@ void CacheMetadata::downloadImpl(FileSegment & file_segment, std::optional= max_file_segment_size) break; - if (download_max_file_segment_size + size > max_file_segment_size) + if (downloaded_size + size > max_file_segment_size) { /// Do not download more than download_max_file_segment_size /// because we want to leave right boundary of file segment aligned. From 37c24838693e573428414016a619fa70de61823a Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Fri, 8 Nov 2024 17:09:23 +0000 Subject: [PATCH 095/244] Do not randomize min_bytes_to_use_direct_io --- tests/clickhouse-test | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/tests/clickhouse-test b/tests/clickhouse-test index 9c035b7cc35..a1ffcc2030f 100755 --- a/tests/clickhouse-test +++ b/tests/clickhouse-test @@ -821,9 +821,10 @@ class SettingsRandomizer: "optimize_aggregation_in_order": lambda: random.randint(0, 1), "aggregation_in_order_max_block_bytes": lambda: random.randint(0, 50000000), "use_uncompressed_cache": lambda: random.randint(0, 1), - "min_bytes_to_use_direct_io": threshold_generator( - 0.2, 0.5, 1, 10 * 1024 * 1024 * 1024 - ), + # see https://github.com/ClickHouse/ClickHouse/issues/65690 + # "min_bytes_to_use_direct_io": threshold_generator( + # 0.2, 0.5, 1, 10 * 1024 * 1024 * 1024 + # ), "min_bytes_to_use_mmap_io": threshold_generator( 0.2, 0.5, 1, 10 * 1024 * 1024 * 1024 ), From bf58f468082917f871dc706f8596020d0364b43e Mon Sep 17 00:00:00 2001 From: Amos Bird Date: Sat, 9 Nov 2024 13:04:39 +0800 Subject: [PATCH 096/244] Fix empty tuple ALTER --- src/Functions/FunctionsComparison.h | 3 +++ src/Functions/if.cpp | 3 +++ .../0_stateless/03268_empty_tuple_update.reference | 1 + .../queries/0_stateless/03268_empty_tuple_update.sql | 11 +++++++++++ 4 files changed, 18 insertions(+) create mode 100644 tests/queries/0_stateless/03268_empty_tuple_update.reference create mode 100644 tests/queries/0_stateless/03268_empty_tuple_update.sql diff --git a/src/Functions/FunctionsComparison.h b/src/Functions/FunctionsComparison.h index be0875581a5..9b2328065fc 100644 --- a/src/Functions/FunctionsComparison.h +++ b/src/Functions/FunctionsComparison.h @@ -1033,6 +1033,9 @@ private: size_t tuple_size, size_t input_rows_count) const { + if (0 == tuple_size) + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Comparison of zero-sized tuples is not implemented"); + ColumnsWithTypeAndName less_columns(tuple_size); ColumnsWithTypeAndName equal_columns(tuple_size - 1); ColumnsWithTypeAndName tmp_columns(2); diff --git a/src/Functions/if.cpp b/src/Functions/if.cpp index e03b27b3c39..5e1e7067e86 100644 --- a/src/Functions/if.cpp +++ b/src/Functions/if.cpp @@ -668,6 +668,9 @@ private: temporary_columns[0] = arguments[0]; size_t tuple_size = type1.getElements().size(); + if (tuple_size == 0) + return ColumnTuple::create(input_rows_count); + Columns tuple_columns(tuple_size); for (size_t i = 0; i < tuple_size; ++i) diff --git a/tests/queries/0_stateless/03268_empty_tuple_update.reference b/tests/queries/0_stateless/03268_empty_tuple_update.reference new file mode 100644 index 00000000000..30bc45d7a18 --- /dev/null +++ b/tests/queries/0_stateless/03268_empty_tuple_update.reference @@ -0,0 +1 @@ +() 2 diff --git a/tests/queries/0_stateless/03268_empty_tuple_update.sql b/tests/queries/0_stateless/03268_empty_tuple_update.sql new file mode 100644 index 00000000000..343117719fc --- /dev/null +++ b/tests/queries/0_stateless/03268_empty_tuple_update.sql @@ -0,0 +1,11 @@ +DROP TABLE IF EXISTS t0; + +CREATE TABLE t0 (c0 Tuple(), c1 int) ENGINE = Memory(); + +INSERT INTO t0 VALUES ((), 1); + +ALTER TABLE t0 UPDATE c0 = (), c1 = 2 WHERE EXISTS (SELECT 1); + +SELECT * FROM t0; + +DROP TABLE t0; From a888db338e1c79166a3ff6993b71d2fd17dc8736 Mon Sep 17 00:00:00 2001 From: Pablo Marcos Date: Sat, 9 Nov 2024 08:23:25 +0100 Subject: [PATCH 097/244] Revert "Add a new setting query_metric_log_debug to avoid the noise" This reverts commit 955f537bd5ef2f4a29717ac4999ce2af47b4c039. --- src/Core/Settings.cpp | 5 ----- src/Core/SettingsChangesHistory.cpp | 1 - src/Interpreters/QueryMetricLog.cpp | 18 +++++++----------- src/Interpreters/QueryMetricLog.h | 3 +-- src/Interpreters/executeQuery.cpp | 3 +-- .../03203_system_query_metric_log.sh | 10 +++++----- 6 files changed, 14 insertions(+), 26 deletions(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index 2677bde4d55..6f0109fa300 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -2787,11 +2787,6 @@ If set to any negative value, it will take the value `collect_interval_milliseco To disable the collection of a single query, set `query_metric_log_interval` to 0. Default value: -1 - )", 0) \ - DECLARE(Bool, query_metric_log_debug, false, R"( -Turns on debugging traces for system.query_metric_log - -Default value: false )", 0) \ DECLARE(LogsLevel, send_logs_level, LogsLevel::fatal, R"( Send server text logs with specified minimum level to client. Valid values: 'trace', 'debug', 'information', 'warning', 'error', 'fatal', 'none' diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index 008980aae11..c6223bef2b2 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -76,7 +76,6 @@ static std::initializer_listgetProcessList(); @@ -223,24 +222,21 @@ void QueryMetricLogStatus::scheduleNext(String query_id) } else { - if (debug_traces) - LOG_DEBUG(logger, "The next collecting task for query {} should have already run at {}. Scheduling it right now", - query_id, timePointToString(next_collect_time)); + LOG_TRACE(logger, "The next collecting task for query {} should have already run at {}. Scheduling it right now", + query_id, timePointToString(next_collect_time)); task->schedule(); } } std::optional QueryMetricLogStatus::createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next) { - if (debug_traces) - LOG_DEBUG(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Next collection time: {}", - query_id, interval_milliseconds, timePointToString(query_info_time), - schedule_next ? timePointToString(next_collect_time + std::chrono::milliseconds(interval_milliseconds)) : "finished"); + LOG_TRACE(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Next collection time: {}", + query_id, interval_milliseconds, timePointToString(query_info_time), + schedule_next ? timePointToString(next_collect_time + std::chrono::milliseconds(interval_milliseconds)) : "finished"); if (query_info_time <= last_collect_time) { - if (debug_traces) - LOG_DEBUG(logger, "Query {} has a more recent metrics collected. Skipping this one", query_id); + LOG_TRACE(logger, "Query {} has a more recent metrics collected. Skipping this one", query_id); return {}; } diff --git a/src/Interpreters/QueryMetricLog.h b/src/Interpreters/QueryMetricLog.h index 5f301b2cd13..65764229b0a 100644 --- a/src/Interpreters/QueryMetricLog.h +++ b/src/Interpreters/QueryMetricLog.h @@ -51,7 +51,6 @@ struct QueryMetricLogStatus std::chrono::system_clock::time_point next_collect_time TSA_GUARDED_BY(getMutex()); std::vector last_profile_events TSA_GUARDED_BY(getMutex()) = std::vector(ProfileEvents::end()); BackgroundSchedulePool::TaskHolder task TSA_GUARDED_BY(getMutex()); - bool debug_traces = false; /// We need to be able to move it for the hash map, so we need to add an indirection here. std::unique_ptr mutex = std::make_unique(); @@ -79,7 +78,7 @@ public: void shutdown() final; /// Both startQuery and finishQuery are called from the thread that executes the query. - void startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds, bool debug_traces = false); + void startQuery(const String & query_id, TimePoint start_time, UInt64 interval_milliseconds); void finishQuery(const String & query_id, TimePoint finish_time, QueryStatusInfoPtr query_info = nullptr); private: diff --git a/src/Interpreters/executeQuery.cpp b/src/Interpreters/executeQuery.cpp index 794d3dab0e6..4507126b7b3 100644 --- a/src/Interpreters/executeQuery.cpp +++ b/src/Interpreters/executeQuery.cpp @@ -146,7 +146,6 @@ namespace Setting extern const SettingsQueryCacheSystemTableHandling query_cache_system_table_handling; extern const SettingsSeconds query_cache_ttl; extern const SettingsInt64 query_metric_log_interval; - extern const SettingsBool query_metric_log_debug; extern const SettingsOverflowMode read_overflow_mode; extern const SettingsOverflowMode read_overflow_mode_leaf; extern const SettingsOverflowMode result_overflow_mode; @@ -456,7 +455,7 @@ QueryLogElement logQueryStart( { auto interval_milliseconds = getQueryMetricLogInterval(context); if (interval_milliseconds > 0) - query_metric_log->startQuery(elem.client_info.current_query_id, query_start_time, interval_milliseconds, settings[Setting::query_metric_log_debug]); + query_metric_log->startQuery(elem.client_info.current_query_id, query_start_time, interval_milliseconds); } return elem; diff --git a/tests/queries/0_stateless/03203_system_query_metric_log.sh b/tests/queries/0_stateless/03203_system_query_metric_log.sh index 4bc764b777c..abcd14c8e5d 100755 --- a/tests/queries/0_stateless/03203_system_query_metric_log.sh +++ b/tests/queries/0_stateless/03203_system_query_metric_log.sh @@ -6,11 +6,11 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) readonly query_prefix=$CLICKHOUSE_DATABASE -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_1000" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_400" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true, query_metric_log_interval=400 FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_123" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true, query_metric_log_interval=123 FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_0" -q "SELECT sleep(2.5) SETTINGS query_metric_log_debug=true, query_metric_log_interval=0 FORMAT Null" & -$CLICKHOUSE_CLIENT --query-id="${query_prefix}_fast" -q "SELECT sleep(0.1) SETTINGS query_metric_log_debug=true FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_1000" -q "SELECT sleep(2.5) FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_400" -q "SELECT sleep(2.5) SETTINGS query_metric_log_interval=400 FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_123" -q "SELECT sleep(2.5) SETTINGS query_metric_log_interval=123 FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_0" -q "SELECT sleep(2.5) SETTINGS query_metric_log_interval=0 FORMAT Null" & +$CLICKHOUSE_CLIENT --query-id="${query_prefix}_fast" -q "SELECT sleep(0.1) FORMAT Null" & wait From 516300e733c8b3a116139a0306797f779b818f56 Mon Sep 17 00:00:00 2001 From: Pablo Marcos Date: Sat, 9 Nov 2024 08:28:47 +0100 Subject: [PATCH 098/244] Demote log from warning to debug to avoid failing the test --- src/Interpreters/QueryMetricLog.cpp | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/src/Interpreters/QueryMetricLog.cpp b/src/Interpreters/QueryMetricLog.cpp index 4fbe4f9e1b5..62700f49605 100644 --- a/src/Interpreters/QueryMetricLog.cpp +++ b/src/Interpreters/QueryMetricLog.cpp @@ -107,6 +107,7 @@ void QueryMetricLog::collectMetric(const ProcessList & process_list, String quer const auto query_info = process_list.getQueryInfo(query_id, false, true, false); if (!query_info) { + /// TODO: remove trace before 24.11 release after checking everything is fine on the CI LOG_TRACE(logger, "Query {} is not running anymore, so we couldn't get its QueryStatusInfo", query_id); return; } @@ -118,6 +119,7 @@ void QueryMetricLog::collectMetric(const ProcessList & process_list, String quer if (it == queries.end()) { global_lock.unlock(); + /// TODO: remove trace before 24.11 release after checking everything is fine on the CI LOG_TRACE(logger, "Query {} not found in the list. Finished while this collecting task was running", query_id); return; } @@ -126,6 +128,7 @@ void QueryMetricLog::collectMetric(const ProcessList & process_list, String quer if (!query_status.mutex) { global_lock.unlock(); + /// TODO: remove trace before 24.11 release after checking everything is fine on the CI LOG_TRACE(logger, "Query {} finished while this collecting task was running", query_id); return; } @@ -230,12 +233,14 @@ void QueryMetricLogStatus::scheduleNext(String query_id) std::optional QueryMetricLogStatus::createLogMetricElement(const String & query_id, const QueryStatusInfo & query_info, TimePoint query_info_time, bool schedule_next) { + /// TODO: remove trace before 24.11 release after checking everything is fine on the CI LOG_TRACE(logger, "Collecting query_metric_log for query {} and interval {} ms with QueryStatusInfo from {}. Next collection time: {}", query_id, interval_milliseconds, timePointToString(query_info_time), schedule_next ? timePointToString(next_collect_time + std::chrono::milliseconds(interval_milliseconds)) : "finished"); if (query_info_time <= last_collect_time) { + /// TODO: remove trace before 24.11 release after checking everything is fine on the CI LOG_TRACE(logger, "Query {} has a more recent metrics collected. Skipping this one", query_id); return {}; } @@ -278,7 +283,8 @@ std::optional QueryMetricLogStatus::createLogMetricElemen } else { - LOG_WARNING(logger, "Query {} has no profile counters", query_id); + /// TODO: remove trace before 24.11 release after checking everything is fine on the CI + LOG_DEBUG(logger, "Query {} has no profile counters", query_id); elem.profile_events = std::vector(ProfileEvents::end()); } From c3f42b7bc770e5e8104527011f6bc51d5b8469ff Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 01:25:25 +0100 Subject: [PATCH 099/244] Something --- cmake/linux/default_libs.cmake | 3 +- src/AggregateFunctions/ReservoirSampler.h | 2 +- src/Columns/ColumnUnique.cpp | 1 + src/Columns/ColumnUnique.h | 1 + src/Columns/IColumn.cpp | 1 + src/Common/FieldVisitorConvertToNumber.cpp | 2 +- src/Common/FieldVisitorConvertToNumber.h | 1 + src/DataTypes/DataTypesBinaryEncoding.cpp | 5 + src/DataTypes/DataTypesNumber.cpp | 1 + src/DataTypes/DataTypesNumber.h | 1 + src/Formats/JSONExtractTree.cpp | 4 +- src/Functions/FunctionBinaryArithmetic.h | 1 + src/Functions/FunctionsConversion.cpp | 1 + src/Functions/FunctionsRound.h | 2 +- src/IO/readFloatText.cpp | 9 ++ src/IO/readFloatText.h | 111 ++++++++++++++++-- .../Impl/Parquet/ParquetDataValuesReader.cpp | 2 + .../Impl/Parquet/ParquetLeafColReader.cpp | 1 + 18 files changed, 132 insertions(+), 17 deletions(-) diff --git a/cmake/linux/default_libs.cmake b/cmake/linux/default_libs.cmake index 51620bc9f33..79875e1ed6b 100644 --- a/cmake/linux/default_libs.cmake +++ b/cmake/linux/default_libs.cmake @@ -3,8 +3,7 @@ set (DEFAULT_LIBS "-nodefaultlibs") -# We need builtins from Clang's RT even without libcxx - for ubsan+int128. -# See https://bugs.llvm.org/show_bug.cgi?id=16404 +# We need builtins from Clang execute_process (COMMAND ${CMAKE_CXX_COMPILER} --target=${CMAKE_CXX_COMPILER_TARGET} --print-libgcc-file-name --rtlib=compiler-rt OUTPUT_VARIABLE BUILTINS_LIBRARY diff --git a/src/AggregateFunctions/ReservoirSampler.h b/src/AggregateFunctions/ReservoirSampler.h index 2668e0dc890..870cb429fb7 100644 --- a/src/AggregateFunctions/ReservoirSampler.h +++ b/src/AggregateFunctions/ReservoirSampler.h @@ -276,6 +276,6 @@ private: { if (OnEmpty == ReservoirSamplerOnEmpty::THROW) throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR, "Quantile of empty ReservoirSampler"); - return NanLikeValueConstructor>::getValue(); + return NanLikeValueConstructor>::getValue(); } }; diff --git a/src/Columns/ColumnUnique.cpp b/src/Columns/ColumnUnique.cpp index 54f45204c00..773edbfd590 100644 --- a/src/Columns/ColumnUnique.cpp +++ b/src/Columns/ColumnUnique.cpp @@ -16,6 +16,7 @@ template class ColumnUnique; template class ColumnUnique; template class ColumnUnique; template class ColumnUnique; +template class ColumnUnique; template class ColumnUnique; template class ColumnUnique; template class ColumnUnique; diff --git a/src/Columns/ColumnUnique.h b/src/Columns/ColumnUnique.h index ffa7c311e9e..ce7bbf0766f 100644 --- a/src/Columns/ColumnUnique.h +++ b/src/Columns/ColumnUnique.h @@ -760,6 +760,7 @@ extern template class ColumnUnique; extern template class ColumnUnique; extern template class ColumnUnique; extern template class ColumnUnique; +extern template class ColumnUnique; extern template class ColumnUnique; extern template class ColumnUnique; extern template class ColumnUnique; diff --git a/src/Columns/IColumn.cpp b/src/Columns/IColumn.cpp index c9a0514af4e..4a3886dddb6 100644 --- a/src/Columns/IColumn.cpp +++ b/src/Columns/IColumn.cpp @@ -443,6 +443,7 @@ template class IColumnHelper, ColumnFixedSizeHelper>; template class IColumnHelper, ColumnFixedSizeHelper>; template class IColumnHelper, ColumnFixedSizeHelper>; template class IColumnHelper, ColumnFixedSizeHelper>; +template class IColumnHelper, ColumnFixedSizeHelper>; template class IColumnHelper, ColumnFixedSizeHelper>; template class IColumnHelper, ColumnFixedSizeHelper>; template class IColumnHelper, ColumnFixedSizeHelper>; diff --git a/src/Common/FieldVisitorConvertToNumber.cpp b/src/Common/FieldVisitorConvertToNumber.cpp index 75b3fbfe02a..a5963e3d028 100644 --- a/src/Common/FieldVisitorConvertToNumber.cpp +++ b/src/Common/FieldVisitorConvertToNumber.cpp @@ -1,5 +1,4 @@ #include -#include "base/Decimal.h" namespace DB { @@ -17,6 +16,7 @@ template class FieldVisitorConvertToNumber; template class FieldVisitorConvertToNumber; template class FieldVisitorConvertToNumber; template class FieldVisitorConvertToNumber; +//template class FieldVisitorConvertToNumber; template class FieldVisitorConvertToNumber; template class FieldVisitorConvertToNumber; diff --git a/src/Common/FieldVisitorConvertToNumber.h b/src/Common/FieldVisitorConvertToNumber.h index 638b8805b6a..38d5dc473c4 100644 --- a/src/Common/FieldVisitorConvertToNumber.h +++ b/src/Common/FieldVisitorConvertToNumber.h @@ -129,6 +129,7 @@ extern template class FieldVisitorConvertToNumber; extern template class FieldVisitorConvertToNumber; extern template class FieldVisitorConvertToNumber; extern template class FieldVisitorConvertToNumber; +//extern template class FieldVisitorConvertToNumber; extern template class FieldVisitorConvertToNumber; extern template class FieldVisitorConvertToNumber; diff --git a/src/DataTypes/DataTypesBinaryEncoding.cpp b/src/DataTypes/DataTypesBinaryEncoding.cpp index dc0f2f3f5aa..c3190b462c3 100644 --- a/src/DataTypes/DataTypesBinaryEncoding.cpp +++ b/src/DataTypes/DataTypesBinaryEncoding.cpp @@ -96,6 +96,7 @@ enum class BinaryTypeIndex : uint8_t SimpleAggregateFunction = 0x2E, Nested = 0x2F, JSON = 0x30, + BFloat16 = 0x31, }; /// In future we can introduce more arguments in the JSON data type definition. @@ -151,6 +152,8 @@ BinaryTypeIndex getBinaryTypeIndex(const DataTypePtr & type) return BinaryTypeIndex::Int128; case TypeIndex::Int256: return BinaryTypeIndex::Int256; + case TypeIndex::BFloat16: + return BinaryTypeIndex::BFloat16; case TypeIndex::Float32: return BinaryTypeIndex::Float32; case TypeIndex::Float64: @@ -565,6 +568,8 @@ DataTypePtr decodeDataType(ReadBuffer & buf) return std::make_shared(); case BinaryTypeIndex::Int256: return std::make_shared(); + case BinaryTypeIndex::BFloat16: + return std::make_shared(); case BinaryTypeIndex::Float32: return std::make_shared(); case BinaryTypeIndex::Float64: diff --git a/src/DataTypes/DataTypesNumber.cpp b/src/DataTypes/DataTypesNumber.cpp index 5972cebbca1..4c8918521fe 100644 --- a/src/DataTypes/DataTypesNumber.cpp +++ b/src/DataTypes/DataTypesNumber.cpp @@ -112,6 +112,7 @@ template class DataTypeNumber; template class DataTypeNumber; template class DataTypeNumber; template class DataTypeNumber; +template class DataTypeNumber; template class DataTypeNumber; template class DataTypeNumber; diff --git a/src/DataTypes/DataTypesNumber.h b/src/DataTypes/DataTypesNumber.h index 29899847c4b..a9e77e01b13 100644 --- a/src/DataTypes/DataTypesNumber.h +++ b/src/DataTypes/DataTypesNumber.h @@ -63,6 +63,7 @@ extern template class DataTypeNumber; extern template class DataTypeNumber; extern template class DataTypeNumber; extern template class DataTypeNumber; +extern template class DataTypeNumber; extern template class DataTypeNumber; extern template class DataTypeNumber; diff --git a/src/Formats/JSONExtractTree.cpp b/src/Formats/JSONExtractTree.cpp index ae6051823b7..62905a2e630 100644 --- a/src/Formats/JSONExtractTree.cpp +++ b/src/Formats/JSONExtractTree.cpp @@ -131,7 +131,7 @@ bool tryGetNumericValueFromJSONElement( switch (element.type()) { case ElementType::DOUBLE: - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { /// We permit inaccurate conversion of double to float. /// Example: double 0.1 from JSON is not representable in float. @@ -175,7 +175,7 @@ bool tryGetNumericValueFromJSONElement( return false; auto rb = ReadBufferFromMemory{element.getString()}; - if constexpr (std::is_floating_point_v) + if constexpr (is_floating_point) { if (!tryReadFloatText(value, rb) || !rb.eof()) { diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h index df239b820af..854b40df441 100644 --- a/src/Functions/FunctionBinaryArithmetic.h +++ b/src/Functions/FunctionBinaryArithmetic.h @@ -110,6 +110,7 @@ template constexpr bool IsIntegralOrExtendedOrDecimal = IsDataTypeDecimal; template constexpr bool IsFloatingPoint = false; +template <> inline constexpr bool IsFloatingPoint = true; template <> inline constexpr bool IsFloatingPoint = true; template <> inline constexpr bool IsFloatingPoint = true; diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp index 70ec390b576..1c662dd1d9a 100644 --- a/src/Functions/FunctionsConversion.cpp +++ b/src/Functions/FunctionsConversion.cpp @@ -2930,6 +2930,7 @@ template <> struct FunctionTo { using Type = FunctionToInt32; }; template <> struct FunctionTo { using Type = FunctionToInt64; }; template <> struct FunctionTo { using Type = FunctionToInt128; }; template <> struct FunctionTo { using Type = FunctionToInt256; }; +//template <> struct FunctionTo { using Type = FunctionToBFloat16; }; template <> struct FunctionTo { using Type = FunctionToFloat32; }; template <> struct FunctionTo { using Type = FunctionToFloat64; }; diff --git a/src/Functions/FunctionsRound.h b/src/Functions/FunctionsRound.h index 809905c692e..255eca5b406 100644 --- a/src/Functions/FunctionsRound.h +++ b/src/Functions/FunctionsRound.h @@ -694,7 +694,7 @@ public: if (arguments.size() > 1) { const ColumnWithTypeAndName & scale_column = arguments[1]; - res = Dispatcher::template apply(value_arg.column.get(), scale_column.column.get()); + res = Dispatcher::template apply(value_arg.column.get(), scale_column.column.get()); return true; } res = Dispatcher::template apply(value_arg.column.get()); diff --git a/src/IO/readFloatText.cpp b/src/IO/readFloatText.cpp index 17ccc1b25b7..fb3c86fd7b6 100644 --- a/src/IO/readFloatText.cpp +++ b/src/IO/readFloatText.cpp @@ -47,26 +47,35 @@ void assertNaN(ReadBuffer & buf) } +template void readFloatTextPrecise(BFloat16 &, ReadBuffer &); template void readFloatTextPrecise(Float32 &, ReadBuffer &); template void readFloatTextPrecise(Float64 &, ReadBuffer &); +template bool tryReadFloatTextPrecise(BFloat16 &, ReadBuffer &); template bool tryReadFloatTextPrecise(Float32 &, ReadBuffer &); template bool tryReadFloatTextPrecise(Float64 &, ReadBuffer &); +template void readFloatTextFast(BFloat16 &, ReadBuffer &); template void readFloatTextFast(Float32 &, ReadBuffer &); template void readFloatTextFast(Float64 &, ReadBuffer &); +template bool tryReadFloatTextFast(BFloat16 &, ReadBuffer &); template bool tryReadFloatTextFast(Float32 &, ReadBuffer &); template bool tryReadFloatTextFast(Float64 &, ReadBuffer &); +template void readFloatTextSimple(BFloat16 &, ReadBuffer &); template void readFloatTextSimple(Float32 &, ReadBuffer &); template void readFloatTextSimple(Float64 &, ReadBuffer &); +template bool tryReadFloatTextSimple(BFloat16 &, ReadBuffer &); template bool tryReadFloatTextSimple(Float32 &, ReadBuffer &); template bool tryReadFloatTextSimple(Float64 &, ReadBuffer &); +template void readFloatText(BFloat16 &, ReadBuffer &); template void readFloatText(Float32 &, ReadBuffer &); template void readFloatText(Float64 &, ReadBuffer &); +template bool tryReadFloatText(BFloat16 &, ReadBuffer &); template bool tryReadFloatText(Float32 &, ReadBuffer &); template bool tryReadFloatText(Float64 &, ReadBuffer &); +template bool tryReadFloatTextNoExponent(BFloat16 &, ReadBuffer &); template bool tryReadFloatTextNoExponent(Float32 &, ReadBuffer &); template bool tryReadFloatTextNoExponent(Float64 &, ReadBuffer &); diff --git a/src/IO/readFloatText.h b/src/IO/readFloatText.h index c2fec9d4b0b..a7fd6058dd9 100644 --- a/src/IO/readFloatText.h +++ b/src/IO/readFloatText.h @@ -222,7 +222,6 @@ ReturnType readFloatTextPreciseImpl(T & x, ReadBuffer & buf) break; } - char tmp_buf[MAX_LENGTH]; int num_copied_chars = 0; @@ -597,22 +596,85 @@ ReturnType readFloatTextSimpleImpl(T & x, ReadBuffer & buf) return ReturnType(true); } -template void readFloatTextPrecise(T & x, ReadBuffer & in) { readFloatTextPreciseImpl(x, in); } -template bool tryReadFloatTextPrecise(T & x, ReadBuffer & in) { return readFloatTextPreciseImpl(x, in); } +template void readFloatTextPrecise(T & x, ReadBuffer & in) +{ + if constexpr (std::is_same_v) + { + Float32 tmp; + readFloatTextPreciseImpl(tmp, in); + x = BFloat16(tmp); + } + else + readFloatTextPreciseImpl(x, in); +} + +template bool tryReadFloatTextPrecise(T & x, ReadBuffer & in) +{ + if constexpr (std::is_same_v) + { + Float32 tmp; + bool res = readFloatTextPreciseImpl(tmp, in); + if (res) + x = BFloat16(tmp); + return res; + } + else + return readFloatTextPreciseImpl(x, in); +} template void readFloatTextFast(T & x, ReadBuffer & in) { bool has_fractional; - readFloatTextFastImpl(x, in, has_fractional); + if constexpr (std::is_same_v) + { + Float32 tmp; + readFloatTextFastImpl(tmp, in, has_fractional); + x = BFloat16(tmp); + } + else + readFloatTextFastImpl(x, in, has_fractional); } + template bool tryReadFloatTextFast(T & x, ReadBuffer & in) { bool has_fractional; - return readFloatTextFastImpl(x, in, has_fractional); + if constexpr (std::is_same_v) + { + Float32 tmp; + bool res = readFloatTextFastImpl(tmp, in, has_fractional); + if (res) + x = BFloat16(tmp); + return res; + } + else + return readFloatTextFastImpl(x, in, has_fractional); } -template void readFloatTextSimple(T & x, ReadBuffer & in) { readFloatTextSimpleImpl(x, in); } -template bool tryReadFloatTextSimple(T & x, ReadBuffer & in) { return readFloatTextSimpleImpl(x, in); } +template void readFloatTextSimple(T & x, ReadBuffer & in) +{ + if constexpr (std::is_same_v) + { + Float32 tmp; + readFloatTextSimpleImpl(tmp, in); + x = BFloat16(tmp); + } + else + readFloatTextSimpleImpl(x, in); +} + +template bool tryReadFloatTextSimple(T & x, ReadBuffer & in) +{ + if constexpr (std::is_same_v) + { + Float32 tmp; + bool res = readFloatTextSimpleImpl(tmp, in); + if (res) + x = BFloat16(tmp); + return res; + } + else + return readFloatTextSimpleImpl(x, in); +} /// Implementation that is selected as default. @@ -624,18 +686,47 @@ template bool tryReadFloatText(T & x, ReadBuffer & in) { return try template bool tryReadFloatTextNoExponent(T & x, ReadBuffer & in) { bool has_fractional; - return readFloatTextFastImpl(x, in, has_fractional); + if constexpr (std::is_same_v) + { + Float32 tmp; + bool res = readFloatTextFastImpl(tmp, in, has_fractional); + if (res) + x = BFloat16(tmp); + return res; + + } + else + return readFloatTextFastImpl(x, in, has_fractional); } /// With a @has_fractional flag /// Used for input_format_try_infer_integers template bool tryReadFloatTextExt(T & x, ReadBuffer & in, bool & has_fractional) { - return readFloatTextFastImpl(x, in, has_fractional); + if constexpr (std::is_same_v) + { + Float32 tmp; + bool res = readFloatTextFastImpl(tmp, in, has_fractional); + if (res) + x = BFloat16(tmp); + return res; + } + else + return readFloatTextFastImpl(x, in, has_fractional); } + template bool tryReadFloatTextExtNoExponent(T & x, ReadBuffer & in, bool & has_fractional) { - return readFloatTextFastImpl(x, in, has_fractional); + if constexpr (std::is_same_v) + { + Float32 tmp; + bool res = readFloatTextFastImpl(tmp, in, has_fractional); + if (res) + x = BFloat16(tmp); + return res; + } + else + return readFloatTextFastImpl(x, in, has_fractional); } } diff --git a/src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp b/src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp index b471989076b..4b79be98810 100644 --- a/src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp +++ b/src/Processors/Formats/Impl/Parquet/ParquetDataValuesReader.cpp @@ -580,6 +580,7 @@ template class ParquetPlainValuesReader; template class ParquetPlainValuesReader; template class ParquetPlainValuesReader; template class ParquetPlainValuesReader; +template class ParquetPlainValuesReader; template class ParquetPlainValuesReader; template class ParquetPlainValuesReader; template class ParquetPlainValuesReader>; @@ -602,6 +603,7 @@ template class ParquetRleDictReader; template class ParquetRleDictReader; template class ParquetRleDictReader; template class ParquetRleDictReader; +template class ParquetRleDictReader; template class ParquetRleDictReader; template class ParquetRleDictReader; template class ParquetRleDictReader>; diff --git a/src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.cpp b/src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.cpp index c3c7db510ed..328dd37107e 100644 --- a/src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.cpp +++ b/src/Processors/Formats/Impl/Parquet/ParquetLeafColReader.cpp @@ -644,6 +644,7 @@ template class ParquetLeafColReader; template class ParquetLeafColReader; template class ParquetLeafColReader; template class ParquetLeafColReader; +template class ParquetLeafColReader; template class ParquetLeafColReader; template class ParquetLeafColReader; template class ParquetLeafColReader; From 1da6e1fffa8e5cc40d71fee52d6f2742a59d8f21 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 02:25:29 +0100 Subject: [PATCH 100/244] Conversions --- src/Functions/FunctionsConversion.cpp | 22 ++++++++++++++++++---- 1 file changed, 18 insertions(+), 4 deletions(-) diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp index 1c662dd1d9a..f37dff35862 100644 --- a/src/Functions/FunctionsConversion.cpp +++ b/src/Functions/FunctionsConversion.cpp @@ -7,10 +7,8 @@ #include #include #include -#include #include #include -#include #include #include #include @@ -73,8 +71,10 @@ #include #include + namespace DB { + namespace Setting { extern const SettingsBool cast_ipv4_ipv6_default_on_conversion_error; @@ -1862,6 +1862,11 @@ struct ConvertImpl } } + if constexpr ((std::is_same_v || std::is_same_v) + && !(std::is_same_v || std::is_same_v)) + throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Conversion from {} to {} is not supported", + TypeName, TypeName); + if constexpr (std::is_same_v || std::is_same_v) { @@ -2875,6 +2880,7 @@ struct NameToInt32 { static constexpr auto name = "toInt32"; }; struct NameToInt64 { static constexpr auto name = "toInt64"; }; struct NameToInt128 { static constexpr auto name = "toInt128"; }; struct NameToInt256 { static constexpr auto name = "toInt256"; }; +struct NameToBFloat16 { static constexpr auto name = "toBFloat16"; }; struct NameToFloat32 { static constexpr auto name = "toFloat32"; }; struct NameToFloat64 { static constexpr auto name = "toFloat64"; }; struct NameToUUID { static constexpr auto name = "toUUID"; }; @@ -2893,6 +2899,7 @@ using FunctionToInt32 = FunctionConvert>; using FunctionToInt128 = FunctionConvert>; using FunctionToInt256 = FunctionConvert>; +using FunctionToBFloat16 = FunctionConvert>; using FunctionToFloat32 = FunctionConvert>; using FunctionToFloat64 = FunctionConvert>; @@ -2930,7 +2937,7 @@ template <> struct FunctionTo { using Type = FunctionToInt32; }; template <> struct FunctionTo { using Type = FunctionToInt64; }; template <> struct FunctionTo { using Type = FunctionToInt128; }; template <> struct FunctionTo { using Type = FunctionToInt256; }; -//template <> struct FunctionTo { using Type = FunctionToBFloat16; }; +template <> struct FunctionTo { using Type = FunctionToBFloat16; }; template <> struct FunctionTo { using Type = FunctionToFloat32; }; template <> struct FunctionTo { using Type = FunctionToFloat64; }; @@ -2973,6 +2980,7 @@ struct NameToInt32OrZero { static constexpr auto name = "toInt32OrZero"; }; struct NameToInt64OrZero { static constexpr auto name = "toInt64OrZero"; }; struct NameToInt128OrZero { static constexpr auto name = "toInt128OrZero"; }; struct NameToInt256OrZero { static constexpr auto name = "toInt256OrZero"; }; +struct NameToBFloat16OrZero { static constexpr auto name = "toBFloat16OrZero"; }; struct NameToFloat32OrZero { static constexpr auto name = "toFloat32OrZero"; }; struct NameToFloat64OrZero { static constexpr auto name = "toFloat64OrZero"; }; struct NameToDateOrZero { static constexpr auto name = "toDateOrZero"; }; @@ -2999,6 +3007,7 @@ using FunctionToInt32OrZero = FunctionConvertFromString; using FunctionToInt128OrZero = FunctionConvertFromString; using FunctionToInt256OrZero = FunctionConvertFromString; +using FunctionToBFloat16OrZero = FunctionConvertFromString; using FunctionToFloat32OrZero = FunctionConvertFromString; using FunctionToFloat64OrZero = FunctionConvertFromString; using FunctionToDateOrZero = FunctionConvertFromString; @@ -3025,6 +3034,7 @@ struct NameToInt32OrNull { static constexpr auto name = "toInt32OrNull"; }; struct NameToInt64OrNull { static constexpr auto name = "toInt64OrNull"; }; struct NameToInt128OrNull { static constexpr auto name = "toInt128OrNull"; }; struct NameToInt256OrNull { static constexpr auto name = "toInt256OrNull"; }; +struct NameToBFloat16OrNull { static constexpr auto name = "toBFloat16OrNull"; }; struct NameToFloat32OrNull { static constexpr auto name = "toFloat32OrNull"; }; struct NameToFloat64OrNull { static constexpr auto name = "toFloat64OrNull"; }; struct NameToDateOrNull { static constexpr auto name = "toDateOrNull"; }; @@ -3051,6 +3061,7 @@ using FunctionToInt32OrNull = FunctionConvertFromString; using FunctionToInt128OrNull = FunctionConvertFromString; using FunctionToInt256OrNull = FunctionConvertFromString; +using FunctionToBFloat16OrNull = FunctionConvertFromString; using FunctionToFloat32OrNull = FunctionConvertFromString; using FunctionToFloat64OrNull = FunctionConvertFromString; using FunctionToDateOrNull = FunctionConvertFromString; @@ -5194,7 +5205,7 @@ private: if constexpr (is_any_of) { @@ -5447,6 +5458,7 @@ REGISTER_FUNCTION(Conversion) factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); + factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); @@ -5485,6 +5497,7 @@ REGISTER_FUNCTION(Conversion) factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); + factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); @@ -5513,6 +5526,7 @@ REGISTER_FUNCTION(Conversion) factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); + factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); From e65bb147d553b3fcd5f361366547b2858a122247 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 02:27:53 +0100 Subject: [PATCH 101/244] Style --- src/Functions/exp.cpp | 6 ++++++ src/Functions/log.cpp | 5 +++++ src/Functions/sigmoid.cpp | 6 ++++++ src/Functions/tanh.cpp | 6 ++++++ 4 files changed, 23 insertions(+) diff --git a/src/Functions/exp.cpp b/src/Functions/exp.cpp index 07c9288e8ab..24f1d313831 100644 --- a/src/Functions/exp.cpp +++ b/src/Functions/exp.cpp @@ -3,6 +3,12 @@ namespace DB { + +namespace ErrorCodes +{ +extern const int NOT_IMPLEMENTED; +} + namespace { diff --git a/src/Functions/log.cpp b/src/Functions/log.cpp index beaa8128b2b..49fc509634b 100644 --- a/src/Functions/log.cpp +++ b/src/Functions/log.cpp @@ -4,6 +4,11 @@ namespace DB { +namespace ErrorCodes +{ +extern const int NOT_IMPLEMENTED; +} + namespace { diff --git a/src/Functions/sigmoid.cpp b/src/Functions/sigmoid.cpp index 1179329845d..bb9710a15fe 100644 --- a/src/Functions/sigmoid.cpp +++ b/src/Functions/sigmoid.cpp @@ -3,6 +3,12 @@ namespace DB { + +namespace ErrorCodes +{ +extern const int NOT_IMPLEMENTED; +} + namespace { diff --git a/src/Functions/tanh.cpp b/src/Functions/tanh.cpp index 293318f9bbb..d0e1440485b 100644 --- a/src/Functions/tanh.cpp +++ b/src/Functions/tanh.cpp @@ -3,6 +3,12 @@ namespace DB { + +namespace ErrorCodes +{ +extern const int NOT_IMPLEMENTED; +} + namespace { From b4acc885f35e4cccae818fca477efffbc9332ded Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 02:37:26 +0100 Subject: [PATCH 102/244] Documentation --- docs/en/sql-reference/data-types/float.md | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/docs/en/sql-reference/data-types/float.md b/docs/en/sql-reference/data-types/float.md index 3c789076c1e..7185308bdce 100644 --- a/docs/en/sql-reference/data-types/float.md +++ b/docs/en/sql-reference/data-types/float.md @@ -1,10 +1,10 @@ --- slug: /en/sql-reference/data-types/float sidebar_position: 4 -sidebar_label: Float32, Float64 +sidebar_label: Float32, Float64, BFloat16 --- -# Float32, Float64 +# Float32, Float64, BFloat16 :::note If you need accurate calculations, in particular if you work with financial or business data requiring a high precision, you should consider using [Decimal](../data-types/decimal.md) instead. @@ -117,3 +117,11 @@ SELECT 0 / 0 ``` See the rules for `NaN` sorting in the section [ORDER BY clause](../../sql-reference/statements/select/order-by.md). + +## BFloat16 + +`BFloat16` is a 16-bit floating point data type with 8-bit exponent, sign, and 7-bit mantissa. + +It is useful for machine learning and AI applications. + +ClickHouse supports conversions between `Float32` and `BFloat16`. Most of other operations are not supported. From 6cb083621aece140d08d800620b0e5fe7bdc2da0 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 02:47:59 +0100 Subject: [PATCH 103/244] Documentation --- ci/jobs/scripts/check_style/aspell-ignore/en/aspell-dict.txt | 1 + utils/check-style/aspell-ignore/en/aspell-dict.txt | 1 + 2 files changed, 2 insertions(+) diff --git a/ci/jobs/scripts/check_style/aspell-ignore/en/aspell-dict.txt b/ci/jobs/scripts/check_style/aspell-ignore/en/aspell-dict.txt index e2966898be2..7cae8509b83 100644 --- a/ci/jobs/scripts/check_style/aspell-ignore/en/aspell-dict.txt +++ b/ci/jobs/scripts/check_style/aspell-ignore/en/aspell-dict.txt @@ -3131,3 +3131,4 @@ DistributedCachePoolBehaviourOnLimit SharedJoin ShareSet unacked +BFloat diff --git a/utils/check-style/aspell-ignore/en/aspell-dict.txt b/utils/check-style/aspell-ignore/en/aspell-dict.txt index a08143467cd..9765b45c085 100644 --- a/utils/check-style/aspell-ignore/en/aspell-dict.txt +++ b/utils/check-style/aspell-ignore/en/aspell-dict.txt @@ -3154,3 +3154,4 @@ znode znodes zookeeperSessionUptime zstd +BFloat From bec94da77e8333d64c71b4bf778fbf78d10a8519 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 13:19:08 +0100 Subject: [PATCH 104/244] Progressing --- src/DataTypes/DataTypesDecimal.cpp | 8 +++----- src/DataTypes/DataTypesDecimal.h | 2 -- 2 files changed, 3 insertions(+), 7 deletions(-) diff --git a/src/DataTypes/DataTypesDecimal.cpp b/src/DataTypes/DataTypesDecimal.cpp index fddae052ada..63bd4bf2a59 100644 --- a/src/DataTypes/DataTypesDecimal.cpp +++ b/src/DataTypes/DataTypesDecimal.cpp @@ -262,9 +262,9 @@ FOR_EACH_ARITHMETIC_TYPE(INVOKE); template requires (is_arithmetic_v && IsDataTypeDecimal) -ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & /*value*/, UInt32 /*scale*/, typename ToDataType::FieldType & /*result*/) +ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & value, UInt32 scale, typename ToDataType::FieldType & result) { -/* using FromFieldType = typename FromDataType::FieldType; + using FromFieldType = typename FromDataType::FieldType; using ToFieldType = typename ToDataType::FieldType; using ToNativeType = typename ToFieldType::NativeType; @@ -306,9 +306,7 @@ ReturnType convertToDecimalImpl(const typename FromDataType::FieldType & /*value return ReturnType(convertDecimalsImpl, ToDataType, ReturnType>(static_cast(value), 0, scale, result)); else return ReturnType(convertDecimalsImpl, ToDataType, ReturnType>(static_cast(value), 0, scale, result)); - }*/ - - return ReturnType(); + } } #define DISPATCH(FROM_DATA_TYPE, TO_DATA_TYPE) \ diff --git a/src/DataTypes/DataTypesDecimal.h b/src/DataTypes/DataTypesDecimal.h index e0d49408981..09a25617506 100644 --- a/src/DataTypes/DataTypesDecimal.h +++ b/src/DataTypes/DataTypesDecimal.h @@ -3,9 +3,7 @@ #include #include #include -#include #include -#include #include #include #include From f0dc1330eb9d830161531819432a611a363fdc6b Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 13:53:08 +0100 Subject: [PATCH 105/244] Rounding --- src/Functions/FunctionsRound.h | 42 ++++++++++++++++++++++++++-------- 1 file changed, 33 insertions(+), 9 deletions(-) diff --git a/src/Functions/FunctionsRound.h b/src/Functions/FunctionsRound.h index 255eca5b406..70ad4d17718 100644 --- a/src/Functions/FunctionsRound.h +++ b/src/Functions/FunctionsRound.h @@ -268,6 +268,19 @@ inline double roundWithMode(double x, RoundingMode mode) std::unreachable(); } +inline BFloat16 roundWithMode(BFloat16 x, RoundingMode mode) +{ + switch (mode) + { + case RoundingMode::Round: return BFloat16(nearbyintf(Float32(x))); + case RoundingMode::Floor: return BFloat16(floorf(Float32(x))); + case RoundingMode::Ceil: return BFloat16(ceilf(Float32(x))); + case RoundingMode::Trunc: return BFloat16(truncf(Float32(x))); + } + + std::unreachable(); +} + template class FloatRoundingComputationBase { @@ -289,6 +302,11 @@ public: } }; +template <> +class FloatRoundingComputationBase : public FloatRoundingComputationBase +{ +}; + /** Implementation of low-level round-off functions for floating-point values. */ @@ -688,20 +706,26 @@ public: using Types = std::decay_t; using DataType = typename Types::RightType; - if constexpr ((IsDataTypeNumber || IsDataTypeDecimal) - && !std::is_same_v) + if (arguments.size() > 1) { - if (arguments.size() > 1) + const ColumnWithTypeAndName & scale_column = arguments[1]; + + auto call_scale = [&](const auto & scaleTypes) -> bool { - const ColumnWithTypeAndName & scale_column = arguments[1]; - res = Dispatcher::template apply(value_arg.column.get(), scale_column.column.get()); + using ScaleTypes = std::decay_t; + using ScaleType = typename ScaleTypes::RightType; + + res = Dispatcher::template apply(value_arg.column.get(), scale_column.column.get()); return true; - } - res = Dispatcher::template apply(value_arg.column.get()); + }; + + TypeIndex right_index = scale_column.type->getTypeId(); + if (!callOnBasicType(right_index, call_scale)) + throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Scale argument for rounding functions must have integer type"); return true; } - else - return false; + res = Dispatcher::template apply(value_arg.column.get()); + return true; }; #if !defined(__SSE4_1__) From db98fb4c79252d6305eabc06a749e2082bb1c489 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 14:39:45 +0100 Subject: [PATCH 106/244] Documentation --- src/Functions/FunctionsConversion.cpp | 64 +++++++++++++++++++++++++-- 1 file changed, 61 insertions(+), 3 deletions(-) diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp index f37dff35862..37a4ba30d30 100644 --- a/src/Functions/FunctionsConversion.cpp +++ b/src/Functions/FunctionsConversion.cpp @@ -5458,7 +5458,17 @@ REGISTER_FUNCTION(Conversion) factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); - factory.registerFunction(); + + factory.registerFunction(FunctionDocumentation{.description=R"( +Converts Float32 to BFloat16 with losing the precision. + +Example: +[example:typical] +)", + .examples{ + {"typical", "SELECT toBFloat16(12.3::Float32);", "12.3125"}}, + .categories{"Conversion"}}); + factory.registerFunction(); factory.registerFunction(); @@ -5497,7 +5507,31 @@ REGISTER_FUNCTION(Conversion) factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); - factory.registerFunction(); + + factory.registerFunction(FunctionDocumentation{.description=R"( +Converts String to BFloat16. + +If the string does not represent a floating point value, the function returns zero. + +The function allows a silent loss of precision while converting from the string representation. In that case, it will return the truncated result. + +Example of successful conversion: +[example:typical] + +Examples of not successful conversion: +[example:invalid1] +[example:invalid2] + +Example of a loss of precision: +[example:precision] +)", + .examples{ + {"typical", "SELECT toBFloat16OrZero('12.3');", "12.3125"}}, + {"invalid1", "SELECT toBFloat16OrZero('abc');", "0"}}, + {"invalid2", "SELECT toBFloat16OrZero(' 1');", "0"}}, + {"precision", "SELECT toBFloat16OrZero('12.3456789');", "12.375"}}, + .categories{"Conversion"}}); + factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); @@ -5526,7 +5560,31 @@ REGISTER_FUNCTION(Conversion) factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); - factory.registerFunction(); + + factory.registerFunction(FunctionDocumentation{.description=R"( +Converts String to Nullable(BFloat16). + +If the string does not represent a floating point value, the function returns NULL. + +The function allows a silent loss of precision while converting from the string representation. In that case, it will return the truncated result. + +Example of successful conversion: +[example:typical] + +Examples of not successful conversion: +[example:invalid1] +[example:invalid2] + +Example of a loss of precision: +[example:precision] +)", + .examples{ + {"typical", "SELECT toBFloat16OrNull('12.3');", "12.3125"}}, + {"invalid1", "SELECT toBFloat16OrNull('abc');", "NULL"}}, + {"invalid2", "SELECT toBFloat16OrNull(' 1');", "NULL"}}, + {"precision", "SELECT toBFloat16OrNull('12.3456789');", "12.375"}}, + .categories{"Conversion"}}); + factory.registerFunction(); factory.registerFunction(); factory.registerFunction(); From 1c85a0401fbbddccbd3e310a965ce0eb67079a2b Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 15:14:17 +0100 Subject: [PATCH 107/244] Documentation --- src/Functions/FunctionsConversion.cpp | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp index 37a4ba30d30..7f4ccc338cf 100644 --- a/src/Functions/FunctionsConversion.cpp +++ b/src/Functions/FunctionsConversion.cpp @@ -5526,9 +5526,9 @@ Example of a loss of precision: [example:precision] )", .examples{ - {"typical", "SELECT toBFloat16OrZero('12.3');", "12.3125"}}, - {"invalid1", "SELECT toBFloat16OrZero('abc');", "0"}}, - {"invalid2", "SELECT toBFloat16OrZero(' 1');", "0"}}, + {"typical", "SELECT toBFloat16OrZero('12.3');", "12.3125"}, + {"invalid1", "SELECT toBFloat16OrZero('abc');", "0"}, + {"invalid2", "SELECT toBFloat16OrZero(' 1');", "0"}, {"precision", "SELECT toBFloat16OrZero('12.3456789');", "12.375"}}, .categories{"Conversion"}}); @@ -5579,9 +5579,9 @@ Example of a loss of precision: [example:precision] )", .examples{ - {"typical", "SELECT toBFloat16OrNull('12.3');", "12.3125"}}, - {"invalid1", "SELECT toBFloat16OrNull('abc');", "NULL"}}, - {"invalid2", "SELECT toBFloat16OrNull(' 1');", "NULL"}}, + {"typical", "SELECT toBFloat16OrNull('12.3');", "12.3125"}, + {"invalid1", "SELECT toBFloat16OrNull('abc');", "NULL"}, + {"invalid2", "SELECT toBFloat16OrNull(' 1');", "NULL"}, {"precision", "SELECT toBFloat16OrNull('12.3456789');", "12.375"}}, .categories{"Conversion"}}); From bf8fc60bacbb95e12760b00960115e2a6230c280 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 16:20:44 +0100 Subject: [PATCH 108/244] Arithmetic --- src/Functions/FunctionBinaryArithmetic.h | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/src/Functions/FunctionBinaryArithmetic.h b/src/Functions/FunctionBinaryArithmetic.h index 854b40df441..43140427170 100644 --- a/src/Functions/FunctionBinaryArithmetic.h +++ b/src/Functions/FunctionBinaryArithmetic.h @@ -804,7 +804,7 @@ class FunctionBinaryArithmetic : public IFunction DataTypeFixedString, DataTypeString, DataTypeInterval>; - using Floats = TypeList; + using Floats = TypeList; using ValidTypes = std::conditional_t, @@ -2043,7 +2043,15 @@ ColumnPtr executeStringInteger(const ColumnsWithTypeAndName & arguments, const A using DecimalResultType = typename BinaryOperationTraits::DecimalResultDataType; if constexpr (std::is_same_v) + { return nullptr; + } + else if constexpr ((std::is_same_v || std::is_same_v) + && (sizeof(typename LeftDataType::FieldType) > 8 || sizeof(typename RightDataType::FieldType) > 8)) + { + /// Big integers and BFloat16 are not supported together. + return nullptr; + } else // we can't avoid the else because otherwise the compiler may assume the ResultDataType may be Invalid // and that would produce the compile error. { @@ -2060,7 +2068,7 @@ ColumnPtr executeStringInteger(const ColumnsWithTypeAndName & arguments, const A ColumnPtr left_col = nullptr; ColumnPtr right_col = nullptr; - /// When Decimal op Float32/64, convert both of them into Float64 + /// When Decimal op Float32/64/16, convert both of them into Float64 if constexpr (decimal_with_float) { const auto converted_type = std::make_shared(); @@ -2095,7 +2103,6 @@ ColumnPtr executeStringInteger(const ColumnsWithTypeAndName & arguments, const A /// Here we check if we have `intDiv` or `intDivOrZero` and at least one of the arguments is decimal, because in this case originally we had result as decimal, so we need to convert result into integer after calculations else if constexpr (!decimal_with_float && (is_int_div || is_int_div_or_zero) && (IsDataTypeDecimal || IsDataTypeDecimal)) { - if constexpr (!std::is_same_v) { DataTypePtr type_res; From 62c94a784158274e28cf05136cf4023de47f4f01 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 16:40:24 +0100 Subject: [PATCH 109/244] Maybe better --- cmake/cpu_features.cmake | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/cmake/cpu_features.cmake b/cmake/cpu_features.cmake index 2bb6deb4847..dbc77d835be 100644 --- a/cmake/cpu_features.cmake +++ b/cmake/cpu_features.cmake @@ -85,7 +85,7 @@ elseif (ARCH_AARCH64) # [8] https://developer.arm.com/documentation/102651/a/What-are-dot-product-intructions- # [9] https://developer.arm.com/documentation/dui0801/g/A64-Data-Transfer-Instructions/LDAPR?lang=en # [10] https://github.com/aws/aws-graviton-getting-started/blob/main/README.md - set (COMPILER_FLAGS "${COMPILER_FLAGS} -march=armv8.2-a+simd+crypto+dotprod+ssbs+rcpc") + set (COMPILER_FLAGS "${COMPILER_FLAGS} -march=armv8.2-a+simd+crypto+dotprod+ssbs+rcpc+bf16") endif () # Best-effort check: The build generates and executes intermediate binaries, e.g. protoc and llvm-tablegen. If we build on ARM for ARM From 08e6e598f7c140d0be39a64d933521872716ed2c Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 17:41:37 +0100 Subject: [PATCH 110/244] Better code --- src/Common/findExtreme.h | 2 +- src/DataTypes/IDataType.h | 1 - 2 files changed, 1 insertion(+), 2 deletions(-) diff --git a/src/Common/findExtreme.h b/src/Common/findExtreme.h index c2b31c51e87..68e7360d6e2 100644 --- a/src/Common/findExtreme.h +++ b/src/Common/findExtreme.h @@ -11,7 +11,7 @@ namespace DB { template -concept has_find_extreme_implementation = (is_any_of); +concept has_find_extreme_implementation = (is_any_of); template std::optional findExtremeMin(const T * __restrict ptr, size_t start, size_t end); diff --git a/src/DataTypes/IDataType.h b/src/DataTypes/IDataType.h index 4d64b927d83..1e41d6b2eba 100644 --- a/src/DataTypes/IDataType.h +++ b/src/DataTypes/IDataType.h @@ -606,7 +606,6 @@ template inline constexpr bool IsDataTypeEnum> = tr M(Int16) \ M(Int32) \ M(Int64) \ - M(BFloat16) \ M(Float32) \ M(Float64) From 9baa5911f9183e1652593b5d362545377baeea2a Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Sun, 10 Nov 2024 20:54:59 +0000 Subject: [PATCH 111/244] Debugging stack with PR queries --- src/Planner/findParallelReplicasQuery.cpp | 57 +++++++++++++++++------ 1 file changed, 42 insertions(+), 15 deletions(-) diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index 28e2dd8a0ea..fbe2993b7c6 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -23,6 +23,8 @@ #include #include +#include + namespace DB { namespace Setting @@ -38,12 +40,12 @@ namespace ErrorCodes /// Returns a list of (sub)queries (candidates) which may support parallel replicas. /// The rule is : -/// subquery has only LEFT or ALL INNER JOIN (or none), and left part is MergeTree table or subquery candidate as well. +/// subquery has only LEFT / RIGHT / ALL INNER JOIN (or none), and left / right part is MergeTree table or subquery candidate as well. /// /// Additional checks are required, so we return many candidates. The innermost subquery is on top. -std::stack getSupportingParallelReplicasQuery(const IQueryTreeNode * query_tree_node) +std::vector getSupportingParallelReplicasQuery(const IQueryTreeNode * query_tree_node) { - std::stack res; + std::vector res; while (query_tree_node) { @@ -75,7 +77,7 @@ std::stack getSupportingParallelReplicasQuery(const IQueryTre { const auto & query_node_to_process = query_tree_node->as(); query_tree_node = query_node_to_process.getJoinTree().get(); - res.push(&query_node_to_process); + res.push_back(&query_node_to_process); break; } case QueryTreeNodeType::UNION: @@ -162,14 +164,25 @@ QueryTreeNodePtr replaceTablesWithDummyTables(QueryTreeNodePtr query, const Cont return query->cloneAndReplace(visitor.replacement_map); } +static void dumpStack(const std::vector & stack) +{ + std::ranges::reverse_view rv{stack}; + for (const auto * node : rv) + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "{}\n{}", CityHash_v1_0_2::Hash128to64(node->getTreeHash()), node->dumpTree()); +} + /// Find the best candidate for parallel replicas execution by verifying query plan. -/// If query plan has only Expression, Filter of Join steps, we can execute it fully remotely and check the next query. +/// If query plan has only Expression, Filter or Join steps, we can execute it fully remotely and check the next query. /// Otherwise we can execute current query up to WithMergableStage only. const QueryNode * findQueryForParallelReplicas( - std::stack stack, + std::vector stack, const std::unordered_map & mapping, const Settings & settings) { + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "{}", StackTrace().toString()); + + dumpStack(stack); + struct Frame { const QueryPlan::Node * node = nullptr; @@ -188,14 +201,20 @@ const QueryNode * findQueryForParallelReplicas( while (!stack.empty()) { - const QueryNode * const subquery_node = stack.top(); - stack.pop(); + const QueryNode * const subquery_node = stack.back(); + stack.pop_back(); auto it = mapping.find(subquery_node); /// This should not happen ideally. if (it == mapping.end()) break; + LOG_DEBUG( + getLogger(__PRETTY_FUNCTION__), + "{} : {}", + CityHash_v1_0_2::Hash128to64(it->first->getTreeHash()), + it->second->step->getName()); + std::stack nodes_to_check; nodes_to_check.push({.node = it->second, .inside_join = false}); bool can_distribute_full_node = true; @@ -208,6 +227,8 @@ const QueryNode * findQueryForParallelReplicas( const auto & children = next_node_to_check->children; auto * step = next_node_to_check->step.get(); + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Step {} childrens={}", step->getName(), children.size()); + if (children.empty()) { /// Found a source step. @@ -235,7 +256,7 @@ const QueryNode * findQueryForParallelReplicas( else { const auto * join = typeid_cast(step); - /// We've checked that JOIN is INNER/LEFT in query tree. + /// We've checked that JOIN is INNER/LEFT/RIGHT on query tree level before. /// Don't distribute UNION node. if (!join) return res; @@ -286,7 +307,7 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr return nullptr; /// We don't have any subquery and storage can process parallel replicas by itself. - if (stack.top() == query_tree_node.get()) + if (stack.back() == query_tree_node.get()) return nullptr; /// This is needed to avoid infinite recursion. @@ -309,18 +330,24 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr const auto & mapping = planner.getQueryNodeToPlanStepMapping(); const auto * res = findQueryForParallelReplicas(new_stack, mapping, context->getSettingsRef()); - /// Now, return a query from initial stack. if (res) { + // find query in initial stack while (!new_stack.empty()) { - if (res == new_stack.top()) - return stack.top(); + if (res == new_stack.back()) + { + res = stack.back(); + break; + } - stack.pop(); - new_stack.pop(); + stack.pop_back(); + new_stack.pop_back(); } } + + if (res) + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Choosen query: {}", res->dumpTree()); return res; } From 7877d59ff6e7334cde310b2eec626bc6ba7442fe Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 22:13:22 +0100 Subject: [PATCH 112/244] Manual implementation --- base/base/BFloat16.h | 300 +++++++++++++++++- base/base/DecomposedFloat.h | 2 +- base/base/TypeLists.h | 5 +- src/AggregateFunctions/AggregateFunctionAvg.h | 2 +- .../AggregateFunctionDeltaSum.cpp | 6 +- .../AggregateFunctionDeltaSumTimestamp.cpp | 10 +- .../AggregateFunctionMaxIntersections.cpp | 6 +- .../AggregateFunctionSparkbar.cpp | 12 +- src/AggregateFunctions/AggregateFunctionSum.h | 11 +- src/Core/Types_fwd.h | 2 +- src/Functions/FunctionsRound.h | 2 +- src/Functions/PolygonUtils.h | 4 +- src/Functions/divide.cpp | 2 +- src/IO/WriteHelpers.h | 2 +- 14 files changed, 318 insertions(+), 48 deletions(-) diff --git a/base/base/BFloat16.h b/base/base/BFloat16.h index 99eab5c67cb..9c6196d6aab 100644 --- a/base/base/BFloat16.h +++ b/base/base/BFloat16.h @@ -1,22 +1,294 @@ #pragma once -#include +#include +#include -using BFloat16 = __bf16; +//using BFloat16 = __bf16; + +class BFloat16 +{ +private: + UInt16 x = 0; + +public: + constexpr BFloat16() = default; + constexpr BFloat16(const BFloat16 & other) = default; + constexpr BFloat16 & operator=(const BFloat16 & other) = default; + + explicit constexpr BFloat16(const Float32 & other) + { + x = static_cast(std::bit_cast(other) >> 16); + } + + template + explicit constexpr BFloat16(const T & other) + : BFloat16(Float32(other)) + { + } + + template + constexpr BFloat16 & operator=(const T & other) + { + *this = BFloat16(other); + return *this; + } + + explicit constexpr operator Float32() const + { + return std::bit_cast(static_cast(x) << 16); + } + + template + explicit constexpr operator T() const + { + return T(Float32(*this)); + } + + constexpr bool isFinite() const + { + return (x & 0b0111111110000000) != 0b0111111110000000; + } + + constexpr bool isNaN() const + { + return !isFinite() && (x & 0b0000000001111111) != 0b0000000000000000; + } + + constexpr bool signBit() const + { + return x & 0b1000000000000000; + } + + constexpr bool operator==(const BFloat16 & other) const + { + return x == other.x; + } + + constexpr bool operator!=(const BFloat16 & other) const + { + return x != other.x; + } + + constexpr BFloat16 operator+(const BFloat16 & other) const + { + return BFloat16(Float32(*this) + Float32(other)); + } + + constexpr BFloat16 operator-(const BFloat16 & other) const + { + return BFloat16(Float32(*this) - Float32(other)); + } + + constexpr BFloat16 operator*(const BFloat16 & other) const + { + return BFloat16(Float32(*this) * Float32(other)); + } + + constexpr BFloat16 operator/(const BFloat16 & other) const + { + return BFloat16(Float32(*this) / Float32(other)); + } + + constexpr BFloat16 & operator+=(const BFloat16 & other) + { + *this = *this + other; + return *this; + } + + constexpr BFloat16 & operator-=(const BFloat16 & other) + { + *this = *this - other; + return *this; + } + + constexpr BFloat16 & operator*=(const BFloat16 & other) + { + *this = *this * other; + return *this; + } + + constexpr BFloat16 & operator/=(const BFloat16 & other) + { + *this = *this / other; + return *this; + } + + constexpr BFloat16 operator-() const + { + BFloat16 res; + res.x = x ^ 0b1000000000000000; + return res; + } +}; + + +template +requires(!std::is_same_v) +constexpr bool operator==(const BFloat16 & a, const T & b) +{ + return Float32(a) == b; +} + +template +requires(!std::is_same_v) +constexpr bool operator==(const T & a, const BFloat16 & b) +{ + return a == Float32(b); +} + +template +requires(!std::is_same_v) +constexpr bool operator!=(const BFloat16 & a, const T & b) +{ + return Float32(a) != b; +} + +template +requires(!std::is_same_v) +constexpr bool operator!=(const T & a, const BFloat16 & b) +{ + return a != Float32(b); +} + +template +requires(!std::is_same_v) +constexpr bool operator<(const BFloat16 & a, const T & b) +{ + return Float32(a) < b; +} + +template +requires(!std::is_same_v) +constexpr bool operator<(const T & a, const BFloat16 & b) +{ + return a < Float32(b); +} + +constexpr inline bool operator<(BFloat16 a, BFloat16 b) +{ + return Float32(a) < Float32(b); +} + +template +requires(!std::is_same_v) +constexpr bool operator>(const BFloat16 & a, const T & b) +{ + return Float32(a) > b; +} + +template +requires(!std::is_same_v) +constexpr bool operator>(const T & a, const BFloat16 & b) +{ + return a > Float32(b); +} + +constexpr inline bool operator>(BFloat16 a, BFloat16 b) +{ + return Float32(a) > Float32(b); +} + + +template +requires(!std::is_same_v) +constexpr bool operator<=(const BFloat16 & a, const T & b) +{ + return Float32(a) <= b; +} + +template +requires(!std::is_same_v) +constexpr bool operator<=(const T & a, const BFloat16 & b) +{ + return a <= Float32(b); +} + +constexpr inline bool operator<=(BFloat16 a, BFloat16 b) +{ + return Float32(a) <= Float32(b); +} + +template +requires(!std::is_same_v) +constexpr bool operator>=(const BFloat16 & a, const T & b) +{ + return Float32(a) >= b; +} + +template +requires(!std::is_same_v) +constexpr bool operator>=(const T & a, const BFloat16 & b) +{ + return a >= Float32(b); +} + +constexpr inline bool operator>=(BFloat16 a, BFloat16 b) +{ + return Float32(a) >= Float32(b); +} + + +template +requires(!std::is_same_v) +constexpr inline auto operator+(T a, BFloat16 b) +{ + return a + Float32(b); +} + +template +requires(!std::is_same_v) +constexpr inline auto operator+(BFloat16 a, T b) +{ + return Float32(a) + b; +} + +template +requires(!std::is_same_v) +constexpr inline auto operator-(T a, BFloat16 b) +{ + return a - Float32(b); +} + +template +requires(!std::is_same_v) +constexpr inline auto operator-(BFloat16 a, T b) +{ + return Float32(a) - b; +} + +template +requires(!std::is_same_v) +constexpr inline auto operator*(T a, BFloat16 b) +{ + return a * Float32(b); +} + +template +requires(!std::is_same_v) +constexpr inline auto operator*(BFloat16 a, T b) +{ + return Float32(a) * b; +} + +template +requires(!std::is_same_v) +constexpr inline auto operator/(T a, BFloat16 b) +{ + return a / Float32(b); +} + +template +requires(!std::is_same_v) +constexpr inline auto operator/(BFloat16 a, T b) +{ + return Float32(a) / b; +} + namespace std { - inline constexpr bool isfinite(BFloat16 x) { return (bit_cast(x) & 0b0111111110000000) != 0b0111111110000000; } - inline constexpr bool signbit(BFloat16 x) { return bit_cast(x) & 0b1000000000000000; } -} - -inline Float32 BFloat16ToFloat32(BFloat16 x) -{ - return bit_cast(static_cast(bit_cast(x)) << 16); -} - -inline BFloat16 Float32ToBFloat16(Float32 x) -{ - return bit_cast(std::bit_cast(x) >> 16); + inline constexpr bool isfinite(BFloat16 x) { return x.isFinite(); } + inline constexpr bool isnan(BFloat16 x) { return x.isNaN(); } + inline constexpr bool signbit(BFloat16 x) { return x.signBit(); } } diff --git a/base/base/DecomposedFloat.h b/base/base/DecomposedFloat.h index 26a929b4997..3bd059cb21c 100644 --- a/base/base/DecomposedFloat.h +++ b/base/base/DecomposedFloat.h @@ -11,7 +11,7 @@ template struct FloatTraits; template <> -struct FloatTraits<__bf16> +struct FloatTraits { using UInt = uint16_t; static constexpr size_t bits = 16; diff --git a/base/base/TypeLists.h b/base/base/TypeLists.h index ce3111b1da3..375ea94b5ea 100644 --- a/base/base/TypeLists.h +++ b/base/base/TypeLists.h @@ -9,10 +9,11 @@ namespace DB { using TypeListNativeInt = TypeList; -using TypeListFloat = TypeList; -using TypeListNativeNumber = TypeListConcat; +using TypeListNativeFloat = TypeList; +using TypeListNativeNumber = TypeListConcat; using TypeListWideInt = TypeList; using TypeListInt = TypeListConcat; +using TypeListFloat = TypeListConcat>; using TypeListIntAndFloat = TypeListConcat; using TypeListDecimal = TypeList; using TypeListNumber = TypeListConcat; diff --git a/src/AggregateFunctions/AggregateFunctionAvg.h b/src/AggregateFunctions/AggregateFunctionAvg.h index 6e1e9289565..8d53a081ee0 100644 --- a/src/AggregateFunctions/AggregateFunctionAvg.h +++ b/src/AggregateFunctions/AggregateFunctionAvg.h @@ -231,7 +231,7 @@ public: void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const final { - increment(place, static_cast(*columns[0]).getData()[row_num]); + increment(place, Numerator(static_cast(*columns[0]).getData()[row_num])); ++this->data(place).denominator; } diff --git a/src/AggregateFunctions/AggregateFunctionDeltaSum.cpp b/src/AggregateFunctions/AggregateFunctionDeltaSum.cpp index 42169c34c25..c61b9918a35 100644 --- a/src/AggregateFunctions/AggregateFunctionDeltaSum.cpp +++ b/src/AggregateFunctions/AggregateFunctionDeltaSum.cpp @@ -27,9 +27,9 @@ namespace template struct AggregationFunctionDeltaSumData { - T sum = 0; - T last = 0; - T first = 0; + T sum{}; + T last{}; + T first{}; bool seen = false; }; diff --git a/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp b/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp index 5819c533fd9..dc1adead87c 100644 --- a/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp +++ b/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp @@ -25,11 +25,11 @@ namespace template struct AggregationFunctionDeltaSumTimestampData { - ValueType sum = 0; - ValueType first = 0; - ValueType last = 0; - TimestampType first_ts = 0; - TimestampType last_ts = 0; + ValueType sum{}; + ValueType first{}; + ValueType last{}; + TimestampType first_ts{}; + TimestampType last_ts{}; bool seen = false; }; diff --git a/src/AggregateFunctions/AggregateFunctionMaxIntersections.cpp b/src/AggregateFunctions/AggregateFunctionMaxIntersections.cpp index ca91f960dab..f4edec7f528 100644 --- a/src/AggregateFunctions/AggregateFunctionMaxIntersections.cpp +++ b/src/AggregateFunctions/AggregateFunctionMaxIntersections.cpp @@ -155,9 +155,9 @@ public: void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena *) const override { - Int64 current_intersections = 0; - Int64 max_intersections = 0; - PointType position_of_max_intersections = 0; + Int64 current_intersections{}; + Int64 max_intersections{}; + PointType position_of_max_intersections{}; /// const_cast because we will sort the array auto & array = this->data(place).value; diff --git a/src/AggregateFunctions/AggregateFunctionSparkbar.cpp b/src/AggregateFunctions/AggregateFunctionSparkbar.cpp index 33412d50b21..de2a741e105 100644 --- a/src/AggregateFunctions/AggregateFunctionSparkbar.cpp +++ b/src/AggregateFunctions/AggregateFunctionSparkbar.cpp @@ -45,7 +45,7 @@ struct AggregateFunctionSparkbarData Y insert(const X & x, const Y & y) { if (isNaN(y) || y <= 0) - return 0; + return {}; auto [it, inserted] = points.insert({x, y}); if (!inserted) @@ -173,13 +173,13 @@ private: if (from_x >= to_x) { - size_t sz = updateFrame(values, 8); + size_t sz = updateFrame(values, Y{8}); values.push_back('\0'); offsets.push_back(offsets.empty() ? sz + 1 : offsets.back() + sz + 1); return; } - PaddedPODArray histogram(width, 0); + PaddedPODArray histogram(width, Y{0}); PaddedPODArray count_histogram(width, 0); /// The number of points in each bucket for (const auto & point : data.points) @@ -218,10 +218,10 @@ private: for (size_t i = 0; i < histogram.size(); ++i) { if (count_histogram[i] > 0) - histogram[i] /= count_histogram[i]; + histogram[i] = histogram[i] / count_histogram[i]; } - Y y_max = 0; + Y y_max{}; for (auto & y : histogram) { if (isNaN(y) || y <= 0) @@ -245,7 +245,7 @@ private: continue; } - constexpr auto levels_num = static_cast(BAR_LEVELS - 1); + constexpr auto levels_num = Y{BAR_LEVELS - 1}; if constexpr (is_floating_point) { y = y / (y_max / levels_num) + 1; diff --git a/src/AggregateFunctions/AggregateFunctionSum.h b/src/AggregateFunctions/AggregateFunctionSum.h index f6c51241a5c..7c7fb6338a2 100644 --- a/src/AggregateFunctions/AggregateFunctionSum.h +++ b/src/AggregateFunctions/AggregateFunctionSum.h @@ -83,7 +83,7 @@ struct AggregateFunctionSumData while (ptr < unrolled_end) { for (size_t i = 0; i < unroll_count; ++i) - Impl::add(partial_sums[i], ptr[i]); + Impl::add(partial_sums[i], T(ptr[i])); ptr += unroll_count; } @@ -95,7 +95,7 @@ struct AggregateFunctionSumData T local_sum{}; while (ptr < end_ptr) { - Impl::add(local_sum, *ptr); + Impl::add(local_sum, T(*ptr)); ++ptr; } Impl::add(sum, local_sum); @@ -227,7 +227,7 @@ struct AggregateFunctionSumData while (ptr < end_ptr) { if (!*condition_map == add_if_zero) - Impl::add(local_sum, *ptr); + Impl::add(local_sum, T(*ptr)); ++ptr; ++condition_map; } @@ -488,10 +488,7 @@ public: void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override { const auto & column = assert_cast(*columns[0]); - if constexpr (is_big_int_v) - this->data(place).add(static_cast(column.getData()[row_num])); - else - this->data(place).add(column.getData()[row_num]); + this->data(place).add(static_cast(column.getData()[row_num])); } void addBatchSinglePlace( diff --git a/src/Core/Types_fwd.h b/src/Core/Types_fwd.h index 6d3383ae7ff..b94a29ce72c 100644 --- a/src/Core/Types_fwd.h +++ b/src/Core/Types_fwd.h @@ -21,7 +21,7 @@ using Int128 = wide::integer<128, signed>; using UInt128 = wide::integer<128, unsigned>; using Int256 = wide::integer<256, signed>; using UInt256 = wide::integer<256, unsigned>; -using BFloat16 = __bf16; +class BFloat16; namespace DB { diff --git a/src/Functions/FunctionsRound.h b/src/Functions/FunctionsRound.h index 70ad4d17718..6c9cc8a37b3 100644 --- a/src/Functions/FunctionsRound.h +++ b/src/Functions/FunctionsRound.h @@ -298,7 +298,7 @@ public: static VectorType prepare(size_t scale) { - return load1(scale); + return load1(ScalarType(scale)); } }; diff --git a/src/Functions/PolygonUtils.h b/src/Functions/PolygonUtils.h index bf8241774a6..601ffcb00b4 100644 --- a/src/Functions/PolygonUtils.h +++ b/src/Functions/PolygonUtils.h @@ -583,7 +583,7 @@ struct CallPointInPolygon template static ColumnPtr call(const IColumn & x, const IColumn & y, PointInPolygonImpl && impl) { - using Impl = TypeListChangeRoot; + using Impl = TypeListChangeRoot; if (auto column = typeid_cast *>(&x)) return Impl::template call(*column, y, impl); return CallPointInPolygon::call(x, y, impl); @@ -609,7 +609,7 @@ struct CallPointInPolygon<> template NO_INLINE ColumnPtr pointInPolygon(const IColumn & x, const IColumn & y, PointInPolygonImpl && impl) { - using Impl = TypeListChangeRoot; + using Impl = TypeListChangeRoot; return Impl::call(x, y, impl); } diff --git a/src/Functions/divide.cpp b/src/Functions/divide.cpp index 7c67245c382..3947ba2d142 100644 --- a/src/Functions/divide.cpp +++ b/src/Functions/divide.cpp @@ -18,7 +18,7 @@ struct DivideFloatingImpl template static NO_SANITIZE_UNDEFINED Result apply(A a [[maybe_unused]], B b [[maybe_unused]]) { - return static_cast(a) / b; + return static_cast(a) / static_cast(b); } #if USE_EMBEDDED_COMPILER diff --git a/src/IO/WriteHelpers.h b/src/IO/WriteHelpers.h index f01e09e3f73..0a32c4c5446 100644 --- a/src/IO/WriteHelpers.h +++ b/src/IO/WriteHelpers.h @@ -174,7 +174,7 @@ inline size_t writeFloatTextFastPath(T x, char * buffer) } else if constexpr (std::is_same_v) { - Float32 f32 = BFloat16ToFloat32(x); + Float32 f32 = Float32(x); if (DecomposedFloat32(f32).isIntegerInRepresentableRange()) result = itoa(Int32(f32), buffer) - buffer; From 16d05bbc6d9a1369b393f836d0ccd8ea64fe2057 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 22:41:40 +0100 Subject: [PATCH 113/244] Comparisons --- base/base/BFloat16.h | 22 +++++++++++++++++++++- src/Functions/FunctionsComparison.h | 7 +++++-- 2 files changed, 26 insertions(+), 3 deletions(-) diff --git a/base/base/BFloat16.h b/base/base/BFloat16.h index 9c6196d6aab..f7491b64eb3 100644 --- a/base/base/BFloat16.h +++ b/base/base/BFloat16.h @@ -4,8 +4,28 @@ #include -//using BFloat16 = __bf16; +/** BFloat16 is a 16-bit floating point type, which has the same number (8) of exponent bits as Float32. + * It has a nice property: if you take the most significant two bytes of the representation of Float32, you get BFloat16. + * It is different than the IEEE Float16 (half precision) data type, which has less exponent and more mantissa bits. + * + * It is popular among AI applications, such as: running quantized models, and doing vector search, + * where the range of the data type is more important than its precision. + * + * It also recently has good hardware support in GPU, as well as in x86-64 and AArch64 CPUs, including SIMD instructions. + * But it is rarely utilized by compilers. + * + * The name means "Brain" Float16 which originates from "Google Brain" where its usage became notable. + * It is also known under the name "bf16". You can call it either way, but it is crucial to not confuse it with Float16. + * Here is a manual implementation of this data type. Only required operations are implemented. + * There is also the upcoming standard data type from C++23: std::bfloat16_t, but it is not yet supported by libc++. + * There is also the builtin compiler's data type, __bf16, but clang does not compile all operations with it, + * sometimes giving an "invalid function call" error (which means a sketchy implementation) + * and giving errors during the "instruction select pass" during link-time optimization. + * + * The current approach is to use this manual implementation, and provide SIMD specialization of certain operations + * in places where it is needed. + */ class BFloat16 { private: diff --git a/src/Functions/FunctionsComparison.h b/src/Functions/FunctionsComparison.h index be0875581a5..bcb9e0641b8 100644 --- a/src/Functions/FunctionsComparison.h +++ b/src/Functions/FunctionsComparison.h @@ -721,6 +721,7 @@ private: || (res = executeNumRightType(col_left, col_right_untyped)) || (res = executeNumRightType(col_left, col_right_untyped)) || (res = executeNumRightType(col_left, col_right_untyped)) + || (res = executeNumRightType(col_left, col_right_untyped)) || (res = executeNumRightType(col_left, col_right_untyped)) || (res = executeNumRightType(col_left, col_right_untyped))) return res; @@ -741,6 +742,7 @@ private: || (res = executeNumConstRightType(col_left_const, col_right_untyped)) || (res = executeNumConstRightType(col_left_const, col_right_untyped)) || (res = executeNumConstRightType(col_left_const, col_right_untyped)) + || (res = executeNumConstRightType(col_left_const, col_right_untyped)) || (res = executeNumConstRightType(col_left_const, col_right_untyped)) || (res = executeNumConstRightType(col_left_const, col_right_untyped))) return res; @@ -1289,9 +1291,10 @@ public: || (res = executeNumLeftType(col_left_untyped, col_right_untyped)) || (res = executeNumLeftType(col_left_untyped, col_right_untyped)) || (res = executeNumLeftType(col_left_untyped, col_right_untyped)) + || (res = executeNumLeftType(col_left_untyped, col_right_untyped)) || (res = executeNumLeftType(col_left_untyped, col_right_untyped)) || (res = executeNumLeftType(col_left_untyped, col_right_untyped)))) - throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Illegal column {} of first argument of function {}", + throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Illegal column {} of the first argument of function {}", col_left_untyped->getName(), getName()); return res; @@ -1339,7 +1342,7 @@ public: getName(), left_type->getName(), right_type->getName()); - /// When Decimal comparing to Float32/64, we convert both of them into Float64. + /// When Decimal comparing to Float32/64/16, we convert both of them into Float64. /// Other systems like MySQL and Spark also do as this. if (left_is_float || right_is_float) { From 92e8fa23ba0073f2caa43d66bab5d99475d3c656 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 23:43:10 +0100 Subject: [PATCH 114/244] Remove obsolete setting from tests --- src/Databases/enableAllExperimentalSettings.cpp | 1 - tests/performance/avg_weighted.xml | 1 - tests/performance/reinterpret_as.xml | 1 - tests/queries/0_stateless/01035_avg.sql | 2 -- .../0_stateless/01182_materialized_view_different_structure.sql | 1 - tests/queries/0_stateless/01440_big_int_exotic_casts.sql | 2 -- .../0_stateless/01554_bloom_filter_index_big_integer_uuid.sql | 2 -- tests/queries/0_stateless/01622_byte_size.sql | 2 -- tests/queries/0_stateless/01721_dictionary_decimal_p_s.sql | 2 -- tests/queries/0_stateless/01804_dictionary_decimal256_type.sql | 2 -- .../0_stateless/01875_ssd_cache_dictionary_decimal256_type.sh | 2 -- 11 files changed, 18 deletions(-) diff --git a/src/Databases/enableAllExperimentalSettings.cpp b/src/Databases/enableAllExperimentalSettings.cpp index d51d2671992..bc2dae55f97 100644 --- a/src/Databases/enableAllExperimentalSettings.cpp +++ b/src/Databases/enableAllExperimentalSettings.cpp @@ -24,7 +24,6 @@ void enableAllExperimentalSettings(ContextMutablePtr context) context->setSetting("allow_experimental_dynamic_type", 1); context->setSetting("allow_experimental_json_type", 1); context->setSetting("allow_experimental_vector_similarity_index", 1); - context->setSetting("allow_experimental_bigint_types", 1); context->setSetting("allow_experimental_window_functions", 1); context->setSetting("allow_experimental_geo_types", 1); context->setSetting("allow_experimental_map_type", 1); diff --git a/tests/performance/avg_weighted.xml b/tests/performance/avg_weighted.xml index edf3c19fdfa..ec1b7aae5c2 100644 --- a/tests/performance/avg_weighted.xml +++ b/tests/performance/avg_weighted.xml @@ -1,6 +1,5 @@ - 1 1 8 diff --git a/tests/performance/reinterpret_as.xml b/tests/performance/reinterpret_as.xml index d05ef3bb038..2e0fa0571c3 100644 --- a/tests/performance/reinterpret_as.xml +++ b/tests/performance/reinterpret_as.xml @@ -1,6 +1,5 @@ - 1 15G diff --git a/tests/queries/0_stateless/01035_avg.sql b/tests/queries/0_stateless/01035_avg.sql index a3cb35a80ec..0f7baddaec5 100644 --- a/tests/queries/0_stateless/01035_avg.sql +++ b/tests/queries/0_stateless/01035_avg.sql @@ -1,5 +1,3 @@ -SET allow_experimental_bigint_types=1; - CREATE TABLE IF NOT EXISTS test_01035_avg ( i8 Int8 DEFAULT i64, i16 Int16 DEFAULT i64, diff --git a/tests/queries/0_stateless/01182_materialized_view_different_structure.sql b/tests/queries/0_stateless/01182_materialized_view_different_structure.sql index 485f9985974..7e41172bd0c 100644 --- a/tests/queries/0_stateless/01182_materialized_view_different_structure.sql +++ b/tests/queries/0_stateless/01182_materialized_view_different_structure.sql @@ -20,7 +20,6 @@ SELECT sum(value) FROM (SELECT number, sum(number) AS value FROM (SELECT *, toDe CREATE TABLE src (n UInt64, s FixedString(16)) ENGINE=Memory; CREATE TABLE dst (n UInt8, s String) ENGINE = Memory; CREATE MATERIALIZED VIEW mv TO dst (n String) AS SELECT * FROM src; -SET allow_experimental_bigint_types=1; CREATE TABLE dist (n Int128) ENGINE=Distributed(test_cluster_two_shards, currentDatabase(), mv); INSERT INTO src SELECT number, toString(number) FROM numbers(1000); diff --git a/tests/queries/0_stateless/01440_big_int_exotic_casts.sql b/tests/queries/0_stateless/01440_big_int_exotic_casts.sql index 42fde9da01b..f411af897e8 100644 --- a/tests/queries/0_stateless/01440_big_int_exotic_casts.sql +++ b/tests/queries/0_stateless/01440_big_int_exotic_casts.sql @@ -32,8 +32,6 @@ SELECT number y, toInt128(number) - y, toInt256(number) - y, toUInt256(number) - SELECT -number y, toInt128(number) + y, toInt256(number) + y, toUInt256(number) + y FROM numbers_mt(10) ORDER BY number; -SET allow_experimental_bigint_types = 1; - DROP TABLE IF EXISTS t; CREATE TABLE t (x UInt64, i256 Int256, u256 UInt256, d256 Decimal256(2)) ENGINE = Memory; diff --git a/tests/queries/0_stateless/01554_bloom_filter_index_big_integer_uuid.sql b/tests/queries/0_stateless/01554_bloom_filter_index_big_integer_uuid.sql index 3472f41092d..f82fe39f439 100644 --- a/tests/queries/0_stateless/01554_bloom_filter_index_big_integer_uuid.sql +++ b/tests/queries/0_stateless/01554_bloom_filter_index_big_integer_uuid.sql @@ -1,5 +1,3 @@ -SET allow_experimental_bigint_types = 1; - CREATE TABLE 01154_test (x Int128, INDEX ix_x x TYPE bloom_filter(0.01) GRANULARITY 1) ENGINE = MergeTree() ORDER BY x SETTINGS index_granularity=8192; INSERT INTO 01154_test VALUES (1), (2), (3); SELECT x FROM 01154_test WHERE x = 1; diff --git a/tests/queries/0_stateless/01622_byte_size.sql b/tests/queries/0_stateless/01622_byte_size.sql index 9f9de4e58e9..f73011f4151 100644 --- a/tests/queries/0_stateless/01622_byte_size.sql +++ b/tests/queries/0_stateless/01622_byte_size.sql @@ -4,8 +4,6 @@ select ''; select '# byteSize'; -set allow_experimental_bigint_types = 1; - -- numbers #0 -- select ''; select 'byteSize for numbers #0'; diff --git a/tests/queries/0_stateless/01721_dictionary_decimal_p_s.sql b/tests/queries/0_stateless/01721_dictionary_decimal_p_s.sql index 272bd2d7104..57483430cc0 100644 --- a/tests/queries/0_stateless/01721_dictionary_decimal_p_s.sql +++ b/tests/queries/0_stateless/01721_dictionary_decimal_p_s.sql @@ -1,6 +1,5 @@ -- Tags: no-parallel -set allow_experimental_bigint_types=1; drop database if exists db_01721; drop table if exists db_01721.table_decimal_dict; drop dictionary if exists db_01721.decimal_dict; @@ -77,4 +76,3 @@ SELECT dictGet('db_01721.decimal_dict', 'Decimal32_', toUInt64(5000)), drop table if exists table_decimal_dict; drop dictionary if exists cache_dict; drop database if exists db_01721; - diff --git a/tests/queries/0_stateless/01804_dictionary_decimal256_type.sql b/tests/queries/0_stateless/01804_dictionary_decimal256_type.sql index 08a8d0feb27..32b029442b9 100644 --- a/tests/queries/0_stateless/01804_dictionary_decimal256_type.sql +++ b/tests/queries/0_stateless/01804_dictionary_decimal256_type.sql @@ -1,7 +1,5 @@ -- Tags: no-parallel -SET allow_experimental_bigint_types = 1; - DROP TABLE IF EXISTS dictionary_decimal_source_table; CREATE TABLE dictionary_decimal_source_table ( diff --git a/tests/queries/0_stateless/01875_ssd_cache_dictionary_decimal256_type.sh b/tests/queries/0_stateless/01875_ssd_cache_dictionary_decimal256_type.sh index 1294ba53e82..2a24a931696 100755 --- a/tests/queries/0_stateless/01875_ssd_cache_dictionary_decimal256_type.sh +++ b/tests/queries/0_stateless/01875_ssd_cache_dictionary_decimal256_type.sh @@ -6,8 +6,6 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) . "$CURDIR"/../shell_config.sh $CLICKHOUSE_CLIENT --query=" - SET allow_experimental_bigint_types = 1; - DROP TABLE IF EXISTS dictionary_decimal_source_table; CREATE TABLE dictionary_decimal_source_table ( From 19ab7d484a6d7a2346103c5468bca611df03e3d9 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Sun, 10 Nov 2024 23:50:31 +0100 Subject: [PATCH 115/244] Add an experimental setting --- src/Core/Settings.cpp | 5 ++++- src/Core/SettingsChangesHistory.cpp | 1 + .../parseColumnsListForTableFunction.cpp | 14 ++++++++++++++ .../parseColumnsListForTableFunction.h | 1 + 4 files changed, 20 insertions(+), 1 deletion(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index 01339226c2d..7c2042ee16d 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -5729,7 +5729,10 @@ Enable experimental functions for natural language processing. Enable experimental hash functions )", EXPERIMENTAL) \ DECLARE(Bool, allow_experimental_object_type, false, R"( -Allow Object and JSON data types +Allow the obsolete Object data type +)", EXPERIMENTAL) \ + DECLARE(Bool, allow_experimental_bfloat16_type, false, R"( +Allow BFloat16 data type (under development). )", EXPERIMENTAL) \ DECLARE(Bool, allow_experimental_time_series_table, false, R"( Allows creation of tables with the [TimeSeries](../../engines/table-engines/integrations/time-series.md) table engine. diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index 0ff9d0a6833..23aeeb47224 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -77,6 +77,7 @@ static std::initializer_list(&data_type)) diff --git a/src/Interpreters/parseColumnsListForTableFunction.h b/src/Interpreters/parseColumnsListForTableFunction.h index 6e00492c0ad..39b9f092d89 100644 --- a/src/Interpreters/parseColumnsListForTableFunction.h +++ b/src/Interpreters/parseColumnsListForTableFunction.h @@ -20,6 +20,7 @@ struct DataTypeValidationSettings bool allow_experimental_object_type = true; bool allow_suspicious_fixed_string_types = true; bool allow_experimental_variant_type = true; + bool allow_experimental_bfloat16_type = true; bool allow_suspicious_variant_types = true; bool validate_nested_types = true; bool allow_experimental_dynamic_type = true; From 1a2ee7929e746395a6f0426b6935887af287fd30 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 00:16:09 +0100 Subject: [PATCH 116/244] More conversions --- src/Functions/FunctionsConversion.cpp | 5 ----- 1 file changed, 5 deletions(-) diff --git a/src/Functions/FunctionsConversion.cpp b/src/Functions/FunctionsConversion.cpp index 7f4ccc338cf..effaa6faa6d 100644 --- a/src/Functions/FunctionsConversion.cpp +++ b/src/Functions/FunctionsConversion.cpp @@ -1862,11 +1862,6 @@ struct ConvertImpl } } - if constexpr ((std::is_same_v || std::is_same_v) - && !(std::is_same_v || std::is_same_v)) - throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Conversion from {} to {} is not supported", - TypeName, TypeName); - if constexpr (std::is_same_v || std::is_same_v) { From f042c921ee84ef583f1b76c9d4587b963bd06f45 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 00:16:28 +0100 Subject: [PATCH 117/244] Distances --- base/base/BFloat16.h | 7 ++ src/Common/CPUID.h | 6 ++ src/Common/TargetSpecific.cpp | 3 + src/Common/TargetSpecific.h | 26 +++++- src/Functions/array/arrayDistance.cpp | 112 +++++++++++++++++++------- 5 files changed, 119 insertions(+), 35 deletions(-) diff --git a/base/base/BFloat16.h b/base/base/BFloat16.h index f7491b64eb3..2df84dbc0f2 100644 --- a/base/base/BFloat16.h +++ b/base/base/BFloat16.h @@ -80,6 +80,13 @@ public: return x & 0b1000000000000000; } + constexpr BFloat16 abs() const + { + BFloat16 res; + res.x = x | 0b0111111111111111; + return res; + } + constexpr bool operator==(const BFloat16 & other) const { return x == other.x; diff --git a/src/Common/CPUID.h b/src/Common/CPUID.h index b49f7706904..b5c26e64d1e 100644 --- a/src/Common/CPUID.h +++ b/src/Common/CPUID.h @@ -266,6 +266,11 @@ inline bool haveAVX512VBMI2() noexcept return haveAVX512F() && ((CPUInfo(0x7, 0).registers.ecx >> 6) & 1u); } +inline bool haveAVX512BF16() noexcept +{ + return haveAVX512F() && ((CPUInfo(0x7, 1).registers.eax >> 5) & 1u); +} + inline bool haveRDRAND() noexcept { return CPUInfo(0x0).registers.eax >= 0x7 && ((CPUInfo(0x1).registers.ecx >> 30) & 1u); @@ -326,6 +331,7 @@ inline bool haveAMXINT8() noexcept OP(AVX512VL) \ OP(AVX512VBMI) \ OP(AVX512VBMI2) \ + OP(AVX512BF16) \ OP(PREFETCHWT1) \ OP(SHA) \ OP(ADX) \ diff --git a/src/Common/TargetSpecific.cpp b/src/Common/TargetSpecific.cpp index 8540c9a9986..4400d9a60b3 100644 --- a/src/Common/TargetSpecific.cpp +++ b/src/Common/TargetSpecific.cpp @@ -23,6 +23,8 @@ UInt32 getSupportedArchs() result |= static_cast(TargetArch::AVX512VBMI); if (CPU::CPUFlagsCache::have_AVX512VBMI2) result |= static_cast(TargetArch::AVX512VBMI2); + if (CPU::CPUFlagsCache::have_AVX512BF16) + result |= static_cast(TargetArch::AVX512BF16); if (CPU::CPUFlagsCache::have_AMXBF16) result |= static_cast(TargetArch::AMXBF16); if (CPU::CPUFlagsCache::have_AMXTILE) @@ -50,6 +52,7 @@ String toString(TargetArch arch) case TargetArch::AVX512BW: return "avx512bw"; case TargetArch::AVX512VBMI: return "avx512vbmi"; case TargetArch::AVX512VBMI2: return "avx512vbmi2"; + case TargetArch::AVX512BF16: return "avx512bf16"; case TargetArch::AMXBF16: return "amxbf16"; case TargetArch::AMXTILE: return "amxtile"; case TargetArch::AMXINT8: return "amxint8"; diff --git a/src/Common/TargetSpecific.h b/src/Common/TargetSpecific.h index f9523f667b2..5584bd1f63a 100644 --- a/src/Common/TargetSpecific.h +++ b/src/Common/TargetSpecific.h @@ -83,9 +83,10 @@ enum class TargetArch : UInt32 AVX512BW = (1 << 4), AVX512VBMI = (1 << 5), AVX512VBMI2 = (1 << 6), - AMXBF16 = (1 << 7), - AMXTILE = (1 << 8), - AMXINT8 = (1 << 9), + AVX512BF16 = (1 << 7), + AMXBF16 = (1 << 8), + AMXTILE = (1 << 9), + AMXINT8 = (1 << 10), }; /// Runtime detection. @@ -102,6 +103,7 @@ String toString(TargetArch arch); /// NOLINTNEXTLINE #define USE_MULTITARGET_CODE 1 +#define AVX512BF16_FUNCTION_SPECIFIC_ATTRIBUTE __attribute__((target("sse,sse2,sse3,ssse3,sse4,popcnt,avx,avx2,avx512f,avx512bw,avx512vl,avx512vbmi,avx512vbmi2,avx512bf16"))) #define AVX512VBMI2_FUNCTION_SPECIFIC_ATTRIBUTE __attribute__((target("sse,sse2,sse3,ssse3,sse4,popcnt,avx,avx2,avx512f,avx512bw,avx512vl,avx512vbmi,avx512vbmi2"))) #define AVX512VBMI_FUNCTION_SPECIFIC_ATTRIBUTE __attribute__((target("sse,sse2,sse3,ssse3,sse4,popcnt,avx,avx2,avx512f,avx512bw,avx512vl,avx512vbmi"))) #define AVX512BW_FUNCTION_SPECIFIC_ATTRIBUTE __attribute__((target("sse,sse2,sse3,ssse3,sse4,popcnt,avx,avx2,avx512f,avx512bw"))) @@ -111,6 +113,8 @@ String toString(TargetArch arch); #define SSE42_FUNCTION_SPECIFIC_ATTRIBUTE __attribute__((target("sse,sse2,sse3,ssse3,sse4,popcnt"))) #define DEFAULT_FUNCTION_SPECIFIC_ATTRIBUTE +# define BEGIN_AVX512BF16_SPECIFIC_CODE \ + _Pragma("clang attribute push(__attribute__((target(\"sse,sse2,sse3,ssse3,sse4,popcnt,avx,avx2,avx512f,avx512bw,avx512vl,avx512vbmi,avx512vbmi2,avx512bf16\"))),apply_to=function)") # define BEGIN_AVX512VBMI2_SPECIFIC_CODE \ _Pragma("clang attribute push(__attribute__((target(\"sse,sse2,sse3,ssse3,sse4,popcnt,avx,avx2,avx512f,avx512bw,avx512vl,avx512vbmi,avx512vbmi2\"))),apply_to=function)") # define BEGIN_AVX512VBMI_SPECIFIC_CODE \ @@ -197,6 +201,14 @@ namespace TargetSpecific::AVX512VBMI2 { \ } \ END_TARGET_SPECIFIC_CODE +#define DECLARE_AVX512BF16_SPECIFIC_CODE(...) \ +BEGIN_AVX512BF16_SPECIFIC_CODE \ +namespace TargetSpecific::AVX512BF16 { \ + DUMMY_FUNCTION_DEFINITION \ + using namespace DB::TargetSpecific::AVX512BF16; \ + __VA_ARGS__ \ +} \ +END_TARGET_SPECIFIC_CODE #else @@ -211,6 +223,7 @@ END_TARGET_SPECIFIC_CODE #define DECLARE_AVX512BW_SPECIFIC_CODE(...) #define DECLARE_AVX512VBMI_SPECIFIC_CODE(...) #define DECLARE_AVX512VBMI2_SPECIFIC_CODE(...) +#define DECLARE_AVX512BF16_SPECIFIC_CODE(...) #endif @@ -229,7 +242,8 @@ DECLARE_AVX2_SPECIFIC_CODE (__VA_ARGS__) \ DECLARE_AVX512F_SPECIFIC_CODE(__VA_ARGS__) \ DECLARE_AVX512BW_SPECIFIC_CODE (__VA_ARGS__) \ DECLARE_AVX512VBMI_SPECIFIC_CODE (__VA_ARGS__) \ -DECLARE_AVX512VBMI2_SPECIFIC_CODE (__VA_ARGS__) +DECLARE_AVX512VBMI2_SPECIFIC_CODE (__VA_ARGS__) \ +DECLARE_AVX512BF16_SPECIFIC_CODE (__VA_ARGS__) DECLARE_DEFAULT_CODE( constexpr auto BuildArch = TargetArch::Default; /// NOLINT @@ -263,6 +277,10 @@ DECLARE_AVX512VBMI2_SPECIFIC_CODE( constexpr auto BuildArch = TargetArch::AVX512VBMI2; /// NOLINT ) // DECLARE_AVX512VBMI2_SPECIFIC_CODE +DECLARE_AVX512BF16_SPECIFIC_CODE( + constexpr auto BuildArch = TargetArch::AVX512BF16; /// NOLINT +) // DECLARE_AVX512BF16_SPECIFIC_CODE + /** Runtime Dispatch helpers for class members. * * Example of usage: diff --git a/src/Functions/array/arrayDistance.cpp b/src/Functions/array/arrayDistance.cpp index a1f48747eb6..da49359c422 100644 --- a/src/Functions/array/arrayDistance.cpp +++ b/src/Functions/array/arrayDistance.cpp @@ -14,6 +14,31 @@ #include #endif + +namespace +{ + inline BFloat16 fabs(BFloat16 x) + { + return x.abs(); + } + + inline BFloat16 sqrt(BFloat16 x) + { + return BFloat16(::sqrtf(Float32(x))); + } + + template + inline BFloat16 pow(BFloat16 x, T p) + { + return BFloat16(::powf(Float32(x), Float32(p))); + } + + inline BFloat16 fmax(BFloat16 x, BFloat16 y) + { + return BFloat16(::fmaxf(Float32(x), Float32(y))); + } +} + namespace DB { namespace ErrorCodes @@ -34,7 +59,7 @@ struct L1Distance template struct State { - FloatType sum = 0; + FloatType sum{}; }; template @@ -65,7 +90,7 @@ struct L2Distance template struct State { - FloatType sum = 0; + FloatType sum{}; }; template @@ -82,7 +107,7 @@ struct L2Distance #if USE_MULTITARGET_CODE template - AVX512_FUNCTION_SPECIFIC_ATTRIBUTE static void accumulateCombine( + AVX512BF16_FUNCTION_SPECIFIC_ATTRIBUTE static void accumulateCombine( const ResultType * __restrict data_x, const ResultType * __restrict data_y, size_t i_max, @@ -90,19 +115,29 @@ struct L2Distance size_t & i_y, State & state) { - static constexpr bool is_float32 = std::is_same_v; - __m512 sums; - if constexpr (is_float32) + if constexpr (sizeof(ResultType) <= 4) sums = _mm512_setzero_ps(); else sums = _mm512_setzero_pd(); - constexpr size_t n = is_float32 ? 16 : 8; + constexpr size_t n = sizeof(__m512) / sizeof(ResultType); for (; i_x + n < i_max; i_x += n, i_y += n) { - if constexpr (is_float32) + if constexpr (sizeof(ResultType) == 2) + { + __m512 x_1 = _mm512_cvtpbh_ps(_mm256_loadu_ps(reinterpret_cast(data_x + i_x))); + __m512 x_2 = _mm512_cvtpbh_ps(_mm256_loadu_ps(reinterpret_cast(data_x + i_x + n / 2))); + __m512 y_1 = _mm512_cvtpbh_ps(_mm256_loadu_ps(reinterpret_cast(data_y + i_y))); + __m512 y_2 = _mm512_cvtpbh_ps(_mm256_loadu_ps(reinterpret_cast(data_y + i_y + n / 2))); + + __m512 differences_1 = _mm512_sub_ps(x_1, y_1); + __m512 differences_2 = _mm512_sub_ps(x_2, y_2); + sums = _mm512_fmadd_ps(differences_1, differences_1, sums); + sums = _mm512_fmadd_ps(differences_2, differences_2, sums); + } + else if constexpr (sizeof(ResultType) == 4) { __m512 x = _mm512_loadu_ps(data_x + i_x); __m512 y = _mm512_loadu_ps(data_y + i_y); @@ -118,7 +153,7 @@ struct L2Distance } } - if constexpr (is_float32) + if constexpr (sizeof(ResultType) <= 4) state.sum = _mm512_reduce_add_ps(sums); else state.sum = _mm512_reduce_add_pd(sums); @@ -128,7 +163,7 @@ struct L2Distance template static ResultType finalize(const State & state, const ConstParams &) { - return sqrt(state.sum); + return sqrt(ResultType(state.sum)); } }; @@ -156,13 +191,13 @@ struct LpDistance template struct State { - FloatType sum = 0; + FloatType sum{}; }; template static void accumulate(State & state, ResultType x, ResultType y, const ConstParams & params) { - state.sum += static_cast(std::pow(fabs(x - y), params.power)); + state.sum += static_cast(pow(fabs(x - y), params.power)); } template @@ -174,7 +209,7 @@ struct LpDistance template static ResultType finalize(const State & state, const ConstParams & params) { - return static_cast(std::pow(state.sum, params.inverted_power)); + return static_cast(pow(state.sum, params.inverted_power)); } }; @@ -187,7 +222,7 @@ struct LinfDistance template struct State { - FloatType dist = 0; + FloatType dist{}; }; template @@ -218,9 +253,9 @@ struct CosineDistance template struct State { - FloatType dot_prod = 0; - FloatType x_squared = 0; - FloatType y_squared = 0; + FloatType dot_prod{}; + FloatType x_squared{}; + FloatType y_squared{}; }; template @@ -241,7 +276,7 @@ struct CosineDistance #if USE_MULTITARGET_CODE template - AVX512_FUNCTION_SPECIFIC_ATTRIBUTE static void accumulateCombine( + AVX512BF16_FUNCTION_SPECIFIC_ATTRIBUTE static void accumulateCombine( const ResultType * __restrict data_x, const ResultType * __restrict data_y, size_t i_max, @@ -249,13 +284,11 @@ struct CosineDistance size_t & i_y, State & state) { - static constexpr bool is_float32 = std::is_same_v; - __m512 dot_products; __m512 x_squareds; __m512 y_squareds; - if constexpr (is_float32) + if constexpr (sizeof(ResultType) <= 4) { dot_products = _mm512_setzero_ps(); x_squareds = _mm512_setzero_ps(); @@ -268,11 +301,19 @@ struct CosineDistance y_squareds = _mm512_setzero_pd(); } - constexpr size_t n = is_float32 ? 16 : 8; + constexpr size_t n = sizeof(__m512) / sizeof(ResultType); for (; i_x + n < i_max; i_x += n, i_y += n) { - if constexpr (is_float32) + if constexpr (sizeof(ResultType) == 2) + { + __m512 x = _mm512_loadu_ps(data_x + i_x); + __m512 y = _mm512_loadu_ps(data_y + i_y); + dot_products = _mm512_dpbf16_ps(dot_products, x, y); + x_squareds = _mm512_dpbf16_ps(x_squareds, x, x); + y_squareds = _mm512_dpbf16_ps(y_squareds, y, y); + } + if constexpr (sizeof(ResultType) == 4) { __m512 x = _mm512_loadu_ps(data_x + i_x); __m512 y = _mm512_loadu_ps(data_y + i_y); @@ -290,7 +331,7 @@ struct CosineDistance } } - if constexpr (is_float32) + if constexpr (sizeof(ResultType) == 2 || sizeof(ResultType) == 4) { state.dot_prod = _mm512_reduce_add_ps(dot_products); state.x_squared = _mm512_reduce_add_ps(x_squareds); @@ -308,7 +349,7 @@ struct CosineDistance template static ResultType finalize(const State & state, const ConstParams &) { - return 1 - state.dot_prod / sqrt(state.x_squared * state.y_squared); + return ResultType(1) - state.dot_prod / sqrt(state.x_squared * state.y_squared); } }; @@ -353,11 +394,13 @@ public: return std::make_shared(); case TypeIndex::Float32: return std::make_shared(); + case TypeIndex::BFloat16: + return std::make_shared(); default: throw Exception( ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Arguments of function {} has nested type {}. " - "Supported types: UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, Float32, Float64.", + "Supported types: UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, BFloat16, Float32, Float64.", getName(), common_type->getName()); } @@ -367,6 +410,9 @@ public: { switch (result_type->getTypeId()) { + case TypeIndex::BFloat16: + return executeWithResultType(arguments, input_rows_count); + break; case TypeIndex::Float32: return executeWithResultType(arguments, input_rows_count); break; @@ -388,6 +434,7 @@ public: ACTION(Int16) \ ACTION(Int32) \ ACTION(Int64) \ + ACTION(BFloat16) \ ACTION(Float32) \ ACTION(Float64) @@ -412,7 +459,7 @@ private: throw Exception( ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Arguments of function {} has nested type {}. " - "Supported types: UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, Float32, Float64.", + "Supported types: UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, BFloat16, Float32, Float64.", getName(), type_x->getName()); } @@ -437,7 +484,7 @@ private: throw Exception( ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Arguments of function {} has nested type {}. " - "Supported types: UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, Float32, Float64.", + "Supported types: UInt8, UInt16, UInt32, UInt64, Int8, Int16, Int32, Int64, BFloat16, Float32, Float64.", getName(), type_y->getName()); } @@ -548,13 +595,15 @@ private: /// SIMD optimization: process multiple elements in both input arrays at once. /// To avoid combinatorial explosion of SIMD kernels, focus on - /// - the two most common input/output types (Float32 x Float32) --> Float32 and (Float64 x Float64) --> Float64 instead of 10 x - /// 10 input types x 2 output types, + /// - the three most common input/output types (BFloat16 x BFloat16) --> BFloat16, + /// (Float32 x Float32) --> Float32 and (Float64 x Float64) --> Float64 + /// instead of 10 x 10 input types x 2 output types, /// - const/non-const inputs instead of non-const/non-const inputs /// - the two most common metrics L2 and cosine distance, /// - the most powerful SIMD instruction set (AVX-512F). #if USE_MULTITARGET_CODE - if constexpr (std::is_same_v && std::is_same_v) /// ResultType is Float32 or Float64 + /// ResultType is BFloat16, Float32 or Float64 + if constexpr (std::is_same_v && std::is_same_v) { if constexpr (std::is_same_v || std::is_same_v) @@ -638,4 +687,5 @@ FunctionPtr createFunctionArrayL2SquaredDistance(ContextPtr context_) { return F FunctionPtr createFunctionArrayLpDistance(ContextPtr context_) { return FunctionArrayDistance::create(context_); } FunctionPtr createFunctionArrayLinfDistance(ContextPtr context_) { return FunctionArrayDistance::create(context_); } FunctionPtr createFunctionArrayCosineDistance(ContextPtr context_) { return FunctionArrayDistance::create(context_); } + } From 6dee7e42766177e30712f7c1c341663b3fba2f91 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 00:24:57 +0100 Subject: [PATCH 118/244] Fix style --- src/Databases/enableAllExperimentalSettings.cpp | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/Databases/enableAllExperimentalSettings.cpp b/src/Databases/enableAllExperimentalSettings.cpp index bc2dae55f97..1be54664bc9 100644 --- a/src/Databases/enableAllExperimentalSettings.cpp +++ b/src/Databases/enableAllExperimentalSettings.cpp @@ -27,6 +27,8 @@ void enableAllExperimentalSettings(ContextMutablePtr context) context->setSetting("allow_experimental_window_functions", 1); context->setSetting("allow_experimental_geo_types", 1); context->setSetting("allow_experimental_map_type", 1); + context->setSetting("allow_experimental_bigint_types", 1); + context->setSetting("allow_experimental_bfloat16_type", 1); context->setSetting("allow_deprecated_error_prone_window_functions", 1); context->setSetting("allow_suspicious_low_cardinality_types", 1); From 89b015cecfad9a6a8f44039efa556d209ea50239 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 00:25:11 +0100 Subject: [PATCH 119/244] Do not compile BFloat16 --- src/DataTypes/IDataType.h | 3 ++- src/DataTypes/Native.cpp | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/src/DataTypes/IDataType.h b/src/DataTypes/IDataType.h index 1e41d6b2eba..8f06526ddbb 100644 --- a/src/DataTypes/IDataType.h +++ b/src/DataTypes/IDataType.h @@ -411,7 +411,8 @@ struct WhichDataType constexpr bool isBFloat16() const { return idx == TypeIndex::BFloat16; } constexpr bool isFloat32() const { return idx == TypeIndex::Float32; } constexpr bool isFloat64() const { return idx == TypeIndex::Float64; } - constexpr bool isFloat() const { return isBFloat16() || isFloat32() || isFloat64(); } + constexpr bool isNativeFloat() const { return isFloat32() || isFloat64(); } + constexpr bool isFloat() const { return isNativeFloat() || isBFloat16(); } constexpr bool isNativeNumber() const { return isNativeInteger() || isFloat(); } constexpr bool isNumber() const { return isInteger() || isFloat() || isDecimal(); } diff --git a/src/DataTypes/Native.cpp b/src/DataTypes/Native.cpp index 5dc490b0bd5..53354d7c6e0 100644 --- a/src/DataTypes/Native.cpp +++ b/src/DataTypes/Native.cpp @@ -37,7 +37,7 @@ bool canBeNativeType(const IDataType & type) return canBeNativeType(*data_type_nullable.getNestedType()); } - return data_type.isNativeInt() || data_type.isNativeUInt() || data_type.isFloat() || data_type.isDate() + return data_type.isNativeInt() || data_type.isNativeUInt() || data_type.isNativeFloat() || data_type.isDate() || data_type.isDate32() || data_type.isDateTime() || data_type.isEnum(); } From 968a559917577a63464bbaf87d3e724912cb7d5a Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 00:59:37 +0100 Subject: [PATCH 120/244] Add a test --- .../queries/0_stateless/03269_bf16.reference | 45 ++++++++++ tests/queries/0_stateless/03269_bf16.sql | 88 +++++++++++++++++++ 2 files changed, 133 insertions(+) create mode 100644 tests/queries/0_stateless/03269_bf16.reference create mode 100644 tests/queries/0_stateless/03269_bf16.sql diff --git a/tests/queries/0_stateless/03269_bf16.reference b/tests/queries/0_stateless/03269_bf16.reference new file mode 100644 index 00000000000..daa26cb252f --- /dev/null +++ b/tests/queries/0_stateless/03269_bf16.reference @@ -0,0 +1,45 @@ +1 -1 1.09375 -1.09375 1 -1 1.09375 -1.09375 18446744000000000000 -0 inf -inf nan +1.09375 1.09375 1.09375 1 +1 1 0 1 1 +0 2.1875 1.1962891 1 Float32 Float32 Float32 Float64 +-0.006250000000000089 2.19375 1.203125 1.0057142857142858 Float64 Float64 Float64 Float64 +0 0 1 0 +1000 1000 1 0 +2000 2000 1 0 +3000 2992 0 8 +4000 4000 1 0 +5000 4992 0 8 +6000 5984 0 16 +7000 6976 0 24 +8000 8000 1 0 +9000 8960 0 40 +49995000 49855104 4999.5 4985.5104 0 0 9999 9984 10000 925 10000 925 +0 0 1 0 +1000 1000 1 0 +2000 2000 1 0 +3000 2992 0 8 +4000 4000 1 0 +5000 4992 0 8 +6000 5984 0 16 +7000 6976 0 24 +8000 8000 1 0 +9000 8960 0 40 +49995000 49855104 4999.5 4985.5104 0 0 9999 9984 10000 925 10000 925 +Row 1: +────── +a32: [0,0.5,1,1.5,2,2.5,3,3.5,4,4.5,5,5.5,6,6.5,7,7.5,8,8.5,9,9.5,10,10.5,11,11.5,12,12.5,13,13.5,14,14.5,15,15.5,16,16.5,17,17.5,18,18.5,19,19.5,20,20.5,21,21.5,22,22.5,23,23.5,24,24.5,25,25.5,26,26.5,27,27.5,28,28.5,29,29.5,30,30.5,31,31.5,32,32.5,33,33.5,34,34.5,35,35.5,36,36.5,37,37.5,38,38.5,39,39.5,40,40.5,41,41.5,42,42.5,43,43.5,44,44.5,45,45.5,46,46.5,47,47.5,48,48.5,49,49.5,50,50.5,51,51.5,52,52.5,53,53.5,54,54.5,55,55.5,56,56.5,57,57.5,58,58.5,59,59.5,60,60.5,61,61.5,62,62.5,63,63.5,64,64.5,65,65.5,66,66.5,67,67.5,68,68.5,69,69.5,70,70.5,71,71.5,72,72.5,73,73.5,74,74.5,75,75.5,76,76.5,77,77.5,78,78.5,79,79.5,80,80.5,81,81.5,82,82.5,83,83.5,84,84.5,85,85.5,86,86.5,87,87.5,88,88.5,89,89.5,90,90.5,91,91.5,92,92.5,93,93.5,94,94.5,95,95.5,96,96.5,97,97.5,98,98.5,99,99.5,100,100.5,101,101.5,102,102.5,103,103.5,104,104.5,105,105.5,106,106.5,107,107.5,108,108.5,109,109.5,110,110.5,111,111.5,112,112.5,113,113.5,114,114.5,115,115.5,116,116.5,117,117.5,118,118.5,119,119.5,120,120.5,121,121.5,122,122.5,123,123.5,124,124.5,125,125.5,126,126.5,127,127.5,128,128.5,129,129.5,130,130.5,131,131.5,132,132.5,133,133.5,134,134.5,135,135.5,136,136.5,137,137.5,138,138.5,139,139.5,140,140.5,141,141.5,142,142.5,143,143.5,144,144.5,145,145.5,146,146.5,147,147.5,148,148.5,149,149.5,150,150.5,151,151.5,152,152.5,153,153.5,154,154.5,155,155.5,156,156.5,157,157.5,158,158.5,159,159.5,160,160.5,161,161.5,162,162.5,163,163.5,164,164.5,165,165.5,166,166.5,167,167.5,168,168.5,169,169.5,170,170.5,171,171.5,172,172.5,173,173.5,174,174.5,175,175.5,176,176.5,177,177.5,178,178.5,179,179.5,180,180.5,181,181.5,182,182.5,183,183.5,184,184.5,185,185.5,186,186.5,187,187.5,188,188.5,189,189.5,190,190.5,191,191.5] +a16: [0,0.5,1,1.5,2,2.5,3,3.5,4,4.5,5,5.5,6,6.5,7,7.5,8,8.5,9,9.5,10,10.5,11,11.5,12,12.5,13,13.5,14,14.5,15,15.5,16,16.5,17,17.5,18,18.5,19,19.5,20,20.5,21,21.5,22,22.5,23,23.5,24,24.5,25,25.5,26,26.5,27,27.5,28,28.5,29,29.5,30,30.5,31,31.5,32,32.5,33,33.5,34,34.5,35,35.5,36,36.5,37,37.5,38,38.5,39,39.5,40,40.5,41,41.5,42,42.5,43,43.5,44,44.5,45,45.5,46,46.5,47,47.5,48,48.5,49,49.5,50,50.5,51,51.5,52,52.5,53,53.5,54,54.5,55,55.5,56,56.5,57,57.5,58,58.5,59,59.5,60,60.5,61,61.5,62,62.5,63,63.5,64,64.5,65,65.5,66,66.5,67,67.5,68,68.5,69,69.5,70,70.5,71,71.5,72,72.5,73,73.5,74,74.5,75,75.5,76,76.5,77,77.5,78,78.5,79,79.5,80,80.5,81,81.5,82,82.5,83,83.5,84,84.5,85,85.5,86,86.5,87,87.5,88,88.5,89,89.5,90,90.5,91,91.5,92,92.5,93,93.5,94,94.5,95,95.5,96,96.5,97,97.5,98,98.5,99,99.5,100,100.5,101,101.5,102,102.5,103,103.5,104,104.5,105,105.5,106,106.5,107,107.5,108,108.5,109,109.5,110,110.5,111,111.5,112,112.5,113,113.5,114,114.5,115,115.5,116,116.5,117,117.5,118,118.5,119,119.5,120,120.5,121,121.5,122,122.5,123,123.5,124,124.5,125,125.5,126,126.5,127,127.5,128,128,129,129,130,130,131,131,132,132,133,133,134,134,135,135,136,136,137,137,138,138,139,139,140,140,141,141,142,142,143,143,144,144,145,145,146,146,147,147,148,148,149,149,150,150,151,151,152,152,153,153,154,154,155,155,156,156,157,157,158,158,159,159,160,160,161,161,162,162,163,163,164,164,165,165,166,166,167,167,168,168,169,169,170,170,171,171,172,172,173,173,174,174,175,175,176,176,177,177,178,178,179,179,180,180,181,181,182,182,183,183,184,184,185,185,186,186,187,187,188,188,189,189,190,190,191,191] +a32_1: [1,1.5,2,2.5,3,3.5,4,4.5,5,5.5,6,6.5,7,7.5,8,8.5,9,9.5,10,10.5,11,11.5,12,12.5,13,13.5,14,14.5,15,15.5,16,16.5,17,17.5,18,18.5,19,19.5,20,20.5,21,21.5,22,22.5,23,23.5,24,24.5,25,25.5,26,26.5,27,27.5,28,28.5,29,29.5,30,30.5,31,31.5,32,32.5,33,33.5,34,34.5,35,35.5,36,36.5,37,37.5,38,38.5,39,39.5,40,40.5,41,41.5,42,42.5,43,43.5,44,44.5,45,45.5,46,46.5,47,47.5,48,48.5,49,49.5,50,50.5,51,51.5,52,52.5,53,53.5,54,54.5,55,55.5,56,56.5,57,57.5,58,58.5,59,59.5,60,60.5,61,61.5,62,62.5,63,63.5,64,64.5,65,65.5,66,66.5,67,67.5,68,68.5,69,69.5,70,70.5,71,71.5,72,72.5,73,73.5,74,74.5,75,75.5,76,76.5,77,77.5,78,78.5,79,79.5,80,80.5,81,81.5,82,82.5,83,83.5,84,84.5,85,85.5,86,86.5,87,87.5,88,88.5,89,89.5,90,90.5,91,91.5,92,92.5,93,93.5,94,94.5,95,95.5,96,96.5,97,97.5,98,98.5,99,99.5,100,100.5,101,101.5,102,102.5,103,103.5,104,104.5,105,105.5,106,106.5,107,107.5,108,108.5,109,109.5,110,110.5,111,111.5,112,112.5,113,113.5,114,114.5,115,115.5,116,116.5,117,117.5,118,118.5,119,119.5,120,120.5,121,121.5,122,122.5,123,123.5,124,124.5,125,125.5,126,126.5,127,127.5,128,128.5,129,129.5,130,130.5,131,131.5,132,132.5,133,133.5,134,134.5,135,135.5,136,136.5,137,137.5,138,138.5,139,139.5,140,140.5,141,141.5,142,142.5,143,143.5,144,144.5,145,145.5,146,146.5,147,147.5,148,148.5,149,149.5,150,150.5,151,151.5,152,152.5,153,153.5,154,154.5,155,155.5,156,156.5,157,157.5,158,158.5,159,159.5,160,160.5,161,161.5,162,162.5,163,163.5,164,164.5,165,165.5,166,166.5,167,167.5,168,168.5,169,169.5,170,170.5,171,171.5,172,172.5,173,173.5,174,174.5,175,175.5,176,176.5,177,177.5,178,178.5,179,179.5,180,180.5,181,181.5,182,182.5,183,183.5,184,184.5,185,185.5,186,186.5,187,187.5,188,188.5,189,189.5,190,190.5,191,191.5,192,192.5] +a16_1: [1,1.5,2,2.5,3,3.5,4,4.5,5,5.5,6,6.5,7,7.5,8,8.5,9,9.5,10,10.5,11,11.5,12,12.5,13,13.5,14,14.5,15,15.5,16,16.5,17,17.5,18,18.5,19,19.5,20,20.5,21,21.5,22,22.5,23,23.5,24,24.5,25,25.5,26,26.5,27,27.5,28,28.5,29,29.5,30,30.5,31,31.5,32,32.5,33,33.5,34,34.5,35,35.5,36,36.5,37,37.5,38,38.5,39,39.5,40,40.5,41,41.5,42,42.5,43,43.5,44,44.5,45,45.5,46,46.5,47,47.5,48,48.5,49,49.5,50,50.5,51,51.5,52,52.5,53,53.5,54,54.5,55,55.5,56,56.5,57,57.5,58,58.5,59,59.5,60,60.5,61,61.5,62,62.5,63,63.5,64,64.5,65,65.5,66,66.5,67,67.5,68,68.5,69,69.5,70,70.5,71,71.5,72,72.5,73,73.5,74,74.5,75,75.5,76,76.5,77,77.5,78,78.5,79,79.5,80,80.5,81,81.5,82,82.5,83,83.5,84,84.5,85,85.5,86,86.5,87,87.5,88,88.5,89,89.5,90,90.5,91,91.5,92,92.5,93,93.5,94,94.5,95,95.5,96,96.5,97,97.5,98,98.5,99,99.5,100,100.5,101,101.5,102,102.5,103,103.5,104,104.5,105,105.5,106,106.5,107,107.5,108,108.5,109,109.5,110,110.5,111,111.5,112,112.5,113,113.5,114,114.5,115,115.5,116,116.5,117,117.5,118,118.5,119,119.5,120,120.5,121,121.5,122,122.5,123,123.5,124,124.5,125,125.5,126,126.5,127,127.5,128,128.5,129,129,130,130,131,131,132,132,133,133,134,134,135,135,136,136,137,137,138,138,139,139,140,140,141,141,142,142,143,143,144,144,145,145,146,146,147,147,148,148,149,149,150,150,151,151,152,152,153,153,154,154,155,155,156,156,157,157,158,158,159,159,160,160,161,161,162,162,163,163,164,164,165,165,166,166,167,167,168,168,169,169,170,170,171,171,172,172,173,173,174,174,175,175,176,176,177,177,178,178,179,179,180,180,181,181,182,182,183,183,184,184,185,185,186,186,187,187,188,188,189,189,190,190,191,191,192,192] +dotProduct(a32, a32_1): 4736944 -- 4.74 million +dotProduct(a16, a16_1): 4726688 -- 4.73 million +cosineDistance(a32, a32_1): 0.000010093636084174129 +cosineDistance(a16, a16_1): 0.00001010226319664298 +L2Distance(a32, a32_1): 19.595917942265423 +L2Distance(a16, a16_1): 19.595917942265423 +L1Distance(a32, a32_1): 384 +L1Distance(a16, a16_1): 384 +LinfDistance(a32, a32_1): 1 +LinfDistance(a16, a16_1): 1 +LpDistance(a32, a32_1, 5): 3.2875036590344515 +LpDistance(a16, a16_1, 5): 3.2875036590344515 diff --git a/tests/queries/0_stateless/03269_bf16.sql b/tests/queries/0_stateless/03269_bf16.sql new file mode 100644 index 00000000000..375cca73b62 --- /dev/null +++ b/tests/queries/0_stateless/03269_bf16.sql @@ -0,0 +1,88 @@ +SET allow_experimental_bfloat16_type = 1; + +-- This is a smoke test, non exhaustive. + +-- Conversions + +SELECT + 1::BFloat16, + -1::BFloat16, + 1.1::BFloat16, + -1.1::BFloat16, + CAST(1 AS BFloat16), + CAST(-1 AS BFloat16), + CAST(1.1 AS BFloat16), + CAST(-1.1 AS BFloat16), + CAST(0xFFFFFFFFFFFFFFFF AS BFloat16), + CAST(-0.0 AS BFloat16), + CAST(inf AS BFloat16), + CAST(-inf AS BFloat16), + CAST(nan AS BFloat16); + +-- Conversions back + +SELECT + CAST(1.1::BFloat16 AS BFloat16), + CAST(1.1::BFloat16 AS Float32), + CAST(1.1::BFloat16 AS Float64), + CAST(1.1::BFloat16 AS Int8); + +-- Comparisons + +SELECT + 1.1::BFloat16 = 1.1::BFloat16, + 1.1::BFloat16 < 1.1, + 1.1::BFloat16 > 1.1, + 1.1::BFloat16 > 1, + 1.1::BFloat16 = 1.09375; + +-- Arithmetic + +SELECT + 1.1::BFloat16 - 1.1::BFloat16 AS a, + 1.1::BFloat16 + 1.1::BFloat16 AS b, + 1.1::BFloat16 * 1.1::BFloat16 AS c, + 1.1::BFloat16 / 1.1::BFloat16 AS d, + toTypeName(a), toTypeName(b), toTypeName(c), toTypeName(d); + +SELECT + 1.1::BFloat16 - 1.1 AS a, + 1.1 + 1.1::BFloat16 AS b, + 1.1::BFloat16 * 1.1 AS c, + 1.1 / 1.1::BFloat16 AS d, + toTypeName(a), toTypeName(b), toTypeName(c), toTypeName(d); + +-- Tables + +DROP TABLE IF EXISTS t; +CREATE TEMPORARY TABLE t (n UInt64, x BFloat16); +INSERT INTO t SELECT number, number FROM numbers(10000); +SELECT *, n = x, n - x FROM t WHERE n % 1000 = 0 ORDER BY n; + +-- Aggregate functions + +SELECT sum(n), sum(x), avg(n), avg(x), min(n), min(x), max(n), max(x), uniq(n), uniq(x), uniqExact(n), uniqExact(x) FROM t; + +-- MergeTree + +DROP TABLE t; +CREATE TABLE t (n UInt64, x BFloat16) ENGINE = MergeTree ORDER BY n; +INSERT INTO t SELECT number, number FROM numbers(10000); +SELECT *, n = x, n - x FROM t WHERE n % 1000 = 0 ORDER BY n; +SELECT sum(n), sum(x), avg(n), avg(x), min(n), min(x), max(n), max(x), uniq(n), uniq(x), uniqExact(n), uniqExact(x) FROM t; + +-- Distances + +WITH + arrayMap(x -> toFloat32(x) / 2, range(384)) AS a32, + arrayMap(x -> toBFloat16(x) / 2, range(384)) AS a16, + arrayMap(x -> x + 1, a32) AS a32_1, + arrayMap(x -> x + 1, a16) AS a16_1 +SELECT a32, a16, a32_1, a16_1, + dotProduct(a32, a32_1), dotProduct(a16, a16_1), + cosineDistance(a32, a32_1), cosineDistance(a16, a16_1), + L2Distance(a32, a32_1), L2Distance(a16, a16_1), + L1Distance(a32, a32_1), L1Distance(a16, a16_1), + LinfDistance(a32, a32_1), LinfDistance(a16, a16_1), + LpDistance(a32, a32_1, 5), LpDistance(a16, a16_1, 5) +FORMAT Vertical; From bfeefa2c8a5ce71dd0cc90d68d831d694aef3418 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 01:02:10 +0100 Subject: [PATCH 121/244] Introspection --- src/Functions/FunctionsBinaryRepresentation.cpp | 1 + tests/queries/0_stateless/03269_bf16.reference | 1 + tests/queries/0_stateless/03269_bf16.sql | 7 +++++++ 3 files changed, 9 insertions(+) diff --git a/src/Functions/FunctionsBinaryRepresentation.cpp b/src/Functions/FunctionsBinaryRepresentation.cpp index c8e8f167e4c..50a3c0862f4 100644 --- a/src/Functions/FunctionsBinaryRepresentation.cpp +++ b/src/Functions/FunctionsBinaryRepresentation.cpp @@ -296,6 +296,7 @@ public: tryExecuteUIntOrInt(column, res_column) || tryExecuteString(column, res_column) || tryExecuteFixedString(column, res_column) || + tryExecuteFloat(column, res_column) || tryExecuteFloat(column, res_column) || tryExecuteFloat(column, res_column) || tryExecuteDecimal(column, res_column) || diff --git a/tests/queries/0_stateless/03269_bf16.reference b/tests/queries/0_stateless/03269_bf16.reference index daa26cb252f..31395d92e2b 100644 --- a/tests/queries/0_stateless/03269_bf16.reference +++ b/tests/queries/0_stateless/03269_bf16.reference @@ -43,3 +43,4 @@ LinfDistance(a32, a32_1): 1 LinfDistance(a16, a16_1): 1 LpDistance(a32, a32_1, 5): 3.2875036590344515 LpDistance(a16, a16_1, 5): 3.2875036590344515 +1.09375 8C3F 1000110000111111 2 16268 8C3F diff --git a/tests/queries/0_stateless/03269_bf16.sql b/tests/queries/0_stateless/03269_bf16.sql index 375cca73b62..de4e2f6da47 100644 --- a/tests/queries/0_stateless/03269_bf16.sql +++ b/tests/queries/0_stateless/03269_bf16.sql @@ -86,3 +86,10 @@ SELECT a32, a16, a32_1, a16_1, LinfDistance(a32, a32_1), LinfDistance(a16, a16_1), LpDistance(a32, a32_1, 5), LpDistance(a16, a16_1, 5) FORMAT Vertical; + +-- Introspection + +SELECT 1.1::BFloat16 AS x, + hex(x), bin(x), + byteSize(x), + reinterpretAsUInt16(x), hex(reinterpretAsString(x)); From 3e50cf94fe858e8440ffd69040334356326b97db Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 01:04:55 +0100 Subject: [PATCH 122/244] Rounding --- tests/queries/0_stateless/03269_bf16.reference | 1 + tests/queries/0_stateless/03269_bf16.sql | 5 +++++ 2 files changed, 6 insertions(+) diff --git a/tests/queries/0_stateless/03269_bf16.reference b/tests/queries/0_stateless/03269_bf16.reference index 31395d92e2b..896cc307623 100644 --- a/tests/queries/0_stateless/03269_bf16.reference +++ b/tests/queries/0_stateless/03269_bf16.reference @@ -44,3 +44,4 @@ LinfDistance(a16, a16_1): 1 LpDistance(a32, a32_1, 5): 3.2875036590344515 LpDistance(a16, a16_1, 5): 3.2875036590344515 1.09375 8C3F 1000110000111111 2 16268 8C3F +1.09375 1 1.09375 1.0859375 0 diff --git a/tests/queries/0_stateless/03269_bf16.sql b/tests/queries/0_stateless/03269_bf16.sql index de4e2f6da47..b332a6e3119 100644 --- a/tests/queries/0_stateless/03269_bf16.sql +++ b/tests/queries/0_stateless/03269_bf16.sql @@ -93,3 +93,8 @@ SELECT 1.1::BFloat16 AS x, hex(x), bin(x), byteSize(x), reinterpretAsUInt16(x), hex(reinterpretAsString(x)); + +-- Rounding (this could be not towards the nearest) + +SELECT 1.1::BFloat16 AS x, + round(x), round(x, 1), round(x, 2), round(x, -1); From 3a855f501cd5d16ff97e9dde8b6fcb2d3b7ae497 Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Mon, 11 Nov 2024 02:15:31 +0100 Subject: [PATCH 123/244] Cleanups --- base/base/DecomposedFloat.h | 2 +- base/base/wide_integer.h | 1 - base/base/wide_integer_impl.h | 8 +------- 3 files changed, 2 insertions(+), 9 deletions(-) diff --git a/base/base/DecomposedFloat.h b/base/base/DecomposedFloat.h index 3bd059cb21c..fef91adefb0 100644 --- a/base/base/DecomposedFloat.h +++ b/base/base/DecomposedFloat.h @@ -230,4 +230,4 @@ struct DecomposedFloat using DecomposedFloat64 = DecomposedFloat; using DecomposedFloat32 = DecomposedFloat; -using DecomposedFloat16 = DecomposedFloat<__bf16>; +using DecomposedFloat16 = DecomposedFloat; diff --git a/base/base/wide_integer.h b/base/base/wide_integer.h index baf6e490ada..f3a4dc9e6d5 100644 --- a/base/base/wide_integer.h +++ b/base/base/wide_integer.h @@ -118,7 +118,6 @@ public: constexpr operator long double() const noexcept; constexpr operator double() const noexcept; constexpr operator float() const noexcept; - constexpr operator __bf16() const noexcept; struct _impl; diff --git a/base/base/wide_integer_impl.h b/base/base/wide_integer_impl.h index d0bbd7df9d4..3787971a20e 100644 --- a/base/base/wide_integer_impl.h +++ b/base/base/wide_integer_impl.h @@ -154,7 +154,7 @@ struct common_type, Arithmetic> static_assert(wide::ArithmeticConcept()); using type = std::conditional_t< - std::is_floating_point_v || std::is_same_v, + std::is_floating_point_v, Arithmetic, std::conditional_t< sizeof(Arithmetic) * 8 < Bits, @@ -1300,12 +1300,6 @@ constexpr integer::operator float() const noexcept return static_cast(static_cast(*this)); } -template -constexpr integer::operator __bf16() const noexcept -{ - return static_cast<__bf16>(static_cast(*this)); -} - // Unary operators template constexpr integer operator~(const integer & lhs) noexcept From ab79efe40f8785a7bd947cd5919feafedfb88259 Mon Sep 17 00:00:00 2001 From: kevinyhzou Date: Mon, 11 Nov 2024 14:07:19 +0800 Subject: [PATCH 124/244] make scale argument not optional --- .../functions/type-conversion-functions.md | 4 +- src/Functions/parseDateTime.cpp | 130 +++++++----------- .../03252_parse_datetime64.reference | 13 +- .../0_stateless/03252_parse_datetime64.sql | 38 ++--- ..._parse_datetime64_in_joda_syntax.reference | 5 - .../03252_parse_datetime64_in_joda_syntax.sql | 12 +- 6 files changed, 69 insertions(+), 133 deletions(-) diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md index 8043b21744a..72e6fda03f7 100644 --- a/docs/en/sql-reference/functions/type-conversion-functions.md +++ b/docs/en/sql-reference/functions/type-conversion-functions.md @@ -6880,7 +6880,7 @@ parseDateTime64(str[, [scale, [format[, timezone]]]]) **Arguments** - `str` — The String to be parsed -- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default 6 if not specified. +- `scale` - The scale of [DateTime64](../data-types/datetime64.md). - `format` — The format string. Optional. `%Y-%m-%d %H:%i:%s` if not specified. - `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. @@ -6907,7 +6907,7 @@ parseDateTime64InJodaSyntax(str[, [scale, [format[, timezone]]]]) **Arguments** - `str` — The String to be parsed -- `scale` - The precision for [DateTime64](../data-types/datetime64.md). Optional, default 0 if not specified. +- `scale` - The scale of [DateTime64](../data-types/datetime64.md). - `format` — The format string. Optional. `yyyy-MM-dd HH:mm:ss` if not specified. - `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. diff --git a/src/Functions/parseDateTime.cpp b/src/Functions/parseDateTime.cpp index 7190c1ad6f8..72e3ba661ca 100644 --- a/src/Functions/parseDateTime.cpp +++ b/src/Functions/parseDateTime.cpp @@ -607,87 +607,71 @@ namespace DataTypePtr getReturnTypeImpl(const ColumnsWithTypeAndName & arguments) const override { - FunctionArgumentDescriptors mandatory_args{ - {"time", static_cast(&isString), nullptr, "String"} - }; - + FunctionArgumentDescriptors mandatory_args; FunctionArgumentDescriptors optional_args; if constexpr (return_type == ReturnType::DateTime64) - optional_args = {{"scale/format", static_cast( - [](const IDataType & data_type) -> bool { return isUInt(data_type) || isString(data_type); } - ), nullptr, "UInt or String"}, - {"format", static_cast(&isString), nullptr, "String"}, - {"timezone", static_cast(&isString), &isColumnConst, "const String"} + { + mandatory_args = { + {"time", static_cast(&isString), nullptr, "String"}, + {"scale", static_cast(&isUInt8), nullptr, "UInt8"} }; - else optional_args = { {"format", static_cast(&isString), nullptr, "String"}, {"timezone", static_cast(&isString), &isColumnConst, "const String"} }; + } + else + { + mandatory_args = {{"time", static_cast(&isString), nullptr, "String"}}; + optional_args = { + {"format", static_cast(&isString), nullptr, "String"}, + {"timezone", static_cast(&isString), &isColumnConst, "const String"} + }; + } validateFunctionArguments(*this, arguments, mandatory_args, optional_args); String time_zone_name = getTimeZone(arguments).getTimeZone(); DataTypePtr data_type; if constexpr (return_type == ReturnType::DateTime64) { - UInt32 scale = 0; - if (arguments.size() == 1) + UInt8 scale = 0; + if (isUInt8(arguments[1].type)) { - /// In MySQL parse syntax, the scale of microseond is 6. - if constexpr (parse_syntax == ParseSyntax::MySQL) - scale = 6; - } - else - { - if (isUInt(arguments[1].type)) - { - const auto * col_scale = checkAndGetColumnConst(arguments[1].column.get()); - if (col_scale) - scale = col_scale->getValue(); - else - throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The input scale value may exceed the max scale value of `DateTime64`: {}.", - maxScaleOfDateTime64); - } + const auto * col_scale = checkAndGetColumnConst(arguments[1].column.get()); + if (col_scale) + scale = col_scale->getValue(); else - { - if constexpr (parse_syntax == ParseSyntax::MySQL) - scale = 6; - } + throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale argument is not Const(UInt8) type."); + } + if (parse_syntax == ParseSyntax::MySQL && scale != 6) + throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale value {} of MySQL parse syntax is not 6.", std::to_string(scale)); + if (scale > maxScaleOfDateTime64) + throw Exception(ErrorCodes::BAD_ARGUMENTS, + "The scale argument's value {} exceed the max scale value {}.", std::to_string(scale), std::to_string(maxScaleOfDateTime64)); - /// Construct the return type `DataTypDateTime64` with scale and time zone name. The scale value can be specified or be extracted - /// from the format string by counting how many 'S' characters are contained in the format's microsceond fragment. - String format = getFormat(arguments, scale); - std::vector instructions = parseFormat(format); - for (const auto & instruction : instructions) + String format = getFormat(arguments, scale); + std::vector instructions = parseFormat(format); + for (const auto & instruction : instructions) + { + /// Check scale by counting how may 'S' characters exists in the format string. + const String & fragment = instruction.getFragment(); + UInt32 s_cnt = 0; + for (char ch : fragment) { - const String & fragment = instruction.getFragment(); - UInt32 val = 0; - for (char ch : fragment) + if (ch != 'S') { - if (ch != 'S') - { - val = 0; - break; - } - else - val++; + s_cnt = 0; + break; } - /// If the scale is already specified by the second argument, but it not equals the value that extract from the format string, - /// then we should throw an exception; If the scale is not specified, then we should set its value as the extracted one. - if (val != 0 && scale != 0 && val != scale) - throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The scale of input format string {} not equals the given scale value {}.", - format, - scale); - else if (scale == 0 && val != 0) - scale = val; + else + s_cnt++; } - if (scale > maxScaleOfDateTime64) + /// If the number of 'S' character in format string not euqals the scale, then throw an exception to report error. + if (s_cnt != 0 && s_cnt != scale) throw Exception(ErrorCodes::BAD_ARGUMENTS, - "The scale of the input format string {} exceed the max scale value {}.", + "The scale of input format string {} not equals the given scale value {}.", format, - maxScaleOfDateTime64); + std::to_string(scale)); } data_type = std::make_shared(scale, time_zone_name); } @@ -2267,18 +2251,7 @@ namespace { size_t format_arg_index = 1; if constexpr (return_type == ReturnType::DateTime64) - { - /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22.22.123', 3), then the format is treated - /// as default value `yyyy-MM-dd HH:mm:ss`. - /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 'yyyy-MM-dd HH:mm:ss.SSS')`, - /// then the second argument is the format. - /// When parse `DateTime64` like `parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-05 12:22:22.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS')`, - /// then the third argument is the format. - if (arguments.size() > 1 && isString(removeNullable(arguments[1].type))) - format_arg_index = 1; - else - format_arg_index = 2; - } + format_arg_index = 2; if (arguments.size() <= format_arg_index) { @@ -2311,18 +2284,11 @@ namespace const DateLUTImpl & getTimeZone(const ColumnsWithTypeAndName & arguments) const { - if (arguments.size() < 3) + if (return_type == ReturnType::DateTime && arguments.size() < 3) return DateLUT::instance(); - else if constexpr (return_type == ReturnType::DateTime64) - { - /// If the return type is DateTime64, and the second argument is UInt type for scale, then it has 2 reasonable situations: - /// the first like parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-07 17:27.30.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT+8') - /// the second like parseDateTime64[InJodaSyntax][OrZero/OrNull]('2024-11-07 17:27.30.123456', 6, '%Y-%m-%d %H:%i:%s.%f'). And for the - /// first one, we should return the last argument as its timezone, and for the second one, we should return the default time zone as - /// `DateLUT::instance()`. - if (isUInt(arguments[1].type) && arguments.size() < 4) - return DateLUT::instance(); - } + else if (return_type == ReturnType::DateTime64 && arguments.size() < 4) + return DateLUT::instance(); + size_t timezone_arg_index = arguments.size() - 1; const auto * col = checkAndGetColumnConst(arguments[timezone_arg_index].column.get()); if (!col) diff --git a/tests/queries/0_stateless/03252_parse_datetime64.reference b/tests/queries/0_stateless/03252_parse_datetime64.reference index 27dcef6bf68..263c9b5d8ea 100644 --- a/tests/queries/0_stateless/03252_parse_datetime64.reference +++ b/tests/queries/0_stateless/03252_parse_datetime64.reference @@ -1,17 +1,8 @@ -2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 +2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 1970-01-01 08:00:00.000000 -1970-01-01 08:00:00.000 -1970-01-01 08:00:00.000 2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 -2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 -1970-01-01 08:00:00.000 2024-10-09 10:30:10.123456 -\N -\N -\N -2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 -2024-10-09 10:30:10.123456 2024-10-09 10:30:10.123456 -\N +1970-01-01 08:00:00.000000 diff --git a/tests/queries/0_stateless/03252_parse_datetime64.sql b/tests/queries/0_stateless/03252_parse_datetime64.sql index d28b6e586f7..2a6ef254887 100644 --- a/tests/queries/0_stateless/03252_parse_datetime64.sql +++ b/tests/queries/0_stateless/03252_parse_datetime64.sql @@ -1,32 +1,22 @@ set session_timezone = 'Asia/Shanghai'; -select parseDateTime64('2024-10-09 10:30:10.123456'); -select parseDateTime64('2024-10-09 10:30:10.123'); -- { serverError NOT_ENOUGH_SPACE } -select parseDateTime64('2024-10-09 10:30:10', 3); -- { serverError NOT_ENOUGH_SPACE } -select parseDateTime64('2024-10-09 10:30:10.', 3); -- { serverError CANNOT_PARSE_DATETIME } +select parseDateTime64('2024-10-09 10:30:10.123456'); -- {serverError NUMBER_OF_ARGUMENTS_DOESNT_MATCH} +select parseDateTime64('2024-10-09 10:30:10', 3); -- { serverError BAD_ARGUMENTS } select parseDateTime64('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} -select parseDateTime64('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } -select parseDateTime64('2024-10-09 10:30:10.123456', 6), parseDateTime64('2024-10-09 10:30:10.123456', '%Y-%m-%d %H:%i:%s.%f'); -select parseDateTime64('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'), parseDateTime64('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7'); -select parseDateTime64('2024-10-09 10:30:10.123', 3, '%Y-%m-%d %H:%i:%s.%f'); -- { serverError CANNOT_PARSE_DATETIME } +select parseDateTime64('2024-10-09 10:30:10.123456', 6), parseDateTime64('2024-10-09 10:30:10.123456', 6,'%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7'); select parseDateTime64('2024-10-09 10:30:10.123', 6, '%Y-%m-%d %H:%i:%s.%f'); -- { serverError NOT_ENOUGH_SPACE } -select parseDateTime64OrZero('2024-10-09 10:30:10.123456'); -select parseDateTime64OrZero('2024-10-09 10:30:10.123'); -select parseDateTime64OrZero('2024-10-09 10:30:10', 3); -select parseDateTime64OrZero('2024-10-09 10:30:10.', 3); +select parseDateTime64OrZero('2024-10-09 10:30:10.123456'); -- {serverError NUMBER_OF_ARGUMENTS_DOESNT_MATCH} +select parseDateTime64OrZero('2024-10-09 10:30:10', 3); -- { serverError BAD_ARGUMENTS } select parseDateTime64OrZero('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} -select parseDateTime64OrZero('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } -select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6), parseDateTime64OrZero('2024-10-09 10:30:10.123456', '%Y-%m-%d %H:%i:%s.%f'); -select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'), parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7'); -select parseDateTime64OrZero('2024-10-09 10:30:10.123', 3, '%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6), parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7'); +select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%fffff'); -select parseDateTime64OrNull('2024-10-09 10:30:10.123456'); -select parseDateTime64OrNull('2024-10-09 10:30:10.123'); -select parseDateTime64OrNull('2024-10-09 10:30:10', 3); -select parseDateTime64OrNull('2024-10-09 10:30:10.', 3); +select parseDateTime64OrNull('2024-10-09 10:30:10.123456'); -- {serverError NUMBER_OF_ARGUMENTS_DOESNT_MATCH} +select parseDateTime64OrNull('2024-10-09 10:30:10', 3); -- { serverError BAD_ARGUMENTS } select parseDateTime64OrNull('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} -select parseDateTime64OrNull('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } -select parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6), parseDateTime64OrZero('2024-10-09 10:30:10.123456', '%Y-%m-%d %H:%i:%s.%f'); -select parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f'), parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7');; -select parseDateTime64OrNull('2024-10-09 10:30:10.123', 3, '%Y-%m-%d %H:%i:%s.%f'); \ No newline at end of file +select parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6), parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6,'%Y-%m-%d %H:%i:%s.%f'); +select parseDateTime64OrNull('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%f', 'Etc/GMT-7');; +select parseDateTime64OrZero('2024-10-09 10:30:10.123456', 6, '%Y-%m-%d %H:%i:%s.%fffff'); \ No newline at end of file diff --git a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference index 0b4a28c4b38..99252ce55ca 100644 --- a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference +++ b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.reference @@ -1,12 +1,9 @@ 2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 -2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-10 02:42:10.123456 2024-10-10 01:30:10.123456 2024-10-10 01:30:10.123456 1970-01-01 08:00:00.000 -1970-01-01 08:00:00.000 -2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-10 02:42:10.123456 @@ -15,8 +12,6 @@ 2024-10-10 01:30:10.123456 1970-01-01 08:00:00.000000 \N -\N -2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-09 10:30:10.123 2024-10-09 10:30:10.000123 2024-10-10 02:42:10.123456 diff --git a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql index 8482677e9c9..bcb0fb5a362 100644 --- a/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql +++ b/tests/queries/0_stateless/03252_parse_datetime64_in_joda_syntax.sql @@ -1,11 +1,9 @@ set session_timezone = 'Asia/Shanghai'; select parseDateTime64InJodaSyntax('2024-10-09 10:30:10', 3); -- { serverError NOT_ENOUGH_SPACE } -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.', 3); -- { serverError CANNOT_PARSE_DATETIME } select parseDateTime64InJodaSyntax('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 3), parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 6); -select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSS'); select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSSS'); -- { serverError BAD_ARGUMENTS } select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456-0812', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); @@ -17,11 +15,9 @@ select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456-8000', 6, 'yyyy-M select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123456ABCD', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); -- { serverError BAD_ARGUMENTS } select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10', 3); -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.', 3); select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 3), parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 6); -select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSS'); select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSSS'); -- { serverError BAD_ARGUMENTS } select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456-0812', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); @@ -33,11 +29,9 @@ select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456-8000', 6, ' select parseDateTime64InJodaSyntaxOrZero('2024-10-09 10:30:10.123456ABCD', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSz'); -- { serverError BAD_ARGUMENTS } select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10', 3); -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.', 3); select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10', -3); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10', 9); -- { serverError BAD_ARGUMENTS } +select parseDateTime64InJodaSyntax('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'); -- {serverError ILLEGAL_TYPE_OF_ARGUMENT} select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 3), parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 6); -select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 'yyyy-MM-dd HH:mm:ss.SSSSSS'); select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSS'), parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSS'); select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123', 3, 'yyyy-MM-dd HH:mm:ss.SSSS'); -- { serverError BAD_ARGUMENTS } select parseDateTime64InJodaSyntaxOrNull('2024-10-09 10:30:10.123456-0812', 6, 'yyyy-MM-dd HH:mm:ss.SSSSSSZ'); From 0768e0b265dc3a7d83d3a4c3ea9ba8625fe70994 Mon Sep 17 00:00:00 2001 From: kevinyhzou Date: Mon, 11 Nov 2024 14:26:58 +0800 Subject: [PATCH 125/244] update doc & comments --- .../functions/type-conversion-functions.md | 10 ++++----- src/Functions/parseDateTime.cpp | 22 ++++++------------- 2 files changed, 12 insertions(+), 20 deletions(-) diff --git a/docs/en/sql-reference/functions/type-conversion-functions.md b/docs/en/sql-reference/functions/type-conversion-functions.md index 72e6fda03f7..a92d7055fd5 100644 --- a/docs/en/sql-reference/functions/type-conversion-functions.md +++ b/docs/en/sql-reference/functions/type-conversion-functions.md @@ -6874,14 +6874,14 @@ Converts a [String](../data-types/string.md) to [DateTime64](../data-types/datet **Syntax** ``` sql -parseDateTime64(str[, [scale, [format[, timezone]]]]) +parseDateTime64(str, scale, [format[, timezone]]) ``` **Arguments** -- `str` — The String to be parsed +- `str` — The String to be parsed. - `scale` - The scale of [DateTime64](../data-types/datetime64.md). -- `format` — The format string. Optional. `%Y-%m-%d %H:%i:%s` if not specified. +- `format` — The format string. Optional. `%Y-%m-%d %H:%i:%s.%f` if not specified. - `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. **Returned value(s)** @@ -6901,12 +6901,12 @@ Converts a [String](../data-types/string.md) to [DateTime64](../data-types/datet **Syntax** ``` sql -parseDateTime64InJodaSyntax(str[, [scale, [format[, timezone]]]]) +parseDateTime64InJodaSyntax(str, scale, [format[, timezone]]) ``` **Arguments** -- `str` — The String to be parsed +- `str` — The String to be parsed. - `scale` - The scale of [DateTime64](../data-types/datetime64.md). - `format` — The format string. Optional. `yyyy-MM-dd HH:mm:ss` if not specified. - `timezone` — [Timezone](/docs/en/operations/server-configuration-parameters/settings.md#timezone). Optional. diff --git a/src/Functions/parseDateTime.cpp b/src/Functions/parseDateTime.cpp index 72e3ba661ca..9fea8a4f130 100644 --- a/src/Functions/parseDateTime.cpp +++ b/src/Functions/parseDateTime.cpp @@ -608,26 +608,18 @@ namespace DataTypePtr getReturnTypeImpl(const ColumnsWithTypeAndName & arguments) const override { FunctionArgumentDescriptors mandatory_args; - FunctionArgumentDescriptors optional_args; if constexpr (return_type == ReturnType::DateTime64) - { mandatory_args = { {"time", static_cast(&isString), nullptr, "String"}, - {"scale", static_cast(&isUInt8), nullptr, "UInt8"} + {"scale", static_cast(&isUInt8), &isColumnConst, "UInt8"} }; - optional_args = { - {"format", static_cast(&isString), nullptr, "String"}, - {"timezone", static_cast(&isString), &isColumnConst, "const String"} - }; - } else - { mandatory_args = {{"time", static_cast(&isString), nullptr, "String"}}; - optional_args = { - {"format", static_cast(&isString), nullptr, "String"}, - {"timezone", static_cast(&isString), &isColumnConst, "const String"} - }; - } + + FunctionArgumentDescriptors optional_args{ + {"format", static_cast(&isString), nullptr, "String"}, + {"timezone", static_cast(&isString), &isColumnConst, "const String"} + }; validateFunctionArguments(*this, arguments, mandatory_args, optional_args); String time_zone_name = getTimeZone(arguments).getTimeZone(); @@ -644,7 +636,7 @@ namespace throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale argument is not Const(UInt8) type."); } if (parse_syntax == ParseSyntax::MySQL && scale != 6) - throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale value {} of MySQL parse syntax is not 6.", std::to_string(scale)); + throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale argument's value {} of MySQL parse syntax is not 6.", std::to_string(scale)); if (scale > maxScaleOfDateTime64) throw Exception(ErrorCodes::BAD_ARGUMENTS, "The scale argument's value {} exceed the max scale value {}.", std::to_string(scale), std::to_string(maxScaleOfDateTime64)); From 206bd174c37a7e6ea47eda9c228c2aa6a5f2fff3 Mon Sep 17 00:00:00 2001 From: Vitaly Baranov Date: Sat, 2 Nov 2024 19:30:03 +0100 Subject: [PATCH 126/244] Corrections after reworking backup/restore synchronization. --- src/Backups/BackupConcurrencyCheck.cpp | 16 +- src/Backups/BackupConcurrencyCheck.h | 11 +- src/Backups/BackupCoordinationCleaner.cpp | 36 +- src/Backups/BackupCoordinationCleaner.h | 17 +- src/Backups/BackupCoordinationLocal.cpp | 3 +- src/Backups/BackupCoordinationLocal.h | 13 +- src/Backups/BackupCoordinationOnCluster.cpp | 95 +- src/Backups/BackupCoordinationOnCluster.h | 20 +- src/Backups/BackupCoordinationStageSync.cpp | 895 ++++++++++++------- src/Backups/BackupCoordinationStageSync.h | 97 +- src/Backups/BackupsWorker.cpp | 108 ++- src/Backups/BackupsWorker.h | 2 - src/Backups/IBackupCoordination.h | 28 +- src/Backups/IRestoreCoordination.h | 28 +- src/Backups/RestoreCoordinationLocal.cpp | 4 +- src/Backups/RestoreCoordinationLocal.h | 14 +- src/Backups/RestoreCoordinationOnCluster.cpp | 95 +- src/Backups/RestoreCoordinationOnCluster.h | 20 +- 18 files changed, 887 insertions(+), 615 deletions(-) diff --git a/src/Backups/BackupConcurrencyCheck.cpp b/src/Backups/BackupConcurrencyCheck.cpp index 8b29ae41b53..a67d241845d 100644 --- a/src/Backups/BackupConcurrencyCheck.cpp +++ b/src/Backups/BackupConcurrencyCheck.cpp @@ -14,12 +14,12 @@ namespace ErrorCodes BackupConcurrencyCheck::BackupConcurrencyCheck( - const UUID & backup_or_restore_uuid_, bool is_restore_, bool on_cluster_, + const String & zookeeper_path_, bool allow_concurrency_, BackupConcurrencyCounters & counters_) - : is_restore(is_restore_), backup_or_restore_uuid(backup_or_restore_uuid_), on_cluster(on_cluster_), counters(counters_) + : is_restore(is_restore_), on_cluster(on_cluster_), zookeeper_path(zookeeper_path_), counters(counters_) { std::lock_guard lock{counters.mutex}; @@ -32,7 +32,7 @@ BackupConcurrencyCheck::BackupConcurrencyCheck( size_t num_on_cluster_restores = counters.on_cluster_restores.size(); if (on_cluster) { - if (!counters.on_cluster_restores.contains(backup_or_restore_uuid)) + if (!counters.on_cluster_restores.contains(zookeeper_path)) ++num_on_cluster_restores; } else @@ -47,7 +47,7 @@ BackupConcurrencyCheck::BackupConcurrencyCheck( size_t num_on_cluster_backups = counters.on_cluster_backups.size(); if (on_cluster) { - if (!counters.on_cluster_backups.contains(backup_or_restore_uuid)) + if (!counters.on_cluster_backups.contains(zookeeper_path)) ++num_on_cluster_backups; } else @@ -64,9 +64,9 @@ BackupConcurrencyCheck::BackupConcurrencyCheck( if (on_cluster) { if (is_restore) - ++counters.on_cluster_restores[backup_or_restore_uuid]; + ++counters.on_cluster_restores[zookeeper_path]; else - ++counters.on_cluster_backups[backup_or_restore_uuid]; + ++counters.on_cluster_backups[zookeeper_path]; } else { @@ -86,7 +86,7 @@ BackupConcurrencyCheck::~BackupConcurrencyCheck() { if (is_restore) { - auto it = counters.on_cluster_restores.find(backup_or_restore_uuid); + auto it = counters.on_cluster_restores.find(zookeeper_path); if (it != counters.on_cluster_restores.end()) { if (!--it->second) @@ -95,7 +95,7 @@ BackupConcurrencyCheck::~BackupConcurrencyCheck() } else { - auto it = counters.on_cluster_backups.find(backup_or_restore_uuid); + auto it = counters.on_cluster_backups.find(zookeeper_path); if (it != counters.on_cluster_backups.end()) { if (!--it->second) diff --git a/src/Backups/BackupConcurrencyCheck.h b/src/Backups/BackupConcurrencyCheck.h index 048a23a716a..a1baeff5464 100644 --- a/src/Backups/BackupConcurrencyCheck.h +++ b/src/Backups/BackupConcurrencyCheck.h @@ -1,7 +1,8 @@ #pragma once -#include +#include #include +#include #include #include @@ -19,9 +20,9 @@ public: /// Checks concurrency of a BACKUP operation or a RESTORE operation. /// Keep a constructed instance of BackupConcurrencyCheck until the operation is done. BackupConcurrencyCheck( - const UUID & backup_or_restore_uuid_, bool is_restore_, bool on_cluster_, + const String & zookeeper_path_, bool allow_concurrency_, BackupConcurrencyCounters & counters_); @@ -31,8 +32,8 @@ public: private: const bool is_restore; - const UUID backup_or_restore_uuid; const bool on_cluster; + const String zookeeper_path; BackupConcurrencyCounters & counters; }; @@ -47,8 +48,8 @@ private: friend class BackupConcurrencyCheck; size_t local_backups TSA_GUARDED_BY(mutex) = 0; size_t local_restores TSA_GUARDED_BY(mutex) = 0; - std::unordered_map on_cluster_backups TSA_GUARDED_BY(mutex); - std::unordered_map on_cluster_restores TSA_GUARDED_BY(mutex); + std::unordered_map on_cluster_backups TSA_GUARDED_BY(mutex); + std::unordered_map on_cluster_restores TSA_GUARDED_BY(mutex); std::mutex mutex; }; diff --git a/src/Backups/BackupCoordinationCleaner.cpp b/src/Backups/BackupCoordinationCleaner.cpp index 1f5068a94de..47095f27eb3 100644 --- a/src/Backups/BackupCoordinationCleaner.cpp +++ b/src/Backups/BackupCoordinationCleaner.cpp @@ -4,31 +4,29 @@ namespace DB { -BackupCoordinationCleaner::BackupCoordinationCleaner(const String & zookeeper_path_, const WithRetries & with_retries_, LoggerPtr log_) - : zookeeper_path(zookeeper_path_), with_retries(with_retries_), log(log_) +BackupCoordinationCleaner::BackupCoordinationCleaner(bool is_restore_, const String & zookeeper_path_, const WithRetries & with_retries_, LoggerPtr log_) + : is_restore(is_restore_), zookeeper_path(zookeeper_path_), with_retries(with_retries_), log(log_) { } -void BackupCoordinationCleaner::cleanup() +bool BackupCoordinationCleaner::cleanup(bool throw_if_error) { - tryRemoveAllNodes(/* throw_if_error = */ true, /* retries_kind = */ WithRetries::kNormal); + WithRetries::Kind retries_kind = throw_if_error ? WithRetries::kNormal : WithRetries::kErrorHandling; + return cleanupImpl(throw_if_error, retries_kind); } -bool BackupCoordinationCleaner::tryCleanupAfterError() noexcept -{ - return tryRemoveAllNodes(/* throw_if_error = */ false, /* retries_kind = */ WithRetries::kNormal); -} - -bool BackupCoordinationCleaner::tryRemoveAllNodes(bool throw_if_error, WithRetries::Kind retries_kind) +bool BackupCoordinationCleaner::cleanupImpl(bool throw_if_error, WithRetries::Kind retries_kind) { { std::lock_guard lock{mutex}; - if (cleanup_result.succeeded) - return true; - if (cleanup_result.exception) + if (succeeded) { - if (throw_if_error) - std::rethrow_exception(cleanup_result.exception); + LOG_TRACE(log, "Nodes from ZooKeeper are already removed"); + return true; + } + if (tried) + { + LOG_INFO(log, "Skipped removing nodes from ZooKeeper because because earlier we failed to do that"); return false; } } @@ -44,16 +42,18 @@ bool BackupCoordinationCleaner::tryRemoveAllNodes(bool throw_if_error, WithRetri }); std::lock_guard lock{mutex}; - cleanup_result.succeeded = true; + tried = true; + succeeded = true; return true; } catch (...) { - LOG_TRACE(log, "Caught exception while removing nodes from ZooKeeper for this restore: {}", + LOG_TRACE(log, "Caught exception while removing nodes from ZooKeeper for this {}: {}", + is_restore ? "restore" : "backup", getCurrentExceptionMessage(/* with_stacktrace= */ false, /* check_embedded_stacktrace= */ true)); std::lock_guard lock{mutex}; - cleanup_result.exception = std::current_exception(); + tried = true; if (throw_if_error) throw; diff --git a/src/Backups/BackupCoordinationCleaner.h b/src/Backups/BackupCoordinationCleaner.h index 43e095d9f33..c760a3611f9 100644 --- a/src/Backups/BackupCoordinationCleaner.h +++ b/src/Backups/BackupCoordinationCleaner.h @@ -12,14 +12,14 @@ namespace DB class BackupCoordinationCleaner { public: - BackupCoordinationCleaner(const String & zookeeper_path_, const WithRetries & with_retries_, LoggerPtr log_); + BackupCoordinationCleaner(bool is_restore_, const String & zookeeper_path_, const WithRetries & with_retries_, LoggerPtr log_); - void cleanup(); - bool tryCleanupAfterError() noexcept; + bool cleanup(bool throw_if_error); private: - bool tryRemoveAllNodes(bool throw_if_error, WithRetries::Kind retries_kind); + bool cleanupImpl(bool throw_if_error, WithRetries::Kind retries_kind); + const bool is_restore; const String zookeeper_path; /// A reference to a field of the parent object which is either BackupCoordinationOnCluster or RestoreCoordinationOnCluster. @@ -27,13 +27,8 @@ private: const LoggerPtr log; - struct CleanupResult - { - bool succeeded = false; - std::exception_ptr exception; - }; - CleanupResult cleanup_result TSA_GUARDED_BY(mutex); - + bool tried TSA_GUARDED_BY(mutex) = false; + bool succeeded TSA_GUARDED_BY(mutex) = false; std::mutex mutex; }; diff --git a/src/Backups/BackupCoordinationLocal.cpp b/src/Backups/BackupCoordinationLocal.cpp index 8bd6b4d327d..402e789eacb 100644 --- a/src/Backups/BackupCoordinationLocal.cpp +++ b/src/Backups/BackupCoordinationLocal.cpp @@ -11,12 +11,11 @@ namespace DB { BackupCoordinationLocal::BackupCoordinationLocal( - const UUID & backup_uuid_, bool is_plain_backup_, bool allow_concurrent_backup_, BackupConcurrencyCounters & concurrency_counters_) : log(getLogger("BackupCoordinationLocal")) - , concurrency_check(backup_uuid_, /* is_restore = */ false, /* on_cluster = */ false, allow_concurrent_backup_, concurrency_counters_) + , concurrency_check(/* is_restore = */ false, /* on_cluster = */ false, /* zookeeper_path = */ "", allow_concurrent_backup_, concurrency_counters_) , file_infos(is_plain_backup_) { } diff --git a/src/Backups/BackupCoordinationLocal.h b/src/Backups/BackupCoordinationLocal.h index 09991c0d301..e63fcde981a 100644 --- a/src/Backups/BackupCoordinationLocal.h +++ b/src/Backups/BackupCoordinationLocal.h @@ -23,20 +23,19 @@ class BackupCoordinationLocal : public IBackupCoordination { public: explicit BackupCoordinationLocal( - const UUID & backup_uuid_, bool is_plain_backup_, bool allow_concurrent_backup_, BackupConcurrencyCounters & concurrency_counters_); ~BackupCoordinationLocal() override; + void setBackupQueryIsSentToOtherHosts() override {} + bool isBackupQuerySentToOtherHosts() const override { return false; } Strings setStage(const String &, const String &, bool) override { return {}; } - void setBackupQueryWasSentToOtherHosts() override {} - bool trySetError(std::exception_ptr) override { return true; } - void finish() override {} - bool tryFinishAfterError() noexcept override { return true; } - void waitForOtherHostsToFinish() override {} - bool tryWaitForOtherHostsToFinishAfterError() noexcept override { return true; } + bool setError(std::exception_ptr, bool) override { return true; } + bool waitOtherHostsFinish(bool) const override { return true; } + bool finish(bool) override { return true; } + bool cleanup(bool) override { return true; } void addReplicatedPartNames(const String & table_zk_path, const String & table_name_for_logs, const String & replica_name, const std::vector & part_names_and_checksums) override; diff --git a/src/Backups/BackupCoordinationOnCluster.cpp b/src/Backups/BackupCoordinationOnCluster.cpp index dc34939f805..1b14f226eff 100644 --- a/src/Backups/BackupCoordinationOnCluster.cpp +++ b/src/Backups/BackupCoordinationOnCluster.cpp @@ -184,17 +184,21 @@ BackupCoordinationOnCluster::BackupCoordinationOnCluster( , plain_backup(is_plain_backup_) , log(getLogger("BackupCoordinationOnCluster")) , with_retries(log, get_zookeeper_, keeper_settings, process_list_element_, [root_zookeeper_path_](Coordination::ZooKeeperWithFaultInjection::Ptr zk) { zk->sync(root_zookeeper_path_); }) - , concurrency_check(backup_uuid_, /* is_restore = */ false, /* on_cluster = */ true, allow_concurrent_backup_, concurrency_counters_) - , stage_sync(/* is_restore = */ false, fs::path{zookeeper_path} / "stage", current_host, all_hosts, allow_concurrent_backup_, with_retries, schedule_, process_list_element_, log) - , cleaner(zookeeper_path, with_retries, log) + , cleaner(/* is_restore = */ false, zookeeper_path, with_retries, log) + , stage_sync(/* is_restore = */ false, fs::path{zookeeper_path} / "stage", current_host, all_hosts, allow_concurrent_backup_, concurrency_counters_, with_retries, schedule_, process_list_element_, log) { - createRootNodes(); + try + { + createRootNodes(); + } + catch (...) + { + stage_sync.setError(std::current_exception(), /* throw_if_error = */ false); + throw; + } } -BackupCoordinationOnCluster::~BackupCoordinationOnCluster() -{ - tryFinishImpl(); -} +BackupCoordinationOnCluster::~BackupCoordinationOnCluster() = default; void BackupCoordinationOnCluster::createRootNodes() { @@ -217,69 +221,52 @@ void BackupCoordinationOnCluster::createRootNodes() }); } +void BackupCoordinationOnCluster::setBackupQueryIsSentToOtherHosts() +{ + stage_sync.setQueryIsSentToOtherHosts(); +} + +bool BackupCoordinationOnCluster::isBackupQuerySentToOtherHosts() const +{ + return stage_sync.isQuerySentToOtherHosts(); +} + Strings BackupCoordinationOnCluster::setStage(const String & new_stage, const String & message, bool sync) { stage_sync.setStage(new_stage, message); - - if (!sync) - return {}; - - return stage_sync.waitForHostsToReachStage(new_stage, all_hosts_without_initiator); + if (sync) + return stage_sync.waitHostsReachStage(all_hosts_without_initiator, new_stage); + return {}; } -void BackupCoordinationOnCluster::setBackupQueryWasSentToOtherHosts() +bool BackupCoordinationOnCluster::setError(std::exception_ptr exception, bool throw_if_error) { - backup_query_was_sent_to_other_hosts = true; + return stage_sync.setError(exception, throw_if_error); } -bool BackupCoordinationOnCluster::trySetError(std::exception_ptr exception) +bool BackupCoordinationOnCluster::waitOtherHostsFinish(bool throw_if_error) const { - return stage_sync.trySetError(exception); + return stage_sync.waitOtherHostsFinish(throw_if_error); } -void BackupCoordinationOnCluster::finish() +bool BackupCoordinationOnCluster::finish(bool throw_if_error) { - bool other_hosts_also_finished = false; - stage_sync.finish(other_hosts_also_finished); - - if ((current_host == kInitiator) && (other_hosts_also_finished || !backup_query_was_sent_to_other_hosts)) - cleaner.cleanup(); + return stage_sync.finish(throw_if_error); } -bool BackupCoordinationOnCluster::tryFinishAfterError() noexcept +bool BackupCoordinationOnCluster::cleanup(bool throw_if_error) { - return tryFinishImpl(); -} - -bool BackupCoordinationOnCluster::tryFinishImpl() noexcept -{ - bool other_hosts_also_finished = false; - if (!stage_sync.tryFinishAfterError(other_hosts_also_finished)) - return false; - - if ((current_host == kInitiator) && (other_hosts_also_finished || !backup_query_was_sent_to_other_hosts)) + /// All the hosts must finish before we remove the coordination nodes. + bool expect_other_hosts_finished = stage_sync.isQuerySentToOtherHosts() || !stage_sync.isErrorSet(); + bool all_hosts_finished = stage_sync.finished() && (stage_sync.otherHostsFinished() || !expect_other_hosts_finished); + if (!all_hosts_finished) { - if (!cleaner.tryCleanupAfterError()) - return false; - } - - return true; -} - -void BackupCoordinationOnCluster::waitForOtherHostsToFinish() -{ - if ((current_host != kInitiator) || !backup_query_was_sent_to_other_hosts) - return; - stage_sync.waitForOtherHostsToFinish(); -} - -bool BackupCoordinationOnCluster::tryWaitForOtherHostsToFinishAfterError() noexcept -{ - if (current_host != kInitiator) + auto unfinished_hosts = expect_other_hosts_finished ? stage_sync.getUnfinishedHosts() : Strings{current_host}; + LOG_INFO(log, "Skipping removing nodes from ZooKeeper because hosts {} didn't finish", + BackupCoordinationStageSync::getHostsDesc(unfinished_hosts)); return false; - if (!backup_query_was_sent_to_other_hosts) - return true; - return stage_sync.tryWaitForOtherHostsToFinishAfterError(); + } + return cleaner.cleanup(throw_if_error); } ZooKeeperRetriesInfo BackupCoordinationOnCluster::getOnClusterInitializationKeeperRetriesInfo() const diff --git a/src/Backups/BackupCoordinationOnCluster.h b/src/Backups/BackupCoordinationOnCluster.h index 7369c2cc746..b439ab619d8 100644 --- a/src/Backups/BackupCoordinationOnCluster.h +++ b/src/Backups/BackupCoordinationOnCluster.h @@ -1,7 +1,6 @@ #pragma once #include -#include #include #include #include @@ -20,7 +19,7 @@ class BackupCoordinationOnCluster : public IBackupCoordination { public: /// Empty string as the current host is used to mark the initiator of a BACKUP ON CLUSTER query. - static const constexpr std::string_view kInitiator; + static const constexpr std::string_view kInitiator = BackupCoordinationStageSync::kInitiator; BackupCoordinationOnCluster( const UUID & backup_uuid_, @@ -37,13 +36,13 @@ public: ~BackupCoordinationOnCluster() override; + void setBackupQueryIsSentToOtherHosts() override; + bool isBackupQuerySentToOtherHosts() const override; Strings setStage(const String & new_stage, const String & message, bool sync) override; - void setBackupQueryWasSentToOtherHosts() override; - bool trySetError(std::exception_ptr exception) override; - void finish() override; - bool tryFinishAfterError() noexcept override; - void waitForOtherHostsToFinish() override; - bool tryWaitForOtherHostsToFinishAfterError() noexcept override; + bool setError(std::exception_ptr exception, bool throw_if_error) override; + bool waitOtherHostsFinish(bool throw_if_error) const override; + bool finish(bool throw_if_error) override; + bool cleanup(bool throw_if_error) override; void addReplicatedPartNames( const String & table_zk_path, @@ -110,11 +109,10 @@ private: const bool plain_backup; LoggerPtr const log; + /// The order is important: `stage_sync` must be initialized after `with_retries` and `cleaner`. const WithRetries with_retries; - BackupConcurrencyCheck concurrency_check; - BackupCoordinationStageSync stage_sync; BackupCoordinationCleaner cleaner; - std::atomic backup_query_was_sent_to_other_hosts = false; + BackupCoordinationStageSync stage_sync; mutable std::optional replicated_tables TSA_GUARDED_BY(replicated_tables_mutex); mutable std::optional replicated_access TSA_GUARDED_BY(replicated_access_mutex); diff --git a/src/Backups/BackupCoordinationStageSync.cpp b/src/Backups/BackupCoordinationStageSync.cpp index 9a05f9490c2..fcf09d7c315 100644 --- a/src/Backups/BackupCoordinationStageSync.cpp +++ b/src/Backups/BackupCoordinationStageSync.cpp @@ -42,9 +42,6 @@ namespace kCurrentVersion = 2, }; - - /// Empty string as the current host is used to mark the initiator of a BACKUP ON CLUSTER or RESTORE ON CLUSTER query. - const constexpr std::string_view kInitiator; } bool BackupCoordinationStageSync::HostInfo::operator ==(const HostInfo & other) const @@ -63,12 +60,32 @@ bool BackupCoordinationStageSync::State::operator ==(const State & other) const bool BackupCoordinationStageSync::State::operator !=(const State & other) const = default; +void BackupCoordinationStageSync::State::merge(const State & other) +{ + if (other.host_with_error && !host_with_error) + { + const String & host = *other.host_with_error; + host_with_error = host; + hosts.at(host).exception = other.hosts.at(host).exception; + } + + for (const auto & [host, other_host_info] : other.hosts) + { + auto & host_info = hosts.at(host); + host_info.stages.insert(other_host_info.stages.begin(), other_host_info.stages.end()); + if (other_host_info.finished) + host_info.finished = true; + } +} + + BackupCoordinationStageSync::BackupCoordinationStageSync( bool is_restore_, const String & zookeeper_path_, const String & current_host_, const Strings & all_hosts_, bool allow_concurrency_, + BackupConcurrencyCounters & concurrency_counters_, const WithRetries & with_retries_, ThreadPoolCallbackRunnerUnsafe schedule_, QueryStatusPtr process_list_element_, @@ -89,35 +106,29 @@ BackupCoordinationStageSync::BackupCoordinationStageSync( , max_attempts_after_bad_version(with_retries.getKeeperSettings().max_attempts_after_bad_version) , zookeeper_path(zookeeper_path_) , root_zookeeper_path(zookeeper_path.parent_path().parent_path()) - , operation_node_path(zookeeper_path.parent_path()) + , operation_zookeeper_path(zookeeper_path.parent_path()) , operation_node_name(zookeeper_path.parent_path().filename()) - , stage_node_path(zookeeper_path) , start_node_path(zookeeper_path / ("started|" + current_host)) , finish_node_path(zookeeper_path / ("finished|" + current_host)) , num_hosts_node_path(zookeeper_path / "num_hosts") + , error_node_path(zookeeper_path / "error") , alive_node_path(zookeeper_path / ("alive|" + current_host)) , alive_tracker_node_path(fs::path{root_zookeeper_path} / "alive_tracker") - , error_node_path(zookeeper_path / "error") , zk_nodes_changed(std::make_shared()) { - if ((zookeeper_path.filename() != "stage") || !operation_node_name.starts_with(is_restore ? "restore-" : "backup-") - || (root_zookeeper_path == operation_node_path)) - { - throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected path in ZooKeeper specified: {}", zookeeper_path); - } - initializeState(); createRootNodes(); try { + concurrency_check.emplace(is_restore, /* on_cluster = */ true, zookeeper_path, allow_concurrency, concurrency_counters_); createStartAndAliveNodes(); startWatchingThread(); } catch (...) { - trySetError(std::current_exception()); - tryFinishImpl(); + if (setError(std::current_exception(), /* throw_if_error = */ false)) + finish(/* throw_if_error = */ false); throw; } } @@ -125,7 +136,26 @@ BackupCoordinationStageSync::BackupCoordinationStageSync( BackupCoordinationStageSync::~BackupCoordinationStageSync() { - tryFinishImpl(); + /// Normally either finish() or setError() must be called. + if (!tried_to_finish) + { + if (state.host_with_error) + { + /// setError() was called and succeeded. + finish(/* throw_if_error = */ false); + } + else if (!tried_to_set_error) + { + /// Neither finish() nor setError() were called, it's a bug. + chassert(false, "~BackupCoordinationStageSync() is called without finish() or setError()"); + LOG_ERROR(log, "~BackupCoordinationStageSync() is called without finish() or setError()"); + } + } + + /// Normally the watching thread should be stopped already because the finish() function stops it. + /// However if an error happened then the watching thread can be still running, + /// so here in the destructor we have to ensure that it's stopped. + stopWatchingThread(); } @@ -137,6 +167,12 @@ void BackupCoordinationStageSync::initializeState() for (const String & host : all_hosts) state.hosts.emplace(host, HostInfo{.host = host, .last_connection_time = now, .last_connection_time_monotonic = monotonic_now}); + + if (!state.hosts.contains(current_host)) + throw Exception(ErrorCodes::LOGICAL_ERROR, "List of hosts must contain the current host"); + + if (!state.hosts.contains(String{kInitiator})) + throw Exception(ErrorCodes::LOGICAL_ERROR, "List of hosts must contain the initiator"); } @@ -179,6 +215,12 @@ String BackupCoordinationStageSync::getHostsDesc(const Strings & hosts) void BackupCoordinationStageSync::createRootNodes() { + if ((zookeeper_path.filename() != "stage") || !operation_node_name.starts_with(is_restore ? "restore-" : "backup-") + || (root_zookeeper_path == operation_zookeeper_path)) + { + throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected path in ZooKeeper specified: {}", zookeeper_path); + } + auto holder = with_retries.createRetriesControlHolder("BackupStageSync::createRootNodes", WithRetries::kInitialization); holder.retries_ctl.retryLoop( [&, &zookeeper = holder.faulty_zookeeper]() @@ -252,27 +294,27 @@ void BackupCoordinationStageSync::createStartAndAliveNodes(Coordination::ZooKeep Coordination::Requests requests; requests.reserve(6); - size_t operation_node_path_pos = static_cast(-1); - if (!zookeeper->exists(operation_node_path)) + size_t operation_node_pos = static_cast(-1); + if (!zookeeper->exists(operation_zookeeper_path)) { - operation_node_path_pos = requests.size(); - requests.emplace_back(zkutil::makeCreateRequest(operation_node_path, "", zkutil::CreateMode::Persistent)); + operation_node_pos = requests.size(); + requests.emplace_back(zkutil::makeCreateRequest(operation_zookeeper_path, "", zkutil::CreateMode::Persistent)); } - size_t stage_node_path_pos = static_cast(-1); - if (!zookeeper->exists(stage_node_path)) + size_t zookeeper_path_pos = static_cast(-1); + if (!zookeeper->exists(zookeeper_path)) { - stage_node_path_pos = requests.size(); - requests.emplace_back(zkutil::makeCreateRequest(stage_node_path, "", zkutil::CreateMode::Persistent)); + zookeeper_path_pos = requests.size(); + requests.emplace_back(zkutil::makeCreateRequest(zookeeper_path, "", zkutil::CreateMode::Persistent)); } - size_t num_hosts_node_path_pos = requests.size(); + size_t num_hosts_node_pos = requests.size(); if (num_hosts) requests.emplace_back(zkutil::makeSetRequest(num_hosts_node_path, toString(*num_hosts + 1), num_hosts_version)); else requests.emplace_back(zkutil::makeCreateRequest(num_hosts_node_path, "1", zkutil::CreateMode::Persistent)); - size_t alive_tracker_node_path_pos = requests.size(); + size_t alive_tracker_node_pos = requests.size(); requests.emplace_back(zkutil::makeSetRequest(alive_tracker_node_path, "", alive_tracker_version)); requests.emplace_back(zkutil::makeCreateRequest(start_node_path, std::to_string(kCurrentVersion), zkutil::CreateMode::Persistent)); @@ -284,7 +326,7 @@ void BackupCoordinationStageSync::createStartAndAliveNodes(Coordination::ZooKeep if (code == Coordination::Error::ZOK) { LOG_INFO(log, "Created start node #{} in ZooKeeper for {} (coordination version: {})", - num_hosts.value_or(0) + 1, current_host_desc, kCurrentVersion); + num_hosts.value_or(0) + 1, current_host_desc, static_cast(kCurrentVersion)); return; } @@ -294,40 +336,34 @@ void BackupCoordinationStageSync::createStartAndAliveNodes(Coordination::ZooKeep LOG_TRACE(log, "{} (attempt #{}){}", message, attempt_no, will_try_again ? ", will try again" : ""); }; - if ((responses.size() > operation_node_path_pos) && - (responses[operation_node_path_pos]->error == Coordination::Error::ZNODEEXISTS)) + if ((operation_node_pos < responses.size()) && + (responses[operation_node_pos]->error == Coordination::Error::ZNODEEXISTS)) { - show_error_before_next_attempt(fmt::format("Node {} in ZooKeeper already exists", operation_node_path)); + show_error_before_next_attempt(fmt::format("Node {} already exists", operation_zookeeper_path)); /// needs another attempt } - else if ((responses.size() > stage_node_path_pos) && - (responses[stage_node_path_pos]->error == Coordination::Error::ZNODEEXISTS)) + else if ((zookeeper_path_pos < responses.size()) && + (responses[zookeeper_path_pos]->error == Coordination::Error::ZNODEEXISTS)) { - show_error_before_next_attempt(fmt::format("Node {} in ZooKeeper already exists", stage_node_path)); + show_error_before_next_attempt(fmt::format("Node {} already exists", zookeeper_path)); /// needs another attempt } - else if ((responses.size() > num_hosts_node_path_pos) && num_hosts && - (responses[num_hosts_node_path_pos]->error == Coordination::Error::ZBADVERSION)) + else if ((num_hosts_node_pos < responses.size()) && !num_hosts && + (responses[num_hosts_node_pos]->error == Coordination::Error::ZNODEEXISTS)) { - show_error_before_next_attempt("Other host changed the 'num_hosts' node in ZooKeeper"); + show_error_before_next_attempt(fmt::format("Node {} already exists", num_hosts_node_path)); + /// needs another attempt + } + else if ((num_hosts_node_pos < responses.size()) && num_hosts && + (responses[num_hosts_node_pos]->error == Coordination::Error::ZBADVERSION)) + { + show_error_before_next_attempt(fmt::format("The version of node {} changed", num_hosts_node_path)); num_hosts.reset(); /// needs to reread 'num_hosts' again } - else if ((responses.size() > num_hosts_node_path_pos) && num_hosts && - (responses[num_hosts_node_path_pos]->error == Coordination::Error::ZNONODE)) + else if ((alive_tracker_node_pos < responses.size()) && + (responses[alive_tracker_node_pos]->error == Coordination::Error::ZBADVERSION)) { - show_error_before_next_attempt("Other host removed the 'num_hosts' node in ZooKeeper"); - num_hosts.reset(); /// needs to reread 'num_hosts' again - } - else if ((responses.size() > num_hosts_node_path_pos) && !num_hosts && - (responses[num_hosts_node_path_pos]->error == Coordination::Error::ZNODEEXISTS)) - { - show_error_before_next_attempt("Other host created the 'num_hosts' node in ZooKeeper"); - /// needs another attempt - } - else if ((responses.size() > alive_tracker_node_path_pos) && - (responses[alive_tracker_node_path_pos]->error == Coordination::Error::ZBADVERSION)) - { - show_error_before_next_attempt("Concurrent backup or restore changed some 'alive' nodes in ZooKeeper"); + show_error_before_next_attempt(fmt::format("The version of node {} changed", alive_tracker_node_path)); check_concurrency = true; /// needs to recheck for concurrency again } else @@ -337,8 +373,7 @@ void BackupCoordinationStageSync::createStartAndAliveNodes(Coordination::ZooKeep } throw Exception(ErrorCodes::FAILED_TO_SYNC_BACKUP_OR_RESTORE, - "Couldn't create the 'start' node in ZooKeeper for {} after {} attempts", - current_host_desc, max_attempts_after_bad_version); + "Couldn't create node {} in ZooKeeper after {} attempts", start_node_path, max_attempts_after_bad_version); } @@ -387,36 +422,53 @@ void BackupCoordinationStageSync::startWatchingThread() void BackupCoordinationStageSync::stopWatchingThread() { - should_stop_watching_thread = true; + { + std::lock_guard lock{mutex}; + if (should_stop_watching_thread) + return; + should_stop_watching_thread = true; - /// Wake up waiting threads. - if (zk_nodes_changed) - zk_nodes_changed->set(); - state_changed.notify_all(); + /// Wake up waiting threads. + if (zk_nodes_changed) + zk_nodes_changed->set(); + state_changed.notify_all(); + } if (watching_thread_future.valid()) watching_thread_future.wait(); + + LOG_TRACE(log, "Stopped the watching thread"); } void BackupCoordinationStageSync::watchingThread() { - while (!should_stop_watching_thread) + auto should_stop = [&] + { + std::lock_guard lock{mutex}; + return should_stop_watching_thread; + }; + + while (!should_stop()) { try { /// Check if the current BACKUP or RESTORE command is already cancelled. checkIfQueryCancelled(); + } + catch (...) + { + tryLogCurrentException(log, "Caugth exception while watching"); + } - /// Reset the `connected` flag for each host, we'll set them to true again after we find the 'alive' nodes. - resetConnectedFlag(); - + try + { /// Recreate the 'alive' node if necessary and read a new state from ZooKeeper. auto holder = with_retries.createRetriesControlHolder("BackupStageSync::watchingThread"); auto & zookeeper = holder.faulty_zookeeper; with_retries.renewZooKeeper(zookeeper); - if (should_stop_watching_thread) + if (should_stop()) return; /// Recreate the 'alive' node if it was removed. @@ -427,7 +479,10 @@ void BackupCoordinationStageSync::watchingThread() } catch (...) { - tryLogCurrentException(log, "Caugth exception while watching"); + tryLogCurrentException(log, "Caught exception while watching"); + + /// Reset the `connected` flag for each host, we'll set them to true again after we find the 'alive' nodes. + resetConnectedFlag(); } try @@ -438,7 +493,7 @@ void BackupCoordinationStageSync::watchingThread() } catch (...) { - tryLogCurrentException(log, "Caugth exception while checking if the query should be cancelled"); + tryLogCurrentException(log, "Caught exception while watching"); } zk_nodes_changed->tryWait(sync_period_ms.count()); @@ -473,7 +528,7 @@ void BackupCoordinationStageSync::readCurrentState(Coordination::ZooKeeperWithFa zk_nodes_changed->reset(); /// Get zk nodes and subscribe on their changes. - Strings new_zk_nodes = zookeeper->getChildren(stage_node_path, nullptr, zk_nodes_changed); + Strings new_zk_nodes = zookeeper->getChildren(zookeeper_path, nullptr, zk_nodes_changed); std::sort(new_zk_nodes.begin(), new_zk_nodes.end()); /// Sorting is necessary because we compare the list of zk nodes with its previous versions. State new_state; @@ -492,6 +547,8 @@ void BackupCoordinationStageSync::readCurrentState(Coordination::ZooKeeperWithFa zk_nodes = new_zk_nodes; new_state = state; + for (auto & [_, host_info] : new_state.hosts) + host_info.connected = false; } auto get_host_info = [&](const String & host) -> HostInfo * @@ -514,7 +571,8 @@ void BackupCoordinationStageSync::readCurrentState(Coordination::ZooKeeperWithFa { String serialized_error = zookeeper->get(error_node_path); auto [exception, host] = parseErrorNode(serialized_error); - if (auto * host_info = get_host_info(host)) + auto * host_info = get_host_info(host); + if (exception && host_info) { host_info->exception = exception; new_state.host_with_error = host; @@ -576,6 +634,9 @@ void BackupCoordinationStageSync::readCurrentState(Coordination::ZooKeeperWithFa { std::lock_guard lock{mutex}; + /// We were reading `new_state` from ZooKeeper with `mutex` unlocked, so `state` could get more information during that reading, + /// we don't want to lose that information, that's why we use merge() here. + new_state.merge(state); was_state_changed = (new_state != state); state = std::move(new_state); } @@ -604,26 +665,10 @@ int BackupCoordinationStageSync::parseStartNode(const String & start_node_conten } -std::pair BackupCoordinationStageSync::parseErrorNode(const String & error_node_contents) -{ - ReadBufferFromOwnString buf{error_node_contents}; - String host; - readStringBinary(host, buf); - auto exception = std::make_exception_ptr(readException(buf, fmt::format("Got error from {}", getHostDesc(host)))); - return {exception, host}; -} - - void BackupCoordinationStageSync::checkIfQueryCancelled() { if (process_list_element->checkTimeLimitSoft()) return; /// Not cancelled. - - std::lock_guard lock{mutex}; - if (state.cancelled) - return; /// Already marked as cancelled. - - state.cancelled = true; state_changed.notify_all(); } @@ -634,13 +679,13 @@ void BackupCoordinationStageSync::cancelQueryIfError() { std::lock_guard lock{mutex}; - if (state.cancelled || !state.host_with_error) + if (!state.host_with_error) return; - state.cancelled = true; exception = state.hosts.at(*state.host_with_error).exception; } + chassert(exception); process_list_element->cancelQuery(false, exception); state_changed.notify_all(); } @@ -652,7 +697,7 @@ void BackupCoordinationStageSync::cancelQueryIfDisconnectedTooLong() { std::lock_guard lock{mutex}; - if (state.cancelled || state.host_with_error || ((failure_after_host_disconnected_for_seconds.count() == 0))) + if (state.host_with_error || ((failure_after_host_disconnected_for_seconds.count() == 0))) return; auto monotonic_now = std::chrono::steady_clock::now(); @@ -685,27 +730,92 @@ void BackupCoordinationStageSync::cancelQueryIfDisconnectedTooLong() } } } - - if (!exception) - return; - - state.cancelled = true; } + if (!exception) + return; + process_list_element->cancelQuery(false, exception); state_changed.notify_all(); } +void BackupCoordinationStageSync::setQueryIsSentToOtherHosts() +{ + std::lock_guard lock{mutex}; + query_is_sent_to_other_hosts = true; +} + +bool BackupCoordinationStageSync::isQuerySentToOtherHosts() const +{ + std::lock_guard lock{mutex}; + return query_is_sent_to_other_hosts; +} + + void BackupCoordinationStageSync::setStage(const String & stage, const String & stage_result) { LOG_INFO(log, "{} reached stage {}", current_host_desc, stage); + + { + std::lock_guard lock{mutex}; + if (state.hosts.at(current_host).stages.contains(stage)) + return; /// Already set. + } + + if ((getInitiatorVersion() == kVersionWithoutFinishNode) && (stage == BackupCoordinationStage::COMPLETED)) + { + LOG_TRACE(log, "Stopping the watching thread because the initiator uses outdated version {}", getInitiatorVersion()); + stopWatchingThread(); + } + auto holder = with_retries.createRetriesControlHolder("BackupStageSync::setStage"); holder.retries_ctl.retryLoop([&, &zookeeper = holder.faulty_zookeeper]() { with_retries.renewZooKeeper(zookeeper); - zookeeper->createIfNotExists(getStageNodePath(stage), stage_result); + createStageNode(stage, stage_result, zookeeper); }); + + /// If the initiator of the query has that old version then it doesn't expect us to create the 'finish' node and moreover + /// the initiator can start removing all the nodes immediately after all hosts report about reaching the "completed" status. + /// So to avoid weird errors in the logs we won't create the 'finish' node if the initiator of the query has that old version. + if ((getInitiatorVersion() == kVersionWithoutFinishNode) && (stage == BackupCoordinationStage::COMPLETED)) + { + LOG_INFO(log, "Skipped creating the 'finish' node because the initiator uses outdated version {}", getInitiatorVersion()); + std::lock_guard lock{mutex}; + tried_to_finish = true; + state.hosts.at(current_host).finished = true; + } +} + + +void BackupCoordinationStageSync::createStageNode(const String & stage, const String & stage_result, Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper) +{ + String serialized_error; + if (zookeeper->tryGet(error_node_path, serialized_error)) + { + auto [exception, host] = parseErrorNode(serialized_error); + if (exception) + std::rethrow_exception(exception); + } + + auto code = zookeeper->tryCreate(getStageNodePath(stage), stage_result, zkutil::CreateMode::Persistent); + if (code == Coordination::Error::ZOK) + { + std::lock_guard lock{mutex}; + state.hosts.at(current_host).stages[stage] = stage_result; + return; + } + + if (code == Coordination::Error::ZNODEEXISTS) + { + String another_result = zookeeper->get(getStageNodePath(stage)); + std::lock_guard lock{mutex}; + state.hosts.at(current_host).stages[stage] = another_result; + return; + } + + throw zkutil::KeeperException::fromPath(code, getStageNodePath(stage)); } @@ -715,71 +825,7 @@ String BackupCoordinationStageSync::getStageNodePath(const String & stage) const } -bool BackupCoordinationStageSync::trySetError(std::exception_ptr exception) noexcept -{ - try - { - std::rethrow_exception(exception); - } - catch (const Exception & e) - { - return trySetError(e); - } - catch (...) - { - return trySetError(Exception(getCurrentExceptionMessageAndPattern(true, true), getCurrentExceptionCode())); - } -} - - -bool BackupCoordinationStageSync::trySetError(const Exception & exception) -{ - try - { - setError(exception); - return true; - } - catch (...) - { - return false; - } -} - - -void BackupCoordinationStageSync::setError(const Exception & exception) -{ - /// Most likely this exception has been already logged so here we're logging it without stacktrace. - String exception_message = getExceptionMessage(exception, /* with_stacktrace= */ false, /* check_embedded_stacktrace= */ true); - LOG_INFO(log, "Sending exception from {} to other hosts: {}", current_host_desc, exception_message); - - auto holder = with_retries.createRetriesControlHolder("BackupStageSync::setError", WithRetries::kErrorHandling); - - holder.retries_ctl.retryLoop([&, &zookeeper = holder.faulty_zookeeper]() - { - with_retries.renewZooKeeper(zookeeper); - - WriteBufferFromOwnString buf; - writeStringBinary(current_host, buf); - writeException(exception, buf, true); - auto code = zookeeper->tryCreate(error_node_path, buf.str(), zkutil::CreateMode::Persistent); - - if (code == Coordination::Error::ZOK) - { - LOG_TRACE(log, "Sent exception from {} to other hosts", current_host_desc); - } - else if (code == Coordination::Error::ZNODEEXISTS) - { - LOG_INFO(log, "An error has been already assigned for this {}", operation_name); - } - else - { - throw zkutil::KeeperException::fromPath(code, error_node_path); - } - }); -} - - -Strings BackupCoordinationStageSync::waitForHostsToReachStage(const String & stage_to_wait, const Strings & hosts, std::optional timeout) const +Strings BackupCoordinationStageSync::waitHostsReachStage(const Strings & hosts, const String & stage_to_wait) const { Strings results; results.resize(hosts.size()); @@ -787,44 +833,28 @@ Strings BackupCoordinationStageSync::waitForHostsToReachStage(const String & sta std::unique_lock lock{mutex}; /// TSA_NO_THREAD_SAFETY_ANALYSIS is here because Clang Thread Safety Analysis doesn't understand std::unique_lock. - auto check_if_hosts_ready = [&](bool time_is_out) TSA_NO_THREAD_SAFETY_ANALYSIS + auto check_if_hosts_reach_stage = [&]() TSA_NO_THREAD_SAFETY_ANALYSIS { - return checkIfHostsReachStage(hosts, stage_to_wait, time_is_out, timeout, results); + return checkIfHostsReachStage(hosts, stage_to_wait, results); }; - if (timeout) - { - if (!state_changed.wait_for(lock, *timeout, [&] { return check_if_hosts_ready(/* time_is_out = */ false); })) - check_if_hosts_ready(/* time_is_out = */ true); - } - else - { - state_changed.wait(lock, [&] { return check_if_hosts_ready(/* time_is_out = */ false); }); - } + state_changed.wait(lock, check_if_hosts_reach_stage); return results; } -bool BackupCoordinationStageSync::checkIfHostsReachStage( - const Strings & hosts, - const String & stage_to_wait, - bool time_is_out, - std::optional timeout, - Strings & results) const +bool BackupCoordinationStageSync::checkIfHostsReachStage(const Strings & hosts, const String & stage_to_wait, Strings & results) const { - if (should_stop_watching_thread) - throw Exception(ErrorCodes::LOGICAL_ERROR, "finish() was called while waiting for a stage"); - process_list_element->checkTimeLimit(); for (size_t i = 0; i != hosts.size(); ++i) { const String & host = hosts[i]; auto it = state.hosts.find(host); - if (it == state.hosts.end()) - throw Exception(ErrorCodes::LOGICAL_ERROR, "waitForHostsToReachStage() was called for unexpected {}, all hosts are {}", getHostDesc(host), getHostsDesc(all_hosts)); + throw Exception(ErrorCodes::LOGICAL_ERROR, + "waitHostsReachStage() was called for unexpected {}, all hosts are {}", getHostDesc(host), getHostsDesc(all_hosts)); const HostInfo & host_info = it->second; auto stage_it = host_info.stages.find(stage_to_wait); @@ -835,10 +865,11 @@ bool BackupCoordinationStageSync::checkIfHostsReachStage( } if (host_info.finished) - { throw Exception(ErrorCodes::FAILED_TO_SYNC_BACKUP_OR_RESTORE, "{} finished without coming to stage {}", getHostDesc(host), stage_to_wait); - } + + if (should_stop_watching_thread) + throw Exception(ErrorCodes::LOGICAL_ERROR, "waitHostsReachStage() can't wait for stage {} after the watching thread stopped", stage_to_wait); String host_status; if (!host_info.started) @@ -846,85 +877,73 @@ bool BackupCoordinationStageSync::checkIfHostsReachStage( else if (!host_info.connected) host_status = fmt::format(": the host is currently disconnected, last connection was at {}", host_info.last_connection_time); - if (!time_is_out) - { - LOG_TRACE(log, "Waiting for {} to reach stage {}{}", getHostDesc(host), stage_to_wait, host_status); - return false; - } - else - { - throw Exception(ErrorCodes::FAILED_TO_SYNC_BACKUP_OR_RESTORE, - "Waited longer than timeout {} for {} to reach stage {}{}", - *timeout, getHostDesc(host), stage_to_wait, host_status); - } + LOG_TRACE(log, "Waiting for {} to reach stage {}{}", getHostDesc(host), stage_to_wait, host_status); + return false; /// wait for next change of `state_changed` } LOG_INFO(log, "Hosts {} reached stage {}", getHostsDesc(hosts), stage_to_wait); - return true; + return true; /// stop waiting } -void BackupCoordinationStageSync::finish(bool & other_hosts_also_finished) +bool BackupCoordinationStageSync::finish(bool throw_if_error) { - tryFinishImpl(other_hosts_also_finished, /* throw_if_error = */ true, /* retries_kind = */ WithRetries::kNormal); + WithRetries::Kind retries_kind = WithRetries::kNormal; + if (throw_if_error) + retries_kind = WithRetries::kErrorHandling; + + return finishImpl(throw_if_error, retries_kind); } -bool BackupCoordinationStageSync::tryFinishAfterError(bool & other_hosts_also_finished) noexcept +bool BackupCoordinationStageSync::finishImpl(bool throw_if_error, WithRetries::Kind retries_kind) { - return tryFinishImpl(other_hosts_also_finished, /* throw_if_error = */ false, /* retries_kind = */ WithRetries::kErrorHandling); -} - - -bool BackupCoordinationStageSync::tryFinishImpl() -{ - bool other_hosts_also_finished; - return tryFinishAfterError(other_hosts_also_finished); -} - - -bool BackupCoordinationStageSync::tryFinishImpl(bool & other_hosts_also_finished, bool throw_if_error, WithRetries::Kind retries_kind) -{ - auto get_value_other_hosts_also_finished = [&] TSA_REQUIRES(mutex) - { - other_hosts_also_finished = true; - for (const auto & [host, host_info] : state.hosts) - { - if ((host != current_host) && !host_info.finished) - other_hosts_also_finished = false; - } - }; - { std::lock_guard lock{mutex}; - if (finish_result.succeeded) + + if (finishedNoLock()) { - get_value_other_hosts_also_finished(); + LOG_INFO(log, "The finish node for {} already exists", current_host_desc); return true; } - if (finish_result.exception) + + if (tried_to_finish) { - if (throw_if_error) - std::rethrow_exception(finish_result.exception); + /// We don't repeat creating the finish node, no matter if it was successful or not. + LOG_INFO(log, "Skipped creating the finish node for {} because earlier we failed to do that", current_host_desc); return false; } + + bool failed_to_set_error = tried_to_set_error && !state.host_with_error; + if (failed_to_set_error) + { + /// Tried to create the 'error' node, but failed. + /// Then it's better not to create the 'finish' node in this case because otherwise other hosts might think we've succeeded. + LOG_INFO(log, "Skipping creating the finish node for {} because there was an error which we were unable to send to other hosts", current_host_desc); + return false; + } + + if (current_host == kInitiator) + { + /// Normally the initiator should wait for other hosts to finish before creating its own finish node. + /// We show warning if some of the other hosts didn't finish. + bool expect_other_hosts_finished = query_is_sent_to_other_hosts || !state.host_with_error; + bool other_hosts_finished = otherHostsFinishedNoLock() || !expect_other_hosts_finished; + if (!other_hosts_finished) + LOG_WARNING(log, "Hosts {} didn't finish before the initiator", getHostsDesc(getUnfinishedOtherHostsNoLock())); + } } + stopWatchingThread(); + try { - stopWatchingThread(); - auto holder = with_retries.createRetriesControlHolder("BackupStageSync::finish", retries_kind); holder.retries_ctl.retryLoop([&, &zookeeper = holder.faulty_zookeeper]() { with_retries.renewZooKeeper(zookeeper); - createFinishNodeAndRemoveAliveNode(zookeeper); + createFinishNodeAndRemoveAliveNode(zookeeper, throw_if_error); }); - - std::lock_guard lock{mutex}; - finish_result.succeeded = true; - get_value_other_hosts_also_finished(); - return true; } catch (...) { @@ -933,63 +952,87 @@ bool BackupCoordinationStageSync::tryFinishImpl(bool & other_hosts_also_finished getCurrentExceptionMessage(/* with_stacktrace= */ false, /* check_embedded_stacktrace= */ true)); std::lock_guard lock{mutex}; - finish_result.exception = std::current_exception(); + tried_to_finish = true; + if (throw_if_error) throw; return false; } + + { + std::lock_guard lock{mutex}; + tried_to_finish = true; + state.hosts.at(current_host).finished = true; + } + + return true; } -void BackupCoordinationStageSync::createFinishNodeAndRemoveAliveNode(Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper) +void BackupCoordinationStageSync::createFinishNodeAndRemoveAliveNode(Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper, bool throw_if_error) { - if (zookeeper->exists(finish_node_path)) - return; - - /// If the initiator of the query has that old version then it doesn't expect us to create the 'finish' node and moreover - /// the initiator can start removing all the nodes immediately after all hosts report about reaching the "completed" status. - /// So to avoid weird errors in the logs we won't create the 'finish' node if the initiator of the query has that old version. - if ((getInitiatorVersion() == kVersionWithoutFinishNode) && (current_host != kInitiator)) - { - LOG_INFO(log, "Skipped creating the 'finish' node because the initiator uses outdated version {}", getInitiatorVersion()); - return; - } - std::optional num_hosts; int num_hosts_version = -1; for (size_t attempt_no = 1; attempt_no <= max_attempts_after_bad_version; ++attempt_no) { + /// The 'num_hosts' node may not exist if createStartAndAliveNodes() failed in the constructor. if (!num_hosts) { + String num_hosts_str; Coordination::Stat stat; - num_hosts = parseFromString(zookeeper->get(num_hosts_node_path, &stat)); - num_hosts_version = stat.version; + if (zookeeper->tryGet(num_hosts_node_path, num_hosts_str, &stat)) + { + num_hosts = parseFromString(num_hosts_str); + num_hosts_version = stat.version; + } } + String serialized_error; + if (throw_if_error && zookeeper->tryGet(error_node_path, serialized_error)) + { + auto [exception, host] = parseErrorNode(serialized_error); + if (exception) + std::rethrow_exception(exception); + } + + if (zookeeper->exists(finish_node_path)) + return; + + bool start_node_exists = zookeeper->exists(start_node_path); + Coordination::Requests requests; requests.reserve(3); requests.emplace_back(zkutil::makeCreateRequest(finish_node_path, "", zkutil::CreateMode::Persistent)); - size_t num_hosts_node_path_pos = requests.size(); - requests.emplace_back(zkutil::makeSetRequest(num_hosts_node_path, toString(*num_hosts - 1), num_hosts_version)); - - size_t alive_node_path_pos = static_cast(-1); + size_t alive_node_pos = static_cast(-1); if (zookeeper->exists(alive_node_path)) { - alive_node_path_pos = requests.size(); + alive_node_pos = requests.size(); requests.emplace_back(zkutil::makeRemoveRequest(alive_node_path, -1)); } + size_t num_hosts_node_pos = static_cast(-1); + if (num_hosts) + { + num_hosts_node_pos = requests.size(); + requests.emplace_back(zkutil::makeSetRequest(num_hosts_node_path, toString(start_node_exists ? (*num_hosts - 1) : *num_hosts), num_hosts_version)); + } + Coordination::Responses responses; auto code = zookeeper->tryMulti(requests, responses); if (code == Coordination::Error::ZOK) { - --*num_hosts; - String hosts_left_desc = ((*num_hosts == 0) ? "no hosts left" : fmt::format("{} hosts left", *num_hosts)); - LOG_INFO(log, "Created the 'finish' node in ZooKeeper for {}, {}", current_host_desc, hosts_left_desc); + String hosts_left_desc; + if (num_hosts) + { + if (start_node_exists) + --*num_hosts; + hosts_left_desc = (*num_hosts == 0) ? ", no hosts left" : fmt::format(", {} hosts left", *num_hosts); + } + LOG_INFO(log, "Created the 'finish' node in ZooKeeper for {}{}", current_host_desc, hosts_left_desc); return; } @@ -999,18 +1042,18 @@ void BackupCoordinationStageSync::createFinishNodeAndRemoveAliveNode(Coordinatio LOG_TRACE(log, "{} (attempt #{}){}", message, attempt_no, will_try_again ? ", will try again" : ""); }; - if ((responses.size() > num_hosts_node_path_pos) && - (responses[num_hosts_node_path_pos]->error == Coordination::Error::ZBADVERSION)) + if ((alive_node_pos < responses.size()) && + (responses[alive_node_pos]->error == Coordination::Error::ZNONODE)) { - show_error_before_next_attempt("Other host changed the 'num_hosts' node in ZooKeeper"); - num_hosts.reset(); /// needs to reread 'num_hosts' again - } - else if ((responses.size() > alive_node_path_pos) && - (responses[alive_node_path_pos]->error == Coordination::Error::ZNONODE)) - { - show_error_before_next_attempt(fmt::format("Node {} in ZooKeeper doesn't exist", alive_node_path_pos)); + show_error_before_next_attempt(fmt::format("Node {} doesn't exist", alive_node_path)); /// needs another attempt } + else if ((num_hosts_node_pos < responses.size()) && + (responses[num_hosts_node_pos]->error == Coordination::Error::ZBADVERSION)) + { + show_error_before_next_attempt(fmt::format("The version of node {} changed", num_hosts_node_path)); + num_hosts.reset(); /// needs to reread 'num_hosts' again + } else { zkutil::KeeperMultiException::check(code, requests, responses); @@ -1026,60 +1069,73 @@ void BackupCoordinationStageSync::createFinishNodeAndRemoveAliveNode(Coordinatio int BackupCoordinationStageSync::getInitiatorVersion() const { std::lock_guard lock{mutex}; - auto it = state.hosts.find(String{kInitiator}); - if (it == state.hosts.end()) - throw Exception(ErrorCodes::LOGICAL_ERROR, "There is no initiator of this {} query, it's a bug", operation_name); - const HostInfo & host_info = it->second; - return host_info.version; + return state.hosts.at(String{kInitiator}).version; } -void BackupCoordinationStageSync::waitForOtherHostsToFinish() const -{ - tryWaitForOtherHostsToFinishImpl(/* reason = */ "", /* throw_if_error = */ true, /* timeout = */ {}); -} - - -bool BackupCoordinationStageSync::tryWaitForOtherHostsToFinishAfterError() const noexcept +bool BackupCoordinationStageSync::waitOtherHostsFinish(bool throw_if_error) const { std::optional timeout; - if (finish_timeout_after_error.count() != 0) - timeout = finish_timeout_after_error; + String reason; - String reason = fmt::format("{} needs other hosts to finish before cleanup", current_host_desc); - return tryWaitForOtherHostsToFinishImpl(reason, /* throw_if_error = */ false, timeout); + if (!throw_if_error) + { + if (finish_timeout_after_error.count() != 0) + timeout = finish_timeout_after_error; + reason = "after error before cleanup"; + } + + return waitOtherHostsFinishImpl(reason, timeout, throw_if_error); } -bool BackupCoordinationStageSync::tryWaitForOtherHostsToFinishImpl(const String & reason, bool throw_if_error, std::optional timeout) const +bool BackupCoordinationStageSync::waitOtherHostsFinishImpl(const String & reason, std::optional timeout, bool throw_if_error) const { std::unique_lock lock{mutex}; /// TSA_NO_THREAD_SAFETY_ANALYSIS is here because Clang Thread Safety Analysis doesn't understand std::unique_lock. - auto check_if_other_hosts_finish = [&](bool time_is_out) TSA_NO_THREAD_SAFETY_ANALYSIS + auto other_hosts_finished = [&]() TSA_NO_THREAD_SAFETY_ANALYSIS { return otherHostsFinishedNoLock(); }; + + if (other_hosts_finished()) { - return checkIfOtherHostsFinish(reason, throw_if_error, time_is_out, timeout); + LOG_TRACE(log, "Other hosts have already finished"); + return true; + } + + bool failed_to_set_error = TSA_SUPPRESS_WARNING_FOR_READ(tried_to_set_error) && !TSA_SUPPRESS_WARNING_FOR_READ(state).host_with_error; + if (failed_to_set_error) + { + /// Tried to create the 'error' node, but failed. + /// Then it's better not to wait for other hosts to finish in this case because other hosts don't know they should finish. + LOG_INFO(log, "Skipping waiting for other hosts to finish because there was an error which we were unable to send to other hosts"); + return false; + } + + bool result = false; + + /// TSA_NO_THREAD_SAFETY_ANALYSIS is here because Clang Thread Safety Analysis doesn't understand std::unique_lock. + auto check_if_hosts_finish = [&](bool time_is_out) TSA_NO_THREAD_SAFETY_ANALYSIS + { + return checkIfOtherHostsFinish(reason, timeout, time_is_out, result, throw_if_error); }; if (timeout) { - if (state_changed.wait_for(lock, *timeout, [&] { return check_if_other_hosts_finish(/* time_is_out = */ false); })) - return true; - return check_if_other_hosts_finish(/* time_is_out = */ true); + if (!state_changed.wait_for(lock, *timeout, [&] { return check_if_hosts_finish(/* time_is_out = */ false); })) + check_if_hosts_finish(/* time_is_out = */ true); } else { - state_changed.wait(lock, [&] { return check_if_other_hosts_finish(/* time_is_out = */ false); }); - return true; + state_changed.wait(lock, [&] { return check_if_hosts_finish(/* time_is_out = */ false); }); } + + return result; } -bool BackupCoordinationStageSync::checkIfOtherHostsFinish(const String & reason, bool throw_if_error, bool time_is_out, std::optional timeout) const +bool BackupCoordinationStageSync::checkIfOtherHostsFinish( + const String & reason, std::optional timeout, bool time_is_out, bool & result, bool throw_if_error) const { - if (should_stop_watching_thread) - throw Exception(ErrorCodes::LOGICAL_ERROR, "finish() was called while waiting for other hosts to finish"); - if (throw_if_error) process_list_element->checkTimeLimit(); @@ -1088,38 +1144,261 @@ bool BackupCoordinationStageSync::checkIfOtherHostsFinish(const String & reason, if ((host == current_host) || host_info.finished) continue; + String reason_text = reason.empty() ? "" : (" " + reason); + String host_status; if (!host_info.started) host_status = fmt::format(": the host hasn't started working on this {} yet", operation_name); else if (!host_info.connected) host_status = fmt::format(": the host is currently disconnected, last connection was at {}", host_info.last_connection_time); - if (!time_is_out) + if (time_is_out) { - String reason_text = reason.empty() ? "" : (" because " + reason); - LOG_TRACE(log, "Waiting for {} to finish{}{}", getHostDesc(host), reason_text, host_status); - return false; - } - else - { - String reason_text = reason.empty() ? "" : fmt::format(" (reason of waiting: {})", reason); - if (!throw_if_error) - { - LOG_INFO(log, "Waited longer than timeout {} for {} to finish{}{}", - *timeout, getHostDesc(host), host_status, reason_text); - return false; - } - else + if (throw_if_error) { throw Exception(ErrorCodes::FAILED_TO_SYNC_BACKUP_OR_RESTORE, "Waited longer than timeout {} for {} to finish{}{}", - *timeout, getHostDesc(host), host_status, reason_text); + *timeout, getHostDesc(host), reason_text, host_status); } + LOG_INFO(log, "Waited longer than timeout {} for {} to finish{}{}", + *timeout, getHostDesc(host), reason_text, host_status); + result = false; + return true; /// stop waiting } + + if (should_stop_watching_thread) + { + LOG_ERROR(log, "waitOtherHostFinish({}) can't wait for other hosts to finish after the watching thread stopped", throw_if_error); + chassert(false, "waitOtherHostFinish() can't wait for other hosts to finish after the watching thread stopped"); + if (throw_if_error) + throw Exception(ErrorCodes::LOGICAL_ERROR, "waitOtherHostsFinish() can't wait for other hosts to finish after the watching thread stopped"); + result = false; + return true; /// stop waiting + } + + LOG_TRACE(log, "Waiting for {} to finish{}{}", getHostDesc(host), reason_text, host_status); + return false; /// wait for next change of `state_changed` } LOG_TRACE(log, "Other hosts finished working on this {}", operation_name); + result = true; + return true; /// stop waiting +} + + +bool BackupCoordinationStageSync::finished() const +{ + std::lock_guard lock{mutex}; + return finishedNoLock(); +} + + +bool BackupCoordinationStageSync::finishedNoLock() const +{ + return state.hosts.at(current_host).finished; +} + + +bool BackupCoordinationStageSync::otherHostsFinished() const +{ + std::lock_guard lock{mutex}; + return otherHostsFinishedNoLock(); +} + + +bool BackupCoordinationStageSync::otherHostsFinishedNoLock() const +{ + for (const auto & [host, host_info] : state.hosts) + { + if (!host_info.finished && (host != current_host)) + return false; + } return true; } + +bool BackupCoordinationStageSync::allHostsFinishedNoLock() const +{ + return finishedNoLock() && otherHostsFinishedNoLock(); +} + + +Strings BackupCoordinationStageSync::getUnfinishedHosts() const +{ + std::lock_guard lock{mutex}; + return getUnfinishedHostsNoLock(); +} + + +Strings BackupCoordinationStageSync::getUnfinishedHostsNoLock() const +{ + if (allHostsFinishedNoLock()) + return {}; + + Strings res; + res.reserve(all_hosts.size()); + for (const auto & [host, host_info] : state.hosts) + { + if (!host_info.finished) + res.emplace_back(host); + } + return res; +} + + +Strings BackupCoordinationStageSync::getUnfinishedOtherHosts() const +{ + std::lock_guard lock{mutex}; + return getUnfinishedOtherHostsNoLock(); +} + + +Strings BackupCoordinationStageSync::getUnfinishedOtherHostsNoLock() const +{ + if (otherHostsFinishedNoLock()) + return {}; + + Strings res; + res.reserve(all_hosts.size() - 1); + for (const auto & [host, host_info] : state.hosts) + { + if (!host_info.finished && (host != current_host)) + res.emplace_back(host); + } + return res; +} + + +bool BackupCoordinationStageSync::setError(std::exception_ptr exception, bool throw_if_error) +{ + try + { + std::rethrow_exception(exception); + } + catch (const Exception & e) + { + return setError(e, throw_if_error); + } + catch (...) + { + return setError(Exception{getCurrentExceptionMessageAndPattern(true, true), getCurrentExceptionCode()}, throw_if_error); + } +} + + +bool BackupCoordinationStageSync::setError(const Exception & exception, bool throw_if_error) +{ + try + { + /// Most likely this exception has been already logged so here we're logging it without stacktrace. + String exception_message = getExceptionMessage(exception, /* with_stacktrace= */ false, /* check_embedded_stacktrace= */ true); + LOG_INFO(log, "Sending exception from {} to other hosts: {}", current_host_desc, exception_message); + + { + std::lock_guard lock{mutex}; + if (state.host_with_error) + { + LOG_INFO(log, "The error node already exists"); + return true; + } + + if (tried_to_set_error) + { + LOG_INFO(log, "Skipped creating the error node because earlier we failed to do that"); + return false; + } + } + + auto holder = with_retries.createRetriesControlHolder("BackupStageSync::setError", WithRetries::kErrorHandling); + holder.retries_ctl.retryLoop([&, &zookeeper = holder.faulty_zookeeper]() + { + with_retries.renewZooKeeper(zookeeper); + createErrorNode(exception, zookeeper); + }); + + { + std::lock_guard lock{mutex}; + tried_to_set_error = true; + return true; + } + } + catch (...) + { + LOG_TRACE(log, "Caught exception while removing nodes from ZooKeeper for this {}: {}", + is_restore ? "restore" : "backup", + getCurrentExceptionMessage(/* with_stacktrace= */ false, /* check_embedded_stacktrace= */ true)); + + std::lock_guard lock{mutex}; + tried_to_set_error = true; + + if (throw_if_error) + throw; + return false; + } +} + + +void BackupCoordinationStageSync::createErrorNode(const Exception & exception, Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper) +{ + String serialized_error; + { + WriteBufferFromOwnString buf; + writeStringBinary(current_host, buf); + writeException(exception, buf, true); + serialized_error = buf.str(); + } + + auto code = zookeeper->tryCreate(error_node_path, serialized_error, zkutil::CreateMode::Persistent); + + if (code == Coordination::Error::ZOK) + { + std::lock_guard lock{mutex}; + if (!state.host_with_error) + { + state.host_with_error = current_host; + state.hosts.at(current_host).exception = parseErrorNode(serialized_error).first; + } + LOG_TRACE(log, "Sent exception from {} to other hosts", current_host_desc); + return; + } + + if (code == Coordination::Error::ZNODEEXISTS) + { + String another_error = zookeeper->get(error_node_path); + auto [another_exception, host] = parseErrorNode(another_error); + if (another_exception) + { + std::lock_guard lock{mutex}; + if (!state.host_with_error) + { + state.host_with_error = host; + state.hosts.at(host).exception = another_exception; + } + LOG_INFO(log, "Another error is already assigned for this {}", operation_name); + return; + } + } + + throw zkutil::KeeperException::fromPath(code, error_node_path); +} + + +std::pair BackupCoordinationStageSync::parseErrorNode(const String & error_node_contents) const +{ + ReadBufferFromOwnString buf{error_node_contents}; + String host; + readStringBinary(host, buf); + if (std::find(all_hosts.begin(), all_hosts.end(), host) == all_hosts.end()) + return {}; + auto exception = std::make_exception_ptr(readException(buf, fmt::format("Got error from {}", getHostDesc(host)))); + return {exception, host}; +} + + +bool BackupCoordinationStageSync::isErrorSet() const +{ + std::lock_guard lock{mutex}; + return state.host_with_error.has_value(); +} + } diff --git a/src/Backups/BackupCoordinationStageSync.h b/src/Backups/BackupCoordinationStageSync.h index dc0d3c3c83d..11d3d1cf6f4 100644 --- a/src/Backups/BackupCoordinationStageSync.h +++ b/src/Backups/BackupCoordinationStageSync.h @@ -1,7 +1,9 @@ #pragma once +#include #include + namespace DB { @@ -9,12 +11,16 @@ namespace DB class BackupCoordinationStageSync { public: + /// Empty string as the current host is used to mark the initiator of a BACKUP ON CLUSTER or RESTORE ON CLUSTER query. + static const constexpr std::string_view kInitiator; + BackupCoordinationStageSync( bool is_restore_, /// true if this is a RESTORE ON CLUSTER command, false if this is a BACKUP ON CLUSTER command const String & zookeeper_path_, /// path to the "stage" folder in ZooKeeper const String & current_host_, /// the current host, or an empty string if it's the initiator of the BACKUP/RESTORE ON CLUSTER command const Strings & all_hosts_, /// all the hosts (including the initiator and the current host) performing the BACKUP/RESTORE ON CLUSTER command bool allow_concurrency_, /// whether it's allowed to have concurrent backups or restores. + BackupConcurrencyCounters & concurrency_counters_, const WithRetries & with_retries_, ThreadPoolCallbackRunnerUnsafe schedule_, QueryStatusPtr process_list_element_, @@ -22,30 +28,37 @@ public: ~BackupCoordinationStageSync(); + /// Sets that the BACKUP or RESTORE query was sent to other hosts. + void setQueryIsSentToOtherHosts(); + bool isQuerySentToOtherHosts() const; + /// Sets the stage of the current host and signal other hosts if there were other hosts waiting for that. void setStage(const String & stage, const String & stage_result = {}); - /// Waits until all the specified hosts come to the specified stage. - /// The function returns the results which specified hosts set when they came to the required stage. - /// If it doesn't happen before the timeout then the function will stop waiting and throw an exception. - Strings waitForHostsToReachStage(const String & stage_to_wait, const Strings & hosts, std::optional timeout = {}) const; - - /// Waits until all the other hosts finish their work. - /// Stops waiting and throws an exception if another host encounters an error or if some host gets cancelled. - void waitForOtherHostsToFinish() const; - - /// Lets other host know that the current host has finished its work. - void finish(bool & other_hosts_also_finished); + /// Waits until specified hosts come to the specified stage. + /// The function returns the results which the specified hosts set when they came to the required stage. + Strings waitHostsReachStage(const Strings & hosts, const String & stage_to_wait) const; /// Lets other hosts know that the current host has encountered an error. - bool trySetError(std::exception_ptr exception) noexcept; + /// The function returns true if it successfully created the error node or if the error node was found already exist. + bool setError(std::exception_ptr exception, bool throw_if_error); + bool isErrorSet() const; - /// Waits until all the other hosts finish their work (as a part of error-handling process). - /// Doesn't stops waiting if some host encounters an error or gets cancelled. - bool tryWaitForOtherHostsToFinishAfterError() const noexcept; + /// Waits until the hosts other than the current host finish their work. Must be called before finish(). + /// Stops waiting and throws an exception if another host encounters an error or if some host gets cancelled. + bool waitOtherHostsFinish(bool throw_if_error) const; + bool otherHostsFinished() const; - /// Lets other host know that the current host has finished its work (as a part of error-handling process). - bool tryFinishAfterError(bool & other_hosts_also_finished) noexcept; + /// Lets other hosts know that the current host has finished its work. + bool finish(bool throw_if_error); + bool finished() const; + + /// Returns true if all the hosts have finished. + bool allHostsFinished() const { return finished() && otherHostsFinished(); } + + /// Returns a list of the hosts which haven't finished yet. + Strings getUnfinishedHosts() const; + Strings getUnfinishedOtherHosts() const; /// Returns a printable name of a specific host. For empty host the function returns "initiator". static String getHostDesc(const String & host); @@ -78,14 +91,17 @@ private: /// Reads the current state from ZooKeeper without throwing exceptions. void readCurrentState(Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper); + + /// Creates a stage node to let other hosts know we've reached the specified stage. + void createStageNode(const String & stage, const String & stage_result, Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper); String getStageNodePath(const String & stage) const; /// Lets other hosts know that the current host has encountered an error. - bool trySetError(const Exception & exception); - void setError(const Exception & exception); + bool setError(const Exception & exception, bool throw_if_error); + void createErrorNode(const Exception & exception, Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper); /// Deserializes an error stored in the error node. - static std::pair parseErrorNode(const String & error_node_contents); + std::pair parseErrorNode(const String & error_node_contents) const; /// Reset the `connected` flag for each host. void resetConnectedFlag(); @@ -102,19 +118,27 @@ private: void cancelQueryIfDisconnectedTooLong(); /// Used by waitForHostsToReachStage() to check if everything is ready to return. - bool checkIfHostsReachStage(const Strings & hosts, const String & stage_to_wait, bool time_is_out, std::optional timeout, Strings & results) const TSA_REQUIRES(mutex); + bool checkIfHostsReachStage(const Strings & hosts, const String & stage_to_wait, Strings & results) const TSA_REQUIRES(mutex); /// Creates the 'finish' node. - bool tryFinishImpl(); - bool tryFinishImpl(bool & other_hosts_also_finished, bool throw_if_error, WithRetries::Kind retries_kind); - void createFinishNodeAndRemoveAliveNode(Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper); + bool finishImpl(bool throw_if_error, WithRetries::Kind retries_kind); + void createFinishNodeAndRemoveAliveNode(Coordination::ZooKeeperWithFaultInjection::Ptr zookeeper, bool throw_if_error); /// Returns the version used by the initiator. int getInitiatorVersion() const; /// Waits until all the other hosts finish their work. - bool tryWaitForOtherHostsToFinishImpl(const String & reason, bool throw_if_error, std::optional timeout) const; - bool checkIfOtherHostsFinish(const String & reason, bool throw_if_error, bool time_is_out, std::optional timeout) const TSA_REQUIRES(mutex); + bool waitOtherHostsFinishImpl(const String & reason, std::optional timeout, bool throw_if_error) const; + bool checkIfOtherHostsFinish(const String & reason, std::optional timeout, bool time_is_out, bool & result, bool throw_if_error) const TSA_REQUIRES(mutex); + + /// Returns true if all the hosts have finished. + bool allHostsFinishedNoLock() const TSA_REQUIRES(mutex); + bool finishedNoLock() const TSA_REQUIRES(mutex); + bool otherHostsFinishedNoLock() const TSA_REQUIRES(mutex); + + /// Returns a list of the hosts which haven't finished yet. + Strings getUnfinishedHostsNoLock() const TSA_REQUIRES(mutex); + Strings getUnfinishedOtherHostsNoLock() const TSA_REQUIRES(mutex); const bool is_restore; const String operation_name; @@ -138,15 +162,16 @@ private: /// Paths in ZooKeeper. const std::filesystem::path zookeeper_path; const String root_zookeeper_path; - const String operation_node_path; + const String operation_zookeeper_path; const String operation_node_name; - const String stage_node_path; const String start_node_path; const String finish_node_path; const String num_hosts_node_path; + const String error_node_path; const String alive_node_path; const String alive_tracker_node_path; - const String error_node_path; + + std::optional concurrency_check; std::shared_ptr zk_nodes_changed; @@ -176,25 +201,21 @@ private: { std::map hosts; /// std::map because we need to compare states std::optional host_with_error; - bool cancelled = false; bool operator ==(const State & other) const; bool operator !=(const State & other) const; + void merge(const State & other); }; State state TSA_GUARDED_BY(mutex); mutable std::condition_variable state_changed; std::future watching_thread_future; - std::atomic should_stop_watching_thread = false; + bool should_stop_watching_thread TSA_GUARDED_BY(mutex) = false; - struct FinishResult - { - bool succeeded = false; - std::exception_ptr exception; - bool other_hosts_also_finished = false; - }; - FinishResult finish_result TSA_GUARDED_BY(mutex); + bool query_is_sent_to_other_hosts TSA_GUARDED_BY(mutex) = false; + bool tried_to_finish TSA_GUARDED_BY(mutex) = false; + bool tried_to_set_error TSA_GUARDED_BY(mutex) = false; mutable std::mutex mutex; }; diff --git a/src/Backups/BackupsWorker.cpp b/src/Backups/BackupsWorker.cpp index 8480dc5d64d..88ebf8eef32 100644 --- a/src/Backups/BackupsWorker.cpp +++ b/src/Backups/BackupsWorker.cpp @@ -329,6 +329,7 @@ std::pair BackupsWorker::start(const ASTPtr & backup_ struct BackupsWorker::BackupStarter { BackupsWorker & backups_worker; + LoggerPtr log; std::shared_ptr backup_query; ContextPtr query_context; /// We have to keep `query_context` until the end of the operation because a pointer to it is stored inside the ThreadGroup we're using. ContextMutablePtr backup_context; @@ -345,6 +346,7 @@ struct BackupsWorker::BackupStarter BackupStarter(BackupsWorker & backups_worker_, const ASTPtr & query_, const ContextPtr & context_) : backups_worker(backups_worker_) + , log(backups_worker.log) , backup_query(std::static_pointer_cast(query_->clone())) , query_context(context_) , backup_context(Context::createCopy(query_context)) @@ -399,9 +401,20 @@ struct BackupsWorker::BackupStarter chassert(!backup); backup = backups_worker.openBackupForWriting(backup_info, backup_settings, backup_coordination, backup_context); - backups_worker.doBackup( - backup, backup_query, backup_id, backup_name_for_logging, backup_settings, backup_coordination, backup_context, - on_cluster, cluster); + backups_worker.doBackup(backup, backup_query, backup_id, backup_settings, backup_coordination, backup_context, + on_cluster, cluster); + + backup_coordination->finish(/* throw_if_error = */ true); + backup.reset(); + + /// The backup coordination is not needed anymore. + if (!is_internal_backup) + backup_coordination->cleanup(/* throw_if_error = */ true); + backup_coordination.reset(); + + /// NOTE: setStatus is called after setNumFilesAndSize in order to have actual information in a backup log record + LOG_INFO(log, "{} {} was created successfully", (is_internal_backup ? "Internal backup" : "Backup"), backup_name_for_logging); + backups_worker.setStatus(backup_id, BackupStatus::BACKUP_CREATED); } void onException() @@ -416,16 +429,29 @@ struct BackupsWorker::BackupStarter if (backup && !backup->setIsCorrupted()) should_remove_files_in_backup = false; - if (backup_coordination && backup_coordination->trySetError(std::current_exception())) + bool all_hosts_finished = false; + + if (backup_coordination && backup_coordination->setError(std::current_exception(), /* throw_if_error = */ false)) { - bool other_hosts_finished = backup_coordination->tryWaitForOtherHostsToFinishAfterError(); + bool other_hosts_finished = !is_internal_backup + && (!backup_coordination->isBackupQuerySentToOtherHosts() || backup_coordination->waitOtherHostsFinish(/* throw_if_error = */ false)); - if (should_remove_files_in_backup && other_hosts_finished) - backup->tryRemoveAllFiles(); - - backup_coordination->tryFinishAfterError(); + all_hosts_finished = backup_coordination->finish(/* throw_if_error = */ false) && other_hosts_finished; } + if (!all_hosts_finished) + should_remove_files_in_backup = false; + + if (backup && should_remove_files_in_backup) + backup->tryRemoveAllFiles(); + + backup.reset(); + + if (backup_coordination && all_hosts_finished) + backup_coordination->cleanup(/* throw_if_error = */ false); + + backup_coordination.reset(); + backups_worker.setStatusSafe(backup_id, getBackupStatusFromCurrentException()); } }; @@ -497,7 +523,6 @@ void BackupsWorker::doBackup( BackupMutablePtr backup, const std::shared_ptr & backup_query, const OperationID & backup_id, - const String & backup_name_for_logging, const BackupSettings & backup_settings, std::shared_ptr backup_coordination, ContextMutablePtr context, @@ -521,10 +546,10 @@ void BackupsWorker::doBackup( backup_settings.copySettingsToQuery(*backup_query); sendQueryToOtherHosts(*backup_query, cluster, backup_settings.shard_num, backup_settings.replica_num, context, required_access, backup_coordination->getOnClusterInitializationKeeperRetriesInfo()); - backup_coordination->setBackupQueryWasSentToOtherHosts(); + backup_coordination->setBackupQueryIsSentToOtherHosts(); /// Wait until all the hosts have written their backup entries. - backup_coordination->waitForOtherHostsToFinish(); + backup_coordination->waitOtherHostsFinish(/* throw_if_error = */ true); } else { @@ -569,18 +594,8 @@ void BackupsWorker::doBackup( compressed_size = backup->getCompressedSize(); } - /// Close the backup. - backup.reset(); - - /// The backup coordination is not needed anymore. - backup_coordination->finish(); - /// NOTE: we need to update metadata again after backup->finalizeWriting(), because backup metadata is written there. setNumFilesAndSize(backup_id, num_files, total_size, num_entries, uncompressed_size, compressed_size, 0, 0); - - /// NOTE: setStatus is called after setNumFilesAndSize in order to have actual information in a backup log record - LOG_INFO(log, "{} {} was created successfully", (is_internal_backup ? "Internal backup" : "Backup"), backup_name_for_logging); - setStatus(backup_id, BackupStatus::BACKUP_CREATED); } @@ -687,6 +702,7 @@ void BackupsWorker::writeBackupEntries( struct BackupsWorker::RestoreStarter { BackupsWorker & backups_worker; + LoggerPtr log; std::shared_ptr restore_query; ContextPtr query_context; /// We have to keep `query_context` until the end of the operation because a pointer to it is stored inside the ThreadGroup we're using. ContextMutablePtr restore_context; @@ -702,6 +718,7 @@ struct BackupsWorker::RestoreStarter RestoreStarter(BackupsWorker & backups_worker_, const ASTPtr & query_, const ContextPtr & context_) : backups_worker(backups_worker_) + , log(backups_worker.log) , restore_query(std::static_pointer_cast(query_->clone())) , query_context(context_) , restore_context(Context::createCopy(query_context)) @@ -753,16 +770,17 @@ struct BackupsWorker::RestoreStarter } restore_coordination = backups_worker.makeRestoreCoordination(on_cluster, restore_settings, restore_context); - backups_worker.doRestore( - restore_query, - restore_id, - backup_name_for_logging, - backup_info, - restore_settings, - restore_coordination, - restore_context, - on_cluster, - cluster); + backups_worker.doRestore(restore_query, restore_id, backup_info, restore_settings, restore_coordination, restore_context, + on_cluster, cluster); + + /// The restore coordination is not needed anymore. + restore_coordination->finish(/* throw_if_error = */ true); + if (!is_internal_restore) + restore_coordination->cleanup(/* throw_if_error = */ true); + restore_coordination.reset(); + + LOG_INFO(log, "Restored from {} {} successfully", (is_internal_restore ? "internal backup" : "backup"), backup_name_for_logging); + backups_worker.setStatus(restore_id, BackupStatus::RESTORED); } void onException() @@ -770,12 +788,16 @@ struct BackupsWorker::RestoreStarter /// Something bad happened, some data were not restored. tryLogCurrentException(backups_worker.log, fmt::format("Failed to restore from {} {}", (is_internal_restore ? "internal backup" : "backup"), backup_name_for_logging)); - if (restore_coordination && restore_coordination->trySetError(std::current_exception())) + if (restore_coordination && restore_coordination->setError(std::current_exception(), /* throw_if_error = */ false)) { - restore_coordination->tryWaitForOtherHostsToFinishAfterError(); - restore_coordination->tryFinishAfterError(); + bool other_hosts_finished = !is_internal_restore + && (!restore_coordination->isRestoreQuerySentToOtherHosts() || restore_coordination->waitOtherHostsFinish(/* throw_if_error = */ false)); + if (restore_coordination->finish(/* throw_if_error = */ false) && other_hosts_finished) + restore_coordination->cleanup(/* throw_if_error = */ false); } + restore_coordination.reset(); + backups_worker.setStatusSafe(restore_id, getRestoreStatusFromCurrentException()); } }; @@ -838,7 +860,6 @@ BackupPtr BackupsWorker::openBackupForReading(const BackupInfo & backup_info, co void BackupsWorker::doRestore( const std::shared_ptr & restore_query, const OperationID & restore_id, - const String & backup_name_for_logging, const BackupInfo & backup_info, RestoreSettings restore_settings, std::shared_ptr restore_coordination, @@ -882,10 +903,10 @@ void BackupsWorker::doRestore( restore_settings.copySettingsToQuery(*restore_query); sendQueryToOtherHosts(*restore_query, cluster, restore_settings.shard_num, restore_settings.replica_num, context, {}, restore_coordination->getOnClusterInitializationKeeperRetriesInfo()); - restore_coordination->setRestoreQueryWasSentToOtherHosts(); + restore_coordination->setRestoreQueryIsSentToOtherHosts(); /// Wait until all the hosts have done with their restoring work. - restore_coordination->waitForOtherHostsToFinish(); + restore_coordination->waitOtherHostsFinish(/* throw_if_error = */ true); } else { @@ -905,12 +926,6 @@ void BackupsWorker::doRestore( backup, context, getThreadPool(ThreadPoolId::RESTORE), after_task_callback}; restorer.run(RestorerFromBackup::RESTORE); } - - /// The restore coordination is not needed anymore. - restore_coordination->finish(); - - LOG_INFO(log, "Restored from {} {} successfully", (is_internal_restore ? "internal backup" : "backup"), backup_name_for_logging); - setStatus(restore_id, BackupStatus::RESTORED); } @@ -943,7 +958,7 @@ BackupsWorker::makeBackupCoordination(bool on_cluster, const BackupSettings & ba if (!on_cluster) { return std::make_shared( - *backup_settings.backup_uuid, !backup_settings.deduplicate_files, allow_concurrent_backups, *concurrency_counters); + !backup_settings.deduplicate_files, allow_concurrent_backups, *concurrency_counters); } bool is_internal_backup = backup_settings.internal; @@ -981,8 +996,7 @@ BackupsWorker::makeRestoreCoordination(bool on_cluster, const RestoreSettings & { if (!on_cluster) { - return std::make_shared( - *restore_settings.restore_uuid, allow_concurrent_restores, *concurrency_counters); + return std::make_shared(allow_concurrent_restores, *concurrency_counters); } bool is_internal_restore = restore_settings.internal; diff --git a/src/Backups/BackupsWorker.h b/src/Backups/BackupsWorker.h index 37f91e269a9..2e5ca84f3f6 100644 --- a/src/Backups/BackupsWorker.h +++ b/src/Backups/BackupsWorker.h @@ -81,7 +81,6 @@ private: BackupMutablePtr backup, const std::shared_ptr & backup_query, const BackupOperationID & backup_id, - const String & backup_name_for_logging, const BackupSettings & backup_settings, std::shared_ptr backup_coordination, ContextMutablePtr context, @@ -102,7 +101,6 @@ private: void doRestore( const std::shared_ptr & restore_query, const BackupOperationID & restore_id, - const String & backup_name_for_logging, const BackupInfo & backup_info, RestoreSettings restore_settings, std::shared_ptr restore_coordination, diff --git a/src/Backups/IBackupCoordination.h b/src/Backups/IBackupCoordination.h index c0eb90de89b..8bd874b9d0d 100644 --- a/src/Backups/IBackupCoordination.h +++ b/src/Backups/IBackupCoordination.h @@ -20,29 +20,27 @@ class IBackupCoordination public: virtual ~IBackupCoordination() = default; + /// Sets that the backup query was sent to other hosts. + /// Function waitOtherHostsFinish() will check that to find out if it should really wait or not. + virtual void setBackupQueryIsSentToOtherHosts() = 0; + virtual bool isBackupQuerySentToOtherHosts() const = 0; + /// Sets the current stage and waits for other hosts to come to this stage too. virtual Strings setStage(const String & new_stage, const String & message, bool sync) = 0; - /// Sets that the backup query was sent to other hosts. - /// Function waitForOtherHostsToFinish() will check that to find out if it should really wait or not. - virtual void setBackupQueryWasSentToOtherHosts() = 0; - /// Lets other hosts know that the current host has encountered an error. - virtual bool trySetError(std::exception_ptr exception) = 0; - - /// Lets other hosts know that the current host has finished its work. - virtual void finish() = 0; - - /// Lets other hosts know that the current host has finished its work (as a part of error-handling process). - virtual bool tryFinishAfterError() noexcept = 0; + /// Returns true if the information is successfully passed so other hosts can read it. + virtual bool setError(std::exception_ptr exception, bool throw_if_error) = 0; /// Waits until all the other hosts finish their work. /// Stops waiting and throws an exception if another host encounters an error or if some host gets cancelled. - virtual void waitForOtherHostsToFinish() = 0; + virtual bool waitOtherHostsFinish(bool throw_if_error) const = 0; - /// Waits until all the other hosts finish their work (as a part of error-handling process). - /// Doesn't stops waiting if some host encounters an error or gets cancelled. - virtual bool tryWaitForOtherHostsToFinishAfterError() noexcept = 0; + /// Lets other hosts know that the current host has finished its work. + virtual bool finish(bool throw_if_error) = 0; + + /// Removes temporary nodes in ZooKeeper. + virtual bool cleanup(bool throw_if_error) = 0; struct PartNameAndChecksum { diff --git a/src/Backups/IRestoreCoordination.h b/src/Backups/IRestoreCoordination.h index daabf1745f3..cc7bfd24202 100644 --- a/src/Backups/IRestoreCoordination.h +++ b/src/Backups/IRestoreCoordination.h @@ -18,29 +18,27 @@ class IRestoreCoordination public: virtual ~IRestoreCoordination() = default; + /// Sets that the restore query was sent to other hosts. + /// Function waitOtherHostsFinish() will check that to find out if it should really wait or not. + virtual void setRestoreQueryIsSentToOtherHosts() = 0; + virtual bool isRestoreQuerySentToOtherHosts() const = 0; + /// Sets the current stage and waits for other hosts to come to this stage too. virtual Strings setStage(const String & new_stage, const String & message, bool sync) = 0; - /// Sets that the restore query was sent to other hosts. - /// Function waitForOtherHostsToFinish() will check that to find out if it should really wait or not. - virtual void setRestoreQueryWasSentToOtherHosts() = 0; - /// Lets other hosts know that the current host has encountered an error. - virtual bool trySetError(std::exception_ptr exception) = 0; - - /// Lets other hosts know that the current host has finished its work. - virtual void finish() = 0; - - /// Lets other hosts know that the current host has finished its work (as a part of error-handling process). - virtual bool tryFinishAfterError() noexcept = 0; + /// Returns true if the information is successfully passed so other hosts can read it. + virtual bool setError(std::exception_ptr exception, bool throw_if_error) = 0; /// Waits until all the other hosts finish their work. /// Stops waiting and throws an exception if another host encounters an error or if some host gets cancelled. - virtual void waitForOtherHostsToFinish() = 0; + virtual bool waitOtherHostsFinish(bool throw_if_error) const = 0; - /// Waits until all the other hosts finish their work (as a part of error-handling process). - /// Doesn't stops waiting if some host encounters an error or gets cancelled. - virtual bool tryWaitForOtherHostsToFinishAfterError() noexcept = 0; + /// Lets other hosts know that the current host has finished its work. + virtual bool finish(bool throw_if_error) = 0; + + /// Removes temporary nodes in ZooKeeper. + virtual bool cleanup(bool throw_if_error) = 0; /// Starts creating a table in a replicated database. Returns false if there is another host which is already creating this table. virtual bool acquireCreatingTableInReplicatedDatabase(const String & database_zk_path, const String & table_name) = 0; diff --git a/src/Backups/RestoreCoordinationLocal.cpp b/src/Backups/RestoreCoordinationLocal.cpp index 569f58f1909..a9eee1fb159 100644 --- a/src/Backups/RestoreCoordinationLocal.cpp +++ b/src/Backups/RestoreCoordinationLocal.cpp @@ -10,9 +10,9 @@ namespace DB { RestoreCoordinationLocal::RestoreCoordinationLocal( - const UUID & restore_uuid, bool allow_concurrent_restore_, BackupConcurrencyCounters & concurrency_counters_) + bool allow_concurrent_restore_, BackupConcurrencyCounters & concurrency_counters_) : log(getLogger("RestoreCoordinationLocal")) - , concurrency_check(restore_uuid, /* is_restore = */ true, /* on_cluster = */ false, allow_concurrent_restore_, concurrency_counters_) + , concurrency_check(/* is_restore = */ true, /* on_cluster = */ false, /* zookeeper_path = */ "", allow_concurrent_restore_, concurrency_counters_) { } diff --git a/src/Backups/RestoreCoordinationLocal.h b/src/Backups/RestoreCoordinationLocal.h index 6be357c4b7e..6e3262a8a2e 100644 --- a/src/Backups/RestoreCoordinationLocal.h +++ b/src/Backups/RestoreCoordinationLocal.h @@ -17,16 +17,16 @@ class ASTCreateQuery; class RestoreCoordinationLocal : public IRestoreCoordination { public: - RestoreCoordinationLocal(const UUID & restore_uuid_, bool allow_concurrent_restore_, BackupConcurrencyCounters & concurrency_counters_); + RestoreCoordinationLocal(bool allow_concurrent_restore_, BackupConcurrencyCounters & concurrency_counters_); ~RestoreCoordinationLocal() override; + void setRestoreQueryIsSentToOtherHosts() override {} + bool isRestoreQuerySentToOtherHosts() const override { return false; } Strings setStage(const String &, const String &, bool) override { return {}; } - void setRestoreQueryWasSentToOtherHosts() override {} - bool trySetError(std::exception_ptr) override { return true; } - void finish() override {} - bool tryFinishAfterError() noexcept override { return true; } - void waitForOtherHostsToFinish() override {} - bool tryWaitForOtherHostsToFinishAfterError() noexcept override { return true; } + bool setError(std::exception_ptr, bool) override { return true; } + bool waitOtherHostsFinish(bool) const override { return true; } + bool finish(bool) override { return true; } + bool cleanup(bool) override { return true; } /// Starts creating a table in a replicated database. Returns false if there is another host which is already creating this table. bool acquireCreatingTableInReplicatedDatabase(const String & database_zk_path, const String & table_name) override; diff --git a/src/Backups/RestoreCoordinationOnCluster.cpp b/src/Backups/RestoreCoordinationOnCluster.cpp index 2029ad8b072..fad7341c044 100644 --- a/src/Backups/RestoreCoordinationOnCluster.cpp +++ b/src/Backups/RestoreCoordinationOnCluster.cpp @@ -35,17 +35,21 @@ RestoreCoordinationOnCluster::RestoreCoordinationOnCluster( , current_host_index(BackupCoordinationOnCluster::findCurrentHostIndex(current_host, all_hosts)) , log(getLogger("RestoreCoordinationOnCluster")) , with_retries(log, get_zookeeper_, keeper_settings, process_list_element_, [root_zookeeper_path_](Coordination::ZooKeeperWithFaultInjection::Ptr zk) { zk->sync(root_zookeeper_path_); }) - , concurrency_check(restore_uuid_, /* is_restore = */ true, /* on_cluster = */ true, allow_concurrent_restore_, concurrency_counters_) - , stage_sync(/* is_restore = */ true, fs::path{zookeeper_path} / "stage", current_host, all_hosts, allow_concurrent_restore_, with_retries, schedule_, process_list_element_, log) - , cleaner(zookeeper_path, with_retries, log) + , cleaner(/* is_restore = */ true, zookeeper_path, with_retries, log) + , stage_sync(/* is_restore = */ true, fs::path{zookeeper_path} / "stage", current_host, all_hosts, allow_concurrent_restore_, concurrency_counters_, with_retries, schedule_, process_list_element_, log) { - createRootNodes(); + try + { + createRootNodes(); + } + catch (...) + { + stage_sync.setError(std::current_exception(), /* throw_if_error = */ false); + throw; + } } -RestoreCoordinationOnCluster::~RestoreCoordinationOnCluster() -{ - tryFinishImpl(); -} +RestoreCoordinationOnCluster::~RestoreCoordinationOnCluster() = default; void RestoreCoordinationOnCluster::createRootNodes() { @@ -66,69 +70,52 @@ void RestoreCoordinationOnCluster::createRootNodes() }); } +void RestoreCoordinationOnCluster::setRestoreQueryIsSentToOtherHosts() +{ + stage_sync.setQueryIsSentToOtherHosts(); +} + +bool RestoreCoordinationOnCluster::isRestoreQuerySentToOtherHosts() const +{ + return stage_sync.isQuerySentToOtherHosts(); +} + Strings RestoreCoordinationOnCluster::setStage(const String & new_stage, const String & message, bool sync) { stage_sync.setStage(new_stage, message); - - if (!sync) - return {}; - - return stage_sync.waitForHostsToReachStage(new_stage, all_hosts_without_initiator); + if (sync) + return stage_sync.waitHostsReachStage(all_hosts_without_initiator, new_stage); + return {}; } -void RestoreCoordinationOnCluster::setRestoreQueryWasSentToOtherHosts() +bool RestoreCoordinationOnCluster::setError(std::exception_ptr exception, bool throw_if_error) { - restore_query_was_sent_to_other_hosts = true; + return stage_sync.setError(exception, throw_if_error); } -bool RestoreCoordinationOnCluster::trySetError(std::exception_ptr exception) +bool RestoreCoordinationOnCluster::waitOtherHostsFinish(bool throw_if_error) const { - return stage_sync.trySetError(exception); + return stage_sync.waitOtherHostsFinish(throw_if_error); } -void RestoreCoordinationOnCluster::finish() +bool RestoreCoordinationOnCluster::finish(bool throw_if_error) { - bool other_hosts_also_finished = false; - stage_sync.finish(other_hosts_also_finished); - - if ((current_host == kInitiator) && (other_hosts_also_finished || !restore_query_was_sent_to_other_hosts)) - cleaner.cleanup(); + return stage_sync.finish(throw_if_error); } -bool RestoreCoordinationOnCluster::tryFinishAfterError() noexcept +bool RestoreCoordinationOnCluster::cleanup(bool throw_if_error) { - return tryFinishImpl(); -} - -bool RestoreCoordinationOnCluster::tryFinishImpl() noexcept -{ - bool other_hosts_also_finished = false; - if (!stage_sync.tryFinishAfterError(other_hosts_also_finished)) - return false; - - if ((current_host == kInitiator) && (other_hosts_also_finished || !restore_query_was_sent_to_other_hosts)) + /// All the hosts must finish before we remove the coordination nodes. + bool expect_other_hosts_finished = stage_sync.isQuerySentToOtherHosts() || !stage_sync.isErrorSet(); + bool all_hosts_finished = stage_sync.finished() && (stage_sync.otherHostsFinished() || !expect_other_hosts_finished); + if (!all_hosts_finished) { - if (!cleaner.tryCleanupAfterError()) - return false; - } - - return true; -} - -void RestoreCoordinationOnCluster::waitForOtherHostsToFinish() -{ - if ((current_host != kInitiator) || !restore_query_was_sent_to_other_hosts) - return; - stage_sync.waitForOtherHostsToFinish(); -} - -bool RestoreCoordinationOnCluster::tryWaitForOtherHostsToFinishAfterError() noexcept -{ - if (current_host != kInitiator) + auto unfinished_hosts = expect_other_hosts_finished ? stage_sync.getUnfinishedHosts() : Strings{current_host}; + LOG_INFO(log, "Skipping removing nodes from ZooKeeper because hosts {} didn't finish", + BackupCoordinationStageSync::getHostsDesc(unfinished_hosts)); return false; - if (!restore_query_was_sent_to_other_hosts) - return true; - return stage_sync.tryWaitForOtherHostsToFinishAfterError(); + } + return cleaner.cleanup(throw_if_error); } ZooKeeperRetriesInfo RestoreCoordinationOnCluster::getOnClusterInitializationKeeperRetriesInfo() const diff --git a/src/Backups/RestoreCoordinationOnCluster.h b/src/Backups/RestoreCoordinationOnCluster.h index 87a8dd3ce83..99929cbdac3 100644 --- a/src/Backups/RestoreCoordinationOnCluster.h +++ b/src/Backups/RestoreCoordinationOnCluster.h @@ -1,7 +1,6 @@ #pragma once #include -#include #include #include #include @@ -15,7 +14,7 @@ class RestoreCoordinationOnCluster : public IRestoreCoordination { public: /// Empty string as the current host is used to mark the initiator of a RESTORE ON CLUSTER query. - static const constexpr std::string_view kInitiator; + static const constexpr std::string_view kInitiator = BackupCoordinationStageSync::kInitiator; RestoreCoordinationOnCluster( const UUID & restore_uuid_, @@ -31,13 +30,13 @@ public: ~RestoreCoordinationOnCluster() override; + void setRestoreQueryIsSentToOtherHosts() override; + bool isRestoreQuerySentToOtherHosts() const override; Strings setStage(const String & new_stage, const String & message, bool sync) override; - void setRestoreQueryWasSentToOtherHosts() override; - bool trySetError(std::exception_ptr exception) override; - void finish() override; - bool tryFinishAfterError() noexcept override; - void waitForOtherHostsToFinish() override; - bool tryWaitForOtherHostsToFinishAfterError() noexcept override; + bool setError(std::exception_ptr exception, bool throw_if_error) override; + bool waitOtherHostsFinish(bool throw_if_error) const override; + bool finish(bool throw_if_error) override; + bool cleanup(bool throw_if_error) override; /// Starts creating a table in a replicated database. Returns false if there is another host which is already creating this table. bool acquireCreatingTableInReplicatedDatabase(const String & database_zk_path, const String & table_name) override; @@ -78,11 +77,10 @@ private: const size_t current_host_index; LoggerPtr const log; + /// The order is important: `stage_sync` must be initialized after `with_retries` and `cleaner`. const WithRetries with_retries; - BackupConcurrencyCheck concurrency_check; - BackupCoordinationStageSync stage_sync; BackupCoordinationCleaner cleaner; - std::atomic restore_query_was_sent_to_other_hosts = false; + BackupCoordinationStageSync stage_sync; }; } From 19bcc5550bad0444d652d760edddbe15fe0611da Mon Sep 17 00:00:00 2001 From: Vitaly Baranov Date: Mon, 11 Nov 2024 01:36:20 +0100 Subject: [PATCH 127/244] Fix tests. --- .../test_cancel_backup.py | 29 +++++++------------ 1 file changed, 11 insertions(+), 18 deletions(-) diff --git a/tests/integration/test_backup_restore_on_cluster/test_cancel_backup.py b/tests/integration/test_backup_restore_on_cluster/test_cancel_backup.py index f63dc2aef3d..4ad53acc735 100644 --- a/tests/integration/test_backup_restore_on_cluster/test_cancel_backup.py +++ b/tests/integration/test_backup_restore_on_cluster/test_cancel_backup.py @@ -251,23 +251,16 @@ def kill_query( if is_initial_query is not None else "" ) + old_time = time.monotonic() node.query( f"KILL QUERY WHERE (query_kind='{query_kind}') AND (query LIKE '%{id}%'){filter_for_is_initial_query} SYNC" ) - node.query("SYSTEM FLUSH LOGS") - duration = ( - int( - node.query( - f"SELECT query_duration_ms FROM system.query_log WHERE query_kind='KillQuery' AND query LIKE '%{id}%' AND type='QueryFinish'" - ) - ) - / 1000 - ) + waited = time.monotonic() - old_time print( - f"{get_node_name(node)}: Cancelled {operation_name} {id} after {duration} seconds" + f"{get_node_name(node)}: Cancelled {operation_name} {id} after {waited} seconds" ) if timeout is not None: - assert duration < timeout + assert waited < timeout # Stops all ZooKeeper servers. @@ -305,7 +298,7 @@ def sleep(seconds): class NoTrashChecker: def __init__(self): self.expect_backups = [] - self.expect_unfinished_backups = [] + self.allow_unfinished_backups = [] self.expect_errors = [] self.allow_errors = [] self.check_zookeeper = True @@ -373,7 +366,7 @@ class NoTrashChecker: if unfinished_backups: print(f"Found unfinished backups: {unfinished_backups}") assert new_backups == set(self.expect_backups) - assert unfinished_backups == set(self.expect_unfinished_backups) + assert unfinished_backups.difference(self.allow_unfinished_backups) == set() all_errors = set() start_time = time.strftime( @@ -641,7 +634,7 @@ def test_long_disconnection_stops_backup(): assert get_status(initiator, backup_id=backup_id) == "CREATING_BACKUP" assert get_num_system_processes(initiator, backup_id=backup_id) >= 1 - no_trash_checker.expect_unfinished_backups = [backup_id] + no_trash_checker.allow_unfinished_backups = [backup_id] no_trash_checker.allow_errors = [ "FAILED_TO_SYNC_BACKUP_OR_RESTORE", "KEEPER_EXCEPTION", @@ -674,7 +667,7 @@ def test_long_disconnection_stops_backup(): # A backup is expected to fail, but it isn't expected to fail too soon. print(f"Backup failed after {time_to_fail} seconds disconnection") assert time_to_fail > 3 - assert time_to_fail < 30 + assert time_to_fail < 35 # A backup must NOT be stopped if Zookeeper is disconnected shorter than `failure_after_host_disconnected_for_seconds`. @@ -695,7 +688,7 @@ def test_short_disconnection_doesnt_stop_backup(): backup_id = random_id() initiator.query( f"BACKUP TABLE tbl ON CLUSTER 'cluster' TO {get_backup_name(backup_id)} SETTINGS id='{backup_id}' ASYNC", - settings={"backup_restore_failure_after_host_disconnected_for_seconds": 6}, + settings={"backup_restore_failure_after_host_disconnected_for_seconds": 10}, ) assert get_status(initiator, backup_id=backup_id) == "CREATING_BACKUP" @@ -703,13 +696,13 @@ def test_short_disconnection_doesnt_stop_backup(): # Dropping connection for less than `failure_after_host_disconnected_for_seconds` with PartitionManager() as pm: - random_sleep(3) + random_sleep(4) node_to_drop_zk_connection = random_node() print( f"Dropping connection between {get_node_name(node_to_drop_zk_connection)} and ZooKeeper" ) pm.drop_instance_zk_connections(node_to_drop_zk_connection) - random_sleep(3) + random_sleep(4) print( f"Restoring connection between {get_node_name(node_to_drop_zk_connection)} and ZooKeeper" ) From 386e16bee24f5bbec50d7b99a1e302c2d01df592 Mon Sep 17 00:00:00 2001 From: kssenii Date: Mon, 11 Nov 2024 12:36:09 +0100 Subject: [PATCH 128/244] Better --- src/Interpreters/Cache/FileCache.cpp | 15 +++--------- src/Interpreters/Cache/FileCache.h | 2 ++ src/Interpreters/Cache/FileCacheUtils.h | 17 +++++++++++++ src/Interpreters/Cache/FileSegment.cpp | 32 +++++++++++++++++++++---- src/Interpreters/Cache/FileSegment.h | 3 +++ src/Interpreters/Cache/Metadata.cpp | 30 ++++++++--------------- 6 files changed, 63 insertions(+), 36 deletions(-) create mode 100644 src/Interpreters/Cache/FileCacheUtils.h diff --git a/src/Interpreters/Cache/FileCache.cpp b/src/Interpreters/Cache/FileCache.cpp index bda91b31692..47b5779dd1a 100644 --- a/src/Interpreters/Cache/FileCache.cpp +++ b/src/Interpreters/Cache/FileCache.cpp @@ -8,6 +8,7 @@ #include #include #include +#include #include #include #include @@ -53,16 +54,6 @@ namespace ErrorCodes namespace { - size_t roundDownToMultiple(size_t num, size_t multiple) - { - return (num / multiple) * multiple; - } - - size_t roundUpToMultiple(size_t num, size_t multiple) - { - return roundDownToMultiple(num + multiple - 1, multiple); - } - std::string getCommonUserID() { auto user_from_context = DB::Context::getGlobalContextInstance()->getFilesystemCacheUser(); @@ -605,8 +596,8 @@ FileCache::getOrSet( /// 2. max_file_segments_limit FileSegment::Range result_range = initial_range; - const auto aligned_offset = roundDownToMultiple(initial_range.left, boundary_alignment); - auto aligned_end_offset = std::min(roundUpToMultiple(initial_range.right + 1, boundary_alignment), file_size) - 1; + const auto aligned_offset = FileCacheUtils::roundDownToMultiple(initial_range.left, boundary_alignment); + auto aligned_end_offset = std::min(FileCacheUtils::roundUpToMultiple(initial_range.right + 1, boundary_alignment), file_size) - 1; chassert(aligned_offset <= initial_range.left); chassert(aligned_end_offset >= initial_range.right); diff --git a/src/Interpreters/Cache/FileCache.h b/src/Interpreters/Cache/FileCache.h index 79966e60ad9..3a9241fd351 100644 --- a/src/Interpreters/Cache/FileCache.h +++ b/src/Interpreters/Cache/FileCache.h @@ -163,6 +163,8 @@ public: size_t getBackgroundDownloadMaxFileSegmentSize() const { return metadata.getBackgroundDownloadMaxFileSegmentSize(); } + size_t getBoundaryAlignment() const { return boundary_alignment; } + bool tryReserve( FileSegment & file_segment, size_t size, diff --git a/src/Interpreters/Cache/FileCacheUtils.h b/src/Interpreters/Cache/FileCacheUtils.h new file mode 100644 index 00000000000..b35ce867a79 --- /dev/null +++ b/src/Interpreters/Cache/FileCacheUtils.h @@ -0,0 +1,17 @@ +#pragma once +#include + +namespace FileCacheUtils +{ + +static size_t roundDownToMultiple(size_t num, size_t multiple) +{ + return (num / multiple) * multiple; +} + +static size_t roundUpToMultiple(size_t num, size_t multiple) +{ + return roundDownToMultiple(num + multiple - 1, multiple); +} + +} diff --git a/src/Interpreters/Cache/FileSegment.cpp b/src/Interpreters/Cache/FileSegment.cpp index 2455461435b..05b0853da07 100644 --- a/src/Interpreters/Cache/FileSegment.cpp +++ b/src/Interpreters/Cache/FileSegment.cpp @@ -4,6 +4,7 @@ #include #include #include +#include #include #include #include @@ -629,6 +630,31 @@ void FileSegment::completePartAndResetDownloader() LOG_TEST(log, "Complete batch. ({})", getInfoForLogUnlocked(lk)); } +size_t FileSegment::getSizeForBackgroundDownload() const +{ + auto lk = lock(); + return getSizeForBackgroundDownloadUnlocked(lk); +} + +size_t FileSegment::getSizeForBackgroundDownloadUnlocked(const FileSegmentGuard::Lock &) const +{ + if (!background_download_enabled + || !downloaded_size + || !remote_file_reader) + { + return 0; + } + + const size_t background_download_max_file_segment_size = cache->getBackgroundDownloadMaxFileSegmentSize(); + size_t desired_size; + if (downloaded_size >= background_download_max_file_segment_size) + desired_size = FileCacheUtils::roundUpToMultiple(downloaded_size, cache->getBoundaryAlignment()); + else + desired_size = FileCacheUtils::roundUpToMultiple(background_download_max_file_segment_size, cache->getBoundaryAlignment()); + + return desired_size - downloaded_size; +} + void FileSegment::complete(bool allow_background_download) { ProfileEventTimeIncrement watch(ProfileEvents::FileSegmentCompleteMicroseconds); @@ -708,10 +734,8 @@ void FileSegment::complete(bool allow_background_download) if (is_last_holder) { bool added_to_download_queue = false; - if (allow_background_download - && background_download_enabled - && remote_file_reader - && downloaded_size < cache->getBackgroundDownloadMaxFileSegmentSize()) + size_t background_download_size = allow_background_download ? getSizeForBackgroundDownloadUnlocked(segment_lock) : 0; + if (background_download_size) { ProfileEvents::increment(ProfileEvents::FilesystemCacheBackgroundDownloadQueuePush); added_to_download_queue = locked_key->addToDownloadQueue(offset(), segment_lock); /// Finish download in background. diff --git a/src/Interpreters/Cache/FileSegment.h b/src/Interpreters/Cache/FileSegment.h index 21d5f9dab5f..a6bfb203cec 100644 --- a/src/Interpreters/Cache/FileSegment.h +++ b/src/Interpreters/Cache/FileSegment.h @@ -185,6 +185,8 @@ public: bool assertCorrectness() const; + size_t getSizeForBackgroundDownload() const; + /** * ========== Methods that must do cv.notify() ================== */ @@ -230,6 +232,7 @@ private: String getDownloaderUnlocked(const FileSegmentGuard::Lock &) const; bool isDownloaderUnlocked(const FileSegmentGuard::Lock & segment_lock) const; void resetDownloaderUnlocked(const FileSegmentGuard::Lock &); + size_t getSizeForBackgroundDownloadUnlocked(const FileSegmentGuard::Lock &) const; void setDownloadState(State state, const FileSegmentGuard::Lock &); void resetDownloadingStateUnlocked(const FileSegmentGuard::Lock &); diff --git a/src/Interpreters/Cache/Metadata.cpp b/src/Interpreters/Cache/Metadata.cpp index 2ef8f76aca0..45077f09020 100644 --- a/src/Interpreters/Cache/Metadata.cpp +++ b/src/Interpreters/Cache/Metadata.cpp @@ -632,9 +632,6 @@ void CacheMetadata::downloadThreadFunc(const bool & stop_flag) auto & file_segment = holder->front(); - if (file_segment.getDownloadedSize() >= download_max_file_segment_size) - continue; - if (file_segment.getOrSetDownloader() != FileSegment::getCallerId()) continue; @@ -681,6 +678,10 @@ void CacheMetadata::downloadImpl(FileSegment & file_segment, std::optionalinternalBuffer().empty()) { if (!memory) - memory.emplace(DBMS_DEFAULT_BUFFER_SIZE); + memory.emplace(std::min(size_t(DBMS_DEFAULT_BUFFER_SIZE), size_to_download)); reader->set(memory->data(), memory->size()); } @@ -706,24 +707,13 @@ void CacheMetadata::downloadImpl(FileSegment & file_segment, std::optional(reader->getPosition())) reader->seek(offset, SEEK_SET); - bool stop = false; - const size_t max_file_segment_size = download_max_file_segment_size.load(); - - while (!stop && !reader->eof()) + while (size_to_download && !reader->eof()) { - auto size = reader->available(); + const auto available = reader->available(); + chassert(available); - const size_t downloaded_size = file_segment.getDownloadedSize(); - if (downloaded_size >= max_file_segment_size) - break; - - if (downloaded_size + size > max_file_segment_size) - { - /// Do not download more than download_max_file_segment_size - /// because we want to leave right boundary of file segment aligned. - size = max_file_segment_size - downloaded_size; - stop = true; - } + const auto size = std::min(available, size_to_download); + size_to_download -= size; std::string failure_reason; if (!file_segment.reserve(size, reserve_space_lock_wait_timeout_milliseconds, failure_reason)) From 39e01d47b1892b2049d18fc19803949d1bfcda51 Mon Sep 17 00:00:00 2001 From: Igor Nikonov Date: Mon, 11 Nov 2024 11:54:04 +0000 Subject: [PATCH 129/244] Fix style check --- src/Planner/findParallelReplicasQuery.cpp | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/Planner/findParallelReplicasQuery.cpp b/src/Planner/findParallelReplicasQuery.cpp index fbe2993b7c6..494326c0ed0 100644 --- a/src/Planner/findParallelReplicasQuery.cpp +++ b/src/Planner/findParallelReplicasQuery.cpp @@ -227,7 +227,7 @@ const QueryNode * findQueryForParallelReplicas( const auto & children = next_node_to_check->children; auto * step = next_node_to_check->step.get(); - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Step {} childrens={}", step->getName(), children.size()); + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Step {} children={}", step->getName(), children.size()); if (children.empty()) { @@ -347,7 +347,7 @@ const QueryNode * findQueryForParallelReplicas(const QueryTreeNodePtr & query_tr } if (res) - LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Choosen query: {}", res->dumpTree()); + LOG_DEBUG(getLogger(__PRETTY_FUNCTION__), "Chosen query: {}", res->dumpTree()); return res; } From 33f9e8bc2e5540386e5ccf7fec591eaa1bf5cc24 Mon Sep 17 00:00:00 2001 From: nauu Date: Mon, 11 Nov 2024 20:25:57 +0800 Subject: [PATCH 130/244] fix error --- src/IO/S3/URI.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/IO/S3/URI.cpp b/src/IO/S3/URI.cpp index ad746ff3326..aefe3ff338c 100644 --- a/src/IO/S3/URI.cpp +++ b/src/IO/S3/URI.cpp @@ -117,7 +117,7 @@ URI::URI(const std::string & uri_, bool allow_archive_path_syntax) is_virtual_hosted_style = true; if (name == "oss-data-acc") { - bucket = bucket.substr(0, bucket.find(".")); + bucket = bucket.substr(0, bucket.find('.')); endpoint = uri.getScheme() + "://" + uri.getHost().substr(bucket.length() + 1); } else From 288756bc9aede92c6d005af34be94973a5d78203 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Mon, 11 Nov 2024 13:32:01 +0000 Subject: [PATCH 131/244] Fix for stateful functions. --- .../QueryPlan/BuildQueryPipelineSettings.cpp | 13 ++++++++++++- .../QueryPlan/BuildQueryPipelineSettings.h | 2 ++ src/Processors/QueryPlan/FilterStep.cpp | 12 ++++++++++-- .../queries/0_stateless/03199_merge_filters_bug.sql | 2 +- 4 files changed, 25 insertions(+), 4 deletions(-) diff --git a/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp b/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp index fb3ed7f80fc..ce02ef8b9ba 100644 --- a/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp +++ b/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp @@ -6,12 +6,23 @@ namespace DB { +namespace Setting +{ + extern const SettingsBool query_plan_merge_filters; +} + BuildQueryPipelineSettings BuildQueryPipelineSettings::fromContext(ContextPtr from) { + const auto & query_settings = from->getSettingsRef(); BuildQueryPipelineSettings settings; - settings.actions_settings = ExpressionActionsSettings::fromSettings(from->getSettingsRef(), CompileExpressions::yes); + settings.actions_settings = ExpressionActionsSettings::fromSettings(query_settings, CompileExpressions::yes); settings.process_list_element = from->getProcessListElement(); settings.progress_callback = from->getProgressCallback(); + + /// Setting query_plan_merge_filters is enabled by default. + /// But it can brake short-circuit without splitting fiter step into smaller steps. + /// So, enable and disable this optimizations together. + settings.enable_multiple_filters_transforms_for_and_chain = query_settings[Setting::query_plan_merge_filters]; return settings; } diff --git a/src/Processors/QueryPlan/BuildQueryPipelineSettings.h b/src/Processors/QueryPlan/BuildQueryPipelineSettings.h index d99f9a7d1f1..6219e37db58 100644 --- a/src/Processors/QueryPlan/BuildQueryPipelineSettings.h +++ b/src/Processors/QueryPlan/BuildQueryPipelineSettings.h @@ -17,6 +17,8 @@ using TemporaryFileLookupPtr = std::shared_ptr; struct BuildQueryPipelineSettings { + bool enable_multiple_filters_transforms_for_and_chain = true; + ExpressionActionsSettings actions_settings; QueryStatusPtr process_list_element; ProgressCallback progress_callback = nullptr; diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp index 3d56a2352dc..a6b157cdd1d 100644 --- a/src/Processors/QueryPlan/FilterStep.cpp +++ b/src/Processors/QueryPlan/FilterStep.cpp @@ -139,7 +139,11 @@ FilterStep::FilterStep( void FilterStep::transformPipeline(QueryPipelineBuilder & pipeline, const BuildQueryPipelineSettings & settings) { - auto and_atoms = splitAndChainIntoMultipleFilters(actions_dag, filter_column_name); + std::vector and_atoms; + + if (settings.enable_multiple_filters_transforms_for_and_chain && !actions_dag.hasStatefulFunctions()) + and_atoms = splitAndChainIntoMultipleFilters(actions_dag, filter_column_name); + for (auto & and_atom : and_atoms) { auto expression = std::make_shared(std::move(and_atom.dag), settings.getActionsSettings()); @@ -178,7 +182,11 @@ void FilterStep::describeActions(FormatSettings & settings) const String prefix(settings.offset, settings.indent_char); auto cloned_dag = actions_dag.clone(); - auto and_atoms = splitAndChainIntoMultipleFilters(cloned_dag, filter_column_name); + + std::vector and_atoms; + if (!actions_dag.hasStatefulFunctions()) + and_atoms = splitAndChainIntoMultipleFilters(cloned_dag, filter_column_name); + for (auto & and_atom : and_atoms) { auto expression = std::make_shared(std::move(and_atom.dag)); diff --git a/tests/queries/0_stateless/03199_merge_filters_bug.sql b/tests/queries/0_stateless/03199_merge_filters_bug.sql index ed2ec2ea217..bb2a4255a3d 100644 --- a/tests/queries/0_stateless/03199_merge_filters_bug.sql +++ b/tests/queries/0_stateless/03199_merge_filters_bug.sql @@ -49,7 +49,7 @@ tmp1 AS fs1 FROM t2 LEFT JOIN tmp1 USING (fs1) - WHERE (fs1 IN ('test')) SETTINGS enable_multiple_prewhere_read_steps = 0; + WHERE (fs1 IN ('test')) SETTINGS enable_multiple_prewhere_read_steps = 0, query_plan_merge_filters=0; optimize table t1 final; From 8c2e541392e552343431a6b9b411ee55f37e8fe8 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Mon, 11 Nov 2024 14:27:48 +0000 Subject: [PATCH 132/244] Avoid using manes in multistage prewhere optimization. --- .../MergeTreeSplitPrewhereIntoReadSteps.cpp | 110 ++++++++++-------- 1 file changed, 60 insertions(+), 50 deletions(-) diff --git a/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp b/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp index 9c82817e8cb..73fe2600946 100644 --- a/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp +++ b/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp @@ -4,6 +4,7 @@ #include #include #include +#include #include @@ -57,9 +58,9 @@ struct DAGNodeRef const ActionsDAG::Node * node; }; -/// Result name -> DAGNodeRef -using OriginalToNewNodeMap = std::unordered_map; -using NodeNameToLastUsedStepMap = std::unordered_map; +/// Result -> DAGNodeRef +using OriginalToNewNodeMap = std::unordered_map; +using NodeNameToLastUsedStepMap = std::unordered_map; /// Clones the part of original DAG responsible for computing the original_dag_node and adds it to the new DAG. const ActionsDAG::Node & addClonedDAGToDAG( @@ -69,12 +70,12 @@ const ActionsDAG::Node & addClonedDAGToDAG( OriginalToNewNodeMap & node_remap, NodeNameToLastUsedStepMap & node_to_step_map) { - const String & node_name = original_dag_node->result_name; + //const String & node_name = original_dag_node->result_name; /// Look for the node in the map of already known nodes - if (node_remap.contains(node_name)) + if (node_remap.contains(original_dag_node)) { /// If the node is already in the new DAG, return it - const auto & node_ref = node_remap.at(node_name); + const auto & node_ref = node_remap.at(original_dag_node); if (node_ref.dag == new_dag.get()) return *node_ref.node; @@ -83,11 +84,11 @@ const ActionsDAG::Node & addClonedDAGToDAG( { node_ref.dag->addOrReplaceInOutputs(*node_ref.node); const auto & new_node = new_dag->addInput(node_ref.node->result_name, node_ref.node->result_type); - node_remap[node_name] = {new_dag.get(), &new_node}; /// TODO: here we update the node reference. Is it always correct? + node_remap[original_dag_node] = {new_dag.get(), &new_node}; /// TODO: here we update the node reference. Is it always correct? /// Remember the index of the last step which reuses this node. /// We cannot remove this node from the outputs before that step. - node_to_step_map[node_name] = step; + node_to_step_map[original_dag_node] = step; return new_node; } } @@ -96,7 +97,7 @@ const ActionsDAG::Node & addClonedDAGToDAG( if (original_dag_node->type == ActionsDAG::ActionType::INPUT) { const auto & new_node = new_dag->addInput(original_dag_node->result_name, original_dag_node->result_type); - node_remap[node_name] = {new_dag.get(), &new_node}; + node_remap[original_dag_node] = {new_dag.get(), &new_node}; return new_node; } @@ -105,7 +106,7 @@ const ActionsDAG::Node & addClonedDAGToDAG( { const auto & new_node = new_dag->addColumn( ColumnWithTypeAndName(original_dag_node->column, original_dag_node->result_type, original_dag_node->result_name)); - node_remap[node_name] = {new_dag.get(), &new_node}; + node_remap[original_dag_node] = {new_dag.get(), &new_node}; return new_node; } @@ -113,7 +114,7 @@ const ActionsDAG::Node & addClonedDAGToDAG( { const auto & alias_child = addClonedDAGToDAG(step, original_dag_node->children[0], new_dag, node_remap, node_to_step_map); const auto & new_node = new_dag->addAlias(alias_child, original_dag_node->result_name); - node_remap[node_name] = {new_dag.get(), &new_node}; + node_remap[original_dag_node] = {new_dag.get(), &new_node}; return new_node; } @@ -128,7 +129,7 @@ const ActionsDAG::Node & addClonedDAGToDAG( } const auto & new_node = new_dag->addFunction(original_dag_node->function_base, new_children, original_dag_node->result_name); - node_remap[node_name] = {new_dag.get(), &new_node}; + node_remap[original_dag_node] = {new_dag.get(), &new_node}; return new_node; } @@ -138,11 +139,11 @@ const ActionsDAG::Node & addClonedDAGToDAG( const ActionsDAG::Node & addFunction( const ActionsDAGPtr & new_dag, const FunctionOverloadResolverPtr & function, - ActionsDAG::NodeRawConstPtrs children, - OriginalToNewNodeMap & node_remap) + ActionsDAG::NodeRawConstPtrs children) + //OriginalToNewNodeMap & node_remap) { const auto & new_node = new_dag->addFunction(function, children, ""); - node_remap[new_node.result_name] = {new_dag.get(), &new_node}; + //node_remap[new_node.result_name] = {new_dag.get(), &new_node}; return new_node; } @@ -152,14 +153,14 @@ const ActionsDAG::Node & addFunction( const ActionsDAG::Node & addCast( const ActionsDAGPtr & dag, const ActionsDAG::Node & node_to_cast, - const DataTypePtr & to_type, - OriginalToNewNodeMap & node_remap) + const DataTypePtr & to_type) + //[[maybe_unused]] OriginalToNewNodeMap & node_remap) { if (!node_to_cast.result_type->equals(*to_type)) return node_to_cast; const auto & new_node = dag->addCast(node_to_cast, to_type, {}); - node_remap[new_node.result_name] = {dag.get(), &new_node}; + //node_remap[new_node.result_name] = {dag.get(), &new_node}; return new_node; } @@ -169,8 +170,8 @@ const ActionsDAG::Node & addCast( /// 2. makes sure that the result contains only 0 or 1 values even if the source column contains non-boolean values. const ActionsDAG::Node & addAndTrue( const ActionsDAGPtr & dag, - const ActionsDAG::Node & filter_node_to_normalize, - OriginalToNewNodeMap & node_remap) + const ActionsDAG::Node & filter_node_to_normalize) + //OriginalToNewNodeMap & node_remap) { Field const_true_value(true); @@ -181,7 +182,7 @@ const ActionsDAG::Node & addAndTrue( const auto * const_true_node = &dag->addColumn(std::move(const_true_column)); ActionsDAG::NodeRawConstPtrs children = {&filter_node_to_normalize, const_true_node}; FunctionOverloadResolverPtr func_builder_and = std::make_unique(std::make_shared()); - return addFunction(dag, func_builder_and, children, node_remap); + return addFunction(dag, func_builder_and, children); //, node_remap); } } @@ -243,7 +244,11 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction struct Step { ActionsDAGPtr actions; - String column_name; + /// Original condition, in case if we have only one condition, and it was not casted + const ActionsDAG::Node * original_node; + /// Result condition node + const ActionsDAG::Node * result_node; + //String column_name; }; std::vector steps; @@ -254,7 +259,9 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction { const auto & condition_group = condition_groups[step_index]; ActionsDAGPtr step_dag = std::make_unique(); - String result_name; + const ActionsDAG::Node * original_node = nullptr; + const ActionsDAG::Node * result_node; + //String result_name; std::vector new_condition_nodes; for (const auto * node : condition_group) @@ -267,48 +274,47 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction { /// Add AND function to combine the conditions FunctionOverloadResolverPtr func_builder_and = std::make_unique(std::make_shared()); - const auto & and_function_node = addFunction(step_dag, func_builder_and, new_condition_nodes, node_remap); - step_dag->addOrReplaceInOutputs(and_function_node); - result_name = and_function_node.result_name; + const auto & and_function_node = addFunction(step_dag, func_builder_and, new_condition_nodes); //, node_remap); + //step_dag->addOrReplaceInOutputs(and_function_node); + result_node = &and_function_node; } else { - const auto & result_node = *new_condition_nodes.front(); + result_node = new_condition_nodes.front(); /// Check if explicit cast is needed for the condition to serve as a filter. - const auto result_type_name = result_node.result_type->getName(); - if (result_type_name == "UInt8" || - result_type_name == "Nullable(UInt8)" || - result_type_name == "LowCardinality(UInt8)" || - result_type_name == "LowCardinality(Nullable(UInt8))") + //const auto result_type_name = result_node->result_type->getName(); + if (isUInt8(removeNullable(removeLowCardinality(result_node->result_type)))) { /// No need to cast - step_dag->addOrReplaceInOutputs(result_node); - result_name = result_node.result_name; + //step_dag->addOrReplaceInOutputs(result_node); + //result_name = result_node.result_name; } else { /// Build "condition AND True" expression to "cast" the condition to UInt8 or Nullable(UInt8) depending on its type. - const auto & cast_node = addAndTrue(step_dag, result_node, node_remap); - step_dag->addOrReplaceInOutputs(cast_node); - result_name = cast_node.result_name; + result_node = &addAndTrue(step_dag, *result_node); //, node_remap); + //step_dag->addOrReplaceInOutputs(cast_node); + //result_name = &cast_node.result_name; } } - steps.push_back({std::move(step_dag), result_name}); + step_dag->getOutputs().insert(step_dag->getOutputs().begin(), result_node); + steps.push_back({std::move(step_dag), original_node, result_node}); } /// 6. Find all outputs of the original DAG auto original_outputs = prewhere_info->prewhere_actions.getOutputs(); + steps.back().actions->getOutputs().clear(); /// 7. Find all outputs that were computed in the already built DAGs, mark these nodes as outputs in the steps where they were computed /// 8. Add computation of the remaining outputs to the last step with the procedure similar to 4 - NameSet all_output_names; + std::unordered_set all_outputs; for (const auto * output : original_outputs) { - all_output_names.insert(output->result_name); - if (node_remap.contains(output->result_name)) + all_outputs.insert(output); + if (node_remap.contains(output)) //->result_name)) { - const auto & new_node_info = node_remap[output->result_name]; - new_node_info.dag->addOrReplaceInOutputs(*new_node_info.node); + const auto & new_node_info = node_remap[output]; + new_node_info.dag->getOutputs().push_back(new_node_info.node); } else if (output->result_name == prewhere_info->prewhere_column_name) { @@ -319,20 +325,23 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction /// 1. AND the last condition with constant True. This is needed to make sure that in the last step filter has UInt8 type /// but contains values other than 0 and 1 (e.g. if it is (number%5) it contains 2,3,4) /// 2. CAST the result to the exact type of the PREWHERE column from the original DAG - const auto & last_step_result_node_info = node_remap[steps.back().column_name]; + //const auto & last_step_result_node_info = node_remap[steps.back().column_name]; auto & last_step_dag = steps.back().actions; + auto & last_step_result_node = steps.back().result_node; /// Build AND(last_step_result_node, true) - const auto & and_node = addAndTrue(last_step_dag, *last_step_result_node_info.node, node_remap); + const auto & and_node = addAndTrue(last_step_dag, *last_step_result_node); //, node_remap); /// Build CAST(and_node, type of PREWHERE column) - const auto & cast_node = addCast(last_step_dag, and_node, output->result_type, node_remap); + const auto & cast_node = addCast(last_step_dag, and_node, output->result_type); //, node_remap); /// Add alias for the result with the name of the PREWHERE column const auto & prewhere_result_node = last_step_dag->addAlias(cast_node, output->result_name); - last_step_dag->addOrReplaceInOutputs(prewhere_result_node); + //last_step_dag->addOrReplaceInOutputs(prewhere_result_node); + last_step_dag->getOutputs().push_back(&prewhere_result_node); + steps.back().result_node = &prewhere_result_node; } else { const auto & node_in_new_dag = addClonedDAGToDAG(steps.size() - 1, output, steps.back().actions, node_remap, node_to_step); - steps.back().actions->addOrReplaceInOutputs(node_in_new_dag); + steps.back().actions->getOutputs().push_back(&node_in_new_dag); } } @@ -345,10 +354,10 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction { .type = PrewhereExprStep::Filter, .actions = std::make_shared(std::move(*step.actions), actions_settings), - .filter_column_name = step.column_name, + .filter_column_name = step.result_node->result_name, /// Don't remove if it's in the list of original outputs .remove_filter_column = - !all_output_names.contains(step.column_name) && node_to_step[step.column_name] <= step_index, + step.original_node && !all_outputs.contains(step.original_node) && node_to_step[step.original_node] <= step_index, .need_filter = false, .perform_alter_conversions = true, }; @@ -356,6 +365,7 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction prewhere.steps.push_back(std::make_shared(std::move(new_step))); } + prewhere.steps.back()->remove_filter_column = prewhere_info->remove_prewhere_column; prewhere.steps.back()->need_filter = prewhere_info->need_filter; } From 5c5016218b77cf77323c126bc064d90601beadef Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Mon, 11 Nov 2024 15:05:53 +0000 Subject: [PATCH 133/244] Fixing style. --- src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp b/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp index ce02ef8b9ba..1832cc2ad42 100644 --- a/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp +++ b/src/Processors/QueryPlan/BuildQueryPipelineSettings.cpp @@ -20,7 +20,7 @@ BuildQueryPipelineSettings BuildQueryPipelineSettings::fromContext(ContextPtr fr settings.progress_callback = from->getProgressCallback(); /// Setting query_plan_merge_filters is enabled by default. - /// But it can brake short-circuit without splitting fiter step into smaller steps. + /// But it can brake short-circuit without splitting filter step into smaller steps. /// So, enable and disable this optimizations together. settings.enable_multiple_filters_transforms_for_and_chain = query_settings[Setting::query_plan_merge_filters]; return settings; From 40c4183ae70c720aaca797b165e3cf71aa4d8133 Mon Sep 17 00:00:00 2001 From: Sema Checherinda Date: Mon, 11 Nov 2024 17:26:28 +0100 Subject: [PATCH 134/244] fix tidy build --- src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h | 10 +++++----- src/Disks/ObjectStorages/Local/LocalObjectStorage.cpp | 4 ++-- src/Disks/ObjectStorages/Local/LocalObjectStorage.h | 8 ++++---- 3 files changed, 11 insertions(+), 11 deletions(-) diff --git a/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h b/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h index 317399b4753..7d6c914c398 100644 --- a/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h +++ b/src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.h @@ -77,11 +77,6 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - /// Remove file. Throws exception if file doesn't exists or it's a directory. - void removeObject(const StoredObject & object); - - void removeObjects(const StoredObjects & objects); - void removeObjectIfExists(const StoredObject & object) override; void removeObjectsIfExist(const StoredObjects & objects) override; @@ -117,6 +112,11 @@ private: void initializeHDFSFS() const; std::string extractObjectKeyFromURL(const StoredObject & object) const; + /// Remove file. Throws exception if file doesn't exists or it's a directory. + void removeObject(const StoredObject & object); + + void removeObjects(const StoredObjects & objects); + const Poco::Util::AbstractConfiguration & config; mutable HDFSBuilderWrapper hdfs_builder; diff --git a/src/Disks/ObjectStorages/Local/LocalObjectStorage.cpp b/src/Disks/ObjectStorages/Local/LocalObjectStorage.cpp index 5f1b6aedc72..f24501dc60e 100644 --- a/src/Disks/ObjectStorages/Local/LocalObjectStorage.cpp +++ b/src/Disks/ObjectStorages/Local/LocalObjectStorage.cpp @@ -81,7 +81,7 @@ std::unique_ptr LocalObjectStorage::writeObject( /// NO return std::make_unique(object.remote_path, buf_size); } -void LocalObjectStorage::removeObject(const StoredObject & object) +void LocalObjectStorage::removeObject(const StoredObject & object) const { /// For local object storage files are actually removed when "metadata" is removed. if (!exists(object)) @@ -91,7 +91,7 @@ void LocalObjectStorage::removeObject(const StoredObject & object) ErrnoException::throwFromPath(ErrorCodes::CANNOT_UNLINK, object.remote_path, "Cannot unlink file {}", object.remote_path); } -void LocalObjectStorage::removeObjects(const StoredObjects & objects) +void LocalObjectStorage::removeObjects(const StoredObjects & objects) const { for (const auto & object : objects) removeObject(object); diff --git a/src/Disks/ObjectStorages/Local/LocalObjectStorage.h b/src/Disks/ObjectStorages/Local/LocalObjectStorage.h index ffc151bda04..5b3c3951364 100644 --- a/src/Disks/ObjectStorages/Local/LocalObjectStorage.h +++ b/src/Disks/ObjectStorages/Local/LocalObjectStorage.h @@ -42,10 +42,6 @@ public: size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE, const WriteSettings & write_settings = {}) override; - void removeObject(const StoredObject & object); - - void removeObjects(const StoredObjects & objects); - void removeObjectIfExists(const StoredObject & object) override; void removeObjectsIfExist(const StoredObjects & objects) override; @@ -82,6 +78,10 @@ public: ReadSettings patchSettings(const ReadSettings & read_settings) const override; private: + void removeObject(const StoredObject & object) const; + + void removeObjects(const StoredObjects & objects) const; + String key_prefix; LoggerPtr log; std::string description; From 6f00b490679f9e26159105f095660f6b23ea34c2 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Mon, 11 Nov 2024 16:41:23 +0000 Subject: [PATCH 135/244] Fixing more tests. --- .../MergeTreeSplitPrewhereIntoReadSteps.cpp | 2 +- ...filter_push_down_equivalent_sets.reference | 68 +++++++++++-------- .../0_stateless/03199_merge_filters_bug.sql | 34 +++++++++- 3 files changed, 74 insertions(+), 30 deletions(-) diff --git a/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp b/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp index 73fe2600946..2af9974c870 100644 --- a/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp +++ b/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp @@ -358,7 +358,7 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction /// Don't remove if it's in the list of original outputs .remove_filter_column = step.original_node && !all_outputs.contains(step.original_node) && node_to_step[step.original_node] <= step_index, - .need_filter = false, + .need_filter = true, .perform_alter_conversions = true, }; diff --git a/tests/queries/0_stateless/03036_join_filter_push_down_equivalent_sets.reference b/tests/queries/0_stateless/03036_join_filter_push_down_equivalent_sets.reference index 80f4e309505..d0a3e7b02ae 100644 --- a/tests/queries/0_stateless/03036_join_filter_push_down_equivalent_sets.reference +++ b/tests/queries/0_stateless/03036_join_filter_push_down_equivalent_sets.reference @@ -163,17 +163,21 @@ Positions: 4 2 0 1 Filter (( + (JOIN actions + Change column names to column identifiers))) Header: __table1.id UInt64 __table1.value String - Filter column: and(equals(__table1.id, 5_UInt8), equals(__table1.id, 6_UInt8)) (removed) + AND column: equals(__table1.id, 5_UInt8) Actions: INPUT : 0 -> id UInt64 : 0 - INPUT : 1 -> value String : 1 + COLUMN Const(UInt8) -> 5_UInt8 UInt8 : 1 + FUNCTION equals(id : 0, 5_UInt8 :: 1) -> equals(__table1.id, 5_UInt8) UInt8 : 2 + Positions: 2 0 2 + Filter column: and(equals(__table1.id, 5_UInt8), equals(__table1.id, 6_UInt8)) (removed) + Actions: INPUT : 2 -> value String : 0 + INPUT : 1 -> id UInt64 : 1 COLUMN Const(UInt8) -> 6_UInt8 UInt8 : 2 - COLUMN Const(UInt8) -> 5_UInt8 UInt8 : 3 - ALIAS id : 0 -> __table1.id UInt64 : 4 - ALIAS value :: 1 -> __table1.value String : 5 - FUNCTION equals(id : 0, 6_UInt8 :: 2) -> equals(__table1.id, 6_UInt8) UInt8 : 1 - FUNCTION equals(id :: 0, 5_UInt8 :: 3) -> equals(__table1.id, 5_UInt8) UInt8 : 2 - FUNCTION and(equals(__table1.id, 5_UInt8) :: 2, equals(__table1.id, 6_UInt8) :: 1) -> and(equals(__table1.id, 5_UInt8), equals(__table1.id, 6_UInt8)) UInt8 : 3 - Positions: 3 4 5 + INPUT : 0 -> equals(__table1.id, 5_UInt8) UInt8 : 3 + ALIAS value :: 0 -> __table1.value String : 4 + ALIAS id : 1 -> __table1.id UInt64 : 0 + FUNCTION equals(id :: 1, 6_UInt8 :: 2) -> equals(__table1.id, 6_UInt8) UInt8 : 5 + FUNCTION and(equals(__table1.id, 5_UInt8) :: 3, equals(__table1.id, 6_UInt8) :: 5) -> and(equals(__table1.id, 5_UInt8), equals(__table1.id, 6_UInt8)) UInt8 : 2 + Positions: 2 0 4 ReadFromMergeTree (default.test_table_1) Header: id UInt64 value String @@ -183,17 +187,21 @@ Positions: 4 2 0 1 Filter (( + (JOIN actions + Change column names to column identifiers))) Header: __table2.id UInt64 __table2.value String - Filter column: and(equals(__table2.id, 6_UInt8), equals(__table2.id, 5_UInt8)) (removed) + AND column: equals(__table2.id, 6_UInt8) Actions: INPUT : 0 -> id UInt64 : 0 - INPUT : 1 -> value String : 1 + COLUMN Const(UInt8) -> 6_UInt8 UInt8 : 1 + FUNCTION equals(id : 0, 6_UInt8 :: 1) -> equals(__table2.id, 6_UInt8) UInt8 : 2 + Positions: 2 0 2 + Filter column: and(equals(__table2.id, 6_UInt8), equals(__table2.id, 5_UInt8)) (removed) + Actions: INPUT : 2 -> value String : 0 + INPUT : 1 -> id UInt64 : 1 COLUMN Const(UInt8) -> 5_UInt8 UInt8 : 2 - COLUMN Const(UInt8) -> 6_UInt8 UInt8 : 3 - ALIAS id : 0 -> __table2.id UInt64 : 4 - ALIAS value :: 1 -> __table2.value String : 5 - FUNCTION equals(id : 0, 5_UInt8 :: 2) -> equals(__table2.id, 5_UInt8) UInt8 : 1 - FUNCTION equals(id :: 0, 6_UInt8 :: 3) -> equals(__table2.id, 6_UInt8) UInt8 : 2 - FUNCTION and(equals(__table2.id, 6_UInt8) :: 2, equals(__table2.id, 5_UInt8) :: 1) -> and(equals(__table2.id, 6_UInt8), equals(__table2.id, 5_UInt8)) UInt8 : 3 - Positions: 3 4 5 + INPUT : 0 -> equals(__table2.id, 6_UInt8) UInt8 : 3 + ALIAS value :: 0 -> __table2.value String : 4 + ALIAS id : 1 -> __table2.id UInt64 : 0 + FUNCTION equals(id :: 1, 5_UInt8 :: 2) -> equals(__table2.id, 5_UInt8) UInt8 : 5 + FUNCTION and(equals(__table2.id, 6_UInt8) :: 3, equals(__table2.id, 5_UInt8) :: 5) -> and(equals(__table2.id, 6_UInt8), equals(__table2.id, 5_UInt8)) UInt8 : 2 + Positions: 2 0 4 ReadFromMergeTree (default.test_table_2) Header: id UInt64 value String @@ -656,17 +664,21 @@ Positions: 4 2 0 1 __table1.value String __table2.value String __table2.id UInt64 - Filter column: and(equals(__table1.id, 5_UInt8), equals(__table2.id, 6_UInt8)) (removed) + AND column: equals(__table1.id, 5_UInt8) Actions: INPUT : 0 -> __table1.id UInt64 : 0 - INPUT :: 1 -> __table1.value String : 1 - INPUT :: 2 -> __table2.value String : 2 - INPUT : 3 -> __table2.id UInt64 : 3 - COLUMN Const(UInt8) -> 5_UInt8 UInt8 : 4 - COLUMN Const(UInt8) -> 6_UInt8 UInt8 : 5 - FUNCTION equals(__table1.id : 0, 5_UInt8 :: 4) -> equals(__table1.id, 5_UInt8) UInt8 : 6 - FUNCTION equals(__table2.id : 3, 6_UInt8 :: 5) -> equals(__table2.id, 6_UInt8) UInt8 : 4 - FUNCTION and(equals(__table1.id, 5_UInt8) :: 6, equals(__table2.id, 6_UInt8) :: 4) -> and(equals(__table1.id, 5_UInt8), equals(__table2.id, 6_UInt8)) UInt8 : 5 - Positions: 5 0 1 2 3 + COLUMN Const(UInt8) -> 5_UInt8 UInt8 : 1 + FUNCTION equals(__table1.id : 0, 5_UInt8 :: 1) -> equals(__table1.id, 5_UInt8) UInt8 : 2 + Positions: 2 0 2 + Filter column: and(equals(__table1.id, 5_UInt8), equals(__table2.id, 6_UInt8)) (removed) + Actions: INPUT :: 1 -> __table1.id UInt64 : 0 + INPUT :: 2 -> __table1.value String : 1 + INPUT :: 3 -> __table2.value String : 2 + INPUT : 4 -> __table2.id UInt64 : 3 + COLUMN Const(UInt8) -> 6_UInt8 UInt8 : 4 + INPUT : 0 -> equals(__table1.id, 5_UInt8) UInt8 : 5 + FUNCTION equals(__table2.id : 3, 6_UInt8 :: 4) -> equals(__table2.id, 6_UInt8) UInt8 : 6 + FUNCTION and(equals(__table1.id, 5_UInt8) :: 5, equals(__table2.id, 6_UInt8) :: 6) -> and(equals(__table1.id, 5_UInt8), equals(__table2.id, 6_UInt8)) UInt8 : 4 + Positions: 4 0 1 2 3 Join (JOIN FillRightFirst) Header: __table1.id UInt64 __table1.value String diff --git a/tests/queries/0_stateless/03199_merge_filters_bug.sql b/tests/queries/0_stateless/03199_merge_filters_bug.sql index bb2a4255a3d..2023e0f1d73 100644 --- a/tests/queries/0_stateless/03199_merge_filters_bug.sql +++ b/tests/queries/0_stateless/03199_merge_filters_bug.sql @@ -51,6 +51,22 @@ tmp1 AS LEFT JOIN tmp1 USING (fs1) WHERE (fs1 IN ('test')) SETTINGS enable_multiple_prewhere_read_steps = 0, query_plan_merge_filters=0; +WITH +tmp1 AS +( + SELECT + CAST(s1, 'FixedString(10)') AS fs1, + s2 AS sector, + s3 + FROM t1 + WHERE (s3 != 'test') +) + SELECT + fs1 + FROM t2 + LEFT JOIN tmp1 USING (fs1) + WHERE (fs1 IN ('test')) SETTINGS enable_multiple_prewhere_read_steps = 1, query_plan_merge_filters=1; + optimize table t1 final; WITH @@ -67,4 +83,20 @@ tmp1 AS fs1 FROM t2 LEFT JOIN tmp1 USING (fs1) - WHERE (fs1 IN ('test')); + WHERE (fs1 IN ('test')) SETTINGS enable_multiple_prewhere_read_steps = 0, query_plan_merge_filters=0; + +WITH +tmp1 AS +( + SELECT + CAST(s1, 'FixedString(10)') AS fs1, + s2 AS sector, + s3 + FROM t1 + WHERE (s3 != 'test') +) + SELECT + fs1 + FROM t2 + LEFT JOIN tmp1 USING (fs1) + WHERE (fs1 IN ('test')) SETTINGS enable_multiple_prewhere_read_steps = 1, query_plan_merge_filters=1; From a5318f60846e206b69538a125af76a222d48dbce Mon Sep 17 00:00:00 2001 From: Peter Nguyen Date: Mon, 11 Nov 2024 08:55:53 -0800 Subject: [PATCH 136/244] Change enable_http_compression setting's default value to 1 --- src/Core/Settings.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Core/Settings.cpp b/src/Core/Settings.cpp index d321f5dbdf2..ec07d1cdc4b 100644 --- a/src/Core/Settings.cpp +++ b/src/Core/Settings.cpp @@ -1788,7 +1788,7 @@ Possible values: - 0 — Disabled. - 1 — Enabled. -)", 0) \ +)", 1) \ DECLARE(Int64, http_zlib_compression_level, 3, R"( Sets the level of data compression in the response to an HTTP request if [enable_http_compression = 1](#enable_http_compression). From f60d35161f9d4b34fac0f51ad566906301cb3762 Mon Sep 17 00:00:00 2001 From: Peter Nguyen Date: Mon, 11 Nov 2024 08:56:50 -0800 Subject: [PATCH 137/244] Update docs for ru/ and zh/ --- docs/ru/operations/settings/settings.md | 2 +- docs/zh/operations/settings/settings.md | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/ru/operations/settings/settings.md b/docs/ru/operations/settings/settings.md index 84bbf6c83d3..bbe1f071381 100644 --- a/docs/ru/operations/settings/settings.md +++ b/docs/ru/operations/settings/settings.md @@ -136,7 +136,7 @@ ClickHouse применяет настройку в тех случаях, ко - 0 — выключена. - 1 — включена. -Значение по умолчанию: 0. +Значение по умолчанию: 1. ## http_zlib_compression_level {#settings-http_zlib_compression_level} diff --git a/docs/zh/operations/settings/settings.md b/docs/zh/operations/settings/settings.md index 5e59196f56c..baa4fcb0754 100644 --- a/docs/zh/operations/settings/settings.md +++ b/docs/zh/operations/settings/settings.md @@ -97,7 +97,7 @@ ClickHouse从表的过时副本中选择最相关的副本。 - 0 — Disabled. - 1 — Enabled. -默认值：0。 +默认值：1。 ## http_zlib_compression_level {#settings-http_zlib_compression_level} From a0cc03b175b035e9c52e782811d990a619acc272 Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Mon, 11 Nov 2024 17:50:11 +0000 Subject: [PATCH 138/244] Cleanup. --- src/Processors/QueryPlan/FilterStep.cpp | 17 ++++++++ .../QueryPlan/ReadFromMergeTree.cpp | 2 + .../MergeTree/MergeTreeBlockReadUtils.cpp | 2 +- src/Storages/MergeTree/MergeTreeIOSettings.h | 2 + .../MergeTree/MergeTreeSelectProcessor.cpp | 8 ++-- .../MergeTree/MergeTreeSelectProcessor.h | 3 +- .../MergeTreeSplitPrewhereIntoReadSteps.cpp | 41 +++++++++++-------- 7 files changed, 51 insertions(+), 24 deletions(-) diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp index a6b157cdd1d..5bf55f67208 100644 --- a/src/Processors/QueryPlan/FilterStep.cpp +++ b/src/Processors/QueryPlan/FilterStep.cpp @@ -64,6 +64,7 @@ static ActionsAndName splitSingleAndFilter(ActionsDAG & dag, const ActionsDAG::N return ActionsAndName{std::move(split_result.first), std::move(name)}; } +/// Try to split the left most AND atom to a separate DAG. static std::optional trySplitSingleAndFilter(ActionsDAG & dag, const std::string & filter_name) { const auto * filter = &dag.findInOutputs(filter_name); @@ -83,6 +84,7 @@ static std::optional trySplitSingleAndFilter(ActionsDAG & dag, c if (node->type == ActionsDAG::ActionType::FUNCTION && node->function_base->getName() == "and") { + /// The order is important. We should take the left-most atom, so put conditions on stack in reverse order. for (const auto * child : node->children | std::ranges::views::reverse) nodes.push(child); @@ -141,6 +143,8 @@ void FilterStep::transformPipeline(QueryPipelineBuilder & pipeline, const BuildQ { std::vector and_atoms; + /// Spliting AND filter condition to steps under the setting, which is enabled with merge_filters optimization. + /// This is needed to support short-circuit properly. if (settings.enable_multiple_filters_transforms_for_and_chain && !actions_dag.hasStatefulFunctions()) and_atoms = splitAndChainIntoMultipleFilters(actions_dag, filter_column_name); @@ -206,6 +210,19 @@ void FilterStep::describeActions(FormatSettings & settings) const void FilterStep::describeActions(JSONBuilder::JSONMap & map) const { + auto cloned_dag = actions_dag.clone(); + + std::vector and_atoms; + if (!actions_dag.hasStatefulFunctions()) + and_atoms = splitAndChainIntoMultipleFilters(cloned_dag, filter_column_name); + + for (auto & and_atom : and_atoms) + { + auto expression = std::make_shared(std::move(and_atom.dag)); + map.add("AND column", and_atom.name); + map.add("Expression", expression->toTree()); + } + map.add("Filter Column", filter_column_name); map.add("Removes Filter", remove_filter_column); diff --git a/src/Processors/QueryPlan/ReadFromMergeTree.cpp b/src/Processors/QueryPlan/ReadFromMergeTree.cpp index 3186df6a6b3..d144187821a 100644 --- a/src/Processors/QueryPlan/ReadFromMergeTree.cpp +++ b/src/Processors/QueryPlan/ReadFromMergeTree.cpp @@ -175,6 +175,7 @@ namespace Setting extern const SettingsBool use_skip_indexes; extern const SettingsBool use_skip_indexes_if_final; extern const SettingsBool use_uncompressed_cache; + extern const SettingsBool query_plan_merge_filters; extern const SettingsUInt64 merge_tree_min_read_task_size; } @@ -206,6 +207,7 @@ static MergeTreeReaderSettings getMergeTreeReaderSettings( .use_asynchronous_read_from_pool = settings[Setting::allow_asynchronous_read_from_io_pool_for_merge_tree] && (settings[Setting::max_streams_to_max_threads_ratio] > 1 || settings[Setting::max_streams_for_merge_tree_reading] > 1), .enable_multiple_prewhere_read_steps = settings[Setting::enable_multiple_prewhere_read_steps], + .force_shirt_circuit_execution = settings[Setting::query_plan_merge_filters] }; } diff --git a/src/Storages/MergeTree/MergeTreeBlockReadUtils.cpp b/src/Storages/MergeTree/MergeTreeBlockReadUtils.cpp index 7ba358d2d35..03a0aed80bf 100644 --- a/src/Storages/MergeTree/MergeTreeBlockReadUtils.cpp +++ b/src/Storages/MergeTree/MergeTreeBlockReadUtils.cpp @@ -330,7 +330,7 @@ MergeTreeReadTaskColumns getReadTaskColumns( auto prewhere_actions = MergeTreeSelectProcessor::getPrewhereActions( prewhere_info, actions_settings, - reader_settings.enable_multiple_prewhere_read_steps); + reader_settings.enable_multiple_prewhere_read_steps, reader_settings.force_shirt_circuit_execution); for (const auto & step : prewhere_actions.steps) add_step(*step); diff --git a/src/Storages/MergeTree/MergeTreeIOSettings.h b/src/Storages/MergeTree/MergeTreeIOSettings.h index 4d1d2533729..ecd4ad34961 100644 --- a/src/Storages/MergeTree/MergeTreeIOSettings.h +++ b/src/Storages/MergeTree/MergeTreeIOSettings.h @@ -45,6 +45,8 @@ struct MergeTreeReaderSettings bool use_asynchronous_read_from_pool = false; /// If PREWHERE has multiple conditions combined with AND, execute them in separate read/filtering steps. bool enable_multiple_prewhere_read_steps = false; + /// In case of multiple prewhere steps, execute filtering earlier to support short-circuit properly. + bool force_shirt_circuit_execution = false; /// If true, try to lower size of read buffer according to granule size and compressed block size. bool adjust_read_buffer_size = true; /// If true, it's allowed to read the whole part without reading marks. diff --git a/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp b/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp index 5efd33ce09a..8beff55e698 100644 --- a/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp +++ b/src/Storages/MergeTree/MergeTreeSelectProcessor.cpp @@ -91,7 +91,7 @@ MergeTreeSelectProcessor::MergeTreeSelectProcessor( , algorithm(std::move(algorithm_)) , prewhere_info(prewhere_info_) , actions_settings(actions_settings_) - , prewhere_actions(getPrewhereActions(prewhere_info, actions_settings, reader_settings_.enable_multiple_prewhere_read_steps)) + , prewhere_actions(getPrewhereActions(prewhere_info, actions_settings, reader_settings_.enable_multiple_prewhere_read_steps, reader_settings_.force_shirt_circuit_execution)) , reader_settings(reader_settings_) , result_header(transformHeader(pool->getHeader(), prewhere_info)) { @@ -124,9 +124,9 @@ String MergeTreeSelectProcessor::getName() const return fmt::format("MergeTreeSelect(pool: {}, algorithm: {})", pool->getName(), algorithm->getName()); } -bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionActionsSettings & actions_settings, PrewhereExprInfo & prewhere); +bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionActionsSettings & actions_settings, PrewhereExprInfo & prewhere, bool force_shirt_circuit_execution); -PrewhereExprInfo MergeTreeSelectProcessor::getPrewhereActions(PrewhereInfoPtr prewhere_info, const ExpressionActionsSettings & actions_settings, bool enable_multiple_prewhere_read_steps) +PrewhereExprInfo MergeTreeSelectProcessor::getPrewhereActions(PrewhereInfoPtr prewhere_info, const ExpressionActionsSettings & actions_settings, bool enable_multiple_prewhere_read_steps, bool force_shirt_circuit_execution) { PrewhereExprInfo prewhere_actions; if (prewhere_info) @@ -147,7 +147,7 @@ PrewhereExprInfo MergeTreeSelectProcessor::getPrewhereActions(PrewhereInfoPtr pr } if (!enable_multiple_prewhere_read_steps || - !tryBuildPrewhereSteps(prewhere_info, actions_settings, prewhere_actions)) + !tryBuildPrewhereSteps(prewhere_info, actions_settings, prewhere_actions, force_shirt_circuit_execution)) { PrewhereExprStep prewhere_step { diff --git a/src/Storages/MergeTree/MergeTreeSelectProcessor.h b/src/Storages/MergeTree/MergeTreeSelectProcessor.h index 33069a78e33..afd88116e15 100644 --- a/src/Storages/MergeTree/MergeTreeSelectProcessor.h +++ b/src/Storages/MergeTree/MergeTreeSelectProcessor.h @@ -73,7 +73,8 @@ public: static PrewhereExprInfo getPrewhereActions( PrewhereInfoPtr prewhere_info, const ExpressionActionsSettings & actions_settings, - bool enable_multiple_prewhere_read_steps); + bool enable_multiple_prewhere_read_steps, + bool force_shirt_circuit_execution); void addPartLevelToChunk(bool add_part_level_) { add_part_level = add_part_level_; } diff --git a/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp b/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp index 2af9974c870..c35e356bf18 100644 --- a/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp +++ b/src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp @@ -50,6 +50,17 @@ void fillRequiredColumns(const ActionsDAG::Node * node, std::unordered_map DAGNodeRef +/// ResultNode -> DAGNodeRef using OriginalToNewNodeMap = std::unordered_map; using NodeNameToLastUsedStepMap = std::unordered_map; @@ -70,7 +81,6 @@ const ActionsDAG::Node & addClonedDAGToDAG( OriginalToNewNodeMap & node_remap, NodeNameToLastUsedStepMap & node_to_step_map) { - //const String & node_name = original_dag_node->result_name; /// Look for the node in the map of already known nodes if (node_remap.contains(original_dag_node)) { @@ -82,9 +92,11 @@ const ActionsDAG::Node & addClonedDAGToDAG( /// If the node is known from the previous steps, add it as an input, except for constants if (original_dag_node->type != ActionsDAG::ActionType::COLUMN) { - node_ref.dag->addOrReplaceInOutputs(*node_ref.node); + // addToOutputsIfNotAlreadyAdded(*node_ref.dag, node_ref.node); + node_ref.dag->getOutputs().push_back(node_ref.node); + const auto & new_node = new_dag->addInput(node_ref.node->result_name, node_ref.node->result_type); - node_remap[original_dag_node] = {new_dag.get(), &new_node}; /// TODO: here we update the node reference. Is it always correct? + node_remap[original_dag_node] = {new_dag.get(), &new_node}; /// Remember the index of the last step which reuses this node. /// We cannot remove this node from the outputs before that step. @@ -207,7 +219,11 @@ const ActionsDAG::Node & addAndTrue( /// 6. Find all outputs of the original DAG /// 7. Find all outputs that were computed in the already built DAGs, mark these nodes as outputs in the steps where they were computed /// 8. Add computation of the remaining outputs to the last step with the procedure similar to 4 -bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionActionsSettings & actions_settings, PrewhereExprInfo & prewhere) +bool tryBuildPrewhereSteps( + PrewhereInfoPtr prewhere_info, + const ExpressionActionsSettings & actions_settings, + PrewhereExprInfo & prewhere, + bool force_shirt_circuit_execution) { if (!prewhere_info) return true; @@ -275,26 +291,16 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction /// Add AND function to combine the conditions FunctionOverloadResolverPtr func_builder_and = std::make_unique(std::make_shared()); const auto & and_function_node = addFunction(step_dag, func_builder_and, new_condition_nodes); //, node_remap); - //step_dag->addOrReplaceInOutputs(and_function_node); result_node = &and_function_node; } else { result_node = new_condition_nodes.front(); /// Check if explicit cast is needed for the condition to serve as a filter. - //const auto result_type_name = result_node->result_type->getName(); - if (isUInt8(removeNullable(removeLowCardinality(result_node->result_type)))) - { - /// No need to cast - //step_dag->addOrReplaceInOutputs(result_node); - //result_name = result_node.result_name; - } - else + if (!isUInt8(removeNullable(removeLowCardinality(result_node->result_type)))) { /// Build "condition AND True" expression to "cast" the condition to UInt8 or Nullable(UInt8) depending on its type. result_node = &addAndTrue(step_dag, *result_node); //, node_remap); - //step_dag->addOrReplaceInOutputs(cast_node); - //result_name = &cast_node.result_name; } } @@ -334,7 +340,6 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction const auto & cast_node = addCast(last_step_dag, and_node, output->result_type); //, node_remap); /// Add alias for the result with the name of the PREWHERE column const auto & prewhere_result_node = last_step_dag->addAlias(cast_node, output->result_name); - //last_step_dag->addOrReplaceInOutputs(prewhere_result_node); last_step_dag->getOutputs().push_back(&prewhere_result_node); steps.back().result_node = &prewhere_result_node; } @@ -358,7 +363,7 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction /// Don't remove if it's in the list of original outputs .remove_filter_column = step.original_node && !all_outputs.contains(step.original_node) && node_to_step[step.original_node] <= step_index, - .need_filter = true, + .need_filter = force_shirt_circuit_execution, .perform_alter_conversions = true, }; From 92114f3c749bb78811ece644123b3d81e011e56f Mon Sep 17 00:00:00 2001 From: Nikolai Kochetov Date: Mon, 11 Nov 2024 18:01:24 +0000 Subject: [PATCH 139/244] Fixing typos. --- src/Processors/QueryPlan/FilterStep.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/Processors/QueryPlan/FilterStep.cpp b/src/Processors/QueryPlan/FilterStep.cpp index 5bf55f67208..af9e3f0c515 100644 --- a/src/Processors/QueryPlan/FilterStep.cpp +++ b/src/Processors/QueryPlan/FilterStep.cpp @@ -143,7 +143,7 @@ void FilterStep::transformPipeline(QueryPipelineBuilder & pipeline, const BuildQ { std::vector and_atoms; - /// Spliting AND filter condition to steps under the setting, which is enabled with merge_filters optimization. + /// Splitting AND filter condition to steps under the setting, which is enabled with merge_filters optimization. /// This is needed to support short-circuit properly. if (settings.enable_multiple_filters_transforms_for_and_chain && !actions_dag.hasStatefulFunctions()) and_atoms = splitAndChainIntoMultipleFilters(actions_dag, filter_column_name); From 5f9506cc7d526c0c7aaf5a8318dd125dae612c69 Mon Sep 17 00:00:00 2001 From: kssenii Date: Mon, 11 Nov 2024 21:24:18 +0100 Subject: [PATCH 140/244] Cleanup --- src/Interpreters/Cache/FileCache.cpp | 4 ++-- src/Interpreters/Cache/FileCache.h | 3 ++- src/Interpreters/Cache/Metadata.cpp | 2 -- src/Interpreters/Cache/Metadata.h | 5 ----- 4 files changed, 4 insertions(+), 10 deletions(-) diff --git a/src/Interpreters/Cache/FileCache.cpp b/src/Interpreters/Cache/FileCache.cpp index 47b5779dd1a..8887165a75d 100644 --- a/src/Interpreters/Cache/FileCache.cpp +++ b/src/Interpreters/Cache/FileCache.cpp @@ -87,6 +87,7 @@ FileCache::FileCache(const std::string & cache_name, const FileCacheSettings & s : max_file_segment_size(settings.max_file_segment_size) , bypass_cache_threshold(settings.enable_bypass_cache_with_threshold ? settings.bypass_cache_threshold : 0) , boundary_alignment(settings.boundary_alignment) + , background_download_max_file_segment_size(settings.background_download_max_file_segment_size) , load_metadata_threads(settings.load_metadata_threads) , load_metadata_asynchronously(settings.load_metadata_asynchronously) , write_cache_per_user_directory(settings.write_cache_per_user_id_directory) @@ -97,7 +98,6 @@ FileCache::FileCache(const std::string & cache_name, const FileCacheSettings & s , metadata(settings.base_path, settings.background_download_queue_size_limit, settings.background_download_threads, - settings.background_download_max_file_segment_size, write_cache_per_user_directory) { if (settings.cache_policy == "LRU") @@ -1597,7 +1597,7 @@ void FileCache::applySettingsIfPossible(const FileCacheSettings & new_settings, if (new_settings.background_download_max_file_segment_size != actual_settings.background_download_max_file_segment_size) { - metadata.setBackgroundDownloadMaxFileSegmentSize(new_settings.background_download_max_file_segment_size); + background_download_max_file_segment_size = new_settings.background_download_max_file_segment_size; LOG_INFO(log, "Changed background_download_max_file_segment_size from {} to {}", actual_settings.background_download_max_file_segment_size, diff --git a/src/Interpreters/Cache/FileCache.h b/src/Interpreters/Cache/FileCache.h index 3a9241fd351..bbe8502fec9 100644 --- a/src/Interpreters/Cache/FileCache.h +++ b/src/Interpreters/Cache/FileCache.h @@ -161,7 +161,7 @@ public: size_t getMaxFileSegmentSize() const { return max_file_segment_size; } - size_t getBackgroundDownloadMaxFileSegmentSize() const { return metadata.getBackgroundDownloadMaxFileSegmentSize(); } + size_t getBackgroundDownloadMaxFileSegmentSize() const { return background_download_max_file_segment_size.load(); } size_t getBoundaryAlignment() const { return boundary_alignment; } @@ -203,6 +203,7 @@ private: std::atomic max_file_segment_size; const size_t bypass_cache_threshold; const size_t boundary_alignment; + std::atomic background_download_max_file_segment_size; size_t load_metadata_threads; const bool load_metadata_asynchronously; std::atomic stop_loading_metadata = false; diff --git a/src/Interpreters/Cache/Metadata.cpp b/src/Interpreters/Cache/Metadata.cpp index 45077f09020..dd4a5502638 100644 --- a/src/Interpreters/Cache/Metadata.cpp +++ b/src/Interpreters/Cache/Metadata.cpp @@ -168,7 +168,6 @@ CacheMetadata::CacheMetadata( const std::string & path_, size_t background_download_queue_size_limit_, size_t background_download_threads_, - size_t background_download_max_file_segment_size_, bool write_cache_per_user_directory_) : path(path_) , cleanup_queue(std::make_shared()) @@ -176,7 +175,6 @@ CacheMetadata::CacheMetadata( , write_cache_per_user_directory(write_cache_per_user_directory_) , log(getLogger("CacheMetadata")) , download_threads_num(background_download_threads_) - , download_max_file_segment_size(background_download_max_file_segment_size_) { } diff --git a/src/Interpreters/Cache/Metadata.h b/src/Interpreters/Cache/Metadata.h index 526b82c9a68..24683b2de71 100644 --- a/src/Interpreters/Cache/Metadata.h +++ b/src/Interpreters/Cache/Metadata.h @@ -165,7 +165,6 @@ public: const std::string & path_, size_t background_download_queue_size_limit_, size_t background_download_threads_, - size_t background_download_max_file_segment_size_, bool write_cache_per_user_directory_); void startup(); @@ -212,9 +211,6 @@ public: bool setBackgroundDownloadThreads(size_t threads_num); size_t getBackgroundDownloadThreads() const { return download_threads.size(); } - void setBackgroundDownloadMaxFileSegmentSize(size_t max_file_segment_size) { download_max_file_segment_size = max_file_segment_size; } - size_t getBackgroundDownloadMaxFileSegmentSize() const { return download_max_file_segment_size; } - bool setBackgroundDownloadQueueSizeLimit(size_t size); bool isBackgroundDownloadEnabled(); @@ -246,7 +242,6 @@ private: }; std::atomic download_threads_num; - std::atomic download_max_file_segment_size; std::vector> download_threads; std::unique_ptr cleanup_thread; From 06debdc479bab58f2d1d7fd4b3764e65a8c9fa01 Mon Sep 17 00:00:00 2001 From: Max Kainov Date: Fri, 8 Nov 2024 17:48:17 +0100 Subject: [PATCH 141/244] result with versioning --- ci/docker/stateless-test/Dockerfile | 6 +- ci/jobs/build_clickhouse.py | 1 - ci/jobs/functional_stateless_tests.py | 11 +- ci/jobs/scripts/clickhouse_proc.py | 11 + .../setup_hdfs_minicluster.sh | 19 ++ ci/praktika/__main__.py | 7 + ci/praktika/_environment.py | 14 +- ci/praktika/_settings.py | 128 ---------- ci/praktika/digest.py | 38 +-- ci/praktika/hook_cache.py | 11 +- ci/praktika/hook_html.py | 71 ++---- ci/praktika/json.html | 11 +- ci/praktika/mangle.py | 36 +-- ci/praktika/native_jobs.py | 14 +- ci/praktika/result.py | 240 +++++++++++++++++- ci/praktika/runner.py | 18 +- ci/praktika/runtime.py | 6 + ci/praktika/s3.py | 172 ++----------- ci/praktika/settings.py | 156 +++++++++++- ci/praktika/utils.py | 2 - ci/praktika/validator.py | 8 +- ci/workflows/pull_request.py | 1 + 22 files changed, 551 insertions(+), 430 deletions(-) create mode 100755 ci/jobs/scripts/functional_tests/setup_hdfs_minicluster.sh delete mode 100644 ci/praktika/_settings.py diff --git a/ci/docker/stateless-test/Dockerfile b/ci/docker/stateless-test/Dockerfile index 4abd8204f1d..760fceeebbf 100644 --- a/ci/docker/stateless-test/Dockerfile +++ b/ci/docker/stateless-test/Dockerfile @@ -100,8 +100,12 @@ ENV PATH="/wd/tests:/tmp/praktika/input:$PATH" RUN curl -L --no-verbose -O 'https://archive.apache.org/dist/hadoop/common/hadoop-3.3.1/hadoop-3.3.1.tar.gz' \ && tar -xvf hadoop-3.3.1.tar.gz \ - && rm -rf hadoop-3.3.1.tar.gz + && rm -rf hadoop-3.3.1.tar.gz \ + && chmod 777 /hadoop-3.3.1 RUN npm install -g azurite@3.30.0 \ && npm install -g tslib && npm install -g node + +RUN addgroup --gid 1001 clickhouse && adduser --uid 1001 --gid 1001 --disabled-password clickhouse +USER clickhouse \ No newline at end of file diff --git a/ci/jobs/build_clickhouse.py b/ci/jobs/build_clickhouse.py index 1e6d2c648a7..3bdc23d383c 100644 --- a/ci/jobs/build_clickhouse.py +++ b/ci/jobs/build_clickhouse.py @@ -127,7 +127,6 @@ def main(): Shell.check(f"ls -l {build_dir}/programs/") res = results[-1].is_ok() - Result.create_from(results=results, stopwatch=stop_watch).complete_job() diff --git a/ci/jobs/functional_stateless_tests.py b/ci/jobs/functional_stateless_tests.py index 0481086d80a..390a6336b45 100644 --- a/ci/jobs/functional_stateless_tests.py +++ b/ci/jobs/functional_stateless_tests.py @@ -27,11 +27,12 @@ def parse_args(): default="", ) parser.add_argument("--param", help="Optional job start stage", default=None) + parser.add_argument("--test", help="Optional test name pattern", default="") return parser.parse_args() def run_stateless_test( - no_parallel: bool, no_sequiential: bool, batch_num: int, batch_total: int + no_parallel: bool, no_sequiential: bool, batch_num: int, batch_total: int, test="" ): assert not (no_parallel and no_sequiential) test_output_file = f"{Settings.OUTPUT_DIR}/test_result.txt" @@ -43,7 +44,7 @@ def run_stateless_test( --no-drop-if-fail --capture-client-stacktrace --queries /repo/tests/queries --test-runs 1 --hung-check \ {'--no-parallel' if no_parallel else ''} {'--no-sequential' if no_sequiential else ''} \ --print-time --jobs {nproc} --report-coverage --report-logs-stats {aux} \ - --queries ./tests/queries -- '' | ts '%Y-%m-%d %H:%M:%S' \ + --queries ./tests/queries -- '{test}' | ts '%Y-%m-%d %H:%M:%S' \ | tee -a \"{test_output_file}\"" if Path(test_output_file).exists(): Path(test_output_file).unlink() @@ -119,11 +120,14 @@ def main(): stop_watch_ = Utils.Stopwatch() step_name = "Start ClickHouse Server" print(step_name) + hdfs_log = "/tmp/praktika/output/hdfs_mini.log" minio_log = "/tmp/praktika/output/minio.log" + res = res and CH.start_hdfs(log_file_path=hdfs_log) res = res and CH.start_minio(log_file_path=minio_log) - logs_to_attach += [minio_log] + logs_to_attach += [minio_log, hdfs_log] time.sleep(10) Shell.check("ps -ef | grep minio", verbose=True) + Shell.check("ps -ef | grep hdfs", verbose=True) res = res and Shell.check( "aws s3 ls s3://test --endpoint-url http://localhost:11111/", verbose=True ) @@ -153,6 +157,7 @@ def main(): no_sequiential=no_sequential, batch_num=batch_num, batch_total=total_batches, + test=args.test, ) results.append(FTResultsProcessor(wd=Settings.OUTPUT_DIR).run()) results[-1].set_timing(stopwatch=stop_watch_) diff --git a/ci/jobs/scripts/clickhouse_proc.py b/ci/jobs/scripts/clickhouse_proc.py index c43283e75e0..8f9bef57083 100644 --- a/ci/jobs/scripts/clickhouse_proc.py +++ b/ci/jobs/scripts/clickhouse_proc.py @@ -44,6 +44,17 @@ class ClickHouseProc: self.minio_proc = None + def start_hdfs(self, log_file_path): + command = ["./ci/jobs/scripts/functional_tests/setup_hdfs_minicluster.sh"] + with open(log_file_path, "w") as log_file: + process = subprocess.Popen( + command, stdout=log_file, stderr=subprocess.STDOUT + ) + print( + f"Started setup_hdfs_minicluster.sh asynchronously with PID {process.pid}" + ) + return True + def start_minio(self, log_file_path): command = ["tests/docker_scripts/setup_minio.sh", "stateless", "./tests"] with open(log_file_path, "w") as log_file: diff --git a/ci/jobs/scripts/functional_tests/setup_hdfs_minicluster.sh b/ci/jobs/scripts/functional_tests/setup_hdfs_minicluster.sh new file mode 100755 index 00000000000..b810b27fe2b --- /dev/null +++ b/ci/jobs/scripts/functional_tests/setup_hdfs_minicluster.sh @@ -0,0 +1,19 @@ +#!/bin/bash +# shellcheck disable=SC2024 + +set -e -x -a -u + +ls -lha + +cd /hadoop-3.3.1 + +export JAVA_HOME=/usr +mkdir -p target/test/data + +bin/mapred minicluster -format -nomr -nnport 12222 & + +while ! nc -z localhost 12222; do + sleep 1 +done + +lsof -i :12222 diff --git a/ci/praktika/__main__.py b/ci/praktika/__main__.py index fbb9f92909a..3dfdc26d69d 100644 --- a/ci/praktika/__main__.py +++ b/ci/praktika/__main__.py @@ -37,6 +37,12 @@ def create_parser(): type=str, default=None, ) + run_parser.add_argument( + "--test", + help="Custom parameter to pass into a job script, it's up to job script how to use it, for local test", + type=str, + default="", + ) run_parser.add_argument( "--pr", help="PR number. Optional parameter for local run. Set if you want an required artifact to be uploaded from CI run in that PR", @@ -106,6 +112,7 @@ if __name__ == "__main__": local_run=not args.ci, no_docker=args.no_docker, param=args.param, + test=args.test, pr=args.pr, branch=args.branch, sha=args.sha, diff --git a/ci/praktika/_environment.py b/ci/praktika/_environment.py index 1c6b547ddde..734a4be3176 100644 --- a/ci/praktika/_environment.py +++ b/ci/praktika/_environment.py @@ -6,7 +6,7 @@ from types import SimpleNamespace from typing import Any, Dict, List, Type from praktika import Workflow -from praktika._settings import _Settings +from praktika.settings import Settings from praktika.utils import MetaClasses, T @@ -35,7 +35,7 @@ class _Environment(MetaClasses.Serializable): @classmethod def file_name_static(cls, _name=""): - return f"{_Settings.TEMP_DIR}/{cls.name}.json" + return f"{Settings.TEMP_DIR}/{cls.name}.json" @classmethod def from_dict(cls: Type[T], obj: Dict[str, Any]) -> T: @@ -66,12 +66,12 @@ class _Environment(MetaClasses.Serializable): @staticmethod def get_needs_statuses(): - if Path(_Settings.WORKFLOW_STATUS_FILE).is_file(): - with open(_Settings.WORKFLOW_STATUS_FILE, "r", encoding="utf8") as f: + if Path(Settings.WORKFLOW_STATUS_FILE).is_file(): + with open(Settings.WORKFLOW_STATUS_FILE, "r", encoding="utf8") as f: return json.load(f) else: print( - f"ERROR: Status file [{_Settings.WORKFLOW_STATUS_FILE}] does not exist" + f"ERROR: Status file [{Settings.WORKFLOW_STATUS_FILE}] does not exist" ) raise RuntimeError() @@ -171,7 +171,7 @@ class _Environment(MetaClasses.Serializable): # TODO: find a better place for the function. This file should not import praktika.settings # as it's requires reading users config, that's why imports nested inside the function - def get_report_url(self, settings): + def get_report_url(self, settings, latest=False): import urllib path = settings.HTML_S3_PATH @@ -179,7 +179,7 @@ class _Environment(MetaClasses.Serializable): if bucket in path: path = path.replace(bucket, endpoint) break - REPORT_URL = f"https://{path}/{Path(settings.HTML_PAGE_FILE).name}?PR={self.PR_NUMBER}&sha={self.SHA}&name_0={urllib.parse.quote(self.WORKFLOW_NAME, safe='')}&name_1={urllib.parse.quote(self.JOB_NAME, safe='')}" + REPORT_URL = f"https://{path}/{Path(settings.HTML_PAGE_FILE).name}?PR={self.PR_NUMBER}&sha={'latest' if latest else self.SHA}&name_0={urllib.parse.quote(self.WORKFLOW_NAME, safe='')}&name_1={urllib.parse.quote(self.JOB_NAME, safe='')}" return REPORT_URL def is_local_run(self): diff --git a/ci/praktika/_settings.py b/ci/praktika/_settings.py deleted file mode 100644 index 17da1519e37..00000000000 --- a/ci/praktika/_settings.py +++ /dev/null @@ -1,128 +0,0 @@ -import dataclasses -from typing import Dict, Iterable, List, Optional - - -@dataclasses.dataclass -class _Settings: - ###################################### - # Pipeline generation settings # - ###################################### - MAIN_BRANCH = "main" - CI_PATH = "./ci" - WORKFLOW_PATH_PREFIX: str = "./.github/workflows" - WORKFLOWS_DIRECTORY: str = f"{CI_PATH}/workflows" - SETTINGS_DIRECTORY: str = f"{CI_PATH}/settings" - CI_CONFIG_JOB_NAME = "Config Workflow" - DOCKER_BUILD_JOB_NAME = "Docker Builds" - FINISH_WORKFLOW_JOB_NAME = "Finish Workflow" - READY_FOR_MERGE_STATUS_NAME = "Ready for Merge" - CI_CONFIG_RUNS_ON: Optional[List[str]] = None - DOCKER_BUILD_RUNS_ON: Optional[List[str]] = None - VALIDATE_FILE_PATHS: bool = True - - ###################################### - # Runtime Settings # - ###################################### - MAX_RETRIES_S3 = 3 - MAX_RETRIES_GH = 3 - - ###################################### - # S3 (artifact storage) settings # - ###################################### - S3_ARTIFACT_PATH: str = "" - - ###################################### - # CI workspace settings # - ###################################### - TEMP_DIR: str = "/tmp/praktika" - OUTPUT_DIR: str = f"{TEMP_DIR}/output" - INPUT_DIR: str = f"{TEMP_DIR}/input" - PYTHON_INTERPRETER: str = "python3" - PYTHON_PACKET_MANAGER: str = "pip3" - PYTHON_VERSION: str = "3.9" - INSTALL_PYTHON_FOR_NATIVE_JOBS: bool = False - INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS: str = "./ci/requirements.txt" - ENVIRONMENT_VAR_FILE: str = f"{TEMP_DIR}/environment.json" - RUN_LOG: str = f"{TEMP_DIR}/praktika_run.log" - - SECRET_GH_APP_ID: str = "GH_APP_ID" - SECRET_GH_APP_PEM_KEY: str = "GH_APP_PEM_KEY" - - ENV_SETUP_SCRIPT: str = "/tmp/praktika_setup_env.sh" - WORKFLOW_STATUS_FILE: str = f"{TEMP_DIR}/workflow_status.json" - - ###################################### - # CI Cache settings # - ###################################### - CACHE_VERSION: int = 1 - CACHE_DIGEST_LEN: int = 20 - CACHE_S3_PATH: str = "" - CACHE_LOCAL_PATH: str = f"{TEMP_DIR}/ci_cache" - - ###################################### - # Report settings # - ###################################### - HTML_S3_PATH: str = "" - HTML_PAGE_FILE: str = "./praktika/json.html" - TEXT_CONTENT_EXTENSIONS: Iterable[str] = frozenset([".txt", ".log"]) - S3_BUCKET_TO_HTTP_ENDPOINT: Optional[Dict[str, str]] = None - - DOCKERHUB_USERNAME: str = "" - DOCKERHUB_SECRET: str = "" - DOCKER_WD: str = "/wd" - - ###################################### - # CI DB Settings # - ###################################### - SECRET_CI_DB_URL: str = "CI_DB_URL" - SECRET_CI_DB_PASSWORD: str = "CI_DB_PASSWORD" - CI_DB_DB_NAME = "" - CI_DB_TABLE_NAME = "" - CI_DB_INSERT_TIMEOUT_SEC = 5 - - DISABLE_MERGE_COMMIT = True - - -_USER_DEFINED_SETTINGS = [ - "S3_ARTIFACT_PATH", - "CACHE_S3_PATH", - "HTML_S3_PATH", - "S3_BUCKET_TO_HTTP_ENDPOINT", - "TEXT_CONTENT_EXTENSIONS", - "TEMP_DIR", - "OUTPUT_DIR", - "INPUT_DIR", - "CI_CONFIG_RUNS_ON", - "DOCKER_BUILD_RUNS_ON", - "CI_CONFIG_JOB_NAME", - "PYTHON_INTERPRETER", - "PYTHON_VERSION", - "PYTHON_PACKET_MANAGER", - "INSTALL_PYTHON_FOR_NATIVE_JOBS", - "INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS", - "MAX_RETRIES_S3", - "MAX_RETRIES_GH", - "VALIDATE_FILE_PATHS", - "DOCKERHUB_USERNAME", - "DOCKERHUB_SECRET", - "READY_FOR_MERGE_STATUS_NAME", - "SECRET_CI_DB_URL", - "SECRET_CI_DB_PASSWORD", - "CI_DB_DB_NAME", - "CI_DB_TABLE_NAME", - "CI_DB_INSERT_TIMEOUT_SEC", - "SECRET_GH_APP_PEM_KEY", - "SECRET_GH_APP_ID", - "MAIN_BRANCH", - "DISABLE_MERGE_COMMIT", -] - - -class GHRunners: - ubuntu = "ubuntu-latest" - - -if __name__ == "__main__": - for setting in _USER_DEFINED_SETTINGS: - print(_Settings().__getattribute__(setting)) - # print(dataclasses.asdict(_Settings())) diff --git a/ci/praktika/digest.py b/ci/praktika/digest.py index a1f2eecf9b6..6b7e5eec07b 100644 --- a/ci/praktika/digest.py +++ b/ci/praktika/digest.py @@ -23,7 +23,7 @@ class Digest: hash_string = hash_obj.hexdigest() return hash_string - def calc_job_digest(self, job_config: Job.Config): + def calc_job_digest(self, job_config: Job.Config, docker_digests): config = job_config.digest_config if not config: return "f" * Settings.CACHE_DIGEST_LEN @@ -34,28 +34,28 @@ class Digest: print( f"calc digest for job [{job_config.name}]: hash_key [{cache_key}] - from cache" ) - return self.digest_cache[cache_key] - - included_files = Utils.traverse_paths( - job_config.digest_config.include_paths, - job_config.digest_config.exclude_paths, - sorted=True, - ) - print( - f"calc digest for job [{job_config.name}]: hash_key [{cache_key}], include [{len(included_files)}] files" - ) - - # Calculate MD5 hash - res = "" - if not included_files: - res = "f" * Settings.CACHE_DIGEST_LEN - print(f"NOTE: empty digest config [{config}] - return dummy digest") + digest = self.digest_cache[cache_key] else: + included_files = Utils.traverse_paths( + job_config.digest_config.include_paths, + job_config.digest_config.exclude_paths, + sorted=True, + ) + print( + f"calc digest for job [{job_config.name}]: hash_key [{cache_key}], include [{len(included_files)}] files" + ) + hash_md5 = hashlib.md5() for i, file_path in enumerate(included_files): hash_md5 = self._calc_file_digest(file_path, hash_md5) - digest = hash_md5.hexdigest()[: Settings.CACHE_DIGEST_LEN] - self.digest_cache[cache_key] = digest + digest = hash_md5.hexdigest()[: Settings.CACHE_DIGEST_LEN] + self.digest_cache[cache_key] = digest + + if job_config.run_in_docker: + # respect docker digest in the job digest + docker_digest = docker_digests[job_config.run_in_docker.split("+")[0]] + digest = "-".join([docker_digest, digest]) + return digest def calc_docker_digest( diff --git a/ci/praktika/hook_cache.py b/ci/praktika/hook_cache.py index 5cfedec0144..e001e936a71 100644 --- a/ci/praktika/hook_cache.py +++ b/ci/praktika/hook_cache.py @@ -1,6 +1,5 @@ from praktika._environment import _Environment from praktika.cache import Cache -from praktika.mangle import _get_workflows from praktika.runtime import RunConfig from praktika.settings import Settings from praktika.utils import Utils @@ -10,6 +9,7 @@ class CacheRunnerHooks: @classmethod def configure(cls, workflow): workflow_config = RunConfig.from_fs(workflow.name) + docker_digests = workflow_config.digest_dockers cache = Cache() print(f"Workflow Configure, workflow [{workflow.name}]") assert ( @@ -18,11 +18,13 @@ class CacheRunnerHooks: artifact_digest_map = {} job_digest_map = {} for job in workflow.jobs: + digest = cache.digest.calc_job_digest( + job_config=job, docker_digests=docker_digests + ) if not job.digest_config: print( f"NOTE: job [{job.name}] has no Config.digest_config - skip cache check, always run" ) - digest = cache.digest.calc_job_digest(job_config=job) job_digest_map[job.name] = digest if job.provides: # assign the job digest also to the artifacts it provides @@ -48,7 +50,6 @@ class CacheRunnerHooks: ), f"BUG, Workflow with enabled cache must have job digests after configuration, wf [{workflow.name}]" print("Check remote cache") - job_to_cache_record = {} for job_name, job_digest in workflow_config.digest_jobs.items(): record = cache.fetch_success(job_name=job_name, job_digest=job_digest) if record: @@ -58,7 +59,7 @@ class CacheRunnerHooks: ) workflow_config.cache_success.append(job_name) workflow_config.cache_success_base64.append(Utils.to_base64(job_name)) - job_to_cache_record[job_name] = record + workflow_config.cache_jobs[job_name] = record print("Check artifacts to reuse") for job in workflow.jobs: @@ -66,7 +67,7 @@ class CacheRunnerHooks: if job.provides: for artifact_name in job.provides: workflow_config.cache_artifacts[artifact_name] = ( - job_to_cache_record[job.name] + workflow_config.cache_jobs[job.name] ) print(f"Write config to GH's job output") diff --git a/ci/praktika/hook_html.py b/ci/praktika/hook_html.py index ca2692d1b22..e2faefb2fa9 100644 --- a/ci/praktika/hook_html.py +++ b/ci/praktika/hook_html.py @@ -6,7 +6,7 @@ from typing import List from praktika._environment import _Environment from praktika.gh import GH from praktika.parser import WorkflowConfigParser -from praktika.result import Result, ResultInfo +from praktika.result import Result, ResultInfo, _ResultS3 from praktika.runtime import RunConfig from praktika.s3 import S3 from praktika.settings import Settings @@ -119,6 +119,7 @@ class HtmlRunnerHooks: # generate pending Results for all jobs in the workflow if _workflow.enable_cache: skip_jobs = RunConfig.from_fs(_workflow.name).cache_success + job_cache_records = RunConfig.from_fs(_workflow.name).cache_jobs else: skip_jobs = [] @@ -128,21 +129,14 @@ class HtmlRunnerHooks: if job.name not in skip_jobs: result = Result.generate_pending(job.name) else: - result = Result.generate_skipped(job.name) + result = Result.generate_skipped(job.name, job_cache_records[job.name]) results.append(result) summary_result = Result.generate_pending(_workflow.name, results=results) summary_result.links.append(env.CHANGE_URL) summary_result.links.append(env.RUN_URL) summary_result.start_time = Utils.timestamp() - # clean the previous latest results in PR if any - if env.PR_NUMBER: - S3.clean_latest_result() - S3.copy_result_to_s3( - summary_result, - unlock=False, - ) - + assert _ResultS3.copy_result_to_s3_with_version(summary_result, version=0) page_url = env.get_report_url(settings=Settings) print(f"CI Status page url [{page_url}]") @@ -150,7 +144,7 @@ class HtmlRunnerHooks: name=_workflow.name, status=Result.Status.PENDING, description="", - url=env.get_report_url(settings=Settings), + url=env.get_report_url(settings=Settings, latest=True), ) res2 = GH.post_pr_comment( comment_body=f"Workflow [[{_workflow.name}]({page_url})], commit [{_Environment.get().SHA[:8]}]", @@ -167,14 +161,8 @@ class HtmlRunnerHooks: @classmethod def pre_run(cls, _workflow, _job): result = Result.from_fs(_job.name) - S3.copy_result_from_s3( - Result.file_name_static(_workflow.name), - ) - workflow_result = Result.from_fs(_workflow.name) - workflow_result.update_sub_result(result) - S3.copy_result_to_s3( - workflow_result, - unlock=True, + _ResultS3.update_workflow_results( + workflow_name=_workflow.name, new_sub_results=result ) @classmethod @@ -184,14 +172,13 @@ class HtmlRunnerHooks: @classmethod def post_run(cls, _workflow, _job, info_errors): result = Result.from_fs(_job.name) - env = _Environment.get() - S3.copy_result_from_s3( - Result.file_name_static(_workflow.name), - lock=True, - ) - workflow_result = Result.from_fs(_workflow.name) - print(f"Workflow info [{workflow_result.info}], info_errors [{info_errors}]") + _ResultS3.upload_result_files_to_s3(result) + _ResultS3.copy_result_to_s3(result) + env = _Environment.get() + + new_sub_results = [result] + new_result_info = "" env_info = env.REPORT_INFO if env_info: print( @@ -203,14 +190,8 @@ class HtmlRunnerHooks: info_str = f"{_job.name}:\n" info_str += "\n".join(info_errors) print("Update workflow results with new info") - workflow_result.set_info(info_str) + new_result_info = info_str - old_status = workflow_result.status - - S3.upload_result_files_to_s3(result) - workflow_result.update_sub_result(result) - - skipped_job_results = [] if not result.is_ok(): print( "Current job failed - find dependee jobs in the workflow and set their statuses to skipped" @@ -223,7 +204,7 @@ class HtmlRunnerHooks: print( f"NOTE: Set job [{dependee_job.name}] status to [{Result.Status.SKIPPED}] due to current failure" ) - skipped_job_results.append( + new_sub_results.append( Result( name=dependee_job.name, status=Result.Status.SKIPPED, @@ -231,20 +212,18 @@ class HtmlRunnerHooks: + f" [{_job.name}]", ) ) - for skipped_job_result in skipped_job_results: - workflow_result.update_sub_result(skipped_job_result) - S3.copy_result_to_s3( - workflow_result, - unlock=True, + updated_status = _ResultS3.update_workflow_results( + new_info=new_result_info, + new_sub_results=new_sub_results, + workflow_name=_workflow.name, ) - if workflow_result.status != old_status: - print( - f"Update GH commit status [{result.name}]: [{old_status} -> {workflow_result.status}]" - ) + + if updated_status: + print(f"Update GH commit status [{result.name}]: [{updated_status}]") GH.post_commit_status( - name=workflow_result.name, - status=GH.convert_to_gh_status(workflow_result.status), + name=_workflow.name, + status=GH.convert_to_gh_status(updated_status), description="", - url=env.get_report_url(settings=Settings), + url=env.get_report_url(settings=Settings, latest=True), ) diff --git a/ci/praktika/json.html b/ci/praktika/json.html index 4e15a67ba76..544fd6e68d4 100644 --- a/ci/praktika/json.html +++ b/ci/praktika/json.html @@ -342,7 +342,7 @@ const milliseconds = Math.floor((duration % 1) * 1000); const formattedSeconds = String(seconds); - const formattedMilliseconds = String(milliseconds).padStart(3, '0'); + const formattedMilliseconds = String(milliseconds).padStart(2, '0').slice(-2); return `${formattedSeconds}.${formattedMilliseconds}`; } @@ -600,8 +600,7 @@ td.classList.add('time-column'); td.textContent = value ? formatDuration(value) : ''; } else if (column === 'info') { - // For info and other columns, just display the value - td.textContent = value || ''; + td.textContent = value.includes('\n') ? '↵' : (value || ''); td.classList.add('info-column'); } @@ -675,7 +674,8 @@ } if (targetData) { - infoElement.style.display = 'none'; + //infoElement.style.display = 'none'; + infoElement.innerHTML = (targetData.info || '').replace(/\n/g, '
'); addStatusToStatus(targetData.status, targetData.start_time, targetData.duration) @@ -804,7 +804,8 @@ // Check if all required parameters are present to load JSON if (PR && sha && root_name) { - loadResultsJSON(PR, sha, nameParams); + const shaToLoad = (sha === 'latest') ? commitsArray[commitsArray.length - 1] : sha; + loadResultsJSON(PR, shaToLoad, nameParams); } else { document.getElementById('title').textContent = 'Error: Missing required URL parameters: PR, sha, or name_0'; } diff --git a/ci/praktika/mangle.py b/ci/praktika/mangle.py index b16d52fbbbf..f94b11adad5 100644 --- a/ci/praktika/mangle.py +++ b/ci/praktika/mangle.py @@ -1,11 +1,10 @@ import copy import importlib.util from pathlib import Path -from typing import Any, Dict from praktika import Job -from praktika._settings import _USER_DEFINED_SETTINGS, _Settings -from praktika.utils import ContextManager, Utils +from praktika.settings import Settings +from praktika.utils import Utils def _get_workflows(name=None, file=None): @@ -14,13 +13,13 @@ def _get_workflows(name=None, file=None): """ res = [] - directory = Path(_Settings.WORKFLOWS_DIRECTORY) + directory = Path(Settings.WORKFLOWS_DIRECTORY) for py_file in directory.glob("*.py"): if file and file not in str(py_file): continue module_name = py_file.name.removeprefix(".py") spec = importlib.util.spec_from_file_location( - module_name, f"{_Settings.WORKFLOWS_DIRECTORY}/{module_name}" + module_name, f"{Settings.WORKFLOWS_DIRECTORY}/{module_name}" ) assert spec foo = importlib.util.module_from_spec(spec) @@ -106,30 +105,3 @@ def _update_workflow_with_native_jobs(workflow): for job in workflow.jobs: aux_job.requires.append(job.name) workflow.jobs.append(aux_job) - - -def _get_user_settings() -> Dict[str, Any]: - """ - Gets user's settings - """ - res = {} # type: Dict[str, Any] - - directory = Path(_Settings.SETTINGS_DIRECTORY) - for py_file in directory.glob("*.py"): - module_name = py_file.name.removeprefix(".py") - spec = importlib.util.spec_from_file_location( - module_name, f"{_Settings.SETTINGS_DIRECTORY}/{module_name}" - ) - assert spec - foo = importlib.util.module_from_spec(spec) - assert spec.loader - spec.loader.exec_module(foo) - for setting in _USER_DEFINED_SETTINGS: - try: - value = getattr(foo, setting) - res[setting] = value - print(f"Apply user defined setting [{setting} = {value}]") - except Exception as e: - pass - - return res diff --git a/ci/praktika/native_jobs.py b/ci/praktika/native_jobs.py index 58af211988b..52bf6c6e204 100644 --- a/ci/praktika/native_jobs.py +++ b/ci/praktika/native_jobs.py @@ -10,9 +10,8 @@ from praktika.gh import GH from praktika.hook_cache import CacheRunnerHooks from praktika.hook_html import HtmlRunnerHooks from praktika.mangle import _get_workflows -from praktika.result import Result, ResultInfo +from praktika.result import Result, ResultInfo, _ResultS3 from praktika.runtime import RunConfig -from praktika.s3 import S3 from praktika.settings import Settings from praktika.utils import Shell, Utils @@ -225,6 +224,7 @@ def _config_workflow(workflow: Workflow.Config, job_name): cache_success=[], cache_success_base64=[], cache_artifacts={}, + cache_jobs={}, ).dump() # checks: @@ -310,9 +310,8 @@ def _finish_workflow(workflow, job_name): print(env.get_needs_statuses()) print("Check Workflow results") - S3.copy_result_from_s3( + _ResultS3.copy_result_from_s3( Result.file_name_static(workflow.name), - lock=False, ) workflow_result = Result.from_fs(workflow.name) @@ -345,7 +344,7 @@ def _finish_workflow(workflow, job_name): failed_results.append(result.name) if failed_results: - ready_for_merge_description = f"failed: {', '.join(failed_results)}" + ready_for_merge_description = f"Failed: {', '.join(failed_results)}" if not GH.post_commit_status( name=Settings.READY_FOR_MERGE_STATUS_NAME + f" [{workflow.name}]", @@ -357,10 +356,9 @@ def _finish_workflow(workflow, job_name): env.add_info(ResultInfo.GH_STATUS_ERROR) if update_final_report: - S3.copy_result_to_s3( + _ResultS3.copy_result_to_s3( workflow_result, - unlock=False, - ) # no lock - no unlock + ) Result.from_fs(job_name).set_status(Result.Status.SUCCESS) diff --git a/ci/praktika/result.py b/ci/praktika/result.py index 842deacbcbd..8164b1d1295 100644 --- a/ci/praktika/result.py +++ b/ci/praktika/result.py @@ -2,10 +2,12 @@ import dataclasses import datetime import sys from pathlib import Path -from typing import Any, Dict, List, Optional +from typing import Any, Dict, List, Optional, Union from praktika._environment import _Environment -from praktika._settings import _Settings +from praktika.cache import Cache +from praktika.s3 import S3 +from praktika.settings import Settings from praktika.utils import ContextManager, MetaClasses, Shell, Utils @@ -55,7 +57,7 @@ class Result(MetaClasses.Serializable): stopwatch: Utils.Stopwatch = None, status="", files=None, - info="", + info: Union[List[str], str] = "", with_info_from_results=True, ): if isinstance(status, bool): @@ -149,7 +151,7 @@ class Result(MetaClasses.Serializable): @classmethod def file_name_static(cls, name): - return f"{_Settings.TEMP_DIR}/result_{Utils.normalize_string(name)}.json" + return f"{Settings.TEMP_DIR}/result_{Utils.normalize_string(name)}.json" @classmethod def from_dict(cls, obj: Dict[str, Any]) -> "Result": @@ -232,7 +234,7 @@ class Result(MetaClasses.Serializable): ) @classmethod - def generate_skipped(cls, name, results=None): + def generate_skipped(cls, name, cache_record: Cache.CacheRecord, results=None): return Result( name=name, status=Result.Status.SKIPPED, @@ -241,7 +243,7 @@ class Result(MetaClasses.Serializable): results=results or [], files=[], links=[], - info="from cache", + info=f"from cache: sha [{cache_record.sha}], pr/branch [{cache_record.pr_number or cache_record.branch}]", ) @classmethod @@ -275,7 +277,7 @@ class Result(MetaClasses.Serializable): # Set log file path if logging is enabled log_file = ( - f"{_Settings.TEMP_DIR}/{Utils.normalize_string(name)}.log" + f"{Settings.TEMP_DIR}/{Utils.normalize_string(name)}.log" if with_log else None ) @@ -321,14 +323,31 @@ class Result(MetaClasses.Serializable): self.dump() if not self.is_ok(): print("ERROR: Job Failed") - for result in self.results: - if not result.is_ok(): - print("Failed checks:") - print(" | ", result) + print(self.to_stdout_formatted()) sys.exit(1) else: print("ok") + def to_stdout_formatted(self, indent="", res=""): + if self.is_ok(): + return res + + res += f"{indent}Task [{self.name}] failed.\n" + fail_info = "" + sub_indent = indent + " " + + if not self.results: + if not self.is_ok(): + fail_info += f"{sub_indent}{self.name}:\n" + for line in self.info.splitlines(): + fail_info += f"{sub_indent}{sub_indent}{line}\n" + return res + fail_info + + for sub_result in self.results: + res = sub_result.to_stdout_formatted(sub_indent, res) + + return res + class ResultInfo: SETUP_ENV_JOB_FAILED = ( @@ -351,3 +370,202 @@ class ResultInfo: ) S3_ERROR = "S3 call failure" + + +class _ResultS3: + + @classmethod + def copy_result_to_s3(cls, result, unlock=False): + result.dump() + env = _Environment.get() + s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}" + s3_path_full = f"{s3_path}/{Path(result.file_name()).name}" + url = S3.copy_file_to_s3(s3_path=s3_path, local_path=result.file_name()) + # if unlock: + # if not cls.unlock(s3_path_full): + # print(f"ERROR: File [{s3_path_full}] unlock failure") + # assert False # TODO: investigate + return url + + @classmethod + def copy_result_from_s3(cls, local_path, lock=False): + env = _Environment.get() + file_name = Path(local_path).name + s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}/{file_name}" + # if lock: + # cls.lock(s3_path) + if not S3.copy_file_from_s3(s3_path=s3_path, local_path=local_path): + print(f"ERROR: failed to cp file [{s3_path}] from s3") + raise + + @classmethod + def copy_result_from_s3_with_version(cls, local_path): + env = _Environment.get() + file_name = Path(local_path).name + local_dir = Path(local_path).parent + file_name_pattern = f"{file_name}_*" + for file_path in local_dir.glob(file_name_pattern): + file_path.unlink() + s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}/" + if not S3.copy_file_from_s3_matching_pattern( + s3_path=s3_path, local_path=local_dir, include=file_name_pattern + ): + print(f"ERROR: failed to cp file [{s3_path}] from s3") + raise + result_files = [] + for file_path in local_dir.glob(file_name_pattern): + result_files.append(file_path) + assert result_files, "No result files found" + result_files.sort() + version = int(result_files[-1].name.split("_")[-1]) + Shell.check(f"cp {result_files[-1]} {local_path}", strict=True, verbose=True) + return version + + @classmethod + def copy_result_to_s3_with_version(cls, result, version): + result.dump() + filename = Path(result.file_name()).name + file_name_versioned = f"{filename}_{str(version).zfill(3)}" + env = _Environment.get() + s3_path_versioned = ( + f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}/{file_name_versioned}" + ) + s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}/" + if version == 0: + S3.clean_s3_directory(s3_path=s3_path) + if not S3.put( + s3_path=s3_path_versioned, + local_path=result.file_name(), + if_none_matched=True, + ): + print("Failed to put versioned Result") + return False + if not S3.put(s3_path=s3_path, local_path=result.file_name()): + print("Failed to put non-versioned Result") + return True + + # @classmethod + # def lock(cls, s3_path, level=0): + # env = _Environment.get() + # s3_path_lock = s3_path + f".lock" + # file_path_lock = f"{Settings.TEMP_DIR}/{Path(s3_path_lock).name}" + # assert Shell.check( + # f"echo '''{env.JOB_NAME}''' > {file_path_lock}", verbose=True + # ), "Never" + # + # i = 20 + # meta = S3.head_object(s3_path_lock) + # while meta: + # locked_by_job = meta.get("Metadata", {"job": ""}).get("job", "") + # if locked_by_job: + # decoded_bytes = base64.b64decode(locked_by_job) + # locked_by_job = decoded_bytes.decode("utf-8") + # print( + # f"WARNING: Failed to acquire lock, meta [{meta}], job [{locked_by_job}] - wait" + # ) + # i -= 5 + # if i < 0: + # info = f"ERROR: lock acquire failure - unlock forcefully" + # print(info) + # env.add_info(info) + # break + # time.sleep(5) + # + # metadata = {"job": Utils.to_base64(env.JOB_NAME)} + # S3.put( + # s3_path=s3_path_lock, + # local_path=file_path_lock, + # metadata=metadata, + # if_none_matched=True, + # ) + # time.sleep(1) + # obj = S3.head_object(s3_path_lock) + # if not obj or not obj.has_tags(tags=metadata): + # print(f"WARNING: locked by another job [{obj}]") + # env.add_info("S3 lock file failure") + # cls.lock(s3_path, level=level + 1) + # print("INFO: lock acquired") + # + # @classmethod + # def unlock(cls, s3_path): + # s3_path_lock = s3_path + ".lock" + # env = _Environment.get() + # obj = S3.head_object(s3_path_lock) + # if not obj: + # print("ERROR: lock file is removed") + # assert False # investigate + # elif not obj.has_tags({"job": Utils.to_base64(env.JOB_NAME)}): + # print("ERROR: lock file was acquired by another job") + # assert False # investigate + # + # if not S3.delete(s3_path_lock): + # print(f"ERROR: File [{s3_path_lock}] delete failure") + # print("INFO: lock released") + # return True + + @classmethod + def upload_result_files_to_s3(cls, result): + if result.results: + for result_ in result.results: + cls.upload_result_files_to_s3(result_) + for file in result.files: + if not Path(file).is_file(): + print(f"ERROR: Invalid file [{file}] in [{result.name}] - skip upload") + result.info += f"\nWARNING: Result file [{file}] was not found" + file_link = S3._upload_file_to_s3(file, upload_to_s3=False) + else: + is_text = False + for text_file_suffix in Settings.TEXT_CONTENT_EXTENSIONS: + if file.endswith(text_file_suffix): + print( + f"File [{file}] matches Settings.TEXT_CONTENT_EXTENSIONS [{Settings.TEXT_CONTENT_EXTENSIONS}] - add text attribute for s3 object" + ) + is_text = True + break + file_link = S3._upload_file_to_s3( + file, + upload_to_s3=True, + text=is_text, + s3_subprefix=Utils.normalize_string(result.name), + ) + result.links.append(file_link) + if result.files: + print( + f"Result files [{result.files}] uploaded to s3 [{result.links[-len(result.files):]}] - clean files list" + ) + result.files = [] + result.dump() + + @classmethod + def update_workflow_results(cls, workflow_name, new_info="", new_sub_results=None): + assert new_info or new_sub_results + + attempt = 1 + prev_status = "" + new_status = "" + done = False + while attempt < 10: + version = cls.copy_result_from_s3_with_version( + Result.file_name_static(workflow_name) + ) + workflow_result = Result.from_fs(workflow_name) + prev_status = workflow_result.status + if new_info: + workflow_result.set_info(new_info) + if new_sub_results: + if isinstance(new_sub_results, Result): + new_sub_results = [new_sub_results] + for result_ in new_sub_results: + workflow_result.update_sub_result(result_) + new_status = workflow_result.status + if cls.copy_result_to_s3_with_version(workflow_result, version=version + 1): + done = True + break + print(f"Attempt [{attempt}] to upload workflow result failed") + attempt += 1 + assert done + + if prev_status != new_status: + return new_status + else: + return None diff --git a/ci/praktika/runner.py b/ci/praktika/runner.py index 1ac8748d1c0..38112dd5684 100644 --- a/ci/praktika/runner.py +++ b/ci/praktika/runner.py @@ -52,6 +52,7 @@ class Runner: cache_success=[], cache_success_base64=[], cache_artifacts={}, + cache_jobs={}, ) for docker in workflow.dockers: workflow_config.digest_dockers[docker.name] = Digest().calc_docker_digest( @@ -123,7 +124,7 @@ class Runner: return 0 - def _run(self, workflow, job, docker="", no_docker=False, param=None): + def _run(self, workflow, job, docker="", no_docker=False, param=None, test=""): # re-set envs for local run env = _Environment.get() env.JOB_NAME = job.name @@ -162,6 +163,9 @@ class Runner: if param: print(f"Custom --param [{param}] will be passed to job's script") cmd += f" --param {param}" + if test: + print(f"Custom --test [{test}] will be passed to job's script") + cmd += f" --test {test}" print(f"--- Run command [{cmd}]") with TeePopen(cmd, timeout=job.timeout) as process: @@ -240,10 +244,6 @@ class Runner: result.set_files(files=[Settings.RUN_LOG]) result.update_duration().dump() - if result.info and result.status != Result.Status.SUCCESS: - # provide job info to workflow level - info_errors.append(result.info) - if run_exit_code == 0: providing_artifacts = [] if job.provides and workflow.artifacts: @@ -310,6 +310,7 @@ class Runner: local_run=False, no_docker=False, param=None, + test="", pr=None, sha=None, branch=None, @@ -358,7 +359,12 @@ class Runner: print(f"=== Run script [{job.name}], workflow [{workflow.name}] ===") try: run_code = self._run( - workflow, job, docker=docker, no_docker=no_docker, param=param + workflow, + job, + docker=docker, + no_docker=no_docker, + param=param, + test=test, ) res = run_code == 0 if not res: diff --git a/ci/praktika/runtime.py b/ci/praktika/runtime.py index a87b67c2c79..07c24e0498c 100644 --- a/ci/praktika/runtime.py +++ b/ci/praktika/runtime.py @@ -15,17 +15,23 @@ class RunConfig(MetaClasses.Serializable): # there are might be issue with special characters in job names if used directly in yaml syntax - create base64 encoded list to avoid this cache_success_base64: List[str] cache_artifacts: Dict[str, Cache.CacheRecord] + cache_jobs: Dict[str, Cache.CacheRecord] sha: str @classmethod def from_dict(cls, obj): cache_artifacts = obj["cache_artifacts"] + cache_jobs = obj["cache_jobs"] cache_artifacts_deserialized = {} + cache_jobs_deserialized = {} for artifact_name, cache_artifact in cache_artifacts.items(): cache_artifacts_deserialized[artifact_name] = Cache.CacheRecord.from_dict( cache_artifact ) obj["cache_artifacts"] = cache_artifacts_deserialized + for job_name, cache_jobs in cache_jobs.items(): + cache_jobs_deserialized[job_name] = Cache.CacheRecord.from_dict(cache_jobs) + obj["cache_jobs"] = cache_artifacts_deserialized return RunConfig(**obj) @classmethod diff --git a/ci/praktika/s3.py b/ci/praktika/s3.py index 04a08622dcd..82034b57b80 100644 --- a/ci/praktika/s3.py +++ b/ci/praktika/s3.py @@ -1,12 +1,11 @@ import dataclasses import json -import time from pathlib import Path from typing import Dict from praktika._environment import _Environment from praktika.settings import Settings -from praktika.utils import Shell, Utils +from praktika.utils import Shell class S3: @@ -59,16 +58,15 @@ class S3: return f"https://{s3_full_path}".replace(bucket, endpoint) @classmethod - def put(cls, s3_path, local_path, text=False, metadata=None): + def put(cls, s3_path, local_path, text=False, metadata=None, if_none_matched=False): assert Path(local_path).exists(), f"Path [{local_path}] does not exist" assert Path(s3_path), f"Invalid S3 Path [{s3_path}]" assert Path( local_path ).is_file(), f"Path [{local_path}] is not file. Only files are supported" - file_name = Path(local_path).name s3_full_path = s3_path - if not s3_full_path.endswith(file_name): - s3_full_path = f"{s3_path}/{Path(local_path).name}" + if s3_full_path.endswith("/"): + s3_full_path = f"{s3_path}{Path(local_path).name}" s3_full_path = str(s3_full_path).removeprefix("s3://") bucket, key = s3_full_path.split("/", maxsplit=1) @@ -76,6 +74,8 @@ class S3: command = ( f"aws s3api put-object --bucket {bucket} --key {key} --body {local_path}" ) + if if_none_matched: + command += f' --if-none-match "*"' if metadata: for k, v in metadata.items(): command += f" --metadata {k}={v}" @@ -84,7 +84,7 @@ class S3: if text: cmd += " --content-type text/plain" res = cls.run_command_with_retries(command) - assert res + return res @classmethod def run_command_with_retries(cls, command, retries=Settings.MAX_RETRIES_S3): @@ -101,6 +101,14 @@ class S3: elif "does not exist" in stderr: print("ERROR: requested file does not exist") break + elif "Unknown options" in stderr: + print("ERROR: Invalid AWS CLI command or CLI client version:") + print(f" | awc error: {stderr}") + break + elif "PreconditionFailed" in stderr: + print("ERROR: AWS API Call Precondition Failed") + print(f" | awc error: {stderr}") + break if ret_code != 0: print( f"ERROR: aws s3 cp failed, stdout/stderr err: [{stderr}], out [{stdout}]" @@ -108,13 +116,6 @@ class S3: res = ret_code == 0 return res - @classmethod - def get_link(cls, s3_path, local_path): - s3_full_path = f"{s3_path}/{Path(local_path).name}" - bucket = s3_path.split("/")[0] - endpoint = Settings.S3_BUCKET_TO_HTTP_ENDPOINT[bucket] - return f"https://{s3_full_path}".replace(bucket, endpoint) - @classmethod def copy_file_from_s3(cls, s3_path, local_path): assert Path(s3_path), f"Invalid S3 Path [{s3_path}]" @@ -128,6 +129,19 @@ class S3: res = cls.run_command_with_retries(cmd) return res + @classmethod + def copy_file_from_s3_matching_pattern( + cls, s3_path, local_path, include, exclude="*" + ): + assert Path(s3_path), f"Invalid S3 Path [{s3_path}]" + assert Path( + local_path + ).is_dir(), f"Path [{local_path}] does not exist or not a directory" + assert s3_path.endswith("/"), f"s3 path is invalid [{s3_path}]" + cmd = f'aws s3 cp s3://{s3_path} {local_path} --exclude "{exclude}" --include "{include}" --recursive' + res = cls.run_command_with_retries(cmd) + return res + @classmethod def head_object(cls, s3_path): s3_path = str(s3_path).removeprefix("s3://") @@ -148,103 +162,6 @@ class S3: verbose=True, ) - # TODO: apparently should be placed into separate file to be used only inside praktika - # keeping this module clean from importing Settings, Environment and etc, making it easy for use externally - @classmethod - def copy_result_to_s3(cls, result, unlock=True): - result.dump() - env = _Environment.get() - s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}" - s3_path_full = f"{s3_path}/{Path(result.file_name()).name}" - url = S3.copy_file_to_s3(s3_path=s3_path, local_path=result.file_name()) - if env.PR_NUMBER: - print("Duplicate Result for latest commit alias in PR") - s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix(latest=True)}" - url = S3.copy_file_to_s3(s3_path=s3_path, local_path=result.file_name()) - if unlock: - if not cls.unlock(s3_path_full): - print(f"ERROR: File [{s3_path_full}] unlock failure") - assert False # TODO: investigate - return url - - @classmethod - def copy_result_from_s3(cls, local_path, lock=True): - env = _Environment.get() - file_name = Path(local_path).name - s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}/{file_name}" - if lock: - cls.lock(s3_path) - if not S3.copy_file_from_s3(s3_path=s3_path, local_path=local_path): - print(f"ERROR: failed to cp file [{s3_path}] from s3") - raise - - @classmethod - def lock(cls, s3_path, level=0): - assert level < 3, "Never" - env = _Environment.get() - s3_path_lock = s3_path + f".lock" - file_path_lock = f"{Settings.TEMP_DIR}/{Path(s3_path_lock).name}" - assert Shell.check( - f"echo '''{env.JOB_NAME}''' > {file_path_lock}", verbose=True - ), "Never" - - i = 20 - meta = S3.head_object(s3_path_lock) - while meta: - print(f"WARNING: Failed to acquire lock, meta [{meta}] - wait") - i -= 5 - if i < 0: - info = f"ERROR: lock acquire failure - unlock forcefully" - print(info) - env.add_info(info) - break - time.sleep(5) - - metadata = {"job": Utils.to_base64(env.JOB_NAME)} - S3.put( - s3_path=s3_path_lock, - local_path=file_path_lock, - metadata=metadata, - ) - time.sleep(1) - obj = S3.head_object(s3_path_lock) - if not obj or not obj.has_tags(tags=metadata): - print(f"WARNING: locked by another job [{obj}]") - env.add_info("S3 lock file failure") - cls.lock(s3_path, level=level + 1) - print("INFO: lock acquired") - - @classmethod - def unlock(cls, s3_path): - s3_path_lock = s3_path + ".lock" - env = _Environment.get() - obj = S3.head_object(s3_path_lock) - if not obj: - print("ERROR: lock file is removed") - assert False # investigate - elif not obj.has_tags({"job": Utils.to_base64(env.JOB_NAME)}): - print("ERROR: lock file was acquired by another job") - assert False # investigate - - if not S3.delete(s3_path_lock): - print(f"ERROR: File [{s3_path_lock}] delete failure") - print("INFO: lock released") - return True - - @classmethod - def get_result_link(cls, result): - env = _Environment.get() - s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix(latest=True if env.PR_NUMBER else False)}" - return S3.get_link(s3_path=s3_path, local_path=result.file_name()) - - @classmethod - def clean_latest_result(cls): - env = _Environment.get() - env.SHA = "latest" - assert env.PR_NUMBER - s3_path = f"{Settings.HTML_S3_PATH}/{env.get_s3_prefix()}" - S3.clean_s3_directory(s3_path=s3_path) - @classmethod def _upload_file_to_s3( cls, local_file_path, upload_to_s3: bool, text: bool = False, s3_subprefix="" @@ -260,36 +177,3 @@ class S3: ) return html_link return f"file://{Path(local_file_path).absolute()}" - - @classmethod - def upload_result_files_to_s3(cls, result): - if result.results: - for result_ in result.results: - cls.upload_result_files_to_s3(result_) - for file in result.files: - if not Path(file).is_file(): - print(f"ERROR: Invalid file [{file}] in [{result.name}] - skip upload") - result.info += f"\nWARNING: Result file [{file}] was not found" - file_link = cls._upload_file_to_s3(file, upload_to_s3=False) - else: - is_text = False - for text_file_suffix in Settings.TEXT_CONTENT_EXTENSIONS: - if file.endswith(text_file_suffix): - print( - f"File [{file}] matches Settings.TEXT_CONTENT_EXTENSIONS [{Settings.TEXT_CONTENT_EXTENSIONS}] - add text attribute for s3 object" - ) - is_text = True - break - file_link = cls._upload_file_to_s3( - file, - upload_to_s3=True, - text=is_text, - s3_subprefix=Utils.normalize_string(result.name), - ) - result.links.append(file_link) - if result.files: - print( - f"Result files [{result.files}] uploaded to s3 [{result.links[-len(result.files):]}] - clean files list" - ) - result.files = [] - result.dump() diff --git a/ci/praktika/settings.py b/ci/praktika/settings.py index 1a4068d9398..b281a95370c 100644 --- a/ci/praktika/settings.py +++ b/ci/praktika/settings.py @@ -1,8 +1,152 @@ -from praktika._settings import _Settings -from praktika.mangle import _get_user_settings +import dataclasses +import importlib.util +from pathlib import Path +from typing import Dict, Iterable, List, Optional -Settings = _Settings() -user_settings = _get_user_settings() -for setting, value in user_settings.items(): - Settings.__setattr__(setting, value) +@dataclasses.dataclass +class _Settings: + ###################################### + # Pipeline generation settings # + ###################################### + MAIN_BRANCH = "main" + CI_PATH = "./ci" + WORKFLOW_PATH_PREFIX: str = "./.github/workflows" + WORKFLOWS_DIRECTORY: str = f"{CI_PATH}/workflows" + SETTINGS_DIRECTORY: str = f"{CI_PATH}/settings" + CI_CONFIG_JOB_NAME = "Config Workflow" + DOCKER_BUILD_JOB_NAME = "Docker Builds" + FINISH_WORKFLOW_JOB_NAME = "Finish Workflow" + READY_FOR_MERGE_STATUS_NAME = "Ready for Merge" + CI_CONFIG_RUNS_ON: Optional[List[str]] = None + DOCKER_BUILD_RUNS_ON: Optional[List[str]] = None + VALIDATE_FILE_PATHS: bool = True + + ###################################### + # Runtime Settings # + ###################################### + MAX_RETRIES_S3 = 3 + MAX_RETRIES_GH = 3 + + ###################################### + # S3 (artifact storage) settings # + ###################################### + S3_ARTIFACT_PATH: str = "" + + ###################################### + # CI workspace settings # + ###################################### + TEMP_DIR: str = "/tmp/praktika" + OUTPUT_DIR: str = f"{TEMP_DIR}/output" + INPUT_DIR: str = f"{TEMP_DIR}/input" + PYTHON_INTERPRETER: str = "python3" + PYTHON_PACKET_MANAGER: str = "pip3" + PYTHON_VERSION: str = "3.9" + INSTALL_PYTHON_FOR_NATIVE_JOBS: bool = False + INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS: str = "./ci/requirements.txt" + ENVIRONMENT_VAR_FILE: str = f"{TEMP_DIR}/environment.json" + RUN_LOG: str = f"{TEMP_DIR}/praktika_run.log" + + SECRET_GH_APP_ID: str = "GH_APP_ID" + SECRET_GH_APP_PEM_KEY: str = "GH_APP_PEM_KEY" + + ENV_SETUP_SCRIPT: str = "/tmp/praktika_setup_env.sh" + WORKFLOW_STATUS_FILE: str = f"{TEMP_DIR}/workflow_status.json" + + ###################################### + # CI Cache settings # + ###################################### + CACHE_VERSION: int = 1 + CACHE_DIGEST_LEN: int = 20 + CACHE_S3_PATH: str = "" + CACHE_LOCAL_PATH: str = f"{TEMP_DIR}/ci_cache" + + ###################################### + # Report settings # + ###################################### + HTML_S3_PATH: str = "" + HTML_PAGE_FILE: str = "./praktika/json.html" + TEXT_CONTENT_EXTENSIONS: Iterable[str] = frozenset([".txt", ".log"]) + S3_BUCKET_TO_HTTP_ENDPOINT: Optional[Dict[str, str]] = None + + DOCKERHUB_USERNAME: str = "" + DOCKERHUB_SECRET: str = "" + DOCKER_WD: str = "/wd" + + ###################################### + # CI DB Settings # + ###################################### + SECRET_CI_DB_URL: str = "CI_DB_URL" + SECRET_CI_DB_PASSWORD: str = "CI_DB_PASSWORD" + CI_DB_DB_NAME = "" + CI_DB_TABLE_NAME = "" + CI_DB_INSERT_TIMEOUT_SEC = 5 + + DISABLE_MERGE_COMMIT = True + + +_USER_DEFINED_SETTINGS = [ + "S3_ARTIFACT_PATH", + "CACHE_S3_PATH", + "HTML_S3_PATH", + "S3_BUCKET_TO_HTTP_ENDPOINT", + "TEXT_CONTENT_EXTENSIONS", + "TEMP_DIR", + "OUTPUT_DIR", + "INPUT_DIR", + "CI_CONFIG_RUNS_ON", + "DOCKER_BUILD_RUNS_ON", + "CI_CONFIG_JOB_NAME", + "PYTHON_INTERPRETER", + "PYTHON_VERSION", + "PYTHON_PACKET_MANAGER", + "INSTALL_PYTHON_FOR_NATIVE_JOBS", + "INSTALL_PYTHON_REQS_FOR_NATIVE_JOBS", + "MAX_RETRIES_S3", + "MAX_RETRIES_GH", + "VALIDATE_FILE_PATHS", + "DOCKERHUB_USERNAME", + "DOCKERHUB_SECRET", + "READY_FOR_MERGE_STATUS_NAME", + "SECRET_CI_DB_URL", + "SECRET_CI_DB_PASSWORD", + "CI_DB_DB_NAME", + "CI_DB_TABLE_NAME", + "CI_DB_INSERT_TIMEOUT_SEC", + "SECRET_GH_APP_PEM_KEY", + "SECRET_GH_APP_ID", + "MAIN_BRANCH", + "DISABLE_MERGE_COMMIT", +] + + +def _get_settings() -> _Settings: + res = _Settings() + + directory = Path(_Settings.SETTINGS_DIRECTORY) + for py_file in directory.glob("*.py"): + module_name = py_file.name.removeprefix(".py") + spec = importlib.util.spec_from_file_location( + module_name, f"{_Settings.SETTINGS_DIRECTORY}/{module_name}" + ) + assert spec + foo = importlib.util.module_from_spec(spec) + assert spec.loader + spec.loader.exec_module(foo) + for setting in _USER_DEFINED_SETTINGS: + try: + value = getattr(foo, setting) + res.__setattr__(setting, value) + # print(f"- read user defined setting [{setting} = {value}]") + except Exception as e: + # print(f"Exception while read user settings: {e}") + pass + + return res + + +class GHRunners: + ubuntu = "ubuntu-latest" + + +Settings = _get_settings() diff --git a/ci/praktika/utils.py b/ci/praktika/utils.py index 62eb13b3e19..2bcc94f2559 100644 --- a/ci/praktika/utils.py +++ b/ci/praktika/utils.py @@ -17,8 +17,6 @@ from threading import Thread from types import SimpleNamespace from typing import Any, Dict, Iterator, List, Optional, Type, TypeVar, Union -from praktika._settings import _Settings - T = TypeVar("T", bound="Serializable") diff --git a/ci/praktika/validator.py b/ci/praktika/validator.py index d612881b819..0bb722903e5 100644 --- a/ci/praktika/validator.py +++ b/ci/praktika/validator.py @@ -4,10 +4,8 @@ from itertools import chain from pathlib import Path from praktika import Workflow -from praktika._settings import GHRunners from praktika.mangle import _get_workflows -from praktika.settings import Settings -from praktika.utils import ContextManager +from praktika.settings import GHRunners, Settings class Validator: @@ -168,9 +166,7 @@ class Validator: "\n echo requests==2.32.3 >> ./ci/requirements.txt" ) message += "\n echo https://clickhouse-builds.s3.amazonaws.com/packages/praktika-0.1-py3-none-any.whl >> ./ci/requirements.txt" - cls.evaluate_check( - path.is_file(), message, job.name, workflow.name - ) + cls.evaluate_check(path.is_file(), message, job.name, workflow.name) @classmethod def validate_dockers(cls, workflow: Workflow.Config): diff --git a/ci/workflows/pull_request.py b/ci/workflows/pull_request.py index 0d505ae27c4..707babb1250 100644 --- a/ci/workflows/pull_request.py +++ b/ci/workflows/pull_request.py @@ -68,6 +68,7 @@ stateless_tests_jobs = Job.Config( name=JobNames.STATELESS, runs_on=[RunnerLabels.BUILDER], command="python3 ./ci/jobs/functional_stateless_tests.py --test-options {PARAMETER}", + # many tests expect to see "/var/lib/clickhouse" in various output lines - add mount for now, consider creating this dir in docker file run_in_docker="clickhouse/stateless-test+--security-opt seccomp=unconfined+--volume=/tmp/praktika:/var/lib/clickhouse", digest_config=Job.CacheDigestConfig( include_paths=[ From 7310376413e34ad2c958c5ccde9cddd75e0a5aed Mon Sep 17 00:00:00 2001 From: Alexey Milovidov Date: Tue, 12 Nov 2024 01:23:01 +0100 Subject: [PATCH 142/244] Remove ridiculous code bloat --- .../AggregateFunctionDeltaSumTimestamp.cpp | 69 ++++++++++++++---- src/AggregateFunctions/Helpers.h | 70 +------------------ 2 files changed, 58 insertions(+), 81 deletions(-) diff --git a/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp b/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp index dc1adead87c..0c5b752b539 100644 --- a/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp +++ b/src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp @@ -22,6 +22,13 @@ namespace ErrorCodes namespace { +/** Due to a lack of proper code review, this code was contributed with a multiplication of template instantiations + * over all pairs of data types, and we deeply regret that. + * + * We cannot remove all combinations, because the binary representation of serialized data has to remain the same, + * but we can partially heal the wound by treating unsigned and signed data types in the same way. + */ + template struct AggregationFunctionDeltaSumTimestampData { @@ -37,23 +44,22 @@ template class AggregationFunctionDeltaSumTimestamp final : public IAggregateFunctionDataHelper< AggregationFunctionDeltaSumTimestampData, - AggregationFunctionDeltaSumTimestamp - > + AggregationFunctionDeltaSumTimestamp> { public: AggregationFunctionDeltaSumTimestamp(const DataTypes & arguments, const Array & params) : IAggregateFunctionDataHelper< AggregationFunctionDeltaSumTimestampData, - AggregationFunctionDeltaSumTimestamp - >{arguments, params, createResultType()} - {} + AggregationFunctionDeltaSumTimestamp>{arguments, params, createResultType()} + { + } AggregationFunctionDeltaSumTimestamp() : IAggregateFunctionDataHelper< AggregationFunctionDeltaSumTimestampData, - AggregationFunctionDeltaSumTimestamp - >{} - {} + AggregationFunctionDeltaSumTimestamp>{} + { + } bool allocatesMemoryInArena() const override { return false; } @@ -63,8 +69,8 @@ public: void NO_SANITIZE_UNDEFINED ALWAYS_INLINE add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override { - auto value = assert_cast &>(*columns[0]).getData()[row_num]; - auto ts = assert_cast &>(*columns[1]).getData()[row_num]; + auto value = unalignedLoad(columns[0]->getRawData().data() + row_num * sizeof(ValueType)); + auto ts = unalignedLoad(columns[1]->getRawData().data() + row_num * sizeof(TimestampType)); auto & data = this->data(place); @@ -172,10 +178,49 @@ public: void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena *) const override { - assert_cast &>(to).getData().push_back(this->data(place).sum); + static_cast(to).template insertRawData( + reinterpret_cast(&this->data(place).sum)); } }; + + +template class AggregateFunctionTemplate, typename... TArgs> +static IAggregateFunction * createWithTwoTypesSecond(const IDataType & second_type, TArgs && ... args) +{ + WhichDataType which(second_type); + + if (which.idx == TypeIndex::UInt32) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::UInt64) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::Int32) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::Int64) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::Float32) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::Float64) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::Date) return new AggregateFunctionTemplate(args...); + if (which.idx == TypeIndex::DateTime) return new AggregateFunctionTemplate(args...); + + return nullptr; +} + +template