Merge pull request #22938 from kitaisreal/aggregate-function-allocates-memory-in-arena-default-impl-fix

IAggreagteFunction allocatesMemoryInArena removed default implementation

src/AggregateFunctions/AggregateFunctionAggThrow.cpp

@@ -60,6 +60,8 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void create(AggregateDataPtr __restrict place) const override
     {
         if (std::uniform_real_distribution<>(0.0, 1.0)(thread_local_rng) <= throw_probability)

src/AggregateFunctions/AggregateFunctionAvg.h

@@ -98,6 +98,8 @@ public:
 
     DataTypePtr getReturnType() const final { return std::make_shared<DataTypeNumber<Float64>>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void NO_SANITIZE_UNDEFINED merge(AggregateDataPtr __restrict place, ConstAggregateDataPtr rhs, Arena *) const override
     {
         this->data(place).numerator += this->data(rhs).numerator;

src/AggregateFunctions/AggregateFunctionBitwise.h

@@ -54,6 +54,8 @@ public:
         return std::make_shared<DataTypeNumber<T>>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).update(assert_cast<const ColumnVector<T> &>(*columns[0]).getData()[row_num]);

src/AggregateFunctions/AggregateFunctionBoundingRatio.h

@@ -127,6 +127,8 @@ public:
         return std::make_shared<DataTypeFloat64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, const size_t row_num, Arena *) const override
     {
         /// NOTE Slightly inefficient.

src/AggregateFunctions/AggregateFunctionCategoricalInformationValue.h

@@ -33,6 +33,8 @@ public:
         return "categoricalInformationValue";
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void create(AggregateDataPtr __restrict place) const override
     {
         memset(place, 0, sizeOfData());

src/AggregateFunctions/AggregateFunctionCount.h

@@ -38,6 +38,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn **, size_t, Arena *) const override
     {
         ++data(place).count;
@@ -126,6 +128,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         data(place).count += !assert_cast<const ColumnNullable &>(*columns[0]).isNullAt(row_num);

src/AggregateFunctions/AggregateFunctionDeltaSum.h

@@ -43,6 +43,8 @@ public:
 
     DataTypePtr getReturnType() const override { return std::make_shared<DataTypeNumber<T>>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void NO_SANITIZE_UNDEFINED ALWAYS_INLINE add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         auto value = assert_cast<const ColumnVector<T> &>(*columns[0]).getData()[row_num];

src/AggregateFunctions/AggregateFunctionEntropy.h

@@ -103,6 +103,8 @@ public:
         return std::make_shared<DataTypeNumber<Float64>>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         if constexpr (!std::is_same_v<UInt128, Value>)

src/AggregateFunctions/AggregateFunctionGroupArrayInsertAt.h

@@ -104,6 +104,8 @@ public:
         return std::make_shared<DataTypeArray>(type);
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         /// TODO Do positions need to be 1-based for this function?

src/AggregateFunctions/AggregateFunctionGroupBitmap.h

@@ -22,6 +22,8 @@ public:
 
     DataTypePtr getReturnType() const override { return std::make_shared<DataTypeNumber<T>>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).rbs.add(assert_cast<const ColumnVector<T> &>(*columns[0]).getData()[row_num]);
@@ -56,6 +58,8 @@ public:
 
     DataTypePtr getReturnType() const override { return std::make_shared<DataTypeNumber<T>>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         Data & data_lhs = this->data(place);

src/AggregateFunctions/AggregateFunctionGroupUniqArray.h

@@ -59,6 +59,8 @@ public:
         return std::make_shared<DataTypeArray>(this->argument_types[0]);
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         if (limit_num_elems && this->data(place).value.size() >= max_elems)

src/AggregateFunctions/AggregateFunctionHistogram.h

@@ -332,6 +332,8 @@ public:
         return std::make_shared<DataTypeArray>(tuple);
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         auto val = assert_cast<const ColumnVector<T> &>(*columns[0]).getData()[row_num];

src/AggregateFunctions/AggregateFunctionMLMethod.h

@@ -323,6 +323,8 @@ public:
         return std::make_shared<DataTypeArray>(std::make_shared<DataTypeFloat64>());
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     /// This function is called from evalMLMethod function for correct predictValues call
     DataTypePtr getReturnTypeToPredict() const override
     {

src/AggregateFunctions/AggregateFunctionMaxIntersections.h

@@ -87,6 +87,8 @@ public:
             return std::make_shared<DataTypeNumber<PointType>>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena * arena) const override
     {
         PointType left = assert_cast<const ColumnVector<PointType> &>(*columns[0]).getData()[row_num];

src/AggregateFunctions/AggregateFunctionNothing.h

@@ -28,6 +28,8 @@ public:
         return argument_types.front();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void create(AggregateDataPtr) const override
     {
     }

src/AggregateFunctions/AggregateFunctionQuantile.h

@@ -103,6 +103,8 @@ public:
             return res;
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         auto value = static_cast<const ColVecType &>(*columns[0]).getData()[row_num];

src/AggregateFunctions/AggregateFunctionRetention.h

@@ -94,6 +94,8 @@ public:
         return std::make_shared<DataTypeArray>(std::make_shared<DataTypeUInt8>());
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, const size_t row_num, Arena *) const override
     {
         for (const auto i : ext::range(0, events_size))

src/AggregateFunctions/AggregateFunctionSequenceMatch.h

@@ -560,6 +560,8 @@ public:
 
     DataTypePtr getReturnType() const override { return std::make_shared<DataTypeUInt8>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena *) const override
     {
         this->data(place).sort();
@@ -588,6 +590,8 @@ public:
 
     DataTypePtr getReturnType() const override { return std::make_shared<DataTypeUInt64>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena *) const override
     {
         this->data(place).sort();

src/AggregateFunctions/AggregateFunctionSimpleLinearRegression.h

@@ -168,6 +168,8 @@ public:
         );
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void insertResultInto(
         AggregateDataPtr place,
         IColumn & to,

src/AggregateFunctions/AggregateFunctionStatistics.h

@@ -123,6 +123,8 @@ public:
         return std::make_shared<DataTypeFloat64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).update(*columns[0], row_num);
@@ -375,6 +377,8 @@ public:
         return std::make_shared<DataTypeFloat64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).update(*columns[0], *columns[1], row_num);

src/AggregateFunctions/AggregateFunctionStatisticsSimple.h

@@ -121,6 +121,8 @@ public:
         return std::make_shared<DataTypeNumber<ResultType>>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         if constexpr (StatFunc::num_args == 2)

src/AggregateFunctions/AggregateFunctionSum.h

@@ -314,6 +314,8 @@ public:
             return std::make_shared<ResultDataType>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         const auto & column = assert_cast<const ColVecType &>(*columns[0]);

src/AggregateFunctions/AggregateFunctionSumMap.h

@@ -140,6 +140,8 @@ public:
         return std::make_shared<DataTypeTuple>(types);
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     static const auto & getArgumentColumns(const IColumn**& columns)
     {
         if constexpr (tuple_argument)

src/AggregateFunctions/AggregateFunctionTTest.h

@@ -109,6 +109,8 @@ public:
         );
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         Float64 value = columns[0]->getFloat64(row_num);

src/AggregateFunctions/AggregateFunctionTopK.h

@@ -50,6 +50,8 @@ public:
         return std::make_shared<DataTypeArray>(this->argument_types[0]);
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         auto & set = this->data(place).value;

src/AggregateFunctions/AggregateFunctionUniq.h

@@ -210,6 +210,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     /// ALWAYS_INLINE is required to have better code layout for uniqHLL12 function
     void ALWAYS_INLINE add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
@@ -265,6 +267,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).set.insert(typename Data::Set::value_type(

src/AggregateFunctions/AggregateFunctionUniqCombined.h

@@ -141,6 +141,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         if constexpr (!std::is_same_v<T, String>)
@@ -211,6 +213,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).set.insert(typename AggregateFunctionUniqCombinedData<UInt64, K, HashValueType>::Set::value_type(

src/AggregateFunctions/AggregateFunctionUniqUpTo.h

@@ -184,6 +184,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     /// ALWAYS_INLINE is required to have better code layout for uniqUpTo function
     void ALWAYS_INLINE add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
@@ -247,6 +249,8 @@ public:
         return std::make_shared<DataTypeUInt64>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena *) const override
     {
         this->data(place).insert(UInt64(UniqVariadicHash<is_exact, argument_is_tuple>::apply(num_args, columns, row_num)), threshold);

src/AggregateFunctions/AggregateFunctionWindowFunnel.h

@@ -247,6 +247,8 @@ public:
         return std::make_shared<DataTypeUInt8>();
     }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     AggregateFunctionPtr getOwnNullAdapter(
         const AggregateFunctionPtr & nested_function, const DataTypes & arguments, const Array & params,
         const AggregateFunctionProperties & /*properties*/) const override

src/AggregateFunctions/IAggregateFunction.h

@@ -104,7 +104,7 @@ public:
     virtual void deserialize(AggregateDataPtr __restrict place, ReadBuffer & buf, Arena * arena) const = 0;
 
     /// Returns true if a function requires Arena to handle own states (see add(), merge(), deserialize()).
-    virtual bool allocatesMemoryInArena() const { return false; }
+    virtual bool allocatesMemoryInArena() const = 0;
 
     /// Inserts results into a column. This method might modify the state (e.g.
     /// sort an array), so must be called once, from single thread. The state

src/Processors/Transforms/WindowTransform.cpp

@@ -1377,6 +1377,8 @@ struct WindowFunctionRank final : public WindowFunction
     DataTypePtr getReturnType() const override
     { return std::make_shared<DataTypeUInt64>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void windowInsertResultInto(const WindowTransform * transform,
         size_t function_index) override
     {
@@ -1397,6 +1399,8 @@ struct WindowFunctionDenseRank final : public WindowFunction
     DataTypePtr getReturnType() const override
     { return std::make_shared<DataTypeUInt64>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void windowInsertResultInto(const WindowTransform * transform,
         size_t function_index) override
     {
@@ -1417,6 +1421,8 @@ struct WindowFunctionRowNumber final : public WindowFunction
     DataTypePtr getReturnType() const override
     { return std::make_shared<DataTypeUInt64>(); }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void windowInsertResultInto(const WindowTransform * transform,
         size_t function_index) override
     {
@@ -1483,6 +1489,8 @@ struct WindowFunctionLagLeadInFrame final : public WindowFunction
     DataTypePtr getReturnType() const override
     { return argument_types[0]; }
 
+    bool allocatesMemoryInArena() const override { return false; }
+
     void windowInsertResultInto(const WindowTransform * transform,
         size_t function_index) override
     {