Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

GH-35166: [C++] Increase precision of decimals in aggregate functions #44184

Open
wants to merge 16 commits into
base: main
Choose a base branch
from
Open

GH-35166: [C++] Increase precision of decimals in aggregate functions #44184

Show file tree
Hide file tree
Changes from 15 commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
1acf888
Increase the precision of decimals during compute
khwilson Sep 17, 2024
3069b2d
migrate utility function to codegen
khwilson Sep 22, 2024
85f5226
fix else condition
khwilson Sep 22, 2024
22df961
improve casts
khwilson Sep 22, 2024
c0ec05d
linting
khwilson Sep 23, 2024
07a6a6b
fix _pointer_ in casts
khwilson Sep 24, 2024
fb7a581
extract common code
khwilson Sep 30, 2024
03c0bc6
continue simplifying code
khwilson Oct 1, 2024
12a2749
add doc and export
khwilson Oct 10, 2024
01e53b3
fix c++ tests
khwilson Oct 13, 2024
9901d1f
linting and doc fixes
khwilson Oct 13, 2024
15ead92
include hash aggregate sums
khwilson Oct 13, 2024
2156906
fix more formatting
khwilson Oct 13, 2024
8eb8cb3
fix hash aggregate test
khwilson Oct 13, 2024
0946f65
update c++ docs on decimal aggregates
khwilson Oct 13, 2024
1a3b1c4
Update cpp/src/arrow/compute/kernels/codegen_internal.h
khwilson Oct 14, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions cpp/src/arrow/acero/hash_aggregate_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1798,12 +1798,12 @@ TEST_P(GroupBy, SumMeanProductDecimal) {

AssertDatumsEqual(ArrayFromJSON(struct_({
field("key_0", int64()),
field("hash_sum", decimal128(3, 2)),
field("hash_sum", decimal256(3, 2)),
field("hash_mean", decimal128(3, 2)),
field("hash_mean", decimal256(3, 2)),
field("hash_product", decimal128(3, 2)),
field("hash_product", decimal256(3, 2)),
field("hash_sum", decimal128(38, 2)),
field("hash_sum", decimal256(76, 2)),
field("hash_mean", decimal128(38, 2)),
field("hash_mean", decimal256(76, 2)),
field("hash_product", decimal128(38, 2)),
field("hash_product", decimal256(76, 2)),
}),
R"([
[1, "4.25", "4.25", "2.13", "2.13", "3.25", "3.25"],
Expand Down
50 changes: 35 additions & 15 deletions cpp/src/arrow/compute/kernels/aggregate_basic.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -365,11 +365,16 @@ struct ProductImpl : public ScalarAggregator {
}

Status Finalize(KernelContext*, Datum* out) override {
std::shared_ptr<DataType> out_type_ = this->out_type;
if (is_decimal(this->out_type->id())) {
ARROW_ASSIGN_OR_RAISE(out_type_, WidenDecimalToMaxPrecision(this->out_type));
}

if ((!options.skip_nulls && this->nulls_observed) ||
(this->count < options.min_count)) {
out->value = std::make_shared<OutputType>(out_type);
out->value = std::make_shared<OutputType>(out_type_);
} else {
out->value = std::make_shared<OutputType>(this->product, out_type);
out->value = std::make_shared<OutputType>(this->product, out_type_);
}
return Status::OK();
}
Expand Down Expand Up @@ -1048,10 +1053,14 @@ void RegisterScalarAggregateBasic(FunctionRegistry* registry) {
func = std::make_shared<ScalarAggregateFunction>("sum", Arity::Unary(), sum_doc,
&default_scalar_aggregate_options);
AddArrayScalarAggKernels(SumInit, {boolean()}, uint64(), func.get());
AddAggKernel(KernelSignature::Make({Type::DECIMAL128}, FirstType), SumInit, func.get(),
SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL256}, FirstType), SumInit, func.get(),
SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL32}, MaxPrecisionDecimalType), SumInit,
func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL64}, MaxPrecisionDecimalType), SumInit,
func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL128}, MaxPrecisionDecimalType),
SumInit, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL256}, MaxPrecisionDecimalType),
SumInit, func.get(), SimdLevel::NONE);
AddArrayScalarAggKernels(SumInit, SignedIntTypes(), int64(), func.get());
AddArrayScalarAggKernels(SumInit, UnsignedIntTypes(), uint64(), func.get());
AddArrayScalarAggKernels(SumInit, FloatingPointTypes(), float64(), func.get());
Expand All @@ -1076,10 +1085,14 @@ void RegisterScalarAggregateBasic(FunctionRegistry* registry) {
&default_scalar_aggregate_options);
AddArrayScalarAggKernels(MeanInit, {boolean()}, float64(), func.get());
AddArrayScalarAggKernels(MeanInit, NumericTypes(), float64(), func.get());
AddAggKernel(KernelSignature::Make({Type::DECIMAL128}, FirstType), MeanInit, func.get(),
SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL256}, FirstType), MeanInit, func.get(),
SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL32}, MaxPrecisionDecimalType),
MeanInit, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL64}, MaxPrecisionDecimalType),
MeanInit, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL128}, MaxPrecisionDecimalType),
MeanInit, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL256}, MaxPrecisionDecimalType),
MeanInit, func.get(), SimdLevel::NONE);
AddArrayScalarAggKernels(MeanInit, {null()}, float64(), func.get());
// Add the SIMD variants for mean
#if defined(ARROW_HAVE_RUNTIME_AVX2)
Expand Down Expand Up @@ -1125,6 +1138,8 @@ void RegisterScalarAggregateBasic(FunctionRegistry* registry) {
AddMinMaxKernels(MinMaxInitDefault, BaseBinaryTypes(), func.get());
AddMinMaxKernel(MinMaxInitDefault, Type::FIXED_SIZE_BINARY, func.get());
AddMinMaxKernel(MinMaxInitDefault, Type::INTERVAL_MONTHS, func.get());
AddMinMaxKernel(MinMaxInitDefault, Type::DECIMAL32, func.get());
AddMinMaxKernel(MinMaxInitDefault, Type::DECIMAL64, func.get());
AddMinMaxKernel(MinMaxInitDefault, Type::DECIMAL128, func.get());
AddMinMaxKernel(MinMaxInitDefault, Type::DECIMAL256, func.get());
// Add the SIMD variants for min max
Expand Down Expand Up @@ -1160,10 +1175,14 @@ void RegisterScalarAggregateBasic(FunctionRegistry* registry) {
AddArrayScalarAggKernels(ProductInit::Init, UnsignedIntTypes(), uint64(), func.get());
AddArrayScalarAggKernels(ProductInit::Init, FloatingPointTypes(), float64(),
func.get());
AddAggKernel(KernelSignature::Make({Type::DECIMAL128}, FirstType), ProductInit::Init,
func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL256}, FirstType), ProductInit::Init,
func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL32}, MaxPrecisionDecimalType),
ProductInit::Init, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL64}, MaxPrecisionDecimalType),
ProductInit::Init, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL128}, MaxPrecisionDecimalType),
ProductInit::Init, func.get(), SimdLevel::NONE);
AddAggKernel(KernelSignature::Make({Type::DECIMAL256}, MaxPrecisionDecimalType),
ProductInit::Init, func.get(), SimdLevel::NONE);
AddArrayScalarAggKernels(ProductInit::Init, {null()}, int64(), func.get());
DCHECK_OK(registry->AddFunction(std::move(func)));

Expand All @@ -1185,7 +1204,8 @@ void RegisterScalarAggregateBasic(FunctionRegistry* registry) {
AddBasicAggKernels(IndexInit::Init, PrimitiveTypes(), int64(), func.get());
AddBasicAggKernels(IndexInit::Init, TemporalTypes(), int64(), func.get());
AddBasicAggKernels(IndexInit::Init,
{fixed_size_binary(1), decimal128(1, 0), decimal256(1, 0), null()},
{fixed_size_binary(1), decimal32(1, 0), decimal64(1, 0),
decimal128(1, 0), decimal256(1, 0), null()},
int64(), func.get());
DCHECK_OK(registry->AddFunction(std::move(func)));
}
Expand Down
16 changes: 12 additions & 4 deletions cpp/src/arrow/compute/kernels/aggregate_basic.inc.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,11 +93,16 @@ struct SumImpl : public ScalarAggregator {
}

Status Finalize(KernelContext*, Datum* out) override {
std::shared_ptr<DataType> out_type_ = this->out_type;
if (is_decimal(this->out_type->id())) {
ARROW_ASSIGN_OR_RAISE(out_type_, WidenDecimalToMaxPrecision(this->out_type));
}

if ((!options.skip_nulls && this->nulls_observed) ||
(this->count < options.min_count)) {
out->value = std::make_shared<OutputType>(out_type);
out->value = std::make_shared<OutputType>(out_type_);
} else {
out->value = std::make_shared<OutputType>(this->sum, out_type);
out->value = std::make_shared<OutputType>(this->sum, out_type_);
}
return Status::OK();
}
Expand Down Expand Up @@ -220,9 +225,12 @@ struct MeanImpl<ArrowType, SimdLevel, enable_if_decimal<ArrowType>>

template <typename T = ArrowType>
Status FinalizeImpl(Datum* out) {
std::shared_ptr<DataType> out_type_;
ARROW_ASSIGN_OR_RAISE(out_type_, WidenDecimalToMaxPrecision(this->out_type));

if ((!options.skip_nulls && this->nulls_observed) ||
(this->count < options.min_count) || (this->count == 0)) {
out->value = std::make_shared<OutputType>(this->out_type);
out->value = std::make_shared<OutputType>(out_type_);
} else {
SumCType quotient, remainder;
ARROW_ASSIGN_OR_RAISE(std::tie(quotient, remainder), this->sum.Divide(this->count));
Expand All @@ -235,7 +243,7 @@ struct MeanImpl<ArrowType, SimdLevel, enable_if_decimal<ArrowType>>
quotient -= 1;
}
}
out->value = std::make_shared<OutputType>(quotient, this->out_type);
out->value = std::make_shared<OutputType>(quotient, out_type_);
}
return Status::OK();
}
Expand Down
Loading
Loading