You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@datafusion.apache.org by "davidlghellin (via GitHub)" <gi...@apache.org> on 2025/11/09 18:41:56 UTC
[PR] feat(spark): implement Spark try_sum function [datafusion]
davidlghellin opened a new pull request, #18569:
URL: https://github.com/apache/datafusion/pull/18569
## Which issue does this PR close?
part of https://github.com/apache/datafusion/issues/15914
## Rationale for this change
Migrate spark functions from https://github.com/lakehq/sail/ to datafusion engine to unify codebase
## What changes are included in this PR?
implement spark udaf try_sum
https://spark.apache.org/docs/latest/api/sql/index.html#try_sum
## Are these changes tested?
unit-tests and sqllogictests added
## Are there any user-facing changes?
now can be called in queries
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "davidlghellin (via GitHub)" <gi...@apache.org>.
davidlghellin commented on PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#issuecomment-3662091767
Hi @Jefffrey , sorry for the long delay without being able to push anything.
I’ve made the changes (I hope I understood correctly what you meant).
Thanks!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "coderfender (via GitHub)" <gi...@apache.org>.
coderfender commented on PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#issuecomment-3666303487
Interesting . For reference here is the ANSI support EPIC on comet : https://github.com/apache/datafusion-comet/issues/313 .. I would love to collaborate with you on this and perhaps avoid redundancy ?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2627929878
##########
datafusion/spark/src/function/aggregate/try_sum.rs:
##########
@@ -0,0 +1,650 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use arrow::array::{ArrayRef, ArrowNumericType, AsArray, BooleanArray, PrimitiveArray};
+use arrow::datatypes::{
+ DECIMAL128_MAX_PRECISION, DataType, Decimal128Type, Field, FieldRef, Float64Type,
+ Int64Type, UInt64Type,
+};
+use datafusion_common::{Result, ScalarValue, downcast_value, exec_err, not_impl_err};
+use datafusion_expr::function::{AccumulatorArgs, StateFieldsArgs};
+use datafusion_expr::utils::format_state_name;
+use datafusion_expr::{Accumulator, AggregateUDFImpl, Signature, Volatility};
+use std::any::Any;
+use std::fmt::{Debug, Formatter};
+use std::mem::size_of_val;
+
+#[derive(PartialEq, Eq, Hash)]
+pub struct SparkTrySum {
+ signature: Signature,
+}
+
+impl Default for SparkTrySum {
+ fn default() -> Self {
+ Self::new()
+ }
+}
+
+impl SparkTrySum {
+ pub fn new() -> Self {
+ Self {
+ signature: Signature::user_defined(Volatility::Immutable),
+ }
+ }
+}
+
+impl Debug for SparkTrySum {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ f.debug_struct("SparkTrySum")
+ .field("signature", &self.signature)
+ .finish()
+ }
+}
+
+/// Accumulator for try_sum that detects overflow
+struct TrySumAccumulator<T: ArrowNumericType> {
+ sum: Option<T::Native>,
+ data_type: DataType,
+ failed: bool,
+ // Only used if data_type is Decimal128(p, s)
+ dec_precision: Option<u8>,
+}
+
+impl<T: ArrowNumericType> Debug for TrySumAccumulator<T> {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ write!(f, "TrySumAccumulator({})", self.data_type)
+ }
+}
+
+impl<T: ArrowNumericType> TrySumAccumulator<T> {
+ fn new(data_type: DataType) -> Self {
+ let dec_precision = match &data_type {
+ DataType::Decimal128(p, _) => Some(*p),
+ _ => None,
+ };
+ Self {
+ sum: None,
+ data_type,
+ failed: false,
+ dec_precision,
+ }
+ }
+}
+
+impl<T: ArrowNumericType> Accumulator for TrySumAccumulator<T> {
+ fn state(&mut self) -> Result<Vec<ScalarValue>> {
+ Ok(vec![
+ self.evaluate()?,
+ ScalarValue::Boolean(Some(self.failed)),
+ ])
+ }
+
+ fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+ update_batch_internal(self, values)
+ }
+
+ fn merge_batch(&mut self, states: &[ArrayRef]) -> Result<()> {
+ // Check if any partition has failed
+ let failed_arr = downcast_value!(states[1], BooleanArray);
+ for failed in failed_arr.iter().flatten() {
+ if failed {
+ self.failed = true;
+ return Ok(());
+ }
+ }
+
+ // Merge the sum values using the same logic as update_batch
+ update_batch_internal(self, states)
+ }
+
+ fn evaluate(&mut self) -> Result<ScalarValue> {
+ evaluate_internal(self)
+ }
+
+ fn size(&self) -> usize {
+ size_of_val(self)
+ }
+}
+
+// Specialized implementations for update_batch for each type
+
+fn update_batch_internal<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ values: &[ArrayRef],
+) -> Result<()> {
+ if values.is_empty() || acc.failed {
+ return Ok(());
+ }
+
+ let array = values[0].as_primitive::<T>();
+
+ // Specialize based on the type
+ if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Int64Type>() {
+ update_int64(acc, array)
+ } else if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Float64Type>() {
+ update_float64(acc, array)
+ } else if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Decimal128Type>() {
+ update_decimal128(acc, array)
+ } else {
+ exec_err!(
+ "try_sum: unsupported type in update_batch: {:?}",
+ acc.data_type
+ )
+ }
+}
+
+fn update_int64<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ array: &PrimitiveArray<T>,
+) -> Result<()> {
+ for v in array.iter().flatten() {
+ // Cast to i64 for checked_add
+ let v_i64 = unsafe { std::mem::transmute_copy::<T::Native, i64>(&v) };
+ let sum_i64 = acc
+ .sum
+ .map(|s| unsafe { std::mem::transmute_copy::<T::Native, i64>(&s) });
+
+ let new_sum = match sum_i64 {
+ None => v_i64,
+ Some(s) => match s.checked_add(v_i64) {
+ Some(result) => result,
+ None => {
+ acc.failed = true;
+ return Ok(());
+ }
+ },
+ };
+
+ acc.sum = Some(unsafe { std::mem::transmute_copy::<i64, T::Native>(&new_sum) });
+ }
+ Ok(())
+}
+
+fn update_float64<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ array: &PrimitiveArray<T>,
+) -> Result<()> {
+ for v in array.iter().flatten() {
+ let v_f64 = unsafe { std::mem::transmute_copy::<T::Native, f64>(&v) };
+ let sum_f64 = acc
+ .sum
+ .map(|s| unsafe { std::mem::transmute_copy::<T::Native, f64>(&s) })
+ .unwrap_or(0.0);
+ let new_sum = sum_f64 + v_f64;
+ acc.sum = Some(unsafe { std::mem::transmute_copy::<f64, T::Native>(&new_sum) });
+ }
+ Ok(())
+}
+
+fn update_decimal128<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ array: &PrimitiveArray<T>,
+) -> Result<()> {
+ let precision = acc.dec_precision.unwrap_or(38);
+
+ for v in array.iter().flatten() {
+ let v_i128 = unsafe { std::mem::transmute_copy::<T::Native, i128>(&v) };
+ let sum_i128 = acc
+ .sum
+ .map(|s| unsafe { std::mem::transmute_copy::<T::Native, i128>(&s) });
+
+ let new_sum = match sum_i128 {
+ None => v_i128,
+ Some(s) => match s.checked_add(v_i128) {
+ Some(result) => result,
+ None => {
+ acc.failed = true;
+ return Ok(());
+ }
+ },
+ };
+
+ if exceeds_decimal128_precision(new_sum, precision) {
+ acc.failed = true;
+ return Ok(());
+ }
+
+ acc.sum = Some(unsafe { std::mem::transmute_copy::<i128, T::Native>(&new_sum) });
+ }
+ Ok(())
+}
+
+fn evaluate_internal<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+) -> Result<ScalarValue> {
+ if acc.failed {
+ return ScalarValue::new_primitive::<T>(None, &acc.data_type);
+ }
+ ScalarValue::new_primitive::<T>(acc.sum, &acc.data_type)
+}
+
+// Helpers to determine if it exceeds decimal precision
+fn pow10_i128(p: u8) -> Option<i128> {
+ let mut v: i128 = 1;
+ for _ in 0..p {
+ v = v.checked_mul(10)?;
+ }
+ Some(v)
+}
+
+fn exceeds_decimal128_precision(sum: i128, p: u8) -> bool {
+ if let Some(max_plus_one) = pow10_i128(p) {
+ let max = max_plus_one - 1;
+ sum > max || sum < -max
+ } else {
+ true
+ }
+}
+
+impl AggregateUDFImpl for SparkTrySum {
+ fn as_any(&self) -> &dyn Any {
+ self
+ }
+
+ fn name(&self) -> &str {
+ "try_sum"
+ }
+
+ fn signature(&self) -> &Signature {
+ &self.signature
+ }
+
+ fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
+ use DataType::*;
+
+ let dt = &arg_types[0];
+ let result_type = match dt {
+ Null => Float64,
+ Decimal128(p, s) => {
+ let new_precision = DECIMAL128_MAX_PRECISION.min(p + 10);
+ Decimal128(new_precision, *s)
+ }
+ Int8 | Int16 | Int32 | Int64 | UInt8 | UInt16 | UInt32 | UInt64 => Int64,
+ Float16 | Float32 | Float64 => Float64,
+
+ other => return exec_err!("try_sum: unsupported type: {other:?}"),
+ };
+
+ Ok(result_type)
+ }
+
+ fn accumulator(&self, acc_args: AccumulatorArgs) -> Result<Box<dyn Accumulator>> {
+ macro_rules! helper {
+ ($t:ty, $dt:expr) => {
+ Ok(Box::new(TrySumAccumulator::<$t>::new($dt.clone())))
+ };
+ }
+
+ match acc_args.return_field.data_type() {
+ DataType::Int64 => helper!(Int64Type, acc_args.return_field.data_type()),
+ DataType::UInt64 => helper!(UInt64Type, acc_args.return_field.data_type()),
Review Comment:
Spark doesn't use UInt if i'm not mistaken
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2627896378
##########
datafusion/spark/src/function/aggregate/try_sum.rs:
##########
@@ -0,0 +1,650 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use arrow::array::{ArrayRef, ArrowNumericType, AsArray, BooleanArray, PrimitiveArray};
+use arrow::datatypes::{
+ DECIMAL128_MAX_PRECISION, DataType, Decimal128Type, Field, FieldRef, Float64Type,
+ Int64Type, UInt64Type,
+};
+use datafusion_common::{Result, ScalarValue, downcast_value, exec_err, not_impl_err};
+use datafusion_expr::function::{AccumulatorArgs, StateFieldsArgs};
+use datafusion_expr::utils::format_state_name;
+use datafusion_expr::{Accumulator, AggregateUDFImpl, Signature, Volatility};
+use std::any::Any;
+use std::fmt::{Debug, Formatter};
+use std::mem::size_of_val;
+
+#[derive(PartialEq, Eq, Hash)]
+pub struct SparkTrySum {
+ signature: Signature,
+}
+
+impl Default for SparkTrySum {
+ fn default() -> Self {
+ Self::new()
+ }
+}
+
+impl SparkTrySum {
+ pub fn new() -> Self {
+ Self {
+ signature: Signature::user_defined(Volatility::Immutable),
+ }
+ }
+}
+
+impl Debug for SparkTrySum {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ f.debug_struct("SparkTrySum")
+ .field("signature", &self.signature)
+ .finish()
+ }
+}
+
+/// Accumulator for try_sum that detects overflow
+struct TrySumAccumulator<T: ArrowNumericType> {
+ sum: Option<T::Native>,
+ data_type: DataType,
+ failed: bool,
+ // Only used if data_type is Decimal128(p, s)
+ dec_precision: Option<u8>,
+}
+
+impl<T: ArrowNumericType> Debug for TrySumAccumulator<T> {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ write!(f, "TrySumAccumulator({})", self.data_type)
+ }
+}
+
+impl<T: ArrowNumericType> TrySumAccumulator<T> {
+ fn new(data_type: DataType) -> Self {
+ let dec_precision = match &data_type {
+ DataType::Decimal128(p, _) => Some(*p),
+ _ => None,
+ };
+ Self {
+ sum: None,
+ data_type,
+ failed: false,
+ dec_precision,
+ }
+ }
+}
+
+impl<T: ArrowNumericType> Accumulator for TrySumAccumulator<T> {
+ fn state(&mut self) -> Result<Vec<ScalarValue>> {
+ Ok(vec![
+ self.evaluate()?,
+ ScalarValue::Boolean(Some(self.failed)),
+ ])
+ }
+
+ fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+ update_batch_internal(self, values)
+ }
+
+ fn merge_batch(&mut self, states: &[ArrayRef]) -> Result<()> {
+ // Check if any partition has failed
+ let failed_arr = downcast_value!(states[1], BooleanArray);
+ for failed in failed_arr.iter().flatten() {
+ if failed {
+ self.failed = true;
+ return Ok(());
+ }
+ }
+
+ // Merge the sum values using the same logic as update_batch
+ update_batch_internal(self, states)
+ }
+
+ fn evaluate(&mut self) -> Result<ScalarValue> {
+ evaluate_internal(self)
+ }
+
+ fn size(&self) -> usize {
+ size_of_val(self)
+ }
+}
+
+// Specialized implementations for update_batch for each type
+
+fn update_batch_internal<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ values: &[ArrayRef],
+) -> Result<()> {
+ if values.is_empty() || acc.failed {
+ return Ok(());
+ }
+
+ let array = values[0].as_primitive::<T>();
+
+ // Specialize based on the type
+ if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Int64Type>() {
Review Comment:
wondering can we rewrite into match expression
```
match T::DATA_TYPE {
Int64Type => {}
....
}
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "davidlghellin (via GitHub)" <gi...@apache.org>.
davidlghellin commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2628278058
##########
datafusion/sqllogictest/test_files/spark/aggregate/try_sum.slt:
##########
@@ -0,0 +1,140 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+
+# http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (1), (2), (3) AS tab(x);
+----
+6
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (NULL), (2), (NULL) AS tab(x);
+----
+2
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST(9223372036854775807 AS BIGINT)), (1) AS tab(x);
+----
+NULL
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (1.5), (2.5), (3.0) AS tab(x);
+----
+7
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (1e308), (1e308) AS tab(x);
+----
+Infinity
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST('NaN' AS DOUBLE)), (1.0) AS tab(x);
+----
+NaN
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST('Infinity' AS DOUBLE)), (1.0) AS tab(x);
+----
+Infinity
+
+# Decimal
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(10,2) '1.23'), (DECIMAL(10,2) '4.77') AS tab(x);
+----
+6
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(10,2) '1.00'), (NULL), (DECIMAL(10,2) '2.50') AS tab(x);
+----
+3.5
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(5,0) '90000'), (DECIMAL(5,0) '20000') AS tab(x);
+----
+110000
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111') AS tab(x);
+----
+NULL
+
+#Group By
+query TI
+SELECT g, try_sum(x) AS sum_x
+FROM VALUES
+ ('bad', CAST(9223372036854775807 AS BIGINT)),
+ ('bad', 1),
+ ('ok', 10),
+ ('ok', NULL),
+ ('ok', 5) AS tab(g, x)
+GROUP BY g
+ORDER BY g;
+----
+bad NULL
+ok 15
+
+query R
+SELECT try_sum(col) FROM VALUES (NULL), (NULL) AS tab(col);
+----
+NULL
+
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('-Infinity' AS DOUBLE)), (CAST('Infinity' AS DOUBLE)) AS tab(col);
+----
+NaN
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('-Infinity' AS DOUBLE)), (CAST('-Infinity' AS DOUBLE)) AS tab(col);
+----
+-Infinity
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('Infinity' AS FLOAT)), (CAST(1.0 AS FLOAT)) AS tab(col);
+----
+Infinity
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (-0.0), (0.0) AS tab(col);
+----
+0
+
+# need be 0.0
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST(-0.0 AS DOUBLE)), (CAST(0.0 AS DOUBLE)) AS tab(col);
+----
+0
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST(-5.5 AS DECIMAL(10,2))), (CAST(5.5 AS DECIMAL(10,2))) AS tab(col);
+----
+0
+
+# Compare double 0.0 vs decimal 0.00
+query RR
+SELECT 0.0 AS double_zero, CAST(0.0 AS DECIMAL(10,2)) AS decimal_zero;
+----
+0 0
Review Comment:
Maybe it´s not correct the display,
<img width="1047" height="197" alt="image" src="https://github.com/user-attachments/assets/7a157ab9-61f7-40e3-8056-bf0e6dd7566b" />
the last test can see the same error, in spark:
<img width="488" height="122" alt="image" src="https://github.com/user-attachments/assets/91e8508e-914c-4c2e-957b-45516c22f3b0" />
@Jefffrey @comphead I remember this https://github.com/apache/datafusion/issues/17455#issuecomment-3301244069
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "davidlghellin (via GitHub)" <gi...@apache.org>.
davidlghellin commented on PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#issuecomment-3513351511
> Is there a way we can leverage the existing sum code in DataFusion to achieve this, instead of adding an entirely new implementation?
>
> https://github.com/apache/datafusion/blob/f162fd325565e14be8e4cace17d8a3a8b2764cc8/datafusion/functions-aggregate/src/sum.rs
Thanks, I try
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "coderfender (via GitHub)" <gi...@apache.org>.
coderfender commented on PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#issuecomment-3666176298
@davidlghellin , I have already implemented spark compatible arithmetic functions supporting all three eval modes in data fusion-comet .
cc : @comphead
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead merged PR #18569:
URL: https://github.com/apache/datafusion/pull/18569
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "davidlghellin (via GitHub)" <gi...@apache.org>.
davidlghellin commented on PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#issuecomment-3666216975
> @davidlghellin Ya he implementado funciones aritméticas compatibles con Spark que admiten los tres modos de evaluación en data fusion-comet.
> cc: @comphead
Thanks!
I’m currently migrating the functions I previously sent to Sail into datafusion
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2627869982
##########
datafusion/sqllogictest/test_files/spark/aggregate/try_sum.slt:
##########
@@ -0,0 +1,97 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+
+# http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
Review Comment:
Thanks @davidlghellin would you mind adding more tests
```
> SELECT try_sum(col) FROM VALUES (NULL), (NULL) AS tab(col);
NULL
```
And also for doubles it might be good to test `-0, +0, -Inf, +Inf`, `NaN` is fine, I can see the test.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2627873909
##########
datafusion/spark/src/function/aggregate/try_sum.rs:
##########
@@ -0,0 +1,650 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use arrow::array::{ArrayRef, ArrowNumericType, AsArray, BooleanArray, PrimitiveArray};
+use arrow::datatypes::{
+ DECIMAL128_MAX_PRECISION, DataType, Decimal128Type, Field, FieldRef, Float64Type,
+ Int64Type, UInt64Type,
+};
+use datafusion_common::{Result, ScalarValue, downcast_value, exec_err, not_impl_err};
+use datafusion_expr::function::{AccumulatorArgs, StateFieldsArgs};
+use datafusion_expr::utils::format_state_name;
+use datafusion_expr::{Accumulator, AggregateUDFImpl, Signature, Volatility};
+use std::any::Any;
+use std::fmt::{Debug, Formatter};
+use std::mem::size_of_val;
+
+#[derive(PartialEq, Eq, Hash)]
+pub struct SparkTrySum {
+ signature: Signature,
+}
+
+impl Default for SparkTrySum {
+ fn default() -> Self {
+ Self::new()
+ }
+}
+
+impl SparkTrySum {
+ pub fn new() -> Self {
+ Self {
+ signature: Signature::user_defined(Volatility::Immutable),
+ }
+ }
+}
+
+impl Debug for SparkTrySum {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ f.debug_struct("SparkTrySum")
+ .field("signature", &self.signature)
+ .finish()
+ }
+}
+
+/// Accumulator for try_sum that detects overflow
+struct TrySumAccumulator<T: ArrowNumericType> {
+ sum: Option<T::Native>,
+ data_type: DataType,
+ failed: bool,
+ // Only used if data_type is Decimal128(p, s)
Review Comment:
can datatype a Decimal other than `Decimal128` ?
UPD: this is Spark, I think it is only Decimal128
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "davidlghellin (via GitHub)" <gi...@apache.org>.
davidlghellin commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2628278058
##########
datafusion/sqllogictest/test_files/spark/aggregate/try_sum.slt:
##########
@@ -0,0 +1,140 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+
+# http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (1), (2), (3) AS tab(x);
+----
+6
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (NULL), (2), (NULL) AS tab(x);
+----
+2
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST(9223372036854775807 AS BIGINT)), (1) AS tab(x);
+----
+NULL
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (1.5), (2.5), (3.0) AS tab(x);
+----
+7
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (1e308), (1e308) AS tab(x);
+----
+Infinity
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST('NaN' AS DOUBLE)), (1.0) AS tab(x);
+----
+NaN
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST('Infinity' AS DOUBLE)), (1.0) AS tab(x);
+----
+Infinity
+
+# Decimal
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(10,2) '1.23'), (DECIMAL(10,2) '4.77') AS tab(x);
+----
+6
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(10,2) '1.00'), (NULL), (DECIMAL(10,2) '2.50') AS tab(x);
+----
+3.5
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(5,0) '90000'), (DECIMAL(5,0) '20000') AS tab(x);
+----
+110000
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111') AS tab(x);
+----
+NULL
+
+#Group By
+query TI
+SELECT g, try_sum(x) AS sum_x
+FROM VALUES
+ ('bad', CAST(9223372036854775807 AS BIGINT)),
+ ('bad', 1),
+ ('ok', 10),
+ ('ok', NULL),
+ ('ok', 5) AS tab(g, x)
+GROUP BY g
+ORDER BY g;
+----
+bad NULL
+ok 15
+
+query R
+SELECT try_sum(col) FROM VALUES (NULL), (NULL) AS tab(col);
+----
+NULL
+
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('-Infinity' AS DOUBLE)), (CAST('Infinity' AS DOUBLE)) AS tab(col);
+----
+NaN
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('-Infinity' AS DOUBLE)), (CAST('-Infinity' AS DOUBLE)) AS tab(col);
+----
+-Infinity
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('Infinity' AS FLOAT)), (CAST(1.0 AS FLOAT)) AS tab(col);
+----
+Infinity
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (-0.0), (0.0) AS tab(col);
+----
+0
+
+# need be 0.0
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST(-0.0 AS DOUBLE)), (CAST(0.0 AS DOUBLE)) AS tab(col);
+----
+0
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST(-5.5 AS DECIMAL(10,2))), (CAST(5.5 AS DECIMAL(10,2))) AS tab(col);
+----
+0
+
+# Compare double 0.0 vs decimal 0.00
+query RR
+SELECT 0.0 AS double_zero, CAST(0.0 AS DECIMAL(10,2)) AS decimal_zero;
+----
+0 0
Review Comment:
Maybe it´s not correct the display,
<img width="1047" height="197" alt="image" src="https://github.com/user-attachments/assets/7a157ab9-61f7-40e3-8056-bf0e6dd7566b" />
the last test can see the same error, in spark:
<img width="488" height="122" alt="image" src="https://github.com/user-attachments/assets/91e8508e-914c-4c2e-957b-45516c22f3b0" />
@Jefffrey @comphead I remember this https://github.com/apache/datafusion/issues/17455#issuecomment-3301244069
UPD: in datafusion-cli it´s ok
<img width="715" height="177" alt="image" src="https://github.com/user-attachments/assets/6aeb669e-e37e-4d82-a945-e6161a6717e9" />
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2627883916
##########
datafusion/spark/src/function/aggregate/try_sum.rs:
##########
@@ -0,0 +1,650 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use arrow::array::{ArrayRef, ArrowNumericType, AsArray, BooleanArray, PrimitiveArray};
+use arrow::datatypes::{
+ DECIMAL128_MAX_PRECISION, DataType, Decimal128Type, Field, FieldRef, Float64Type,
+ Int64Type, UInt64Type,
+};
+use datafusion_common::{Result, ScalarValue, downcast_value, exec_err, not_impl_err};
+use datafusion_expr::function::{AccumulatorArgs, StateFieldsArgs};
+use datafusion_expr::utils::format_state_name;
+use datafusion_expr::{Accumulator, AggregateUDFImpl, Signature, Volatility};
+use std::any::Any;
+use std::fmt::{Debug, Formatter};
+use std::mem::size_of_val;
+
+#[derive(PartialEq, Eq, Hash)]
+pub struct SparkTrySum {
+ signature: Signature,
+}
+
+impl Default for SparkTrySum {
+ fn default() -> Self {
+ Self::new()
+ }
+}
+
+impl SparkTrySum {
+ pub fn new() -> Self {
+ Self {
+ signature: Signature::user_defined(Volatility::Immutable),
+ }
+ }
+}
+
+impl Debug for SparkTrySum {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ f.debug_struct("SparkTrySum")
+ .field("signature", &self.signature)
+ .finish()
+ }
+}
+
+/// Accumulator for try_sum that detects overflow
+struct TrySumAccumulator<T: ArrowNumericType> {
+ sum: Option<T::Native>,
+ data_type: DataType,
+ failed: bool,
+ // Only used if data_type is Decimal128(p, s)
+ dec_precision: Option<u8>,
+}
+
+impl<T: ArrowNumericType> Debug for TrySumAccumulator<T> {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ write!(f, "TrySumAccumulator({})", self.data_type)
+ }
+}
+
+impl<T: ArrowNumericType> TrySumAccumulator<T> {
+ fn new(data_type: DataType) -> Self {
+ let dec_precision = match &data_type {
+ DataType::Decimal128(p, _) => Some(*p),
+ _ => None,
+ };
+ Self {
+ sum: None,
+ data_type,
+ failed: false,
+ dec_precision,
+ }
+ }
+}
+
+impl<T: ArrowNumericType> Accumulator for TrySumAccumulator<T> {
+ fn state(&mut self) -> Result<Vec<ScalarValue>> {
+ Ok(vec![
+ self.evaluate()?,
+ ScalarValue::Boolean(Some(self.failed)),
+ ])
+ }
+
+ fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+ update_batch_internal(self, values)
+ }
+
+ fn merge_batch(&mut self, states: &[ArrayRef]) -> Result<()> {
+ // Check if any partition has failed
+ let failed_arr = downcast_value!(states[1], BooleanArray);
Review Comment:
maybe try more idiomatic
```
if downcast_value!(states[1], BooleanArray).iter().flatten().any(|&f| f) {
self.failed = true;
return Ok(());
}
```
?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "comphead (via GitHub)" <gi...@apache.org>.
comphead commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2627927890
##########
datafusion/spark/src/function/aggregate/try_sum.rs:
##########
@@ -0,0 +1,650 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use arrow::array::{ArrayRef, ArrowNumericType, AsArray, BooleanArray, PrimitiveArray};
+use arrow::datatypes::{
+ DECIMAL128_MAX_PRECISION, DataType, Decimal128Type, Field, FieldRef, Float64Type,
+ Int64Type, UInt64Type,
+};
+use datafusion_common::{Result, ScalarValue, downcast_value, exec_err, not_impl_err};
+use datafusion_expr::function::{AccumulatorArgs, StateFieldsArgs};
+use datafusion_expr::utils::format_state_name;
+use datafusion_expr::{Accumulator, AggregateUDFImpl, Signature, Volatility};
+use std::any::Any;
+use std::fmt::{Debug, Formatter};
+use std::mem::size_of_val;
+
+#[derive(PartialEq, Eq, Hash)]
+pub struct SparkTrySum {
+ signature: Signature,
+}
+
+impl Default for SparkTrySum {
+ fn default() -> Self {
+ Self::new()
+ }
+}
+
+impl SparkTrySum {
+ pub fn new() -> Self {
+ Self {
+ signature: Signature::user_defined(Volatility::Immutable),
+ }
+ }
+}
+
+impl Debug for SparkTrySum {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ f.debug_struct("SparkTrySum")
+ .field("signature", &self.signature)
+ .finish()
+ }
+}
+
+/// Accumulator for try_sum that detects overflow
+struct TrySumAccumulator<T: ArrowNumericType> {
+ sum: Option<T::Native>,
+ data_type: DataType,
+ failed: bool,
+ // Only used if data_type is Decimal128(p, s)
+ dec_precision: Option<u8>,
+}
+
+impl<T: ArrowNumericType> Debug for TrySumAccumulator<T> {
+ fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+ write!(f, "TrySumAccumulator({})", self.data_type)
+ }
+}
+
+impl<T: ArrowNumericType> TrySumAccumulator<T> {
+ fn new(data_type: DataType) -> Self {
+ let dec_precision = match &data_type {
+ DataType::Decimal128(p, _) => Some(*p),
+ _ => None,
+ };
+ Self {
+ sum: None,
+ data_type,
+ failed: false,
+ dec_precision,
+ }
+ }
+}
+
+impl<T: ArrowNumericType> Accumulator for TrySumAccumulator<T> {
+ fn state(&mut self) -> Result<Vec<ScalarValue>> {
+ Ok(vec![
+ self.evaluate()?,
+ ScalarValue::Boolean(Some(self.failed)),
+ ])
+ }
+
+ fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+ update_batch_internal(self, values)
+ }
+
+ fn merge_batch(&mut self, states: &[ArrayRef]) -> Result<()> {
+ // Check if any partition has failed
+ let failed_arr = downcast_value!(states[1], BooleanArray);
+ for failed in failed_arr.iter().flatten() {
+ if failed {
+ self.failed = true;
+ return Ok(());
+ }
+ }
+
+ // Merge the sum values using the same logic as update_batch
+ update_batch_internal(self, states)
+ }
+
+ fn evaluate(&mut self) -> Result<ScalarValue> {
+ evaluate_internal(self)
+ }
+
+ fn size(&self) -> usize {
+ size_of_val(self)
+ }
+}
+
+// Specialized implementations for update_batch for each type
+
+fn update_batch_internal<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ values: &[ArrayRef],
+) -> Result<()> {
+ if values.is_empty() || acc.failed {
+ return Ok(());
+ }
+
+ let array = values[0].as_primitive::<T>();
+
+ // Specialize based on the type
+ if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Int64Type>() {
+ update_int64(acc, array)
+ } else if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Float64Type>() {
+ update_float64(acc, array)
+ } else if std::any::TypeId::of::<T>() == std::any::TypeId::of::<Decimal128Type>() {
+ update_decimal128(acc, array)
+ } else {
+ exec_err!(
+ "try_sum: unsupported type in update_batch: {:?}",
+ acc.data_type
+ )
+ }
+}
+
+fn update_int64<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ array: &PrimitiveArray<T>,
+) -> Result<()> {
+ for v in array.iter().flatten() {
+ // Cast to i64 for checked_add
+ let v_i64 = unsafe { std::mem::transmute_copy::<T::Native, i64>(&v) };
+ let sum_i64 = acc
+ .sum
+ .map(|s| unsafe { std::mem::transmute_copy::<T::Native, i64>(&s) });
+
+ let new_sum = match sum_i64 {
+ None => v_i64,
+ Some(s) => match s.checked_add(v_i64) {
+ Some(result) => result,
+ None => {
+ acc.failed = true;
+ return Ok(());
+ }
+ },
+ };
+
+ acc.sum = Some(unsafe { std::mem::transmute_copy::<i64, T::Native>(&new_sum) });
+ }
+ Ok(())
+}
+
+fn update_float64<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ array: &PrimitiveArray<T>,
+) -> Result<()> {
+ for v in array.iter().flatten() {
+ let v_f64 = unsafe { std::mem::transmute_copy::<T::Native, f64>(&v) };
+ let sum_f64 = acc
+ .sum
+ .map(|s| unsafe { std::mem::transmute_copy::<T::Native, f64>(&s) })
+ .unwrap_or(0.0);
+ let new_sum = sum_f64 + v_f64;
+ acc.sum = Some(unsafe { std::mem::transmute_copy::<f64, T::Native>(&new_sum) });
+ }
+ Ok(())
+}
+
+fn update_decimal128<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+ array: &PrimitiveArray<T>,
+) -> Result<()> {
+ let precision = acc.dec_precision.unwrap_or(38);
+
+ for v in array.iter().flatten() {
+ let v_i128 = unsafe { std::mem::transmute_copy::<T::Native, i128>(&v) };
+ let sum_i128 = acc
+ .sum
+ .map(|s| unsafe { std::mem::transmute_copy::<T::Native, i128>(&s) });
+
+ let new_sum = match sum_i128 {
+ None => v_i128,
+ Some(s) => match s.checked_add(v_i128) {
+ Some(result) => result,
+ None => {
+ acc.failed = true;
+ return Ok(());
+ }
+ },
+ };
+
+ if exceeds_decimal128_precision(new_sum, precision) {
+ acc.failed = true;
+ return Ok(());
+ }
+
+ acc.sum = Some(unsafe { std::mem::transmute_copy::<i128, T::Native>(&new_sum) });
+ }
+ Ok(())
+}
+
+fn evaluate_internal<T: ArrowNumericType>(
+ acc: &mut TrySumAccumulator<T>,
+) -> Result<ScalarValue> {
+ if acc.failed {
+ return ScalarValue::new_primitive::<T>(None, &acc.data_type);
+ }
+ ScalarValue::new_primitive::<T>(acc.sum, &acc.data_type)
+}
+
+// Helpers to determine if it exceeds decimal precision
+fn pow10_i128(p: u8) -> Option<i128> {
+ let mut v: i128 = 1;
+ for _ in 0..p {
+ v = v.checked_mul(10)?;
+ }
+ Some(v)
+}
+
+fn exceeds_decimal128_precision(sum: i128, p: u8) -> bool {
+ if let Some(max_plus_one) = pow10_i128(p) {
+ let max = max_plus_one - 1;
+ sum > max || sum < -max
+ } else {
+ true
+ }
+}
+
+impl AggregateUDFImpl for SparkTrySum {
+ fn as_any(&self) -> &dyn Any {
+ self
+ }
+
+ fn name(&self) -> &str {
+ "try_sum"
+ }
+
+ fn signature(&self) -> &Signature {
+ &self.signature
+ }
+
+ fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
+ use DataType::*;
+
+ let dt = &arg_types[0];
+ let result_type = match dt {
+ Null => Float64,
Review Comment:
nice, I double checked it
```
scala> spark.sql("select try_sum(null)").printSchema()
root
|-- try_sum(NULL): double (nullable = true)
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] feat(spark): implement Spark `try_sum` function [datafusion]
Posted by "davidlghellin (via GitHub)" <gi...@apache.org>.
davidlghellin commented on code in PR #18569:
URL: https://github.com/apache/datafusion/pull/18569#discussion_r2628278058
##########
datafusion/sqllogictest/test_files/spark/aggregate/try_sum.slt:
##########
@@ -0,0 +1,140 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+
+# http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (1), (2), (3) AS tab(x);
+----
+6
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (NULL), (2), (NULL) AS tab(x);
+----
+2
+
+query I
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST(9223372036854775807 AS BIGINT)), (1) AS tab(x);
+----
+NULL
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (1.5), (2.5), (3.0) AS tab(x);
+----
+7
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (1e308), (1e308) AS tab(x);
+----
+Infinity
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST('NaN' AS DOUBLE)), (1.0) AS tab(x);
+----
+NaN
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (CAST('Infinity' AS DOUBLE)), (1.0) AS tab(x);
+----
+Infinity
+
+# Decimal
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(10,2) '1.23'), (DECIMAL(10,2) '4.77') AS tab(x);
+----
+6
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(10,2) '1.00'), (NULL), (DECIMAL(10,2) '2.50') AS tab(x);
+----
+3.5
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(5,0) '90000'), (DECIMAL(5,0) '20000') AS tab(x);
+----
+110000
+
+query R
+SELECT try_sum(x) AS sum_x FROM VALUES (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111'),
+ (DECIMAL(38,0) '11111111111111111111111111111111111111') AS tab(x);
+----
+NULL
+
+#Group By
+query TI
+SELECT g, try_sum(x) AS sum_x
+FROM VALUES
+ ('bad', CAST(9223372036854775807 AS BIGINT)),
+ ('bad', 1),
+ ('ok', 10),
+ ('ok', NULL),
+ ('ok', 5) AS tab(g, x)
+GROUP BY g
+ORDER BY g;
+----
+bad NULL
+ok 15
+
+query R
+SELECT try_sum(col) FROM VALUES (NULL), (NULL) AS tab(col);
+----
+NULL
+
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('-Infinity' AS DOUBLE)), (CAST('Infinity' AS DOUBLE)) AS tab(col);
+----
+NaN
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('-Infinity' AS DOUBLE)), (CAST('-Infinity' AS DOUBLE)) AS tab(col);
+----
+-Infinity
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST('Infinity' AS FLOAT)), (CAST(1.0 AS FLOAT)) AS tab(col);
+----
+Infinity
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (-0.0), (0.0) AS tab(col);
+----
+0
+
+# need be 0.0
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST(-0.0 AS DOUBLE)), (CAST(0.0 AS DOUBLE)) AS tab(col);
+----
+0
+
+query R
+SELECT try_sum(col) AS sum_x FROM VALUES (CAST(-5.5 AS DECIMAL(10,2))), (CAST(5.5 AS DECIMAL(10,2))) AS tab(col);
+----
+0
+
+# Compare double 0.0 vs decimal 0.00
+query RR
+SELECT 0.0 AS double_zero, CAST(0.0 AS DECIMAL(10,2)) AS decimal_zero;
+----
+0 0
Review Comment:
Maybe it´s not correct the display,
<img width="1047" height="197" alt="image" src="https://github.com/user-attachments/assets/7a157ab9-61f7-40e3-8056-bf0e6dd7566b" />
the last test can see the same error, in spark:
<img width="488" height="122" alt="image" src="https://github.com/user-attachments/assets/91e8508e-914c-4c2e-957b-45516c22f3b0" />
@Jefffrey I remember this https://github.com/apache/datafusion/issues/17455#issuecomment-3301244069
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org