This is an automated email from the ASF dual-hosted git repository. github-merge-queue[bot] pushed a commit to branch gh-readonly-queue/main/pr-22755-6e61cd01fe186955ba57e36292e8166a0c633474 in repository https://gitbox.apache.org/repos/asf/datafusion.git
commit 71d22b73e8c2c2df4d1edfec3e72fdc446553785 Author: Kumar Ujjawal <[email protected]> AuthorDate: Sat Jun 6 04:53:48 2026 +0530 fix: handle NULLs in sliding SUM(DISTINCT) window frames (#22755) ## Which issue does this PR close? <!-- We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123. --> - Closes #22754 ## Rationale for this change `SUM(DISTINCT)` over bounded/sliding window frames did not handle `NULL` values correctly. It could read values from null slots in the Arrow buffer, and it returned `0` instead of `NULL` when a frame had no non-null values. <!-- Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed. Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes. --> ## What changes are included in this PR? - skip `NULL` rows in sliding `SUM(DISTINCT)` update - skip `NULL` rows in sliding `SUM(DISTINCT)` retract - return `NULL` when the frame has no non-null distinct values - reuse the same helper for distinct-value updates during merge - add regression tests for accumulator behavior with null slots - add sqllogictest coverage for sliding window SQL behavior, including when a frame becomes all `NULL` <!-- There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR. --> ## Are these changes tested? Yes <!-- We typically require tests for all PRs in order to: 1. Prevent the code from being accidentally broken by subsequent changes 2. Serve as another way to document the expected behavior of the code If tests are not included in your PR, please explain why (for example, are they covered by existing tests)? --> ## Are there any user-facing changes? `SUM(DISTINCT)` over bounded/sliding window frames now handles `NULL` values correctly: - `NULL` inputs are ignored - frames with no non-null values return `NULL` instead of `0` - No API Change <!-- If there are user-facing changes then we may require documentation to be updated before approving the PR. --> <!-- If there are any breaking changes to public APIs, please add the `api change` label. --> --- datafusion/functions-aggregate/src/sum.rs | 120 ++++++++++++++++++++------ datafusion/sqllogictest/test_files/window.slt | 22 +++++ 2 files changed, 118 insertions(+), 24 deletions(-) diff --git a/datafusion/functions-aggregate/src/sum.rs b/datafusion/functions-aggregate/src/sum.rs index 81efea1df2..c3c2e5e0b9 100644 --- a/datafusion/functions-aggregate/src/sum.rs +++ b/datafusion/functions-aggregate/src/sum.rs @@ -533,25 +533,60 @@ impl SlidingDistinctSumAccumulator { data_type: data_type.clone(), }) } + + fn update_value(&mut self, value: i64) { + let cnt = self.counts.entry(value).or_insert(0); + if *cnt == 0 { + // first occurrence in window + self.sum = self.sum.wrapping_add(value); + } + *cnt += 1; + } + + fn retract_value(&mut self, value: i64) { + if let Some(cnt) = self.counts.get_mut(&value) { + *cnt -= 1; + if *cnt == 0 { + // last copy leaving window + self.sum = self.sum.wrapping_sub(value); + self.counts.remove(&value); + } + } + } + + fn apply_valid_values<F>( + &mut self, + arr: &arrow::array::PrimitiveArray<Int64Type>, + mut op: F, + ) where + F: FnMut(&mut Self, i64), + { + if arr.null_count() == 0 { + for &value in arr.values() { + op(self, value); + } + } else { + for (idx, &value) in arr.values().iter().enumerate() { + if arr.is_valid(idx) { + op(self, value); + } + } + } + } } impl Accumulator for SlidingDistinctSumAccumulator { fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> { let arr = values[0].as_primitive::<Int64Type>(); - for &v in arr.values() { - let cnt = self.counts.entry(v).or_insert(0); - if *cnt == 0 { - // first occurrence in window - self.sum = self.sum.wrapping_add(v); - } - *cnt += 1; - } + self.apply_valid_values(arr, Self::update_value); Ok(()) } fn evaluate(&mut self) -> Result<ScalarValue> { // O(1) wrap of running sum - Ok(ScalarValue::Int64(Some(self.sum))) + Ok(ScalarValue::Int64( + (!self.counts.is_empty()).then_some(self.sum), + )) } fn size(&self) -> usize { @@ -581,11 +616,7 @@ impl Accumulator for SlidingDistinctSumAccumulator { if let ScalarValue::Int64(Some(v)) = ScalarValue::try_from_array(&*maybe_inner, idx)? { - let cnt = self.counts.entry(v).or_insert(0); - if *cnt == 0 { - self.sum = self.sum.wrapping_add(v); - } - *cnt += 1; + self.update_value(v); } } } @@ -594,16 +625,7 @@ impl Accumulator for SlidingDistinctSumAccumulator { fn retract_batch(&mut self, values: &[ArrayRef]) -> Result<()> { let arr = values[0].as_primitive::<Int64Type>(); - for &v in arr.values() { - if let Some(cnt) = self.counts.get_mut(&v) { - *cnt -= 1; - if *cnt == 0 { - // last copy leaving window - self.sum = self.sum.wrapping_sub(v); - self.counts.remove(&v); - } - } - } + self.apply_valid_values(arr, Self::retract_value); Ok(()) } @@ -611,3 +633,53 @@ impl Accumulator for SlidingDistinctSumAccumulator { true } } + +#[cfg(test)] +mod tests { + use super::*; + use arrow::{ + array::Int64Array, + buffer::{NullBuffer, ScalarBuffer}, + }; + use std::sync::Arc; + + #[test] + fn sliding_distinct_sum_ignores_null_slots() -> Result<()> { + let mut acc = SlidingDistinctSumAccumulator::try_new(&DataType::Int64)?; + + let values: ArrayRef = Arc::new(Int64Array::new( + ScalarBuffer::from(vec![42, 5, 5]), + Some(NullBuffer::from(vec![false, true, true])), + )); + acc.update_batch(&[values])?; + assert_eq!(acc.evaluate()?, ScalarValue::Int64(Some(5))); + + let retract: ArrayRef = Arc::new(Int64Array::new( + ScalarBuffer::from(vec![42, 5]), + Some(NullBuffer::from(vec![false, true])), + )); + acc.retract_batch(&[retract])?; + assert_eq!(acc.evaluate()?, ScalarValue::Int64(Some(5))); + + let retract_last: ArrayRef = + Arc::new(Int64Array::new(ScalarBuffer::from(vec![5]), None)); + acc.retract_batch(&[retract_last])?; + assert_eq!(acc.evaluate()?, ScalarValue::Int64(None)); + + Ok(()) + } + + #[test] + fn sliding_distinct_sum_returns_null_for_all_null_frame() -> Result<()> { + let mut acc = SlidingDistinctSumAccumulator::try_new(&DataType::Int64)?; + + let values: ArrayRef = Arc::new(Int64Array::new( + ScalarBuffer::from(vec![99]), + Some(NullBuffer::from(vec![false])), + )); + acc.update_batch(&[values])?; + assert_eq!(acc.evaluate()?, ScalarValue::Int64(None)); + + Ok(()) + } +} diff --git a/datafusion/sqllogictest/test_files/window.slt b/datafusion/sqllogictest/test_files/window.slt index bc2f1bfcbc..1b51950a70 100644 --- a/datafusion/sqllogictest/test_files/window.slt +++ b/datafusion/sqllogictest/test_files/window.slt @@ -5959,6 +5959,28 @@ physical_plan 07)------------DataSourceExec: partitions=2, partition_sizes=[5, 4] +# SUM(DISTINCT) over sliding frames must skip NULLs and return NULL +# for frames containing no non-null values. +statement ok +CREATE TABLE table_distinct_sum_nulls(ts INT, v BIGINT) AS VALUES + (1, NULL), (2, 3), (3, NULL), (4, NULL), (5, 5); + +query II +SELECT + ts, + SUM(DISTINCT v) OVER ( + ORDER BY ts + ROWS BETWEEN 1 PRECEDING AND CURRENT ROW + ) AS s +FROM table_distinct_sum_nulls; +---- +1 NULL +2 3 +3 3 +4 NULL +5 5 + + # FILTER clause with window functions # Verify FILTER clause with non-aggregate window functions fails with a clear message --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
