Phoenix500526 opened a new pull request, #27076: URL: https://github.com/apache/datafusion/pull/27076
## Which issue does this PR close? Closes #27074. ## Rationale for this change With two rows in the same group, applying `WHERE random() < 0.5` to the result of `COUNT(*)` should return either `2` or no rows. Pushing the filter below the aggregate evaluates it for each input row, so the count can incorrectly be `1`. ## What changes are included in this PR? `PushDownFilter` keeps volatile predicates above an aggregate. It can still push a separate predicate on a grouping key below the aggregate, even when the two predicates appear in the same `WHERE` clause. ## What is the testing strategy for this PR? The regression test in `push_down_filter_regression.slt` checks the logical plan for `column1 = 1 AND random() < 0.5`. It verifies that the grouping-key filter moves below the aggregate and the random filter stays above it. This reproduces the bug without relying on a particular random draw. ## Are there any user-facing changes? Yes. A volatile filter on grouped results now decides which groups to return while preserving each group's count. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
