wirybeaver opened a new issue, #2700: URL: https://github.com/apache/datafusion-ballista/issues/2700
Enable opt-in cross-stage runtime filtering in Ballista AQE so probe tasks can avoid non-matching rows and Parquet row groups. - [ ] **Task 1 — Filter operators and task reports:** Add bounded Bloom/IN/MinMax summaries, build/apply operators, plan serialization, and executor reporting. - [ ] **Task 2 — AQE coordination:** Place filters on eligible joins, merge complete build-side reports, and dispatch probe stages with filters or timeout fallback. [Personal stacked PR](https://github.com/wirybeaver/datafusion-ballista/pull/11). - [ ] **Task 3 — Parquet pruning:** Push merged MinMax predicates into eligible scans and verify row-group pruning. [Personal stacked PR](https://github.com/wirybeaver/datafusion-ballista/pull/3). Implementation order: **1 → 2 → 3**. Tasks 2 and 3 remain stacked in the personal fork for review until their dependencies land upstream. Check off each task when its upstream PR is merged. -- 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]
