xumingming commented on PR #58144: URL: https://github.com/apache/spark/pull/58144#issuecomment-6172792091
Thanks for the thorough re-review, @cloud-fan! I've pushed a follow-up round addressing the findings. Summary of the changes in this round: 1. Make `InferWindowGroupLimit.support` private; fix test conf scoping 2. Keep the rank-filter-over-Window shape so `InferWindowGroupLimit` still fires under a pin 3. Rewrite the `windowReorder.enabled` conf doc in plainer terms 4. Rank required-ordering classes per group 5. Fix the benchmark doc comments 6. Fall back to adjacent-pair transposition on rejected stacks 7. Align `TransposeWindow` comments with the code 8. Prefer a pin-ridable group as the last group 9. Exercise `refsAbove` in the two-link test 10. Cut chains at derived window links 11. Peel transparent projects in the pinned merge check 12. Decide pin merge safety directly 13. Cover the pinned tail and strict specs 14. Test that the later rideable group stays last On the benchmark results: I've triggered the benchmark workflow on the fork for JDK 17, 21 and 25 with auto-commit enabled, so the locally-generated Mac numbers will be replaced by the reference-environment results and the `-jdk21`/`-jdk25` variants will be committed. -- 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]
