sarutak commented on PR #55912: URL: https://github.com/apache/spark/pull/55912#issuecomment-4594268126
@cloud-fan Thank you for the feedback. > Two caveats: (a) a shared scanner base gives robustness/maintenance parity but not whole-stage codegen — that lives in the operator's doProduce and stays future work; (b) the Backward "reverse scan needs random access" concern (the cited blocker to spill) is avoidable — since the group is sorted ascending and the as-of predicate is monotone, a forward scan keeping the last as-of-satisfying row finds the same nearest match, so ExternalAppendOnlyUnsafeRowArray's forward-only generateIterator(startIndex) suffices. Non-blocking — an architecture direction; the null-key fix below should land regardless of whether this refactor happens here or in a follow-up. Let me do them in follow-up PRs. -- 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]
