JingsongLi commented on PR #10143: URL: https://github.com/apache/paimon/pull/10143#issuecomment-5805700830
I reviewed the selected-key reconstruction path and its fallback against the existing semantics. The Arrow path preserves candidate ordering, parent nulls, sliced mapping offsets, physical null values, and overflow lookup; the added candidate bounds check addresses the earlier review finding. This has clear end-to-end value: the PR reports 65–80% shorter reads on large selected-key workloads while keeping fallback cases close to baseline. Local verification on the PR patch: 30 related Python tests passed, with 39 subtests; 6 ORC cases were excluded because this macOS host’s PyArrow ORC path fails while querying `hw.l1dcachesize`. I did not reproduce the full published benchmark or the author’s PyArrow 6/19 matrix here. No blocking correctness issue found in the reviewed path. Please keep the full CI and benchmark artifacts green before merge. -- 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]
