JustinBinber commented on PR #11285: URL: https://github.com/apache/paimon/pull/11285#issuecomment-7030478692
Reproduced and addressed both P1 reports in `8a51ed6ac5`. * DIFF now intersects its selected file ranges with the active RowRangeIndex and suppresses empty intersections. Both SnapshotReader range setters share state, including replacing ranges, empty selections and null resets. Real-table regressions cover strict append/partial-update subsets, DVs and a dedicated BLOB subrange; a deleted requested row cannot expose an unrequested live row. * Incremental reads bypass automatic global-index lookup until the index, coverage and data plan can share a resolved historical endpoint. This covers eager/deferred index modes and tag ranges. The regression returns `(200,20)` for historical end snapshot 2 after the current value became `(200,30)`, while the latest read still uses its index and returns `(200,30)`. * Explicit external index results have selection/scoring semantics and no endpoint provenance, so incremental reads reject them clearly rather than ignoring them. Explicit RowID selection remains supported through withRowRanges; latest external selections remain unchanged. The normal Maven Core reactor passes all 32 focused tests; six targeted Spark cases also pass with the current Core classes. Fresh upstream CI is automatic for this new head and is still pending; the earlier green run belongs to adad80e295. -- 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]
