JingsongLi opened a new pull request, #10682: URL: https://github.com/apache/paimon/pull/10682
Data Evolution clustering changes RowIDs even when indexed values remain unchanged. Flink clustering now carries old RowIDs through sorting and rewrites B-tree and bitmap postings after the committer assigns final RowIDs, reusing stored keys instead of rebuilding from table data. Index coverage follows the exact sorted rows: unaffected ranges and source baselines survive, including other ranges in the same partition. Concurrent index creation and optimistic retries re-read the latest index state and rebind the mapping to the final RowIDs. Stale indexed values and unsupported affected index types are invalidated; Spark retains its existing invalidation behavior. The contiguous-range protocol bounds coverage fragmentation to 64 fragments and rejects an over-bound rewrite before commit. Validation: core index/RowID/DV/concurrency regressions and Flink real-index global/local sort cases passed locally, with normal static checks enabled. Tests include partial/interleaved coverage, null and negative predicates, same-partition retained ranges, index creation during retry, reverse commit order and failed-output cleanup. Independent review completed. Split from #10667 after #10681. -- 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]
