thswlsqls opened a new pull request, #10562: URL: https://github.com/apache/paimon/pull/10562
### Purpose fix #10559 - `metric()` returns the index meta's persisted metric after `createReader`, so raw fallback and refinement match the index, like eslib (#8618) and pypaimon (#9754). - Conflicting shards or unparsable meta keep today's configured metric; can throw like eslib if preferred. - Not covered: index splits evaluated on remote engine tasks (Flink `vector_search` when splits ≥ 2 × min(parallelism, `global-index.thread-num`), Spark when ≥ 2 × `global-index.thread-num`), where the coordinator still uses the configured metric, and raw-only searches. Those need a core change, left for a follow-up. ### Tests - `LuminaVectorGlobalIndexerFactoryTest` (6/6 pass): persisted metric (red before fix), no reader, conflicting shards, corrupt meta. - `mvn -pl paimon-lumina clean verify` (JDK 11): 49 run, 0 failures; 27 native-library tests were skipped locally (arm64), so CI covers them. -- 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]
