srielau commented on PR #58317: URL: https://github.com/apache/spark/pull/58317#issuecomment-5873902929
Scope freeze and validation update (bc6b789fe8b): I documented the cache lifecycle contract in `CacheManager` and mapped it to the existing focused tests: - Capture: first-class CHAR/VARCHAR relations bind `PreserveNative` or `SparkStandard`; the generated read-side `Project` is part of that policy. - Identity: ordinary cache substitution remains mode-sensitive; non-CHAR relations do not split. - Mutation matching: write/refresh/rename discovery ignores only scan mode, using `BaseRelation`, catalog+identifier, or catalog-less name+path identity as appropriate. Rename restores direct caches only. - Replay: cache rebuild and rename restoration use an isolated session configured for the captured mode; direct V2 recache rebuilds relation output from refreshed catalog metadata before padding is re-analyzed. I ran the focused cache lifecycle cross-product locally: V1/V2 mutation, catalog-less batch and micro-batch paths, V1 refresh, V2 rename/direct/dependent/time-travel behavior, temp-view rebinding, Hive converted/native ORC cache identity, and the SPARK-54022/SPARK-54424 schema-refresh suites. All 26 tests passed. This freezes the scope of SPARK-58814 around format round-trips and the two bound scan modes. Explicit Legacy identity and any newly discovered sibling cache-lifecycle expansion remain deferred to [SPARK-59751](https://issues.apache.org/jira/browse/SPARK-59751). -- 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]
