suxiaogang223 commented on PR #65334: URL: https://github.com/apache/doris/pull/65334#issuecomment-4931317771
Re-reviewed latest head `d2288a9da66efb6509551a4a6ba6b6745c581da1`. No new blocking issue found beyond the existing table-id-scoped fallback discussion. The patch fixes the partition-values-cache eviction gap by deriving values directly from `HiveUtil.toPartitionValues(partitionName)`, and the changed paths are limited to Hive partition/file cache invalidation plus the new unit test. Remaining concern/tradeoff: in the partition-cache-miss branch, the fallback invalidates only `k.isSameTable(tableId) && Objects.equals(k.getPartitionValues(), values)`. Ordinary `FileCacheKey` identity ignores `id` and reuses entries by `catalogId + location + inputFormat + partitionValues`, so if another table with the same physical partition location/input format/values populated the entry first, a later scan of this refreshed table can still hit that stale listing. This is the same residual case discussed in the existing review thread; please either make that an explicit accepted tradeoff, or switch to a more conservative fallback. If the current behavior is kept, a shared-key test would make the intended boundary clear. Validation: static diff/surrounding-code review completed; `git diff --check origin/master...origin/pr/65334` passed. I did not run FE UT locally. -- 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]
