dwangatt opened a new pull request, #10052:
URL: https://github.com/apache/paimon/pull/10052
## Purpose
Extracted from #9370 as a prerequisite for supporting different bucket
counts across partitions.
A partition's current bucket layout must be derived from the files that are
active in the current snapshot. It must not be inferred from data-file creation
timestamps, because deleted files from a later rescale can otherwise override
the layout restored by a rollback.
## Changes
- Update `AbstractFileStoreScan#readPartitionEntries` to merge manifest
entries first and discard deleted files before aggregating partition entries.
- Preserve the existing generic `PartitionEntry` aggregation semantics;
layout resolution is handled at the snapshot scan boundary where file liveness
is known.
- Add a regression test covering:
1. a partition written with 2 buckets;
2. overwrite with 4 buckets;
3. rollback to the first snapshot;
4. verification that the visible partition layout returns to 2 buckets.
## Testing
-
`FileStoreScanPartitionBucketEntryTest#testReadPartitionEntriesRestoresBucketLayoutAfterRollback`
## Related
- Extracted from #9370
- Prerequisite for the follow-up layout-aware `BatchWriteBuilder` overwrite
routing work.
--
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]
