SEZ9 commented on PR #11932: URL: https://github.com/apache/seatunnel/pull/11932#issuecomment-6407268958
Thanks for the update at f2ef35145d8fbaed3d73c7c384b5793dcd6b4cd9. The description of the changes lines up with the earlier findings, but I'd like to confirm against the code before closing them out. Could you point me to the relevant hunks/tests for each: 1. **Schema-restore guard** — the switch to selecting only non-empty checkpoint/legacy tables (so a non-null-but-empty list no longer rolls the deserializer/collector back to an empty schema). 2. **Deserializer vs. collector restore** — both restores now gated behind the same resolver guard before `super.pollNext`. 3. **`restoreWriter` schema selection** — how "one consistent schema" is chosen when restored states diverge (and the test covering it). 4. **Null-Xid states** — where null Xids are filtered on restore and where `checkpointEnabled` is kept independent. 5. **Non-exactly-once writer schema state** — emitting schema state only after a successful evolution narrows the exposure; is the non-Serializable column payload case handled or tested? 6. **Primary-key lookup on restored schema** — I don't see this covered yet. How is a restored `TableSchema` without a `PrimaryKey` handled now (fall back to the configured/catalog PK, or fail loudly)? A test for this would be great. 7. **Checkpoint state format docs** — please confirm the docs now describe the changed JDBC sink state (schema persisted, non-exactly-once writer now emitting state) and any upgrade considerations. 8. **Flink/Spark `Collector.restoreSchema` no-op** — since this is a bounded follow-up, please note in the PR description (and ideally a code comment at the call site) that the connector-level restore is Zeta-only today, and link whatever tracks the follow-up. Once I can see those, I'm happy to do a final pass. <!-- streview-comment:2042 --> -- 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]
