DanielLeens commented on PR #11516: URL: https://github.com/apache/seatunnel/pull/11516#issuecomment-6210066473
Thanks @SEZ9 for the recap, and thanks @fatmanverse for your patience. I re-checked head `10e70d2257a05acb289cf863ea3699aa413ba93a`: there is no new commit and the PR description still reads as Spark-only, so SEZ9's four follow-ups are all still open, and my view from the last round is unchanged. There is no source-level blocker from my side, and the apache-side `Build` check is green. To make the next push cheap, here is the shortest path I would suggest for each item: - Description: add one short paragraph saying that Hive multi-table Parquet rows on Zeta and Flink now carry the split's real `tableId` instead of `null` (`MultipleTableHiveSourceReader` passes `split.getTableId()` into `ParquetReadStrategy#read`). This is the user-visible behavior change reviewers need to see. - Guard in `InternalRowConverter.parcel`: since the root cause in `ParquetReadStrategy` is fixed, I would drop the guard, which also settles the question of applying it in `SeaTunnelRowConverter` and `MultiTableManager`. If you prefer to keep it as a defensive layer, then please reword the comment as SEZ9 suggested. - Tests: assert the emitted `tableId` in `testParquetRead1` and add one multi-table Parquet case asserting the id. These are the assertions that would turn red if the one-line fix in `ParquetReadStrategy` were reverted. None of this blocks merge from my side; the remaining gate is maintainer approval. I will re-check as soon as a new commit lands. -- 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]
