DanielLeens commented on PR #11077:
URL: https://github.com/apache/seatunnel/pull/11077#issuecomment-6730675267

   Thanks @hesam-oxe, and no rush at all. One correction to the plan before you 
spend days on it, because it would duplicate work that is already on this head 
and miss the two items that are actually blocking.
   
   **Please do not re-implement these; they already exist on `8b383c04ee`:**
   - Writer creation outside `computeIfAbsent`, with `IOException` preserved: 
`MultiTableSink.java:224-229`.
   - A context registered for every alias, with a direct lookup: 
`MultiTableSink.java:248-251`.
   - Alias row-layout compatibility check that fails fast: 
`validateSharedDestinationSchemas()`, `MultiTableSink.java:142-174`.
   - `MultiTableSinkWriter` already groups shared writer instances by identity 
for snapshot and close, and merges legacy alias records on restore.
   
   If you believe a gap still remains in destination-key collisions (F1) or 
aliased state on restore (F2), the cheapest way to settle it is a focused test 
that fails on the current code, not a rewrite. Please keep the existing 
coverage.
   
   **What is still blocking from my head review (please make sure these are in 
the push):**
   1. Issue 2: fix the EN and ZH `incompatible-changes.md` entries and the 
release note, which promise a startup `IllegalStateException` that cannot 
happen for the file sink. Describe the actual behavior.
   2. Issue 3: tests that pin the new destination-identifier components (next 
to `HdfsFileSinkTest`), plus a sharing test with metadata-only schema 
differences (in `MultiTableSinkWriterTest`).
   3. Resolve the conflicts against `dev` keeping both sides, then trigger the 
existing fork `Build` on the new head.
   
   Docs/Javadoc for the writer-sharing and `restoreWriter` contract (F4/F6/F8) 
are welcome in the same push. Once the new head SHA is up, ping me and I will 
do a full re-review.


-- 
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]

Reply via email to