DanielLeens commented on PR #11932:
URL: https://github.com/apache/seatunnel/pull/11932#issuecomment-5601151226
@SEZ9 agreed with keeping the `Collector.restoreSchema` thread open with
that qualification rather than marking it resolved — that's the right call, and
to directly answer the intent question: yes, "known Zeta-only limitation for
now" is the actual design intent here, not an accidental gap. I re-pulled the
diff on `b6c96b97356d` directly to confirm rather than relying on my own
earlier summary: `Collector.restoreSchema(List<CatalogTable>)` is added as a
`default void ... {}` no-op on the shared `Collector` interface, and the only
override anywhere in this PR's diff is in `SeaTunnelSourceCollector` (Zeta's
own engine-side collector). No Flink or Spark translation-layer collector class
is touched by this PR at all, so it does not add or remove any capability on
those engines — they simply keep the interface's pre-existing no-op default,
exactly as before this PR. That's a scope boundary consistent with the PR's own
`[Fix][Zeta]` title, not a regression on Flink/Spark, and it fol
lows the same default-method extension pattern already used elsewhere on this
interface (`markSchemaChangeBeforeCheckpoint()` / `collect(SchemaChangeEvent)`
are no-ops outside Zeta too).
Given that, I'd treat "extend `restoreSchema` to Flink/Spark's
translation-layer collectors" as a legitimate separate enhancement rather than
a blocker on this PR. Happy to see it tracked as a follow-up issue if you or
@davidzollo want to file one, but I don't think it should hold up this
checkpoint-recovery fix, which is scoped to Zeta from the title down.
On the docs/upgrade note: no action needed from my side — you said you'll
verify the "Checkpoint restore compatibility" section against the diff yourself
and resolve that thread once confirmed, which is the right way to close it out;
the file/line pointers are in my 2026-09-08 comment if useful.
One correction to my own 2026-09-09T00:35 comment: I said "nothing further
outstanding, ready to merge," which was premature given this thread was still
open when I posted it. To update that: from the core-logic side (Findings A/B/C
from my 2026-08-23 review) I still have nothing further to add, but the PR
isn't clear of all reviewer threads until your `restoreSchema` qualification
above and your remaining re-check of the earlier points against the current
head are both resolved on your end. No rush — happy to take another pass once
you've closed those out.
--
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]
