SEZ9 commented on PR #12578:
URL: https://github.com/apache/seatunnel/pull/12578#issuecomment-6206758625
Thanks @hutiefang76 for merging current `dev` in `ef53a70c6` and recording
the results on the merged head separately.
Re-checked at `ef53a70c6f72` against my earlier points:
- **F1**: the production change is unchanged from `ec14f3a3` — `setField`
does the array store and then `if (size != 0) { size = 0; }`, so freshly
assembled rows pay only a volatile read. Resolved.
- **F2**: Javadoc is now on `setField`, `getFields` and both `getBytesSize`
overloads, and the "Row Data Contract" section is in the EN/ZH architecture
guides. Resolved.
- **F3 / F4 / F5**: the merge kept the three cache-invalidation tests, and
the only test delta since `ec14f3a3` is `testRowTypeCopiesConstructorArrays`
from `dev`. The two cases I'd still like to see in `SeaTunnelRowTest` are:
1. cross-overload on a single row — `getBytesSize(rowType)`, then
`setField`, then untyped `getBytesSize()` returns the recomputed value (and
vice versa), matching the collector → metrics flow;
2. the documented limitation — mutating through `getFields()` or inside a
nested row does **not** clear the cache, so the stale value is returned until a
top-level `setField`.
Each should be a few lines and needs no production change.
Nothing else from my side.
<!-- streview-comment:1785 -->
--
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]
