QuakeWang opened a new pull request, #8632:
URL: https://github.com/apache/paimon/pull/8632
### Purpose
Ray 2.53+ defers `Datasink.on_write_start()` until the first input bundle.
A zero-block dataset has no bundle, but `on_write_complete()` still runs,
leaving Paimon's writer builder uninitialized. Whole-table and static-partition
overwrites therefore fail before `commit([])`.
Initialize the writer builder on demand when completion needs to commit.
Zero-block append writes remain no-ops.
### Tests
- Added unit coverage for a skipped `on_write_start()` callback.
- Added an integration test using `ray.data.range(0)`.
- Passed `RaySinkTest` and `RayIntegrationTest` on Ray 2.54.0.
- Verified the regression tests on Ray 2.52.1, 2.53.0, and 2.54.0.
- Passed Python license and Flake8 checks.
--
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]
