databird07 opened a new pull request, #20357: URL: https://github.com/apache/hudi/pull/20357
### Describe the issue this Pull Request addresses `TestHoodieFileSystemViews#testFileSystemViewConsistency` was one of the most expensive classes in the Java UT GitHub job. Each MDT-on table was written twice, once per reader view type, even though the view type only changes how the finished table is read. The matrix also only covered table versions 6 and 8, so versions 9 and 10 were never tested here. ### Summary and Changelog - Every case now writes with the metadata table enabled. The four MDT-off cases are removed. Turning MDT off on a table that has one stays covered by `TestHoodieBackedMetadata#testTurnOffMetadataTableAfterEnable` and `TestHoodieTable`. - Each table is written once and checked against both the in-memory and spillable views at every checkpoint. - Writer-side view storage: `SPILLABLE_DISK` for versions 6, 8 and 9, and both `MEMORY` and `SPILLABLE_DISK` for the current version (10), for COW and MOR. That's 10 cases instead of 12. - Views are closed in a `finally` block, and assertion failures name the view type that diverged. ### Impact Test-only. Runtime is about the same as before (+~10% in interleaved runs). About 750 more lines are covered in every run, mostly version 9/10 write and read paths. All 20 local runs passed. The only lines lost from the old test's always-covered set are the MDT-off paths listed in the commit message, and other tests still cover them. Measured locally on JDK 17 with 4 pinned CPUs and JaCoCo on; full numbers are in the commit message. ### Risk Level low. Test-only change; no production code touched. ### Documentation Update none ### Contributor's checklist - [x] Read through [contributor's guide](https://hudi.apache.org/contribute/how-to-contribute) - [x] Enough context is provided in the sections above - [x] Adequate tests were added if applicable -- 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]
