+1 Using Spark’s ready-made test suites to maintain data consistency between Auron and the Spark engine is a good idea. Apart from the question of whether different Spark versions can use macros to replace the multi-directory approach, I’ve observed in the PRs that include Gluten-related modules that, besides introducing Spark unit tests, there are many newly added classes and pieces of code that are very similar to Spark’s existing tests. This usage seems somewhat redundant.
王磊 <[email protected]> 于2025年12月17日周三 21:09写道： > +1 > > Glad to see this discussion. The correctness of Auron is very important. > Looking forward to the implementation of this proposal. > > > Zhang Li <[email protected]> 于2025年12月17日周三 20:58写道： > > > +1 > > i'd like to see someone working on improving the correctness of the > > project. > > one suggestion: as we already have spark-version-annotation-macros to > > distinguish different spark versions. it is better to have just one > testing > > module, and use the `@sparkver` annotation to activate/deactivate > specified > > spark versions. > > > > James <[email protected]> 于2025年12月17日周三 00:05写道： > > > > > Hi, everyone I'd like to start a discussion about AIP-2: Enhance > Auron’s > > > Correctness Testing [1]. Looking forward to your feedback. > > > > > > [1]. > > > > > > > > > https://docs.google.com/document/d/1v8wMyLZXuA7tmDSJysAo8CRqWX36SWR5ZgjyOMSfm94/edit?tab=t.0 > > > > > > > > -- > Best regards > > Your sincerely, > Lei Wang(王磊）,Ph.D. Candidate > Institute of Computing Technology > Chinese Academy of Sciences > Beijing, China, 100190 > E-mail: [email protected] <[email protected]> >
