yadavay-amzn commented on PR #56244: URL: https://github.com/apache/spark/pull/56244#issuecomment-4851281487
Thanks @cloud-fan, both addressed: - **V2 wiring:** moved the expansion into `FileScanBuilder.pushFilters` (`SupportsPushDownCatalystFilters`, which all V2 file sources use). Only the translate-and-push loop sees the expanded filters; the returned post-scan filters stay the original struct predicates, so the pushed field predicates remain a sound over-approximation like the V1 path. Removed the now-redundant expansion from `PushDownUtils.pushFilters`. - **V2 test gap:** added a DSv2 case (`useV1SourceList=''`) asserting the decomposed field predicates reach `ParquetScan`'s pushed filters, plus a DSv2 on/off parity test. Confirmed the new test fails without the `FileScanBuilder` change and passes with it. Also refreshed the PR description to reflect the V2 path via `FileScanBuilder`. PTAL. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
