sam-1112 commented on PR #5409: URL: https://github.com/apache/datafusion-comet/pull/5409#issuecomment-5385194839
Thanks for the follow-up and the detailed reproducer. I’ve addressed both remaining boundaries in the latest commit. For the source subtree, native eligibility now only accepts a column reference, `NULL`, or a short byte-preserving UTF-8 literal. Nested expressions such as `substring` and `concat` remain on the dispatcher. I added regressions for the ANSI short-circuit case, malformed source literals (including under `concat`), and oversized source literals. For the `explode` overflow, native `UnnestExec` is now wrapped in `BatchSplitExec`, which uses the runtime `TaskContext` batch size to split oversized output before downstream projections. The exact 8,192 → 16,384 row reproducer with the 131,072-byte replacement now passes. The regression test verifies all 16,384 results, confirms the plan uses `CometExplodeExec`, and confirms `replace` remains native. I also added a Rust unit test for runtime batch splitting. Both focused Rust and Spark 4.1 tests pass. -- 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]
