andygrove commented on PR #4746: URL: https://github.com/apache/datafusion-comet/pull/4746#issuecomment-5876346209
This is a light fully automated review since there are so many PRs open. With `hasEmptyRelationInput` gone, a few places still describe it. `docs/source/user-guide/latest/compatibility/operators.md:32` tells users that Parquet writes whose input contains an empty relation use Spark's writer to preserve readable empty output files, and the `EmptyRelationExec` row at `docs/source/user-guide/latest/operators.md:57` links there for "writer fallback". That sentence documents the guard this PR removes, and `CometEmptyRelationParquetWriterSuite` asserts the opposite: over a native empty relation the write stays native and the zero-partition swap writes the schema-only file. The comment at `spark/src/main/scala/org/apache/spark/sql/comet/CometWriteFilesExec.scala:134` also still says `CometDataWritingCommand` declines empty-relation inputs on 3.x because its writer "has nowhere to put this swap", but `runNativeWriteJob` now does the same swap. And `native/proto/src/proto/operator.proto:785` explains the `IcebergWrite` per-task fields by pointing at how `Parque tWriter.task_attempt_id` is filled in `CometNativeWriteExec`, a field this PR reserves. Could these be updated here too, so the user guide says a write over a native empty relation stays native and still leaves a readable schema-only file? -- 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]
