sunchao opened a new issue, #7113: URL: https://github.com/apache/datafusion-comet/issues/7113
Join runtime filters cannot currently attach to a Parquet reader through any standalone projection. This includes a column-only projection inserted by Spark to remove a residual-only column, for example selecting `p.item_id` while filtering `p.quantity > 0`. Follow-up from [PR #6431 review](https://github.com/apache/datafusion-comet/pull/6431#discussion_r4243806638): allow reader attachment through column-only `CometProjectionExec` nodes, remapping the live runtime-filter expression to the input columns. The early decoded-batch consumer path already handles column-only projection remapping and may offer reusable logic. The implementation should preserve projection metric ownership and runtime-filter updates, retain barriers for computed projections and limits, and keep the per-file Parquet schema-conversion safeguards. Coverage should include reordered/aliased/subset columns, the residual-only-column query above, retained residual evaluation, and unsupported projection shapes. -- 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]
