dhruv-dagar commented on issue #7092: URL: https://github.com/apache/datafusion-comet/issues/7092#issuecomment-6268669296
I traced `BoundedWindowAggExec` in the pinned DataFusion 55.2.0 dependency. The stream retains rows across input_buffer, per-partition `partition_buffers`, and window output state, then prunes these at different points. There is no MemoryConsumer or MemoryReservation in the executor itself. A wrapper around the input/output stream seems likely to misaccount memory because it cannot observe internal pruning, and Arrow slices may retain shared backing buffers. Would you prefer that this be addressed upstream in Apache DataFusion by adding reservation accounting to `BoundedWindowAggExec`, or should we implement a Comet-side bounded-window executor? I’d like to align on the intended ownership before starting a patch. -- 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]
