korowa commented on issue #7848: URL: https://github.com/apache/arrow-datafusion/issues/7848#issuecomment-1769145148
I was able to reproduce it with cross join (all records in both build and probe sides have the same key, and build side is small enough to fit in memory limit) -- it's true, "working" batch is not tracked at all. I guess the most proper solution would be to produce partially joined batch (in the same fashion like it [works](https://github.com/apache/arrow-datafusion/blob/7acd8833cc5d03ba7643d4ae424553c7681ccce8/datafusion/physical-plan/src/joins/sort_merge_join.rs#L724) in MergeJoin), rather than accumulating output in memory until memory limit violated. Any thoughts on this option? -- 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]
