aokolnychyi commented on PR #7558: URL: https://github.com/apache/iceberg/pull/7558#issuecomment-1541336715
I took a closer look. If I remember correctly, we added `NoStatsUnaryNode` to prevent issues with broadcast nested loop joins when the target would be the broadcasted side. It is an extremely edge case as the target must be smaller and there must be no equality condition in the ON clause. That means we should probably keep it for now unless it causes a performance degradation. Then we may workaround it. I am reluctant to provide an explicit hint to prefer a shuffle hash join as such joins can lead to OOM if the size of each partition after the shuffle on the build size is too big to fit into memory. We highly advise our users to configure Spark to use shuffle joins but that's not always possible. That said, let me think more about this tomorrow. -- 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]
