gene-bordegaray commented on PR #21832: URL: https://github.com/apache/datafusion/pull/21832#issuecomment-4314056110
> I have not read the code in detail (1.6k LOC) but I wonder how this will work in a world where the partition index of an opener is much less clear. Morselized scans (#20529) mean that filters will get applied at the morsel stage. Morsels won't belong to partitions per se / will be shared across partitions. So it seems to me that this entire feature is incompatible with the direction in which DataFusion is headed, unless I'm missing something. The important thing to point out is that this only uses partition-index routing when the scan still has a constant partition identity, which is the preserved file partitioned case we are targetting here. If future morselized scans no longer have that identity for a morsel, they can fall back to the global dynamic filter, so this is still very valuable use case for partitioned data wee are seeing in production and should not be blocking the morselized driven scans that are being pursued. Let me know thoughts. -- 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]
