peter-toth commented on issue #23802: [SPARK-26893][SQL] Allow partition pruning with subquery filters on file source URL: https://github.com/apache/spark/pull/23802#issuecomment-464736310 > > if subquery filter can be a partition filter, why can't it be a data filter? > > `DataSourceStrategy.translateFilter` can't transform them to `Filter` so they will not be pushed down to data source. > (Partition filters are handled in `FileSourceScanExec.selectedPartitions` and there is no need to transform them to `Filter`.) Actually I don't see a reason we can't defer their translation to filters until `inputRDD` creation. Obviously they need to be removed from `metadata` as I did for `PartitionCount` in case they contain subquery. This could work if we can be sure that `Filter`s pushed down are enforced and we can remove the `FilterExec` parent of `FileSourceScanExec`.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
