stczwd commented on pull request #35669: URL: https://github.com/apache/spark/pull/35669#issuecomment-1054879956
> +1 on what @viirya said - the PR description is a bit confusing, for instance I don't really know what column `a` and `c` are for when first looking at them. It'd be better if you can also put the original SQL query there. > > Also I'm curious how Parquet handles pushed filters like `(a < 10 and c == 0) or ((a >= 10 and c >= 1) and c < 3))`, since the partition columns are not materialized. Thanks for your attension, @sunchao @viirya @c21. Just as @cloud-fan said, we can push all filters, data filters and partition filters, to ParquetScan, and evaluate and prune partition filter with partition values at runtime, before pushed into parquet. With this, we can read `a < 10` in partition `c=0` and read `a>=10` in partition `c>=1 and c<2`, instead of reading `a<10 or a>=10` in partition `c=0` and partition `c>=1 and c<2`. -- 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]
