adriangb commented on PR #19639: URL: https://github.com/apache/datafusion/pull/19639#issuecomment-3715746801
@alamb I've come to the finding here that the non-filter pushdown version of this implementation is slower than `main`, hence why the benchmarks are bad: https://github.com/pydantic/datafusion/blob/dc3593ff95ebfa5d27efff71b6a0ad8d36d16063/datafusion/datasource-parquet/src/opener.rs#L778-L846 I don't see anything obvious this is doing worse than `FilterExec`, I wonder if it has to do with batch coalescing or repartitions of the data. If it is indeed one of those things I think that's another thing we need to think / rethink with filter pushdown, otherwise we're fighting an uphill battle. -- 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]
