pchintar commented on PR #4591: URL: https://github.com/apache/datafusion-comet/pull/4591#issuecomment-4858418419
Hi @andygrove I've addressed your additional read-path and pruning concerns: - For cached batches written through Spark's default cache format, I added explicit handling in `ArrowCachedBatchSerializer` instead of letting the read path throw. The serializer now detects `DefaultCachedBatch` in both columnar and row read paths and decodes it locally using Spark's default cache layout, including the selected-column indexes and `ColumnAccessor.decompress`. This keeps tables cached while `spark.comet.exec.inMemoryCache.enabled=false` readable later, even though the Comet cache serializer is installed for the session. - Added a `count(*)` cached-table regression test for empty projection scans. - Added a NaN floating-point pruning regression test. - Added a comment explaining why the per-column size stat is currently left as 0. I also updated the PR description by including the results there. -- 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]
