Github user rdblue commented on the issue: https://github.com/apache/spark/pull/21230 @cloud-fan, I opened #21262 that is similar to this, but does pushdown when converting to a physical plan. You might like that as an alternative because it cleans up `DataSourceV2Relation` quite a bit and adds `output` to the case class arguments like other relations. The drawback to that approach that I had forgotten about is that it breaks `computeStats` because that runs on the optimized plan (but this affects all the other code paths as well). Up to you how to continue with this work, I just think we should consider the other approach since it solves a few problems. And `computeStats` is something we should update to work on physical plans anyway, right? Just let me know how you want to move forward. If you want to pull that commit into this PR, I'll close the other one.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org