Github user cloud-fan commented on the issue:
https://github.com/apache/spark/pull/21143
Hi @rdblue , you can also create a codegen-ed filter in your data source,
but the real problem is whole-stage-codegen. The code is generated at the
driver side, sent to and compiled at the executor side. It is non-trivial to
change it to generating code at executor side. If union doesn't work, let's
think more about this and write a design doc first. I'm willing to join the
discussion and help to make a design, thanks!
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
