Github user metasim commented on the issue:
https://github.com/apache/spark/pull/16478
Does this PR remove the ability to read directly from Tungsten memory, in
cases where one wants to avoid the serialization/deserialization costs? IOW, in
an `Expression` over a UDT, can one still pull from select subfields without
deserializing the whole structure into JVM memory?
This is an important optimization for us in RasterFrames:
https://github.com/locationtech/rasterframes/blob/b731865d9d170fedbf0ffbe1e257694f4030db85/core/src/main/scala/org/apache/spark/sql/rf/InternalRowTile.scala#L43-L53
Just want to look ahead to what we might need to change down the road.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
