Github user cloud-fan commented on the issue:
https://github.com/apache/spark/pull/20085
Hi @bdrillard , sorry for the late reply, as I was thinking hard about this
problem. I think we all agree that we should have more object-related
expressions, so that it's more flexible for Spark and other projects to do many
things with the codegen-able expressions.
However we should think hard about what object-related expressions Spark
should provide, and make sure they are orthogonal and composable. I'm OK with
most of the expressions you added, but have some other thoughts about
`InitializeObject`.
I propose to improve the existing `NewObject`, and introduce a new phase:
initialize phase. So `NewObject` should have a list of expression as its
constructor parameters, and another list of expressions as post-hoc
initializaion. To create a case class, the construct parameters expressions are
not empty and initializing expressions are empty. To create a java bean, it's
the opposite.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
