Github user kiszk commented on the issue:
https://github.com/apache/spark/pull/21537
For 2. and 3., it is harder to say my opinion in the comment. Let me say
short comments at first.
For 2., if I remember correctly, @viirya once wrote the API document in a
JIRA entry. it would be good to update and add some thoughts about design as a
first step. I understand that it is hard to keep the up-to-date design
document, in particular, during the open-source development. This is because we
have a lot of excellent comments in a PR.
For 3., at the early implementation of SQL codegen (i.e. use `s""` to
represent code), I thought there are two problems.
1. lack of type of an expression (in other words, `ExprCode` did not have
the type of `value`)
2. lack of a structure of statements
Now, we meet a problem that it is hard to rewrite a method argument due to
problem 1. In my understanding, the current effort led by @viirya is trying to
resolve problem 1.
It is hard to rewrite a set of statements due to Problem 2. To resolve
problem 2, we need more effort. In my opinion, we are addressing them step by
step.
Of course, it would be happy for me to co-lead a discussion of the IR
design for 2.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
