Github user HyukjinKwon commented on the issue:
https://github.com/apache/spark/pull/20280
If the renaming scenario works in most of cases as expected, I think it'd
be worthwhile to have a configuration; however, the previous behaviour looks
actually odd because it's going to work only in certain weird conditions when
fields in `Row` and fields in the given schema are in the same alphabetical
order (https://github.com/apache/spark/pull/20280#discussion_r182569705).
Otherwise this case fails already as well.
The test case modified in
https://github.com/apache/spark/pull/20280#discussion_r182569705 actually works
only because `key` and `value` in `Row` and `a` and `b` in the schema are in
the same order. I think the test case should be invalid ..
I thought about this for a while and failed to describe what the
configuration does .. It sounded describing a bug like it was a proper
behaviour that can be controlled by a configuration ..
I think this one sounds more like a bug fix to me so far. Workaround should
be relatively easy. Maybe, would it be good enough to describe workaround in
the guide instead? I think it should be fine if we just use a map to convert
`Row` to things like a tuple.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
