Github user mallman commented on the issue:
https://github.com/apache/spark/pull/21889
> select id, name.middle, address from temp - Works
> select name.middle, address from temp - Fails
> select name.middle from temp - Works
> select name.middle, id, address from temp - Works
> select name.middle, address, id from temp - Works
Removing the `order by` clause from your test query caused it to fail, but
it has nothing to do with ordering. It appears that the failure in this case is
manifested when the file scan schema is exactly the `name.middle` and `address`
columns. Introducing the `order by` clauses in the test suite queries gave them
necessary determinism for checking query answers, but these modifications also
altered the file scan schema.
I need to fix the tests, but I think that the failure underlying the
previously ignored test case has not been resolved after all. It was just a
case of confusing coincidence. Unfortunately we're still not ready to merge
this PR yet.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
