scalar function reverse is not compatible with Spark #2612
Description
Describe the bug
I don't have a simple repro yet, but I found a bug while experimenting with some fuzz testing.
The query was select c0, reverse(c0) FROM t1 ORDER BY c0.
Two of the result mismatches were:
== Results ==
!== Correct Answer - 1000 == == Spark Answer - 1000 ==
struct<c29:array<binary>,reverse(c29):array<binary>> struct<c29:array<binary>,reverse(c29):array<binary>>
![ArrayBuffer(null),ArrayBuffer(null)] [ArrayBuffer(null),ArrayBuffer([B@1814c14)]
![ArrayBuffer(null, null),ArrayBuffer(null, null)] [ArrayBuffer(null, null),ArrayBuffer([B@589e926b, [B@6aaa0ef7)]
Perhaps we are creating an empty array instead of null, but it is hard to tell because the output here does not show the contents of the byte array.
Steps to reproduce
No response
Expected behavior
No response
Additional context
No response