Inconsistent Null handling in array_max/min functions in presence of NaN

[bikramSingh91]

presto returns a NaN, even if there is a NULL in the array. However, for other types, it always returns NULL if one exists. As a part of fixing the NaN behavior under this RFC (https://github.com/prestodb/rfcs/blob/main/RFC-0001-nan-definition.md), it would be good to also make the behavior consistent across types for nulls and always return null if it exists.

select ARRAY_MIN(array[1, 3, NULL , 3]);  ---- NULL
select ARRAY_MIN(array['a','b', NULL, 'c']);  ---- NULL
select ARRAY_MIN(array[1, nan(), NULL , 3]);  ---- nan

Similarly for array_max

select ARRAY_MAX(array['a','b', NULL, 'c']);  ---- NULL
select ARRAY_MAX(array[1, 3, NULL , 3]);  ---- NULL
select ARRAY_MAX(array[1, nan(), NULL , 3]);  ---- nan

[elharo]

NaN's a weird one. I'm not sure we should do this. I'm not sure we shouldn't. However, I do notice that the RFC doesn't address the case of null comparisons with NaN and whichever way we go it likely should.

[rschlussel]

with the new plan, nan is treated like a regular number, so null comparisons with nan would like any other comparison with null (i.e. unknown).