Andrew Lamb

@Dandandan -- I had not thought of the "parallel sorts with limit" as a good intermediate state -- you are exactly right and that makes perfect sense

Thank you @milenkovicm -- I plan to review this more carefully tomorrow morning. cc @Dandandan and @yjshen

> is it line 442 which is "unbounded" ? Yes that looks correct

> One question before hand, which will save me some time, which aggregation operators will end up using that hash? I think it is based on the type of the...

Thanks again @milenkovicm

This PR is still marked as a draft -- but it looks like maybe it is ready for review again?

I actually think in general the coercion rules for `uniform` should attempt to coerce all arguments to the same type as the first argument

@comphead I think https://github.com/apache/arrow-datafusion/issues/3123 and this ticket are different This ticket is about the input types of the arguments to `array[]` https://github.com/apache/arrow-datafusion/issues/3123 is about the output type produced by the...

> here I can see nonconsistent behaviuor. I suppose all 3 should try coerce to first datatype and fail if its not possible? I think that would make sense --...

> All those constructions do the same work, but they can be out of sync like now. do we need to support all 2 branches? I am sorry for the...