On Tue, 8 Sept 2020 at 13:46, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> I've been doing some handwaving about changing the representation
> of Vars, with an eye to making it clear by inspection whether a
> given Var is nullable by some lower outer join [2]. If that work
> ever comes to fruition then the need for "check_null_side" would
> go away. So maybe we should put this idea on the back burner
> until that happens.
I looked at this patch too. I agree that we should delay adding any
new smarts in regards to NULL or NOT NULL until we have some more
robust infrastructure to make this sort of patch easier and cheaper.
My vote is to just return this patch with feedback. Maybe Surafel
will be interested in pursuing this later when we have better
infrastructure or perhaps helping review the patch you're talking
about.
David