Re: Minor difference in behavior between +/-

At Wed, 25 Jan 2023 07:12:18 -0700, "David G. Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com> wrote in
> On Tuesday, January 24, 2023, Ryan Murphy <ryanmurf(at)gmail(dot)com> wrote:
>
> > SELECT NULL - date '2023-01-05';
> > -> result is null
> > SELECT NULL + date '2023-01-05';
> > -> result is [42725] ERROR: operator is not unique: unknown + date Hint:
> > Could not choose a best candidate operator. You might need to add explicit
> > type casts. Position: 13
> >
> >
> There is no bug here.
>
>
> > I would expect both of those to be null.
> >
>
> I get why you expect that but we expend little to no effort to handle null
> in a type agnostic way. Type inference has to happen first and regardless
> of whether the literal is null. Type inference failed in the + case but
> not the - case. That is just happenstance of what operators happen to
> exist in the system.

In a bit more deatil, the type inference mechanism assumes the null as
a date since the other input is a date. In this case, the operator
"date - date" exists but "date + date" doesn't.

I'm not sure whether the SQL standard defines a binary operator should
be "strict/or return null on null input", but it seems to to me that
we assume a case of non-strict binary operators, or maybe we simply
dont' care about the case of null operands.

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center