Re: POC, WIP: OR-clause support for indexes

On Fri, Oct 4, 2024 at 2:20 PM Peter Geoghegan <pg(at)bowt(dot)ie> wrote:
> On Fri, Oct 4, 2024 at 2:00 PM Alexander Korotkov <aekorotkov(at)gmail(dot)com> wrote:
> > Yes, transformAExprIn() does the work to coerce all the expressions in
> > the right part to the same type. Similar logic could be implemented
> > in match_orclause_to_indexcol(). What worries me is whether it's
> > quite late stage for this kind of work. transformAExprIn() works
> > during parse stage, when we need to to resolve types, operators etc.
> > And we do that once.
>
> I agree that it would be a bit awkward. Especially having spent so
> much time talking about doing this later on, not during parsing. That
> doesn't mean that it's necessarily the wrong thing to do, though.

True, but we also can't realistically use select_common_type() here. I
mean, it thinks that we have a ParseState and that there might be
values with type UNKNOWNOID floating around. By the time we reach the
planner, neither thing is true. And honestly, it looks to me like
that's pointing to a deeper problem with your idea. When someone
writes foo IN (1, 2222222222222222222222222), we have to make up our
mind what type of literal each of those is. select_common_type()
allows us to decide that since the second value is big, we're going to
consider both to be literals of type int8. But that is completely
different than the situation this patch faces. We're now much further
down the road; we have already decided that, say, 1, is and int4 and
2222222222222222222222222 is an int8. It's possible to cast a value to
a different type if we don't mind failing or have some principled way
to avoid doing so, but it's way too late to reverse our previous
decision about how to parse the characters the user entered. The
original "char *" value is lost to us and the type OID we picked may
already be stored in the catalogs or something.

--
Robert Haas
EDB: http://www.enterprisedb.com