Andres Freund <andres@anarazel.de> writes:
> On October 10, 2018 10:37:40 AM PDT, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
>>> So, which part of this supposedly does not work in PostgreSQL?
>> The part where it infers that b.sno is unique based solely on it having
>> been equated to a.sno.
> Isn't the spec compliant thing that's missing dealing with unique not null?
IIRC, the spec has a whole bunch of "functional dependency" proof rules,
of which the only one we implement at the moment is the one about the
other columns of a table all being functionally dependent on its pkey.
I don't know if any of the spec's rules are at all close to this one.
regards, tom lane
