From: Tom Lane <tgl@sss.pgh.pa.us>
To: Richard Guo <guofenglinux@gmail.com>
cc: Tender Wang <tndrwang@gmail.com>, exclusion@gmail.com,
        pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #18422: Assert in expandTupleDesc() fails on row mismatch
 with additional SRF
In-reply-to: 
 <CAMbWs4-xmohkFzVaT8rxEjpkB35JUNxpQtyEaPD+OafFMm5W8g@mail.gmail.com>
References: <18422-89ca86c8eac5246d@postgresql.org>
 <CAHewXNkfDfKw80kgNvtC0ZvNz=hGxXLyEwiGJzhr3r_fFAkVeQ@mail.gmail.com>
 <3861246.1712691263@sss.pgh.pa.us>
 <CAMbWs4-3O34h5_T_OXRsAzGKg+cQEp1w9aq5i=U3EbxQzvo5VA@mail.gmail.com>
 <3917908.1712719937@sss.pgh.pa.us>
 <CAMbWs4_j1wq+nUK4t1oROOj2Qgvds=3G4RWA3NBNygf1wkha0A@mail.gmail.com>
 <4019337.1712758601@sss.pgh.pa.us>
 <CAMbWs4_QgZogwYUgXb=7K_GQ-w2JC3JxONZygkQiXBu11CyXCQ@mail.gmail.com>
 <170712.1712844795@sss.pgh.pa.us>
 <CAMbWs4-xmohkFzVaT8rxEjpkB35JUNxpQtyEaPD+OafFMm5W8g@mail.gmail.com>
Comments: In-reply-to Richard Guo <guofenglinux@gmail.com>
	message dated "Fri, 12 Apr 2024 11:25:49 +0800"
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-ID: <330607.1712893609.1@sss.pgh.pa.us>
Content-Transfer-Encoding: 8bit
Date: Thu, 11 Apr 2024 23:46:49 -0400
Message-ID: <330608.1712893609@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/330608.1712893609%40sss.pgh.pa.us>
Precedence: bulk

Richard Guo <guofenglinux@gmail.com> writes:
> On Thu, Apr 11, 2024 at 10:13 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> I did think about that, but it seems mighty weird.  The semantics of
>> the flag would have to be something like "I want a tupdesc when the
>> result type is COMPOSITE, but not when it's RECORD", which seems
>> rather arbitrary.

> Indeed.

I did actually spend some time today trying to code that up, to see
what it would look like.  The attempt failed though, because there
are existing cases in which get_expr_result_type() returns
TYPEFUNC_RECORD, so we can't use that result code as a positive
indicator that "this RTE has a coldeflist".  In a green field
we could invent another TYPEFUNC_XXX code, but that's not a
back-patchable idea.  So the new function would need some independent
indicator that it saw a coldeflist, and at that point there's
basically nothing we're hiding from the callers.

>> Perhaps it'd be sufficient to add a note to the header comment of
>> get_expr_result_type warning about when not to use it.

> Works for me.

OK, barring other objections I'll push forward on that basis.

			regards, tom lane