MIME-Version: 1.0
References: <18652-deaa782ebcca85d1@postgresql.org>
 <2527690.1728510183@sss.pgh.pa.us>
 <CAMbWs49Pcnm_cX=7azYY9d8+qRrOOKDjJDCa0KGRk9kWWMOz3A@mail.gmail.com>
 <40c8a5c2-0040-4fe6-ab11-6dca23353a07@gmail.com>
 <3001952.1728680364@sss.pgh.pa.us>
In-Reply-To: <3001952.1728680364@sss.pgh.pa.us>
From: Richard Guo <guofenglinux@gmail.com>
Date: Sat, 12 Oct 2024 12:13:51 +0800
Message-ID: 
 <CAMbWs48t0LBsPf4bZ9UdGousuTNmrauD_01NrauRFXTrwt++Jg@mail.gmail.com>
Subject: Re: BUG #18652: Planner can not find pathkey item to sort for query
 with expression and expression index
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Andrei Lepikhov <lepihov@gmail.com>, exclusion@gmail.com,
 pgsql-bugs@lists.postgresql.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CAMbWs48t0LBsPf4bZ9UdGousuTNmrauD_01NrauRFXTrwt%2B%2BJg%40mail.gmail.com>
Precedence: bulk

On Sat, Oct 12, 2024 at 4:59=E2=80=AFAM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> This means that it will work so long as all of the Vars needed by the
> EC member expression are available from the plan node's input, which
> they surely are if they are referenced in the existing tlist.  That
> is, even if we wanted to compute "t2.i + 2" it'd be fine.
>
> (This would fall down perhaps if there are volatile functions in
> the sort expression, but I believe we already reject using volatile
> expressions for merge append, so it's not a problem.)
>
> So I conclude that Andrei's patch will fix it, although I don't like
> the way that that requires (potentially) multiple re-executions of
> pull_var_clause.  I think we should refactor the code to avoid that,
> which actually ends up being less code, as in the attached draft.

+1.  I think it's conceptually correct that if 'Var_A + 2' is
computable in a plan node's targetlist, then 'Var_A + 1' must be also
computable in the same targetlist.

Thanks
Richard