Message-ID: <fca8fe78-c67a-48ac-9164-60138e122f6c@postgrespro.ru>
Date: Wed, 14 Feb 2024 13:18:02 +0700
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: Propagate pathkeys from CTEs up to the outer query
To: Richard Guo <guofenglinux@gmail.com>,
 PostgreSQL-development <pgsql-hackers@postgresql.org>
Cc: Tom Lane <tgl@sss.pgh.pa.us>
References: <CAMbWs49xYd3f8CrE8-WW3--dV1zH_sDSDn-vs2DzHj81Wcnsew@mail.gmail.com>
Content-Language: en-US
From: Andrei Lepikhov <a.lepikhov@postgrespro.ru>
Organization: Postgres Professional
In-Reply-To: <CAMbWs49xYd3f8CrE8-WW3--dV1zH_sDSDn-vs2DzHj81Wcnsew@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: <https://www.postgresql.org/message-id/fca8fe78-c67a-48ac-9164-60138e122f6c%40postgrespro.ru>
Precedence: bulk

On 29/1/2024 10:18, Richard Guo wrote:
> In [1] we've reached a conclusion that for a MATERIALIZED CTE it's okay
> to 'allow our statistics or guesses for the sub-query to subsequently
> influence what the upper planner does'.  Commit f7816aec23 exposes
> column statistics to the upper planner.  In the light of that, here is a
> patch that exposes info about the ordering of the CTE result to the
> upper planner.
> 
> This patch was initially posted in that same thread and has received
> some comments from Tom in [2].  Due to the presence of multiple patches
> in that thread, it has led to confusion.  So fork a new thread here
> specifically dedicated to discussing the patch about exposing pathkeys
> from CTEs to the upper planner.
I like this approach. It looks good initially, but such features need 
more opinions/views/time to analyse corner cases.
It goes alongside my current backburner - pull parameterisation through 
the GROUP-BY and the query block fence up to the JOIN searching code of 
the parent query.

-- 
regards,
Andrei Lepikhov
Postgres Professional