From: Dmitry Ivanov <d.ivanov@postgrespro.ru>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Dmitry Ivanov <d.ivanov@postgrespro.ru>,
 pgsql-hackers <pgsql-hackers@postgresql.org>
Subject: Re: Possible problem in Custom Scan API
Date: Tue, 11 Apr 2017 23:28:17 +0300
MIME-Version: 1.0
Message-ID: <f7590622-182b-4362-b148-c21de109ad2a@postgrespro.ru>
In-Reply-To: <367.1491940070@sss.pgh.pa.us>
References: <e29ddd30-8ef9-4da5-a50b-2bb7b8c7198d@postgrespro.ru>
 <20021.1491933424@sss.pgh.pa.us>
 <1ef85b46-1385-4810-bb80-97cd977da1d5@postgrespro.ru>
 <367.1491940070@sss.pgh.pa.us>
Organization: Postgres Professional
User-Agent: Trojita/0.7; Qt/5.8.0; xcb; Linux; 
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: quoted-printable
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

Tom Lane wrote:
> Reading between the lines, I think the problem may be that you're not
> being careful about how you set up custom_scan_tlist.  But since the
> core code has zero involvement in that decision, it's hard to see why
> it would be a core code bug.

I'm sorry, but you didn't understand. It's *the core code* that builds the=
=20
targetlist, and sometimes it may decide to provide a physical targetlist,=20
not the one that's *really needed*. The broader targetlist has Vars that=20
cannot be supplied by the IndexOnlyScan node, hence the error.

We cannot come up with our own targetlist because we don't know if it's a=20
top node and we should return exactly the same tuple (CP_EXACT_TLIST) or=20
it's just the stray optimization that made us think so.

Append works only because it doesn't allow projections, and it can never=20
get a physical targetlist if an index doesn't contain all columns.

But everything changes when we use CustomScan: PostgreSQL will not pay=20
attention. For example, if our CustomScan is a child of NestLoop, the=20
former will call this (create_nestloop_plan):

inner_plan =3D create_plan_recurse(root, best_path->innerjoinpath, 0);

Since NestLoop can make projections, it doesn't enforce CP_EXACT_TLIST=20
flag, and our CustomScan may end up with a full physical targetlist (see=20
the code of create_scan_plan() if you don't believe me), INSTEAD OF the one=
=20
it really has been asked to produce. Meanwhile, Append will behave as it=20
should, since it doesn't care about physical tlists.

It's not just my imagination.


--=20
Dmitry Ivanov
Postgres Professional: http://www.postgrespro.com
Russian Postgres Company


--=20
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers