Message-ID: <0db12127e1bfc939a66a6ee959346b28cf1c6838.camel@cybertec.at>
Subject: Re: Parallel workers via functions?
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Jim Vanns <jvanns@ilm.com>, pgsql-general <pgsql-general@postgresql.org>
Date: Tue, 28 Jan 2025 05:47:11 +0100
In-Reply-To: <CAH7vdhPr9BJ2TjbVU0RepN92eUFmWorukAGEyhBqpx8gF67xdQ@mail.gmail.com>
References: 
	<CAH7vdhPr9BJ2TjbVU0RepN92eUFmWorukAGEyhBqpx8gF67xdQ@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.54.3 (3.54.3-1.fc41) 
MIME-Version: 1.0
Archived-At: <https://www.postgresql.org/message-id/0db12127e1bfc939a66a6ee959346b28cf1c6838.camel%40cybertec.at>
Precedence: bulk

On Mon, 2025-01-27 at 18:08 +0000, Jim Vanns wrote:
> If I have a function that is marked 'stable parallel safe' and returns
> a table, can a calling function or procedure (marked volatile parallel
> unsafe) still take advantage of the parallel workers from the first
> function - as the data source. I.e.
>=20
> func_a(); // selects, returns table, parallel safe
> func_b() {
>    insert into foo
>    select * from func_a(); // Will func_a still execute parallel
> workers to fetch the data?
> }
>=20
> Or even if func_b() uses 'create temporary table as select * from
> func_a()' and then insert?
>=20
> I ask because when I simply call func_a() from a psql shell, I see the
> parallel workers run and everything is nice and swift. But when called
> from a data-modifying function like func_b(), no workers are spawned
> :( Even from the read-part of the code.
>=20
> Are there differences in functions vs. stored procedures that might
> affect the behaviour of the planner to disregard workers?

See https://www.postgresql.org/docs/current/when-can-parallel-query-be-used=
.html

The problem here is the INSERT.  Data modifying statements won't use
parallel query.

There are exceptions: CREATE TABLE ... AS SELECT ... should be able
to use parallel query.

Yours,
Laurenz Albe