Message-ID: <b1b9330fed3738135aba8664a55a6fd8239985c8.camel@cybertec.at>
Subject: Re: Do blocks support transaction control?
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: depesz@depesz.com, pgsql-docs@lists.postgresql.org
Date: Sat, 24 Sep 2022 17:14:42 +0200
In-Reply-To: <166394003436.654.15293085988078557076@wrigleys.postgresql.org>
References: <166394003436.654.15293085988078557076@wrigleys.postgresql.org>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.42.4 (3.42.4-2.fc35) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Archived-At: 
 <https://www.postgresql.org/message-id/b1b9330fed3738135aba8664a55a6fd8239985c8.camel%40cybertec.at>
Precedence: bulk

On Fri, 2022-09-23 at 13:33 +0000, PG Doc comments form wrote:
> Page: https://www.postgresql.org/docs/14/sql-do.html
> Description:
> 
> Apparently now DO blocks support COMMIT; - which make them more like
> procedures than functions.
> 
> Tried it with:
> 
> create table z (a int4, b int4);
> insert into z (a,b) select i, i from generate_Series(1,10) i;
> do $$
> declare
> begin
> update z set b = 2;
> commit;
> perform pg_sleep(120);
> end;
> $$ language plpgsql;
> 
> And while it was running, in another psql sessions, I:
> 
> 1. could see b= 2
> 2. could update any of the rows in z.
> 
> Is it documented anywhere? DO docs say that do is like function, which it
> doesn't seem to be?

The documentation says:

  If DO is executed in a transaction block, then the procedure code cannot execute
  transaction control statements. Transaction control statements are only allowed
  if DO is executed in its own transaction.

That sentence would not make sense if COMMIT were not allowed in a DO statement.
So it is not spelled out, but implicitly clear.

Yours,
Laurenz Albe