Date: Wed, 18 Nov 2020 14:51:34 -0300
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Daniel Verite <daniel@manitou-mail.org>
Cc: "David G. Johnston" <david.g.johnston@gmail.com>,
	Matthieu Garrigues <matthieu.garrigues@gmail.com>,
	Michael Paquier <michael@paquier.xyz>,
	Dave Cramer <davecramer@postgres.rocks>,
	PostgreSQL-development <pgsql-hackers@postgresql.org>,
	Craig Ringer <craig@2ndquadrant.com>,
	Vaishnavi Prabakaran <vaishnaviprabakaran@gmail.com>,
	Andres Freund <andres@anarazel.de>,
	Michael Paquier <michael.paquier@gmail.com>,
	Claudio Freire <klaussfreire@gmail.com>,
	Heikki Linnakangas <heikki.linnakangas@iki.fi>
Subject: Re: PATCH: Batch/pipelining support for libpq
Message-ID: <20201118175134.GA30739@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <b27523b8-6e41-44e8-8cc6-ea136316cdcd@manitou-mail.org>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

On 2020-Nov-14, Daniel Verite wrote:

> The patch I posted in [1] was pretty simple, but at the time, query
> results were always discarded. Now that pgbench can instantiate
> variables from query results, a script can do:
>   select 1 as var \gset
>   select :var;
> This kind of sequence wouldn't work in batch mode since it
> sends queries before getting results of previous queries.
> 
> So maybe \gset should be rejected when inside a batch section.

Hah.

Hacking pgbench extensively is beyond what I'm willing to do for this
feature at this time.  Making \gset rejected in a batch section sounds
simple enough and supports \beginbatch et al sufficiently to compare
performance, so I'm OK with a patch that does that.  That'd be a small
extension to your previous patch, if I understand correctly.

If you or others want to send patches to extend batch support with
read-write tracking for variables, feel free, but I hereby declare that
I'm not taking immediate responsibility for getting them committed.