MIME-Version: 1.0
References: <eb5cc5d20c520e32b5963a4044b70af1@paolodestefani.it>
 <CA+mi_8bXjHTrzDLQ+n_6K2MHVsr0J32X1HA28Ey9O+-Zm6JaeQ@mail.gmail.com>
 <9a99d2d110b121542d1ea0a05066858b@paolodestefani.it>
 <CA+mi_8afF2jMifiLSvL99+Ztupy1me7xDNAv0g47WxHWbwN9vw@mail.gmail.com>
In-Reply-To: 
 <CA+mi_8afF2jMifiLSvL99+Ztupy1me7xDNAv0g47WxHWbwN9vw@mail.gmail.com>
From: Daniel Fortunov <postgresql@danielfortunov.com>
Date: Thu, 14 Oct 2021 00:28:11 +0100
Message-ID: 
 <CAH1rg6ZoQm=STwMR6AQ_FggxRxUn7qH_tLkRTfJTVYYdbTeL6w@mail.gmail.com>
Subject: Re: psycopg3 transactions
To: Daniele Varrazzo <daniele.varrazzo@gmail.com>
Cc: Paolo De Stefani <paolo@paolodestefani.it>,
 Psycopg <psycopg@postgresql.org>
Content-Type: multipart/alternative; boundary="000000000000887e4a05ce445116"
Archived-At: 
 <https://www.postgresql.org/message-id/CAH1rg6ZoQm%3DSTwMR6AQ_FggxRxUn7qH_tLkRTfJTVYYdbTeL6w%40mail.gmail.com>
Precedence: bulk

--000000000000887e4a05ce445116
Content-Type: text/plain; charset="UTF-8"

On Wed, 13 Oct 2021 at 20:06, Daniele Varrazzo <daniele.varrazzo@gmail.com>
wrote:

>
> I think we can improve the documentation there by extending more about
> the effects of the interaction between DBAPI transactions and the
> transaction() blocks. And things are definitely more intuitive if
> transaction() is used in autocommit: we might want to advise people to
> do that.
>

Agree. The DBAPI choice to mandate that autocommit be off by default is a
strange and unfortunate choice that in my experience leads to a never
ending series of "surprises" such as this one.

My conclusion is that the only sane thing to do is:
1. Only ever create connections in autocommit mode.
2. Only ever use `with connection.transaction()` to control transactions.
3. Forget that `connection.commit()` and `connection.rollback()` exist, and
never use them.

I'm in favour of recommending this as strongly as reasonably possible in
the documentation.

It's a shame that 1 is not the default. So you still have to remember to
not forget to do this explicitly, every time you create a connection.

Dani

--000000000000887e4a05ce445116
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr">On Wed, 13 Oct 2021 at 20:06, Daniele Var=
razzo &lt;<a href=3D"mailto:daniele.varrazzo@gmail.com">daniele.varrazzo@gm=
ail.com</a>&gt; wrote:<br></div><div class=3D"gmail_quote"><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex"><br>
I think we can improve the documentation there by extending more about<br>
the effects of the interaction between DBAPI transactions and the<br>
transaction() blocks. And things are definitely more intuitive if<br>
transaction() is used in autocommit: we might want to advise people to<br>
do that.<br></blockquote><div><br></div><div>Agree. The DBAPI choice to man=
date that autocommit be off by default is a strange and unfortunate choice =
that in my experience leads to a never ending series of &quot;surprises&quo=
t; such as this one.</div><div><br></div><div>My conclusion is that the onl=
y sane thing to do is:</div><div>1. Only ever create connections in autocom=
mit mode.</div><div>2. Only ever use `with connection.transaction()` to con=
trol transactions.</div><div>3. Forget that `connection.commit()` and `conn=
ection.rollback()` exist, and never use them.</div><div><br></div><div>I=
9;m in favour of recommending this as strongly=C2=A0as reasonably possible =
in the documentation.</div><div><br></div><div>It&#39;s a shame that 1 is n=
ot the default. So you still have to remember to not forget to do this expl=
icitly, every time you create a connection.</div><div><br></div><div>Dani</=
div></div></div>

--000000000000887e4a05ce445116--