MIME-Version: 1.0
References: 
 <CA+mi_8Yc92bX3qhTqXq4LkGyn5VmQj6uRAcVebpqt398Tm4EJA@mail.gmail.com>
 <CA+mi_8Y+fuzZcT+boRwcQKPd58B=bc324MA2=TAd6xuAvfuMmw@mail.gmail.com>
 <7171e07d-bb96-dfc6-47e9-351d85f4a973@dalibo.com>
In-Reply-To: <7171e07d-bb96-dfc6-47e9-351d85f4a973@dalibo.com>
From: Daniele Varrazzo <daniele.varrazzo@gmail.com>
Date: Tue, 8 Dec 2020 14:56:14 +0000
Message-ID: 
 <CA+mi_8YgOjGA3P9sBY4nVsE=_f6BbCi3OE-ZPz=LZoP5yEtedA@mail.gmail.com>
Subject: Re: Executing on the connection?
To: Denis Laxalde <denis.laxalde@dalibo.com>
Cc: psycopg@lists.postgresql.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, 8 Dec 2020 at 13:51, Denis Laxalde <denis.laxalde@dalibo.com> wrote:

> It did not strike me earlier, but it looks a bit weird to have
> connection.execute() return a "cursor" to read results while this
> "cursor" can also be used to execute commands. So, perhaps, another
> object, with only the interface for result retrieval would be more
> appropriate?

It thought about that, and it would be weird if people used a cursor
for further queries after consuming it, yes. But hey, who are we to
deny them to do that? :D I wouldn't go about creating and maintaining
new objects and interfaces only to limit the possibility.

My idea is that those who are interested in using `conn.execute()` are
exactly the people who don't care about having execute() on a cursor.
I think there's no harm in leaving it there :)

-- Daniele