Re: Chatter on DROP SOMETHING IF EXISTS - Mailing list pgsql-hackers

From Bruce Momjian
Subject Re: Chatter on DROP SOMETHING IF EXISTS
Date
Msg-id 200702180202.l1I22Xc13630@momjian.us
Whole thread Raw
In response to Re: Chatter on DROP SOMETHING IF EXISTS  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: Chatter on DROP SOMETHING IF EXISTS  (Alvaro Herrera <alvherre@commandprompt.com>)
List pgsql-hackers
Tom Lane wrote:
> "Jim C. Nasby" <jim@nasby.net> writes:
> > On Thu, Feb 08, 2007 at 01:54:13PM -0500, Tom Lane wrote:
> >> I would be satisfied if the returned command tag were something else,
> >> maybe "NO OPERATION".
> 
> > "TABLE blah DID NOT EXIST" might be less confusing...
> 
> You're confusing a command tag with a notice.  In the first place,
> we shouldn't assume that applications are ready to deal with
> indefinitely long command tags (the backend itself doesn't think they
> can be longer than 64 bytes); in the second place, they should be
> constant strings for the most part so that simple strcmp()s suffice
> to see what happened.  Command tags are meant for programs to deal
> with, more than humans.

Yep.  Because IF EXISTS is in a lot of object destruction commands,
adding a modified tag seems very confusing, because in fact the DROP
TABLE did succeed, so to give any other tag seems incorrect.

I think the only option would be to use INFO instead of NOTICE, but
because the output is optional based on whether the object exists, you
might say NOTICE is the right level.

I am afraid we might just need to live with the current behavior.

--  Bruce Momjian  <bruce@momjian.us>          http://momjian.us EnterpriseDB
http://www.enterprisedb.com
 + If your life is a hard drive, Christ can be your backup. +


pgsql-hackers by date:

Previous
From: Lukas Kahwe Smith
Date:
Subject: Re: Plan invalidation design
Next
From: Tatsuo Ishii
Date:
Subject: Re: Re: [GENERAL] [ANNOUNCE] Advisory on possibly insecure security definer functions