Date: Thu, 12 Oct 2006 14:34:15 -0500
From: "Jim C. Nasby" <jim@nasby.net>
To: Jeff Davis <pgsql@j-davis.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: Hints proposal
Message-ID: <20061012193414.GV28647@nasby.net>
References: <20061012151439.GT28647@nasby.net>
	<1160671375.31966.84.camel@dogma.v10.wvs>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1160671375.31966.84.camel@dogma.v10.wvs>
User-Agent: Mutt/1.5.11

On Thu, Oct 12, 2006 at 09:42:55AM -0700, Jeff Davis wrote:
> On Thu, 2006-10-12 at 10:14 -0500, Jim C. Nasby wrote:
> > The syntax these hints is something arbitrary. I'm borrowing Oracle's
> > idea of embedding hints in comments, but we can use some other method if
> > desired. Right now I'm more concerned with getting the general idea
> > across.
> > 
> 
> Is there any advantage to having the hints in the queries? To me that's
> asking for trouble with no benefit at all. It would seem to me to be
> better to have a system catalog that defined hints as something like:
> 
> "If user A executes a query matching regex R, then coerce (or force) the
> planner in this way."
> 
> I'm not suggesting that we do that, but it seems better then embedding
> the hints in the queries themselves.

My experience is that on the occasions when I want to beat the planner
into submission, it's usually a pretty complex query that's the issue,
and that it's unlikely to have more than a handful of them in the
application. That makes me think a regex facility would just get in the
way, but perhaps others have much more extensive need of hinting.

I also suspect that writing that regex could become a real bear.

Having said that... I see no reason why it couldn't work... but the real
challenge is defining the hints.
-- 
Jim Nasby                                            jim@nasby.net
EnterpriseDB      http://enterprisedb.com      512.569.9461 (cell)