Re: Using Index-only scans to speed up count(*)

Le samedi 7 juin 2014 09:09:00 Gurjeet Singh a écrit :
> On Sat, Jun 7, 2014 at 8:56 AM, Cédric Villemain 
 wrote:
> > Le samedi 7 juin 2014 08:35:27 Gurjeet Singh a écrit :
> >> PS: Please note that I am not proposing to add support for the
> >> optimizer hint embedded in Mitsuru's query.
> >> 
> > :-)
> 
> Even though I (sometimes) favor hints, and developed the optimizer
> hints feature in EDB (PPAS), I know how much Postgres **hates** [1]
> optimizer hints :) So just trying to wade off potential flamewar-ish
> comments.

There is a large benefits to users in preventing HINT in core: it makes 
it mandatory for PostgreSQL to keep improving, and it makes it mandatory 
for developers to find solutions around this constraint. There is at 
least planner_hint contribution (search Oleg website), and added to a 
postgresql hook on the parser you're done implementing HINT in userland.

I'm not arguing pro/cons about the feature (as you said the question has 
been answered already) but arguing that arbitrary constraints challenge 
us and produces good things for PostgreSQL in return.

-- 
Cédric Villemain +33 (0)6 20 30 22 52
http://2ndQuadrant.fr/
PostgreSQL: Support 24x7 - Développement, Expertise et Formation