Re: Query planner isn't using my indices - Mailing list pgsql-general

From Tom Lane
Subject Re: Query planner isn't using my indices
Date
Msg-id 14952.1010609404@sss.pgh.pa.us
Whole thread Raw
In response to Re: Query planner isn't using my indices  (Shaun Thomas <sthomas@townnews.com>)
List pgsql-general
Shaun Thomas <sthomas@townnews.com> writes:
> Can anyone explain this to me?  Anyone at all?  I mean, what logical
> reason is there to leave the database so crippled?

Feel free to step right up and fix it.  This is an open-source project.

Whining aside, the reason why these restrictions still exist is that
we have an extensible type system and so we don't really want to put
hard-wired knowledge about different types into the parser or optimizer.
The difficult part is to design a general-purpose, extensible framework
in which these desired conversions can be specified.

There have been prior discussions of how to do this (see the pghackers
archives); but no one's put forward a really satisfactory proposal yet.
If anyone had, it'd probably have been implemented by now --- there
isn't anyone here who likes the current behavior any more than you do.

            regards, tom lane

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: A Nit RE overlaps
Next
From: Tom Lane
Date:
Subject: Re: Question regarding VACUUM ANALYZE ERROR . . .