Re: Dynamic Partial Index - Mailing list pgsql-general

From Jeff Davis
Subject Re: Dynamic Partial Index
Date
Msg-id 1155861311.26280.19.camel@dogma.v10.wvs
Whole thread Raw
In response to Re: Dynamic Partial Index  (gustavo halperin <ggh.develop@gmail.com>)
Responses Re: Dynamic Partial Index  (gustavo halperin <ggh.develop@gmail.com>)
List pgsql-general
On Fri, 2006-08-18 at 00:19 +0300, gustavo halperin wrote:
> > Create an index on the table, and then periodically move records into a
> > separate archive table.
> >
> > Regards,
> >     Jeff Dave
> Thanks, but I have a question. If the table is a BIIIIIIG table, use
> your solution is still a good idea ?? What about to create a partial
> INDEX for the really current date (and not using the function
> current_date) and periodically dropped and created it with the current
> day again and again ??
>

If you continue to drop and recreate an index like that, it will need to
scan the table during the creation of the index. That will take a while
on a large table.

You might be better off just using a normal index. To search the index
only takes log(n) time. What problem are you currently having with a
normal index?

The downside of a normal index on a large table is that the index will
grow large and consume space. I think the solution is to move seldom-
accessed records to a separate archive table. That way, you don't ever
have to scan the archive table unless you do a search in the archives.

Regards,
    Jeff Davis



pgsql-general by date:

Previous
From: Chris
Date:
Subject: Re: count and limit
Next
From: Tom Lane
Date:
Subject: Re: select * from users where user_id NOT in (select