Re: Fastest char datatype - Mailing list pgsql-performance

From Robert James
Subject Re: Fastest char datatype
Date
Msg-id e09785e00907200623o11237fa9s1ad983d180841d66@mail.gmail.com
Whole thread Raw
In response to Re: Fastest char datatype  (Peter Eisentraut <peter_e@gmx.net>)
List pgsql-performance
Is there a way to use a more compact encoding? I only need 4 bits per char - that would certainly help caching.  (I have indexes tuned very well, already).

On Mon, Jul 20, 2009 at 2:02 AM, Peter Eisentraut <peter_e@gmx.net> wrote:
On Monday 20 July 2009 04:46:53 Robert James wrote:
> I'm storing a lot of words in a database.  What's the fastest format for
> finding them? I'm going to be doing a lot of WHERE w LIKE 'marsh%' and
> WHERE w IN ('m', 'ma').  All characters are lowercase a-z, no punctuation,
> no other alphabets.  By default I'm using varchar in utf-8 encoding, but
> was wondering if I could specificy something else (perhaps 7bit ascii,
> perhaps lowercase only) that would speed things up even further.

If your data is only lowercase a-z, as you say, then the binary representation
will be the same in all server-side encodings, because they are all supersets
of ASCII.

pgsql-performance by date:

Previous
From: Marcin Stępnicki
Date:
Subject: Re: Full text search with ORDER BY performance issue
Next
From: Robert James
Date:
Subject: Re: Can Postgres use an INDEX over an OR?