Re: Very slow search using basic pattern matching

On 18 May 2004, at 17:12, Bruno Wolff III wrote:

> On Tue, May 18, 2004 at 16:47:11 +0100,
> Dan Field <dof(at)llgc(dot)org(dot)uk> wrote:
>> FROM
>> da_records
>> WHERE
>> da_records.DESCRIPTION_CY ~* '.*Aberystwyth*.'
>> OR
>> da_records.TITLE_CY ~* '.*Aberystwyth*.'
>> limit 100
>>
>> Is there a better way of matching the string? (Must be case
>> insensitive)
>
> This is going to require a sequential scan and if you aren't going to
> hit 100 entries near the beginning of the table, you will be
> effectively
> doing a full scan of the table.
>
> You might try looking at the tsearch2 contrib entry (a package for
> full text
> searching) as that will likely be able to help you out.

OK It looks like the full text search is the way forward here, but I
can only find the tsearch.sql file in contrib. I'm using RedHat EL 3.0
Advanced Server with PG version 7.3-RH.

Is there any documentation available on how to go about creating these
indexes on my text fields?

I'm at a bit of a loss here, Google is failing me sadly

--
Dan Field <dof(at)llgc(dot)org(dot)uk> - Support Programmer: Cymru ar y we
cy_GB: http://www.cymruarywe.org
en_GB: http://www.walesontheweb.org