Skip to main content.
home | support | download

Back to List Archive

Re: Indexing DB data

From: Bill Moseley <moseley(at)not-real.hank.org>
Date: Wed Mar 22 2006 - 13:41:15 GMT
On Wed, Mar 22, 2006 at 04:54:05AM -0800, Julian Perry wrote:
> 
> We have a product database (about 1.5m products) - currently
> in Oracle, in simple terms, like this:
> - product_id
> - price
> - description
> - category
> - shop_location
> 
> So a query might be:
> - find all products including the words:
>    "green hat" where shop_location is
>    "england" and category is "clothing"
> 
> The free-text searching element is not suited to Oracle
> and so I was considering a structured query perhaps
> using meta-tags?  I was figuring that I could define
> price, category and shop_location as meta-tags.
> 
> I would guess there is no way to do a price range,
> as in:
> - products between 20 and 50 dollars
> Is that right?

Kind of.

    swish-e -L price 20 50 description="(red or blue) coat"

> Another question ... if I want to spit out data
> from Oracle and effectively build a dummy page
> for each record for swish-e to then index - is
> that feasible for 1.5m records?  I guess I would
> use something like MySQL.pl from prog-bin?

Pushing it.  The searches would probably be ok, but indexing might
take a while, so changes to the database would take a while to show
up in searches.  Swish has an incremental option that has worked for
some, you might try that, but it's not that well tested.

And the -L limit is a hack and not really a designed in feature.
It works by sorting an integer table the size of your complete index
- twice - for each search.  Not very efficient.

It might be faster to have swish return a list of possible results,
and if less than 100 or so then pass those to the database and with a
WHERE to limit.  You would just need to test.

Sure there aren't any full-text options for Oracle?

If you do some testing, it would be great for you to post back here
so others with similar situation can benefit.

Thanks,



-- 
Bill Moseley
moseley@hank.org

Unsubscribe from or help with the swish-e list: 
   http://swish-e.org/Discussion/

Help with Swish-e:
   http://swish-e.org/current/docs
   swish-e@sunsite.berkeley.edu
Received on Wed Mar 22 05:41:17 2006