Unless Bill wants to add this feature, I think it's likely beyond the 
scope of the script. Thomas is right in suggesting that web log anaysis 
is really the starting place. But only the starting place. It would be 
easy to log swish.cgi use to another log file, as to whether queries are 
successful or not, but what you *do* with that information is way beyond 
the scope of that script. That's really a whole other issue (mess) that 
requires some kind of intelligent semantic analysis. Still, I've found 
it useful to look at logs like that to identify common misspellings, etc.

I personally believe that good log analysis might lend itself to some 
good swish-e feature in the future (like Bill's suggestion for weighting 
rank higher based on proximity or 'virtual phrases'), but swish.cgi (and 
swish-e) just lets you retrieve *data* -- not *information*. It's up to 
you to analyze the logs and deduce whatever patterns you may see in the 
tea leaves...


John Wentworth wrote on 7/16/04 12:11 PM:

> Thomas Dowling wrote:
> 
> 
>>I basically like the contents of the swish.cgi script for search
>>and I do not want to modify it alot but I would like to add one feature =
> 
> =3D
> 
>>I would like the drop the search terms and the result of the search =
> 
> into =3D
> 
>>a database to analyze what our customers are searching for
>>I don't really know perl but I figure I should be able to hack =
> 
> something =3D
> 
>>together with some tutorials I found
>>what I basically need to information as to which perl variables in this =
> 
> =3D
> 
>>script correspond to that information and where would be a good place =
> 
> to =3D
> 
>>insert this code since I'm having trouble following the code
>>
>>=20
>>
> 
> 
> I don't use swish.cgi myself, so apologies if I'm off base.  But if=20
> swish.cgi uses HTTP GET requests, the search terms have probably in your =
> 
> web transaction log all along.  Log analysis software, a simple Perl=20
> script, or grep could all fish them out for you without editing the CGI=20
> script.  I'm guessing you're looking for lines that match something like =
> 
> "GET /somewhere/swish.cgi?query=3Dmy+search+terms....".
> 
> 
> Thomas Dowling
> tdowling@ohiolink.edu
> 
> This is a good suggestion however doesn't fully solve the problem =
> because it doesn't record whether the search was successful or not. =
> Which is why I thought I had to try to accomplish this in the cgi =
> script.
> 
> 
> 
> *********************************************************************
> Due to deletion of content types excluded from this list by policy,
> this multipart message was reduced to a single part, and from there
> to a plain text message.
> *********************************************************************

-- 
Peter Karman - Software Publications Engineer - Cray Inc
phone: 651-605-9009 - mailto:karman@cray.com