I think the issue is the nature of the information you're dealing with. As long as the queries are closely interconnected with the relations you've defined, the SQL queries to an RDMBS will be better. But for free-form text or other unstructured data, I think it makes to view your data as it's presented, rather than as it's stored.
When your users want to find the info in a page context, rather than in a collection of structured records, then you need to create another way of viewing the info. You could create a VIEW (if your RDBMS supports this) and query the view. Or build a separate index in swish-e or some other tool.
On Tuesday, October 19, 2004, at 08:55 AM, Roland Dumas wrote:
Hmmmm....
Looking at swish-e, it's clear that it's a more powerful approach, but...
If I want to keep life simple and have everything running from witango -
even if I won't get all the swish-e bells and whistles, is there a better
approach? Or am I hearing that witango isn't really a good tool for this
task?
On 10/19/04 8:15 AM, "Bill Conlon" <[EMAIL PROTECTED]> wrote:
Then, maybe spidering the forum content is the easiest way to do this. It will let you do free form searching. The cgi script included with swish-e will highlight found terms, and let the user apply various restrictions.
The other thing is you don't have to worry about a SQL injection attack.
On Tuesday, October 19, 2004, at 08:06 AM, Roland Dumas wrote:
It's a search function for a forum. Dynamic popups won't work here,
because
it's a search of subject and content, which would overwhelm a selection
menu.
My thoughts were like this:
User's Search string = first argument
Search string parsed (by space and comma) and the articles tossed out.
That
leaves an array of words within the first argument. The first argument
and
the remaning substrings comprise all the OR conditions you want.
Option 1: Generate SQL from a <@ROWS> that just appends a series of OR statements to the SELECT command.(easiest to do, but least secure)
Option 2: Write a taf in XML, using the <@ROWS> to create a custom <Criteria> section in a temporary taf that is just the search action, and then call that action with a branch/return. (typos will crash tango server - venturing into deep unknown)
Option 3: Do a For loop for each of the substrings and glue all the resultsets together (slow and painful)
Then, when you've got the amalgamated found set of records with whole string or substrings, figure a way to bubble up to the top the whole strings or items of greater value.
On 10/19/04 7:13 AM, "Bill Conlon" <[EMAIL PROTECTED]> wrote:
Full text indexing can be expensive if your application does a lot of
inserts/updates into columns that are indexed, but things like the
winery/varietal shouldn't be a problem if you want to get it out of
the
db.
If it's a problem for your users, then maybe you can build a selection
list from the available choices. Or maybe you need something akin
auto-complete: run a javascript keyboard event handler that populates
your input field based on the characters typed so far.
On Tuesday, October 19, 2004, at 06:59 AM, John McGowan wrote:
Don't some newer databases have full text indexing now. (I believe MSSQL calls the feature "Full-Text Search") Wouldn't the best solution be to use a database that supports that type of searching?
If this functionality isn't available to you in your DB then I would
suggest you still use swish-e like Bill suggests...
1. create a "dummy site" that will have a unique page for every record in the table that you're looking for. www.mysite.com/dbindex/main.taf 2. when you hit main.taf it generates a link to each record in the table you care about www.mysite.com/dbindex/main.taf?id=xxxxx www.mysite.com/dbindex/main.taf?id=yyyyy www.mysite.com/dbindex/main.taf?id=zzzzzz
(if you're familiar with witango this should take you about 5 minutes
to accomplish)
3. Tell swish-e to index the site by hitting the initial main.taf url.
4. now when you want to do a full text search of the table, you call
swish-e's searching functionality. it will return a list of the
matching entries.
www.mysite.com/dbindex/main.taf?id=aaaaa
www.mysite.com/dbindex/main.taf?id=bbbbb
www.mysite.com/dbindex/main.taf?id=cccccc
5. Of course at this point you know that if you strip out the
"www.mysite.com/dbindex/main.taf?id=" You will have the part of the
url that you care about, the aaaaa,bbbbb,ccccc which should be in a
ranked order, and now you can do with that information whatever you
want.
6. Schedule the re running of step 3. at some interval that satisfies
your need for accuracy vs. performance.
Of course this all assumes you're doing this for 1 particular table.
However, if you had more than 1 table you could still do it all by
adding a little more code to your main.taf and some more logic to the
part that stripps the url to get the important part.
/John
Roland Dumas wrote:
___________________________________________________________________ __But, we're talking about a search of a database.
On 10/18/04 5:59 PM, "Bill Conlon" <[EMAIL PROTECTED]> wrote:
Roland,
You've heard this from me before on this list. Take a look at
swish-e.
You could use its built-in spider to index your site, and then use
the
built-in cgi-script to highlight your results. It's really a great
piece of software.
Now if you take the swish-e approach, here's what I would do to solve this.
Dyanmically create metatags for the key parameters you want to search:
<meta name="vineyard" content="Chateau Lafite, Chateau, Lafite"> <meta name="varietal" content="Pinot Noir, Pinot, Noir"> etc.
Use witango to tokenize while creating the HTML pages for the various wines.
Then use swish-e's meta name search.
On Monday, October 18, 2004, at 05:39 PM, Roland Dumas wrote:
In search engines, when you submit a search string, the search_________________________________________________________________ __
engine
first
tokenizes and then searches for each substring string separately
and
then
brings them together as your found set. So if I search for 1961
Chateau
Lafite, I'll get items with 1961, others with Chateau or Chateu
Lafite, and
on top will be the found records with 1961 Chateau Lafite (I know,
if
you
put it in quotes, it forces it to find only the whole string. That
part is
easy)
They will also rank a find of the full set of terms above ones with one or two terms in the documents.
Questions:
What's the approach with witango that will enable the search of tokenized strings.
Any ideas on how to do a crude ranking, such that the full term comes up on top of the found set?
________________________________________________________________ __
__
___
_
TO UNSUBSCRIBE: Go to
http://www.witango.com/developer/maillist.taf
__
___
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
----------------------------------------- Roland Dumas Roberts Information Services 310 W. Bellevue Avenue San Mateo CA 94402 650-347-1373 415-412-9300 (cell) [EMAIL PROTECTED] SMS: http://new.servqual.com/html/sms.tml
__________________________________________________________________ __
__
__
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
__
_
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
____________________________________________________________________ __
__
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
----------------------------------------- Roland Dumas Roberts Information Services 310 W. Bellevue Avenue San Mateo CA 94402 650-347-1373 415-412-9300 (cell) [EMAIL PROTECTED] SMS: http://new.servqual.com/html/sms.tml
_____________________________________________________________________ __
_
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
______________________________________________________________________ __
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
----------------------------------------- Roland Dumas Roberts Information Services 310 W. Bellevue Avenue San Mateo CA 94402 650-347-1373 415-412-9300 (cell) [EMAIL PROTECTED] SMS: http://new.servqual.com/html/sms.tml
_______________________________________________________________________ _
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
________________________________________________________________________ TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
