Re: search engines

From: tj bandrowsky <tbandrow_at_unitedsoftworks.com>
Date: 4 Mar 2002 13:38:01 -0800
Message-ID: <52e031b9.0203041338.1de57271_at_posting.google.com>


>
> You are talking about loading the table into memory? Right?

No, inserting all of the keywords into the index.

>
> "Reference to a file"? Are you talking about pointers to binaries/text
> files?

Yes. You want to search for things, so you need the index for the search plus the content that will be found.

>

> I'm really talking about searches, not loading.

But, the point that I'm making is that populating the table is so horrifically slow with an RDBMS that you are best using either a separate indexing facility or an RDBMS that has its own bundled with it.

>
> I don't know what a 'bound' statement is.

Instead of:

INSERT NAME, AGE INTO PEOPLE VALUES ('TODD', 35 ) INSERT NAME, AGE INTO PEOPLE VALUES ('SUE', 23 ) you say something conceptually like

statement = INSERT NAME, AGE INTO PEOPLE VALUES (:1, :2 )

and then

statement[1]='TODD';
statement[2]=42;
statement.execute

statement[1]='SUE';
statement[2]=25;
statement.execute

This sort of thing can be faster. ADO calls this "parameterized queries", Perl calls it bind, but they all do the same thing.

> This seems more likely. How do you go about writing such an index
> scheme and how is it implmeneted? Is there documentation/literature on
> this. I'm using Informix.

I think an off the rack solution is probably better, and it depends on what you want. Still, if you wanted to roll your own full text index, I guess you'd start with trees and go from there...?? Received on Mon Mar 04 2002 - 22:38:01 CET

Original text of this message