Re: full text searching
Thomas T. Thai <tom@minnesota.com>
From: "Thomas T. Thai" <tom@minnesota.com>
To: Gunnar R|nning <gunnar@candleweb.no>
Cc: Mitch Vincent <mitch@venux.net>, pgsql-general@postgresql.org
Date: 2001-02-07T17:40:30Z
Lists: pgsql-hackers
On 7 Feb 2001, Gunnar R|nning wrote: > It is really not based on the FTI code in PostgreSQL, since with we started > out with a Progress database last year before porting to PostgreSQL. The > idea is the same though, a separate lookup table containing the words for > exact matching. Last time I had a look at the clients database it had about > 50-60K rows in the content table, which amounted to about ~3500K rows in > the lookup table. Searches return results instantly even though most of > them are joins involving 3-4 tables. The database(7.0.2) is running on a > Sun 220R with one 450MHZ processor, 10000RPM disks, 1GB RAM and Solaris > 7. (As a curiosity my P466 laptop with Linux is actually running PostgreSQL > faster...) > > Since we're only doing exact searches, the index is utilized. But the > quality isn't good enough - I would love to have language sensitive > searches. "car" should match "cars" but not cartography and "ship" > should/could match "boat" etc. you can use ispell prefix/suffix for searching base words.