IRC Discussion about fulltext searches: http://www.haz.cmsd.bc.ca/cgi-bin/kohalog.pl?date=15Jan03&refresh=&firstline=24&lastline=90&eventid=18 The one caveat I found is that search terms of three characters or less are discarded. So if you do a fulltext search for "rabbit fur" you will get results regarding rabbits, but with no relevance placed on the word fur. Also, doing a fulltext search for the word "fur" will always return 0 hits. Searching for "furs" may return some hits, because it is a four letter word. In my current setup, I am using "like" searches if one or more search terms are three letter words, and fulltext searching if all search terms are longer than 3 letters. Just something to be aware of. The database I'm working on has about 600,000 records (almost 2,000,000 items), and initial "like" style searches complete in about 10 seconds or less, while fulltext searches complete in under 2 seconds. I don't know how fair this comparison really is as I used "car" for my like-style query nd "cars" for my fulltext query, but it gives you an idea of the difference in speed. I've also got caching working, so subsequent searches on the same term (ie looking at page 2 of the results) complete in under a second. Steve.
Tonnesen Steve a écrit:
IRC Discussion about fulltext searches: http://www.haz.cmsd.bc.ca/cgi-bin/kohalog.pl?date=15Jan03&refresh=&firstline=24&lastline=90&eventid=18
The one caveat I found is that search terms of three characters or less are discarded. So if you do a fulltext search for "rabbit fur" you will get results regarding rabbits, but with no relevance placed on the word fur. Also, doing a fulltext search for the word "fur" will always return 0 hits. Searching for "furs" may return some hits, because it is a four letter word.
In my current setup, I am using "like" searches if one or more search terms are three letter words, and fulltext searching if all search terms are longer than 3 letters.
good idea. I'll use it !
Just something to be aware of.
yep, thx to point this, which i ignored !
The database I'm working on has about 600,000 records (almost 2,000,000 items), and initial "like" style searches complete in about 10 seconds or less, while fulltext searches complete in under 2 seconds. I don't know how fair this comparison really is as I used "car" for my like-style query nd "cars" for my fulltext query, but it gives you an idea of the difference in speed. I've also got caching working, so subsequent searches on the same term (ie looking at page 2 of the results) complete in under a second.
More than 2seconds => a user find it's "too long". I think that 10 seconds is really too long for a opac member, which can't be aware of the technical constraints :-) -- Paul POULAIN Consultant indépendant en logiciels libres responsable francophone de koha (SIGB libre http://www.koha-fr.org)
participants (2)
-
paul POULAIN -
Tonnesen Steve