Bug to think about when working on the MARC database
Hi all Currently there is a bug/feature in koha when it comes to searching for titles that have a ' in them. For example on http://intradev.katipo.co.nz/catalogue/ Title search for asta's Finds asta's book. Which is fine... searching for astas finds nothing. The HLT librarians and suspect most librarians would like the a title search for astas to find asta's book. An idea I had was to have another field in the biblio table, called search title, which has the ' and - , plus stopwords such as and, the etc removed. This is the field searched . Then I thought we might not want to restrict this to just the title, we probably want it for author, and subtitle etc. So I was thinking of having a separate table, that we used for searching, which was keyed by biblionumber back to the biblio table and had entrys like biblionumber,entry,type where type = author, or subtitle or title etc. Does this fit with the work being done with MARC ? are there better ways to do this? Chris -- Chris Cormack chris@katipo.co.nz Programmer 025 500 579 Katipo Communications www.katipo.co.nz
On Tue, Jun 11, 2002 at 02:30:53PM +1200, Chris Cormack wrote:
So I was thinking of having a separate table, that we used for searching, which was keyed by biblionumber back to the biblio table and had entrys like biblionumber,entry,type where type = author, or subtitle or title etc.
Does this fit with the work being done with MARC ? are there better ways to do this?
Interesting, I was thinking of the same type of search table today. I don't know if there are better ways of doing it, but in my opinion it will be necessary for the MARC version. I'm still a little skeptical about the "pure MARC" idea, but I'm going with it on the assumption it will all magically work out. The idea on pure Marc is that Koha will retain exactly what was in a Marc record without "squashing it" to fit into another normalized set of entity relationships. The example with the apostrophe highlights the ongoing need to fit data into a normalized form. ISBN will be exactly the same. When a Marc record has an ISBN field of "0-395-12345-1 (pbk.)", a search on "039512345" must match. Presumably same for LCCN "GB74-123456" matching "74123456", and so on. The trick is that we want to have it both ways. If we dump the current biblio/biblioitems/items tables, I think we will simply have to replace them with another substitute like the (biblionumber,searchvalue,valuetype) idea instead. - Alan ---- Alan Millar --==> am12@bolis.com <==--
participants (2)
-
Alan Millar -
Chris Cormack