Re: [Tracker] text search is implicitly always OR

[Jamie McCracken <jamiemcc blueyonder co uk>]

Samuel wrote:
> Also, as I said earlier, there's a difference when matching
> "wallpaper" vs "wallpapers", or "pictures" and "picture", or
> "accommodate" vs "accomodate". Ideally, there should be some sort of
> stemming (Porter, for example) and maybe the double metaphone
> algorithm to do fuzzier matching. Of course, the Porter stemming would
> bring up issues when using Tracker in other languages, though I think
> there are also algorithms for French and German (that I know of).

yeah until we have language independent stemming it would be tricky.

For now you can use the * wildcard at the end of a search term to get 
plurals and other related stuff

I dont really like the fuzzy sort of searches and they are best left to 
a dedicated indexer (which will do a better faster job)

Long term I may provide the option to store text contents in CLucene(1) 
which has all those extra bells and whistles for those with bigger 
machines (the duping of metadata but not contents and the increased size 
of lucene's indexes means a lot more disk space and RAM would be 
required though)

(1) http://clucene.sourceforge.net/index.php/Main_Page

(this would make a nice project for any budding c++ programmer)


--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/