I've spent alot of time looking through various approaches at FTI. The biggest problem with Sphinx is that it requires you to provide a unique 32 or 64bit integer as a document id. The impedance mismatch with CouchDB's arbitrary string document id's has provided a bit of a blocker for full on integration with Sphinx in the arbitrary case.
I've got entirely too many different attempts for full text indexing on github [1], but the real one to keep an eye on is probably Robert Newson's fork of my CouchDB-Lucene [2] branch. He's taken it quite a long way and as soon as he's integrated Rhino I'm gonna close up all of my rather half baked attempts. HTH, Paul Davis [1] http://github.com/davisp [2] http://github.com/rnewson/couchdb-lucene/tree/master On Tue, Feb 24, 2009 at 10:28 AM, Adam Groves <[email protected]> wrote: > Hi all, > > I read somewhere on the mailing list that Sphinx could be used to > provide fti for couchdb views. Has anyone implemented this? Would you > recommend it? Care to share some details? :) > > Cheers > > Adam >
