Link membase to sphinx

I was wondering if anyone ever tried to link membase's (or should I say couchbase's now?) TAP output protocol to sphinx (the open source full te开发者_如何学Goxt search engine).

How did you do it? and how well did it work? Any good references on the web (I can't seem to find any)

Thanks!

Regards, ldx

I've never actually done this before, but it looks like sphinx has an xml piping mechanism. They put this in place so people wouldn't have to write a new data source drivers themselves.

More info on xmlpipe: http://sphinxsearch.com/docs/2.0.1/xmlpipe.html

What you could do is grab one of the java tap implementations. One is solely a tap project.

https://github.com/mikewied/jtap

and the other is the latest version of spymemcached (which at the moment is a pre-release with tap support)

You can find it on the Couchbase website here:

http://files.couchbase.com/maven2/spy/spymemcached/2.7/

I would use the spymemcached one though because it tracks topology changes and with jtap you would have to start a separate tap stream for each node. Both are very similar though.

The best thing to do is probably to start up a tap stream and then create little xml snippets like in the example on the sphinx page I linked to above. Then you could send the xml over to sphinx pipe for indexing.

Also, if you have any problems with the tap clients just post to the Couchbase forums. They are pretty quick at answering questions.