I’m interested in working with Lucene from a Cocoa application. I’m aware that there are many ways to do this, but my question is, “which way is best?” My investigations so far:
LuceneKit is an Objective-C port of Lucene, but is based on a version of Lucene that is ancient at this point, and in trying to use it, I’ve run into several major issues from the get go. (Improper subclass of NSDate; A basic query that works in Luke doesn’t work with LuceneKit;) It appears to be a non-starter.
CLucene looked like it might be viable, but it fails a bunch of it’s own tests on build, including an intermittent concurrency related problem where half the time I run the tests they deadlock. Not inspiring. This still may be the answer, but I’m very nervous considering my experience just building it and running its own tests.
Current Apache Lucene via JNI – Having simply never called a Java library from C, I’m unsure what’s involved here. I certainly feel like the official Apache-curated incarnation of Lucene is likely to be the most mature and functional, but having not done the C <-> Java JNI thing before, I’m unclear how the effort involved would compare to working with CLucene.
Maybe there are other options. I’m not necessarily looking for a first-class Objective-C interface (although I wouldn’t turn one down, either) just something functional and hopefully reasonably mature and reasonably performant. Anyone have any sage advice?
From my experience using JNI (although, not with Lucene), it’s not too tricky to get something simple working, but you can wind up writing a lot of fairly monotonous code wiring everything up.
Another option you may want to consider is JCC, which is used by the PyLucene project to generate a boilerplate C++ wrapping around the JNI itnerface, which they then use to wrap a Python API around.