As you may remember, at Greplin we have built some of our search features on top of the excellent Lucene project. As avid users, we've built a fair number of tools that help us use Lucene to the fullest. Today we're happy to announce that we'll be open sourcing a few more of them in the greplin-lucene-utils GitHub project.

Some noteworthy features include:

• A class that construct BooleanQueries in an Effective Java-style Builder pattern 
• A query that matches no documents (sounds odd, but it's come in handy more than once for us)
• A query that matches phrase prefixes - for example "Epic w" will match both "Epic win" and "Epic wonder". This is particularly useful for implementing Google Instant style searches
• A Collector that collects all matching documents in no specified order.
• A Collector that only collects the first matching document.
• A Collector that counts all matching documents.

We're also excited to introduce an open source persistent Bloom Filter implementation - the greplin-bloom-filter project. Storing 50,000 2.5KB items in a traditional hash set requires over 125MB. But if you're willing to accept a 1-in-10,000 false positive rate on lookups, a comparable counting bloom filter requires under 500KB (over 250x space savings) - see our benchmark for details. Some noteworthy implementation details:

• Instead of using N-distnct hashes, we use a linear combination of two hashes (as pioneered by Kirsch and Mitzenmacher)
• Our base hash is a Java implementation of Appleby's MurmurHash (maybe an ambitious contributor can try to build a Java implementation of Google's CityHash though!)
• We use 4-bit counting buckets, which means items can be deleted from the BloomFilter
• Our bloom filters can be efficiently written to disk to support persistence

As always, pull-requests, issue reports, and comments are welcome!