Hydra 1.2

A high performance persistent bloom filter

Hydra: The Python Bloom Filter.

Compile with Cython 0.13 or higher. That means trunk as of 2010/04/21.

You should get no warnings, errors or other crap when compiling this module. If you do - something has gone wrong.

—

Hydra is a high performance bloom filter. It’s basically a port of the Cassandra bloom filter with some fun Cython hackery.

1) It’s persistent using memory mapped io. On Linux, the mmap uses the MAP_POPULATE flag so the entire file is loaded into kernel space virtual memory. In other words - fast.

2) The hash function uses the MurmurHash2A algorithm. I’m using the original code base in raw C - so it’s blazing fast and has excellent key distribution and avalanche properties.

3) The filter exports a set-like interface. Use .add(..), .contains() or use the “in” operator.

4. Correctness is built-in - with test cases.

5) Did I say it’s fast? It’s stupidly fast. 200,000 lookups per second on my 1.6Ghz netbook.

The filter supports periodic forced synchronization to disk using fdatasync(), or you can just let the deallocator flush everything to disk when your filter goes out of scope, or your process terminates.

Hydras are snakes with multiple heads. They’re also bad dudes with snake logos on their chest who regularly try to beat on Nick Fury. Now it’s a bloom filter.

Mostly, I couldn’t bear to make this yet another PySomeLibraryName library.

Running tests:

Install nose.

Go into the tests directory and do :

nosetests