This bloom filter implementation:
1) Has a constructor that accepts a maximum comfortable number of members and maximum appropriate error (false positive) rate, and
   derives the fiddly bits from that; most bloom filter modules ask the enduser to specify the fiddly bits themselves.
2) Has a nice test suite, including checks for error rate.
3) Is in pure Python that'll run on CPython 2.x, CPython 3.x, PyPy or Jython.
4) Has a pair of simple, fast hash functions that give a good error rate - they're better than many of the alternatives.
   They're not Murmur or Jenkins, but the tests strongly suggest that they're working well.
5) Passes pylint and pep8.
6) Supports adding elements, testing for membership, and'ing sets and or'ing sets.

The code is derived from http://code.activestate.com/recipes/577686-bloom-filter/ and inherits that code's license (MIT).

For more about Bloom Filters:
http://en.wikipedia.org/wiki/Bloom_filter
http://spyced.blogspot.com/2009/01/all-you-ever-wanted-to-know-about.html