Bloom Filter

This JavaScript bloom filter implementation uses the non-cryptographic Fowler–Noll–Vo hash function for speed.

Usage

var bloom = new BloomFilter(
  32 * 256, // number of bits to allocate.
  16        // number of hash functions.
);
 
// Add some elements to the filter.
bloom.add("foo");
bloom.add("bar");
 
// Test if an item is in our filter.
// Returns true if an item is probably in the set,
// or false if an item is definitely not in the set.
bloom.test("foo");
bloom.test("bar");
bloom.test("blah");
 
// Serialisation. Note that bloom.buckets may be a typed array,
// so we convert to a normal array first.
var array = [].slice.call(bloom.buckets),
    json = JSON.stringify(array);
 
// Deserialisation. Note that the any array-like object is supported, but
// this will be used directly, so you may wish to use a typed array for
// performance.
var bloom = new BloomFilter(array, 16);

Implementation

Although the bloom filter requires k hash functions, we can simulate this using only two hash functions. In fact, we can use the same FNV algorithm for both hash functions, using only different base offsets for the two hashes.

Thanks to Will Fitzgerald for his help and inspiration with the hashing optimisation.