go-bloom

go-bloom is a collection of bloom filters for Go, including a standard bloom filter, a counting bloom filter, and a layered bloom filter (for counting occurrences of the same item.) All of them use the 64-bit FNV-1a hash function and an efficient bitset.

A bloom filter is a space-efficient probabilistic data structure that is used to test whether an element is a member of a set. False positives are possible, but false negatives are not; i.e. a query returns either “inside set (may be wrong)” or “definitely not in set”. Elements can be added to the set, but not removed (though this can be addressed with a counting filter.)

Installation

go get github.com/patrickmn/go-bloom

Usage

// Normal bloom filter

// Create a bloom filter which will contain an expected 100,000 items, and which
// allows a false positive rate of 1%.
f := bloom.New(100000, 0.01)

// Add an item to the filter
f.Add([]byte("foo"))

// Check if an item has been added to the filter (if true, subject to the
// false positive chance; if false, then the item definitely has not been
// added to the filter.)
f.Test([]byte("foo"))


// Counting bloom filter

// Create a counting bloom filter which will contain an expected 100,000 items,
// and which allows a false positive rate of 1%.
f := bloom.NewCounting(100000, 0.01)

// Add an item to the counting filter
f.Add([]byte("foo"))

// Remove an item from the counting filter
f.Remove([]byte("foo"))


// Layered bloom filter

// Create a layered bloom filter which will contain an expected 100,000 items,
// and which allows a false positive rate of 1%.
f := bloom.NewLayered(100000, 0.01)

// Add an item to the layered filter
f.Add([]byte("foo"))

// Add an item to the layered filter again
f.Add([]byte("foo"))

// Find out how many times an item appears in the layered filter
count := f.Test([]byte("foo")
count == 2

To use go-bloom in multiple goroutines, use a sync.RWMutex, and surround test calls with RLock()/RUnlock(), and set calls with Lock()/()Unlock.

go-bloom is based on bloom by William Fitzgerald.