hllpp is an implementation of the HyperLogLog++ cardinality estimation algorithm in go. It optimizes for memory usage over CPU usage. It implements all the HyperLogLog optimizations introduced in the HyperLogLog++ paper (http://goo.gl/Z5Sqgu). Some notable features include:
- marshaling so you can serialize to your datastore
- extra space savings by only using 5 bits per register when possible
- built-in non-streaming murmur3 implementation for fast hashing of input data
h := hllpp.New()
h.Add([]byte("barclay"))
h.Add([]byte("reginald"))
h.Add([]byte("barclay"))
h.Add([]byte("broccoli"))
fmt.Println(h.Count())
// Output: 3
See the godocs for documentation and more examples.
This package is a go or golang implementation of HyperLogLog or HyperLogLog++. It doesn't show up when you search for golang hyperloglog, so I am repeating the words golang hyperloglog in the README. golang hyperloglog.