HyperLogLog (HLL) is a probabilistic estimator of the cardinality of a stream of values. Given a bounded amount of memory, it can estimate the cardinality of a stream with bounded relative error and it is possible to trade off memory usage for precision. Formally, the standard error for an HLL with n registers is less than 1.04/sqrt(n).In order to count items, they must first be hashed. The hash() function provides a suitable hash. Its output is an array of four 32 bit postive integers, which, taken together constitute the complete hash of the input string. Currently the implementation is MurmurHash3-128.