Rust implementation of a fast, discrete, bounded,
Zipf-distributed random
number generator. Compared to the implementation provided by
randomkit (which binds
to NumPy's fork of RandomKit), this crate is approximately twice as
fast:
console
$ cargo +nightly bench
test tests::bench_randomkit ... bench: 344 ns/iter (+/- 15)
test tests::bench_us ... bench: 66 ns/iter (+/- 2)
test tests::bench_threadrng ... bench: 9 ns/iter (+/- 0)
It is also both driven by, and provides, a Rust random number generator.
This implementation is effectively a direct port of Apache Common's RejectionInversionZipfSampler, written in Java. It is based on the method described by Wolfgang Hörmann and Gerhard Derflinger in Rejection-inversion to generate variates from monotone discrete distributions from ACM Transactions on Modeling and Computer Simulation (TOMACS) 6.3 (1996).