The tf-random package
This package contains an implementation of a high-quality splittable pseudorandom number generator. The generator is based on a cryptographic hash function built on top of the ThreeFish block cipher. See the paper Splittable Pseudorandom Number Generators Using Cryptographic Hashing by Claessen, Pałka for details and the rationale of the design.
The package provides the following:
A splittable PRNG that implements the standard System.Random.RandomGen class.
The generator also implements an alternative version of the System.Random.TF.Gen.RandomGen class (exported from System.Random.TF.Gen), which requires the generator to return pseudorandom integers from the full 32-bit range, and contains an n-way split function.
An alternative version of the Random class is provided, which is linked to the new RandomGen class, together with Random instances for some integral types.
Two functions for initialising the generator with a non-deterministic seed: one using the system time, and one using the /dev/urandom UNIX special file.
The package uses an adapted version of the reference C implementation of ThreeFish from the reference package of the Skein hash function (https://www.schneier.com/skein.html), originally written by Doug Whiting.
Please note that even though the generator provides very high-quality pseudorandom numbers, it has not been designed with cryptographic applications in mind.
|Versions||0.1, 0.2, 0.3, 0.4, 0.5|
|Dependencies||base (>=4.2 && <5), primitive (>=0.3), random, time [details]|
|Author||Michał Pałka <email@example.com>|
|Maintainer||Michał Pałka <firstname.lastname@example.org>|
|Source repository||head: darcs get http://hub.darcs.net/michal.palka/tf-random|
|Uploaded||Wed Apr 9 13:57:37 UTC 2014 by MichalPalka|
|Distributions||Arch:0.5, Debian:0.5, Fedora:0.5, FreeBSD:0.5, LTSHaskell:0.5, NixOS:0.5, Stackage:0.5|
|Downloads||109724 total (60 in last 30 days)|
|Status||Docs available [build log]|
Successful builds reported [all 1 reports]
For package maintainers and hackage trustees