CheapRandom is a set of tools for pseudo random number generation from arbitrary data. The properties of the CheapRandom seed make convenient random number generation possible -- useful for easily repeatable software testing. The CheapRandom algorithm is information conserving and generally appears to produce lower chi-squared statistics than Kernel::rand i.e. it appears to be more random. The CheapRandom algorithm, an original work by Bardi Einarsson, has been in use for 6 years.

Why should one use CheapRandom?

Simple and powerful: The CheapRandom algorithm is fast, fast enough to be practical to use in ruby as ruby. The properties of the CheapRandom seed (see below) make management of seeds and random data easy and verifiable.


v0.9.4 with comprehensive tests, developed using ruby 1.9.3 and rspec 2.10.0 is the same as v0.9.1 (gem v0.9.2) aside from the addition of the examples2 directory (for gem v0.9.3) and the examples3 directory (for MS Windows).


Get the v0.9.4 gem or clone the repository. Create a directory called random in the repository root. Use examples/make_seed.rb to make a the.seed file - (a 256 byte permutation file) from arbitrary data using the CheapRandom default seed.

Usage and documentation

Study the programs in the examples and spec directories and use the .rb files in the lib directory. See below for more information. Additionally, the examples2 directory contains tools for the crx bash script, using the c programs cr and rpb, for a very convenient way of working with cheap random files.


Copyright (c) 2006 - 2012 Bardi Einarsson. Released under the MIT License. See the LICENSE file for further details.

Create a large file of random bytes

Copy (or link) some large file (with a known hash) to the random directory. Use examples/cr.rb to randomize it into a .random file (the rspec tests expect and

Use a large file of random bytes as a source of random numbers

Use examples/cb.rb to see how to use lib/cheap_bits.rb to create a random number generator which uses that large file. spec/using_cheap_bits_cheap_random_spec.rb uses lib/cheap_bits.rb to generate random numbers. It should be compared with spec/cheap_random_spec.rb which uses Kernel::rand.

Manage a large file used as the source of random numbers

The large file does not need to be kept if it is something like For example, one can keep a reference to its internet location, its sha1 hash and the CheapRandom seed file used when creating its .random file.

Manage the seed file used when generating .random files

The CheapRandom seed file does not need to be kept if it is generated from the CheapRandom default seed and some arbitrary file, say a picture of a pet cat. One can simply keep the picture of the pet cat.

CheapBits#random(n) in lib/cheap_bits.rb

random(n), where n is an integer greater than zero, behaves like Kernel::rand(n). It is only dependent on the .random file, not on how the .random file was generated. Note that the .random file is treated like a ring buffer of random bits.

CheapRandom::cheap_random3!(is_randomizing, perm, s)

cheap_random3! updates s, a byte string. perm is a CheapRandom seed, a byte string of 256 bytes all different. is_randomizing is a boolean which determines whether or s is being randomized or un-randomized. cheap_random3! returns another perm / CheapRandom seed. spec/cheap_random_spec.rb is used to test cheap_random3!. spec/using_cheap_bits_cheap_random_spec.rb is also used to test cheap_random3! and to demonstrate the use of lib/cheap_bits.rb.

Properties of the CheapRandom seed

CheapRandom seeds are easy to identify. All the CheapRandom seeds are 256 bytes, all different.

CheapRandom seeds can be used to identify files. CheapRandom seeds are a type of hash function result. When is processed into, is also produced. is the result of cheap_random3! on

Given the same start seed -- the.seed, the result seed files and are always identical. ( is the result of cheap_random3! on


ruby examples/make_seed.rb pet_cat.png => random/pet_cat.png.seed which should be copied to random/the.seed


ruby examples/cr.rb => random/ and random/


ruby examples/cb.rb => listing of byte frequencies for random/


ruby examples/chi_squared.rb => a listing of a chi-squared statistic for random/ followed by a listing of a chi-squared statistic for the same amount of data generated by Kernel::rand 256


Make sure that the random directory exists and contains the files pet_cat.png, pet_cat.png.seed, the.seed, and as described above. Run rspec. The tests are quite comprehensive.

Other possible uses of CheapRandom

There are a number of intriguing possible uses for the CheapRandom algorithm and the CheapRandom seed properties beyond random number generation.

Additional Examples - C code and In Place File Processing

The examples2 directory contains examples related to example2/cr.c These examples are designed to be on the PATH and require the use of a C compiler. The examples3 directory contains examples suitable for use on Microsoft Windows.

cr.c Cheap Random Filter and Friends

gcc -Wall cr.c -o cr creates the cr cheap random filter. Running cr h writes out some help text. cr <somefile to stdout and cr u <somefile.random to stdout are the most basic use of cr. the.seed is hard coded into cr. To see how, use chex_pipe.rb <the.seed and compare the console output to the definition of SEED in cr.c. the.seed was generated by running the bash script make_seed pet_cat.png

crf.rb Cheap Random File, In Place File Processing

crf.rb somefile => somefile.random - somefile over-written in place, somefile.the.seed, crf.rb somefile.random => somefile - somefile.random over-written in place, somefile.random.the.seed, The seed used by crf.rb is a file designated by the environmental variable CR_SEED if defined, otherwise crf.rb any one file located in ~/.cheap_random. If the designated file is a .seed file it will be used directly, otherwise a seed will be generated internally from the designated file. ~/.cheap_random/pet_cat.png can serve as a seed file for crf.rb.

crx and crxs Cheap Random File, In Place File Processing

gcc -Wall rpb.c -o rpb, rpb makes in place file processing possible for the bash scripts crx and crxs. crx somefile => somefile.random - somefile over-written in place. crx somefile.random => somefile - somefile.random over-written in place. crxs behaves like crx with the addition of the creation of the seed files the.seed and new.seed.