Warning: This is the manual of the legacy Guile 2.0 series. You may want to read the manual of the current stable series instead.

7.5.19.3 Obtaining random number generator procedures

Function: random-source-make-integers source

Obtains a procedure to generate random integers using the random source source. The returned procedure takes a single argument n, which must be a positive integer, and returns the next uniformly distributed random integer from the interval {0, ..., n-1} by advancing the state of source.

If an application obtains and uses several generators for the same random source source, a call to any of these generators advances the state of source. Hence, the generators do not produce the same sequence of random integers each but rather share a state. This also holds for all other types of generators derived from a fixed random sources.

While the SRFI text specifies that “Implementations that support concurrency make sure that the state of a generator is properly advanced”, this is currently not the case in Guile’s implementation of SRFI-27, as it would cause a severe performance penalty. So in multi-threaded programs, you either must perform locking on random sources shared between threads yourself, or use different random sources for multiple threads.

Function: random-source-make-reals source

Function: random-source-make-reals source unit

Obtains a procedure to generate random real numbers 0 < x < 1 using the random source source. The procedure rand is called without arguments.

The optional parameter unit determines the type of numbers being produced by the returned procedure and the quantization of the output. unit must be a number such that 0 < unit < 1. The numbers created by the returned procedure are of the same numerical type as unit and the potential output values are spaced by at most unit. One can imagine rand to create numbers as x * unit where x is a random integer in {1, ..., floor(1/unit)-1}. Note, however, that this need not be the way the values are actually created and that the actual resolution of rand can be much higher than unit. In case unit is absent it defaults to a reasonably small value (related to the width of the mantissa of an efficient number format).