SML/NJ Library Manual

The Random structure

Synopsis

signature RANDOM
structure Random :> RANDOM


The Random structure implements a random number generator using a subtract-with-borrow (SWB) generator as described in Marsaglia and Zaman, "A New Class of Random Number Generators," Ann. Applied Prob. 1(3), 1991, pp. 462-480.

The SWB generator is a 31-bit generator with lags 48 and 8. It has period (2⁽¹⁴⁸⁷⁾ - 2⁽²⁴⁷⁾)/105 or about 10⁽⁴⁴⁵⁾. Although generally excellent, they act locally like a lagged Fibonacci generator and thus have troubles with the birthday test. To repair this, the implementation provided here incorporates a suggestion of Marsaglia and Zaman, and combines the SWB generator with a linear congruential generator, specifically (48271*a)mod(2⁽³¹⁾-1).

Interface

type rand
val rand : (int * int) -> rand
val toString : rand -> string
val fromString : string -> rand
val randInt : rand -> int
val randNat : rand -> int
val randReal : rand -> real
val randRange : (int * int) -> rand -> int


Description

type rand

encapsulates state for the random number generator.

rand (i, j)

creates the initial seed for the random number generator based on the integers i and j.

val toString

val fromString

These functions convert between generator state and a string representation of that state. The string representation is not necessarily printable. fromString is only guaranteed to work on the output of toString, in which case fromString o toString is the identity function. A string in an incorrect format will cause fromString to raise the exception.

These functions allow the state of a generator to be saved, in order that the generator can be replicated at a later time.

randInt s

generates random integers uniformly in the range [Int.minInt,Int.maxInt] based on the information in s. As a side-effect, the function updates the internal state of s.

randNat s

generates random integers uniformly in the range [0,Int.maxInt] based on the information in s. As a side-effect, the function updates the internal state of s.

randReal s

generates random integers uniformly in the range (0.0,1.0) based on the information in s. As a side-effect, the function updates the internal state of s.

randRange (i, j) s

generates random integers uniformly in the range [i, j] based on the information in s. As a side-effect, the function updates the internal state of s. This raises the exception Fail if j < i.

See Also

Rand

Last Modified May 6, 1998
Comments to John Reppy
Copyright © 1998 Bell Labs, Lucent Technologies