Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!cs.utexas.edu!samsung!usc!snorkelwacker!husc6!spdcc!ima!haddock!karl
From: karl@haddock.ima.isc.com (Karl Heuer)
Newsgroups: comp.lang.c
Subject: Re: random() error in Turbo C++
Message-ID: <17607@haddock.ima.isc.com>
Date: 31 Aug 90 22:25:27 GMT
References: <118662@linus.mitre.org>
Reply-To: karl@kelp.ima.isc.com (Karl Heuer)
Organization: Interactive Systems, Cambridge, MA 02138-5302
Lines: 16

In article <118662@linus.mitre.org> jrv@sdimax2.mitre.org (VanZandt) writes:
>     #define random(n) ((int)((long)rand()*(n))/((long)RAND_MAX+1))
>p.s.  Note that the common device of using rand()%n is guaranteed to
>yield a number in the range 0...n-1, but the probabilities are not uniform.

This is equally true of your version (proof: there are RAND_MAX+1 equally
likely outcomes but n possible results); only the particular "hot" values have
been redistributed.  If it's important, here is a correct algorithm.
	#include <stdlib.h>
	int random(int n) {
	    register int r;
	    do r = rand() / ((RAND_MAX + 1UL) / n); while (r >= n);
	    return (r);
	}

Karl W. Z. Heuer (karl@kelp.ima.isc.com or ima!kelp!karl), The Walking Lint