Date: Tue, 4 Nov 1997 18:19:01 -0800 (PST)
Message-Id: <199711050219.SAA15044@adit.ap.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: eblazecka AT bc DOT sympatico DOT ca, djgpp AT delorie DOT com
From: Nate Eldredge <eldredge AT ap DOT net>
Subject: Re: rand() or random()
Precedence: bulk

At 11:20  11/3/1997 -0800, Ryan Blazecka wrote:
>Mark Phillips wrote:
>> Just for the heck of it, say I want to get a random number from 0 to
>> 19.  rand() and random() appear to return numbers from 0 to 2^31.  I
>> could always do rand() * 20 / 2^31 but there must be a better way.
rand() returns values from 0 to RAND_MAX (#defined in <stdlib.h>). random()
goes from 0 to MAXINT (#defined in <values.h>). On DJGPP they happen to both
be the same, and equal to 2^31.
>
>How about rand() % 19?   :)
That does work, but perhaps not so well with some other random number
generator where the low-order bits aren't so random. 
My personal favorite method is this:
(int)((double)rand() / (double)RAND_MAX)) * 20
I realize this does involve floating point, but you don't have to worry
about integer overflow or low-order non-randomness. Also, I find it clearer
mathematically.
>
>(BTW, you should use random() instead of rand(), its numbers are more
>"random")
Agreed. Compare the sources and be amazed. :) I just wonder why ANSI
programs get punished by having to use the worse generator, while
incompatible BSD-isms are rewarded.

Nate Eldredge
eldredge AT ap DOT net