Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!apple!sun-barr!lll-winken!uunet!munnari.oz.au!goanna!ok
From: ok@goanna.cs.rmit.oz.au (Richard A. O'Keefe)
Newsgroups: comp.lang.prolog
Subject: Re: random number generator
Message-ID: <4682@goanna.cs.rmit.oz.au>
Date: 31 Jan 91 09:00:54 GMT
References: <27A20DBF.15393@ics.uci.edu> <4024@skye.ed.ac.uk>
Organization: Comp Sci, RMIT, Melbourne, Australia
Lines: 51

In article <4024@skye.ed.ac.uk>, ken@aiai.ed.ac.uk (Ken Johnson) writes:
> The following is in the Prolog Library and (as far as I know!!) is
> public domain.  Please acknowledge ROK's authorship.  The algorithm is
> very clever. 

Oh dear.  Did the version of RANDOM.PL I left at Edinburgh *really* omit
the attribution?  The algorithm used there for generating uniform random
floats is algorithm AS 163 from the journal Applied Statistics.  (In
effect it is a 48-bit linear congruential method.)

> random(N, I) :-
> 	(	recorded(seed,[A0,A1,A2],Key)
> 		->	erase(Key)
> 	;
> 			A0 = 3172, A1 = 9814, A2 = 20125
> 	),
>	...
> 	record(seed,[B0,B1,B2],_),
> 	...

I'm quite sure *I* didn't leave it like that.  The third argument of
recorded/1 is NOT a "key".  It's a "data base reference", "dbref" or
"ref" for short.  The FIRST argument is the "key".

Further, the thing that is recorded is not a seed, it is a random state,
and it always contains precisely three numbers, so a list is not appropriate.
What's more, if you haven't modules, it's a good idea to choose key names
that are unlikely to picked by users.  Hence

	recorded('random state', 'random state'(A0,A1,A2), Ref)
	...
	recorda('random state', 'random state'(B1,B1,B3), _)

> %   rand_perm(List, Perm) unifies Perm with a random permutation
> %   of List.  What good this may be I'm not sure, and there's bound
> %   to be a more efficient way of doing it.  Oh well.

> rand_perm([], []).
> rand_perm([H1|T1], [H2|T2]) :-
> 	random([H1|T1], H2, T3),
> 	rand_perm(T3, T2).

I'm sure that comment can't be mine.  There are *much* more efficient
ways of doing it, and the Quintus library uses one of them.  What's more,
the operation is there because it is extremely useful; I found that most
of the time that I was generating random numbers was to shuffle lists
(generate random permuations), so it seemed as well to do it directly.
(Think about generating test data.)
-- 
The Marxists have merely _interpreted_ Marxism in various ways;
the point, however, is to _change_ it.		-- R. Hochhuth.