bug#9103: 23.3.50; random* from cl does not support the whole range of integers

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Leo]

As far as I can see, (random* most-positive-fixnum) only returns a 32bit
integer on 64bit Emacs. This might be an oversight given the age of
cl.el.

Leo

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Glenn Morris]

Leo wrote:

> As far as I can see, (random* most-positive-fixnum) only returns a 32bit
> integer on 64bit Emacs.

Why does random* even need to exist? What can it do that plain random
cannot?

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Glenn Morris]

Glenn Morris wrote:

> Why does random* even need to exist? What can it do that plain random
> cannot?

...non-integer results, it seems.

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Leo]

On 2011-07-19 05:58 +0800, Glenn Morris wrote:
> Glenn Morris wrote:
>
>> Why does random* even need to exist? What can it do that plain random
>> cannot?
>
> ...non-integer results, it seems.

We need to extend plain random to support random-states. I plan to do it
once the random module in gnulib is usable.

Another option is to use Mersenne Twister PRNG, which I have in my local
tree (patch: http://paste.pocoo.org/show/ZYhLUPeXlkj405100KX4/). The
files mt19937.[ch] are from its authors'. Do you know if that can be
included in Emacs?

Leo

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Juanma Barranquero]

On Tue, Jul 19, 2011 at 03:21, Leo <sdl.web@gmail.com> wrote:

> The
> files mt19937.[ch] are from its authors'. Do you know if that can be
> included in Emacs?

From http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/elicense.html :

"Until 2001/4/6, MT had been distributed under GNU Public License, but
after 2001/4/6, we decided to let MT be used for any purpose,
including commercial use. 2002-versions mt19937ar.c, mt19937ar-cok.c
are considered to be usable freely."

    Juanma

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Lars Magne Ingebrigtsen]

Glenn Morris <rgm@gnu.org> writes:

> Glenn Morris wrote:
>
>> Why does random* even need to exist? What can it do that plain random
>> cannot?
>
> ...non-integer results, it seems.

`random*' ends with

(* (/ n '8388608e0) lim)

if LIM isn't an integer.  Anybody willing to hazard a guess why
8388608.0 was chosen?

And would it make sense to just replace it all with a call to `random',
and then make it into a floating point if it floats?

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Andreas Schwab]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> `random*' ends with
>
> (* (/ n '8388608e0) lim)
>
> if LIM isn't an integer.  Anybody willing to hazard a guess why
> 8388608.0 was chosen?

Because that's the range of n.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Stefan Kangas]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

>> Glenn Morris wrote:
>>
>>> Why does random* even need to exist? What can it do that plain random
>>> cannot?
>>
>> ...non-integer results, it seems.

(That's `cl-random' these days.)

> and then make it into a floating point if it floats?

The Common Lisp version of `random' supports float and integer values
depending on what type its argument has.  I guess that's what
`cl-random' is based on.

I'm not sure if we can change `random' to do the same.  But if there is
a need for random floating point numbers, it does seem unfortunate that
we only have that functionality in CL-Lib.

bug#9103: 23.3.50; random* from cl does not support the whole range of integers

[Stefan Kangas]

Leo <sdl.web@gmail.com> writes:

> As far as I can see, (random* most-positive-fixnum) only returns a 32bit
> integer on 64bit Emacs. This might be an oversight given the age of
> cl.el.

This seems to be an issue also with cl-random on current master, AFAICT.