Guy Decoux writes:

> I was not against Math::Randon but my remark was for class
> RandomNumberGenerator (it was used only _once_ after you have *always*
> used RNG). 
>
Ok, I see. Sorry.

> I don't really want, one day, to have a class FileTransfertProtocol :-)
>
Point taken. Maybe we should stick to Rng and Mt19937 after all...

> I know this, this is, for me, the internal of Marshal. For me, an user
>don't need to know that it exist a method #_dump, or #_load, it just need
> to know that it can write :
>
>   Marshal.dump(object)
>
> Why have you an 'include Marshal' ?
>
I'd misunderstood how to use Marshal...

So it suffices to add load and dump methods?

> Yes, for each of this distribution you can also define a probability
>density 
>
> The module function will return the probability density
> and if you include this module in Random, it will return a random number
> in this distribution.
>
I'm not sure I know exactly what you mean (even though it sounds as
though I should!). Can you give short code (well, structure will
do) example (for gaussian)? (I had planned to use something like
Box-Mueller method for generating gaussian from uniform(0,1))

> I make simulation (Monte-Carlo, GA), but they are not written in ruby,
> sorry :-) :-) :-)
>
IMHO, doing a GA in Ruby is not as weird as it might sound if you do the
time-critical parts in a C extension. I do this with GP and the
performance penalty is in the range 30-40% (after profiling-based
optimizations). Using Ruby as glue gives me more flexible experimenting
with extensions and additions.

Regards,

Robert