Re: comments needed for Random class

On Wed, 13 Jun 2007, gga wrote:

> Different distributions like random, uniform, etc. are excellent ideas
> too.  This is a standard naming conventions used in math (and pretty
> common, for example, in the 3d industry, when you create a renderer).

Problem with this is that gaussian distributions need the centre and
variance specified.  How about a method that given a block uses that
block as a cumulative probability distribution funtion, so performing
a 1:1 mapping from uniform input to non-uniform output?  The existence
of such a function would act as documentation on how to get nonuniform
distributions.  Then a collection of mapping functions could be provided
to support this, Rayleigh, Poisson, Gaussian.

A method to create a cumulative frequency table may be useful in
Monte Carlo methods, and save people having to rewrite one each time.
It would be very difficult to get this wrong if there were

inject_with_index(initial_accumulator=nil) {|accumulator, value, index| ... }

Then it would be (untested)  
  f.inject_with_index(0){|a,v,i| cf[i] = a + v }


        Hugh