[boost] Boost Random Number Library Framework Proposal

From: John Maddock (John_Maddock_at_[hidden])
Date: 1999-11-07 07:25:36

Jeet -

This is fairly trivial criticism but some of the out of line declarations
could have been declared class-inline to avoid the problems that some
compilers have with static template members:

template<typename Generator>
class uniform_deviate{
public:
enum{
   rand_max = 0xffffffff,
   uint_div = rand_max,
   sint_div = rand_max/2
};
// etc....
};

 or:

template<typename Generator>
class uniform_deviate{
public:
   static const randnum_t rand_max = 0xffffffff;
   static const uinteger_t uint_div = rand_max;
   static const uinteger_t sint_div = rand_max/2;
// etc....
};

rather than:

 template <typename Generator>
        const uniform_deviate<Generator>::randnum_t
        uniform_deviate<Generator>::rand_max = 0xffffffff;

    template <typename Generator>
        const uniform_deviate<Generator>::uinteger_t
        uniform_deviate<Generator>::uint_div = rand_max;

    template <typename Generator>
        const uniform_deviate<Generator>::uinteger_t
        uniform_deviate<Generator>::sint_div = rand_max/2;

Also, if you want to investigate the theory in this area further... by
coincidence I was reading an article on random numbers in New Scientist
this morning, and that led to: http://www.cwi.nl/~paulv/ which has a wide
selection of papers on Kolmogorov complexity and its application to both
random number theory and the "average case" analysis of algorithms.

Anyway, I hope this helps rather than hinders...

- John.