Prelude's Random Numbers Corner

As a starter I have no formal instruction in programming, my background is mathematics. While doing some browsing on the subject of generating x distinct random ints from 0 to y (with y>=x of course) I took a look in the FAQ and in Prelude's Corner about the subject. In the Prelude's Corner entry the second to last example (the part included here) really confuses me.

Code:

---

int nrand ( int n )
{
  int r;
  int bucket_size = RAND_MAX / n;

  if ( n <= 0 || n > RAND_MAX )
    return 0;
  do {
    r = std::rand() / bucket_size;
  } while ( r >= n );

  return r;
}

---

This (correct me if I'm wrong) generates a random int between 0 and n-1. Prelude states that the idea is to split the range of random function into buckets of equal size (Is this actually equal or just as near equal as is possible without adding extra code to worry about a few small values? by extreme example if RAND_MAX happened to be 2,would the one bucket be of length 1 and the other of length 2 to split the values 0,1,2 as evenly as possible?). Am I missing something when I think the following implementation of the nrand function is just as good but shorter and without the need for a loop:

Code:

---

int nrand (int n)
{
  if ( n <= 0 || n > RAND_MAX )
    return 0;
 
  return (int)(((double)std::rand() * n) / ((double)RAND_MAX +1));
}

---

---
Thraen

>This (correct me if I'm wrong) generates a random int between 0 and n-1.
For the most part, yes. Though it's not exactly great code (written in haste and has a few bugs) and I should rewrite it. It's on my list of things to do.

>Is this actually equal or just as near equal as is possible
It's about as equal as it gets considering that it's impossible to divide an uneven set of numbers into an even number of buckets, such as when RAND_MAX isn't an exact multiple of n.

>Am I missing something
Yes. The problem with your suggestion is probability. If the value of n begins to near RAND_MAX then the probability of certain values appearing more often than others increases. The bucket trick avoids this situation and you have a much more uniform set of pseudo-random numbers.

Thanks for your additional explanation Prelude, I now understand what was confusing me.

---
Thraen