You need a Fisher-Yates shuffle.
This is a very efficient “select n from m” solution that gives you a subset of your values with zero possibility of duplicates (and no unnecessary up-front sorting). Pseudo-code to do this follows:
dim n[N] // gives n[0] through n[N-1]
for each i in 0..N-1:
n[i] = i // initialise them to their indexes
nsize = N // starting pool size
do N times:
i = rnd(nsize) // give a number between 0 and nsize-1
print n[i]
nsize = nsize - 1 // these two lines effectively remove the used number
n[i] = n[nsize]
By simply selecting a random number from the pool (based on the current pool size), replacing it with the top number from that pool, then reducing the size of the pool, you get a shuffle without having to worry about a large number of swaps up front.
This is important if the number is high in that it doesn’t introduce an unnecessary startup delay.
For example, examine the following bench-check, choosing 10-from-10:
<------ n[] ------>
0 1 2 3 4 5 6 7 8 9 nsize rnd(nsize) output
------------------- ----- ---------- ------
0 1 2 3 4 5 6 7 8 9 10 4 4
0 1 2 3 9 5 6 7 8 9 7 7
0 1 2 3 9 5 6 8 8 2 2
0 1 8 3 9 5 6 7 6 6
0 1 8 3 9 5 6 0 0
5 1 8 3 9 5 2 8
5 1 9 3 4 1 1
5 3 9 3 0 5
9 3 2 1 3
9 1 0 9
You can see the pool reducing as you go and, because you’re always replacing the used one with an unused one, you’ll never have a repeat.
Here’s a little Java program that shows this in action:
import java.util.Random;
public class testprog {
private int[] pool; // The pool of numbers.
private int size; // The current "size".
private Random rnd; // A random number generator.
// Constructor: just initilise the pool.
public testprog (int sz) {
pool = new int[sz];
size = sz;
rnd = new Random();
for (int i = 0; i < size; i++) pool[i] = i;
}
// Get next random number in pool (or -1 if exhausted).
public int next() {
if (size < 1) return -1;
int idx = rnd.nextInt(size--);
int rval = pool[idx];
pool[idx] = pool[size];
return rval;
}
// Test program for the pool.
public static void main(String[] args) {
testprog tp = new testprog (10);
for (int i = 0; i < 11; i++) System.out.println (tp.next());
}
}
The output is (for one particular run):
3
5
1
0
6
4
9
2
8
7
-1
The -1 is there just to show you what happens when you exhaust the list. Since you’ve explicitly stated you don’t want duplicates, it returns a sentinel value. You could also choose other options such as throwing an exception or just restarting the pool.