rand() function in C is not random even when seede

问题:

This is most likely a machine dependent issue but I can't figure out what could be wrong.

#include <stdio.h>
#include <stdlib.h>
#include <time.h>

int main(int argc, char** argv) {

  srand(time(NULL));
  int r1 = rand();
  int r2 = rand();
  printf("%d %d\n", r1, r2);
}

I compile the above piece of code using

gcc randd.c

Then running it a few times manually and the first numbers seem incredibly similar while the second ones seem random:

1025720610 1435057801
1025737417 1717533050
1025754224 2000008299
1025771031 134999901
1025787838 417475150

This first call to rand() seems strongly co-related to the time and is strictly increasing as time passes. Any ideas as to why this occurs or how to resolve it?

This happens on OSX 10.11

回答1:

1. rand() is quite bad, avoid it if possible. In any good RNG the first values will be indistinguishable from random even when the seed is close (hamming distance). In rand this is not the case.
2. If you must use rand then seed it, preferably with something higher entropy than time, and call rand() multiple times instead of reseeding-calling-reseeding.

For example of 2, consider:

#include <stdio.h>
#include <stdlib.h>
#include <time.h>

int main(int argc, char** argv) {
  int t=time(NULL);
  srand(t);

  for(int i=0; i < 10; i++) {
    float r = (float)rand()/(float)(RAND_MAX);
    printf("%f\n", r);
  }
}

With the result:

0.460600
0.310486
0.339473
0.519799
0.258825
0.072276
0.749423
0.552250
0.665374
0.939103

It's still a bad RNG but at least the range is better when you allow it to use the internal state instead of giving it another similar seed.

回答2:

This is exactly what you should expect. There's no such thing as "a random number". There are only sequences of numbers with a random distribution. The rand() function generates such sequences, but you're not giving it a chance to, because you keep re-seeding it. The first number generated by rand() may very well be just some function of the seed, or the seed itself. Some rand() functions might hash the seed to hide this, but that doesn't really make them any better, because the contract of rand() is to produce a random sequence.

If you need a sequence of random numbers that survives running multiple programs, you'll have to do something like (a) Write a program that calls srand() once, then calls rand() many times, and have your other programs ask for random numbers from that program over IPC; (b) Use something like /dev/urandom; (c) Use something like random.org.