seo's blog

Just what!? Standart rand() in C++ or undeniable proof of the existence of magic!
By seo, history, 6 years ago, In English

Until recently, I knew that rand() in C++ works awfully with random_shuffle and it was not credible, but continued to use it for most tasks.. But recently I wrote some code and I cannot explain the result. Maybe you can?

I advise you to read to the end, it is really very interesting!

At each step, this code generates a random value from 0 to 1. As soon as 11 (for example) identical values equal 1 were received in a row, index number of current value is added to the vector. This operation is repeated a number of times and each time it starts from the beginning.

Let's output vector values and look at the last 15 of them for convenience.

Code here:

CODE
OUTPUT1
OUTPUT2
OUTPUT3
OUTPUT4

At first glance it seems that the values are random. But let's try to sort the vector.

Code here:

CODE
OUTPUT1
OUTPUT2
OUTPUT3
OUTPUT4

It should be noted that the output is different each time, but the individual values are the same. Moreover, these individual values are repeated three to four times.

  • Using of mt19937 rnd(chrono::steady_clock::now().time_since_epoch().count()); instead of rand() gives a random output array.

  • If we add a line that will not affect the answer in any way, but will affect the time difference between the rand() calls we also get random values in the output array. In doing so, we have to use random delay. If the delay is fixed, we will not get random values.

CODE1, rnd(), normal random values
CODE2, random delay between calls rand(), normal random values

With a fixed delay, we also get magic values, even if the delay is large.

CODE3, fixed delay between calls rand(), magic values

Now let's instead of an empty loop in the third code call rand().

CODE
OUTPUT1
OUTPUT2
OUTPUT3

Values become repeated much more often..

CODE
OUTPUT

If we call the method twice, the values are repeated more often.

I believe that this is not quite a trivial algorithm and the result is rather strange. If you have any idea how to explain this, please state your thoughts in the comments. Thanks for reading!

  • Vote: I like it
  • +41
  • Vote: I do not like it

»
6 years ago, # |
  Vote: I like it +2 Vote: I do not like it

My guess is: it's due to time(0). But I don't really know how or why :S

→ Reply
  • »
    »
    6 years ago, # ^ |
    ← Rev. 2   Vote: I like it +3 Vote: I do not like it

    Checked without using srand(time(0)); You should understand that every time the output array will be the same. But the question is whether we get groups of identical values in a sorted vector.

    Result:

    CODE
    OUTPUT, same every time

    Therefore, srand(time(0)); does not affect the oddity of the result.

    P.S. Using srand(time(nullptr)); instead of srand(time(0)); gives same result as using of srand(time(0));.

    → Reply
    • »
      »
      »
      6 years ago, # ^ |
        Vote: I like it +2 Vote: I do not like it

      No, I mean, I know you need to seed the rand, but maybe there's a pattern on time(0) everytime you run it. I read that time(0) gets time in seconds since the Unix epoch, I thought it might depend on how codeforces or your computer get that value, but I think I'm wrong.

      → Reply
      • »
        »
        »
        »
        6 years ago, # ^ |
          Vote: I like it +3 Vote: I do not like it

        Tried to use srand(56164989), same result.

        → Reply
»
6 years ago, # |
  Vote: I like it +23 Vote: I do not like it

It isn't magic. It's small period of std::rand

→ Reply
  • »
    »
    6 years ago, # ^ |
      Vote: I like it +8 Vote: I do not like it

    So how I should write this code with rand() and without crutches to get a normal result?

    → Reply
  • »
    »
    6 years ago, # ^ |
    ← Rev. 2   Vote: I like it +3 Vote: I do not like it

    Why std::rand? rand is part of the C standard library, is it not? std:: prefix is not needed.

    Nevermind, apparently everything contained in libraries is part of the std namespace. But compilers are allowed to inject things in the global namespace as well.

    → Reply
»
6 years ago, # |
  Vote: I like it +23 Vote: I do not like it

You also shouldn't really expect any differences between CODE and CODE3. You are simply doing a loop that doesn't change the state of the PRNG.

→ Reply
  • »
    »
    6 years ago, # ^ |
      Vote: I like it +3 Vote: I do not like it

    The topic has been updated. If we call the method a fixed number of times

    //
rand();
//

    or

    //
for(int j = 0; j < 2; ++ j) rand();
//

    or

    //
for(int j = 0; j < 10; ++ j) rand();
//

    the repeating becomes more frequent. But if we use

    //
int u = 1;
for(int j = 0; j < rand() % 10; ++ j) u *= 3;
//

    values become random.

    Can you explain this in some way?

    → Reply
»
6 years ago, # |
  Vote: I like it 0 Vote: I do not like it

Auto comment: topic has been updated by seo (previous revision, new revision, compare).

→ Reply
»
6 years ago, # |
  Vote: I like it +9 Vote: I do not like it

No magic at all here. That's just an implementation of rand() function.

The only fact you should know that it is Linear congruential generator with some parameters.

So what you are saying (11 times odd numbers in a row) are actually just some equations which need to be satisfied, so you have not many solutions. You can increase this number to 17 odd numbers in a row and will eventually find that on Windows it will fail to find any number (I've waited for five minutes and haven't found any number). That means that no value satisfies the system of equations.

→ Reply
  • »
    »
    6 years ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    Very interesting. Thanks for the reply. I understood how it works approximately, but for me it remains unclear how to explain it exactly.

    → Reply