Generating number from binomial distribution using C++ TR1

Question

I am trying to use the following code (taken from the internet) to generate numbers from binomial distribution. It compiles but one execution it hangs. (I am using g++ on mac.)

Could someone suggest a working code to generate numbers from binomial distribution using C++ TR1 library features?

#include <tr1/random>
#include <iostream>
#include <cstdlib>

using namespace std;
using namespace std::tr1;

int main()
{
  std::tr1::mt19937 eng; 
  eng.seed(time(NULL));
  std::tr1::binomial_distribution<int, double> roll(5, 1.0/6.0);
  std::cout << roll(eng) << std::endl;
  return 0;
}

Lemniscate

Did you check in which line your program hangs? (Did you use a debugger?) – Sonyasoo

It hangs in the cout line. More precisely calling roll(eng) does not seem to return. – Haematoxylon

Works fine using clang's C++11 std::binomial_distribution (after removing the extra template argument), FYI – Kenishakenison

How interesting. Seems to be a bug in the TR1 implementation. There's an infinite loop because an accumulator is -inf. I think it happens when -std::log(__urng()) draws a 0. The C++11 version uses slightly different code, which appears to avoid this problem, or at least avoid it almost surely. – Gaspard

It looks like the GCC implementation is trying to be too clever. The Boost version simply performs n Bernoulli trials. I guess when n is large, that's annoying, so the GCC standard implementation uses a trick. – Gaspard

@Nate How did you compile it using C++ 11? On my machine neither "g++ -std=c++0x" nor "g++ -std=c++11" works. It says "cc1plus: error: unrecognized command line option "-std=c++0x"". Thanks! – Lemniscate

If your copy of gcc is too old, it might predate C++11 features. I used libc++ and macports' clang-3.2, compiled using the --stdlib=libc++ -std=c++11 flags. – Kenishakenison

Using the latest gcc compiler I got the code (given below by DennisL) to work. Thanks. – Lemniscate

Answer 1

Here is working code:

#include <iostream>
#include <random>

int main() {
  std::random_device rd;
  std::mt19937 gen(rd());
  std::binomial_distribution<> d(5, 1.0/6.0);
  std::cout << d(gen) << std::endl;
}

You can check its result here, and it works with recent GCC and Clang versions. Please note that it is generally better to use the random_device instead of time to get a seed.

Compile it with the --std=c++11.

Mclendon