I'm trying to write some code that will use random generators but allow you to seed it (for reproducability).
The code looks something like the following (trying to create a snippet that can be run)
#include <cstdio>
#include <functional>
#include <random>
class CarterWegmanHash {
private:
unsigned int a_;
unsigned int b_;
public:
// What should the signature of this be?
CarterWegmanHash(int k, std::function<unsigned int()> unif_rand) {
// Other stuff
a_ = unif_rand();
b_ = unif_rand();
}
unsigned int get_a() { return a_; }
unsigned int get_b() { return b_; }
};
int main() {
// Construct our uniform generator
int n = 8;
std::random_device rd;
auto unif_rand = std::bind(
// uniform rand from [0, n - 1]
std::uniform_int_distribution<unsigned int>(0, pow(2, n)),
std::mt19937(rd()));
for (int i = 0; i < 5; ++i) {
// How do I call this?
CarterWegmanHash h_k(n, unif_rand);
printf("seeds for h_%d are %u, %u\n", i, h_k.get_a(), h_k.get_b());
}
}
I'd like each of the CarterWegmanHash objects to be able to have different seeds from unif_rand
, but (not suprisingly), they're all the same.
Passing by pointer gives me error: no matching function for call to ‘CarterWegmanHash::CarterWegmanHash
, and passing by ref gives me cannot bind non-const lvalue reference of type ‘std::function<unsigned int()>&’ to an rvalue of type ‘std::function<unsigned int()>’
Is there a way to pass a random number generator like uniform_int_distribution
so the state is maintained?
You can just create your distribution and mt19937
in main
and capture them in a lambda:
std::uniform_int_distribution<unsigned int> distribution(0, pow(2, n));
std::mt19937 mt(rd());
auto unif_rand = [&](){ return distribution(mt); };
Or with bind
:
auto unif_rand = std::bind(std::ref(distribution), std::ref(mt));