Error in STL Random Generator program

[aligh]

I'm trying out the random generation facilities of STL.
The following code produces a histogram for a normal distribution for 30 rows over 200 tries.

It works, and the output seems to be correct, but the program ends with the following error:

"munmap_chunk(): invalid pointer

RUN FINISHED; Aborted; core dumped; real time: 130ms; user: 0ms; system: 0ms"

  #include <random>
#include <iostream>

using namespace std;

class Rand_int {
public:
    Rand_int(int low, int high): dist{low, high}{}
    int operator()(){return dist(re);}
private:
    default_random_engine re;
    normal_distribution<> dist;
};

/*
 * 
 */
int main(int argc, char** argv) {
    Rand_int rnd{0, 29};
    
    vector<int> histogram(30);
    for (int i = 0; i < 2000; ++i) {
        ++histogram[rnd()];
    }
    
    for (int i = 0; i < histogram.size(); ++i) {
        cout << i << "\t";
        for (int j = 0; j < histogram[i]; ++j)
            cout << '*';
        cout << endl;
    }
    return 0;
}

g++ --version:
"g++ (Ubuntu 7.4.0-1ubuntu1~18.04) 7.4.0"

compile arguments:
"-g -std=c++14"

Using NetBeans 11.0

Why do I get such an error?

Thank you for your help!

Edit: Running the code on cpp.sh with "Standard Input option" set to "None" gives the same error, while setting it to "Interactive" works fine without any errors.

[Peter87]

A normal distribution is not restricted within some finite limits. The first argument to the std::normal_distribution constructor is the mean, and the second argument is the standard deviation.

[aligh]

Ahh, thank you! I think I understand the problem now.

I was using std::normal_distribution constructor like std::uniform_int_distribution with a high and a low value but that's not how standard deviation works.

I was writing outside of my vector<int> and that was the cause of the error.

I've corrected my code, both the standard_distribution constructor and my for loop to make sure I'm not accessing invalid elements.

#include <random>
#include <iostream>

using namespace std;

class Rand_int {
public:
    Rand_int(double low, double high): dist{low, high}{}
    int operator()(){return dist(re);}
private:
    default_random_engine re;
    normal_distribution<> dist;
};

/*
 * 
 */
int main(int argc, char** argv) {
    Rand_int rnd{15, 5};
    
    vector<int> histogram(30);
    for (int i = 0; i < 2000; ++i) {
        int d{rnd()};
        if (d < histogram.size())
            ++histogram[d];
    }
    
    for (int i = 0; i < histogram.size(); ++i) {
        cout << i << "\t";
        for (int j = 0; j < histogram[i]; ++j)
            cout << '*';
        cout << endl;
    }
    return 0;
}

Please, let me know if I've missed anything or there are other important things that I need to look after.

Thanks!

[Peter87]

This code "works" but I just want to point out two things.

1. The std::normal_distribution can give you negative values. Your condition (d < histogram.size()) only happens to reject negative values because histogram.size() returns an unsigned integer type which forces d to be converted to an unsigned integer type (which turns negative values into large positive values) before the comparison takes place.

2. The std::normal_distribution returns floating-point values which you convert to integers by just throwing away the decimal part. This means that all values in the range (-1, 1) will map to 0 which makes 0 about twice as likely as it should be. For your current parameters it isn't easily noticeable but if you change line 19 to something like Rand_int rnd{5, 5}; you'll see a spike at 0 that shouldn't be there.

[aligh]

Thank you so much!