自定义排序比较功能引发编译错误

Question

A custom sort function in C++ on VC++ 2012 throws a compile error for the following code.

class Segmenter(){
public:
    vector<vector<float>> scanned;
    void modifyAndSort();
    bool sort_fn(const vector<float>&, const vector<float>&);
};

void Segmenter::modifyAndSort(){
    // Modify scanned
    // ...

    sort(scanned.begin(), scanned.end(), sort_fn);
}

bool Segmenter::sort_fn(const vector<float>& x, const vector<float>& y){
    return ((x[0]*x[1]) < (y[0]*y[1]));
}

The error thrown is:

Error   3   error C3867: 'Segmenter::sort_fn': function call missing argument list; use '&Segmenter::sort_fn' to create a pointer to member

Answer 1

You need:

sort(scanned.begin(), scanned.end(), std::bind(&Segmenter::sort_fn, this));
//                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Don't forget to #include <functional> .

Answer 2

A couple ways other ways to do this.

We can't just pass in a class method without binding the instance as Kerrek SB does in his answer. A method will have a hidden this parameter whether you use it or not.

In this case it looks like you use it not, so a static method is probably the easiest:

class Segmenter{
public:
    vector<vector<float>> scanned;
    void modifyAndSort();
    static bool sort_fn(const vector<float>&, const vector<float>&);
};

void Segmenter::modifyAndSort(){
    // Modify scanned
    // ...

    sort(scanned.begin(), scanned.end(), &sort_fn);
    //                                   ^ note: we need a pointer to the method.
}

bool Segmenter::sort_fn(const vector<float>& x, const vector<float>& y){
    return ((x[0]*x[1]) < (y[0]*y[1]));
}

Slightly more and less complicated at the same time option is everybody's close, personal friend the Lambda Function :

class Segmenter
{
public:
    vector<vector<float>> scanned;
    void modifyAndSort();
};

void Segmenter::modifyAndSort()
{
    // Modify scanned
    // ...

    sort(scanned.begin(),
         scanned.end(),
         [](const vector<float>& x, const vector<float>& y)->bool
         {
             return ((x[0]*x[1]) < (y[0]*y[1]));
         });
}

Note: you will need two more Lambdas to get an electric violin solo .

Answer 3

std::bind() definitely works.

A simple method with just a struct.

void Segmenter::modifyAndSort(){
    // Modify scanned
    // ...

    sort(scanned.begin(), scanned.end(), Compare());
}

struct Compare {

    bool operator()(vector<int> x, vector<int> y) const { 
        return ((x[0]*x[1]) < (y[0]*y[1]));
    }
};