使用模板确定要调用的函数

Question

Let's assume, that a class can use the following data-types to pass data to and from it:

std::vector<complex> // where "complex" is a struct 
and
std::vector DOUBLE or INT or FLOAT or STRING

If the user passes through a std::vector<double> this implies that a calculation has already been carried out and therefore only a small process has to be completed. If the user, however, passes through std::vector<complex> this means that the calculation has to be carried out.

In a class I can do the following:

class Foo {

  template<typename T>
  Foo(T begin, T end)
  {
      this->transform(begin, end);
  }
  template<typename T>
  void transform(T begin, T end)
  {
       std::cout << "This has been called";

  }
};

But this will have to still invoke me having to have a data member of one specific type and always assume they are doing that.

For example, is it possible to override the function transform for 2 different cases, and, have the constructor decide which method to call depending on type of data being passed to the constructor.

It would work something like this:

int main()
{
    std::vector<double> vals = {1,1,1,1,1,1,1,1};

    Foo(std::begin(vals), std::end(vals)); // this will call the method that handles dbls

    std::vector<complex> vals = {};

    Foo(std::begin(vals), std::end(vals)); // This will call the method that handles complex numbers

}

I hope this makes sense

Answer 1

This can be solved through template specialization .

Considere the following function:

template<typename T>
void Foo(T arg)
{
    std::cout << "General" << std::endl;
}

We can now specialize this function for char types, this means providing another implementation for those types:

template<>
void Foo(char arg)
{
    std::cout << arg << std::endl;
}

Notice that the template<> can be omitted in this case.

If we now call our functions like this:

Foo(1);
Foo('a');

The output will be:

General
a

Answer 2

One way would be tag dispatching (I used struct X instead of complex here):

#include <iostream>
#include <vector>
#include <iterator>

struct X {};

struct Foo {
  template<typename T>
  Foo(T begin, T end)
  {
      transform(begin, end, typename std::iterator_traits<T>::value_type());
  }
  template<typename T, typename U>
  void transform(T begin, T end, U) // third parameter is unused and serves as a tag
  {
    std::cout << "floating point\n";
  }

  template<typename T>
  void transform(T begin, T end, X) // same here
  {
     std::cout << "X\n";
  }
};

int main()
{
    std::vector<double> v1 = {1,1,1,1,1,1,1,1};

    Foo(std::begin(v1), std::end(v1)); // floating point

    std::vector<X> v2 = {};

    Foo(std::begin(v2), std::end(v2));  // X
}