Adding etc. different vectors

[korczurekk]

Hi, currently to add 2 different vectors I have to cast one of them:

sf::Vector2f vect1;
sf::Vector2i vect2;

vect1 += static_cast<sf::Vector2f>(vect2);
 

What about adding templates to those operators?

[Laurent]

This was already asked not so long ago. The problem is that you can't deduce the type of the result with C++03 code, and SFML still can't use C++11.

[korczurekk]

Sorry, I couldn't find simillar topic.

dynamic_cast inside operator's code wouldn't solve this problem?

EDIT: Here's simple example compiled with -std=c++03 and it works, why sfml can't do this?

#include <iostream>
#include <memory>

template<typename T>
struct MyVector
{
    T x;
    T y;

    MyVector()
        : x(0),
          y(0)
    { }
    MyVector(T x, T y)
    {
        this->x = x;
        this->y = y;
    }

    template<typename U>
    MyVector<T> operator +(MyVector<U> vect)
    {
        return MyVector<T>(this->x + vect.x, this->y + vect.y);
    }

    template<typename U>
    MyVector<T> & operator=(MyVector<U> vect)
    {
        x = static_cast<T>(vect.x);
        y = static_cast<T>(vect.y);
    }

    template<typename U>
    MyVector<T> & operator+=(MyVector<U> vect)
    {
        return (*this) = (*this) + vect;
    }
};

int main()
{
    MyVector<float> fVec(2.4f, 3.9f);
    MyVector<int>   iVec(5, 5);

    iVec += fVec;

    std::cout << "[" << iVec.x << " " << iVec.y << "]" << std::endl;


    return 0;
}
 

[Laurent]

You can't assume that T + U = T. The rules are clearly defined by the language and depend on the types involved. For example, int + float = float. That's why you need C++11 with its decltype keyword.

[Nexus]

Apart from the technical limitations mentioned by Laurent, I'm not sure if it would be wise to enable such semantics. Allowing operations on different types of vectors is similar to implicit conversions, which have been disabled by design.

How often do you need to add float and integer vectors? Such requests often indicate flaws in the logic, e.g. trying to add pixels and world coordinates. I prefer if such attempts yield a compiler error, and people who know what they're doing can still use explicit conversions. Type safety is a powerful feature.

[FRex]

There is no need for a static cast to add two different types of vectors right now:

sf::Vector2f vect1;
sf::Vector2i vect2;

vect1 += sf::Vector2f(vect2);