The other answers here will solve the problem, but the following is the pattern I use when I'm doing this:
class Num
{
public:
Num(int i) // Not explicit, allows implicit conversion to Num
: i_ (i)
{
}
Num (Num const & rhs)
: i_ (rhs.i_)
{
}
Num & operator+= (Num const & rhs) // Implement +=
{
i_ += rhs.i_;
return *this;
}
private:
int i_;
};
//
// Because of Num(int), any number on the LHS or RHS will implicitly
// convert to Num - so no need to have lots of overloads
Num operator+(Num const & lhs, Num const & rhs)
{
//
// Implement '+' using '+='
Num tmp (lhs);
tmp+=rhs;
return tmp;
}
One of the key benefits of this approach is that your functions can be implemented in terms of each other reducing the amount of overall code you need.
UPDATE:
To keep performance concerns at bay, I would probably define the non member operator+ as an inline function something like:
inline Num operator+(Num lhs, Num const & rhs)
{
lhs+=rhs;
return lhs;
}
The member operations are also inline (as they're declared in the class body) and so in all the code should be very close to the cost of adding two raw int
objects.
Finally, as pointed out by jalf, the consequences of allowing implicit conversions in general needs to be considered. The above example assumes that it's sensible to convert from an integral type to a 'Num'.