Why does C# not consider the return type of a function in polymorphism?

Question

There are two functions with same name and same set of parameters but with different return types. Why is it not a form of polymorphism i.e. method overloading? Why is it not allowed by compiler?

Answer 1

Because C# is designed so that types can be analyzed from inside to outside. Imagine if you have

int N() {}
float N()() {}

and then a call

float x = N();

OK, great, obviously we could say that the float version was wanted. But then you say:

void M(int x) {}
void M(float x) {}

M(N());

OK, now which version was wanted? The rule is figure out what N() means and then figure out what the best overload of M is once you know what N() means. You reason from inside to outside.

Overload resolution based on return type requires reasoning from outside to inside and that can be a lot harder.

Answer 2

Consider the following:

int combine(int a, int b) {
    return a + b;
}

float combine(int a, int b) {
    return a - b;
}

If I were to call combine(1, 2), there is no way for the compiler to know which of the two methods I want to call (it's ambiguous).

You could almost make a case for checking return types, but what about:

var c = combine(1, 2);
dynamic d = combine(1, 2);
combine(1, 2);

In the above, what should the value of c or d be? 3? -1? It's impossible to tell. How about the last statement, where there's no value being assigned? I didn't define an overload that returns void, so which of the two methods should it call?