Interface with generic parameter vs Interface with

Question

Let's say I have such interface and concrete implementation

public interface IMyInterface<T>
{
    T My();
}

public class MyConcrete : IMyInterface<string>
{
    public string My()
    {
        return string.Empty;
    }
}

So I create MyConcrete implementation for strings, I can have one more concrete implementation for int. And that's ok. But let's say, that I want to do the same thing, but with generic methods, so I have

public interface IMyInterface2
{
    T My<T>();
}

public class MyConcrete2 : IMyInterface2
{
    public string My<string>()
    {
        throw new NotImplementedException();
    }
}

So I have the same IMyInterface2, but which defines generic behavior by means of T My<T>(). In my concrete class I want to implement My behavior, but for concrete data type - string. But C# doesn't allow me to do that.

My question is why I cannot do that? In other words, if i can create concrete implementation of MyInterface<T> as MyClass : MyInterface<string> and stop genericness at this point, why I can't do that with generic method - T My<T>()?

Answer 1

Your generic method implementation has to be generic as well, so it has to be:

public class MyConcrete2 : IMyInterface2
{
    public T My<T>()
    {
        throw new NotImplementedException();
    }
}

Why you can't do My<string>() here? Because interface contract needs a method, that could be called with any type parameter T and you have to fulfill that contract.

Why you can't stop genericness in this point? Because it would cause situations like following:

Class declarations:

public interface IMyInterface2
{
    T My<T>(T value);
}

public class MyClass21 : IMyInterface2
{
    public string My<string>(string value) { return value; }
}

public class MyClass22 : IMyInterface2
{
    public int My<int>(int value) { return value; }
}

Usage:

var item1 = new MyClass21();
var item2 = new MyClass22();

// they both implement IMyInterface2, so we can put them into list
var list = new List<IMyInterface2>();
list.Add(item1);
list.Add(item2);

// iterate the list and call My method
foreach(IMyInterface2 item in list)
{
    // item is IMyInterface2, so we have My<T>() method. Choose T to be int and call with value 2:
    item.My<int>(2);

    // how would it work with item1, which has My<string> implemented?
}

Answer 2

Your solution does not work for two reasons.

First, an interface is a contract. When you implement IMyInterface2 you guarantee that you will implement a function named My that takes a generic type parameter and returns that type. MyConcrete2 does not do this.

Second, C# generics do not allow any kind of type parameter specialization. (I do wish C# supported this.) This is a common thing in C++ templates where your example would compile, but any usages of MyConcrete2 would fail to compile if they don't call My with a string.

Answer 3

when you write the Generic Method the Definition is for keeping the placeholder. Actual Type comes into picture when you call the method. so instead you should write

public T My<T>()
{
    throw new NotImplementedException();
}

and when you call the method you can use the string there.

Answer 4

Because your interface declares a generic method T My<T>(), but you implementation does not implement a function with that specific signature.

To achieve what you want, you need to provide the T generic parameter to the interface instead, in your first example:

public interface IMyInterface2<T>
{
        T My();
}

public class MyConcrete2 : IMyInterface2<string>
{
    public string My()
    {
        throw new NotImplementedException();
    }
}