Search code examples
c#genericscovariancecontravariancevariance

Type parameters defining each other? class A<T1, T2> where T1 : Foo where T2 : T1

Does 

class A<T1, T2>
    where T1 : Foo
    where T2 : T1

have an actual use case?

What's the difference to

class A<T1, T2>
    where T1 : Foo
    where T2 : Foo

? What does actually change?

Is it the same when variance is involved?

Solution

  • The difference is that T2 cannot be just any Foo it has to be a Foo that is derived from T1.

    For instance

    public class Foo{}
public class Foo1 : Foo {}
public class Foo2 : Foo {}
public class Foo12 : Foo1 {}
public class A<T1,T2> where T1: Foo where T2 : T1 {}

    will allow

    var a = new A<Foo1, Foo12>()

    but not

    var a  = new A<Foo1, Foo2>()

    This also means that you can safely cast an object of type T2 to T1.

    Is it the same when variance is involved?

    Variance only comes into play with interfaces.