[Proposal] Unique generic constraint

[petarpetrovt]

Add support for unique generic constraint that ensures some or all generic type parameters are unique at compile time.

public View<T1, T2> View<T1, T2>()
    // where T1 != T2
    // OR: where T1 is not T2
    // OR: where typeof(T1) != typeof(T2)
    // OR: where unique
{
    // Runtime checks are the only current alternative.
    Assert.NotSame<T1, T2>();
}

[ZacharyPatten]

How would this work with polymorphism? Would generic types be considered equal if they are parts of the same branch in an inheritance tree?

If this feature would not respect polymorphism, it should never use the "is" keyword, because the current usage of "is" does respect polymorphism.

using System;

public class Program
{
	public static void Main()
	{
		A a = new A();
		B b = new B();
		Method(a, b);
	}

	public static void Method<T1, T2>(T1 a, T2 b)
	{
		Console.WriteLine(a is T2);
		Console.WriteLine(b is T1);
	}
}

public class A { }
public class B : A { }

Mentioning this because you proposed the possible syntax:

// OR: where T1 is not T2

[svick]

Why?

[petarpetrovt]

@ZacharyPatten
I don't really care about polymorphism support (my use case all types are value types with struct constraint). The proposed syntax is optional.

[PathogenDavid]

Generic type constraints are encoded in CIL and are understood and enforced by the runtime, which means this request would most likely require a runtime change, which raises the bar significantly.

This is something you could/should pretty easily enforce with an analyzer.

Constraints are meant to enable you to use new capabilities with a generic type, not to enable your domain-specific usage restrictions.

[PathogenDavid]

Also this is a duplicate of #1706 (which also presents a scenario where such a constraint enables you to do something new that an analyzer can't help with.)

Somewhat related: #707, #3560

[333fred]

Closing as a duplicate.