If an identity conversion exists from S to T, must it be that S and T are same type?

Question

In 6.1.6. of the C# language specification, there is:

The implicit reference conversions are:

(...)
From any reference-type to a reference-type T if it has an implicit identity or reference conversion to a reference-type T0 and T0 has an identity conversion to T.

Why don't they say instead, more simply:

From any reference-type to a reference-type T if it has an implicit identity or reference conversion to T.

Is there any factual difference?

EDIT: I realized I mistyped the specification and the error could potentially be significant for the question (the specification says "The implicit reference conversion are" rather than "An implicit conversion exists")

Apuleius

"reference conversion to T0 AND T0 has an identity conversion to T.".. sounds like there is a difference. – Guggenheim

I think the last part could be important. As type T could implicitly convert to type U due to its implicit identity or reference conversion to U. However U may not necessarily have an identity conversion to U. It doesn't mention U having a reference conversion to T though... confuzzed. – Guggenheim

@LewsTherin It sounds like that, sure, but what is it? I'm confused why the "middle-man" T0 is necessary. By my reading, all types have an identity conversion to themselves and no type has an identity conversion to any other type. – Glasper

I won't pretend to understand this shiz. But I just found this.. might help: #3737289 – Guggenheim

I think you mistake the and, it should be (from S to T0) AND (from T0 to T). So if S can be converted to T0 and T0 to T then S can be converted to T. – Mandiemandingo

@Mandiemandingo Like transitivity? if a is b and b is c then a is c? – Guggenheim

@Lews Therin that's how I understand it, except that is slightly more complicated because from S to T0 is a reference conversion and from T0 to T an indentity conversion. – Mandiemandingo

S to T0 can be an identity or reference conversion.. so I guess the confusion starts when S to T0 has no identity conversion but only reference conversion. Maybe the spec should first define what "identity conversion" and "reference conversion" is. – Guggenheim

I think you need it to convert Derived<dynamic> to Base<object>. – Precess

@LewsTherin: I think you'll find that the spec does define both those things. – Perloff

@EricLippert I was able to find one for "reference conversion" not identity conversion though. – Guggenheim

@LewsTherin: Consider looking in the table of contents. – Perloff

@EricLippert I did find that. But it is over complicated. "An identity conversion converts from any type to the same type. This conversion exists such that an entity that already has a required type can be said to be convertible to that type." "Same type", "that type" means little to me. I like your explanation better though. – Guggenheim

@LewsTherin: The specification does not even define "type". It's not intended to be a formal description of the type system; you won't find any Hoare logic in the specification. – Perloff

Answer 1

If an identity conversion exists from S to T, must it be that S and T are same type?

The oddity you've discovered in the spec arose as a result of adding dynamic to the language in C# 4.0. At runtime there is no such thing as dynamic; rather, dynamic is just a type that means "I'm really object; please defer analysis of this portion of the program until runtime".

Therefore there is an identity conversion between, say, List<object> and List<dynamic>. From the C# compiler's perspective they are different types because myList[0].Frob() would give an error for the former but not the latter. But from the runtime's perspective they are identical. Therefore the C# language classifies the conversion from one to the other as an identity conversion. At compile time the types can be different for the purposes of the C# language, but from the runtime's perspective they will be identical.

Perloff