An Implementation Model for Rational Protocol Conformance Behavior

[mattrips]

@Douglas_Gregor

For example, to satisfy the requirements of containsDuplicates() or doSomething() , we look up the conformance of X to Equatable so the witness table can be passed in the function. This affects type identity when the conformances are part of the type's generic requirements ( Hashable conformances affect Set<X> identity, but not [X] identity). If there's an ambiguity, the compiler should produce an error. The model is consistent, but the implementation is buggy . [emphasis added]

I believe part of the motivation behind this thread is to define and document the model so that users are able to make this sort of judgment for themselves.

As things stand, now, we have multiple sources of truth: (1) StackOverflow answers, (2) blog posts, (3) WWDC videos, (4) reading between the lines in The Swift Programming Language, (5) asking questions in the Using Swift part of this forum, and (6) whatever the compiler does.

That last one is a special one. The common wisdom seems to be that the compiler defines the language. If the compiler let's you do something, then that something is allowed (unless someone from the Core Team or at Apple says otherwise). But that really isn't true, and shouldn't be true.

This is not meant as a complaint. Rather, it is intended as a call to please continue down the path this thread already is on.