I'm inclined to agree it's a bug, but fixing it would be source-breaking. Might be one of those warning-in-Swift-4, error-in-Swift-5 things.

@rudkx?

The generic argument could be a disfavored overload instead of outright disallowed. It'd still be technically a potentially source-breaking change but in practice I'd imagine it should work pretty much as users intend.

Comment by Anders Ha (JIRA)

Even if we provide a more specific variant that accepts `@escaping (Value) -> U?`, the compiler still picks the most generic variant that accepts just `U`.

The original version should already count as "more specific". That's somewhat separable from the trailing-closure-for-non-closure-parameter thing, though.

The above code sample works on master as of 44531db. The first overload is chosen.

But I think trailing closure syntax should be disallowed when the argument isn't specifically a parameter. i.e., this should error:

public final class Object<T> {  
    public func map<U>(value: U) -> Object<U> {
        return Object<U>()
    }
}

_ = Object<Int>().map { $0 as Int? }

@jckarter, do you think such a change would have to go through swift-evolution? What about if it were just a warning?