[swift-users] Workarounds for SR-6039?

[Michael Gottesman]

> On Jan 8, 2018, at 11:02 AM, Marc Palmer via swift-users <swift-users at swift.org> wrote:
> 
> Hi,
> 
> It’s hard to be 100% certain because of bugs.swift.org being down currently, but I believe I have hit this bug:

Just an FYI, I think it is back up now (it is for me). Did you just try?

+CC swift-infrastructure.

> 
> https://bugs.swift.org/browse/SR-6039 "A `protocol A: class` is not any `AnyObject`”
> 
> This is marked as a duplicate of a presumably very old issue judging by its low number:
> 
> https://bugs.swift.org/browse/SR-55 "non- at objc protocol existentials do not conform to their own protocol type"
> 
> So, I have a protocol:
> 
> public protocol ActionAuditorObserver: AnyObject {
>    func actionAuditorInserted(entries: [ActionAuditEntry])
> }
> 
> …where ActionAuditEntry is a struct.
> 
> Trying to use a generic collection of my own for these, of the form:
> 
> class ObserverSet<T: AnyObject> {
> …
> }
> 
> Using:
> 
> let observers = ObserverSet<ActionAuditorObserver>()
> 
> …results in "'ActionAuditorObserver' is not convertible to ‘AnyObject’”.
> 
> I think this is indeed SR-6039, which makes me really sad as it implies this is not going to be fixed imminently.
> 
> I am trying to write this as a pure-swift framework as so far have succeeded. If I add @objc to the protocol as a workaround for this, I then can no longer have the observer pass in an array of structs.
> 
> So then I have to change the struct to a class, and I’m pushed farther away from my goals :(
> 
> Is there anything I am missing, or a better way to workaround this?
> 
> I have to have the ObserverSet require AnyObject because it needs to do identity checking when removing them. Equatable does not make sense for observers. Losing the struct and forcing @objc feels like such a bitter pill for something so simple.
> 
> Thanks in advance
> Marc
> 
> _______________________________________________
> swift-users mailing list
> swift-users at swift.org
> https://lists.swift.org/mailman/listinfo/swift-users