let timer = Timer(timeInterval: 1, target: object,
userInfo: nil, repeats: false)
button.addTarget(object, action: #selector(MyClass.buttonTapped),
with: button, with: otherButton)
var foo: Int
Multiple functions with the same base name can be differentiated by their parameter labels using the aforementioned syntax for function references (e.g. insertSubview(_:at:) vs insertSubview(_:aboveSubview:)). But if a function has no parameters, the only way to disambiguate it is to use an as cast with the function's type signature (e.g. foo as () -> () vs foo(_:)).
(This is actually an improvement over ObjC's @selector() directive, because the compiler's -Wundeclared-selector check verifies only that the named selector exists. The Swift function reference you pass to #selector checks existence, membership in a class, and type signature.)
But there are still a number of important ObjC-based APIs that use selectors, including timers and the target/action pattern. Swift provides the Selector type for working with these. (Swift automatically uses this in place of ObjC's SEL type.)
Cases where #selector doesn't work, and naming: Sometimes you don't have a function reference to make a selector with (for example, with methods dynamically registered in the ObjC runtime). In that case, you can construct a Selector from a string: e.g. Selector("dynamicMethod:") though you lose the compiler's validity checking. When you do that, you need to follow ObjC naming rules, including colons (:) for each parameter.
I'd like to mention that while "Interacting with Objective-C APIs" is on the website, it is NOT in 'The Swift Programming Language' book.
Is there anyway to add validation around passing the "selector" as a string? IE compiler warn us when we misspell, etc.
It should also be pointed out that the Cocoa frameworks expect an Objective-C style method name. If your method takes an argument you will need a ':' if it takes 2 arguments, size:andShape:, if the first argument is named you may need a With, i.e. initWithData: for func init(Data data: NSData)
Key paths: These are related to but not quite the same as selectors. There's a special syntax for these in Swift 3, too: e.g. chris.valueForKeyPath(#keyPath(Person.friends.firstName)). See SE-0062 for details. And even more KeyPath stuff in Swift 4, so make sure you're using the right KeyPath-based API instead of selectors if appropriate.
Putting a string with the function name worked, NSSelectorFromString() works also.
Remember that private symbols aren't exposed to the runtime, too your method needs to have at least internal visibility.
Selector availability: The method referenced by the selector must be exposed to the ObjC runtime. In Swift 4, every method exposed to ObjC must have its declaration prefaced with the @objc attribute. (In previous versions you got that attribute for free in some cases, but now you have to explicitly declare it.)
Swift itself doesn't use selectors several design patterns that in Objective-C make use of selectors work differently in Swift. (For example, use optional chaining on protocol types or is/as tests instead of respondsToSelector:, and use closures wherever you can instead of performSelector: for better type/memory safety.)
The great thing about this approach? A function reference is checked by the Swift compiler, so you can use the #selector expression only with class/method pairs that actually exist and are eligible for use as selectors (see "Selector availability" below). You're also free to make your function reference only as specific as you need, as per the Swift 2.2+ rules for function-type naming.
There are a couple of extra caveats for the function references you pass to the #selector expression:
This should probably mention that the selector needs a ":" at the end if it takes an argument. (E.g. test() -> "test" & test(this:String) -> "test:")
You can construct a Selector from a Swift function type using the #selector expression.