I'm trying to convert some old WWDC swift code to Swift 4. I think that I have everything done, except for this last bit that does some KVO. This has been pretty difficult to narrow it down to this last bit because everything appears to function like the example code - but these KVO methods do not get called in Swift 4. I found that out here: Open Radar Bug
What would be the Swift 4 way to represent the following?
// use the KVO mechanism to indicate that changes to "state" affect other properties as well
class func keyPathsForValuesAffectingIsReady() -> Set<NSObject> {
return ["state" as NSObject]
}
class func keyPathsForValuesAffectingIsExecuting() -> Set<NSObject> {
return ["state" as NSObject]
}
class func keyPathsForValuesAffectingIsFinished() -> Set<NSObject> {
return ["state" as NSObject]
}
And here are the variable definitions from the example:
override var isReady: Bool {
switch state {
case .initialized:
// If the operation has been cancelled, "isReady" should return true
return isCancelled
case .pending:
// If the operation has been cancelled, "isReady" should return true
guard !isCancelled else {
return true
}
// If super isReady, conditions can be evaluated
if super.isReady {
evaluateConditions()
}
// Until conditions have been evaluated, "isReady" returns false
return false
case .ready:
return super.isReady || isCancelled
default:
return false
}
}
override var isExecuting: Bool {
return state == .executing
}
override var isFinished: Bool {
return state == .finished
}
If more code is needed, please let me know.
If this is a duplicate question, please link to the duplicate here. I've been unable to find a solution.
The keyPathsForValuesAffecting… members can be properties instead of methods.
They must be declared #objc because the KVO system accesses the properties using the Objective-C runtime.
The properties should have type Set<String>.
If you use the #keyPath directive, the compiler can tell you when you've used an invalid key path (for example because of a spelling error or a change to the property name).
Thus:
#objc class var keyPathsForValuesAffectingIsReady: Set<String> {
return [#keyPath(state)]
}
#objc class var keyPathsForValuesAffectingIsExecuting: Set<String> {
return [#keyPath(state)]
}
#objc class var keyPathsForValuesAffectingIsFinished: Set<String> {
return [#keyPath(state)]
}
You also need to make sure your state property is declared #objc dynamic.
The main problem is that KVO is built using Objective-C, and it uses the Objective-C runtime to detect the existence of the keyPathsForValuesAffecting methods. In Swift 4, methods are no longer exposed to Objective-C by default if you don't include an #objc annotation on them. So, in a nutshell, adding the #objc annotation will probably fix your problem.
Another thing that I do—not strictly necessary, but it makes the code look a bit nicer—is to declare these as static constants. The #objc will cause these to get exposed to Objective-C as class methods, so it all works, and it's slightly cleaner. I like to put private on them, too, since these will never get called by Swift code, and there's no point cluttering your class's internal and/or public interface.
You also need to make sure that your state property is KVO-compliant, and sending the notifications when it is changed. You can either do this by making the property dynamic, which will cause the KVO system to automatically generate the notification calls for you, or you can manually call willChangeValue(for:) and didChangeValue(for:) (or the string-based versions, willChangeValue(forKey:) and didChangeValue(forKey:)) in your willSet and didSet handlers for the property.
Finally, don't use raw string key paths in Swift if you can avoid it. The #keyPath() mechanism is the preferred way to get string-based key paths (and for uses other than these legacy Objective-C methods that need to take strings, you should use the new KeyPath type which is better still). If your state property is not an Objective-C-compatible type, though, you're stuck with the old string key paths (in which case you'll fire the notifications in your willSet and didSet as described in the previous paragraph). Alternatively, you can create a dummy Any-typed object that mirrors your state property, purely for KVO purposes.
So, something like this:
#objc private static let keyPathsForValuesAffectingIsReady: Set<String> = [
#keyPath(state)
]
Now, the state property. If it's an Objective-C-compatible type, it's easy:
#objc dynamic var state: ...
Or, if it's not:
#objc var state: SomeNonObjCThing {
willSet { self.willChangeValue(forKey: "state") }
didSet { self.didChangeValue(forKey: "state") }
}
OR:
#objc private var _stateKVO: Any { return self.state }
var state: SomeNonObjCThing {
willSet { self.willChangeValue(for: \.stateKVO) }
didSet { self.didChangeValue(for: \.stateKVO) }
}
// you can now use #keyPath(_stateKVO) in keyPathsForValuesAffecting...
(NS)Operation relies heavily on NSObject-KVO
The closest Swift syntax is
#objc private class func keyPathsForValuesAffectingIsReady() -> Set<String> {
return [#keyPath(state)]
}
#objc private class func keyPathsForValuesAffectingIsExecuting() -> Set<String> {
return [#keyPath(state)]
}
#objc private class func keyPathsForValuesAffectingIsFinished() -> Set<String> {
return [#keyPath(state)]
}
Side note: You might need to make state thread-safe.
I have a Swift protocol defined as follows:
protocol SmartContract {
func apply(transaction :Transaction)
func addBrokenRule(_ brokenRule: BrokenRule)
var brokenRules :[BrokenRule] { get set }
}
I have an extension on SmartContract defined as follows:
extension SmartContract {
mutating func addBrokenRule(_ brokenRule :BrokenRule) {
if self.brokenRules == nil {
self.brokenRules = [BrokenRule]()
}
self.brokenRules.append(brokenRule)
}
}
I also have a class MoneyTransferContract which conforms to the protocol but does not define brokenRules. This is because I have defined the brokenRules inside the extension.
class MoneyTransferContract : SmartContract {
func apply(transaction :Transaction) { // do something here }
}
My question is how can I make sure that MoneyTransformContract conforms to the SmartContract protocol. Is there anyway to have BrokenRule available to MoneyTransformContract without defining it again and again in different SmartContracts.
john doe wrote:
Is there anyway to have BrokenRule available to MoneyTransformContract without defining it again and again in different SmartContracts.
What you want is a superclass/subclass relationship for that behavior.
class SmartContract {
func apply(transaction :Transaction) {
//implemention
}
var brokenRules: [BrokenRule] = []
func addBrokenRule(_ brokenRule :BrokenRule) {
brokenRules.append(brokenRule)
}
}
class MoneyTransferContract : SmartContract {
// Gets `brokenRules` for free from superclass.
}
class BitCoinTransferContract : SmartContract {
// Gets `brokenRules` for free from superclass.
}
If I'm understanding correctly, there is no way to do what you want. (EDIT: as Jeff notes, if you want to use inheritance as opposed to a protocol, this is possible. See his answer for how). Protocols in Swift are just lists of requirements that it is up to implementing types to properly define. Protocols generally don't have any say over the actual behavior or implementation of implementing types, and only guarantees that the properties and functions exist. Your SmartContract protocol says that every SmartContract must have a function apply(transaction:), a function addBrokenRule(_:), and a property brokenRules which can be accessed and modified.
When a type implements SmartContract it has to define each one of these. Just as you have to write out the signature to tell the compiler that you are implementing apply(transaction:), you also have to tell the compiler how the property brokenRules will be implemented. You could obtain this functionality in a somewhat useless way, by defining an extension which has brokenRules as a computed property which is essentially a no-op:
extension SmartContract {
var brokenRules: [BrokenRule] {
get { return [] }
set(newRules) { }
}
}
But this means that any implementing type which forgets to specify their own implementation for brokenRules will have a brokenRules property which always resolves to an empty array.
One reason for this is Swift's computed properties. For some implementing type, it might not make sense for brokenRules to be stored as an array--and a protocol can't force that. That's an implementation detail that a protocol author can't (and shouldn't) worry about. For instance, if BrokenRules were easily convertible to and from strings, you could imagine some class which implemented SmartContract like so:
class StringContract: SmartContract {
var ruleString: String
var brokenRules: [BrokenRule] {
get {
let stringArray = ruleString.split(separator: ",")
return stringArray.map { BrokenRule(string:String($0)) }
}
set(newRules) {
let stringArray = newRules.map { $0.stringValue }
ruleString = stringArray.joined(separator: ",")
}
}
func apply(transaction: Transaction) {
// do something here...
}
init() {
ruleString = ""
}
}
Note that we don't have to specify an implementation for addBrokenRule(_:). That's what your protocol extension gets you. By providing an implementation in the extension, you ensure that all implementing types have a default implementation of the function, and so they can choose to forego defining their own.
Example:
internal protocol PropertyProtocol {
var property: Self {
get
}
}
The only option I see to implement it, let us say in a class is
internal final class PropertyClass: PropertyProtocol {
let property: PropertyClass
internal init(otherOne pOtherOne: PropertyClass) {
self.property = pOtherOne
}
}
But then I do not see a possibility to use it.
let test: PropertyProtocol = PropertyProtocol(...) // hmm, how?
Does Self in a protocol property type declaration always have to be optional?
As a stored property, indeed it would have to be optional for you to create an instance, as each instance would require the stored property to be assigned during initialisation – leading to recursive behaviour. Therefore Self doesn't make too much sense as a stored property; it's really more designed to be used with methods or calculated properties.
Depending on what you're using this for (seems like a fairly hypothetical example), you could implement a calculated property like so:
protocol PropertyProtocol {
var property : Self { get }
}
final class PropertyClass : PropertyProtocol {
var property : PropertyClass {
get {
return // ...
}
set {
// ...
}
}
}
That way the class itself can manage the creation of the property when it's accessed, preventing the recursive behaviour of requiring it to be assigned during initialisation.
While migrating some projects from Objective-C to Swift, I find myself replacing the #property/#synthesize syntax with a lot more boiler-plate code.
For example, for every #property I am implementing, I currently use this pattern:
var foo: Foo? {
get {
return _foo
}
set {
_foo = newValue
}
}
var _foo: Foo?
It ends up adding quickly. Compare this with the 2 lines in Objective-C. That doesn't feel right.
Should I just replace the code using _foo in my ported classes with foo and skip the abstraction of each public var behind an interface entirely?
As #Amit89 pointed out, you can simply use _foo.
In your example, adding foo does exactly nothing, as you simply recreated the get and set functionality already supplied with _foo.
If you want to control how _foo is set you can use property observers:
var _foo: Foo? {
didSet { }
willSet { }
}
If you want to control how _foo is retrieved, then it would indeed make sense to create a private version of _foo, and then accessing it via a computed property:
private var _foo: Foo?
var foo: Foo? {
get {
// return `_foo` in some special way
}
// add a setter if you want to set `_foo` in some special way
}
Swift has a property declaration syntax very similar to C#'s:
var foo: Int {
get { return getFoo() }
set { setFoo(newValue) }
}
However, it also has willSet and didSet actions. These are called before and after the setter is called, respectively. What is their purpose, considering that you could just have the same code inside the setter?
The point seems to be that sometimes, you need a property that has automatic storage and some behavior, for instance to notify other objects that the property just changed. When all you have is get/set, you need another field to hold the value. With willSet and didSet, you can take action when the value is modified without needing another field. For instance, in that example:
class Foo {
var myProperty: Int = 0 {
didSet {
print("The value of myProperty changed from \(oldValue) to \(myProperty)")
}
}
}
myProperty prints its old and new value every time it is modified. With just getters and setters, I would need this instead:
class Foo {
var myPropertyValue: Int = 0
var myProperty: Int {
get { return myPropertyValue }
set {
print("The value of myProperty changed from \(myPropertyValue) to \(newValue)")
myPropertyValue = newValue
}
}
}
So willSet and didSet represent an economy of a couple of lines, and less noise in the field list.
My understanding is that set and get are for computed properties (no backing from stored properties)
if you are coming from an Objective-C bare in mind that the naming conventions have changed. In Swift an iVar or instance variable is named stored property
Example 1 (read only property) - with warning:
var test : Int {
get {
return test
}
}
This will result in a warning because this results in a recursive function call (the getter calls itself).The warning in this case is "Attempting to modify 'test' within its own getter".
Example 2. Conditional read/write - with warning
var test : Int {
get {
return test
}
set (aNewValue) {
//I've contrived some condition on which this property can be set
//(prevents same value being set)
if (aNewValue != test) {
test = aNewValue
}
}
}
Similar problem - you cannot do this as it's recursively calling the setter.
Also, note this code will not complain about no initialisers as there is no stored property to initialise.
Example 3. read/write computed property - with backing store
Here is a pattern that allows conditional setting of an actual stored property
//True model data
var _test : Int = 0
var test : Int {
get {
return _test
}
set (aNewValue) {
//I've contrived some condition on which this property can be set
if (aNewValue != test) {
_test = aNewValue
}
}
}
Note The actual data is called _test (although it could be any data or combination of data)
Note also the need to provide an initial value (alternatively you need to use an init method) because _test is actually an instance variable
Example 4. Using will and did set
//True model data
var _test : Int = 0 {
//First this
willSet {
println("Old value is \(_test), new value is \(newValue)")
}
//value is set
//Finaly this
didSet {
println("Old value is \(oldValue), new value is \(_test)")
}
}
var test : Int {
get {
return _test
}
set (aNewValue) {
//I've contrived some condition on which this property can be set
if (aNewValue != test) {
_test = aNewValue
}
}
}
Here we see willSet and didSet intercepting a change in an actual stored property.
This is useful for sending notifications, synchronisation etc... (see example below)
Example 5. Concrete Example - ViewController Container
//Underlying instance variable (would ideally be private)
var _childVC : UIViewController? {
willSet {
//REMOVE OLD VC
println("Property will set")
if (_childVC != nil) {
_childVC!.willMoveToParentViewController(nil)
self.setOverrideTraitCollection(nil, forChildViewController: _childVC)
_childVC!.view.removeFromSuperview()
_childVC!.removeFromParentViewController()
}
if (newValue) {
self.addChildViewController(newValue)
}
}
//I can't see a way to 'stop' the value being set to the same controller - hence the computed property
didSet {
//ADD NEW VC
println("Property did set")
if (_childVC) {
// var views = NSDictionaryOfVariableBindings(self.view) .. NOT YET SUPPORTED (NSDictionary bridging not yet available)
//Add subviews + constraints
_childVC!.view.setTranslatesAutoresizingMaskIntoConstraints(false) //For now - until I add my own constraints
self.view.addSubview(_childVC!.view)
let views = ["view" : _childVC!.view] as NSMutableDictionary
let layoutOpts = NSLayoutFormatOptions(0)
let lc1 : AnyObject[] = NSLayoutConstraint.constraintsWithVisualFormat("|[view]|", options: layoutOpts, metrics: NSDictionary(), views: views)
let lc2 : AnyObject[] = NSLayoutConstraint.constraintsWithVisualFormat("V:|[view]|", options: layoutOpts, metrics: NSDictionary(), views: views)
self.view.addConstraints(lc1)
self.view.addConstraints(lc2)
//Forward messages to child
_childVC!.didMoveToParentViewController(self)
}
}
}
//Computed property - this is the property that must be used to prevent setting the same value twice
//unless there is another way of doing this?
var childVC : UIViewController? {
get {
return _childVC
}
set(suggestedVC) {
if (suggestedVC != _childVC) {
_childVC = suggestedVC
}
}
}
Note the use of BOTH computed and stored properties. I've used a computed property to prevent setting the same value twice (to avoid bad things happening!); I've used willSet and didSet to forward notifications to viewControllers (see UIViewController documentation and info on viewController containers)
If I've made a mistake anywhere, please edit to fix it!
You can also use the didSet to set the variable to a different value. This does not cause the observer to be called again as stated in Properties guide. For example, it is useful when you want to limit the value as below:
let minValue = 1
var value = 1 {
didSet {
if value < minValue {
value = minValue
}
}
}
value = -10 // value is minValue now.
These are called Property Observers:
Property observers observe and respond to changes in a property’s
value. Property observers are called every time a property’s value is
set, even if the new value is the same as the property’s current
value.
Excerpt From: Apple Inc. “The Swift Programming Language.” iBooks. https://itun.es/ca/jEUH0.l
I suspect it's to allow for things we would traditionally do with KVO such as data binding with UI elements, or triggering side effects of changing a property, triggering a sync process, background processing, etc, etc.
NOTE
willSet and didSet observers are not called when a property is set in an initializer before delegation takes place
The many well-written existing answers cover the question well, but I'll mention, in some detail, an addition that I believe is worth covering.
The willSet and didSet property observers can be used to call delegates, e.g., for class properties that are only ever updated by user interaction, but where you want to avoid calling the delegate at object initialization.
I'll cite Klaas up-voted comment to the accepted answer:
willSet and didSet observers are not called when a property is first
initialized. They are only called when the property’s value is set
outside of an initialization context.
This is a quite neat as it means e.g. the didSet property is a good choice of launch point for delegate callbacks & functions, for your own custom classes.
As an example, consider some custom user control object, with some key property value (e.g. position in rating control), implemented as a subclass of UIView:
// CustomUserControl.swift
protocol CustomUserControlDelegate {
func didChangeValue(value: Int)
// func didChangeValue(newValue: Int, oldValue: Int)
// func didChangeValue(customUserControl: CustomUserControl)
// ... other more sophisticated delegate functions
}
class CustomUserControl: UIView {
// Properties
// ...
private var value = 0 {
didSet {
// Possibly do something ...
// Call delegate.
delegate?.didChangeValue(value)
// delegate?.didChangeValue(value, oldValue: oldValue)
// delegate?.didChangeValue(self)
}
}
var delegate: CustomUserControlDelegate?
// Initialization
required init?(...) {
// Initialise something ...
// E.g. 'value = 1' would not call didSet at this point
}
// ... some methods/actions associated with your user control.
}
After which your delegate functions can be used in, say, some view controller to observe key changes in the model for CustomViewController, much like you'd use the inherent delegate functions of the UITextFieldDelegate for UITextField objects (e.g. textFieldDidEndEditing(...)).
For this simple example, use a delegate callback from the didSet of the class property value to tell a view controller that one of it's outlets have had associated model update:
// ViewController.swift
Import UIKit
// ...
class ViewController: UIViewController, CustomUserControlDelegate {
// Properties
// ...
#IBOutlet weak var customUserControl: CustomUserControl!
override func viewDidLoad() {
super.viewDidLoad()
// ...
// Custom user control, handle through delegate callbacks.
customUserControl = self
}
// ...
// CustomUserControlDelegate
func didChangeValue(value: Int) {
// do some stuff with 'value' ...
}
// func didChangeValue(newValue: Int, oldValue: Int) {
// do some stuff with new as well as old 'value' ...
// custom transitions? :)
//}
//func didChangeValue(customUserControl: CustomUserControl) {
// // Do more advanced stuff ...
//}
}
Here, the value property has been encapsulated, but generally: in situations like these, be careful not to update the value property of the customUserControl object in the scope of the associated delegate function (here: didChangeValue()) in the view controller, or you'll end up with infinite recursion.
The willSet and didSet observers for the properties whenever the property is assigned a new value. This is true even if the new value is the same as the current value.
And note that willSet needs a parameter name to work around, on the other hand, didSet does not.
The didSet observer is called after the value of property is updated. It compares against the old value. If the total number of steps has increased, a message is printed to indicate how many new steps have been taken. The didSet observer does not provide a custom parameter name for the old value, and the default name of oldValue is used instead.
Getter and setter are sometimes too heavy to implement just to observe proper value changes. Usually this needs extra temporary variable handling and extra checks, and you will want to avoid even those tiny labour if you write hundreds of getters and setters. These stuffs are for the situation.
In your own (base) class, willSet and didSet are quite reduntant , as you could instead define a calculated property (i.e get- and set- methods) that access a _propertyVariable and does the desired pre- and post- prosessing.
If, however, you override a class where the property is already defined, then the willSet and didSet are useful and not redundant!
One thing where didSet is really handy is when you use outlets to add additional configuration.
#IBOutlet weak var loginOrSignupButton: UIButton! {
didSet {
let title = NSLocalizedString("signup_required_button")
loginOrSignupButton.setTitle(title, for: .normal)
loginOrSignupButton.setTitle(title, for: .highlighted)
}
I do not know C#, but with a little guesswork I think I understand what
foo : int {
get { return getFoo(); }
set { setFoo(newValue); }
}
does. It looks very similar to what you have in Swift, but it's not the same: in Swift you do not have the getFoo and setFoo. That is not a little difference: it means you do not have any underlying storage for your value.
Swift has stored and computed properties.
A computed property has get and may have set (if it's writable). But the code in the getter and setter, if they need to actually store some data, must do it in other properties. There is no backing storage.
A stored property, on the other hand, does have backing storage. But it does not have get and set. Instead it has willSet and didSet which you can use to observe variable changes and, eventually, trigger side effects and/or modify the stored value. You do not have willSet and didSet for computed properties, and you do not need them because for computed properties you can use the code in set to control changes.