Collection of carefully-prepared Classes and Protocols designed to imbue your inheriting Object Types with efficient, protocol-driven Observer Pattern Behaviour. As of version 2.0.0, this includes support for Keyed Observers (see usage examples below for details)
Select File -> Swift Packages -> Add Package Dependency and enter https://github.com/Flowduino/Observable.git
You can use Observable as a Package Dependency in your own Packages' Package.swift file:
let package = Package(
//...
dependencies: [
.package(
url: "https://github.com/Flowduino/Observable.git",
.upToNextMajor(from: "2.0.0")
),
],
//...
)From there, refer to Observable as a "target dependency" in any of your package's targets that need it.
targets: [
.target(
name: "YourLibrary",
dependencies: [
"Observable",
],
//...
),
//...
]You can then do import Observable in any code that requires it.
Here are some quick and easy usage examples for the features provided by Observable:
You can inherit from ObservableClass in your own Class Types to provide out-of-the-box Observer Pattern support.
This not only works for @ObservedObject decorated Variables in a SwiftUI View, but also between your Classes (e.g. between Services, or Repositories etc.)
First, you would define a Protocol describing the Methods implemented in your Observer Class Type that your Observable Class can invoke:
/// Protocol defining what Methods the Obverable Class can invoke on any Observer
protocol DummyObservable: AnyObject { // It's extremely important that this Protocol be constrained to AnyObject
func onFooChanged(oldValue: String, newValue: String)
func onBarChanged(oldValue: String, newValue: String)
}Note - It is important that our Protocol define the AnyObject conformity-constraint as shown above.
Now, we can define our Observable, inheriting from ObservableClass:
/// Class that can be Observed
class Dummy: ObservableClass {
private var _foo: String = "Hello"
public var foo: String {
get {
return _foo
}
set {
// Invoke onFooChanged for all current Observers
withObservers { (observer: DummyObservable) in
observer.onFooChanged(oldValue: _foo, newValue: newValue)
}
_foo = newValue
objectWillChange.send() // This is for the standard ObservableObject behaviour (both are supported together)
}
}
private var _bar: String = "World"
public var bar: String {
get {
return _bar
}
set {
// Invoke onBarChanged for all current Observers
withObservers { (observer: DummyObservable) in
observer.onBarChanged(oldValue: _bar, newValue: newValue)
}
_bar = newValue
objectWillChange.send() // This is for the standard ObservableObject behaviour (both are supported together)
}
}
}We can now define an Observer to register with the Observable, ensuring that we specify that it implements our DummyObservable protocol:
class DummyObserver: DummyObservable {
/* Implementations for DummyObservable */
func onFooChanged(oldValue: String, newValue: String) {
print("Foo Changed from \(oldValue) to \(newValue)")
}
func onBarChanged(oldValue: String, newValue: String) {
print("Bar Changed from \(oldValue) to \(newValue)")
}
}We can now produce some simple code (such as in a Playground) to put it all together:
// Playground Code to use the above
var observable = Dummy() // This is the Object that we can Observe
var observer = DummyObserver() // This is an Object that will Observe the Observable
observable.addObserver(observer) // This is what registers the Observer with the Observable!
observable.foo = "Test 1"
observable.bar = "Test 2"ObservableThreadSafeClass works exactly the same way as ObservableClass. The internal implementation simply encapsulates the Observer collections behind a DispatchSemaphore, and provides a Revolving Door mechanism to ensure unobstructed access is available to addObserver and removeObserver, even when withObservers is in execution.
Its usage is exactly as shown above in ObservableClass, only you would substitute the inheritence of ObservableClass to instead inherit from ObservableThreadSafeClass.
ObservableThread provides you with a Base Type for any Thread Types you would want to Observe.
Note - ObservableThread does implement the ObservableObject protocol, and is technically compatible with the @ObservedObject property decorator in a SwiftUI View. However, to use it in this way, anywhere you would invoke objectWillUpdate.send() you must instead use notifyChange(). Internally, ObservableThread will execute objectWillChange.send() but enforce that it must execute on the MainActor (as required by Swift)
Let's now begin taking a look at how we can use ObservableThread in your code.
The example is intentionally simplistic, and simply generates a random number every 60 seconds within an endless loop in the Thread.
Let's begin by defining our Observation Protocol:
protocol RandomNumberObserver: AnyObject {
func onRandomNumber(_ randomNumber: Int)
}Any Observer for our Thread will need to conform to the RandomNumberObserver protocol above.
Now, let's define our RandomNumberObservableThread class:
class RandomNumberObservableThread: ObservableThread {
init() {
self.start() // This will start the thread on creation. You aren't required to do it this way, I'm just choosing to!
}
public override func main() { // We must override this method
while self.isExecuting { // This creates a loop that will continue for as long as the Thread is running!
let randomNumber = Int.random(in: -9000..<9001) // We'll generate a random number between -9000 and +9000
// Now let's notify all of our Observers!
withObservers { (observer: RandomNumberObserver) in
observer.onRandomNumber(randomNumber)
}
Self.sleep(forTimeInterval: 60.00) // This will cause our Thread to sleep for 60 seconds
}
}
}So, we now have a Thread that can be Observed, and will notify all Observers every minute when it generates a random Integer.
Let's now implement a Class intended to Observe this Thread:
class RandomNumberObserverClass: RandomNumberObserver {
public func onRandomNumber(_ randomNumber: Int) {
print("Random Number is: \(randomNumber)")
}
We can now tie this all together in a simple Playground:
var myThread = RandomNumberObservableThread()
var myObserver = RandomNumberObserverClass()
myThread.addObserver(myObserver)That's it! The Playground program will now simply print out the new Random Number notice message into the console output every 60 seconds.
You can adopt this approach for any Observation-Based Thread Behaviour you require, because ObservableThread will always invoke the Observer callback methods in the execution context their own threads! This means that, for example, you can safely instantiate an Observer class on the UI Thread, while the code execution being observed resides in its own threads (one or many, per your requirements).
As of version 1.1.0, you can now register and notify Keyed Observers. Note: Version 2.0.0 modified the Interface significantly to eliminate the need for Generic Typing of the Key. Key Types are now inferred for you.
This functionality is an extension of the standard Observer Pattern, and is implemented in the following classes from which you can extend:
KeyedObservableClassinstead ofObservableClassKeyedObservableThreadinstead ofObservableThreadKeyedObservableThreadSafeClassinstead ofObservableThreadSafeClass
Remember, Keyed Observation is an extension of the basic Observation Pattern, so any Keyed Observable is also inherently able to register and notify non-Keyed Observers
You would use Keyed Observation whenever your Observers care about a specific context of change. A good example would be for a Model Repository, where an Observer may only care about changes to a specific Model contained in the Repository. In this scenario, you would used Keyed Observation to ensure the Observer is only being notified about changes corresponding to the given Key.
Key Types must always conform to the Hashable protocol, just as must any Key Type used for a Dictionary collection.
Let's take a look at a basic usage example.
We shall provide a basic usage example to synchronize an Observer's internal Dictionary for specific keys only with the values from the Observable's internal Dictionary.
First, we would begin with an Observation Protocol:
protocol TestKeyedObservable: AnyObject {
func onValueChanged(key: String, oldValue: String, newValue: String)
}The above Observation Protocol provides the method onValueChanged which takes the key (in this case a String value) and provides the corresponding oldValue and newValue values for that key.
Our Observer will implement TestKeyedObservable to provide an implementation for this function.
Now, let's define a simple Keyed Observable to house the master Dictionary we will be selectively-synchronizing with one or more Observers.
class TestKeyedObservableClass: KeyedObservableClass<String> {
private var keyValues: [String:String] = ["A":"Hello", "B":"Foo", "C":"Ping"]
func setValue(key: String, value: String) {
withKeyedObservers(key: key) { (key, observer: TestKeyedObservable) in
observer.onValueChanged(key: key, oldValue: self.keyValues[key]!, newValue: value)
}
self.keyValues[key] = value
}
}The above class inherits from KeyedObservableClass and specializes the TKey generic to be a String. In other words, the Keys for this Observable must always be String values.
It includes a simple String:String dictionary (String key with a String value)
The setValue method will simply notify all observers using withKeyedObservers any time a specific key the Obsever(s) is(are) observing is updated, passing along the oldValue and newValue values. It will then update its internal Dictionary (keyValues) so that it always contains the latest value.
Note the use of withKeyedObservers instead of withObservers. You will use this syntax in your own Keyed Observables, changing only the declared Observer Protocol (TestKeyedObservable in this example) with the Observer Protocol representing your own observation methods.
Now that we have a Keyed Observable that will notify Observers each time the value of a key changes, let's define an Observer.
class TestKeyedObserverClass: TestKeyedObservable {
public var keyValues: [String:String] = ["A":"Hello", "B":"Foo"]
func onValueChanged(key: String, oldValue: String, newValue: String) {
keyValues[key] = newValue
}
}So, TestKeyedObserverClass is a simple class, implementing our TestKeyedObservable Observer Protocol.
For this example, we are going to presume that there are 2 pre-defined Keys with known initial values (there do not have to be... you can have as many keys as you wish)
You will notice that we initialized both the Observable and Observer classes to have identical keyValues dictionaries. This is solely for the sake of simplifying this example by ensuring there is always an oldValue. You don't need to do this in your own implementations.
So, now that we have the Observable and the Observer types, let's produce a simple bit of Playground code to tie it together.
let observable = TestKeyedObservableClass() // Creates our Observable
let observer = TestKeyedObserverClass // Creates a single Observer instanceAt this point, we need to consider what Key or Keys our observer is going to Observe.
For example, we can Observe just one key:
observable.addKeyedObserver(key: "A", observer)The above means that observer would only have its onValueChanged method invoked when the value of key A is modified in observable.
Likewise, if we only care about key B, we can do:
observable.addKeyedObserver(key: "B", observer)If we care about both known keys, we can simply register them both:
observable.addKeyedObserver(keys: ["A", "B"], observer)Also, we can do something particularly clever and basically register the Observer for every Key known to its own Dictionary:
observable.addKeyedObserver(keys: Array(observer.keyValues.keys), observer)The above would register observer with observable for every key contained in observer's keyValues dictionary.
Ultimately, you can register the observer with the observable for any keys you want:
observable.addKeyedObserver(key: "Foo", observer)Let's output the initial values of all of our keys before we invoke any code that would modify their values:
for (key, value) in observer.keyValues {
print("Key: '\(key)' has a value of '\(value)'")
}This would output:
Key: 'A' has a value of 'Hello' Key: 'B' has a value of 'Foo'
So, now that we can register the Keyed Observer with the Observer for whatever key or keys we wish, let's trigger the Observer Pattern in the observer:
observable.setValue(key: "A", "World")The above will then update the value if A from "Hello" to "World".
If we repeat the following code:
for (key, value) in observer.keyValues {
print("Key: '\(key)' has a value of '\(value)'")
}This would output:
Key: 'A' has a value of 'World' Key: 'B' has a value of 'Foo'
Okay, so what if we change the value for key "C"? What will happen?
observable.setValue(key: "C", "Pong")Now, if we repeat the following code:
for (key, value) in observer.keyValues {
print("Key: '\(key)' has a value of '\(value)'")
}This would output:
Key: 'A' has a value of 'World' Key: 'B' has a value of 'Foo'
Note that the observer was not notified about the change to the value of key C. This is because observer is not observing observable for changes to key C.
This is the value of Keyed Observation Pattern. Put simply: not all Observations are meaningful to all Observers. So, as you have now seen, Keyed Observeration enables our Observers to be notified specifically of changes relevant to that Observer.
As of version 1.1.0, all useful combination overloads for the above-specified methods of ObservableClass, ObservableThread, ObservableThreadSafeClass, KeyedObservableClass, KeyedObservableThread, and KeyedObservableThreadSafeClass have been provided to streamline the adding and removal of Observers with/from an Observable.
observable.addObserver(myObserver)observable.addObserver([myObserver1, myObserver2, myObserver3])keyedObservable.addKeyedObserver("MyKey", myKeyedObserver)keyedObservable.addKeyedObserver(["Key1", "Key2", "Key3"], myKeyedObserver)keyedObservable.addKeyedObserver("MyKey", [myKeyedObserver1, myKeyedObserver2, myKeyedObserver3])keyedObservable.addKeyedObserver(["Key1", "Key2", "Key3"], [myKeyedObserver1, myKeyedObserver2, myKeyedObserver3])removeObserver and removeKeyedObserver also provide the same overloads as shown above.
There are a few additional useful things you should know about this Package.
This library intentionally performs run-time type checks against each registered Observer to ensure that it conforms to the explicitly-defined Observer Protocol being requested by your withObservers Closure method.
Simple example protocols:
protocol ObserverProtocolA: AnyObject {
func doSomethingForProtocolA()
}
protocol ObserverProtocolB: AnyObject {
func doSomethingForProtocolB()
}Which can then both be used by the same ObservableClass, ObservableThreadSafeClass, or ObservableThread descendant:
withObservers { (observer: ObserverProtocolA) in
observer.doSomethingForProtocolA()
}
withObservers { (observer: ObserverProtocolB) in
observer.doSomethingForProtocolB()
}Any number of Observer Protocols can be marshalled by any of our Observable types, and only Observers conforming to the explicitly-specified Observer Protocol will be passed into your withObservers Closure method.
Observable is available under the MIT license. See the LICENSE file for more info.
If you require additional support, or would like to discuss Observable, Swift, or any other topics related to Flowduino, you can join us on Discord.