I am writing a chess GUI in Swift 3 and use nvzqz/Sage as the chess model/library. Now I face a problem with a Sage closure used for piece promotion.
Sage uses (in its game class) the execute(move: promotion:) method for promotion move execution which has a closure that returns a promotion piece kind. This allows to prompt the user for a promotion piece or perform any other operations before choosing a promotion piece kind, as follows:
try game.execute(move: move) {
...
return .queen
}
I implemented a promotion view controller ("pvc") which is called in this closure so that the player may select the new piece:
// This is in the main View Controller class
/// The piece selected in promotionViewController into which a pawn shall promote
var newPiece: Piece.Kind = ._queen // default value = Queen
try game.execute(move: move) {
boardView.isUserInteractionEnabled = false
// promotionview controller appears to select new promoted piece
let pvc = PromotionViewController(nibName: nil, bundle: nil)
pvc.delegate = self
pvc.modalPresentationStyle = .overCurrentContext
self.present(pvc, animated:true)
return newPiece
}
When the button for the new piece in the pvc is pressed, the pvc dismisses itself and the data of the selected piece (the constant selectedType) is transferred back to the main view controller via delegation:
// This is in the sending PVC class
protocol PromotionViewControllerDelegate {
func processPromotion(selectedType: Piece.Kind)
}
func buttonPressed(sender: UIButton) {
let selectedType = bla bla bla ...
delegate?.processPromotion(selectedType: selectedType)
presentingViewController!.dismiss(animated:true)
}
// This is in the receiving main View Controller class
extension GameViewController: PromotionViewControllerDelegate {
func processPromotion(selectedType: Piece.Kind) {
defer {
boardView.isUserInteractionEnabled = true
}
newPiece = selectedType
}
The problem I have is that the closure (in the game.execute method) does not wait until the player made his selection in the pvc (and immediately returns the newPiece variable which is still the default value) so that I never get another promotion piece processed other than the default value.
How do I make the closure wait until the player pressed a button in the pvc?
Of course, I tried to find a solution and read about callbacks, completion handlers or property observers. I do not know which is the best way forward, some thoughts:
Completion handler: the pvc dismisses itself upon button-press event so the completion handler is not in the receiving (main) view controller. How do I deal with this?
Property observer: I could call the try game.execute(move) method only after the promotion piece was set (with didset) but that would make the code difficult to read and not use the nice closure the game.execute method provides.
Callbacks may be related to completion handlers, but am not sure.
So your block in game.execute(move: move) will fully execute which is so designed by the Sage API. You can not pause it as easy but it is doable, still let's try to solve it the other way;
Why do you need to call the presentation of the view controller within this block? By all means try to move that away. The call try game.execute(move: move) { should only be called within processPromotion delegate method. You did not post any code but wherever this try game.execute(move: move) { code is it needs to be replaced by presenting a view controller alone.
Then on delegate you do not even need to preserve the value newPiece = selectedType but rather just call try game.execute(move: move) { return selectedType }.
So about pausing a block:
It is not possible to directly "pause" a block because it is a part of execution which means the whole operation needs to pause which in the end means you need to pause your whole thread. That means you need to move the call to a separate thread and pause that one. Still this will only work if the API supports the multithreading, if the callback is called on the same tread as its execute call... So there are many tools and ways on how to lock a thread so let me just use the most primitive one which is making the thread sleep:
var executionLocked: Bool = false
func foo() {
DispatchQueue(label: "confiramtion queue").async {
self.executionLocked = true
game.execute(move: move) {
// Assuming this is still on the "confiramtion queue" queue
DispatchQueue.main.async {
// UI code needs to be executed on main thread
let pvc = PromotionViewController(nibName: nil, bundle: nil)
pvc.delegate = self
pvc.modalPresentationStyle = .overCurrentContext
self.present(pvc, animated:true)
}
while self.executionLocked {
Thread.sleep(forTimeInterval: 1.0/5.0) // Check 5 times per second if unlocked
}
return self.newPiece // Or whatever the code is
}
}
}
Now in your delegate you need:
func processPromotion(selectedType: Piece.Kind) {
defer {
boardView.isUserInteractionEnabled = true
}
newPiece = selectedType
self.executionLocked = false
}
So what happens here is we start a new thread. Then lock the execution and start execution on game instance. In the block we now execute our code on main thread and then create an "endless" loop in which a thread sleeps a bit every time (the sleep is not really needed but it prevents the loop to take too much CPU power). Now all the stuff is happening on main thread which is that a new controller is presented and user may do stuff with it... Then once done a delegate will unlock the execution lock which will make the "endless" loop exit (on another thread) and return a value to your game instance.
I do not expect you to implement this but if you will then ensure you make all precautions to correctly release the loop if needed. Like if view controller is dismissed it should unlock it, if a delegate has a "cancel" version it should exit...
Related
I am using ReactiveSwift for a while but suddenly encountered a strange bug or something.
I am using an MVVM architecture and have a simple view controller and a view model for it. The VM has a single property of type Signal<Void, NoError>. In the VM's init I send to its observer an empty value. On the view controller's viewDidLoad I listen for that signal and print some text.
Here is the view model:
class SomeVCModel {
let (someSignal, someSignalObserver) = Signal<Void, NoError>.pipe()
init() {
print (Thread.isMainThread)
DispatchQueue.main.async {
self.someSignalObserver.send(value: ())
}
}
}
And here is the view controller:
class SomeVC {
var model: SomeVCModel!
override func viewDidLoad() {
super.viewDidLoad()
model = SomeVCModel()
model.someSignal.observeValues { (_) in
print ("Some Value")
}
}
}
In this case the print statement doesn't get called. However, when I send a value to an observer by explicitly defining the main thread from DispatchQueue everything works:
DispatchQueue.main.async {
self.someSignalObserver.send(value: ())
}
I can't figure out why is this happening. Init for the view model itself is called on the main thread, so what could cause this behaviour?
Thanks in advance.
Your model sends the value during initialization.
However, your viewDidLoad function sets the observer after it is initialized, so it is not ready to observe it. Subsequent calls will work since the observer is now configured.
You have a few options, but basically you need to change the sequence.
I feel like I'm missing something and this should not be too hard.
I'm reading in some data in the initial scene in my app.
I've got a singleton and I make the call in viewDidLoad to singleton.getData().
This initial scene is part of a tab controller. And while I thought viewDidLoad would only get called once for each scene I'm pretty sure it's being called a few times during the lifecycle of my app.
So just wondering if there is a way to ensure a function call to retrieve some data only happens once.
viewDidLoad will be called when selected tab is changed, you can change the place you call getData.
If you want to call getData in viewDidLoad and be sure it won't be called multiple times you can create a flag and check, if it is previously called or not.
class Singleton {
static let sharedInstance = Singleton()
private static var getDataCalled = false
func getData() {
if Singleton.getDataCalled {
return
}
Singleton.getDataCalled = true
// request data
print("data requested")
}
}
Singleton.sharedInstance.getData()
Singleton.sharedInstance.getData()
Calling getData multiple times print data requested only once.
I use DispatchQueue.main.async in UIViewController and i need to cancel async task when dismiss UIViewController
Grand Central Dispatch does not allow tasks to be cancelled from the outside when already running.
You basically have to check inside the asynchronously running task if your view controller still exists.
Assuming your call to GCD is directly in your UIViewController, so self refers to that view controller:
DispatchQueue.global().async { [weak self] in
// Do work
// Check if self still exists:
guard let _ = self else {
return // cancels the task
}
// Continue working
}
As self is only a weak reference to your view controller it will not stop the view controller from getting deallocated when dismissed. When it gets deallocated, self inside your GCD block becomes nil and you known that you can stop your task.
So you just have to check if self is nil every once in a while in your asynchronous block.
Note: Do not perform long running tasks on the main queue but in a global queue or even a private queue. Using the main queue blocks the main thread from performing other work like UI updates so your app freezes.
Let's say I have a ViewController A and a class B.
when I press some button inside A, it calls an IBAction that calls a function B.foo() which returns an Int
B.foo() takes 8~10 seconds to finish and while it runs I'd like to put an Loading... animation on A, and when B.foo() finishes, the animation would stop.
How can I do this? this is an pseudo-code example of what I wish:
#IBAction func buttonPressed(_ sender: UIButton){
UIView.animate(blablabla......)
DO({
self.answer = B.foo()
}, andWhenItFinishesDo: {
self.someone.layer.removeAllAnimation()
})
}
This is a very common problem. One way to solve it would be to use different queues (You can think of them as lines of work that can happen in parallel).
The the basic idea is that once your button is pressed, you show your loading indicator and "dispatch" the long work to a secondary queue, that will operate in the background and do the work. This ensures that your main queue does not block while the work happens and the user interface stays responsive.
The trick is now that you want to get notified when the long work is finished so that you can stop showing the loading indicator (and possibly do even more).
While you actually could use some kind of notification system, there are other, sometimes more appropriate ways. It would actually be even more convenient, if you could just tell the long running function to call you back specifically with code that you provide.
That would be the basic concept of a "completion handler" or "callback".
The whole thing would look something like that:
// Some made up class here
class B {
// This is only static because I do not have an instance of B around.
static func foo(completion: #escaping (Int) -> Void ) {
// The method now does all of its work on a background queue and returns immediately
DispatchQueue.global(qos: .background).async {
// In the background this may take as long as it wants
let result = longCalculation()
// VERY important. The caller of this function might have a certain
// expectation about on which queue the completion handler runs.
// Here I just use the main queue because this is relatively safe.
// You could let the caller provide a queue in the function
// parameters and use it here
DispatchQueue.main.async {
// The completion handler is a function that takes an Int.
// That is exactly what you are providing here
completion(result)
}
}
}
}
#IBAction func buttonPressed(_ sender: UIButton){
self.showLoadingIndicator()
// The foo function now takes a completion handler that gets the result in.
// You have to provide this function here and do something with the result
//
// The completion handler will only be run when the foo function calls it
// (which is after the computation as you can see in the method above.
//
// I am also telling the completion handler here that self should not be
// held on to as the view controller might already have gone away when the
// long calculation finished. The `[weak self]` thingy makes that inside
// your completion handler self is an optional and might be nil (and it
// doesn't hold a strong reference to self, but that's a whole other topic)
B.foo(completion: { [weak self] result in
// Do something with the result
// Since we are called back on the main queue we can also do UI stuff safely
self?.hideLoadingIndicator()
})
}
I hope this helps a bit.
Asynchronous programming can be quite difficult to learn but there are tons of tutorials and examples you can find on this topic.
Hey Hamish you can do this in two simple ways,
First one is using the defer statements provided for functions.
Defer statement block is executed after the functions goes out of scope.
here is a simple example to describe the same.
func print1000000() {
//start displaying the loading indicator
defer {
// hide the loading indicator and move to the next ViewController
let seVC = storyboard?.instantiateViewController(withIdentifier: "SecondVC") as! SecondVC
self.navigationController?.pushViewController(seVC, animated: true)
}
// here goes the task you want to execute such as downloading a file or the one i did here
for index in 0...1000000 {
print(index)
}
}
The above function prints numbers upto 1000000 and then pushes the control to another ViewController
=========================================================================
Second way of doing it is by using closures, as described by Thomas in his answer.
I am trying to detect clicks on an UIElement like a button using Reactive Cocoa, using RAC for the first time in MVVM architecture.
I set the rac_command for my button in my ViewController.
addContactBtn.rac_command = viewModel.addContact
My ViewModel does the following:
func init(){
self.addContact = RACCommand() {
(any:AnyObject!) -> RACSignal in
return RACSignal.createSignal({
(subscriber: RACSubscriber!) -> RACDisposable! in
print("creating viewModel")
let viewModel = ContactAddViewModel(services: self.services)
self.services.pushViewModel(viewModel)
return RACDisposable(block: {
})
})
}
}
However, the command is executed only once and then the button is in disabled state when I pop the view controller and come back to original viewController. How can detect the button click any number of times?
Obviously, you missed something and had a simple mistake.
RACCommand expects to a signal which will be alive when the button clicked and be disposed after click-business-logic (like create viewModel, then pushViewModel in the above) executed. That is to say each button clicking-event associates a signal, not shares one unique signal, but has common signal inner logic. If a signal does not achieve completion or error, the responding clicking-event is not finished such that the button is disabled.
The below modified codes could be correct.
func init(){
self.addContact = RACCommand() {
(any:AnyObject!) -> RACSignal in
return RACSignal.createSignal({
(subscriber: RACSubscriber!) -> RACDisposable! in
print("creating viewModel")
let viewModel = ContactAddViewModel(services: self.services)
self.services.pushViewModel(viewModel)
// if you like to expose some value
// subscriber.sendNext(value)
subscriber.sendCompleted() // which makes clicking-event finished
// maybe error occurs
// subscriber.sendError()
return RACDisposable(block: {
})
})
}
}
I would like to advise you to checkout CocoaAction and Action in ReactiveSwift, which are replacement for RACCommand of legacy ReactiveObjC.