I have made some search here and there but I didn't find (or understand) how to make a delay in a loop in SpriteKit.
This is the idea : I have some SKSpriteNodes sorted in an array and I want to display them on the screen, one every second. The problem is... I simply don't manage to do it (sorry, I'm quite new to this).
let sprite1 = SKSpriteNode(color: .red, size: CGSize(width: 20, height: 20))
sprite1.position = CGPoint(x: 100, y: 100)
let sprite2 = SKSpriteNode(color: .red, size: CGSize(width: 20, height: 20))
sprite2.position = CGPoint(x: 100, y: 300)
let sprite3 = SKSpriteNode(color: .red, size: CGSize(width: 20, height: 20))
sprite3.position = CGPoint(x: 300, y: 100)
let sprite4 = SKSpriteNode(color: .red, size: CGSize(width: 20, height: 20))
sprite4.position = CGPoint(x: 300, y: 300)
let allSprites : [SKSpriteNode] = [sprite1, sprite2, sprite3, sprite4]
let delay = 1.0
var i = 0
while i <= 3
{
let oneSprite = allSprites[i]
self.addChild(oneSprite)
DispatchQueue.main.asyncAfter(deadline : .now() + delay) {
i = i + 1
}
}
If you're asking : this doesn't work at all. It seems that what is inside the DispatchQueue.main.asyncAfter isn't read.
So, if you can help me to understand why, that would be great. I'm not picky, I can't take the answer with a "for" loop.
Regards,
This is a common beginner misunderstanding of event-driven / run-loop-based / GUI programming systems. Some key points to help get on the right track:
SpriteKit is trying to render the scene 60 (or so) times per second.
That means SpriteKit internally has a loop where, once per frame, it calls your code asking what's new (update), then runs its own code to draw the changes (render).
Setup code, or things that happen in response to events (clicks/taps/buttons) are external to this loop, but feed into it: events change state, which the render loop reacts to.
So, if you have a loop in an update method, an event handler, or initial setup code, everything that happens in that loop will happen before SpriteKit gets a chance to draw any of it. That is, if we ignore for a moment the "delay" part of your question, a loop like this...
var i = 0
while i <= 3 {
let oneSprite = allSprites[i]
self.addChild(oneSprite)
}
... will result in none of the sprites being visible right before the loop starts, and all of the sprites being visible after it completes. No amount of introducing delays within the loop will change that, because SpriteKit doesn't get its first chance to draw until after your loop is finished.
Fixing the problem
The best way to do animations and delays in a system like SpriteKit is to make use of the tools it gives you for treating animation declaratively. That is, you tell SpriteKit what you want to have happen over the next several (hundred) frames, and SpriteKit makes that happen — each time it goes through that update/render loop, SpriteKit determines what changes need to be made in the scene to accomplish your animation. For example, if you're running at 60 fps, and you ask for a fade-out animation on some sprite lasting one second, then each frame it'll reduce that sprite's opacity by 1/60.
SpriteKit's tool for declarative animations is SKAction. There are actions for most of the things you'd want to animate, and actions for composing other actions into groups and sequences. In your case, you probably want some combination of the wait and unhide and run(_:onChildWithName) actions:
Give each of your sprites a unique name:
let sprite1 = // ...
sprite1.name = "sprite1"
// etc
Add all the nodes to your scene, but keep the ones you don't want visible yet hidden:
let allSprites : [SKSpriteNode] = [sprite1, sprite2, sprite3, sprite4]
for sprite in allSprites {
sprite.isHidden = true
self.addChild(sprite)
}
Create a sequence action that alternates delays with telling each node to unhide, and run that action on your scene:
let action = SKAction.sequence([
run(.unhide(), onChildNodeWithName: "sprite1"),
wait(forDuration: 1.0),
run(.unhide(), onChildNodeWithName: "sprite2"),
wait(forDuration: 1.0),
// etc
])
self.run(action)
That should accomplish what you're looking for. (Disclaimer: code written in web browser, may require tweaking.) Read on if you want to better understand the situation...
Other alternatives
If you really want to get your head around how the update/render loop works, you could participate directly in that loop. I wouldn't recommend it in general, because it's a lot more code and bookkeeping, but it's useful to know.
Keep a queue (array) of nodes that you want to add or unhide.
In you scene's update method, keep track of how much time has passed.
If it's been at least 1.0 second (or whatever delay you're after) since you last showed a node, add the first one in the array, and remove it from the array.
When the array is empty, you're done.
About DispatchQueue.asyncAfter...
This isn't essential to completing your task, but helpful to understand so you don't get into similar problem later: the word "async" in the call you were using to try delaying things is short for "asynchronous". Asynchronous means that the code you're writing doesn't execute in the order you write it in.
A simplified explanation: Remember that modern CPUs have multiple cores? When CPU 0 is executing your while loop, it gets to the asyncAfter call and passes a note to CPU 1 saying to wait one second and then execute the closure body attached to that asyncAfter call. As soon as it passes that note, CPU 0 continues merrily on its way — it doesn't wait for CPU 1 to receive the node and complete the work that note specifies.
Just from reading the code, I'm not 100% clear on how this approach fails, but fail it certainly does. Either you have an infinite loop, because the while condition never changes, because the work to update i is happening elsewhere and not propagating back to local scope. Or the the changes to i do propagate back, but only after the loop spins an indeterminate number of times waiting for the four asyncAfter calls to finish waiting and execute.
Related
I have a game which is running well but as soon as I introduce a while loop my entire UI goes blank.
I have some code which generates a sprite from an array and moves it down the screen
func addALetter() {
let randomX = CGFloat.random(in: 50...size.width - 50)
let shrink = SKAction.scale(to: 0.1, duration: 0.01)
let grow = SKAction.scale(to: 1, duration: 0.5)
let wait = SKAction.wait(forDuration: 0.7)
let spawn = SKAction.move(to: CGPoint(x: randomX, y: size.height - tileSize), duration: 0.001)
let move = SKAction.moveTo(y: -500, duration: 7)
let sequence = SKAction.sequence([shrink, spawn, grow, move, wait])
// scroll through the lettersArray
if activeLetter < lettersArray.count - 1 {
bubbles[activeLetter].removeAllActions()
bubbles[activeLetter].run(sequence)
activeLetter += 1
} else {
// go back to first in letter array
activeLetter = 0
bubbles[activeLetter].removeAllActions()
bubbles[activeLetter].run(sequence)
}
}
It is working fine triggered using an SKAction in my didMove to view run(SKAction.repeatForever(SKAction.sequence([SKAction.run(addALetter), SKAction.wait(forDuration: spawnTime)])))
but I have problems with that as I get to the end of the array because the action repeats too frequently making sprites disappear before I want them too.
So I tried using a while loop instead...
while gameOn == true {
addALetter()
}
to repeat the action. But then I get a completely blank UI - I assume because it's then stuck in the loop there's no time to update the UI?
Looking for a solid way to repeat the function that I can vary the frequency as the array gets to low numbers
It seems likely that the problem is that spawnTime is small enough that you're wrapping around through the letters too quickly and start doing bubbles[activeLetter].removeAllActions() before the bubble's previous action sequence has finished.
I think the most best way to deal with this sort of situation is to coordinate through completion blocks. I.e., use the run method with a closure to be called after the action finishes, https://developer.apple.com/documentation/spritekit/sknode/1483103-run. That way you don't wind up trying to adjust delays explicitly in an effort to keep actions from being prematurely cancelled. The completion blocks will update state that you can use to coordinate, or they can run other actions directly.
It's not clear from your question what behavior you want. But, e.g., you might have addALetter set a flag for the bubble when it starts the action sequence and include a completion block for the sequence to clear the flag. Then before addALetter restarts the sequence for a bubble it can make sure that the flag is clear; if it's not (the bubble is still running the previous sequence), just return without doing anything.
Still new, but slowly building my first app/game, and am slowly getting there.
I'd like to be able to add a fading streak or trail as one of my SKSpriteNodes moves. Whether it is moving due to touchesMoved or being sent back to its original spot by code. I just want to add a nice visual effect.
The only thing I can think of is calculating the distance, breaking it down into x movements, then gradually move the Sprite a little, with some fade options, and then repeat in a loop till it gets back to home base, using a lot of nested SKActions and sequences.
I know that has to be wrong because it's just so ugly.
When I look at the Sprite Kit's Particle File, it has so few options. And I'm not really sure that's what I should be looking at. I have also looked at SKAction's options, and if there's an option there I'm missing it.
Surely in Swift's huge animation library there has to be something?
Let's create a basic sprite and an emitter, and make the emitter a child of the sprite so that it follows it:
let sprite = SKSpriteNode(color: .white, size: CGSize(width: 20, height: 10))
let emitter = SKEmitterNode() // better to use the visual designer in Xcode...
emitter.particleLifetime = 1.0
emitter.particleBirthRate = 50.0
emitter.particleSpeed = 100.0
emitter.emissionAngleRange = .pi / 5
emitter.particleTexture = SKTexture(imageNamed: "spark")
emitter.particleScale = 0.1
emitter.particleAlphaSpeed = -1.0
emitter.emissionAngle = -.pi
sprite.addChild(emitter) // attach it to the sprite
emitter.position.x = -15 // but not to the center
scene.addChild(sprite)
sprite.run(SKAction.group([ // let's run some actions to test it:
SKAction.sequence([
SKAction.move(to: CGPoint(x: 200, y: 200), duration: 5),
SKAction.move(to: CGPoint(x: 50, y: 50), duration: 5),
]),
SKAction.rotate(byAngle: .pi * 2.0, duration: 10)
]))
(Click to open animated GIF if it doesn't display correctly:)
To the casual observer, it looks fine… Except that, after some scrutiny, you'll realize what's off: the particles emitted live in the universe of the parent sprite, moving and rotating with it, even long after they were emitted! That's not right!
That's because the targetNode of the emitter is its parent, and it should be the scene!
So let's insert this line somewhere:
emitter.targetNode = scene // otherwise the particles would keep moving with the sprite
(Click to open animated GIF if it doesn't display correctly:)
Okay, this is a no-go: the particles now stay in the "universe" of the scene, but apparently their emission angle fails to follow that of the parent (which looks like a bug to me).
Luckily, we can attach a custom action to the emitter which keeps aligning this angle to the parent sprite:
emitter.run(SKAction.repeatForever(
SKAction.customAction(withDuration: 1) {
($0 as! SKEmitterNode).emissionAngle = sprite.zRotation + .pi
_ = $1
}))
(Click to open animated GIF if it doesn't display correctly:)
Okay, now new particles are launched in the correct direction, and keep moving that way even if the sprite moves or rotates in the meantime. This seems to be the most realistic simulation so far, though there may still be ways to improve it by modifying the behavior of the emitter on the fly.
Sorry for the jaggy GIFs, apparently my Mac is too slow to render and capture at the same time. The animations themselves run just fine.
I'm making app, that can build graphs.
I'm working with UIBezierPath, and everything was fine until I decided to make it possible to move and scale the graph using gestures.
here's how i'm drawing:
var isFirstPoint = true
color.set()
funcPath.lineWidth = width
for i in 0...Int(bounds.size.width * contentScaleFactor) {
let cgX = CGFloat(i) / contentScaleFactor
let funcX = Double ((cgX - bounds.width / 2) / scale)
let funcY = function(funcX) * Double(scale)
guard !funcY.isNaN else { continue }
let cgY = -( CGFloat(funcY) - bounds.height / 2)
if isFirstPoint {
funcPath.move(to: CGPoint(x: cgX, y: cgY))
isFirstPoint = false
} else {
if cgY > max(bounds.size.width, bounds.size.height) * 2 {
isFirstPoint = true
} else {
funcPath.addLine(to: CGPoint(x: cgX, y: cgY) )
}
}
}
funcPath.stroke()
is there is a faster way to do it?
A couple of thoughts:
Rebuilding the path for every pixel can be an expensive process. One way to reduce this cost is to reduce the number of data points you render. For example, if you’re adjusting scale and rebuilding the whole path, you might want to use some step factor when updating the path mid-gesture, and replace your for loop with:
let step = 4
for i in stride(from: 0, through: Int(bounds.size.width * contentScaleFactor), by: step) {
...
}
Then, when the gesture ends, you might update the path one more time, with a step of 1.
The other way to do this is to not update the scale and call setNeedsDisplay each time, but rather to just update the transform of the view. (This can result in super fast behavior when zooming in, but is obviously problematic when zooming out, as portions excluded from the path won’t be rendered.)
If you’re testing this, make sure to test release (i.e. optimized) build on physical device. Debug builds, with all of their safety checks, can really slow down the process, which will stand out with something as computationally intensive as this.
By the way, it looks like the building of the path is buried in draw(rect:). I’d decouple the building of the UIBezierPath from the stroking of the path because while, yes, every time you update the function, you may want to update and draw the path, you definitely don’t want re-build the path every time draw(rect:) is called.
Once you do that, the draw(rect:) might not even be needed any. You might just add a CAShapeLayer as sublayer of the view’s layer, set the strokeColor and strokeWidth of that shape layer, and then update its path.
Generally moving and scaling is done by applying transforms to layers. Look at CAShapeLayer to hold your path. As the user gestures, apply transforms; then recreate the path when the gesture completes. In more advanced usage, you may recalculate the actual path more often than just when the user stops (such as if they pause without letting go), but these are refinements.
If you're able to describe the translation (pan) and scale as a CGAffineTransform then you can apply it to your UIBezierPath in one go using apply(_:).
Hello I'm trying to spawn bullets at the bottom of my screen to travel upwards but the current code that I have spawns the bullets at the top of the screen. I've tried making the height negative and nothing happened. Here's the code I'm working with, thanks.
let randomBulletPosition = GKRandomDistribution(lowestValue: -300, highestValue: 300)
let position = CGFloat(randomBulletPosition.nextInt())
bullet.position = CGPoint(x: position, y: self.frame.size.height + bullet.size.height)
Some nice conversions will help you.
Now, do not do this all the time, this should be a one and done type deal, like in a lazy property.
First, we want to get the bottom of our view
let viewBottom = CGPoint(x:scene!.view!.midX,y:scene!.view!.frame.maxY) //In a UIView, 0,0 is the top left corner, so we look to bottom middle
Second, we want to convert the position to the scene
let sceneBottom = scene!.view!.convert(viewBottom, to:scene!)
Finally we want to convert to whatever node you need it to be a part of. (This is optional if you want to place it on the scene)
let nodeBottom = scene!.convert(sceneBottom,to:node)
Code should look like this:
let viewBottom = CGPoint(x:scene!.view!.midX,y:scene!.view!.frame.maxY)
let sceneBottom = scene!.view!.convert(viewBottom!, to:scene!)
let nodeBottom = scene!.convert(sceneBottom,to:node)
Of course, this is a little ugly.
Thankfully we have convertPoint and convert(_from:) to clean things up a little bit
let sceneBottom = scene.convertPoint(from:viewBottom)
Which means we can clean up the code to look like this:
let sceneBottom = scene.convertPoint(from:CGPoint(x:scene!.view!.midX,y:scene!.view!.frame.maxY))
let nodeBottom = node.convert(sceneBottom,from:scene!)
Then we can make it 1 line as:
let nodeBottom = node.convert(scene.convertPoint(from:CGPoint(x:scene!.view!.midX,y:scene!.view!.frame.maxY),from:scene!)
As long as the node is available to the class, we can make it lazy:
lazy var nodeBottom = self.node.convert(self.scene!.convertPoint(CGPoint(x:self.scene!.view!.midX,y:self.scene!.view!.frame.maxY),from:self.scene!)
This means the first time you call nodeBottom, it will do these calculations for you and store it into memory. Everytime after that, the number is preserved.
Now that you know where the bottom of the screen is in the coordinate system you want to use, you can assign the x value to whatever your random is producing, and you can subtract the (node.height * (1 - node.anchorPoint.y)) to fully hide your node from the scene.
Now keep in mind, if your node moves between various parents, this lazy will not update.
Also note, I unwrapped all optionals with !, you may want to be using ? and checking if it exists first.
Running Xcode 8.2 on a Swift Playground. It's also making Xcode not respond. This is the error I'm getting.
Context leak detected, msgtracer returned -1
I've traced the problem to these lines of code
while (true){
let rotate = SKAction.rotate(toAngle: angle - CGFloat(M_PI_2), duration: 1)
crank.run(rotate)
}
When I comment out the while loop and leave just the inside it works fine.
The while (true){ part will make that loop execute forever and your code probably will not proceed beyond that point. If you want an action to repeat forever there is an SKAction method for that - take a look at repeat.
If you remove the while (true){ condition and instead use a repeat action, I believe your code should run fine.
#Fahim is 100% correct, by the code posted you are stuck in an endless loop. So the question for you is how to get out of it?
The question for you is - how to make this loop do what you need. I'll guess that you want to change the condition to something like while (someVar == true) and put in a condition inside your loop to change someVar to false. But only you know what that condition is. The syntax would be:
let someVar = true
while (someVar == true){
let rotate = SKAction.rotate(toAngle: angle - CGFloat(M_PI_2), duration: 1)
crank.run(rotate)
if someCondition == isMet { // THIS IS FOR YOU TO DECIDE!
someVar = false
}
}
I ended up rewriting it completely and using SKAction.repeatForever to run an SKAction (to rotate) and that terminated only if the condition was false.
crank.run(
SKAction.repeatForever (
SKAction.sequence([
SKAction.wait(forDuration: 0.1),
SKAction.run({
let angle = atan2(self.point.y - crank.position.y , self.point.x - crank.position.x)
let rotate = SKAction.rotate(toAngle: angle - CGFloat(M_PI_2), duration: 0.25)
crank.run(rotate)
if(!(crank.frame.contains(self.point))) {
self.removeAction(forKey: "New Thread")
}
})
])
),
withKey: "New Thread"
)
I think the existing answers are missing the point here. You have two "threads" you are dealing with: the active thread running the routine (which will keep pumping new rotate actions onto the actions list for the node as fast as it can) and the action processing thread (which is really not a "thread" per se but can be imagined as one for the purpose of this question) which is where the rotate actions are actually executed.
Essentially, in a given timeslice, the "main" execution thread might pump 100k "rotate" actions onto your node's action list, and the SpriteKit engine will process all of them at the same time as efficiently as it can (I believe by removing duplicate actions but I'm not positive on that).
The effect is that you are doing a lot more work than you have to, and burning CPU in both your spin loop and in the SKAction processing code trying to recover from the millions of actions added to the node every time it comes up to process.
Replace it with a repeat action, as Fahim suggested, and you'll be good. Then you just need a trigger of some sort to stop the action (give it a name to end just the one action, or stop all actions on this node when the trigger fires).
I solve this problem using more singletong. For example SKView() inside the loop in singletong.