Rewriting SpeakerClock in SwiftUI | Cocoanetics

After I began out creating iOS apps, 11 years in the past I put a number of apps on the App Retailer. Since they turned fewer and fewer because the earnings from them didn’t warrant updating them. Amongst these my most profitable one was iWoman, which I offered in 2015. My second-most-valuable (when it comes to income) remained my beloved SpeakerClock, the final app standing.

I had left SpeakerClock on-line for the principle purpose that it stored producing like a mean of $100 per 30 days, even with out me doing something on it. For that purpose, I didn’t wish to make it free, however fairly put it to a comparatively excessive price ticket of $5. There’s additionally an In-App-Buy of one other $5. I figured “why kill the cow whereas it nonetheless produces some tasty milk”.

The opposite aspect impact of those value tags was that – I imagine – solely individuals who actually wished what the app was providing would truly buy it. My philosophy with this talking timer was to have the most important LED digits attainable, with the performance that helps the talking type of TED Talks, which traditionally have defaulted to a most size of 18 minutes.

Some crashes launched by new iOS variations prompted me to do small bug fixing releases (for iOS 3 in 2010, iOS 5 in 2011, and 2017 for iOS 10). Additionally, trying again on the launch notes of these variations, I had made this precise promise:

“Now we have completely modernised the code base in order that we will carry you some thrilling new options within the subsequent main launch”

However I didn’t lie with this assertion, a “subsequent main” launch would have been model 2.0. However I didn’t ever dare to show the model quantity up that prime. I solely elevated the third digit of the model quantity.

Apple did pressure me to do a brand new construct ultimately, once they cracked down on apps which weren’t up to date in too lengthy a time. And the newest replace they did themselves, when the Apple certificates had expired they usually re-signed my app on their servers with out me doing something.

Enter SwiftUI

Over the previous couple of months, I’ve grown very keen on SwiftUI. Being a developer on Apple platforms for greater than a decade made me fairly bored with having to maintain writing the identical MVC code numerous instances. And that may solely get you want commonplace performance, nothing actually thrilling. So I jumped on the likelihood when one in every of my purchasers requested me to implement a brand new iOS Widget in SwiftUI, within the fall of 2020. Apple had turned to SwiftUI as the one approach you can create such widgets due to SwiftUIs capacity to supply and protect a static view hierarchy which the system might present to the consumer at sure factors in a timeline with out substantial energy utilization.

My shopper was comfortable concerning the consequence and so I used to be tasked with the following degree of SwiftUI improvement. I wanted to implement a watchOS app, additionally totally in SwiftUI. Growth was fairly just like the widget, however this time I additionally wanted to cope with consumer interplay and communication with the iOS counterpart app. That every one took some a number of months greater than the widget, however once more elevated my SwiftUI abilities tremendously.

After having delivered the watch app, I had somewhat additional time accessible to do one thing for myself. I do have another concepts for apps, however my ideas turned to SpeakerClock. I figured that this extremely customized UI would lend itself properly to be carried out in SwiftUI.

Paths in Shapes

A very powerful asset within the legacy code was the drawing of the large pink LED digits and the way they organize themselves in portrait versus panorama, in a pleasant animation. So my first SwiftUI view was one which had a Path component with the SwiftUI instructions including the trail parts to make up the person bars of the LED. My first error right here involved utilizing a GeometryReader to find out the size of the trail. The LED digits have a hard and fast side ratio and the drawing coordinates are based mostly on these.

struct LEDDigit: View
{
   var digit: Int? = nil
	
   var physique: some View
   {
      GeometryReader { proxy in
         let (w, h) = proxy.unitSize

         // high horizontal line
         Path { path in
            path.transfer(to: CGPoint(x: 24 * w, y: 7 * h))
            path.addLine(to: CGPoint(x: 60 * w, y: 7 * h))
            path.addLine(to: CGPoint(x: 62 * w, y: 10 * h))
            path.addLine(to: CGPoint(x: 57 * w, y: 15 * h))
            path.addLine(to: CGPoint(x: 24 * w, y: 15 * h))
            path.addLine(to: CGPoint(x: 21 * w, y: 10 * h))
            path.closeSubpath()
         }
         .activeLEDEffect(when: [0, 2, 3, 5, 7, 8, 9].incorporates(digit))
         ...
}

Whereas this produces the proper output, it causes the person Paths to animate individually when rotating the gadget. I solved this drawback by transferring the person path’s code right into a Form the place I’m including the bars solely based mostly on whether or not I’m on the lookout for the energetic or inactive LED parts. The trail(in rect: CGRect) operate palms us the required measurement, so we don’t a GeometryReader any extra.

struct LEDDigitShape: Form
{
   var digit: Int? = nil
   var isActive: Bool
	
   func path(in rect: CGRect) -> Path
   {
      let w = rect.measurement.width / 73
      let h = rect.measurement.top / 110
		
      var path = Path()
		
      // high horizontal line
		
      if [0, 2, 3, 5, 7, 8, 9].incorporates(digit) == isActive
      {
         path.transfer(to: CGPoint(x: 24 * w, y: 7 * h))
         path.addLine(to: CGPoint(x: 60 * w, y: 7 * h))
         path.addLine(to: CGPoint(x: 62 * w, y: 10 * h))
         path.addLine(to: CGPoint(x: 57 * w, y: 15 * h))
         path.addLine(to: CGPoint(x: 24 * w, y: 15 * h))
         path.addLine(to: CGPoint(x: 21 * w, y: 10 * h))
         path.closeSubpath()
      }
      ...
}

That is used such:

struct LEDDigit: View
{
   var digit: Int? = nil
	
   var physique: some View
   {
   ZStack
   {
      LEDDigitShape(digit: digit, dot: dot, isActive: false)
         .activeLEDEffect(isActive: false)
      LEDDigitShape(digit: digit, dot: dot, isActive: true)
         .activeLEDEffect(isActive: true)
   }
}

The 2 members of the ZStack draw all of the inactive LED parts behind the energetic LED parts. It nonetheless wanted to be two Shapes as a result of one form can solely have a single drawing type. The inactive parts are merely stuffed in a grey. The energetic parts are crammed with pink and have a pink glow round them simulating some radiance.

With this strategy a digit is all the time drawn in its entirety which lends itself to easy resizing.

Structure and Orientation Woes

The following step was to combination a number of LED digits and lay them out over the display screen with totally different positions for panorama and portrait orientations, with a easy animation once you rotate the gadget.

I’ve principally two layouts:

1. Hour digits, Colon, Minute digits (in a HStack)- in horizontal format with the outer sides touching the secure space insets
2. A VStack of Hour digits and Minute digits – in vertical format

Sounds simple, however my makes an attempt with HStacks and VStacks failed miserably. In the beginning of the rotation animation the digits would all the time get a really small body increasing into the ultimate one.

I can solely think about that someway the SwiftUI format system doesn’t do not forget that these are the identical views. So I attempted giving them static identifiers and I additionally tried geometry matching. However I couldn’t shake these animation artefacts. There have to be some piece lacking in my understanding about view id.

Ultimately I got here again to doing my very own format inside a GeometryReader, setting body’s width/top and acceptable offsets (i.e. translation) for particular person parts. This works very properly and in addition lets me have a separate animation for the opacity of the colon.

The colon sticks to the proper aspect of the hour digits and disappears in portrait format. By sorting view modifiers in a sure approach I used to be in a position to get this impact that the colon fades in with a slight delay.

var physique: some View
{
   GeometryReader { proxy in
			
   let digitSize = self.digitSize(proxy: proxy)
   let colonSize = self.colonSize(proxy: proxy)
   let centeringOffset = self.centeringOffset(proxy: proxy)
   let isLandscape = proxy.isLandscape
			
   let timerSize = self.timerSize(proxy: proxy)
			
   Group
   {
      LEDNumber(worth: mannequin.countdown.minutes)
      .body(width: digitSize.width * 2, top: digitSize.top)
      .animation(nil)
				
      LEDColon()
      .body(width: colonSize.width, top: colonSize.top)
      .offset(x: digitSize.width * 2, y: 0)
      .animation(nil)
      .opacity(isLandscape ? 1 : 0)
      .animation(isPadOrPhone ? (isLandscape ? .easeInOut.delay(0.2) 
                              : .easeInOut) : nil)
				
      LEDNumber(worth: mannequin.countdown.seconds)
      .body(width: digitSize.width * 2, top: digitSize.top)
      .offset(x: isLandscape ? digitSize.width * 2 + colonSize.width : 0,
              y: isLandscape ? 0 : digitSize.top)
      .animation(nil)
   }
   .offset(x: centeringOffset.width,
           y: centeringOffset.top)

You may see that I’m particularly disabling animation with .animation(nil) for probably the most elements as a result of I discovered that the animation in any other case is all the time out of sync with the rotation resizing animation. The LED colon however has its personal animation with a further delay of 0.2 seconds.

The second purpose why I explicitly disabled animations is as a result of on the Mac model these animations would lag behind the resizing of the app’s window. This resizing additionally switches between each layouts relying on the way you drag the window nook, kind of like “responsive design” as we have now seen on HTML net pages. Extra on Mac issues additional down under.

Multi-Modal Buttons

One other problem that had me attempt a number of approaches involved the preset buttons (high left) and site visitors gentle buttons (heart backside). These buttons have a distinct operate for a single faucet (choose) versus a protracted press (set).

The primary drawback is that you simply can’t have a easy .onLongPressGesture as a result of this prevents the traditional faucets from being dealt with. One strategy is to have a .simultaneousGesture for the lengthy press, however then the faucet motion is executed proper (i.e. “simultaneous”) after the lengthy press motion in the event you elevate the finger over the button. The opposite strategy is to make use of a .highPriorityGesture which once more disables the built-in faucet.

I ended up with the next strategy which makes use of the gesture masks to selectively disable the lengthy press gesture if there isn’t any lengthy press motion and to disable the faucet gesture if a protracted press was detected.

struct LEDButton: View
{
   var motion: ()->()
   var longPressAction: (()->())?
   @ViewBuilder var content material: ()->Content material
	
   @State fileprivate var didLongPress = false
	
   var physique: some View
   {
      Button(motion: {}, label: content material)  // will need to have empty motion
      .contentShape(Circle())
      .buttonStyle(PlainButtonStyle())   // wanted for Mac
      .simultaneousGesture(LongPressGesture().onEnded({ _ in
         didLongPress = true
         longPressAction!()
         didLongPress = false
      }), together with: longPressAction != nil ? .all : .subviews)
      .highPriorityGesture(TapGesture().onEnded({ _ in
         motion()
      }), together with: didLongPress ? .subviews : .all)
   }
}

This strategy makes use of a customized TapGesture in tandem with the LongPressGesture. A @State variable retains observe of the lengthy press. We do must reset didLongPress to false or else all subsequent faucets would proceed to be ignored. I discovered that I don’t want a dispatch async for placing it again to false.

I imagine that the explanation for that’s that the primary setting of the variable causes the physique to be up to date and thus the together with: to disable the faucet gesture whereas in progress. Thus the faucet doesn’t hearth upon releasing the lengthy press. Good to know: The .all allows the gesture and the .subviews disables a gesture.

Opposite to different approaches I’ve seen on the web this strategy preserves the usual habits of Button for highlighting, Whilst you press a customized button like this, it makes it barely clear.

A Mac Model – For Free?

The massive promise of SwiftUI is that you’d get a Mac model of your app for little additional work, successfully “without cost”. So I made a decision to place this to the take a look at additionally produce a macOS model. I set the focused units to iPhone, iPad, Mac and selected the “Optimize Interface for Mac” as a result of that sounded to me like the higher consequence.

This optimized mode prompted some points for my customized buttons, as a result of they bought changed with empty spherical rects destroying my customized look. You may stop this modification by including .buttonStyle(PlainButtonStyle()).

Aside from this my code actually did run as a local Mac app fairly properly. Behind the scenes although it’s all Mac Catalyst. As I perceive it, meaning UIKit remains to be on the helm, on Mac only a macOS model of it.

I left the code signing settings alone as I wished to have customers be capable of set up the Mac and iOS variations with the identical buy. This “common buy” is enabled by having the identical bundle identifier for each variations.

Some very minor tweaks had been required for adjusting some minimal and most button sizes. There’s a bug on macOS that stumped me for some time. Solely on Mac I discovered that once I tapped in sure spots in my app this may trigger gestures to cease working. Then once I triggered a brand new format by resizing the window, all the things returned again to regular.

My workaround for this was to connect the Pan Gesture (for setting the timer) solely to the LED digits. This fashion there isn’t any interference and all buttons proceed to work usually. The system may get confused by having too many conflicting gestures on high of one another.

A side-effect of the Mac model is that you simply begin to connect keyboard shortcuts to buttons. This was additionally a purpose why I wished to get Button to work with faucet and lengthy press versus making a customized view that’s not a button.

let title = "(index+1)"

PresetButton()
.keyboardShortcut(KeyEquivalent(title.first!), modifiers: [.command])

This fashion you may set off the preset buttons additionally with COMMAND plus quantity. And never only for the Mac app, however that works for iPads with hooked up keyboard as properly.

That bought me considering, that perhaps it will be nice to permit the house bar to cease/begin the timer, like we’re used to from video gamers. For that goal I’ve an empty utterly black button behind the LED digits:

Button(motion: { mannequin.isTimerActive.toggle() },
       label: {
          Rectangle()
          .foregroundColor(.black)
          .body(width: timerSize.width, top: timerSize.top)
          .onTapGesture(depend: 2) { mannequin.restoreGreenTime() }
       })
.keyboardShortcut(.house, modifiers: [])
.buttonStyle(PlainButtonStyle())

This button permits me so as to add a keyboard shortcut for house to behave the identical as a faucet. Curiously having a two-tap gesture hooked up to the Rectangle() poses no drawback.

I submitted the Mac construct proper after the one for iOS however initially bought a stunning rejection:

The consumer interface of your app shouldn’t be in line with the macOS Human Interface Tips. Particularly:

We discovered that the app incorporates iOS contact management directions equivalent to faucet and swipe.

The explanation for that was that I put again the assistance display screen with a textual content I had beforehand written with iOS in thoughts. I wanted to exchange mentions of swiping with dragging and as an alternative of tapping you’re clicking. I’ve onerous coded the textual content and formatting for now and with and #if I can change the textual content between a model for Mac and one for iOS.

Group
{
   Textual content("Setting the Timer")
   .font(.headline)
   .padding(.backside, 5)
						
#if targetEnvironment(macCatalyst)
   Textual content("To regulate the timer, click on on the LED digits and drag horizontally.")
   .font(.physique)
   .padding(.backside, 5)
#else
   Textual content("To regulate the timer swipe left and proper.")
   .font(.physique)
   .padding(.backside, 5)
#endif					
}

As soon as I had made these adjustments the Mac app was permitted in a short time.

Conclusion

I’ve skilled first hand how I can rewrite an app in SwiftUI and the good pleasure that may be had from deleting all of your crufty Goal-C code when doing so.

SwiftUI is my new love and this manner my app is not a “baby from one other mom”. This restores some enthusiasm in me to truly lastly actually add some long-promised “thrilling new options”. For starters I’m considering of getting a watchOS companion app which reveals the timer and permits you to distant management it. One other concept could be to retailer my presets on iCloud in order that they’re the identical on all my units.

I might love to listen to from you what you concentrate on the method of re-implementing elements of apps and even entire apps in SwiftUI.

Additionally revealed on Medium.

Tagged as: SpeakerClock

Classes: Updates