Incorporating Predicted Touches into an App

Learn how to create a simple app that incorporates predicted touches into its drawing code.


The sample app Speed Sketch (see Leveraging Touch Input for Drawing Apps) uses predicted touches to minimize latency when drawing using either Apple Pencil or a finger. The key class for gathering touches is the StrokeGestureRecognizer class. Each new sequence of touch events results in the creation of a Stroke object to the app’s drawing canvas. Stroke objects store the touch data needed to do stylized line drawing and can render that data using a calligraphy pen or a regular pen, or in a special debug mode that draws line segments for each distinct touch event.

Figure 1

Speed Sketch drawing modes

An illustration of the calligraphy, pen, and debug drawing modes in Speed Sketch.

Collect the Touch Input

The StrokeGestureRecognizer class collects drawing-related touch input and uses it to create a Stroke object representing the path to render. In addition to the touches that actually occurred, the class also gathers any predicted touches. Listing 1 shows the portion of the gesture recognizer’s append method that is responsible for gathering the predicted touches. The collector block called by this code processes each touch event. The parameters to that block indicate whether the touch is an actual touch or a predicted touch.

Listing 1

Collecting predicted touches in Speed Sketch

// Collect predicted touches only while the gesture is ongoing. 
if (usesPredictedSamples && stroke.state == .active) {
   if let predictedTouches = event?.predictedTouches(for: touchToAppend) {
      for touch in predictedTouches {
         collector(stroke, touch, view, false, true)

The collection of touch input results in the creation of StrokeSample objects, which are then added to the current Stroke object. Stroke objects store predicted touches separately from other touches. Keeping them separate makes it easier to remove them later, and keeps them from being accidentally confused with the real touch input. Each time the app adds a new set of actual touches, it discards the preceding set of predicted samples.

Listing 2 shows a portion of the Stroke class, which represents the touches associated with a single drawn line. For each new set of touches, the class adds the actual touches to its primary list of samples. Any predicted touches are then stored in the predictedSamples property. Each time StrokeGestureRecognizer calls the Stroke method add, the method moves the last set of predicted touches to the previousPredictedSamples property and are ultimately discarded. Thus, Stroke maintains only the last set of predicted touches.

Listing 2

Managing predicted samples in the Stroke class

class Stroke {
    static let calligraphyFallbackAzimuthUnitVector = CGVector(dx: 1.0, dy:1.0).normalize! 
    var samples: [StrokeSample] = []
    var predictedSamples: [StrokeSample] = []
    var previousPredictedSamples: [StrokeSample]?
    var state: StrokeState = .active
    var sampleIndicesExpectingUpdates = Set<Int>()
    var expectsAltitudeAzimuthBackfill = false
    var hasUpdatesFromStartTo: Int?
    var hasUpdatesAtEndFrom: Int? 
    var receivedAllNeededUpdatesBlock: (() -> ())?
    func add(sample: StrokeSample) -> Int {
        let resultIndex = samples.count
        if hasUpdatesAtEndFrom == nil {
            hasUpdatesAtEndFrom = resultIndex
        if previousPredictedSamples == nil {
            previousPredictedSamples = predictedSamples
        if sample.estimatedPropertiesExpectingUpdates != [] {
        return resultIndex
    func addPredicted(sample: StrokeSample) {
    func clearUpdateInfo() {
        hasUpdatesFromStartTo = nil
        hasUpdatesAtEndFrom = nil
        previousPredictedSamples = nil
    // Other methods...

Render the Predicted Touches

During rendering, the app treats predicted touches like actual touches. It breaks down the contents of each Stroke object into one or more StrokeSegment objects, which the drawing code fetches using a StrokeSegmentIterator object. Listing 3 shows the implementation of this class. As the drawing code iterates over the stroke samples, the sampleAt method returns the samples for the actual touches first. Only after the method returns all of the actual touch samples does the iterator return the samples for any predicted touches. Thus, the predicted touches are always located at the end of the stroked line.

Listing 3

Fetching predicted touches during drawing

class StrokeSegmentIterator: IteratorProtocol {
    private let stroke: Stroke
    private var nextIndex: Int
    private let sampleCount: Int
    private let predictedSampleCount: Int
    private var segment: StrokeSegment!
    init(stroke: Stroke) {
        self.stroke = stroke
        nextIndex = 1
        sampleCount = stroke.samples.count
        predictedSampleCount = stroke.predictedSamples.count
        if (predictedSampleCount + sampleCount > 1) {
            segment = StrokeSegment(sample: sampleAt(0)!)
            segment.advanceWithSample(incomingSample: sampleAt(1))
    func sampleAt(_ index: Int) -> StrokeSample? {
        if (index < sampleCount) {
            return stroke.samples[index]
        let predictedIndex = index - sampleCount
        if predictedIndex < predictedSampleCount {
            return stroke.predictedSamples[predictedIndex]
        } else {
            return nil
    func next() -> StrokeSegment? {
        nextIndex += 1
        if let segment = self.segment {
            if segment.advanceWithSample(incomingSample: sampleAt(nextIndex)) {
                return segment
        return nil