Wrapping View Controllers in Signals

Signals aren’t just only great for handling model state - you can use them as well to plan your applications’ flow.

Most people think that functional reactive programming is only useful to control your applications’ model. But it’s also a pretty handy concept to layout your applications’ flow as well. You can achieve this by wrapping your view controllers in an async transform.

For this example we’ll build a transform that let’s the user take a picture and returns a Result<UIImage>. We’ll use UIImagePickerController and implement it’s delegate in a class:

class ImageCapture: NSObject, UIImagePickerControllerDelegate, UINavigationControllerDelegate {
    private var completion: (Result<UIImage> -> Void)?
    let vc: UIViewController

    init(viewController: UIViewController) {
        vc = viewController
    }

    func takePicture(source: UIImagePickerControllerSourceType, completion: (Result<UIImage>->Void)){
        self.completion = completion
        let picker = UIImagePickerController()
        picker.delegate = self
        picker.sourceType = source
        vc.presentViewController(picker, animated: true, completion: nil)
    }

    func imagePickerControllerDidCancel(picker: UIImagePickerController) {
        let error = NSError(domain: "User did cancel", code: 401, userInfo: nil)
        completion?(Result.Error(error))
        picker.presentingViewController?.dismissViewControllerAnimated(true, completion: nil)
    }

    func imagePickerController(picker: UIImagePickerController, didFinishPickingImage image: UIImage!, editingInfo: [NSObject : AnyObject]!) {
        completion?(Result.Success(image))
        picker.presentingViewController?.dismissViewControllerAnimated(true, completion: nil)
    }
}

Let’s explore this code step by step:

1. ImageCapture is an NSObject so it can conform to the UIImagePickerControllerDelegate. It’s also a UINavigationControllerDelegate for the same reason.
2. It’s inited with a UIViewController that’s later used to present the picker.
3. takePicture is the actual transform. It saves the completion handler for later use and presents the picker.
4. If the picker is canceled by the user the completion is called with an error and the picker is dismissed.
5. If the user picks an image, the completion handler is called with the image object.

Now it’s pretty easy to take a picture. If you have a signal that executes because of a button, you can attach the picker like this:

let buttonSignal = Signal<Event>()
buttonSignal
.map {event in
  .Camera
}
.flatMap(ImageCapture(viewController: self).takePicture)
.next { image in
  // here's the picture!
}
.error { error in
  // the user is not really into taking pictures.
}

There we are: A selection from the user can be seen as a transform (just like picking stuff from a list, selecting a picture, select a receiver of a message or edit details of an object). The moment your signal is triggered, the picker will be presented on screen and automatically dismissed once the user has decided. Your application logic doesn’t need to know about the details: It just cares about the result of picking an image (either an image or an error).

Laying out the flow of your application in transforms can greatly simplify view controller presentation logic because all dependencies are explicit and there’s no need to use segues (as always: Functional programming is not a silver bullet. Segues can be extremly handy).