Drawing diagrams in Swift using a recursive enum data structure
Branch: master
Clone or download
Latest commit ef2f090 Sep 21, 2016
Type Name Latest commit message Commit time
Failed to load latest commit information.
Crustacean.playground Updated `Crustacean` playground Sep 20, 2016
CrustaceanEnumOriented.playground Updated `CrustaceanEnumOriented` playground Sep 20, 2016
Diagrams.playground Updated `Diagrams` playground Sep 20, 2016
.gitignore Add .gitignore Jun 11, 2015
README.md Updated README Sep 20, 2016



Swift 3.0

This is an adaption of Apple’s sample code for the Protocol-Oriented Programming in Swift talk given during WWDC 2015.

Included is Apple’s original example playground file Crustacean.playground that uses a Protocol-oriented design (updated for Swift 3). In addition there's an alternative version CrustaceanEnumOriented.playground that uses a recursive enum as the data structure.

Finally there's the Diagrams.playground which adds a bit more functionality and includes several pages of example diagrams.

The playgrounds demonstrate two different approaches to creating Diagrams as value types and show how to draw them into a CGContext.


Apple’s version uses a variety of structs that conform to the Drawable protocol to represent different shapes. The alternative approach uses a recursive enum to achieve the same result. It looks like this:

public enum Diagram {
  case Polygon([CGPoint])
  case Line([CGPoint])
  case Arc(radius: CGFloat, startAngle: CGFloat, endAngle: CGFloat)
  case Circle(radius: CGFloat)
  indirect case Scale(x: CGFloat, y: CGFloat, diagram: Diagram)
  indirect case Translate(x: CGFloat, y: CGFloat, diagram: Diagram)
  indirect case Rotate(angle: CGFloat, diagram: Diagram)
  case Diagrams([Diagram])

Note: livePreview will merrily consume processing power to continuously redraw still images, therefore it's recommended to manually stop execution of the Playground after the images have rendered.

Protocol-Oriented or Enum-Oriented – which is better?

The two approaches are a good demonstration of the expression problem. Which approach is easier to extend? Using a protocol-oriented technique allows you to add new types without too much hassle. In Apple’s example code a Bubble struct is added by implementing the Drawable protocol and Equatable (no pre-existing code needs to be adjusted). If a Bubble case were added to the enum version it would necessitate the altering of pre-existing functions (Equatable for Diagram and the drawDiagram function) this is more hassle and more error prone. However, we don't need to add a new case to the enum to draw Bubbles, we can simply add a function that constructs a bubble and returns a Diagram, in that case no code needs to be altered.

The use of a Renderer protocol makes it much easier to add a TestRenderer to log drawing. But using the Renderer protocol to add diagram transformation functionality is potentially very cumbersome. It is easy to add a ScaledRenderer type, but it would be more complicated to add a TranslateRenderer, or a RotateRenderer and duplicates functionality that is already provided by CGContext. The enum approach doesn't attempt to provide the logic for Diagram transformation, it simply stores the information needed and uses CGContext functions to do the hard work.

Which approach is better? I dunno ¯\_(ツ)_/¯