What is protocol-oriented programming?

Swift version: 5.6

Paul Hudson @twostraws May 28th 2019

One powerful feature of Swift is its ability to extend protocols – to be able to add new functionality not only to one type, but to a whole collection of types that all conform to the same protocol.

Protocol-oriented programming takes that feature and encourages you to craft your app’s architecture around it so that the first thing you do is sketch out one or more protocols rather than get straight into concrete types.

If you’re coming from a more traditional object-oriented system where inheritance is more common, try to think as your first protocol as being a base class. You can then create new protocols by inheriting from that initial protocol, and write extensions so they have default implementations.

However, a better approach is to write lots of small protocols that each do specific, individual things: one to make products purchasable, one to make them serializable, one to make them searchable, and so on. You can then add default implementations to those extensions, which means you can add functionality to existing types just by making them conform to your protocol.

In this respect protocol-oriented programming is sort of a similar approach to multiple inheritance from languages such as C++. However, because protocol extensions can’t include state you don’t get any of the cruft, and Swift’s constraint-based conflict resolution is easy enough for everyone to understand.