What’s New in Swift 2

At WWDC we found out the Swift team hasn’t been messing around sending crude watch drawings to each other like the rest of us — they’ve been hard at work on Swift 2.

We’ll have plenty of written and video Swift 2 tutorials for you soon, but in the meantime I wanted to highlight the most exciting changes so you can be prepared for the great migration to Swift 2 in the fall.

Error Handling

As Ray mentioned in his WWDC 2015 Initial Impressions post, error handling has been revamped in Swift 2. Instead of NSError objects and double pointers, we’re moving to a new system that looks similar to exception handling.

You may be familiar with code like this:

if drinkWithError(nil) {
  print("Could not drink beer! :[")
  return
}

Generally in Cocoa, you pass in a reference to an NSError object (an inout parameter in Swift) and then the method will assign the variable if there was a problem. However, the problem is that you can pass nil here to completely ignore the error; or, you can pass in the NSError but then never check it.

Swift 2 adds additional safety to your error checking. You use the throws keyword to specify which functions and methods could throw an error. Then you have the do, try, and catch keywords for when you call something that could throw:

// 1
enum DrinkError: ErrorType {
  case NoBeerRemainingError
}

// 2
func drinkWithError() throws {
  if beer.isAvailable() {
    // party!
  } else {
    // 3
    throw DrinkError.NoBeerRemainingError
  }
}

func tryToDrink() {
  // 4
  do {
    try drinkWithError()
  } catch {
    print("Could not drink beer! :[")
    return
  }
}

There are a few things to highlight here:

1. To create an error to throw, simply create an enum that derives from ErrorType.
2. You need to use the throws keyword to mark any function that can throw an error.
3. This throws an error, which will be caught in section 4.
4. Instead of try blocks, which might be familiar from other languages, you wrap any code that can throw an error in a do block. Then, you add the try keyword to each function call that could throw an error.

The new syntax is pretty lightweight and readable. Any API that currently uses NSError now uses this system, so we’ll be seeing a lot of it!

Binding

With Swift 1.2, we lost the "pyramid of doom" and gained the ability to test binding multiple optionals in one line:

if let pants = pants, frog = frog {
  // good stuff here!
}

That works just fine, but the issue for some people is that the "preferred" code path where all the optionals have some value is indented. That means you need to keep looking inside indented code blocks for the mainline part of your code, while the error conditions are outside.

If only there were some way to check that some of the optionals don’t have a value, and then exit early! That’s exactly what Swift 2 offers with its guard statement:

guard let pants = pants, frog = frog else {
  // sorry, no frog pants here :[
  return
}

// at this point, frog and pants are both unwrapped and bound!

Using guard means you can perform the optional binding (or any other operation, really) and provide a code block in the else to run if the condition fails. Then, you can continue on – in this case, the optionals frog and pants are now bound and are no longer optionals in scope.

This should make for much clearer code since guard lets you specify the state you’re actually expecting rather than checking for the error case.

Protocol Extensions

Object-oriented? Functional? There’s one more to add to the front of the line of what Swift is: a protocol-oriented programming language!

In Swift 1, protocols were like interfaces to specify a set of properties and methods that a class, struct, or enum would then conform to.

Now in Swift 2, you can extend protocols and add default implementations for properties and methods. You can already do this with classes and structs — adding new methods to String or Array, for example — but adding these to protocols now gives you a much wider reach.

extension CustomStringConvertible {
  var shoutyDescription: String {
    return "\(self.description.uppercaseString)!!!"
  }
}

let greetings = ["Hello", "Hi", "Yo yo yo"]

// prints ["Hello", "Hi", "Yo yo yo"]
print("\(greetings.description)")

// prints [HELLO, HI, YO YO YO]!!!
print("\(greetings.shoutyDescription)")

Note that the Printable protocol is now called CustomStringConvertible, which most Foundation objects conform to. With protocol extensions, you can extend wide swaths of the system with your own custom functionality. And rather than adding bits of custom code to many classes and structs and enums, you can write one general implementation and have it apply across a set of types.

The Swift team has been busy doing this already — if you’ve ever used map or filter in Swift, you may have thought they would do better as methods rather than as global functions. Thanks to the power of protocol extensions, there are a new set of methods on collection types such as map, filter, indexOf, and more!

let numbers = [1, 5, 6, 10, 16, 42, 45]

// Swift 1
find(filter(map(numbers, { $0 * 2}), { $0 % 3 == 0 }), 90)

// Swift 2
numbers.map { $0 * 2 }.filter { $0 % 3 == 0 }.indexOf(90) // returns 2

Thanks to protocol conformance, your Swift 2 code can be more concise and readable. With the Swift 1 version, you need to look at the calls inside out to understand what’s going on; in the Swift 2 version, the chain of functions is clear.

You’re also leveraging the power of protocol-oriented programming — check out the WWDC session on this topic and keep an eye out for future tutorials and articles here on the site!