As we’ve seen previously, Swift will happily auto-wrap your non-optional value inside an optional temporarily if it helps make the types involved compatible. We’ve seen the occasional instance where this can be surprising. But for the most part, this implicit wrapping of non-optionals is your friend.¹ Here’s a few more examples.

Equating Optionals

The Swift standard library defines an == operator for optionals:

func ==<T : Equatable>(lhs: T?, rhs: T?) -> Bool

So long as your optional wraps a type that’s Equatable, that means you can compare two optionals of the same underlying type:

let i: Int? = 1
let j: Int? = 1
let k: Int? = 2
let m: Int? = nil
let n: Int? = nil

i == j  // true,  1 == 1
i == k  // false, 1 != 2
i == m  // false, non-nil != nil
m == n  // true,  nil == nil

let a: [Int]? = [1,2,3]
let b: [Int]? = [1,2,3]

// error, [Int] is not equatable
// (it does have an == operator, but that's not the same thing)
a == b

let f: Double? = 1.0

// ignore the weird compiler error,
// the problem is Int and Double aren't
// the same type...
i == f

Well, this is useful, but even more useful is this one definition of == allows you to compare optionals against non-optionals too. Suppose you want to know if an optional string contains a specific value. You don’t care if it’s nil or not – just whether it contains that value.

You can just pass your non-optional into that same == that takes optionals:

let x = "Elsa"

let s1: String? = "Elsa"
let s2: String? = "Anna"
let s3: String? = nil

s1 == x  // true
s2 == x  // false
s3 == x  // false

This works because Swift is implicitly wrapping x in an optional each time. If Swift didn’t do this, and you implemented an optional-to-non-optional operator yourself, you’d have to define == an extra two times, to make sure it was symmetric (which is a requirement whenever you implement ==):

// definitions if there weren't implicit optionals
func ==<T:Equatable>(lhs:T, rhs:T?)->Bool { return rhs != nil && lhs == rhs! }
func ==<T:Equatable>(lhs:T?, rhs:T)->Bool { return lhs != nil && lhs! == rhs }

Beware though, optional == isn’t all sunshine and unicorn giggles. Imagine the following code:

let a: Any = "Hello"
let b: Any = "Goodbye"

// false, of course
(a as? String) == (b as? String)

// oops, this is true...
(a as? Int) == (b as? Int)

Depending on how you think, this is either perfectly reasonable or immensley confusing, but either way you should bear it in mind. And maybe it’s another reason to go on your long list “Why not to use Any and as unless you really really have to.”

Comparing Optionals

Unsurprisingly, there’s a similar definition of < that will compare two optionals so long as what they wrap is Comparable:

func <<T : _Comparable>(lhs: T?, rhs: T?) -> Bool

If lhs and rhs are both some value, it compares them. Two nils are considered equal just like before. But a nil is always considered less than any non-nil value, (Comparable must always enforce a strict total order so they have to go somewhere well-defined in the ordering).

So if you sort a set of optionals, the nil values all end up at one end:

let a: [Int?] = [1, 3, nil, 2]
sorted(a, <)
// results in [nil, 1, 2, 3]

This is fairly intuitive, but it can lead to other subtle bugs. Suppose you want to filter a list like so:²

let ages = [
    "Tim":  53,  "Angela": 54,  "Craig":   44,
    "Jony": 47,  "Chris":  37,  "Michael": 34,
]

let people = ["Tim", "Johny", "Chris", "Michael"]

// don't trust anyone over 40
let trust = people.filter { ages[$0] < 40 }

// trust will contain ["Johny", "Chris", "Michael"]

The string "Johny" was not in the ages dictionary so you get a nil – but the way < works, nil is less than 40, so included in the filtered array.

Optional Subscripts

While we're talking about dictionaries, did you know that Dictionary.subscript(key) doesn't just return an optional, it also takes an optional?

var d = ["1":1]

let four = "4"
let nonsense = "nonsense"

// this will add ["4":4] to d
// remember toInt returns an optional
// (the string might not have been a number)
d[four] = four.toInt()

// nothing will be added, because
// the rhs is nil
d[nonsense] = nonsense.toInt()

// this will remove the "1" entry from d
d["1"] = nil

That's quite useful – but also, there wasn't much choice in it working like this.³ Dictionary.subscript(key) { set } has to take an optional, because when you define a setter, you must define a getter. And Dictionary's getter has to be optional (in case the key isn't present). So the setter's input value has to be.

This means every time you assign a value to a dictionary using a key subscript, your non-optional value is getting auto-wrapped in an optional. If this didn’t happen automatically, it would make assigning values to your dictionary a real pain, as you’d have to type d[key] = .Some(value) every time.

To see this in practice, here's a very bare-bones (very inefficient) implementation of a dictionary. Take a look at the subscript to see how the optional is handled:

public struct RubbishDictionary<Key: Equatable, Value> {

    public typealias Element = (Key, Value)

    // just store the key/value pairs in an unsorted
    // array (see, told you it was rubbish)
    private typealias Storage = [Element]
    private var _store: Storage = []

    // there's a reason why this is a private method,
    // which I'll get to later...
    private func _indexForKey(key: Key) -> Storage.Index? {
        // new year's resolution: see if I can avoid
        // calling array[i] completely in 2015
        for (idx, (k, _)) in Zip2(indices(_store),_store) {
            if key == k { return idx }
        }
        return nil
    }

    // Subscript for key-based lookup/storage
    public subscript(key: Key) -> Value? {
        // "get" is pretty simple, either key can be
        // found, or not...
        get {
            if let idx = _indexForKey(key) {
                // implicit optional wrap
                return _store[idx].1
            }
            else {
                return nil
            }
        }
        // remember, newValue is a Value?
        set(newValue) {
            switch (_indexForKey(key), newValue) {

            // existing index, so replace value
            case let (.Some(idx), .Some(value)):
                _store[idx].1 = value

            // existing index, nil means remove value
            case let (.Some(idx), .None):
                _store.removeAtIndex(idx)

            // new index, add non-nil value
            case let (.None, .Some(value)):
                _store.append((key, value))

            // new index, nil value is no-op
            case (.None,.None):
                break
            }
        }
    }
}

Finally, just a reminder that an optional containing an optional containing nil is not the same as nil:

var d: [Int:Int?] = [1:1, 2:nil]
// d now contains two entries, 1:1 and 2:nil
d[2] = nil  // removes nil
d[3] = .Some(nil)  // adds 3:nil

Full imlementation of RubbishDictionary

So, in case you've made it all the way down here, and you're interested, I thought I'd end with a fuller but minimal-as-possible implementation of RubbishDictionary. It has almost-feature-parity with Dictionary (it's just missing getMirror and init(minimumCapatity)). It ought to do everything Dictionary can, just slower.

If reading raw code's not your thing, skip it. But it shows how to simply implement several of the standard library's protocols such as CollectionType and DictionaryLiteralConvertible, as well as a few other features like lazily-evaluated key and value collections. I like how Swift extensions allow you to build up the type in logical groups (e.g. each protocol implementation one-by-one). Since this is a post about implicit optional wrapping, I've flagged where it’s also helping keep the code consise.

I have a nagging feeling there are some neater ways to handle some of the optionals – if you spot one, let me know. Oh and bugs. I'm sure there are a couple of them.

// building on the initial definition given above...

// An index type for RubbishDictionary.  In theory, you could perhaps just
// expose an alias of the storage type’s index, however, if the Key type were
// an Int, this would mean subscript(key) and subscript(index) would be
// ambiguous. So instead it's simple wrapper struct that does little more than
// pass-through to the real index.
public struct RubbishDictionaryIndex<Key: Equatable, Value>: BidirectionalIndexType {
    private let _idx: RubbishDictionary<Key,Value>.Storage.Index

    public typealias Index = RubbishDictionaryIndex<Key,Value>

    public func predecessor() -> Index { return Index(_idx: _idx.predecessor()) }
    public func successor() -> Index { return Index(_idx: _idx.successor()) }
}

// Required to make the index conform to Equatable
// (indirectly required by BidirectionalIndexType)
public func ==<Key: Equatable, Value>
  (lhs: RubbishDictionaryIndex<Key,Value>, rhs: RubbishDictionaryIndex<Key,Value>)
   -> Bool { 
    return lhs._idx == rhs._idx 
}

// Conformance to CollectionType
extension RubbishDictionary: CollectionType {
    public typealias Index = RubbishDictionaryIndex<Key,Value>

    // Add a subscript that takes an Index.  This one
    // does not need to be nil, since an indexed entry
    // _must_ exist.  Read-only.
    public subscript(idx: Index) -> Element { return _store[idx._idx] }

    public var startIndex: Index { return Index(_idx: _store.startIndex) }
    public var endIndex: Index { return Index(_idx: _store.endIndex) }

    public typealias Generator = IndexingGenerator<RubbishDictionary>
    public func generate() -> Generator { return Generator(self) }
}

// Conformance to Printable and DebugPrintable
extension RubbishDictionary: Printable, DebugPrintable {
    public var description: String {
        let elements = map(self) { "($0.0):($0.1)" }
        return "[" + ",".join(elements) + "]"
    }

    public var debugDescription: String {
        let elements = map(self) { "(toDebugString($0.0)):(toDebugString($0.1))" }
        return "[" + ",".join(elements) + "]"
    }
}

// Conformance to DictionaryLiteralConvertible
extension RubbishDictionary: DictionaryLiteralConvertible {
    public init(dictionaryLiteral elements: (Key, Value)...) {
        for (k,v) in elements {
            self[k] = v
        }
    }
}

// Remaining features supported by Dictionary
extension RubbishDictionary {

    public var count: Int { return _store.count }
    public var isEmpty: Bool { return _store.isEmpty }

    // now for the public version of indexForKey
    public func indexForKey(key: Key) -> Index? {
        // this maps non-nil results from _indexForKey to
        // the wrapper type, or just returns nil if 
        // _indexForKey did...
        return _indexForKey(key).map { Index(_idx: $0) }
    }

    // a lazily-evaluated collection of keys in the dictionary
    public var keys: LazyBidirectionalCollection<MapCollectionView<RubbishDictionary<Key,Value>,Key>> {
        return lazy(self).map { (k,v) in k }
    }

    // and the same for the values
    public var values: LazyBidirectionalCollection<MapCollectionView<RubbishDictionary<Key,Value>,Value>> {
        return lazy(self).map { (k,v) in v }
    }

    // unlike its subscript equivalent, updateValue:forKey does not take an optional value...
    public mutating func updateValue(value: Value, forKey key: Key) -> Value? {
        if let idx = _indexForKey(key) {
            let old = _store[idx].1
            _store[idx].1 = value
            // implicit optional wrap
            return old
        }
        else {
            _store.append((key,value))
            return nil
        }
    }

    // returns the value removed, if there was one
    public mutating func removeValueForKey(key: Key) -> Value? {
        if let idx = _indexForKey(key) {
            // implicit optional wrap
            return _store.removeAtIndex(idx).1
        }
        else {
            return nil
        }
    }

    public mutating func removeAtIndex(index: Index) {
        _store.removeAtIndex(index._idx)
    }

    public mutating func removeAll(keepCapacity: Bool = false) { 
        _store.removeAll(keepCapacity: keepCapacity)
    }

}