Swift – Why convenience keyword is even needed in Swift

initializationswift

Since Swift supports method and initializer overloading, you can put multiple init alongside each other and use whichever you deem convenient:

class Person {
    var name:String

    init(name: String) {
        self.name = name
    }

    init() {
        self.name = "John"
    }
}

So why would convenience keyword even exist? What makes the following substantially better?

class Person {
    var name:String

    init(name: String) {
        self.name = name
    }

    convenience init() {
        self.init(name: "John")
    }
}

Best Solution

The existing answers only tell half of the convenience story. The other half of the story, the half that none of the existing answers cover, answers the question Desmond has posted in the comments:

Why would Swift force me to put convenience in front of my initializer just because I need to call self.init from it?`

I touched on it slightly in this answer, in which I cover several of Swift's initializer rules in details, but the main focus there was on the required word. But that answer was still addressing something that's relevant to this question and this answer. We have to understand how Swift initializer inheritance works.

Because Swift does not allow for uninitialized variables, you are not guaranteed to inherit all (or any) initializers from the class you inherit from. If we subclass and add any uninitialized instance variables to our subclass, we have stopped inheriting initializers. And until we add initializers of our own, the compiler will yell at us.

To be clear, an uninitialized instance variable is any instance variable which isn't given a default value (keeping in mind that optionals and implicitly unwrapped optionals automatically assume a default value of nil).

So in this case:

class Foo {
    var a: Int
}

a is an uninitialized instance variable. This will not compile unless we give a a default value:

class Foo {
    var a: Int = 0
}

or initialize a in an initializer method:

class Foo {
    var a: Int

    init(a: Int) {
        self.a = a
    }
}

Now, let's see what happens if we subclass Foo, shall we?

class Bar: Foo {
    var b: Int

    init(a: Int, b: Int) {
        self.b = b
        super.init(a: a)
    }
}

Right? We added a variable, and we added an initializer to set a value to b so it'll compile. Depending on what language you're coming from, you might expect that Bar has inherited Foo's initializer, init(a: Int). But it doesn't. And how could it? How does Foo's init(a: Int) know how to assign a value to the b variable that Bar added? It doesn't. So we can't initialize a Bar instance with an initializer that can't initialize all of our values.

What does any of this have to do with convenience?

Well, let's look at the rules on initializer inheritance:

Rule 1

If your subclass doesn’t define any designated initializers, it automatically inherits all of its superclass designated initializers.

Rule 2

If your subclass provides an implementation of all of its superclass designated initializers—either by inheriting them as per rule 1, or by providing a custom implementation as part of its definition—then it automatically inherits all of the superclass convenience initializers.

Notice Rule 2, which mentions convenience initializers.

So what the convenience keyword does do is indicate to us which initializers can be inherited by subclasses that add instance variables without default values.

Let's take this example Base class:

class Base {
    let a: Int
    let b: Int

    init(a: Int, b: Int) {
        self.a = a
        self.b = b
    }

    convenience init() {
        self.init(a: 0, b: 0)
    }

    convenience init(a: Int) {
        self.init(a: a, b: 0)
    }

    convenience init(b: Int) {
        self.init(a: 0, b: b)
    }
}

Notice we have three convenience initializers here. That means we have three initializers that can be inherited. And we have one designated initializer (a designated initializer is simply any initializer that's not a convenience initializer).

We can instantiate instances of the base class in four different ways:

enter image description here

So, let's create a subclass.

class NonInheritor: Base {
    let c: Int

    init(a: Int, b: Int, c: Int) {
        self.c = c
        super.init(a: a, b: b)
    }
}

We're inheriting from Base. We added our own instance variable and we didn't give it a default value, so we must add our own initializers. We added one, init(a: Int, b: Int, c: Int), but it doesn't match the signature of the Base class's designated initializer: init(a: Int, b: Int). That means, we're not inheriting any initializers from Base:

enter image description here

So, what would happen if we inherited from Base, but we went ahead and implemented an initializer that matched the designated initializer from Base?

class Inheritor: Base {
    let c: Int

    init(a: Int, b: Int, c: Int) {
        self.c = c
        super.init(a: a, b: b)
    }

    convenience override init(a: Int, b: Int) {
        self.init(a: a, b: b, c: 0)
    }
}

Now, in addition to the two initializers we implemented directly in this class, because we implemented an initializer matching Base class's designated initializer, we get to inherit all of Base class's convenience initializers:

enter image description here

The fact that the initializer with the matching signature is marked as convenience makes no difference here. It only means that Inheritor has just one designated initializer. So if we inherit from Inheritor, we'd just have to implement that one designated initializer, and then we'd inherit Inheritor's convenience initializer, which in turn means we've implemented all of Base's designated initializers and can inherit its convenience initializers.