What’s the difference between init?() and init()?

Swift version: 5.10

Paul Hudson @twostraws May 28th 2019

It’s the job of a regular Swift initializer to create a fully fledged instance of a new type, however sometimes the data that has been provided is insufficient or incorrect, and creation can’t proceed.

For example, consider this code:

struct Person {
    var ssn: String

    init(socialSecurityNumber: String) {
        self.ssn = socialSecurityNumber
    }
}

let person = Person(socialSecurityNumber: "111-11-1111")
print(person)

That defines a Person struct that can be created using a nine-digit social security number, then creates an instance of that struct.

But what should happen here?

let person = Person(socialSecurityNumber: "FISH")

In that instance we’re passing an invalid social security number, so really we expect creating a Person to fail.

This is where failable initializers come in: they are written as init?(), and can return nil rather than a value if something goes wrong during creation. For example, we could write a quick check to make sure the social security number is more or less correct like this:

struct Person {
    var ssn: String

    init?(socialSecurityNumber: String) {
        if socialSecurityNumber.count < 11 {
            return nil
        } else {
            self.ssn = socialSecurityNumber
        }
    }
}

Notice the initializer is now called init?() to reflect that it returns an optional – the process might return nil if the creation fails. The logic is pretty simple: if there are 11 digits we assume it’s correct, otherwise we return nil. Note: if you really wanted to validate that number you’d need to use a regular expression.