Why does the type of the result of the `left()` fu...

# getting-started

[Lukasz Kalnik]

Why does the type of the result of the left() function here must have R?, although R is already by definition nullable?

class Either<L, R> private constructor(
    private val left: L,
    private val right: R,
) {

    companion object {

        fun <L, R> left(left: L): Either<L, R?> {
            return Either(left, null)
        }
    }
}

[Joffrey]

What makes you say that R is by definition nullable? R is only nullable if the caller of this function specifies it as such.

This signature forces Either.left<Int, String>(42) to return an Either<Int, String?>. Were it not for the so called "definitely nullable" R?, it would be expected to return Either<Int, String> which is not possible given the body of the function (so defining the function without ? would not compile)

[Lukasz Kalnik]

You're right, I was thinking, because R : Any? that it's guaranteed to be nullable.
But indeed it depends on the concrete instantiation of the class.
It can be nullable, but doesn't have to.

[Lukasz Kalnik]

Thanks!

[Joffrey]

You're welcome 🙂

[Marc]

Why are you defining such an algebra for Either? you could just do something like:

sealed interface Either<out L, out R> {

    data class Left<L>(val value: L) : Either<L, Nothing>
    data class Right<R>(val value: R) : Either<Nothing, R>

    companion object {

        fun <L, R> left(left: L): Either<L, R> {
            return Left(left)
        }

        fun <L, R> right(right: R): Either<L, R> {
            return Right(right)
        }
    }
}

And then type each of them in isolation so you don’t need to pass null for non used branches

[Marc]

if you make it Either<L, R?> you could not make it work with other non nullable R perhaps

[Lukasz Kalnik]

It's not me, it's legacy code. I agree that your approach is better.