1

I've seen a lot of times code like this:

...
val aString: String = someFunctionDataReturnsAString()
...
if (someCondition) Some(aString)
else 
...

Is it right to create a new Option using Some(aString)? What if for some reason turns out the value of aString is null? Wouldn't be good replacing the if sentence with:

if (someCondition) Option(aString)

Because:

val a = Some("hello")
val b: String = null
val c = Some(b)
val d = Option(b)
println(a)
println(c)
println(d)

Will print in the console:

Some(hello)
Some(null)
None

Looks like a better idea to use Option when passing a string as parameter, so in case the string is null the value of the Option will be None.

I've seen code to be reviewed and I think with cases like this a comment should be added asking for replacing Some(aString) with Option(aString).

M.G.
  • 369
  • 1
  • 14

2 Answers2

5

You are right. Assuming your string can be null, it's recommended to use Option(aString):

def apply[A](x: A): Option[A]

An Option factory which creates Some(x) if the argument is not null, and None if it is null.

x - the value

returns - Some(value) if value != null, None if value == null

Class Option mentions that both have their use case, so the programmer has the last say which one is better suited for his particular need:

Interacting with code that can occasionally return null can be safely wrapped in scala.Option to become None and scala.Some otherwise.

On a personal note, since strings are immutable and you can't really have uninitialized strings (because Scala always requires you to initialize them - unless you use var variables, which is not the recommended way to program in Scala anyway), you'll rarely actually encounter null strings in your code.

That being said, one can still decide to go for Option(aString) every time, and it's perfectly fine.

Alin Gabriel Arhip
  • 2,568
  • 1
  • 14
  • 24
3

Normally it's ok to replace Some.apply with Option.apply (and safer).

But, although normally this is not a big deal (and let's avoid premature optimizations), there can be minor runtime overhead for Option.apply, since we add another branching instead of just wrapping.

Generally, you could split your project into pure FP part (it's ok to assume that your strings are not null there and use Some.apply, also Some.apply is canonical monadic pure for Option) and part where you interact with the outer world, with Java libraries etc. (it's more safely to use Option.apply there).

But we should keep in mind that Some.apply returns Some[A] while Option.apply returns its supertype Option[A]. Normally it's ok but sometimes in type-level calculations this can be too rough if you need to know precise type at compile time. See for example Shapeless and annotations , how to use shapeless to detect field type annotation

Dmytro Mitin
  • 48,194
  • 3
  • 28
  • 66