Singleton types should not be inferred for final vars

[smarter]

class Test {
  final var x = false
  x = true
}

error: type mismatch:
 found   : Boolean(true)
 required: Boolean(false)
  x = true
      ^

[sjrd]

So there is such a thing as a final var ... 😧

[DarkDimius]

What does final var mean in the first place? Aren't final and var contradicting?

[smarter]

Good question, I just found it in the wild in https://github.com/molecule-labs/molecule/search?utf8=%E2%9C%93&q=%22final+var%22 but didn't stop to think if it made any sense :). http://stackoverflow.com/questions/18222233/what-is-the-usage-of-a-final-var-in-scala says it's to prevent overriding.

[DarkDimius]

says it's to prevent overriding.

Mutable variables can't be overridden even if they are not final:

scala> class A{var s = 3}; class B extends A{override var s = 4}
<console>:10: error: overriding variable s in class A of type Int;
 variable s cannot override a mutable variable
       class A{var s = 3}; class B extends A{ override var s = 4}

[smarter]

@adriaanm: any idea if the semantics of final var are different from the semantics of var ?

[adriaanm]

As @DarkDimius said, currently, final is implied for a var. I can't think of any difference implied by the presence of the modifier. I guess you could defend a design where you can override a var so you can introduce validation logic in the setter, but you couldn't change the type and the storage would still have to be in the super class, so, not sure...

[adriaanm]

The logic around widening inferred types is here: https://github.com/scala/scala/blob/2.12.x/src/compiler/scala/tools/nsc/typechecker/Namers.scala#L839