Question

Consider this simple Scala class:

class A(val d: Int)

Is there a difference in Scala (either in behaviour or generated bytecode) between

class B(d: Int) extends A(d)

and

class B(override val d: Int) extends A(d)

or are both equivalent? If they are different, what would be the specific usecase for each of them?

Would it be different if A was defined as class A(var d: Int)?

Answer 1

For vals, there is no semantic difference. However, there may be a difference in the generated bytecode. In particular, if a method defined in the derived class refers to d, it refers to the constructor parameter d rather than to the val of the same name. This is implemented via an additional private field generated for the derived class.

For vars, there is a difference in behavior. Without an override, any methods that refer to d from within the derived class will be referring to the constructor parameter, while callers referencing d from outside the class will get the field. In this case, the two values may differ (if the value has changed since construction).

Here’s a session that demonstrates the behavior with a var:

scala> class A(var d: Int)
defined class A

scala> class B(d: Int) extends A(d) { override def toString = "d: " + d }
defined class B

scala> val b = new B(1)
b: B = d: 1

scala> b.d = 2

scala> b.d
res1: Int = 2

scala> b
res2: B = d: 1

This question is related: Idiomatic Scala way to deal with base vs derived class field names?.