Reputation: 848
As per the definition of contravariance (super class instances will be accepted), my last statement in the below code snippet, should be accepted; but it is thrown with type error. Can you please correct my understanding.
class A
class B extends A
class C extends B
abstract class Box[-T] {
def set(x : T) :Unit
}
val x = new Box[B] {
def set(b:B) = Console println "B"
}
val y = new Box[A] {
def set(a:A) = Console println "A"
}
val z = new Box[C] {
def set(c:C) = Console println "C"
}
x.set(new A) <-- Type error
But x.set(new C)
is fine! So even though "contravariant parameters are accepted as method parameters" is in fact, are covariant parameters actually.
Upvotes: 0
Views: 76
Reputation: 4991
It looks like this not how it would work and that's from a conceptual stand point: a B
is an A
but not the other way around.
As indicated in the scala doc for variances, contravariant type parameters allow to pass a "super-type" instead of the "sub-type", but this applies to what is used within the body (e.g. If the "super-type" already has defined what's used) otherwise I guess it just considers the type and we're back to what I explained before.
Upvotes: 2
Reputation: 22895
You are confusing how contravariance works. x
is a Box[B] as such, set
accepts values of type B
(or any subtype of B
because that is what Liskvo says).
However, Box[A] is a subtype of a Box[B]. So you could have passed y
where x
is expected. Because a Box that can store any A
can store a B
.
But, a Box that can store only B
s can not store and arbitrary A
.
Upvotes: 2