Suppose an outer class with an inner composite of two inner classes:
class Outer(val times: Int) {
class Inner(val count: Int)
object Inner {
def apply(i: Int) = new Inner(i)
}
class InnerWrap(val i: Inner)
object InnerWrap {
def apply(i: Inner) = new InnerWrap(i)
}
def method(i: Inner) = i.count * times
}
object Outer {
def apply(times: Int) = new Outer(times)
}
class PathDependentExpl {
val o = new Outer(3)
val i = o.Inner(2)
val p = new Outer(5)
val j = p.Inner(3)
println(o.method(i))
println(p.method(j))
}
All nice and dandy. We even have compiler protection against feeding inners from one Outer
to methods of another Outer
through path-dependent types.
But the syntax is a little annoying. I want to be able to write something like
implicit val p = new Outer(5)
val k = Inner(3) // same as p.Inner(3)
val l = InnerWrap(Inner(3))
...omitting the outer instance for the block where the implicit is valid. So I rush adding delegate methods with the Outer object as an implicit parameter.
object Outer {
def apply(times: Int) = new Outer(times)
def Inner(c: Int)(implicit o: Outer) = o.Inner(c)
def InnerWrap(i: o.Inner)(implicit o: Outer) = o.InnerWrap(i)
}
That last InnerWrap
definition yields: "illegal dependent method type: parameter appears in the type of another parameter in the same section or an earlier one", which actually makes sense. I tried def InnerWrap[O <: Outer#Inner](i: O)(implicit o: Outer) = o.InnerWrap(i)
and other variants to no avail.
Question is, how can I still get to use the cleaner syntax? How can the declaration of InnerWrap
in object Outer
can be made to accept o.Inner
s or equivalent?
As I noted in a comment above, it's possible to import the objects (or methods) you need from p
, which may give you the clean syntax you're looking for:
scala> val p = new Outer(5)
p: Outer = Outer@4fe2fe5d
scala> import p._
import p._
scala> val k = Inner(3)
k: p.Inner = Outer$Inner@b90ffa7
scala> val l = InnerWrap(Inner(3))
l: p.InnerWrap = Outer$InnerWrap@3bd6bff7