Reputation: 45
I have a question about implicits resolution. Say, I have the following type class:
trait Foo[In <: Base, Out <: Base] {
def factor : Double
}
whereas
sealed trait Base
object Base {
implicit def symmetricFoo[In <: Base, Out <: Base](implicit foo : Foo[In, Out]) : Foo[Out, In] =
new Foo[Out, In] {
def factor : Double = 1.0 / foo.factor
}
implicit def transitiveFoo[In <: Base, Mid <: Base, Out <: Base](implicit foo1 : Foo[In, Mid], foo2 : Foo[Mid, Out]) : Foo[In, Out] =
new Foo[In, Out] {
def factor : Double = foo1.factor * foo2.factor
}
}
case object A extends Base
case object B extends Base {
implicit def bFoo : Foo[B.type, A.type] =
new Foo[B.type, A.type] {
def factor : Double = 2.0
}
}
case object C extends Base {
implicit def cFoo : Foo[C.type, A.type] =
new Foo[C.type, A.type] {
def factor : Double = 3.0
}
}
case object D extends Base {
implicit def dFoo : Foo[D.type, C.type] =
new Foo[D.type, C.type] {
def factor : Double = 5.0
}
}
I would like to be able to get instances of Foo[X, Y] if I have intermediary Foo's along the path from X to Y. Sometimes it does work, e. g.
println(implicitly[Foo[D.type, A.type]].factor) // 15.0 (D->C, C->A = 5 * 3)
println(implicitly[Foo[D.type, B.type]].factor) // 7.5 (D->C, C->A, A->B = 5 * 3 * 1/2)
but if I change the order of lines, it does not:
println(implicitly[Foo[D.type, B.type]].factor) // 7.5
println(implicitly[Foo[D.type, A.type]].factor) // does not compile
There are a lot of more or less identical error messages like
test.this.Base.transitiveFoo is not a valid implicit value for test.Foo[test.C.type,test.A.type] because: hasMatchingSymbol reported error: diverging implicit expansion for type test.Foo[test.C.type,Mid] starting with method transitiveFoo in object Base
diverging implicit expansion for type test.Foo[test.A.type,Mid] starting with method transitiveFoo in object Base
not enough arguments for method implicitly: (implicit e: test.Foo[test.D.type,test.A.type])test.Foo[test.D.type,test.A.type]. Unspecified value parameter e.
The full log is a bit longer.
What is the correct way, given Foo[X, Y], to get Foo[Y, X], and given Foo[X, Y] and Foo[Y, Z], to get Foo[X, Z] for any combinations of defined Foo's? Would something like shapeless' Lazy help here? Thanks in advance!
Upvotes: 2
Views: 651
Reputation: 51683
This variant seems to work:
import shapeless.Lazy
trait Foo[In <: Base, Out <: Base] {
def factor : Double
}
sealed trait Base
case object A extends Base
case object B extends Base
case object C extends Base
case object D extends Base
trait LowPriorityImplicits {
implicit def symmetricFoo[In <: Base, Out <: Base](implicit foo : Lazy[Foo[In, Out]]) : Foo[Out, In] =
new Foo[Out, In] {
def factor : Double = 1.0 / foo.value.factor
}
}
object Base extends LowPriorityImplicits {
implicit def bFoo : Foo[B.type, A.type] =
new Foo[B.type, A.type] {
def factor : Double = 2.0
}
implicit def cFoo : Foo[C.type, A.type] =
new Foo[C.type, A.type] {
def factor : Double = 3.0
}
implicit def dFoo : Foo[D.type, C.type] =
new Foo[D.type, C.type] {
def factor : Double = 5.0
}
implicit def transitiveFoo[In <: Base, Out <: Base, Mid <: Base](implicit foo1 : Foo[In, Mid], foo2 : Foo[Mid, Out]) : Foo[In, Out] =
new Foo[In, Out] {
def factor : Double = foo1.factor * foo2.factor
}
}
def main(args: Array[String]): Unit = {
println(implicitly[Foo[D.type, B.type]].factor)
println(implicitly[Foo[D.type, A.type]].factor)
}
7.5
15.0
def main(args: Array[String]): Unit = {
println(implicitly[Foo[D.type, A.type]].factor)
println(implicitly[Foo[D.type, B.type]].factor)
}
15.0
7.5
Just in case my build.sbt:
scalaOrganization := "org.typelevel"
scalaVersion := "2.12.3-bin-typelevel-4"
libraryDependencies += "com.chuusai" %% "shapeless" % "2.3.2"
Upvotes: 2