问题
I have an implicit conversion - below - which feels like it should definitely be working but is definitely not. Can anyone shed any light? I know implicitly
can sometimes fail when type refinements are used - is that the issue here?
trait GetItem[A[_], T, R] {
type Out
def ret(a: A[T], ref: R): Out
}
object GetItem {
implicit def ifRefIsInt[A[_], T]: GetItem[A, T, Int] { type Out = A[T] } = new GetItem[A, T, Int] {
type Out = A[T]
def ret(a: A[T], ref: Int): Out = a
}
}
import GetItem._
//this works fine:
val t: GetItem[List, Double, Int] { type Out = List[Double] } = ifRefIsInt[List, Double]
// so does this:
implicitly[GetItem[List, Double, Int] { type Out = List[Double] }](t)
// this does not:
implicitly[GetItem[List, Double, Int] { type Out = List[Double] }]
// Could not find implicit parameter for value e: Example.Main.GetItem[List, Double, Int]{type Out = List[Double]}
Any help much appreciated, I have been staring at this for some time with little success.
回答1:
Seems to be another example of over-constrained implicits (1 2 3 4 5 6). This seems to be too much work for implicits in a single step. Compiler doesn't like complex types like (A, B)
, H :: L
, and A[T]
(in our case) in a type refinement. If we replace
implicit def ifRefIsInt[A[_], T]: GetItem[A, T, Int] { type Out = A[T] } =
new GetItem[A, T, Int] {
type Out = A[T]
def ret(a: A[T], ref: Int): Out = a
}
with
implicit def ifRefIsInt[A[_], T, O](implicit
ev: A[T] =:= O
): GetItem[A, T, Int] { type Out = O } = new GetItem[A, T, Int] {
type Out = O
def ret(a: A[T], ref: Int): Out = a
}
then
implicitly[GetItem.Aux[List, Double, Int, List[Double]]]
implicitly[GetItem[List, Double, Int] { type Out = List[Double] }]
compile: https://scastie.scala-lang.org/P5iXP2ZfQUCKEIMukYyqIg (Scala 2.13.3)
For some reason compiler swallows a warning (with -Xlog-implicits
switched on). If I trigger implicit search manually
import scala.language.experimental.macros
import scala.reflect.macros.{whitebox, contexts}
def foo[A]: Unit = macro fooImpl[A]
def fooImpl[A: c.WeakTypeTag](c: whitebox.Context): c.Tree = {
import c.universe._
val context = c.asInstanceOf[contexts.Context]
val global: context.universe.type = context.universe
val analyzer: global.analyzer.type = global.analyzer
val callsiteContext = context.callsiteTyper.context
val typ = weakTypeOf[A]
val search = new analyzer.ImplicitSearch(
tree = EmptyTree.asInstanceOf[global.Tree],
pt = typ.asInstanceOf[global.Type],
isView = false,
context0 = callsiteContext.makeImplicit(reportAmbiguousErrors = true),
pos0 = c.enclosingPosition.asInstanceOf[scala.reflect.internal.util.Position]
)
println(s"allImplicits=${search.allImplicits}")
q""
}
then
foo[GetItem[List, Double, Int] { type Out = List[Double] }]
produces a warning
App.this.GetItem.ifRefIsInt is not a valid implicit value for App.GetItem[List,Double,Int]{type Out = List[Double]} because:
hasMatchingSymbol reported error: polymorphic expression cannot be instantiated to expected type;
found : [A[_], T]App.GetItem[A,T,Int]{type Out = A[T]}
required: App.GetItem[List,Double,Int]{type Out = List[Double]}
scalac: allImplicits=List()
i.e. A
, T
are not inferred.
回答2:
Not sure why it doesn't work like that, however a good technique to solve this kind of problems is to use the Aux pattern to lift the type members into a type parameter, which improves resolution.
trait GetItem[A[_], T, R] {
type Out
def ret(a: A[T], ref: R): Out
}
object GetItem {
type Aux[A[_], T, R, O] = GetItem[A, T, R] { type Out = O }
implicit def ifRefIsInt[A[_], T]: Aux[A, T, Int, A[T]] = new GetItem[A, T, Int] {
type Out = A[T]
def ret(a: A[T], ref: Int): Out = a
}
}
Which you can test like this:
implicitly[GetItem.Aux[List, Double, Int, List[Double]]]
// res: GetItem.Aux[List, Double, Int, List[Double]] = ...
implicitly[GetItem[List, Double, Int]]
// res: GetItem[List, Double, Int] = ...
来源:https://stackoverflow.com/questions/64814539/why-is-this-implicit-resolution-failing