When I try to compile the small example:
trait Foo[A,B] {
type F[_,_]
def foo(): F[A,B]
}
class Bar[A,B] extends Foo[A,B] {
type F[D,E] = Bar[D,E]
def foo() = this
}
object Helper {
def callFoo[A,B,FF <: Foo[A,B]]( f: FF ): FF#F[A,B] =
f.foo()
}
object Run extends App {
val x = new Bar[Int,Double]
val y = Helper.callFoo(x)
println( y.getClass )
}
I get the error:
[error] src/Issue.scala:20: inferred type arguments
[Nothing,Nothing,issue.Bar[Int开发者_如何学编程,Double]] do not conform to method callFoo's type
parameter bounds [A,B,FF <: issue.Foo[A,B]]
[error] val y = Helper.callFoo(x)
Apparently, the type inference mechanism is not able to infer A and B out of Bar[A,B]. However, it works if I pass all the types by hand:
val y = Helper.callFoo[Int,Double,Bar[Int,Double]](x)
I there a way to avoid passing types explicitly?
You'll have to change the signature of callFoo
to this:
def callFoo[A, B, FF[A, B] <: Foo[A, B]](f: FF[A, B]): FF[A, B]#F[A, B] =
You have to tell the compiler that FF
is actually a parametrized type.
Would it work to use type members instead of parameters?
trait Foo {
type A
type B
type F
def foo(): F
}
class Bar extends Foo {
type F = Bar
def foo() = this
}
object Helper {
def callFoo[FF <: Foo]( f: FF ): FF#F =
f.foo()
}
object Run extends App {
val x = new Bar{type A=Int; type B=Double}
val y = Helper.callFoo(x)
println( y.getClass )
}
When using type members, it's useful to know that they can be surfaced as type parameters using refinement, as in Miles Sabin's answer to: Why is this cyclic reference with a type projection illegal?
See also this recent question, which seems similar to yours: Scala fails to infer the right type arguments
精彩评论