Demo of generalised type constraints in Scala 2.8

generalised-type-constraints.scala

// scala 2.7 simple type constraint. This can only constrain a type parameter of this function.
// Below, in ListW.sumint28, we can't use this approach because we want to constrain T, 
// a type param of the enclosing trait.
def sumint27A[T <: Int](l: List[T]) : Int = l.reduceLeft((a: Int, b: Int) => a + b)

trait IntLike[X] extends (X => Int)
object IntLike {
  implicit val intIntLike: IntLike[Int] = new IntLike[Int] { def apply(x: Int) = identity(x) }
}

trait ListW[T] {
  val l: List[T]

  // T can be constrained within the context of this function by requiring an implicit parameter
  // to certify that it is IntLike.
  def sumint27B(implicit ev: IntLike[T]) : Int = l.map(ev).reduceLeft(_ + _)

  // In scala 2.8, we can require an implicit parameter of type T <:< Int (that is, <:<[T, Int])
  // The compiler will provide this if T conforms to Int. The parameter `ev` can be used to
  // safely cast a value T to Int.
  def sumint28(implicit ev: T <:< Int) : Int = l.map(ev).reduceLeft(_ + _)

}

implicit def ListW[T](list: List[T]): ListW[T] = new ListW[T] {
  val l = list
}

val intList = List(1, 2, 3)
val stringList = List("1", "2", "3")

sumint27A(intList)

{
  import IntLike._
  intList.sumint27B
}

intList.sumint28

/* Compiler errors:
sumint27A(stringList) // type arguments [java.lang.String] do not conform to method sumint27's type parameter bounds [T <: Int]
stringList.sumint27 // compile error: could not find implicit value for parameter ev: IntLike[Any]
stringList.sumint28 // compile error: could not find implicit value for parameter ev: <:<[java.lang.String,Int]
*/

Just out of curiosity, when would you use this sort of subclassing constraint rather than a type class?

Just out of curiosity, when would you use this sort of subclassing constraint rather than a type class?

It comes in handy when pimping generic types, e.g:

class Tuple2W[A, B](t2: (A, B)) {
    def toSeq[C](implicit ac: A <:< C, bc: B <:
Here we aren't relying on any properties of the types, just finding the Least Upper Bound C.
This can also be used at times when a methods with a regular type bound fails type inference. No hard and fast rules, just can try it as a method of last resort if inference isn't playing nicely.
def foo[X, Y <: FooX = ...

def foo[X, Y <: Foo[X](implicit yfx: Y <:< Foo[X]) = ...
e.g. http://gist.github.com/445874

It comes in handy when pimping generic types, e.g:

class Tuple2W[A, B](t2: (A, B)) {
    def toSeq[C](implicit ac: A <:< C, bc: B <:
Here we aren't relying on any properties of the types, just finding the Least Upper Bound C.
This can also be used at times when a methods with a regular type bound fails type inference. No hard and fast rules, just can try it as a method of last resort if inference isn't playing nicely.
def foo[X, Y <: FooX = ...

def foo[X, Y <: Foo[X](implicit yfx: Y <:< Foo[X]) = ...
e.g. http://gist.github.com/445874