Edit this page on GitHub

Implicit Conversions - More Details

Implementation

An implicit conversion, or view, from type S to type T is defined by either:

  • An implicit def which has type S => T or (=> S) => T
  • An implicit value which has type Conversion[S, T]

The standard library defines an abstract class Conversion:

package scala
@java.lang.FunctionalInterface
abstract class Conversion[-T, +U] extends Function1[T, U]:
  def apply(x: T): U

Function literals are automatically converted to Conversion values.

Views are applied in three situations:

  1. If an expression e is of type T, and T does not conform to the expression's expected type pt. In this case, an implicit v which is applicable to e and whose result type conforms to pt is searched. The search proceeds as in the case of implicit parameters, where the implicit scope is the one of T => pt. If such a view is found, the expression e is converted to v(e).
  2. In a selection e.m with e of type T, if the selector m does not denote an accessible member of T. In this case, a view v which is applicable to e and whose result contains an accessible member named m is searched. The search proceeds as in the case of implicit parameters, where the implicit scope is the one of T. If such a view is found, the selection e.m is converted to v(e).m.
  3. In an application e.m(args) with e of type T, if the selector m denotes some accessible member(s) of T, but none of these members is applicable to the arguments args. In this case, a view v which is applicable to e and whose result contains a method m which is applicable to args is searched. The search proceeds as in the case of implicit parameters, where the implicit scope is the one of T. If such a view is found, the application e.m(args) is converted to v(e).m(args).

Differences with Scala 2 implicit conversions

In Scala 2, views whose parameters are passed by-value take precedence over views whose parameters are passed by-name. This is no longer the case in Scala 3. A type error reporting the ambiguous conversions will be emitted in cases where this rule would be applied in Scala 2:

implicit def conv1(x: Int): String = x.toString
implicit def conv2(x: => Int): String = x.toString

val x: String = 0 // Compiles in Scala2 (uses `conv1`),
                  // type error in Scala 3 because of ambiguity.

In Scala 2, implicit values of a function type would be considered as potential views. In Scala 3, these implicit value need to have type Conversion:

// Scala 2:
def foo(x: Int)(implicit conv: Int => String): String = x

// Becomes with Scala 3:
def foo(x: Int)(implicit conv: Conversion[Int, String]): String = x

// Call site is unchanged:
foo(4)(_.toString)

// Scala 2:
implicit val myConverter: Int => String = _.toString

// Becomes with Scala 3:
implicit val myConverter: Conversion[Int, String] = _.toString

Note that implicit conversions are also affected by the changes to implicit resolution between Scala 2 and Scala 3.

Motivation for the changes

The introduction of scala.Conversion in Scala 3 and the decision to restrict implicit values of this type to be considered as potential views comes from the desire to remove surprising behavior from the language:

implicit val m: Map[Int, String] = Map(1 -> "abc")

val x: String = 1  // Scala 2: assigns "abc" to x
                   // Scala 3: type error

This snippet contains a type error. The right-hand side of val x does not conform to type String. In Scala 2, the compiler will use m as an implicit conversion from Int to String, whereas Scala 3 will report a type error, because Map isn't an instance of Conversion.

Migration path

Implicit values that are used as views should see their type changed to Conversion.

For the migration of implicit conversions that are affected by the changes to implicit resolution, refer to the Changes in Implicit Resolution for more information.

Reference

For more information about implicit resolution, see Changes in Implicit Resolution. Other details are available in PR #2065.