scala.annotation

A base class for annotations.

Annotations extending this class directly are not preserved in the classfile. To enable storing annotations in the classfile's Scala signature and make it available to Scala reflection and other tools, the annotation needs to inherit from scala.annotation.StaticAnnotation.

Annotation classes defined in Scala are not stored in classfiles in a Java-compatible manner and therefore not visible in Java reflection. In order to achieve this, the annotation has to be written in Java.

Annotation classes extending this trait only accept constant values as arguments.

Note that this trait extends StaticAnnotation, so constant annotations are persisted in the classfile.

The implementation requires arguments of constant annotations to be passed as named arguments, except if there is a single argument, which then defines the annotation's parameter named value.

Constant annotations may use default arguments. Note that the internal representation of an annotation usage (which is visible for compiler plugins, for example) only contains arguments that are explicitly provided.

Constant annotations are not allowed to define auxiliary constructors, and the primary constructor is required to have a single parameter list.

Example:

class Ann(value: Int, x: Int = 0) extends scala.annotation.ConstantAnnotation
class Test {
  def someInt = 0
  @Ann(value = 0, x = 1) def g = 0
  @Ann(0) def f = 0                 // Internal representation contains `@Ann(value = 0)`
  @Ann(someInt)                     // error: argument needs to be a compile-time constant
}

A base class for static annotations. These are available to the Scala type checker or Scala reflection, even across different compilation units.

Annotation classes defined in Scala are not stored in classfiles in a Java-compatible manner and therefore not visible in Java reflection. In order to achieve this, the annotation has to be written in Java.

A marker for annotations that, when applied to a type, should be treated as a constraint on the annotated type.

A proper constraint should restrict the type based only on information mentioned within the type. A Scala compiler can use this assumption to rewrite the contents of the constraint as necessary. To contrast, a type annotation whose meaning depends on the context where it is written down is not a proper constrained type, and this marker should not be applied. A Scala compiler will drop such annotations in cases where it would rewrite a type constraint.

An annotation that designates that an annottee should not be referred to after type checking (which includes macro expansion).

Examples of potential use: 1) The annottee can only appear in the arguments of some other macro that will eliminate it from the AST during expansion. 2) The annottee is a macro and should have been expanded away, so if hasn't, something wrong has happened. (Comes in handy to provide better support for new macro flavors, e.g. macro annotations, that can't be expanded by the vanilla compiler).

An annotation for methods whose bodies may be excluded from compiler-generated bytecode.

Behavior is influenced by passing -Xelide-below <arg> to scalac. Calls to methods marked elidable (as well as the method body) will be omitted from generated code if the priority given the annotation is lower than that given on the command line.

@elidable(123)           // annotation priority
scalac -Xelide-below 456 // command line priority

The method call will be replaced with an expression which depends on the type of the elided expression. In decreasing order of precedence:

Unit            ()
Boolean         false
T <: AnyVal     0
T >: Null       null
T >: Nothing    Predef.???

Complete example:

import scala.annotation._, elidable._
object Test extends App {
  def expensiveComputation(): Int = { Thread.sleep(1000) ; 172 }

  @elidable(WARNING) def warning(msg: String) = println(msg)
  @elidable(FINE) def debug(msg: String)      = println(msg)
  @elidable(FINE) def computedValue           = expensiveComputation()

  warning("Warning! Danger! Warning!")
  debug("Debug! Danger! Debug!")
  println("I computed a value: " + computedValue)
}
% scalac example.scala && scala Test
Warning! Danger! Warning!
Debug! Danger! Debug!
I computed a value: 172

// INFO lies between WARNING and FINE
% scalac -Xelide-below INFO example.scala && scala Test
Warning! Danger! Warning!
I computed a value: 0

Note that only concrete methods can be marked @elidable. A non-annotated method is not elided, even if it overrides / implements a method that has the annotation.

Also note that the static type determines which annotations are considered:

import scala.annotation._, elidable._
class C { @elidable(0) def f(): Unit = ??? }
object O extends C { override def f(): Unit = println("O.f") }
object Test extends App {
  O.f()      // not elided
  (O: C).f() // elided if compiled with `-Xelide-below 1`
}

This useless appearing code was necessary to allow people to use named constants for the elidable annotation. This is what it takes to convince the compiler to fold the constants: otherwise when it's time to check an elision level it's staring at a tree like

(Select(Level, Select(FINEST, Apply(intValue, Nil))))

instead of the number 300.

To customize the error message that's emitted when an implicit search finds multiple ambiguous values, annotate at least one of the implicit values @implicitAmbiguous. Assuming the implicit value is a method with type parameters X1,..., XN, the error message will be the result of replacing all occurrences of ${Xi} in the string msg with the string representation of the corresponding type argument Ti.

If more than one @implicitAmbiguous annotation is collected, the compiler is free to pick any of them to display.

Nice errors can direct users to fix imports or even tell them why code intentionally doesn't compile.

trait =!=[C, D]

implicit def neq[E, F] : E =!= F = null

@annotation.implicitAmbiguous("Could not prove ${J} =!= ${J}")
implicit def neqAmbig1[G, H, J] : J =!= J = null
implicit def neqAmbig2[I] : I =!= I = null

implicitly[Int =!= Int]

To customize the error message that's emitted when an implicit of type C[T1,..., TN] cannot be found, annotate the class C with @implicitNotFound. Assuming C has type parameters X1, ..., XN, the error message will be the result of replacing all occurrences of ${Xi} in the string msg with the string representation of the corresponding type argument Ti. The annotation is effectively inherited by subtypes if they are not annotated.

The annotation can also be attached to implicit parameters. In this case, ${Xi} can refer to type parameters in the current scope. The @implicitNotFound message on the parameter takes precedence over the one on the parameter's type.

import scala.annotation.implicitNotFound

@implicitNotFound("Could not find an implicit C[${T}, ${U}]")
class C[T, U]

class K[A] {
  def m[B](implicit c: C[List[A], B]) = 0
  def n[B](implicit @implicitNotFound("Specific message for C of list of ${A} and ${B}") c: C[List[A], B]) = 1
}

object Test {
  val k = new K[Int]
  k.m[String]
  k.n[String]
}

The compiler issues the following error messages:

Test.scala:13: error: Could not find an implicit C[List[Int], String]
 k.m[String]
    ^
Test.scala:14: error: Specific message for C of list of Int and String
 k.n[String]
    ^

An annotation for local warning suppression.

The optional value parameter allows selectively silencing messages, see scalac -Wconf:help for help. Examples:

def f = {
  1: @nowarn // don't warn "a pure expression does nothing in statement position"
  2
}

@nowarn def f = { 1; deprecated() } // don't warn

@nowarn("msg=pure expression does nothing")
def f = { 1; deprecated() } // show deprecation warning

To ensure that a @nowarn annotation actually suppresses a warning, enable -Xlint:unused or -Wunused:nowarn. The unused annotation warning is emitted in category unused-nowarn and can be selectively managed using -Wconf:cat=unused-nowarn:s.

This annotation configures how Scala prints two-parameter generic types.

By default, types with symbolic names are printed infix; while types without them are printed using the regular generic type syntax.

Example of usage:

scala> class Map[T, U]
defined class Map

scala> def foo: Int Map Int = ???
foo: Map[Int,Int]

scala> @showAsInfix class Map[T, U]
defined class Map

scala> def foo: Int Map Int = ???
foo: Int Map Int

An annotation to be applied to a match expression. If present, the compiler will verify that the match has been compiled to a tableswitch or lookupswitch and issue a warning if it instead compiles into a series of conditional expressions. Example usage:

val Constant = 'Q'
def tokenMe(ch: Char) = (ch: @switch) match {
  case ' ' | '\t' | '\n'  => 1
  case 'A' | 'Z' | '$'    => 2
  case '5' | Constant     => 3  // a non-literal may prevent switch generation: this would not compile
  case _                  => 4
}

Note: for pattern matches with one or two cases, the compiler generates jump instructions. Annotating such a match with @switch does not issue any warning.

A method annotation which verifies that the method will be compiled with tail call optimization.

If it is present, the compiler will issue an error if the method cannot be optimized into a loop.

Mark an element unused for a given context.

Unused warnings are suppressed for elements known to be unused.

For example, a method parameter may be marked @unused because the method is designed to be overridden by an implementation that does use the parameter.