language

The experimental object contains features that are known to have unstable API or behavior that may change in future releases.

Experimental features may undergo API changes in future releases, so production code should not rely on them.

Programmers are encouraged to try out experimental features and report any bugs or API inconsistencies they encounter so they can be improved in future releases.

Only where this feature is enabled, can direct or indirect subclasses of trait scala.Dynamic be defined. If dynamics is not enabled, a definition of a class, trait, or object that has Dynamic as a base trait is rejected by the compiler.

Selections of dynamic members of existing subclasses of trait Dynamic are unaffected; they can be used anywhere.

Why introduce the feature? To enable flexible DSLs and convenient interfacing with dynamic languages.

Why control it? Dynamic member selection can undermine static checkability of programs. Furthermore, dynamic member selection often relies on reflection, which is not available on all platforms.

Where this feature is enabled, existential types that cannot be expressed as wildcard types can be written and are allowed in inferred types of values or return types of methods. If existentials is not enabled, those cases will trigger a warning from the compiler.

Existential types with wildcard type syntax such as List[?], or Map[String, ?] are not affected.

Why keep the feature? Existential types are needed to make sense of Java’s wildcard types and raw types and the erased types of run-time values.

Why control it? Having complex existential types in a code base usually makes application code very brittle, with a tendency to produce type errors with obscure error messages. Therefore, going overboard with existential types is generally perceived not to be a good idea. Also, complicated existential types might be no longer supported in a future simplification of the language.

Where this feature is enabled, higher-kinded types can be written. If higherKinds is not enabled, a higher-kinded type such as F[A] will trigger a warning from the compiler.

Why keep the feature? Higher-kinded types enable the definition of very general abstractions such as functor, monad, or arrow. A significant set of advanced libraries relies on them. Higher-kinded types are also at the core of the scala-virtualized effort to produce high-performance parallel DSLs through staging.

Why control it? Higher kinded types in Scala lead to a Turing-complete type system, where compiler termination is no longer guaranteed. They tend to be useful mostly for type-level computation and for highly generic design patterns. The level of abstraction implied by these design patterns is often a barrier to understanding for newcomers to a Scala codebase. Some syntactic aspects of higher-kinded types are hard to understand for the uninitiated and type inference is less effective for them than for normal types. Because we are not completely happy with them yet, it is possible that some aspects of higher-kinded types will change in future versions of Scala. So an explicit enabling also serves as a warning that code involving higher-kinded types might have to be slightly revised in the future.

There are two kinds of implicit conversions, depending on whether we are in Scala 2 or 3. This language import has different meanings depending on which kind of conversion is used.

In Scala 2, an implicit conversion is an implicit method that has in its first parameter section a single, non-implicit parameter, or it is an implicit value of function type A => B. Examples:

   implicit def intToString(i: Int): String = s"$i"
   implicit val conv: Int => String = i => s"$i"
   implicit val numerals: List[String] = List("zero", "one", "two", "three")
   implicit val strlen: String => Int = _.length
   implicit def listToInt[T](xs: List[T])(implicit f: T => Int): Int = xs.map(f).sum

The language import controls whether definitions of these Scala 2 conversions are allowed. If implicitConversions is not enabled, the definition of an implicit conversion method will trigger a warning from the compiler. The warning is only for implicit conversions introduced by methods, implicit values are unaffected. Implicit class definitions, which introduce a conversion to the wrapping class, also do not warn.

In Scala 3, an implicit conversion is a given of type scala.Conversion. Examples:

    given Conversion[Int, BigInt] = (i: Int) => BigInt(i)
    given [T] => Conversion[T, Option[T]] = Some(_)

As of Scala 3.9, the language still supports Scala-2 style conversions, but these are slated to be phased out.

In Scala 3, the use of a (new style) implicit conversion triggers a warning or an error if the implicitConversions language import is not given. Exempted are only uses that map a value into an into[...] type, since into[...] signifies that an implicit conversion is allowed. Everwhere else the language import has to be given. Example:

    import Conversion.into
    given [T] => Conversion[T, Option[T]] = Some(_)

    def strict(x: Option[String]) = ()
    def lenient(x: into[Option[String]]) = ()

    strict("abc")    // error or warning: language import needed
    lenient("abc")   // ok

The precise kind of diagnostic depends on the Scala version:

• For Scala 3.0 - 3.9: feature warning.
• For Scala 3.10: regular warning.
• For Scala 3.11 and higher: error.

Why keep the feature? Implicit conversions can add flexibility. They are widely used in many libraries including Scala’s collection library.

Why control it? Implicit conversions may have unexpected global effects, potentially causing unexpected behavior or performance degradation. Their presence also imposes a tax on precise type inference. It's therefore better to limit their use to specific situations where a library explicitly allows them by wrapping expected argument types in into.

Only where this feature is enabled, is postfix operator notation (expr op) permitted. If postfixOps is not enabled, an expression using postfix notation is rejected by the compiler.

Why keep the feature? Postfix notation is preserved for backward compatibility only. Historically, several DSLs written in Scala need the notation.

Why control it? Postfix operators interact poorly with semicolon inference. Most programmers avoid them for this reason alone. Postfix syntax is associated with an abuse of infix notation, a op1 b op2 c op3, that can be harder to read than ordinary method invocation with judicious use of parentheses. It is recommended not to enable this feature except for legacy code.

Where this feature is enabled, accesses to members of structural types that need reflection are supported. If reflectiveCalls is not enabled, an expression requiring reflection will trigger a warning from the compiler.

A structural type is a type of the form Parents { Decls } where Decls contains declarations of new members that do not override any member in Parents. To access one of these members, a reflective call is needed.

Why keep the feature? Structural types provide great flexibility because they avoid the need to define inheritance hierarchies a priori. Besides, their definition falls out quite naturally from Scala’s concept of type refinement.

Why control it? Reflection is not available on all platforms. Popular tools such as ProGuard have problems dealing with it. Even where reflection is available, reflective dispatch can lead to surprising performance degradations.