[ 2.10.0 [1] | 2.9.0 [2] | 2.8.0 [3] | 2.7.2 [4] | 2.7.1 [5] | 2.7.0 [6] | 2.6.1 [7] | 2.6.0 [8] | 2.5.0 [9] | 2.4.0 [10] | 2.3.2 [11] | 2.3.0 [12] | 2.1.8 [13] | 2.1.7 [14] | 2.1.5 [15] | 2.0 [16] ]
The Scala 2.10.0 codebase includes the following new features and changes:
AnyVal to make it behave like a struct type (restrictions apply).val what = "awesome"; println(s"string interpolation is ${what.toUpperCase}!")for (req <- WS.url(restApiUrl).get()) yield (req.json \ "users").as[List[User]] (uses play!)x.foo becomes x.applyDynamic("foo") if x's type does not define a foo, but is a subtype of Dynamicdef identity(x: AnyRef): x.type = x // the return type says we return exactly what we got-Xno-patmat-analysis)TreeSet/TreeMap??? and NotImplementedErrorIsTraversableOnce + IsTraversableLike type classes for extension methodsThe following exciting -- experimental -- features are part of 2.10.0:
The API is subject to (possibly major) changes in the 2.11.x series, but don't let that stop you from experimenting with them! A lot of developers have already come up with very cool applications for them.
No language changes were introduced in Scala 2.9.1, 2.9.1-1, and 2.9.2.
The Scala 2.9.0 codebase includes the following new features and changes:
Every collection may be converted into a corresponding parallel collection with the new `par` method. Parallel collections utilize multicore processors by implementing bulk operations such as `foreach`, `map`, `filter` etc. in parallel. Parallel collections are located in the package `scala.collection.parallel`.
Depending on the collection in question, `par` may require copying the underlying dataset to create a parallel collection. However, specific collections share their underlying dataset with a parallel collection, making `par` a constant time operation.
Currently available parallel collections are:
The method `seq` is used to convert from a parallel collection to a corresponding sequential collection. This method is always efficient (O(1)).
The App trait is a safer, more powerful alternative to the previous Application trait, which has now been deprecated. The new recommended way to write a top-level application is like this:
object Echo extends App {
println("Echo" + (args mkString " "))
}
Objects inheriting from the old Application trait were almost as convenient to write, but were not thread-safe and were often not optimized by the VM, since the application’s body was execited as part of of the object’s initialization sequence. Objects inheriting the App trait instead make use of Scala 2.9’s delayed initialization feature to execute the whole body as part of an inherited main method.
Another new feature of the App scheme is that command line arguments are now accessible via the args value (which is inherited from trait App)
The DelayedInit trait provides another tool to customize initialization sequences of classes and objects. If a class or object inherits from this trait, all its initialization code is packed in a closure and forwarded as an argument to a method named delayedInit which is defined as an abstract method in trait DelayedInit.
Implementations of delayedInit have thus full freedom when to execute the initialization code. For instance, Scala’s new App trait stores all initialization sequences in an internal buffer and executes them when the object’s main method is called.
Note that only initialization code contained in classes and objects is passed to DelayedInit; initialization code contained in traits is not affected.
Improvements in jline, the repl input handler. More robust cursor handling, bash-style ctrl-R history search, new commands like :imports, :implicits, :keybindings. On platforms with the necessary runtime support, :javap will disassemble any class including repl-defined ones. A long-running repl command can now be interrupted via ctrl-C without terminating the repl session. Improved programmability: the repl classloader exposes repl-defined classes via their given names.
Scala code can now be executed in any of the following ways:
The @strictfp annotation is now supported.
Various fixes in JavaConverters and JavaConversions for smoother interoperation.
Primitive types and their boxed versions are now implicitly converted bidirectionally.
try body catch handler finally cleanup
Here, body and cleanup can be arbitrary expressions, and handler can be any expression which evaluates to a valid exception handler (which is: PartialFunction[Throwable, T]).
No language changes were introduced in Scala 2.8.1 and 2.8.2.
Scala 2.8.0 is a significantly innovative release, which contains a large amount of fixes and introduces many new features:
No language changes were introduced in 2.7.3 through 2.7.7.
The Scala compiler now generates Java's generic signatures, so that Scala generics are visible to Java.
The compiler can now parse (but not translate) Java source files. This makes it possible to have mixed Java/Scala projects with recursive dependencies between them. In such a project, you can submit first all the Java and Scala sources to the Scala compiler. In a second step, the Java sources are compiled using the Scala generated .class files and the Scala sources are compiled again using the Java generated .class files.
Another major addition is the first beta version of the ScalaSwing library, which is now bundled with the distribution.
There are new implementations of collection classes, contributed by David MacIver: IntMap, LongMap, and TreeHashMap (immutable), ArrayStack and OpenHashMap (mutable).
A wildcard in a type now binds to the closest enclosing type application.
For example List[List[_]] is now equivalent to the existential type
List[List[t] forSome { type t }]
In version 2.7.0, the type expanded instead to
List[List[t]] forSome { type t }
The new convention corresponds exactly to the way wildcards in Java are interpreted.
The contractiveness requirement for implicit method definitions has been dropped. Instead it is checked for each implicit expansion individually that the expansion does not result in a cycle or a tree of infinitely growing types.
Scala now supports Java generic types [36] by default:
ArrayList<String> is translated to a generic type in Scala: ArrayList[String].ArrayList<? extends Number> is translated to ArrayList[_ <: Number]. This is itself a shorthand for the existential type ArrayList[T] forSome { type T <: Number }.ArrayList is translated to ArrayList[_], which is a shorthand for ArrayList[T] forSome { type T }.This translation works if -target:jvm-1.5 is specified, which is the new default. For any other target, Java generics are not recognized. To ensure upgradability of Scala codebases, extraneous type parameters for Java classes under -target:jvm-1.4 [37] are simply ignored. For instance, when compiling with -target:jvm-1.4, a Scala type such as ArrayList[String] is simply treated as the unparameterized type ArrayList.
The Scala compiler generates now for every case class a companion extractor object. For instance, given the case class:
case class X(elem: String)
the following companion object is generated:
object X { def unapply(x: X): Some[String] = Some(x.elem) def apply(s: String): X = new X(s) }
If the object exists already, only the apply and unapply methods are added to it.
Three restrictions on case classes have been removed:
case class Foo(x: Int) case class Bar(override val x: Int, y: Int) extends Foo(x) object test extends Application { println(Bar(1, 2).x) (Bar(1, 2): Foo) match { case Foo(x) => println(x) } }
abstract.case class Foo(x: Int) object Foo { val x = 2 val y = Foo(2) } object test extends Application { println(Foo.x) println(Foo.y match { case Foo(x) => x } ) }
The following deprecated features have been removed from the standard Scala library:
| Removed | Use instead |
|---|---|
All and AllRef
(object
scala.Predef [38]) |
Nothing and Null (available since 2.3.0 [39]) |
element and arity
(class
scala.Product) |
productElement and productArity |
scala.compat.Math |
scala.Math [40] |
scala.testing.UnitTest |
scala.testing.SUnit [41] |
assertNotSame and assertSame
(class
scala.testing.SUnit.Assert [42]) |
assertNotEq and assertEq |
scala.util.Fluid |
scala.util.DynamicVariable [43] |
Mutable variables can now be introduced by a pattern matching definition, just like values can. For example:
var (x, y) = if (positive) (1, 2) else (-1, -3) var hd :: tl = mylist
Self types can now be introduced without defining an alias name for this. For example:
class C { type T <: Trait trait Trait { this: T => ... } }
It is now possible to define existential types using the new keyword forSome. An existential type has the form T forSome {Q} where Q is a sequence of value and/or type declarations. Given the class definitions
class Ref[T] abstract class Outer { type T }
one may for example write the following existential types
Ref[T] forSome { type T <: java.lang.Number } Ref[x.T] forSome { val x: Outer }
It is now possible to define lazy value declarations using the new modifier lazy.
import compat.Platform._ val t0 = currentTime lazy val t1 = currentTime val t2 = currentTime println("t0 <= t2: " + (t0 <= t2)) //true println("t1 <= t2: " + (t1 <= t2)) //false (lazy evaluation of t1)
It is now possible to declare structural types using type refinements. For example:
class File(name: String) { def getName(): String = name def open() { /*..*/ } def close() { println("close file") } } def test(f: { def getName(): String }) { println(f.getName) } test(new File("test.txt")) test(new java.io.File("test.txt"))
for-comprehensions has been deprecated.requires clause has been deprecated; use { self: T => ... } instead.&f for unapplied methods has been deprecated; use f _ instead.Type parameters and abstract type members can now also abstract over type constructors. This allows a more precise Iterable interface:
trait Iterable[+t] { type MyType[+t] <: Iterable[t] // MyType is a type constructor def filter(p: t => Boolean): MyType[t] = //... def map[s](f: t => s): MyType[s] = //... } abstract class List[+t] extends Iterable[t] { type MyType[+t] = List[t] }
This definition of Iterable makes explicit that mapping a function over a certain structure (e.g., a List) will yield the same structure (containing different elements).
It is now possible to initialize some fields of an object before any parent constructors are called. This is particularly useful for traits, which do not have normal constructor parameters. For example:
trait Greeting { val name: String val msg = "How are you, " + name } class C extends { val name = "Bob" } with Greeting { println(msg) }
In the code above, the field name is initialized before the constructor of Greeting is called. Therefore, field msg in class Greeting is properly initialized to "How are you, Bob".
for-comprehensions, revisedThe syntax of for-comprehensions has been changed. For example:
for (val x <- List(1, 2, 3); x % 2 == 0) println(x)
is now written
for (x <- List(1, 2, 3) if x % 2 == 0) println(x)
Thus a for-comprehension now starts with a (possibly guarded) generator followed by one or more enumerators which can be either a (possibly guarded) generator, a guard or a local value definition.
The old syntax is still available but will be deprecated in the future.
It is now possible to define anonymous functions using underscores in parameter position. For instance, the expressions in the left column are each function values which expand to the anonymous functions on their right.
_ + 1 x => x + 1 _ * _ (x1, x2) => x1 * x2 (_: int) * 2 (x: int) => (x: int) * 2 if (_) x else y z => if (z) x else y _.map(f) x => x.map(f) _.map(_ + 1) x => x.map(y => y + 1)
As a special case, a partially unapplied method is now designated m _ instead of the previous notation &m.
The new notation will displace the special syntax forms .m() for abstracting over method receivers and &m for treating an unapplied method as a function value. For the time being, the old syntax forms are still available, but they will be deprecated in the future.
It is now possible to use case clauses to define a function value directly for functions of arities greater than one. Previously, only unary functions could be defined that way. For example:
def scalarProduct(xs: Array[Double], ys: Array[Double]) = (0.0 /: (xs zip ys)) { case (a, (b, c)) => a + b * c }
The private and protected modifiers now accept a [this] qualifier . A definition M which is labelled private[this] is private, and in addition can be accessed only from within the current object. That is, the only legal prefixes for M are this or C.this. Analogously, a definition M which is labelled protected[this] is protected, and in addition can be accessed only from within the current object.
The syntax for tuples has been changed from {...} to (...) . For any sequence of types T1, ..., Tn,
(T1, ..., Tn) is a shorthand for Tuplen[T1, ..., Tn].
Analogously, for any sequence of expressions or patterns x1, ..., xn,
(x1, ..., xn) is a shorthand for Tuplen(x1, ..., xn).
The primary constructor of a class can now be marked private or protected . If such an access modifier is given, it comes between the name of the class and its value parameters. For example:
class C[T] private (x: T) { ... }
The support for attributes has been extended and its syntax changed . Attributes are now called annotations. The syntax has been changed to follow Java's conventions, e.g. @annotation instead of [attribute]. The old syntax is still available but will be deprecated in the future.
Annotations are now serialized so that they can be read by compile-time or run-time tools. Class scala.Annotation has two sub-traits which are used to indicate how annotations are retained. Instances of an annotation class inheriting from trait scala.ClassfileAnnotation [44] will be stored in the generated class files. Instances of an annotation class inheriting from trait scala.StaticAnnotation [45] will be visible to the Scala type-checker in every compilation unit where the annotated symbol is accessed.
The implementation of subtyping has been changed to prevent infinite recursions. Termination of subtyping is now ensured by a new restriction of class graphs to be finitary .
It is now explicitly ruled out that case classes can be abstract . The specification was silent on this point before, but did not explain how abstract case classes were treated. The Scala compiler allowed the idiom.
It is now possible to give an explicit alias name and/or type for the self reference this . For instance, in
class C { self: D => ... }
the name self is introduced as an alias for this within C and the self type of C is assumed to be D. This construct is introduced now in order to replace eventually both the qualified this construct C.this and the requires clause in Scala.
It is now possible to combine operators with assignments . For example:
var x: int = 0 x += 1
It is now possible to define patterns independently of case classes, using unapply methods in extractor objects . Here is an example:
object Twice { def apply(x: Int): int = x*2 def unapply(z: Int): Option[int] = if (z%2 == 0) Some(z/2) else None } val x = Twice(21) x match { case Twice(n) => Console.println(n) } // prints 21
In the example, Twice is an extractor object with two methods:
apply method is used to build even numbers.unapply method is used to decompose an even number; it is in a sense the reverse of apply. unapply methods return option types: Some(...) for a match that suceeds, None for a match that fails. Pattern variables are returned as the elements of Some. If there are several variables, they are grouped in a tuple.In the second-to-last line, Twice's apply method is used to construct a number x. In the last line, x is tested against the pattern Twice(n). This pattern succeeds for even numbers and assigns to the variable n one half of the number that was tested. The pattern match makes use of the unapply method of object Twice. More details on extractors can be found in the paper "Matching Objects with Patterns" [46] by Emir, Odersky and Williams.
A new lightweight syntax for tuples has been introduced . For any sequence of types {T1 ,.., Tn},
{T1 ,.., Tn} is a shorthand for Tuplen[T1 ,.., Tn].
Analogously, for any sequence of expressions or patterns x1,.., xn,
{x1 ,.., xn} is a shorthand for Tuplen(x1 ,.., xn).
It is now possible to use methods which have more than one parameter as infix operators . In this case, all method arguments are written as a normal parameter list in parentheses. Example:
class C { def +(x: int, y: String) = ... } val c = new C c + (1, "abc")
A new standard attribute deprecated is available . If a member definition is marked with this attribute, any reference to the member will cause a "deprecated" warning message to be emitted.
A simplified syntax for functions returning unit has been introduced . Scala now allows the following shorthands:
def f(params) |
for | def f(params): unit |
def f(params) { ... } |
for | def f(params): unit = { ... } |
The syntax of types in patterns has been refined . Scala now distinguishes between type variables (starting with a lower case letter) and types as type arguments in patterns. Type variables are bound in the pattern. Other type arguments are, as in previous versions, erased. The Scala compiler will now issue an "unchecked" warning at places where type erasure might compromise type-safety.
The recommended names for the two bottom classes in Scala's type hierarchy have changed as follows:
All ==> Nothing AllRef ==> Null
The old names are still available as type aliases.
Protected members can now have a visibility qualifier , e.g. protected[<qualifier>]. In particular, one can now simulate package protected access as in Java writing
protected[P] def X ...
where P would name the package containing X.
Private members of a class can now be referenced from the companion module of the class and vice versa .
The lookup method for implicit definitions has been generalized . When searching for an implicit definition matching a type T, now are considered
T.(The second clause is more general than before). Here, a class is associated with a type T if it is referenced by some part of T, or if it is a base class of some part of T. For instance, to find implicit members corresponding to the type
HashSet[List[Int], String]
one would now look in the companion modules (aka static parts) of HashSet, List, Int, and String. Before, it was just the static part of HashSet.
A typed pattern match with a singleton type p.type now tests whether the selector value is reference-equal to p . For example:
val p = List(1, 2, 3) val q = List(1, 2) val r = q r match { case _: p.type => Console.println("p") case _: q.type => Console.println("q") }
This will match the second case and hence will print "q". Before, the singleton types were erased to List, and therefore the first case would have matched, which is non-sensical.
It is now possible to write multi-line string-literals enclosed in triple quotes . Example
"""this is a
multi-line
string literal"""
No escape substitutions except for unicode escapes are performed in such string literals.
There is a new syntax for class literals : For any class type C, classOf[C] designates the run-time representation of C.
Scala in its second version is different in some details from the first version of the language. There have been several additions and some old idioms are no longer supported. This section summarizes the main changes.
The following three words are now reserved; they cannot be used as identifiers
implicit match requires
Newlines can now be used as statement separators in place of semicolons
There are some other situations where old constructs no longer work:
Pattern matching expressions The match keyword now appears only as infix operator between a selector expression and a number of cases, as in:
expr match { case Some(x) => ... case None => ... }
Variants such as expr.match {...} or just match {...} are no longer supported.
"With" in extends clauses. The idiom
class C with M { ... }
is no longer supported. A with connective is only allowed following an extends clause. For instance, the line above would have to be written
class C extends AnyRef with M { ... }
However, assuming M is a trait , it is also legal to write
class C extends M { ... }
The latter expression is treated as equivalent to
class C extends S with M { ... }
where S is the superclass of M.
Regular Expression Patterns The only form of regular expression pattern that is currently supported is a sequence pattern, which might end in a sequence wildcard _*. Example:
case List(1, 2, _*) => ... // will match all lists starting with 1,2
It is at current not clear whether this is a permanent restriction. We are evaluating the possibility of re-introducing full regular expression patterns in Scala.
The recommended syntax of selftype annotations has changed.
class C: T extends B { ... }
becomes
class C requires T extends B { ... }
That is, selftypes are now indicated by the new requires keyword. The old syntax is still available but is considered deprecated.
For-comprehensions now admit value and pattern definitions. For example:
for { val x <- List.range(1, 100) val y <- List.range(1, x) val z = x + y isPrime(z) } yield Pair(x, y)
Note the definition val z = x + y as the third item in the for-comprehension.
The rules for implicit conversions of methods to functions have been tightened. Previously, a parameterized method used as a value was always implicitly converted to a function. This could lead to unexpected results when method arguments where forgotten. Consider for instance the statement below:
show(x.toString)
where show is defined as follows:
def show(x: String) = Console.println(x)
Most likely, the programmer forgot to supply an empty argument list () to toString. The previous Scala version would treat this code as a partially applied method, and expand it to:
show(() => x.toString())
As a result, the address of a closure would be printed instead of the value of s.
Scala version 2.0 will apply a conversion from partially applied method to function value only if the expected type of the expression is indeed a function type. For instance, the conversion would not be applied in the code above because the expected type of show's parameter is String, not a function type.
The new convention disallows some previously legal code. Example:
def sum(f: int => double)(a: int, b: int): double = if (a > b) 0 else f(a) + sum(f)(a + 1, b) val sumInts = sum(x => x) // error: missing arguments
The partial application of sum in the last line of the code above will not be converted to a function type. Instead, the compiler will produce an error message which states that arguments for method sum are missing. The problem can be fixed by providing an expected type for the partial application, for instance by annotating the definition of sumInts with its type:
val sumInts: (int, int) => double = sum(x => x) // OK
On the other hand, Scala version 2.0 now automatically applies methods with empty parameter lists to () argument lists when necessary. For instance, the show expression above will now be expanded to
show(x.toString())
Scala version 2.0 also relaxes the rules of overriding with respect to empty parameter lists. The revised definition of matching members makes it now possible to override a method with an explicit, but empty parameter list () with a parameterless method, and vice versa. For instance, the following class definition is now legal:
class C { override def toString: String = //... }
Previously this definition would have been rejected, because the toString method as inherited from java.lang.Object takes an empty parameter list.
A class parameter may now be prefixed by val or var .
Previously, Scala had three levels of visibility: private, protected and public. There was no way to restrict accesses to members of the current package, as in Java. Scala 2 now defines access qualifiers that let one express this level of visibility, among others. In the definition
private[C] def f(...)
access to f is restricted to all code within the class or package C (which must contain the definition of f)
The model which details mixin composition of classes has changed significantly. The main differences are:
The new mixin model is explained in more detail in the Scala Language Specification [47].
Views in Scala 1.0 have been replaced by the more general concept of implicit parameters
The new version of Scala implements more flexible typing rules when it comes to pattern matching over heterogeneous class hierarchies . A heterogeneous class hierarchy is one where subclasses inherit a common superclass with different parameter types. With the new rules in Scala version 2.0 one can perform pattern matches over such hierarchies with more precise typings that keep track of the information gained by comparing the types of a selector and a matching pattern . This gives Scala capabilities analogous to guarded algebraic data types.
Links:
[1] http://www.scala-lang.org/node/43#2.10.0
[2] http://www.scala-lang.org/node/43#2.9.0
[3] http://www.scala-lang.org/node/43#2.8.0
[4] http://www.scala-lang.org/node/43#2.7.2
[5] http://www.scala-lang.org/node/43#2.7.1
[6] http://www.scala-lang.org/node/43#2.7.0
[7] http://www.scala-lang.org/node/43#2.6.1
[8] http://www.scala-lang.org/node/43#2.6.0
[9] http://www.scala-lang.org/node/43#2.5.0
[10] http://www.scala-lang.org/node/43#2.4.0
[11] http://www.scala-lang.org/node/43#2.3.2
[12] http://www.scala-lang.org/node/43#2.3.0
[13] http://www.scala-lang.org/node/43#2.1.8
[14] http://www.scala-lang.org/node/43#2.1.7
[15] http://www.scala-lang.org/node/43#2.1.5
[16] http://www.scala-lang.org/node/43#2.0
[17] http://docs.scala-lang.org/overviews/core/value-classes.html
[18] http://docs.scala-lang.org/sips/pending/implicit-classes.html
[19] http://docs.scala-lang.org/overviews/core/string-interpolation.html
[20] http://docs.scala-lang.org/overviews/core/futures.html
[21] http://docs.scala-lang.org/sips/pending/type-dynamic.html
[22] https://issues.scala-lang.org/browse/SI-1133
[23] http://docs.scala-lang.org/sips/pending/modularizing-language-features.html
[24] http://docs.scala-lang.org/overviews/parallel-collections/overview.html
[25] http://docs.scala-lang.org/actors-migration/
[26] https://docs.google.com/document/d/1Z1VhhNPplbUpaZPIYdc0_EUv5RiGQ2X4oqp0i-vz1qw/edit#heading=h.pqwdkl
[27] http://docs.scala-lang.org/overviews/macros/overview.html
[28] http://www.scala-lang.org/sid/3
[29] http://www.scala-lang.org/sid/7
[30] http://www.scala-lang.org/sid/9
[31] http://www.scala-lang.org/sid/1
[32] http://www.scala-lang.org/sid/8
[33] http://www.scala-lang.org/sid/5
[34] http://lampsvn.epfl.ch/trac/scala/wiki/Scaladoc/
[35] http://www.scala-ide.org
[36] http://en.wikipedia.org/wiki/Generics_in_Java
[37] http://www.scala-lang.org/docu/files/tools/scalac.html#
[38] http://www.scala-lang.org/docu/files/api/scala/Predef$object.html#
[39] http://www.scala-lang.org/?q=node/43#2.3.0
[40] http://www.scala-lang.org/docu/files/api/scala/Math$object.html#
[41] http://www.scala-lang.org/docu/files/api/scala/testing/SUnit$object.html#
[42] http://www.scala-lang.org/docu/files/api/scala/testing/SUnit$object.Assert.html#
[43] http://www.scala-lang.org/docu/files/api/scala/util/DynamicVariable.html#
[44] http://www.scala-lang.org/downloads/distrib/files/nightly/scala/doc/scala/api/scala/ClassfileAnnotation.html#
[45] http://www.scala-lang.org/downloads/distrib/files/nightly/scala/doc/scala/api/scala/StaticAnnotation.html#
[46] http://www.scala-lang.org/docu/files/MatchingObjectsWithPatterns-TR.pdf#
[47] http://www.scala-lang.org/?q=node/35