Scala primer

This Scala primer is not intended as a study material for someone new to Scala but as a refresher for someone who hasn't worked on the technology for a while.

  • If the type parameter starts with a lowercase then it means the type parameter is unknown. it is semantically equal to a type parameter as underscore. ie ClassName[t] == ClassName[_]. use this if you want to use the classname without specifying the type parameter.

  • == operator (which is a method) simply calls equals which is logical equality. there is a separate method called eq which looks for object identity.

  • empty parantheses () is a synonym for Unit data type. Unlike other languages like Python empty data is not Falsy. ie "" != false

  • Tuple is not a collection in Scala and they can no more then 22 elements. And since it is not a collection it is not iterable.

  • { expression block } can contain a series of expression and it returns the last expression. It can be used inlieu of a single expression. expression blocks can be nested too.

  • for an if statement without the else counterpart, on failure condition the return type is Unit (descendant of AnyVal). This explains the below.
scala> val a = if (true) { 10 } // LHS is AnyVal and not Int
a: AnyVal = 10

scala> val a = if (true) { "" } // LHS is Any and not String
a: Any = "" 

  • case match supports multiple patterns with Pipe operator.
scala> val kind = day match {
 case "MON" | "TUE" | "WED" | "THU" | "FRI" =>
 "weekday"
 case "SAT" | "SUN" =>
 "weekend"
 }
kind: String = weekday

  • procedure is a function that has a return type of Unit. which means it a fully side-effecting function. A function definition without the equal sign means it is a procedure and would ignore the last expression and always return Unit data type.

  • if a function is defined without paramaters/parentheses then it has to be called without parentheses. i.e. def test = { } has to be called test and test() is not allowed. in these cases if you want to access the function object without invoking the function then you will have to do something like val a = greet _. Here you are creating a partially applied function for greet. Another alternative here is to explicitly set the type of val a. ie. val a:Int=>String = greet.similarly consider this curried function def add(a:Int)(b:Int). now to create a partially applied function you would have make the call as something like this val a = add(10)(_).

  • @annotation.tailrec use this annotation to ensure tail recursion in a function. If compiler cannot enforce tail recursion optimization then a compile time failure is launched.

  • varargs parameter are supported. eg: def test(a:Int*).

  • Even type parameter of a function is optional if it can be inferred. for eg: def identity[A](a:A) = a can be called as val a = identity(“string”). here both the type parameter of the function and as well as the return type of val a are both inferred.

  • The infix or operator notation famous for one argument object methods like a + b = a.+(b) can also be used for methods with two parameters like this a + (b,c) = a.+(b,c)

  • A function with single parameter can be expressed with type Int => Int. But function with no parameters require an empty parentheses () => Int.

  • function definition and literal forms
 // one parameter example
      def act(s:String,f:String=>String) = f(s)
      act("Hello",(a:String) => a.reverse) full form
      act("Hello",a => a.reverse) shorter type inferred form
      act("Hello",_.reverse) placeholder form
 // two parameter example
     def act(s1:String,s2:String,f:(String,String)=>String) = f(s1,s2)
     act("Hello","World",(a:String,b:String) => a+b) 
     act("Hello","World",(a,b) => a+b) 
     act("Hello","World",_+_) 

  • Partial applied functions can be created with placeholder syntax. for eg
def add(a:Int,b:Long) = a+b
val p1 = add(10,_:Int)
p1(20)
val p2 = add(20,_:Long)
p2(30)

Here the palceholder parameters requires explicit type since there is chance of ambiguity due to overloaded method. ie we must unambigously select add(Int,Int) instead of a potential overloaded method add(Int,String)


  • By-name function parameters are denoted with an additional arrow like def func1(a: => Int) = println(a). func1 is not a higher order function that takes function value as argument that returns int as the function definition seems to suggest. The argument can either be a function that returns an Int or can be a direct value Int. if it is a function then the function is evaluated (ie called) everytime when the function is referenced in the func1 method body.
    def func1(a: => Int) = println(a)
    val a = () => 10
    val b = (a:Int) => a
    func1(a()) # prints 10
    func1(b(10)) # prints 10
    func1(10) # prints 10
    

  • Partially applied functions and Partial functions are two different things in Scala. former is a function which has currently defined only a partial list of parameters and yet to define the other parameters. The latter is a function which has restrictions on the values it can accept for the defined parameters. A Partial Function object must have two methods apply and isDefinedAt defined.

  • a case block statement without a match is a partial function literal. { case i: Int if i != 0 => i} is equivalent to new PartialFunction[Int,Int] { def apply(a:Int) = a; def isDefinedAt(a:Int) = a!=0 }

  • Nil is singleton instance of List[Nothing]. And since Nothing is a subtype of every other type Nil is an empty list of any type

  • apply,unapply and update are special methods in scala. if class called Test had these methods defined on them then

    apply => called on Test()

    unapply => called when used with pattern matching

    update => called on assignment Test(10) = “Some Value”


  • The Scala collections hierarchy as shown below.

Class Hierarchy

Comments