Rocky Warren

Scala: The Good Parts

Scala allows you to accomplish tasks in different ways. I recently gave a talk explaining some of the more readable, maintainable approaches while attempting to add detail so the slides could stand on their own. The excellent reveal.js framework lets you create presentations in Markdown. This blog is also in Markdown, so through the power of copy/paste, here it is.

Scala Overview, Benefits Over Java

  • Created by Martin Odersky 17 years ago

  • Runs on the JVM

  • Can create Java objects, call their methods, and inherit from them in Scala and vice versa

  • Java collection interoperability

    import scala.jdk.CollectionConverters._

val javaList = java.util.Arrays.asList("Hi")
val scalaList = javaList.asScala
scalaList.foreach(println(_)) // Use any Scala collection methods
val backToJava = scalaList.asJava

  • Type inference

    List<String> msgs = Arrays.asList("Hi"); // This Java code...

    val msgs = List("Hi") // ..becomes this Scala code

  • Concise

    // This Java code...
public class Person {
  private String name;
  public String getName() { return name; }
  public String toString() { ... }
  public boolean equals() { ... }
  public int hashCode() { ... }
}
Map<String, Integer> map  = new HashMap<String, Integer>() {{
  put("a", 0);
  put("b", 2);
}};

    // ...becomes this Scala code
case class Person(name: String)
val map = Map("a" -> 0, "b" -> 1)

  • Concurrency

    • Immutability by default enables safe concurrency and parallelism
    • Parallel collections, concurrent distribution with actors, asynchrony with futures

  • Traits

    • Java interfaces with optional behavior, allows for safe multiple-inheritance

  • Pattern Matching

    • More powerful switch, match against class hierarchies, sequences, and more

  • Higher-order Functions

    • Use functions anywhere and pass them to anything with type safety

  • Option instead of null, no more NullPointerException!

  • Functional paradigms and excellent collection methods

    • map, flatMap, filter, head, headOption, isEmpty, foreach, contains, find, zip, zipWithIndex, reduce, fold, sum, count, groupBy, mkString, sortBy

  • Automatically returns last statement without explicit return, no semicolons

    def doubleIt(n: Int) = n * 2 // Returns `n * 2`, infers return Int type

  • if statements are expressions, can have results

    // Java's ternary (`int x = condition ? result : defaultVal`) in Scala
val x = if (condition) result else defaultVal

  • Powerful imports

    // Rename import
import java.util.{HashMap => JavaHashMap}
// Ignore HashMap, import everything else
import java.util.{HashMap => _, _}

// Follows scoping rules
def someFn() = {
    import cats.implicits.toShow
    ??? // Aside: This is valid Scala, throws NotImplementedError if hit
}

Declaring variables

val items = Seq("1", "2") // Type inferred
val typed: Seq[String] = Seq("1", "2") // Can optionally provide type
var mutable = 5 // Int, prefer val over var
var small: Short = 5 // Short, need type or inferred as Int
val big = 5L // Long, need L or inferred as Int
val bigger = 5.0F // Float, need F or inferred as Double
val biggest = 5.0 // Double
val javaType = new java.math.BigDecimal(5) // Java interop
val byte: Byte = 0xa // Byte using hex notation, need type or inferred as Int
val char = 'D' // Char
val nothing = () // Unit, similar to void in Java
lazy val deferred = ??? // Initialization deferred until first access

Classes

class Counter {
  private val myValue = 2 // No accessors generated
  val value = 1 // Only getter generated
  var mutableValue = 0 // Getter and setter generated, prefer val to var
}

class Person(val name: String) { // Primary constructor
  // Constructor body, initialize things here

  def this() { // Auxiliary constructor
    this("unnamed") // Must call primary constructor
  }
}

Traits

  • Like Java interfaces, but can have implementation
  • Only difference between Scala class: cannot have constructor parameters
// Optionally sealed, barring extension outside file, useful for enums
trait Logger {
  def log(msg: String): Unit // Abstract method
  def info(msg: String): Unit = println(msg) // Implementation provided
}

// Mix as many traits into class as you like, constructed left to right
class Logged(name: String) extends Person(name) with Logger {
  // Must implement abstract methods
 override def log(msg: String): Unit = ???
 def logName(): Unit = info(name) // Using implementation in trait
}

Objects

  • Use for Singletons or home for misc values/functions
  • Can extend classes or traits, cannot have constructor parameters
  • Commonly used as "companion object" to classes for static functions
object Accounts { // Singleton
  private var lastNum = 0
  def uniqueNum(): Int = {
    lastNum += 1;
    lastNum
  }
}

class Person private(val name: String)
// Companion object
object Person {
  // apply is special, called as `val p = Person("Rocky")`
  def apply(name: String): Person = new Person(name)
  // Called as Person.staticFn() as in Java
  def staticFn() = ???
}

Trait Initialization Order Gotcha

trait MyTrait {
 // scalafix error: abstract val in trait
 // Use def, lazy val, or move to object instead
 val foo: Int // Defaults to 0
 val bar = foo * 2 // Initialized prior to foo getting set to 20 by MyObject
}

object MyObject extends MyTrait {
 val foo = 20
}

MyObject.bar // 0 instead of 40 as expected!

Case Classes

  • Special kind of class, immutable with implementations for
    • toString, equals, hashCode
    • apply, making new unnecessary on creation
    • unapply for pattern matching
    • copy for immutable modifications
case class Person(name: String, age: Option[Int]) // No new required
val r = Person("Rocky", None) // Calls apply
println(r) // Person(Rocky,None)
val b = r.copy(name = "Bob") // New object, r unchanged
println(r == b) // false
println(r == r.copy()) // true
Scala's class hierarchy
Scala's class hierarchy

Option

  • A much better null
  • Calling get if value is None results in exception, no better than NullPointerException
  • Instead, handle both Some and None cases explicitly
val maybeName: Option[String] = Some("Rocky")
maybeName match { // Good, but verbose
  case Some(n) => println(n.toUppercase)
  case None => println("N/a")
}
println(maybeName.fold("N/a")(_.toUpperCase)) // Better, but can be unclear
println(maybeName.map(_.toUpperCase).getOrElse("N/a")) // Best for clarity

// Wrap potentially null values in Option
val maybeVal: Option[Int] = Option(javaMethodThatMayReturnNull())

// Conditionally set val with Option.when
val maybeId: Option[UUID] = Option.when(req.hasId)(req.getId)

Try

try someFn() catch { // Standard try with pattern matching catch
  case ex: IOException => ???
  case t: Throwable => ???
}

// Try type can be chained
val res: Try[Int] = Try(someFn())
  .flatMap(d => Try(someFn()) // Without flatMap, res would be Try[Try[Int]]
    .map(d / _)
  )
res match { // Good, but verbose
 case Success(a) => println(a)
 case Failure(ex) => println(ex.getMessage)
}
res // Better, but can be unclear
 .fold(ex => println(ex.getMessage), println(_))
res // Best for clarity and to pattern match
 .map(println(_))
 .recover { case ex: IOException => println(ex.getMessage) }

// To only need to handle failed case
res.failed.map(ex => println(ex.getMessage))

Either

// Return either an error or Widget
def getById(id: UUID): Future[Either[NotFoundError, Widget]] =
  database
    .run("...")
    .map(_.toRight(NotFoundError()))

// Handling the Either at service boundary
implicit class FromEither[T](future: Future[Either[NotFoundError, T]]) {
  def completeWith(obj: T) =
    future.onComplete { // With pattern matching
      case Success(Right(res))  => // Successful future with Widget
        res
      case Success(Left(error)) => // Successful future with Error
        throw fromServiceEx(error)
      case Failure(error)       => // Failed future
        throw error
    }
  // Same as above using map/fold
  def completeWith(obj: T) =
   future
    .map(_.fold(
      error => throw fromServiceEx(error)), // Successful future with Error
      res => res) // Successful future with Widget
    )
    .recover { case error => throw error } // Failed future
}
Scala immutable collections hierarchy
Scala immutable collections hierarchy

  • Prefer immutable, they're safer and still performant

  • Map: key/value pairs, similar to Java's HashMap

    val myMap = Map("a" -> 0, "b" -> 1)
myMap("b") // 1
// Looping over map, also shows string interpolation
myMap.foreach{ case (k, v) => println(s"$k -> $v") }

  • Tuple: aggregates of values, useful for multiple returns in class's private methods. Prefer case class return types for public methods for readability.

    val myTuple = (1, 3.14, "Fred")
myTuple._2 // 3.14

  • Seq: general purpose, similar to Java's List

    val mySeq = Seq(3, 2, 1)
mySeq
  .filter(_ > 1)
  .sorted
  .zipWithIndex
  .foreach { case (v, idx) =>  print(s"$v at $idx ") }
  // 2 at 0 3 at 1

  • Set: distinct elements, similar to Java's HashSet

  • IndexedSeq: fast random access and count, similar to Java's Array

  • List: fast beginning and end operations, similar to Java's LinkedList

Pattern Matching

  • Preferable to casting via asInstanceOf, isInstanceOf
obj match { // Type matching
  case x: Int | Long if x < 10 => x // Multi-match with if guards
  case s: String => Integer.parseInt(s) // No cast needed, s is now String
  case _ => 0 // Default case, often required to avoid MatchError
}

Seq(0, 1, 2) match { // Collection matching
  case Seq(x, y) => s"$x $y" // List equivalent, case x :: y :: Nil
  case Seq(0, _*) => "0 ..." // List equivalent, case 0 :: tail
  case _ => "Not matched"
}

person match { // Case class/Option matching
  case Person("Rocky", age) => println(s"Rocky is $age")
  case Person(name, Some(_)) => println(name)
  case _ => println("Not matched")
}

Additional Reading

Scala for the Impatient - Cay S. Horstmann
Scala for the Impatient - Cay S. Horstmann

For more on learning Scala, check out my Scala Learning Resources.

Stay up to date

Get notified when I publish. Unsubscribe at any time.