Scala LanguageStreams

Remarks

Streams are lazily-evaluated, meaning they can be used to implement generators, which will provide or 'generate' a new item of the specified type on-demand, rather than before the fact. This ensures only the computations necessary are done.

Using a Stream to Generate a Random Sequence

genRandom creates a stream of random numbers that has a one in four chance of terminating each time it's called.

def genRandom: Stream[String] = {
  val random = scala.util.Random.nextFloat()
  println(s"Random value is: $random")
  if (random < 0.25) {
    Stream.empty[String]
  } else {
    ("%.3f : A random number" format random) #:: genRandom
  }
}

lazy val randos = genRandom  // getRandom is lazily evaluated as randos is iterated through

for {
  x <- randos
} println(x) // The number of times this prints is effectively randomized.

Note the #:: construct, which lazily recurses: because it is prepending the current random number to a stream, it does not evaluate the remainder of the stream until it is iterated through.

Infinite Streams via Recursion

Streams can be built that reference themselves and thus become infinitely recursive.

// factorial
val fact: Stream[BigInt] = 1 #:: fact.zipWithIndex.map{case (p,x)=>p*(x+1)}
fact.take(10)  // (1, 1, 2, 6, 24, 120, 720, 5040, 40320, 362880)
fact(24)       // 620448401733239439360000

// the Fibonacci series
val fib: Stream[BigInt] = 0 #:: fib.scan(1:BigInt)(_+_)
fib.take(10)  // (0, 1, 1, 2, 3, 5, 8, 13, 21, 34)
fib(124)      // 36726740705505779255899443

// random Ints between 10 and 99 (inclusive)
def rndInt: Stream[Int] = (util.Random.nextInt(90)+10) #:: rndInt
rndInt.take(10)  // (20, 95, 14, 44, 42, 78, 85, 24, 99, 85)

In this context the difference between Var, Val, and Def is interesting. As a def each element is recalculated every time it is referenced. As a val each element is retained and reused after it's been calculated. This can be demonstrated by creating a side-effect with each calculation.

// def with extra output per calculation
def fact: Stream[Int] = 1 #:: fact.zipWithIndex.map{case (p,x)=>print("!");p*(x+1)}
fact(5)  // !!!!!!!!!!!!!!! 120
fact(4)  // !!!!!!!!!! 24
fact(7)  // !!!!!!!!!!!!!!!!!!!!!!!!!!!! 5040

// now as val
val fact: Stream[Int] = 1 #:: fact.zipWithIndex.map{case (p,x)=>print("!");p*(x+1)}
fact(5)  // !!!!! 120
fact(4)  // 24
fact(7)  // !! 5040

This also explains why the random number Stream doesn't work as a val.

val rndInt: Stream[Int] = (util.Random.nextInt(90)+10) #:: rndInt
rndInt.take(5)  // (79, 79, 79, 79, 79)

Infinite self-referent stream

// Generate stream that references itself in its evaluation
lazy val primes: Stream[Int] =
  2 #:: Stream.from(3, 2)
    .filter { i => primes.takeWhile(p => p * p <= i).forall(i % _ != 0) }
    .takeWhile(_ > 0) // prevent overflowing

// Get list of 10 primes
assert(primes.take(10).toList == List(2, 3, 5, 7, 11, 13, 17, 19, 23, 29))

// Previously calculated values were memoized, as shown by toString
assert(primes.toString == "Stream(2, 3, 5, 7, 11, 13, 17, 19, 23, 29, ?)")