CODEWITHSUNDEEP
design-patternsscalaaggregate-functions
yura
yura

Reputation: 14655

How to write Aggregate pattern in Scala?

Suppose I have Iterator[A](size is infinite) and I want to get Iterator[B] from it where some subsequent values of type A are aggregated.

Example: I have list of strings:

Iterator(
    "START",
    "DATA1",
    "DATA2",
    "DATA3",
    "START",
    "DATA1",
    "DATA2",
    //.. 10^10 more records
)

I want to join strings from START to NEXT START excluding. I.e. write parser.

Iterator(
"START DATA1 DATA2 DATA3",
"START DATA1 DATA2",
    //.. 10^10 / 5 more records
)

I know how to do this imperatively, but I want to accomplish it with scala higher order functions. Any ideas?

PS EIP Aggregate http://camel.apache.org/aggregator2.html.

Upvotes: 3

Views: 603

Answers (4)

Derek Wyatt
Derek Wyatt

Reputation: 2717

Well, an infinite stream changes things rather dramatically. Assuming I understand the rest of your situation, this should work:

def aggregate(it: Iterator[String]) = new Iterator[String] {
  if (it.hasNext) it.next
  def hasNext = it.hasNext
  def next = "START " + (it.takeWhile(_ != "START")).mkString(" ")
}

So that you can:

val i = aggregate(yourStream.iterator)
i.take(20).foreach(println) // or whatever

Upvotes: 5

ron
ron

Reputation: 9374

With Streams:

object Iter {
  def main(args: Array[String]) {
    val es = List("START", "DATA1", "DATA2", "START", "DATA1", "START")
    val bit = batched(es.iterator, "START")
    println(bit.head.toList)
    println(bit.tail.head.toList)
  }

  def batched[T](it: Iterator[T], start: T) = { 
    def nextBatch(): Stream[List[T]] = { 
      (it takeWhile { _ != start }).toList match {
        case Nil => nextBatch()
        case es => Stream.cons(start :: es, nextBatch())
      }
    }
    nextBatch()
  }

}

Upvotes: 0

paradigmatic
paradigmatic

Reputation: 40461

If you want a functional solution, you should use Streams rather than iterators (streams are immutable). Here's one possible approach:

def aggregate(strs: Stream[String] ) = { 
  aggregateRec( strs )
}

def aggregateRec( strs: Stream[String] ): Stream[String] = {
  val tail = strs.drop(1)
  if( tail.nonEmpty ) {
    val (str, rest ) = accumulate( tail )
    Stream.cons( str, aggregateRec( rest ) )
  }
  else Stream.empty
}

def accumulate( strs: Stream[String] ): (String, Stream[String])  = {
  val first = "START " + strs.takeWhile( _ != "START").mkString(" ")
  val rest = strs.dropWhile( _ != "START" )
  ( first, rest )
}

It works as expected:

val strs = Stream( "START", "1", "2", "3", "START", "A", "B" )
val strs2 = aggregate( strs )
strs2 foreach println

Upvotes: 6

Jean-Philippe Pellet
Jean-Philippe Pellet

Reputation: 60006

You could try it with a fold:

val ls = List(
  "START",
  "DATA1",
  "DATA2",
  "DATA3",
  "START",
  "DATA1",
  "DATA2"
)

(List[List[String]]() /: ls) { (acc, elem) =>
  if (elem == "START")
    List(elem) :: acc // new head list
  else
    (elem :: acc.head) :: acc.tail // prepend to current head list
} map (_.reverse mkString " ") reverse;

Upvotes: 1

Related Questions