Akka streams reduce to smaller stream

Question

I have an ordered stream of data

A A A B B C C C C ... (very long)

And I want to transform it to a stream of aggregates in the form (item, count):

(A, 3) (B, 2) (C, 4)

What operators could I use in Akka Streams for this?

Source.fromPublisher(publisher)
    .aggregateSomehow()  // ?
    .runWith(sink)

I've looked into .groupBy but it requires that I know the number of categories in advance which I don't. Also I believe it will keep all groups in memory which I'd like to avoid. I should be able to discard (A, 3) after it has been processed and free up resources it consumes.

Edit: This question ask for similar functionality but using SubFlows. However using SubFlows doesn't seem to be required because I have a solution using the statefulMapConcat combinator.

Answer 1

One option is to use the statefulMapConcat combinator:

Source(List("A", "A", "B", "B", "B", "C", "C", ""))
      .statefulMapConcat({ () =>
        var lastChar = ""
        var count = 0

        char => if(lastChar == char) {
            count += 1
            List.empty
          } else {
            val charCount = (lastChar, count)
            lastChar = char
            count = 1
            List(charCount)
          }
      })
    .runForeach(println)

However that required appending an element to the input stream to mark the end.

Output:

(,0)
(A,2)
(B,3)
(C,2)

Thanks to @chunjef for the suggestion in comments