Accumulators as counters

TextAnalyser.scala

class TextAnalyser(val sc: SparkContext, ...) {
  ...
  // Instance variables
  val _totalChars = sc.accumulator(0, "Total Characters")
  val _totalWords = sc.accumulator(0, "Total Words")
  ...
  // In a worker thread
  def analyse(rdd: RDD[String]): TextStats = {
    ...
    // This limits serialization to the reference variables only
    val totalChars = _totalChars
    val totalWords = _totalWords
    ...
    // Accumulators can be safely used in parallel computations
    .map(x => {totalWords += 1; x})
    ...