Play 2 Scala - Best way to upload a big CSV file with Iteratee in order to process each line reactively

Question

I want to use Play2 to upload a very big CSV files (milions of lines) on elasticsearch. I have written the following code that works fine.

I'm not pleased with the way I skip the http response header in the first chunk. There should be a way to chain this iteratee with a first Iteratee that skip the http header and directly switch to the Done state, but I haven't found how.

If anyone can help

object ReactiveFileUpload extends Controller {
  def upload = Action(BodyParser(rh => new CsvIteratee(isFirst = true))) {
    request =>
      Ok("File Processed")
  }
}

case class CsvIteratee(state: Symbol = 'Cont, input: Input[Array[Byte]] = Empty, lastChunk: String = "", isFirst: Boolean = false) extends Iteratee[Array[Byte], Either[Result, String]] {
  def fold[B](
               done: (Either[Result, String], Input[Array[Byte]]) => Promise[B],
               cont: (Input[Array[Byte]] => Iteratee[Array[Byte], Either[Result, String]]) => Promise[B],
               error: (String, Input[Array[Byte]]) => Promise[B]
               ): Promise[B] = state match {
    case 'Done =>
      done(Right(lastChunk), Input.Empty)

    case 'Cont => cont(in => in match {
      case in: El[Array[Byte]] => {
        // Retrieve the part that has not been processed in the previous chunk and copy it in front of the current chunk
        val content = lastChunk + new String(in.e)
        val csvBody =
          if (isFirst)
            // Skip http header if it is the first chunk
            content.drop(content.indexOf("\r\n\r\n") + 4)
          else content
        val csv = new CSVReader(new StringReader(csvBody), ';')
        val lines = csv.readAll
        // Process all lines excepted the last one since it is cut by the chunk
        for (line <- lines.init)
          processLine(line)
        // Put forward the part that has not been processed
        val last = lines.last.toList.mkString(";")
        copy(input = in, lastChunk = last, isFirst = false)
      }
      case Empty => copy(input = in, isFirst = false)
      case EOF => copy(state = 'Done, input = in, isFirst = false)
      case _ => copy(state = 'Error, input = in, isFirst = false)
    })

    case _ =>
      error("Unexpected state", input)

  }

  def processLine(line: Array[String]) = WS.url("http://localhost:9200/affa/na/").post(
    toJson(
      Map(
        "date" -> toJson(line(0)),
        "trig" -> toJson(line(1)),
        "code" -> toJson(line(2)),
        "nbjours" -> toJson(line(3).toDouble)
      )
    )
  )
}

Answer 1

To chain two iteratees together, use flatMap.

val combinedIteratee = firstIteratee.flatMap(firstResult => secondIteratee)

Or, using a for comprehension:

val combinedIteratee = for {
  firstResult <- firstIteratee
  secondResult <- secondIteratee
} yield secondResult

You can use flatMap to sequence as many iteratees together as you like.

You might want to do something like:

val headerAndCsvIteratee = for {
  headerResult <- headerIteratee
  csvResult <- csvIteratee
} yield csvResult