This answer applies to Play 2.5.x and higher since it uses the Akka streams API that replaced Play's Iteratee-based streaming in that version.
Basically, you can create a body parser that returns a Source[T]
that you can pass to Ok.chunked(...)
. One way to do this is to use Accumulator.source[T]
in the body parser. For example, an action that just returned data sent to it verbatim might look like this:
def verbatimBodyParser: BodyParser[Source[ByteString, _]] = BodyParser { _ =>
// Return the source directly. We need to return
// an Accumulator[Either[Result, T]], so if we were
// handling any errors we could map to something like
// a Left(BadRequest("error")). Since we're not
// we just wrap the source in a Right(...)
Accumulator.source[ByteString]
.map(Right.apply)
}
def stream = Action(verbatimBodyParser) { implicit request =>
Ok.chunked(request.body)
}
If you want to do something like transform a TSV file you can use a Flow
to transform the source, e.g:
val tsvToCsv: BodyParser[Source[ByteString, _]] = BodyParser { req =>
val transformFlow: Flow[ByteString, ByteString, NotUsed] = Flow[ByteString]
// Chunk incoming bytes by newlines, truncating them if the lines
// are longer than 1000 bytes...
.via(Framing.delimiter(ByteString("\n"), 1000, allowTruncation = true))
// Replace tabs by commas. This is just a silly example and
// you could obviously do something more clever here...
.map(s => ByteString(s.utf8String.split('\t').mkString(",") + "\n"))
Accumulator.source[ByteString]
.map(_.via(transformFlow))
.map(Right.apply)
}
def convert = Action(tsvToCsv) { implicit request =>
Ok.chunked(request.body).as("text/csv")
}
There may be more inspiration in the Directing the Body Elsewhere section of the Play docs.