1. Home
2. Working with Play
3. Play for Java developers
4. Main concepts
5. Asynchronous HTTP programming

§Streaming HTTP responses

§Standard responses and Content-Length header

Since HTTP 1.1, to keep a single connection open to serve several HTTP requests and responses, the server must send the appropriate Content-Length HTTP header along with the response.

By default, when you send a simple result, such as:

public Result index() {
    return ok("Hello World");
}

You are not specifying a Content-Length header. Of course, because the content you are sending is well known, Play is able to compute the content size for you and to generate the appropriate header.

Note: for text-based content this is not as simple as it looks, since the Content-Length header must be computed according the encoding used to translate characters to bytes.

To be able to compute the Content-Length header properly, Play must consume the whole response data and load its content into memory.

§Serving files

If it’s not a problem to load the whole content into memory for simple content what about a large data set? Let’s say we want to send back a large file to the web client.

Play provides easy to use helpers to this common task of serving a local file:

public Result index() {
    return ok(new java.io.File("/tmp/fileToServe.pdf"));
}

Additionally this helper will also compute the Content-Type header from the file name. And it will also add the Content-Disposition header to specify how the web browser should handle this response. The default is to ask the web browser to download this file by using Content-Disposition: attachment; filename=fileToServe.pdf.

§Chunked responses

For now, this works well with streaming file content, since we are able to compute the content length before streaming it. But what about dynamically-computed content with no content size available?

For this kind of response we have to use Chunked transfer encoding.

Chunked transfer encoding is a data transfer mechanism in version HTTP 1.1 in which a web server serves content in a series of chunks. This uses the Transfer-Encoding HTTP response header instead of the Content-Length header, which the protocol would otherwise require. Because the Content-Length header is not used, the server does not need to know the length of the content before it starts transmitting a response to the client (usually a web browser). Web servers can begin transmitting responses with dynamically-generated content before knowing the total size of that content.

The size of each chunk is sent right before the chunk itself so that a client can tell when it has finished receiving data for that chunk. The data transfer is terminated by a final chunk of length zero.

https://en.wikipedia.org/wiki/Chunked_transfer_encoding

The advantage is that we can serve data live, meaning that we send chunks of data as soon as they are available. The drawback is that since the web browser doesn’t know the content size, it is not able to display a proper download progress bar.

Let’s say that we have a service somewhere that provides a dynamic InputStream that computes some data. We can ask Play to stream this content directly using a chunked response:

public Result index() {
    InputStream is = getDynamicStreamSomewhere();
    return ok(is);
}

You can also set up your own chunked response builder:

public Result index() {
    // Prepare a chunked text stream
    Source<ByteString, ?> source = Source.<ByteString>actorRef(256, OverflowStrategy.dropNew())
        .mapMaterializedValue(sourceActor -> {
            sourceActor.tell(ByteString.fromString("kiki"), null);
            sourceActor.tell(ByteString.fromString("foo"), null);
            sourceActor.tell(ByteString.fromString("bar"), null);
            sourceActor.tell(new Status.Success(NotUsed.getInstance()), null);
            return NotUsed.getInstance();
        });
    // Serves this stream with 200 OK
    return ok().chunked(source);
}

The method Source.actorRef creates an Akka Streams Source that materializes to an ActorRef. You can then publish elements to the stream by sending messages to the actor. An alternative approach is to create an actor that extends ActorPublisher and use the Stream.actorPublisher method to create it.

We can inspect the HTTP response sent by the server:

HTTP/1.1 200 OK
Content-Type: text/plain; charset=utf-8
Transfer-Encoding: chunked

4
kiki
3
foo
3
bar
0

We get three chunks and one final empty chunk that closes the response.

For more information on using Akka Streams, you can reference the Akka Streams documentation.

Next: Comet