streams.adoc

Streams

There are several objects in Vert.x that allow items to be read from and written.

In previous versions the io.vertx.core.streams package was manipulating Buffer objects exclusively. From now, streams are not coupled to buffers anymore and they work with any kind of objects.

In Vert.x, write calls return immediately, and writes are queued internally.

It’s not hard to see that if you write to an object faster than it can actually write the data to its underlying resource, then the write queue can grow unbounded - eventually resulting in memory exhaustion.

To solve this problem a simple flow control (back-pressure) capability is provided by some objects in the Vert.x API.

Any flow control aware object that can be written-to implements WriteStream, while any flow control object that can be read-from is said to implement ReadStream.

Let’s take an example where we want to read from a ReadStream then write the data to a WriteStream.

A very simple example would be reading from a NetSocket then writing back to the same NetSocket - since NetSocket implements both ReadStream and WriteStream. Note that this works between any ReadStream and WriteStream compliant object, including HTTP requests, HTTP responses, async files I/O, WebSockets, etc.

A naive way to do this would be to directly take the data that has been read and immediately write it to the NetSocket:

NetServer server = vertx.createNetServer(
    new NetServerOptions().setPort(1234).setHost("localhost")
);
server.connectHandler(sock -> {
  sock.handler(buffer -> {
    // Write the data straight back
    sock.write(buffer);
  });
}).listen();

There is a problem with the example above: if data is read from the socket faster than it can be written back to the socket, it will build up in the write queue of the NetSocket, eventually running out of RAM. This might happen, for example if the client at the other end of the socket wasn’t reading fast enough, effectively putting back-pressure on the connection.

Since NetSocket implements WriteStream, we can check if the WriteStream is full before writing to it:

NetServer server = vertx.createNetServer(
    new NetServerOptions().setPort(1234).setHost("localhost")
);
server.connectHandler(sock -> {
  sock.handler(buffer -> {
    if (!sock.writeQueueFull()) {
      sock.write(buffer);
    }
  });

}).listen();

This example won’t run out of RAM but we’ll end up losing data if the write queue gets full. What we really want to do is pause the NetSocket when the write queue is full:

NetServer server = vertx.createNetServer(
    new NetServerOptions().setPort(1234).setHost("localhost")
);
server.connectHandler(sock -> {
  sock.handler(buffer -> {
    sock.write(buffer);
    if (sock.writeQueueFull()) {
      sock.pause();
    }
  });
}).listen();

We’re almost there, but not quite. The NetSocket now gets paused when the file is full, but we also need to unpause it when the write queue has processed its backlog:

NetServer server = vertx.createNetServer(
    new NetServerOptions().setPort(1234).setHost("localhost")
);
server.connectHandler(sock -> {
  sock.handler(buffer -> {
    sock.write(buffer);
    if (sock.writeQueueFull()) {
      sock.pause();
      sock.drainHandler(done -> {
        sock.resume();
      });
    }
  });
}).listen();

And there we have it. The drainHandler event handler will get called when the write queue is ready to accept more data, this resumes the NetSocket that allows more data to be read.

Wanting to do this is quite common while writing Vert.x applications, so we added the pipeTo method that does all of this hard work for you. You just feed it the WriteStream and use it:

NetServer server = vertx.createNetServer(
  new NetServerOptions().setPort(1234).setHost("localhost")
);
server.connectHandler(sock -> {
  sock.pipeTo(sock);
}).listen();

This does exactly the same thing as the more verbose example, plus it handles stream failures and termination: the destination WriteStream is ended when the pipe completes with success or a failure.

You can be notified when the operation completes:

server.connectHandler(sock -> {

  // Pipe the socket providing an handler to be notified of the result
  sock.pipeTo(sock, ar -> {
    if (ar.succeeded()) {
      System.out.println("Pipe succeeded");
    } else {
      System.out.println("Pipe failed");
    }
  });
}).listen();

When you deal with an asynchronous destination, you can create a Pipe instance that pauses the source and resumes it when the source is piped to the destination:

server.connectHandler(sock -> {

  // Create a pipe that to use asynchronously
  Pipe<Buffer> pipe = sock.pipe();

  // Open a destination file
  fs.open("/path/to/file", new OpenOptions(), ar -> {
    if (ar.succeeded()) {
      AsyncFile file = ar.result();

      // Pipe the socket to the file and close the file at the end
      pipe.to(file);
    } else {
      sock.close();
    }
  });
}).listen();

When you need to abort the transfer, you need to close it:

vertx.createHttpServer()
  .requestHandler(request -> {

    // Create a pipe that to use asynchronously
    Pipe<Buffer> pipe = request.pipe();

    // Open a destination file
    fs.open("/path/to/file", new OpenOptions(), ar -> {
      if (ar.succeeded()) {
        AsyncFile file = ar.result();

        // Pipe the socket to the file and close the file at the end
        pipe.to(file);
      } else {
        // Close the pipe and resume the request, the body buffers will be discarded
        pipe.close();

        // Send an error response
        request.response().setStatusCode(500).end();
      }
    });
  }).listen(8080);

When the pipe is closed, the streams handlers are unset and the ReadStream resumed.

As seen above, by default the destination is always ended when the stream completes, you can control this behavior on the pipe object:

• endOnFailure controls the behavior when a failure happens

• endOnSuccess controls the behavior when the read stream ends

• endOnComplete controls the behavior in all cases

Let’s now look at the methods on ReadStream and WriteStream in more detail:

ReadStream

ReadStream is implemented by HttpClientResponse, DatagramSocket, HttpClientRequest, HttpServerFileUpload, HttpServerRequest, MessageConsumer, NetSocket, WebSocket, TimeoutStream, AsyncFile.

Functions:

• handler: set a handler which will receive items from the ReadStream.

• pause: pause the handler. When paused no items will be received in the handler.

• resume: resume the handler. The handler will be called if any item arrives.

• exceptionHandler: Will be called if an exception occurs on the ReadStream.

• endHandler: Will be called when end of stream is reached. This might be when EOF is reached if the ReadStream represents a file, or when end of request is reached if it’s an HTTP request, or when the connection is closed if it’s a TCP socket.

WriteStream

WriteStream is implemented by HttpClientRequest, HttpServerResponse WebSocket, NetSocket, AsyncFile, and MessageProducer

Functions:

• write: write an object to the WriteStream. This method will never block. Writes are queued internally and asynchronously written to the underlying resource.

• setWriteQueueMaxSize: set the number of object at which the write queue is considered full, and the method writeQueueFull returns true. Note that, when the write queue is considered full, if write is called the data will still be accepted and queued. The actual number depends on the stream implementation, for Buffer the size represents the actual number of bytes written and not the number of buffers.

• writeQueueFull: returns true if the write queue is considered full.

• exceptionHandler: Will be called if an exception occurs on the WriteStream.

• drainHandler: The handler will be called if the WriteStream is considered no longer full.

Pump

Instances of Pump have the following methods:

• start: Start the pump.

• stop: Stops the pump. When the pump starts it is in stopped mode.

• setWriteQueueMaxSize: This has the same meaning as setWriteQueueMaxSize on the WriteStream.

A pump can be started and stopped multiple times.

When a pump is first created it is not started. You need to call the start() method to start it.