EchoHandler.ashx.cs

This is an example of how to use the new low level WebSockets APIs introduced in ASP.NET 4.5. In practice we expect that few developers will need to use these low level APIs and will instead use the Microsoft.WebSockets NuGet package.

The function of this handler is simple. It receive text messages over the WebSocket connection, appends "You said " to the front of the message and sends it back over the WebSocket connection.

This handler takes advantage of the new asynchrony features in C# 5. Explaining these features is beyond the scope of this documentation - to learn more visit the async homepage or read the async articles on Eric Lippert's blog.

The source for this sample is on GitHub.

This HTML documentation was generated using Nocco.


Imports

Some standard imports, but note the last one is the new System.Net.WebSockets namespace.

using System;
using System.Web;
using System.Text;
using System.Threading;
using System.Threading.Tasks;
using System.Net.WebSockets;

namespace AspNetWebSocketEcho
{

The EchoHandler class

The most low-level way to utilize the new WebSocket features in ASP.NET is to implement your own IHttpHandler.

    public class EchoHandler : IHttpHandler
    {        

Accepting the connection

If the incoming request is a valid WebSocket request then accept the request and use the HandleWebSocket method to handle the WebSocket connection. Defining a seperate method (HandleWebSocket in this case) isn't necessary but it helps with readability by reducing the amount of code nesting.

        public void ProcessRequest(HttpContext context)
        {
            if (context.IsWebSocketRequest)
                context.AcceptWebSocketRequest(HandleWebSocket);
            else
                context.Response.StatusCode = 400;

        }
                     

When a WebSocket connection request is accepted, HandleWebSocket is invoked and ASP.NET passes in an instance of WebSocketContext. The WebSocketContext class captures relevant information available at the time of the request. For example:

  • WebSocketContext.RequestUri is the Uri used to establish the WebSocket connection. This might include query string parameters.
  • WebSocketContext.CookieCollection contains any cookies that were included in the connection request.

WebSocketContext is a read-only type - you cannot perform any actual IO operations such as sending or receiving using the WebSocketContext. These operations can be performed by accessing the System.Net.WebSocket instance via the WebSocketContext.WebSocket property.

        private async Task HandleWebSocket(WebSocketContext wsContext)
        {

Receiving

Define a maximum message size this handler can receive (1K in this case) and allocate a buffer to contain the received message. This buffer will be reused for each receive operation.

            const int maxMessageSize = 1024;
            byte[] receiveBuffer = new byte[maxMessageSize];            
            WebSocket socket = wsContext.WebSocket;            

While the WebSocket connection remains open we run a simple loop that receives messages and then sends them back.

            while (socket.State == WebSocketState.Open)
            {                

The first step is to begin a receive operation on the WebSocket. ReceiveAsync takes two parameters:

  • An ArraySegment to write the received data to. In this particular case we are passing an ArraySegment that points to our entire receive buffer. You might be wondering what the point of ArraySegment is and why ReceiveAsync doesn't just accept a byte[]. This will be explained below.
  • A cancellation token. In this example we are not using any timeouts so we use CancellationToken.None.

ReceiveAsync returns a Task<WebSocketReceiveResult>. We use the await keyword to "asynchronously wait" for the receive operation to complete and extract the WebSocketReceiveResult from the completed task. The WebSocketReceiveResult provides information on the receive operation that was just completed, such as:

  • WebSocketReceiveResult.MessageType - What type of data was received and written to the provided buffer. Was it binary, utf8, or a close message?
  • WebSocketReceiveResult.Count - How many bytes were read?
  • WebSocketReceiveResult.EndOfMessage - Have we finished reading the data for this message or is there more coming?
                WebSocketReceiveResult receiveResult = await socket.ReceiveAsync(new ArraySegment<byte>(receiveBuffer), CancellationToken.None);
                

The WebSocket protocol defines a close handshake that allows a party to send a close frame when they wish to gracefully shut down the connection. The party on the other end can complete the close handshake by sending back a close frame.

If we received a close frame then lets participate in the handshake by sending a close frame back. This is acheived by calling CloseAsync. CloseAsync will also terminate the underlying TCP connection once the close handshake is complete.

The WebSocket protocol defines different status codes that can be sent as part of a close frame and also allows a close message to be sent. If we are just responding to the client's request to close we can just use WebSocketCloseStatus.NormalClosure and omit the close message.

                if (receiveResult.MessageType == WebSocketMessageType.Close)
                {                    
                    await socket.CloseAsync(WebSocketCloseStatus.NormalClosure, string.Empty, CancellationToken.None);
                }

This echo server can't handle binary messages so if we receive one we close the connection with an appropriate status code and message.

                else if (receiveResult.MessageType == WebSocketMessageType.Binary)
                {                    
                    await socket.CloseAsync(WebSocketCloseStatus.InvalidMessageType, "Cannot accept binary frame", CancellationToken.None);
                }
                else
                {

At this point we know we are receiving UTF-8 data, but we don't know if we've received the entire message or if we need to call ReceiveAsync again. We can use the EndOfMessage property to determine this and start a loop that will continue to call ReceiveAsync until we have read the entire message or exceeded our maximum message size (in which case we close the connection using an appropriate status code). The local count variable is used to keep track of the total number of bytes read for this message.

                    int count = receiveResult.Count;

                    while (receiveResult.EndOfMessage == false)
                    {
                        if (count >= maxMessageSize)
                        {                        
                            string closeMessage = string.Format("Maximum message size: {0} bytes.", maxMessageSize);
                            await socket.CloseAsync(WebSocketCloseStatus.MessageTooLarge, closeMessage, CancellationToken.None);
                            return;
                        }

Our first call to ReceiveAsync didn't return the data for the complete message so we are calling it again. Now the benefits of using ArraySegment over byte[] are clear - we can pass an ArraySegment that exposes the next section of our receive buffer by using the count of bytes that we have already read as an offset.

                        receiveResult = await socket.ReceiveAsync(new ArraySegment<byte>(receiveBuffer, count, maxMessageSize - count), CancellationToken.None);
                        count += receiveResult.Count;
                    }

Decoding the message and sending a response

The complete message is now in the receive buffer so we must convert the data to a string by decoding it as UTF-8. We know that we can do this because the WebSocket protocol only defines two different data formats - binary and UTF-8. If we wanted our server to receive text with a different encoding then we would expect to receive binary messages instead. After appending some text to the front of the message, we UTF-8 encode the string so that can get an ArraySegment with the appropriate data to send.

                    var receivedString = Encoding.UTF8.GetString(receiveBuffer, 0, count);
                    var echoString = "You said " + receivedString;
                    ArraySegment<byte> outputBuffer = new ArraySegment<byte>(Encoding.UTF8.GetBytes(echoString));

Now send the data using SendAsync using WebSocketMessageType.Text as the message type.

                    await socket.SendAsync(outputBuffer, WebSocketMessageType.Text, true, CancellationToken.None);                    

The echo operation is complete. The loop will resume and ReceiveAsync is called again to wait for the next message.

You might be wondering about exception handling - for example, what if the client aborts the connection after sending a message but before receiving the echo? Won't an exception be thrown when SendAsync is called? Yes, but ASP.NET will catch the exception that bubbles up from our handler, dispose of the WebSocket for us and log the error in the Application event log. Of course a real application will certainly need error handling but its unnecessary for this simple example.

                }
            }
        }

Since this IHttpHandler contains no state, it is safe for IIS to reuse for multiple requests.

        public bool IsReusable
        {
            get { return true; }
        }

    }
}