A multi-participant Paint in HTML5

De $1

Version de 22:19, 21 Nov 2024

cette version.

Revenir à liste des archives.

Voir la version actuelle

_

Introduction

In this tutorial, you will learn how to write a multi-participant Paint program in pure HTML5, using the <canvas> element, WebSockets, the getUserMedia API pour live video capture, the NodeJS server, etc.

Recommended tools

  • A recent web browser that supports the <canvas> tag, most do,
  • For live web cam capture, you will need a special Opera build (Windows, Linux, Mac) or Chrome Canary build (Windows, better than Opera for Mac)
  • Check that you have a JavaScript debugger in your browser (Firefox users will need to install Firebug.
  • Editor of your choice for HTML/Javascript (recommended : WebStorm from JetBrains, free licence for researchers/education, 30 days free trial for others).
  • For steps 7 and superior that involves WebSockets, you will need the NodeJS web server and some modules. See NodeJS installation guide for Windows, Linux, Mac.

Download lab exercices

All the source files you will need are included in this zip archive (4Mbytes). Unzip, go in the "Paint_HTML5_Multi_Tutorial". You will find subdirectories step1, step2, etc... corresponding to each step described in this document.

Step 1 : a very simple paint !

Go into the step 1 directory. You may  try the example by opening paint.html in your browser.

Notice the first lines of this file :

<!DOCTYPE HTML>
<!-- Note : in HTML5 the DOCTYPE can be memorized by a normal human being -->
<!-- Common mistake : if we put a comment before the doctype element, Internet Explorer will not
     consider the document as a HTML5 document and the canvas tag for example,
     will not be displayed ! -->

The DOCTYPE is the brand new ultra-simple HTML5 DOCTYPE, yeah !

Notice that if you write ANYTHING before this DOCTYPE line, IE9 will not consider the file as HTML5, and you will no be able to run the example ! The <canvas> tag in the page will be ignored. Other browsers will however do nothing nasty if you write some comment at the top of the document.

Then, we declare a CSS and some JavaScript scripts. Yep, we will use jQuery for most DOM manipulation and event handling. HTML5 proposes the new document.querySelector(...) that behaves closely like jQuery selectors but is so much longer than $(...),  and jQuery uses querySelector internally...

<link rel="stylesheet" href="css/paint.css"/>

<script src="js/jquery-1.7.2.min.js"></script>
<script src="js/utils.js"></script>
<script src="js/paint.js"></script>

paint.js holds most of the code for a small "draw when mouse moves" paint program, while utils.js contains a utility function for getting correct mouse cursor position.

When the document if fully loaded and the DOM ready, we create a Paint object, that will hold most of our JavaScript code.

<script type="text/javascript">
// Run when the DOM is ready
$(document).ready(function () {
  // Create the pseudo object which will handle the main canvas
  paint = new PaintObject("canvasMain");
});
</script>

The parameter ("canvasMain") is the Id of the <canvas> tag you will find in the HTML part of the page :

<canvas id="canvasMain">
    <p>Canvas tag not supported by your browser</p>
</canvas>

Let' have a look at the PaintObject code in the paint.js file... First we get a reference on the canvas element and secondly we get a context object in order to draw in the canvas :

function PaintObject(canvas) {
    // get handle of the main canvas, as a DOM object, not as a jQuery Object.
    var mainCanvas = $("#"+canvas).get(0);
    var mainContext = mainCanvas.getContext('2d');
    ...

We also bind a mouse move listener to the canvas :

// Bind event on the canvas
$("#canvasMain").mousemove(this.mouseMoveListener);

The mouse listener is a method of the PaintObject that draw a line at each mouse move, from the previous mouse position to the current one :

// Mouse motion listener for drawing lines
    this.mouseMoveListener = function (event) {
        // we delegate the computation of the mouse position
        // to a utility function (in utils.js) as this is not so trivial
        // we must take into account insets, etc.
        var mousePos = getMousePos(mainCanvas, event);

        // Let's draw some lines that follow the mouse pos
        if (!started) {
            previousMousePos = mousePos;
            started = true;
        } else {
            mainContext.beginPath();
            mainContext.moveTo(previousMousePos.x, previousMousePos.y);
            mainContext.lineTo(mousePos.x, mousePos.y);
            mainContext.closePath();
            // draws as wireframe the drawing orders between beginPath() and closePath()
            mainContext.stroke();

            previousMousePos = mousePos;
        }
    };

Notice that in order to draw a line we declare a path, we move to a position, we draw a line to another position, we close the path. The path is drawn at the mainContext.stroke() call. It will be drawn with the current mainContext.strokeStyle value, which is defaulted to Black.

Be careful : just forget to close the path and at every mouse move you will draw the whole path from the beginning ! Indeed : mainContext.stroke() draws every draw primitive declared since the context.beginPath() call... if a beginPath() is called while a path is not closed, it is ... ignored.

Notice the trick we use in order to set the size of the canvas depending on the size of the browser. Instead of using CSS percentages on canvas width and height, that produce ugly scaling of canvas content, we adjust the size from a JavaScript piece of code :

// Canvas doesnt scale well with '%' in CSS so we use a little trick.
// We give them the size of one of their parent node which can be scalable.
mainCanvas.height = $("#content")[0].clientHeight;
mainCanvas.width = $("#content")[0].clientWidth;

Step 2 : manage mouseUp, mouseMove, mouseDown, define a pencil drawing object

Go into the step 2 directory. You may  try the example by opening paint.html in your browser.

This time we draw only when the mouse moves while a mouse button is pressed !

We also separated the code that draws from the initialisation of the PaintObject. the file drawingtools.js will hold "drawing objects" like pencil, line, rectangle, circle and so on... while paint.js will just do the job for creating contexts, binding events, etc.

We used jQuery to bind different mouse events to the canvas (paint.js)

// Create the drawing tool
var drawingTool = new pencilDrawingTool(); 

// Bind events on the canvas
$("#canvasMain").mousedown(drawingTool.mouseDownListener);
$("#canvasMain").mousemove(drawingTool.mouseMoveListener);
$("#canvasMain").mouseup(drawingTool.mouseUpListener);

 and here is the drawingtool.js content with only a tool that draws like a pencil (notice the 3 listener methods for mouseDown, mouseMove and mouseUp events) :

// Previous position of the mouse
var previousMousePos;

// Drawing tool object
function pencilDrawingTool() {
    this.mouseDownListener = function (event) {
        paint.started = true;
        previousMousePos = getMousePos(paint.getFrontCanvas(), event);
    };

    this.mouseMoveListener = function (event) {
        // we delegate the computation of the mouse position
        // to a utility function as this is not so trivial
        var mousePos = getMousePos(paint.getMainCanvas(), event);

        // Let's draw some lines that follow the mouse pos
        if (paint.started) {
            paint.getMainContext().beginPath();
            paint.getMainContext().moveTo(previousMousePos.x, previousMousePos.y);
            paint.getMainContext().lineTo(mousePos.x, mousePos.y);
            paint.getMainContext().closePath();
            paint.getMainContext().stroke();
        }
        previousMousePos = mousePos;
    };

    this.mouseUpListener = function (event) {
        paint.started = false;
    }
};

Step 3 : add a second, transparent canvas on top of the main canvas, for drawing elastic lines

Go into the step 3 directory. You may  try the example by opening paint.html in your browser.

This example shows an interesting use of canvases : they can be layered one on top of another. Using relative position and z-index with different values, they can act as layers in photoshop. By default they are transparent, and if you use the context.clearRect(x, y, width, height) function you will make a transparent rectangle and be able to "see through" it.

We will use this feature to draw an elastic line (lines that follow the mouse position) on a transparent front canvas, without modifying the main canvas.

In order to animate the line when the mouse moves, we clear the front canvas before drawing a line in its new position.

Adding a new canvas on top of the previous one

This is done in paint.js :

// Prepare a second canvas on top of the previous one, kind of second "layer" 
 var frontCanvas = document.createElement('canvas');
 frontCanvas.id = 'canvasFront';

 // Add the temporary canvas as a second child of the mainCanvas parent.
 mainCanvas.parentNode.appendChild(frontCanvas);

// Get the context for drawing in the canvas
var frontContext = frontCanvas.getContext('2d');

this.getFrontCanvas = function () {
        return frontCanvas;
}

this.getFrontContext = function () {
        return frontContext;
}

We added a line drawing tool in drawingtools.js

We declared an array that will hold the different drawing tools (pencil, line, etc) :

var DEFAULT_TOOL = 'pencil';
...
// Create the drawing tool
var drawingTool = new setOfDrawingTools[DEFAULT_TOOL]();

And the event binding has been modified in order to call the current drawing tool mouse listener method :

// bind events. We use a function multiplexEvent that will call the proper listeners
// methods of the currentTool.
this.bindMultiplexEvents = function () {
    $("#canvasFront").mousedown(this.multiplexEvents);
    $("#canvasFront").mousemove(this.multiplexEvents);
    $("#canvasFront").mouseup(this.multiplexEvents);
}

// if currentTool is pencil, and event.type is mousemove, will
// call pencil.mousemouve(event), if currentTool is line and
// event.type is mouseup, will call line.mouseup(event) etc.
this.multiplexEvents = function (event) {
    drawingTool[event.type](event);
}

...notice that in the line drawing tool (drawingtools.js) we draw only in the front canvas, and before we clear the front canvas :

// the Line Drawing Tool Object
setOfDrawingTools.line = function () {
    this.mousedown = function (event) {
        paint.started = true;
        previousMousePos = getMousePos(paint.getFrontCanvas(), event);
    };

    this.mousemove = function (event) {
        var mousePos = getMousePos(paint.getFrontCanvas(), event);
        if (paint.started) {
            // Clear front canvas before drawing the elastic line in a new position
            paint.getFrontContext().clearRect(0, 0, paint.getFrontCanvas().width, 
                                                    paint.getFrontCanvas().height);

            paint.getFrontContext().beginPath();
            paint.getFrontContext().moveTo(previousMousePos.x, previousMousePos.y);
            paint.getFrontContext().lineTo(mousePos.x, mousePos.y);
            paint.getFrontContext().stroke();
            paint.getFrontContext().closePath();
        }
    };

    this.mouseup = function (event) {
        paint.started = false;
    }
};

 In the paint.html file, we added some spans in order to select the current tool (l

<div id="drawCommands">
    <h6>Pick a tool</h6>
    <span id="line">line</span>
    <span id="pencil">pencil</span>
</div>

And in paint.js we detect clicks on them :

// Handle the drawing tools menu. The selected entry value can be 'Pencil',
// 'Line' etc.
this.changeDrawingTool = function () {
    // this.id is the id of the selected menu item
    drawingTool = new setOfDrawingTools[this.id]();
}

// Bind the changeDrawingTool function onClick to every menu items.
$("#drawCommands").find("span").click(this.changeDrawingTool);

Small problem : how can we make the line stay on the drawing when we release the mouse button ?

Step 4 : draw the front canvas on the main canvas on mouse release when drawing lines

Go into the step 4 directory. You may try the example by opening paint.html in your browser.

This time when we release the mouse button the line is drawn and will stay in the main drawing, even if we draw other elastic lines.

We just draw the front canvas on the the main canvas on mouseUp events.

In paint.js we added a drawFrontCanvasOnMainCanvas() method on the PaintObject :

this.drawFrontCanvasOnMainCanvas = function () {
    mainContext.drawImage(frontCanvas, 0, 0);
    // after copying the front canvas content, we clear it !
    frontContext.clearRect(0, 0, frontCanvas.width, frontCanvas.height);
}

Notice that we use the drawImage() call from a canvas context object. Indeed, we can draw a canvas on a canvas !

And we call this method from the mouseUp() listener method of the line drawing tool :

// The Line Drawing Tool Object
setOfDrawingTools.line = function () {
    ...
    this.mouseup = function (event) {
        paint.started = false;
        paint.drawFrontCanvasOnMainCanvas();
    }
};

Step 5 : adding a rectangle and a Circle tool

Go into the step 5 directory. You may try the example by opening paint.html in your browser.

This time, the same way we did the line tool, we added two new tools : one for drawing elastic rectangles and one for drawing elastic circles.

You may look at these tools in drawingtools.js.

Now that we have a small framework, you can try to add your own tools for drawing ellipses, images etc.

Step 6 : playing with stroke and fill styles, lineWidth, HTML5 new input types (color, range)

Go into the step 6 directory. You may try the example by opening paint.html in your browser.

This time we added new HTML5 input types in the graphic user interface :

<ul>
  <li>Fill color:
      <!-- if running opera you may try : <input id="fillColor" type="color" 
           value="FFFFFF"/> instead of this jscolor input field -->
      <input id="fillColor" class="color" value="FFFFFF"/>
   </li>
   <li>Stroke color:
       <!--  if running opera you may try :<input id="strokeColor" type="color" 
             value="FFFFFF"/> -->
       <input id="strokeColor" class="color" value="000000"/>
   </li>
   <li>
       <!-- works only in Chrome and Opera, will display a standard input text in Firefox -->
        Stroke size: <input id="strokeSize" type="range" 
                            min="0" max="20" value="1" step="0.5" 
                            style="position:relative;top:6px;"/>
   </li>
   <li>
       Fill shapes : <input id="fillShapes" type="checkbox" checked/>
   </li>
</ul>

We used a <input type="range".../> for selecting the line width. This will work in Opera, Webkit based browser. It is not yet supported by Firefox, but will work nevertheless if you enter a numeric value.

For color selection we used a "polyfill", a <input class="color" /> that works with the jscolor.js library. If you are running Opera, the HTML5 <input type="color"/> will work and display a nice color chooser natively. Just modify the HTML code and open in Opera.

In paint.js we handle the values of these fields in order to keep the contexts for both canvases (mainCanvas and frontCanvas) up to date :

// Handle the color menus
    mainContext.strokeStyle = frontContext.strokeStyle = "#" + $("#strokeColor").val();
    $("#strokeColor").change(function() {
        mainContext.strokeStyle = frontContext.strokeStyle = "#" + $("#strokeColor").val();
    });

    mainContext.fillStyle = frontContext.fillStyle = "#" + $("#fillColor").val();
    $("#fillColor").change(function() {
        mainContext.fillStyle = frontContext.fillStyle = "#" + $("#fillColor").val();
    });

    // handle the stroke size
    mainContext.lineWidth = frontContext.lineWidth = $("#strokeSize").val();
    $("#strokeSize").change(function() {
        mainContext.lineWidth = frontContext.lineWidth = $("#strokeSize").val();
    });

    var fillShapes = true;
    // handle the check box that specifies if we fill shapes
    //this.fillShapes = $("#fillShapes").attr('checked');
    $("#fillShapes").change(function(){
        fillShapes = $(this).attr("checked");
    });

Like that, all the calls to stroke() or fill() using any of the contexts will draw a shape using the selected corresponding color.

Step 7 : adding a chat in our application, starting point for a multi-participant experience

We are going to use HTML5 WebSockets in order to send and receive messages from a small chat application we will embed in our graphic user interface.

We will use the NodeJS web server

In order to use WebSockets we need a WebSocket server. We propose in this tutorial to use the NodeJS web server. NodeJS is a lightweight server that has the particularity of running server applications written in JavaScript (it embeds the V8 Javascript interpreter, the same that is embedded in the Chrome Browser).

Numerous modules are available :

  • express brings the hability to run a HTTP web server using a MVC approach in nodeJS. We will use it to serve static HTML pages that use CSS and Javascript files,
  • socket.io : brings WebSockets over HTTP + proposes several transparent fallbacks in case the browser you use does not support WebSockets (it will use Flash WebSockets or AJAX Long Calling for example...). Socket.io also takes care of deconnexions, etc.
  • nowJS : built on top of socket.io, this module brings distributed objects over WebSockets, making the development of multi-participant applications much easier.

Installing NodeJS

Go to http://www.nodejs.org and follow the installation procedure. If you are running Linux, look carefuly at nodeJS home page, there is a link to precompiled builds :

Linux NodeJs.png

Update your PATH variable so that the node executable is found when you type "node --version" in your favorite terminal/dos command window (note : C:\Program Files(x86)\nodejs by default on windows).

Test now NodeJS using a small application

  • Create a test.js file with this content :
var http = require('http');
http.createServer(function (req, res) {
  res.writeHead(200, {'Content-Type': 'text/plain'});
  res.end('Hello World\n');
}).listen(8124, "127.0.0.1");
console.log('Server running at http://127.0.0.1:8124/');

Save it somewhere. Then run from the command line :

node test.js


You just ran a nodeJS server that runs an application at this URL : http://localhost:8124

Snap5.jpg

Open it using your web browser !

You should get :

Snap6.jpg

This was our first NodeJS application ;-)

Adding node modules for WebSockets

There are numerous ways to add modules to NodeJS. If you are using Linux, it is pretty straightforward, normally it should go fine if you have a Mac, and if you are running Windows it may be a bit more complicated for some modules like NowJS that require VC++ redistribuable to be installed.

Most of the time we install modules using the npm command (Node Package Manager). And most of the time we run this command in the directory of our projet. This will create a node_modules subdir in the project.

The chat application we will run now comes in this archive : ChatNowJs.rar

Note that the archive contains already a node_module directory, but it will not suffice as the installation of modules may also compile native parts of some modules.

Unzip the archive somewhere, go in the ChatNowJS directory and run the following commands :

  • If on windows : install VC++ runtime (3Mo) before going any further !
     
  • Then run the following commands :
    1. npm install socket.io,
    2. npm install node-proxy"
    3. npm install now
    4. npm install express

Run the simple chat application

Run the following command : "node simpleChat_server.js", then open http://localhost:8080 in your browser. Open two tabs at the same URL, you will see an input field. Type in something, and it should be sent to the other tab (as both are clients connected to the same WebSocket server).

In case you have no errors when you launch"node  simpleChat_server.js", and the chat is not working, open the JavaScript debugger console. If you get a  "now.distributeMessage no such object/function", then your browser is "too old but not so much", it supports WebSockets, but not the "good ones" ! Indeed some Firefox version from one year/one year and a half had an implementation for an older versions of the protocol. Update or change your browser. If it still does not work even with a recent browser (problem only encountered under windows and exotic antivirus/anti spywares), you may tell socket.io (used by nowJS) not to use HTML5 WebSockets but another  mean of transmission instead: modify one line in the simpleChat_server.js code :

  • Put var everyone = nowjs.initialize(server, {socketio: {transports: ['xhr-polling', 'jsonp-polling', 'htmlfile']}}); instead of var everyone = nowjs.initialize(server);.

Study the simple chat application

Two files are involved in the simple chat application :

  1. simpleChat_server.js
  2. simpleChat.html

Open these files in your favorite editor : the application is ...wow, just a few lines of code... Let's talk about nowJS.... now !

NowJS defines a "magic pocket" called "now" : a JavaScript variable you can manipulate from client code running in your browser, that will hold variables and functions shared between the client and the server code.

The server code can use a "everyone.now" variable in order to define variables and functions shared with all clients.

For example, in the server side, defining a function distributeMessage in the everyone.now variable will make it callable by any client who will use it with his own "now" magic pocket.

Server :

everyone.now.distributeMessage = function(message){
  everyone.now.receiveMessage(this.now.name, message);
};

Example of call by a client (see the simpleChat.html file) :

$("#send-button").click(function(){
    now.distributeMessage($("#text-input").val());
    $("#text-input").val("");
});

So one clic on the send button executes now.distributeMessage() that will be run on the server, passing the message typed in the input field. The server gets the message in its everyone.now.distributeMessage function, that calls in turn everyone.now.receiveMessage, a function defined.... in the client ! The "everyone" parts means that it will call this function for all the connected clients !

Client :

now.receiveMessage = function(name, message){
    $("#messages").append("<br>" + name + ": " + message);
}

And this function, located on the client side, will just append to the chat log the message that just arrived !

The same way, a variable defined in the client code,

now.name = prompt("What's your name?", "");

May be used in the server code as a kind of "session" variable, using the this.now namespace :

Server :

everyone.now.distributeMessage = function(message){
  everyone.now.receiveMessage(this.now.name, message);
};

Using this simple paradigm, NowJS enables rapid development of collaborative logic over WebSockets. Notice that objects are send as is, as pure javascript objects, they are serialized/deserialized in JSON under the hood.

Adding the Chat application to our Paint Program

Go into the step 7 directory. Stop nodeJS if it is still running.

Look at the server.js file :

// Use the express module for nodeJS
// Express is a framework for implementing a more
// complete web server with nodeJS
var express = require('express');

var app = express.createServer();

// Indicate where static files are located
app.configure(function () {
    app.use(express.static(__dirname + '/client/'));
});
app.get('/', function (req, res) {
    //res.render('client/paint.html');
    res.redirect("/paint.html");
});
app.listen(80);

// Creation of a web socket server using nowJS
var nowjs = require("now");
var everyone = nowjs.initialize(app);


nowjs.on("connect", function () {
    console.log("Joined: " + this.user.clientId);
});

nowjs.on("disconnect", function () {
    console.log("Left: " + this.user.clientId);
});

// Function called by a chat client, passes the message to each client
// connected by calling its processIncomingChatMessage shared function
everyone.now.distributeMessage = function (message) {
    // send the message to everyone
    everyone.now.processIncomingChatMessage(this.now.name, message);
};

This server code is very similar to the one of the chat program, instead  this time we use the "express" NodeJS module to run the HTTP web server part (13 first lines). This enables to serve properly HTML pages that include CSS or JavaScript files. The rest is very similar to the simpleChat_server.js file.

Run the following command : "node server.js", and open "http://localhost" in your browser.

This time you should see the same paint program we had in step6, augmented by a chat application at the bottom of the user interface.

Open two tabs with the same URL, try to chat... Both the paint and the chat should work at the same time. Paint actions are not transferred however... we will add that in the next step !

You may look at the source code as well. We isolated the JavaScript code for the chat in a chat.js file while we added the input field and a div for handling log messages in the paint.html file.

Here are the additions in the paint.html file :

<head>
    ...
    <!-- NowJS is a High level lib built on top of socket.io. 
         Uses WebSockets if available. The page needs to
         be served by the NodeJS server -->
    <script src="js/chat.js"></script>
    <script src="/nowjs/now.js"></script>

    <script type="text/javascript">
        // Run when the DOM is ready
        $(document).ready(function () {
            // Create the pseudo object which will handle the main canvas
            paint = new PaintObject("canvasMain");
            // Bind events to the canvas
            paint.bindMultiplexEvents();

            // Init the chat system
            now.ready(function () {
                // We must be sure "now" exists. So we prompt the user only
                // when the shared variable "now" is ready !

                // prompts a dialog and asks for a nickname
                now.name = prompt("What's your name?", "");

                //Create the pseudo object which will handle the chat
                chat = new ChatObject();
            });
        });
    </script>
</head>

Notice that the paint.html file includes a now/now.js file that is served by NodeJS and that is not part of the paint project. It is located in the node_modules/now directory. This is why we had to install/configure the different modules for NodeJS.

We also create the chat object only when the nowJS is fully initialized. NowJS calls the now.ready() method once it is ready, so we create the chat object in the body of this method.

At the end of the paint.html file we added the HTML necessary for the chat :

<div id="chat">
    <div id="chatMessages"></div>
     <input type="text" id="chatTextInput">
     <button class="menuButton" type="button" id="chatSendButton">Send</button>
</div>

The code for handling a press on the send button (or the enter key in the input field), sending the message, and displaying received messages is in the chat.js file :

var chat;

function ChatObject() {
    // Handle message input for the chat
    // When the send button has been clicked... or when the enter key has been 
    // pressed -> send content to the chat
    // server
    $("#chatTextInput").keypress(function (e) {
        var code = (e.keyCode ? e.keyCode : e.which);
        if (code == 13) {
            sendChatMessage();
        }
    });

    $("#chatSendButton").click(function () {
        sendChatMessage();
    });

    function sendChatMessage() {
        if ($("#chatTextInput").val() != "") {
            // This function is defined on the JavaScript code that runs on the server
            now.distributeMessage($("#chatTextInput").val());
            $("#chatTextInput").val("");
        }
    }

    // This function is called when a chat message arrives. Called by the server !
    now.processIncomingChatMessage = function (username, message) {
        // appends the incoming message to the messageLogs
        $("#chatMessages").append("<p><strong>" + username + "</strong> : " + 
                                   message + "</p>");
        $("#chatMessages").get(0).scrollTop = $("#chatMessages").get(0).scrollHeight;
    }
};

 Step 8 : broadcast paint events

Go into the step 8 directory. Stop nodeJS if it is still running from the previous step.

Server side : broadcasting paint event to every client except the sender

This time, without touching anything from the chat code, we added one single method on the server side for broadcasting paintCommand objects that will hold a description of a paint event (like "drawing a line with this color, from this position to this position")... :

Server (see server.js file) :

everyone.now.distributePaintCommand = function (paintCommand) {
    // send the paint command to everyone except sender
    everyone.exclude(this.user.clientId).now.processPaintCommand(paintCommand);
};

This time we used another goodie from nowJS : the feature for excluding some clients when broadcasting. The everyone.exclude(this.user.clientId).now.processPaintCommand(paintCommand) part means "call the processPaintCommand of every client except myself !

Indeed, when one client draws a line, this line must be drawn on other clients' screens but not on on the sender's screen. This is the only difference on the server side, if we compare to the distributeMessage we used for broadcasting chat messages.

Sending paint events

In the client code, the sending of messages is done in each drawing tool. Let's have a look at one of them : the line tool (drawingtools.js file) :

// The Line Drawing Tool Object
setOfDrawingTools.line = function () {
    ...

    this.mousemove = function (event) {
        ...
        sendLineMove(mousePos);
    };

    // Send the line to other clients.
    function sendLineMove(mousePos) {
        var paintCommand = {};
        paintCommand.type = 'lineMove';
        paintCommand.previousMousePos = previousMousePos;
        paintCommand.currentMousePos = mousePos;
        paintCommand.properties = paint.getCurrentContextProperties();
        now.distributePaintCommand(paintCommand);
    }

    this.mouseup = function (event) {
        ...
        paint.sendOrderToDrawFrontCanvasOnMainCanvas();
    }
};

We added a sendLineMove method that is called at every mousemove event. This method builds a paintCommand object with a type attribute equal to 'lineMove', and that will hold the coordinates of the extremity of the line, and a field name 'properties' for holding the current stroke color, fill color, line width etc. These properties are returned by   paint.getCurrentContextProperties() method.

After this object is built, we send it to the server for broadcasting, by calling the now.distributeCommand(paintCommand) method, running on the server.

Processing incoming messages

First, we defined in the paint.html file a now.processIncomingMessage method. As it is defined as a method of the now object, it is callable by the server code.

<script type="text/javascript">
        // Run when the DOM is ready
        $(document).ready(function () {

            // prompts a dialog and asks for a nickname
            now.ready(function () {
                ...

                now.processPaintCommand = function (paintCommand) {
                    paint.processPaintCommand(paintCommand);
                };

               ...
            });
        });
    </script>

So, everytime the server calls processPaintCommand of the connected clients, this will in turn call paint.processPaintCommand(paintCommand), a method from the paint Object, located in the paint.js file :

this.processPaintCommand = function (paintCommand) {
    // save contexts on a stack, method provided by the canvas API
    mainContext.save();
    frontContext.save();

    // change current contexts so that they are same as sender contexts
    this.changeContextsProperties(paintCommand.properties);

    switch (paintCommand.type) {
        ...
        case 'lineMove' :
            // clear the content of the front canvas
            frontContext.clearRect(0, 0, frontCanvas.width, frontCanvas.height);

            // draw the line, using values received in the paintCommand...
            ...
            break;
            ...
        case 'drawFrontCanvasOnMainCanvas' :
            // Got the order to draw front canvas on main canvas
            this.drawFrontCanvasOnMainCanvas();
            break;
        }

        // restore contexts, current color, etc. From canvas API
        mainContext.restore();
        frontContext.restore();
}

This is very simple processing, we just had to take care of the contexts : when we get an incoming paint event, we need to :

  1. save current contexts (i.e : draw in blue, wireframe, line width = 1)
  2. switch to sender contexts (i.e : draw in red, fill shapes, fill color = green, line width = 5)
  3. draw
  4. restore contexts to previous values

We use the save() and restore() methods from the context object, from the canvas API. These methods save and restore from a stack so we can do multiple save/restores.

You may look at other drawing tools and see the way we processed the different type of paint commands.

Using live webcam capture : the getUserMedia API

Just for fun, we added support for drawing rectangles filled with the current image seen by your webcam. This feature works only with a few experimental browsers :

  • Windows : Chrome Canary build, Opera special lab build for camera support,
  • Mac : Chrome Canary build (recommended), Opera special lab build for camera support, bugs sometimes...
  • Linux : ?

See the links to these browsers in the recommend tools section. You may need to activate support for the getUserMedia API in the config options for Chrome Canary.

Init the getUserMedia API

In the paint.html file, we have a call to a initWebCam() function :

<script src="js/webcam.js"></script>
...    
<script type="text/javascript">
        // Run when the DOM is ready
        $(document).ready(function () {
            ...
            // init web cam, works only in last test versions of chrome and opera.
            initWebCam();
        });
    </script>

The initWebCam function is located in the webcam.js file :

function initWebCam() {
    var n = navigator,
    is_webkit = false;

    function onSuccess(stream) {
        var output = document.getElementById('output'),
            source;

        output.autoplay = true;

        if (!is_webkit) {
            source = stream;
        }
        else {
            source = window.webkitURL.createObjectURL(stream);
        }

        output.src = source;
    }

    function onError() {
        // womp, womp :(
    }

    if (n.getUserMedia) {
        // opera users (hopefully everyone else at some point)
        n.getUserMedia({video:true, audio:true}, onSuccess, onError);
    }
    else if (n.webkitGetUserMedia) {
        // webkit users
        is_webkit = true;
        n.webkitGetUserMedia('video, audio', onSuccess, onError);
    }
}

As you may notice, we've got here several tests for checking if we are running opera or a webkit based browser. Indeed, the getUserMedia is still in beta implementation stages and API implementations may differ.

The principle is :

  1. Make a call to navigator.getUserMedia(params, onSuccess, onError), this is done at the end of the code and we take care of Opera/Webkit differences.
  2. in the onSuccess callback, notice that a stream object is passed as a parameter. This is the live video stream coming from the webcam. We will need to set the src attribute of a <video> element with this stream...
  3. To do so : get a reference to a video element :
    /div[12]/div/ol/li[3]/pre, reference to undefined name 'document'
  4. Set the src attribute :
    /div[12]/div/ol/li[4]/pre, line 1, column 12: EOF expected
  5. Or if you are running a WebKit based browser :
output.src = window.webkitURL.createObjectURL(stream);

and you are done !

Now, the live video stream should be displayed in your application (at bottom right of the GUI, in our paint program)

Painting with the live images

Like we did earlier for other tools, we added a 'webcam' drawing tool in the drawintools.js file :

setOfDrawingTools.webcam = function() {
    var mousePos, previousMousePos, x, y;
    // ref to the video element that displays webcam real time content
    var video =document.getElementById('output');
    ...

    this.mousedown = function (event) {
        previousMousePos = getMousePos(paint.getFrontCanvas(), event);
        paint.started = true;
    }

    this.mousemove = function (event) {
        mousePos = getMousePos(paint.getFrontCanvas(), event);
        // Draw only if we clicked somewhere
        if (paint.started) {
            // clear the content of the front canvas
            paint.getFrontContext().clearRect(0, 0, paint.getFrontCanvas().width, 
                                              paint.getFrontCanvas().height);

            // Size and pos of the elastic rectangle with video snapshot inside
            var imageProperties = computeProperties(previousMousePos, mousePos);

            // Draw video content on front canvas
            paint.getFrontContext().drawImage(video,imageProperties.x,imageProperties.y, 
                  imageProperties.width,imageProperties.height);
        }
    }

    // Compute the coordinates of the top left corner and the size of the image drawn.
	function computeProperties(previousMousePos, mousePos){
	    var properties = {};
        properties.x = Math.min(previousMousePos.x, mousePos.x);
        properties.y = Math.min(previousMousePos.y, mousePos.y);
        properties.width = Math.abs(previousMousePos.x - mousePos.x);
        properties.height = Math.abs(previousMousePos.y - mousePos.y);
        return properties;
    }

    ...

    this.mouseup = function (event) {
        paint.started = false;
        paint.drawFrontCanvasOnMainCanvas();
        ...
    }
};

We just draw the live video stream current displayed image with a call to the drawImage method of a context object. In that cas, we draw elastic rectangles filled with a live image using the frontContext object.

Sending live images as base64 pictures

The trick here is to draw the current video frame into an offscreenCanvas of the size of the rectangle we are drawing, then get its content as a base64 image, broadcast the image with position, size, etc. to other clients...

We use the offScreenCanvas.toDataURL("image/png") method in order to get the canvas content as a png imaghe encoded in base64. WebSockets are HTTP based so we can only send text ! And base64 is a text String!

So... we added several lines in our webcam drawing tool in order to grab and send the live pictures :

setOfDrawingTools.webcam = function() {
    var mousePos, previousMousePos, x, y;
    // ref to the video element that displays webcam real time content
    var video =document.getElementById('output');

    // An off screen canvas for getting webcam data
    var offScreenCanvas= document.createElement('canvas');
    var offScreenContext = offScreenCanvas.getContext('2d');

    this.mousedown = function (event) {
        previousMousePos = getMousePos(paint.getFrontCanvas(), event);
        paint.started = true;
    }

    this.mousemove = function (event) {
        mousePos = getMousePos(paint.getFrontCanvas(), event);
        // Draw only if we clicked somewhere
        if (paint.started) {
            ...

            // Size and pos of the elastic rectangle with video snapshot inside
            var imageProperties = computeProperties(previousMousePos, mousePos);

            // Draw video content on front canvas
            paint.getFrontContext().drawImage(video,imageProperties.x,imageProperties.y, 
                                     imageProperties.width,imageProperties.height);

            // draw in the offscreen canvas a snapshot of current picture 
            // displayed in the video element
            offScreenCanvas.width = imageProperties.width;
            offScreenCanvas.height = imageProperties.height;
            offScreenContext.drawImage(video, 0,0, imageProperties.width,
                                                   imageProperties.height);
            // Get this snapshot as a base64 picture
            var imageData = offScreenCanvas.toDataURL("image/png");

           // send image data through websocket
           sendWebcamMove(imageData,imageProperties);
        }
    }

   ...

    // Send the rectangle to other clients.
    function sendWebcamMove(imageData,properties) {
         var paintCommand = {};
        paintCommand.type = 'webcamMove';
        paintCommand.imageData = imageData;
        paintCommand.x = properties.x;
        paintCommand.y = properties.y;
        paintCommand.width = properties.width;
        paintCommand.height = properties.height;
        now.distributePaintCommand(paintCommand);
    }

    this.mouseup = function (event) {
        ...
        paint.sendOrderToDrawFrontCanvasOnMainCanvas();
    }
};

Processing images received as base64 data objects

We added some lines in the paint.js file, where we process incoming paint events. Notice the way we build an image from the received data :

this.processPaintCommand = function (paintCommand) {
       ...

        switch (paintCommand.type) {
            
            case 'webcamMove' :
                // clear the content of the front canvas
                frontContext.clearRect(0, 0, frontCanvas.width, frontCanvas.height);
                // build a temporary image of the right size
                var img = new Image();
                img.width = paintCommand.width;
                img.height = paintCommand.height;

                // Listener called when the image is ready to be drawn
                img.onload = function () {
                    // draw the received picture
                    frontContext.drawImage(img, paintCommand.x, paintCommand.y);
                };

                // will call the onload listener
                img.src = paintCommand.imageData;
                break;
            ...
        }

        ...
};