Node.JS

AJAX is a useful technology for web applications that need to pull periodic updates from your server. But what do you do when you have thousands of clients all requesting data that changes real-time? A PHP+AJAX approach would choke. It is situations like these that Node.JS comes to the rescue.

Non-Blocking I/O

You dabbled in the concept of non-blocking I/O when you implemented AJAX in the previous JavaScript module. Node.JS uses this execution model at its core.

The core idea is the use of anonymous functions as callbacks for processing the results of intensive operations. Consider the pseudo-code:

// Blocking I/O Pseudo-Code

var data = database.query("select * from news");
print data[0].title;

The above pseudo-code queries a database, saves the result of the query in a variable named data, and then prints out the first title in data. The problem arises when multiple clients all want to access the server at the same time. In a blocking I/O model, a client needs to wait until all other clients are finished before it can start its process. With intensive operations and thousands of clients, this can easily crash your server!

The solution is to use non-blocking I/O. Consider this revised pseudo-code:

// Non-Blocking I/O Pseudo-Code

database.query("select * from news", function(data){
	print data[0].title;
});

The revised pseudo-code achieves the same end result as the first version. However, it allows other clients to perform tasks while this process is waiting for the database to be queried! The code occurs in two parts: first, we start the query. Second, when the query is complete, our anonymous callback function is executed. In Node.JS, callback functions are usually passed one or more parameters.

Better yet, with non-blocking I/O, you have no concurrency issues, because everything runs in just one thread. Nifty!

Installing Node.JS

You can get Node.JS from yum or apt-get, but the versions are going to be out dated. Remember that Node.JS is a rapidly changing language, so a version of Node from last year is going to be drastically different than a version of Node today.

The unofficial instructions for installing Node.JS can be found here: https://gist.github.com/isaacs/579814

There are several choices, but my favorite is probably use-nave-no-shell.sh.

Your First Node.JS Application: Hello World

Let's try running the example application given on the front page of nodejs.org:

var http = require('http');
http.createServer(function (req, res) {
	res.writeHead(200, {
		'Content-Type': 'text/plain'
	});
	res.end('Hello World\n');
}).listen(1337);
console.log('Server running at http://localhost:1337/');

Paste the example into a file named `hello.js`, and then run it by calling `node hello.js`. You should now be able to see your Hello World by visiting your server at port 1337! (Press ^C to stop the application.) Note: If you're running the code from your EC2 instance, you will need to open port 1337.

Let's walk through each line in the above example.

• var http = require('http'); enables us to use Node's built-in HTTP server. We've been spoiled that languages like PHP automatically load all of the libraries that we need. However, most functionality in Node.JS is NOT loaded by default. When you want to use any non-fundamental feature of Node, you need to require it into a variable, as this example does with the HTTP package. You should be familiar with this concept when you needed to use ArrayList and other specialty packages in Java in CSE 132.
• http.createServer(function (req, res) { ... }).listen(1337); uses the HTTP server API to listen for HTTP connections on port 1337. The anonymous callback function will be called whenever a connection is made.
• res.writeHead(200, { 'Content-Type': 'text/plain' }); specifies the Content-Type header to be "text/plain".
• res.end('Hello World\n'); writes "Hello World" to the response.

Note the use of non-blocking I/O, even in this one little example. Whenever a connection is made, the callback function will be called and passed parameters associated with that connection.

A Static Fileserver Using Node.JS

Node.JS does not automatically serve up static files in the way that an Apache/PHP does. We need to either use an extension that handles this for us, like Express, or we need to write our own code to make a simple static fileserver. The latter option is really not all that difficult, and it gives good insight into how a static fileserver works.

Installing Packages

In order for the example code below to work, you need to install the mime package. To do this, cd to the working directory where you will save your Node.JS JavaScript file and run:

$ npm install mime

npm is the package manager for Node.JS. Unlike all of the package managers we've used up to this point, npm installs dependencies locally, not globally. This means that you can have two different projects using different versions of the same package. npm will create a node_modules directory in the CWD and will install the packages into it.

Note: When using npm to install local dependencies, do not use sudo. It will screw up your file permissions.

Static Fileserver Code

The following code will make an HTTP server on port 3456 that serves up all static files in <STATIC DIRECTORY NAME> relative to app.js.

Make an attempt to understand this code. It's important that you are comfortable with the non-blocking nature, as well as how we handle file I/O. There may be quiz questions about this code.

// Require the functionality we need to use:
var http = require('http'),
	url = require('url'),
	path = require('path'),
	mime = require('mime'),
	path = require('path'),
	fs = require('fs');

// Make a simple fileserver for all of our static content.
// Everything underneath <STATIC DIRECTORY NAME> will be served.
var app = http.createServer(function(req, resp){
	var filename = path.join(__dirname, "<STATIC DIRECTORY NAME>", url.parse(req.url).pathname);
	(fs.exists || path.exists)(filename, function(exists){
		if (exists) {
			fs.readFile(filename, function(err, data){
				if (err) {
					// File exists but is not readable (permissions issue?)
					resp.writeHead(500, {
						"Content-Type": "text/plain"
					});
					resp.write("Internal server error: could not read file");
					resp.end();
					return;
				}
				
				// File exists and is readable
				var mimetype = mime.lookup(filename);
				resp.writeHead(200, {
					"Content-Type": mimetype
				});
				resp.write(data);
				resp.end();
				return;
			});
		}else{
			// File does not exist
			resp.writeHead(404, {
				"Content-Type": "text/plain"
			});
			resp.write("Requested file not found: "+filename);
			resp.end();
			return;
		}
	});
});
app.listen(3456);

Port 3456

We are running our server on port 3456 as opposed to the default web server port, which of course is 80. We therefore need to open our Node.JS application like this:

http://ec2-xx-xx-xx-xx.compute-1.amazonaws.com:3456/path.extension

However, this won't work on the first time. Why? (Try to figure this out on your own before reading the next paragraph.)

We need to open port 3456 on our Amazon EC2 instance. Remember when you opened port 80 way back in Module 2? We need to do the same sort of procedure again now.

1. Log into your EC2 management console
2. Go to "Security Groups"
3. Select your security group from Module 2
4. Under "Inbound", create a custom TCP rule for port 3456, and apply your changes.

Node.JS Security

Just because we're writing in JavaScript doesn't mean that we can forget about the security practices you've been learning in the last three months of CSE 330. XSS is still a problem, as is the use of eval, CSRF attacks, and so on.

Here is a good slideshow introducing Node.JS as well as some security vulnerabilities to look out for: http://www.slideshare.net/ASF-WS/asfws-2012-nodejs-security-old-vulnerabilities-in-new-dresses-par-sven-vetsch

Node.JS Conventions

There ought to be a standard way to save metadata about your app, right? Well, npm introduces a straightforward way to do this in the form of a package.json file.

To make your npm file with a step-by-step prompt, run in your project root directory:

$ npm init

One nifty feature of package.json is the way it automatically handles dependencies. For example, you can put the following lines in your package.json:

  "dependencies": {
    "mime": "~1.2.11",
    "socket.io": "~0.9.16",
  },

Then, back on the command line, when you run

$ npm install

it automatically installs the dependencies for you! No more headaches of trying to figure out which modules you need to install; all that information is saved for you in your code base.