Nerdy Weekly

Thoughts on programming and other stuff that matters

Setting up your development environment for a node-webkit project

Node-webkit lets you write cross-platform (Mac + Linux + Winbloze) desktop applications using HTML5 and nodejs. That’s a fancy way of saying “a webkit wrapper that also gives you filesystem access, disables same-origin policy and has nodejs embedded”. It’s the lazy web developers’ dream come true. Who needs Qt, GTK, or any other legitimate battle-tested, performant cross-platform GUI toolkit when you can slap in Bootstrap/Foundation with some hip JavaScript framework instead?

Joking aside, this is an interesting take on desktop development. Firefox OS is on the horizon, and even Ubuntu is pushing the “HTML5 apps as first-class citizens” paradigm. Why not give it a try?

If you already have experience in client-side web development tools, you can just clone node-webkit-hipster-seed and digest the code. That’s a project skeleton that integrates all kinds of stuff: Jade/Coffescript/LESS automatic compiler, node-webkit, grunt tasks…

This tutorial is like a stripped down version of that. In the end we’ll have:

  • Automatic app reload when source code changes with livereload
  • Single command to build binaries for mac + linux + win with grunt
  • That’s it, really

Install node-webkit on your machine

Follow the README on node-webkit’s GitHub page to download a precompiled nw binary for your platform. If you’re using Arch Linux, you’re in luck since there’s already an AUR package:

$ yaourt -S node-webkit

If you’re on Ubuntu or some other repo and you get some error about, read this for a hotfix.

The rest of this tutorial will assume that you have nw accessible as an executable in your $PATH.

Running an app

First, take a look at nw’s quickstart guide. We’ll make a somewhat different structure, allowing the dist directory to store our binary releases:

├── app/
│  ├── css/
│  ├── js/
│  ├── index.html
│  └── package.json
└── dist/

package.json stores the information that nw requires. Its content goes like this:

  "name": "your-project-name",
  "version": "0.0.1",
  "main": "index.html",
  "window": {
      "toolbar": true

window.toolbar is true by default so we don’t actually need it. Switch it to false if you want to hide the browser-like address bar.

To embed css/js files, don’t use absolute urls. Use relative ones like this:

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

There’s no point in using CDNs because we’re distributing the whole app with its assets as a one-time download. Things like automatic css/js minification or concatination aren’t needed either. For simplicity’s sake, we’ll just download a minified version of whatever js/css library that we need and stuff them into css or js dir. For example: (dist directory omitted)

├── css
│   └── bootstrap.min.css
├── js
│   ├── bootstrap.min.js
│   ├── knockout.min.js
│   ├── jquery.min.js
│   └── app.js
├── index.html
└── package.json

If you’re experienced in front-end web development tools, feel free to go wild with bower, grunt/gulp/whatever. Again, check out the node-webkit-hipster-seed project if you know what you’re doing.

You can now test run your app with the nw <directory> command. In our case: nw app. Notice the weird url, which is why we can’t use absolute urls in the first place:


Once you’ve packaged your app to a single executable, the url will be something like this:


But let’s not get ahead of ourselves. Let’s solve the most obvious dev issue first:

Automatic reload

Sure enough, at first glance your app is just another html page. You may be tempted to run some simple http server and open localhost in Google Chrome (python2 -m SimpleHTTPServer 8080 anyone?). There are tons of ways to make Google Chrome automatically reload a page, right?

But then, the true strength of node-webkit is the ability to use nodejs modules, which Google Chrome certainly doesn’t offer. Therefore, the only way to check out how the app really works is using nw app.

We’ll use livereload to make automatic reloading possible. The idea is quite simple: we fire off a livereload daemon that watches for any change in our app/ directory. In our app, we embed a certain piece of javascript that connects to that livereload daemon and refreshes the page whenever a “change” event is broadcast.

There are many livereload daemon implementations. Considering the fact that most of us developers have python and pip installed, let’s go with the livereload pip package (it’s only compatible with python2, by the way). If you’re on Ubuntu and don’t know what I’m talking about:

$ sudo apt-get install python-pip
$ sudo pip install livereload

There are ruby/javascript implementations too. Google them if you prefer those things.

Either way, we can now fire off a livereload server:

# I don't know why but seems like the python implementation doesn't work
# when I type `livereload app`. Weird.
$ cd app
$ livereload .

Now how do we inject the livereload javascript? On Google Chrome there is an official livereload plugin, but we’re using node-webkit so that’s not possible. No problem! The livereload daemon we fired off earlier is actually a web server which also serves the necessary livereload client JavaScript snippet too. Simply embed it to your index.html:

<script src="http://localhost:35729/livereload.js"></script>

Fire off the app with nw app again and you’ll have automatic reloading. Cool eh?

Another problem: we only want livereload in our development version, not in the released app. Let’s modify our javascript snippet in index.html to only load livereload when a certain environment variable is set to 1:

// Load livereload if in dev environment
if (process.env.NW_DEV_MY_AWESOME_PROJECT == 1) {
    var script = document.createElement('script');
    script.type = 'text/javascript';
    script.src = 'http://localhost:35729/livereload.js';

Now to start the app:

$ nw app

Simple cross-platform build command

To be honest, you can manually write shell scripts to build for each platform. Check out this wiki article if you prefer the do-it-yourself way.

But if you’re lazy (like me) and don’t have a problem using nodejs/grunt, just use the excellent grunt-node-webkit-builder. Again, for newcomers using Ubuntu:

$ sudo add-apt-repository ppa:chris-lea/node.js
$ sudo apt-get update
$ sudo apt-get install python-software-properties python g++ make nodejs
$ sudo npm install -g grunt-cli
$ # cd to your project root (the one containing app/ and dist/)

Create package.json and Gruntfile.js in project root:

├── app/
│   ├── css/
│   ├── js/
│   ├── index.html
│   └── package.json
├── dist/
├── Gruntfile.js
└── package.json


  "name": "whatever",
  "version": "0.0.1",
  "description": "Whatever",
  "author": "Bui Thanh Nhan"


module.exports = function(grunt) {
        pkg: grunt.file.readJSON('package.json'),
        nodewebkit: {
            options: {
                build_dir: './dist',
                // choose what platforms to compile for here
                mac: true,
                win: true,
                linux32: true,
                linux64: true
            src: ['./app/**/*']

    grunt.registerTask('default', ['nodewebkit']);


$ npm install grunt grunt-node-webkit-builder --save-dev
$ grunt

The first time will be slow because grunt will download precompiled nw binaries for all supported platforms, which will be stored in dist/cache/. From now you can compile for mac + linux + win with a simple grunt command. The compiled binaries will be stored in dist/releases/.

Congratulations! You now know how to use yet another weird stack born out of the HTML5 craze that isn’t guaranteed to still be alive the next year (or even next month). For extra credit, use it to piss off long-time UNIX Philosophy believers. The suckless community would be a good place to start. ;)