cujo resource loader (CommonJS compatible AMD loader)
Switch branches/tags
Pull request Compare This branch is 835 commits behind cujojs:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

curl (cujo resource loader)

version 0.5.1

What's New?

  • 0.5.1:
    • fixes to domReady! in the compiled dist/ versions and
    • fixes for commonjs compatibility caused by google-closure in dist/ versions
    • support for parent module ids (../sibling-of-parent)
  • 0.5:
    • dojo 1.6 support has been moved to separate module (curl/dojo16Compat)
    • curl/domReady now returns a callback function (not a promise)
    • new async! plugin to allow a module to defer definition
    • new css! plugin that inlines css into javascript when used with cram
    • cram (AMD builder) support (css! and async! plugins)
    • require is no longer an alias for curl unless you set the apiName config param to "require"
    • configuration parameters for plugins are now defined in a sub-object of the main config object: { css: { cssOption: true } }
  • earlier:
    • Fixed !order option for js! plugin in non-Firefox browsers (0.4.3)
    • Fixed the compiled version in 0.4.2 (dist/ folder)
    • Several fixes to path and package mapping were made in 0.4.1


  • configuration options per package
  • finish i18n plugin (eta: August 1)
  • create dojo 1.6 tests
  • document plugin configuration options and how to use plugins
  • notes about using JSONP (it works for objects, arrays, functions, numbers and strings! use ?callback=define)
  • use CommonJS file structure (lib/ instead of src/)

What is curl.js?

curl.js is a small and very fast AMD-compliant asynchronous loader. Size: 4.6KB (2.5KB gzipped) using Google's Closure Compiler.

If you'd like to use curl.js for non-AMD modules (ordinary javascript files), you'll want to use the version with the js! plugin built in. You may also want to build-in the domReady module. The combined curl+js+domReady loader is still only 6.4KB (3KB gzipped).

What the heck is "cujo"? cujo.js is a web app development platform. See the bottom of this file for more info.

What is "cram"? cram (cujo resource assembler) is the build tool companion to curl.js. You use cram to compile all of your modules into a small number of javascript files which are loaded much faster into the browsers.

Features at a glance:

  • Loads CommonJS AMD-formatted javascript modules in parallel (fast!)
  • Loads CommonJS Modules (v1.1 when wrapped in a define()) (fast!)
  • Loads CommonJS Packages (v1.1 modules wrapped in a define()) (fast!)
  • Loads non-AMD javascript files in parallel, too (fast! via js! plugin)
  • Loads CSS files and text files in parallel (fast! via plugins)
  • Waits for dependencies (js, css, text, etc) before executing javascript
  • Waits for domReady, if/when desired
  • Allows for virtually limitless combinations of files and dependencies
  • Tested with Chrome, FF3+, Safari 3.2+, IE6-8, Opera 9.5+

Oh, did we mention? It's fast! It's even faster than the leading non-AMD script loaders.

API at a glance

curl(['dep1', 'dep2', 'dep3' /* etc */], callback);

Loads dependencies and the executes callback.

  • ['dep1', 'dep2', 'dep3']: Module names or plugin-prefixed resource files
  • callback: Function to receive modules or resources. This is where you'd typically start up your app.

curl(['dep1', 'dep2', 'dep3' /* etc */])
	.then(callback, errorback);

Promises-based API for executing callbacks.

  • ['dep1', 'dep2', 'dep3']: Module names or plugin-prefixed resource files
  • callback: Function to receive modules or resources
  • errorback: Function to call if an exception occurred while loading

curl(config, ['dep1', 'dep2', 'dep3' /* etc */], callback);

Specify configuration options, load dependencies, and execute callback.

  • config: Object containing curl configuration options (paths, etc.)
  • ['dep1', 'dep2', 'dep3']: Module names or plugin-prefixed resource files
  • callback: Function to receive modules or resources

curl(['domReady!', 'dep2', 'dep3' /* etc */])
curl(['dep1', 'dep2', 'domReady!' /* etc */], function (dep1, dep2) {
	// do something here

Executes the callback when the dom is ready for manipulation AND all dependencies have loaded.

  • callback: No parameters except the domReady object
  • errorback: Function to call if an exception occurred while loading

curl(['domReady!', 'js!nonAMD.js!order', 'js!another.js!order']), function () {
	/* do something cool here */

Executes the function when the non-AMD javascript files are loaded and the dom is ready. The another.js file will wait for the nonAMD.js file before executing.

	.next(['dep1', 'dep2', 'dep3'], function (dep1, dep2, dep3) {
		// do something before the dom is ready
		function () {
			// do something after the dom is ready
		function (ex) {
			// show an error to the user

Executes callbacks in stages using .next(deps, callback).

curl = {
	baseUrl: '/path/to/my/js',
	pluginPath: 'for/some/reason/plugins/r/here',
	paths: {
		curl: 'curl/src/curl',
		cssx: 'cssx/src/cssx'
		my: '../../my-lib/'
	apiName: 'someOtherName'

If called before the <script> that loads curl.js, configures curl.js. All of the configuration parameters are optional. curl.js tries to do something sensible in their absence. :)

  • baseUrl: the root folder to find all modules, default is the document's folder
  • paths: a mapping of module paths to relative paths (from baseUrl)
  • pluginPath: the place to find plugins when they are specified without a path (e.g. "css!myCssFile" vs. "cssx/css!myCssFile") and there is no paths mapping that applies.
  • apiName: an alternate name to curl and require for curl.js's global variable

define(['dep1', 'dep2', 'dep3' /* etc */], definition);
define(['dep1', 'dep2', 'dep3' /* etc */], module);
define(['dep1', 'dep2', 'dep3' /* etc */], promise);
define(name, ['dep1', 'dep2', 'dep3' /* etc */], definition);
define(name, ['dep1', 'dep2', 'dep3' /* etc */], module);
define(name, ['dep1', 'dep2', 'dep3' /* etc */], promise);
define(name, module);
define(name, promise);

Defines a module per the CommonJS AMD proposed specification.

  • ['dep1', 'dep2', 'dep3']: Module names or plugin-prefixed resource files. Dependencies may be named 'require', 'exports', or 'module' and will behave as defined in the CommonJS Modules 1.1 proposal.
  • definition: Function called to define the module
  • module: Any javascript object, function, constructor, or primitive
  • promise: Object compatible with CommonJS Promises/A. Useful for further deferring resolution of the module.
  • name: String used to name a module (not necessary nor recommended)

Very Simple Example


	// configure curl
	curl = {
		paths: {
			cssx: 'cssx/src/cssx/',
			stuff: 'my/stuff/

<script src="../js/curl.js" type="text/javascript"></script>
<script type="text/javascript">

		// fetch all of these resources ("dependencies")
			'stuff/three', // an AMD module
			'cssx/css!stuff/base', // a css file
			'i18n!stuff/nls/strings', // a translation file
			'text!stuff/template.html', // an html template
	// when they are loaded
		// execute this callback, passing all dependencies as params
		function (three, link, strings, template) {
			var body = document.body;
			if (body) {
				body.appendChild(document.createTextNode('three == ' + three.toString() + ' '));
				body.appendChild(document.createElement('div')).innerHTML = template;
		// execute this callback if there was a problem
		function (ex) {
			var msg = 'OH SNAP: ' + ex.message;


The file structure for this example would look as follows:


What is an asynchronous loader?

Web apps, especially large ones, require many modules and resources. Most of these modules and resources need to be loaded at page load, but some may be loaded later, either in the background or "just in time". They also need to be loaded as quickly as possible.

The traditional way to load javascript modules is via a <SCRIPT> element in an HTML page. Similarly, CSS files are loaded via a <LINK> element, and text resources are either loaded in the page or via XHR calls.

The problem with <SCRIPT> and <LINK> elements is that a browser must execute them sequentially since it has no idea if one may depend on another. It just assumes the developer has placed them in the correct order and that there are dependencies. (The term "synchronous loading" is used to describe this process since the elements are executed in a single timeline.)

If there are no dependencies between two files, loading them sequentially is a waste of time. These files could be loaded and executed in parallel (i.e at the same time).

An asynchronous loader does just that: it loads javascript files (and other types of files) in parallel whenever possible.

curl.js has lots of company. Other async loaders include LABjs, Steal.js, yepnope.js, $script.js, the Backdraft loader (bdLoad), and RequireJS.

(a more complete list)

What is AMD?

Asynchronous Module Definition is the CommonJS proposed standard for javascript modules that can be loaded by asynchronous loaders. It defines a simple API that developers can use to write their javascript modules so that they may be loaded by any AMD-compliant loader.

CommonJS AMD Proposal

The AMD proposal follows the CommonJS Modules proposal as much as possible. Because of the way browsers load and evaluate scripts, AMD can't follow it completely without causing significant processing overhead. Instead, AMD allows us to place a lightweight wrapper around javascript modules to help work around the shortcomings.

Ultimately, both proposals (AMD and Modules 1.1) are in preparation for an official javascript modules specification and eventual implementation in browsers.

If you don't want to wait for official javascript modules, then don't. The future is now. AMD works now -- and it's awesome.

AMD's API focuses on two globally-available functions: require() and define(). require() specifies a list of dependent modules or resources that must be loaded before running a set of code. This code resides in a callback function that is executed asynchronously, i.e. it runs later, not in the current "thread". Specifically, it executes when all of the dependencies are loaded and ready.

Actually, the proposal says that the public require() function could have a different name -- or could even be implemented differently. To keep confusion to a minimum curl.js uses curl() for the public API. You may rename this API back to require() by supplying the apiName config param (apiName: "require").

It's only important that the define() method be consistent. This is the method that tells the loader what modules have been loaded by a script. define() also specifies a list of dependencies and a callback function that defines and/or creates the resource when the dependencies are ready. Optionally, define() also takes a name parameter, but this is mainly for build tools and optimizers.

Inside the define(), the require() method acts like other AMD loaders.

AMD's API also helps code reuse by providing compatibility with CommonJS server modules. AMD-compliant loaders support the same require() syntax and argument signatures as server-side javascript (ssjs) modules.

Not all async loaders are AMD-compliant. Of the list above, only the following are AMD-compliant:



backdraft loader

The beauty of AMD loaders is their ability to remove the drudgery of manually managing dependencies. Since all dependencies are listed within the modules, the loader will ensure that everything is loaded into the browser -- and in the right order.

What makes curl different from other AMD loaders?

curl.js is much smaller than other AMD loaders. Less than 1/2 the size of the others in the list above. It's able to achieve this via a Promises-based design. (Promises are another CommonJS proposed standard.)

curl.js communicates with it's plugins via Promises, rather than a simple callback function. This allows proactive error handling, rather than detecting problems via a timeout, which can be tricky to set correctly. curl does this in a backwards-compatible way so AMD-compliant plugins will still work in curl.

curl.js will also return a promise from curl() calls. This allows you to write code like this:

	function success (A, B) {
		// load myApp here!
	function failure (ex) {
		alert('myApp didn't load. reason: ' + ex.message);

(When using require as a dependency, it does not return a promise. This is so that 100% CommonJS compliance is assured inside modules.)

Can curl.js work with non-AMD javascript files?

Yes, but why would you? Once you start using AMD, you'll never go back! :)

You may use non-AMD javascript files by specifying the js! plugin prefix like this:

	function () {
		/* do something with your plain, boring javascript files */
	function () {
		/* do something if any fail to load */

The !order suffix instructs curl.js to wait for previous scripts to execute before executing the current script. All scripts download in parallel, though, unless you specify jsPrefetch:false in the config. Be sure to have proper cache headers set if you plan to use jsPrefetch:true or scripts will get downloaded twice in browsers that don't support async=false (only Firefox 4 supports it so far).

Can curl.js load non-javascript files?

Yes, curl.js follows the CommonJS Loader Plugin specification, so you can use any compatible plugin. The following plugins are included:

js! -- loads non-AMD javascript files

text! -- loads text files

You can also load css files by using the AMD plugin at the following repo:

The following plugins are in progress:

i18n! -- loads text strings and other locale-specific constants

cssx! -- loads and automatically shims css files for older browsers

How are modules loaded?

curl.js uses <script> element injection rather than XHR. This allows curl.js to load cross-domain scripts as well as local scripts.

To find scripts and other resources, curl uses module names. A module name looks just like a file path, but typically without the file extension. If a module requires a plugin in order to load correctly, it will have a prefix delimited by a "!" and will also often have a file extension when a plugin may load different types of files.

Some examples of module names:

By default, curl.js will look in the same folder as the current document's location. For instance, if your web page is located at http://my-domain/apps/myApp.html, curl.js will look for the JsonRest module at http://my-domain/apps/dojo/store/JsonRest.js.

You can tell curl.js to find modules in other locations by specifying a baseUrl or individual paths for each of your libraries. For example, if you specify a baseUrl of /resources/ and the following paths:

paths: {
	dojo: "third-party/dojo",
	css: "third-party/cssx/css",
	my: "my-cool-app-v1.3",
	"my/lib/js": "old-js-libs"

Then the modules listed above will be sought in the following locations:

  • /resources/third-party/dojo/store/JsonRest.js
  • /resources/my-cool-app-v1.3/lib/string/format.js
  • /resources/old-js-libs/plain-old-js.js
  • /resources/my-cool-app-v1.3/styles/reset.css
  • http://some-cdn/uber/module.js

Note: you will need to create a path to curl's plugins and other modules if the curl folder isn't directly under the same folder as your web page. curl.js uses the same mechanism to find its own modules.

TODO: explain the pluginPath configuration parameter.

What are AMD plugins?

AMD supports the notion of plugins. Plugins are AMD modules that can be used to load javascript modules -- or other types of resources. curl comes with several plugins already, including a text plugin (for templates or other text resources), a css plugin, a sync plugin (for loading modules synchronously), and a debug plugin (for collecting and logging details of the inner workings of curl).

Plugins are designated by a prefix on the name of the module or resource to be loaded. They are delineated by a ! symbol. The following example shows the use of some plugins:

	function (templateString, cssLinkNode) {
		// do something with the template and css here

Since plugins are just AMD modules, they would typically be referenced using their fully-pathed names. curl provides a pluginPath configuration option that allows you to specify the folder where [most of] your plugins reside so you don't have to specify their full paths. This also helps with compatibility with other AMD loaders that assume that certain plugins are bundled and internally mapped.

If one or more of your plugins does not reside in the folder specified by the pluginPath config option, you can use its full path or you can specify a path for it in curl's paths config object.

// example of a fully-pathed plugin under the cssx folder
define(['/css!myCssFile'], function (cssxDef) {
	// do some awesome css stuff here

Plugins can also have configuration options. Global options can be specified on curl's configuration object. Options can also be supplied to plugins via suffixes. Suffixes are also delineated by the ! symbol. Here's an example of a plugin using options:

// don't try to repair IE6-8 opacity issues in my css file
define(['css!myCssFile!ignore:opacity'], function (cssxDef) {
	// do some awesome css stuff here

How do I use curl.js?

  1. Optional: Learn about AMD-formatted javascript modules if you don't already know how.
  2. Clone or download curl to your local machine or server.
  3. Figure out the baseUrl and paths configuration that makes sense for your project.
  4. Check out the "API at a glance" section above to figure out which loading methodology you want to use.
  5. Study the "Very Simple Example" and some of the test files.
  6. Try it on your own files.

Too Many Modules!

I have dozens (or hundreds) of modules. Even with parallel loading, the performance sucks! What can I do about that?

True! No parallel loader can lessen the latency required to create an HTTP connection. If you have dozens or hundreds of files to download, it's going to take time to initiate each of the connections.

However, there are tools to that are designed to fix this problem! There are builders and compilers. dojo users are probably already familiar with dojo's build tool and optimizer. RequireJS comes with a build tool and Google's Closure compiler.

The build tool is used to concatenate several modules (and/or resources) into just a few files. It does this by following the dependency chain specified in the define() and require() calls. You can specify which top-level modules or resources are in each file and the build tool finds the rest.

After the build tool creates the concatenated files, the files can be passed into a compiler (also called a shrinker or compressor).

We're writing curl to be compatible with RequireJS's build tool, but there's also another cujo project in the pipeline: cram. Cram is the Cujo Resource Assembler. cram will be ready by mid 2011, so use another build tool or a custom shell script in the mean time.

CommonJS Package Support

cujo.js supports the CommonJS Packages 1.1 specification. Packages are defined in the packages configuration parameter:

cujo = {
	baseUrl: 'path/to/js',
	packages: {
		'my-package': {
			path: 'path/to/my-package',
			main: 'main/main-module-file',
			lib: 'location/of/other/modules'

The path property describes where to find the package in relation to the baseUrl parameter. The main and lib properties describe where to find modules inside the package. The main property gives the relative path to the pacage's main module. The lib property reflects the path to all other modules in the package.

The main module is always executed before any other modules are executed in the package. Essentially, the main module becomes an automatic dependency.

In the example above, the main module of the package can be obtained as follows

curl(['my-package'], callback);

and will be fetched from the following path:


Some other file in the package would be obtained as follows:

curl(['my-package/other-module'], callback);

and will be fetched from the following path:


What is cujo?

cujo.js is a web app development platform. It employs MVC, IOC, AMD and lots of other TLAs. :) curl.js is one of the many micro-libs we're pulling out of cujo.js. Our goal is to make the advanced concepts in cujo.js more palatable by breaking them down into easier-to-grok chunks. Other cujo.js libs include:

canhaz: a project and code bootstrapping tool that will save you tons of typing. wire: an application bootstrap, configuration, and assembly tool based on the principles of Inversion of Control, and Dependency Injection. cram: a [forthcoming] javascript compressor, concatenator, and optimizer meant to be used with curl.js


Many thanks to Bryan Forbes (@bryanforbes) for helping to clean up my code and for making cujo's domReady much more robust. More about Bryan:

Kudos also to James Burke (@jrburke) who instigated the CommonJS AMD proposal and paved the way to create AMD-style loaders. More about James:

Shout out to Kris Zyp (@kriszyp) for excellent ideas and feedback and to Kyle Simpson (@getify) who is inarguably the godfather of javascript loading.