IMD is an implementation of the AMD specification that performs absolutely no loading. The primary goal of it is to play nice with HTML Imports, but it should work well with any code loader that doesn't mandate a particular module registry.
JavaScript module systems generally perform three different tasks:
- Definition Allow module authors to define an encapsulated unit of code, including what other modules it depends on.
- Resolving Given a module reference, resolve it to an actual module instance. This typically involves a central module registry that maps module names to instances.
- Loading Given a module reference that is not yet loaded into the registry, load it from some source, usually a URL (via an XHR).
IMD only performs the first two tasks, in a way that's fully compliant with the AMD spec.
This is the minimal module system needed if you're already loaded your script resources via HTML Imports. HTML Imports handily takes care of loading resources and their transitive dependencies, de-duplicating imports, and executing scripts in the correct order.
Together, IMD and HTML Imports form a really great module system. Some of the benefits over RequireJS include:
- IMD is extremely small. With no loading code, IMD is almost entirely just a AMD-compiant registry in just 150 lines of code (with comments).
- A module defined in an HTML Import can be sure its non-script resources, like CSS and HTML templates, have been loaded as well.
- Browsers with native HTML Import support can aggressively pre-scan and
pre-load HTML Imports by reading the
<link>
tags. - Servers that support HTTP/2 Push can easily read
<link>
tags and preemptively push dependencies, resulting in very fast load times.
They're great. Really, we can't wait for them to be implemented natively in browsers. And until then, compilers and loaders like SystemJS and Babel do allow developers to use ES6 modules now, if they're willing to deal with a build step.
Not every project can or wishes to do so, and HTML Imports is not a JavaScript module definition system, so IMD is an extremely lightweight way to use modules on top of imports.
When the ES6 module loader spec is further along, we will look into ways of integrating the two so that JavaScript can import HTML "modules" and vice-versa. We're very excited about the possibilities for interop.
IMD is distributed as an HTML Import-able HTML file, naturally. Install and uselike this:
Install:
> bower install PolymerLabs/IMD --save
Import:
<link rel="import" href="../imd/imd.html">
We recommend that HTML files that require or define do not directly import
IMD, but rather let the main page import IMD, usually as the very first import.
This is so that an application that really requires RequireJS to load modules
(this can be true when loading pure JS modules with no corresponding <script>
tags), can setup RequireJS instead of IMD. Since IMD is fully AMD compliant,
all modules defined in HTML imports will work just fine.
Modules are defined exactly as in other AMD systems like RequireJS: Public modules are defined by name:
Here's the definition of a mythical module, 'squidbits', that depnds on the modules 'tentacles.html', and 'ink':
define('squidbits', ['./tentacles.html', 'ink'], function(tentacles, ink) {
return {tentacles: tentacles, ink: ink, squidbits: true};
});
Now, since IMD is targeted at projects already using HTML Imports, it's likely that the module will be defined inside an HTML file, like so:
<link rel="import" href="./tentacles.html">
<script src="ink.js"></script>
<script>
define('squidbits', ['./tentacles.html', 'ink'], function(tentacles, ink) {
return {tentacles: tentacles, ink: ink, squidbits: true};
});
</script>
Notice the <link>
and <script>
tags. These tell the browser to load
'tentacles.html' and 'ink.js'. Because scripts are blocked until imports and
other scripts have loaded and run, and 'tentacles.html' and 'ink.js' define
modules, squidbits
's dependencies are guaranteed to be loaded and registered
when the inline script runs.
Note that we're assuming that ink.js
defines a ink
module here, and
tentacles.html
defines an anonymous module.
Private (relative) module names are inferred from the current import:
tentacles.html
:
<script>
define(function() {
return /* module contents */;
})
</script>
This module is available at a relative URL, as shown in the above squidbits
example. There can be only one anonymous module per HTML document.
If you want to use an existing anonymous module (say, something defined
following the UMD pattern), give it a name with
the as
attribute on a <script>
tag:
<script src="thingo.js" as="thingy"></script>
Imagine:
foo.html
:
<template id="bar">...</template>
You could import a document:
index.html
:
define(['foo.html'], function(foo) {
// foo is a document
let bar = foo.querySelector('#bar');
document.appendChild(document.importNode(bar.content));
})
Or a node directly:
index.html
:
define(['foo.html#bar'], function(bar) {
// bar is a HTMLTemplate node!
document.appendChild(document.importNode(bar.content));
})
The one downside of IMD relative to RequireJS is that dependencies often appear
twice in a file: once as a HTML Import and one in the define()
call. We could
conceivably infer the dependency list from the imports so that there's no
duplication.
There is a lot to explore with ES6 module loading and HTML Imports. The loading and ordering semantics appear to be fully compatible, and given a configurable enough loader we should be able to let ES6 modules import HTML nodes and vice-versa.