As our application grows bigger, we want to split it into multiple files, so called “modules”. A module may contain a class or a library of functions for a specific purpose.
For a long time, JavaScript existed without a language-level module syntax. That wasn’t a problem, because initially scripts were small and simple, so there was no need.
But eventually scripts became more and more complex, so the community invented a variety of ways to organize code into modules, special libraries to load modules on demand.
To name some (for historical reasons):
- AMD – one of the most ancient module systems, initially implemented by the library require.js.
- CommonJS – the module system created for Node.js server.
- UMD – one more module system, suggested as a universal one, compatible with AMD and CommonJS.
Now these all slowly became a part of history, but we still can find them in old scripts.
The language-level module system appeared in the standard in 2015, gradually evolved since then, and is now supported by all major browsers and in Node.js. So we’ll study the modern JavaScript modules from now on.
What is a module?
A module is just a file. One script is one module. As simple as that.
Modules can load each other and use special directives export
and import
to interchange functionality, call functions of one module from another one:
export
keyword labels variables and functions that should be accessible from outside the current module.import
allows the import of functionality from other modules.
For instance, if we have a file sayHi.js
exporting a function:
// 📁 sayHi.js
export function sayHi(user) {
alert(`Hello, ${user}!`);
}
…Then another file may import and use it:
// 📁 main.js
import {sayHi} from './sayHi.js';
alert(sayHi); // function...
sayHi('John'); // Hello, John!
The import
directive loads the module by path ./sayHi.js
relative to the current file, and assigns exported function sayHi
to the corresponding variable.
Let’s run the example in-browser.
As modules support special keywords and features, we must tell the browser that a script should be treated as a module, by using the attribute <script type="module">
.
Like this:
export function sayHi(user) {
return `Hello, ${user}!`;
}
<!doctype html>
<script type="module">
import {sayHi} from './say.js';
document.body.innerHTML = sayHi('John');
</script>
The browser automatically fetches and evaluates the imported module (and its imports if needed), and then runs the script.
If you try to open a web-page locally, via file://
protocol, you’ll find that import/export
directives don’t work. Use a local web-server, such as static-server or use the “live server” capability of your editor, such as VS Code Live Server Extension to test modules.
Core module features
What’s different in modules, compared to “regular” scripts?
There are core features, valid both for browser and server-side JavaScript.
Always “use strict”
Modules always work in strict mode. E.g. assigning to an undeclared variable will give an error.
<script type="module">
a = 5; // error
</script>
Module-level scope
Each module has its own top-level scope. In other words, top-level variables and functions from a module are not seen in other scripts.
In the example below, two scripts are imported, and hello.js
tries to use user
variable declared in user.js
. It fails, because it’s a separate module (you’ll see the error in the console):
alert(user); // no such variable (each module has independent variables)
let user = "John";
<!doctype html>
<script type="module" src="user.js"></script>
<script type="module" src="hello.js"></script>
Modules should export
what they want to be accessible from outside and import
what they need.
user.js
should export theuser
variable.hello.js
should import it fromuser.js
module.
In other words, with modules we use import/export instead of relying on global variables.
This is the correct variant:
import {user} from './user.js';
document.body.innerHTML = user; // John
export let user = "John";
<!doctype html>
<script type="module" src="hello.js"></script>
In the browser, if we talk about HTML pages, independent top-level scope also exists for each <script type="module">
.
Here are two scripts on the same page, both type="module"
. They don’t see each other’s top-level variables:
<script type="module">
// The variable is only visible in this module script
let user = "John";
</script>
<script type="module">
alert(user); // Error: user is not defined
</script>
In the browser, we can make a variable window-level global by explicitly assigning it to a window
property, e.g. window.user = "John"
.
Then all scripts will see it, both with type="module"
and without it.
That said, making such global variables is frowned upon. Please try to avoid them.
A module code is evaluated only the first time when imported
If the same module is imported into multiple other modules, its code is executed only once, upon the first import. Then its exports are given to all further importers.
The one-time evaluation has important consequences, that we should be aware of.
Let’s see a couple of examples.
First, if executing a module code brings side-effects, like showing a message, then importing it multiple times will trigger it only once – the first time:
// 📁 alert.js
alert("Module is evaluated!");
// Import the same module from different files
// 📁 1.js
import `./alert.js`; // Module is evaluated!
// 📁 2.js
import `./alert.js`; // (shows nothing)
The second import shows nothing, because the module has already been evaluated.
There’s a rule: top-level module code should be used for initialization, creation of module-specific internal data structures. If we need to make something callable multiple times – we should export it as a function, like we did with sayHi
above.
Now, let’s consider a deeper example.
Let’s say, a module exports an object:
// 📁 admin.js
export let admin = {
name: "John"
};
If this module is imported from multiple files, the module is only evaluated the first time, admin
object is created, and then passed to all further importers.
All importers get exactly the one and only admin
object:
// 📁 1.js
import {admin} from './admin.js';
admin.name = "Pete";
// 📁 2.js
import {admin} from './admin.js';
alert(admin.name); // Pete
// Both 1.js and 2.js reference the same admin object
// Changes made in 1.js are visible in 2.js
As you can see, when 1.js
changes the name
property in the imported admin
, then 2.js
can see the new admin.name
.
That’s exactly because the module is executed only once. Exports are generated, and then they are shared between importers, so if something changes the admin
object, other importers will see that.
Such behavior is actually very convenient, because it allows us to configure modules.
In other words, a module can provide a generic functionality that needs a setup. E.g. authentication needs credentials. Then it can export a configuration object expecting the outer code to assign to it.
Here’s the classical pattern:
- A module exports some means of configuration, e.g. a configuration object.
- On the first import we initialize it, write to its properties. The top-level application script may do that.
- Further imports use the module.
For instance, the admin.js
module may provide certain functionality (e.g. authentication), but expect the credentials to come into the config
object from outside:
// 📁 admin.js
export let config = { };
export function sayHi() {
alert(`Ready to serve, ${config.user}!`);
}
Here, admin.js
exports the config
object (initially empty, but may have default properties too).
Then in init.js
, the first script of our app, we import config
from it and set config.user
:
// 📁 init.js
import {config} from './admin.js';
config.user = "Pete";
…Now the module admin.js
is configured.
Further importers can call it, and it correctly shows the current user:
// 📁 another.js
import {sayHi} from './admin.js';
sayHi(); // Ready to serve, Pete!
import.meta
The object import.meta
contains the information about the current module.
Its content depends on the environment. In the browser, it contains the URL of the script, or a current webpage URL if inside HTML:
<script type="module">
alert(import.meta.url); // script URL
// for an inline script - the URL of the current HTML-page
</script>
In a module, “this” is undefined
That’s kind of a minor feature, but for completeness we should mention it.
In a module, top-level this
is undefined.
Compare it to non-module scripts, where this
is a global object:
<script>
alert(this); // window
</script>
<script type="module">
alert(this); // undefined
</script>
Browser-specific features
There are also several browser-specific differences of scripts with type="module"
compared to regular ones.
You may want to skip this section for now if you’re reading for the first time, or if you don’t use JavaScript in a browser.
Module scripts are deferred
Module scripts are always deferred, same effect as defer
attribute (described in the chapter Scripts: async, defer), for both external and inline scripts.
In other words:
- downloading external module scripts
<script type="module" src="...">
doesn’t block HTML processing, they load in parallel with other resources. - module scripts wait until the HTML document is fully ready (even if they are tiny and load faster than HTML), and then run.
- relative order of scripts is maintained: scripts that go first in the document, execute first.
As a side effect, module scripts always “see” the fully loaded HTML-page, including HTML elements below them.
For instance:
<script type="module">
alert(typeof button); // object: the script can 'see' the button below
// as modules are deferred, the script runs after the whole page is loaded
</script>
Compare to regular script below:
<script>
alert(typeof button); // button is undefined, the script can't see elements below
// regular scripts run immediately, before the rest of the page is processed
</script>
<button id="button">Button</button>
Please note: the second script actually runs before the first! So we’ll see undefined
first, and then object
.
That’s because modules are deferred, so we wait for the document to be processed. The regular script runs immediately, so we see its output first.
When using modules, we should be aware that the HTML page shows up as it loads, and JavaScript modules run after that, so the user may see the page before the JavaScript application is ready. Some functionality may not work yet. We should put “loading indicators”, or otherwise ensure that the visitor won’t be confused by that.
Async works on inline scripts
For non-module scripts, the async
attribute only works on external scripts. Async scripts run immediately when ready, independently of other scripts or the HTML document.
For module scripts, it works on inline scripts as well.
For example, the inline script below has async
, so it doesn’t wait for anything.
It performs the import (fetches ./analytics.js
) and runs when ready, even if the HTML document is not finished yet, or if other scripts are still pending.
That’s good for functionality that doesn’t depend on anything, like counters, ads, document-level event listeners.
<!-- all dependencies are fetched (analytics.js), and the script runs -->
<!-- doesn't wait for the document or other <script> tags -->
<script async type="module">
import {counter} from './analytics.js';
counter.count();
</script>
External scripts
External scripts that have type="module"
are different in two aspects:
-
External scripts with the same
src
run only once:<!-- the script my.js is fetched and executed only once --> <script type="module" src="my.js"></script> <script type="module" src="my.js"></script>
-
External scripts that are fetched from another origin (e.g. another site) require CORS headers, as described in the chapter Fetch: Cross-Origin Requests. In other words, if a module script is fetched from another origin, the remote server must supply a header
Access-Control-Allow-Origin
allowing the fetch.<!-- another-site.com must supply Access-Control-Allow-Origin --> <!-- otherwise, the script won't execute --> <script type="module" src="http://another-site.com/their.js"></script>
That ensures better security by default.
No “bare” modules allowed
In the browser, import
must get either a relative or absolute URL. Modules without any path are called “bare” modules. Such modules are not allowed in import
.
For instance, this import
is invalid:
import {sayHi} from 'sayHi'; // Error, "bare" module
// the module must have a path, e.g. './sayHi.js' or wherever the module is
Certain environments, like Node.js or bundle tools allow bare modules, without any path, as they have their own ways for finding modules and hooks to fine-tune them. But browsers do not support bare modules yet.
Compatibility, “nomodule”
Old browsers do not understand type="module"
. Scripts of an unknown type are just ignored. For them, it’s possible to provide a fallback using the nomodule
attribute:
<script type="module">
alert("Runs in modern browsers");
</script>
<script nomodule>
alert("Modern browsers know both type=module and nomodule, so skip this")
alert("Old browsers ignore script with unknown type=module, but execute this.");
</script>
Build tools
In real-life, browser modules are rarely used in their “raw” form. Usually, we bundle them together with a special tool such as Webpack and deploy to the production server.
One of the benefits of using bundlers – they give more control over how modules are resolved, allowing bare modules and much more, like CSS/HTML modules.
Build tools do the following:
- Take a “main” module, the one intended to be put in
<script type="module">
in HTML. - Analyze its dependencies: imports and then imports of imports etc.
- Build a single file with all modules (or multiple files, that’s tunable), replacing native
import
calls with bundler functions, so that it works. “Special” module types like HTML/CSS modules are also supported. - In the process, other transformations and optimizations may be applied:
- Unreachable code removed.
- Unused exports removed (“tree-shaking”).
- Development-specific statements like
console
anddebugger
removed. - Modern, bleeding-edge JavaScript syntax may be transformed to older one with similar functionality using Babel.
- The resulting file is minified (spaces removed, variables replaced with shorter names, etc).
If we use bundle tools, then as scripts are bundled together into a single file (or few files), import/export
statements inside those scripts are replaced by special bundler functions. So the resulting “bundled” script does not contain any import/export
, it doesn’t require type="module"
, and we can put it into a regular script:
<!-- Assuming we got bundle.js from a tool like Webpack -->
<script src="bundle.js"></script>
That said, native modules are also usable. So we won’t be using Webpack here: you can configure it later.
Summary
To summarize, the core concepts are:
- A module is a file. To make
import/export
work, browsers need<script type="module">
. Modules have several differences:- Deferred by default.
- Async works on inline scripts.
- To load external scripts from another origin (domain/protocol/port), CORS headers are needed.
- Duplicate external scripts are ignored.
- Modules have their own, local top-level scope and interchange functionality via
import/export
. - Modules always
use strict
. - Module code is executed only once. Exports are created once and shared between importers.
When we use modules, each module implements the functionality and exports it. Then we use import
to directly import it where it’s needed. The browser loads and evaluates the scripts automatically.
In production, people often use bundlers such as Webpack to bundle modules together for performance and other reasons.
In the next chapter we’ll see more examples of modules, and how things can be exported/imported.