Related
Lets say I'm writing a module in JavaScript which can be used on both the browser and the server (with Node). Lets call it Module. And lets say that that Module would benefit from methods in another module called Dependancy. Both of these modules have been written to be used by both the browser and the server, à la CommonJS style:
module.js
if (typeof module !== 'undefined' && module.exports)
module.exports = Module; /* server */
else
this.Module = Module; /* browser */
dependancy.js
if (typeof module !== 'undefined' && module.exports)
module.exports = Dependancy; /* server */
else
this.Dependancy = Dependancy; /* browser */
Obviously, Dependancy can be used straight-out-of-the-box in a browser. But if Module contains a var dependancy = require('dependency'); directive in it, it becomes more of a hassle to 'maintain' the module.
I know that I could perform a global check for Dependancy within Module, like this:
var dependancy = this.Dependancy || require('dependancy');
But that means my Module has two added requirements for browser installation:
the user must include the dependency.js file as a <script> in their document
and the user must make sure this script is loaded before module.js
Adding those two requirements throws the idea of an easy-going modular framework like CommonJS.
The other option for me is that I include a second, compiled script in my Module package with the dependency.js bundled using browserify. I then instruct users who are using the script in the browser to include this script, while server-side users use the un-bundled entry script outlined in the package.json. This is preferable to the first way, but it requires a pre-compilation process which I would have to run every time I changed the library (for example, before uploading to GitHub).
Is there any other way of doing this that I haven't thought of?
The two answers currently given are both very useful, and have helped me to arrive at my current solution. But, as per my comments, they don't quite satisfy my particular requirements of both portability vs ease-of-use (both for the client and the module maintainer).
What I found, in the end, was a particular flag in the browserify command line interface that can bundle the modules and expose them as global variables AND be used within RequireJS (if needed). Browserify (and others) call this Universal Module Definition (UMD). Some more about that here.
By passing the --standalone flag in a browserify command, I can set my module up for UMD easily.
So...
Here's the package.js for Module:
{
"name": "module",
"version": "0.0.1",
"description": "My module that requires another module (dependancy)",
"main": "index.js",
"scripts": {
"bundle": "browserify --standalone module index.js > module.js"
},
"author": "shennan",
"devDependencies": {
"dependancy": "*",
"browserify": "*"
}
}
So, when at the root of my module, I can run this in the command line:
$ npm run-script bundle
Which bundles up the dependancies into one file, and exposes them as per the UMD methodology. This means I can bootstrap the module in three different ways:
NodeJS
var Module = require('module');
/* use Module */
Browser Vanilla
<script src="module.js"></script>
<script>
var Module = module;
/* use Module */
</script>
Browser with RequireJS
<script src="require.js"></script>
<script>
requirejs(['module.js'], function (Module) {
/* use Module */
});
</script>
Thanks again for everybody's input. All of the answers are valid and I encourage everyone to try them all as different use-cases will require different solutions.
Of course you could use the same module with dependency on both sides. You just need to specify it better. This is the way I use:
(function (name, definition){
if (typeof define === 'function'){ // AMD
define(definition);
} else if (typeof module !== 'undefined' && module.exports) { // Node.js
module.exports = definition();
} else { // Browser
var theModule = definition(), global = this, old = global[name];
theModule.noConflict = function () {
global[name] = old;
return theModule;
};
global[name] = theModule;
}
})('Dependency', function () {
// return the module's API
return {
'key': 'value'
};
});
This is just a very basic sample - you can return function, instantiate function or do whatever you like. In my case I'm returning an object.
Now let's say this is the Dependency class. Your Module class should look pretty much the same, but it should have a dependency to Dependency like:
function (require, exports, module) {
var dependency = require('Dependency');
}
In RequireJS this is called Simplified CommonJS Wrapper: http://requirejs.org/docs/api.html#cjsmodule
Because there is a require statement at the beginning of your code, it will be matched as a dependency and therefore it will either be lazy loaded or if you optimize it - marked as a dependency early on (it will convert define(definition) to define(['Dependency'], definition) automatically).
The only problem here is to keep the same path to the files. Keep in mind that nested requires (if-else) won't work in Require (read the docs), so I had to do something like:
var dependency;
try {
dependency = require('./Dependency'); // node module in the same folder
} catch(err) { // it's not node
try {
dependency = require('Dependency'); // requirejs
} catch(err) { }
}
This worked perfectly for me. It's a bit tricky with all those paths, but at the end of the day, you get your two separate modules in different files which can be used on both ends without any kind of checks or hacks - they have all their dependencies are work like a charm :)
Take a look at webpack bundler.
You can write module and export it via module exports. Then You can in server use that where you have module.export and for browser build with webpack. Configuration file usage would be the best option
module.exports = {
entry: "./myModule",
output: {
path: "dist",
filename: "myModule.js",
library: "myModule",
libraryTarget: "var"
}
};
This will take myModule and export it to myModule.js file. Inside module will be assigned to var (libraryTarget flag) named myModule (library flag).
It can be exported as commonJS module, war, this, function
Since bundling is node script, this flag values can be grammatically set.
Take a look at externals flag. It is used if you want to have special behavior for some dependencies. For example you are creating react component and in your module you want to require it but not when you are bundling for web because it already is there.
Hope it is what you are looking for.
I am trying to optimize a durandal app into a single file using r.js via grunt and the durandal grunt task.
I am using js libraries that register with require inside a top level self executing function. The code below is from the breeze.js source but the code is conceptually the same in knockout, moment + others
if (typeof exports === "object" && typeof module === "object") {
module.exports = def();
} else if (typeof define === "function" && define["amd"]) {
define(def);
} else {
breeze = def();
}
The problem is that the r.js optimizer obviously cant insert a module id into the anonymous define.
THis means that when the optimized build file is loaded, breeze is registered as a module without a name so it be 'required' into other modules.
Whats the best way to handle this ? (modifying the source to define('breeze', [], def) would work for example).
Maybe patch the code so it 'knows' its running with almond and registers with a configurable name ?
I am trying to understand how to develop stand-alone Javascript code. I want to write Javscript code with tests and modules, running from the command line. So I have installed node.js and npm along with the libraries requirejs, underscore, and mocha.
My directory structure looks like this:
> tree .
.
├── node_modules
├── src
│ └── utils.js
└── test
└── utils.js
where src/utils.js is a little module that I am writing, with the following code:
> cat src/utils.js
define(['underscore'], function () {
"use strict";
if ('function' !== typeof Object.beget) {
Object.beget = function (o) {
var f = function () {
};
f.prototype = o;
return new f();
};
}
});
and test/utils.js is the test:
> cat test/utils.js
var requirejs = require('requirejs');
requirejs.config({nodeRequire: require});
requirejs(['../src/utils'], function(utils) {
suite('utils', function() {
test('should always work', function() {
assert.equal(1, 1);
})
})
});
which I then try to run from the top level directory (so mocha sees the test directory):
> mocha
node.js:201
throw e; // process.nextTick error, or 'error' event on first tick
^
Error: Calling node's require("../src/utils") failed with error: ReferenceError: define is not defined
at /.../node_modules/requirejs/bin/r.js:2276:27
at Function.execCb (/.../node_modules/requirejs/bin/r.js:1872:25)
at execManager (/.../node_modules/requirejs/bin/r.js:541:31)
...
So my questions are:
Is this the correct way to structure code?
Why is my test not running?
What is the best way to learn this kind of thing? I am having a hard time finding good examples with Google.
Thanks...
[sorry - momentarily posted results from wrong code; fixed now]
PS I am using requirejs because I also want to run this code (or some of it) from a browser, later.
Update / Solution
Something that is not in the answers below is that I needed to use mocha -u tdd for the test style above. Here is the final test (which also requires assert) and its use:
> cat test/utils.js
var requirejs = require('requirejs');
requirejs.config({nodeRequire: require});
requirejs(['../src/utils', 'assert'], function(utils, assert) {
suite('utils', function() {
test('should always work', function() {
assert.equal(1, 1);
})
})
});
> mocha -u tdd
.
✔ 1 tests complete (1ms)
The reason your test isn't running is because src/utils.js is not a valid Node.js library.
According to the RequireJS documentation, in order to co-exist with Node.js and the CommonJS require standard, you need to add a bit of boilerplate to the top of your src/utils.js file so RequireJS's define function is loaded.
However, since RequireJS was designed to be able to require "classic" web browser-oriented source code, I tend to use the following pattern with my Node.js libraries that I also want running in the browser:
if(typeof require != 'undefined') {
// Require server-side-specific modules
}
// Insert code here
if(typeof module != 'undefined') {
module.exports = whateverImExporting;
}
This has the advantage of not requiring an extra library for other Node.js users and generally works well with RequireJS on the client.
Once you get your code running in Node.js, you can start testing. I personally still prefer expresso over mocha, even though its the successor test framework.
The Mocha documentation is lacking on how to set this stuff up, and it's perplexing to figure out because of all the magic tricks it does under the hood.
I found the keys to getting browser files using require.js to work in Mocha under Node: Mocha has to have the files added to its suites with addFile:
mocha.addFile('lib/tests/Main_spec_node');
And second, use beforeEach with the optional callback to load your modules asynchronously:
describe('Testing "Other"', function(done){
var Other;
beforeEach(function(done){
requirejs(['lib/Other'], function(_File){
Other = _File;
done(); // #1 Other Suite will run after this is called
});
});
describe('#1 Other Suite:', function(){
it('Other.test', function(){
chai.expect(Other.test).to.equal(true);
});
});
});
I created a bootstrap for how to get this all working: https://github.com/clubajax/mocha-bootstrap
You are trying to run JS modules designed for browsers (AMD), but in the backend it might not work (as modules are loaded the commonjs way). Because of this, you will face two issues:
define is not defined
0 tests run
In the browserdefine will be defined. It will be set when you require something with requirejs. But nodejs loads modules the commonjs way. define in this case is not defined. But it will be defined when we require with requirejs!
This means that now we are requiring code asynchronously, and it brings the second problem, a problem with async execution.
https://github.com/mochajs/mocha/issues/362
Here is a full working example.
Look that I had to configure requirejs (amd) to load the modules, we are not using require (node/commonjs) to load our modules.
> cat $PROJECT_HOME/test/test.js
var requirejs = require('requirejs');
var path = require('path')
var project_directory = path.resolve(__dirname, '..')
requirejs.config({
nodeRequire: require,
paths: {
'widget': project_directory + '/src/js/some/widget'
}
});
describe("Mocha needs one test in order to wait on requirejs tests", function() {
it('should wait for other tests', function(){
require('assert').ok(true);
});
});
requirejs(['widget/viewModel', 'assert'], function(model, assert){
describe('MyViewModel', function() {
it("should be 4 when 2", function () {
assert.equal(model.square(2),4)
})
});
})
And for the module that you want to test:
> cat $PROJECT_HOME/src/js/some/widget/viewModel.js
define(["knockout"], function (ko) {
function VideModel() {
var self = this;
self.square = function(n){
return n*n;
}
}
return new VideModel();
})
Just in case David's answer was not clear enough, I just needed to add this:
if (typeof define !== 'function') {
var define = require('amdefine')(module);
}
To the top of the js file where I use define, as described in RequireJS docs ("Building node modules with AMD or RequireJS") and in the same folder add the amdefine package:
npm install amdefine
This creates the node_modules folder with the amdefine module inside.
I don't use requirejs so I'm not sure what that syntax looks like, but this is what I do to run code both within node and the browser:
For imports, determine if we are running in node or the browser:
var root = typeof exports !== "undefined" && exports !== null ? exports : window;
Then we can grab any dependencies correctly (they will either be available already if in the browser or we use require):
var foo = root.foo;
if (!foo && (typeof require !== 'undefined')) {
foo = require('./foo');
}
var Bar = function() {
// do something with foo
}
And then any functionality that needs to be used by other files, we export it to root:
root.bar = Bar;
As for examples, GitHub is a great source. Just go and check out the code for your favorite library to see how they did it :) I used mocha to test a javascript library that can be used in both the browser and node. The code is available at https://github.com/bunkat/later.
Currently, I use a few defines via the Google Closure Compiler along the lines of IS_CJS and IS_BROWSER, and just have different files that get built (browser.myproject.js, cjs.myproject.js, etc).
Is this the standard way of doing things? If not, what is it and what are the advantages?
I've been using the following preamble in all my projects, for libraries that are loaded by both browser and server code:
if (define === undefined) {
var define = function(f) {
require.paths.unshift('.');
f(require, exports, module);
};
}
define(function(require, exports, module) {
...
// main library here
...
// use require to import dependencies
var v = require(something);
...
// use exports to return library functions
exports.<stuff> = { some stuff };
...
});
This works to load the library with a require(<library>) call running on my node server, as well as with a require(<library>) call with RequireJS. On the browser, nested require calls are pre-fetched by RequireJS prior to library execution, on Node these dependencies are loaded synchronously. Since I'm not using my libraries as stand-alone scripts (via a script tag in the html), and only as dependencies for scripts loaded via the script tag, this works well for me.
However, looking at stand-alone libraries, it looks like the following preamble seems to be the most flexible. (cut and paste from the Q promise library
(function (definition, undefined) {
// This file will function properly as a <script> tag, or a module
// using CommonJS and NodeJS or RequireJS module formats. In
// Common/Node/RequireJS, the module exports the Q API and when
// executed as a simple <script>, it creates a Q global instead.
// The use of "undefined" in the arguments is a
// micro-optmization for compression systems, permitting
// every occurrence of the "undefined" variable to be
// replaced with a single-character.
// RequireJS
if (typeof define === "function") {
define(function (require, exports, module) {
definition(require, exports, module);
});
// CommonJS
} else if (typeof exports === "object") {
definition(require, exports, module);
// <script>
} else {
Q = definition(undefined, {}, {});
}
})(function (serverSideRequire, exports, module, undefined) {
...
main library here
...
/*
* In module systems that support ``module.exports`` assignment or exports
* return, allow the ``ref`` function to be used as the ``Q`` constructor
* exported by the "q" module.
*/
for (var name in exports)
ref[name] = exports[name];
module.exports = ref;
return ref;
});
While wordy, it's impressively flexible, and simply works, no matter what your execution environment is.
You can use uRequire that converts modules written in either AMD or CommonJS to either AMD, CommonJS or UMD through a template system.
Optionally uRequire builds your whole bundle as a combinedFile.js that runs in ALL environments (nodejs, AMD or module-less browser < script/>) thats using rjs optimizer and almond under the hood.
uRequire saves you from having to maintain any boilerplate in each module - just write plain AMD or CommonJS modules (as .js, .coffee, .coco, .ls etc) without gimmicks.
Plus you can declaratively add standard functionality such as exporting a module to global such as window.myModule along with a noConflict() method, or have runtimeInfo (eg __isNode, __isAMD) selectively or replace/remove/inject a dependency while building, automatically minify, manipulate module code and much more.
All of these configuration options can be turned on and off per bundle OR per module, and you can have different build profiles (development, test, production etc) that derive(inherit) from each other.
It works great with grunt through grunt-urequire or standalone and it has a great watch option that rebuilds ONLY changed files.
Have you tried this: https://github.com/medikoo/modules-webmake#modules-webmake ?
It's the approach I'm taking, and it works really well. No boilerplate in a code and you can run same modules on both server and client side
Is it possible to use require() (or something similar) on client side?
Example
var myClass = require('./js/myclass.js');
You should look into require.js or head.js for this.
I've been using browserify for that. It also lets me integrate Node.js modules into my client-side code.
I blogged about it here: Add node.js/CommonJS style require() to client-side JavaScript with browserify
If you want to have Node.js style require you can use something like this:
var require = (function () {
var cache = {};
function loadScript(url) {
var xhr = new XMLHttpRequest(),
fnBody;
xhr.open('get', url, false);
xhr.send();
if (xhr.status === 200 && xhr.getResponseHeader('Content-Type') === 'application/x-javascript') {
fnBody = 'var exports = {};\n' + xhr.responseText + '\nreturn exports;';
cache[url] = (new Function(fnBody)).call({});
}
}
function resolve(module) {
//TODO resolve urls
return module;
}
function require(module) {
var url = resolve(module);
if (!Object.prototype.hasOwnProperty.call(cache, url)) {
loadScript(url);
}
return cache[url];
}
require.cache = cache;
require.resolve = resolve;
return require;
}());
Beware: this code works but is incomplete (especially url resolving) and does not implement all Node.js features (I just put this together last night).
YOU SHOULD NOT USE THIS CODE in real apps but it gives you a starting point. I tested it with this simple module and it works:
function hello() {
console.log('Hello world!');
}
exports.hello = hello;
I asked myself the very same questions. When I looked into it I found the choices overwhelming.
Fortunately I found this excellent spreadsheet that helps you choice the best loader based on your requirements:
https://spreadsheets.google.com/lv?key=tDdcrv9wNQRCNCRCflWxhYQ
Take a look at requirejs project.
I have found that in general it is recommended to preprocess scripts at compile time and bundle them in one (or very few) packages with the require being rewritten to some "lightweight shim" also at compile time.
I've Googled out following "new" tools that should be able to do it
http://mixu.net/gluejs/
https://github.com/jrburke/almond
https://github.com/component/builder2.js
And the already mentioned browserify should also fit quite well - http://esa-matti.suuronen.org/blog/2013/04/15/asynchronous-module-loading-with-browserify/
What are the module systems all about?
Older Stack Overflow explanation - Relation between CommonJS, AMD and RequireJS?
Detailed discussion of various module frameworks and the require() they need is in Addy Osmani - Writing Modular JavaScript With AMD, CommonJS & ES Harmony
You can create elements to the DOM, which loads items.
Like such:
var myScript = document.createElement('script'); // Create new script element
myScript.type = 'text/javascript'; // Set appropriate type
myScript.src = './js/myclass.js'; // Load javascript file
Simply use Browserify, what is something like a compiler that process your files before it go into production and packs the file in bundles.
Think you have a main.js file that require the files of your project, when you run browserify in it, it simply process all and creates a bundle with all your files, allowing the use of the require calls synchronously in the browser without HTTP requests and with very little overhead for the performance and for the size of the bundle, for example.
See the link for more info: http://browserify.org/
Some answers already - but I would like to point you to YUI3 and its on-demand module loading. It works on both server (node.js) and client, too - I have a demo website using the exact same JS code running on either client or server to build the pages, but that's another topic.
YUI3: http://developer.yahoo.com/yui/3/
Videos: http://developer.yahoo.com/yui/theater/
Example:
(precondition: the basic YUI3 functions in 7k yui.js have been loaded)
YUI({
//configuration for the loader
}).use('node','io','own-app-module1', function (Y) {
//sandboxed application code
//...
//If you already have a "Y" instance you can use that instead
//of creating a new (sandbox) Y:
// Y.use('moduleX','moduleY', function (Y) {
// });
//difference to YUI().use(): uses the existing "Y"-sandbox
}
This code loads the YUI3 modules "node" and "io", and the module "own-app-module1", and then the callback function is run. A new sandbox "Y" with all the YUI3 and own-app-module1 functions is created. Nothing appears in the global namespace. The loading of the modules (.js files) is handled by the YUI3 loader. It also uses (optional, not show here) configuration to select a -debug or -min(ified) version of the modules to load.
Here's a solution that takes a very different approach: package up all the modules into a JSON object and require modules by reading and executing the file content without additional requests.
https://github.com/STRd6/require/blob/master/main.coffee.md
STRd6/require depends on having a JSON package available at runtime. The require function is generated for that package. The package contains all the files your app could require. No further http requests are made because the package bundles all dependencies. This is as close as one can get to the Node.js style require on the client.
The structure of the package is as follows:
entryPoint: "main"
distribution:
main:
content: "alert(\"It worked!\")"
...
dependencies:
<name>: <a package>
Unlike Node a package doesn't know it's external name. It is up to the pacakge including the dependency to name it. This provides complete encapsulation.
Given all that setup here's a function that loads a file from within a package:
loadModule = (pkg, path) ->
unless (file = pkg.distribution[path])
throw "Could not find file at #{path} in #{pkg.name}"
program = file.content
dirname = path.split(fileSeparator)[0...-1].join(fileSeparator)
module =
path: dirname
exports: {}
context =
require: generateRequireFn(pkg, module)
global: global
module: module
exports: module.exports
PACKAGE: pkg
__filename: path
__dirname: dirname
args = Object.keys(context)
values = args.map (name) -> context[name]
Function(args..., program).apply(module, values)
return module
This external context provides some variable that modules have access to.
A require function is exposed to modules so they may require other modules.
Additional properties such as a reference to the global object and some metadata
are also exposed.
Finally we execute the program within the module and given context.
This answer will be most helpful to those who wish to have a synchronous node.js style require statement in the browser and are not interested in remote script loading solutions.
I find the component project giving a much more streamlined workflow than other solutions (including require.js), so I'd advise checking out https://github.com/component/component . I know this is a bit late answer but may be useful to someone.
Here's a light weight way to use require and exports in your web client. It's a simple wrapper that creates a "namespace" global variable, and you wrap your CommonJS compatible code in a "define" function like this:
namespace.lookup('org.mydomain.mymodule').define(function (exports, require) {
var extern = require('org.other.module');
exports.foo = function foo() { ... };
});
More docs here:
https://github.com/mckoss/namespace
The clientside-require library provides an asynchronous load() function that can be used to load any JS file or NPM module (which uses module.exports), any .css file, any .json, any .html, any any other file as text.
e.g.,
npm install clientside-require --save
<script src = '/node_modules/clientside-require/dist/bundle.js'></script>
<script>
load('color-name') // an npm module
.then(color_name=>{
console.log(color_name.blue); // outputs [0, 0, 255]
})
</script>
A really cool part of this project is that inside of any load()ed script, you can use the synchronous require() function the same way you would expect in node.js!
e.g.,
load('/path/to/functionality.js')
and inside /path/to/functionality.js:
var query_string = require("qs") // an npm module
module.exports = function(name){
return qs.stringify({
name:name,
time:new Date()
}
}
That last part, implementing the synchronous require() method, is what enables it to utilize NPM packages built to run on the server.
This module was designed to implement the require functionality as closely as possible in the browser. Disclaimer: I have written this module.
Yes it is very easy to use, but you need to load javascript file in browser by script tag
<script src="module.js"></script>
and then user in js file like
var moduel = require('./module');
I am making a app using electron and it works as expected.