module-loader-tdd
v0.2.1
Published
Module loader that lets you test each module with private methods and stubbed dependencies
Downloads
11
Readme
module-loader-tdd
An easy to use and easy to test module loader
Table of contents
- Install
- The spec
- Creating a module
- Adding dependencies
- Templates
- Adding private methods
- Using resources
- Initializing
- Testing a module
- Node: Creating a module
- Node: Initializing
- Node: Testing a module
- Why build it?
I strongly urge you to have a look at the following Grunt plugin: https://github.com/christianalfoni/grunt-tdd
The module-loader-tdd is built to more easily test code and grunt-tdd will help you get going with that.
CHANGE LOG
2.0.1:
- Added resources (see tutorial below)
- No timeout, but dependency tracking instead
- modules.initialize callback will be called on DOMContentLoaded, does not matter when initialize is called anymore
- Added configurable compiled Handlebar templates property
Install
- Browser: Download the file from lib folder
- Node: npm install module-loader-tdd
Spec
The basics
- Create closed modules that does not pollute the global scope
- A module can import other modules
- Module loading order is irrelevant. A dependency not yet loaded should postpone the initializing of the module until the dependency has been initialized
- It supports concatinating all module files, compress and uglify without breaking any code
- Give good error indications if you are using it wrong
- Freedom in naming modules (namespaces etc.)
- Short stack trace
- Supports Node JS
- Supports HTML templates
Test additions
- Close the module completely, no code execution beyond the tested module
- A custom module context, giving access to setting private methods and deps that are only exposed during testing
- A "modules.test()" method which exposes private functions and deps. The deps are stubbed causing the module to be completely unaffected by other parts of the project
How to use it in the browser
You are required to load the scripts with script tags. They should be loaded in the body of your HTML so that any CSS and/or HTML content can be displayed while loading the scripts.
module-loader exposes a global variable called: modules. It has three methods, create, initialize and test.
Creating a module
Pass two arguments to the create method. A name for the module, which can contain any character, and a function. The returned value from the function will be available to other modules.
// FILE: logger.js
modules.create('logger', function () {
'use strict';
return {
log: function (message) {
console.log(message);
}
};
});
TIP: Use namespace to divide your modules. E.g. modules.create('helper.logger'... or modules.create('model.Todo'...
Adding dependencies
The first argument passed to your module function is require. Use it to fetch other modules defined.
// FILE: helloWorld.js
modules.create('helloWorld', function (require) {
'use strict';
var logger = require('logger');
return {
hello: function () {
logger.log('Hello world!');
}
};
});
Even if helloWorld.js is loaded after logger.js it will still work.
Adding templates
The second argument passed to the module, require, also has its own method template. Currently it only supports Handlebars. Use it to require a template from a /templates path or a predefined path.
// FILE: helloWorld.js
modules.create('helloWorld', function (require, p) {
'use strict';
var logger = require('logger'),
template = require.template('message');
return {
hello: function () {
logger.log('Hello world!');
document.body.innerHMTL = template({ content: 'Hello world!' });
}
};
});
You can configure the template directory, take a look at "Initializing the project"
Creating private methods
The second argument passed is an object of private methods. These methods are not exposed, but will be during testing of the module. This gives you a clear definition of which methods are public and which are private to the module.
// FILE: helloWorld.js
modules.create('helloWorld', function (require, p) {
'use strict';
var logger = require('logger');
p.sayToWorld = function (say) {
return say + ' world!';
};
return {
hello: function () {
logger.log(p.sayToWorld('Hello'));
}
};
});
Why create these private methods? When developing test-driven and functional, a core concept is creating small input-output functions that are easily testable. You normally do not want to expose these methods to the rest of the application. By using a "module-context" these private methods can be exposed only during testing.
The argument passed to the module function is also available in the execution context itself. An alternative convention on defining privates can be done by replacing the privates object, which has to be done like this:
modules.create('helloWorld', function (require) {
'use strict';
var logger = require('logger');
// The p variable lets you point to the privates
// from any public method, or other private method
var p = this.privates = {
sayToWorld: function (say) {
return say + ' world!';
}
};
return {
hello: function () {
logger.log(p.sayToWorld('Hello'));
}
};
});
Using resources
In large projects it is often necessary to share certain resources. This could be configurations, a user, language resources or maybe some items from a collection.
Instead of loading this as a direct dependency it can be loaded as a resource instead. A resource is kinda event-based, meaning that the module just shouts out "I want this", and it is up to something else to have registered a handler for it. This makes it easier to test your code and it decreases the magnitude of your dependencies.
// FILE: helloWorld.js
modules.create('helloWorld', function (require, p, resource) {
'use strict';
var logger = require('logger');
return {
hello: function () {
var user = resource.fetch('user');
logger.log(user.name + ' says: Hello world!');
}
};
});
// FILE: user.js
modules.create('user', function (require, p, resource) {
'use strict';
var user = {
name: 'Bubba'
};
resource.register('user', function () {
return user;
});
return {};
});
The registered resource can return anything. If the resource requires server connection it is possible to return a promise instead, or a callback.
// FILE: helloWorld.js
modules.create('helloWorld', function (require, p, resource) {
'use strict';
var logger = require('logger');
return {
hello: function () {
var user = resource.fetch('user', function (user) {
logger.log(user.name + ' says: Hello world!');
});
}
};
});
// FILE: user.js
modules.create('user', function (require, p, resource) {
'use strict';
resource.register('user', function (callback) {
$.get('/user', callback);
});
return {};
});
Initializing the project
<!-- FILE: index.html -->
<!DOCTYPE html>
<html>
<head>
</head>
<body>
<script src="lib/module-loader-tdd.js"></script>
<script src="src/myModule.js"></script>
<script src="src/myDep.js"></script>
<script>
modules.templatesPath = 'src/templates/'; // Sets the path to the templates, default is 'templates/'
modules.templates = 'JST'; // Sets the global property where templates will be when compiled
modules.initialize(function (require) {
var helloWorld = require('helloWorld');
helloWorld.hello(); // -> Hello world!
});
</script>
</body>
</html>
Note that in production all the modules will be concatinated and optimized into one single javascript file, avoiding any unneccesary fetching of files.
Note that you could of course load a main.js as the last script instead of putting it inside your html file/template
TIP: When using Node JS and the HTML is delivered with a template, you can easily dynamically add the script tags based on available .js files in your source folder. Look at the demo.
Testing a module
Now, this is where the module-loader-tdd shines. Running a test on a module requires you to pass the name of the module and a function for testing. The function receives three arguments. The first being the the module, the second being its private methods and the last argument is the required dependencies inside the module being tested.
The deps object is a map of the dependencies, e.g. deps.logger. If the dependency is an object with methods they are all automatically stubbed, which basically means that they are verifiable empty functions. We do this to isolate the test to only the module and not trigger code that should be tested elsewhere.
Any required templates will not be fetched, it will return an empty string, as you will be testing the code that puts content into your HTML, not the HTML itself.
// FILE: helloWorld-test.js
modules.test('helloWorld', function (helloWorld, p, deps, resource) {
'use strict';
buster.testCase('helloWorld test', {
setUp: function () {
resource.register('user', function () {
return {
name: 'bilbo'
};
};
},
tearDown: function () {
resource.remove('user');
},
'hello()': {
'is a function': function () {
assert.isFunction(helloWorld.hello);
},
'calls dependency with message': function () {
helloWorld.hello();
// Deps are stubbed methods (Sinon JS), which lets us verify their usage
// without actually executing the code
assert(deps.logger.log.calledOnce); // Has the log method been called?
assert(deps.logger.log.calledWith('bilbo says: Hello world!')); // Was it called with the expected message?
}
},
'p.sayToWorld()': {
'is a function': function () {
assert.isFunction(p.sayToWorld);
},
'returns passed argument with " world!" appended to the string': function () {
assert.equals(p.sayToWorld('Hello'), 'Hello world!');
assert.equals(p.sayToWorld('Good evening'), 'Good evening world!');
}
}
});
});
Creating a module with Node js
Node JS has a module loader, but it does not have the privates and dep stubbing that module-loader-tdd offers. If you want that functionality also in Node you wrap each file the same way as in the browser.
Since Node JS already has a require function and a convention for loading files, module-loader-tdd does not mess with that. It only works as a "middle-man" registering all loaded modules to create a context with privates and stubbed dependencies for testing. Because of this, you do not set a name for the module.
The require.template method is not implemented on Node JS as you normally do not load templates directly and Node JS is not only web related.
// FILE: mainModule.js
modules.create(function (require, p) {
'use strict';
var fs = require('fs'), // Loads the built in fs module in Node JS
myModule = require('./myModule'); // Loads one of your own modules, relative to the file you are in
p.log = function () {
console.log('test');
};
return {
log: function (message) {
p.log();
}
};
});
NOTE That if a dependency returns an empty object you have probably forgotten to use the passed require function. Modules required with Node JS global require will not be registered and handled by module-loader-tdd
Initializing the modules
In your main .js file for the Node project, add the following:
require('module-loader-tdd'); // Will add "modules" to the global scope
modules.initialize(function (require) {
var module = require('./main');
module.log();
});
Testing Node JS modules
There is little difference in testing a Node JS module.
// FILE: myModule-test.js
require('module-loader-tdd');
// Going to parent folder of tests/ and into the modules folder
modules.test('./../modules/myModule', function (myModule, p, deps) {
'use strict';
buster.testCase('helloWorld test', {
'hello()': {
'is a function': function () {
assert.isFunction(myModule.log);
}
}
});
Why build it?
I have been working a lot with RequireJS and have a good experience with it. When going TDD though, I started having problems with test libraries, test related plugins and testing Node JS. A very typical search result on Google is: "How to make X work with requirejs". So I went into the specific details about why I was actually using RequireJS and identify through my experience what I really needed and if I could solve it myself.
The parts of RequireJS that does not really work for me
- RequireJS requires a lot of configuration in big projects and the configuration is duplicated when running multiple applications against the same modules. The tests also needs a RequireJS configuration
- "How to make X work with requirejs" gives a feeling of "hacking" together parts of the project
- A lot of logic is running to load all the modules and "only load needed dependencies" does not really make sense to me. When you load the project you normally want to load all the modules
- It is hard to test the modules. RequireJS modules gives a public interface, but the modules private functions and dependencies used makes them difficult to test
- All libraries are built for normal script loading, not all are built for RequireJS. Shimmed libraries are global, which kinda breaks the concept
- The loading of files is based on directory structuring. You can overwrite with "paths", but that just adds complexity
- It is very complex with lots of functionality
- Have not successfully accomplished code-coverage with it
- Stack trace is way too long, and does not make much sense
In RequireJS defense I am not using all the functionality and maybe the projects I work on would benefit more with a different structure etc, but that also indicates its complexity.