Chapter 3. Modularizing and Managing JavaScript

One of the great aspects of writing Node.js applications is the built-in modularity the environment provides. As demonstrated in [link unavailable], it’s simple to download and install any number of Node modules, and using them is equally simple: just include a single require() statement naming the module, and you’re off and running.

The ease with which the modules can be incorporated is one of the benefits of JavaScript modularization. Modularizing ensures that external functionality is created in such a way that it isn’t dependent on other external functionality, a concept known as loose coupling. This means I can use a Foo module, without having to include a Bar module, because Foo is tightly dependent on having Bar included.

JavaScript modularization is both a discipline and a contract. The discipline comes in by having to follow certain mandated criteria in order for external code to participate in the module system. The contract is between you, me, and other JavaScript developers: we’re following an agreed on path when we produce (or consume) external functionality in a module system, and we all have expectations based on the module system.

Note

ECMAScript 6 provides native support for modules, but the specification is still undergoing change and there is no implementation support yet. There is some support for it in Traceur, as well as a polyfill, which can at least provide an idea of how they’ll be implemented in the future.

Chances are you have used modularized JavaScript. If you have used jQuery with RequireJS or Dojo, you’ve used modularized JavaScript. If you’ve used Node, you’ve used a modular system. They don’t look the same, but they work the same: ensuring that functionality developed by disparate parties works together seamlessly. The modular system that RequireJS and Dojo support is the Asynchronous Module Definition (AMD), while Node’s system is based on CommonJS. One major difference between the two is that AMD is asynchronous, while CommonJS is synchronous.

Even if you don’t use a formal modular system, you can still improve the performance of script loading with script loaders and using new HTML5 async functionality. You can also improve the management of your entire application process using tools such as Grunt, or ensuring your own code is packaged for ease of use and innovation.

Note

One major dependency on virtually all aspects of application and library management and publication is the use of Git, a source control system, and GitHub, an extremely popular Git endpoint. How Git works and using Git with GitHub are beyond the scope of this book. I recommend The Git Pocket Guide (O’Reilly) to get more familiar with Git, and GitHub’s own documentation for more on using this service.

Loading Scripts with a Script Loader

Problem

You need to use several different JavaScript libraries in your web pages, and they’re starting to slow the page loads.

Solution

One solution is to use a script loader to load your JavaScript files asynchronously and concurrently. Examples of use are documented in the discussion.

Discussion

There are several techniques you can use to load JavaScript files. One is the traditional method of using a script element for each file, and just loading each in turn. The issue that people have had with this approach is the inefficiency of having to access each file individually, the problems that can occur if scripts are loaded out of order (with one script being dependent on another already loaded), and the fact that the entire page is blocked while the scripts load.

Some solutions are to compile all the individual JavaScript files into a single file, which is what the content management system (CMS) Drupal does. This eliminates the multiple file access and even the issues with ordering, but it still leaves us with the fact that the page is blocked from loading until the scripts are loaded.

Script loaders were created to provide a way of loading JavaScript files asynchronously, which means the rest of the page can continue loading while the script is loading. They use script injection: creating a script element in a script block that loads the JavaScript file, and then appending that block to the page. The inline JavaScript is executed asynchronously and does not block the page from loading like the use of the traditional script element does.

The code to do so can be similar to the script block shown in the following minimal HTML5 page:

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<title>title</title>
</head>
<body>
  <script>
    var scrpt = document.querySelector("script");
    var t = document.createElement("script");
    t.src = "test1.js";
    scrpt.parentNode.insertBefore(t,scrpt);
  </script>
</body>
</html>

To prevent the variables from cluttering up the global namespace, they can be included in an Immediately-Invoked Function Expression (IIFE):

<script>
  (function() {
     var scrpt = document.querySelector("script");
     var t = document.createElement("script");
     t.src = "test1.js";
     scrpt.parentNode.insertBefore(t,scrpt);
  }());
</script>

If you need to use a pathname for the script, you can use a protocol-relative URL (sometimes referred to as a protocol-less URL) so that the code adapts whether the page is accessed with http or https:

t.src = "//somecompany.com/scriptfolder/test1.js";

With this, the client application uses the same protocol (http or https) used to access the parent page.

Multiple scripts can be loaded into the page using this approach. It can also be used to load CSS files, as well as larger images or other media files. However, we don’t have to do the work ourselves: we can use a script loading library, such as HeadJS.

According to the HeadJS documentation, the best approach to including support for the library is to include a link to the library in the head element:

<html>
  <head>
    <script src="head.min.js"></script>
    <script>
      head.load("file1.js", "file2.js");
    </script>
  </head>
  <body>
    <!-- my content-->

    <script>
      head.ready(function () {
            // some callback stuff
     });
    </script>
  </body>
</html>

Note the head.load() function call. All of the script files to be loaded are listed in the function call. In addition, any ready state functionality can be provided in the head.ready() function call.

If you do have JavaScript, you want to load right away; rather than using another script element, you can use a data- attribute on the script element loading HeadJS:

<script src="head.min.js" data-headjs-load="init.js"></script>

Any immediately invoked functionality is then listed in init.js.

Note

HeadJS has other functionality, including assistance for responsive design and browser version support. Read more about setting it up in the set up documentation.

Another script loader with an interesting twist is Basket.js. It also loads JavaScript files asynchronously, but it goes a step further: it caches the script using localStorage, which means if the JavaScript has already been accessed once, a second access loads the JavaScript from cache rather than loading the file again.

Once you include the Basket.js JavaScript file, you can then define the JavaScript files to be loaded:

<!DOCTYPE html>
<html lang="en">
  <head>
    <meta charset="utf-8">
    <title>title</title>
  </head>
<body>
  <script src="basket.full.min.js"></script>
  <script>
    basket.require({ url: 'test1.js'},
                   { url: 'test2.js'});
  </script>
</body>
</html>

If you monitor the page using your browser’s debugger/development tools, and reload the page, you’ll note that the files aren’t accessed again after the first load.

To handle source dependencies, Basket.js returns a promise from require(), and the then() callback is executed. You can then list the second JavaScript file in the callback:

<script>
    basket.require({ url: 'test2.js'}).then(function() {
        basket.require({ url: 'test1.js'});
    });
</script>
Note

Access Basket.js and read how to use it in the library’s home page.

Loading Scripts Asynchronously the HTML5 Way

Problem

You’re interested in processing scripts asynchronously—not blocking the page from loading while the scripts load—but you have discovered that the script injection technique has one problem: the CSS Object Model (CSSOM) blocks inline scripts because these scripts typically operate on the CSSOM. Since the CSSOM doesn’t know what the script is going to do, it blocks the script until all of the CSS is loaded. This, then, delays the network access of the script until all CSS files have been loaded.

Solution

Use the new HTML5 async script element attribute instead of script injection:

<script src="//cdnjs.cloudflare.com/ajax/libs/mathjs/0.26.0/math.min.js" async>
</script>
<script
src="//cdnjs.cloudflare.com/ajax/libs/backbone.js/1.1.2/backbone-min.js" async>
</script>

Discussion

There are two script element attributes: defer, which defers script loading until the rest of the page is loaded, and the newest async. The latter tells the browser to load the script asynchronously, as the page is being parsed. It only works with external scripts; the page still blocks with inline scripts.

The async attribute prevents many of the problems we’ve had with blocked scripts and having to use tricks such as script injection. The only reason script injection is still being used is there are older versions of browsers, such as IE9 and older, that don’t support it.

Converting Your JavaScript to AMD and RequireJS

Problem

You’re interested in taking advantage of modularization and controlled dependencies by converting your libraries to the Asynchronous Module Definition (AMD) format, implemented with RequireJS, but you’re not sure where to start and what to do.

Solution

RequireJS is integrated into the following three small JavaScript libraries:

one.js

define(function() {
        return {
          hi: function() {
            console.log('hello from one');
          }
        }
});

two.js

define(function() {
        return {
          hi: function(val) {
            console.log('hello ' + val + ' from two');
          }
        }
});

mylib.js

require(["./one","./two"],function(one,two) {
         one.hi();
         two.hi('world');
         console.log("And that's all");
});

And the web page, index.html:

<!DOCTYPE html>
<html>
  <head>
    <title>Hello Modularization</title>
    <script data-main="scripts/mylib" src="scripts/require.js"></script>
  </head>
  <body>
    <h1>Stuff</h1>
  </body>
</html>

Discussion

Consider the following three very basic JavaScript libraries:

one.js

function oneHi() {
  console.log('hello from one');
}

two.js

function twoHi(val) {
  console.log('hello ' + val + ' from two');
}

mylib.js

function allThat() {
  oneHi();
  twoHi('world');
  console.log("And that's all");
}

They could be included in a simple web page as demonstrated in the following code, assuming all the JavaScript libraries are in a subdirectory named scripts/:

<!DOCTYPE html>
<html>
  <head>
    <title>Hello Modularization</title>
    <script src="scripts/one.js" type="text/javascript"></script>
    <script src="scripts/two.js" type="text/javascript"></script>
    <script src="scripts/mylib.js" type="text/javascript"></script>
    <script type="text/javascript">
      allThat();
    </script>
  </head>
  <body>
    <h1>Stuff</h1>
  </body>
</html>

And you might expect the application to work, with the messages printed out in the right order. However, if you make a modest change, such as use the async attribute with all of the scripts:

<script src="scripts/one.js" async type="text/javascript"></script>
<script src="scripts/two.js" async type="text/javascript"></script>
<script src="scripts/mylib.js" async type="text/javascript"></script>

You’ll be hosed, because the browser no longer blocks program execution, waiting for each script to load, in turn, before going to the next. Other challenges that can occur are that you’re using other people’s libraries and you don’t know the correct order to list the source scripts, or you forget one or more of them. The problem with this common approach from the past is that nothing enforces both order and dependencies. That’s where RequireJS comes in.

In the solution, you’ll notice two key words: define and require. The define keyword is used to define a module, while require is used to list dependencies with a callback function that’s called when all dependencies are loaded.

In the solution, two of the libraries are defined as modules, each return a function. The third library, mylib.js, declares the two modules as dependencies and in the callback function, invokes the returned module functions. All of this is pulled into the HTML page with the following line:

<script data-main="scripts/mylib" src="scripts/require.js"></script>

The actual source is the RequireJS library. The custom attribute data-main specifies the JavaScript source to load after RequireJS is loaded.

The modules can return more than one function, or can return data objects, functions, or a combination of both:

define(function() {
    return {
        value1: 'one',
        value2: 'two',
        doSomething: function() {
           // do something
        }
    }
})

Modules can also have dependencies. The following code version of two.js creates a dependency on one.js in two.js and removes it as a dependency in mylib.js:

two.js

define(['one'], function(one) {
        return {
           hi: function(val) {
                 one.hi();
                 console.log('hello ' + val + ' from two');
               }
        }
});

mylib.js

require(["./two"],function(two) {
         two.hi('world');
         console.log("And that's all");
});
Note

Typically after you create your JavaScript files, you’ll want to optimize them. RequireJS provides the tools and documentation for optimizing your source at http://requirejs.org/docs/optimization.html.

See Also

Your library can still exist as a standard JavaScript library and an AMD-compliant module, as discussed in [link unavailable].

Using RequireJS with jQuery or Another Library

Problem

Your applications uses jQuery (or Underscore.js or Backbone). How can the library fit into the use of RequireJS to manage dependencies?

Solution

If the library can work with AMD (as jQuery can), and you save the jQuery file as jquery.js and load it in the same directory as your application JavaScript, you can use the jQuery functionality easily, as shown in the following small code snippet:

require(["./jquery"],function($) {
    $('h1').css('color','red');
});

However, if the jQuery file is named something else, or you’re accessing the library from a CDN, then you’ll need to use a RequireJS shim:

requirejs.config({
   baseUrl: 'scripts/lib',
   paths: {
      jquery: '//ajax.googleapis.com/ajax/libs/jquery/2.1.1/jquery.min'
   },
});

Discussion

As the solution demonstrates, if your application code already incorporates jQuery’s dollar sign ($) and the jQuery file is local to the script, you can incorporate its use in your application in the same manner used for any other module. The jQuery library can recognize that it’s within a RequireJS environment, and respond accordingly. Where things get a little more complicated is if the library is not accessed locally, is accessed from a CDN, or the library doesn’t support AMD.

To demonstrate, I modified the source files discussed in “Converting Your JavaScript to AMD and RequireJS”. The source files are now organized in the following directory structure:

www
  app
     main.js
  index.html
  scripts
     app.js
     lib
        one.js
        require.js
        two.js

In addition, I removed the define() in the source library two.js, making it into an anonymous closure—an IIFE object that is added to the Window object as two:

(function (){
   window.two = this;
   this.hi =  function(val) {
      console.log('hello ' + val + ' from two');
   }
}());

The one.js file still contains the AMD define() statement, meaning it requires no special handling to use:

define(function() {
        return {
           hi: function() {
                  console.log('hello from one');
               }
        }
});

The app.js file contains a RequireJS config block that, among other things, sets a baseUrl for all loaded modules, defines a CDN path for both jQuery and the app subdirectory, and creates a shim for the non-AMD compliant two. It also loads the app/main module:

requirejs.config({
   baseUrl: 'scripts/lib',
   paths: {
      app: '../../app',
      jquery: '//ajax.googleapis.com/ajax/libs/jquery/2.1.1/jquery.min'
   },
   shim: {
      two: {
        exports: 'two'
      }
   }
});

requirejs(["app/main"]);

The shim for two defines an exported object (an object defined on Window in the browser), since the library doesn’t use define() to identify the object.

Lastly, the main.js module lays out the dependency on jQuery, one, and two, and runs the application:

define(["jquery","one","two"],function($,one, two) {
  one.hi();
  two.hi('world');
  console.log("And that's all");
  $('h1').css('color','red');
});

If two had been dependent on one of the modules or other libraries, such as one, the dependency would have been noted in the shim:

requirejs.config({
   baseUrl: 'scripts/lib',
   paths: {
      app: '../../app',
      jquery: '//ajax.googleapis.com/ajax/libs/jquery/2.1.1/jquery.min'
   },
   shim: {
      two: {
        deps: ['one'],
        exports: 'two'
      }
   }
});

If you’d like to make your JavaScript library into an AMD-compliant module, but still allow it to be used in other contexts, you can add a small amount of code to ensure both:

(function (){
   window.two = this;
   this.hi =  function(val) {
      console.log('hello ' + val + ' from two');
   }

}());

The tiny library is now redesigned into an IIFE. Any private data and methods would be fully enclosed in the closure, and the only public method is exposed by adding it as a property to the object. The object itself is given global access via assignment to the Window property.

A variation on this would be the following, where the exposed methods and data are returned as an object to the assigned variable:

var two = (function (){
   return {
      hi: function (val) {
         console.log('hello ' + val + ' from two');
      }
   }

}());

The code now meets the module pattern, ensuring both public and private data and functions are encapsulated using the closure, and globally accessible methods and data are returned in the object. Another variation of the module pattern is the following:

var two = (function() {
   var my = {};
   my.hi = function(val) {
              console.log('hello ' + val + ' from two');
           };
   return my;
}());

I modified the original form of the object to make it AMD compliant:

(function (){
   window.two = this;
   this.hi =  function(val) {
      console.log('hello ' + val + ' from two');
   }

   if ( typeof define === "function" && define.amd ) {
      define( "two", [], function() {
         return two;
      });
   }
}());

The code tests to see if the define() function exists. If so, then it’s invoked, passing in the name of the exported library object and in the callback, returning the exported library object. This is how a library such as jQuery can work in AMD, but still work in other traditional JavaScript environments.

A variation, using the more established module pattern, is the following:

var two = (function (){
   var two = {};

   two.hi =  function(val) {
      console.log('hello ' + val + ' from two');
   }

   if ( typeof define === "function" && define.amd ) {
      define( "two", [], function() {
         return two;
      });
   }

   return two;
}());
Note

jQuery also supports the CommonJS modular system.

Loading and Using Dojo Modules

Problem

You’re interested in using some of the Dojo functionality, but you’re not sure how to load the associated modules.

Solution

Dojo has implemented the AMD architecture for its functionality. When you add the main Dojo script to your page, what you’re loading is the module loader, rather than all of its various functions:

<script src="http://ajax.googleapis.com/ajax/libs/dojo/1.10.0/dojo/dojo.js"
        data-dojo-config="async: true"></script>

The library can be accessed at a CDN, as the code snippet demonstrates. The custom data attribute data-dojo-config specifies that the Dojo asynchronous AMD loader should be used.

To use the Dojo functionality, specify the dependencies in the require() method:

    <script>
        require([
            'dojo/dom',
            'dojo/dom-construct'
        ], function (dom, domConstruct) {
            var ph = dom.byId("placeholder");
            ph.innerHTML = "Using Dojo";
            domConstruct.create("h1", {innerHTML: "<i>Howdy!</i>"},ph,"before");
        });
    </script>

Discussion

Dojo is a sophisticated library system providing functionality similar to that provided in the jQuery environment. It does require a little time to become familiar with its implementation of AMD, though, before jumping in.

In the solution, the Dojo asynchronous loader is sourced from a CDN. The solution then imports two Dojo modules: dojo/dom and dojo/dom-construct. Both provide much of the basic DOM functionality, such as the ability to access an existing element by an identifier (dom.byId()), and create and place a new element (domConstruct.create()). To give you a better idea how it all holds together, a complete page example is given in Example 3-1.

Example 3-1. A complete Dojo example accessing one page element and adding another
<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <title>Dojo</title>
  <script src="http://ajax.googleapis.com/ajax/libs/dojo/1.10.0/dojo/dojo.js"
            data-dojo-config="async: true"></script>
</head>
<body>
  <div id="placeholder"></div>
  <script>
    require([
        'dojo/dom',
        'dojo/dom-construct'
    ], function (dom, domConstruct) {
        var ph = dom.byId("placeholder");
        ph.innerHTML = "Using Dojo";
        domConstruct.create("h1", {innerHTML: "<i>Howdy!</i>"},ph,"before");
    });
  </script>
</body>
</html>
Caution

Though Dojo is generally AMD-compatible, there’s still some funkiness with the implementation that makes it incompatible with a module loader like RequireJS. The concepts of a module loader, the require() and define() functions, and creating a configuration object are the same, but implementation compatibility fails.

Dojo does provide a decent set of tutorials to help you understand more fully how the framework operates.

Installing and Maintaining Node Modules with npm

Problem

You’re new to Node. You’ve installed it, and played around with the core Node modules installed with Node. But now, you need something more.

Solution

The glue that holds the Node universe together is npm, the Node package manager. To install a specific module, use the following on the command line:

npm install packagename

If you want to install the package globally, so it’s accessible from all locations in the computer, use the following:

npm install -g packagename

When to install locally or globally is dependent on whether you’re going to require() the module, or if you need to run it from the command line. Typically you install require() modules locally, and executables are installed globally, though you don’t have to follow this typical usage. If you do install a module globally, you might need administrative privileges:

sudo npm install -g packagename

Discussion

The solution demonstrated the most common use of npm: installing a registered npm module locally or globally on your system. However, you can install modules that are located in GitHub, downloaded as a tar file, or located in a folder. If you type:

npm install --help

you’ll get a list of allowable approaches for installing a module:

npm install
npm install <pkg>
npm install <pkg>@<tag>
npm install <pkg>@<version>
npm install <pkg>@<version range>
npm install <folder>
npm install <tarball file>
npm install <tarball url>
npm install <git:// url>
npm install <github username>/<github project>

If your current directory contains a npm-shrinkwrap.json or package.json file, the dependencies in the files are installed by typing npm install.

To remove an installed Node module, use:

npm rm packagename

The package and any dependencies are removed. To update existing packages, use:

npm update [g] [packagename [packagename ...]]

You can update locally or globally installed modules. When updating, you can list all modules to be updated, or just type the command to update all locally-installed modules relative to your current location.

Searching for a Specific Node Module via npm

Problem

You’re creating a Node application and want to reuse existing modules, but you don’t know how to discover them.

Solution

In most cases, you’ll discover modules via recommendations from your friends and co-developers, but sometimes you need something new.

You can search for new modules directly at the npm website. The front page also lists the most popular modules, which are worth an exploratory look.

You can also use npm directly to search for a module. For instance, if you’re interested in modules that do something with PDFs, run the following search at the command line:

npm search pdf

Discussion

The npm website provides more than just good documentation for using npm; it also provides a listing of newly updated modules, as well as those modules most depended on. Regardless of what you’re looking for, you definitely should spend time exploring these essential modules. In addition, if you access each module’s page at npm, you can see how popular the module is, what other modules are dependent on it, the license, and other relevant information.

However, you can also search for modules, directly, using npm.

The first time you perform a search with npm, you’ll get the following feedback:

npm WARN Building the local index for the first time, please be patient

The process can take a fair amount of time, too. Luckily, the index build only needs to be performed the first time you do a search. And when it finishes, you’re likely to get a huge number of modules in return, especially with a broader topic such as modules that work with PDFs.

You can refine the results by listing multiple terms:

npm search PDF generation

This query returns a much smaller list of modules, specific to PDF generation. You can also use a regular expression to search:

npm search \/Firefox\\sOS

Now I’m getting all modules that reference Firefox OS. However, as the example demonstrates, you have to incorporate escape characters specific to your environment, as I did with the beginning of the regular expression, and the use of \s for white space.

Once you do find a module that sounds interesting, you can get detailed information about it with:

npm view node-firefoxos-cli

You’ll get the package.json file for the module, which can tell you what it’s dependent on, who wrote it, and when it was created. I still recommend checking out the module’s GitHub page directly. There you’ll be able to determine if the module is being actively maintained or not. If you access the npm website page for the module, you’ll also get an idea of how popular the module is.

Converting Your Library into a Node Module

Problem

You want to use one of your libraries in Node.

Solution

Convert the library into a Node module. For example, if the library is designed as the following IIFE:

(function () {
   var val = 'world';
   console.log('Hello ' + val + ' from two');
}());

You can convert it to work with Node by the simple addition of an exports keyword:

module.exports = (function () {
            return {
              hi: function(val) {
                    console.log('Hello ' + val + ' from two');
                  }
              };
            }());

You can then use the module in your application:

var two = require('./two.js');

two.hi('world');

Discussion

Node’s module system is based on CommonJS, the second modular system covered in this chapter. CommonJS uses three constructs: exports to define what’s exported from the library, require() to include the module in the application, and module, which includes information about the module but also can be used to export a function, directly.

Though the solution maintains the IIFE, it’s not really required in the CommonJS environment, because every module operates in its own module space. The following is also acceptable:

module.exports.hi = function (val) {
   console.log('hello ' + val + ' from two');
}

If your library returns an object with several functions and data objects, you can assign each to the comparably named property on module.exports, or you could return an object from a function:

module.exports = function () {
            return {
              somedata: 'some data',
              hi: function(val) {
                    console.log('Hello ' + val + ' from two');
                  }
              };
            };

And then invoke the object in the application:

var twoObj = require('./two.js');

var two = twoObj();
two.hi(two.somedata);

Or you can access the object property directly:

var hi = require('./twob.js').hi;

hi('world');

Because the module isn’t installed using npm, and just resides in the directory where the application resides, it’s accessed by the location and name, not just the name.

See Also

In the next section, I cover how to make sure your library code works in all of the environments: CommonJS, Node, AMD, and as a traditional JavaScript library.

Taking Your Code Across All Module Environments

Problem

You’ve written a library that you’d like to share with others, but folks are using a variety of module systems to incorporate external JavaScript. How can you ensure your library works in all of the various environments?

Solution

The following library with two functions:

function concatArray(str, array) {
  return array.map(function(element) {
       return str + ' ' + element;
  });
}

function splitArray(str,array) {
  return array.map(function(element) {
       var len = str.length + 1;
       return element.substring(len);
  });
}

Will work with RequireJS, Node, as a plain script, and CommonJS in the browser when converted to:

(function(global) {
     'use strict';

     var bbArray = {};

     bbArray.concatArray = function (str, array) {
             return array.map(function(element) {
               return str + ' ' + element;
             });
          };

     bbArray.splitArray = function (str,array) {
             return array.map(function(element) {
               var len = str.length + 1;
               return element.substring(len);
             });
          };

     if (typeof module != 'undefined' && module.exports) {
        module.exports = bbArray;
     } else if ( typeof define === "function" && define.amd ) {
         define( "bbArray", [], function() {
            return bbArray;
         });
     } else {
        global.bbArray = bbArray;
     }

}(this));

Discussion

To ensure your library works in a traditional scripting environment, you should encapsulate your functionality in an IIFE, to minimize leak between private and public functionality and data. You’ll also want to limit pollution of the global space:

(function(global) {
     'use strict';

     var bbArray = {};

     bbArray.concatArray = function (str, array) {
             return array.map(function(element) {
               return str + ' ' + element;
             });
          };

     bbArray.splitArray = function (str,array) {
             return array.map(function(element) {
               var len = str.length + 1;
               return element.substring(len);
             });
          };

     global.bbArray = bbArray;

}(this));

The object is being used in an environment that may not have access to a window object, so the global object (global in Node, window in the browser) is passed as an argument to the object as this, and then defined as global in the library.

At this point, the library can work as a traditional library in a browser application:

<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <title>Array test</title>
  <script src="bbarray.js" type="text/javascript">
  </script>
  <script type="text/javascript">
     var a = ['one', 'two', 'three'];
     var b = bbArray.concatArray('number is ',a);
     console.log(b);
     var c = bbArray.splitArray('number is ', b);
     console.log(c);
  </script>
</head>
<body>
</body>
</html>

The result is two print outs to the console:

[ 'number is  one', 'number is  two', 'number is  three' ]
[ 'one', 'two', 'three' ]

Next, we’ll add the Node support. We add this using the following lines of code:

if (typeof module != 'undefined' && module.exports) {
  module.exports = bbArray;
}

This code checks whether the module object is defined and if it is, whether the module.exports object exists. If the tests succeed, then the object is assigned to module.exports, no different than defining exported functionality (covered earlier in “Converting Your Library into a Node Module”). It can now be accessed in a Node application like the following:

var bbArray = require('./bbarray.js');

var a = ['one', 'two', 'three'];
var b = bbArray.concatArray('number is ',a);
console.log(b);
var c = bbArray.splitArray('number is ', b);
console.log(c);

Now we add support for CommonJS, specifically RequireJS. From “Using RequireJS with jQuery or Another Library”, we know to check if define exists, and if so, to add support for RequireJS. After adding this modification, the library module now looks like this:

(function(global) {
     'use strict';

     var bbArray = {};

     bbArray.concatArray = function (str, array) {
             return array.map(function(element) {
               return str + ' ' + element;
             });
          };

     bbArray.splitArray = function (str,array) {
             return array.map(function(element) {
               var len = str.length + 1;
               return element.substring(len);
             });
          };


     if (typeof module != 'undefined' && module.exports) {
        module.exports = bbArray;
     } else if ( typeof define === "function" && define.amd ) {
         define( "bbArray", [], function() {
            return bbArray;
         });
     } else {
        global.bbArray = bbAarray;
     }

}(this));

The module can now be used in a web application that incorporates RequireJS for module support. Following RequireJS’s suggestion that all inline scripts be pulled into a separate file, the JavaScript application to test the library is created in a file named main.js:

require(["./bbarray"], function(bbArray) {
    var a = ['one', 'two', 'three'];
    var b = bbArray.concatArray('number is ',a);
    console.log(b);
    var c = bbArray.splitArray('number is ', b);
    console.log(c);
});

And the web page incorporates the RequireJS script, loaded via CDN:

<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <title>Array test</title>
  <script src="//cdnjs.cloudflare.com/ajax/libs/require.js/2.1.14/require.min.js"
    data-main="main">
  </script>
</head>
<body>

</body>
</html>

Modify the URL for Require.js to match what’s available at the CDN when you run the test.

See Also

The example covered in this recipe works in all of our environments but it has one limitation: it’s not using any other libraries. So what happens when you need to include libraries?

This is where things can get ugly. We know that CommonJS/Node import dependencies with require:

var library = require('somelib');

While AMD incorporates dependencies in require or define:

define(['./somelib'], function(library) {

// rest of the code
});

Not compatible. At all. The workaround for this problem has been either to use Browserify (covered in “Compiling Node.js Modules for Use in the Browser with Browserify”) or to incorporate a Universal Module Definition (UMD). You can see examples of a UMD online, and it’s covered in detail in Addy Osmani’s “Writing Modular JavaScript with AMD, CommonJS, and ES Harmony”.

Creating an Installable Node Module

Problem

You’ve either created a Node module from scratch, or converted an existing library to one that will work in the browser or in Node. Now, you want to know how to modify it into a module that can be installed using npm.

Solution

Once you’ve created your Node module and any supporting functionality (including module tests), you can package the entire directory. The key to packaging and publishing the Node module is creating a package.json file that describes the module, any dependencies, the directory structure, what to ignore, and so on.

The following is a relatively basic package.json file:

{
  "name": "bbArray",
  "version": "0.1.0",
  "description": "A description of what my module is about",
  "main": "./lib/bbArray",
  "author": {
    "name": "Shelley Powers"
  },
  "keywords": [
    "array",
    "utility"
  ],
  "repository": {
    "type": "git",
    "url": "https://github.com/accountname/bbarray.git"
  },
  "engines" : {
    "node" : ">=0.10.3 <0.12"
  },
  "bugs": {
    "url": "https://github.com/accountname/bbarray/issues"
  },
  "licenses": [
    {
      "type": "MIT",
      "url": "https://github.com/accountname/bbarray/raw/master/LICENSE"
    }
  ],
  "dependencies": {
     "some-module": "~0.1.0"
  },
  "directories":{
     "doc":"./doc",
     "man":"./man",
     "lib":"./lib",
     "bin":"./bin"
  },
  "scripts": {
    "test": "nodeunit test/test-bbarray.js"
  }
 }

Once you’ve created package.json, package all the source directories and the package.json file as a gzipped tarball. Then install the package locally, or install it in npm for public access.

Discussion

The package.json file is key to packaging a Node module up for local installation or uploading to npm for management. At a minimum, it requires a name and a version. The other fields given in the solution are:

  • description: A description of what the module is and does

  • main: Entry module for application

  • author: Author(s) of the module

  • keywords: List of keywords appropriate for module

  • repository: Place where code lives, typically GitHub

  • engines: Node version you know your module works with

  • bugs: Where to file bugs

  • licenses: License for your module

  • dependencies: Any module dependencies

  • directories: A hash describing directory structure for your module

  • scripts: A hash of object commands that are run during module lifecycle

There are a host of other options, which are described at the npm website. You can also use a tool to help you fill in many of these fields. Typing the following at the command line runs the tool that asks questions and then generates a basic package.json file:

npm init

Once you have your source set up and your package.json file, you can test whether everything works by running the following command in the top-level directory of your module:

npm install . -g

If you have no errors, then you can package the file as a gzipped tarball. At this point, if you want to publish the module, you’ll first need to add yourself as a user in the npm registry:

npm add-user

To publish the Node module to the npm registry, use the following in the root directory of the module, specifying a URL to the tarball, a filename for the tarball, or a path:

npm publish ./

If you have development dependencies for your module, such as using a testing framework like Mocha, one excellent shortcut to ensure these are added to your package.json file is to use the following, in the same directory as the package.json file, when you’re installing the dependent module:

npm install -g mocha --save-dev

Not only does this install Mocha (discussed later, in “Unit Testing Your Node Modules”), this command also updates your package.json file with the following:

 "devDependencies": {
    "grunt": "^0.4.5",
    "grunt-contrib-jshint": "^0.10.0",
    "mocha": "^1.21.4"
  }

You can also use this same type of option to add a module to dependencies in package.json. The following:

npm install d3 --save

adds the following to the package.json file:

"dependencies": {
    "d3": "^3.4.11"
  }

If the module is no longer needed and shouldn’t be listed in package.json, remove it from the devDependencies with:

npm remove mocha --save-dev

And remove a module from dependencies with:

npm remove d3 --save

If the module is the last in either dependencies or devDependencies, the property isn’t removed. It’s just set to an empty value:

"dependencies": {}
Note

npm provides a decent developer guide for creating and installing a Node module. You should consider the use of an .npmignore file for keeping stuff out of your module. And though this is beyond the scope of the book, you should also become familiar with Git and GitHub, and make use of it for your applications/modules.

Extra: The README File and Markdown Syntax

When you package your module or library for reuse and upload it to a source repository such as GitHub, you’ll need to provide how-to information about installing the module/library and basic information about how to use it. For this, you need a README file.

You’ve seen files named README.md or readme.md with applications and Node modules. They’re text-based with some odd, unobtrusive markup that you’re not sure is useful, until you see it in a site like GitHub, where the README file provides all of the project page installation and usage information. The markup translates into HTML, making for readable Web-based help.

The content for the README is marked up with annotation known as Markdown. The popular website Daring Fireball calls Markdown easy to read and write, but “Readability, however, is emphasized above all else.” Unlike with HTML, the Markdown markup doesn’t get in the way of reading the text.

Note

Daring Fireball also provides an overview of generic Markdown, but if you’re working with GitHub files, you might also want to check out GitHub’s Flavored Markdown.

In [link unavailable], I created a simple Firefox OS mobile app named “Where Am I?” Part of its installation is a README.md file that provides information about using the app. The following is a brief excerpt from the file:

# Where Am I?

This is a simple demonstration Firefox OS app that uses the Geolocation API
to get the user's current location, and then loads a static map into the page.

## Obtaining

The Where Am I? app is hosted on the web, in a [Burningbird work directory]
(http://burningbird.net/work/whereami)

## Usage

Import it into the Mozilla WebIDE using the hosted app option, and then run
the app in one or more simulators.

When I use a CLI tool like Pandoc, I can covert the README.md file into readable HTML:

pandoc README.md -o readme.html

Figure 3-1 displays the generated content. It’s not fancy, but it is imminently readable.

jscb 1201
Figure 3-1. Generated HTML from README.md text and Markdown annotation

When you install your source in a site such as GitHub (discussed in [link unavailable]), GitHub uses the README.md file to generate the cover page for the repository.

Packaging and Managing Your Client-Side Dependencies with Bower

Problem

You really like how npm manages dependencies and wish there was something comparable for the client.

Solution

Bower can help you manage client dependencies. To use it you must have Node, npm, and support for Git installed on your client or server.

Once your environment is set up, install Bower using npm:

npm install -g bower

Now, to add packages to the bower-components subdirectory, install them with bower:

bower install jquery

Then you can create a bower.json file by typing the following in the root directory of your library or application:

bower init

The application asks a set of questions and generates a bower.json file, which can be used to install the dependencies with another simple command:

bower install

Discussion

Bower is a way of keeping your script and other dependencies collected and up to date. Unlike npm, it can work with a variety of file extensions, including CSS, images, as well as script. You can use it to install dependencies in bower-components, and then access the dependencies directly in your web applications:

<script src="path/to/bower_components/d3/d3.min.js"></script>

You can package all of your application’s dependencies in a bower.json file, and reinstall them in a fresh directory with a simple command (in the same directory as the bower.json file):

bower install

To ensure you’re using the latest and greatest version of the module and library, update your dependencies:

bower update

If your application is publicly available on GitHub, you can register its dependencies in Bower by, first, ensuring the bower.json file for the application is accurate, you’re using semantic versioning with your Git tags, your application is publicly available as a Git end point (such as GitHub), and the package name adheres to the bower.json specification. Once these dependencies are met, register the application:

bower register <package-name> <git-endpoint>

If you’re wondering why you can’t use something like require directly with Bower, remember that it’s a dependency management tool, just like npm. It’s the libraries and infrastructure in place, such as RequireJS, that allows you to use modular AMD or CommonJS techniques.

Note

You can read more about using Bower at the application’s website.

Bower can be used with other tools, such as Grunt, demonstrated later in “Running Tasks with Grunt”.

Compiling Node.js Modules for Use in the Browser with Browserify

Problem

Node has a lot of really great modules that you’d really like to use in your browser.

Solution

You can use Browserify to compile the Node module into browser accessible code. If it’s one of the Node core modules, many are already compiled into shims that can be used in your browser application.

For instance, if you’re interested in using the Node querystring module functionality, you create a client JavaScript bundle using the following Browserify command:

browserify -r querystring > bundle.js

Then use the module in your browser app:

 <script src="bundle.js" type="text/javascript">
  </script>
  <script type="text/javascript">
   var qs = require('querystring');

   var str = qs.stringify({ first: 'apple', second: 'pear', third: 'pineapple'})
;
   console.log(str); //first=apple&second=pear&third=pineapple
  </script

Discussion

Browserify is a tool that basically moves Node functionality to the browser, as long as doing so makes sense. Of course, some functionality won’t work (think input/output) but a surprising amount of functionality, including that in Node core, can work in the browser.

Browserify is installed via npm:

npm install -g browserify

It runs at the command line, as shown in the solution. In the solution, the -r flag triggers Browserify into creating a require() function to wrap the module’s functionality, so we can use it in a similar manner in the browser app. The querystring module is one of the many Node core modules already compiled as a shim. The others are:

  • assert

  • buffer

  • console

  • constants

  • crypto

  • domain

  • events

  • http

  • https

  • os

  • path

  • punycode

  • querystring

  • stream

  • string_decoder

  • timers

  • tty

  • url

  • util

  • vm

  • zlib

You can also compile other Node modules into browser code, including your own. As an example, let’s say I have the following three Node files:

one.js

module.exports = function() {
   console.log('hi from one');
};

two.js

var one = require ('./one');

module.exports = function(val) {
   one();
   console.log('hi ' + val + ' from two');
};

index.js

var two = require ('./two');

module.exports = function() {
  two('world');
  console.log("And that's all");
}

I compiled it into an appl.js file using the following:

browserify ./index.js -o ./appl.js

Including the library in a web page results in the same three console log() function calls as you would see if you ran the original index.js file with Node, as soon as the generated script file is loaded.

Unit Testing Your Node Modules

Problem

You want to know the best way to ensure your module is ready for others to try.

Solution

Add unit tests as part of your production process.

Given the following module, named bbarray, and created in a file named index.js in the module directory:

var util = require('util');

(function(global) {
     'use strict';

     var bbarray = {};

     bbarray.concatArray = function (str, array) {
         if (!util.isArray(array) || array.length === 0) {
             return -1;
         } else if (typeof str != 'string') {
             return -1;
         } else {
             return array.map(function(element) {
               return str + ' ' + element;
             });
         }
     };
     bbarray.splitArray = function (str,array) {
         if (!util.isArray(array) || array.length === 0) {
             return -1;
         } else if (typeof str != 'string') {
             return -1;
         } else {
             return array.map(function(element) {
               var len = str.length + 1;
               return element.substring(len);
             });
         }
      };
     if (typeof module != 'undefined' && module.exports) {
        module.exports = bbarray;
     } else if ( typeof define === "function" && define.amd ) {
         define( "bbarray", [], function() {
            return bbarray;
         });
     } else {
        global.bbarray = bbaarray;
     }

}(this));

Using Mocha, a JavaScript testing framework, and Node’s built-in assert module, the following unit test (created as index.js and located in the project’s test subdirectory) should result in the successful pass of six tests:

var assert = require('assert');
var bbarray = require('../index.js');

describe('bbarray',function() {
   describe('#concatArray()', function() {
     it('should return -1 when not using array', function() {
        assert.equal(-1, bbarray.concatArray(9,'str'));
     });
     it('should return -1 when not using string', function() {
        assert.equal(-1, bbarray.concatArray(9,['test','two']));
     });
     it('should return an array with proper args', function() {
        assert.deepEqual(['is test','is three'],
                      bbarray.concatArray('is',['test','three']));
     });
   });
   describe('#splitArray()', function() {
     it('should return -1 when not using array', function() {
       assert.equal(-1, bbarray.splitArray(9,'str'));
     });
     it('should return -1 when not using string', function() {
       assert.equal(-1, bbarray.splitArray(9,['test','two']));
     });
     it('should return an array with proper args', function() {
       assert.deepEqual(['test','three'],
                        bbarray.splitArray('is',['is test','is three']));
     });
   });
});

The result of the test is shown in Figure 3-2, run using npm test.

jscb 1202
Figure 3-2. Running unit tests based on Node Assert and Mocha

Discussion

Unit testing is one of those development tasks that may seem like a pain when you first start, but can soon become second nature. I don’t necessarily agree with the folks that believe we should write the unit tests (test-driven development) first, before writing the code. But developing both test and code in parallel to each other should be a goal.

A unit test is a way that developers test their code to ensure it meets the specifications. It involves testing functional behavior, and seeing what happens when you send bad arguments—or no arguments at all. It’s called unit testing because it’s used with individual units of code, such as testing one module in a Node application, as compared to testing the entire Node application. It becomes one part of integration testing, where all the pieces are plugged together, before going to user acceptance testing: testing to ensure that the application does what users expect it to do (and that they generally don’t hate it when they use it).

In the solution, I use two different functionalities for testing: Node’s built-in assert module, and Mocha, a sophisticated testing framework. My module is simple, so I’m not using some of the more complex Mocha testing mechanisms. However, I think you’ll get a feel for what’s happening.

To install Mocha, use the following:

npm install mocha --save-dep

I’m using the --save-dep flag, because I’m installing Mocha into the module’s Node dependencies. In addition, I modify the module’s package.json file to add the following section:

 "scripts": {
    "test": "node_modules/mocha/bin/mocha test"
  },

The test script is saved as index.js in the test subdirectory under the project. The following command runs the test:

npm test

The Mocha unit test makes use of assertion tests from Node’s assert module.

Running Tasks with Grunt

Problem

Pulling your Node module together is getting more complex—too complex to manually manage all of the elements.

Solution

Use a task runner like Grunt to manage all the bits for you.

For the following bbarray module:

var util = require('util');

(function(global) {
     'use strict';

     var bbarray = {};

     bbarray.concatArray = function (str, array) {
         if (!util.isArray(array) || array.length === 0) {
             return -1;
         } else if (typeof str != 'string') {
             return -1;
         } else {
             return array.map(function(element) {
               return str + ' ' + element;
             });
         }
     };

     bbarray.splitArray = function (str,array) {
         if (!util.isArray(array) || array.length === 0) {
             return -1;
         } else if (typeof str != 'string') {
             return -1;
         } else {
             return array.map(function(element) {
               var len = str.length + 1;
               return element.substring(len);
             });
         }
      };

     if (typeof module != 'undefined' && module.exports) {
        module.exports = bbarray;
     } else if ( typeof define === 'function' && define.amd ) {
         define( 'bbarray', [], function() {
            return bbarray;
         });
     } else {
        global.bbarray = bbaarray;
     }

}(this));

Saved as bbarray.js in the root directory, with a Mocha test file:

var assert = require('assert');
var bbarray = require('../bbarray.js');

describe('bbarray',function() {
   describe('#concatArray()', function() {
     it('should return -1 when not using array', function() {
        assert.equal(-1, bbarray.concatArray(9,'str'));
     });
     it('should return -1 when not using string', function() {
        assert.equal(-1, bbarray.concatArray(9,['test','two']));
     });
     it('should return an array with proper args', function() {
        assert.deepEqual(['is test','is three'],
                      bbarray.concatArray('is',['test','three']));
     });
   });
   describe('#splitArray()', function() {
     it('should return -1 when not using array', function() {
       assert.equal(-1, bbarray.splitArray(9,'str'));
     });
     it('should return -1 when not using string', function() {
       assert.equal(-1, bbarray.splitArray(9,['test','two']));
     });
     it('should return an array with proper args', function() {
       assert.deepEqual(['test','three'],
                        bbarray.splitArray('is',['is test','is three']));
     });
   });
});

Saved as index.js in a test subdirectory, the Grunt file is:

module.exports = function(grunt) {
    var banner = '/*\n<%= pkg.name %> <%= pkg.version %>';
    banner += '- <%= pkg.description %>\n<%= pkg.repository.url %>\n';
    banner += 'Built on <%= grunt.template.today("yyyy-mm-dd") %>\n*/\n';

    grunt.initConfig({
        pkg: grunt.file.readJSON('package.json'),
        jshint: {
            files: ['gruntfile.js', 'src/*.js'],
            options: {
                maxlen: 80,
                quotmark: 'single'
            }
        },
        uglify: {
            options: {
                banner: banner,
            },
            build: {
                files: {
                    'build/<%= pkg.name %>.min.js':
                        ['build/<%= pkg.name %>.js'],
                }
            }
        },
        simplemocha: {
            options: {
                globals: ['assert'],
                timeout: 3000,
                ignoreLeaks: false,
                ui: 'bdd',
                reporter: 'tap'
            },
            all: { src: ['test/*.js'] }
        }
    });

    grunt.loadNpmTasks('grunt-contrib-jshint');
    grunt.loadNpmTasks('grunt-contrib-uglify');
    grunt.loadNpmTasks('grunt-simple-mocha');

    grunt.registerTask('default',
        ['jshint', 'simplemocha', 'uglify']);
};

When the file is saved as gruntfile.js, Grunt runs all the tasks defined in the file:

grunt

Discussion

Grunt is a task runner. Its only purpose is to consistently run a series of tasks. It’s similar to the old Makefile, but without the decades of musty history.

To use Grunt, install it first:

npm install -g grunt-cli

Grunt needs to run in the same directory as your application/module’s package.json file, as it works with the file. You can create either a JavaScript or Coffee-based Grunt file, but I’m focusing on the JS version.

Create the file by using the grunt-init CLI, with a given template, or you can use the example file given in the Getting Started Guide.

A module needs to run within a certain framework to work with Grunt. Luckily, plugins have been created for many of the commonly used modules, such as the plugins used in the example for JSHint, Uglify, and Mocha. To ensure they’re listed in the package.json file, they need to be installed using --save-dev:

npm install grunt-contrib-jshint --save-dev
npm install grunt-simple-mocha --save-dev
npm install grunt-contrib-uglify --save-dev

Each plugin also provides instructions about how to modify the Gruntfile to use the plugin and process your files.

Once you have both the package.json and gruntfile.js files running, the following will install any of the dependencies in the file, and run the Grunt tasks:

npm install
grunt

The result of running Grunt with the file in the solution is:

Running "jshint:files" (jshint) task
>> 1 file lint free.

Running "simplemocha:all" (simplemocha) task
1..6
ok 1 bbarray concatArray() should return -1 when not using array
ok 2 bbarray concatArray() should return -1 when not using string
ok 3 bbarray concatArray() should return an array with proper args
ok 4 bbarray splitArray() should return -1 when not using array
ok 5 bbarray splitArray() should return -1 when not using string
ok 6 bbarray splitArray() should return an array with proper args
# tests 6
# pass 6
# fail 0

Running "uglify:build" (uglify) task
>> Destination build/bbarray.min.js not written because src files were empty.

Done, without errors.

There are no files in the src directory, but I left the instructions in the Grunt file, for future expansion of the module.

See Also

Read all about Grunt, and check out the available plugins, at the application’s website.

Another popular build system is Gulp.

Get Modern JavaScript now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.