citizen

citizen

citizen is an event-driven MVC framework for Node.js web applications. Its purpose is to handle serving, routing, and event emitter creation, while providing some useful helpers to get you on your way; the nuts and bolts of your application are up to you. citizen favors convention over configuration, and those conventions are covered throughout this guide.

Version 0.2.0 contains many breaking changes and new features. For example, configuration files are parsed differently and I've added caching capabilities. If you've built an app based on citizen, you'll want to read this documentation thoroughly before upgrading.

citizen is in beta. Your comments, criticisms, and requests are appreciated.

citizen's static file serving is just a hack to get your dev environment up and running quickly. I recommend something like nginx as a front end for static file serving in your production environment.

Getting Started with citizen

Installing citizen

$ npm install citizen

Application Directory Structure

Here's the most basic directory structure of a citizen web app:

app/
  patterns/
    controllers/
      index.js
    models/
      index.js      // Optional
    views/
      index/
        index.jade  // You can use Jade (.jade), Handlebars (.hbs), or HTML files
  start.js
web/
  // public static assets

Here's a more complex app example (more about config and on directories later):

app/
  config/
    local.json
    qa.json
    production.json
  logs/
  on/
    application.js
    request.js
    response.js
    session.js
  patterns/
    controllers/
      index.js
    models/
      index.js
    views/
      index/
        index.jade
        index-alt.jade
  start.js
web/

Initializing citizen and starting the web server

The start.js file in your app directory can be as simple as this:

// start.js

global.app = require('citizen');

app.start();

Run start.js from the command line:

$ node start.js

Objects created by citizen
app.start()	Starts the web server
app.cache() app.exists() app.retrieve() app.clear() app.listen() app.copy() app.extend() app.isNumeric() app.dashes() app.size()	Helpers used internally by citizen, exposed publicly since you might find them useful
app.models	Contains models from your supplied patterns, which you can use instead of require. Controllers and views aren't exposed this way because you don't need to access them directly.
app.handlebars	A pointer to the citizen Handlebars global, allowing you full access to Handlebars methods such as app.handlebars.registerHelper()
app.jade	A pointer to the citizen Jade global
app.config	The configuration settings you supplied at startup
CTZN	The global namespace used by citizen for internal objects, user sessions, and cache. You should not access or modify this namespace directly; anything you might need in your application will be exposed by the server to your controllers through local scopes.

Configuration

citizen follows convention over configuration, but some things are best handled by a config file.

The config directory is optional and contains configuration files that drive both citizen and your app in JSON format. You can have multiple citizen configuration files within this directory, allowing different configurations based on environment. citizen retrieves its configuration file from this directory based on the following logic:

1. citizen parses each JSON file looking for a hostname key that matches the machine's hostname. If it finds one, it loads that configuration.
2. If it can't find a matching hostname key, it looks for a file named citizen.json and loads that configuration.
3. If it can't find citizen.json or you don't have a config directory, it runs under its default configuration.

The following represents citizen's default configuration, which is extended by your citizen configuration file:

{
  "citizen": {
    "mode":               "production",
    "directories": {
      "app":              "[absolute path to start.js]",
      "logs":             "[directories.app]/logs",
      "on":               "[directories.app]/on",
      "controllers":      "[directories.app]/patterns/controllers",
      "models":           "[directories.app]/patterns/models",
      "views":            "[directories.app]/patterns/views",
      "web":              "[directories.app]../web"
    },
    "urlPaths": {
      "app":              "",
      "fileNotFound":     "/404.html"
    },
    "httpPort":           80,
    "hostname":           "localhost", // Hostname for accepting requests
    "connectionQueue":    undefined,
    "logs": {
      "console":          true,
      "file":             false
    },
    "sessions":           false,
    "sessionTimeout":     1200000, // In ms (20 minutes)
    "requestTimeout":     30000, // In ms (30 seconds)
    "mimetypes":          [parsed from internal config],
    "debug": {
      "output":           "console",
      "depth":            2,
      "jade":             false
    }
  }
}

These settings are exposed publicly via app.config.citizen.

If you want to add a database configuration for your local dev environment, you could do it like this:

{
  "hostname":             "My-MacBook-Pro",
  "citizen": {
    // your custom citizen config
  },
  "db": {
    "server":             "127.0.0.1",
    "username":           "dbuser",
    "password":           "dbpassword"
  }
}

This config file would extend the default configuration when running on your local machine. The database settings would be accessible within your app via app.config.db.

urlPaths.app is the path name in your app's web address. If your app's URL is:

http://www.website.com/to/my-app

urlPaths.app should be "/to/my-app". This is necessary for citizen's router to work.

Routing and URLs

Apps using citizen have a simple URL structure that determines which controller to fire, passes URL parameters, and makes a bit of room for SEO-friendly content that can double as a unique identifier. The structure looks like this:

http://www.site.com/controller/content-description/param/val/param2/val2

For example, let's say your site's base URL is:

http://www.cleverna.me

Requesting that URL would cause the index controller to fire, because the index controller is the default. The following URL would also cause the index controller to fire:

http://www.cleverna.me/index

If you have an article controller, you'd request it like this:

http://www.cleverna.me/article

Instead of query strings, citizen uses an SEO-friendly method of passing URL parameters consisting of name/value pairs. If you had to pass an article ID of 237 and a page number of 2, you'd append name/value pairs to the URL:

http://www.cleverna.me/article/id/237/page/2

Valid parameter names may contain letters, numbers, underscores, and dashes, but must start with a letter or underscore.

citizen also lets you optionally insert relevant content into your URLs, like so:

http://www.cleverna.me/article/My-Clever-Article-Title/page/2

This SEO content must always follow the pattern name and precede any name/value pairs. You can access it generically via route.descriptor or specifically via the url scope (url.article in this case), which means you can use it as a unique identifier (more on URL parameters in the Controllers section).

MVC Patterns

citizen relies on a simple model-view-controller convention. The article pattern mentioned above might use the following structure:

app/
  patterns/
    controllers/
      article.js
    models/
      article.js
    views/
      article/        // Matches the controller name
        article.jade  // Matches the controller name, making it the default view
        edit.jade     // Secondary view for editing an article

At least one controller is required for a given URL, and a controller's default view directory and default view file must share its name. Additional views should reside in this same directory. More on views in the Views section.

Models and views are optional and don't necessarily need to be associated with a particular controller. If your controller doesn't need a model, you don't need to create one. If your controller is going to pass its output to another controller for further processing and final rendering, you don't need to include a matching view. (See the controller handoff directive.)

Controllers

Each controller requires at least one public function. The default action is named handler(), which is called by citizen when no action is specified in the URL.

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {

  // do some stuff

  emitter.emit('ready', {
    // content and directives for the server
  });
}

The citizen server calls handler() after it processes the initial request and passes it 3 arguments: an object containing the parameters of the request, the current request's context generated by the app up to this point, and an emitter for the controller to emit when it's ready to pass the results to the server.

Contents of the params argument
request	The request object generated by the server, just in case you need direct access
response	The response object generated by the server
route	Details of the route, such as the requested URL and the name of the route (controller)
url	Any URL parameters that were passed including the descriptor, if provided
form	Data collected from a POST
payload	Data collected from a PUT
cookie	An object containing any cookies that were sent with the request
session	An object containing any session variables, if sessions are enabled

In addition to having access to these objects within your controller, they are also included in your view context automatically so you can reference them within your view templates as local variables (more details in the Views section).

For example, based on the previous article URL...

http://www.cleverna.me/article/My-Clever-Article-Title/page/2

...you'll have the following params.url object passed to your controller:

{
  article: 'My-Clever-Article-Title',
  page: 2
}

The controller name becomes a property in the URL scope that contains the SEO content, which makes it well-suited for use as a unique identifier. This content is also available in the route object as route.descriptor.

The context argument contains any output that's been generated by the request up to this point. There are various events that can populate this argument with content and directives, which are then passed to your controller so you can access that content or see what directives have been set by previous events.

The emitter argument is the method by which the controller lets the server know that it's done with its tasks and ready to render the result. The emitter should emit a "ready" event when the controller has accomplished its task. This lets the server know it's okay to proceed. As part of this event, the emitter should include any view content and directives for the server.

Using the above URL parameters, I can retrieve the article content from the model and pass it back to the server:

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  // Get the article content
  var article = app.models.article.getArticle(params.url.article, params.url.page);

  // Emit the 'ready' event and pass any objects you want added to the view
  // context via the content object
  emitter.emit('ready', {
    content: {
      article: article
    }
  });
};

Alternate actions can be requested using the action URL parameter. For example, maybe we want a different action and view to edit an article:

http://www.cleverna.me/article/My-Clever-Article-Title/page/2/action/edit

// article controller

module.exports = {
  handler: handler,
  edit: edit
};

function handler(params, context, emitter) {
  // Get the article content
  var article = app.models.article.getArticle(params.url.article, params.url.page);

  // Emit the 'ready' event and pass any objects you want added to the view
  // context via the content object
  emitter.emit('ready', {
    content: {
      article: article
    }
  });
};

function edit(params, context, emitter) {
  // Get the article content
  var article = app.models.article.getArticle(params.url.article, params.url.page);

  // Emit the 'ready' event and pass any objects you want added to the view
  // context via the content object. Use the /patterns/views/article/edit.jade
  // view for this action (more on alternate views in later sections).
  emitter.emit('ready', {
    content: {
      article: article
    },
    view: 'edit'
  });
};

The second argument in emitter.emit is an object containing any data you want to pass back to citizen. All the content you want to render in your view should be passed to citizen within an object called content, as shown above. Additional objects can be passed to citizen to set directives that provide instructions to the server (explained later in the Emitter Directives section). You can even add your own objects to the request context and pass them from controller to controller (more in the Controller Handoff section.)

Private controllers

To make a controller private—inaccessible via HTTP, but accessible within your app—add a plus sign (+) to the beginning of the file name:

app/
  patterns/
    controllers/
      +_header.js  // Partial, only accessible internally
      _head.js     // Partial, accessible via www.cleverna.me/_head
      article.js   // Accessible via www.cleverna.me/article

Models

Models are optional and their structure is completely up to you. citizen doesn't talk to your models directly; it only stores them in app.models for your convenience.

Here's a simple static model for the article pattern (just an example, because storing content this way is awful):

// article model

exports.getArticle = getArticle;

function getArticle(article, page) {
  var articles = {
    'My-Clever-Article-Title': {
      title: 'My Clever Article Title',
      summary: 'Am I not terribly clever?',
      pages: {
        '1': 'First page content',
        '2': 'Second page content'
      }
    },
    'Clever-Part-II-The-Sequel': {
      title: 'Clever Part II: The Sequel',
      summary: 'Too clever for just one article.',
      pages: {
        '1': 'First page content',
        '2': 'Second page content'
      }
    }
  };

  return {
    title: articles[article]['title'],
    summary: articles[article]['summary'],
    text: articles[article]['pages'][page]
  };
};

Views

citizen supports Jade and Handlebars templates, as well as good old HTML. You can even mix and match Jade, Handlebars, and HTML templates as you see fit; just use the appropriate file extensions (.jade, .hbs, or .html) and citizen will compile and render each view with the appropriate engine.

You have direct access to each engine's methods via app.handlebars and app.jade, allowing you to use methods like app.handlebars.registerHelper() to create global helpers. Keep in mind that you're extending the global Handlebars and Jade objects, potentially affecting citizen's view rendering if you do anything wacky because citizen relies on these same objects.

In article.jade, you can reference objects you placed within the content object passed by the emitter. citizen also injects the params object into your view context automatically, so you have access to those objects as local variables (such as the url scope):

// article.jade

doctype html
html
  body
    main
      h1 #{article.title} - Page #{url.page}
      p#summary #{article.summary}
      #text #{article.text}

citizen sends HTML to the client by default, but you can also return JSON and JSONP with no extra work on your part.

JSON

You don't need a custom view just for JSON. You can tell a controller to return its content as plain text JSON by adding the format URL parameter.

http://www.cleverna.me/article/My-Clever-Article-Title/page/2/format/json

Returns...

{
  "article": {
    "My-Clever-Article-Title": {
      "title": "My Clever Article Title",
      "summary": "Am I not terribly clever?",
      "text": "Second page content"
    }
  }
}

Whatever you've added to the controller's emitter content object will be returned. (The line breaks are just for readability. The actual output is compressed.)

JSONP

JSONP is pretty much the same. Use format and callback in the URL:

http://www.cleverna.me/article/My-Clever-Article-Title/format/jsonp/callback/foo

Returns:

foo({
  "article": {
    "My-Clever-Article-Title": {
      "title": "My Clever Article Title",
      "summary": "Am I not terribly clever?",
      "text": "Second page content"
    }
  }
});

If you want to make a controller available to third party sites, see the CORS section.

Rendering alternate views

By default, the server renders the view whose name matches that of the controller. To render a different view, use the view directive in your emitter.

Emitter Directives

In addition to the view content, the controller's ready emitter can also pass directives to render alternate views, set cookies and session variables, initiate redirects, call and render includes, cache views (or entire routes), and hand off the request to another controller for further processing.

Alternate Views

By default, the server renders the view whose name matches that of the controller. To render a different view, use the view directive in your emitter:

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  var article = app.models.article.getArticle(params.url.article, params.url.page);

  emitter.emit('ready', {
    content: {
      article: article
    },
    // This tells the server to render app/patterns/views/article/edit.jade
    view: 'edit'
  });
}

Cookies

You set cookies by appending a cookie object to the emitter context.

Here's an example of a complete cookie object's default values:

cookie.foo = {
  value: '',
  // Valid expiration options are:
  // 'now' - deletes an existing cookie
  // 'never' - current time plus 30 years, so effectively never
  // 'session' - expires at the end of the browser session (default)
  // [time in milliseconds] - added to current time for a specific expiration date
  expires: 'session',
  path: '/',
  httpOnly: true,
  secure: false
}

The following sample login controller tells the server to set username and passwordHash cookies that never expire:

// login controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  var authenticate = app.models.login.authenticate({
        // Form values, just like URL parameters, are passed via the params
        // argument
        username: params.form.username,
        password: params.form.password
      }),
      // If a directive is an empty object, that's fine. citizen just ignores it.
      cookie = {};

  if ( authenticate.success ) {
    cookie = {
      // The cookie gets its name from the property name
      username: {
        value: authenticate.username,
        expires: 'never'
      },
      passwordHash: {
        value: authenticate.passwordHash,
        expires: 'never'
      }
    };
  }

  emitter.emit('ready', {
    content: {
      authenticate: authenticate
    },
    cookie: cookie
  });
};

The following code sets the same cookies, but they expire at the end of the browser session:

cookie.username.value = authenticate.username;
cookie.passwordHash.value = authenticate.passwordHash;

Cookies sent by the client are available in params.cookie within the controller and simply cookie within the view context:

doctype html
html
  body
    #welcome
      if cookie.username
        | Welcome, #{cookie.username}.
      else
        a(href="/login") Login

Cookie variables you set within your controller aren't immediately available within the params.cookie scope. citizen's server has to receive the emitter's ready event from the controller before it can send the cookie to the client, so use a local instance of the variable if you need to access it during the same request.

Session Variables

If sessions are enabled, you can access session variables via params.session in your controller or simply session within views. These local scopes reference the current user's session without having to pass a session ID.

By default, a session has four properties: id, started, expires, and timer. The session ID is also sent to the client as a cookie called ctznSessionID.

Setting session variables is pretty much the same as setting cookie variables:

emitter.emit('ready', {
  session: {
    username: 'Danny',
    nickname: 'Doc'
  }
});

Sessions expire based on the sessionTimeout config property, which represents the length of a session in milliseconds. The default is 1200000 (20 minutes). The timer is reset with each request from the user. When the timer runs out, the session is deleted. Any client requests after that time will generate a new session ID and send a new session ID cookie to the client. Remember that the browser's session is separate from the server's session, so any cookies you've set with an expiration of session are untouched if the user's session expires on the server. You need to clear those cookies manually at the start of the next server session if you don't want them hanging around.

To forcibly clear and expire the current user's session:

session.expires = 'now';

This won't end the session immediately. citizen has to receive your controller's response before it can act on this directive.

Like cookies, session variables you've just assigned aren't available during the same request within the params.session scope, so use a local instance if you need to access this data right away.

Redirects

You can pass redirect instructions to the server that will be initiated after the request is complete. Redirects using this method within the controller are not immediate, so the controller will do everything it's been asked to do before the redirect is processed. The user agent won't receive a full response, however. No view content will be rendered or sent, but cookies and session variables will be set if specified.

The redirect object takes two properties: statusCode and url. If you don't provide a status code, citizen uses 302 (temporary redirect).

emitter.emit('ready', {
  redirect: {
    statusCode: 301,
    url: 'http://redirect.com'
  }
});

Include Patterns

citizen lets you use complete MVC patterns as includes. These includes are more than just chunks of code that you can reuse because each has its own controller, model, and view(s).

Let's say our article pattern's Jade template has the following contents. The head section contains dynamic meta data, and the header nav's content changes depending on whether the user is logged in or not:

doctype html
html
  head
    title #{metaData.title}
    meta(name="description" content="#{metaData.description}")
    meta(name="keywords" content="#{metaData.keywords}")
    link(rel="stylesheet" type="text/css" href="app.css")
  body
    header
      a#logo Home page
      if cookie.username
        p Welcome, #{cookie.username}
      nav
        ul
          li
            a(href="/") Home
          li
            a(href="/articles") Articles
          if cookie.username
            li
              a(href="/admin") Site Administration
    main
      h1 #{article.title} - Page #{url.page}
      p#summary #{article.summary}
      #text #{article.text}

It probably makes sense to use includes for the head section and header because you'll use that code everywhere, but rather than simple partials, let's create citizen includes. The head section can use its own model for populating the meta data, and since the header is different for authenticated users, let's pull that logic out of the template and set up different header views in our controller. I like to follow the convention of starting partials with an underscore, but that's up to you:

app/
  patterns/
    controllers/
      _head.js
      _header.js   // Doesn't generate data, so it doesn't need a model
      article.js
    models/
      _head.js
      article.js
    views/
      _head/
        _head.jade
      _header/
        _header.jade
        _header-authenticated.jade // A different header for logged in users
      article/
        article.jade

citizen include patterns have the same requirements as regular patterns, including a controller with a handler() function. When the article controller is fired, it has to tell citizen which includes it needs. We do that with the include directive, which we pass via the context in the emitter:

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  var article = app.models.article.getArticle(params.url.article, params.url.page),
      // We'll use the standard header by default, _header.jade
      headerAction = 'handler',
      headerView = '_header';

  // If the user is logged in, use _header-authenticated.jade
  if ( params.cookie.username ) {
    headerAction = 'authenticated';
    headerView = '_header-authenticated';
  }

  emitter.emit('ready', {
    content: {
      article: article
    },
    include: {
      _head: {
        // If only the controller is specified, the default action handler() is
        // called and the default view is rendered (_head.jade in this case).
        controller: '_head'
      },
      _header: {
        controller: '_header',
        action: headerAction,
        view: headerView
      }
    }
  });
}

This tells citizen to call the _head and _header controllers, pass them the existing request context, render their respective views, and add them to the view context.

The rendered includes are stored in the include scope. The include object contains rendered HTML views, so you need to skip escaping (!= in Jade, {{{...}}} in Handlebars):

doctype html
html
  != include._head
  body
    != include._header
    main
      h1 #{article.title} - Page #{url.page}
      p#summary #{article.summary}
      #text #{article.text}

citizen includes can generate content and add it to the view context of your primary controller (article.js in this example) because the primary view is the last to be rendered. However, includes are called and rendered asynchronously, so while your _header controller can generate content and add it to the view context of your article controller, don't assume that the other includes' controllers will have access to that data. (The option of waterfall execution is being worked on, so this is only true for the time being.)

citizen includes can also pass all emitter directives except for handoff.

A pattern meant to be used as an include can be accessed via HTTP just like any other controller. You could request the _head controller like so:

http://cleverna.me/_head

Perhaps you'd have it return meta data as JSON for the article pattern:

// http://cleverna.me/_head/article/My-Clever-Article-Title/format/json

{
  "metaData": {
    "title": "My Clever Article Title",
    "description": "My article's description.",
    "keywords": "clever, article, keywords"
  }
}

Here's an example of the _head controller written as both an include and a handler of direct requests:

// _head controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  var metaData = {},
      // If the article URL param exists, use that. Otherwise, assume _head is
      // being used as an include and use the requested route.
      getMetaDataFor = params.url.article || params.route.controller;

  if ( app.models[getMetaDataFor] && app.models[getMetaDataFor].getMetaData ) {
    metaData = app.models[getMetaDataFor].getMetaData();
  }

  emitter.emit('ready', {
    content: {
      metaData: metaData
    }
  });
}

Of course, if you don't write the controller in a manner to accept direct requests and return content, it'll return nothing (or throw an error).

Reminder: To make a controller private—inaccessible via HTTP, but accessible within your app—add a plus sign (+) to the beginning of the file name:

app/
  patterns/
    controllers/
      +_header.js  // Only accessible internally
      _head.js     // Accessible via www.cleverna.me/_head
      article.js   // Accessible via www.cleverna.me/article

Should I use a citizen include or a Jade include/Handlebars partial?

citizen includes provide rich functionality, but they do have limitations and can be overkill in certain situations.

• Do you only need to share a chunk of markup across different views? Use a standard Handlebars partial, Jade template, or HTML document. The syntax is easy and you don't have to create a full MVC pattern like you would with a citizen include.
• Do you need to loop over a chunk of markup to render a data set? The server processes citizen includes and returns them as fully-rendered HTML (or JSON), not compiled templates. You can't loop over them and inject data like you can with Handlebars partials or Jade includes.
• Do you need to retrieve additional content that isn't in the parent view's context? A citizen include can do anything that a standard MVC pattern can do except set the handoff directive. If you want to retrieve additional data and add it to the view context or set cookies and session variables, a citizen include is the way to go.
• Do you need the ability to render different includes based on business logic? citizen includes can have multiple views because they're full MVC patterns. Using a citizen include, you can place logic in the include's controller and request different views based on that logic. Using Handlebars partials or Jade includes would require registering multiple partials and putting the logic in the view template.
• Do you want the include to be accessible from the web? Since a citizen include has a controller, you can request it like any other controller and get back HTML, JSON, or JSONP, which is great for AJAX requests and single page apps.

Controller Handoff

citizen allows the requested controller to give another controller the responsibility of handling the request and rendering its own view via a directive called handoff. The requested controller passes its content and directives to a secondary controller that assumes responsibility for the request, adding its own content and directives and rendering its own view. This is also a method for passing your own custom content and directives to the receiving controller.

A common use case for handoff would be to create a layout controller that serves as a template for every page on your site, rendering all the includes necessary and leaving only the core content and markup to the initially requested controller. Let's modify the article controller and view so it hands off rendering responsibility to a separate layout controller:

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  var article = app.models.article.getArticle(params.url.article, params.url.page);

  emitter.emit('ready', {
    content: {
      article: article
    },
    handoff: {
      // Pass this request to app/patterns/controller/layout.js
      controller: 'layout',

      // Specifying the action is optional. The layout controller will use the
      // default action, handler(), unless you specify a different action here.
      action: 'handler',

      // Specifying the view is optional. The layout controller will use its
      // default view unless you tell it to use a different one.
      view: 'layout',

      // Rendering the requested controller's view is optional.
      // Using includeThisView tells citizen to render the article.jade view and
      // store it in the include scope. If you don't specify includeThisView,
      // the article controller's view won't be rendered.
      includeThisView: true
    },

    // A custom directive to drive some logic in the layout controller.
    // You could even pass a function here for layout to call.
    // Just don't use any reserved citizen directive names!
    myDirective: {
      doSomething: true
    }
  });
}

When you use the handoff directive and specify includeThisView like we did above, the originally requested view (article.jade in this case) is rendered as an include whose name matches its controller:

// article.jade, which is stored in the include scope as include.article

h1 #{article.title}
p#summary #{article.summary}
#text #{article.text}

The layout controller handles the includes, follows your custom directive, and renders its own view:

// layout controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  // We'll use the standard header by default
  var headerView = '_header';

  // If the user is logged in, use a different header view
  if ( params.cookie.username ) {
    headerView = '_header-authenticated';
  }

  // Access my custom directive using the context argument
  if ( context.myDirective && context.myDirective.doSomething ) {
    doSomething();
  }

  emitter.emit('ready', {
    // No need to specify previous directives, such as content, here because
    // citizen keeps track of the article pattern's request context throughout
    // the handoff process.
    include: {
      _head: {
        controller: '_head'
      },
      _header: {
        controller: '_header',
        view: headerView
      }
    }
  });
}

function doSomething() {
  // do something
}

And our layout.jade file:

// layout.jade

doctype html
html
  != include._head
  body
    != include._header
    main
      // You could use include.article here, but remember the route object?
      // It contains useful details about the route, like the original
      // controller's name. Now you can use this layout for any pattern.
      != include[route.controller]

Chaining controllers

You can use handoff to chain requests across as many controllers as you want, with each controller's directives added to the request context and each controller's view optionally added to the include scope. The initially requested controller's name and all following handoff controllers' names, along with their view names, are stored in the route object as an array called route.chain. You can loop over this object to render all the included views:

// layout.jade

doctype html
html
  != include._head
  body
    != include._header
    main
      // Include every view that was rendered via handoff
      each val in route.chain
        != include[val.controller]

Cache

In many cases, a requested route or controller will generate the same view every time based on the same input parameters, so it doesn't make sense to run the controller and render the view from scratch for each request. citizen provides flexible caching capabilities to speed up your server side rendering via the cache directive.

Here's an example cache directive (more details after the code sample):

emitter.emit('ready', {
  cache: {
    // Required. Valid values are 'controller', 'route', and 'global'
    scope: 'route',

    // Optional. List of valid URL parameters that protects against
    // accidental caching of malformed URLs.
    urlParams: ['article', 'page'],

    // Optional. List of directives to cache with the controller.
    directives: ['handoff', 'cookie'],

    // Optional. Life of cached item in milliseconds. Default is the life of
    // the application (no expiration).
    lifespan: 600000

    // Reset the cached item's expiration timer whenever the item is
    // accessed, keeping it in the cache until traffic subsides.
    resetOnAccess: true
});

cache.scope

The scope property determines how the controller and its resulting view are cached.

Values for cache.scope
route	A cache scope of "route" caches the entire rendered view for a given route. If a route's view doesn't vary across requests, use this scope to render it once when it's first requested and then serve it from the cache for every following request.
controller	Setting cache scope to "controller" caches an instance of the controller and the resulting view for every unique route that calls the controller. If the following URLs are requested and the article controller's cache scope is set to "controller", each URL will get its own unique cached instance of the article controller: http://cleverna.me/article/My-Clever-Article http://cleverna.me/article/My-Clever-Article/page/2 http://cleverna.me/article/Another-Article This may sound similar to the "route" scope, but a good use of the "controller" scope is when you're calling multiple controllers using citizen includes or the handoff directive and you want to cache each of those controllers based on the route, but not cache the final rendered view (like the previous layout controller example).
global	A cache scope of "global" caches a single instance of a given controller and uses it everywhere, regardless of context or the requested route. If you have a controller whose output and rendering won't change across requests regardless of the context or route, "global" is a good option.

cache.urlParams

The urlParams property helps protect against invalid cache items (or worse: an attack meant to flood your server's resources by overloading the cache). If we used the example above in our article controller, the following URLs would be cached because the "article" and "page" URL paramters are permitted:

http://cleverna.me/article
http://cleverna.me/article/My-Article-Title
http://cleverna.me/article/My-Article-Title/page/2

These URLs wouldn't be cached, which is a good thing because it wouldn't take long for an attacker's script to loop over a URL and flood the cache:

// "dosattack" isn't a valid URL parameter
http://cleverna.me/article/My-Article-Title/dosattack/1
http://cleverna.me/article/My-Article-Title/dosattack/2

// "page" is valid, but "dosattack" isn't, so it's not cached
http://cleverna.me/article/My-Article-Title/page/2/dosattack/3

The server will throw an error when an invalid URL is requested with a cache directive. Additionally, any URL that results in an error won't be cached, whether it's valid or not.

cache.directives

By default, any directives you specify in a cached controller aren't cached; they're implemented the first time the controller is called and then ignored after that. This is to prevent accidental storage of private data in the cache through session or cookie directives.

If you want directives to persist within the cache, include them in the directives property as an array:

emitter.emit('ready', {
  handoff: {
    controller: 'layout',
    includeThisView: true
  },
  cookie: {
    myCookie: {
      value: 'howdy'
    }
  },
  myCustomDirective: {
    doSomething: true
  },
  cache: {
    scope: 'controller',

    // Cache handoff and myCustomDirective so that if this controller is
    // called from the cache, it hands off to the layout controller and acts
    // upon myCustomDirective every time. The cookie directive will only be
    // acted upon the first time the controller is called, however.
    directives: ['handoff', 'myCustomDirective']
  }
});

Cache Limitations and Warnings

Controllers that use the include directive can't use global or controller cache scopes due to the way citizen renders includes, but it's on the roadmap for a future release. In the meantime, you can get around it by using the cache directive within the included controllers. The route scope works fine.

If you use the handoff directive to call a series of controllers and any one of those controllers sets the cache directive with the route scope, it takes priority over any cache settings in the following controllers. This is because the route scope caches the entire controller chain as a single cache object.

citizen's cache is a RAM cache, so be careful with your caching strategy. You could very quickly find yourself out of RAM. Use the lifespan option so URLs that aren't receiving a ton of traffic naturally fall out of the cache and free up resources for frequently accessed pages.

Cache defensively. Place logic in your controllers that combines the urlParams validation with some simple checks so invalid URLs don't result in junk pages clogging your cache:

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  var article = app.model.article.getArticle(params.url.article, params.url.page);

  // If the article exists, cache the result. citizen will compare the
  // existing URL parameters against the urlParams list you provide. If
  // there's a mismatch, citizen won't cache the result.
  if ( article.title ) {
    emitter.emit('ready', {
      content: {
        article: article
      },
      cache: {
        scope: 'controller',
        urlParams: ['article', 'page']
      }
    });
  // Throw a 404 if the article doesn't exist. Nothing will be cached.
  } else {
    throw {
      statusCode: 404
    };
  }

}

Application Events and the Context Argument

Certain events will occur throughout the life of your citizen application. You can act on these application events, execute functions, set directives, and pass the results to the next event or your controller via the context argument. For example, you might set a custom cookie at the beginning of every new session, or check for cookies at the beginning of every request and redirect the user to a login page if they're not authenticated.

To take advantage of these events, include a directory called "on" in your app with any or all of follwowing modules and exports:

app/
  on/
    application.js // exports start(), end(), and error()
    request.js     // exports start() and end()
    response.js    // exports start() and end()
    session.js     // exports start() and end()

request.start(), request.end(), and response.start() are called before your controller is fired, so the output from those events is passed from each one to the next, and ultimately to your controller via the context argument. Exactly what they output—content, citizen directives, custom directives—is up to you.

All files and exports are optional. citizen only calls them if they exist. For example, you could have only a request.js module that exports start().

Note: As of this version, session end() and application end() aren't functional. They'll be in a future version.

Here's an example of a request module that checks for a username cookie at the beginning of every request and redirects the user to the login page if it doesn't exist. We also avoid a redirect loop by making sure the requested controller isn't the login controller:

// app/on/request.js

exports.start = start;

function start(params, context, emitter) {
  var redirect = {};

  if ( !params.cookie.username && params.route.controller !== 'login' ) {
    redirect.url = '/login';
  }

  emitter.emit('ready', {
    redirect: redirect
  });
};

Cross-Origin Resource Sharing (CORS)

citizen supports cross-domain HTTP requests via access control headers. By default, all controllers respond to requests from the host only. This includes POST requests, which makes any controller that accepts form input safe from cross-site form submissions.

To enable cross-domain access, add an access object with the necessary headers to your controller's exports:

module.exports = {
  handler: handler,
  access: {
    // citizen expects header names in lowercase (per the spec, HTTP headers
    // are case-insensitive)
    'access-control-allow-origin': 'http://www.foreignhost.com',
    'access-control-expose-headers': 'X-My-Custom-Header, X-Another-Custom-Header',
    'access-control-max-age': 1728000,
    'access-control-allow-credentials': 'true',
    'access-control-allow-methods': 'OPTIONS, PUT',
    'access-control-allow-headers': 'Content-Type',
    'vary': 'Origin'
  }
};

For more details on CORS, check out the W3C spec and the Mozilla Developer Network.

Helpers

citizen has a few basic helper functions that it uses internally, but might be of use to you, so it returns them for public use.

cache(options)

You can store any object in citizen's cache.

// Cache a string for the life of the application.
app.cache({
  key: 'welcomeMessage',
  value: 'Welcome to my site.'
});

// Cache a string for the life of the application, and overwrite the
// existing key. The overwrite property is required any time you want to
// write to an existing key. This prevents accidental overwrites.
app.cache({
  key: 'welcomeMessage',
  value: 'Welcome to our site.',
  overwrite: true
});

// Cache a file buffer using the file path as the key. This is a wrapper for
// fs.readFile and fs.readFileSync paired with citizen's cache function.
// Optionally, tell citizen to perform a synchronous file read operation and
// use an encoding different from the default (UTF-8).
app.cache({
  file: '/path/to/articles.txt',
  synchronous: true,
  encoding: 'CP-1252'
});

// Cache a file with a custom key. Optionally, parse the JSON and store the
// parsed object in the cache instead of the raw buffer. Expire the cache
// after 60000 ms (60 seconds), but reset the timer whenever the key is
// retrieved.
app.cache({
  file: '/path/to/articles.json',
  key: 'articles',
  parseJSON: true,
  lifespan: 60000,
  resetOnAccess: true
});

exists(key)

This is a way to check for the existence of a given key in the cache without resetting the cache timer on that key.

app.exists('welcomeMessage')          // true
app.exists('/path/to/articles.txt')   // true
app.exists('articles')                // true

retrieve(key)

Retrieve a cached object using the key name. Returns false if the requested key doesn't exist. If resetOnAccess was true when the item was cached, using retrieve() will reset the cache clock and extend the life of the cached item.

app.retrieve('welcomeMessage')
app.retrieve('/path/to/articles.txt')
app.retrieve('articles')

clear(key)

Clear a cache object using the key name.

app.clear('welcomeMessage')
app.clear('/path/to/articles.txt')
app.clear('articles')

listen({ functions }, callback)

The article example we've been using has only simple methods that return static content immediately, but things are rarely that simple. The listen() function takes advantage of the asynchronous, event-driven nature of Node.js, letting you wrap a single function or multiple asynchronous functions within it and firing a callback when they're done. You can also chain and nest multiple listen() functions for very powerful asynchronous function calls.

listen() takes two arguments: an object containing one or more methods you want to call, and a callback to handle the output. listen() requires that your functions be written to accept an emitter argument, which is how your function notifies listen() that it's ready.

Let's say our article model has two methods that need to be called before returning the results to the controller. One is called getArticle() and the other is getViewers(). Assume both methods make an asynchronous call to a database and can't be relied upon to return their output immediately, so we have to listen for when they're ready and then react.

// article controller

module.exports = {
  handler: handler
};

function handler(params, context, emitter) {
  app.listen({
    // The property contains the action you want to listen for, which is
    // wrapped in an anonymous function in order to pass the emitter
    article: function (emitter) {
      app.models.article.getArticle(params.url.article, params.url.page, emitter);
    },
    viewers: function (emitter) {
      app.models.article.getViewers(params.url.article, emitter);
    }
  }, function (output) {
    // Emit `ready` now that we have the output from article and viewers and
    // pass the context back to the server
    emitter.emit('ready', {
      content: {
        // The property names you assign to the methods above become the names
        // of the output objects
        article: output.article,
        viewers: output.viewers
      }
    });
  });
}

And the model:

// article model

module.exports = {
  getArticle: getArticle,
  getViewers: getViewers
};

// Methods called via listen() must be written to accept an emitter
function getArticle(article, page, emitter) {
  app.db.article({ article: article, page: page }, function (data) {
    // When the database returns the data, emit `ready` and pass the
    // data back to listen()
    emitter.emit('ready', data);
  });
};

function getViewers(article, emitter) {
  app.db.viewers({ article: article }, function (data) {
    emitter.emit('ready', data);
  });
};

listen() currently fires all functions asynchronously and returns the results for every function in a single output object after all functions have completed. A waterfall-type execution is being worked on, but in the meantime, you can nest listen() functions to achieve the same effect:

app.listen({
  first: function (emitter) {
    doSomething(emitter);
  }
}, function (output) {
  app.listen({
    second: function (emitter) {
      doNextThing(output.first, emitter);
    }
  }, function (output) {
    app.listen({
      third: function (emitter) {
        doOneMoreThing(output.second, emitter);
      }
    }, function (output) {
      thisIsExhausting(output.third);
    });
  });
});

copy(object)

Creates a deep copy of an object.

var myCopy = app.copy(myObject);

extend(object, extension)

Extends an object with another object, effectively merging the two objects. extend() creates a copy of the original before extending it, creating a new object. Nested objects are also merged, but properties in the original object containing arrays are overwritten by the extension (arrays aren't merged).

var newObject = app.extend(originalObject, extensionObject);

isNumeric(object)

Returns true if the object is a number, false if not.

if ( app.isNumeric(params.url.id) ) {
  // pass it to the db
} else {
  return 'Naughty naughty...';
}

dashes(string)

Convert strings into SEO-friendly versions that you can use in your URLs.

var seoTitle = app.dashes("Won't You Read My Article?"); // 'Wont-You-Read-My-Article'

size(object)

Returns the number of objects contained within an object literal. Uses hasOwnProperty() to return a valid count.

var cars = {
      first: 'Volkswagen',
      second: 'Ford',
      third: 'BMW'
    },
    count = app.size(cars); // count equals 3

Debugging

Warning: debug and development modes are inherently insecure. Don't use them in a production environment.

If you set "mode": "debug" in your config file, citizen dumps the current pattern's output to the console by default. You can also dump it to the view by setting debug.output in your config file to view, or use the ctzn_dump URL parameter on a per-request basis:

// config file: always dumps debug output in the view
{
  "debug": {
    "output": "view"
  }
}

// URL
http://www.cleverna.me/article/id/237/page/2/ctzn_dump/view

By default, the pattern's complete output is dumped. You can specify the exact object to debug with the ctzn_debug URL parameter. You can access globals, pattern, and server params:

// Dumps pattern.content to the console
http://www.cleverna.me/article/id/237/page/2/ctzn_debug/pattern.content

// Dumps the server params object to the console
http://www.cleverna.me/article/id/237/page/2/ctzn_debug/params

// Dumps the user's session to the console
http://www.cleverna.me/article/id/237/page/2/ctzn_debug/params.session

// Dumps the user's session to the view
http://www.cleverna.me/article/id/237/page/2/ctzn_debug/params.session/ctzn_dump/view

In development mode, you must specify the ctzn_debug URL parameter to enable debug output. Debug output is disabled in production mode.