sprockets 3.0.0 → 4.0.2

data/README.md

@@ -5,7 +5,6 @@ It features declarative dependency management for JavaScript and CSS
 assets, as well as a powerful preprocessor pipeline that allows you to
 write assets in languages like CoffeeScript, Sass and SCSS.
 
-
 ## Installation
 
 Install Sprockets from RubyGems:
@@ -17,62 +16,37 @@ $ gem install sprockets
 Or include it in your project's `Gemfile` with Bundler:
 
 ``` ruby
-gem 'sprockets', '~> 3.0'
+gem 'sprockets', '~> 4.0'
 ```
 
+## Upgrading to Sprockets 4.x
 
-## Understanding the Sprockets Environment
-
-You'll need an instance of the `Sprockets::Environment` class to
-access and serve assets from your application. Under Rails 4.0 and
-later, `YourApp::Application.assets` is a preconfigured
-`Sprockets::Environment` instance. For Rack-based applications, create
-an instance in `config.ru`.
-
-The Sprockets `Environment` has methods for retrieving and serving
-assets, manipulating the load path, and registering processors. It is
-also a Rack application that can be mounted at a URL to serve assets
-over HTTP.
+These are the major features in Sprockets 4.x
 
-### The Load Path
+- Source Maps
+- Manifest.js
+- ES6 support
+- Deprecated processor interface in 3.x is removed in 4.x
 
-The *load path* is an ordered list of directories that Sprockets uses
-to search for assets.
+Read more about them by referencing [Upgrading document](UPGRADING.md)
 
-In the simplest case, a Sprockets environment's load path will consist
-of a single directory containing your application's asset source
-files. When mounted, the environment will serve assets from this
-directory as if they were static files in your public root.
+## Guides
 
-The power of the load path is that it lets you organize your source
-files into multiple directories -- even directories that live outside
-your application -- and combine those directories into a single
-virtual filesystem. That means you can easily bundle JavaScript, CSS
-and images into a Ruby library or [Bower](http://bower.io) package and import them into your application.
+For most people interested in using Sprockets, you will want to see the README below.
 
-#### Manipulating the Load Path
+If you are a framework developer that is using Sprockets, see [Building an Asset Processing Framework](guides/building_an_asset_processing_framework.md).
 
-To add a directory to your environment's load path, use the
-`append_path` and `prepend_path` methods. Directories at the beginning
-of the load path have precedence over subsequent directories.
+If you are a library developer who is extending the functionality of Sprockets, see [Extending Sprockets](guides/extending_sprockets.md).
 
-``` ruby
-environment = Sprockets::Environment.new
-environment.append_path 'app/assets/javascripts'
-environment.append_path 'lib/assets/javascripts'
-environment.append_path 'vendor/assets/bower_components'
-```
+If you want to work on Sprockets or better understand how it works read [How Sprockets Works](guides/how_sprockets_works.md)
 
-In general, you should append to the path by default and reserve
-prepending for cases where you need to override existing assets.
+## Behavior Overview
 
-### Accessing Assets
+You can interact with Sprockets primarily through directives and file extensions. This section covers how to use each of these things, and the defaults that ship with Sprockets.
 
-Once you've set up your environment's load path, you can mount the
-environment as a Rack server and request assets via HTTP. You can also
-access assets programmatically from within your application.
+Since you are likely using Sprockets through another framework (such as the [the Rails asset pipeline](http://guides.rubyonrails.org/asset_pipeline.html)), there will be configuration options you can toggle that will change behavior such as what directories or files get compiled. For that documentation you should see your framework's documentation.
 
-#### Logical Paths
+#### Accessing Assets
 
 Assets in Sprockets are always referenced by their *logical path*.
 
@@ -82,562 +56,610 @@ contains the directory `app/assets/javascripts`:
 
 <table>
   <tr>
-    <th>Asset source file</th>
     <th>Logical path</th>
+    <th>Source file on disk</th>
   </tr>
   <tr>
-    <td>app/assets/javascripts/application.js</td>
     <td>application.js</td>
+    <td>app/assets/javascripts/application.js</td>
   </tr>
   <tr>
-    <td>app/assets/javascripts/models/project.js</td>
     <td>models/project.js</td>
+    <td>app/assets/javascripts/models/project.js</td>
+  </tr>
+  <tr>
+    <td>hello.js</td>
+    <td>app/assets/javascripts/hello.coffee</td>
   </tr>
 </table>
 
-In this way, all directories in the load path are merged to create a
-virtual filesystem whose entries are logical paths.
+> Note: For assets that are compiled or transpiled, you want to specify the extension that you want, not the extension on disk. For example we specified `hello.js` even if the file on disk is a coffeescript file, since the asset it will generate is javascript.
 
-#### Serving Assets Over HTTP
+### Directives
 
-When you mount an environment, all of its assets are accessible as
-logical paths underneath the *mount point*. For example, if you mount
-your environment at `/assets` and request the URL
-`/assets/application.js`, Sprockets will search your load path for the
-file named `application.js` and serve it.
+Directives are special comments in your asset file and the main way of interacting with processors. What kind of interactions? You can use these directives to tell Sprockets to load other files, or specify dependencies on other assets.
 
-Under Rails 4.0 and later, your Sprockets environment is automatically
-mounted at `/assets`. If you are using Sprockets with a Rack
-application, you will need to mount the environment yourself. A good
-way to do this is with the `map` method in `config.ru`:
+For example, let's say you have custom JavaScript that you've written. You put this javascript in a file called `beta.js`. The javascript makes heavy use of jQuery, so you need to load that before your code executes. You could add a `require` directive to the top of `beta.js`:
 
-``` ruby
-require 'sprockets'
-map '/assets' do
-  environment = Sprockets::Environment.new
-  environment.append_path 'app/assets/javascripts'
-  environment.append_path 'app/assets/stylesheets'
-  run environment
-end
+```js
+//= require jquery
 
-map '/' do
-  run YourRackApp
-end
+$().ready({
+  // my custom code here
+})
 ```
 
-#### Accessing Assets Programmatically
+The directive processor understands comment blocks in three formats:
 
-You can use the `find_asset` method (aliased as `[]`) to retrieve an
-asset from a Sprockets environment. Pass it a logical path and you'll
-get a `Sprockets::Asset` instance back:
+``` css
+/* Multi-line comment blocks (CSS, SCSS, JavaScript)
+ *= require foo
+ */
+```
 
-``` ruby
-environment['application.js']
-# => #<Sprockets::Asset ...>
+``` js
+// Single-line comment blocks (SCSS, JavaScript)
+//= require foo
+```
+
+``` coffee
+# Single-line comment blocks (CoffeeScript)
+#= require foo
 ```
 
-Call `to_s` on the resulting asset to access its contents, `length` to
-get its length in bytes, `mtime` to query its last-modified time, and
-`filename` to get its full path on the filesystem.
+> Note: Directives are only processed if they come before any application code. Once you have a line that does not include a comment or whitespace then Sprockets will stop looking for directives. If you use a directive outside of the "header" of the document it will not do anything, and won't raise any errors.
 
+Here is a list of the available directives:
 
-## Using Processors
+- [`require`](#require) - Add the contents of a file to current
+- [`require_self`](#require_self) - Change order of where current contents are concatenated to current
+- [`require_directory`](#require_directory) - Add contents of each file in a folder to current
+- [`require_tree`](#require_tree) - Add contents of all files in all directories in a path to current
+- [`link`](#link) - Make target file compile and be publicly available without adding contents to current
+- [`link_directory`](#link_directory) - Make target directory compile and be publicly available without adding contents to current
+- [`link_tree`](#link_tree) - Make target tree compile and be publicly available without adding contents to current
+- [`depend_on`](#depend_on) - Recompile current file if target has changed
+- [`stub`](#stub) - Ignore target file
 
-Asset source files can be written in another format, like SCSS or
-CoffeeScript, and automatically compiled to CSS or JavaScript by
-Sprockets. Processors that convert a file from one format to another are called *transformers*.
+You can see what each of these does below.
 
-### Minifying Assets
+### Specifying Processors through File Extensions
 
-Several JavaScript and CSS minifiers are available through shorthand.
+Sprockets uses the filename extensions to determine what processors to run on your file and in what order. For example if you have a file:
 
-``` ruby
-environment.js_compressor  = :uglify
-environment.css_compressor = :scss
+```
+application.scss
 ```
 
-### Styling with Sass and SCSS
+Then Sprockets will by default run the sass processor (which implements scss). The output file will be converted to css.
 
-[Sass](http://sass-lang.com/) is a language that compiles to CSS and
-adds features like nested rules, variables, mixins and selector
-inheritance.
+You can specify multiple processors by specifying multiple file extensions. For example you can use Ruby's [ERB template language](#invoking-ruby-with-erb) to embed content in your doc before running the sass processor. To accomplish this you would need to name your file
 
-If the `sass` gem is available to your application, you can use Sass
-to write CSS assets in Sprockets.
+```
+application.scss.erb
+```
 
-Sprockets supports both Sass syntaxes. For the original
-whitespace-sensitive syntax, use the extension `.sass`. For the
-new SCSS syntax, use the extension `.scss`.
+Processors are run from right to left (tail to head), so in the above example the processor associated with `erb` will be run before the processor associated with `scss` extension.
 
-### Scripting with CoffeeScript
+For a description of the processors that Sprockets has by default see the "default processors" section below. Other libraries may register additional processors.
 
-[CoffeeScript](http://jashkenas.github.com/coffee-script/) is a
-language that compiles to the "good parts" of JavaScript, featuring a
-cleaner syntax with array comprehensions, classes, and function
-binding.
+When "asking" for a compiled file, you always ask for the extension you want. For example if you're using Rails, to get the contents of `application.scss.erb` you would use
 
-If the `coffee-script` gem is available to your application, you can
-use CoffeeScript to write JavaScript assets in Sprockets. Note that
-the CoffeeScript compiler is written in JavaScript, and you will need
-an [ExecJS](https://github.com/rails/execjs)-supported runtime
-on your system to invoke it.
+```
+asset_path("application.css")
+```
 
-To write JavaScript assets with CoffeeScript, use the extension
-`.coffee`.
+Sprockets understands that `application.scss.erb` will compile down to a `application.css`. Ask for what you need, not what you have.
 
-### JavaScript Templating with EJS and Eco
+If this isn't working like you expect, make sure you didn't typo an extension, and make sure the file is on a "load path" (see framework docs for adding new load paths).
 
-Sprockets supports *JavaScript templates* for client-side rendering of
-strings or markup. JavaScript templates have the special format
-extension `.jst` and are compiled to JavaScript functions.
+## File Order Processing
 
-When loaded, a JavaScript template function can be accessed by its
-logical path as a property on the global `JST` object. Invoke a
-template function to render the template as a string. The resulting
-string can then be inserted into the DOM.
+By default files are processed in alphabetical order. This behavior can impact your asset compilation when one asset needs to be loaded before another.
 
+For example if you have an `application.js` and it loads another directory
+
+```js
+//= require_directory my_javascript
 ```
-<!-- templates/hello.jst.ejs -->
-<div>Hello, <span><%= name %></span>!</div>
 
-// application.js
-//= require templates/hello
-$("#hello").html(JST["templates/hello"]({ name: "Sam" }));
+The files in that directory will be loaded in alphabetical order. If the directory looks like this:
+
+```sh
+$ ls -1 my_javascript/
+
+alpha.js
+beta.js
+jquery.js
 ```
 
-Sprockets supports two JavaScript template languages:
-[EJS](https://github.com/sstephenson/ruby-ejs), for embedded
-JavaScript, and [Eco](https://github.com/sstephenson/ruby-eco), for
-embedded CoffeeScript. Both languages use the familiar `<% … %>`
-syntax for embedding logic in templates.
+Then `alpha.js` will be loaded before either of the other two. This can be a problem if `alpha.js` uses jquery. For this reason it is not recommend to use `require_directory` with files that are ordering dependent. You can either require individual files manually:
 
-If the `ejs` gem is available to your application, you can use EJS
-templates in Sprockets. EJS templates have the extension `.jst.ejs`.
+```js
+//= require jquery
+//= require alpha
+//= require beta
+```
 
-If the `eco` gem is available to your application, you can use [Eco
-templates](https://github.com/sstephenson/eco) in Sprockets. Eco
-templates have the extension `.jst.eco`. Note that the `eco` gem
-depends on the CoffeeScript compiler, so the same caveats apply as
-outlined above for the CoffeeScript engine.
+Or you can use index files to proxy your folders.
 
-### Invoking Ruby with ERB
+### Index files are proxies for folders
 
-Sprockets provides an ERB engine for preprocessing assets using
-embedded Ruby code. Append `.erb` to a CSS or JavaScript asset's
-filename to enable the ERB engine.
+In Sprockets index files such as `index.js` or `index.css` files inside of a folder will generate a file with the folder's name. So if you have a `foo/index.js` file it will compile down to `foo.js`. This is similar to NPM's behavior of using [folders as modules](https://nodejs.org/api/modules.html#modules_folders_as_modules). It is also somewhat similar to the way that a file in `public/my_folder/index.html` can be reached by a request to `/my_folder`. This means that you cannot directly use an index file. For example this would not work:
 
-Ruby code embedded in an asset is evaluated in the context of a
-`Sprockets::Context` instance for the given asset. Common uses for ERB
-include:
+```erb
+<%= asset_path("foo/index.js") %>
+```
 
-- embedding another asset as a Base64-encoded `data:` URI with the
-  `asset_data_uri` helper
-- inserting the URL to another asset, such as with the `asset_path`
-  helper provided by the Sprockets Rails plugin
-- embedding other application resources, such as a localized string
-  database, in a JavaScript asset via JSON
-- embedding version constants loaded from another file
+Instead you would need to use:
 
-See the [Helper Methods](lib/sprockets/context.rb) section for more information about
-interacting with `Sprockets::Context` instances via ERB.
+```erb
+<%= asset_path("foo.js") %>
+```
 
+Why would you want to use this behavior?  It is common behavior where you might want to include an entire directory of files in a top level JavaScript. You can do this in Sprockets using `require_tree .`
 
-## Managing and Bundling Dependencies
+```js
+//= require_tree .
+```
 
-You can create *asset bundles* -- ordered concatenations of asset
-source files -- by specifying dependencies in a special comment syntax
-at the top of each source file.
+This has the problem that files are required alphabetically. If your directory has `jquery-ui.js` and `jquery.min.js` then Sprockets will require `jquery-ui.js` before `jquery` is required which won't work (because jquery-ui depends on jquery). Previously the only way to get the correct ordering would be to rename your files, something like `0-jquery-ui.js`. Instead of doing that you can use an index file.
 
-Sprockets reads these comments, called *directives*, and processes
-them to recursively build a dependency graph. When you request an
-asset with dependencies, the dependencies will be included in order at
-the top of the file.
+For example, if you have an `application.js` and want all the files in the `foo/` folder you could do this:
 
-### The Directive Processor
+```js
+//= require foo.js
+```
 
-Sprockets runs the *directive processor* on each CSS and JavaScript
-source file. The directive processor scans for comment lines beginning
-with `=` in comment blocks at the top of the file.
+Then create a file `foo/index.js` that requires all the files in that folder in any order you want using relative references:
 
-``` js
-//= require jquery
-//= require jquery-ui
-//= require backbone
-//= require_tree .
+```js
+//= require ./foo.min.js
+//= require ./foo-ui.js
 ```
 
-The first word immediately following `=` specifies the directive
-name. Any words following the directive name are treated as
-arguments. Arguments may be placed in single or double quotes if they
-contain spaces, similar to commands in the Unix shell.
+Now in your `application.js` will correctly load the `foo.min.js` before `foo-ui.js`. If you used `require_tree` it would not work correctly.
 
-**Note**: Non-directive comment lines will be preserved in the final
-  asset, but directive comments are stripped after
-  processing. Sprockets will not look for directives in comment blocks
-  that occur after the first line of code.
+## Cache
 
-#### Supported Comment Types
+Compiling assets is slow. It requires a lot of disk use to pull assets off of hard drives, a lot of RAM to manipulate those files in memory, and a lot of CPU for compilation operations. Because of this Sprockets has a cache to speed up asset compilation times. That's the good news. The bad news, is that sprockets has a cache and if you've found a bug it's likely going to involve the cache.
 
-The directive processor understands comment blocks in three formats:
+By default Sprockets uses the file system to cache assets. It makes sense that Sprockets does not want to generate assets that already exist on disk in `public/assets`, what might not be as intuitive is that Sprockets needs to cache "partial" assets.
 
-``` css
-/* Multi-line comment blocks (CSS, SCSS, JavaScript)
- *= require foo
- */
+For example if you have an `application.js` and it is made up of `a.js`, `b.js`, all the way to `z.js`
+
+```js
+//= require a.js
+//= require b.js
+# ...
+//= require z.js
 ```
 
-``` js
-// Single-line comment blocks (SCSS, JavaScript)
-//= require foo
+The first time this file is compiled the `application.js` output will be written to disk, but also intermediary compiled files for `a.js` etc. will be written to the cache directory (usually `tmp/cache/assets`).
+
+So, if `b.js` changes it will get recompiled. However instead of having to recompile the other files from `a.js` to `z.js` since they did not change, we can use the prior intermediary files stored in the cached values . If these files were expensive to generate, then this "partial" asset cache strategy can save a lot of time.
+
+Directives such as `require`, `link`, and `depend_on` tell Sprockets what assets need to be re-compiled when a file changes. Files are considered "fresh" based on their mtime on disk and a combination of cache keys.
+
+On Rails you can force a "clean" install by clearing the `public/assets` and `tmp/cache/assets` directories.
+
+
+## Default Directives
+
+Directives take a path or a path to a file. Paths for directive can be relative to the current file, for example:
+
+```js
+//= require ../foo.js
 ```
 
-``` coffee
-# Single-line comment blocks (CoffeeScript)
-#= require foo
+This would load the file up one directory and named `foo.js`. However this isn't required if `foo.js` is on one of Sprocket's load paths. You can simply use
+
+```js
+//= require foo.js
 ```
 
-### Sprockets Directives
+Without any prepended dots and sprockets will search for the asset. If the asset is on a sub-path of the load path, you can specify it without using a relative path as well:
 
-You can use the following directives to declare dependencies in asset
-source files.
+```js
+//= require sub/path/foo.js
+```
+
+You can also use an absolute path, but this is discouraged unless you know the directory structure of every machine you plan on running code on.
 
-For directives that take a *path* argument, you may specify either a
-logical path or a relative path. Relative paths begin with `./` and
-reference files relative to the location of the current file.
+Below is a section for each of the built in directive types supported by Sprockets.
 
-#### The `require` Directive
+### require
 
 `require` *path* inserts the contents of the asset source file
 specified by *path*. If the file is required multiple times, it will
 appear in the bundle only once.
 
-### The `require_directory` Directive ###
+**Example:**
 
-`require_directory` *path* requires all source files of the same
-format in the directory specified by *path*. Files are required in
-alphabetical order.
+If you've got an `a.js`:
 
-#### The `require_tree` Directive
+```js
+var a = "A";
+```
 
-`require_tree` *path* works like `require_directory`, but operates
-recursively to require all files in all subdirectories of the
-directory specified by *path*.
+and a `b.js`;
+
+```js
+var b = "B";
+```
 
-#### The `require_self` Directive
+Then you could require both of these in an `application.js`
+
+```js
+//= require a.js
+//= require b.js
+```
+
+Which would generate one concatenated file:
+
+```js
+var a = "A";
+var b = "B";
+```
+
+### require_self
 
 `require_self` tells Sprockets to insert the body of the current
 source file before any subsequent `require` directives.
 
-#### The `link` Directive
+**Example:**
 
-`link` *path* declares a dependency on the target *path* and adds it to a list
-of subdependencies to automatically be compiled when the asset is written out to
-disk.
+If you've got an `a.js`:
 
-For an example, in a CSS file you might reference an external image that always
-needs to be compiled along with the css file.
+```js
+var a = "A";
+```
 
-``` css
-/*= link "logo.png" */
-.logo {
-  background-image: url(logo.png)
-}
+And an `application.js`
+
+```js
+//= require_self
+//= require 'a.js'
+
+var app_name = "Sprockets";
 ```
 
-However, if you use a `asset-path` or `asset-url` SCSS helper, these links will
-automatically be defined for you.
+Then this will take the contents of `application.js` (that come after the last require) and put them at the beginning of the file:
 
-``` css
-.logo {
-  background-image: asset-url("logo.png")
-}
+```js
+var app_name = "Sprockets";
+var a = "A";
 ```
 
-#### The `depend_on` Directive
+### require_directory
 
-`depend_on` *path* declares a dependency on the given *path* without
-including it in the bundle. This is useful when you need to expire an
-asset's cache in response to a change in another file.
+`require_directory` *path* requires all source files of the same
+format in the directory specified by *path*. Files are required in
+alphabetical order.
 
-#### The `depend_on_asset` Directive
+**Example:**
 
-`depend_on_asset` *path* works like `depend_on`, but operates
-recursively reading the file and following the directives found. This is automatically implied if you use `link`, so consider if it just makes sense using `link` instead of `depend_on_asset`.
+If we've got a directory called `alphabet` with an `a.js` and `b.js` files like before, then our `application.js`
 
-#### The `stub` Directive
+```js
+//= require_directory alphabet
+```
 
-`stub` *path* allows dependency to be excluded from the asset bundle.
-The *path* must be a valid asset and may or may not already be part
-of the bundle. `stub` should only be used at the top level bundle, not
-within any subdependencies.
+Would produce:
 
+```js
+var a = "A";
+var b = "B";
+```
 
-## Processor Interface
+You can also see [Index files are proxies for folders](#index-files-are-proxies-for-folders) for another method of organizing folders that will give you more control.
 
-Sprockets 2.x was originally design around [Tilt](https://github.com/rtomayko/tilt)'s engine interface. However, starting with 3.x, a new interface has been introduced deprecating Tilt.
+### require_tree
 
-Similar to Rack, a processor is a any "callable" (an object that responds to `call`). This maybe a simple Proc or a full class that defines a `def self.call(input)` method. The `call` method accepts an `input` Hash and returns a Hash of metadata.
+`require_tree` *path* works like `require_directory`, but operates
+recursively to require all files in all subdirectories of the
+directory specified by *path*.
 
-Also see [`Sprockets::ProcessorUtils`](https://github.com/rails/sprockets/blob/master/lib/sprockets/processor_utils.rb) for public helper methods.
+### link
 
-### input Hash
+`link` *path* declares a dependency on the target *path* and adds it to a list
+of subdependencies to be compiled when the asset is written out to
+disk.
 
-The `input` Hash defines the following public fields.
+Example:
 
-* `:data` - String asset contents
-* `:environment` - Current `Sprockets::Environment` instance.
-* `:cache` - A `Sprockets::Cache` instance. See [`Sprockets::Cache#fetch`](https://github.com/rails/sprockets/blob/master/lib/sprockets/cache.rb).
-* `:uri` - String Asset URI.
-* `:filename` - String full path to original file.
-* `:load_path` - String current load path for filename.
-* `:name` - String logical path for filename.
-* `:content_type` - String content type of the output asset.
-* `:metadata` - Hash of processor metadata.
+If you've got a `manifest.js` file and you want to specify that a `admin.js` source file should be
+generated and made available to the public you can link it by including this in the `manifest.js` file:
 
-``` ruby
-def self.call(input)
-  input[:cache].fetch("my:cache:key:v1") do
-    # Remove all semicolons from source
-    input[:data].gsub(";", "")
-  end
-end
+```
+//= link admin.js
 ```
 
-### return Hash
+The argument to `link` is a _logical path_, that is it will be resolved according to the
+configured asset load paths. See [Accesing Assets](#accessing-assets) above. A path relative to
+the current file won't work, it must be a logical path.
 
-The processor should return metadata `Hash`. With the exception of the `:data` key, the processor can store arbitrary JSON valid values in this Hash. The data will be stored and exposed on `Asset#metadata`.
+**Caution**: the "link" directive should always have an explicit extension on the end.
 
-The returned `:data` replaces the assets `input[:data]` to the next processor in the chain. Returning a `String` is shorthand for returning `{ data: str }`. And returning `nil` is shorthand for a no-op where the input data is not transformed, `{ data: input[:data] }`.
+### link_directory
 
-### metadata
+`link_directory` *path* links all the files inside the directory specified by the *path*. By "link", we mean they are specified as compilation targets to be written out to disk, and made available to be served to user-agents.
 
-The metadata Hash provides an open format for processors to extend the pipeline processor. Internally, built-in processors use it for passing data to each other.
+Files in subdirectories will not be linked (Compare to [link_tree](#link_tree)).
 
-* `:required` - A `Set` of String Asset URIs that the Bundle processor should concatenate together.
-* `:stubbed` - A `Set` of String Asset URIs that will be omitted from the `:required` set.
-* `:links` - A `Set` of String Asset URIs that should be compiled along with this asset.
-* `:dependencies` - A `Set` of String Cache URIs that should be monitored for caching.
+The *path* argument to `link_directory` is _not_ a logical path (it does not use the asset load paths), but is a path relative to the file the `link_directory` directive is found in, and can use `..` to  . For instance, you might want:
 
-``` ruby
-def self.call(input)
-  # Any metadata may start off as nil, so initialize it the value
-  required = Set.new(input[:metadata][:required])
+```js
+//= link_directory ../stylesheets
+```
 
-  # Manually add "foo.js" asset uri to our bundle
-  required << input[:environment].resolve("foo.js")
+`link_directory` can take an optional second argument with an extension or content-type, with the
+two arguments separated by a space:
 
-  { required: required }
-end
+```js
+//= link_directory ../stylesheets text/css
+//= link_directory ../more_stylesheets .css
 ```
 
+This will limit the matching files to link to only files recognized as that type. An extension is
+just a shortcut for the type referenced, it does not need to match the source file exactly, but
+instead identifies the content-type the source file must be recognized as.
 
-## Development
+### link_tree
 
-### Contributing
+`link_tree` *path* works like [link_directory](#link_directory), but operates
+recursively to link all files in all subdirectories of the
+directory specified by *path*.
 
-The Sprockets source code is [hosted on
-GitHub](https://github.com/rails/sprockets). You can check out a
-copy of the latest code using Git:
+Example:
 
-    $ git clone https://github.com/rails/sprockets
+```js
+//= link_tree ./path/to/folder
+```
 
-If you've found a bug or have a question, please open an issue on the
-[Sprockets issue
-tracker](https://github.com/rails/sprockets/issues). Or, clone
-the Sprockets repository, write a failing test case, fix the bug and
-submit a pull request.
+Like `link_directory`, the argument is path relative to the current file, it is *not* a 'logical path' tresolved against load paths.
 
-### Version History
 
-**3.0.0**
+As with `link_directory`, you can also specify a second argument -- separated by a space --  so any extra files not matching the content-type specified will be ignored:
 
-* New processor API. Tilt interface is deprecated.
-* Improved file store caching backend.
-* MIME Types now accept charset custom charset detecters. Improves support for UTF-16/32 files.
-* Environment#version no longer affects asset digests. Only used for busting the asset cache.
-* Removed builtin support for LESS.
-* Removed `//= include` directive support.
-* Deprecated `BundledAsset#to_a`. Use `BundledAsset#included` to access debugging subcomponents.
-* Support circular dependencies. For parity with ES6 modules.
-* Manifest compilation will no longer generate .gz files by default. [Mixing
-  Content-Encoding and ETags is just a bad
-  idea](https://issues.apache.org/bugzilla/show_bug.cgi?id=39727)
-* Added linked or referenced assets. When an asset is compiled, any of its links will be compiled as well.
-* Introduce some limitations around enumerating all logical paths. 4.x will deprecate it and favor linked manifests for compliation.
-* Add Asset integrity attribute for Subresource Integrity
-* Default digest changed to SHA256. Configuring `digest_class` is deprecated.
-* Rename `Asset#digest` to `Asset#hexdigest`. `Asset#digest` is deprecated and will
-  return a raw byte String in 4.x.
-* Added transitional compatibility flag to `Environment#resolve(path, compat: true)`. 2.x mode operates with `compat: true` and 4.x with `compat: false`
-* `manifest-abc123.json` renamed to `.sprockets-abc123.json`
+```js
+//= link_tree ./path/to/folder text/javascript
+//= link_tree ./path/to/other_folder .js
+```
 
-**2.12.3** (October 28, 2014)
 
-* Security: Fix directory traversal bug in development mode server.
+### depend_on
 
-**2.12.2** (September 5, 2014)
+`depend_on` *path* declares a dependency on the given *path* without
+including it in the bundle. This is useful when you need to expire an
+asset's cache in response to a change in another file.
 
-* Ensure internal asset lookups calls are still restricted to load paths within
-  asset compiles. Though, you should not depend on internal asset resolves to be
-  completely restricted for security reasons. Assets themselves should be
-  considered full scripting environments with filesystem access.
+**Example:**
 
-**2.12.1** (April 17, 2014)
+If you have a file such as `bar.data` and you're using data from that file in another file, then
+you need to tell sprockets that it needs to re-compile the file if `bar.data` changes:
 
-* Fix making manifest target directory when its different than the output directory.
+```js
+//= depend_on "bar.data"
 
-**2.12.0** (March 13, 2014)
+var bar = '<%= File.read("bar.data") %>'
+```
 
-* Avoid context reference in SassImporter hack so its Marshallable. Fixes
- issues with Sass 3.3.x.
+### depend_on_asset
 
-**2.11.0** (February 19, 2014)
+`depend_on_asset` *path* works like `depend_on`, but operates
+recursively reading the file and following the directives found. This is automatically implied if you use `link`, so consider if it just makes sense using `link` instead of `depend_on_asset`.
 
-* Cache store must now be an LRU implementation.
-* Default digest changed to SHA1. To continue using MD5.
-  `env.digest_class = Digest::MD5`.
+### stub
 
-**2.10.0** (May 24, 2013)
+`stub` *path* excludes that asset and its dependencies from the asset bundle.
+The *path* must be a valid asset and may or may not already be part
+of the bundle. `stub` should only be used at the top level bundle, not
+within any subdependencies.
 
-* Support for `bower.json`
+### Invoking Ruby with ERB
 
-**2.9.3** (April 20, 2013)
+Sprockets provides an ERB engine for preprocessing assets using
+embedded Ruby code. Append `.erb` to a CSS or JavaScript asset's
+filename to enable the ERB engine.
 
-* Fixed sass caching bug
+For example if you have an `app/application/javascripts/app_name.js.erb`
+you could have this in the template
 
-**2.9.2** (April 8, 2013)
+```js
+var app_name = "<%= ENV['APP_NAME'] %>";
+```
 
-* Improve file freshness check performance
-* Directive processor encoding fixes
+Generated files are cached. If you're using an `ENV` var then
+when you change then ENV var the asset will be forced to
+recompile. This behavior is only true for environment variables,
+if you are pulling a value from somewhere else, such as a database,
+must manually invalidate the cache to see the change.
 
-**2.9.1** (April 6, 2013)
+If you're using Rails, there are helpers you can use such as `asset_url`
+that will cause a recompile if the value changes.
 
-* Support for Uglifier 2.x
+For example if you have this in your `application.css`
 
-**2.9.0** (February 25, 2013)
+``` css
+.logo {
+  background: url(<%= asset_url("logo.png") %>)
+}
+```
 
-* Write out gzipped variants of bundled assets.
+When you modify the `logo.png` on disk, it will force `application.css` to be
+recompiled so that the fingerprint will be correct in the generated asset.
 
-**2.8.2** (December 10, 2012)
+You can manually make sprockets depend on any other file that is generated
+by sprockets by using the `depend_on` directive. Rails implements the above
+feature by auto calling `depend_on` on the original asset when the `asset_url`
+is used inside of an asset.
 
-* Fixed top level Sass constant references
-* Fixed manifest logger when environment is disabled
+### Styling with Sass and SCSS
 
-**2.8.1** (October 31, 2012)
+[Sass](http://sass-lang.com/) is a language that compiles to CSS and
+adds features like nested rules, variables, mixins and selector
+inheritance.
 
-* Fixed Sass importer bug
+If the `sass` gem is available to your application, you can use Sass
+to write CSS assets in Sprockets.
 
-**2.8.0** (October 16, 2012)
+Sprockets supports both Sass syntaxes. For the original
+whitespace-sensitive syntax, use the extension `.sass`. For the
+new SCSS syntax, use the extension `.scss`.
 
-* Allow manifest location to be separated from output directory
-* Pass logical path and absolute path to each_logical_path iterator
+In Rails if you have `app/application/stylesheets/foo.scss` it can
+be referenced with `<%= asset_path("foo.css") %>`. When referencing
+an asset in Rails, always specify the extension you want. Sprockets will
+convert `foo.scss` to `foo.css`.
 
-**2.7.0** (October 10, 2012)
+### Scripting with CoffeeScript
 
-* Added --css-compressor and --js-compressor command line flags
-* Added css/js compressor shorthand
-* Change default manifest.json filename to be a randomized manifest-16HEXBYTES.json
-* Allow nil environment to be passed to manifest
-* Allow manifest instance to be set on rake task
+[CoffeeScript](http://jashkenas.github.io/coffeescript/) is a
+language that compiles to the "good parts" of JavaScript, featuring a
+cleaner syntax with array comprehensions, classes, and function
+binding.
 
-**2.6.0** (September 19, 2012)
+If the `coffee-script` gem is available to your application, you can
+use CoffeeScript to write JavaScript assets in Sprockets. Note that
+the CoffeeScript compiler is written in JavaScript, and you will need
+an [ExecJS](https://github.com/rails/execjs)-supported runtime
+on your system to invoke it.
 
-* Added bower component.json require support
+To write JavaScript assets with CoffeeScript, use the extension
+`.coffee`.
 
-**2.5.0** (September 4, 2012)
+In Rails if you have `app/application/javascripts/foo.coffee` it can
+be referenced with `<%= asset_path("foo.js") %>`. When referencing
+an asset in Rails, always specify the extension you want. Sprockets will
+convert `foo.coffee` to `foo.js`.
 
-* Fixed Ruby 2.0 RegExp warning
-* Provide stubbed implementation of context *_path helpers
-* Add SassCompressor
 
-**2.4.5** (July 10, 2012)
+## ES6 Support
 
-* Tweaked some logger levels
+Sprockets 4 ships with a Babel processor. This allows you to transpile ECMAScript6 to JavaScript just like you would transpile CoffeeScript to JavaScript. To use this, modify your Gemfile:
 
-**2.4.4** (July 2, 2012)
+```ruby
+gem 'babel-transpiler'
+```
 
-* Canonicalize logical path extensions
-* Check absolute paths passed to depend_on
+Any asset with the extension `es6` will be treated as an ES6 file:
 
-**2.4.3** (May 16, 2012)
+```es6
+// app/assets/javascript/application.es6
 
-* Exposed :sprockets in sass options
-* Include dependency paths in asset mtime
+var square = (n) => n * n
 
-**2.4.2** (May 7, 2012)
+console.log(square);
+```
 
-* Fixed MultiJson feature detect
+Start a Rails server in development mode and visit `localhost:3000/assets/application.js`, and this asset will be transpiled to JavaScript:
 
-**2.4.1** (April 26, 2012)
+```js
+var square = function square(n) {
+  return n * n;
+};
 
-* Fixed MultiJson API change
-* Fixed gzip mtime
+console.log(square);
+```
 
-**2.4.0** (March 27, 2012)
 
-* Added global path registry
-* Added global processor registry
+### JavaScript Templating with EJS and Eco
 
-**2.3.2** (March 26, 2012)
+Sprockets supports *JavaScript templates* for client-side rendering of
+strings or markup. JavaScript templates have the special format
+extension `.jst` and are compiled to JavaScript functions.
 
-* Fix Context#logical_path with dots
+When loaded, a JavaScript template function can be accessed by its
+logical path as a property on the global `JST` object. Invoke a
+template function to render the template as a string. The resulting
+string can then be inserted into the DOM.
 
-**2.3.1** (February 11, 2012)
+```
+<!-- templates/hello.jst.ejs -->
+<div>Hello, <span><%= name %></span>!</div>
 
-* Added bytesize to manifest
-* Added Asset#bytesize alias
-* Security: Check path for forbidden access after unescaping
+// application.js
+//= require templates/hello
+$("#hello").html(JST["templates/hello"]({ name: "Sam" }));
+```
 
-**2.3.0** (January 16, 2012)
+Sprockets supports two JavaScript template languages:
+[EJS](https://github.com/sstephenson/ruby-ejs), for embedded
+JavaScript, and [Eco](https://github.com/sstephenson/ruby-eco), for
+embedded CoffeeScript. Both languages use the familiar `<% … %>`
+syntax for embedding logic in templates.
+
+If the `ejs` gem is available to your application, you can use EJS
+templates in Sprockets. EJS templates have the extension `.jst.ejs`.
+
+If the `eco` gem is available to your application, you can use [Eco
+templates](https://github.com/sstephenson/eco) in Sprockets. Eco
+templates have the extension `.jst.eco`. Note that the `eco` gem
+depends on the CoffeeScript compiler, so the same caveats apply as
+outlined above for the CoffeeScript engine.
+
+### Minifying Assets
+
+Several JavaScript and CSS minifiers are available through shorthand.
+
+In Rails you will specify them with:
+
+```ruby
+config.assets.js_compressor  = :uglify
+config.assets.css_compressor = :scss
+```
 
-* Added special Sass importer that automatically tracks any `@import`ed files.
+If you're not using Rails, configure this directly on the "environment".
 
-**2.2.0** (January 10, 2012)
+``` ruby
+environment.js_compressor  = :uglify
+environment.css_compressor = :scss
+```
 
-* Added `sprockets` command line utility.
-* Added rake/sprocketstask.
-* Added json manifest log of compiled assets.
-* Added `stub` directive that allows you to exclude files from the bundle.
-* Added per environment external encoding (Environment#default_external_encoding). Defaults to UTF-8. Fixes issues where LANG is not set correctly and Rubys default external is set to ASCII.
+If you are using Sprockets directly with a Rack app, don't forget to add
+the `uglifier` and `sass` gems to your Gemfile when using above options.
 
-**2.1.2** (November 20, 2011)
+### Gzip
 
-* Disabled If-Modified-Since server checks. Fixes some browser caching issues when serving the asset body only. If-None-Match caching is sufficient.
+By default when Sprockets generates a compiled asset file it will also produce a gzipped copy of that file. Sprockets only gzips non-binary files such as CSS, javascript, and SVG files.
 
-**2.1.1** (November 18, 2011)
+For example if Sprockets is generating
 
-* Fix windows absolute path check bug.
+```
+application-12345.css
+```
 
-**2.1.0** (November 11, 2011)
+Then it will also generate a compressed copy in
 
-* Directive comment lines are now turned into empty lines instead of removed. This way line numbers in
-  CoffeeScript syntax errors are correct.
-* Performance and caching bug fixes.
+```
+application-12345.css.gz
+```
 
-**2.0.3** (October 17, 2011)
+This behavior can be disabled, refer to your framework specific documentation.
 
-* Detect format extensions from right to left.
-* Make JST namespace configurable.
+### Serving Assets
 
-**2.0.2** (October 4, 2011)
+In production you should generate your assets to a directory on disk and serve them either via Nginx or a feature like Rail's `config.public_file_server.enabled = true`.
 
-* Fixed loading stale cache from bundler gems.
+On Rails you can generate assets by running:
 
-**2.0.1** (September 30, 2011)
+```term
+$ RAILS_ENV=production rake assets:precompile
+```
 
-* Fixed bug with fingerprinting file names with multiple dots.
-* Decode URIs as default internal.
-* Fix symlinked asset directories.
+In development Rails will serve assets from `Sprockets::Server`.
 
-**2.0.0** (August 29, 2011)
+## Contributing to Sprockets
 
-* Initial public release.
+Sprockets is the work of hundreds of contributors. You're encouraged to submit pull requests, propose
+features and discuss issues.
 
-## License
+See [CONTRIBUTING](CONTRIBUTING.md).
 
-Copyright &copy; 2014 Sam Stephenson <<sstephenson@gmail.com>>
+### Version History
 
-Copyright &copy; 2014 Joshua Peek <<josh@joshpeek.com>>
+Please see the [CHANGELOG](https://github.com/rails/sprockets/tree/master/CHANGELOG.md)
 
-Sprockets is distributed under an MIT-style license. See LICENSE for
-details.
+## License
+Sprockets is released under the [MIT License](MIT-LICENSE).