Action Controller is the C in MVC. After routing has determined which controller to use for a request, your controller is responsible for making sense of the request and producing the appropriate output. Luckily, Action Controller does most of the groundwork for you and uses smart conventions to make this as straight-forward as possible.

For most conventional RESTful applications, the controller will receive the request (this is invisible to you as the developer), fetch or save data from a model and use a view to create HTML output. If your controller needs to do things a little differently, that's not a problem, this is just the most common way for a controller to work.

A controller can thus be thought of as a middle man between models and views. It makes the model data available to the view so it can display that data to the user, and it saves or updates data from the user to the model.

A controller is a Ruby class which inherits from ApplicationController and has methods just like any other class. When your application receives a request, the routing will determine which controller and action to run, then Rails creates an instance of that controller and runs the public method with the same name as the action.

There's no rule saying a method on a controller has to be an action; they may well be used for other purposes such as filters, which will be covered later in this guide.

As an example, if a user goes to /clients/new in your application to add a new client, Rails will create an instance of ClientsController and run the new method. Note that the empty method from the example above could work just fine because Rails will by default render the new.html.erb view unless the action says otherwise. The new method could make available to the view a @client instance variable by creating a new Client:

The Layouts & rendering guide explains this in more detail.

ApplicationController inherits from ActionController::Base, which defines a number of helpful methods. This guide will cover some of these, but if you're curious to see what's in there, you can see all of them in the API documentation or in the source itself.

You will probably want to access data sent in by the user or other parameters in your controller actions. There are two kinds of parameters possible in a web application. The first are parameters that are sent as part of the URL, called query string parameters. The query string is everything after "?" in the URL. The second type of parameter is usually referred to as POST data. This information usually comes from a HTML form which has been filled in by the user. It's called POST data because it can only be sent as part of an HTTP POST request. Rails does not make any distinction between query string parameters and POST parameters, and both are available in the params hash in your controller:

The params hash is not limited to one-dimensional keys and values. It can contain arrays and (nested) hashes. To send an array of values, append "[]" to the key name:

The value of params[:ids] will now be ["1", "2", "3"]. Note that parameter values are always strings; Rails makes no attempt to guess or cast the type.

To send a hash you include the key name inside the brackets:

The value of params[:client] when this form is submitted will be {"name" ⇒ "Acme", "phone" ⇒ "12345", "address" ⇒ {"postcode" ⇒ "12345", "city" ⇒ "Carrot City"}}. Note the nested hash in params[:client][:address].

Note that the params hash is actually an instance of HashWithIndifferentAccess from Active Support which is a subclass of Hash which lets you use symbols and strings interchangeably as keys.

The params hash will always contain the :controller and :action keys, but you should use the methods controller_name and action_name instead to access these values. Any other parameters defined by the routing, such as :id will also be available. As an example, consider a listing of clients where the list can show either active or inactive clients. We can add a route which captures the :status parameter in a "pretty" URL:

In this case, when a user opens the URL /clients/active, params[:status] will be set to "active". When this route is used, params[:foo] will also be set to "bar" just like it was passed in the query string in the same way params[:action] will contain "index".

You can set global default parameters that will be used when generating URLs with default_url_options. To do this, define a method with that name in your controller:

These options will be used as a starting-point when generating, so it's possible they'll be overridden by url_for. Because this method is defined in the controller, you can define it on ApplicationController so it would be used for all URL generation, or you could define it on only one controller for all URLs generated there.

Your application has a session for each user in which you can store small amounts of data that will be persisted between requests. The session is only available in the controller and the view and can use one of a number of different storage mechanisms:

All session stores use a cookie - this is required and Rails does not allow any part of the session to be passed in any other way (e.g. you can't use the query string to pass a session ID) because of security concerns (it's easier to hijack a session when the ID is part of the URL).

Most stores use a cookie to store the session ID which is then used to look up the session data on the server. The default and recommended store, the CookieStore, does not store session data on the server, but in the cookie itself. The data is cryptographically signed to make it tamper-proof, but it is not encrypted, so anyone with access to it can read its contents but not edit it (Rails will not accept it if it has been edited). It can only store about 4kB of data - much less than the others - but this is usually enough. Storing large amounts of data is discouraged no matter which session store your application uses. You should especially avoid storing complex objects (anything other than basic Ruby objects, the most common example being model instances) in the session, as the server might not be able to reassemble them between requests, which will result in an error. The CookieStore has the added advantage that it does not require any setting up beforehand - Rails will generate a "secret key" which will be used to sign the cookie when you create the application.

Read more about session storage in the Security Guide.

If you need a different session storage mechanism, you can change it in the config/environment.rb file:

Sometimes you don't need a session. In this case, you can turn it off to avoid the unnecessary overhead. To do this, use the session class method in your controller:

You can also turn the session on or off for a single controller:

Or even for specified actions:

In your controller you can access the session through the session instance method.

Session values are stored using key/value pairs like a hash:

To store something in the session, just assign it to the key like a hash:

To remove something from the session, assign that key to be nil:

To reset the entire session, use reset_session.

The flash is a special part of the session which is cleared with each request. This means that values stored there will only be available in the next request, which is useful for storing error messages etc. It is accessed in much the same way as the session, like a hash. Let's use the act of logging out as an example. The controller can send a message which will be displayed to the user on the next request:

The destroy action redirects to the application's root_url, where the message will be displayed. Note that it's entirely up to the next action to decide what, if anything, it will do with what the previous action put in the flash. It's conventional to display eventual errors or notices from the flash in the application's layout:

This way, if an action sets an error or a notice message, the layout will display it automatically.

If you want a flash value to be carried over to another request, use the keep method:

By default, adding values to the flash will make them available to the next request, but sometimes you may want to access those values in the same request. For example, if the create action fails to save a resource and you render the new template directly, that's not going to result in a new request, but you may still want to display a message using the flash. To do this, you can use flash.now in the same way you use the normal flash:

Your application can store small amounts of data on the client - called cookies - that will be persisted across requests and even sessions. Rails provides easy access to cookies via the cookies method, which - much like the session - works like a hash:

Note that while for session values, you set the key to nil, to delete a cookie value, you should use cookies.delete(:key).

Filters are methods that are run before, after or "around" a controller action. For example, one filter might check to see if the logged in user has the right credentials to access that particular controller or action. Filters are inherited, so if you set a filter on ApplicationController, it will be run on every controller in your application. A common, simple filter is one which requires that a user is logged in for an action to be run. You can define the filter method this way:

The method simply stores an error message in the flash and redirects to the login form if the user is not logged in. If a before filter (a filter which is run before the action) renders or redirects, the action will not run. If there are additional filters scheduled to run after the rendering or redirecting filter, they are also cancelled. To use this filter in a controller, use the before_filter method:

In this example, the filter is added to ApplicationController and thus all controllers in the application. This will make everything in the application require the user to be logged in in order to use it. For obvious reasons (the user wouldn't be able to log in in the first place!), not all controllers or actions should require this. You can prevent this filter from running before particular actions with skip_before_filter :

Now, the LoginsController's "new" and "create" actions will work as before without requiring the user to be logged in. The :only option is used to only skip this filter for these actions, and there is also an :except option which works the other way. These options can be used when adding filters too, so you can add a filter which only runs for selected actions in the first place.

In addition to the before filters, you can run filters after an action has run or both before and after. The after filter is similar to the before filter, but because the action has already been run it has access to the response data that's about to be sent to the client. Obviously, after filters can not stop the action from running. Around filters are responsible for running the action, but they can choose not to, which is the around filter's way of stopping it.

While the most common way to use filters is by creating private methods and using *_filter to add them, there are two other ways to do the same thing.

The first is to use a block directly with the *_filter methods. The block receives the controller as an argument, and the require_login filter from above could be rewritten to use a block:

Note that the filter in this case uses send because the logged_in? method is private and the filter is not run in the scope of the controller. This is not the recommended way to implement this particular filter, but in more simple cases it might be useful.

The second way is to use a class (actually, any object that responds to the right methods will do) to handle the filtering. This is useful in cases that are more complex than can not be implemented in a readable and reusable way using the two other methods. As an example, you could rewrite the login filter again to use a class:

Again, this is not an ideal example for this filter, because it's not run in the scope of the controller but gets the controller passed as an argument. The filter class has a class method filter which gets run before or after the action, depending on if it's a before or after filter. Classes used as around filters can also use the same filter method, which will get run in the same way. The method must yield to execute the action. Alternatively, it can have both a before and an after method that are run before and after the action.

The Rails API documentation has more information on using filters.

Verifications make sure certain criteria are met in order for a controller or action to run. They can specify that a certain key (or several keys in the form of an array) is present in the params, session or flash hashes or that a certain HTTP method was used or that the request was made using XMLHTTPRequest (Ajax). The default action taken when these criteria are not met is to render a 400 Bad Request response, but you can customize this by specifying a redirect URL or rendering something else and you can also add flash messages and HTTP headers to the response. It is described in the API documentation as "essentially a special kind of before_filter".

Here's an example of using verification to make sure the user supplies a username and a password in order to log in:

Now the create action won't run unless the "username" and "password" parameters are present, and if they're not, an error message will be added to the flash and the "new" action will be rendered. But there's something rather important missing from the verification above: It will be used for every action in LoginsController, which is not what we want. You can limit which actions it will be used for with the :only and :except options just like a filter:

Cross-site request forgery is a type of attack in which a site tricks a user into making requests on another site, possibly adding, modifying or deleting data on that site without the user's knowledge or permission. The first step to avoid this is to make sure all "destructive" actions (create, update and destroy) can only be accessed with non-GET requests. If you're following RESTful conventions you're already doing this. However, a malicious site can still send a non-GET request to your site quite easily, and that's where the request forgery protection comes in. As the name says, it protects from forged requests. The way this is done is to add a non-guessable token which is only known to your server to each request. This way, if a request comes in without the proper token, it will be denied access.

If you generate a form like this:

You will see how the token gets added as a hidden field:

Rails adds this token to every form that's generated using the form helpers, so most of the time you don't have to worry about it. If you're writing a form manually or need to add the token for another reason, it's available through the method form_authenticity_token:

The Security Guide has more about this and a lot of other security-related issues that you should be aware of when developing a web application.

In every controller there are two accessor methods pointing to the request and the response objects associated with the request cycle that is currently in execution. The request method contains an instance of AbstractRequest and the response method returns a response object representing what is going to be sent back to the client.

The request object contains a lot of useful information about the request coming in from the client. To get a full list of the available methods, refer to the API documentation. Among the properties that you can access on this object are:

Rails collects all of the parameters sent along with the request in the params hash, whether they are sent as part of the query string or the post body. The request object has three accessors that give you access to these parameters depending on where they came from. The query_parameters hash contains parameters that were sent as part of the query string while the request_parameters hash contains parameters sent as part of the post body. The path_parameters hash contains parameters that were recognized by the routing as being part of the path leading to this particular controller and action.

The response object is not usually used directly, but is built up during the execution of the action and rendering of the data that is being sent back to the user, but sometimes - like in an after filter - it can be useful to access the response directly. Some of these accessor methods also have setters, allowing you to change their values.

If you want to set custom headers for a response then response.headers is the place to do it. The headers attribute is a hash which maps header names to their values, and Rails will set some of them - like "Content-Type" - automatically. If you want to add or change a header, just assign it to headers with the name and value:

Rails comes with built-in HTTP Basic authentication. This is an authentication scheme that is supported by the majority of browsers and other HTTP clients. As an example, consider an administration section which will only be available by entering a username and a password into the browser's HTTP Basic dialog window. Using the built-in authentication is quite easy and only requires you to use one method, authenticate_or_request_with_http_basic.

With this in place, you can create namespaced controllers that inherit from AdminController. The before filter will thus be run for all actions in those controllers, protecting them with HTTP Basic authentication.

Sometimes you may want to send a file to the user instead of rendering an HTML page. All controllers in Rails have the send_data and the send_file methods, that will both stream data to the client. send_file is a convenience method which lets you provide the name of a file on the disk and it will stream the contents of that file for you.

To stream data to the client, use send_data:

The download_pdf action in the example above will call a private method which actually generates the file (a PDF document) and returns it as a string. This string will then be streamed to the client as a file download and a filename will be suggested to the user. Sometimes when streaming files to the user, you may not want them to download the file. Take images, for example, which can be embedded into HTML pages. To tell the browser a file is not meant to be downloaded, you can set the :disposition option to "inline". The opposite and default value for this option is "attachment".

If you want to send a file that already exists on disk, use the send_file method. This is usually not recommended, but can be useful if you want to perform some authentication before letting the user download the file.

This will read and stream the file 4Kb at the time, avoiding loading the entire file into memory at once. You can turn off streaming with the :stream option or adjust the block size with the :buffer_size option.

While send_data works just fine, if you are creating a RESTful application having separate actions for file downloads is usually not necessary. In REST terminology, the PDF file from the example above can be considered just another representation of the client resource. Rails provides an easy and quite sleek way of doing "RESTful downloads". Here's how you can rewrite the example so that the PDF download is a part of the show action, without any streaming:

In order for this example to work, you have to add the PDF MIME type to Rails. This can be done by adding the following line to the file config/initializers/mime_types.rb:

Now the user can request to get a PDF version of a client just by adding ".pdf" to the URL:

Rails keeps a log file for each environment (development, test and production) in the "log" folder. These are extremely useful when debugging what's actually going on in your application, but in a live application you may not want every bit of information to be stored in the log file. The filter_parameter_logging method can be used to filter out sensitive information from the log. It works by replacing certain values in the params hash with "[FILTERED]" as they are written to the log. As an example, let's see how to filter all parameters with keys that include "password":

The method works recursively through all levels of the params hash and takes an optional second parameter which is used as the replacement string if present. It can also take a block which receives each key in return and replaces those for which the block returns true.

Most likely your application is going to contain bugs or otherwise throw an exception that needs to be handled. For example, if the user follows a link to a resource that no longer exists in the database, Active Record will throw the ActiveRecord::RecordNotFound exception. Rails' default exception handling displays a 500 Server Error message for all exceptions. If the request was made locally, a nice traceback and some added information gets displayed so you can figure out what went wrong and deal with it. If the request was remote Rails will just display a simple "500 Server Error" message to the user, or a "404 Not Found" if there was a routing error or a record could not be found. Sometimes you might want to customize how these errors are caught and how they're displayed to the user. There are several levels of exception handling available in a Rails application:

By default a production application will render either a 404 or a 500 error message. These messages are contained in static HTML files in the public folder, in 404.html and 500.html respectively. You can customize these files to add some extra information and layout, but remember that they are static; i.e. you can't use RHTML or layouts in them, just plain HTML.

If you want to do something a bit more elaborate when catching errors, you can use rescue_from, which handles exceptions of a certain type (or multiple types) in an entire controller and its subclasses. When an exception occurs which is caught by a rescue_from directive, the exception object is passed to the handler. The handler can be a method or a Proc object passed to the :with option. You can also use a block directly instead of an explicit Proc object.

Here's how you can use rescue_from to intercept all ActiveRecord::RecordNotFound errors and do something with them.

Of course, this example is anything but elaborate and doesn't improve on the default exception handling at all, but once you can catch all those exceptions you're free to do whatever you want with them. For example, you could create custom exception classes that will be thrown when a user doesn't have access to a certain section of your application:

Lighthouse ticket