README.md in actioncable-5.0.0.beta2 vs README.md in actioncable-5.0.0.beta3

- old
+ new

@@ -15,11 +15,11 @@ WebSockets open to your application if they use multiple browser tabs or devices. The client of a WebSocket connection is called the consumer. Each consumer can in turn subscribe to multiple cable channels. Each channel encapsulates a logical unit of work, similar to what a controller does in a regular MVC setup. For example, -you could have a `ChatChannel` and a `AppearancesChannel`, and a consumer could be subscribed to either +you could have a `ChatChannel` and an `AppearancesChannel`, and a consumer could be subscribed to either or to both of these channels. At the very least, a consumer should be subscribed to one channel. When the consumer is subscribed to a channel, they act as a subscriber. The connection between the subscriber and the channel is, surprise-surprise, called a subscription. A consumer can act as a subscriber to a given channel any number of times. For example, a consumer @@ -37,11 +37,11 @@ ## Examples ### A full-stack example The first thing you must do is define your `ApplicationCable::Connection` class in Ruby. This -is the place where you authorize the incoming connection, and proceed to establish it +is the place where you authorize the incoming connection, and proceed to establish it, if all is well. Here's the simplest example starting with the server-side connection class: ```ruby # app/channels/application_cable/connection.rb module ApplicationCable @@ -71,11 +71,11 @@ automatically sent to the connection instance when a new connection is attempted, and you use that to set the `current_user`. By identifying the connection by this same current_user, you're also ensuring that you can later retrieve all open connections by a given user (and potentially disconnect them all if the user is deleted or deauthorized). -Then you should define your `ApplicationCable::Channel` class in Ruby. This is the place where you put +Next, you should define your `ApplicationCable::Channel` class in Ruby. This is the place where you put shared logic between your channels. ```ruby # app/channels/application_cable/channel.rb module ApplicationCable @@ -92,23 +92,23 @@ @App = {} App.cable = ActionCable.createConsumer("ws://cable.example.com") ``` -The ws://cable.example.com address must point to your set of Action Cable servers, and it +The `ws://cable.example.com` address must point to your Action Cable server(s), and it must share a cookie namespace with the rest of the application (which may live under http://example.com). This ensures that the signed cookie will be correctly sent. That's all you need to establish the connection! But of course, this isn't very useful in itself. This just gives you the plumbing. To make stuff happen, you need content. That content is defined by declaring channels on the server and allowing the consumer to subscribe to them. ### Channel example 1: User appearances -Here's a simple example of a channel that tracks whether a user is online or not and what page they're on. -(This is useful for creating presence features like showing a green dot next to a user name if they're online). +Here's a simple example of a channel that tracks whether a user is online or not, and also what page they are currently on. +(This is useful for creating presence features like showing a green dot next to a user's name if they're online). First you declare the server-side channel: ```ruby # app/channels/appearance_channel.rb @@ -178,19 +178,19 @@ ``` Simply calling `App.cable.subscriptions.create` will setup the subscription, which will call `AppearanceChannel#subscribed`, which in turn is linked to original `App.cable` -> `ApplicationCable::Connection` instances. -We then link the client-side `appear` method to `AppearanceChannel#appear(data)`. This is possible because the server-side +Next, we link the client-side `appear` method to `AppearanceChannel#appear(data)`. This is possible because the server-side channel instance will automatically expose the public methods declared on the class (minus the callbacks), so that these can be reached as remote procedure calls via a subscription's `perform` method. ### Channel example 2: Receiving new web notifications The appearance example was all about exposing server functionality to client-side invocation over the WebSocket connection. But the great thing about WebSockets is that it's a two-way street. So now let's show an example where the server invokes -action on the client. +an action on the client. This is a web notification channel that allows you to trigger client-side web notifications when you broadcast to the right streams: ```ruby @@ -213,11 +213,11 @@ # Somewhere in your app this is called, perhaps from a NewCommentJob ActionCable.server.broadcast \ "web_notifications_#{current_user.id}", { title: 'New things!', body: 'All the news that is fit to print' } ``` -The `ActionCable.server.broadcast` call places a message in the Redis' pubsub queue under a separate broadcasting name for each user. For a user with an ID of 1, the broadcasting name would be `web_notifications_1`. +The `ActionCable.server.broadcast` call places a message in the Action Cable pubsub queue under a separate broadcasting name for each user. For a user with an ID of 1, the broadcasting name would be `web_notifications_1`. The channel has been instructed to stream everything that arrives at `web_notifications_1` directly to the client by invoking the `#received(data)` callback. The data is the hash sent as the second parameter to the server-side broadcast call, JSON encoded for the trip across the wire, and unpacked for the data argument arriving to `#received`. @@ -232,11 +232,11 @@ stream_from "chat_#{params[:room]}" end end ``` -Pass an object as the first argument to `subscriptions.create`, and that object will become your params hash in your cable channel. The keyword `channel` is required. +If you pass an object as the first argument to `subscriptions.create`, that object will become the params hash in your cable channel. The keyword `channel` is required. ```coffeescript # Client-side, which assumes you've already requested the right to send web notifications App.cable.subscriptions.create { channel: "ChatChannel", room: "Best Room" }, received: (data) -> @@ -291,50 +291,52 @@ The rebroadcast will be received by all connected clients, _including_ the client that sent the message. Note that params are the same as they were when you subscribed to the channel. ### More complete examples -See the [rails/actioncable-examples](http://github.com/rails/actioncable-examples) repository for a full example of how to setup Action Cable in a Rails app and adding channels. +See the [rails/actioncable-examples](http://github.com/rails/actioncable-examples) repository for a full example of how to setup Action Cable in a Rails app, and how to add channels. ## Configuration -Action Cable has three required configurations: the Redis connection, allowed request origins, and the cable server url (which can optionally be set on the client side). +Action Cable has three required configurations: a subscription adapter, allowed request origins, and the cable server URL (which can optionally be set on the client side). ### Redis By default, `ActionCable::Server::Base` will look for a configuration file in `Rails.root.join('config/cable.yml')`. -This file must specify a Redis url for each Rails environment. It may use the following format: +This file must specify an adapter and a URL for each Rails environment. It may use the following format: ```yaml production: &production + adapter: redis url: redis://10.10.3.153:6381 development: &development + adapter: redis url: redis://localhost:6379 test: *development ``` -You can also change the location of the Redis config file in a Rails initializer with something like: +You can also change the location of the Action Cable config file in a Rails initializer with something like: ```ruby Rails.application.paths.add "config/cable", with: "somewhere/else/cable.yml" ``` ### Allowed Request Origins Action Cable will only accept requests from specified origins, which are passed to the server config as an array. The origins can be instances of strings or regular expressions, against which a check for match will be performed. ```ruby -ActionCable.server.config.allowed_request_origins = ['http://rubyonrails.com', /http:\/\/ruby.*/] +Rails.application.config.action_cable.allowed_request_origins = ['http://rubyonrails.com', /http:\/\/ruby.*/] ``` When running in the development environment, this defaults to "http://localhost:3000". To disable and allow requests from any origin: ```ruby -ActionCable.server.config.disable_request_forgery_protection = true +Rails.application.config.action_cable.disable_request_forgery_protection = true ``` ### Consumer Configuration Once you have decided how to run your cable server (see below), you must provide the server url (or path) to your client-side setup. @@ -345,15 +347,15 @@ something like: `App.cable = ActionCable.createConsumer("/cable")`. The second option is to pass the server url through the `action_cable_meta_tag` in your layout. This uses a url or path typically set via `config.action_cable.url` in the environment configuration files, or defaults to "/cable". -This method is especially useful if your websocket url might change between environments. If you host your production server via https, you will need to use the wss scheme -for your ActionCable server, but development might remain http and use the ws scheme. You might use localhost in development and your +This method is especially useful if your WebSocket url might change between environments. If you host your production server via https, you will need to use the wss scheme +for your Action Cable server, but development might remain http and use the ws scheme. You might use localhost in development and your domain in production. -In any case, to vary the websocket url between environments, add the following configuration to each environment: +In any case, to vary the WebSocket url between environments, add the following configuration to each environment: ```ruby config.action_cable.url = "ws://example.com:28080" ``` @@ -372,11 +374,11 @@ ### Other Configurations The other common option to configure is the log tags applied to the per-connection logger. Here's close to what we're using in Basecamp: ```ruby -ActionCable.server.config.log_tags = [ +Rails.application.config.action_cable.log_tags = [ -> request { request.env['bc.account_id'] || "no-account" }, :action_cable, -> request { request.uuid } ] ``` @@ -387,11 +389,11 @@ ## Running the cable server ### Standalone -The cable server(s) is separated from your normal application server. It's still a rack application, but it is its own rack +The cable server(s) is separated from your normal application server. It's still a Rack application, but it is its own Rack application. The recommended basic setup is as follows: ```ruby # cable/config.ru require ::File.expand_path('../../config/environment', __FILE__) @@ -408,20 +410,20 @@ The above will start a cable server on port 28080. ### In app -If you are using a threaded server like Puma or Thin, the current implementation of ActionCable can run side-along with your Rails application. For example, to listen for WebSocket requests on `/cable`, mount the server at that path: +If you are using a threaded server like Puma or Thin, the current implementation of Action Cable can run side-along with your Rails application. For example, to listen for WebSocket requests on `/cable`, mount the server at that path: ```ruby # config/routes.rb Example::Application.routes.draw do mount ActionCable.server => '/cable' end ``` -For every instance of your server you create and for every worker your server spawns, you will also have a new instance of ActionCable, but the use of Redis keeps messages synced across connections. +For every instance of your server you create and for every worker your server spawns, you will also have a new instance of Action Cable, but the use of Redis keeps messages synced across connections. ### Notes Beware that currently the cable server will _not_ auto-reload any changes in the framework. As we've discussed, long-running cable connections mean long-running objects. We don't yet have a way of reloading the classes of those objects in a safe manner. So when you change your channels, or the model your channels use, you must restart the cable server. @@ -429,28 +431,34 @@ The WebSocket server doesn't have access to the session, but it has access to the cookies. This can be used when you need to handle authentication. You can see one way of doing that with Devise in this [article](http://www.rubytutorial.io/actioncable-devise-authentication). ## Dependencies -Action Cable is currently tied to Redis through its use of the pubsub feature to route -messages back and forth over the WebSocket cable connection. This dependency may well -be alleviated in the future, but for the moment that's what it is. So be sure to have -Redis installed and running. +Action Cable provides a subscription adapter interface to process its pubsub internals. By default, asynchronous, inline, PostgreSQL, evented Redis, and non-evented Redis adapters are included. The default adapter in new Rails applications is the asynchronous (`async`) adapter. To create your own adapter, you can look at `ActionCable::SubscriptionAdapter::Base` for all methods that must be implemented, and any of the adapters included within Action Cable as example implementations. -The Ruby side of things is built on top of [faye-websocket](https://github.com/faye/faye-websocket-ruby) and [concurrent-ruby](https://github.com/ruby-concurrency/concurrent-ruby). +The Ruby side of things is built on top of [websocket-driver](https://github.com/faye/websocket-driver-ruby), [nio4r](https://github.com/celluloid/nio4r), and [concurrent-ruby](https://github.com/ruby-concurrency/concurrent-ruby). ## Deployment -Action Cable is powered by a combination of websockets and threads. All of the +Action Cable is powered by a combination of WebSockets and threads. All of the connection management is handled internally by utilizing Ruby’s native thread support, which means you can use all your regular Rails models with no problems as long as you haven’t committed any thread-safety sins. But this also means that Action Cable needs to run in its own server process. So you'll have one set of server processes for your normal web work, and another -set of server processes for the Action Cable. The former can be single-threaded, -like Unicorn, but the latter must be multi-threaded, like Puma. +set of server processes for the Action Cable. + +The Action Cable server does _not_ need to be a multi-threaded application server. +This is because Action Cable uses the [Rack socket hijacking API](http://old.blog.phusion.nl/2013/01/23/the-new-rack-socket-hijacking-api/) +to take over control of connections from the application server. Action Cable +then manages connections internally, in a multithreaded manner, regardless of +whether the application server is multi-threaded or not. So Action Cable works +with all the popular application servers -- Unicorn, Puma and Passenger. + +Action Cable does not work with WEBrick, because WEBrick does not support the +Rack socket hijacking API. ## License Action Cable is released under the MIT license: