Real time is important. Users have come to expect real time behavior from every website, because they want to see the latest data without having to reload the page. Real time increases their engagement, provides better context for the data they're seeing, and makes collaboration easier.
Entangled stores and syncs data from ActiveRecord instantly across every device. It is a layer behind your models and controllers that pushes updates to all connected clients in real time. It is cross-browser compatible and offers real time validations.
Currently, Entangled runs on Rails 4.2 and Ruby 2.0 in the back end, and Angular 1.3 in the front end.
Add this line to your application's Gemfile:
gem 'entangled'
Note that Redis and Puma are required as well. Redis is needed to build the channels clients subscribe to, Puma is needed to handle websockets concurrently.
Entangled comes with Redis, but you need to add Puma to your Gemfile:
gem 'puma'
And then execute:
$ bundle
Or install it yourself as:
$ gem install entangled
Entangled is needed in three parts of your app: Routes, models, and controllers. Given the example of a MessagesController
and a Message
model for a chat app, you will need:
Add the following to your routes file:
sockets_for :messages
Under the hood, this creates the following routes:
Prefix Verb URI Pattern Controller#Action
messages GET /messages(.:format) messages#index
message GET /messages/:id(.:format) messages#show
create_messages GET /messages/create(.:format) messages#create
update_message GET /messages/:id/update(.:format) messages#update
destroy_message GET /messages/:id/destroy(.:format) messages#destroy
Note that websockets don't speak HTTP, so only GET requests are available. That's why these routes deviate slightly from restful routes. Also note that there are no edit
and new
actions, since an Entangled controller is only concerned with rendering data, not views.
You can use sockets_for
just like resources
, including the following features:
# Inclusion/exclusion
sockets_for :messages, only: :index
sockets_for :messages, only: [:index, :show]
sockets_for :messages, except: :index
sockets_for :messages, except: [:index, :show]
# Nesting
sockets_for :parents do
sockets_for :children
end
# Multiple routes at once
sockets_for :foos, :bars
# ...etc
Add the following to the top of your model (e.g., a Message
model):
class Message < ActiveRecord::Base
include Entangled::Model
entangle
end
This will create the callbacks needed to push changes to data to all clients who are subscribed. This is essentially where the data binding is set up.
By default, the following callbacks will be added:
after_create
after_update
after_destroy
You can limit this behavior by specifying :only
or :except
options. For example, if you don't want to propagate the destruction or update of an object to all connected clients, you can do the following:
entangle only: :create
entangle only: [:create, :update]
Your controllers will be a little more lightweight than in a standard restful Rails app. A restful-style controller is expected and should look like this:
class MessagesController < ApplicationController
include Entangled::Controller
def index
broadcast do
@messages = Message.all
end
end
def show
broadcast do
@message = Message.find(params[:id])
end
end
def create
broadcast do
@message = Message.create(message_params)
end
end
def update
broadcast do
@message = Message.find(params[:id])
@message.update(message_params)
end
end
def destroy
broadcast do
@message = Message.find(params[:id]).destroy
end
end
private
def message_params
# params logic here
end
end
Note the following:
- All methods are wrapped in a new
broadcast
block needed to receive and send data to connected clients - The
index
action will expect an instance variable with the same name as your controller in the plural form (e.g.@messages
in aMessagesController
) - The
show
,create
,update
, anddestroy
actions will expect an instance variable with the singular name of your controller (e.g.@message
in aMessagesController
) - The instance variables are sent to clients as stringified JSON
- Strong parameters are expected
Remember to run Redis whenever you run your server:
$ redis-server
Otherwise the channels won't work.
If you store your Redis instance in $redis
or REDIS
(e.g. in an initializer), Entangled will use that assigned instance so that you can configure Redis just like you're used to. Otherwise, Entangled will instantiate Redis itself and use its default settings.
You will need to configure your client to create Websockets and understand incoming requests on those sockets. In order to use the helper methods for the front end provided by the Entangled Angular library, you must use Angular in your front end. The use of Angular as counterpart of this gem is highly recommended, since its inherent two way data binding complements the real time functionality of this gem nicely.
You can either download or reference the file entangled.js
from this repository, or simply install it with Bower:
$ bower install entangled
Then include it in your HTML.
Lastly, add the Entangled module as a dependency to your Angular app:
angular.module('appName', ['entangled']);
Entangled is best used within Angular services. For example, consider a Message
service for a chat app:
app.factory('Message', function(Entangled) {
return new Entangled('ws://localhost:3000/messages');
});
In the above example, first we inject Entangled into our service, then instantiate a new Entangled object and return it. The Entangled object takes one argument when instantiated: the URL of your resource's index action (in this case, /messages
). Note that the socket URL looks just like a standard restful URL with http, except that the protocol part has been switched with ws
to use the websocket protocol. Also note that you need to use wss
instead if you want to use SSL.
The Entangled service comes with these functions:
new(params)
create(params, callback)
find(id, callback)
all(callback)
...and the following functions on returned objects:
$save(callback)
$update(params, callback)
$destroy(callback)
They're just like class and instance methods in Active Record.
In your controller, you could then inject that Message
service and use it like so:
// To instantiate a blank message, e.g. for a form;
// You can optionally pass in an object to new() to
// set some default values
$scope.message = Message.new();
// To instantiate and save a message in one go
Message.create({ body: 'text' }, function(message) {
$scope.$apply(function() {
$scope.message = message;
});
});
// To retrieve a specific message from the server
// with id 1 and subscribe to its channel
Message.find(1, function(message) {
$scope.$apply(function() {
$scope.message = message;
});
});
// To retrieve all messages from the server and
// subscribe to the collection's channel
Message.all(function(messages) {
$scope.$apply(function() {
$scope.messages = messages;
});
});
// To store a newly instantiated or update an existing message.
// If saved successfully, $scope.message is updated in place
// with the attributes id, created_at and updated_at
$scope.message.body = 'new body';
$scope.message.$save(function() {
// Do stuff after save
});
// To update a newly instantiated or existing message in place.
// If updated successfully, $scope.message is updated in place
// with the attributes id, created_at and updated_at
$scope.message.$update({ body: 'new body' }, function() {
// Do stuff after update
});
// To destroy a message
$scope.message.$destroy(function() {
// Do stuff after destroy
});
All functions above will interact with your server's controllers in real time. Your scope variables will always reflect your server's most current data.
Objects from the Entangled service automatically receive ActiveRecord's error messages from your model when you $save()
. An additional property called errors
containing the error messages is available, formatted the same way you're used to from calling .errors
on a model in Rails.
For example, consider the following scenario:
# Message model (Rails)
validates :body, presence: true
// Controller (Angular)
$scope.message.$save(function() {
console.log($scope.message.errors);
// => { body: ["can't be blank"] }
});
You could then display these error messages to your users.
To check if a resource is valid, you can use $valid()
and $invalid()
. Both functions return booleans. For example:
$scope.message.$save(function() {
// Check if record has no errors
if ($scope.message.$valid()) { // similar to ActiveRecord's .valid?
alert('Yay!');
}
// Check if record errors
if ($scope.message.$invalid()) { // similar to ActiveRecord's .invalid?
alert('Nay!');
}
});
Note that $valid()
and $invalid()
should only be used after $saving a resource, i.e. in the callback of $save
, since they don't actually invoke server side validations. They only check if a resource contains errors.
Just as with ActiveRecord's persisted?
method, you can use $persisted()
on an object to check if it was successfully stored in the database.
$scope.message.$persisted();
// => true or false
What if you want to only fetch and subscribe to children that belong to a specific parent? Or maybe you want to create a child in your front end and assign it to a specific parent?
Entangled currently supports one belongs_to
association per model.
For example, imagine the following Parent > Children relationship in your models:
class Parent < ActiveRecord::Base
include Entangled::Model
entangle
has_many :children
end
class Child < ActiveRecord::Base
include Entangled::Model
entangle
belongs_to :parent
end
To reflect this in your front end, you just need to add three things to your app:
- Nest your routes so that they resemble the parent/child relationship:
sockets_for :parents do
sockets_for :children
end
- Adjust the
index
andcreate
actions in yourChildrenController
so that they look like this:
class ChildrenController < ApplicationController
include Entangled::Controller
# Fetch children of specific parent
def index
broadcast do
@children = Parent.find(params[:parent_id]).children
end
end
# Create child of specific parent
def create
broadcast do
@child = Parent.find(params[:parent_id]).children.create(child_params)
end
end
# show, update and destroy don't need to be nested
end
- Lastly, inform your Angular parent service about the association:
app.factory('Parent', function(Entangled) {
// Instantiate Entangled service
var Parent = new Entangled('ws://localhost:3000/parents');
// Set up association
Parent.hasMany('children');
return Parent;
});
This makes a children()
function available on your parent records on which you can chain all other functions to fetch/manipulate data:
Parent.find(1, function(parent) {
parent.children().all(function(children) {
// children here all belong to parent with id 1
});
parent.children().find(1, function(child) {
// child has id 1 and belongs to parent with id 1
});
parent.children().create({ foo: 'bar' }, function(child) {
// child has been persisted and associated with parent
});
// etc
});
This is the way to go if you want to fetch records that only belong to a certain record, or create records that should belong to a parent record. In short, it is ideal to scope records to parent records.
Naturally, all nested records are also synced in real time across all connected clients.
This gem is best used for Rails apps that serve as APIs only and are not concerned with rendering views, since Entangled controllers cannot render views. A front end separate from your Rails app is recommended, either in your Rails app's public directory, or a separate front end app altogether.
The gem relies heavily on convention over configuration and currently only works with restful style controllers as shown above. More features will be available soon, such as associations, authentication, and more.
The following features are to be implemented next:
- Check if routes really allow options right now. For example, what if I pass in shallow: true? Run rake routes to check!
- Allow for more than one level of nesting of
#channels
inEntangled::Model
- Support
belongsTo
in front end - Support deeply nested
belongs_to
, e.g.Parent > Child > Grandchild
- Support
has_one
association in back end and front end - Add offline capabilities
- Add authentication - with JWT?
- On Heroku, tasks are always in different order depending on which ones are checked off and not
- Add
$onChange
function to objects - or could a simple $watch and $watchCollection suffice? - Add diagram on how it works to Readme
- Check if Rails 4.0.0 supported too
- GNU instead of MIT? Or something else? How to switch?
- Contact Jessy to tweet about it!
- Handle errors gracefully (e.g. finding a non-existent id, etc, authorization error in the back end, timeouts, etc)
- Test controllers (see ngauthier/tubesock#41)
- Freeze destroyed object
- Set
$persisted()
to false on a destroyed object - Add
.destroyAll()
function toResources
- Add support for plain JavaScript usage (without Angular) and add section about that to Readme
- Fork it - you will notice that the repo comes with a back end and a front end part to test both parts of the gem
- Run
$ bundle install
in the root of the repo - Run
$ bower install
and$ npm install
in spec/dummy/public - The back end example app resides in spec/dummy. You can run
rails
andrake
commands in there if you prefix them withbin/
, i.e.$ bin/rails s
or$ bin/rake db:schema:load
. Run your Rails tests in the root of the repo by running$ rspec
- The front end example app resides in spec/dummy/public. To look at it in your browser, cd into spec/dummy/public and run
$ bin/rails s
. Tests for this part of the app can be located under spec/dummy/public/test and are written with Jasmine. To run the tests, first run$ bin/rails -e test
to start up the server in test mode, and then run$ grunt test
in a new terminal tab. It's important to remember that changes you make to the server will not take effect until you restart the server since you're running it in the test environment. Also remember to prepare the test database by running$ bin/rake db:test:prepare
- The Entangled Angular service resides in spec/dummy/public/app/entangled/entangled.js. This is where you can make changes to the service. A copy of it, living in /entangled.js at the root of the repo, should be kept in sync for it to be available with Bower. Once you're done editing spec/dummy/public/app/entangled/entangled.js, copy it over to /entangled.js
- Write your tests. Test coverage is required
- Write your feature to make the tests pass
- Stage and commit your changes
- Push to a new feature branch in your repo
- Send me a pull request!
Thanks to Ilias Tsangaris for inspiring the name "Entanglement" based on Quantum Entanglement where pairs or groups of particles always react to changes as a whole, i.e. changes to one particle will result in immediate change of all particles in the group.