DEPREATED

This gem was kind of an experiment to automate API creation for easy integration with javascript clients (primarly sproutcore, now known as Ember). Unfortunately after using it in reality, it turns out that this kind of automation of bulk operations can be hard with complicated APIs. Currently I have no plans to continue development and I use presenters and regular controllers for creating APIs for Ember. One example of such approach is active model serializers. If you want to maintaint bulk_api, please fork and let me know.

Bulk Rails API

Bulk Rails API plugin makes integrating Sproutcore applications with Rails applications dead simple. It handles all the communication and allows to take advantage of bulk operations, which can make your application much faster. To use that plugin you will also need BulkDataSource, which will handle Sproutcore side of communcation.

Example

https://github.com/drnic/todos-bulk-api-demo

Installing

After following this guide you will have a setup like this:

$RAILS_ROOT/app/sproutcore is the path where your sproutcore application lives
/api/bulk is routed to serve the API
/_sproutcore is routed to serve your sproutcore application

Rails

Add this line to Gemfile and run bundle install:

gem 'bulk_api'

To set up Bulk API in your Rails app:

rails generate bulk:install

Sproutcore

Now you need to configure your application. Install BulkDataSource:

cd <sproutcore_root>
mkdir frameworks # In case it does not exist yet
git clone https://github.com/drogus/bulk_data_source.git frameworks/bulk_data_source

Change your Buildfile to include it:

config :all, :required => [:sproutcore, "sproutcore/core_foundation", "bulk_data_source"]

Configure your store:

YourApp = SC.Application.create({
  store: SC.Store.create().from('BulkApi.BulkDataSource')
});

The last thing that you need to do is to set resource names for your models. BulkDataSource assumes that you have resourceName attribute set on your model. If you don't have such attributes, you can add it like that:

Todos.Todo = SC.Record.extend({
  // your code here
});
Todos.Todo.resourceName = 'todo';

Usage

By default Bulk Api plugin handles all of the models, in production you will probably want to filter it:

# app/bulk/abstract_resource.rb
class ApplicationResource < Bulk::Resource
  resources :tasks, :projects
end

If you don't have any specific needs like authentication or authorization, you're good to go with such simple configuration. In other cases you will need to do a bit more to integrate your application.

Bulk API approach is a bit different than standard REST APIs that you're probably used to, thus it needs to be handled differently. The point of using bulk API is to cut the requests number - it can handle many records (and many types of records) with one request. If you want to read more about how the API looks from HTTP point of view, please scroll to "HTTP Api" section. For now let's focus on what you need to know to implement it in ruby.

When using bulk api you can handle things on 3 levels: 1) All records level 2) Particular record type level 3) Individual record level

Let's see how to handle things on all of the 3 levels to add your own logic (like authentication or authorization).

If some of your logic is common for all the record types, you can use ApplicationResource that lives in app/bulk/abstract_resource.rb. This is base class for all of the resources, just like ApplicationController is a base class for all of your controllers (this may not be true for some applications, but let's agree that's the most common scenario). To allow easy integration with application, you have access to several application objects in ApplicationResource and its subclasses:

session
controller
params

The methods used for records manipulation are:

get
create
update
delete

Authentication callbacks

There are 3 kind of authorization callbacks that you can use. Each of them represents differnet level of handling records:

authenticate(action) - that callback is executed before handling the request, if it returns false, the entire response gets 401 status
authenticate_records(action, klass) - this callback is run before handling each type of resource, if it returns false, not_authenticated error is added to all of the records from given resource type. The class of the resource is passed as an argument.
authenticate_record(action, record) - this callback is run for each of the records, if it returns false, not_authenticated error is added to the given record

Let's see example usage of each of this callbacks types:

class ApplicationResource < Bulk::Resource
  # delegate all the things that we need from controller
  delegate current_user, :can?, :to => :controller

  def authenticate(action)
    current_user.logged_in?
  end

  def authenticate_records(action, klass)
  end

  def authenticate_record(action, record)
  end
end

Authorization callbacks

Authorization callbacks are very similar to authentication callbacks. Notice that authorization callbacks will only be run when authentication callback succeeds.

authorize(action) - that callback is executed before handling the request, if it returns false, the entire response gets 403 status
authorize_records(action, klass) - this callback is run before handling each type of resource, if it returns false, 403 error is added to all of the records from given resource type. The class of the resource is passed as an argument.
authorize_record(action, record) - this callback is run for each of the records, if it returns false, 403 error is added to the given record

Let's see example usage of each of this callbacks types.

class ApplicationResource < Bulk::Resource
  # delegate all the things that we need from controller
  delegate current_user, :can?, :to => :controller

  def authorize(action)
    current_user.is_admin?
  end

  def authorize_records(action, klass)
    # klass can be for example Project
    if action == :update
      can? :update, klass
    end
  end

  def authorize_record(action, record)
    can? action, record # action returns one of the 4 actions (get, create, update, delete),
                        # so this will check if user can perform given type of action on
                        # the record
  end
end

Params filtering

While preparing your API, you will probably need to filter parameters that user can set on your models. The easiest way to do it is to set params_accessible or params_protected callbacks:

class ApplicationResource < Bulk::Resource
  def params_accessible(klass)
    { :tasks    => [:title, :done],
                    :projects => [:name] }
  end

  # or:

  def params_protected(klass)
    { :tasks => [:created_at, :updated_at] }
  end
end

You can also set it for individual resource classes. In such case this will overwrite the one that's set in ApplicationResource.

Attributes filtering

If you want to filter the attributes that are sent in a response, the easiest way to do it is to use standard Rails mechanism for that - override as_json method in your model:

class MyModel < ActiveRecord::Base
  def as_json(options={})
    super(:only => [:email, :avatar], :include =>[:addresses])
  end
end

With some applications that's not enough, though. If you have several user roles, the chances are that you will need to differentiate responses based on user rights. In that case you can use as_json callback. Value returned from that callback will be passed to the record's as_json method:

class ApplicationResource < Bulk::Resource
  def as_json(record)
    # return hash that will be passed to record's as_json
    { :only => [:email] }
  end
end

You can also override that method in individual resource classes.

Specific resource classes

Sometimes you may want to implement specific application logic to one of the resources. Or you don't want to end up with Switch Driven Development in one of you authenticate callbacs. In such cases, the easiest way to handle resource specific code is to create an ApplicationResouce subclass that you can use to override standard behavior. There is a generator to make things easy for you:

rails g bulk:resource task

This will create the following file:

# app/bulk/task_resource.rb
class TaskResource < ApplicationResource
end

If you do nothing else, Rails will automatically return records by retrieving them using ActiveRecord. You can customize any of the default behavior by overriding methods on the resource class. As you already now the main methods that are used to fetch and modify records are: get, create, update, delete. Let's see how you can override those methods:

# app/resources/task_resource.rb
class TaskResource < Sproutcore::Resource

  def get(ids)
    # ids is an array with records that we need to fetch

    collection = super(ids)

    # collection is an instance of Bulk::Collection class that keeps
    # fetched records, please check the rest of the README and the docs
    # to see how you can manipulate records in collection
    collection
  end

  def create(records)
    # records is an array of hashes with data that will be used
    # to create new records e.g.:
    # [{:title => "First", :done => false, :_local_id => 1},
    #  {:title => "", :done => true, :_local_id => 3}]
    # _local_id is needed to identify the records in sproutcore
    # application, since they do not have an id yet

    collection = super(records)

    collection
  end

  def update(records)
    # records array is very similar to the array from create method,
    # but this time we should get data with id, like:
    # [{:id => 1, :title => "First (changed!)", :done => false},
    #  {:id => 2, :title => ""}]

    collection = super(records)

    collection
  end

  def delete(ids)
    # similarly to get method, we get array of ids to delete

    collection = super(ids)

    collection
  end
end

While overriding records you can use super to handle the actions with default behavior or reimplement them yourself. In latter case you just need to make sure that you properly construct collection.

Bulk::Collection

Bulk::Collection is a container for records and is used to construct response from. It has a few handy methods to easily modify collection, for more please refer to documantation.

collection = Bulk::Collection.new
collection.set(1, record) # add record with identifier '1', identifier is then used while constructing response
                          # most of the time it's id or _local_id (the latter one is mainly for create)

collection.errors.set(1, :access_denied) # add error that will be passed to DataSource
                                         # notice that these errors are not the same as validation errors,
                                         # this is more general way to tell DataSource what's going on
collection.delete(1, record) # remove the record

Advanced usage

If you want to change class or resource name that will be send, you can use resource_class and resource_name methods:

class TaskResource < Sproutcore::Resource
  resource_class Todo
  resource_name 'todo'
end

Http Bulk API

The point of using bulk API is to cut the requests number. Because of its nature it can't be handled efficiently using standard REST API. The bulk API is designed to handle many records and record types in one request. Let's look how does GET request can look like with bulk API;

POST /bulk/api
{
  'todos': [
    {'title': "First todo", 'done': false, '_local_id': '3'},
    {'title': "Second todo", 'done': true, '_local_id': '10'}
  ],
  'projects': [
    {'name': "Sproutcore todolist", '_local_id': '12'}
  ]
}

As you can see we POST some new items to our application. Rails application will then respond with list of created records:

{
  'todos': [
    {'id': 1, 'title': "First todo", 'done': false, '_local_id': '3'}
  ],
  'projects': [
    {'id': 1, 'name': "Sproutcore todolist", '_local_id': '12'}
  ]
  'errors': {
    'todos': {
    }
  }
}

As you can see, all the records that were created have id attached now and there is additional attribute 'errors' that tells us what went wrong during validation.

drogus / bulk_api

readme