RABL for Rails Build Status

rabl-rails is a ruby templating system for rendering your objects in different format (JSON, XML, PLIST).

This gem aims for speed and little memory footprint while letting you build complex response with a very intuitive DSL.

rabl-rails targets Rails 4.2/5/6 application and have been testing with MRI and jRuby.

Installation

Install as a gem :

gem install rabl-rails

or add directly to your Gemfile

gem 'rabl-rails', '~> 0.6.0'

Overview

The gem enables you to build responses using views like you would using HTML/erb/haml. As example, assuming you have a Post model filled with blog posts, and a PostController that look like this:

class PostController < ApplicationController
  def index
     @posts = Post.order('created_at DESC')
  end
end

You can create the following RABL-rails template to express the API output of @posts

# app/views/post/index.rabl
collection :@posts

attributes :id, :title, :subject
child(:user) { attributes :full_name }
node(:read) { |post| post.read_by?(@user) }

This would output the following JSON when visiting http://localhost:3000/posts.json

[{
  "id" : 5, title: "...", subject: "...",
  "user" : { full_name : "..." },
  "read" : true
}]

How it works

This gem separates compiling, ie. transforming a RABL-rails template into a Ruby hash, and the actual rendering of the object or collection. This allows to only compile the template once (when template caching is enabled) which is the slow part, and only use hashes during rendering.

The drawback of compiling the template outside of any rendering context is that we can't access instance variables like usual. Instead, you'll mostly use symbols representing your variables and the gem will retrieve them when needed.

There are places where the gem allows for "dynamic code" -- code that is evaluated at each rendering, such as within node or condition blocks.

# We reference the @posts varibles that will be used at rendering time
collection :@posts

# Here you can use directly the instance variable because it
# will be evaluated when rendering the object
node(:read) { |post| post.read_by?(@user) }

The same rule applies for view helpers such as current_user

After the template is compiled into a hash, rabl-rails will use a renderer to create the actual output. Currently, JSON, XML and PList formats are supported.

Configuration

RablRails works out of the box, with default options and fastest engine available (oj, libxml). But depending on your needs, you might want to change that or how your output looks like. You can set global configuration in your application:

# config/initializers/rabl_rails.rb

RablRails.configure do |config|
  # These are the default
  # config.cache_templates = true
  # config.include_json_root = true
  # config.json_engine = ::Oj
  # config.xml_options = { :dasherize => true, :skip_types => false }
  # config.enable_jsonp_callbacks = false
  # config.replace_nil_values_with_empty_strings = false
  # config.replace_empty_string_values_with_nil = false
  # config.exclude_nil_values = false
  # config.non_collection_classes = Set.new(['Struct'])
end

Usage

Data declaration

To declare data to use in the template, you can use either object or collection with the symbol name or your data.

# app/views/users/show.json.rabl
object :@user

# app/views/users/index.json.rabl
collection :@users

You can specify root label for the collection using hash or :root option

collection :@posts, root: :articles
#is equivalent to
collection :@posts => :articles

# => { "articles" : [{...}, {...}] }

There are rares cases when the template doesn't map directly to any object. In these cases, you can set data to false.

object false
node(:some_count) { |_| @user.posts.count }
child(:@user) { attribute :name }

If you use gems like decent_exposure or focused_controller, you can use your variable directly without the leading @

object :object_exposed

Attributes / Methods

Adds a new field to the response object, calling the method on the object being rendered. Methods called this way should return natives types from the format you're using (such as String, integer, etc for JSON). For more complex objects, see child nodes.

attributes :id, :title, :to_s

You can aliases these attributes in your response

attributes :my_custom_method, as: :title
# => { "title" : <result of my_custom_method> }

or show attributes based on a condition. The currently rendered object is given to the proc condition.

attributes :published_at, :anchor, if: ->(post) { post.published? }

Child nodes

Changes the object being rendered for the duration of the block. Depending on if you use node or glue, the result will be added as a new field or merged respectively.

Data passed can be a method or a reference to an instance variable.

For example if you have a Post model that belongs to a User and want to add the user's name to your response.

object :@post

child(:user, as: :author) do
    attributes :name
end
# => { "post": { "author" : { "name" : "John D." } } }

If instead of having an author node in your response you wanted the name at the root level, you can use glue:

object :@post

glue(:user) do
  attributes :name, as: :author_name
end
# => { "post": { "author_name" : "John D." } }

Arbitrary data source can also be passed:

# in your controller
# @custom_data = [...]

# in the view
child(:@custom_data) do
    attributes :id, :name
end
# => { "custom_data": [...] }

You can use a Hash-like data source, as long as keys match a method or attribute of your main resource, using the fetch keyword:

# assuming you have something similar in your controller
# @users_hash = { 1 => User.new(pseudo: 'Batman') }

# in the view
object :@post

fetch(:@users_hash, as: :user, field: :user_id) do
  attributes :pseudo
end
# => { user: { pseudo: 'Batman' } }

This comes very handy when adding attributes from external queries not really bound to a relation, like statistics.

Constants

Adds a new field to the response using an immutable value.

const(:api_version, API::VERSION)
const(:locale, 'fr_FR')

Lookups

Adds a new field to the response, using rendered resource's id by default or any method to fetch a value from the given hash variable.

collection :@posts

lookup(:comments_count, :@comments_count, field: :uuid, cast: false)
# => [{ "comments_count": 3 }, { "comments_count": 6 }]

In the example above, for each post it will fetch the value from @comments_count using the post's uuid as key. When the cast value is set to true (it is false by default), the value will be casted to a boolean using !!.

Custom nodes

Adds a new field to the response with block's result as value.

object :@user
node(:full_name) { |u| u.first_name + " " + u.last_name }
# => { "user" : { "full_name" : "John Doe" } }

You can add condition on your custom nodes. If the condition evaluates to a falsey value, the node will not added to the response at all.

node(:email, if: ->(u) { u.valid_email? }) do |u|
    u.email
end

Nodes are evaluated at rendering time, so you can use any instance variables or view helpers within them

node(:url) { |post| post_url(post) }

If the result of the block is a Hash, it can be directly merge into the response using merge instead of node

object :@user
merge { |u| { name: u.first_name + " " + u.last_name } }
# => { "user" : { "name" : "John Doe" } }

Extends & Partials

Often objects have a basic representation that is shared accross different views and enriched according to it. To avoid code redundancy you can extend your template from any other RABL template.

# app/views/shared/_user.rabl
attributes :id, :name

# app/views/users/show.rabl
object :@user

extends('shared/_user')
attributes :super_secret_attribute

#=> { "id": 1, "name": "John", "super_secret_attribute": "Doe" }

When used with child node, if they are the only thing added you can instead use the partial option directly.

child(:user, partial: 'shared/_user')

# is equivalent to

child(:user) do
  extends('shared/_user')
end

Extends can be used dynamically using rendered object and lambdas.

extends ->(user) { "shared/_#{user.client_type}_infos" }

Partials can also be used inside custom nodes. When using partial this way, you MUST declare the object associated to the partial

node(:location) do |user|
    { city: user.city, address: partial('users/address', object: m.address) }
end

When used this way, partials can take locals variables that can be accessed in the included template.

# _credit_card.rabl
node(:credit_card, if: ->(u) { locals[:display_credit_card] }) do |user|
  user.credit_card_info
end

# user.json.rabl
merge { |u| partial('_credit_card', object: u, locals: { display_credit_card: true }) }

Putting it all together

rabl-rails allows you to format your responses easily, from simple objects to hierarchy of 2 or 3 levels.

object :@thread

attribute :caption, as: :title

child(:@sorted_posts, as: :posts) do
  attributes :title, :slug

  child :comments do
        extends 'shared/_comment'
    lookup(:upvotes, :@upvotes_per_comment)
    end
end

Other features

And more in the WIKI

Performance

Benchmarks have been made using this application, with rabl 0.13.1 and rabl-rails 0.5.0

Overall, rabl-rails is 10% faster and use 10% less memory, but these numbers skyrockets to 50% when using extends with collection of objects.

You can see full tests on test application repository.

Authors and contributors

Want to add another format to Rabl-rails ? Checkout JSON renderer for reference Want to make another change ? Just fork and contribute, any help is very much appreciated. If you found a bug, you can report it via the Github issues.

Original idea

  • RABL Standart RABL gem. I used it a lot but I needed to improve my API response time, and since most of the time was spent in view rendering, I decided to implement a faster rabl gem.

Copyright © 2012-2020 Christopher Cocchi-Perrier. See MIT-LICENSE for details.