RABL (Ruby API Builder Language) is a Rails and Padrino ruby templating system for generating JSON and XML. When using the ActiveRecord 'to_json' method, I tend to quickly find myself wanting a more expressive and powerful system for generating APIs. This is especially frustrating when the json representation is complex or doesn't match the exact schema defined in the database itself.
I wanted a simple, and flexible system for generating APIs. In particular, I wanted to easily:
- Create arbitrary nodes named based on combining data in an object
- Pass arguments to methods and store the result as a child node
- Partial templates and inheritance to reduce code duplication
- Easily renaming attributes from their name in the model
- Simple way to append attributes from a child into the parent
- Include nodes only if a certain condition is met
Anyone who has tried the 'to_json' method used in ActiveRecord for generating a json response has felt the pain of this restrictive approach. RABL is a general templating system created to solve all of those problems.
Install RABL as a gem:
gem install rabl
or add to your Gemfile:
# Gemfile
gem 'rabl'
# Also add either `json` or `yajl-ruby` as the JSON parser
gem 'yajl-ruby'
and run bundle install
to install the dependency.
If you are using Rails 2.X, Rails 3, Rails 3.X or Padrino, RABL works without configuration.
With Sinatra, or any other tilt-based framework, simply register:
Rabl.register!
and RABL will be initialized and ready for use. For usage with Sinatra, check out the Sinatra Usage guide.
The quick idea here is that you can use RABL to generate JSON and XML API based on any arbitrary data source. With RABL, the data is expected to come primarily from a model (ORM-agnostic) and the representation of the API output is described in the view with a simple ruby DSL. This allows you to keep your data separate from the JSON or XML you wish to output.
Once you have installed RABL (explained above), you can construct a RABL view template and then render the template from your Sinatra, Padrino or Rails applications from the controller (or route) very easily. Using Padrino as an example, assuming you have a Post
model filled with blog posts, you can render an API representation (both JSON and XML) by creating a route:
# app/app.rb
get "/posts", :provides => [:json, :xml] do
@user = current_user
@posts = Post.order("id DESC")
render "posts/index"
end
Then we can create the following RABL template to express the API output of @posts
:
# app/views/posts/index.rabl
collection @posts
attributes :id, :title, :subject
child(:user) { attributes :full_name }
node(:read) { |post| post.read_by?(@user) }
Which would output the following JSON or XML when visiting http://localhost:3000/posts.json
[{ post :
{
id : 5, title: "...", subject: "...",
user : { full_name : "..." },
read : true
}
}]
That's the basic overview but there is a lot more (partials, inheritance, custom nodes, etc). Read the full details below of RABL below.
RABL is intended to require little to no configuration to get working. This is the case in most scenarios, but depending on your needs you may want to set the following global configurations in your application (this block is completely optional):
# config/initializers/rabl_init.rb
Rabl.configure do |config|
# Commented as these are the defaults
# config.json_engine = nil # Any multi\_json engines
# config.msgpack_engine = nil # Defaults to ::MessagePack
# config.include_json_root = true
# config.include_msgpack_root = true
# config.include_xml_root = false
# config.enable_json_callbacks = false
# config.xml_options = { :dasherize => true, :skip_types => false }
end
Each option specifies behavior related to RABL's output. If include_json_root
is disabled that removes the root node for each child in the output, and enable_json_callbacks
enables support for 'jsonp' style callback output if the incoming request has a 'callback' parameter.
Note that the json_engine
option uses the multi_json intelligent engine defaults so in most cases you don't need to configure this directly. If you wish to use yajl as the primary JSON encoding engine simply add that to your Gemfile:
# Gemfile
gem 'yajl-ruby', :require => "yajl"
and RABL will automatically start using that engine for encoding your JSON responses!
Rabl also includes optional support for Message Pack serialization format using the msgpack gem.
To enable, include the msgpack gem in your project's Gemfile. Then use Rabl as normal with the msgpack
format (akin to json and xml formats).
# Gemfile
gem 'msgpack', '~> 0.4.5'
One can additionally use a custom Message Pack implementation by setting the Rabl msgpack_engine
configuration attribute. This custom message pack engine must conform to the MessagePack#pack method signature.
class CustomEncodeEngine
def self.pack string
# Custom Encoding by your own engine.
end
end
Rabl.configure do |config|
config.msgpack_engine = CustomEncodeEngine
end
NOTE: Attempting to render the msgpack format without either including the msgpack gem or setting a msgpack_engine
will cause an exception to be raised.
To declare the data object for use in the template:
# app/views/users/show.json.rabl
object @user
or specify an alias for the object:
object @user => :person
# => { "person" : { ... } }
or pass a collection of objects:
collection @users
# => [ { "user" : { ... } } ]
or even specify a root node label for the collection:
collection @users => :people
# => { "people" : [ { "person" : { ... } } ] }
and this will be used as the default data for the rendering.
There can also be odd cases where the root-level of the response doesn't map directly to any object:
object false
node(:some_count) { |m| @user.posts.count }
child(@user) { attribute :name }
In those cases, object can be assigned to 'false' and nodes can be constructed free-form.
Basic usage of the templater to define a few simple attributes for the response:
# app/views/users/show.json.rabl
attributes :id, :foo, :bar
or use with aliased attributes:
# Take the value of model attribute `foo` and name the node `bar`
attribute :foo => :bar
# => { bar : 5 }
or even multiple aliased attributes:
attributes :bar => :baz, :dog => :animal
# => # { baz : <bar value>, animal : <dog value> }
Often a response requires including nested information from data associated with the parent model:
child :address do
attributes :street, :city, :zip, :state
end
You can also add child nodes from an arbitrary data source:
child @posts => :foobar do
attributes :id, :title
end
or use model associations with an alias:
# Renders all the 'posts' association
# from the model into a node called 'foobar'
child :posts => :foobar do
attributes :id, :title
end
You can also append child attributes back to the root node:
# Appends post_id and post_name to parent json object
glue @post do
attributes :id => :post_id, :name => :post_name
end
Use glue to add additional attributes to the parent object.
This will generate a json response based on the result of the node
block:
# app/views/users/show.json.rabl
node :full_name do |u|
u.first_name + " " + u.last_name
end
or a custom node that exists only if a condition is true:
# m is the object being rendered, also supports :unless
node(:foo, :if => lambda { |m| m.has_foo? }) do |m|
m.foo
end
or don't pass a name and have the node block merged into the response:
node do |u|
{ :full_name => u.first_name + " " + u.last_name }
# => { full_name : "Bob Johnson" }
end
You can use custom nodes like these to create flexible representations of a value utilizing all the data from the model.
Often you need to access other data objects in order to construct custom nodes in more complex associations. You can get access to the rabl representation of another data object by rendering a RABL partial:
node :location do
{ :city => @city, :address => partial("users/address", :object => @address) }
end
or event access an object associated with the parent model:
node :location do |m|
{ :city => m.city, :address => partial("users/address", :object => m.address) }
end
You can use this method to construct arbitrarily complex nodes for your APIs. Note that you need to have RABL templates defined for each of the objects you wish to construct representations for in this manner.
Another common issue of many template builders is unnecessary code redundancy. Typically many representations of an object across multiple endpoints share common attributes or nodes. The nodes for a 'post' object are probably the same or similar in most references throughout the various endpoints.
RABL has the ability to extend other "base" rabl templates and additional attributes:
# app/views/users/advanced.json.rabl
extends "users/base" # another RABL template in "app/views/users/base.json.rabl"
node :can_drink do |m|
m.age > 21
end
You can also extend other rabl templates while constructing child nodes to reduce duplication:
# app/views/users/show.json.rabl
child @address do
extends "address/item"
end
Using partials and inheritance can significantly reduce code duplication in your templates.
In RABL, you have access to everything you need to build an API response. Each RABL template has full access to the controllers instance variables as well as all view helpers and routing urls.
# app/some/template.rabl
object @post
# Access instance variables
child(@user => :user) { ... }
# or Rails helpers
node(:formatted_body) { |post| simple_format(post.body) }
There should be no problem fetching the appropriate data to construct a response.
In APIs, you can often need to construct 2nd or 3rd level nodes. Let's suppose we have a 'quiz' model that has many 'questions' and then each question has many 'answers'. We can display this hierarchy in RABL quite easily:
# app/views/quizzes/show.json.rabl
object @quiz
attribute :title
child :questions do
attribute :caption
child :answers do
# Use inheritance to reduce duplication
extends "answers/item"
end
end
This will display the quiz object with nested questions and answers as you would expect with a quiz node, and embedded questions and answers. Note that RABL can be nested arbitrarily deep within child nodes to allow for these representations to be defined.
Links to resources for advanced usage:
- Rendering JSON for a tree structure using RABL: nesquena#70
- Layouts (erb, haml and rabl) in RABL: https://github.com/nesquena/rabl/wiki/Using-Layouts
- Backbone or Ember.js Integration: https://github.com/nesquena/rabl/wiki/Backbone-Integration
Please add your own usages and let me know so I can add them here!
Tutorials can always be helped when getting started:
- http://blog.joshsoftware.com/2011/12/23/designing-rails-api-using-rabl-and-devise/
- http://engineering.gomiso.com/2011/06/27/building-a-platform-api-on-rails/
Let me know if there's any other useful resources.
- Redundant calls for a collection: nesquena#142 (comment)
Check out the Issues tab for a full list:
- Better Tilt template support (precompiling templates)
- Benchmarks and performance optimizations
CI is hosted by travis-ci.org.
Thanks to Miso for allowing me to create this for our applications and release this project!
- Nathan Esquenazi - Creator of the project
- Arthur Chiu - Core Maintainer, Riot Testing Guru
- Tim Lee - RABL is an awesome name and was chosen by the Miso CTO.
- Rick Thomas - Added options for extends and Sinatra testing
- Chris Kimpton - Helping with documentation and wiki
- Marjun - Added xml option configurations
- Anton Orel - Added Rails 3.1 compatibility
- Sasha Koss - Added multi_json support
- Matthew Schulkind - Cleanup of configuration and tests
- Luke van der Hoeven - Support non-ORM objects in templates
and many more contributors listed in the CHANGELOG. Please fork and contribute, any help in making this project better is appreciated!
There are a few excellent libraries that helped inspire RABL and they are listed below:
Thanks again for all of these great projects.
See the examples directory.
Copyright © 2011 Nathan Esquenazi. See MIT-LICENSE for details.