RABL (Ruby API Builder Language) is a Rails and Padrino ruby templating system for generating JSON, XML, MessagePack, PList and BSON. When using the ActiveRecord 'to_json' method, I find myself wanting a more expressive and powerful solution for generating APIs. This is especially true when the JSON representation is complex or doesn't match the exact schema defined within the database.
In particular, I want 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
- Render partial templates and inherit to reduce code duplication
- Rename or alias attributes to change the name from the model
- Append attributes from a child into a parent node
- Include nodes only if a certain condition has been 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 these problems by approaching API response generation in an entirely new way.
RABL at the core is all about adhering to MVC principles by deferring API data representations to the view layer of your application. For a breakdown of common misconceptions about RABL, please check out our guide to understanding RABL which can help clear up any confusion about this project.
-
v0.9.0 (released Oct 14, 2013) changes the default node name for certain associations especially around STI models. You might want to verify for any breakages as a result and be more explicit by specifying an alias i.e
@users => :users
-
v0.8.0 (released Feb 14, 2013) removes multi_json dependency and relies on Oj (or JSON) as the json parser. Simplifies code, removes a dependency but you might want to remove any references to MultiJson.
-
v0.6.14 (released June 28, 2012) requires the use of render_views with RSpec to test templates. Otherwise, the controller will simply pass through the render command as it does with ERB templates.
Install RABL as a gem:
gem install rabl
or add to your Gemfile:
# Gemfile
gem 'rabl'
# Also add either `oj` or `yajl-ruby` as the JSON parser
gem 'oj'
and run bundle install
to install the dependency.
If you are using Rails 2.3.8 (and up), Rails 3.X or Padrino, RABL works without configuration.
Important: With Padrino, be sure that the rabl gem is listed after the padrino gem in your Gemfile, otherwise Rabl will not register properly as a template engine.
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.
You can use RABL to generate JSON and XML based APIs from any ruby object. With RABL, the data typically is derived primarily from models (ORM-agnostic) and the representation of the API output is described within a view template using a simple ruby DSL. This allows you to keep your data separated 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 a basic overview but there is a lot more to see such as partials, inheritance, custom nodes, etc. Read the full details 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
require 'rabl'
Rabl.configure do |config|
# Commented as these are defaults
# config.cache_all_output = false
# config.cache_sources = Rails.env != 'development' # Defaults to false
# config.cache_engine = Rabl::CacheEngine.new # Defaults to Rails cache
# config.perform_caching = false
# config.escape_all_output = false
# config.json_engine = nil # Class with #dump class method (defaults JSON)
# config.msgpack_engine = nil # Defaults to ::MessagePack
# config.bson_engine = nil # Defaults to ::BSON
# config.plist_engine = nil # Defaults to ::Plist::Emit
# config.include_json_root = true
# config.include_msgpack_root = true
# config.include_bson_root = true
# config.include_plist_root = true
# config.include_xml_root = false
# config.include_child_root = true
# config.enable_json_callbacks = false
# config.xml_options = { :dasherize => true, :skip_types => false }
# config.view_paths = []
# config.raise_on_missing_attribute = true # Defaults to false
# config.replace_nil_values_with_empty_strings = true # Defaults to false
# config.replace_empty_string_values_with_nil_values = true # Defaults to false
# config.exclude_nil_values = true # Defaults to false
# config.exclude_empty_values_in_collections = true # Defaults to false
end
Each option specifies behavior related to RABL's output.
If include_json_root
is disabled that removes the root node for each root object in the output, and enable_json_callbacks
enables support for 'jsonp' style callback
output if the incoming request has a 'callback' parameter.
If include_child_root
is set to false then child objects in the response will not include
a root node by default. This allows you to further fine-tune your desired response structure.
If cache_engine
is set, you should assign it to a class with a fetch
method. See the default engine for an example.
If perform_caching
is set to true
then it will perform caching. You
can ignore this option if you are using Rails, it's same to Rails
config.action_controller.perform_caching
If cache_sources
is set to true
, template lookups will be cached for improved performance.
The cache can be reset manually by running Rabl.reset_source_cache!
within your application.
If cache_all_output
is set to true
, every template including each individual template used as part of a collection will be cached separately.
Additionally, anything within child, glue and partial will also be cached separately.
To cache just a single template, see the section titled 'Caching' below.
If escape_all_output
is set to true
and ActiveSupport is available, attribute output will be escaped using ERB::Util.html_escape.
If view_paths
is set to a path, this view path will be checked for every rabl template within your application.
Add to this path especially when including Rabl in an engine and using view paths within a another Rails app.
If raise_on_missing_attribute
is set to true
, a RuntimeError will be raised whenever Rabl
attempts to render an attribute that does not exist. Otherwise, the attribute will simply be omitted.
Setting this to true during development may help increase the robustness of your code, but using true
in
production code is not recommended.
If replace_nil_values_with_empty_strings
is set to true
, all values that are nil
and would normally be displayed as null
in the response are converted to empty strings.
If exclude_nil_values
is set to true
, all values that are nil
and would normally be displayed as null
in the response are not included in the response.
if exclude_empty_values_in_collections
is set to true
, all vaules in a collection that are {}
and would normally be displayed as {}
in the response are not included in the response.
If you wish to use oj as the primary JSON encoding engine simply add that to your Gemfile:
# Gemfile
gem 'oj'
and RABL will use that engine automatically for encoding your JSON responses.
Set your own custom json_engine which define a dump
or encode
method for converting to JSON from ruby data:
config.json_engine = ActiveSupport::JSON
RABL supports configuration for MessagePack, BSON, and Plist. Check the Format Configuration page for more details.
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 specify a root node label for the collection:
collection @users => :people
# => { "people" : [ { "person" : { ... } } ] }
or even specify both the child and root labels for a collection:
collection @users, :root => "people", :object_root => "user"
# => { "people" : [ { "user" : { ... } } ] }
and this will be used as the default data for the rendering, or disable the object root explicitly:
collection @users, :root => "people", :object_root => false
# => { "people" : [ { ... }, { ... } ] }
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> }
or show attributes only if a condition is true:
# m is the object being rendered, also supports :unless
attributes :foo, :bar, :if => lambda { |m| m.condition? }
Named and aliased attributes can not be combined on the same line. This currently does not work:
attributes :foo, :bar => :baz # throws exception
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 disable object root for child node:
child :posts, :object_root => false do
attributes :id, :title
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 pass in the current object:
object @user
child :posts do |user|
attribute :title unless user.suspended?
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.
You can also pass in the current object:
object @user
glue(@post) {|user| attribute :title if user.active? }
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 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 even 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.
You can see more examples on the Reusing Templates wiki page.
You can pass an arbitrary set of locals when rendering partials or extending templates.
For example, if we want to show on posts/:id.json
any information regarding particular post and associated comments
but in other cases we want to hide those comments. We can use locals to do this:
# app/views/posts/index.json.rabl
collection @posts
extends('posts/show', :locals => { :hide_comments => true })
# or using partial instead of extends
# node(false) { |post| partial('posts/show', :object => :post, :locals => { :hide_comments => true })}
and then access locals in the sub-template:
# app/views/posts/show.json.rabl
object @post
attributes :id, :title, :body, :created_at
node(:comments) { |post| post.comments } unless locals[:hide_comments]
This can be useful as an advanced tool when extending or rendering partials.
You can provide conditions to all kinds of nodes, attributes, extends, etc. which includes a given element only if the specified condition is true.
collection @posts
# m is the object being rendered, also supports :unless
node(:coolness, :if => lambda { |m| m.coolness > 5 }) do |m|
m.coolness
end
Because attributes take conditional options as well, we could simplify the example with:
collection @posts
# m is the object being rendered, also supports :unless
attribute(:coolness, :if => lambda { |m| m.coolness > 5 })
The value for the :if
and :unless
options may be a simple Boolean
, Proc
or a Symbol
. If it is a Symbol
and the specific @object
responds to its, the method will be called. Thus the example above can be rewritten as:
class Post
def cool?
coolness > 5
end
end
and then:
collection @posts
attribute :coolness, if: :cool?
Using conditions allows for easy control over when certain elements render.
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.
RABL has built-in caching support for templates leveraging fragment caching strategies. Note that caching is currently only available for Rails but support for other frameworks is planned in a future release. Simplest caching usage is:
# app/views/users/show.json.rabl
object @quiz
cache @quiz # key = rabl/quiz/[cache_key]
attribute :title
Caching can significantly speed up the rendering of RABL templates in production and is strongly recommended when possible. For more a more detailed look at caching, check out the Caching guide on the wiki.
There are situations where an application requires RABL templates to be rendered outside
a traditional view context. For instance, to render RABL within a Rake task or to create
message queue payloads. For this case, Rabl.render
can be used as show below:
Rabl.render(object, template, :view_path => 'app/views', :format => :json) #=> "{...json...}"
You can use convenience methods on Rabl::Renderer
to render the objects as well:
Rabl::Renderer.json(@post, 'posts/show')
Rabl::Renderer.xml(@post, 'posts/show')
These methods allow RABL to be used for arbitrary conversions of an object into a desired format.
Rabl::Renderer.new('posts/show', @post, :view_path => 'app/views', :format => 'hash').render
You can also pass in other instance variables to be used in your template as:
Rabl::Renderer.new('posts/show', @post, :locals => { :custom_title => "Hello world!" })
Then, in your template, you can use @custom_title
as:
attribute :content
node(:title) { @custom_title }
Currently in RABL, the content-type of your response is not set automatically. This is because RABL is intended to work for any Rack-based framework and as agnostic to format as possible. Check this issue for more details, and if you have any ideas or patches please let me know.
In the meantime, be sure to set the proper content-types if needed. This is usually pretty simple in both Rails and Padrino. I recommend a before_filter on that controller or directly specified in an action.
There are many resources available relating to RABL including the RABL Wiki, and many tutorials and guides detailed below. You can check out the RABL Site as well.
Links to resources for advanced usage:
- Managing Complexity
- Production Optimizations
- Grape Integration
- Rendering JSON for a tree structure using RABL
- Layouts (erb, haml and rabl) in RABL
- Backbone or Ember.js Integration
- RABL with Rails Engines
Please add your own usages and let me know so we can add them here! Also be sure to check out the RABL Wiki for other usages.
Tutorials can always be helpful when first getting started:
- Railscasts #322 - Ryan Bates explains RABL
- BackboneRails - Great screencasts by Brian Mann
- Creating an API with RABL and Padrino
- 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/
- http://blog.lawrencenorton.com/better-json-requests-with-rabl
- http://www.rodrigoalvesvieira.com/developing-json-api-rails-rabl/
- http://tech.favoritemedium.com/2011/06/using-rabl-in-rails-json-web-api.html
- http://seesparkbox.com/foundry/better_rails_apis_with_rabl
- http://blog.dcxn.com/2011/06/22/rails-json-templates-through-rabl
- http://teohm.github.com/blog/2011/05/31/using-rabl-in-rails-json-web-api
Let me know if there's any other useful resources not listed here.
There are other libraries that can either complement or extend the functionality of RABL:
- versioncake - Excellent library for easily versioning your RABL APIs
- gon - Exposes your Rails variables in JS with RABL support integrated.
- rabl-rails - Reimplementation for RABL and Rails focused on speed.
Let me know if there's any other related libraries not listed here.
See the examples directory.
Check out the Issues tab for a full list:
- Rigorous benchmarking and performance optimizations
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 was a great name chosen by the Miso CTO.
- David Sommers - Template resolution, caching support, and much more
- Rick Thomas - Added options for extends and Sinatra testing
- Benjamin Yu - Added msgpack format support
- 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
- Andrey Voronkov - Added BSON format support
- Alli Witheford - Added Plist format support
- Ryan Bigg - Improved template resolution code
- Ivan Vanderbyl - Added general purpose renderer
- Cyril Mougel - Added cache_engine pluggable support and renderer tweaks
- Teng Siong Ong - Improved renderer interface
- Brad Dunbar - Pass current object into blocks
and many more contributors listed in the CHANGELOG.
Want to contribute support for another format? Check out the patches for msgpack support, plist support and BSON support for reference.
Please fork and contribute, any help in making this project better is appreciated!
This project is a member of the OSS Manifesto.
There are a few excellent libraries that helped inspire RABL and they are listed below:
Thanks again for all of these great projects.
Copyright © 2011-2013 Nathan Esquenazi. See MIT-LICENSE for details.