TRB JOBS: Deutschsprachiger Ruby Developer in Hannover - Jaaaa!

Roar

JSON API

Last updated 21 January 2017

Roar JSON API supports rendering and parsing documents using the JSON API 1.0 specification.

N.B. Roar’s JSON API support is now provided as a separate gem.

Overview

Here’s a basic example of a JSON API representer.

class ArticleDecorator < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :articles

  # top-level link.
  link :self, toplevel: true do
    "//articles"
  end

  attributes do
    property :title
  end

  # resource object links
  link(:self) { "http://#{represented.class}/#{represented.id}" }

  # relationships
  has_one :author, class: Author, populator: ::Representable::FindOrInstantiate do # populator is for parsing, only.
    type :authors

    attributes do
      property :email
    end

    link(:self) { "http://authors/#{represented.id}" }
  end

  has_many :comments, class: Comment, decorator: CommentDecorator
end

Basic Usage

By including Roar::JSON::JSONAPI.resource into your representer a handful of additional DSL methods get imported.

As JSON API per definition can represent singular models and collections you have two entry points.

SongsRepresenter.new(Song.find(1)).to_json
SongsRepresenter.new(Song.new).from_json("..")
SongsRepresenter.for_collection.new([Song.find(1), Song.find(2)]).to_json
SongsRepresenter.for_collection.new([Song.new, Song.new]).from_json("..")

Resource Objects

→ specification

JSON API Resource Objects must contain an id and type member. You specify the type when you include Roar::JSON::JSONAPI.resource(type):

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  # ...
end

You do not need to specify an id. An id property will be created for you automatically. However, if your represented object uses a method other than id to represent its id, you must specify this with the id_key: option on inclusion:

class Song
  attr_accessor :song_id
end

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs, id_key: :song_id

  # ...
end

Attributes

→ specification

Attributes should be defined with ::property in an ::attributes block.

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  attributes do
    property :title
  end
end

Relationships

→ specification

To define relationships, use ::has_one or ::has_many with either an inline representer or an explicit, standalone representer (specified with the decorates: or extend: option).

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  has_one :album, class: Album do
    property :title
  end

  has_many :musicians, class: Musician, decorator: MusiciansRepresenter
end

You are able to define links and meta information for a given relationship in a ::relationship block:

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  has_many :musicians, class: Musician, decorator: MusiciansRepresenter do
    relationship do
      link(:self)     { "/songs/#{represented.id}/relationships/musicians" }
      link(:related)  { "/songs/#{represented.id}/musicians" }
    end
  end
end

Member Names

→ specification

By default, member names will be rendered according to JSON API recommendations: only non-reserved, URL safe characters specified in RFC 3986 will be used. Following JSON API conventions, underscores will also be replaced by hyphens.

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  attributes do
    property :song_title        # rendered as song-title
    property :lyric_writer, as: :lyricist
  end

  has_one :album, class: Album do
    property :release_date    # rendered as release-date
  end
end

If you want less-strict behaviour (such as allowing non-ASCII Unicode characters), you can override this default:

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  defaults do |name, _|
    { as: JSONAPI::MemberName.(name, strict: false) }
  end

  attributes do
    property :titel_des_Liedes    # rendered as titel_des_Liedes
    property :Klavierstück        # rendered as Klavierstück
  end
end

Meta information

→ specification

Meta information can be included into rendered singular and collection documents in two ways.

You can define meta information on your collection object and then let Roar compile it.

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  meta toplevel: true do
    property :page
    property :total
  end
end

Your collection object must expose the respective methods.

collection.page  #=> 1
collection.total #=> 12

This will render the {"meta": {"page": 1, "total": 12}} hash into the JSON API document.

Alternatively, you can provide meta information as a hash when rendering. Any values also defined on your object will be overriden.

collection.to_json(meta: {page: params["page"], total: collection.size})

Both methods work for singular documents too.

class SongsRepresenter < Roar::Decorator
  include Roar::JSON::JSONAPI.resource :songs

  meta do
    property :label
    property :format
  end
end
song.to_json(meta: { label: 'EMI' })

Compound Document

You can suppress rendering of the compound document using the :included option.

decorator.to_json(included: false)

Sparse Fieldsets

As per specification, JSON API allows the rendering of arbitrary fields to be suppressed. This feature is called sparse fieldsets.

With Roar, you can do that with all kinds of resource objects: Filtering attributes works for the top “primary data” and for the compound object.

For the primary data, you simply provide what to include via the :fields option.

decorator.to_hash(
  fields: { articles: [:title] }
)

This will only render the two mandatory :id, :type members and the :title attribute.

"data": {
  "type": "articles",
  "id": "1",
  "attributes": {"title": "My Article"}
}

To include particular fields, only, in the compound document, use :fields.

decorator.to_hash(
  include: [:author],
  fields: { articles: [:title], author: [:email] }
)

You can control what types should be included in the compound document by adding them to :include.

The :fields option allows to specify what attributes for what type to include.

"data": {
  "type": "articles",
  "id": "1",
  "attributes": {"title": "My Article"}
},
"included": [
  {
    "type": "author",
    "id":   "a:1",
    "attributes": {
      "email": "celsito@trb.to"
    }
  }
]

Note that the author fragment only contains the email in its attributes.

The to_json API is designed to be able to process params directly, where JSON API parameters like fields from the request URL are parsed into hashes and arrays.

uri = Addressable::URI.parse('/articles?include=author&fields%5Barticles%5D=title,body&fields%5Bpeople%5D=name')
=> {"include"=>"author", "fields[articles]"=>"title,body", "fields[people]"=>"name"}

query = Rack::Utils.parse_nested_query(uri.query)
=> {"include"=>"author", "fields"=>{"articles"=>"title,body", "people"=>"name"}}

representer.to_json(include: query['include'], fields: query['fields'])