Girders Blog
Notes on building internet applications

Building Rails API's with JSONAPI and JSONAPI-Resources

Aug 26, 2016

Sooner or later, every application needs an API. It will need to integrate with your microservices, peer client service, or your new front-end mobile and web apps. With JSONAPI, and JSONAPI-Resources, building your API platform is easy. You just add your own application magic. Along with one of these JSONAPI Client Libraries, your application can rule the world.

There is a lot of background that I won’t be covering. I’ll assume you are familiar with HTTP, REST, JSON, Rails Routing, and have some basic Rails experience. I’m writing the overview I wish I had when I started. Not all examples are complete or polished, but should be far enough along so you can “load” the problem into your head when starting work on your API.

Install & Setup

Create your rails application if don’t have one. If your new app will be API-Only (no HTML pages rendered), use the --api option to pare down the libraries Rails won’t need.

Add the jsonapi-resources gem to your Gemfile and run the bundle install command to pull it in. I’ll refer to jsonapi-resources as JR from now on.

What is a resource? Usually, we think of resources as a term used in REST web apps matching the data model or underlying database table. This isn’t always the case. They can also refer to a service that performs and action or return data that is a composite view of your data. Each resource has its own URL or “endpoint” which responds to requests on that resource.

JR provides two rails generators to build out your resource. It saves a little typing, but doesn’t provide anything special. You’ll need one of each to put the resource online.

In your routes file, use the jsonapi_resources :resource_name helper to build routes for the resource.

JSONAPI Requests

Now we have the tools we need, so next we need to understand how JSONAPI works. We need to first plan out how our API URL’s will look like. Let’s break this down.

Dreaming in URL’s

We’ll put the API URL’s into its own namespace. This lets our web server identify API requests to perform special processing, like sending them to a separate application instance or rate limiting.

Also, we’ll need to version our API to designate breaking changes. JSONAPI guarantees that future changes must be backwards compatible with existing services, but our application may need to evolve beyond what we do now. Conventionally, tags like “v1”, “v2”, etc. are used to identify the version. There are multiple methods of versioning, such as adding it to your Content-Type or other header, but putting it in the URL is the simplest for our application structure. Each version becomes a namespace on top of the API namespace.

https://example.com/api/v1/

Each resource name is an endpoint in this namespace. Resources should not be nested in JSONAPI, and instead are more like Rails’ shallow nesting. The URL endpoints will look like this:

https://example.com/api/v1/posts
https://example.com/api/v1/posts/123
https://example.com/api/v1/posts/123/relationships/comments
https://example.com/api/v1/comments/456

Fetch (GET) Requests

The JSONAPI standard provides query string parameters for filtering, pagination, sorting, sparse fieldsets (limiting attributes returned), and including related resources (parent, children, or related objects).

A nearly complete example request using these features follows. I’ll format across multiple lines for easy reading, but the URL and query string consists of a single string sent via HTTP. Also, I won’t be percent encoding these URL examples as required.

GET https://example.com/api/v1/posts
    ?page[number]=2
    &page[size]=50
    &sort=author,-created
    &fields[posts]=id,title,created
    &include=comments,author
    &filter[posts]=1,2
    &filter[comments]=1,2
    &filter[status]=posted

The specification doesn’t cover the details of pagination and filtering, leaving that decision up to the library implementation and application, but does provide these examples on how these are expected to be used. This usage reflects the JR basic implementation.

Check out the full details on the JSONAPI Specification page.

JSONAPI POST and PATCH Request

POST (create) and PATCH (update) requests require a payload body of JSON in the JSONAPI format. This is a JSON object with key:

A resource object is a JSON object (hash) with the keys:

    {
      "data": {
        "type": "photos",
        "attributes": {
          "title": "Ember Hamster",
          "src": "http://example.com/images/productivity.png"
        },
        "relationships": {
          "photographer": {
            "data": { "type": "people", "id": "9" }
          }
        }
      }
    }

JSONAPI Delete Request

DELETE is the simplest request. Give it a record endpoint with an id.

DELETE /api/v1/posts/1

I can’t find any more information on deleting all records or using a filter parameter to the delete request to delete a list of records. That could be nice if you can do something like:

DELETE /api/v1/posts?filter[author]=1,2,3 /* Not real */

JSONAPI HTTP Request

JSONAPI has its own MIME Type that must be sent on the Accept and Content-Type (when sending data) HTTP Headers. The HTTP Request could look like this:

GET /api/v1/posts?page[number]=2 HTTP/1.1
Accept: application/vnd.api+json

POST /api/v1/posts HTTP/1.1
Accept: application/vnd.api+json
Content-Type: application/vnd.api+json

{"data":{"type":"posts","attributes":{...}}}

JSONAPI Response

The HTTP Response is a JSON document returned with the JSONAPI Mime Type. The HTTP Status Code is also important, and can also be found in the “errors” section when an error occurred.

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{"data":{"type":"posts","attributes":{...}}}

The JSONAPI Response data structure has the following structure:

links: {self: url, next: url, last: url, related: url, ...}
data: [
  { type: name,            /* Resource Object */
    id:123,
    attributes: {name: value,...},
    relationships: {
      name: {links: {}, data: [
        {type: name, id:234}, ... /* Resource Identifier Object */
      ],
  }, ...
],
included: [
  { type: name,            /* Resource Object */
    id:123,
    attributes: {name: value,...},
    relationships: {},
    links: {self:url, ...}
  }, ...
],
errors: [
  { id: uniqueIdentifier,        /* Error Object */
    links: {about: url},
    status: http_status_code,
    code: application_error_code,
    title: "Short human readable, no details",
    detail: "Human readable with details"
    source: {
      pointer: "/data/attributes/title",
      parameter: which URI param caused error},
    meta: { ... non-standard response details here ... }
  }, ...
],
meta: { ... non-standard response details here ... }
jsonapi: { ... server implemtation deatils ... }

Notes:

Building the API

Now we know how to speak and read JSONAPI, so the implementation details can make sense as we implement our API. As it turns out, our app is the next hottest blogging platform, so our API will use the objects and relationships you already know! Let’s get started with posts and comments. I’ll create them both at once for clarity, but I know you’ll build them properly on your own.

rails g jsonapi:resource Api::V1::Post
rails g jsonapi:controller Api::V1::Post
rails g jsonapi:resource Api::V1::Comment
rails g jsonapi:controller Api::V1::Comment

Wait, what’s that “Api::V1” for? That’s the namespace we discussed earlier. Your JR Resource and Controller have to be in the same namespace. Now we’ll add them to our config/routes.rb file:

namespace :api do
  namespace :v1 do
    jsonapi_resources :posts
    jsonapi_resources :comments
  end
end

Our JR Resources inherit from JSONAPI::Resource and JSONAPI::ResourceController, but I like to share application-specific methods like we do with ApplicationController, ApplicationHelper, and ApplicationRecord, so let’s create our own class to inherit from, and change the generated files to inherit from ours. We’ll also stick the application level classes in the versioned API namespace because they may change by version.

    # app/controllers/api/v1/api_resource_controller.rb
    class Api::V1::ApiResourceController < JSONAPI::ResourceController
    end

    # app/resources/api/v1/application_resource.rb
    class Api::V1::ApplicationResource < JSONAPI::Resource
      # An optional Meta object added to every request
      def meta(options)
        { copyright: #{Time.now.year} WAAS - WeblogAsAService, Inc." }
      end
    end

    # app/resources/api/v1/post_resource.rb
    class Api::V1::PostResource < Api::V1::ApplicationResource
    end

    # app/controllers/api/v1/posts_controller.rb
    class Api::V1::PostsController < Api::V1::ApiResourceController
    end

Configure JSONAPI

I’ll create a basic configuration inside the JR initializer file.

    # config/initializers/jsonapi_resources.rb:
    JSONAPI.configure do |config|
      # built in paginators are :none, :offset, :paged
      config.default_paginator = :paged
      config.default_page_size = 50
      config.maximum_page_size = 1000

      # Do this if you use UUID's instead of Integers for id's
      config.resource_key_type = :uuid
    end

Configuring our Resources

Your resource magically finds your model when you are using the same name. Otherwise, there are a few configurations you can use to be more explicit. JR’s DSL is straightforward.

Here’s a simple version of our resources:

    # app/resources/api/v1/post_resource.rb
    class Api::V1::PostResource < Api::V1::ApplicationResource
      attributes :title, :body, :status
      has_many   :comments
      filters    :id, :title
      filter     :status, default: "published,pending"
    end

    # app/resources/api/v1/comment_resource.rb
    class Api::V1::CommentResource < Api::V1::ApplicationResource
      attributes :created_at, :body, :author
      has_one    :post
      filters    :id, :author
    end

A full read of the JSONAPI Resources README will show you all the customization you can do. You will have a lot of requirements as you proceed and can find most of your answers there.

Authentication

We’ll need authentication on our API. Authentication is a big subject, but I’ll just touch on the relevant parts for this use case. There are several authorization techniques, but API’s usually use an “API Key” for access, usually a 32-character hexadecimal string, probably from a randomly generated UUID, like SecureRandom.uuid.gsub("-",""). We’ll add an api_key column of type UUID to the users table and assign the key at create time. (Note: This is not secure, it is the same as putting clear-text passwords in your database, but I know you know better. Check out devise_token_auth.) Token-based access like this needs the HTTP Authorization header such as:

Authorization: Token token="2253b04477254110b3ea30997b71a38a"

If your app is servicing a Javascript front-end app like Ember, React, Angular, etc., you may also want to handle HTTP Basic Authentication as a login method. It could return a session token to you to use until it times out or logs out. Check out the best solution for your platform.

You can setup a before_action hook to check the header. Something like this (hand-wavy code):

    # app/controllers/api/v1/api_resource_controller.rb
    before_action :authorize!

    def authorize!
      hdr = request.headers["Authorization"]
      if hdr && hdr =~ /\AToken\s+(token="?)?(.+?)"?\s*\z/
        return true if valid_apikey?($2)
      end
      render(status: :unauthorized, json:{errors:[{
        status:401, code:"unauthorized", title:"Unauthorized"
      }]})
    end

    def valid_apikey?(key)
      @user = User.find_by(api_key:key)
      !!@user # Make boolean
    end

Context

Every request carries some context, such as who is making the request. We’ll usually want to restrict records to those the current user (identified by an Authentication header) owns. The controller’s context() method returns a Hash of key/value pairs that could be used to limit the returned result to just the user’s data.

At the start of the request, the controller identifies the user and sets up the context hash as follows.

    # api_resource_controller.rb
    def context
      # @user assigned in #authorize!()
      { api_user: @user }
    end

Now your Resource can scope the records returned. The records() class method should return an Arel query object. Individual record finds are chained off this object, so it will find the record by key only if the user owns it. The context hash is passed to this method.

    # app/resources/api/v1/post_resource.rb
    def self.records(options={})
      options[:context][:api_user].posts
    end

Rate Limiting

If you need to limit the frequency requests to your service, you should check out rack-throttle.

Webhook and Non-REST-ful Requests

Depending on your use case, standard REST requests can’t always be sent. I’ve worked with systems that can only send GET requests, so all payload information and authentication must appear in the query string.

If your application needs to respond to these Webhooks requests that are used by a limited client that doesn’t understand REST, you need to build a non-REST-ful interface.

For that case, I create a special endpoint and set up a controller that maps the get request into a processing request. Here, I accept the request and send it off into an ActiveJob so we can quickly return for the client.

    # config/routes: setup as "resources :webhook" under api/v1 namespace
    # /api/v1/webhook?apikey=foo&bar=baz&...
    # app/controllers/api/v1/webhook_controller.rb
    class Api::V1::WebhookController < Api::V1::ApiResourceController
      def index;  webhook; end
      def create; webhook; end
      def show;   webhook; end
      def update; webhook; end
      def delete; webhook; end

      private
      def webhook
        WebhookJob.perform_later(params.to_json)
        render(status: :accepted, plain:"OK")
      end
    end

Try it out

Let’s make REST Requests using your favorite REST posting tool. (I use Postman.) Remember to set your Accept and Content-Type headers! I’ll let you play around with this as working through the details is that will really help you learn JSONAPI. Start your app up locally and connect to localhost.

POST http://localhost:3000/api/v1/posts
Accept: application/vnd.api+json
Content-Type: application/vnd.api+json
Authorization: Token token="2253b04477254110b3ea30997b71a38a"

{"data":{"type":"posts","attributes":{...}}}

Fetch some data:

GET http://localhost:3000/api/v1/comments?filter[author]=pat
Accept: application/vnd.api+json
Authorization: Token token="2253b04477254110b3ea30997b71a38a"

Notes:

Using a JSONAPI Client

Check out the list of JSONAPI Client Libraries, as a starting point for your client. As I’ve shown, once you know how JSONAPI works, the client isn’t so hard, but a well-tested library will help you integrate your API into your client.

Summary

So that’s about it. Now we have demystified JSONAPI, and understand how to talk to it. Building API services that speak it is straightforward, and so is consuming the results. JASONAPI is a fairly new protocol, but has been methodically designed to handle most use cases. JR was developed along side of it, so shows the best of Ruby and Rail’s strengths and a great data interchange format. Good luck!