Him

Him is an ORM (Object Relational Mapper) that maps REST resources to Ruby objects. It is designed to build applications that are powered by a RESTful API instead of a database.

The spiritual successor of Her, modernized for current Ruby/Rails.

Requirements:

Ruby >= 3.1
ActiveModel >= 6.1
Faraday >= 2.0

Installation

In your Gemfile, add:

gem "him"

That’s it! Him provides a Her = Him alias, so code using Her::Model will continue to work.

Usage

First, you have to define which API your models will be bound to. For example, with Rails, you would create a new config/initializers/him.rb file with these lines:

# config/initializers/him.rb
Him::API.setup url: "https://api.example.com" do |c|
  # Request
  c.use Faraday::Request::UrlEncoded

  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

And then to add the ORM behavior to a class, you just have to include Him::Model in it:

class User
  include Him::Model
end

After that, using Him is very similar to many ActiveRecord-like ORMs:

User.all
# GET "https://api.example.com/users" and return an array of User objects

User.find(1)
# GET "https://api.example.com/users/1" and return a User object

@user = User.create(fullname: "Tobias Fünke")
# POST "https://api.example.com/users" with `fullname=Tobias+Fünke` and return the saved User object

@user = User.new(fullname: "Tobias Fünke")
@user.occupation = "actor"
@user.save
# POST "https://api.example.com/users" with `fullname=Tobias+Fünke&occupation=actor` and return the saved User object

@user = User.find(1)
@user.fullname = "Lindsay Fünke"
@user.save
# PUT "https://api.example.com/users/1" with `fullname=Lindsay+Fünke` and return the updated User object

ActiveRecord-like methods

These are the basic ActiveRecord-like methods you can use with your models:

class User
  include Him::Model
end

# Update a fetched resource
user = User.find(1)
user.fullname = "Lindsay Fünke" # OR user.assign_attributes(fullname: "Lindsay Fünke")
user.save # returns false if it fails, errors in user.response_errors array
# PUT "/users/1" with `fullname=Lindsay+Fünke`

# Update a resource without fetching it
User.save_existing(1, fullname: "Lindsay Fünke")
# PUT "/users/1" with `fullname=Lindsay+Fünke`

# Destroy a fetched resource
user = User.find(1)
user.destroy
# DELETE "/users/1"

# Destroy a resource without fetching it
User.destroy_existing(1)
# DELETE "/users/1"

# Fetching a collection of resources
User.all
# GET "/users"
User.where(moderator: 1).all
# GET "/users?moderator=1"

# Create a new resource
User.create(fullname: "Maeby Fünke")
# POST "/users" with `fullname=Maeby+Fünke`

# Save a new resource
user = User.new(fullname: "Maeby Fünke")
user.save! # raises Him::Errors::ResourceInvalid if it fails
# POST "/users" with `fullname=Maeby+Fünke`

Middleware

Since Him relies on Faraday to send HTTP requests, you can choose the middleware used to handle requests and responses. Using the block in the setup call, you have access to Faraday’s connection object and are able to customize the middleware stack used on each request and response.

Authentication

Him doesn’t support authentication by default. However, it’s easy to implement one with request middleware. Using the setup block, we can add it to the middleware stack.

For example, to add a token header to your API requests in a Rails application, you could use the excellent request_store gem like this:

# app/controllers/application_controller.rb
class ApplicationController < ActionController::Base
  before_action :set_user_api_token

  protected
  def set_user_api_token
    RequestStore.store[:my_api_token] = current_user.api_token # or something similar based on `session`
  end
end

# lib/my_token_authentication.rb
class MyTokenAuthentication < Faraday::Middleware
  def call(env)
    env[:request_headers]["X-API-Token"] = RequestStore.store[:my_api_token]
    @app.call(env)
  end
end

# config/initializers/him.rb
require "lib/my_token_authentication"

Him::API.setup url: "https://api.example.com" do |c|
  # Request
  c.use MyTokenAuthentication
  c.use Faraday::Request::UrlEncoded

  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

Now, each HTTP request made by Him will have the X-API-Token header.

Basic HTTP Authentication

Him can use basic HTTP auth by adding a line to your initializer:

# config/initializers/him.rb
Him::API.setup url: "https://api.example.com" do |c|
  # Request
  c.request :authorization, :basic, 'myusername', 'mypassword'
  c.use Faraday::Request::UrlEncoded

  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

OAuth

Using the faraday_middleware and simple_oauth gems, it’s fairly easy to use OAuth authentication with Him.

In your Gemfile:

gem "him"
gem "faraday_middleware"
gem "simple_oauth"

In your Ruby code:

# Create an application on `https://dev.twitter.com/apps` to set these values
TWITTER_CREDENTIALS = {
  consumer_key: "",
  consumer_secret: "",
  token: "",
  token_secret: ""
}

Him::API.setup url: "https://api.twitter.com/1/" do |c|
  # Request
  c.use FaradayMiddleware::OAuth, TWITTER_CREDENTIALS

  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

class Tweet
  include Him::Model
end

@tweets = Tweet.get("/statuses/home_timeline.json")

See the Authentication middleware section for an example of how to pass different credentials based on the current user.

Parsing JSON data

By default, Him handles JSON data. It expects the resource/collection data to be returned at the first level.

// The response of GET /users/1
{ "id" : 1, "name" : "Tobias Fünke" }

// The response of GET /users
[{ "id" : 1, "name" : "Tobias Fünke" }]

However, if you want Him to be able to parse the data from a single root element (usually based on the model name), you’ll have to use the parse_root_in_json method (See the JSON attributes-wrapping section).

Also, you can define your own parsing method using a response middleware. The middleware should set env[:body] to a hash with three symbol keys: :data, :errors and :metadata. The following code uses a custom middleware to parse the JSON data:

# Expects responses like:
#
#     {
#       "result": { "id": 1, "name": "Tobias Fünke" },
#       "errors": []
#     }
#
class MyCustomParser < Faraday::Middleware
  def on_complete(env)
    json = JSON.parse(env[:body], symbolize_names: true)
    env[:body] = {
      data: json[:result],
      errors: json[:errors],
      metadata: json[:metadata]
    }
  end
end

Him::API.setup url: "https://api.example.com" do |c|
  # Response
  c.use MyCustomParser

  # Adapter
  c.adapter :net_http
end

Caching

Again, using the faraday_middleware and memcached gems makes it very easy to cache requests and responses.

In your Gemfile:

gem "him"
gem "faraday_middleware"
gem "memcached"

In your Ruby code:

Him::API.setup url: "https://api.example.com" do |c|
  # Request
  c.use FaradayMiddleware::Caching, Memcached::Rails.new('127.0.0.1:11211')

  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

class User
  include Him::Model
end

@user = User.find(1)
# GET "/users/1"

@user = User.find(1)
# This request will be fetched from memcached

Advanced Features

Here’s a list of several useful features available in Him.

Associations

Examples use this code:

class User
  include Him::Model
  has_many :comments
  has_one :role
  belongs_to :organization
end

class Comment
  include Him::Model
end

class Role
  include Him::Model
end

class Organization
  include Him::Model
end

Fetching data

You can define has_many, has_one and belongs_to associations in your models. The association data is handled in two different ways.

If Him finds association data when parsing a resource, that data will be used to create the associated model objects on the resource.
If no association data was included when parsing a resource, calling a method with the same name as the association will fetch the data (providing there’s an HTTP request available for it in the API).

For example, if there’s association data in the resource, no extra HTTP request is made when calling the #comments method and an array of resources is returned:

@user = User.find(1)
# GET "/users/1", response is:
# {
#   "id": 1,
#   "name": "George Michael Bluth",
#   "comments": [
#     { "id": 1, "text": "Foo" },
#     { "id": 2, "text": "Bar" }
#   ],
#   "role": { "id": 1, "name": "Admin" },
#   "organization": { "id": 2, "name": "Bluth Company" }
# }

@user.comments
# => [#<Comment id=1 text="Foo">, #<Comment id=2 text="Bar">]

@user.role
# => #<Role id=1 name="Admin">

@user.organization
# => #<Organization id=2 name="Bluth Company">

If there’s no association data in the resource, Him makes an HTTP request to retrieve the data.

@user = User.find(1)
# GET "/users/1", response is { "id": 1, "name": "George Michael Bluth", "organization_id": 2 }

# has_many association:
@user.comments
# GET "/users/1/comments"
# => [#<Comment id=1>, #<Comment id=2>]

@user.comments.where(approved: 1)
# GET "/users/1/comments?approved=1"
# => [#<Comment id=1>]

# has_one association:
@user.role
# GET "/users/1/role"
# => #<Role id=1>

# belongs_to association:
@user.organization
# (the organization id comes from :organization_id, by default)
# GET "/organizations/2"
# => #<Organization id=2>

Subsequent calls to #comments, #role and #organization will not trigger extra HTTP requests and will return the cached objects.

Creating data

You can use the association methods to build new objects and save them.

@user = User.find(1)
@user.comments.build(body: "Just a draft")
# => [#<Comment body="Just a draft" user_id=1>]

@user.comments.create(body: "Hello world.", user_id: 1)
# POST "/comments" with `body=Hello+world.&user_id=1`
# => [#<Comment id=3 body="Hello world." user_id=1>]

You can also explicitly request a new object via the API when using build. This is useful if you're dealing with default attributes.

class Comment
  include Him::Model
  request_new_object_on_build true
end

@user = User.find(1)
@user.comments.build(body: "Just a draft")
# GET "/users/1/comments/new" with `body=Just+a+draft.`
# => [#<Comment id=nil body="Just a draft" archived=false user_id=1>]

Notes about paths

Resources must always have all the required attributes to build their complete path. For example, if you have these models:

class User
  include Him::Model
  collection_path "organizations/:organization_id/users"
end

class Organization
  include Him::Model
  has_many :users
end

Him expects all User resources to have an :organization_id (or :_organization_id) attribute. Otherwise, calling mostly all methods, like User.all, will throw an exception like this one:

Him::Errors::PathError: Missing :_organization_id parameter to build the request path. Path is `organizations/:organization_id/users`. Parameters are `{ … }`.

Associations with custom attributes

Associations can also be made using custom attributes:

class User
  include Him::Model
  belongs_to :owns, class_name: "Organization"
end

class Organization
  include Him::Model
  has_many :owners, class_name: "User"
end

Validations

Him includes ActiveModel::Validations so you can declare validations the same way you do in Rails.

However, validations must be triggered manually — they are not run, for example, when calling #save on an object, or #create on a model class.

class User
  include Him::Model

  attributes :fullname, :email
  validates :fullname, presence: true
  validates :email, presence: true
end

@user = User.new(fullname: "Tobias Fünke")
@user.valid? # => false

@user.save
# POST "/users" with `fullname=Tobias+Fünke` will still be called, even if the user is not valid

Dirty attributes

Him includes ActiveModel::Dirty so you can keep track of the attributes that have changed in an object.

class User
  include Him::Model

  attributes :fullname, :email
end

@user = User.new(fullname: "Tobias Fünke")
@user.fullname_changed? # => true
@user.changes # => { :fullname => [nil, "Tobias Fünke"] }

@user.save
# POST "/users" with `fullname=Tobias+Fünke`

@user.fullname_changed? # => false
@user.changes # => {}

To update only the modified attributes specify :send_only_modified_attributes => true in the setup.

Callbacks

You can add before and after callbacks to your models that are triggered on specific actions. You can use symbols or blocks.

class User
  include Him::Model
  before_save :set_internal_id
  after_find { |u| u.fullname.upcase! }

  def set_internal_id
    self.internal_id = 42 # Will be passed in the HTTP request
  end
end

@user = User.create(fullname: "Tobias Fünke")
# POST "/users" with `fullname=Tobias+Fünke&internal_id=42`

@user = User.find(1)
@user.fullname # => "TOBIAS FUNKE"

The available callbacks are:

before_save
before_create
before_update
before_destroy
after_save
after_create
after_update
after_destroy
after_find
after_initialize

JSON attributes-wrapping

Him supports sending and parsing JSON data wrapped in a root element (to be compatible with Rails’ include_root_in_json setting), like so:

Sending

If you want to send all data to your API wrapped in a root element based on the model name.

class User
  include Him::Model
  include_root_in_json true
end

class Article
  include Him::Model
  include_root_in_json :post
end

User.create(fullname: "Tobias Fünke")
# POST "/users" with `user[fullname]=Tobias+Fünke`

Article.create(title: "Hello world.")
# POST "/articles" with `post[title]=Hello+world`

Parsing

If the API returns data wrapped in a root element based on the model name.

class User
  include Him::Model
  parse_root_in_json true
end

class Article
  include Him::Model
  parse_root_in_json :post
end

user = User.create(fullname: "Tobias Fünke")
# POST "/users" with `fullname=Tobias+Fünke`, response is { "user": { "fullname": "Tobias Fünke" } }
user.fullname # => "Tobias Fünke"

article = Article.create(title: "Hello world.")
# POST "/articles" with `title=Hello+world.`, response is { "post": { "title": "Hello world." } }
article.title # => "Hello world."

Of course, you can use both include_root_in_json and parse_root_in_json at the same time.

ActiveModel::Serializers support

If the API returns data in the default format used by the ActiveModel::Serializers project you need to configure Him as follows:

class User
  include Him::Model
  parse_root_in_json true, format: :active_model_serializers
end

user = Users.find(1)
# GET "/users/1", response is { "user": { "id": 1, "fullname": "Lindsay Fünke" } }

users = Users.all
# GET "/users", response is { "users": [{ "id": 1, "fullname": "Lindsay Fünke" }, { "id": 1, "fullname": "Tobias Fünke" }] }

JSON API support

To consume a JSON API 1.0 compliant service, it must return data in accordance with the JSON API spec. The general format of the data is as follows:

{ "data": {
  "type": "developers",
  "id": "6ab79c8c-ec5a-4426-ad38-8763bbede5a7",
  "attributes": {
    "language": "ruby",
    "name": "avdi grimm",
  }
}

Then to setup your models:

class Contributor
  include Him::JsonApi::Model

  # defaults to demodulized, pluralized class name, e.g. contributors
  type :developers
end

Finally, you'll need to use the included JsonApiParser Him middleware:

Him::API.setup url: 'https://my_awesome_json_api_service' do |c|
  # Request
  c.use FaradayMiddleware::EncodeJson

  # Response
  c.use Him::Middleware::JsonApiParser

  # Adapter
  c.adapter :net_http
end

Custom requests

You can easily define custom requests for your models using custom_get, custom_post, etc.

class User
  include Him::Model

  custom_get :popular, :unpopular
  custom_post :from_default, :activate
end

User.popular
# GET "/users/popular"
# => [#<User id=1>, #<User id=2>]

User.unpopular
# GET "/users/unpopular"
# => [#<User id=3>, #<User id=4>]

User.from_default(name: "Maeby Fünke")
# POST "/users/from_default" with `name=Maeby+Fünke`
# => #<User id=5 name="Maeby Fünke">

User.activate(id: 6)
# POST "/users/6/activate"
# => #<User id=6>

You can also use get, post, put or delete (which maps the returned data to either a collection or a resource).

class User
  include Him::Model
end

User.get(:popular)
# GET "/users/popular"
# => [#<User id=1>, #<User id=2>]

User.get(:single_best)
# GET "/users/single_best"
# => #<User id=1>

You can also use get_raw which yields the parsed data and the raw response from the HTTP request. Other HTTP methods are supported (post_raw, put_raw, etc.).

class User
  include Him::Model

  def self.total
    get_raw(:stats) do |parsed_data, response|
      parsed_data[:data][:total_users]
    end
  end
end

User.total
# GET "/users/stats"
# => 42

You can also use full request paths (with strings instead of symbols).

class User
  include Him::Model
end

User.get("/users/popular")
# GET "/users/popular"
# => [#<User id=1>, #<User id=2>]

Custom paths

You can define custom HTTP paths for your models:

class User
  include Him::Model
  collection_path "/hello_users/:id"
end

@user = User.find(1)
# GET "/hello_users/1"

You can also include custom variables in your paths:

class User
  include Him::Model
  collection_path "/organizations/:organization_id/users"
end

@user = User.find(1, _organization_id: 2)
# GET "/organizations/2/users/1"

@user = User.all(_organization_id: 2)
# GET "/organizations/2/users"

@user = User.new(fullname: "Tobias Fünke", organization_id: 2)
@user.save
# POST "/organizations/2/users" with `fullname=Tobias+Fünke`

Custom primary keys

If your record uses an attribute other than :id to identify itself, specify it using the primary_key method:

class User
  include Him::Model
  primary_key :_id
end

user = User.find("4fd89a42ff204b03a905c535")
# GET "/users/4fd89a42ff204b03a905c535", response is { "_id": "4fd89a42ff204b03a905c535", "name": "Tobias" }

user.destroy
# DELETE "/users/4fd89a42ff204b03a905c535"

Inheritance

If all your models share the same settings, you might want to make them children of a class and only include Him::Model in that class. However, there are a few settings that don’t get passed to the children classes:

root_element
collection_path and resource_path

Those settings are based on the class name, so you don’t have to redefine them each time you create a new children class (but you still can). Every other setting is inherited from the parent (associations, scopes, JSON settings, etc.).

module MyAPI
  class Model
    include Him::Model

    parse_root_in_json true
    include_root_in_json true
  end
end

class User < MyAPI::Model
end

User.find(1)
# GET "/users/1"

Scopes

Just like with ActiveRecord, you can define named scopes for your models. Scopes are chainable and can be used within other scopes.

class User
  include Him::Model

  scope :by_role, ->(role) { where(role: role) }
  scope :admins, -> { by_role('admin') }
  scope :active, -> { where(active: 1) }
end

@admins = User.admins
# GET "/users?role=admin"

@moderators = User.by_role('moderator')
# GET "/users?role=moderator"

@active_admins = User.active.admins # @admins.active would have worked here too
# GET "/users?role=admin&active=1"

A neat trick you can do with scopes is interact with complex paths.

class User
  include Him::Model

  collection_path "organizations/:organization_id/users"
  scope :for_organization, ->(id) { where(organization_id: id) }
end

@user = User.for_organization(3).find(2)
# GET "/organizations/3/users/2"

@user = User.for_organization(3).create(fullname: "Tobias Fünke")
# POST "/organizations/3" with `fullname=Tobias+Fünke`

Multiple APIs

It is possible to use different APIs for different models. Instead of calling Him::API.setup, you can create instances of Him::API:

# config/initializers/him.rb
MY_API = Him::API.new
MY_API.setup url: "https://my-api.example.com" do |c|
  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

OTHER_API = Him::API.new
OTHER_API.setup url: "https://other-api.example.com" do |c|
  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

You can then define which API a model will use:

class User
  include Him::Model
  use_api MY_API
end

class Category
  include Him::Model
  use_api OTHER_API
end

User.all
# GET "https://my-api.example.com/users"

Category.all
# GET "https://other-api.example.com/categories"

SSL

When initializing Him::API, you can pass any parameter supported by Faraday.new. So to use HTTPS, you can use Faraday’s :ssl option.

ssl_options = { ca_path: "/usr/lib/ssl/certs" }
Him::API.setup url: "https://api.example.com", ssl: ssl_options do |c|
  # Response
  c.use Him::Middleware::DefaultParseJSON

  # Adapter
  c.adapter :net_http
end

Testing

Suppose we have these two models bound to your API:

# app/models/user.rb
class User
  include Him::Model
  custom_get :popular
end

# app/models/post.rb
class Post
  include Him::Model
  custom_get :recent, :archived
end

In order to test them, we’ll have to stub the remote API requests. With RSpec, we can do this like so:

# spec/spec_helper.rb
RSpec.configure do |config|
  config.include(Module.new do
    def stub_api_for(klass)
      klass.use_api (api = Him::API.new)

      # Here, you would customize this for your own API (URL, middleware, etc)
      # like you have done in your application’s initializer
      api.setup url: "http://api.example.com" do |c|
        c.use Him::Middleware::FirstLevelParseJSON
        c.adapter(:test) { |s| yield(s) }
      end
    end
  end)
end

Then, in your tests, we can specify what (fake) HTTP requests will return:

# spec/models/user.rb
describe User do
  before do
    stub_api_for(User) do |stub|
      stub.get("/users/popular") { |env| [200, {}, [{ id: 1, name: "Tobias Fünke" }, { id: 2, name: "Lindsay Fünke" }].to_json] }
    end
  end

  describe :popular do
    subject { User.popular }
    its(:length) { should == 2 }
    its(:errors) { should be_empty }
  end
end

We can redefine the API for a model as many times as we want, like for more complex tests:

# spec/models/user.rb
describe Post do
  describe :recent do
    before do
      stub_api_for(Post) do |stub|
        stub.get("/posts/recent") { |env| [200, {}, [{ id: 1 }, { id: 2 }].to_json] }
      end
    end

    subject { Post.recent }
    its(:length) { should == 2 }
    its(:errors) { should be_empty }
  end

  describe :archived do
    before do
      stub_api_for(Post) do |stub|
        stub.get("/posts/archived") { |env| [200, {}, [{ id: 1 }, { id: 2 }].to_json] }
      end
    end

    subject { Post.archived }
    its(:length) { should == 2 }
    its(:errors) { should be_empty }
  end
end

Contribute

Feel free to contribute and submit issues/pull requests on GitHub.

him

Development

Runtime

Him