Alba

Alba is the fastest JSON serializer for Ruby, JRuby, and TruffleRuby.

Discussions

Alba uses GitHub Discussions to openly discuss the project.

If you’ve already used Alba, please consider posting your thoughts and feelings on Feedback. The fact that you enjoy using Alba gives me energy to keep developing Alba!

If you have feature requests or interesting ideas, join us with Ideas. Let’s make Alba even better, together!

Why Alba?

Because it’s fast, easy-to-use and extensible!

Fast

Alba is faster than most of the alternatives. We have a benchmark.

Easy to use

Alba provides four DSLs, attributes to fetch attribute with its name, attribute to execute block for the attribute, one to seriaize single attribute with another resource, and many to serialize collection attribute with another resource. When you want to do something complex, there are many examples in this README so you can mimic them to get started.

Extensible

Alba embraces extensibility through common techniques such as class inheritance and module inclusion. Alba provides its capacity with one module so you can still have your own class hierarchy.

Installation

Add this line to your application’s Gemfile:

gem 'alba'

And then execute:

$ bundle install

Or install it yourself as:

$ gem install alba

Supported Ruby versions

Alba supports CRuby 2.5 and higher and latest JRuby and TruffleRuby.

Documentation

You can find the documentation on RubyDoc.

Features

• Conditional attributes and associations
• Selectable backend
• Key transformation
• Root key inference
• Error handling
• Nil handling
• Resource name inflection based on association name
• Circular associations control
• [Experimental] Types for validation and conversion
• Layout
• No runtime dependencies

Usage

Configuration

Alba’s configuration is fairly simple.

Backend configuration

Backend is the actual part serializing an object into JSON. Alba supports these backends.

• Oj, the fastest. Gem installation required.
• active_support, mostly for Rails. Gem installation required.
• default or json, with no external dependencies.

You can set a backend like this:

Alba.backend = :oj

Encoder configuration

You can also set JSON encoder directly with a Proc.

Alba.encoder = ->(object) { JSON.generate(object) }

You can consider setting a backend with Symbol as a shortcut to set encoder.

Inference configuration

You can enable inference feature using enable_inference! method.

Alba.enable_inference!(with: :active_support)

You can choose which inflector Alba uses for inference. Possible value for with option are:

• :active_support for ActiveSupport::Inflector
• :dry for Dry::Inflector
• any object which responds to some methods (see below)

For the details, see Error handling section

Simple serialization with root key

You can define attributes with (yes) attributes macro with attribute names. If your attribute need some calculations, you can use attribute with block.

class User
  attr_accessor :id, :name, :email, :created_at, :updated_at
  def initialize(id, name, email)
    @id = id
    @name = name
    @email = email
    @created_at = Time.now
    @updated_at = Time.now
  end
end

class UserResource
  include Alba::Resource

  root_key :user

  attributes :id, :name

  attribute :name_with_email do |resource|
    "#{resource.name}: #{resource.email}"
  end
end

user = User.new(1, 'Masafumi OKURA', 'masafumi@example.com')
UserResource.new(user).serialize
# => "{\"user\":{\"id\":1,\"name\":\"Masafumi OKURA\",\"name_with_email\":\"Masafumi OKURA: masafumi@example.com\"}}"

You can define instance methods on resources so that you can use it as attribute name in attributes.

# The serialization result is the same as above
class UserResource
  include Alba::Resource

  root_key :user, :users # Later is for plural

  attributes :id, :name, :name_with_email

  # Attribute methods must accept one argument for each serialized object
  def name_with_email(user)
    "#{user.name}: #{user.email}"
  end
end

This even works with users collection.

user1 = User.new(1, 'Masafumi OKURA', 'masafumi@example.com')
user2 = User.new(2, 'Test User', 'test@example.com')
UserResource.new([user1, user2]).serialize
# => "{\"users\":[{\"id\":1,\"name\":\"Masafumi OKURA\",\"name_with_email\":\"Masafumi OKURA: masafumi@example.com\"},{\"id\":2,\"name\":\"Test User\",\"name_with_email\":\"Test User: test@example.com\"}]}"

Serialization with associations

class User
  attr_reader :id, :created_at, :updated_at
  attr_accessor :articles

  def initialize(id)
    @id = id
    @created_at = Time.now
    @updated_at = Time.now
    @articles = []
  end
end

class Article
  attr_accessor :user_id, :title, :body

  def initialize(user_id, title, body)
    @user_id = user_id
    @title = title
    @body = body
  end
end

class ArticleResource
  include Alba::Resource

  attributes :title
end

class UserResource
  include Alba::Resource

  attributes :id

  many :articles, resource: ArticleResource
end

user = User.new(1)
article1 = Article.new(1, 'Hello World!', 'Hello World!!!')
user.articles << article1
article2 = Article.new(2, 'Super nice', 'Really nice!')
user.articles << article2

UserResource.new(user).serialize
# => '{"id":1,"articles":[{"title":"Hello World!"},{"title":"Super nice"}]}'

You can define associations inline if you don’t need a class for association.

class ArticleResource
  include Alba::Resource

  attributes :title
end

class UserResource
  include Alba::Resource

  attributes :id

  many :articles, resource: ArticleResource
end

# This class works the same as `UserResource`
class AnotherUserResource
  include Alba::Resource

  attributes :id

  many :articles do
    attributes :title
  end
end

You can “filter” association using second proc argument. This proc takes association object and params.

This feature is useful when you want to modify association, such as adding includes or order to ActiveRecord relations.

class User
  attr_reader :id
  attr_accessor :articles

  def initialize(id)
    @id = id
    @articles = []
  end
end

class Article
  attr_accessor :id, :title, :body

  def initialize(id, title, body)
    @id = id
    @title = title
    @body = body
  end
end

class ArticleResource
  include Alba::Resource

  attributes :title
end

class UserResource
  include Alba::Resource

  attributes :id

  # Second proc works as a filter
  many :articles,
    proc { |articles, params|
      filter = params[:filter] || :odd?
      articles.select {|a| a.id.send(filter) }
    },
    resource: ArticleResource
end

user = User.new(1)
article1 = Article.new(1, 'Hello World!', 'Hello World!!!')
user.articles << article1
article2 = Article.new(2, 'Super nice', 'Really nice!')
user.articles << article2

UserResource.new(user).serialize
# => '{"id":1,"articles":[{"title":"Hello World!"}]}'
UserResource.new(user, params: {filter: :even?}).serialize
# => '{"id":1,"articles":[{"title":"Super nice"}]}'

Inline definition with Alba.serialize

Alba.serialize method is a shortcut to define everything inline.

Alba.serialize(user, root_key: :foo) do
  attributes :id
  many :articles do
    attributes :title, :body
  end
end
# => '{"foo":{"id":1,"articles":[{"title":"Hello World!","body":"Hello World!!!"},{"title":"Super nice","body":"Really nice!"}]}}'

Alba.serialize can be used when you don’t know what kind of object you serialize. For example:

Alba.serialize(something)
# => Same as `FooResource.new(something).serialize` when `something` is an instance of `Foo`.

Although this might be useful sometimes, it’s generally recommended to define a class for Resource.

Inheritance and attributes filter

You can filter out certain attributes by overriding attributes instance method. This is useful when you want to customize existing resource with inheritance.

You can access raw attributes via super call. It returns a Hash whose keys are the name of the attribute and whose values are the body. Usually you need only keys to filter out, like below.

class Foo
  attr_accessor :id, :name, :body

  def initialize(id, name, body)
    @id = id
    @name = name
    @body = body
  end
end

class GenericFooResource
  include Alba::Resource

  attributes :id, :name, :body
end

class RestrictedFooResource < GenericFooResource
  def attributes
    super.select { |key, _| key.to_sym == :name }
  end
end

RestrictedFooResource.new(foo).serialize
# => '{"name":"my foo"}'

Key transformation

If you want to use transform_keys DSL and you already have active_support installed, key transformation will work out of the box, using ActiveSupport::Inflector. If active_support is not around, you have 2 possibilities:

• install it
• use a custom inflector

With transform_keys DSL, you can transform attribute keys.

class User
  attr_reader :id, :first_name, :last_name

  def initialize(id, first_name, last_name)
    @id = id
    @first_name = first_name
    @last_name = last_name
  end
end

class UserResource
  include Alba::Resource

  attributes :id, :first_name, :last_name

  transform_keys :lower_camel
end

user = User.new(1, 'Masafumi', 'Okura')
UserResourceCamel.new(user).serialize
# => '{"id":1,"firstName":"Masafumi","lastName":"Okura"}'

Possible values for transform_keys argument are:

• :camel for CamelCase
• :lower_camel for lowerCamelCase
• :dash for dash-case
• :snake for snake_case
• :none for not transforming keys

You can also transform root key when:

• Alba.enable_inference! is called
• root_key! is called in Resource class
• root option of transform_keys is set to true

Alba.enable_inference!(with: :active_support) # with :dry also works

class BankAccount
  attr_reader :account_number

  def initialize(account_number)
    @account_number = account_number
  end
end

class BankAccountResource
  include Alba::Resource

  root_key!

  attributes :account_number
  transform_keys :dash, root: true
end

bank_account = BankAccount.new(123_456_789)
BankAccountResource.new(bank_account).serialize
# => '{"bank-account":{"account-number":123456789}}'

This behavior to transform root key will become default at version 2.

Supported transformation types are :camel, :lower_camel and :dash.

Custom inflector

A custom inflector can be plugged in as follows.

module CustomInflector
  module_function

  def camelize(string)
  end

  def camelize_lower(string)
  end

  def dasherize(string)
  end

  def underscore(string)
  end

  def classify(string)
  end
end

Alba.enable_inference!(with: CustomInflector)

Filtering attributes

You can filter attributes by overriding Alba::Resource#converter method, but it’s a bit tricky.

class User
  attr_accessor :id, :name, :email, :created_at, :updated_at

  def initialize(id, name, email)
    @id = id
    @name = name
    @email = email
  end
end

class UserResource
  include Alba::Resource

  attributes :id, :name, :email

  private

  # Here using `Proc#>>` method to compose a proc from `super`
  def converter
    super >> proc { |hash| hash.compact }
  end
end

user = User.new(1, nil, nil)
UserResource.new(user).serialize # => '{"id":1}'

The key part is the use of Proc#>> since Alba::Resource#converter returns a Proc which contains the basic logic and it’s impossible to change its behavior by just overriding the method.

It’s not recommended to swap the whole conversion logic. It’s recommended to always call super when you override converter.

Conditional attributes

Filtering attributes with overriding convert works well for simple cases. However, It’s cumbersome when we want to filter various attributes based on different conditions for keys.

In these cases, conditional attributes works well. We can pass if option to attributes, attribute, one and many. Below is an example for the same effect as filtering attributes section.

class User
  attr_accessor :id, :name, :email, :created_at, :updated_at

  def initialize(id, name, email)
    @id = id
    @name = name
    @email = email
  end
end

class UserResource
  include Alba::Resource

  attributes :id, :name, :email, if: proc { |user, attribute| !attribute.nil? }
end

user = User.new(1, nil, nil)
UserResource.new(user).serialize # => '{"id":1}'

Default

Alba doesn’t support default value for attributes, but it’s easy to set a default value.

class FooResource
  attribute :bar do |foo|
    foo.bar || 'default bar'
  end
end

We believe this is clearer than using some (not implemented yet) DSL such as default because there are some conditions where default values should be applied (nil, blank?, empty? etc.)

Inference

After Alba.enable_inference! called, Alba tries to infer root key and association resource name.

Alba.enable_inference!(with: :active_support) # with :dry also works

class User
  attr_reader :id
  attr_accessor :articles

  def initialize(id)
    @id = id
    @articles = []
  end
end

class Article
  attr_accessor :id, :title

  def initialize(id, title)
    @id = id
    @title = title
  end
end

class ArticleResource
  include Alba::Resource

  attributes :title
end

class UserResource
  include Alba::Resource

  key!

  attributes :id

  many :articles
end

user = User.new(1)
user.articles << Article.new(1, 'The title')

UserResource.new(user).serialize # => '{"user":{"id":1,"articles":[{"title":"The title"}]}}'
UserResource.new([user]).serialize # => '{"users":[{"id":1,"articles":[{"title":"The title"}]}]}'

This resource automatically sets its root key to either “users” or “user”, depending on the given object is collection or not.

Also, you don’t have to specify which resource class to use with many. Alba infers it from association name.

Error handling

You can set error handler globally or per resource using on_error.

class User
  attr_accessor :id, :name

  def initialize(id, name, email)
    @id = id
    @name = name
    @email = email
  end

  def email
    raise RuntimeError, 'Error!'
  end
end

class UserResource
  include Alba::Resource

  attributes :id, :name, :email

  on_error :ignore
end

user = User.new(1, 'Test', 'email@example.com')
UserResource.new(user).serialize # => '{"id":1,"name":"Test"}'

This way you can exclude an entry when fetching an attribute gives an exception.

There are four possible arguments on_error method accepts.

• :raise re-raises an error. This is the default behavior.
• :ignore ignores the entry with the error.
• :nullify sets the attribute with the error to nil.
• Block gives you more control over what to be returned.

The block receives five arguments, error, object, key, attribute and resource class and must return a two-element array. Below is an example.

class ExampleResource
  include Alba::Resource
  on_error do |error, object, key, attribute, resource_class|
    if resource_class == MyResource
      ['error_fallback', object.error_fallback]
    else
      [key, error.message]
    end
  end
end

Nil handling

Sometimes we want to convert nil to different values such as empty string. Alba provides a flexible way to handle nil.

class User
  attr_reader :id, :name, :age

  def initialize(id, name = nil, age = nil)
    @id = id
    @name = name
    @age = age
  end
end

class UserResource
  include Alba::Resource

  on_nil { '' }

  root_key :user, :users

  attributes :id, :name, :age
end

UserResource.new(User.new(1)).serialize
# => '{"user":{"id":1,"name":"","age":""}}'

You can get various information via block parameters.

class UserResource
  include Alba::Resource

  on_nil do |object, key|
    if key == age
      20
    else
      "User#{object.id}"
    end
  end

  root_key :user, :users

  attributes :id, :name, :age
end

UserResource.new(User.new(1)).serialize
# => '{"user":{"id":1,"name":"User1","age":20}}'

Metadata

You can set a metadata with meta DSL or meta option.

class UserResource
  include Alba::Resource

  root_key :user, :users

  attributes :id, :name

  meta do
    if object.is_a?(Enumerable)
      {size: object.size}
    else
      {foo: :bar}
    end
  end
end

user = User.new(1, 'Masafumi OKURA', 'masafumi@example.com')
UserResource.new([user]).serialize
# => '{"users":[{"id":1,"name":"Masafumi OKURA"}],"meta":{"size":1}}'

# You can merge metadata with `meta` option

UserResource.new([user]).serialize(meta: {foo: :bar})
# => '{"users":[{"id":1,"name":"Masafumi OKURA"}],"meta":{"size":1,"foo":"bar"}}'

# You can set metadata with `meta` option alone

class UserResourceWithoutMeta
  include Alba::Resource

  root_key :user, :users

  attributes :id, :name
end

UserResource.new([user]).serialize(meta: {foo: :bar})
# => '{"users":[{"id":1,"name":"Masafumi OKURA"}],"meta":{"foo":"bar"}}'

You can use object method to access the underlying object and params to access the params in meta block.

Note that setting root key is required when setting a metadata.

Circular associations control

Note that this feature works correctly since version 1.3. In previous versions it doesn’t work as expected.

You can control circular associations with within option. within option is a nested Hash such as {book: {authors: books}}. In this example, Alba serializes a book’s authors’ books. This means you can reference BookResource from AuthorResource and vice versa. This is really powerful when you have a complex data structure and serialize certain parts of it.

For more details, please refer to test code

Experimental support of types

You can validate and convert input with types.

class User
  attr_reader :id, :name, :age, :bio, :admin, :created_at

  def initialize(id, name, age, bio = '', admin = false) # rubocop:disable Style/OptionalBooleanParameter
    @id = id
    @name = name
    @age = age
    @admin = admin
    @bio = bio
    @created_at = Time.new(2020, 10, 10)
  end
end

class UserResource
  include Alba::Resource

  attributes :name, id: [String, true], age: [Integer, true], bio: String, admin: [:Boolean, true], created_at: [String, ->(object) { object.strftime('%F') }]
end

user = User.new(1, 'Masafumi OKURA', '32', 'Ruby dev')
UserResource.new(user).serialize
# => '{"name":"Masafumi OKURA","id":"1","age":32,"bio":"Ruby dev","admin":false,"created_at":"2020-10-10"}'

Notice that id and created_at are converted to String and age is converted to Integer.

If type is not correct and auto conversion is disabled (default), TypeError occurs.

user = User.new(1, 'Masafumi OKURA', '32', nil) # bio is nil and auto conversion is disabled for bio
UserResource.new(user).serialize
# => TypeError, 'Attribute bio is expected to be String but actually nil.'

Note that this feature is experimental and interfaces are subject to change.

Layout

Sometimes we’d like to serialize JSON into a template. In other words, we need some structure OUTSIDE OF serialized JSON. IN HTML world, we call it a “layout”.

Alba supports serializing JSON in a layout. You need a file for layout and then to specify file with layout method.

{
  "header": "my_header",
  "body": <%= serialized_json %>
}

class FooResource
  include Alba::Resource
  layout file: 'my_layout.json.erb'
end

Note that layout files are treated as json and erb and evaluated in a context of the resource, meaning

• A layout file must be a valid JSON
• You must write <%= serialized_json %> in a layout to put serialized JSON string into a layout
• You can access params in a layout so that you can add virtually any objects to a layout
  • When you access params, it’s usually a Hash. You can use encode method in a layout to convert params Hash into a JSON with the backend you use
• You can also access object, the underlying object for the resource

In case you don’t want to have a file for layout, Alba lets you define and apply layouts inline:

class FooResource
  include Alba::Resource
  layout inline: proc do
    {
      header: 'my header',
      body: serializable_hash
    }
  end
end

In the example above, we specify a Proc which returns a Hash as an inline layout. In the Proc we can use serializable_hash method to access a Hash right before serialization.

You can also use a Proc which returns String, not a Hash, for an inline layout.

class FooResource
  include Alba::Resource
  layout inline: proc do
    %({
      "header": "my header",
      "body": #{serialized_json}
    })
  end
end

It looks similar to file layout but you must use string interpolation for method calls since it’s not an ERB.

Also note that we use percentage notation here to use double quotes. Using single quotes in inline string layout causes the error which might be resolved in other ways.

Caching

Currently, Alba doesn’t support caching, primarily due to the behavior of ActiveRecord::Relation’s cache. See the issue.

Extend Alba

Sometimes we have shared behaviors across resources. In such cases we can have a module for common logic.

In attribute block we can call instance method so we can improve the code below:

class FooResource
  include Alba::Resource
  # other attributes
  attribute :created_at do |foo|
    foo.created_at.strftime('%m/%d/%Y')
  end

  attribute :updated_at do |foo|
    foo.updated_at.strftime('%m/%d/%Y')
  end
end

class BarResource
  include Alba::Resource
  # other attributes
  attribute :created_at do |bar|
    bar.created_at.strftime('%m/%d/%Y')
  end

  attribute :updated_at do |bar|
    bar.updated_at.strftime('%m/%d/%Y')
  end
end

to:

module SharedLogic
  def format_time(time)
    time.strftime('%m/%d/%Y')
  end
end

class FooResource
  include Alba::Resource
  include SharedLogic
  # other attributes
  attribute :created_at do |foo|
    format_time(foo.created_at)
  end

  attribute :updated_at do |foo|
    format_time(foo.updated_at)
  end
end

class BarResource
  include Alba::Resource
  include SharedLogic
  # other attributes
  attribute :created_at do |bar|
    format_time(bar.created_at)
  end

  attribute :updated_at do |bar|
    format_time(bar.updated_at)
  end
end

We can even add our own DSL to serialize attributes for readability and removing code duplications.

To do so, we need to extend our module. Let’s see how we can achieve the same goal with this approach.

module AlbaExtension
  # Here attrs are an Array of Symbol
  def formatted_time_attributes(*attrs)
    attrs.each do |attr|
      attribute attr do |object|
        time = object.send(attr)
        time.strftime('%m/%d/%Y')
      end
    end
  end
end

class FooResource
  include Alba::Resource
  extend AlbaExtension
  # other attributes
  formatted_time_attributes :created_at, :updated_at
end

class BarResource
  include Alba::Resource
  extend AlbaExtension
  # other attributes
  formatted_time_attributes :created_at, :updated_at
end

In this way we have shorter and cleaner code. Note that we need to use send or public_send in attribute block to get attribute data.

Rails

When you use Alba in Rails, you can create an initializer file with the line below for compatibility with Rails JSON encoder.

Alba.backend = :active_support
# or
Alba.backend = :oj_rails

Why named “Alba”?

The name “Alba” comes from “albatross”, a kind of birds. In Japanese, this bird is called “Aho-dori”, which means “stupid bird”. I find it funny because in fact albatrosses fly really fast. I hope Alba looks stupid but in fact it does its job quick.

Pioneers

There are great pioneers in Ruby’s ecosystem which does basically the same thing as Alba does. To name a few:

• ActiveModelSerializers a.k.a AMS, the most famous implementation of JSON serializer for Ruby
• Blueprinter shares some concepts with Alba

Development

After checking out the repo, run bin/setup to install dependencies. Then, run rake test to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/okuramasafumi/alba. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the code of conduct.

License

The gem is available as open source under the terms of the MIT License.

Code of Conduct

Everyone interacting in the Alba project’s codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.