Disposable

Decorators on top of your ORM layer.

Introduction

Disposable is the missing API of ActiveRecord*. The mission:

Maintain a manipulatable object graph that is a copy/map of a persistent structure.
Prevent any write to the persistence layer until you say sync.
Help designing your domain layer without being restricted to database layouts (renaming, compositions, hash fields).
Provide additional behavior like change tracking, imperative callbacks and collection semantics.

Disposable gives you "Twins": non-persistent domain objects. That is reflected in the name of the gem. They can read from and write values to a persistent object and abstract the persistence layer until data is synced to the model.

API

The public twin API is unbelievably simple.

Twin::new creates and populates the twin.
Twin#"reader" returns the value or nested twin of the property.
Twin#"writer"=(v) writes the value to the twin, not the model.
Twin#sync writes all values to the model.
Twin#save writes all values to the model and calls save on configured models.

Twin

Twins are only # FIXME % slower than AR alone.

Twins implement light-weight decorators objects with a unified interface. They map objects, hashes, and compositions of objects, along with optional hashes to inject additional options.

Every twin is based on a defined schema.

class AlbumTwin < Disposable::Twin
  property :title
  property :playable?, virtual: true # context-sensitive, e.g. current_user dependent.

  collection :songs do
    property :name
    property :index
  end

  property :artist do
    property :full_name
  end
end

Constructor

Twins get populated from the decorated models.

Song   = Struct.new(:name, :index)
Artist = Struct.new(:full_name)
Album  = Struct.new(:title, :songs, :artist)

You need to pass model and the facultative options to the twin constructor.

album = Album.new("Nice Try")
twin  = AlbumTwin.new(album, playable?: current_user.can?(:play))

Readers

This will create a composition object of the actual model and the hash.

twin.title     #=> "Nice Try"
twin.playable? #=> true

You can also override property values in the constructor:

twin = AlbumTwin.new(album, title: "Plasticash")
twin.title #=> "Plasticash"

Writers

Writers change values on the twin and are not propagated to the model.

twin.title = "Skamobile"
twin.title  #=> "Skamobile"
album.title #=> "Nice Try"

Writers on nested twins will "twin" the value.

twin.songs #=> []
twin.songs << Song.new("Adondo", 1)
twin.songs  #=> [<Twin::Song name="Adondo" index=1 model=<Song ..>>]
album.songs #=> []

The added twin is not passed to the model. Note that the nested song is a twin, not the model itself.

Sync

Given the above state change on the twin, here is what happens after calling #sync.

album.title  #=> "Nice Try"
album.songs #=> []

twin.sync

album.title  #=> "Skamobile"
album.songs #=> [<Song name="Adondo" index=1>]

#sync writes all configured attributes back to the models using public setters as album.name= or album.songs=. This is recursive and will sync the entire object graph.

Note that sync might already trigger saving the model as persistence layers like ActiveRecord can't deal with collection= [] and instantly persist that.

You may implement your syncing manually by passing a block to sync.

twin.sync do |hash|
  hash #=> {
  #  "title"     => "Skamobile",
  #  "playable?" => true,
  #  "songs"     => [{"name"=>"Adondo"...}..]
  # }
end

Invoking sync with block will not write anything to the models.

Needs to be included explicitly (Sync).

Save

Calling #save will do sync plus calling save on all nested models. This implies that the models need to implement #save.

twin.save
#=> album.save
#=>      .songs[0].save

Needs to be included explicitly (Save).

Nested Twin

Nested objects can be declared with an inline twin.

property :artist do
  property :full_name
end

The setter will automatically "twin" the model.

twin.artist = Artist.new
twin.artist #=> <Twin::Artist model=<Artist ..>>

You can also specify nested objects with an explicit class.

property :artist, twin: TwinArtist

Unnest

todo: document

Features

You can simply include feature modules into twins. If you want a feature to be included into all inline twins of your schema, use ::feature.

class AlbumTwin < Disposable::Twin
  feature Coercion

  property :artist do
    # this will now include Coercion, too.

Coercion

Twins can use dry-types coercion. This will override the setter in your twin, coerce the incoming value, and call the original setter. Nothing more will happen.

Disposable already defines a module Disposable::Twin::Coercion::Types with all the Dry::Types built-in types. So you can use any of the types documented in http://dry-rb.org/gems/dry-types/built-in-types/.

class AlbumTwin < Disposable::Twin
  feature Coercion
  feature Setup::SkipSetter

  property :id, type: Types::Form::Int

The :type option defines the coercion type. You may incluce Setup::SkipSetter, too, as otherwise the coercion will happen at initialization time and in the setter.

twin.id = "1"
twin.id #=> 1

Again, coercion only happens in the setter.

Defaults

Default values can be set via :default.

class AlbumTwin < Disposable::Twin
  feature Default

  property :title, default: "The Greatest Songs Ever Written"
  property :composer, default: Composer.new do
    property :name, default: -> { "Object-#{id}" }
  end
end

Default value is applied when the model's getter returns nil when initializing the twin.

Note that :default also works with :virtual and readable: false. :default can also be a lambda which is then executed in twin context.

Collections

Collections can be defined analogue to property. The exposed API is the Array API.

twin.songs = [..] will override the existing value and "twin" every item.
twin.songs << Song.new will add and twin.
twin.insert(0, Song.new) will insert at the specified position and twin.

You can also delete, replace and move items.

twin.songs.delete( twin.songs[0] )

None of these operations are propagated to the model.

Collection Semantics

In addition to the standard Array API the collection adds a handful of additional semantics.

songs=, songs<< and songs.insert track twin via #added.
songs.delete tracks via #deleted.
twin.destroy( twin.songs[0] ) deletes the twin and marks it for destruction in #to_destroy.
twin.songs.save will call destroy on all models marked for destruction in to_destroy. Tracks destruction via #destroyed.

Again, the model is left alone until you call sync or save.

Twin Collections

To twin a collection of models, you can use ::from_collection.

SongTwin.from_collection([song, song])

This will decorate every song instance using a fresh twin.

Change Tracking

The Changed module will allow tracking of state changes in all properties, even nested structures.

class AlbumTwin < Disposable::Twin
  feature Changed

Now, consider the following operations.

twin.name = "Skamobile"
twin.songs << Song.new("Skate", 2) # this adds second song.

This results in the following tracking results.

twin.changed?             #=> true
twin.changed?(:name)      #=> true
twin.changed?(:playable?) #=> false
twin.songs.changed?       #=> true
twin.songs[0].changed?    #=> false
twin.songs[1].changed?    #=> true

Assignments from the constructor are not tracked as changes.

twin = AlbumTwin.new(album)
twin.changed? #=> false

Persistance Tracking

The Persisted module will track the persisted? field of the model, implying that your model exposes this field.

twin.persisted? #=> false
twin.save
twin.persisted? #=> true

The persisted? field is a copy of the model's persisted? flag.

You can also use created? to find out whether a twin's model was already persisted or just got created in this session.

twin = AlbumTwin.new(Album.create) # assuming we were using ActiveRecord.
twin.created? #=> false
twin.save
twin.created? #=> false

This will only return true when the persisted? field has flipped.

Renaming

The Expose module allows renaming properties.

class AlbumTwin < Disposable::Twin
  feature Expose

  property :song_title, from: :title

The public accessor is now song_title whereas the model's accessor needs to be title.

album = OpenStruct.new(title: "Run For Cover")
AlbumTwin.new(album).song_title #=> "Run For Cover"

Composition

Compositions of objects can be mapped, too.

class AlbumTwin < Disposable::Twin
  include Composition

  property :id,    on: :album
  property :title, on: :album
  property :songs, on: :cd
  property :cd_id, on: :cd, from: :id

When initializing a composition, you have to pass a hash that contains the composees.

AlbumTwin.new(album: album, cd: CD.find(1))

Note that renaming works here, too.

Struct

Twins can also map hash properties, e.g. from a deeply nested serialized JSON column.

album.permissions #=> {admin: {read: true, write: true}, user: {destroy: false}}

Map that using the Struct module.

class AlbumTwin < Disposable::Twin
  property :permissions do
     include Struct
    property :admin do
      include Struct
      property :read
      property :write
    end

    property :user # you don't have to use Struct everywhere!
  end

You get fully object-oriented access to your properties.

twin.permissions.admin.read #=> true

Note that you do not have to use Struct everywhere.

twin.permissions.user #=> {destroy: false}

Of course, this works for writing, too.

twin.permissions.admin.read = :MAYBE

After syncing, you will find a hash in the model.

album.permissions #=> {admin: {read: :MAYBE, write: true}, user: {destroy: false}}

With Representers

they indirect data, the twin's attributes get assigned without writing to the persistence layer, yet.

With Contracts

Overriding Getter for Presentation

You can override getters for presentation.

class AlbumTwin < Disposable::Twin
    property :title

    def title
      super.upcase
    end
  end

Be careful, though. The getter normally is also called in sync when writing properties to the models.

You can skip invocation of getters in sync and read values from @fields directly by including Sync::SkipGetter.

class AlbumTwin < Disposable::Twin
  feature Sync
  feature Sync::SkipGetter

Manual Coercion

You can override setters for manual coercion.

class AlbumTwin < Disposable::Twin
    property :title

    def title=(v)
      super(v.trim)
    end
  end

Be careful, though. The setter normally is also called in setup when copying properties from the models to the twin.

Analogue to SkipGetter, include Setup::SkipSetter to write values directly to @fields.

class AlbumTwin < Disposable::Twin
  feature Setup::SkipSetter

Imperative Callbacks

Please refer to the full documentation.

Note: Chapter 8 of the Trailblazer book is dedicated to callbacks and discusses them in great detail.

Callbacks use the fact that twins track state changes. This allows to execute callbacks on certain conditions.

Callback.new(twin).on_create { |twin| .. }
Callback.new(twin.songs).on_add { |twin| .. }
Callback.new(twin.songs).on_add { |twin| .. }

It works as follows.

Twins track state changes, like "item added to collection (on_add)" or "property changed (on_change)".
You decide when to invoke one or a group of callbacks. This is why there's no before_save and the like anymore.
You also decide what events to consider by calling the respective events only, like on_add.
The Callback will now find out which properties of the twin are affected and exectue your passed code for each of them.

This is called Imperative Callback and the opposite of what you've learned from Rails.

By inversing the control, we don't need before_ or after_. This is in your hands now and depends on where you invoke your callbacks.

Events

The following events are available in Callback.

Don't confuse that with event triggering, though! Callbacks are passive, calling an event method means the callback will look for twins that have tracked the respective event (e.g. an twin has changed).

on_update: Invoked when the underlying model was persisted, yet, at twin initialization and attributes have changed since then.
on_add: For every twin that has been added to a collection.
on_add(:create): For every twin that has been added to a collection and got persisted. This will only pick up collection items after sync or save.
on_delete: For every item that has been deleted from a collection.
on_destroy: For every item that has been removed from a collection and physically destroyed.
on_change: For every item that has changed attributes. When persisted? has flippend, this will be triggered, too.
on_change(:email): When the scalar field changed.

Callback Groups

Callback::Group simplifies grouping callbacks and allows nesting.

class AfterSave < Disposable::Callback::Group
  on_change :expire_cache!

  collection :songs do
    on_add :notify_album!
    on_add :reset_song!
  end

  on_update :rehash_name!, property: :title

  property :artist do
    on_change :sing!
  end
end

Calling that group on a twin will invoke all callbacks that apply, in the order they were added.

AfterSave.new(twin).(context: self)

Methods like :sing! will be invoked on the :context object. Likewise, nested properties will be retrieved by simply calling the getter on the twin, like twin.songs.

An options hash is passed as the second argument. # TODO: document Group.(operation: Object.new).

Again, only the events that match will be invoked. If the top level twin hasn't changed, expire_cache! won't be invoked. This works by simply using Callback under the hood.

Callback Inheritance

You can inherit groups, add and remove callbacks.

class EnhancedAfterSave < AfterSave
  on_change :redo!

  collection :songs do
    on_add :rewind!
  end

  remove! :on_change, :expire_cache!
end

The callbacks will be appended to the existing chain.

Instead of appending, you may also refine existing callbacks.

class EnhancedAfterSave < AfterSave
  collection :songs, inherit: true do
    on_delete :rewind!
  end
end

This will add the rewind! callback to the songs property, resulting in the following chain.

collection :songs do
  on_add    :notify_album!
  on_add    :reset_song!
  on_delete :rewind!
end

Readable, Writeable, Virtual

Properties can have various access settings.

readable: false won't read from the model in Setup.
writeable: false won't write to model in Sync.
virtual: true is both settings above combined.

Options

To inject context data into a twin that is not part of any model, you can simply use :virtual properties.

class AlbumTwin < Disposable::Twin
  property :title
  property :current_user, virtual: true
end

You can now pass the current_user as an option into the constructor and then access it via the reader.

twin = AlbumTwin.new(album, current_user: User.find(1))
twin.current_user #=> <User id:1>

Parent

By using the Parent feature you can access the parent twin of a nested one.

class AlbumTwin < Disposable::Twin
  feature Parent

  property :artist do
    property :name
  end
end

Use parent to grab the nested's container twin.

twin = AlbumTwin.new(Album.new(artist: Artist.new))

twin.artist.parent #=> twin

Note that this will internally add a parent property.

Builders

Used In

Reform forms are based on twins and add a little bit of form decoration on top. Every nested form is a twin.
Trailblazer uses twins as decorators and callbacks in operations to structure business logic.

swamp09/disposable