Pattern

A collection of lightweight, standardized, rails-oriented patterns used by RubyOnRails Developers @ Selleo

• Query - complex querying on active record relation
• Service - useful for handling processes involving multiple steps
• Collection - when in need to add a method that relates to the collection as whole
• Form - when you need a place for callbacks, want to replace strong parameters or handle virtual/composite resources
• Calculation - when you need a place for calculating a simple value (numeric, array, hash) and/or cache it
• Rule and Ruleset - when you need a place for conditional logic

# Gemfile

#...
gem "rails-patterns"
#...

Then bundle install

One should consider using query objects pattern when in need to perform complex querying on active record relation. Usually one should avoid using scopes for such purpose. As a rule of thumb, if scope interacts with more than one column and/or joins in other tables, it should be moved to query object. Also whenever a chain of scopes is to be used, one should consider using query object too. Some more information on using query objects can be found in this article.

• Query objects are always used by calling class-level .call method
• Query objects require ActiveRecord::Relation or ActiveRecord::Base as constructor argument
• Default relation (see above) can be defined by using queries macro
• Query objects have to implement #query method that returns ActiveRecord::Relation
• Query objects provide access to consecutive keyword arguments using #options hash

Because of the fact, that QueryObject implements .call method, those can be used to construct scopes if required. (read more...)

class RecentlyActivatedUsersQuery < Patterns::Query
  queries User

  private

  def query
    relation.active.where(activated_at: date_range)
  end

  def date_range
    options.fetch(:date_range, default_date_range)
  end

  def default_date_range
    Date.yesterday.beginning_of_day..Date.today.end_of_day
  end
end

RecentlyActivatedUsersQuery.call
RecentlyActivatedUsersQuery.call(User.without_test_users)
RecentlyActivatedUsersQuery.call(date_range: Date.today.beginning_of_day..Date.today.end_of_day)
RecentlyActivatedUsersQuery.call(User.without_test_users, date_range: Date.today.beginning_of_day..Date.today.end_of_day)

class User < ApplicationRecord
  scope :recently_activated, RecentlyActivatedUsersQuery
end

Service objects are commonly used to mitigate problems with model callbacks that interact with external classes (read more...). Service objects are also useful for handling processes involving multiple steps. E.g. a controller that performs more than one operation on its subject (usually a model instance) is a possible candidate for Extract ServiceObject (or Extract FormObject) refactoring. In many cases service object can be used as scaffolding for replace method with object refactoring. Some more information on using services can be found in this article.

• Service objects are always used by calling class-level .call method
• Service objects have to implement #call method
• Calling service object's .call method executes #call and returns service object instance
• A result of #call method is accessible through #result method
• It is recommended for #call method to be the only public method of service object (besides state readers)
• It is recommended to name service object classes after commands (e.g. ActivateUser instead of UserActivation)

A bit higher level of abstraction is provided by business_process gem.

class ActivateUser < Patterns::Service
  def initialize(user)
    @user = user
  end

  def call
    user.activate!
    NotificationsMailer.user_activation_notification(user).deliver_now
    user
  end

  private

  attr_reader :user
end

  user_activation = ActivateUser.call(user)
  user_activation.result # <User id: 5803143, email: "tony@patterns.dev ...

One should consider using collection pattern when in need to add a method that relates to the collection a whole. Popular example for such situation is for paginated collections, where for instance #current_page getter makes sense only in collection context. Also collections can be used as a container for mapping or grouping logic (especially if the mapping is not 1-1 in terms of size). Collection might also act as a replacement for models not inheriting from ActiveRecord::Base (e.g. StatusesCollection, ColorsCollection etc.). What is more, collections can be used if we need to encapsulate "flagging" logic - for instance if we need to render a separator element between collection elements based on some specific logic, we can move this logic from view layer to collection and yield an additional flag to control rendering in view.

• Collections include Enumerable
• Collections can be initialized using .new, .from and .for (aliases)
• Collections have to implement #collection method that returns object responding to #each
• Collections provide access to consecutive keyword arguments using #options hash
• Collections provide access to first argument using #subject

class ColorsCollection < Patterns::Collection
  AVAILABLE_COLORS = { red: "#FF0000", green: "#00FF00", blue: "#0000FF" }

  private

  def collection
    AVAILABLE_COLORS
  end
end

class CustomerEventsByTypeCollection < Patterns::Collection
  private

  def collection
    subject.
    events.
    group_by(&:type).
    transform_values{ |events| events.map{ |e| e.public_send(options.fetch(:label_method, "description")) }}
  end
end

ColorsCollection.new
CustomerEventsByTypeCollection.for(customer)
CustomerEventsByTypeCollection.for(customer, label_method: "name")

Form objects, just like service objects, are commonly used to mitigate problems with model callbacks that interact with external classes (read more...). Form objects can also be used as replacement for ActionController::StrongParameters strategy, as all writable attributes are re-defined within each form. Finally form objects can be used as wrappers for virtual (with no model representation) or composite (saving multiple models at once) resources. In the latter case this may act as replacement for ActiveRecord::NestedAttributes. In some cases FormObject can be used as scaffolding for replace method with object refactoring. Some more information on using form objects can be found in this article.

• Forms include ActiveModel::Validations to support validation.
• Forms include Virtus.model to support attribute static method with all corresponding capabilities.
• Forms can be initialized using .new.
• Forms accept optional resource object as first constructor argument.
• Forms accept optional attributes hash as latter constructor argument.
• Forms have to implement #persist method that returns falsey (if failed) or truthy (if succeeded) value.
• Forms provide access to first constructor argument using #resource.
• Forms are saved using their #save or #save! methods.
• Forms will attempt to pre-populate their fields using resource#attributes and public getters for resource
• Form's fields are populated with passed-in attributes hash reverse-merged with pre-populated attributes if possible.
• Forms provide #as builder method that populates internal @form_owner variable (can be used to store current user).
• Forms allow defining/overriding their #param_key method result by using .param_key static method. This defaults to #resource#model_name#param_key.
• Forms delegate #persisted? method to #resource if possible.
• Forms do handle ActionController::Parameters as attributes hash (using to_unsafe_h)
• It is recommended to wrap #persist method in transaction if possible and if multiple model are affected.

class UserForm < Patterns::Form
  param_key "person"

  attribute :first_name, String
  attribute :last_name, String
  attribute :age, Integer
  attribute :full_address, String
  attribute :skip_notification, Boolean

  validate :first_name, :last_name, presence: true

  private

  def persist
    update_user and
      update_address and
      deliver_notification
  end

  def update_user
    resource.update_attributes(attributes.except(:full_address, :skip_notification))
  end

  def update_address
    resource.address.update_attributes(full_address: full_address)
  end

  def deliver_notification
    skip_notification || UserNotifier.user_update_notification(user, form_owner).deliver
  end
end

class ReportConfigurationForm < Patterns::Form
  param_key "report"

  attribute :include_extra_data, Boolean
  attribute :dump_as_csv, Boolean
  attribute :comma_separated_column_names, String
  attribute :date_start, Date
  attribute :date_end, Date

  private

  def persist
    SendReport.call(attributes)
  end
end

form = UserForm.new(User.find(1), params[:person])
form.save

form = UserForm.new(User.new, params[:person]).as(current_user)
form.save!

ReportConfigurationForm.new
ReportConfigurationForm.new({ include_extra_data: true, dump_as_csv: true })

Calculation objects provide a place to calculate simple values (i.e. numeric, arrays, hashes), especially when calculations require interacting with multiple classes, and thus do not fit into any particular one. Calculation objects also provide simple abstraction for caching their results.

• Calculations have to implement #result method that returns any value (result of calculation).
• Calculations do provide .set_cache_expiry_every method, that allows defining caching period.
• When .set_cache_expiry_every is not used, result is not being cached.
• Calculations return result by calling any of following methods: .calculate, .result_for or .result.
• First argument passed to calculation is accessible by #subject private method.
• Arguments hash passed to calculation is accessible by #options private method.
• Caching takes into account arguments passed when building cache key.
• To build cache key, #cache_key of each argument value is used if possible.
• By default Rails.cache is used as cache store.

class AverageHotelDailyRevenue < Patterns::Calculation
  set_cache_expiry_every 1.day

  private

  def result
    reservations.sum(:price) / days_in_year
  end

  def reservations
    Reservation.where(
      date: (beginning_of_year..end_of_year),
      hotel_id: subject.id
    )
  end

  def days_in_year
    end_of_year.yday
  end

  def year
    options.fetch(:year, Date.current.year)
  end

  def beginning_of_year
    Date.new(year).beginning_of_year
  end

  def end_of_year
    Date.new(year).end_of_year
  end
end

hotel = Hotel.find(123)
AverageHotelDailyRevenue.result_for(hotel)
AverageHotelDailyRevenue.result_for(hotel, year: 2015)

TotalCurrentRevenue.calculate
AverageDailyRevenue.result

Rule objects provide a place for dislocating/extracting conditional logic.

Use it when:

• given complex condition is duplicated in multiple places in your codebase
• part of condition logic can be reused in some other place
• there is a need to instantiate condition itself for some reason (i.e. to represent it in the interface)
• responsibility of your class is blurred by complex conditional logic, and as a result...
• ...tests for your class require multiple condition branches / nested contexts

• Rule has #satisfied?, #applicable?, #not_applicable? and #forceable? methods available.
• Rule has to implement at least #satisfied? method. #not_applicable? and #forceable? are meant to be overridable.
• #forceable? makes sense in scenario where condition is capable of being force-satisfied regardless if its actually satisfied or not. Is true by default.
• Override #not_applicable? when method is applicable only under some specific conditions. Is false by default.
• Rule requires a subject as first argument.
• Multiple rules and rulesets can be combined into new ruleset as both share same interface and can be used interchangeably (composite pattern).
• By default empty ruleset is satisfied.

On some occasions there is a situation in which some condition should be overridable. Let's say we may want send shipping notification even though given order was not paid for and under regular circumstances such notification should not be sent. In this case, while regular logic with some automated process would not trigger delivery, an action triggered by user from UI could do it, by passing force: true option to #satisified? methods.

It might be good idea to test for #forceable? on the UI level to control visibility of such link/button.

Overriding #forceable can be useful to prevent some edge cases, i.e. ContactInformationProvidedRule might check if customer for given order has provided any contact means by which a notification could be delivered. If not, ruleset containing such rule (and the rule itself) would not be "forceable" and UI could reflect that by querying #forceable?.

While regular Ruleset can be satisfied or forced if any of its rules in not applicable, the StrongRuleset is not satisfied and not "forceable" if any of its rules is not applicable.

It might be surprising that is is the negated version of the #applicable? predicate methods that is overridable. However, from the actual usage perspective, it usually easier to conceptually define when condition makes no sense than other way around.

class OrderIsSentRule < Patterns::Rule
  def satisfied?
    subject.sent?
  end
end

class OrderIsPaidRule < Patterns::Rule
  def satisfied?
    subject.paid?
  end

  def forceable?
    true
  end
end

OrderCompletedNotificationRuleset = Class.new(Patterns::Ruleset)
OrderCompletedNotificationRuleset.
  add_rule(:order_is_sent_rule).
  add_rule(:order_is_paid_rule)

OrderIsPaidRule.new(order).satisfied?
OrderCompletedNotificationRuleset.new(order).satisfied?

ResendOrderNotification.call(order) if OrderCompletedNotificationRuleset.new(order).satisfied?(force: true)

• 7 ways to decompose fat active record models

Software development teams with an entrepreneurial sense of ownership at their core delivering great digital products and building culture people want to belong to. We are a community of engaged co-workers passionate about crafting impactful web solutions which transform the way our clients do business.

All names and logos for Selleo are trademark of Selleo Labs Sp. z o.o. (formerly Selleo Sp. z o.o. Sp.k.)