GroupedScope: Has Many Associations IN (GROUPS)

GroupedScope provides an easy way to group objects and to allow those groups to share association collections via existing has_many relationships. You may enjoy my original article titled Jack has_many :things.

Install the gem with bundler. We follow a semantic versioning format that tracks ActiveRecord's minor version. So this means to use the latest 3.2.x version of GroupedScope with any ActiveRecord 3.2 version.

gem 'grouped_scope', '~> 3.2.0'

To use GroupedScope on a model it must have a :group_id column.

class AddGroupId < ActiveRecord::Migration
  def up
    add_column :employees, :group_id, :integer
  end
  def down
    remove_column :employees, :group_id
  end
end

Assume the following model.

class Employee < ActiveRecord::Base
  has_many :reports
  grouped_scope :reports
end

By calling grouped_scope on any association you create a new group accessor for each instance. The object returned will act just like an array and at least include the current object that called it.

@employee_one.group   # => [#<Employee id: 1, group_id: nil>]

To group resources, just assign the same :group_id to each record in that group.

@employee_one.update_attribute :group_id, 1
@employee_two.update_attribute :group_id, 1
@employee_one.group   # => [#<Employee id: 1, group_id: 1>, #<Employee id: 2, group_id: 1>]

Calling grouped_scope on the :reports association leaves the existing association intact.

@employee_one.reports  # => [#<Report id: 2, employee_id: 1>]
@employee_two.reports  # => [#<Report id: 18, employee_id: 2>, #<Report id: 36, employee_id: 2>]

Now the good part, all associations passed to the grouped_scope method can be called on the group proxy. The collection will return resources shared by the group.

@employee_one.group.reports # => [#<Report id: 2, employee_id: 1>, 
                                  #<Report id: 18, employee_id: 2>, 
                                  #<Report id: 36, employee_id: 2>]

You can even call scopes or association extensions defined on the objects in the collection defined on the original association. For instance:

@employee.group.reports.urgent.assigned_to(user)

The group scoped object can respond to either blank? or present? which checks the group's target group_id presence or not. We use this internally so that grouped scopes only use grouping SQL when absolutely needed.

@employee_one = Employee.create :group_id => nil
@employee_two = Employee.create :group_id => 38

@employee_one.group.blank?   # => true
@employee_two.group.present? # => true

The object returned by the #group method is an ActiveRecord relation on the targets class, in this case Employee. Given this, you can further scope the grouped proxy if needed. Below, we use the :email_present scope to refine the group down.

class Employee < ActiveRecord::Base
  has_many :reports
  grouped_scope :reports
  scope :email_present, where("email IS NOT NULL")
end

@employee_one = Employee.create :group_id => 5, :name => 'Ken'
@employee_two = Employee.create :group_id => 5, :name => 'MetaSkills', :email => 'ken@metaskills.net'

# Only one employee is returned now.
@employee_one.group.email_present # => [#<Employee id: 1, group_id: 5, name: 'MetaSkills', email: 'ken@metaskills.net']

We always use raw SQL to get the group ids vs. mapping them to an array and using those in scopes. This means that large groups can avoid pushing down hundreds of keys in SQL form. So given an employee with a group_id of 43 and calling @employee.group.reports, you would get something similar to the following SQL.

SELECT "reports".* 
FROM "reports"  
WHERE "reports"."employee_id" IN (
  SELECT "employees"."id" 
  FROM "employees"  
  WHERE "employees"."group_id" = 43
)

You can pass the group scoped object as a predicate to ActiveRecord's relation interface. In past versions, this would have treated the group object as an array of IDs. The new behavior is to return a SQL literal to be used with IN statements. So note, the following would generate SQL similar to the one above.

Employee.where(:group_id => @employee.group).all

If you need more control and you are working with the group at a lower level, you can always use the #ids or #ids_sql methods on the group.

# Returns primary key array.
@employee.group.ids # => [33, 58, 240]

# Returns a Arel::Nodes::SqlLiteral object.
@employee.group.ids_sql # => 'SELECT "employees"."id" FROM "employees"  WHERE "employees"."group_id" = 33'

• Raise errors for :finder_sql/:counter_sql.
• Add a user definable group_id schema.
• Remove SelfGrouping#with_relation, has not yet proved useful.

Simple! Just clone the repo, then run bundle install and bundle exec rake. The tests will begin to run. We also use Travis CI to run our tests too. Current build status is:

Released under the MIT license. Copyright (c) 2011 Ken Collins