module Sequel::Plugins::ClassTableInheritance

  1. lib/sequel/plugins/class_table_inheritance.rb


The class_table_inheritance plugin uses the single_table_inheritance plugin, so it supports all of the single_table_inheritance features, but it additionally supports subclasses that have additional columns, which are stored in a separate table with a key referencing the primary table.


For example, with this hierarchy:

   /        \
Staff     Manager
  |          |
Cook      Executive

the following database schema may be used (table - columns):


id, name, kind


id, manager_id


id, num_staff


id, num_managers

The class_table_inheritance plugin assumes that the root table (e.g. employees) has a primary key column (usually autoincrementing), and all other tables have a foreign key of the same name that points to the same column in their superclass's table, which is also the primary key for that table. In this example, the employees table has an id column is a primary key and the id column in every other table is a foreign key referencing, which is also the primary key of that table.

Additionally, note that other than the primary key column, no subclass table has a column with the same name as any superclass table. This plugin does not support cases where the column names in a subclass table overlap with any column names in a superclass table.

In this example the staff table also stores Cook model objects and the executives table also stores CEO model objects.

When using the class_table_inheritance plugin, subclasses that have additional columns use joined datasets:

# SELECT * FROM employees

# SELECT,, employees.kind,
#        managers.num_staff
# FROM employees
# JOIN managers ON ( =

# SELECT,, employees.kind,
#        managers.num_staff, executives.num_managers
# FROM employees
# JOIN managers ON ( =
# JOIN executives ON ( =
# WHERE (employees.kind IN ('CEO'))

This allows CEO.all to return instances with all attributes loaded. The plugin overrides the deleting, inserting, and updating in the model to work with multiple tables, by handling each table individually.

Subclass loading

When model objects are retrieved for a superclass the result can contain subclass instances that only have column entries for the columns in the superclass table. Calling the column method on the subclass instance for a column not in the superclass table will cause a query to the database to get the value for that column. If the subclass instance was retreived using Dataset#all, the query to the database will attempt to load the column values for all subclass instances that were retrieved. For example:

a = Employee.all # [<#Staff>, <#Manager>, <#Executive>]
a.first.values # {:id=>1, name=>'S', :kind=>'Staff'}
a.first.manager_id # Loads the manager_id attribute from the database

If you want to get all columns in a subclass instance after loading via the superclass, call Model#refresh.

a = Employee.first
a.values # {:id=>1, name=>'S', :kind=>'CEO'}
a.refresh.values # {:id=>1, name=>'S', :kind=>'CEO', :num_staff=>4, :num_managers=>2}

You can also load directly from a subclass:

a = Executive.first
a.values # {:id=>1, name=>'S', :kind=>'Executive', :num_staff=>4, :num_managers=>2}

Note that when loading from a subclass, because the subclass dataset uses a JOIN, if you are referencing the primary key column, you need to disambiguate the reference by explicitly qualifying it:

a = Executive.where(:id=>1).first # database error
a = Executive.where(:executives__id=>1).first # no error


# Use the default of storing the class name in the sti_key
# column (:kind in this case)
class Employee < Sequel::Model
  plugin :class_table_inheritance, :key=>:kind

# Have subclasses inherit from the appropriate class
class Staff < Employee; end    # uses staff table
class Cook < Staff; end        # cooks table doesn't exist so uses staff table
class Manager < Employee; end  # uses managers table
class Executive < Manager; end # uses executives table
class CEO < Executive; end     # ceos table doesn't exist so uses executives table

# Some examples of using these options:

# Specifying the tables with a :table_map hash
Employee.plugin :class_table_inheritance,
  :table_map=>{:Employee  => :employees,
               :Staff     => :staff,
               :Cook      => :staff,
               :Manager   => :managers,
               :Executive => :executives,
               :CEO       => :executives }

# Using integers to store the class type, with a :model_map hash
# and an sti_key of :type
Employee.plugin :class_table_inheritance, :type,
  :model_map=>{1=>:Staff, 2=>:Cook, 3=>:Manager, 4=>:Executive, 5=>:CEO}

# Using non-class name strings
Employee.plugin :class_table_inheritance, :key=>:type,
  :model_map=>{'staff'=>:Staff, 'cook staff'=>:Cook, 'supervisor'=>:Manager}

# By default the plugin sets the respective column value
# when a new instance is created.
Cook.create.type == 'cook staff'
Manager.create.type == 'supervisor'

# You can customize this behavior with the :key_chooser option.
# This is most useful when using a non-bijective mapping.
Employee.plugin :class_table_inheritance, :key=>:type,
  :model_map=>{'cook staff'=>:Cook, 'supervisor'=>:Manager},
  :key_chooser=>proc{|instance| instance.model.sti_key_map[instance.model.to_s].first || 'stranger' }

# Using custom procs, with :model_map taking column values
# and yielding either a class, string, symbol, or nil,
# and :key_map taking a class object and returning the column
# value to use
Employee.plugin :single_table_inheritance, :key=>:type,
  :model_map=>proc{|v| v.reverse},

# You can use the same class for multiple values.
# This is mainly useful when the sti_key column contains multiple values
# which are different but do not require different code.
Employee.plugin :single_table_inheritance, :key=>:type,
  :model_map=>{'staff' => "Staff",
               'manager' => "Manager",
               'overpayed staff' => "Staff",
               'underpayed staff' => "Staff"}

One minor issue to note is that if you specify the :key_map option as a hash, instead of having it inferred from the :model_map, you should only use class name strings as keys, you should not use symbols as keys.


Public Class

  1. apply
  2. configure

Public Class methods

apply (model, opts = OPTS)

The class_table_inheritance plugin requires the single_table_inheritance plugin and the lazy_attributes plugin to handle lazily-loaded attributes for subclass instances returned by superclass methods.

[show source]
# File lib/sequel/plugins/class_table_inheritance.rb, line 175
def self.apply(model, opts = OPTS)
  model.plugin :single_table_inheritance, nil
  model.plugin :lazy_attributes
configure (model, opts = OPTS)

Initialize the plugin using the following options:


Column symbol that holds the key that identifies the class to use. Necessary if you want to call model methods on a superclass that return subclass instances


Hash or proc mapping the key column values to model class names.


Hash or proc mapping model class names to key column values. Each value or return is an array of possible key column values.


proc returning key for the provided model instance


Hash with class name symbols keys mapping to table name symbol values Overrides implicit table names

[show source]
# File lib/sequel/plugins/class_table_inheritance.rb, line 190
def self.configure(model, opts = OPTS)
  SingleTableInheritance.configure model, opts[:key], opts

  model.instance_eval do
    @cti_models = [self]
    @cti_tables = [table_name]
    @cti_instance_dataset = @instance_dataset
    @cti_table_columns = columns
    @cti_table_map = opts[:table_map] || {}