If you’re developing an application in Ruby on Rails framework, there are many situations when you should consider using tableless models. The advantages you’d get are:

• consistent application structure, because you’re using models to represent objects in your app
• routes available for free if you also define controllers dedicated for those models, resulting in RESTful application
• easy validation (just like with normal models) and other goodies shipped with Active Record
• easier testing with unit tests
• form building as easy as with normal models

If you’re familiar with “fat model, skinny controller” concept you’ll find more reasons.

An example situation: you give the users an ability to recommends object they find on the site to their friends. They can recommend photos and articles. You do not want to track those recommendations in database. Your model file would look like this.

class Recommendation < ActiveRecord::Base
  class_inheritable_accessor :columns
  self.columns = []
 
  def self.column(name, sql_type = nil, default = nil, null = true)
    columns << ActiveRecord::ConnectionAdapters::Column.new(name.to_s, default, sql_type.to_s, null)
  end
 
  column :recommendable_type, :string
  column :recommendable_id, :integer
  column :email, :string
  column :body, :text
 
  belongs_to :recommendable, :polymorphic => true
 
  validates_presence_of :recommendable
  validates_associated :recommendable
  validates_format_of :email, :with => /^$|^\S+\@(\[?)[a-zA-Z0-9\-\.]+\.([a-zA-Z]{2,4}|[0-9]{1,4})(\]?)$/ix
  validates_presence_of :body
end

As you can see the only difference in model definition is that you need to provide the columns yourself.

The idea was borrowed from this snippet.

In the next article I’ll cover polymorphic controllers which will help to convert above example into complete one.

[edit: Thanks to Curran Schiefelbein for making this work when tableless model is inherited by another model.]