I want to have two layers of STI. The first layer is working but am unsure how to have the first model lookup to another table also using STI. Example:
class Instrument < ActiveRecord::Base
end
class Guitar < Instrument
end
class Piano < Instrument
# and so on...
Ok. But I want to track categories of instruments so I can tell their type:
class InstrumentType < ActiveRecord::Base
end
class StringInstrumentType < InstrumentType
end
# etc.
So... our Guitar model becomes:
class Guitar < Instrument
belongs_to :string_intrument_type
end
class StringInstrumentType < InstrumentType
has_many :guitars
end
How do I implement this? What do I name the foreign-key column and how do I tell Rails/AR what to do?
Am I way off base here?
This scenario is invented but we are finding some real-world refactoring wants us to move in this direction.
Your first STI makes sense, the other one does not. On the abstract level, strings are rather an instance of an InstrumentType, not its subclass. It seems to me that you might rather use sth like:
class Instrument < ActiveRecord::Base
# abstract
end
class StringInstrument < Instrument
# abstract
end
module Instrument::Keyboard
end
class Guitar < StringInstrument
end
class Piano < StringInstrument
include Instrument::Keyboard
end
or similar. Obviously it depends A LOT on your actual code. In example you gave instrument types are rather constant in time, hence it is possible to use inheritance.
My guess is however, that your types are dynamic and can be added by clients. What you are looking for then is probably a tree structure. So every type has its parent and each parent type might have couple of children. You can then simply do Instrument belongs_to instrument_type and point to some InstrumentType tree leaf.
Related
I was recently working on a project where I faced a dilemma of choosing between two ways of getting same results. Here is the class structure:
class Book < ApplicationRecord
belongs_to :author
end
class Author < ApplicationRecord
has_many :books
end
An author has first name, last name. I want to get the full name of the author for a given book as an instance method.
In simple active record terms, since book is associated with author, we can get the author name for a book as follows:
For example in Book class, we have:
class Book < ApplicationRecord
belongs_to :author
def author_name
"#{author.first_name} #{author.last_name}"
end
end
And we get the result!
But, according to the target of minimizing dependencies (POODR Book), future ease of change and better object oriented design, the book should not know properties of an author. It should interact with an author object by interfaces.
So Book should not be the one responsible for getting the Author name. The author class should.
class Book < ApplicationRecord
belongs_to :author
def author_name
get_author_name(self.author_id)
end
private
#minimizing class dependecies by providing private methods as external interfaces
def get_author_name(author_id)
Author.get_author_name_from_id(author_id)
end
end
class Author < ApplicationRecord
has_many :books
#class methods which provides a gate-way for other classes to communicate through interfaces, thus reducing coupling.
def self.get_author_name_from_id(id)
author = self.find_by_id(id)
author == nil ? "Author Record Not Found" : "#{author.first_name.titleize} #{author.last_name.titleize}"
end
end
Now, book is just interacting with the public interface provided by Author and Author is handling the responsibility of getting full name from its properties which is a better design for sure.
I tried running the queries as two separate methods in my console:
class Book < ApplicationRecord
def author_name
get_author_name(self.author_id)
end
def author_name2
"#{author.last_name} + #{author.first_name}"
end
end
The results are shown below:
Looks like both run the same queries.
My questions are
Does rails convert author.last_name called inside the Book class to
the same SQL query as Author.find_by_id(author_id).last_name called inside
Author class (through message passing from Book class) in case of bigger data size?
Which one is more performant in case of bigger data size?
Doesn't calling author.last_name from Book class violates design
principles ?
It's actually much more common and simplier to use delegation.
class Book < ApplicationRecord
belongs_to :author
delegate :name, to: :author, prefix: true, allow_nil: true
end
class Author < ApplicationRecord
has_many :books
def name
"#{first_name.titleize} #(last_name.titleize}"
end
end
As to performance, if you join the authors at the time of the book query you end up doing a single query.
#books = Book.joins(:author)
Now when you iterate through #books and you call individually book.author_name no SQL query needs to be made to the authors table.
1) Obviously not, it performs JOIN of books & authors tables. What you've made requires 2 queries, instead of 1 join you'll have book.find(id) and author.find(book.author_id).
2) JOIN should be faster.
3) Since last_name is a public interface, it absolutely doesn't violate design principles. It would violate principles if you were accessing author's last name from outside like that: Book.find(1).author.last_name - that's a bad thing. Correct is: Book.find(1).authors_last_name - and accessing author's name inside Model class.
Your provided example seems to be overcomplicated to me.
According to the example you shared, you only want to get full name of the book's author. So, the idea of splitting responsibility is correct, but in Author class should be simple instance method full_name, like:
class Author < ApplicationRecord
has_many :books
def full_name
"#{author.first_name.titleize} #{author.last_name.titleize}"
end
end
class Book < ActiveRecord::Base
belongs_to :author
def author_name
author.full_name
end
end
Note, there're no direct queries in this code. Once you'll need the author's name somewhere (in a view, in api response, etc), Rails will make the most optimized query possible (depends on your use case though, it may be ineffective for example, if you call iterate over books and call author in a loop)
I prefer the second approach because the full_name is property of author not a book. If the book wants to access that information, it can using book.author&.full_name (& is for handling cases of books with no authors).
but I would suggest a refactoring as below:
class Book < ApplicationRecord
belongs_to :author
end
class Author < ApplicationRecord
has_many :books
def full_name
"#{firstname} #{lastname}"
end
end
Does rails convert author.last_name called inside the Book class to the same SQL query as Author.find_by_id(author_id).last_name called inside Author class (through message passing from Book class) in case of bigger data size?
Depend upon the calling factor, like in your example both will generate the same query. But if you have a include\join clause while getting the Book/Author, both will generate different queries.
As per the rails convention, Author.find_by_id(author_id).last_name is not recommended as it will always fire a query on database whenever the method is called. One should use the rails' association interface to call the method on related object which is smart to identify the object from memory or fetch it from database if not in memory.
Which one is more performant in case of bigger data size?
author.last_name is better because it will take care of joins, include, and memoization clauses if used and avoid the N+1 query problem.
Doesn't calling author.last_name from Book class violates design principles?
No, you can even use delegate like #Steve Suggested.
In my experience, it's a balancing act between minimizing code complexity and minimizing scalability issues.
However, in this case, I think the simplest solution that would separate class concerns and minimize code would be to simply use: #book.author.full_name
And in your Author.rb define full_name in Author.rb:
def full_name
"#{self.first_name} #{self.last_name}"
end
This will simplify your code a lot. For example, if in the future you had another model called Magazine that has an Author, you don't have to go define author_name in the Magazine model as well. You simply use #magazine.author.full_name. This will DRY up your code nicely.
I have a set up where I have multiple models inheriting from a Base model - standard Single Table Inheritance:
class Parent < ActiveRecord::Base
end
class A < Parent
end
class B < Parent
end
My STI setup is correct and works great! However, I want to add :type specific attributes such as description.
For example, I want all A types of Parent to have the description, "I am the A type of Parent. My function is for..."
I want to avoid replicating data over and over (having each instance of A store the same description for example).
The first thing that came to mind for this was to have a model specific method on the Subclass. So something like:
class A < Parent
def self.description
"I am the A type of Parent. My function is for..."
end
end
I don't like this solution because this really is data on the specific type of subclass (rather than on the subclass instance itself) and you get all the problems that come with making this behavior (deployments to change data, etc.)
Is this the only way to do it or are there alternatives I just am not seeing?
What about creating a model for the description?
class Description < ActiveRecord::Base
has_many :as
end
class A < Parent
belongs_to :description
before_save { description_id = 1 }
end
This way, you manage the content of description in the database and can modify it through either a web interface or migrations. Furthermore, you can easily add different descriptions for the different subclasses, or even change them per instance if that is ever required.
One downside of this approach is that you need to create the model with the correct description. One potential solution could be the before_save or before_create hook, but I'm sure those are not the only way to do it.
for your case I prefer to use ruby Duck typing as follow
class ParentAll
def talk(object1)
object1.talk
end
end
class A < ParentAll
def talk
puts 'I am the A type of Parent. My function is for...'
end
end
class B < ParentAll
def talk
puts 'I am the B type of Parent. My function is for...'
end
end
#parent = ParentAll.new
puts 'Using the A'
#parent.talk(A.new)
# this will output talk from A
puts 'Using the B'
#parent.talk(B.new)
# this will output talk from B
I have an application I'm building where I need one model to create instances of another model. I want every Car to have 4 tires.
Car model
class Car < ActiveRecord::Base
has_many :tires
after_create :make_tires
def make_tires
4.times { Tire.create(car: self.id) }
end
end
Tire model
class Tire < ActiveRecord::Base
belongs_to :car
end
However, inside of make_tires there is an error that there is no activerecord method for create or new if I try it for Tire. When I inspect Tire it doesn't have those methods.
How can I remedy this?
The error is this: undefined method 'create' for ActiveRecord::AttributeMethods::Serialization::Tire::Module
I have tested two environments: Testing and Development and they both fail for the same error.
It is a name conflict. Sit down and relax while I explain.
Solution with explanation:
In Ruby classes are just instances of class Class (which is a subclass of class Module). Instances of Module (including instances of Class) are quite weird objects, especially weird is their connection with ruby constants. You can create a new class at any point using standard ruby notation:
my_class = Class.new { attr_accessor :a }
instance = my_class.new
instance.a = 3
insatnce.a #=>
instance.class.name #=> nil
Well, our class has no name. It is just an anonymous class. How do classes get their name? By assigning it to a constant (for the first time):
MyClass = my_class
my_class.name #=> 'MyClass'
When you define class using a class keyword:
class MyClass
...
end
You just create a new instance of Class and assign it to a constant. Because of that, Ruby compiler seeing a constant has no idea whether it is a class or a number under it - it has to make a full search for that constant.
The logic behind finding a constant is quite complex and depends on the current nesting. Your case is quite simple (as there is no nesting), so ruby will try to find Tire class inside your class first and when failed it's subclasses and included modules.
Your problem is that your class inherits from ActiveRecord::Base (which is correct), which includes ActiveRecord::AttributeMethods::Serialization module, which defines Tire constant already. Hence, ruby will use this constant instead, as this is the best match for that name in given context.
To fix it, you must tell the compiler not to look within the current class but directly in the "top namespace" (which in ruby is Object. Seriously, try Object.constants) - you can do that using :: in front of your constant, like ::Tire.
Note: even though it works, this issue is a first warning for you that your code starts to smell. You should look after this ActiveRecord::AttributeMethods::Serialization::Tire::Module thingy as it seems you will encounter it more than once in the future.
Other stuff:
You can simplify your method slightly:
def make_tires
4.times { tires.create }
end
At that point you might encounter some error you had initially. If you do, then please find what is going on with that Tire::Module thing. If you don't care about the smell:
has_many :tires, class_name: '::Tire'
I'm not sure what's causing the exception you are seeing but you have a number of issues. First, you need to pass in a car instance instead of the id in make_tires. Like this:
def make_tires
4.times { Tire.create(car: self) }
end
You also need to have attr_accessible :car in the Tire model. Like this:
class Tire < ActiveRecord::Base
belongs_to :car
attr_accessible :car
end
I'm trying to create a mixin that allows an ActiveRecord model to act as a delegate for another model. So, doing it the typical way:
class Apple < ActiveRecord::Base
def foo_species
"Red delicious"
end
end
class AppleWrapper < ActiveRecord::Base
belongs_to :apple
# some meta delegation code here so that AppleWrapper
# has all the same interface as Apple
end
a = Apple.create
w = AppleWrapper.create
w.apple = a
w.foo_species
# => 'Red delicious'
What I want is to abstract this behavior into a Mixin, so that given a bunch of data models, I can create "Wrapper" classes that are also ActiveRecords, but that each wrapper corresponds to a specific class. Why? Each of the data models have calculations, aggregations with other models, and I want the "Wrapper" classes to contain fields (in the schema) that correspond to these calculations...so in effect. the Wrapper acts as a cached version of the original data model, with the same interface.
I will have to write out each Wrapper...so for Apple, Orange, Pear, there is a different Wrapper model for each of them. However, I just want to abstract out the wrapper behavior...so that there's a class level method that sets what the Wrapper points to, a la:
module WrapperMixin
extend ActiveSupport::Concern
module ClassMethods
def set_wrapped_class(klass)
# this sets the relation to another data model and does the meta delegation
end
end
end
class AppleWrapper < ActiveRecord::Base
include WrapperMixin
set_wrapped_class Apple
end
class OrangeWrapper < ActiveRecord::Base
include WrapperMixin
set_wrapped_class Orange
end
How would I set this up? And would this have to be a STI type relation? That is, does the Wrapper class have to have a wrapped_record_id and a wrapped_record_type?
You can use belongs_to in your set_wrapped_class method.
def set_wrapped_class(klass)
belongs_to klass.to_s.downcase.to_sym
end
Short version: Where should I store environment-specific IDs? ENV['some-variable']? Somewhere else?
Long version:
Let's say I have a model called Books and a book has a Category. (For the sake of this question, let's say a book only has one category.)
class Book < ActiveRecord::Base
belongs_to :category
end
class Category < ActiveRecord::Base
has_many :books
end
Now let's say one category is called 'erotica.' And I want to suppress erotica books in my type ahead. That seems straight forward. But in production and in development 'erotica' has a different ID. I don't want my code to be ID dependent. I don't want it to be string dependent (in case 'erotica' is renamed pr0n or whatever).
I think I should have something like
def suppress_method
suppress_category_id = look_up_suppression_id
...
end
but where should 'look up' look?
Thanks!
Taking this approach will be brittle, what if you want to suppress multiple categories? Erotica and Politics? The best design here is for you to actually add 'suppressed' as a boolean to category in a migration, and maintain that in your application's administration interface. After you've done that you can add a named scope like:
class Category < ActiveRecord::Base
named_scope :not_suppressed, :conditions=>{:suppressed=>false}
# or for rails 3
scope :not_suppressed, where(:suppressed=>false)
end
Then just update your type ahead code to do:
Category.not_suppressed.find ...
Rather than
Category.find