Has_one :through

mlittle · October 2, 2009, 12:18am

Another newbie question by me…

I have this:

class User < ActiveRecord::Base
…
has_one :book, :through => :library
…
end

class Library < ActiveRecord::Base
has_one :book, :dependent => :destroy
end

class Library < ActiveRecord::Base
belongs_to :library
validates_existence_of :library
end

I just don’t know what my route will look like. Any help or
suggestions? Are my models correct in how I used the has_one :through

Mike

Thanks

mlittle · October 2, 2009, 2:19am

mlittle wrote:

Another newbie question by me…

First, do yourself a favor and not start off by advertising your
“newbie” status. Trust that experienced programmers will know right away
where you are in your learning by reading your question.

Before attempting to answer this question I would like some
confirmations. Read on…

I have this:

class User < ActiveRecord::Base
…
has_one :book, :through => :library
…
end

Are you wanting to say here that a user can only check out one book at a
time from the library?

class Library < ActiveRecord::Base
has_one :book, :dependent => :destroy
end

And a library can only ever contain one book?

class Library < ActiveRecord::Base
belongs_to :library
validates_existence_of :library
end

Did you mean to define Book here instead of reopening the Library class
and adding stuff to it?

Assuming you meant Book it seems fine for a book to belong to a library.

There is no validates_existence_of. Assuming you meant
validates_presence_of :library:
I’m even a little bit torn on this one, but I validate the presence of
the foreign key (:library_id in this case) rather than validate the
association. From my experience so far it seems to make for cleaner test
scripts and also does not directly depend on the actual association
object making model tests easier to manage.

I just don’t know what my route will look like.

Once you get your model actually right your routes will make sense.

Any help or suggestions? Are my models correct in how I used the has_one :through

Keep this in mind as a general rule of thumb: The reason has_one exists
is for creating one-to-one associations. Thing A can have only one B and
thing B can only ever have one A.

So why would you ever need one-to-one associations? Can’t you just add
all the needed fields to one table? Technically the answer is “Yes you
can.” However, there are some very useful and practical uses for
one-to-one associations.

Sometimes groups of fields, which actually are functionally depend on
the primary key of a row, are related in other ways and are good
candidate to be broken out into a separate table.
Often a table may contain a set of “optional” fields that don’t need
to have values for every row in a table. These can be broken out to a
separate table for effeciency’s sake and to clean up all those NULLs you
would see if there was only one table.
There are also cases where a table may contain a large character or
binary blob. Most of the time it’s convenient to just use the wildcard
when selecting data: (select * from my_table;). If that table contained
a column of type (TEXT or BLOB). Then you’re going to suffer a
significant performance hit. That could be fixed by always specifying
the list of columns with something like: (select Col1, Col2, Coln from
my_table;), but you could also use a one-to-many and put the TEXT or
BLOB in a separate table. Then (select * from my_table;) would only
fetch the foreign key (i.e. my_blob_id), which is a simple integer and
you won’t suffer the performance hit. When you actually want to get the
large TEXT or BLOB you can either grab them individually thought the
association, or include them if you need to grab a list of objects:
@posts = Post.find(:all, :include => :post_narrative).

So why has_one and not belongs_to? Well to answer this you need to
consider how a one-to-on association is built in Rails. It is really
just like a one-to-many association with a validation that only allows
one object to be associated on the “many” side. So think of has_one as a
“special” has_many than includes the proper validation (to ensure
one-to-one) and also provides the one object instead of an array
containing the one object.

So which side gets the has_one? Well that’s easy to answer as well: use
belongs_to on the model that contains the foreign key and use has_one on
the model that doesn’t. You get to decide which model should get the
foreign key. Usually objects have a natural association that makes that
fairly easy. Say that a Person can only have one Residence (within this
design). You wouldn’t think about a Person belonging to the Residence.
You would rather say the Residence belongs to the Person. So Residence
would get the foreign key, an hence the belongs_to :person. So then
Person gets the has_one: residence.

I apologize for the long reply. Maybe this belongs_to :blog.

mlittle · October 2, 2009, 2:49am

Robert, thanks for the reply. validates_existence_of is a plugin →
GitHub - nilaccount/validates_existence_of: A small plugin to validate the existence of an AR record. I’ll try to make
this as simple as possible. Here is what i’m trying to do:

A’s has many B’s and B’s have 1 C.

A = Users
B = Books
C = Libraries

Hope this clarifies everything for you. Sorry about the mistakes in my
first post. Thanks again for your help

On Oct 1, 5:19 pm, Robert W. [email protected]

mlittle · October 2, 2009, 3:56am

mlittle wrote:

Robert, thanks for the reply. validates_existence_of is a plugin →
GitHub - nilaccount/validates_existence_of: A small plugin to validate the existence of an AR record. I’ll try to make
this as simple as possible. Here is what i’m trying to do:

A’s has many B’s and B’s have 1 C.

A = Users
B = Books
C = Libraries

Then your associations will be as follows:
class User < AR::B
has_many :books
end

class Book < AR::B
belongs_to :user
belongs_to :library
end

class Library < AR::B
has_many :books
end

Hope this clarifies everything for you. Sorry about the mistakes in my
first post. Thanks again for your help

On Oct 1, 5:19ï¿½pm, Robert W. [email protected]

Best,

Marnen Laibow-Koser
http://www.marnen.org
[email protected]

mlittle · October 2, 2009, 4:20am

mlittle wrote:

Robert, thanks for the reply. validates_existence_of is a plugin →
GitHub - nilaccount/validates_existence_of: A small plugin to validate the existence of an AR record. I’ll try to make
this as simple as possible. Here is what i’m trying to do:

Got it. I though maybe that was the case, and might explain some of my
previous frustrations with testing these associations.

A’s has many B’s and B’s have 1 C.

A = Users
B = Books
C = Libraries

Okay, That makes more sense. Let me first write this as a user story:

Users may check out many books from a library. Users may borrow from
many libraries. Only one library may own a book (in this design let’s
consider two books with the same title separate books). Books loaned to
users must be returned to the owning library.

So I see four models here not three:

User
Library
Book
Loan

Feature: Book check out
In order to learn cool stuff, or be entertained
As a user
I want to be able to checkout many books from many libraries

Scenario: Check from 2 libraries
Given I select ‘Programming Ruby’ from the Public Library
And I select ‘The Rails Way’ from the University Library
When I check out from the Public Library
And I check out from the University Library
Then I should have 2 books

Feature: Book check in
In order to prevent late fees
As a user
I want to be able to check in books

Scenario: I must return a book to the correct library
Given I have ‘Programming Ruby’
When I check in to the University Library
Then the check in should be denied
And I should have 2 books

There may be other scenarios to consider, but I wanted to focus on these
two to help describe the model:

It looks to me like you need a standard many-to-many association with
some additional validation to make sure the books get returned back to
the library that owns them.

class User < ActiveRecord::Base
has_many :loans
has_many :books, :through => :loans
end

class Loan < ActiveRecord::Base
belongs_to :user
belongs_to :book
end

class Book < ActiveRecord::Base
has_many :loans
has_many :users, :through => :loans
belongs_to :library
end

class Library < ActiveRecord::Base
has_many :books
end

loans_controller.rb

DELETE /loans/1

DELETE /loans/1.xml

def destroy
@loan = Loan.find(params[:id])
@loan.destroy if @loan.book.library == current_library

respond_to do |format|
  flash[:error] = "This book was
  format.html { redirect_to(loans_url) }
  format.xml  { head :ok }
end

end

I’ll leave it as an exercise for you to figure out how to determine the
“current library” that the user is attempting to return the book to.

Notice that in this design the act of checking in a book is to destroy
the loan of the book. So loan is a resource in it’s own right. Of course
checking out a book would be to create a new loan.

Note: this code is incomplete and untested with no warrantee of any
kind.

mlittle · October 2, 2009, 6:50am

Robert,

Thanks so much for the help. Question, though, could you help with
with the routing that would be involved?

On Oct 1, 7:20 pm, Robert W. [email protected]

mlittle · October 2, 2009, 8:08am

mlittle wrote:

Robert,

Thanks so much for the help. Question, though, could you help with
with the routing that would be involved?

On Oct 1, 7:20ï¿½pm, Robert W. [email protected]

I can’t think of any “special” routing that would be required. This
design uses all standard REST actions so you should need only the
standard resource routes:

map.resources :users, :loans, :books, :libraries

I suppose you could use some nested routes, but that would not be
required for this to work. You would need some way to let the
loans_controller know what library the book is attempting to be returned
to, but I don’t think that should affect routing in any way.