Breakdowns in has_many abstraction


#1

I discovered an interesting aspect of has_many behavior that I’m
struggling to work around. I’m not sure if I’m doing something wrong,
or if it’s a legitimate bug, or if it’s an inherent part of Rails that
I just have to learn to deal with.

It boils down to these two problems:

  • changes in collection objects (i.e. models that belong_to a
    container model) don’t propagate to the container object’s collection
    view until the collection objects are saved and the association
    reloaded
  • the collection= method doesn’t do anything when you give it updated
    objects

I encountered these problems when I was updating both the container
object and some or all of the collection objects at the same time, and
I wanted to use the container.collection before everything is saved
(i.e. for collective validation).

The first problem comes from the fact that
container.collection.find(id) returns a completely different Ruby
object than the objects contained in the container.collection array.
When you modify the object returned by find, those changes don’t show
up in the container.collection array until the modified objects are
saved and reloaded from the database. How could these be better
synced up?

I tried using container.collection= to overwrite the ruby objects in
the collection with those that were generated by find, but that
doesn’t work, because of the second problem. The collection= method
ignores objects that match the id of an object already in the
collection! So even if the objects you pass in to
container.collection= are completely different except for their ids,
the collection remains unchanged. So that’s a weird behavior, too.
Why is it like that?

Here’s my simplified test case, part of a new Web 2.0 app, Rain’d.
:slight_smile: The umbrellas are all named, and we want to make sure that each
person has uniquely-named umbrellas.

class Person < ActiveRecord::Base
has_many :umbrellas

def validate
  errors.add("umbrellas", "must be uniquely named") unless

no_duplicate_names?
end

def no_duplicate_names?
  names = umbrellas.collect {|u| u.name}
  names.length == names.uniq.length
end

end

class Umbrella < ActiveRecord::Base
belongs_to :person
end

In people_controller.rb (note the comment in the middle):

def update
@person = Person.find(params[:person][:id], :include=>:umbrellas)
@person.attributes = params[:person]
@umbrellas = params[:umbrellas].values.collect do |u|
umbrella = u[:id] ? @person.umbrellas.find(u[:id]):
@person.umbrellas.build(u)
umbrella.attributes = u
umbrella
end
begin
Person.transaction do
@umbrellas.each {|u| u.save!}

    # in order for the validation to work, you have to save the

umbrellas first,
# then call this reloading trick here:
@person.umbrellas(true)

    @person.save!    # now this will validate
  end
  flash[:notice] = 'Person was successfully updated.'
  redirect_to :action => 'show', :id=>@person.id
rescue => exception
  flash[:notice] = exception.to_s
  render_scaffold('edit')
end

end

Here’s a transcript showing the second problem, from a breakpoint I
set in the update method. It illustrates some of the weirdness that’s
going on here.

@umbrellas = @person.umbrellas
=> [#<Umbrella:0x38d7218 @attributes={“name”=>“U. Horatio”, “id”=>“1”,
“person_i
d”=>“1”}>, #<Umbrella:0x38d5ca0 @attributes={“name”=>“U. Glavellus”,
“id”=>“2”,
“person_id”=>“1”}>]

@umbrellas.each{|u| u.name = u.id.to_s}
=> [#<Umbrella:0x38d7218 @attributes={“name”=>“1”, “id”=>“1”,
“person_id”=>“1”}>
, #<Umbrella:0x38d5ca0 @attributes={“name”=>“2”, “id”=>“2”,
“person_id”=>“1”}>]

@person.umbrellas
=> [#<Umbrella:0x38d7218 @attributes={“name”=>“1”, “id”=>“1”,
“person_id”=>“1”}>
, #<Umbrella:0x38d5ca0 @attributes={“name”=>“2”, “id”=>“2”,
“person_id”=>“1”}>]

@umbrellas = [@person.umbrellas.find(1), @person.umbrellas.find(2)]
=> [#<Umbrella:0x389b578 @attributes={“name”=>“U. Horatio”, “id”=>“1”,
“person_i
d”=>“1”}>, #<Umbrella:0x38977b0 @attributes={“name”=>“U. Glavellus”,
“id”=>“2”,
“person_id”=>“1”}>]

@person.umbrellas
=> [#<Umbrella:0x38d7218 @attributes={“name”=>“1”, “id”=>“1”,
“person_id”=>“1”}>
, #<Umbrella:0x38d5ca0 @attributes={“name”=>“2”, “id”=>“2”,
“person_id”=>“1”}>]

@person.umbrellas = @umbrellas
=> [#<Umbrella:0x389b578 @attributes={“name”=>“U. Horatio”, “id”=>“1”,
“person_i
d”=>“1”}>, #<Umbrella:0x38977b0 @attributes={“name”=>“U. Glavellus”,
“id”=>“2”,
“person_id”=>“1”}>]

@person.umbrellas
=> [#<Umbrella:0x38d7218 @attributes={“name”=>“1”, “id”=>“1”,
“person_id”=>“1”}>
, #<Umbrella:0x38d5ca0 @attributes={“name”=>“2”, “id”=>“2”,
“person_id”=>“1”}>]

Sorry for all the text, and thanks if you’ve gotten this far. It
seems like this is a relatively simple thing to want to accomplish,
and I’m surprised that in doing so I’ve run into a bunch of Rails
weirdnesses. I can work around them fine, I’m just wondering if
there’s a Better Way™.

-RYaN


#2

So one way of getting around the first problem is to return the actual
object from within the collection.

@umbrellas = params[:umbrellas].values.collect do |u|
umbrella = u[:id] ? @person.umbrellas.detect{|u| u.id ==
u[:id].to_i}: @person.umbrellas.build(u)
umbrella.attributes = u
umbrella
end

This means that you don’t have to reload the umbrellas collection in
the transaction anymore. However, this solution seems inelegant, and
slow when you have a lot of umbrellas.

Is there no better way?

-RYaN


#3

Hello Ryan,

The no_duplicate_names? method pulls umbrellas from the database with
this line:
names = umbrellas.collect {|u| u.name}

Try switching that line to use the instance variable @umbrellas such
that:
names = @umbrellas.collect {|u| u.name}


#4

Huh, I didn’t think of that. But it doesn’t seem to make a
difference. I think that the umbrellas method uses the @umbrellas
instance variable if it exists, and otherwise creates it from the
database. So in my case, the @umbrellas array is out-of-sync.

-RYaN