parent doesn't work when building associationg without saving to database

[brodock]

I have a really specific model validation that must ensure parent is not defined to enable some validations.

The problem is that if I create both the parent and the children without saving to the database (via build), it fails. Here is a code snippet:

class Category < ActiveRecord::Base
  acts_as_tree

  validates :code, :presence => {:if => first_level?}

  def first_level?
    parent.nil?
  end
end

here is how to make it fail:

c = Category.new(:code => '123')
c1 = c.children.build
c1.valid?
=> false

I've searched for the problem and found this patch submmited to rails 2.3.6 (it's old but still there): https://rails.lighthouseapp.com/projects/8994/tickets/2815-nested-models-build-should-directly-assign-the-parent

The solution is to use "inverse_of" while defining the associations. In theory it will fix this error.

[brodock]

There is one more thing about it, that can help create the tests and solve the problem:

c1.root?
=> true
c1.child?
=> false

c1.root? should return false and c1.child? should return true...

I've double checked my example, and by using "inverse_of" on both sides of the relations definition, solved it.

[brodock]

I can write the tests if you point me which spec you would like it to be in...

here is a patch to solve the problem I reported:

diff --git a/lib/closure_tree/model.rb b/lib/closure_tree/model.rb
index 866d2e8..07526c9 100644
--- a/lib/closure_tree/model.rb
+++ b/lib/closure_tree/model.rb
@@ -7,7 +7,8 @@ module ClosureTree
     included do
       belongs_to :parent,
         :class_name => _ct.model_class.to_s,
-        :foreign_key => _ct.parent_column_name
+        :foreign_key => _ct.parent_column_name,
+        :inverse_of => :children

       attr_accessible :parent if _ct.use_attr_accessible?

@@ -16,7 +17,8 @@ module ClosureTree
       has_many :children, *_ct.has_many_with_order_option(
         :class_name => _ct.model_class.to_s,
         :foreign_key => _ct.parent_column_name,
-        :dependent => _ct.options[:dependent])
+        :dependent => _ct.options[:dependent],
+        :inverse_of => :parent)

       has_many :ancestor_hierarchies, *_ct.has_many_without_order_option(
         :class_name => _ct.hierarchy_class_name,

[mceachen]

Sweet, thanks, but it seems like that doesn't quite make it work. Am I missing something?

Check out ce25cf1 — and the travis build: https://travis-ci.org/mceachen/closure_tree/builds/8485398

[mceachen]

In running this in the debugger, I think inverse is a red-herring.

I don't believe Rails supports bi-directional references when the parent doesn't have an ID.

This, for example, works without your patch—but note that I'm doing a create! on the parent, not just new:

  context "Parent/child inverse relationships" do
    it "should associate both sides of the parent and child relationships" do
      parent = Label.create!(:name => 'parent')
      child = parent.children.build(:name => 'child')
      parent.should be_root
      parent.should_not be_leaf
      child.should_not be_root
      child.should be_leaf
    end
  end

[brodock]

Oh, sorry, I missed one thing... you can't check for "parent_id", the id will only be filled after the database persistence, but "parent" will refer to the correct object.

You will probably want to check if it's "persisted" (.persisted?) to prevent a "useless" database query.

The idea is close to the code below:

def root?
  if persisted?
    _ct_parent_id.nil?
  else
    parent.nil?
  end
end

[mceachen]

Nice, I'll change it to that.

On Wednesday, June 26, 2013, Gabriel Mazetto wrote:

Oh, sorry, I missed one thing... you can't check for "parent_id", the id
will only be filled after the database persistence, but "parent" will
refer to the correct object.

You will probably want to check if it's "persisted" (.persisted?) to
prevent a "useless" database query.

The idea is close to the code below:

def root?
if persisted?
_ct_parent_id.nil?
else
parent.nil?
endend

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/64#issuecomment-20092166
.

[mceachen]

https://travis-ci.org/mceachen/closure_tree/builds/8493892 and a35b612

[mceachen]

Released with v4.2.4. Thanks for your help!