Expected behaviour of polymorphic has_many relationship

[Ashley Moran]

Hi

I posted something similar to this on the Rails-takl list but I've
just discovered the core list and Iguess this is a better place to
ask. I've been doing a bit of work on inheritance in ActiveRecord
and my changes broke this test from associations/join_model_test.rb:

def
test_polymorphic_has_many_create_model_with_inheritance_and_custom_base_
class
post = SubStiPost.create :title => 'SubStiPost', :body =>
'SubStiPost body'
assert_instance_of SubStiPost, post

tagging = tags(:misc).taggings.create(:taggable => post)
assert_equal "SubStiPost", tagging.taggable_type
end

I was getting '<"SubStiPost"> expected but was <"Post">'. To fix
this I modified a line in this section of the belongs_to method in
associations.rb:

if reflection.options[:polymorphic]
association_accessor_methods(reflection,
BelongsToPolymorphicAssociation)

module_eval do
before_save <<-EOF
association = instance_variable_get("@#{reflection.name}")
if association && association.target
if association.new_record?
association.save(true)
end

if association.updated?
self["#{reflection.primary_key_name}"] = association.id
- self["#{reflection.options[:foreign_type]}"] =
association.class.base_class.name.to_s
+ self["#{reflection.options[:foreign_type]}"] =
association.class.name.to_s
end
end
EOF
end
else
...
end

But that change breaks these tests:

def test_polymorphic_has_many_create_model_with_inheritance
post = posts(:thinking)
assert_instance_of SpecialPost, post

tagging = tags(:misc).taggings.create(:taggable => post)
assert_equal "Post", tagging.taggable_type
end

def test_polymorphic_has_one_create_model_with_inheritance
tagging = tags(:misc).create_tagging(:taggable => posts(:thinking))
assert_equal "Post", tagging.taggable_type
end

Both now fail on <"Post"> expected but was <"SpecialPost">.

What I don't understand is that the fixture Post :thinking is
actually a SpecialPost, so why is that not the correct taggable_type
value? I can't figure out what the intended behaviour of this
association.

I'd be really grateful for an explanation of this, it's the only
thing I can't make work.

Ashley

[evn]

What are your original changes, that broke the tests initially?

Evan

[Ashley Moran]

On 20 Nov 2006, at 21:41, evn wrote:
> What are your original changes, that broke the tests initially?
>
> Evan

Hi Evan

It was my attempt at class-table inheritance here: http://
dev.rubyonrails.org/ticket/6566

It's not finished yet, because I haven't made it work with
associations, but the ONLY thing it breaks in the existing tests (ie
if you use a CTI-enabled ActiveRecord but don't actually use CTI) is
polymorphic has_many relationships. I think this is because I've
redefined the meaning of "abstract class".

Any ideas?

Ashley