[PATCH] Make nested models using STI work correctly

A simple example from the patch:

class Post < ActiveRecord::Base
end

class Post::NestedStiPost < Post
end

nested_sti_post = Post::NestedStiPost.create!
Post.find(nested_sti_post.id) # raises an exception because it tries to instantiate NestedStiPost, not Post::NestedStiPost

This patch may be backwards incompatible because of the changes to what is stored in the inheritance column. It breaks a couple of existing tests because of this problem.

This is a problem I've been running into for a long, long time now. There a few places where this patch doesn't address, primarily eager loading. I'll attach a modified patch with the additional fixes.

+1 for this being a Rails 2.0 patch. It is much needed.

Agreed. It'd be nice to provide a sample migration to translate old -> new types.

#6792

Is this really a duplicate of #6792 ?

It seems to me that ticket #6792 is advocating using demodulize in inheritance columns, while this ticket advocates the opposite: removing demodulized class names from inheritance columns.

If it were up to me we would NOT use demodulize in inheritance columns, so I agree with this ticket. I've run into my share of problems with the current approach.

Perhaps there would be a way to make this configurable. We could set the following by default:

 ActiveRecord::Base.demodulize_class_names_in_associations = true
}}

But display deprecation warnings telling people to explicitly set it to false and run a migration script.  For Rails 2.0 it could be flipped to false by default.

Reopening to get some comment on whether or not this is really a duplicate of ticket #6792.

The approach this ticket takes appears to be the opposite of that one, so I am not sure why it was considered a dupe. Would someone care to enlighten me on why?

Replying to dkubb:

Reopening to get some comment on whether or not this is really a duplicate of ticket #6792. The approach this ticket takes appears to be the opposite of that one, so I am not sure why it was considered a dupe. Would someone care to enlighten me on why?

I believe it was because the author changed his mind in #6792 and decided against using demodulize, making it then a duplicate of this ticket since they are both now proposing removing .demodulize from STI. At least .. I think that's why :)

Ahh obrie, I re-read that ticket thread and you're right. Closing this ticket.