I've created a sample app here: http://github.com/jchupp/is_paranoid_nested_attributes_test/tree/master and punted back to drawohara to attempt to reproduce it

You couldn't reproduce the problem in your sample app?

I could not. And drawohara didn't get back to me on it. I'd love to close this if someone else will confirm that they can't replicate it.

I can see why you might want to do this, but it looks like a fringe use case. I'll give this some though and if I come up with a quick fix, I'll see what I can do.

However, I'm super busy, so I wouldn't recommend holding your breath. If you want to get your hands dirty with the code, I'll gladly entertain a pull request. Good luck :)

You're definitely right about #1 being messy. I'll have to put some thought into a workaround for #2. Its definitely fixable, just a matter of finding the right solution.

This is why we test :)

To add to that comment, I should be able to push a fix this weekend. Stay tuned

I'm going to back-pedal on this. Unfortunately there's no good way to "undo" scopes with ActiveRecord. Basically the default_scope (is_paranoid) gets mashed into your named scope (in this case, "senior") and ends up just being string. So removing it is somewhat complicated.

I'm going to keep looking into this, but, to be frank, I just don't use the xxx_with_destroyed methods with named_scopes (I generally just ignore anything that's deleted unless I'm doing statistics or purging old records, etc.), so I'm not heavily personally motivated. If its your itch, though, please scratch it and fork away. I'd love to pull in a fix.

I understand your lazyless :P

I almost found a workaround but I'm a bit blocked. I don't want to undo with AR, just not do: I was thinking about removing the default scope and defining 3 named scopes: :without_deleted, :with_deleted and :only_deleted. Then wrapping the find method, check if :with_deleted or :only_deleted scopes are applied, and apply :without_deleted if none of these 2 scopes are already sent. Unfortunately, the #scopes method only returns which scopes are already available for the class and not which ones are already applied to the AR instance. Keeping with_destroyed methods is piece of cake of course for backward compatibility (we can also add with_deleted for the technoweenie legacy). That would be the trick. What do you think ?

Oh btw, what about the sql conditions duplicated ?

I think that could definitely work. If you can fork it and pass the specs plus add a spec covering named scopes working in the with_destroyed, etc. instances, I'll almost certainly merge it.

In regards to the sql conditions being doubled up, I'm not yet sure where that's occurring (I did try to recreate it but failed). What DBMS are you using? can you provide a spec that exhibits this behavior?

SCOPES: ok first I need to learn rspec :p, next I can fork and write what I have in mind but only if you know how to get the list of named scopes a model instance has ! Do you ?

SQL: I'm using postgresql, but I do experience the same issue with a blank sqlite app.

ModelName.scopes will return a list.

Yes, that's what I wrote in a previous message. But it's a class method returning available named scopes, not an instance method returning applied named scopes.

yeah, and you'll probably agree that this is why it isn't easy.

you can use #current_scoped_methods which is the currently applied scope, or #scoped_methods which returns an array of applies scopes. However, those aren't nice symbols or anything of the sort, it is a flattened set of conditions, etc.

Example: [{:find=>{:conditions=>{:deleted_at=>nil}}, :create=>{:deleted_at=>nil}}, {:find=>{:conditions=>"("androids"."deleted_at" IS NULL) AND (name like '%C%')"}, :create=>{:deleted_at=>nil}}]

so, not much nice to work with

I helped myself with this snipped... but it is not DRY

def exists_with_destroyed id
  self.with_exclusive_scope{ exists?(id)}
end

it definitely could be. As i've noted in the readme and repo description, though, this library is defunct. I'll leave this open for other people to be aware of (and maybe offer solutions) but i can't promise it will be fixed.

good luck