Pyramid Hybrid application (traversal + url dispatch)

Question

I've started using pyramid recently and I have some best-practice/general-concept questions about it's hybrid application approach. Let me try to mock a small but complex scenario for what I'm trying to do...

scenario

Say I'm building a website about movies, a place where users can go to add movies, like them and do other stuff with them. Here's a desired functionality:

• / renders templates/root.html
• /movies renders templates/default.html, shows a list of movies
• /movies/123456 shows details about a movie
• /movies/123456/users shows list of users that edited this movie
• /users renders templates/default.html, shows a list of users
• /users/654321 shows details about a user
• /users/654321/movies shows a list of favorite movies for this user
• /reviews/movies renders templates/reviews.html, shows url_fetched movie reviews
• /admin place where you can log in and change stuff
• /admin/movies renders templates/admin.html, shows a list of editable movies
• /admin/movies/123456 renders templates/form.html, edit this movie

extras:

• ability to handle nested resources, for example movies/123/similar/234/similar/345, where similar is a property of movie class that lists ids of related movies. Or maybe more clear example would be: companies/123/partners/234/clients/345...
• /movies/123456.json details about a movie in JSON format
• /movies/123456.xml details about a movie in XML format
• RESTFull methods (GET, PUT, DELETE..) with authorization headers for resource handling

approach

This is what I've done so far (my views are class based, for simplicity, I'll list just decorators...):

# resources.py
    class Root(object):
        __name__ = __parent__ = None
        def __getitem__(self, name):
            return None

    def factory(request):
        # pseudo code...
        # checks for database model that's also a resource
        if ModelResource:
            return ModelResource()
        return Root()

# main.py
    def notfound(request):
        return HTTPNotFound()

    def wsgi_app():
        config = Configurator()
        config.add_translation_dirs('locale',)
        config.add_renderer('.html', Jinja2Renderer)
        # Static
        config.add_static_view('static', 'static')
        # Root
        config.add_route('root', '/')
        # Admin
        config.add_route('admin',   '/admin/*traverse',     factory='factory')
        config.add_route('default', '/{path}/*traverse',    factory='factory')

        config.add_notfound_view(notfound, append_slash=True)
        config.scan('views')
        return config.make_wsgi_app()


# views/root.py
    @view_config(route_name='root', renderer="root.html")
    def __call__():

# views/default.py
    @view_defaults(route_name='default')
    class DefaultView(BaseView):
        @view_config(context=ModelResource, renderer="default.html")
        def with_context():
        @view_config()
        def __call__():

# views/admin.py
    @view_defaults(route_name='admin')
    class AdminView(BaseView):
        @view_config(context=ModelResource, renderer="admin.html")
        def default(self):
        @view_config(renderer="admin.html")
        def __call__(self):

And following piece of code is from my real app. ModelResource is used as context for view lookup, and this is basically the reason for this post... Since all my models (I'm working with Google App Engine datastore) have same basic functionality they extend specific superclass. My first instinct was to add traversal functionality at this level, that's why I created ModelResource (additional explanation in code comments), but I'm begining to regret it :) So I'm looking for some insight and ideas how to handle this.

class ModelResource(ndb.Model):
    def __getitem__(self, name):
        module = next(module for module in application.modules if module.get('class') == self.__class__.__name__)

        # this can be read as: if name == 'movies' or name == 'users' or .....
        if name in module.get('nodes'):
            # with this setup I can set application.modules[0]['nodes'] = [u'movies', u'films', u'фильми' ]
            # and use any of the defined alternatives to find a resource

            # return an empty instance of the resource which can be queried in view
            return self

        # my models use unique numerical IDs 
        elif re.match(r'\d+', name):
            # return instance with supplied ID
            cls = self.__class__
            return cls.get_by_id(name)

        # this is an attempt to solve nesting issue
        # if this model has repeated property it should return the class
        #       of the child item(s)

        # for now I'm just displaying the property's value...
        elif hasattr(self, name):
            # TODO: ndb.KeyProperty, ndb.StructuredProperty
            return getattr(self, name)

        # this is an option to find resource by slug instead of ID...
        elif re.match(r'\w+', name):
            cls = self.__class__
            instance = cls.query(cls.slug == name).get()
            if instance:
                return instance

        raise KeyError

questions

I have basically two big questions (widh some sub-questions):

• How should I handle traversal in described scenario?

  1. Should I extend some class, or implement an interface, or something else?
  2. Should traversal be handled at model level (movie model, or it's superclass would have __getitem__ method)? Should resource be a model instance?
  3. What's the best practice for __getitem__ that returns collection (list) of resources? In this scenario /movies/
  4. My ModelResource in the code above is not location aware... I'm having trouble getting __public__ and __name__ to work, probably because this is the wrong way to do traversal :) I'm guessing once I find the right way, nested resources shouldn't be a problem.

• How can I switch back from traversal to url dispatching?

  1. For the extra stuff: How can I set a different renderer to handle JSON requests? Before I added traversal I configured config.add_route('json', '/{path:.*}.json') and it worked with appropriate view, but now this route is never matched...
  2. Same problem as above - I added a view that handles HTTP methods @view_config(request_method='GET'), but it's never matched in this configuration.

Sorry for the long post, but it's fairly complex scenario. I should probably mention that real app has 10-15 models atm, so I'm looking for solution that would handle everything in one place - I want to avoid manually setting up Resource Tree...