hugo wrote:
[snip]
>
> 1) I'd prefer "manager = PersonManager('name_of_manager_attribute')
> over the "an attribute named objects will magically turn into objects_"
> - the default for the name should be 'objects', but if there already is
> an attribute named that, the validator should barf. And the user would
> just have to add a name to give to the manager attribute.+1 > 2) I really, really dislike the > Person.objects.get_list(Person.q.first_name == 'Adrian') - it a) > doesn't look any better or more concise than > Person.objects.get_list(first_name__exact='Adrian) and I especially > dislike the repetition of Person - did you look at #980? A much nicer > approach for the same problem, and it takes up with combined queries, > joins and other stuff by making it look more like handling of Python > objects and sequences - which is what ORMs are about. +1 A Python ORM is supposed to make the interface to a relational db as Python-like as possible - the closer we get to the expected behaviour of ordinary Python modules, classes, objects, methods and operators the better. I fear that this principle is getting lost a little in the heat of the discussion.
