do not autogenerate classes when loading namespace

[ajtritt]

Motivation

Fix #553

How to test the behavior?

Set breakpoints or print statements deep in TypeMap.get_container_cls.

Checklist

• Did you update CHANGELOG.md with your changes?
• Have you checked our Contributing document?
• Have you ensured the PR clearly describes the problem and the solution?
• Is your contribution compliant with our coding style? This can be checked running flake8 from the source directory.
• Have you checked to ensure that there aren't other open Pull Requests for the same change?
• Have you included the relevant issue number using "Fix #XXX" notation where XXX is the issue number? By including "Fix #XXX" you allow GitHub to close issue #XXX when the PR is merged.

[rly]

Looks good, though one test is failing. IIRC that test was the following:
'test-core' namespace contains type Bar
'ndx-qux' namespace contains types Qux, Spam
'ndx-test' namespace contains type Baz which extends Bar and includes types Qux and Bar

I think I figured out the core issue.

1. some dependencies (data_type_inc and links) were not being resolved. That can be resolved with the following new get_container_cls and renamed __resolve_child_container_classes:

    def get_container_cls(self, **kwargs):
        """Get the container class from data type specification.
        If no class has been associated with the ``data_type`` from ``namespace``, a class will be dynamically
        created and returned.
        """
        namespace, data_type, autogen = getargs('namespace', 'data_type', 'autogen', kwargs)
        cls = self.__get_container_cls(namespace, data_type)
        if cls is None and autogen:  # dynamically generate a class
            spec = self.__ns_catalog.get_spec(namespace, data_type)
            self.__resolve_dependent_container_classes(spec, namespace)
            parent_cls = self.__get_parent_cls(namespace, data_type, spec)
            attr_names = self.__default_mapper_cls.get_attr_names(spec)
            cls = self.__class_generator.generate_class(data_type, spec, parent_cls, attr_names, self)
            self.register_container_type(namespace, data_type, cls)
        return cls

    def __resolve_dependent_container_classes(self, spec, namespace):
        """Trigger class generation and registration for dependent data types."""
        if spec.data_type_inc is not None:
            self.get_container_cls(namespace, spec.data_type_inc)
        if isinstance(spec, GroupSpec):
            for child_spec in (spec.groups + spec.datasets):
                if child_spec.data_type_inc is not None:
                    self.get_container_cls(namespace, child_spec.data_type_inc)
                elif child_spec.data_type_def is not None:
                    self.get_container_cls(namespace, child_spec.data_type_def)
            for child_spec in spec.links:
                if child_spec.target_type is not None:
                    self.get_container_cls(namespace, child_spec.target_type)

2. The test generates classes from different namespaces and expects the TypeMap to resolve between them, BUT when a data type from one namespace references types from another namespace, those type classes should be included/added in the first namespace and it is those generated classes with the first namespace that will be referred to in the docval and other properties of the original generated class.

So the test can be modified to generate classes all from the ndx-test namespace

baz_cls = self.type_map.get_container_cls('ndx-test', 'Baz')
bar_cls = self.type_map.get_container_cls('ndx-test', 'Bar')
qux_cls = self.type_map.get_container_cls('ndx-test', 'Qux')

This cross-namespace strategy might not work fully though.
Let's say:
Namespace N1 contains group type A
Namespace N2 contains group type B extends A. And so N2 will also include its own version of type A.
Namespace N3 contains group type C which includes subgroup type A. And so N3 will also include its own version of type A.

Type B inherits from type A built with namespace N2
Type C allows inclusion of classes of type A built with namespace N3. So the constructor for type C will not allow Type B to be passed in.

I'm not sure what the solution to this is yet, but this situation is pretty rare. It has happened already with the Allen Institute so we should test that before releasing this (but we can merge this).

[rly]

This change will fix #554 as well.

[rly]

Never mind my comments above. I just realized the purpose of TypeSource and made a proposed change in #557 to fix the failing test.

[oruebel]

Looks good to me. I'm wondering whether this will also help speed-up importing HDMF.

[ajtritt]

@rly @oruebel This change will require a change in pynwb to use the new form of TypeMap.get_container_cls