CUBAType traits should reference Ontology attributes

[flongford]

At the moment our CUBAType traits are unformatted strings that act as placeholders to reference objects in an osp-core ontology. It is expected that they would match up to an OntologyAttribute object, which also defies the python basic data type, and holds an internal reference to the OntologyNamespace that it resides in.

Therefore we should develop a method to allow either:

• The value of an CUBAType trait to be of type OntologyAttribute
• Include a reference to the OntologyNamespace in the CUBAType string formatting (i.e. 'my_namspace.my_attribute'. An internal method should then be able to parse this string and 'lookup' the OntologyAttribute object

(This system resembles string references to python modules)

It is also important not to break any existing implementations of the CUBAType trait if possible. So far none of our plugins would contain string values that contain references to an ontology namespace. The 'lookup' act should be optional at first, whilst depreciating the unformatted strings, before removing them in a later version.

[flongford]

Something like:

#: Identifies a CUBA type with its key.
 CUBAType = String()

 class CUBATypeMixin(HasStrictTraits):

     #: A CUBA key describing the type of the parameter
     type = CUBAType('Value', visible=False)

     def get_ontology_attribute(self):
         """Return the OntologyAttribute for an CUBAType trait"""

[flongford]

The following code could be used to get a reference to the CUDS ontology class for a CUBAType correctly formatted string.

from osp.core import get_entity

# Create some instance of a class that inherits CUBATypeMixin
data_value = CUBATypeMixin(type='my_namespace.my_attribute`)

# Look up the CUDS class associated with that string
entity = get_entity(data_value.type)

# Obtain the python data type associated with this class
data_type = entity.datatype

However, if the entity variable is a complex object with no associated datatype, then an error will be raised. If we just wish to attempt to look up the expected basic python type of a particular class, then we could simply handle this exception accordingly.