…e.py

The new attribute system will involve meta classes other helper objects
I prefer to keep separate from the objects that use them for neatness.

Rather than specify which data model the collection is a 'collection'
of in an overridden __init__ - now subclasses override the
object_class class attribute with the correct data model.

Not hugely important, stylistic choice - prefer avoiding super calls.

Previously a MembershipCollection object would be
generated on request on accessing the membership property.
Now this is generated and assigned at creation.

Reason: collections need to be permanent so that changes to
their objects can be repackaged into a file.

In new system, rather than the properties of a PopoloObject
being defined as @property-ed functions accessing properties
of self.data - this will be replaced by the 'Attribute' class
 - that given a attribute will control getting and setting
 to self.data.

Subclasses of Attribute control different ways of talking to data
e.g. talking to data that is a date.

A small meta-class is used for PopoloObject to avoid repetition
in definitions.
e.g. person = Property(attr=person) can be expressed as
person = Property()

Almost all previous @property-ed functions replaced with
Attribute object.

Functionality for 'getting' and default values are identical
to old arrangement.

When a new 'twitter' value is set, if a url - the link value is
set - and the username is also extracted for contact details.

If not a url, just updates the contact details.

Date fields for Person and Organization were set to default to
None rather than the vague .PAST and .FUTURE from approxdate.

This has been changed to bring it into sync with other Membership
and Event objects.

Person Test updated to reflect this change.

Popolo objects can now export to json (to_json) or
direct to a file name.

Two popolo objects can now be merged. Each collectiontype has
a value that must be made 'unique' (either name or id).

In a collision, the 'larger' will be kept (assumed more information)
and all refs to discarded ids replaced.

More complex functionality e.g. the bigger person object picking
up all possible alternate names of the smaller one, can be
specified by overriding the absorb function.

Datetime objects will be reduced to just their date, then converted to isoformat

Exposed json_data as something that can be accessed,
but that is not independent of child objects.
New objects are based on a deepcopy of the incoming dict.

Updated person tests with test for __gt__ and __lt__

…ssigning 'none'.

…dAttributes

Originally this was so you could build RelatedAttributes on something other than an ID field.
This was never necessary, so removed to tidy up.

These features will also be useful in the approx_dates package directly,
and will be ported there at some point.

Returns either unique time from membership or that from
overall legislative period.

Nicer way of getting a wikidata, parlparse, etc
for an individual.

Replaces with a cross-platform compatible hashing method
so membership ids are generated consistently

Will absorb gender of 'other' when self has none.