Interlally it was using add_term which did just what its name said:
it add a term for the item. There was no check that you could only
have one if the thing was a category. So I added a rudimentary support
for that and creted some more test debt. (list copied in below in case
it gets lost)

Then added the apply_term method on Item (see note about how it sucks
for this to be there) which checks the multiplicity of the taxonomy
of the incoming term and acts apporporately. Finally all taxonomies
currently say they are have optional multiplicity (i.e. categories)
whichi s fine for the question type story.

- [ ] Add tests for data_layer.models.Item.apply_Term
- [ ] refactor tests in rest api to use the api
- [ ] Are the updates to the model tags being saved and output okay? Why does that fail?
- [ ] hid/views/add_Categories should only update category if it has actually changed
- [ ] Test add_Categories (in the hid view) with multiple id/term pairs?
- [ ] Add error reporting to the bit of View/Edit screen that adds categories to items.
- [ ] Add tests for adding categories to more than one item at once via browser
- [ ] We probably need a default value in the category selector to show  that there's initailly no value selected
- [ ] Add tests for what happens when add_categories gets called with nonsense items and category names, should fail gracefully
- [ ] Add transport wrappers for listing the terms in a taxonomy, and update View/Editr screen to use these in place of modls directly
- [ ] Add transport wrappers for creating taxnonmies and terms
- [ ] add_Term in api views should retrn json of the serialized version of the new item -- currently empty dict
- [ ] Term.objects.by_taxonomy() should propbably do select related to get the term back as well, seewhere used in api.add_term

I think the version of the models you get for migrations doesn't include
the methods we defined, so the save method doesn't set slug
from name. So I have added explicit name and made the old
name, which was in slug format, into the value of slug.

Use accessor to fetch out the name of the first term
(assuming there will be at most one)

And the get_category function isn't needed any more
because its the shrot name that gets compared in the template
to decide which is selected.

Categories on the view/edit screen no longer have hard coded values
but instead read the available terms from the database

Add add categories method to hid/views and call from process items

Has a test, but so far only for the single category case.

add_term now throws TransportException if term does not exist

when adding to item

Throws exception if important fields are missing.

To create "tagged" items with terms
to add tags to items via the api

We don't want to see this, the name and taxonomy should be enough to
idenfiy them.

Had to make the nested serializer writable to stop other tests failing
(and because I think we need this later for creating tagged items on
import)
Also moved some meta classes up to top of classes
to make them easier to parse.

Add passingt est for adding a Term to an Item
Add unicode methods for Taxonomy and Term
Change the unique contstraints for Term so that
Term name needs be unique only for a given Taxonomy

Removed taxonomy and term classes from data_layer - just reference the ones
in taxonomies app directly for now.

Will need to run "deploy/tasks.py clean_db" and "deploy/tasks.py update_db"

Ive been working through some options for how the REST
api will work with categories in the readme file.

The last error from running the tests was that the TErm table didnt
exist in the test database. So I'm running

    manage.py makemigrations

And it fails with

    AttributeError: 'NoneType' object has no attribute 'unique'

Following the [django docs](https://docs.djangoproject.com/en/1.8/topics/db/models/#abstract-related-name)
I put in:

        related_name="%(app_label)s_%(class)s_term"

But still can't even get a helpful error message.

Tidy up transport delete items

Make the url generation work proerly (though I'm fairly sure we don't
actually need the url as an argument to delete) and remove
debug print statement *blush*

Make bulk_delete use delete() instead of bulk delete from REST