Consider making Blueprint.fields() update fields instead of overwriting

I think Blueprint.fields() should return a new Blueprint, with updated fields, instead of overwriting all existing fields. This makes for what I think is a more intuitive API when using blueprints.

I'm happy to work up a PR for this if the direction seems appropriate, but I want to validate the direction before I code it up.

Steps to reproduce

I have a blueprint:

reports = factory.blueprint(Report).fields(
    contract_number=lambda n, f: f.bothify("K#####T####"),
    ... many more fields ...
)

This model happens to have a FK to another model (Contractor) , so I tried something that seemed intuitive:

c = contractors.make()
r = reports.fields(contractor=c).make()

This feels like a fairly intuitive API for making "sub-blueprints", but it doesn't work.

Expected behavior

I expected the fields on r to be set according to my field callbacks.

Actual behavior

Instead, not only are the field callbacks not called, subsequent calls to the the report blueprint don't get it either! This is because Blueprint.fields overwrites self._fields

If Blueprint.fields was instead something like:

def fields(self, **kwargs):
    new_fields = self._fields.copy()
    new_fields.update(kwargs)
    return Blueprint(self.model, new_fields, pre_save=self.pre_save, post_save=self.post_save, seed=self.seed)

Then what I expected to happen, would in fact happen. Again I think this makes for a more natural, easy-to-use API -- but maybe I'm wrong? LMK if you'd like to see a patch!

fcurella / django-fakery