python providers should use value_types as a keyword argument

Some of the python providers use a slightly unusual function call signature like this:

fake.pydict(nb_elements=10, variable_nb_elements=True, *value_types)
fake.pylist(nb_elements=10, variable_nb_elements=True, *value_types)

This leads to some usage problems with factory_boy (https://github.com/FactoryBoy/factory_boy/issues/387).

IMHO it would be much more convenient to be able to call the python providers with value_types as a keyword argument like so:

fake.pydict(nb_elements=10, variable_nb_elements=True, value_types=['str'])
fake.pylist(nb_elements=10, variable_nb_elements=True, value_types=['str'])

I know this would introduce a braking change to the python provider api, but if this is acceptable I can try to provide a PR.

[fodra]

I agree this confused me a little bit.

At first I was doing this

>>> fake.pydict(nb_elements=5, variable_nb_elements=False, ['str', 'str', 'str', 'str', 'str'])
  File "<input>", line 1
SyntaxError: positional argument follows keyword argument

Figured out you should not use the keywords if you want to specify value_types

>>> fake.pydict(5, False, 'str')
{'kind': 'NNxcLDYxjKXcCypxOanT', 'social': 'RfNzeHzBFhcHdEbUMiGL', 'article': 'HrTMTBGFOAcsUAibdVRo', 'per': 'gwZMuksrnrvMyVPXAkJB', '
imagine': 'RkvSHFrUMESRLIQDgkxL'}

I hope this helps someone going through the same issue.

[joke2k]

Hi @mhubig, it seems reasonable. To handle the backcompatibility we could use some signature like that:

fake.pydict(nb_elements=10, variable_nb_elements=True, value_types=None, *allowed_types)
fake.pylist(nb_elements=10, variable_nb_elements=True, value_types=None, *allowed_types)

Combining values_types and allowed_types carefully.