uber / charlatan Goto Github PK

FixtureManager's get_fixtures method accepts the builder instance but does not pass it further in nested call to get_fixture method.

This could be a problem with pyyaml itself.

When creating fixtures without models and nesting !now only first level is evaluated

This yaml definition:

simple_dict:
  fields:
    created_at: !now
    nested:
      created_at: !now

produces this output:

{ 
    'created_at': datetime.datetime(2015, 11, 25, 17, 27, 3, 119923, tzinfo=<UTC>), 
    'nested': {
        'created_at': <function configure_yaml.<locals>.now_constructor.<locals>.get_now at 0x1052cbbf8>
    }
}

Cheers,
Jakub

The doc for FixturesManager states that it accepts "list_or_str" of filenames. However, it always calls extend on the given parameter. If the input is a list, appending a list to a list gives you a list of lists. In this case self.filenames should be extending the given input.

When I reference a fixture in a post_creation section using !rel, it is created and added to relationship but I have to manually run an extra commit afterwards.

Results in code like:

fix.install_fixture(name)
db.session.commit()

Feels like I shouldn't need to do this...?

Currently if we want to override a field for a fixture,
we have to use install_fixture like so

self.install_fixture('foo', overrides={'a': 'baz'})
self.install_fixtures((
'foo',
'bar',
'baz'
))

however, the ability to do this inline would be nice:

self.install_fixtures((
'foo',
'bar',
'baz',
('foo', overrides={'a': 'baz'})
))

Error resolving anonymous references based on list of objects:

(...)
charlatan/utils.pyc in richgetter(obj, path)
    174             obj = getattr(obj, name)
    175             continue
--> 176         obj = obj[name]
    177 
    178     return obj

TypeError: list indices must be integers, not str

This happens loading the fixture

>>> import charlatan
>>> fm = charlatan.FixturesManager(session)
>>> fm.load('fixtures.yaml', 'app.models')

# Specific index is OK
>>> fm.get_fixture('department.0')
Department

# Whole collection raises error
>>> fm.get_fixture('departments')
(...)
TypeError: list indices must be integers, not str

office:
  model: Office

offices:
  inherit_from: office
  objects:
    -
       name: 'Office0'

    -
       name: 'Office1'

department:
  model: Department

departments:
  inherit_from: department
  objects:
    -
       name: 'Department0'
       office: !rel offices.0

    -
       name: 'Department1'
       office: !rel offices.1

It works using keys 0, 1 for offices instead of a list, but it changes the get_fixture return type from list to dict.

Multiple file support for fixtures (c5d027b) is great!

I'd like to go one step further and be able to specify a directory containing all my fixtures and have that loaded

The default SQLAlchemy builder (InstallAndSave) commits the session on each individual fixture.

In some cases, it may be more desirable to allow all fixtures to be instantiated added to the session before committing the entire session. For example, if any semantic error occurs (valid YAML, but invalid for the SQLAlchemy model), the application may wish to roll back all fixture changes rather than commit partial data.

This could be done by adding a autocommit boolean kwarg to the __call__ method that would skip this session commit call.

Currently depend_on appears to be ignored on collections.

I can't see how to use depend_on for either the parent collection or the child objects. Ideally I'd like the whole collection to depend on another collection.

Charlatan seems like a very nice way to manage and load fixture data. Is it possible to go the other way, and generate fixtures from an existing database?

Django's fixture management lets you do this via manage.py dumpdata. It's a useful way to capture database state to something that's human readable/editable and less brittle than sqldump.

Thanks.

The install_all_fixtures() method ignores fixtures declared with the inherit_from attribute.
I'm using version 0.4.5