addcall does not allow binding of keyword args in parametrized test

[pytestbot]

Originally reported by: BitBucket: blep, GitHub: blep

---

When calling metafunc.addcall in pytest_generate_tests and providing keyword args with default value, addcall fails with a message such as:

bug\pytest_generator_kwargs.py:15: in pytest_generate_tests
>               metafunc.addcall( funcargs=kwargs )
C:\Python32\lib\site-packages\pytest-2.0.3-py3.2.egg\_pytest\python.py:550: in addcall
>                   pytest.fail("funcarg %r not used in this %r function." % (name, self.function))
E                   Failed: funcarg 'dbo_class' not used in this function.

Code example to reproduce the issue:

#!python

META_DBO_SCENARIOS = [
    dict( dbo_class=int,
          key_properties=['id'],
          value_properties=['content', 'name']
          )
    ]

def pytest_generate_tests(metafunc):
    if 'meta_dbo_iter' in metafunc.funcargnames:
        for scenario in META_DBO_SCENARIOS:
            kwargs = dict(meta_dbo_iter=None) # dummy parameter
            kwargs.update( scenario )
            metafunc.addcall( funcargs=kwargs )

def test_meta_dbo( meta_dbo_iter, dbo_class=None,
                   key_names=(), value_names=(),
                   auto_id=None ):
    """Checks the properties meta-data of a dbo class.
    """
    # ...

---

• Bitbucket: https://bitbucket.org/pytest-dev/pytest/issue/42

[pytestbot]

Original comment by holger krekel (BitBucket: hpk42, GitHub: hpk42):

---

it's true that only arguments without defaults are currently considered as "funcargs". how comes you need to have such a signature for your test functions? It's not impossible to extend the mechanism but i'd like to see the need.

[pytestbot]

Original comment by Anonymous:

---

It is frequent for data driven tests to have a few mandatory parameters and some optional parameters. Keyword parameters provide a convenient way to provide default value for optional parameters. This makes it easy to add a new optional parameter to handle the need of a few new specific test without impacting existing tests.

I currently work around this by forwarding to a nested function as in the example below. dbo_class is the only required parameters, though the code below does not express this.

{{{
#!python

def test_meta_dbo( data ):
def do_test( dbo_class=None, key_properties=(), value_properties=(), auto_id=None ):
pass
do_test( **data )

}}}

[pytestbot]

Original comment by holger krekel (BitBucket: hpk42, GitHub: hpk42):

---

i think this has been fixed long ago (and you now call metafunc.parametrize()).