this can be parameterized or a fixture

can remove this entirely and use fixtures

these can all be parameterized

you can make this a module level function

can parameterize over data_types as well

for tests that xfail, pls provide a reason

Should I add comments in the code with the reason?

yes if they are xfailing, reference the issue number and a brief description. (when the bug is fixed will remove the xfails, but that's in a subsequent PR).

you can parameterize on the sep

I would just move gen_df inside the fixture itself, I think makes it more clear.

gen_df is called directly from some of the tests. The fixture is called using a a pytest object and gen_df needs to be called with a string argument. Is there a better workaround than separating the two functions?

I can change it to something like this, but I think separating the two out is still preferable.

@pytest.fixture(params=['delims', 'utf8', 'string', 'long', 'nonascii',
                        'colwidth', 'mixed', 'float', 'int'])
def df(request):
    if type(request) is str:
        data_type = request
    else:
        data_type = request.param

    if data_type == 'delims':
        return pd.DataFrame({'a': ['"a,\t"b|c', 'd\tef´'],
                             'b': ['hi\'j', 'k\'\'lm']})
    elif data_type == 'utf8':
        return pd.DataFrame({'a': ['µasd', 'Ωœ∑´'],
                             'b': ['øπ∆˚¬', 'œ∑´®']})

oh didn't see that, though then this begs the question of why you are not simply passing the constructed df in?

That's a good point, I'll update

every xfail needs a reason

Can you take a look at the latest version? Is that what you meant for reason?

so instead of calling gen_df inside a test, just create a nother fixture, simplier that way I think
e.g.

@pytest.fixture
def df_string():
    return gen_df('string')

Thanks, that makes a lot of sense and I'll keep that in mind for future cases. I went a different way and just ran that test function on the entire df fixture, which is a little cleaner.