Hi,
I am reading this thread with growing fascination, because I have the
feeling I am missing something, so can I summarize?
If I had a test, where I don't care about output and only about whther
exceptions are thrown or not I would normally write this:
@pytest.mark.parametrize(
"req, exc", (
('foo', None),
('bar', None),
('BLARF', ValueError),
('PLONK', IndexError),
))
def test_something(self, arg, exc):
if exc:
with pytest.raises(exc):
something(arg)
else:
something(arg)
So the new feature would save me two lines of code:
@pytest.mark.parametrize(
"req, exc", (
('foo', None),
('bar', None),
('BLARF', ValueError),
('PLONK', IndexError),
))
def test_something(arg, exc):
with pytest.raises(exc):
something(arg)
... and this would be a bit harder to understand, because as Bruno found
out already, there are quite different expectations about what is supposed
to happen if I write pytest.raises(None).
Am I missing something?
I usually don't write tests like that anyway but rather tests where I
expect either a concrete value of a certain type or I expect an Exception
to be raised. E.g.
@pytest.mark.parametrize(
"req, exp", (
('foo', True),
('bar', False),
('BLARF', ValueError),
('PLONK', IndexError),
))
def test_something(arg, exp):
if not isinstance(exp, bool):
with pytest.raises(exp):
something(arg)
else:
assert something(arg) == exp
And there this feature would not help me anyway or one would have to make a
new bastard context manager like
@pytest.mark.parametrize(
"req, exp", (
('foo', True),
('bar', False),
('BLARF', ValueError),
('PLONK', IndexError),
))
def test_something(arg, exp):
with pytest.raises_or_returns_or_yields_or_whatever(exp):
something(arg)
**shudder**
So I am not sure if this is worth implementing unless I am just not able to
grasp the real benfits.
Cheers,
Oliver
_______________________________________________
pytest-dev mailing list
[email protected]
https://mail.python.org/mailman/listinfo/pytest-dev
