Repository navigation
Parametrize question: how to make 1 selenium-webdriver session per 1 test_*() function call irregardless of the number of parametrized sets of data? #2627
Description
I have conftest.py with a class that starts a selenium webdriver function.
I have test_something.py module with test_a() and test_b() functions.
Each of the functions is parametrized (uses @pytest.mark.parametrize fixture), so that each of them runs twice with different sets of parameters.
Problem: I need to make 1 webdriver session match 2 parametrized calls for each test function separately. In other words: webdriver_session1: test_a(param1), test_a(param2); webdriver_session2: test_b(param1), test_b(param2).
If I create an instance of the class outside the test functions - then I get 1 webdriver session per MODULE (1 session per 2 functions, instead of 1 session per 1 function).
If I create an instance of the class inside the test functions - then I get 1 webdriver sessions per PARAMETRIZED CALL (2 sessions per 1 function, instead of 1 session per 1 function).
I thought that wrapping test functions into test_wrapper_a() and test_wrapper_b() would help, but this way parametrizing doesn't work.
pip list:
appdirs (1.4.3)
certifi (2017.4.17)
chardet (3.0.4)
EasyProcess (0.2.3)
idna (2.5)
lxml (3.8.0)
packaging (16.8)
pip (9.0.1)
py (1.4.33)
pyparsing (2.2.0)
pytest (3.0.7)
pytest-base-url (1.3.0)
pytest-html (1.14.2)
pytest-metadata (1.3.0)
pytest-selenium (1.9.1)
pytest-variables (1.6.0)
PyVirtualDisplay (0.2.1)
requests (2.18.1)
selenium (3.4.1)
setuptools (35.0.1)
six (1.10.0)
urllib3 (1.21.1)
wheel (0.29.0)
CentOS 7x64.
conftest.py:
class Driver():
def __init__(self):
self.wd = webdriver.Chrome()
def __del__(self):
self.wd.quit()
test_module_fail_1.py:
from conftest import Driver
@pytest.mark.parametrize('arg', [('set1'),('set2')])
def test_a(arg):
o = Driver() # I want 1 call of that per the whole function, not per 1 parametrized call of this function.
assert True
@pytest.mark.parametrize('arg', [('set1'),('set2')])
def test_b(arg):
o = Driver() # I want 1 call of that per the whole function, not per 1 parametrized call of this function.
assert True
test_module_fail_2.py:
from conftest import Driver
def test_a_wrapper():
o = Driver()
@pytest.mark.parametrize('arg', [('set1'),('set')])
def test_a(arg): # This way pytest doesn't call this function, even if I rename test_a_wrapper to a_wrapper.
assert True
def test_b_wrapper():
o = Driver()
@pytest.mark.parametrize('arg', [('set1'),('set')])
def test_b(arg): # This way pytest doesn't call this function, even if I rename test_b_wrapper to b_wrapper.
assert True
Activity
RonnyPfannschmidt commentedon Jul 28, 2017
currently that is not supported, #2522 would enable it with a new scope
inl-pd-autotest commentedon Jul 29, 2017
Thank you.
I could replace the currently non-working parametrizing mechanism inside the wrapper function with a simple for loop that makes calls of the wrapped function as many times as I need, but all the calls would together count only as a single test. Could you provide some mechanism so that I could artificially mark successful tests finishes?
nicoddemus commentedon Sep 2, 2018
@inl-pd-autotest has this advanced or is still an issue?
inl-pd-autotest commentedon Sep 3, 2018
@nicoddemus Well, I got no answer, so it is still an issue.
RonnyPfannschmidt commentedon Sep 3, 2018
@inl-pd-autotest you could fake it by having a fixture that hold a browser and a key and at the lowert level request the browser with the "key" and drop it every time the key doesnt match
inl-pd-autotest commentedon Sep 3, 2018
I just noticed that ticket #2522 referenced in the first comment to this ticket has now been resolved.
Does it mean that what I initially wanted by this ticket is now possible to do via some new scope?
How?
RonnyPfannschmidt commentedon Sep 3, 2018
@inl-pd-autotest unfortunately its only half-fixed - we need a major release to fix the rest as far as i can tell (its not availiable as scope or in the node tree, so it cant help you with fixtures
inl-pd-autotest commentedon Sep 3, 2018
@RonnyPfannschmidt not sure I understand: is it already fixed, but just not in a release yet, or is it only half-fixed and there are still things to fix first before a new release with a new scope could come out?
If the latter is the case - could you, please, reference related tickets numbers?
So that I'd know what to wait for.
RonnyPfannschmidt commentedon Sep 3, 2018
its the later and actually good call out - i need to make an issue