I'm testing a REST API, which can process different requests, e.g., dated and dateless. A request has a field called request_type. I'm wondering what's the best way to write test in hypothesis:
I can write two testes, one for dated, and the other is for dateless.
But for a common property, can I write one test which combine with pytest.mark.parametrize. The problem is how the request strategy uses that parameter req_type in @given.
@pytest.mark.parameterize('req_type', ['dated', 'dateless'])
@given(req=requests())
def test_process_request(req, req_type):
# override the parameter req_type in req with input req_type
pass
Is there a way to parametrize like @given(req=requests(req_type))? Or shall I just generate requests with dated and dateless randomly and pass into the test?
You can't do it all in external decorators, since they're (lazily) evaluated at import time, but you can use the data() strategy to draw a value inside your test function.
@pytest.mark.parametrize('req_type', ['dated', 'dateless'])
@given(data=st.data())
def test_process_request(req_type, data):
req = data.draw(requests(req_type))
...
Alternatively, if you can get the req_type from the req object, it is probably more elegant to generate either type and pass them into the test:
@given(req=st.one_of(requests(dateless), requests(dated)))
def test_process_request(req):
req_type = req.type
...
I'd prefer the latter as it's a bit less magical, but the former does work if you need it.