target() cannot be called outside a hypothesis test

[aarchiba]

I have a test assert_allclose_phase which checks whether two arrays are almost equal modulo one. When running under hypothesis, this seems like a sensible place to put target, albeit with a(n optional?) name argument in case I want to call it twice in the same test. Unfortunately if I do so then I can't use the same function in non-hypothesis tests:

E           hypothesis.errors.InvalidArgument: Calling target() outside of a test is invalid.

My current workaround is to call target only if name is provided, but this clutters simple tests that wouldn't need a name.

[Zac-HD]

hypothesis/hypothesis-python/src/hypothesis/control.py

Lines 173 to 175 in caccabb

	context = _current_build_context.value
	if context is None:
	raise InvalidArgument("Calling target() outside of a test is invalid.")

I'd be happy to add an optional keyword-only argument like allow_outside_given_test=False, and if True add an early return inside this check. The idea would be to continue catching accidental use outside @given-tests, which would be a noop, but also support this kind of helper function which may be used outside them too.

[PurpleDevAu]

I'll give this a shot, if no one has yet

[Zac-HD]

Thanks to @PiDelport in #2675 (review), I've come around to the idea that instead of a special-case argument, we should provide a general predicate. For example:

if hypothesis.currently_in_test_context():
    # We're in an @given or stateful test, and can safely use e.g.
    target(x)
    event("hello")
    assume(foo)
    if bar: reject()

This should still be documented as "only use this in test helper functions", but I think it would be of sufficient if niche interest to justify its existence - and more orthogonal/composable than supporting similar capabilities on a function-by-function basis.