hypothesis-crosshair

Add the power of solver-based symbolic execution to your Hypothesis tests with CrossHair.

Just

pip install hypothesis-crosshair

and then add a backend="crosshair" setting, like so:

from hypothesis import given, settings, strategies as st

@settings(backend="crosshair")
@given(st.integers())
def test_needs_solver(x):
    assert x != 123456789

Docs hopefully coming soon. In the meantime, start a discussion or file an issue.

FAQ

Can I try using crosshair for ALL my hypothesis tests?

Yes! Create or edit your pytest conftest.py file to register a profile like the following:

from hypothesis import settings

settings.register_profile(
    "crosshair",
    backend="crosshair",
)

And then run pytest using the profile you've defined:

pytest . --hypothesis-profile=crosshair 

Changelog

0.0.18

• Ensure drawn floats respect hypothesis signed-zero semantics for min_value/max_value.

0.0.17

• Do not interpret Unsatisfiable errors as user exceptions; just re-throw, so that hypothesis can act appropriately.
• Report CrossHair path abortions to hypothesis as discard_test_case instead of verified. This lets Hypothesis report unsatisfiable strategies correctly when run under crosshair.

0.0.16

• Integrate hypothesis's new BackCannotProceed exception, which will reduce the likelihood of FlakeyReplay errors.
• Validate suspected counterexamples with concrete executions.
• Treat nondeterminism as an unexplored path rather than a user error. (though we might change this back later)
• Ensure realization logic called by hypothesis cannot grow the path tree.
• Allow for collapsing more SMT expressions when drawing strings and floats.

0.0.15

• (was never released)

0.0.14

• Support the revised hypothesis provider draw interfaces as of hypothesis v6.112.0.

0.0.13

• Integrate with the hypothesis observability system.

0.0.12

• Error early when trying to nest hypothesis tests. (which will otherwise put CrossHair into a bad state)

0.0.11

• Address errors when the solver can't keep up (fixes #20)

0.0.10

• Reduce the numebr of iterations required to generate valid datetimes

0.0.9

• Quietly ignore iterations that appear to be failing due to symbolic intolerance.