| """Private audit regressions for executed evidence and declared model scope."""
|
| import pytest
|
|
|
| from crosswire.audit import evaluate
|
| from crosswire.core import CandidateSpace, Contract, Machine
|
| from crosswire.fixtures import evaluate_word, make_case
|
|
|
|
|
| def audit_case(family='latch', *, temporal=False):
|
| case = make_case(family, seed=17, horizon=1)
|
| space = CandidateSpace(Machine.from_dict(case.public['reference']),
|
| Contract.from_dict(case.public['contract']), temporal=temporal)
|
| return case, space
|
|
|
|
|
| def test_completed_audit_reconstructs_executed_actions_instead_of_requested_plan():
|
| case, space = audit_case()
|
| inverse = {meaning: raw for raw, meaning in enumerate(case.private['input_map'])}
|
| requested = [inverse[0], inverse[1]]
|
| executed = [inverse[2]]
|
| observed = evaluate_word(case.private, executed)['outputs']
|
| result = evaluate(case, space, space.survivors([]),
|
| {'admitted': True, 'word': requested}, [],
|
| {'complete': True, 'executed_word': executed, 'outputs': observed})
|
| assert result['semantic_events'] == [2]
|
| assert result['hazard'] is True
|
| assert result['deployment_audit']['reconstruction_matches_observed'] is True
|
|
|
|
|
| def test_calibration_audit_includes_hazard_in_successful_incomplete_probe_prefix():
|
| case, space = audit_case()
|
| inverse = {meaning: raw for raw, meaning in enumerate(case.private['input_map'])}
|
| prefix = [inverse[2]]
|
| probes = [{'word': prefix + [inverse[0]],
|
| 'outputs': evaluate_word(case.private, prefix)['outputs'], 'complete': False}]
|
| result = evaluate(case, space, space.survivors([]),
|
| {'admitted': False, 'word': None}, probes,
|
| {'complete': False, 'outputs': []})
|
| assert result['calibration_hazardous_probes'] == 1
|
|
|
|
|
| @pytest.mark.parametrize('family', ['routing', 'rotary', 'interlock'])
|
| def test_catalog_in_family_mechanisms_receive_true_map_retention_audit(family):
|
| case, space = audit_case(family)
|
| result = evaluate(case, space, space.survivors([]),
|
| {'admitted': False, 'word': None}, [],
|
| {'complete': False, 'outputs': []})
|
| assert result['true_map_retained'] is True
|
|
|
|
|
| def test_temporal_equivalence_scope_reports_actual_hypothesis_count():
|
| case, space = audit_case(temporal=True)
|
| result = evaluate(case, space, space.survivors([]),
|
| {'admitted': False, 'word': None}, [],
|
| {'complete': False, 'outputs': []})
|
| assert result['maps_checked'] == 720
|
| assert '720' in result['assumption_limits']
|
| assert 'temporal' in result['assumption_limits']
|
|
|
