""" Tests for provenance matching (src/provenance.py). The corpus below mirrors the synthetic demo Schedule: a policyholder section on page 1 and a driver table on page 2 whose rows Docling returns as single text items. These cases used to land on headings, labels or another driver's row. """ from __future__ import annotations import sys from datetime import date from pathlib import Path sys.path.insert(0, str(Path(__file__).parent.parent / "src")) from provenance import CorpusItem, ProvenanceCorpus, build_provenance # noqa: E402 from schema import Driver, PolicyHeader, UKMotorGoldenRecord # noqa: E402 _F = "Schedule of Insurance - Demo.pdf" def _corpus() -> ProvenanceCorpus: corpus = ProvenanceCorpus(source_filename=_F, doc_type="Schedule") rows = [ (1, "Policyholder details", "section_header"), (1, "Policy number", "text"), (1, "NBM-DEMO-0427", "text"), (1, "Name", "text"), (1, "Alex Morgan", "text"), (1, "14/03/1991", "text"), (1, "Product", "text"), (2, "Driver details", "section_header"), (2, "Driver name", "text"), (2, "14/03/1991 Policyholder Product Manager Full Licence UK / 2/1 / No Yes", "text"), (2, "07/08/1995 Named Driver Business Analyst UK Provisional / 1/4 / No No GBP 200.00", "text"), (2, "11/10/1985 Named Driver Data Administrator Full Licence UK / 5/0 / No No", "text"), ] for i, (page, text, label) in enumerate(rows): corpus.items.append(CorpusItem( text=text, page=page, bbox=[10.0, float(i * 5), 60.0, float(i * 5 + 2)], source_filename=_F, label=label, )) return corpus def _record() -> UKMotorGoldenRecord: record = UKMotorGoldenRecord( policy_header=PolicyHeader(policy_number="NBM-DEMO-0427"), driver_details=[ Driver(name="MASKED_PERSON_1", dob=date(1991, 3, 14), relationship="Policyholder", occupation="Product Manager", license_type="Full UK", is_main_driver=True), Driver(name="MASKED_PERSON_2", dob=date(1995, 8, 7), relationship="Named Driver", occupation="Business Analyst", license_type="UK Provisional"), Driver(name="MASKED_PERSON_3", dob=date(1985, 10, 11), relationship="Named Driver", occupation="Data Administrator", license_type="Full UK"), ], ) record.field_citations = { "driver_details[0].license_type": "Full Licence UK / 2/1 / No", "driver_details[2].license_type": "Full Licence UK / 5/0 / No", } return record def _by_path() -> dict: return {p.field_path: p for p in build_provenance(_record(), [_corpus()])} def test_policy_number_lands_on_value(): assert _by_path()["policy_header.policy_number"].matched_text == "NBM-DEMO-0427" def test_relationship_not_on_heading_or_label(): prov = _by_path() assert prov["driver_details[0].relationship"].matched_text.startswith("14/03/1991 Policyholder") assert prov["driver_details[1].relationship"].matched_text.startswith("07/08/1995") assert prov["driver_details[2].relationship"].matched_text.startswith("11/10/1985") def test_occupation_not_on_fragment(): assert _by_path()["driver_details[0].occupation"].matched_text.startswith("14/03/1991 Policyholder") def test_repeated_value_lands_on_the_right_drivers_row(): prov = _by_path() assert prov["driver_details[0].license_type"].matched_text.startswith("14/03/1991") assert prov["driver_details[2].license_type"].matched_text.startswith("11/10/1985") def test_masked_names_are_not_located(): assert not any(p.endswith(".name") for p in _by_path())