breakpilot-compliance/backend-compliance/tests/test_impressum_multi_entity_check.py

"""Tests for B9 Multi-Entity-Impressum-Check.

GT-Anker: Elli IMPRESSUM-001 — bei mehreren Entitäten USt-IdNr nur
bei einer von zwei vorhanden → Finding muss per-Entity erkennen.
"""

from compliance.services.impressum_multi_entity_check import (
    _clean_entity_name,
    _slice_entities,
    check_multi_entity_impressum,
)


_ELLI_LIKE = """
Impressum

Volkswagen Group Charging GmbH
Karl-Liebknecht-Str. 32
10178 Berlin
Amtsgericht Charlottenburg HRB 208967 B
Geschäftsführer: Giovanni Palazzo, Mark Möller
Telefon: 00800 3554 1111
E-Mail: info@elli.eco

Elli Mobility GmbH
Karl-Liebknecht-Str. 32
10178 Berlin
Amtsgericht Charlottenburg HRB 274616 B
USt-IdNr.: DE814424009
Geschäftsführer: Joschi Jennermann
Telefon: 00800 00002030
E-Mail: ellimobility@elli.eco
"""


_ELLI_REAL_WORLD = """
eco Volkswagen Group Charging GmbH ist im Handelsregister des
Amtsgerichtes Charlottenburg unter der Nummer HRB 208967 B eingetragen.
Verantwortlich für den Inhalt nach § 55 Abs. 2 RStV: Giovanni Palazzo.

eco Die Elli Mobility GmbH ist im Handelsregister des Amtsgerichtes
Charlottenburg unter der Nummer HRB 274616 B eingetragen. Die
Umsatzsteueridentifikationsnummer der Elli Mobility GmbH ist
DE814424009. Postanschrift: Karl-Liebknecht-Str. 32.
Geschäftsführer: Joschi Jennermann Sebastian Steffen.
"""


class TestRealWorldElliPattern:
    """Regression: Elli's reale HTML→Text-extrahierte Form mit
    leading-noise-Artefakten (eco/Die), HRB-Boundary, und USt-IdNr-
    Vollform statt Abkürzung."""

    def test_slice_finds_two_clean_entities(self):
        slices = _slice_entities(_ELLI_REAL_WORLD)
        names = [n for n, _ in slices]
        assert names == [
            "Volkswagen Group Charging GmbH",
            "Elli Mobility GmbH",
        ]

    def test_ust_id_long_form_detected(self):
        # "Umsatzsteueridentifikationsnummer der ... ist DE..."
        findings = check_multi_entity_impressum(
            {"doc_texts": {"impressum": _ELLI_REAL_WORLD}}
        )
        ust = [f for f in findings if f["check_id"] == "IMPRESSUM-MULTI-UST_ID"]
        assert len(ust) == 1
        assert ust[0]["entities_missing"] == [
            "Volkswagen Group Charging GmbH",
        ]
        assert ust[0]["entities_present"] == ["Elli Mobility GmbH"]

    def test_blocklist_filters_false_positives(self):
        # "Programmierung der Webseite Elli Mobility GmbH" was an
        # over-match before the blocklist. With it, the only "entity"
        # candidates per HRB-block are the actual GmbHs.
        noisy = (
            "Acme GmbH HRB 1000 B Berlin.\n"
            "Foo Holding GmbH HRB 2000 B München.\n"
            "Programmierung der Webseite Acme GmbH.\n"
            "Umsatzsteueridentifikationsnummer der Foo Holding GmbH "
            "ist DE111111111."
        )
        slices = _slice_entities(noisy)
        # Both real entities, no false positives.
        names = sorted(n for n, _ in slices)
        assert names == ["Acme GmbH", "Foo Holding GmbH"]


class TestCleanEntityName:
    def test_strips_header_prefix(self):
        assert _clean_entity_name(
            "Impressum\n\nVolkswagen Group Charging GmbH"
        ) == "Volkswagen Group Charging GmbH"

    def test_strips_email_artifact(self):
        # mid-text matches may capture an email TLD before the next entity
        assert _clean_entity_name(
            "eco\n\nElli Mobility GmbH"
        ) == "Elli Mobility GmbH"

    def test_collapses_whitespace(self):
        assert _clean_entity_name(
            "Acme   Holding   GmbH"
        ) == "Acme Holding GmbH"

    def test_clean_name_passes_through(self):
        assert _clean_entity_name("Elli Mobility GmbH") == "Elli Mobility GmbH"


class TestSliceEntities:
    def test_no_entities_empty(self):
        assert _slice_entities("Random text without companies.") == []

    def test_single_entity_no_slice(self):
        # Single-entity impressum is the normal case — multi-entity-only.
        assert _slice_entities("Acme GmbH\nMusterstr 1\n") == []

    def test_two_entities_clean_names(self):
        slices = _slice_entities(_ELLI_LIKE)
        names = [n for n, _ in slices]
        assert names == [
            "Volkswagen Group Charging GmbH",
            "Elli Mobility GmbH",
        ]


class TestCheck:
    def test_elli_pattern_finds_missing_ust_id(self):
        findings = check_multi_entity_impressum(
            {"doc_texts": {"impressum": _ELLI_LIKE}}
        )
        ust_findings = [
            f for f in findings if f["check_id"] == "IMPRESSUM-MULTI-UST_ID"
        ]
        assert len(ust_findings) == 1
        f = ust_findings[0]
        assert f["entities_missing"] == ["Volkswagen Group Charging GmbH"]
        assert f["entities_present"] == ["Elli Mobility GmbH"]
        assert f["severity"] == "MEDIUM"
        assert "TMG" in f["norm"]

    def test_no_impressum_no_findings(self):
        assert check_multi_entity_impressum({"doc_texts": {}}) == []

    def test_single_entity_no_findings(self):
        text = "Acme GmbH\nMusterstr 1\nUSt-IdNr DE123456789"
        assert check_multi_entity_impressum(
            {"doc_texts": {"impressum": text}}
        ) == []

    def test_both_entities_have_ust_id_no_finding(self):
        text = (
            "Acme GmbH\nUSt-IdNr DE111111111\n\n"
            "Foo Holding GmbH\nUSt-IdNr DE222222222\n"
        )
        ust_findings = [
            f for f in check_multi_entity_impressum(
                {"doc_texts": {"impressum": text}}
            )
            if f["check_id"] == "IMPRESSUM-MULTI-UST_ID"
        ]
        assert ust_findings == []