spb/company-atlas

1"""Industry taxonomy file + deterministic label → slug mapping."""2from __future__ import annotations34import pytest56from companyatlas.registry.industries import (7    industry_index,8    is_valid_slug,9    load_industries,10    map_industry,11    map_label,12    map_sic,13    top_level_of,14    top_level_slugs,15)1617EXPECTED = {18    "software industry": ["software"],19    "Software": ["software"],20    "banking": ["banking"],21    "bank": ["banking"],22    "video game industry": ["gaming"],23    "automotive industry": ["automotive"],24    "insurance": ["insurance"],25    "pharmaceutical industry": ["pharmaceuticals"],26    "semiconductor industry": ["semiconductors"],27    "e-commerce": ["e-commerce"],28    "retail": ["retail"],29    "telecommunications industry": ["telecommunications"],30    "airline": ["airlines"],31    "film industry": ["entertainment"],32    "financial services": ["financial-services"],33    "oil and gas industry": ["oil-gas"],34    "renewable energy": ["renewables"],35    "biotechnology": ["biotechnology"],36    "hospitality industry": ["hospitality"],37    "aerospace manufacturer": ["aerospace-defense"],38    "arms industry": ["defense"],39    "food industry": ["food-beverage"],40    "fashion": ["apparel"],41    "mass media": ["media"],42    "real estate": ["real-estate"],43    "construction industry": ["construction"],44    "steel industry": ["mining"],45    "chemical industry": ["chemicals"],46    "cloud computing": ["cloud-infrastructure"],47    "artificial intelligence": ["artificial-intelligence"],48    "payment system": ["payments"],49    "asset management": ["asset-management"],50    "management consulting": ["consulting"],51    "agriculture": ["agriculture"],52    "robotics": ["robotics"],53    "machinery industry": ["industrial-machinery"],54    "public utility": ["utilities"],55    "glass industry": ["materials"],56    "health care industry": ["healthcare"],57    "medical device": ["medical-devices"],58    "toy industry": ["consumer-goods"],59    "logistics": ["logistics"],60    "shipping": ["shipping"],61    "education": ["education"],62    "tourism": ["travel"],63    "think tank": ["education"],          # longest keyword wins over "tank" (defense)64    "cybersecurity": ["cybersecurity"],65    "social media": ["internet"],66}676869@pytest.mark.parametrize(("label", "slugs"), sorted(EXPECTED.items()))70def test_map_label(label: str, slugs: list[str]) -> None:71    assert map_industry(label) == slugs727374def test_unknown_and_empty() -> None:75    assert map_industry("") == []76    assert map_industry(None) == []77    assert map_industry("zzz nothing here") == []78    assert map_label("post-production") == []      # hyphenated words are not split into keywords798081def test_multi_label_order_and_dedupe() -> None:82    assert map_industry(["banking", "insurance", "banking", "software industry"]) == ["banking", "insurance", "software"]83    assert map_industry(["a", "b", "c", "d", "e"] + ["banking"] * 3, limit=1) == ["banking"]848586def test_taxonomy_integrity() -> None:87    inds = load_industries()88    slugs = [i.slug for i in inds]89    assert len(slugs) == len(set(slugs)) >= 4590    for i in inds:91        assert i.name and i.description92        assert len(i.keywords) >= 3, i.slug93        assert i.parent is None or i.parent in slugs94        assert i.parent is None or industry_index()[i.parent].parent is None, "max two levels"95    required = {"technology", "software", "artificial-intelligence", "semiconductors", "cloud-infrastructure", "cybersecurity", "internet",96                "e-commerce", "fintech", "payments", "banking", "insurance", "asset-management", "real-estate", "construction", "retail",97                "consumer-goods", "food-beverage", "apparel", "energy", "oil-gas", "renewables", "utilities", "mining", "chemicals", "materials",98                "manufacturing", "industrial-machinery", "automotive", "aerospace-defense", "transportation", "logistics", "airlines", "shipping",99                "telecommunications", "media", "entertainment", "gaming", "healthcare", "biotechnology", "pharmaceuticals", "medical-devices",100                "hospitality", "travel", "education", "professional-services", "consulting", "agriculture", "robotics", "defense"}101    assert required <= set(slugs)102    assert "technology" in top_level_slugs() and "software" not in top_level_slugs()103    assert top_level_of("software") == "technology" and top_level_of("technology") == "technology"104    assert is_valid_slug("banking") and not is_valid_slug("banks")105106107def test_map_sic() -> None:108    assert map_sic(7372) == "software"109    assert map_sic("3674") == "semiconductors"110    assert map_sic(6021) == "banking"111    assert map_sic(2836) == "biotechnology"112    assert map_sic(4512) == "airlines"113    assert map_sic(1311) == "oil-gas"114    assert map_sic(9999) is None115    assert map_sic(None) is None and map_sic("") is None and map_sic("abc") is None116