tests/privacy/test_recognizers.py

"""
Tests for the three custom Presidio PatternRecognizers in src/privacy/redactor.py.

No spaCy model required — these tests exercise pure regex pattern matching only.
Run fast subset: pytest tests/privacy/test_recognizers.py
"""

import pytest
from presidio_analyzer import AnalyzerEngine
from presidio_analyzer.nlp_engine import NlpEngineProvider

from src.privacy.redactor import (
    _aadhaar_recognizer,
    _pan_recognizer,
    _vehicle_registration_recognizer,
)


# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------

def _matches(recognizer, text):
    """Return analyzer results from a single recognizer against text."""
    return recognizer.analyze(text=text, entities=[recognizer.supported_entities[0]])


def _engine_with(*recognizers):
    """Build a minimal AnalyzerEngine with only the given recognizers (no spaCy)."""
    nlp_config = {
        "nlp_engine_name": "spacy",
        "models": [{"lang_code": "en", "model_name": "en_core_web_sm"}],
    }
    try:
        nlp_engine = NlpEngineProvider(nlp_configuration=nlp_config).create_engine()
    except Exception:
        pytest.skip("spaCy en_core_web_sm not available for collision test")

    engine = AnalyzerEngine(nlp_engine=nlp_engine, supported_languages=["en"])
    # Remove all default recognizers, add only ours
    engine.registry.recognizers.clear()
    for r in recognizers:
        engine.registry.add_recognizer(r)
    return engine


# ---------------------------------------------------------------------------
# IN_AADHAAR — valid formats
# ---------------------------------------------------------------------------

def test_aadhaar_valid_formats():
    """Space-separated, hyphen-separated, and continuous formats all match."""
    recognizer = _aadhaar_recognizer()

    space_result = _matches(recognizer, "My aadhaar is 2345 6789 0123")
    assert len(space_result) == 1
    assert space_result[0].entity_type == "IN_AADHAAR"

    hyphen_result = _matches(recognizer, "UID: 2345-6789-0123")
    assert len(hyphen_result) == 1
    assert hyphen_result[0].entity_type == "IN_AADHAAR"

    continuous_result = _matches(recognizer, "234567890123")
    assert len(continuous_result) == 1
    assert continuous_result[0].entity_type == "IN_AADHAAR"


# ---------------------------------------------------------------------------
# IN_AADHAAR — invalid formats
# ---------------------------------------------------------------------------

def test_aadhaar_invalid_formats():
    """First digit 1, 11-digit, and 13-digit inputs must not match."""
    recognizer = _aadhaar_recognizer()

    assert len(_matches(recognizer, "1345 6789 0123")) == 0, "first digit 1 should not match"
    assert len(_matches(recognizer, "2345 6789 012")) == 0, "11 digits should not match"
    assert len(_matches(recognizer, "2345 6789 01234")) == 0, "13 digits should not match"


# ---------------------------------------------------------------------------
# IN_PAN — valid format
# ---------------------------------------------------------------------------

def test_pan_valid_format():
    """Canonical PAN format (5 uppercase + 4 digits + 1 uppercase) matches."""
    recognizer = _pan_recognizer()
    result = _matches(recognizer, "My PAN is ABCDE1234F")
    assert len(result) == 1
    assert result[0].entity_type == "IN_PAN"


# ---------------------------------------------------------------------------
# IN_PAN — invalid formats
# ---------------------------------------------------------------------------

def test_pan_invalid_formats():
    """Lowercase, digit-ending, and 4-letter-prefix inputs must not match."""
    recognizer = _pan_recognizer()

    assert len(_matches(recognizer, "abcde1234f")) == 0, "lowercase PAN should not match"
    assert len(_matches(recognizer, "ABCDE12345")) == 0, "PAN ending in digit should not match"
    assert len(_matches(recognizer, "ABCD1234F")) == 0, "only 4 leading letters should not match"


# ---------------------------------------------------------------------------
# IN_VEHICLE_REGISTRATION — valid formats
# ---------------------------------------------------------------------------

def test_vehicle_registration_valid_formats():
    """2-digit district, 1-digit district, and 1-letter series all match."""
    recognizer = _vehicle_registration_recognizer()

    result_std = _matches(recognizer, "vehicle number MH01AB1234")
    assert len(result_std) == 1
    assert result_std[0].entity_type == "IN_VEHICLE_REGISTRATION"

    result_1d = _matches(recognizer, "reg no DL3CAF0001")
    assert len(result_1d) == 1
    assert result_1d[0].entity_type == "IN_VEHICLE_REGISTRATION"

    result_1s = _matches(recognizer, "number plate KA01A1234")
    assert len(result_1s) == 1
    assert result_1s[0].entity_type == "IN_VEHICLE_REGISTRATION"


# ---------------------------------------------------------------------------
# PAN vs vehicle registration collision — PAN score (0.85) wins
# ---------------------------------------------------------------------------

def test_pan_wins_over_vehicle_on_collision():
    """When a span matches both IN_PAN and IN_VEHICLE_REGISTRATION, IN_PAN wins."""
    # ABCDE1234F — 5 letters + 4 digits + 1 letter — matches PAN regex exactly.
    # It also satisfies the vehicle reg pattern (2+1+3+4 split: AB·C·DE·1234F is
    # borderline, but Presidio score resolves it). We assert IN_PAN dominates.
    pan_r = _pan_recognizer()
    veh_r = _vehicle_registration_recognizer()
    engine = _engine_with(pan_r, veh_r)

    results = engine.analyze(
        text="My PAN is ABCDE1234F",
        entities=["IN_PAN", "IN_VEHICLE_REGISTRATION"],
        language="en",
    )
    entity_types = {r.entity_type for r in results}
    assert "IN_PAN" in entity_types
    assert "IN_VEHICLE_REGISTRATION" not in entity_types