tests/test_profanity.py

# ============================================================
# FILE  : test_profanity.py
# FUNGSI: Unit test detektor kata kasar
# AUTHOR: Ariel Jonathan
# ============================================================

from __future__ import annotations

import pytest

from profanity.profanity_detector import (
    ProfanityChecker,
    _consonant_skeleton,
    _normalize,
)


@pytest.fixture(scope="module")
def checker():
    chk = ProfanityChecker()
    chk.load()
    return chk


class _FakePipe:
    """Pipeline tiruan meniru keluaran text-classification toksisitas."""

    def __init__(self, label: str, score: float) -> None:
        self._label, self._score = label, score

    def __call__(self, text, **kwargs):
        return [{"label": self._label, "score": self._score}]


class TestLayer2Toksisitas:
    """Layer 2 ML opsional — diuji dengan pipe tiruan (tanpa unduh model)."""

    def _chk(self, label, score):
        chk = ProfanityChecker(use_ml=False)
        chk.load()
        chk._ml_pipe = _FakePipe(label, score)  # suntik pipe tiruan
        return chk

    def test_teks_toksik_ditandai_layer_ml(self):
        chk = self._chk("HateSpeech", 0.93)
        out = chk.check("kamu benar benar menyebalkan sekali hari ini")
        assert out and out[0].layer == "ml"

    def test_label_negasi_tidak_ditandai(self):
        # "Non_HateSpeech" mengandung "HATE" tetapi negatif → tidak toksik.
        chk = self._chk("Non_HateSpeech", 0.99)
        assert chk.check("kamu benar benar menyebalkan sekali hari ini") == []

    def test_skor_rendah_tidak_ditandai(self):
        chk = self._chk("HateSpeech", 0.40)
        assert chk.check("kamu benar benar menyebalkan sekali hari ini") == []

    def test_leksikon_menang_atas_ml(self):
        # Bila Layer 1 menemukan, Layer 2 tidak dijalankan (tidak ada temuan ml).
        chk = self._chk("HateSpeech", 0.99)
        out = chk.check("dasar anjing kamu ini")
        assert out and all(f.layer == "lexicon" for f in out)


# ============================================================
# Fungsi normalisasi (fungsi murni)
# ============================================================

class TestNormalize:
    def test_leetspeak_dikonversi(self):
        assert _normalize("4nj1ng") == "anjing"

    def test_karakter_berulang_dirapikan(self):
        assert _normalize("anjiiing") == "anjing"

    def test_huruf_besar_menjadi_kecil(self):
        assert _normalize("GOBLOK") == "goblok"


class TestConsonantSkeleton:
    def test_vokal_dihapus(self):
        assert _consonant_skeleton("anjing") == "njng"

    def test_kata_tanpa_vokal_tetap(self):
        assert _consonant_skeleton("bngst") == "bngst"


# ============================================================
# Deteksi end-to-end
# ============================================================

class TestDeteksi:
    def test_umpatan_langsung_terdeteksi(self, checker):
        words = {f.normalized for f in checker.check("dasar anjing goblok")}
        assert {"anjing", "goblok"} <= words

    def test_kata_vulgar_severity_high(self, checker):
        findings = checker.check("kontol itu kata vulgar")
        assert findings and findings[0].severity == "HIGH"

    def test_singkatan_tanpa_vokal_terdeteksi(self, checker):
        findings = checker.check("anjng bngst")
        assert len(findings) == 2

    def test_leetspeak_terdeteksi(self, checker):
        findings = checker.check("4nj1ng lu")
        assert findings and findings[0].normalized == "anjing"

    def test_teks_bersih_tanpa_temuan(self, checker):
        assert checker.check("Tolong jelaskan materi fotosintesis") == []


class TestRegresiKataMedisReligius:
    """Regresi B5: kata medis dibersihkan dari SUMBER lexicon (bukan whitelist).

    Catatan: setelah whitelist dihapus, kata religius/budaya yang masih ada di
    lexicon eksternal (mis. "setan", "iblis") kembali ditandai sebagai kata kasar.
    Itu konsekuensi yang diterima dari penghapusan whitelist.
    """

    def test_payudara_konteks_kesehatan_bersih(self, checker):
        text = "jelaskan deteksi dini kanker payudara untuk siswa SMA"
        assert checker.check(text) == []