# ============================================================
# FILE  : test_ner.py
# FUNGSI: Unit test rule-based NER dan fungsi pendukungnya
# AUTHOR: Ariel Jonathan
# ============================================================

from __future__ import annotations

import pytest

from ner.ner_detector import (
    IndonesianNER,
    NEREntity,
    _chunk_text,
    _has_preceding_title,
    _merge_entities,
)


class TestChunkText:
    def test_teks_pendek_satu_potongan(self):
        chunks = _chunk_text("Ahmad bekerja di Jakarta.")
        assert len(chunks) == 1
        assert chunks[0][1] == 0

    def test_teks_panjang_beberapa_potongan(self):
        text = "Budi pergi ke pasar. " * 120
        chunks = _chunk_text(text)
        assert len(chunks) > 1

    def test_offset_potongan_cocok_dengan_teks_asli(self):
        text = "Budi pergi ke pasar. " * 120
        for chunk, offset in _chunk_text(text):
            assert text[offset:offset + len(chunk)] == chunk


class TestPrecedingTitle:
    def test_gelar_terdeteksi(self):
        assert _has_preceding_title("Pak ", 4) is True
        assert _has_preceding_title("Ibu ", 4) is True

    def test_kata_biasa_bukan_gelar(self):
        assert _has_preceding_title("Bunga ", 6) is False


@pytest.fixture(scope="module")
def ner():
    # Rule-only, tanpa model ML, agar test cepat dan deterministik
    return IndonesianNER()


class TestRegresiNamaAwalKalimat:
    """Regresi B8: nama satu kata yang juga kata umum tidak ditandai di awal kalimat."""

    def test_bunga_di_awal_kalimat_bersih(self, ner):
        orang = [e for e in ner.predict("Bunga bank ditetapkan lima persen.")
                 if e.label == "ORANG"]
        assert orang == []

    def test_nama_dua_kata_tetap_terdeteksi(self, ner):
        orang = [e.word for e in ner.predict("Ahmad Santoso memimpin rapat.")
                 if e.label == "ORANG"]
        assert "Ahmad Santoso" in orang

    def test_nama_dengan_gelar_terdeteksi(self, ner):
        orang = [e.word for e in ner.predict("Ibu Bunga mengajar kelas VIII.")
                 if e.label == "ORANG"]
        assert any("Bunga" in w for w in orang)


class TestRegresiOrganisasiInggris:
    """Regresi B11: organisasi Inggris terdeteksi via akronim dan sufiks."""

    def test_akronim_global_terdeteksi(self, ner):
        org = [e.word for e in ner.predict("UNESCO mengeluarkan pedoman baru.")
               if e.label == "ORGANISASI"]
        assert "UNESCO" in org

    def test_sufiks_badan_usaha_terdeteksi(self, ner):
        org = [e.word for e in ner.predict("Dia bekerja di OpenAI Inc. mulai besok.")
               if e.label == "ORGANISASI"]
        assert any("OpenAI" in w for w in org)


class TestRegresiSpanEntitas:
    """Regresi: entitas tidak boleh membawa tanda baca tepi."""

    def test_rule_organisasi_tidak_membawa_koma(self, ner):
        org = [e.word for e in ner.predict("Dia bekerja di PT Maju Bersama, sejak 2024.")
               if e.label == "ORGANISASI"]
        assert "PT Maju Bersama" in org
        assert all(not w.endswith(",") for w in org)

    def test_rule_spesifik_mengalahkan_fragmen_ml(self):
        ml = [NEREntity("Universitas", "ORGANISASI", 0.99, 20, 31, "ml")]
        rules = [NEREntity("Universitas Indonesia", "ORGANISASI", 0.90, 20, 41, "rule")]
        merged = _merge_entities(ml, rules)
        assert [e.word for e in merged] == ["Universitas Indonesia"]