from __future__ import annotations

from pathlib import Path

from tools.graphrag.evaluation.source_oracle import load_source_oracle, supporting_documents
from tools.graphrag.evaluation.benchmark_v2 import make_case, validate_case


def source_fixture(tmp_path: Path):
    root = tmp_path / 'wiki'
    profile = 'pages/students/2020/eva_nova/README.md'
    path = root / profile
    path.parent.mkdir(parents=True)
    path.write_text('''---
title: Eva Nová
taxonomy:
  category: [bp2025]
  tag: [rag]
  author: Autor Správny
---
Rok začiatku štúdia: 2020
# Diplomový projekt 2024
Téma: Výskum jazyka pomocou modelov
## Bakalárska práca 2025
Názov: Overovanie jazykových modelov
''')
    child = root / 'pages/students/2020/eva_nova/projekt/README.md'
    child.parent.mkdir(parents=True)
    child.write_text('''---
title: Pomocný projekt
taxonomy:
  category: [tp2023]
  tag: [python]
  author: Autor Iný
---
Pomocný dokument projektu.
''')
    return load_source_oracle(root), profile


def test_raw_oracle_distinguishes_year_roles_and_academic_types(tmp_path):
    oracle, _ = source_fixture(tmp_path)
    person = oracle['people']['Eva Nová']
    assert person['start_year'] == 2020
    assert {(row['academic_kind'], row['academic_type'], row['year']) for row in person['academics']} == {
        ('Work', 'diploma_project', 2024), ('Thesis', 'bachelor_thesis', 2025)}
    for row in person['academics']:
        lines = oracle['documents'][row['document_path']]['raw_text'].splitlines()
        assert row['title'] in lines[row['title_line'] - 1]
        assert str(row['year']) in lines[row['heading_line'] - 1]


def test_categories_include_child_documents_without_becoming_work_keywords(tmp_path):
    oracle, profile = source_fixture(tmp_path)
    assert oracle['people']['Eva Nová']['categories'] == ['bp2025', 'tp2023']
    assert set(oracle['people']['Eva Nová']['tags']) == {'rag', 'python'}
    assert not supporting_documents(oracle, 'Eva Nová', keyword='jazykové modely')
    assert supporting_documents(oracle, 'Eva Nová', category='bp2025', document_author='Autor Správny') == [profile]
    assert not supporting_documents(oracle, 'Eva Nová', category='bp2025', document_author='Autor Iný')


def test_source_audit_rejects_falsely_negative_existing_tag(tmp_path):
    oracle, _ = source_fixture(tmp_path)
    case = make_case('example', 'dev', 'negative_lookup', 'Má Eva Nová kľúčové slovo rag?',
                     ['Eva Nová'], oracle, [], no_answer=True,
                     negative_constraints={'keyword': 'rag'})
    assert 'negative_conjunction_actually_satisfied' in validate_case(case, oracle)
    case['graph']['negative_constraints']['keyword'] = 'neexistujucitag'
    assert validate_case(case, oracle) == []


def test_source_audit_rejects_wrong_count(tmp_path):
    oracle, _ = source_fixture(tmp_path)
    case = make_case('count', 'dev', 'candidate_set_aggregation',
                     'Z kandidátov Eva Nová vyber osoby s prácou v roku 2025. Koľko ich je?',
                     ['Eva Nová'], oracle, [], expected_set=[], set_field='person', count=0)
    assert 'incorrect_candidate_set_or_count' in validate_case(case, oracle)


def test_source_audit_rejects_faked_partition_profiles(tmp_path):
    oracle, _ = source_fixture(tmp_path)
    case = make_case('example', 'dev', 'person_title_year', 'Aká práca?', ['Eva Nová'], oracle,
                     [{'person': 'Eva Nová', 'year': 2025}])
    case['graph']['source_profiles'] = ['pages/students/fake/README.md']
    assert 'incorrect_source_profile_partition' in validate_case(case, oracle)


def test_academic_range_uses_supported_category_end_year(tmp_path):
    oracle, profile = source_fixture(tmp_path)
    path = Path(oracle['wiki_root']) / profile
    path.write_text(path.read_text().replace('bp2025', 'bp2026').replace(
        'Bakalárska práca 2025', 'Bakalárska práca 2025/26'))
    person = load_source_oracle(Path(oracle['wiki_root']))['people']['Eva Nová']
    thesis = next(row for row in person['academics'] if row['academic_kind'] == 'Thesis')
    assert thesis['year'] == 2026