| from __future__ import annotations |
|
|
| from pathlib import Path |
|
|
| import pytest |
|
|
| from ragsolver.rag.parse_structure import pair_problems_and_solutions, parse_markdown |
|
|
| SAMPLE_MD = """\ |
| <!-- page 1 of 3 --> |
| ## Problemario |
| |
| 1. Enunciado del problema uno. |
| |
| 2. Enunciado del problema dos con una figura. |
| |
|  |
| |
| <!-- page 2 of 3 --> |
| ## Soluciones |
| |
| 1. Solución del problema uno. |
| |
| 2. |
| |
| $$ |
| x = 2 |
| $$ |
| |
| <!-- page 3 of 3 --> |
| ## Bibliografía |
| |
| Referencia 1. |
| """ |
|
|
|
|
| def test_parse_markdown_splits_problems_and_solutions(): |
| parsed = parse_markdown(SAMPLE_MD) |
| assert parsed["warnings"] == [] |
| assert [p.number for p in parsed["problems"]] == [1, 2] |
| assert [s.number for s in parsed["solutions"]] == [1, 2] |
|
|
|
|
| def test_problem_page_and_image_locator_captured(): |
| parsed = parse_markdown(SAMPLE_MD) |
| p2 = next(p for p in parsed["problems"] if p.number == 2) |
| assert p2.page_start == 1 |
| assert p2.image_locators == ["doc:abc123/tier:standard/page:1/block:5"] |
|
|
|
|
| def test_solution_starting_as_bare_number_then_formula_is_captured(): |
| """Caso real encontrado en el PDF de ejemplo: 'N.' solo en su línea, la |
| fórmula sigue después (docs/DECISIONS.md, solución 13).""" |
| parsed = parse_markdown(SAMPLE_MD) |
| s2 = next(s for s in parsed["solutions"] if s.number == 2) |
| assert "x = 2" in s2.text_md |
|
|
|
|
| def test_pair_problems_and_solutions_matches_by_number(): |
| parsed = parse_markdown(SAMPLE_MD) |
| cards, warnings = pair_problems_and_solutions(parsed) |
| assert len(cards) == 2 |
| assert warnings == [] |
| assert cards[0]["problem"].number == 1 |
| assert cards[0]["solution"].number == 1 |
|
|
|
|
| def test_embedded_subcase_numbering_does_not_hijack_an_earlier_item(): |
| """Regresión: una solución posterior puede traer sub-casos numerados en su |
| propio texto ("1. Prolongar...", "2. Prolongar...") que un split por línea |
| confundiría con el inicio de un ítem nuevo, pisando el ítem 1 real (pasó |
| de verdad con la solución 1 del PDF de ejemplo -- ver docs/DECISIONS.md).""" |
| md = """\ |
| ## Soluciones |
| |
| Contenido de la solución 1 sin numeral propio. |
| |
| 2. Solución dos, con dos sub-casos: |
| |
| 1. Prolongar el segmento DC por el lado de C. |
| |
| 2. Prolongar el segmento DA por el lado de A. |
| |
| 3. Solución tres. |
| """ |
| parsed = parse_markdown(md) |
| numbers = [s.number for s in parsed["solutions"]] |
| assert numbers == [1, 2, 3] |
| sol1 = next(s for s in parsed["solutions"] if s.number == 1) |
| assert sol1.text_md == "Contenido de la solución 1 sin numeral propio." |
| sol2 = next(s for s in parsed["solutions"] if s.number == 2) |
| assert "Prolongar el segmento DC" in sol2.text_md |
| assert "Prolongar el segmento DA" in sol2.text_md |
|
|
|
|
| def test_pair_reports_missing_solution_without_dropping_the_problem(): |
| parsed = parse_markdown(SAMPLE_MD.replace("2.\n\n$$\nx = 2\n$$\n\n", "")) |
| cards, warnings = pair_problems_and_solutions(parsed) |
| assert len(cards) == 2 |
| assert any("problema 2 sin solución" in w for w in warnings) |
| assert cards[1]["problem"].number == 2 |
| assert cards[1]["solution"] is None |
|
|
|
|
| |
|
|
| REAL_MD_PATH = Path(__file__).resolve().parent.parent / "data/parsed/edd3436ca148/full.md" |
|
|
| pytestmark_real = pytest.mark.skipif(not REAL_MD_PATH.exists(), reason="requiere haber corrido mineru parse sobre el PDF de ejemplo") |
|
|
|
|
| @pytestmark_real |
| def test_real_document_pairs_at_least_90_percent_of_100_problems(): |
| md = REAL_MD_PATH.read_text(encoding="utf-8") |
| parsed = parse_markdown(md) |
| cards, warnings = pair_problems_and_solutions(parsed) |
|
|
| numbers = {c["item_number"] for c in cards} |
| assert numbers == set(range(1, 101)), "todos los problemas 1-100 deben aparecer como card, con o sin solución" |
|
|
| paired = sum(1 for c in cards if c["problem"] is not None and c["solution"] is not None) |
| |
| |
| |
| |
| |
| assert paired >= 90 |
|
|
