chatkit/backend/tests/test_native_parsers.py

"""Integration tests for native PDF table extraction.

Covers the public `native_pdf_tables` facade (registered) and exercises the
underlying engine classes directly to keep regression coverage on each.
"""

from __future__ import annotations

from pathlib import Path

import pytest

from app.parsers import get_parser, pick_auto
from app.parsers.base import ParserFile
from app.parsers.pdfplumber_tables import PdfPlumberTablesParser
from app.parsers.pymupdf_tables import PyMuPDFTablesParser

SAMPLE_PDF = Path(
    "/Users/yu.c/Desktop/npo-ca-classlens-data copy/example-2/9A作答一覽表.pdf"
)


@pytest.fixture
def sample_file() -> ParserFile:
    if not SAMPLE_PDF.exists():
        pytest.skip(f"Sample PDF not available: {SAMPLE_PDF}")
    return ParserFile(filename=SAMPLE_PDF.name, content=SAMPLE_PDF.read_bytes())


# ---------------------------------------------------------------------------
# Engine-level coverage (PyMuPDF + pdfplumber)
# ---------------------------------------------------------------------------


def _engines():
    return [PyMuPDFTablesParser(), PdfPlumberTablesParser()]


@pytest.mark.integration
@pytest.mark.parametrize("engine", _engines(), ids=lambda e: e.name)
async def test_engine_parses_student_answers(sample_file: ParserFile, engine):
    result = await engine.parse([sample_file], "student_answers")
    students = result.data["students"]
    assert len(students) == 7, f"expected 7 students, got {len(students)}"
    for s in students:
        assert 40 <= len(s["answers"]) <= 50
        assert s["name"]
        assert s["id"]
    by_name = {s["name"]: s for s in students}
    assert "陳筱琳" in by_name
    assert by_name["陳筱琳"]["id"] == "902-13"
    first_10 = [a["answer"] for a in by_name["陳筱琳"]["answers"][:10]]
    # From the PDF: "==CBBBA==C"
    assert first_10 == ["=", "=", "C", "B", "B", "B", "A", "=", "=", "C"]


@pytest.mark.integration
@pytest.mark.parametrize("engine", _engines(), ids=lambda e: e.name)
async def test_engine_parses_teacher_answers(sample_file: ParserFile, engine):
    result = await engine.parse([sample_file], "teacher_answers")
    answers = result.data["answers"]
    assert 40 <= len(answers) <= 50
    first_10 = "".join(a["correct_answer"] for a in answers[:10])
    assert first_10 == "BAACCACACB"


# ---------------------------------------------------------------------------
# Facade behavior
# ---------------------------------------------------------------------------


@pytest.mark.integration
async def test_native_facade_succeeds_via_first_engine(sample_file: ParserFile):
    parser = get_parser("native_pdf_tables")
    result = await parser.parse([sample_file], "student_answers")
    assert result.parser_name == "native_pdf_tables"
    assert any("native PDF engine" in n for n in result.notes)
    assert len(result.data["students"]) == 7


@pytest.mark.unit
async def test_native_facade_raises_csv_hint_when_all_engines_fail():
    parser = get_parser("native_pdf_tables")
    fake = ParserFile(filename="not-a.pdf", content=b"%PDF-1.4 garbage")
    with pytest.raises(ValueError, match=r"(?i)please convert to csv"):
        await parser.parse([fake], "student_answers")


@pytest.mark.unit
def test_native_facade_refuses_image_file():
    parser = get_parser("native_pdf_tables")
    png_bytes = b"\x89PNG\r\n\x1a\n"
    assert parser.can_handle(png_bytes, "scan.png", "student_answers") is False


# ---------------------------------------------------------------------------
# Auto-router
# ---------------------------------------------------------------------------


@pytest.mark.integration
async def test_auto_picks_native_facade_for_text_pdf(sample_file: ParserFile):
    chosen = pick_auto(sample_file.content, sample_file.filename, "student_answers")
    assert chosen.name == "native_pdf_tables"


@pytest.mark.integration
async def test_auto_falls_back_to_llm_for_questions(sample_file: ParserFile):
    chosen = pick_auto(sample_file.content, sample_file.filename, "questions")
    assert chosen.name == "llm_vision"