UmbraAudio/tests/test_translation_resolution.py

"""Tests for resolving a translation from the DB and the on-disk sample files.

The behaviours pinned here are the ones that made the review UI lie: a valid
Chinese translation existed on disk but the page showed English, and the DB
held an English string under a ``zh-Hans`` code that the page would have
rendered as "translated" if it had simply trusted the row.
"""

from __future__ import annotations

import json
import sqlite3
from pathlib import Path

import pytest

from umbraaudio import translations as tr


def _seed(conn: sqlite3.Connection) -> None:
    conn.executescript(
        """
        CREATE TABLE project (id INTEGER PRIMARY KEY, name TEXT);
        CREATE TABLE chapters (id INTEGER PRIMARY KEY, chapter_index INTEGER);
        CREATE TABLE utterances (
            id INTEGER PRIMARY KEY,
            chapter_id INTEGER,
            utterance_index INTEGER,
            speaker_name TEXT,
            text TEXT,
            context_before TEXT,
            context_after TEXT,
            language_code TEXT,
            language_origin TEXT
        );
        INSERT INTO project (id, name) VALUES (1, 'Test');
        INSERT INTO chapters (id, chapter_index) VALUES (1, 1);
        INSERT INTO utterances (id, chapter_id, utterance_index, speaker_name, text) VALUES
            (1, 1, 0, 'Narrator', 'Welcome to the book.'),
            (2, 1, 1, 'Kainda',  'Hello, I am the main character.');
        """
    )
    tr.ensure_tables(conn)
    conn.commit()


@pytest.fixture
def conn() -> sqlite3.Connection:
    c = sqlite3.connect(":memory:")
    c.row_factory = sqlite3.Row
    _seed(c)
    yield c
    c.close()


def _write_sample(directory: Path, code: str, texts: list[str], chapter_id: int = 1) -> None:
    directory.mkdir(parents=True, exist_ok=True)
    payload = {
        "project": "Test",
        "chapter_id": chapter_id,
        "language_code": code,
        "utterances": [
            {"index": str(i), "speaker": "Narrator", "type": "narration", "text": text}
            for i, text in enumerate(texts)
        ],
    }
    (directory / f"sample_ch1_{code.upper()}.json").write_text(
        json.dumps(payload, ensure_ascii=False), encoding="utf-8"
    )


# --- code normalisation ---------------------------------------------------


@pytest.mark.parametrize(
    ("value", "expected"),
    [
        ("zh", "zh"),
        ("zh-Hans", "zh"),
        ("zh_CN", "zh"),
        ("EN", "en"),
        ("en-US", "en"),
        ("Russian", "ru"),
        ("nonsense", None),
        ("", None),
        (None, None),
    ],
)
def test_canonical_code_tolerates_script_subtags(value: str | None, expected: str | None) -> None:
    assert tr.canonical_code(value) == expected


# --- reading the sample files ---------------------------------------------


def test_sample_translations_dir_is_beside_the_project(tmp_path: Path) -> None:
    # .umbra is gitignored runtime state; the translations are project content.
    assert tr.sample_translations_dir(tmp_path) == tmp_path / "samples" / "translations"


def test_load_sample_translations_reads_every_file(tmp_path: Path) -> None:
    directory = tr.sample_translations_dir(tmp_path)
    _write_sample(directory, "en", ["Welcome.", "Hello."])
    _write_sample(directory, "zh", ["欢迎。", "你好。"])

    files = tr.load_sample_translations(tmp_path)

    assert set(files) == {"en", "zh"}
    assert files["zh"][(1, 0)] == "欢迎。"
    assert files["zh"][(1, 1)] == "你好。"


def test_load_sample_translations_survives_a_broken_file(tmp_path: Path) -> None:
    directory = tr.sample_translations_dir(tmp_path)
    _write_sample(directory, "zh", ["欢迎。", "你好。"])
    (directory / "sample_ch1_BROKEN.json").write_text("{not json", encoding="utf-8")

    files = tr.load_sample_translations(tmp_path)

    assert files["zh"][(1, 0)] == "欢迎。"


def test_load_sample_translations_missing_directory_is_empty(tmp_path: Path) -> None:
    assert tr.load_sample_translations(tmp_path) == {}


# --- resolving ------------------------------------------------------------


def test_sample_file_supplies_a_translation_the_db_does_not_have(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    directory = tr.sample_translations_dir(tmp_path)
    _write_sample(directory, "en", ["Welcome.", "Hello."])
    _write_sample(directory, "zh", ["欢迎。", "你好。"])

    resolved = tr.resolved_translations(conn, tmp_path, "zh")

    assert resolved[(1, 0)].text == "欢迎。"
    assert resolved[(1, 1)].text == "你好。"
    assert resolved[(1, 0)].origin == "sample"


def test_english_is_not_reported_as_a_translation(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    # The English file is the reference the other files are compared against,
    # so English must never come back as a translation of itself.
    directory = tr.sample_translations_dir(tmp_path)
    _write_sample(directory, "en", ["Welcome.", "Hello."])

    resolved = tr.resolved_translations(conn, tmp_path, "en")

    assert resolved == {}


def test_db_row_wins_over_the_sample_file(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    directory = tr.sample_translations_dir(tmp_path)
    _write_sample(directory, "en", ["Welcome.", "Hello."])
    _write_sample(directory, "zh", ["欢迎。", "你好。"])
    conn.execute(
        """
        INSERT INTO utterance_translations
            (project_id, utterance_id, language_code, language_name, text, status, origin)
        VALUES (1, 1, 'zh', 'Chinese', '从数据库来的欢迎。', 'confirmed', 'umbratrans')
        """
    )
    conn.commit()

    resolved = tr.resolved_translations(conn, tmp_path, "zh")

    assert resolved[(1, 0)].text == "从数据库来的欢迎。"
    assert resolved[(1, 0)].status == "confirmed"
    assert resolved[(1, 0)].origin == "umbratrans"
    # The second utterance still comes from the file.
    assert resolved[(1, 1)].text == "你好。"


def test_db_row_holding_the_english_source_is_a_placeholder_not_a_translation(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    # This is the shape of the one row that was actually in Deal Hunter's DB:
    # the English text stored under a zh-Hans code. Rendering it would show
    # English under a Chinese heading and claim it was translated.
    conn.execute(
        """
        INSERT INTO utterance_translations
            (project_id, utterance_id, language_code, language_name, text, status, origin)
        VALUES (1, 1, 'zh-Hans', 'Chinese', 'Welcome to the book.', 'draft', 'umbratrans')
        """
    )
    conn.commit()

    resolved = tr.resolved_translations(conn, tmp_path, "zh")

    assert resolved == {}


def test_script_subtag_rows_are_matched_to_the_dropdown_code(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    conn.execute(
        """
        INSERT INTO utterance_translations
            (project_id, utterance_id, language_code, language_name, text, status, origin)
        VALUES (1, 1, 'zh-Hans', 'Chinese', '从数据库来的欢迎。', 'draft', 'umbratrans')
        """
    )
    conn.commit()

    resolved = tr.resolved_translations(conn, tmp_path, "zh")

    assert resolved[(1, 0)].text == "从数据库来的欢迎。"


def test_a_row_for_another_language_is_not_returned(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    conn.execute(
        """
        INSERT INTO utterance_translations
            (project_id, utterance_id, language_code, language_name, text, status, origin)
        VALUES (1, 1, 'ru', 'Russian', 'Добро пожаловать.', 'draft', 'umbratrans')
        """
    )
    conn.commit()

    resolved = tr.resolved_translations(conn, tmp_path, "zh")

    assert resolved == {}


def test_missing_translations_table_is_not_an_error(
    tmp_path: Path,
) -> None:
    c = sqlite3.connect(":memory:")
    c.row_factory = sqlite3.Row
    c.executescript(
        """
        CREATE TABLE utterances (
            id INTEGER PRIMARY KEY, chapter_id INTEGER, utterance_index INTEGER,
            speaker_name TEXT, text TEXT
        );
        """
    )
    c.commit()
    try:
        assert tr.resolved_translations(c, tmp_path, "zh") == {}
    finally:
        c.close()


def test_untranslated_target_file_is_not_counted_as_translated(
    conn: sqlite3.Connection, tmp_path: Path
) -> None:
    # A target file whose text is identical to the source file has not been
    # translated; reporting it as translated would hide the gap.
    directory = tr.sample_translations_dir(tmp_path)
    _write_sample(directory, "en", ["Welcome.", "Hello."])
    _write_sample(directory, "ru", ["Welcome.", "Hello."])

    resolved = tr.resolved_translations(conn, tmp_path, "ru")

    assert resolved == {}