japanese/tests/test_two_section_review.py

"""
测试新的两段式待确认文件 + done 信号机制。
"""
import pytest
import tempfile
import shutil
from pathlib import Path
from pl_japanese.cleaner import TaskManager
from pl_japanese.cleaner.workflow import CleanerWorkflow


@pytest.fixture
def temp_dir():
    """每个测试独立的临时目录"""
    tmp = Path(tempfile.mkdtemp())
    yield tmp
    shutil.rmtree(tmp, ignore_errors=True)


@pytest.fixture(autouse=True)
def reset_pinyin_dict():
    """每个测试前重置拼音字典单例"""
    from pl_japanese.cleaner.pinyin_overrides import DEFAULT_DICT
    original_path = DEFAULT_DICT.source_path
    original_word = DEFAULT_DICT.word_override.copy()
    original_kana = DEFAULT_DICT.kana_hint.copy()
    
    yield
    
    DEFAULT_DICT.source_path = original_path
    DEFAULT_DICT.word_override = original_word
    DEFAULT_DICT.kana_hint = original_kana


def test_two_section_file_generation(temp_dir):
    """测试两段式待确认文件生成"""
    tmp = temp_dir
    tm = TaskManager(tasks_root=str(tmp / 'tasks'))
    src = tmp / 'src.txt'
    src.write_text(
        '議会:ぎかい:\n'       # 多音字 → 待确认
        '社会:しゃかい:\n'     # 多音字 → 待确认
        '飲料:いんりょう:\n'   # 自动通过
        'IT:アイティー:\n',    # 无汉字 → 建议skip
        encoding='utf-8')
    
    task = tm.create_task('t1', source=str(src))
    main_path = str(tmp/'voc.txt')
    skipped_path = str(tmp/'skip.txt')
    wf = CleanerWorkflow(task, main=main_path, skipped=skipped_path)
    
    result = wf.run()
    assert result['status'] == 'reviewing'
    assert result['process']['auto_pass'] == 1
    assert result['process']['need_review'] == 3  # 2多音字 + 1skip建议
    
    # 检查文件结构
    review_file = Path(task.review_file)
    content = review_file.read_text(encoding='utf-8')
    lines = content.splitlines()
    
    # 第一段标记
    assert '自动通过' in lines[0]
    # 飲料在第一段
    assert '飲|料:いん|りょう:yin|liao' in content
    # 第二段标记
    assert '待确认' in content
    # 待确认条目
    assert '議|会:ぎ|かい:yi|hui' in content
    assert '# 多音字' in content
    # skip建议（无汉字条目拼音为空）
    assert 'skip IT:アイティー:' in content


def test_apply_review_with_modifications(temp_dir):
    """测试人工review后应用修改"""
    tmp = temp_dir
    tm = TaskManager(tasks_root=str(tmp / 'tasks'))
    src = tmp / 'src.txt'
    src.write_text('議会:ぎかい:\n社会:しゃかい:\n飲料:いんりょう:\n', encoding='utf-8')
    
    task = tm.create_task('t1', source=str(src))
    main_path = str(tmp/'voc.txt')
    skipped_path = str(tmp/'skip.txt')
    wf = CleanerWorkflow(task, main=main_path, skipped=skipped_path)
    wf.run()
    
    # 模拟人工review：修改第一条拼音，第二条加skip
    review_file = Path(task.review_file)
    content = review_file.read_text(encoding='utf-8')
    # 修改拼音
    content = content.replace('議|会:ぎ|かい:yi|hui', '議|会:ぎ|かい:yi|kuai')
    # 加skip标记
    content = content.replace('社|会:しゃ|かい:she|hui', 'skip 社|会:しゃ|かい:she|hui')
    content += 'done\n'
    review_file.write_text(content, encoding='utf-8')
    
    # 应用
    result = wf.run()
    assert result['status'] == 'ready'
    assert result['apply_result']['accepted'] == 2  # 飲料 + 議会
    assert result['apply_result']['skipped'] == 1   # 社会
    
    # 检查落地文件
    proc = wf.processor
    auto = proc._read_clean_lines(proc.auto_done)
    assert len(auto) == 2
    assert '飲|料:いん|りょう:yin|liao' in auto
    assert '議|会:ぎ|かい:yi|kuai' in auto  # 修正后的拼音
    
    skip = proc._read_clean_lines(proc.skip)
    assert len(skip) == 1
    assert '社|会:しゃ|かい:she|hui' in skip


def test_format_error_rollback_to_third_section(temp_dir):
    """测试格式错误回退到第三段"""
    tmp = temp_dir
    tm = TaskManager(tasks_root=str(tmp / 'tasks'))
    src = tmp / 'src.txt'
    src.write_text('議会:ぎかい:\n社会:しゃかい:\n', encoding='utf-8')
    
    task = tm.create_task('t1', source=str(src))
    main_path = str(tmp/'voc.txt')
    skipped_path = str(tmp/'skip.txt')
    wf = CleanerWorkflow(task, main=main_path, skipped=skipped_path)
    wf.run()
    
    # 模拟人工错误修改（误删分隔符）
    review_file = Path(task.review_file)
    content = review_file.read_text(encoding='utf-8')
    content = content.replace('議|会:ぎ|かい:yi|hui', '議|会ぎ|かいyi|hui')
    content += 'done\n'
    review_file.write_text(content, encoding='utf-8')
    
    # 尝试应用 → 格式错误
    result = wf.run()
    assert result['action'] == 'need_human'
    assert '格式错误' in result['message']
    
    # 检查回退结果：三段结构
    content_after = review_file.read_text(encoding='utf-8')
    lines = content_after.splitlines()
    
    # 第一段：自动通过（无）
    assert '自动通过' in content_after
    # 第二段：待确认（正确的）
    assert '待确认' in content_after
    assert '社|会:しゃ|かい:she|hui' in content_after
    # 第三段：格式错误
    assert '格式错误，请修正' in content_after
    assert '議|会ぎ|かいyi|hui    # 格式错误:' in content_after
    # done 被删除
    assert 'done' not in content_after


def test_skip_suggestion_can_be_accepted_or_modified(temp_dir):
    """测试skip建议可以接受或修改"""
    tmp = temp_dir
    tm = TaskManager(tasks_root=str(tmp / 'tasks'))
    src = tmp / 'src.txt'
    src.write_text('IT:アイティー:\nCS:シーエス:\n', encoding='utf-8')
    
    task = tm.create_task('t1', source=str(src))
    main_path = str(tmp/'voc.txt')
    skipped_path = str(tmp/'skip.txt')
    wf = CleanerWorkflow(task, main=main_path, skipped=skipped_path)
    wf.run()
    
    review_file = Path(task.review_file)
    content = review_file.read_text(encoding='utf-8')
    
    # IT保持skip，CS删掉skip前缀（改为保留）
    # 注意：无汉字条目的拼音段为空，格式是 CS:シーエス:
    content = content.replace('skip CS:シーエス:', 'CS:シーエス:')
    content += 'done\n'
    review_file.write_text(content, encoding='utf-8')
    
    result = wf.run()
    assert result['status'] == 'ready'
    
    proc = wf.processor
    auto = proc._read_clean_lines(proc.auto_done)
    skip = proc._read_clean_lines(proc.skip)
    
    assert 'CS:シーエス:' in auto  # 删skip后进auto
    assert 'IT:アイティー:' in skip  # 保持skip


def test_tilde_preserved_in_output(temp_dir):
    """测试～被保留在输出中（占一段，拼音留空）"""
    tmp = temp_dir
    tm = TaskManager(tasks_root=str(tmp / 'tasks'))
    src = tmp / 'src.txt'
    src.write_text('～人:アメリカじん:\n～員:かいいん:\n', encoding='utf-8')
    
    task = tm.create_task('t1', source=str(src))
    main_path = str(tmp/'voc.txt')
    skipped_path = str(tmp/'skip.txt')
    wf = CleanerWorkflow(task, main=main_path, skipped=skipped_path)
    result = wf.run()
    
    # ～ 条目全部自动通过，直接进 auto_done（不经过待确认）
    assert result['status'] == 'ready'
    
    proc = wf.processor
    auto = proc._read_clean_lines(proc.auto_done)
    
    # ～ 占一段，拼音留空
    assert '～|人:アメリカ|じん:|ren' in auto
    assert '～|員:かい|いん:|yuan' in auto