blank/tests/cleaning/test_splitter.py

"""
Unit tests for ChapterSplitter.
"""

import pytest
from src.cleaning.splitter import ChapterSplitter, ChapterSplitterError


class TestChapterSplitter:
    """Test suite for ChapterSplitter."""

    @pytest.fixture
    def splitter(self):
        """Create a ChapterSplitter instance with low min length for testing."""
        return ChapterSplitter(min_chapter_length=10, merge_short_chapters=False)

    def test_split_chinese_numerals(self, splitter):
        """Test splitting Chinese chapter titles (第一章, etc.)."""
        text = """第一章 开始

这是第一章的内容。

第二章 继续

这是第二章的内容。"""

        chapters = splitter.split(text)
        assert len(chapters) == 2
        assert "第一章" in chapters[0].title
        assert "第二章" in chapters[1].title

    def test_split_arabic_chinese(self, splitter):
        """Test splitting Arabic-Chinese chapter titles (第1章, etc.)."""
        text = """第1章 开始

这是第一章的内容。

第2章 继续

这是第二章的内容。"""

        chapters = splitter.split(text)
        assert len(chapters) == 2
        assert "第1章" in chapters[0].title
        assert "第2章" in chapters[1].title

    def test_split_english_format(self, splitter):
        """Test splitting English chapter titles."""
        text = """Chapter 1 The Beginning

This is chapter one.

Chapter 2 The Journey Continues

This is chapter two."""

        chapters = splitter.split(text)
        assert len(chapters) >= 2
        assert "Chapter 1" in chapters[0].title

    def test_split_numbered_format(self, splitter):
        """Test splitting numbered chapter titles."""
        text = """1. The Start

Content here.

2. The Middle

More content."""

        chapters = splitter.split(text)
        assert len(chapters) >= 2
        assert "1." in chapters[0].title

    def test_split_date_format(self, splitter):
        """Test splitting date format chapters."""
        text = """2024年3月15日

这是第一天的内容。

2024年3月16日

这是第二天的内容。"""

        chapters = splitter.split(text)
        assert len(chapters) >= 2

    def test_split_volume_format(self, splitter):
        """Test splitting volume format (第一卷, etc.)."""
        text = """第一卷 命运的开始

这是第一卷的内容。

第二卷 奇遇

这是第二卷的内容。"""

        chapters = splitter.split(text)
        assert len(chapters) >= 2

    def test_split_with_brackets(self, splitter):
        """Test splitting bracket format chapters."""
        text = """[1] 开始

内容。

[2] 继续

更多内容。"""

        chapters = splitter.split(text)
        assert len(chapters) >= 2

    def test_empty_text_returns_empty_list(self, splitter):
        """Test that empty text returns empty chapter list."""
        chapters = splitter.split("")
        assert chapters == []

    def test_no_chapter_titles_returns_single_chapter(self, splitter):
        """Test that text without chapter titles becomes one chapter."""
        text = "这是一段没有章节标题的文本。\n第二行内容。"
        chapters = splitter.split(text)
        assert len(chapters) == 1
        assert chapters[0].title == "全文"

    def test_chapter_char_count(self, splitter):
        """Test that chapter character count is correct."""
        text = """第一章 测试章节

这是第一章的内容，包含一些文字。

第二章 第二个章节

这是第二章的内容。"""

        chapters = splitter.split(text)
        assert chapters[0].char_count > 0
        assert len(chapters[0].content) == chapters[0].char_count

    def test_chapter_positions(self, splitter):
        """Test that chapter start/end positions are correct."""
        text = "第一章\n内容1\n\n第二章\n内容2"
        chapters = splitter.split(text)
        assert len(chapters) >= 2
        if chapters[0].start_pos is not None:
            assert chapters[0].start_pos == 0

    def test_detect_chapter_title_chinese(self, splitter):
        """Test chapter title detection for Chinese format."""
        assert splitter.is_chapter_title("第一章 开始")
        assert splitter.is_chapter_title("第123章")
        assert splitter.is_chapter_title("第十章 约战")

    def test_detect_chapter_title_english(self, splitter):
        """Test chapter title detection for English format."""
        assert splitter.is_chapter_title("Chapter 1")
        assert splitter.is_chapter_title("Chapter One - The Beginning")

    def test_detect_chapter_title_numbered(self, splitter):
        """Test chapter title detection for numbered format."""
        assert splitter.is_chapter_title("1. Start")
        assert splitter.is_chapter_title("123. End")

    def test_detect_chapter_title_date(self, splitter):
        """Test chapter title detection for date format."""
        assert splitter.is_chapter_title("2024年3月15日")
        assert splitter.is_chapter_title("2024年12月1日")

    def test_detect_not_chapter_title(self, splitter):
        """Test that non-titles are correctly identified."""
        assert not splitter.is_chapter_title("这是一个普通的句子")
        assert not splitter.is_chapter_title("")
        assert not splitter.is_chapter_title("hello world")

    def test_get_chapter_count(self, splitter):
        """Test getting chapter count without full split."""
        text = """第一章 开始

内容。

第二章 继续

更多内容。

第三章 结束

最后内容。"""

        count = splitter.get_chapter_count(text)
        assert count == 3

    def test_get_chapter_count_no_chapters(self, splitter):
        """Test getting chapter count for text without chapters."""
        text = "这是一段没有章节的文本。"
        count = splitter.get_chapter_count(text)
        assert count == 1

    def test_preview_chapters(self, splitter):
        """Test getting chapter previews."""
        text = """第一章 开始

这是第一章的内容，包含一些文字。

第二章 继续

这是第二章的内容，包含更多文字。"""

        previews = splitter.preview_chapters(text, preview_length=50)
        assert len(previews) >= 2
        assert "第一章" in previews[0]
        assert "第二章" in previews[1]

    def test_merge_short_chapters_enabled(self):
        """Test that short chapters are merged when enabled."""
        text = """第一章 开始

短。

第二章 中间

这是第二章较长的内容。

第三章 结尾

也短。"""

        splitter = ChapterSplitter(min_chapter_length=50, merge_short_chapters=True)
        chapters = splitter.split(text)
        # Short chapters should be merged with adjacent ones
        assert len(chapters) <= 3

    def test_merge_short_chapters_disabled(self):
        """Test that short chapters are kept when merging disabled."""
        text = """第一章 开始

短内容。

第二章 继续

更多内容。"""

        splitter = ChapterSplitter(min_chapter_length=1000, merge_short_chapters=False)
        chapters = splitter.split(text)
        # All chapters should be kept
        assert len(chapters) == 2

    def test_custom_patterns(self):
        """Test using custom chapter patterns."""
        text = """EPISODE 1 Start

Content.

EPISODE 2 Middle

More content."""

        custom_patterns = [(r'^EPISODE\s+\d+', 1)]
        splitter = ChapterSplitter(
            min_chapter_length=10,
            merge_short_chapters=False,
            custom_patterns=custom_patterns
        )
        chapters = splitter.split(text)
        assert len(chapters) >= 2
        assert "EPISODE 1" in chapters[0].title

    def test_mixed_pattern_types(self, splitter):
        """Test handling mixed chapter pattern types."""
        text = """第一章 开始

内容。

Chapter 2 Middle

English content.

第三章 End

中文内容。"""

        chapters = splitter.split(text)
        # Should detect all chapters despite mixed formats
        assert len(chapters) >= 3

    def test_chapter_with_special_characters(self, splitter):
        """Test chapters with special characters in title."""
        text = """第一章：命运的齿轮！

内容。

第二章 - 新的开始

更多内容。"""

        chapters = splitter.split(text)
        assert len(chapters) >= 2

    def test_large_chapter_count(self, splitter):
        """Test handling many chapters."""
        # Create text with 100 chapters
        lines = []
        for i in range(1, 101):
            lines.append(f"第{i}章")
            lines.append(f"这是第{i}章的内容。\n")

        text = "\n".join(lines)
        chapters = splitter.split(text)
        assert len(chapters) == 100

    def test_consecutive_chapter_titles(self, splitter):
        """Test handling consecutive chapter titles without content."""
        text = """第一章

第二章

这是第二章的内容。

第三章

这是第三章的内容。"""

        chapters = splitter.split(text)
        # Should handle empty chapters gracefully
        assert len(chapters) >= 2

    def test_chapter_with_leading_whitespace(self, splitter):
        """Test chapter titles with leading whitespace."""
        text = """  第一章 开始

内容。

  第二章 继续

更多内容。"""

        chapters = splitter.split(text)
        assert len(chapters) >= 2

    def test_detect_chapter_title_returns_priority(self, splitter):
        """Test that detect_chapter_title returns priority."""
        result = splitter.detect_chapter_title("第一章 开始")
        assert result is not None
        priority, title = result
        assert isinstance(priority, int)
        assert isinstance(title, str)

    def test_word_count_property(self, splitter):
        """Test chapter word_count property."""
        text = """第一章 测试

这是测试内容。"""

        chapters = splitter.split(text)
        assert chapters[0].word_count > 0

    def test_len_operator(self, splitter):
        """Test len() operator on Chapter."""
        text = """第一章 测试

这是测试内容。"""

        chapters = splitter.split(text)
        assert len(chapters[0]) == chapters[0].char_count