fix(txt): merge scene-break sections into the preceding chapter (#4063) (#4207)

The TXT-to-EPUB segment regex splits on dash dividers (`-{8,}`), which
authors commonly use as in-chapter scene breaks. Each heading-less section
after such a divider was emitted as its own chapter — a numbered paragraph
fallback chapter, or a chapter titled after a stray sentence — flooding the
generated TOC with entries that aren't real chapters.

Mark chapters with whether their title came from a detected heading, and
merge heading-less chapters into the preceding detected chapter instead of
pushing them as separate TOC entries. Fully heading-less text still chunks
into numbered fallback chapters as before.

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

apps/readest-app/src/__tests__/utils/txt-converter.test.ts

@@ -346,6 +346,86 @@ describe('TxtToEpubConverter', () => {
   });
 });
 
+describe('scene-break dividers do not pollute the TOC (issue #4063)', () => {
+  const zhMetadata: TestMetadata = {
+    bookTitle: 'Test',
+    author: '',
+    language: 'zh',
+    identifier: 'test',
+  };
+  const option = { linesBetweenSegments: 8, fallbackParagraphsPerChapter: 100 };
+  const divider = '-'.repeat(37);
+
+  it('keeps a single chapter when dash dividers split scene breaks', () => {
+    const converter = new TxtToEpubConverter() as unknown as TxtConverterFlowPrivateAPI;
+    const text = [
+      '第15章',
+      '这是第十五章的开头内容。',
+      divider,
+      '场景切换后的内容继续。',
+      divider,
+      '又一个场景的内容。',
+    ].join('\n');
+
+    const chapters = converter.extractChapters(text, zhMetadata, option);
+
+    expect(chapters.length).toBe(1);
+    expect(chapters[0]!.title).toContain('第15章');
+    expect(chapters[0]!.content).toContain('场景切换后的内容继续');
+    expect(chapters[0]!.content).toContain('又一个场景的内容');
+  });
+
+  it('merges mid-chapter content split off before the next heading', () => {
+    const converter = new TxtToEpubConverter() as unknown as TxtConverterFlowPrivateAPI;
+    const text = [
+      '第1章',
+      '绿水青山就是金山银山。',
+      divider,
+      '分隔符之后继续讲述的内容。',
+      '第2章',
+      '新的章节正式开始了。',
+    ].join('\n');
+
+    const chapters = converter.extractChapters(text, zhMetadata, option);
+
+    expect(chapters.map((c) => c.title)).toEqual(['第1章', '第2章']);
+    expect(chapters[0]!.content).toContain('分隔符之后继续讲述的内容');
+    expect(chapters[0]!.content).not.toContain('<h3>');
+  });
+
+  it('still chunks heading-less plain text by paragraph fallback', () => {
+    const converter = new TxtToEpubConverter() as unknown as TxtConverterFlowPrivateAPI;
+    const paragraphs = Array.from({ length: 250 }, (_, i) => `段落${i + 1}`).join('\n');
+
+    const chapters = converter.extractChapters(paragraphs, zhMetadata, option);
+
+    expect(chapters.length).toBe(3);
+  });
+
+  it('merges divider-split scene breaks in the chunked file path', async () => {
+    const converter = new TxtToEpubConverter() as unknown as TxtConverterFlowPrivateAPI;
+    const text = [
+      '第15章',
+      '这是第十五章的开头内容。',
+      divider,
+      '场景切换后的内容继续。',
+      divider,
+      '又一个场景的内容。',
+    ].join('\n');
+    const file = new File([text], 'sample.txt');
+
+    const chapters = await converter.extractChaptersFromFileBySegments(
+      file,
+      'utf-8',
+      zhMetadata,
+      option,
+    );
+
+    expect(chapters.length).toBe(1);
+    expect(chapters[0]!.title).toContain('第15章');
+  });
+});
+
 describe('extractTxtFilenameMetadata', () => {
   it('extracts the title from CJK 《》 brackets', () => {
     expect(extractTxtFilenameMetadata('《三体》.txt')).toEqual({ title: '三体' });

apps/readest-app/src/utils/txt.ts

@@ -54,6 +54,11 @@ interface Chapter {
   title: string;
   content: string;
   isVolume: boolean;
+  // True when the title came from a detected chapter heading. Chapters whose
+  // content was not found under a heading (paragraph fallback, or stray text
+  // split off by the segment regex) are merged into the preceding detected
+  // chapter instead of becoming bogus TOC entries. See issue #4063.
+  detected?: boolean;
 }
 
 interface Txt2EpubOptions {
@@ -242,12 +247,30 @@ export class TxtToEpubConverter {
         option,
         chapters.length,
       );
-      chapters.push(...segmentChapters);
+      this.appendSegmentChapters(chapters, segmentChapters);
     }
 
     return chapters;
   }
 
+  /**
+   * Append a segment's chapters to the running list. The segment regex also
+   * splits on dash dividers, which authors frequently use as in-chapter scene
+   * breaks; the content after such a divider has no heading of its own. When a
+   * heading-less chapter follows a detected chapter, merge its content into
+   * that chapter instead of emitting a separate (bogus) TOC entry. See #4063.
+   */
+  private appendSegmentChapters(chapters: Chapter[], segmentChapters: Chapter[]): void {
+    for (const chapter of segmentChapters) {
+      const previous = chapters[chapters.length - 1];
+      if (!chapter.detected && previous?.detected) {
+        previous.content += chapter.content.replace(/^<h[1-6][^>]*>[\s\S]*?<\/h[1-6]>/, '');
+      } else {
+        chapters.push(chapter);
+      }
+    }
+  }
+
   private probeChapterCount(
     txtContent: string,
     metadata: Metadata,
@@ -286,7 +309,7 @@ export class TxtToEpubConverter {
         option,
         chapters.length,
       );
-      chapters.push(...segmentChapters);
+      this.appendSegmentChapters(chapters, segmentChapters);
     }
     return chapters;
   }
@@ -545,7 +568,7 @@ export class TxtToEpubConverter {
         const formattedSegment = this.formatSegment(chunks.join('\n'));
         const title = `${chapterOffset + chapters.length + 1}`;
         const content = `<h2>${title}</h2><p>${formattedSegment}</p>`;
-        chapters.push({ title, content, isVolume: false });
+        chapters.push({ title, content, isVolume: false, detected: false });
       }
       return chapters;
     }
@@ -568,6 +591,7 @@ export class TxtToEpubConverter {
         title: escapeXml(title),
         content: `${headTitle}<p>${formattedSegment}</p>`,
         isVolume,
+        detected: true,
       });
     }
 
@@ -582,6 +606,7 @@ export class TxtToEpubConverter {
         title: escapeXml(segmentTitle),
         content: `<h3></h3><p>${formattedSegment}</p>`,
         isVolume: false,
+        detected: false,
       });
     }