openclaw-config/skill-workshop/proposals/calibre-cleaner-20260624-cdbeb219a3/rollback.json

{"schema":"openclaw.skill-workshop.rollback.v1","proposalId":"calibre-cleaner-20260624-cdbeb219a3","writtenAt":"2026-06-24T03:03:30.908Z","targetSkillFile":"/home/yangxuan/.openclaw/workspace/storage/skills/calibre-cleaner/SKILL.md","action":"update","previousContent":"---\nname: \"calibre-cleaner\"\ndescription: \"清洗 Calibre 书库：修复损坏书名/提取作者/检测编码损坏、小文件的批量删除\"\n---\n\n# Calibre 书库清洗 Skill\n\n## 概述\n\nCalibre 书库整理三件套：\n\n1. **书名清洗** —— 修复从 txt 导入产生的 HTML 标签/长文本/乱码书名\n2. **作者提取** —— 从书名中自动提取作者名（匹配 `作者：xxx`、`-xxx`、`BY xxx` 等格式）\n3. **脏数据清理** —— 检测并删除编码损坏的 txt、小于 10KB 的垃圾文件\n\n## 前置条件\n\n- 书库路径：`/home/yangxuan/calibre/books`\n- 元数据文件：`/home/yangxuan/calibre/books/metadata.db`\n- 运行前须停止其他 Calibre 进程（桌面版、Docker Calibre-Web）\n- 运行前 **必须备份** `metadata.db`\n\n---\n\n## 一、书名清洗 & 作者提取\n\n（详见 skill 原内容，此处保留完整清洗逻辑）\n\n---\n\n## 二、脏数据清理\n\n### 2.1 检测编码损坏的 TXT\n\n检测条件：中文比例 < 10% 且含有乱码字符（`Ŀ¼ĴƿȿɿʼÈÌÒÙ（` 等 UTF-8 编码错误字）\n\n```python\nimport sqlite3, os\n\nbad_char_set = set('Ŀ¼ĴƿȿɿʼÈÌÒÙ（')\nbad_ids = []\n\nfor bid, title, bpath, fname in rows:\n    full_path = os.path.join('/home/yangxuan/calibre/books', bpath, fname + '.txt')\n    if not os.path.exists(full_path):\n        continue\n    with open(full_path, 'rb') as f:\n        raw = f.read(20480)\n    text = raw.decode('utf-8', errors='replace')\n    chinese = sum(1 for c in text if '\\u4e00' <= c <= '\\u9fff')\n    total = len(text)\n    ratio = chinese / total if total > 0 else 0\n    has_bad = any(c in bad_char_set for c in text)\n    if ratio < 0.10 and has_bad:\n        bad_ids.append(bid)\n```\n\n### 2.2 检测小文件\n\n```python\nc.execute('''\n    SELECT b.id, b.path FROM books b \n    JOIN data d ON d.book = b.id \n    WHERE d.format = 'TXT' AND d.uncompressed_size < 10240\n''')\n```\n\n### 2.3 批量删除（通过 SQLite 直写）\n\n停掉 calibre 进程后，通过 SQLite 删除记录并清理文件系统：\n\n```sql\nDELETE FROM data WHERE book = ?\nDELETE FROM books_authors_link WHERE book = ?\nDELETE FROM books_languages_link WHERE book = ?\nDELETE FROM books_publishers_link WHERE book = ?\nDELETE FROM books_series_link WHERE book = ?\nDELETE FROM books_tags_link WHERE book = ?\nDELETE FROM comments WHERE book = ?\nDELETE FROM identifiers WHERE book = ?\nDELETE FROM books WHERE id = ?\n```\n\n同时删除对应目录：`shutil.rmtree(os.path.join('/home/yangxuan/calibre/books', bpath))`\n\n### 2.4 完整删除脚本\n\n```python\nimport sqlite3, os, shutil\n\ndb_path = '/home/yangxuan/calibre/books/metadata.db'\nconn = sqlite3.connect(db_path)\nc = conn.cursor()\n\nc.execute('''\n    SELECT b.id, b.title, b.path, d.name, d.uncompressed_size\n    FROM books b \n    JOIN data d ON d.book = b.id \n    WHERE d.format = 'TXT'\n''')\nrows = c.fetchall()\n\nbad_char_set = set('Ŀ¼ĴƿȿɿʼÈÌÒÙ（')\nto_delete = []\n\nfor bid, title, bpath, fname, size in rows:\n    full_path = os.path.join('/home/yangxuan/calibre/books', bpath, fname + '.txt')\n    if not os.path.exists(full_path):\n        continue\n    with open(full_path, 'rb') as f:\n        raw = f.read(20480)\n    text = raw.decode('utf-8', errors='replace')\n    chinese = sum(1 for c in text if '\\u4e00' <= c <= '\\u9fff')\n    ratio = chinese / len(text) if len(text) > 0 else 0\n    has_bad = any(c in bad_char_set for c in text)\n    if (ratio < 0.10 and has_bad) or size < 10240:\n        to_delete.append((bid, bpath))\n\nprint(f\"待删除: {len(to_delete)}\")\nfor bid, bpath in to_delete:\n    c.execute('DELETE FROM data WHERE book=?', (bid,))\n    c.execute('DELETE FROM books_authors_link WHERE book=?', (bid,))\n    c.execute('DELETE FROM books_languages_link WHERE book=?', (bid,))\n    c.execute('DELETE FROM books_publishers_link WHERE book=?', (bid,))\n    c.execute('DELETE FROM books_series_link WHERE book=?', (bid,))\n    c.execute('DELETE FROM books_tags_link WHERE book=?', (bid,))\n    c.execute('DELETE FROM comments WHERE book=?', (bid,))\n    c.execute('DELETE FROM identifiers WHERE book=?', (bid,))\n    c.execute('DELETE FROM books WHERE id=?', (bid,))\n    full_dir = os.path.join('/home/yangxuan/calibre/books', bpath)\n    if os.path.exists(full_dir):\n        shutil.rmtree(full_dir)\n\nconn.commit()\nconn.close()\nprint(f\"完成，删除: {len(to_delete)}\")\n```\n\n---\n\n## 三、常用操作\n\n### 3.1 查看 XX 作者的书\n\n```bash\nsqlite3 /home/yangxuan/calibre/books/metadata.db \"\nSELECT b.id, b.title\nFROM books b JOIN books_authors_link l ON b.id = l.book\nJOIN authors a ON l.author = a.id\nWHERE a.name = 'Unknown'\nLIMIT 20;\n\"\n```\n\n### 3.2 统计作者分布\n\n```bash\nsqlite3 /home/yangxuan/calibre/books/metadata.db \"\nSELECT a.name, COUNT(*) as cnt\nFROM books b JOIN books_authors_link l ON b.id = l.book\nJOIN authors a ON l.author = a.id\nGROUP BY a.name ORDER BY cnt DESC LIMIT 20;\n\"\n```\n\n### 3.3 书库统计\n\n```bash\npython3 -c \"\nimport sqlite3\nconn = sqlite3.connect('/home/yangxuan/calibre/books/metadata.db')\nc = conn.cursor()\nc.execute('SELECT COUNT(*) FROM books')\nprint(f'总书籍: {c.fetchone()[0]}')\nc.execute('SELECT COUNT(*) FROM data WHERE format=\\\"TXT\\\"')\nprint(f'TXT 文件: {c.fetchone()[0]}')\nconn.close()\n\"\n```\n\n---\n\n## 四、操作流程（推荐顺序）\n\n```bash\ncd /home/yangxuan/calibre\n\n# 1. 停服务 + 备份\ndocker compose down\ncp books/metadata.db books/metadata.db.backup.$(date +%Y%m%d)\nkillall calibre 2>/dev/null\n\n# 2. 脏数据清理（编码损坏 + 小文件）\npython3 /path/to/clean_dirty_data.py\n\n# 3. 书名清洗 & 作者提取\ncalibre-debug /path/to/fix_titles.py\n\n# 4. 重启\ndocker compose up -d\n```\n\n---\n\n## 五、已知问题\n\n### `set_metadata` 报 `Errno 36 文件名过长`\n\n当 `author_sort` 字段超 80 字节时，Calibre 路径会触及 ext4 255 字节限制。先清理极长作者名：\n\n```bash\nsqlite3 /home/yangxuan/calibre/books/metadata.db \"\nUPDATE authors SET name='Unknown' WHERE LENGTH(name) > 30;\n\"\n```\n\n### WSL `/mnt/f` 删除 NTFS 文件权限不够\n\nWSL 的 drvfs 默认不带 metadata 选项，`sudo rm`、`os.remove`、`cmd.exe /c del` 均无法删除 NTFS 文件。不要在 WSL 下尝试跨分区删除操作，直接让 Windows 侧（文件资源管理器/PowerShell）操作。\n","previousContentHash":"dd59723daea61a10edc128f446e41d71c07b03cbe6064ea5140e260a21fce51e"}