This repository has been archived on 2026-10-02. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
NeteaseUrl/music_downloader/cover.py
T
snowy-yang 6288a98ecc refactor: 拆分 4 个巨石模块为包,清理过时注释并删除作废文档
为什么:单文件 1100~1900 行的巨石模块让「改一处要通读全文」;同时注释里残留
大量已删台账的编号,以及「Flask/FastAPI 双栈迁移期」这类已失实的架构描述,
会误导后来者按不复存在的架构去改代码。

拆分(对外导入路径不变,行为零变更):
- music_downloader.py  -> music_downloader/(models/downloader/cover/_http/_log/_concurrency)
- auth_system.py       -> auth_system/(db/security/users/credits/ledger/mail/cards/backup...)
- music_api.py         -> music_api/(api/client/crypto/qr/cookies/compat/constants/errors)
- services/runtime.py  -> services/(api_service/_shared/_taskcore/downloads/batch/podcast/
  streaming/tasks);services/runtime.py 保留为兼容门面

兼容性做法:各包用 __init__.py 以 __all__ 原样再导出全部符号,既有调用点
(auth_system.xxx / services.xxx / from music_api import Y)零改动;路径常量改为经包属性
动态读取(auth_system.DB_PATH),否则测试的 monkeypatch 会静默失效并写到生产库。

修复 9 处代码损伤(上一轮注释清理把 `()` 当「空括号残留」删掉所致;用「与拆分前逐字
等价」校验逐一定位,静态检查只会报出其中 4 处):
- core/config.py  ).strip() or '*'
- core/state.py   threading.Lock() x2
- qr_login.py     datetime.now() / .strip()
- music_downloader/downloader.py  .lower()
- services/podcast.py  set() / time.time() x2

清理与删除:
- 删除 部署文档.md / 功能介绍.md / README_私有包警告.txt
- 清理指向已删台账与文档的过时编号注释;把「迁移期 / 两栈 / 双框架 / 已迁移」等
  失实描述改写为现状描述(保留 `<-> Flask /path` 映射表等仍有价值的契约来源说明)
- _mkpkg.py:公开包剔除清单去掉已不存在的文档名;私有包警告改为自包含
- pyproject.toml:补 pythonpath = ["."],使裸跑 pytest 不再 import core 失败

验证:ruff 全通过;pyright 0 errors / 0 warnings(102 文件);pytest 144 passed;
真实 uvicorn 冒烟 6 条路由符合预期且生产 data/ 指纹未变;忽略注释的逐字等价性校验通过。
2026-09-26 12:37:48 +08:00

69 lines
2.4 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# -*- coding: utf-8 -*-
"""封面字节缓存(同专辑只从网络拉一次)。
(自 music_downloader.py 拆出,行为与原实现一致。)
"""
from __future__ import annotations
import threading
from collections import deque
from ._http import (
get_http_session,
)
# 封面字节缓存:同一专辑封面只从网络拉取一次(ffmpeg 写元数据 与 ZIP 打包 共用)
# 容量抽成常量:淘汰循环按它判定,不再读 deque.maxlen(该属性类型为 int|None,
# 一旦有人把 deque 改成无界,len() >= None 会直接 TypeError)
_COVER_CACHE_MAX = 128
_COVER_CACHE: dict[str, bytes] = {}
_COVER_CACHE_ORDER: deque[str] = deque(maxlen=_COVER_CACHE_MAX)
_COVER_CACHE_LOCK = threading.Lock()
def fetch_cover_bytes(pic_url: str) -> bytes | None:
"""下载并缓存封面图片字节;失败返回 None(字节预算 10MB,拒绝超大盘点阵/伪装文件)"""
if not pic_url:
return None
with _COVER_CACHE_LOCK:
cached = _COVER_CACHE.get(pic_url)
if cached:
return cached
try:
resp = get_http_session().get(pic_url, timeout=15, stream=True)
try:
resp.raise_for_status()
try:
declared = int(resp.headers.get('Content-Length') or 0)
except ValueError:
declared = 0 # 垃圾 Content-Length 不炸,按未知长度走流式
if declared > 10 * 1024 * 1024:
return None
# bytearray 就地追加(原 bytes += 是 O(n²) 拷贝,10MB 封面最坏多拷上百 MB)
data = bytearray()
for chunk in resp.iter_content(chunk_size=64 * 1024):
data += chunk
if len(data) > 10 * 1024 * 1024:
return None
if not data:
return None
data = bytes(data)
finally:
# 所有路径统一关闭响应(原 Content-Length 异常路径漏关)
resp.close()
except Exception:
return None
with _COVER_CACHE_LOCK:
if pic_url not in _COVER_CACHE:
# 简单 FIFO 淘汰,防止缓存无界增长
while len(_COVER_CACHE_ORDER) >= _COVER_CACHE_MAX:
evict = _COVER_CACHE_ORDER.popleft()
_COVER_CACHE.pop(evict, None)
_COVER_CACHE[pic_url] = data
_COVER_CACHE_ORDER.append(pic_url)
return data