306 lines
14 KiB
Python
306 lines
14 KiB
Python
# -*- coding: utf-8 -*-
|
||
r"""
|
||
YouTube Studio 内容管理器「导出当前视图 → 逗号分隔值 (.csv)」下载流程 —— 拦截/解码/自动保存/重名去重 参考实现。
|
||
|
||
## 已实证的下载生成机制(2026-08-21 抓包确认)
|
||
|
||
1. 前端点击「导出当前视图 → 逗号分隔值 (.csv)」后,向后端发起:
|
||
POST https://studio.youtube.com/youtubei/v1/yta_web/csv_export?alt=json
|
||
请求体 `exportQuery` 内含 joinRequest 各节点(表格数据/图表数据/总计),以及
|
||
日期范围 `dateIdRange.inclusiveStart` / `dateIdRange.exclusiveEnd`(都是 YYYYMMDD)。
|
||
|
||
2. 后端**不在服务器上生成一个可下载的 URL**,而是直接把打好的 zip 以
|
||
**base64 字符串**内联在响应里:
|
||
{ "responseContext": {...}, "zippedData": "<base64>" }
|
||
实测 `zippedData` 以 `UEsDBBQ...` 开头(即 `PK\\x03\\x04`,ZIP 魔数),
|
||
base64 解码后得到 zip,内含 `表格数据.csv`、`图表数据.csv`、`总计.csv`。
|
||
|
||
3. 前端把 `zippedData` base64 解码 → Blob → 触发浏览器下载。
|
||
由于没有出现指向下载文件的 GET/跳转,判定为「客户端 Blob 下载」而非服务端重定向。
|
||
|
||
4. 浏览器自身已自动保存到本机默认下载目录(本机为 `D:\\Downloads`),无需"另存为"确认;
|
||
且 Chromium 对重名文件会自动追加 ` (1)`、` (2)` …后缀。
|
||
|
||
## 结论:可以跨越下载流程
|
||
- 在响应层拦截 `csv_export`,拿到 `zippedData`,自行 base64 解码并写盘,
|
||
即可完全掌控「保存目录 + 文件名 + 重名去重」,不依赖浏览器的下载管理器和弹窗。
|
||
- 或者只用 Chromium 的下载偏好(auto-download + 内置去重)让它自动落盘。
|
||
|
||
## 文件名约定(与实测 D:\\Downloads 中产物一致)
|
||
<维度标签> <inclusiveStart>_<exclusiveEnd> <账号名>.zip
|
||
例:内容 2026-07-23_2026-08-20 WL Media.zip
|
||
- 维度标签:VIDEO -> 内容;USER -> 频道(生产中建议从页面"维度"按钮文本读取)
|
||
- 日期格式 YYYY-MM-DD:inclusiveStart 与 exclusiveEnd 各取 YYYYMMDD 转 YYYY-MM-DD
|
||
- 账号名:右上角账号按钮文本
|
||
|
||
自测(无需 Playwright): python youtube_export_interceptor.py --selftest
|
||
|
||
实际运行(需 Playwright;必须复用已登录 YouTube Studio 的浏览器会话,否则跳登录页):
|
||
方式 A(用已登录的用户数据目录启动,需先关闭 Chrome/Edge):
|
||
python youtube_export_interceptor.py --url "<explore URL>" --channel chrome ^
|
||
--user-data-dir "%LOCALAPPDATA%\Google\Chrome\User Data"
|
||
方式 B(附加到已在调试端口运行的浏览器,无需关闭):
|
||
python youtube_export_interceptor.py --url "<explore URL>" --connect http://localhost:9222
|
||
# 先启动: chrome.exe --remote-debugging-port=9222 或 msedge.exe --remote-debugging-port=9222
|
||
"""
|
||
|
||
import argparse
|
||
import base64
|
||
import json
|
||
import os
|
||
import re
|
||
import sys
|
||
|
||
# 本机 Windows 下载目录。可改为 os.path.expanduser("~") / "Downloads"。
|
||
DOWNLOAD_DIR = r"D:\Downloads"
|
||
|
||
# 维度类型 -> 文件名前缀标签(生产环境请从页面「维度」按钮文本读取,这里兜底映射)。
|
||
DIMENSION_LABEL = {
|
||
"VIDEO": "内容",
|
||
"USER": "频道",
|
||
"CONTENT_OWNER": "内容",
|
||
}
|
||
|
||
CSV_EXPORT_PATH = "/youtubei/v1/yta_web/csv_export"
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# 重名去重:需求文件.zip -> 需求文件 (1).zip -> 需求文件 (2).zip ... #
|
||
# --------------------------------------------------------------------------- #
|
||
def dedup_path(directory, filename):
|
||
"""返回不冲突的落盘路径。重名时按 `名称 (n).后缀` 递增,n 从 1 开始。
|
||
|
||
规则与用户要求一致:第二次同名保存 `xx (1).zip`,第三次 `xx (2).zip`,以此类推;
|
||
若 `xx (1).zip` 也已存在,则继续找 `xx (2).zip`(即取最小无冲突的 n)。
|
||
"""
|
||
directory = os.path.abspath(directory)
|
||
base, ext = os.path.splitext(filename)
|
||
candidate = os.path.join(directory, filename)
|
||
n = 1
|
||
while os.path.exists(candidate):
|
||
candidate = os.path.join(directory, f"{base} ({n}){ext}")
|
||
n += 1
|
||
return candidate
|
||
|
||
|
||
def build_export_filename(export_query, account_name, dimension_label=None):
|
||
"""从 csv_export 请求体 `exportQuery` 反推导出文件名。
|
||
|
||
与实测产物命名一致:`<维度标签> <inclusiveStart>_<exclusiveEnd> <账号名>.zip`
|
||
"""
|
||
def fmt_dateid(yyyymmdd):
|
||
s = str(yyyymmdd)
|
||
return f"{s[0:4]}-{s[4:6]}-{s[6:8]}"
|
||
|
||
# 从任意一个 joinRequest 节点取 dateIdRange
|
||
date_range = None
|
||
dimension = None
|
||
nodes = (export_query.get("joinRequest", {}).get("nodes") or [])
|
||
for node in nodes:
|
||
q = (node.get("value", {}).get("query") or {})
|
||
if not date_range:
|
||
tr = q.get("timeRange", {}).get("dateIdRange")
|
||
if tr and tr.get("inclusiveStart"):
|
||
date_range = (tr["inclusiveStart"], tr.get("exclusiveEnd"))
|
||
dims = q.get("dimensions") or []
|
||
if dimension is None and dims:
|
||
dimension = dims[0].get("type")
|
||
|
||
if not date_range:
|
||
raise ValueError("无法从 exportQuery 解析日期范围")
|
||
|
||
if dimension_label is None:
|
||
dimension_label = DIMENSION_LABEL.get(dimension or "", "")
|
||
|
||
start, end = date_range
|
||
return f"{dimension_label} {fmt_dateid(start)}_{fmt_dateid(end)} {account_name}.zip"
|
||
|
||
|
||
def decode_zipped_data(payload):
|
||
"""把 csv_export 响应 payload 里的 zippedData 解码为 zip 字节流。"""
|
||
zipped = payload.get("zippedData")
|
||
if not zipped:
|
||
raise ValueError("响应中缺少 zippedData 字段")
|
||
return base64.b64decode(zipped)
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# Playwright 主流程(拦截响应 -> 解码 -> 去重 -> 落盘) #
|
||
# --------------------------------------------------------------------------- #
|
||
def intercept_and_save(page, account_name):
|
||
"""给 page 绑定 response 拦截器:命中 csv_export 就把 zip 保存到下载目录。"""
|
||
import pathlib
|
||
|
||
saved = []
|
||
|
||
def on_response(response):
|
||
if CSV_EXPORT_PATH not in response.url:
|
||
return
|
||
try:
|
||
payload = response.json()
|
||
data = decode_zipped_data(payload)
|
||
|
||
# 反推文件名:请求体在 response.request.post_data 里不总是可读,
|
||
# 这里从已捕获的 body 兜底;找不到就用时间戳命名,保证不误覆盖。
|
||
filename = None
|
||
try:
|
||
body = json.loads(response.request.post_data or "{}")
|
||
filename = build_export_filename(body.get("exportQuery", {}), account_name)
|
||
except Exception:
|
||
filename = f"export-{response.request.headers.get('date', '')}.zip"
|
||
|
||
filename = re.sub(r"[\\/:*?\"<>|]", "_", filename) # Windows 非法字符
|
||
path = dedup_path(DOWNLOAD_DIR, filename)
|
||
pathlib.Path(path).write_bytes(data)
|
||
saved.append((filename, len(data), path))
|
||
except Exception as e: # noqa: BLE001
|
||
print(f"[interceptor] 处理 csv_export 响应失败: {e}", file=sys.stderr)
|
||
|
||
page.on("response", on_response)
|
||
return saved
|
||
|
||
|
||
def _default_user_data_dir(channel):
|
||
"""返回指定浏览器的默认用户数据目录(Windows),用于复用已登录会话。"""
|
||
_local = os.environ.get("LOCALAPPDATA") or os.path.expanduser(r"~\AppData\Local")
|
||
if channel == "msedge":
|
||
return os.path.join(_local, "Microsoft", "Edge", "User Data")
|
||
return os.path.join(_local, "Google", "Chrome", "User Data")
|
||
|
||
|
||
def run(url, user_data_dir=None, channel=None, cdp_url=None):
|
||
"""启动/连接浏览器并触发导出。
|
||
|
||
关键:必须复用「已登录 YouTube Studio」的浏览器会话,否则会跳 Google 登录页。
|
||
- cdp_url: 附加到已在调试端口运行的浏览器(推荐,无需关闭浏览器)
|
||
- user_data_dir:用已登录的用户数据目录启动持久化上下文(需先关闭该浏览器)
|
||
- 两者都不传: 新建空白会话(大概率未登录,仅作占位/调试)
|
||
"""
|
||
from playwright.sync_api import sync_playwright
|
||
|
||
with sync_playwright() as p:
|
||
browser = None
|
||
context = None
|
||
|
||
if cdp_url:
|
||
# 方式 B:附加到已打开、已登录的浏览器(先以调试端口启动浏览器)
|
||
browser = p.chromium.connect_over_cdp(cdp_url)
|
||
context = browser.contexts[0] if browser.contexts else \
|
||
browser.new_context(accept_downloads=True)
|
||
page = context.new_page()
|
||
page.goto(url, wait_until="domcontentloaded")
|
||
elif user_data_dir:
|
||
# 方式 A:用已登录的用户数据目录启动(cookies 复用;必须先关闭同名浏览器)
|
||
context = p.chromium.launch_persistent_context(
|
||
user_data_dir=user_data_dir or _default_user_data_dir(channel),
|
||
channel=channel,
|
||
headless=False,
|
||
accept_downloads=True,
|
||
args=["--disable-blink-features=AutomationControlled"],
|
||
)
|
||
page = context.new_page()
|
||
page.goto(url, wait_until="domcontentloaded")
|
||
else:
|
||
browser = p.chromium.launch(headless=False, channel=channel)
|
||
context = browser.new_context(accept_downloads=True)
|
||
page = context.new_page()
|
||
page.goto(url, wait_until="domcontentloaded")
|
||
|
||
if "studio.youtube.com" not in page.url and "accounts.google" in page.url:
|
||
print("[!] 当前会话未登录,已跳转到 Google 登录页。", file=sys.stderr)
|
||
print(" 请用 --user-data-dir 复用已登录浏览器,或先手动登录后再试。",
|
||
file=sys.stderr)
|
||
|
||
# 账号名:右上角账号按钮(示例选择器,按实际页面微调)
|
||
account_name = "WL Media"
|
||
try:
|
||
account_name = page.locator(
|
||
"ytcp-account-item button, .account-switcher button"
|
||
).first.inner_text(timeout=5000).strip() or account_name
|
||
except Exception:
|
||
pass
|
||
|
||
saved = intercept_and_save(page, account_name)
|
||
|
||
# 触发导出:点「导出当前视图」→「逗号分隔值 (.csv)」
|
||
page.get_by_text("导出当前视图").click()
|
||
page.get_by_text("逗号分隔值 (.csv)").click()
|
||
|
||
page.wait_for_timeout(3000)
|
||
if not saved:
|
||
print("未捕获到 csv_export 响应,请确认已点击导出且登录态有效。")
|
||
else:
|
||
for name, size, path in saved:
|
||
print(f"已保存: {path} ({size} bytes)")
|
||
|
||
if browser is not None:
|
||
browser.close()
|
||
elif context is not None:
|
||
context.close()
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# 自测 #
|
||
# --------------------------------------------------------------------------- #
|
||
def selftest():
|
||
import tempfile
|
||
from pathlib import Path
|
||
|
||
with tempfile.TemporaryDirectory() as td:
|
||
# 1) 用户示例:需求文件.zip 第二次 -> 需求文件 (1).zip,第三次 -> (2).zip
|
||
p0 = Path(dedup_path(td, "需求文件.zip"))
|
||
assert p0.name == "需求文件.zip", p0.name
|
||
p0.write_bytes(b"a")
|
||
|
||
p1 = Path(dedup_path(td, "需求文件.zip"))
|
||
assert p1.name == "需求文件 (1).zip", p1.name
|
||
p1.write_bytes(b"b")
|
||
|
||
p2 = Path(dedup_path(td, "需求文件.zip"))
|
||
assert p2.name == "需求文件 (2).zip", p2.name
|
||
p2.write_bytes(b"c")
|
||
|
||
# 2) 若 (1) 已存在,也应跳到 (2)
|
||
assert Path(dedup_path(td, "需求文件.zip")).name == "需求文件 (3).zip"
|
||
|
||
# 3) 无冲突时不加后缀
|
||
p3 = Path(dedup_path(td, "其他.zip"))
|
||
assert p3.name == "其他.zip", p3.name
|
||
|
||
# 4) 文件名反推(对应实测产物「内容 2026-07-23_2026-08-20 WL Media.zip」)
|
||
export_query = {
|
||
"joinRequest": {"nodes": [{"value": {"query": {
|
||
"dimensions": [{"type": "VIDEO"}],
|
||
"timeRange": {"dateIdRange": {
|
||
"inclusiveStart": 20260723, "exclusiveEnd": 20260820}},
|
||
}}}]},
|
||
}
|
||
name = build_export_filename(export_query, "WL Media")
|
||
assert name == "内容 2026-07-23_2026-08-20 WL Media.zip", name
|
||
|
||
print("selftest OK:去重与文件名反推逻辑全部通过")
|
||
|
||
|
||
if __name__ == "__main__":
|
||
ap = argparse.ArgumentParser()
|
||
ap.add_argument("--selftest", action="store_true", help="仅跑去重/命名自测")
|
||
ap.add_argument("--url", help="explore URL")
|
||
ap.add_argument("--user-data-dir", help="浏览器用户数据目录(复用登录态,需先关闭该浏览器)")
|
||
ap.add_argument("--channel", choices=["chrome", "msedge"],
|
||
help="浏览器品牌:chrome / msedge(复用登录态时必填其一)")
|
||
ap.add_argument("--connect", help="通过 CDP 附加到已打开的浏览器,如 http://localhost:9222")
|
||
args = ap.parse_args()
|
||
|
||
if args.selftest:
|
||
selftest()
|
||
elif args.url:
|
||
run(args.url, user_data_dir=args.user_data_dir,
|
||
channel=args.channel, cdp_url=args.connect)
|
||
else:
|
||
selftest()
|
||
print("\n实际运行请先: pip install playwright\n"
|
||
"复用登录态(必选其一,详见脚本顶部 docstring):\n"
|
||
" 方式 A: python youtube_export_interceptor.py --url \"<explore URL>\" "
|
||
"--channel chrome --user-data-dir \"%LOCALAPPDATA%\\Google\\Chrome\\User Data\"\n"
|
||
" 方式 B: python youtube_export_interceptor.py --url \"<explore URL>\" "
|
||
"--connect http://localhost:9222") |