Fix CHSI catalogue session verification and pagination

This commit is contained in:
Codex
2026-10-11 15:28:14 +08:00
parent 8001cd86b1
commit 2e3439dbb9
11 changed files with 496 additions and 85 deletions
+2 -1
View File
@@ -28,7 +28,7 @@ def create_app(data_dir=None):
router = ProxyRouter(settings)
transport = Transport(settings, router)
auth = AuthService(transport, CookieVault(directory))
jobs = JobManager(db, settings, transport)
jobs = JobManager(db, settings, transport, auth=auth)
if (ROOT / "data/schools.json").exists():
db.index_schools(json.loads((ROOT / "data/schools.json").read_text(encoding="utf-8")))
@@ -41,6 +41,7 @@ def create_app(data_dir=None):
app.state.db, app.state.jobs = db, jobs
app.state.settings, app.state.router = settings, router
app.state.pool = pool
app.state.auth = auth
@app.middleware("http")
async def local_access(request: Request, call_next):
+98 -39
View File
@@ -6,8 +6,12 @@ import threading
import time
from pathlib import Path
from bs4 import BeautifulSoup
from studio.connectors import CatalogClient, LoginRequired
from studio.domain import now
from studio.network import ProxyUnavailable, RemoteUnavailable, RateLimited
LOGIN_URL = "https://account.chsi.com.cn/passport/login?entrytype=yzgr&service=https%3A%2F%2Fyz.chsi.com.cn%2Fj_spring_cas_security_check"
CATALOG_ENTRY = "https://yz.chsi.com.cn/zsml/a/dw.do"
class Blob(ctypes.Structure):
@@ -55,9 +59,16 @@ class AuthService:
self.transport, self.vault = transport, vault
self.message = "尚未验证登录状态"
self.valid = False
self.state = "unverified"
self.checked_at = None
self.verified_route = None
self.revision = 0
self.state_lock = threading.RLock()
self.check_lock = threading.RLock()
self.browser_state = "idle"
self.browser_message = ""
self.browser_thread = None
self.browser_cancel = None
try:
self._load(self.vault.load())
except (RuntimeError, OSError, ValueError):
@@ -80,7 +91,19 @@ class AuthService:
return [{"name": c.name, "value": c.value, "domain": c.domain, "path": c.path, "secure": c.secure, "expires": c.expires} for c in self.transport.session.cookies]
def status(self):
return {"saved": self.vault.path.exists(), "valid": self.valid, "message": self.message, "browser_state": self.browser_state, "browser_message": self.browser_message}
with self.state_lock:
if self.valid and self.verified_route != self.transport.router.status()["route"]:
self.invalidate("网络出口已变化,请验证目录访问后继续采集", state="unverified")
return {"saved": self.vault.path.exists(), "valid": self.valid, "state": self.state, "message": self.message, "checked_at": self.checked_at, "browser_state": self.browser_state, "browser_message": self.browser_message}
def invalidate(self, message, state="login_required"):
with self.state_lock:
self.revision += 1
self.valid = False
self.state, self.message = state, message
self.verified_route = None
if self.browser_state == "completed":
self.browser_state, self.browser_message = "idle", ""
def cookie(self, text):
if not text.strip() or "\n" in text or "\r" in text:
@@ -94,50 +117,76 @@ class AuthService:
cookies[name] = value
if not cookies:
raise ValueError("未识别到 Cookie 键值")
self._load(cookies)
self.vault.save(cookies)
self.valid = False
self.message = "Cookie 已加密保存,点击验证确认目录访问权限"
return self.status()
if self.browser_cancel:
self.browser_cancel.set()
with self.check_lock:
self._load(cookies)
self.vault.save(cookies)
self.invalidate("Cookie 已加密保存,点击验证确认目录访问权限", state="unverified")
self.checked_at = None
self.browser_state, self.browser_message = "idle", ""
return self.status()
def check(self):
# The school/major indexes can be public while direction details require login.
response = self.transport.request("POST", "https://yz.chsi.com.cn/zsml/rs/yjfxs.do", data={"dwdm": "10001", "zydm": "081200", "zymc": "计算机科学与技术", "curPage": 1, "start": 0, "pageSize": 1, "dwlxs[0]": "all"})
try:
body = response.json()
except ValueError:
self.valid = False
self.message = "目录接口未返回有效数据,可能需要登录或验证码"
with self.check_lock:
with self.state_lock:
self.invalidate("正在同步研招目录会话并验证访问权限", state="unverified")
revision = self.revision
try:
# The first ten rows are available anonymously, even for directions.
# Enter the CAS-protected page to bind cookies, then prove page-two access.
self.transport.request("GET", CATALOG_ENTRY)
CatalogClient(self.transport).fetch_page("yjfxs", {"dwdm": "10001", "zydm": "081200", "zymc": "计算机科学与技术", "curPage": 2, "start": 10, "pageSize": 10, "dwlxs[0]": "all"})
proof_route = self.transport.local.response_route
self.vault.save(self._snapshot())
except LoginRequired as exc:
self.invalidate(str(exc), state="captcha_required" if exc.reason == "captcha" else "login_required")
except (ProxyUnavailable, RemoteUnavailable, RateLimited):
self.invalidate("暂时无法验证研招目录,请检查网络或稍后重试;已保存的 Cookie 保留", state="unverified")
raise
else:
with self.state_lock:
# A later crawler login/captcha failure supersedes this older proof.
if self.revision == revision:
self.valid, self.state = True, "verified"
self.message = "研招目录登录会话已同步,第二页访问验证成功"
self.verified_route = proof_route
finally:
self.checked_at = now()
return self.status()
self.valid = body.get("flag") is True and isinstance(body.get("msg"), dict)
self.message = "研招目录接口验证成功" if self.valid else "登录未生效或数据源限流,请在浏览器完成登录后重试"
if self.valid:
self.vault.save(self._snapshot())
return self.status()
def login(self, username, password):
if not username or not password:
raise ValueError("请输入学信网账号和密码")
response = self.transport.request("GET", LOGIN_URL)
soup = BeautifulSoup(response.text, "html.parser")
form = {node.get("name"): node.get("value", "") for node in soup.select("input[type=hidden][name]")}
if "lt" not in form or "execution" not in form:
raise ValueError("登录页面需要人工验证,请使用浏览器登录")
form.update(username=username, password=password, _eventId="submit")
self.transport.request("POST", LOGIN_URL, data=form)
return self.check()
if self.browser_cancel:
self.browser_cancel.set()
with self.check_lock:
response = self.transport.request("GET", LOGIN_URL)
soup = BeautifulSoup(response.text, "html.parser")
form = {node.get("name"): node.get("value", "") for node in soup.select("input[type=hidden][name]")}
if "lt" not in form or "execution" not in form:
raise ValueError("登录页面需要人工验证,请使用浏览器登录")
form.update(username=username, password=password, _eventId="submit")
self.transport.request("POST", LOGIN_URL, data=form)
return self.check()
def clear(self):
self._load({})
self.vault.clear()
self.valid = False
self.message = "已清除本项目保存的登录凭据"
return self.status()
if self.browser_cancel:
self.browser_cancel.set()
with self.check_lock:
self._load({})
self.vault.clear()
self.invalidate("已清除本项目保存的登录凭据", state="unverified")
self.checked_at = None
self.browser_state, self.browser_message = "idle", ""
return self.status()
def browser(self):
if self.browser_thread and self.browser_thread.is_alive():
if self.browser_thread and self.browser_thread.is_alive() and not self.browser_cancel.is_set():
return self.status()
proxy = self.transport.router.resolve()
cancelled = threading.Event()
self.browser_cancel = cancelled
self.browser_state = "opening"
self.browser_message = "正在打开独立 Edge 登录窗口,完成登录后可关闭窗口"
@@ -149,24 +198,34 @@ class AuthService:
context = browser.new_context()
page = context.new_page()
page.goto(LOGIN_URL, wait_until="domcontentloaded", timeout=60000)
if cancelled.is_set():
browser.close()
return
self.browser_state = "waiting"
self.browser_message = "请在 Edge 完成学信网登录和验证码;窗口保持打开,工作台自动检测"
deadline = time.monotonic() + 300
while time.monotonic() < deadline:
while time.monotonic() < deadline and not cancelled.is_set():
page.wait_for_timeout(2000)
if "yz.chsi.com.cn" in page.url:
cookies = [c for c in context.cookies() if c["domain"].lstrip(".") == "chsi.com.cn" or c["domain"].endswith(".chsi.com.cn")]
if cookies:
self._load(cookies)
if self.check()["valid"]:
self.browser_state = "completed"
self.browser_message = "已获取登录 Cookie,目录接口验证成功"
browser.close()
return
with self.check_lock:
if cancelled.is_set():
break
self._load(cookies)
if self.check()["valid"] and not cancelled.is_set():
self.browser_state = "completed"
self.browser_message = "已获取登录 Cookie,目录接口验证成功"
browser.close()
return
browser.close()
if cancelled.is_set():
return
self.browser_state = "expired"
self.browser_message = "登录等待超时,可再次打开浏览器"
except Exception:
if cancelled.is_set():
return
self.browser_state = "failed"
self.browser_message = "登录窗口已关闭或连接失败,可重试或粘贴 Cookie"
self.browser_thread = threading.Thread(target=run, daemon=True, name="browser-login")
+54 -25
View File
@@ -6,7 +6,9 @@ from studio.network import RemoteUnavailable, RateLimited
class LoginRequired(RuntimeError):
pass
def __init__(self, message, reason="login"):
super().__init__(message)
self.reason = reason
class SourceYearMismatch(RuntimeError):
@@ -16,8 +18,8 @@ class SourceYearMismatch(RuntimeError):
class CatalogClient:
BASE = "https://yz.chsi.com.cn/zsml/"
def __init__(self, transport):
self.transport = transport
def __init__(self, transport, auth=None):
self.transport, self.auth = transport, auth
def verify_year(self, year):
response = self.transport.request("GET", self.BASE)
@@ -32,32 +34,53 @@ class CatalogClient:
raise SourceYearMismatch(f"研招网当前公开目录为 {current} 年,任务选择了 {year} 年;历史目录请导入已有快照")
return current
def pages(self, endpoint, payload, page_size=50):
seen = set()
for page in range(1, 10001):
for attempt in range(3):
response = self.transport.request("POST", self.BASE + "rs/" + endpoint + ".do", data={**payload, "curPage": page, "start": (page - 1) * page_size, "pageSize": page_size})
try:
body = response.json()
except ValueError:
raise LoginRequired("研招网返回登录/验证页面,请先完成登录") from None
message = str(body.get('msg', ''))
if body.get('flag') is not True and '验证码' in message:
raise LoginRequired('研招网需要人工验证码,请在浏览器完成验证后恢复')
if body.get('flag') is not True and any(token in message for token in ('频繁', '稍后')):
if attempt == 2:
raise RateLimited('研招网业务限流持续出现,已暂停;请增加请求间隔后恢复')
self.transport.sleep(5 * 2 ** attempt)
continue
break
def _fetch_page(self, endpoint, payload):
for attempt in range(3):
response = self.transport.request("POST", self.BASE + "rs/" + endpoint + ".do", data=payload)
try:
body = response.json()
except ValueError:
if '验证码' in response.text:
raise LoginRequired('研招网需要人工验证码,请在浏览器完成验证后恢复', reason="captcha") from None
if re.search(r"passport/login|j_spring_cas_security_check|name=[\"']password[\"']", response.text) or '/passport/login' in response.url:
raise LoginRequired("研招网返回登录页面,请在浏览器完成登录后恢复") from None
raise RemoteUnavailable("研招目录接口未返回有效 JSON,请稍后重试") from None
if not isinstance(body, dict):
raise RemoteUnavailable("研招网目录响应结构已变化,未写入不完整数据")
message = str(body.get('msg', ''))
if body.get('flag') is not True and '验证码' in message:
raise LoginRequired('研招网需要人工验证码,请在浏览器完成验证后恢复', reason="captcha")
if body.get('flag') is not True and any(token in message for token in ('频繁', '稍后')):
if attempt == 2:
raise RateLimited('研招网业务限流持续出现,已暂停;请增加请求间隔后恢复')
self.transport.sleep(5 * 2 ** attempt)
continue
if body.get("flag") is not True:
message = str(body.get("msg", ""))
if any(token in message for token in ("登录", "登陆", "认证")):
raise LoginRequired("研招网要求登录,请更新 Cookie 后恢复任务")
raise LoginRequired("研招网要求登录,请重新验证或在浏览器登录后恢复任务")
raise RemoteUnavailable("研招网目录接口未成功返回数据")
data = body.get("msg") or {}
data = body.get("msg")
if not isinstance(data, dict) or not isinstance(data.get("list"), list):
raise RemoteUnavailable("研招网目录响应结构已变化,未写入不完整数据")
return data
def fetch_page(self, endpoint, payload):
try:
return self._fetch_page(endpoint, payload)
except LoginRequired as exc:
if not self.auth or exc.reason == "captcha":
raise
# A CAS cookie may still be valid while the catalogue session has expired.
# Refresh once, then retry exactly the same page; never skip its records.
status = self.auth.check()
if not status["valid"]:
raise LoginRequired(status["message"], reason="captcha" if status["state"] == "captcha_required" else "login") from None
return self._fetch_page(endpoint, payload)
def pages(self, endpoint, payload, page_size=10):
seen, start = set(), 0
for page in range(1, 10001):
data = self.fetch_page(endpoint, {**payload, "curPage": page, "start": start, "pageSize": page_size})
rows = data["list"]
signature = json.dumps(rows, ensure_ascii=False, sort_keys=True)
if rows and signature in seen:
@@ -66,8 +89,14 @@ class CatalogClient:
yield from rows
more = data.get("nextPageAvailable")
total = integer(data.get("totalCount"))
if not rows or more is False or more in ("false", "0", 0) or more is None and ((total is not None and page * page_size >= total) or len(rows) < page_size):
size = integer(data.get("size")) or page_size
if not rows or more is False or more in ("false", "0", 0) or more is None and ((total is not None and start + len(rows) >= total) or len(rows) < size):
return
next_start = integer(data.get("startOfNextPage"))
next_start = start + len(rows) if next_start is None else next_start
if next_start <= start:
raise RemoteUnavailable("研招目录下一页位置无效,已停止以避免遗漏分页")
start, page_size = next_start, size
raise RemoteUnavailable("目录分页超过安全上限")
def schools(self, province, keyword=""):
+10 -2
View File
@@ -15,8 +15,9 @@ class Shutdown(Exception):
class JobManager:
def __init__(self, database, settings, transport=None, runner=None):
def __init__(self, database, settings, transport=None, runner=None, auth=None):
self.db, self.settings, self.transport = database, settings, transport
self.auth = auth
self.runner = runner or self._collect
self.controls = {}
self.lock = threading.RLock()
@@ -80,6 +81,8 @@ class JobManager:
except Cancelled:
self.db.job_update(jid, status="cancelled", message="已取消,已保存数据保留")
except LoginRequired as exc:
if self.auth:
self.auth.invalidate(str(exc), state="captcha_required" if exc.reason == "captcha" else "login_required")
self.db.job_update(jid, status="waiting_login", message=str(exc))
self.db.log(jid, "warning", str(exc))
except ProxyUnavailable as exc:
@@ -186,8 +189,13 @@ class JobManager:
self._scores(jid, params)
def _catalog(self, jid, params):
client = CatalogClient(self.transport)
client = CatalogClient(self.transport, auth=self.auth)
client.verify_year(params["year"])
if self.auth:
status = self.auth.check()
if not status["valid"]:
raise LoginRequired(status["message"], reason="captcha" if status["state"] == "captcha_required" else "login")
self.db.log(jid, "info", "研招目录会话已同步,第二页访问验证成功")
for province in params["provinces"]:
self.checkpoint(jid)
self.db.log(jid, "info", "查询 " + PROVINCES[province] + " 院校")
+1
View File
@@ -142,6 +142,7 @@ class Transport:
self.sleep(min(30, 5 * (attempt + 1)))
continue
response.raise_for_status()
self.local.response_route = proxy or "direct"
return response
except (requests.exceptions.ProxyError, requests.exceptions.ConnectionError, requests.exceptions.Timeout):
if proxy: