feat(gateway): 账号只读查询通道——从已登录账号取真实订单

上游要的不只是网关记的任务状态镜像,还有「已登录账号在站点上的真实订单」,
但账号只在 NAT 后本地机上,只能经网关队列走。新增独立查询通道(§11):

- gateway 单开 account_queries 表 + QueryStatus 状态机,接口
  POST /api/account/queries(幂等)/ lease / {id}/result / {id}
- 不复用下单任务队列:查询是只读,租约过期可安全重投(与下单「绝不自动
  重投」相反),且不该被全局并发度 1 堵死、task_reports 是订单镜像不能污染
- 本地交易服务起第二条常驻循环 query_runner,领到即调 SiteInteractor 真读:
  order_list 复用已实测的 list_recent_orders(规范化字段 + 站点
  orderListData 原文),order_detail 复用 fetch_order_detail(配送阶段 +
  页面 __INITIAL_STATE__ 原样透传,结构未经真实样本,不抽字段)
- 账号级串行仍由 SiteInteractor 的锁保证;每次执行套超时按失败回报
- 错误码 6005/6006(查询通道,可重试只读区别于 6001-6004);/health 暴露
  queued_query_count;结果体积上限先丢原始 JSON

openapi.json 重导,docs/order-gateway.md §11、README、.env.example 补全
配置与实测边界。全量测试 404→454 通过。

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-08-16 22:46:01 +08:00
co-authored by Claude Opus 5
parent b01c9659d1
commit c03158488b
22 changed files with 2575 additions and 43 deletions
+390
View File
@@ -0,0 +1,390 @@
"""账号只读查询通道测试(docs/order-gateway.md §11)
分两层:
- HTTP 层:鉴权、幂等、6005/6006、lease → result 全链路、列表与 /health
- 队列语义层:直接驱动 QueryQueue,测那些靠 HTTP 不好造的时序——租约超时重投、
重投次数用尽、整体 TTL 过期、迟到结果被拒、过保留期清理
这里最关键的一条是「只读可重投」:它与下单任务的「绝不自动重投」正好相反,
两条通道的语义分歧全在这几个用例里,改任何一边前先看它们。
"""
from __future__ import annotations
from datetime import timedelta
from pathlib import Path
import pytest
from fastapi.testclient import TestClient
from app.gateway.db import GatewayDB, QueryRow
from app.gateway.query_queue import QueryQueue, to_iso, utcnow
from app.shared.config import get_settings
from app.shared.errors import QueryLeaseInvalidError
from app.shared.task_state import AccountQueryKind, QueryStatus
TOKEN = get_settings().bearer_token
AUTH = {"Authorization": f"Bearer {TOKEN}"}
QUERIES = "/api/account/queries"
@pytest.fixture
def gateway_client(tmp_path: Path, monkeypatch):
"""起一个独立 DB 的网关应用(与 test_gateway_api.py 同一套装配方式)"""
monkeypatch.setenv("RAKUTEN_GATEWAY_DB_PATH", str(tmp_path / "gw.db"))
get_settings.cache_clear()
try:
from app.gateway.main import create_app
app = create_app()
with TestClient(app) as client:
yield client
finally:
get_settings.cache_clear()
@pytest.fixture
async def queue(tmp_path: Path):
"""直接可用的 (QueryQueue, GatewayDB) 对
队列语义用例要模拟「租约已过期」「结果早就写完了」这类时间流逝,直接改库里的
时间戳比 sleep 快得多,所以把 db 一并交出去。
"""
db = GatewayDB(tmp_path / "q.db")
await db.start()
try:
yield (
QueryQueue(
db,
lease_ttl_seconds=60,
query_ttl_seconds=600,
max_attempts=3,
retention_seconds=7 * 24 * 3600,
),
db,
)
finally:
await db.close()
def _submit(client: TestClient, **overrides) -> dict:
payload = {
"query_id": "q1",
"site": "rakuten",
"kind": AccountQueryKind.ORDER_LIST.value,
"params": {"max_pages": 1},
}
payload.update(overrides)
return client.post(QUERIES, json=payload, headers=AUTH).json()
# ---- HTTP:鉴权与参数校验 ----
@pytest.mark.parametrize(
"path,method",
[
(QUERIES, "POST"),
(f"{QUERIES}/lease", "GET"),
(f"{QUERIES}/q1", "GET"),
],
)
def test_query_endpoints_reject_missing_token(gateway_client, path, method):
response = gateway_client.request(method, path)
assert response.status_code == 401
assert response.json()["code"] == 1001
def test_unknown_kind_is_rejected_at_submit(gateway_client):
"""kind 拼错当场 422,而不是等 worker 领走、执行、回报失败才知道"""
response = gateway_client.post(
QUERIES,
json={"site": "rakuten", "kind": "order_lst", "params": {}},
headers=AUTH,
)
assert response.status_code == 422
assert response.json()["code"] == 1002
# ---- HTTP:提交、幂等、领取、回结果 ----
def test_submit_returns_query_id_and_queued(gateway_client):
body = _submit(gateway_client)
assert body["success"] is True
assert body["data"]["query_id"] == "q1"
assert body["data"]["status"] == QueryStatus.QUEUED.value
assert body["data"]["created"] is True
def test_submit_is_idempotent_on_same_query_id(gateway_client):
assert _submit(gateway_client)["data"]["created"] is True
assert _submit(gateway_client)["data"]["created"] is False
def test_submit_generates_query_id_when_absent(gateway_client):
body = gateway_client.post(
QUERIES,
json={"site": "rakuten", "kind": AccountQueryKind.ORDER_DETAIL.value, "params": {}},
headers=AUTH,
).json()
assert body["data"]["query_id"].startswith("q-")
def test_lease_returns_null_when_no_query(gateway_client):
response = gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH)
assert response.status_code == 200
assert response.json()["data"] is None
def test_lease_then_result_completes_the_query(gateway_client):
_submit(gateway_client)
leased = gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH).json()
assert leased["data"]["query_id"] == "q1"
assert leased["data"]["kind"] == AccountQueryKind.ORDER_LIST.value
assert leased["data"]["params"] == {"max_pages": 1}
assert leased["data"]["attempt"] == 1
result = {"kind": "order_list", "orders": [{"order_number": "306087-20260813-0863947697"}]}
reported = gateway_client.post(
f"{QUERIES}/q1/result",
json={"worker_id": "w1", "success": True, "result": result},
headers=AUTH,
).json()
assert reported["data"]["status"] == QueryStatus.SUCCEEDED.value
detail = gateway_client.get(f"{QUERIES}/q1", headers=AUTH).json()["data"]
assert detail["status"] == QueryStatus.SUCCEEDED.value
assert detail["result"] == result
assert detail["error"] is None
assert detail["lease_owner"] is None
def test_failed_result_carries_error_code_and_message(gateway_client):
_submit(gateway_client)
gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH)
gateway_client.post(
f"{QUERIES}/q1/result",
json={
"worker_id": "w1",
"success": False,
"error_code": 5001,
"error_message": "账号未登录",
},
headers=AUTH,
)
detail = gateway_client.get(f"{QUERIES}/q1", headers=AUTH).json()["data"]
assert detail["status"] == QueryStatus.FAILED.value
assert detail["error"] == {"code": 5001, "message": "账号未登录"}
def test_two_queries_can_be_in_flight_at_once(gateway_client):
"""与下单任务不同:只读查询没有「全局并发度 1」的闸门"""
_submit(gateway_client, query_id="q1")
_submit(gateway_client, query_id="q2")
first = gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH).json()
second = gateway_client.get(f"{QUERIES}/lease?worker_id=w2&wait=0", headers=AUTH).json()
assert first["data"]["query_id"] == "q1"
assert second["data"]["query_id"] == "q2"
def test_order_lease_is_not_disturbed_by_pending_query(gateway_client):
"""两条通道互不干扰:有查询在飞,下单任务照领"""
_submit(gateway_client)
gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH)
gateway_client.post(
"/api/orders",
json={"task_id": "t1", "site": "rakuten", "intent": {}},
headers=AUTH,
)
leased = gateway_client.get("/api/orders/lease?worker_id=w1&wait=0", headers=AUTH).json()
assert leased["data"]["task_id"] == "t1"
# ---- HTTP:错误码 ----
def test_get_unknown_query_returns_6005(gateway_client):
response = gateway_client.get(f"{QUERIES}/no-such", headers=AUTH)
assert response.status_code == 404
assert response.json()["code"] == 6005
def test_result_from_non_owner_returns_6006(gateway_client):
_submit(gateway_client)
gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH)
response = gateway_client.post(
f"{QUERIES}/q1/result",
json={"worker_id": "w2", "success": True, "result": {}},
headers=AUTH,
)
assert response.status_code == 409
assert response.json()["code"] == 6006
def test_result_on_terminal_query_returns_6006(gateway_client):
_submit(gateway_client)
gateway_client.get(f"{QUERIES}/lease?worker_id=w1&wait=0", headers=AUTH)
gateway_client.post(
f"{QUERIES}/q1/result",
json={"worker_id": "w1", "success": True, "result": {"a": 1}},
headers=AUTH,
)
response = gateway_client.post(
f"{QUERIES}/q1/result",
json={"worker_id": "w1", "success": True, "result": {"a": 2}},
headers=AUTH,
)
assert response.status_code == 409
assert response.json()["code"] == 6006
# 结果没有被第二次回报覆盖
detail = gateway_client.get(f"{QUERIES}/q1", headers=AUTH).json()["data"]
assert detail["result"] == {"a": 1}
# ---- HTTP:列表与健康检查 ----
def test_list_queries_filters_by_kind(gateway_client):
_submit(gateway_client, query_id="q1", kind=AccountQueryKind.ORDER_LIST.value)
_submit(gateway_client, query_id="q2", kind=AccountQueryKind.ORDER_DETAIL.value)
body = gateway_client.get(
f"{QUERIES}?kind={AccountQueryKind.ORDER_DETAIL.value}", headers=AUTH
).json()
assert body["data"]["total"] == 1
assert body["data"]["items"][0]["query_id"] == "q2"
def test_health_reports_queued_query_count(gateway_client):
_submit(gateway_client)
body = gateway_client.get("/health").json()
assert body["data"]["queued_query_count"] == 1
# ---- 队列语义:只读可重投(与下单任务的核心分歧)----
async def test_expired_lease_goes_back_to_queued(queue):
"""租约过期→回 queued 自动重投。下单任务在这里是 stale 且绝不重投"""
q, db = queue
await q.submit(query_id="q1", site="rakuten", kind="order_list", params={})
lease = await q.lease(worker_id="w1", wait=0, site=None, max_wait=60)
assert lease is not None
await db.update_query("q1", lease_expires_at=to_iso(utcnow() - timedelta(seconds=1)))
assert await q.sweep() == 1
detail = await q.get_detail("q1")
assert detail.status == QueryStatus.QUEUED
assert detail.lease_owner is None
again = await q.lease(worker_id="w2", wait=0, site=None, max_wait=60)
assert again is not None
assert again.attempt == 2
async def test_late_result_after_reinvest_is_rejected(queue):
"""超时重投后,上一轮 worker 迟到的结果必须被拒,不能覆盖新一轮"""
q, db = queue
await q.submit(query_id="q1", site="rakuten", kind="order_list", params={})
await q.lease(worker_id="w1", wait=0, site=None, max_wait=60)
await db.update_query("q1", lease_expires_at=to_iso(utcnow() - timedelta(seconds=1)))
await q.sweep()
await q.lease(worker_id="w2", wait=0, site=None, max_wait=60)
with pytest.raises(QueryLeaseInvalidError):
await q.submit_result(
"q1", worker_id="w1", success=True, result={"stale": True},
error_code=None, error_message="",
)
async def test_reinvest_stops_at_max_attempts(queue):
"""重投次数用尽 → failed,不再无限重投同一张卡死的单"""
q, db = queue
await q.submit(query_id="q1", site="rakuten", kind="order_list", params={})
for _ in range(3):
await q.lease(worker_id="w1", wait=0, site=None, max_wait=60)
await db.update_query("q1", lease_expires_at=to_iso(utcnow() - timedelta(seconds=1)))
await q.sweep()
detail = await q.get_detail("q1")
assert detail.status == QueryStatus.FAILED
assert detail.attempts == 3
assert "不再重投" in detail.error.message
async def test_query_expires_after_ttl(tmp_path):
"""整体 TTL 到点 → expired(本地 worker 不在线的典型表现)"""
db = GatewayDB(tmp_path / "q.db")
await db.start()
try:
queue = QueryQueue(
db, lease_ttl_seconds=60, query_ttl_seconds=60, max_attempts=3,
retention_seconds=3600,
)
# 直接插一张「10 分钟前创建」的单,比 sleep 现实
old = to_iso(utcnow() - timedelta(minutes=10))
await db.insert_query(
QueryRow(
query_id="q1", site="rakuten", kind="order_list", params_json="{}",
status=QueryStatus.QUEUED.value, lease_owner=None, lease_expires_at=None,
attempts=0, result_json=None, error_code=None, error_message=None,
created_at=old, updated_at=old, completed_at=None,
)
)
assert await queue.sweep() == 1
detail = await queue.get_detail("q1")
assert detail.status == QueryStatus.EXPIRED
assert "worker 不在线" in detail.error.message
# 过期的单不会再被领走
assert await queue.lease(worker_id="w1", wait=0, site=None, max_wait=60) is None
finally:
await db.close()
async def test_sweep_purges_queries_past_retention(tmp_path):
"""终态查询单过保留期被清掉——结果里带站点原始 JSON,不清会一直涨"""
db = GatewayDB(tmp_path / "q.db")
await db.start()
try:
queue = QueryQueue(
db, lease_ttl_seconds=60, query_ttl_seconds=600, max_attempts=3,
retention_seconds=1,
)
await queue.submit(query_id="q1", site="rakuten", kind="order_list", params={})
await queue.lease(worker_id="w1", wait=0, site=None, max_wait=60)
await queue.submit_result(
"q1", worker_id="w1", success=True, result={"a": 1},
error_code=None, error_message="",
)
await db.update_query("q1", completed_at=to_iso(utcnow() - timedelta(seconds=60)))
await queue.sweep()
assert await db.get_query("q1") is None
finally:
await db.close()
async def test_sweep_keeps_unfinished_queries(tmp_path):
"""保留期清理只动终态单,进行中的一律不碰"""
db = GatewayDB(tmp_path / "q.db")
await db.start()
try:
queue = QueryQueue(
db, lease_ttl_seconds=60, query_ttl_seconds=600, max_attempts=3,
retention_seconds=1,
)
await queue.submit(query_id="q1", site="rakuten", kind="order_list", params={})
await queue.sweep()
assert await db.get_query("q1") is not None
finally:
await db.close()
+417
View File
@@ -0,0 +1,417 @@
"""账号只读查询 worker 测试(app/trading/worker/query_runner.py)
站点交互与网关都用桩替换,测的是 QueryRunner 自己的编排:
- 两种 kind 各自把站点结果整理成什么样的 result(原始 JSON 有没有原样带出来)
- 失败怎么转成一次「失败回报」——超时、站点 AppError、参数不合法、未知 kind
- 结果体积闸门:先丢站点原始 JSON,仍超限就判失败让上游缩小窗口
一条贯穿始终的约定:**worker 永远不把异常抛回主循环**,每张查询单都要有一次
回报(成功或失败),否则网关侧那张单只能干等到租约超时。
"""
from __future__ import annotations
import asyncio
from datetime import datetime, timezone
from typing import Any
import pytest
from app.shared.config import Settings
from app.shared.errors import AppError, NotLoggedInError
from app.shared.task_state import AccountQueryKind, OrderState
from app.trading.worker.models import QueryTask
from app.trading.worker.query_runner import QueryRunner
from app.trading.worker.site_interact import (
OrderDetailSnapshot,
OrderListEntry,
OrderListItem,
OrderListWindow,
OrderStatusSnapshot,
)
class _FakeGateway:
"""记录回报内容的网关客户端替身
`lease_query` 里那句 `sleep(0)` 是必需的:真实实现是 HTTP 长轮询,一定会把
控制权交回事件循环;纯内存替身不 await 任何东西的话,run() 会一直霸占循环,
同一个 event loop 里的停机任务永远排不上,测试直接挂死。
"""
def __init__(self, queries: list[QueryTask | None] | None = None):
self._queries = list(queries or [])
self.reports: list[dict[str, Any]] = []
self.lease_calls = 0
async def lease_query(self, worker_id: str, *, wait: int = 30) -> QueryTask | None:
self.lease_calls += 1
await asyncio.sleep(0)
if not self._queries:
return None
return self._queries.pop(0)
async def report_query_result(
self,
query_id: str,
worker_id: str,
*,
success: bool,
result: dict[str, Any] | None = None,
error_code: int | None = None,
error_message: str = "",
) -> dict[str, Any]:
self.reports.append(
{
"query_id": query_id,
"worker_id": worker_id,
"success": success,
"result": result,
"error_code": error_code,
"error_message": error_message,
}
)
return {"query_id": query_id, "status": "succeeded" if success else "failed"}
class _FakeSite:
"""SiteInteractor 替身:按脚本返回窗口/详情,或抛出指定异常"""
def __init__(self, *, window=None, detail=None, error: Exception | None = None, delay: float = 0):
self._window = window
self._detail = detail
self._error = error
self._delay = delay
self.list_calls: list[dict[str, Any]] = []
self.detail_calls: list[str] = []
async def list_recent_orders(self, *, since, max_pages=None):
self.list_calls.append({"since": since, "max_pages": max_pages})
await asyncio.sleep(self._delay)
if self._error:
raise self._error
return self._window
async def fetch_order_detail(self, site_order_id: str):
self.detail_calls.append(site_order_id)
await asyncio.sleep(self._delay)
if self._error:
raise self._error
return self._detail
def _runner(site: _FakeSite, gateway: _FakeGateway, **settings_kwargs) -> QueryRunner:
settings = Settings(worker_id="w1", **settings_kwargs)
return QueryRunner(settings=settings, gateway_client=gateway, site=site) # type: ignore[arg-type]
def _query(kind: str, params: dict[str, Any] | None = None, **kwargs) -> QueryTask:
return QueryTask(
query_id=kwargs.get("query_id", "q1"),
site=kwargs.get("site", "rakuten"),
kind=kind,
params=params or {},
attempt=kwargs.get("attempt", 1),
)
def _window(*, covered: bool = True) -> OrderListWindow:
return OrderListWindow(
entries=[
OrderListEntry(
order_number="306087-20260813-0863947697",
order_date="2026-08-13T09:47:28.000Z",
shop_id=306087,
shop_name="BACKYARD FAMILY インテリアタウン",
items=[
OrderListItem(
item_url="https://item.rakuten.co.jp/backyard/x/",
item_name="壁掛けフック",
item_id=10012345,
)
],
)
],
window_fully_covered=covered,
raw_pages=[{"ordersFound": 1, "orderList": [{"orderNumber": "306087-20260813-0863947697"}]}],
)
# ---- order_list ----
async def test_order_list_returns_normalized_and_raw():
site = _FakeSite(window=_window())
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value, {"max_pages": 2}))
report = gateway.reports[0]
assert report["success"] is True
result = report["result"]
assert result["kind"] == "order_list"
assert result["window_fully_covered"] is True
assert result["orders"][0]["order_number"] == "306087-20260813-0863947697"
assert result["orders"][0]["items"][0]["item_id"] == 10012345
# 站点原文原样带出,规范化模型没覆盖的字段上游还能自己取
assert result["raw_pages"][0]["ordersFound"] == 1
assert site.list_calls[0]["max_pages"] == 2
async def test_order_list_since_is_passed_through():
site = _FakeSite(window=_window())
runner = _runner(site, _FakeGateway())
await runner.handle(
_query(AccountQueryKind.ORDER_LIST.value, {"since": "2026-08-01T00:00:00Z"})
)
assert site.list_calls[0]["since"] == datetime(2026, 8, 1, tzinfo=timezone.utc)
async def test_order_list_without_since_uses_epoch_and_default_max_pages():
site = _FakeSite(window=_window())
runner = _runner(site, _FakeGateway(), account_query_default_max_pages=5)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
assert site.list_calls[0]["since"] == datetime(1970, 1, 1, tzinfo=timezone.utc)
assert site.list_calls[0]["max_pages"] == 5
async def test_order_list_reports_partial_coverage_honestly():
"""翻页没覆盖完窗口时必须如实标出来——上游据此知道「没查到」不等于「没有」"""
site = _FakeSite(window=_window(covered=False))
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
assert gateway.reports[0]["result"]["window_fully_covered"] is False
async def test_invalid_since_is_reported_as_failure():
site = _FakeSite(window=_window())
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value, {"since": "上周"}))
assert gateway.reports[0]["success"] is False
assert gateway.reports[0]["error_code"] == 1003
assert site.list_calls == [] # 参数不合法就没去碰站点
async def test_invalid_max_pages_is_reported_as_failure():
gateway = _FakeGateway()
runner = _runner(_FakeSite(window=_window()), gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value, {"max_pages": 0}))
assert gateway.reports[0]["success"] is False
assert gateway.reports[0]["error_code"] == 1003
# ---- order_detail ----
async def test_order_detail_returns_stage_and_raw():
detail = OrderDetailSnapshot(
status=OrderStatusSnapshot(
found=True, stage_label="出荷", order_state=OrderState.SHIPPED, html="<html/>"
),
raw={"pageType": "ph-detail"},
)
site = _FakeSite(detail=detail)
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(
_query(AccountQueryKind.ORDER_DETAIL.value, {"order_number": "306087-20260813-0863947697"})
)
result = gateway.reports[0]["result"]
assert result["found"] is True
assert result["stage_label"] == "出荷"
assert result["order_state"] == OrderState.SHIPPED.value
assert result["raw"] == {"pageType": "ph-detail"}
assert result["raw_available"] is True
assert site.detail_calls == ["306087-20260813-0863947697"]
async def test_order_detail_not_found_is_success_not_failure():
"""站点自己说订单要 10 分钟才反映出来:查不到是正常结果,不是查询失败"""
site = _FakeSite(detail=OrderDetailSnapshot(status=OrderStatusSnapshot(found=False), raw=None))
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(_query(AccountQueryKind.ORDER_DETAIL.value, {"order_number": "x-1-2"}))
assert gateway.reports[0]["success"] is True
assert gateway.reports[0]["result"]["found"] is False
assert gateway.reports[0]["result"]["raw_available"] is False
async def test_order_detail_requires_order_number():
gateway = _FakeGateway()
runner = _runner(_FakeSite(), gateway)
await runner.handle(_query(AccountQueryKind.ORDER_DETAIL.value, {}))
assert gateway.reports[0]["success"] is False
assert gateway.reports[0]["error_code"] == 1003
# ---- 失败路径 ----
async def test_unknown_kind_is_reported_as_failure():
gateway = _FakeGateway()
runner = _runner(_FakeSite(), gateway)
await runner.handle(_query("order_everything"))
assert gateway.reports[0]["success"] is False
assert "未知的查询种类" in gateway.reports[0]["error_message"]
async def test_non_rakuten_site_is_rejected():
"""交易服务只覆盖乐天市场,ラクマ 永不进入账号链路"""
gateway = _FakeGateway()
runner = _runner(_FakeSite(), gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value, site="rakuma"))
assert gateway.reports[0]["success"] is False
assert "rakuma" in gateway.reports[0]["error_message"]
async def test_site_app_error_keeps_its_error_code():
"""站点侧错误码原样回给上游(掉登录=5001),不被翻译成一个笼统的失败"""
site = _FakeSite(error=NotLoggedInError(site="rakuten"))
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
assert gateway.reports[0]["success"] is False
assert gateway.reports[0]["error_code"] == 5001
async def test_unexpected_exception_is_reported_not_raised():
site = _FakeSite(error=RuntimeError("boom"))
gateway = _FakeGateway()
runner = _runner(site, gateway)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
assert gateway.reports[0]["success"] is False
assert gateway.reports[0]["error_code"] == 1500
assert "RuntimeError" in gateway.reports[0]["error_message"]
async def test_timeout_is_reported_as_retryable_failure():
"""账号锁被下单占着 → 超时按失败回报(上游重发即可),不挂到租约过期"""
site = _FakeSite(window=_window(), delay=0.2)
gateway = _FakeGateway()
runner = _runner(site, gateway, account_query_timeout_seconds=0)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
assert gateway.reports[0]["success"] is False
assert gateway.reports[0]["error_code"] == 2002
assert "可稍后重发" in gateway.reports[0]["error_message"]
async def test_report_failure_does_not_escape():
"""回报本身失败(比如租约已被重投)只记日志,不能把循环带崩"""
class _RejectingGateway(_FakeGateway):
async def report_query_result(self, *args: Any, **kwargs: Any) -> dict[str, Any]:
raise AppError(message="lease invalid", code="X", err_code=6006, status_code=409)
runner = _runner(_FakeSite(window=_window()), _RejectingGateway())
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value)) # 不抛异常即通过
# ---- 结果体积闸门 ----
async def test_oversized_result_drops_raw_first():
big_window = _window()
big_window.raw_pages = [{"padding": "x" * 5000}]
gateway = _FakeGateway()
runner = _runner(_FakeSite(window=big_window), gateway, query_result_max_bytes=2000)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
report = gateway.reports[0]
assert report["success"] is True
assert report["result"]["raw_pages"] == []
assert "raw_omitted" in report["result"]
# 规范化字段还在——上游至少拿得到订单号
assert report["result"]["orders"][0]["order_number"] == "306087-20260813-0863947697"
async def test_result_still_oversized_after_dropping_raw_fails():
"""丢掉原始 JSON 仍超限:判失败并告诉上游缩小窗口,不硬塞进网关库"""
window = _window()
window.entries = [
OrderListEntry(order_number=f"o{i}", order_date="2026-08-13T09:47:28.000Z")
for i in range(200)
]
gateway = _FakeGateway()
runner = _runner(_FakeSite(window=window), gateway, query_result_max_bytes=500)
await runner.handle(_query(AccountQueryKind.ORDER_LIST.value))
assert gateway.reports[0]["success"] is False
assert "缩小查询窗口" in gateway.reports[0]["error_message"]
# ---- 主循环 ----
async def test_run_loop_handles_then_stops():
"""领到单就处理;stop() 后循环退出,不吃掉后续任务"""
gateway = _FakeGateway([_query(AccountQueryKind.ORDER_LIST.value), None])
runner = _runner(_FakeSite(window=_window()), gateway)
async def stop_soon():
# 让 run() 先跑几轮(领单 → 处理 → 再领一次拿到 None),再停
for _ in range(20):
await asyncio.sleep(0)
runner.stop()
await asyncio.wait_for(asyncio.gather(runner.run(), stop_soon()), timeout=5)
assert [r["query_id"] for r in gateway.reports] == ["q1"]
async def test_run_loop_survives_lease_error(monkeypatch):
"""lease 报错不能让循环退出(网关重启、网络抖动都算正常)"""
real_sleep = asyncio.sleep
class _FlakyGateway(_FakeGateway):
"""前两次 lease 都报错,第二次顺便把循环停掉,避免测试无限转"""
def __init__(self, stopper):
super().__init__()
self._stopper = stopper
async def lease_query(self, worker_id: str, *, wait: int = 30):
self.lease_calls += 1
if self.lease_calls >= 2:
self._stopper()
raise AppError(message="gateway down", code="X", err_code=3001)
holder: dict[str, QueryRunner] = {}
gateway = _FlakyGateway(lambda: holder["runner"].stop())
runner = _runner(_FakeSite(), gateway)
holder["runner"] = runner
# 出错后那句 sleep(5) 在测试里没必要真等;捕获原函数再替换,否则递归自调
monkeypatch.setattr(asyncio, "sleep", lambda *_: real_sleep(0))
await runner.run()
assert gateway.lease_calls == 2
+103
View File
@@ -754,6 +754,109 @@ async def test_list_recent_orders_without_start_raises():
await site.list_recent_orders(since=_SINCE)
# ---- 账号只读查询通道用到的两条只读路径(docs/order-gateway.md §11)----
#
# 查询接口不新写解析,复用的就是上面这两条已实测路径;这里补的是「站点原始
# JSON 有没有被完整带出来」和「翻页上限有没有生效」——这两点是查询接口独有的,
# 恢复核对那条老路径不关心。
def test_parse_order_list_keeps_raw_order_list_data():
"""规范化字段之外,站点 orderListData 原文要原样留着供上游取用"""
html = _wrap_state(_real_order_list_state())
page = _parse_order_list(html)
assert page.raw is not None
assert page.raw["ordersFound"] == 1
# 规范化模型里没有的字段也在(这正是「原样透传」的意义)
assert page.raw["orderList"][0]["shopName"] == "BACKYARD FAMILY インテリアタウン"
def test_parse_order_list_raw_is_none_when_page_type_wrong():
"""不是 ph-list(改版 / 掉登录 / act 分支不对):raw 为 None,不给上游半截数据"""
html = _wrap_state(_real_order_list_state(page_type="ph-detail"))
assert _parse_order_list(html).raw is None
def test_accumulate_collects_raw_pages_in_order():
"""多页翻页时,每页的原始 JSON 按顺序累积"""
acc = _OrderListAccumulator()
page1 = OrderListPage(
entries=[_entry("o1", "2026-08-10T00:00:00Z")], orders_found=3, raw={"page": 1}
)
acc, stop = _accumulate_order_list_page(acc, page1, since=_SINCE)
assert stop is False
page2 = OrderListPage(
entries=[_entry("o2", "2026-07-01T00:00:00Z")], orders_found=3, raw={"page": 2}
)
acc, stop = _accumulate_order_list_page(acc, page2, since=_SINCE)
assert stop is True
assert acc.raw_pages == [{"page": 1}, {"page": 2}]
async def test_list_recent_orders_respects_max_pages(tmp_path):
"""max_pages 是硬上限:翻到上限就停,且必须如实报 window_fully_covered=False"""
# 每页都还有更多订单(orders_found 远大于已取回数),正常会一直翻下去
state = _real_order_list_state(orders_found=99)
page = _FakeOrderPage([(_ORDER_LIST_LANDED_URL, _wrap_state(state))])
site = _build_site(tmp_path, _FakeContext([page]), _FakeAuthSession())
window = await site.list_recent_orders(since=_SINCE, max_pages=2)
assert len(page.goto_urls) == 2
assert page.goto_urls[1].endswith("?page=2")
assert window.window_fully_covered is False
assert len(window.raw_pages) == 2
async def test_list_recent_orders_window_carries_raw_pages(tmp_path):
html = _wrap_state(_real_order_list_state())
page = _FakeOrderPage([(_ORDER_LIST_LANDED_URL, html)])
site = _build_site(tmp_path, _FakeContext([page]), _FakeAuthSession())
window = await site.list_recent_orders(since=_SINCE)
assert window.window_fully_covered is True
assert [raw["ordersFound"] for raw in window.raw_pages] == [1]
async def test_fetch_order_detail_returns_status_and_raw_state(tmp_path):
"""详情页:已实测的配送阶段照常解析,页面原始状态原样带出"""
html = _wrap_state({"pageType": "ph-detail", "whatever": {"a": 1}}) + _stepper_html(
active_stage="出荷"
)
page = _FakeOrderPage([(_ORDER_DETAIL_LANDED_URL, html)])
site = _build_site(tmp_path, _FakeContext([page]), _FakeAuthSession())
detail = await site.fetch_order_detail(_REAL_ORDER_ID)
assert detail.status.found is True
assert detail.status.order_state == OrderState.SHIPPED
assert detail.raw == {"pageType": "ph-detail", "whatever": {"a": 1}}
async def test_fetch_order_detail_raw_is_none_without_inline_state(tmp_path):
"""页面没有内联状态时 raw=None——不编造,也不因此把这次读取判成失败"""
page = _FakeOrderPage([(_ORDER_DETAIL_LANDED_URL, _stepper_html(active_stage="出荷"))])
site = _build_site(tmp_path, _FakeContext([page]), _FakeAuthSession())
detail = await site.fetch_order_detail(_REAL_ORDER_ID)
assert detail.raw is None
assert detail.status.found is True
async def test_check_order_status_still_returns_only_status(tmp_path):
"""付款后监控那条老路径不受影响:仍然拿到 OrderStatusSnapshot"""
page = _FakeOrderPage([(_ORDER_DETAIL_LANDED_URL, _stepper_html(active_stage="配達完了"))])
site = _build_site(tmp_path, _FakeContext([page]), _FakeAuthSession())
snapshot = await site.check_order_status(_REAL_ORDER_ID)
assert isinstance(snapshot, OrderStatusSnapshot)
assert snapshot.order_state == OrderState.DELIVERED
# ---- submit_order / pay:没有 enter_checkout 留存的确认页会话时应报错,不静默成功 ----