fix: 修复采购尺码候选扫描面板误判 (#272)

client/src/pdd_collect_service.py

@@ -1186,12 +1186,18 @@ class PddCollectService:
         )
 
     def collect_second_dimension_candidates(
-        self, device: Any
+        self, device: Any, *, initial_xml: Optional[str] = None
     ) -> Optional[SpecDimension]:
-        """复用采集流程完整遍历当前颜色的第二规格，不点击任何选项。"""
+        """完整遍历当前颜色的第二规格，不点击任何选项。
+
+        ``initial_xml`` 是调用方刚确认的最新规格面板首帧，用于避免滚动
+        动画期间第一次重新读取控件树不完整而误判面板已经消失。
+        """
 
         return self._collect_size_dimension(
-            device, require_confirmed_edges=True
+            device,
+            require_confirmed_edges=True,
+            initial_xml=initial_xml,
         )
 
     def _check_cancelled(self) -> None:
@@ -1500,17 +1506,44 @@ class PddCollectService:
         deadline = self._monotonic() + min(2.0, self._spec_panel_timeout)
         xml_data = initial_xml
         transient_reads = 0
+        last_page_kind = "unknown"
+        last_parse_error = ""
         # 次数上限避免测试时钟或设备时钟异常导致无限循环。
         for attempt in range(20):
             self._check_cancelled()
             if xml_data is None:
                 xml_data = self._dump_hierarchy(device)
             try:
+                root = _parse_xml(xml_data)
+                observation = classify_pdd_page(root, "")
+                last_page_kind = observation.kind
+                if observation.kind == PAGE_CAPTCHA:
+                    raise PddCollectError(
+                        "PDD_PAGE_CAPTCHA", "PDD 出现安全验证，需要人工处理"
+                    )
+                if observation.kind == PAGE_LOGIN_REQUIRED:
+                    raise PddCollectError(
+                        "PDD_PAGE_LOGIN_REQUIRED", "PDD 登录已失效，需要人工重新登录"
+                    )
+                if observation.kind == PAGE_RISK_CONTROL:
+                    raise PddCollectError(
+                        "PDD_PAGE_RISK_CONTROL", "PDD 出现风控页面，已停止采购"
+                    )
+                if observation.kind == PAGE_PAYMENT:
+                    raise PddCollectError(
+                        "PDD_PAGE_PAYMENT", "PDD 出现支付页面，已停止采购"
+                    )
                 snapshot = parse_spec_panel(xml_data)
                 panel_open = _is_spec_panel_open(xml_data)
             except PddCollectError as exc:
-                if exc.code != "PDD_DATA_SPEC_INCOMPLETE":
+                if exc.code in {
+                    "PDD_PAGE_CAPTCHA",
+                    "PDD_PAGE_LOGIN_REQUIRED",
+                    "PDD_PAGE_RISK_CONTROL",
+                    "PDD_PAGE_PAYMENT",
+                }:
                     raise
+                last_parse_error = exc.code
                 panel_open = False
                 snapshot = None
 
@@ -1525,10 +1558,21 @@ class PddCollectService:
             self._sleep(0.1)
             xml_data = None
 
+        diagnostics = {
+            "transient_spec_reads": transient_reads,
+            "last_page_kind": last_page_kind,
+            "last_parse_error": last_parse_error,
+        }
+        if last_page_kind == PAGE_ORDER_CONFIRMATION:
+            raise PddCollectError(
+                "PDD_DATA_SPEC_INCOMPLETE",
+                "规格确认页仍在，但规格区域持续无法解析",
+                diagnostics,
+            )
         raise PddCollectError(
             "PDD_PAGE_SPEC_PANEL_LOST",
-            "规格面板操作期间控件树持续为空或面板已经消失",
+            "规格面板已经消失或已离开确认页",
-            {"transient_spec_reads": transient_reads},
+            diagnostics,
         )
 
     def _save_xml(self, label: str, xml_data: str) -> Optional[Mapping[str, Any]]:
@@ -1965,12 +2009,13 @@ class PddCollectService:
         device: Any,
         *,
         require_confirmed_edges: bool = False,
+        initial_xml: Optional[str] = None,
     ) -> Optional[SpecDimension]:
         """完整遍历第二规格并收集文字，全程不点击尺码或套餐。"""
 
         sizes: dict[str, bool] = {}
         size_name = "尺码"
-        xml_data: Optional[str] = None
+        xml_data: Optional[str] = initial_xml
         size_found = False
         previous_signature: Optional[
             tuple[tuple[str, tuple[str, ...]], ...]
@@ -1981,7 +2026,9 @@ class PddCollectService:
         # 保留原有行为，避免改变普通衣服“颜色 + 尺码”的稳定流程。
         for swipe_count in range(self._max_spec_swipes + 1):
             self._check_cancelled()
-            xml_data, snapshot = self._read_valid_spec_panel(device)
+            xml_data, snapshot = self._read_valid_spec_panel(
+                device, initial_xml=xml_data
+            )
             for dimension in snapshot.dimensions:
                 if dimension.key not in ("color", "size"):
                     raise PddCollectError(
@@ -2022,6 +2069,7 @@ class PddCollectService:
                 device, region, horizontal=False, reverse=False
             )
             self._sleep(0.35)
+            xml_data = None
 
         if not size_found or xml_data is None:
             return None

client/src/pdd_u2_purchase_adapter.py

@@ -1004,9 +1004,7 @@ class U2PddPurchaseAdapter(PddPurchaseAdapter):
         self._panel_timeout = panel_timeout
         self._select_color = select_color_fn
         self._select_size = select_size_fn
-        self._size_candidate_collector = (
+        self._size_candidate_collector = size_candidate_collector
-            size_candidate_collector or self._collect_size_candidates
-        )
         self._now = now
         self._artifact_directory = artifact_directory
         self._last_xml: Optional[str] = None
@@ -1232,8 +1230,10 @@ class U2PddPurchaseAdapter(PddPurchaseAdapter):
         """完整只读遍历后，仅为页面确实不存在的目标返回候选快照。"""
 
         try:
-            self._restore_purchase_panel_color_region(failed_xml)
+            restored_xml = self._restore_purchase_panel_color_region(failed_xml)
-            dimension = self._size_candidate_collector(self._require_device())
+            dimension = self._read_size_candidates(
+                self._require_device(), restored_xml
+            )
         except PddPurchaseError:
             raise
         except PddCollectError as exc:
@@ -1337,7 +1337,18 @@ class U2PddPurchaseAdapter(PddPurchaseAdapter):
                 },
             )
 
-    def _collect_size_candidates(self, device: Any) -> Optional[SpecDimension]:
+    def _read_size_candidates(
+        self, device: Any, initial_xml: Optional[str] = None
+    ) -> Optional[SpecDimension]:
+        """读取第二规格候选；注入测试函数保持原有单参数接口。"""
+
+        if self._size_candidate_collector is not None:
+            return self._size_candidate_collector(device)
+        return self._collect_size_candidates(device, initial_xml=initial_xml)
+
+    def _collect_size_candidates(
+        self, device: Any, *, initial_xml: Optional[str] = None
+    ) -> Optional[SpecDimension]:
         """使用采集模块同一套第二规格遍历，不建立新连接也不保存 XML。"""
 
         service = PddCollectService(
@@ -1349,7 +1360,9 @@ class U2PddPurchaseAdapter(PddPurchaseAdapter):
             cancelled=self._cancelled,
             artifact_directory=None,
         )
-        return service.collect_second_dimension_candidates(device)
+        return service.collect_second_dimension_candidates(
+            device, initial_xml=initial_xml
+        )
 
     def apply_resolved_size(
         self,
@@ -1386,7 +1399,7 @@ class U2PddPurchaseAdapter(PddPurchaseAdapter):
                 },
             )
         try:
-            dimension = self._size_candidate_collector(self._require_device())
+            dimension = self._read_size_candidates(self._require_device())
         except PddCollectError as exc:
             raise PddPurchaseError(
                 "PURCHASE_SPEC_CANDIDATES_CHANGED",

client/test/test_pdd_collect_service.py

@@ -2184,6 +2184,25 @@ class PddCollectParserTest(unittest.TestCase):
         self.assertEqual("PDD_PAGE_CAPTCHA", raised.exception.code)
         self.assertEqual([], clock.sleeps)
 
+    def test_valid_initial_spec_frame_is_used_before_reading_device_again(self):
+        device = FakeCollectDevice(self.home_xml, self.spec_xml)
+        device.panel_open = True
+        device.opened_url = "https://mobile.yangkeduo.com/goods.html?goods_id=123"
+        service = PddCollectService(
+            PddDeviceService(lambda _serial: device),
+            "USB-001",
+            "client-001",
+            sleeper=lambda _seconds: None,
+        )
+
+        xml_data, snapshot = service._read_valid_spec_panel(
+            device, initial_xml=self.spec_xml
+        )
+
+        self.assertEqual(self.spec_xml, xml_data)
+        self.assertTrue(snapshot.dimensions)
+        self.assertEqual(0, service._goods_screens_checked)
+
 
 if __name__ == "__main__":
     unittest.main()