fix: 阻止售罄页误入规格采集 (#159)

client/src/pdd_collect_service.py

@@ -48,6 +48,13 @@ _DIMENSION_NAMES = ("颜色分类", "颜色", "尺码", "尺寸", "规格", "型
 _LOGIN_MARKERS = ("手机号登录", "登录后继续", "验证码登录", "账号登录")
 _CAPTCHA_MARKERS = ("请完成验证", "拖动滑块", "安全验证", "点击图中")
 _READY_MARKERS = ("发起拼单", "立即购买", "单独购买", "免拼购买", "快要抢光")
+_OUT_OF_STOCK_PAGE_MARKERS = (
+    "商品已售罄，推荐以下相似商品",
+    "商品已售罄，看看相似商品",
+    "该商品已售罄",
+    "该商品暂时缺货",
+    "商品暂时缺货",
+)
 _SHOP_NAME_EXCLUDES = frozenset({"进店", "关注", "店铺", "收藏", "客服"})
 _SHOP_ROW_TOLERANCE = 40
 
@@ -509,18 +516,19 @@ def _is_spec_panel_open(xml_data: str | bytes) -> bool:
     parents = {child: parent for parent in root.iter() for child in parent}
     if _find_non_scrollable_spec_panel(root, parents) is not None:
         return True
-    has_scrollable_region = any(
-        node.get("scrollable") == "true"
-        and _parse_bounds(node.get("bounds", "")) is not None
+    scrollable_regions = [
+        node
         for node in root.iter("node")
-    )
+        if node.get("scrollable") == "true"
+        and _parse_bounds(node.get("bounds", "")) is not None
+    ]
+    has_scrollable_region = bool(scrollable_regions)
     if not has_scrollable_region:
         return False
 
     compact_labels = [label.replace(" ", "") for label in labels]
     has_selection_summary = any(
-        label.startswith(("请选择", "已选"))
-        and any(word in label for word in ("颜色", "尺码", "规格", "款式"))
+        label.startswith(("请选择", "請選擇", "已选", "已選"))
         for label in compact_labels
     )
     has_submit_hint = any(
@@ -528,7 +536,15 @@ def _is_spec_panel_open(xml_data: str | bytes) -> bool:
         and any(word in label for word in ("选择", "颜色", "尺码", "规格"))
         for label in compact_labels
     )
-    return has_selection_summary or has_submit_hint
+    has_dimension_heading = any(
+        _is_dimension_heading(_preferred_node_label(node).strip())
+        and any(
+            region in _ancestors(node, parents)
+            for region in scrollable_regions
+        )
+        for node in root.iter("node")
+    )
+    return has_selection_summary or has_submit_hint or has_dimension_heading
 
 
 def _top_level_clickable_options(
@@ -600,6 +616,10 @@ def parse_spec_panel(xml_data: str | bytes) -> SpecSnapshot:
         if node.get("scrollable") == "true" and _parse_bounds(node.get("bounds", ""))
     ]
     non_scrollable_outer = _find_non_scrollable_spec_panel(root, parents)
+    if non_scrollable_outer is None and not _is_spec_panel_open(xml_data):
+        raise PddCollectError(
+            "PDD_DATA_SPEC_INCOMPLETE", "规格面板没有可识别的规格区域"
+        )
     panel_scrollables = []
     if non_scrollable_outer is not None:
         panel_scrollables = [
@@ -646,10 +666,12 @@ def parse_spec_panel(xml_data: str | bytes) -> SpecSnapshot:
         bounds = _parse_bounds(nested.get("bounds", ""))
         if bounds is None:
             continue
-        name = "颜色分类"
         preceding = [item for item in heading_nodes if item[0] <= bounds[1] + 30]
-        if preceding:
-            name = preceding[-1][1]
+        if not preceding:
+            # 普通推荐列表也是 RecyclerView。没有真实规格标题时，不能
+            # 默认把其中的商品卡片解释成颜色。
+            continue
+        name = preceding[-1][1]
         options = _top_level_clickable_options(nested, parents)
         if options:
             groups.append((name, options))
@@ -713,6 +735,17 @@ def _ancestors(
 
 
 def _raise_special_page(labels: Sequence[str]) -> None:
+    compact_labels = [label.replace(" ", "") for label in labels]
+    if any(
+        marker in label
+        for label in compact_labels
+        for marker in _OUT_OF_STOCK_PAGE_MARKERS
+    ):
+        raise PddCollectError(
+            "PDD_GOODS_UNAVAILABLE",
+            "商品已售罄，PDD 没有提供可采集的规格",
+            {"page_reason": "out_of_stock"},
+        )
     combined = " ".join(labels)
     if any(marker in combined for marker in _CAPTCHA_MARKERS):
         raise PddCollectError("PDD_PAGE_CAPTCHA", "PDD 出现安全验证，需要人工处理")
@@ -1134,7 +1167,7 @@ class PddCollectService:
                 if exc.code != "PDD_DATA_SPEC_INCOMPLETE":
                     raise
             else:
-                if snapshot.dimensions or _is_spec_panel_open(xml_data):
+                if _is_spec_panel_open(xml_data):
                     return snapshot
             self._sleep(0.25)
         raise PddCollectError(