diff --git a/client/src/collect_task_service.py b/client/src/collect_task_service.py index 6987de9..51dc890 100644 --- a/client/src/collect_task_service.py +++ b/client/src/collect_task_service.py @@ -206,6 +206,7 @@ class CollectTaskService: return { "PDD_DATA_SKU_NAME_TRUNCATED": "SKU_NAME_TRUNCATED", "PDD_PAGE_SPEC_ENTRY_MISSING": "SKU_PANEL_NOT_FOUND", + "PDD_PAGE_SPEC_PANEL_TIMEOUT": "SKU_PANEL_NOT_FOUND", "PDD_DATA_TITLE_MISSING": "TITLE_TOO_SHORT", "PDD_PAGE_OVERALL_TIMEOUT": "COLLECT_TIMEOUT", "DEVICE_OFFLINE": "DEVICE_DISCONNECTED", diff --git a/client/src/pdd_collect_service.py b/client/src/pdd_collect_service.py index 9c77711..4a16d0d 100644 --- a/client/src/pdd_collect_service.py +++ b/client/src/pdd_collect_service.py @@ -241,22 +241,6 @@ def _all_labels(root: ET.Element) -> list[str]: return [label for node in root.iter("node") if (label := _node_label(node))] -def _screen_bounds(root: ET.Element) -> Optional[Bounds]: - bounds_list = [ - bounds - for node in root.iter("node") - if (bounds := _parse_bounds(node.get("bounds", ""))) is not None - ] - if not bounds_list: - return None - return ( - min(item[0] for item in bounds_list), - min(item[1] for item in bounds_list), - max(item[2] for item in bounds_list), - max(item[3] for item in bounds_list), - ) - - def parse_quantity(raw: Optional[str]) -> QuantityMetric: """把“已拼1.2万+”等文字转成整数，同时保留原文。""" @@ -600,21 +584,6 @@ def _validate_goods_url(goods_url: str, task_goods_id: str) -> str: return value -def _combine_goods_snapshots(snapshots: Sequence[GoodsSnapshot]) -> GoodsSnapshot: - return GoodsSnapshot( - next((item.title for item in snapshots if item.title), None), - next((item.shop_name for item in snapshots if item.shop_name), None), - next( - (item.sales for item in snapshots if item.sales.raw), - QuantityMetric(None, None, False), - ), - next( - (item.reviews for item in snapshots if item.reviews.raw), - QuantityMetric(None, None, False), - ), - ) - - class PddCollectService: """打开商品页并采集结构化商品与 SKU 数据。""" @@ -629,8 +598,8 @@ class PddCollectService: now: Callable[[], datetime] = lambda: datetime.now(timezone.utc), cancelled: Callable[[], bool] = lambda: False, page_timeout: float = 30.0, + spec_panel_timeout: float = 10.0, overall_timeout: float = 600.0, - max_page_swipes: int = 12, max_spec_swipes: int = 12, max_sku_count: int = 200, artifact_directory: Optional[Path] = None, @@ -643,9 +612,9 @@ class PddCollectService: self._now = now self._cancelled = cancelled self._page_timeout = page_timeout + self._spec_panel_timeout = spec_panel_timeout self._overall_timeout = overall_timeout self._overall_deadline: Optional[float] = None - self._max_page_swipes = max_page_swipes self._max_spec_swipes = max_spec_swipes self._max_sku_count = max_sku_count self._artifact_directory = artifact_directory @@ -672,10 +641,13 @@ class PddCollectService: raise PddCollectError("PDD_DATA_TITLE_MISSING", "商品页没有可识别的标题") if not goods.sales.raw: raise PddCollectError("PDD_DATA_SALES_MISSING", "商品页没有采集到已拼数量") - if not goods.reviews.raw: - raise PddCollectError("PDD_DATA_REVIEWS_MISSING", "商品页没有采集到评价数量") - if not goods.shop_name and self._last_goods_xml: - artifact = self._save_xml("shop-not-found", self._last_goods_xml) + if ( + (not goods.shop_name or not goods.reviews.raw) + and self._last_goods_xml + ): + artifact = self._save_xml( + "goods-metadata-incomplete", self._last_goods_xml + ) if artifact: self._artifacts.append(artifact) @@ -687,7 +659,7 @@ class PddCollectService: "商品页没有找到可靠的规格入口", ) device.click(*coordinate) - self._sleep(0.5) + self._wait_spec_panel(device) snapshots = self._discover_dimensions(device) dimensions = merge_dimensions(snapshots) @@ -791,46 +763,36 @@ class PddCollectService: raise PddCollectError("PDD_PAGE_TIMEOUT", "等待 PDD 商品详情页加载超时") def _collect_goods_details(self, device: Any) -> GoodsSnapshot: - snapshots: list[GoodsSnapshot] = [] - previous_signature: Optional[tuple[str, ...]] = None - unchanged = 0 - for _ in range(self._max_page_swipes + 1): + """只读商品首页当前视口，不在打开规格面板前滚动详情页。""" + + self._check_cancelled() + xml_data = device.dump_hierarchy() + self._last_goods_xml = str(xml_data) + self._goods_screens_checked += 1 + return parse_goods_page(xml_data) + + def _wait_spec_panel(self, device: Any) -> SpecSnapshot: + """等待点击后的规格面板真正出现，不能只依赖固定延时。""" + + deadline = self._monotonic() + self._spec_panel_timeout + while self._monotonic() < deadline: self._check_cancelled() xml_data = device.dump_hierarchy() self._last_goods_xml = str(xml_data) self._goods_screens_checked += 1 - root = _parse_xml(xml_data) - labels = tuple(_all_labels(root)) - snapshots.append(parse_goods_page(xml_data)) - combined = _combine_goods_snapshots(snapshots) - is_complete = ( - combined.title - and combined.sales.raw - and combined.reviews.raw - ) - if is_complete: - break - if labels == previous_signature: - unchanged += 1 + try: + snapshot = parse_spec_panel(xml_data) + except PddCollectError as exc: + if exc.code != "PDD_DATA_SPEC_INCOMPLETE": + raise else: - unchanged = 0 - if unchanged >= 1: - break - previous_signature = labels - screen = _screen_bounds(root) - if screen is None: - break - left, top, right, bottom = screen - x = (left + right) // 2 - device.swipe( - x, - top + int((bottom - top) * 0.75), - x, - top + int((bottom - top) * 0.30), - duration=0.35, - ) - self._sleep(0.35) - return _combine_goods_snapshots(snapshots) + if snapshot.dimensions: + return snapshot + self._sleep(0.25) + raise PddCollectError( + "PDD_PAGE_SPEC_PANEL_TIMEOUT", + "点击规格入口后，等待规格面板加载超时", + ) def _save_xml(self, label: str, xml_data: str) -> Optional[Mapping[str, Any]]: """保存本地诊断 XML；测试未提供目录时不写文件。""" diff --git a/client/test/test_pdd_collect_service.py b/client/test/test_pdd_collect_service.py index b27ef91..bc2878c 100644 --- a/client/test/test_pdd_collect_service.py +++ b/client/test/test_pdd_collect_service.py @@ -33,6 +33,7 @@ class FakeCollectDevice: self.panel_open = False self.clicks = [] self.swipes = [] + self.swipe_panel_states = [] def app_current(self): return {"package": "com.xunmeng.pinduoduo"} @@ -56,6 +57,7 @@ class FakeCollectDevice: def swipe(self, *args, **kwargs): self.swipes.append((args, kwargs)) + self.swipe_panel_states.append(self.panel_open) class LoadingDevice(FakeCollectDevice): @@ -85,6 +87,11 @@ class FocusMismatchDevice(FakeCollectDevice): self.app_started = True +class PanelDoesNotOpenDevice(FakeCollectDevice): + def click(self, x, y): + self.clicks.append((x, y)) + + def keep_only_one_sku(xml_data: str) -> str: """从脱敏固件中删除蓝色和 L，只保留一个组合。""" @@ -176,7 +183,6 @@ class PddCollectParserTest(unittest.TestCase): "client-001", sleeper=lambda _seconds: None, now=lambda: datetime(2026, 8, 7, 8, 0, tzinfo=timezone.utc), - max_page_swipes=1, max_spec_swipes=1, ) @@ -213,7 +219,6 @@ class PddCollectParserTest(unittest.TestCase): "USB-001", "client-001", sleeper=lambda _seconds: None, - max_page_swipes=0, max_spec_swipes=0, ) @@ -231,7 +236,6 @@ class PddCollectParserTest(unittest.TestCase): "USB-001", "client-001", sleeper=lambda _seconds: None, - max_page_swipes=0, max_spec_swipes=0, ) @@ -375,7 +379,6 @@ class PddCollectParserTest(unittest.TestCase): "USB-001", "client-001", sleeper=lambda _seconds: None, - max_page_swipes=0, ) result = service.collect( @@ -383,6 +386,62 @@ class PddCollectParserTest(unittest.TestCase): ) self.assertIsNone(result.shop_name) + def test_missing_reviews_does_not_scroll_before_opening_spec_panel(self): + home_without_reviews = self.home_xml.replace( + '', + "", + ) + device = FakeCollectDevice( + home_without_reviews, keep_only_one_sku(self.spec_xml) + ) + service = PddCollectService( + PddDeviceService(lambda _serial: device), + "USB-001", + "client-001", + sleeper=lambda _seconds: None, + max_spec_swipes=0, + ) + + result = service.collect( + FakeTask("https://mobile.yangkeduo.com/goods.html?goods_id=123") + ) + + self.assertIsNone(result.reviews.raw) + self.assertTrue(device.clicks) + self.assertTrue(all(device.swipe_panel_states)) + + def test_spec_panel_timeout_has_stable_code_and_xml_evidence(self): + with tempfile.TemporaryDirectory() as directory: + device = PanelDoesNotOpenDevice(self.home_xml, self.spec_xml) + tick = {"value": -0.5} + + def clock(): + tick["value"] += 0.5 + return tick["value"] + + service = PddCollectService( + PddDeviceService(lambda _serial: device), + "USB-001", + "client-001", + sleeper=lambda _seconds: None, + monotonic=clock, + spec_panel_timeout=1.0, + artifact_directory=Path(directory), + ) + + with self.assertRaises(PddCollectError) as raised: + service.collect( + FakeTask("https://mobile.yangkeduo.com/goods.html?goods_id=123") + ) + + self.assertEqual( + raised.exception.code, "PDD_PAGE_SPEC_PANEL_TIMEOUT" + ) + artifacts = raised.exception.diagnostics["artifacts"] + self.assertTrue(Path(artifacts[0]["path"]).is_file()) + if __name__ == "__main__": unittest.main() diff --git a/client/test/test_pdd_real_xml_fixtures.py b/client/test/test_pdd_real_xml_fixtures.py index 70cf9ad..93c1848 100644 --- a/client/test/test_pdd_real_xml_fixtures.py +++ b/client/test/test_pdd_real_xml_fixtures.py @@ -9,6 +9,7 @@ from src.pdd_collect_service import ( parse_spec_panel, ) from src.pdd_device_service import PddDeviceService +from src.util.get_size_panle_coord import get_size_panel_coord REAL_XML = Path(__file__).parents[1] / "image_xml" @@ -74,6 +75,14 @@ class PddRealXmlFixtureTest(unittest.TestCase): self.assertEqual(result.price_cent, 470) self.assertEqual(result.list_price_cent, 1990) + def test_real_home_xml_has_immediate_spec_entry(self): + xml_data = (REAL_XML / "737116531267_home.xml").read_text( + encoding="utf-8" + ) + coordinate = get_size_panel_coord(xml_data) + self.assertIsNotNone(coordinate) + self.assertGreater(coordinate[1], 2000) + if __name__ == "__main__": unittest.main() diff --git a/docs/client/01-requirements.md b/docs/client/01-requirements.md index 676e3d6..5c23d0d 100644 --- a/docs/client/01-requirements.md +++ b/docs/client/01-requirements.md @@ -46,9 +46,9 @@ Client 顶级导航仅包含： 输入至少包含远程任务编号和商品链接。Client 应采集： - 商品编号、商品链接和商品标题； -- 店铺名称； +- 店铺名称（无障碍树未提供时允许留空并保存诊断证据）； - 已拼数量及原始显示文字； -- 评价数量及原始显示文字； +- 评价数量及原始显示文字（当前首页未提供时允许留空，不能为此反复滚动并阻塞规格采集）； - 所有规格维度及其可见值； - 每个规格组合对应的价格、币种和可用状态； - 采集时间、设备及必要诊断产物引用。 diff --git a/docs/client/02-architecture.md b/docs/client/02-architecture.md index 0bc9ec7..609c1e7 100644 --- a/docs/client/02-architecture.md +++ b/docs/client/02-architecture.md @@ -354,6 +354,10 @@ reconcile_purchase(task, run) -> PurchaseResult | ManualReview PDD 页面可能出现登录失效、验证码、控件树不完整、A/B 页面、库存变化和价格变化。适配层必须返回结构化错误，不得把这些情况统一返回 `False`。 +采集商品时按“首页就绪 → 读取当前首页摘要 → 点击规格入口 → 确认规格面板 +出现 → 扫描颜色和尺码”的顺序执行。打开规格面板前不得为了寻找评价或店铺 +连续滚动商品详情；这两个可选字段缺失时保存证据，但不阻塞核心 SKU 采集。 + ## 10. 关键架构决策 1. 使用 PyQt5、Qt Widgets 和 PyQt-Fluent-Widgets，不混用其他 Qt 绑定。