From 7f2a746f66d2214035e68c5e34af07bb2e959b5e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B5=85=E5=BF=B5?= <email@sample.com>
Date: Wed, 16 Sep 2026 18:39:02 +0800
Subject: [PATCH] =?UTF-8?q?fix:=20=E6=8E=92=E5=BA=8F=E4=BD=9C=E7=94=A8?=
 =?UTF-8?q?=E4=BA=8E=E7=BB=84=E5=86=85=E4=BA=8B=E4=BB=B6+=E6=97=B6?=
 =?UTF-8?q?=E9=97=B4=E7=BA=BF=E9=99=90=E9=AB=98=E6=BB=9A=E5=8A=A8+UTM?=
 =?UTF-8?q?=E5=88=AB=E5=90=8D=E5=BD=92=E4=B8=80=E6=B6=88=E5=86=B2=E7=AA=81?=
 =?UTF-8?q?=E8=AF=AF=E6=8A=A5?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- attribution.py：新增 _canon 归一（小写/去协议与域名后缀/别名表 fb→facebook、百度→baidu 等/仅留字母数字汉字），
  冲突判定改为两路各自归一后比较——同源升级 combined 并在 detail 记 utm_source 原文，仍不同才 conflict
- detail.vue：orderedDays 组内会话/事件同步翻转（desc=组与组内都最新在前，asc=都最早在前）；
  .tl-body 限高 calc(100vh - 380px) 卡内滚动，日期组头随内容滚动
- test_attribution.py：+4 用例（fb↔facebook、chatgpt.com↔ChatGPT、google↔facebook 仍冲突、百度↔baidu）

Co-Authored-By: Claude Code <noreply@anthropic.com>
---
 admin/src/views/tracker/visitor/detail.vue | 20 ++++++++++----
 server/tracker/services/attribution.py     | 24 +++++++++++++++-
 server/tracker/tests/test_attribution.py   | 32 ++++++++++++++++++++++
 3 files changed, 70 insertions(+), 6 deletions(-)

diff --git a/admin/src/views/tracker/visitor/detail.vue b/admin/src/views/tracker/visitor/detail.vue
index face236..5ae9eca 100644
--- a/admin/src/views/tracker/visitor/detail.vue
+++ b/admin/src/views/tracker/visitor/detail.vue
@@ -90,7 +90,7 @@
                 <ElRadioButton :value="30">近 30 天</ElRadioButton>
                 <ElRadioButton :value="90">近 90 天</ElRadioButton>
               </ElRadioGroup>
-              <!-- 正序/倒序真实切换：倒序=日期组最新在前；正序=整体按时间正序 -->
+              <!-- 正序/倒序真实切换：倒序=日期组与组内事件都最新在前；正序=都最早在前 -->
               <div class="vd-order-toggle" role="group" aria-label="时间线排序">
                 <button
                   type="button"
@@ -640,7 +640,7 @@
   // ---------- 跨天时间线 ----------
   const activeTab = ref<'sessions' | 'raw'>('sessions')
   const days = ref(30)
-  /** 排序：desc=日期组最新在前（现行为）；asc=整体按时间正序（最早日期组在前） */
+  /** 排序：desc=日期组最新在前+组内事件最新在前；asc=都最早在前 */
   const order = ref<'desc' | 'asc'>('desc')
   const timeline = ref<TimelineResp | null>(null)
   const timelineLoading = ref(false)
@@ -662,10 +662,17 @@
     }
   }
 
-  /** 展示用日期组：倒序原样（后端倒序），正序整体翻转 */
+  /** 展示用日期组：后端=日期组倒序+组内正序，前端按口径组装（浅拷贝，不动原数据）。
+   *  desc=日期组最新在前+组内（会话/事件）最新在前；asc=日期组最早在前+组内最早在前 */
   const orderedDays = computed<TimelineDay[]>(() => {
     const list = timeline.value?.days ?? []
-    return order.value === 'desc' ? list : [...list].reverse()
+    if (order.value === 'asc') return [...list].reverse()
+    return list.map((day) => ({
+      ...day,
+      sessions: [...day.sessions]
+        .reverse()
+        .map((s) => ({ ...s, events: [...s.events].reverse() }))
+    }))
   })
 
   /** ISO 日期（YYYY-MM-DD）→ 「09月16日」 */
@@ -1113,8 +1120,11 @@
     padding-bottom: 12px;
   }
 
+  // 时间线滚动容器：长轨迹卡内滚动（日期组头随内容滚动，不做 sticky）
   .tl-body {
-    min-height: 200px;
+    max-height: calc(100vh - 380px);
+    min-height: 320px;
+    overflow-y: auto;
     padding: 4px 20px 8px;
   }
 
diff --git a/server/tracker/services/attribution.py b/server/tracker/services/attribution.py
index 796ac85..7ae9824 100644
--- a/server/tracker/services/attribution.py
+++ b/server/tracker/services/attribution.py
@@ -19,6 +19,10 @@ from tracker import constants as C
 _CTRL_CHARS = re.compile(r"[\x00-\x1f\x7f]")
 _GENERIC_TLDS = ("com",)   # 引擎 host 落在这些 TLD 时不作为国家 variant 显示
 
+# UTM 标签 ↔ 平台名别名（冲突判定前归一）：缩写/中文品牌名 → 标准名；baidu 恒等仅为可读
+_UTM_ALIASES = {"fb": "facebook", "ig": "instagram", "tt": "tiktok",
+                "wa": "whatsapp", "g": "google", "baidu": "baidu", "百度": "baidu"}
+
 
 def _host(url: str) -> str:
     """提取 referrer host；异常/空 → ""。hostname 自动去端口并小写。"""
@@ -88,6 +92,22 @@ def _in_app_name(ua: str) -> str:
     return ""
 
 
+def _canon(name: str) -> str:
+    """平台名/UTM 标签宽松归一（仅用于冲突判定）：小写 → 去协议与路径 → 去 www. 前缀
+    → 去最后一段域名后缀（chatgpt.com→chatgpt、google.ru→google）→ 别名表 → 仅留
+    字母数字（含汉字）。fb↔Facebook、百度↔baidu 归一后同源，不再误报 conflict。"""
+    s = (name or "").lower().strip()
+    if "://" in s:
+        s = s.split("://", 1)[1]
+    s = s.split("/", 1)[0]
+    if s.startswith("www."):
+        s = s[4:]
+    if "." in s:
+        s = s.rsplit(".", 1)[0]
+    s = _UTM_ALIASES.get(s, s)
+    return re.sub(r"[^0-9a-z一-鿿]", "", s)
+
+
 def _utm_channel(label: str) -> str:
     """UTM 标签尽力归类：已知 AI/引擎/社交标签（裸标签 "baidu"/显示名 "Google" 皆可）
     映射到对应渠道，未知 → referral（有名来源，可信度由 label_only 单独表达）。"""
@@ -150,8 +170,10 @@ def classify(*, referrer_url: str = "", utm: dict | None = None,
     # UTM 证据叠加：观察结论之上只做印证/并列，不覆盖
     if utm_source:
         if out["resolution"] == "resolved":
-            if utm_source.lower() == out["platform"].lower():
+            if _canon(utm_source) == _canon(out["platform"]):
+                # 归一后同源 → 互相印证；UTM 原文并记进 detail（顶层展示观察结论）
                 out["evidence_kind"] = "combined"
+                out["detail"] = {**out["detail"], "utm_source": utm_source}
             else:   # 矛盾：观察证据保留在顶层，两路证据各自完整并列进 detail
                 out["resolution"] = "conflict"
                 out["detail"] = {**out["detail"],
diff --git a/server/tracker/tests/test_attribution.py b/server/tracker/tests/test_attribution.py
index 66e5a65..bb85900 100644
--- a/server/tracker/tests/test_attribution.py
+++ b/server/tracker/tests/test_attribution.py
@@ -113,6 +113,38 @@ def test_utm_contradicting_referrer_is_conflict_with_both_evidences():
     assert utm_ev["platform"] == "google" and utm_ev["evidence_kind"] == "utm_label"
 
 
+# ------------------------------------------- 别名归一（消 conflict 误报）
+def test_utm_alias_fb_matches_facebook_referrer():
+    r = classify(referrer_url="https://facebook.com/l/abc", utm={"utm_source": "fb"})
+    assert r["resolution"] != "conflict"                # fb↔Facebook 归一后同源，不是矛盾
+    assert r["channel"] == "social" and r["platform"] == "Facebook"
+    assert r["evidence_kind"] == "combined"
+    assert r["detail"].get("utm_source") == "fb"        # 归一对齐时原文并记进 detail
+
+
+def test_utm_with_domain_suffix_matches_bare_platform():
+    r = classify(referrer_url="https://chatgpt.com/c/abc", utm={"utm_source": "chatgpt.com"})
+    assert r["evidence_kind"] == "combined"             # chatgpt.com → chatgpt 与 ChatGPT 同源
+    assert r["platform"] == "ChatGPT"
+    assert r["detail"].get("utm_source") == "chatgpt.com"
+    r2 = classify(referrer_url="https://www.google.com/search?q=x",
+                  utm={"utm_source": "google.ru"})      # 国家域名后缀 → google
+    assert r2["evidence_kind"] == "combined" and r2["platform"] == "Google"
+
+
+def test_utm_alias_still_conflict_when_platforms_differ():
+    r = classify(referrer_url="https://facebook.com/l/abc", utm={"utm_source": "google"})
+    assert r["resolution"] == "conflict"                # 归一后仍不同 → 保持矛盾并列
+
+
+def test_utm_chinese_alias_baidu_matches():
+    r = classify(referrer_url="https://www.baidu.com/s?wd=x", utm={"utm_source": "百度"})
+    assert r["resolution"] != "conflict"                # 百度 ↔ baidu.com(平台"百度") 同源
+    assert r["channel"] == "natural_search" and r["platform"] == "百度"
+    assert r["evidence_kind"] == "combined"
+    assert r["detail"].get("utm_source") == "百度"
+
+
 def test_utm_only_is_label_only():
     r = classify(utm={"utm_source": "newsletter"})
     assert r["channel"] == "referral"               # 未知标签 → 有名 referral，但状态诚实标注