From 169e7987188acb59a165d2fd8730d6f03ea7a856 Mon Sep 17 00:00:00 2001
From: stefanfeng <stefan@autopas.cn>
Date: Tue, 7 Apr 2026 13:24:42 +0800
Subject: [PATCH] =?UTF-8?q?fix:=20=E5=8E=86=E5=8F=B2=E6=96=87=E7=AB=A0?=
 =?UTF-8?q?=E6=94=B9=E4=B8=BA=E4=BB=8E=E6=96=B0=E5=88=B0=E6=97=A7=E9=A1=BA?=
 =?UTF-8?q?=E5=BA=8F=E7=BF=BB=E9=A1=B5=EF=BC=8C=E8=80=8C=E9=9D=9E=E9=9A=8F?=
 =?UTF-8?q?=E6=9C=BA?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

修复原因：
- 需求要求从最新文章开始往旧的方向依次互动
- 原实现随机选页，导致每次跳到不同页，无法体现「从新到旧」
- 列表API返回的文章已按 createTime 降序排列（第1页最新）

新Phase 2逻辑：
- 计算总页数（最多10页）
- 用当前小时 % 总页数 决定起始页（同小时内分散到不同页）
- 若起始页为空，顺序往后再折回，直到找到有文章的页
- 在同一页内随机打散，保证同一时段不同用户不总是抢相同文章
- validate_article 校验，不够则从剩余补充

Phase 1（今日新文章）逻辑不变：
- 仍从第1页抓取当天文章，按 createTime 降序（新→旧）排列后校验返回
---
 backend/app/services/news_service.py | 91 +++++++++++++---------------
 1 file changed, 42 insertions(+), 49 deletions(-)

diff --git a/backend/app/services/news_service.py b/backend/app/services/news_service.py
index f815007..9bb1027 100755
--- a/backend/app/services/news_service.py
+++ b/backend/app/services/news_service.py
@@ -488,8 +488,12 @@ class NewsPlatformService:
                 return valid
             logger.info(f"[广场新闻] {user.account} 今日文章校验后全部无效，转历史")
 
-        # ── Phase 2: 无今日新文章 → 历史随机翻页（热度+新鲜度加权）────────
-        logger.info(f"[广场新闻] {user.account} 无今日新文章，随机历史翻页")
+        # ── Phase 2: 无今日新文章 → 从最新到旧顺序翻页（随机从某页开始）────
+        # 规则：从第1页（最新）开始依次往后，轮到哪页由当前小时决定，
+        #       保证不同时段覆盖不同页，模拟"从新到旧"逐步互动
+        logger.info(f"[广场新闻] {user.account} 无今日新文章，从最新向旧顺序翻页")
+
+        # 获取总页数
         total_pages = 1
         try:
             async with httpx.AsyncClient(timeout=10) as _c:
@@ -505,58 +509,48 @@ class NewsPlatformService:
         except Exception:
             pass
 
-        rand_page = _rand.randint(1, min(total_pages, 10))
-        items = []
-        try:
-            async with httpx.AsyncClient(timeout=15) as c:
-                r = await c.get(
-                    f"{biz}/business/member/square/list",
-                    headers=self._bearer(token),
-                    params=_build(rand_page),
-                )
-            if r.status_code == 200:
-                d = r.json()
-                if d.get("code") in [0, 200]:
-                    nd    = d.get("data", {})
-                    items = nd.get("data") or nd.get("list") or nd.get("records") or []
-                    items = _filter(items)
-        except Exception as e:
-            logger.error(f"[广场新闻-历史] {user.account}: {e}")
+        max_pages = min(total_pages, 10)   # 最多翻前10页
+        # 用小时数取模决定起始页，保证同一小时内不同用户分散在不同页
+        # 从最新(page=1)开始往旧的方向走
+        hour_slot = _dt.now().hour % max_pages
+        start_page = hour_slot + 1  # 1-based
 
-        logger.info(f"[广场新闻] {user.account} 历史第{rand_page}页获取到 {len(items)} 条")
+        items = []
+        # 尝试从 start_page 开始，若该页为空则顺序往后再往前找
+        pages_to_try = list(range(start_page, max_pages + 1)) + list(range(1, start_page))
+        tried_page = start_page
+        for page in pages_to_try:
+            try:
+                async with httpx.AsyncClient(timeout=15) as c:
+                    r = await c.get(
+                        f"{biz}/business/member/square/list",
+                        headers=self._bearer(token),
+                        params=_build(page),
+                    )
+                if r.status_code == 200:
+                    d = r.json()
+                    if d.get("code") in [0, 200]:
+                        nd     = d.get("data", {})
+                        _items = nd.get("data") or nd.get("list") or nd.get("records") or []
+                        _items = _filter(_items)
+                        if _items:
+                            items = _items
+                            tried_page = page
+                            break
+            except Exception as e:
+                logger.error(f"[广场新闻-历史] {user.account} page={page}: {e}")
+
+        logger.info(f"[广场新闻] {user.account} 历史第{tried_page}页获取到 {len(items)} 条")
 
         if not items:
             return []
 
-        # 热度 + 新鲜度加权采样
-        def _hot_weight(a):
-            hot = (int(a.get("commentNum") or 0) * 3 +
-                   int(a.get("praiseNum")  or 0) * 2 +
-                   int(a.get("readNum")    or 0))
-            freshness = 1.0
-            t = a.get("createTime") or a.get("publishTime") or ""
-            if t:
-                try:
-                    pub = _dt.strptime(t[:19], "%Y-%m-%d %H:%M:%S")
-                    h   = (_dt.now() - pub).total_seconds() / 3600
-                    freshness = max(1.0, 3.0 - h / 36.0)
-                except Exception:
-                    pass
-            return max(1.0, (hot + 1) * freshness)
-
-        weights  = [_hot_weight(a) for a in items]
-        pool_idx = list(range(len(items)))
-        selected = []
-        for _ in range(min(count * 2, len(items))):
-            if not pool_idx:
-                break
-            ci = _rand.choices(pool_idx, weights=[weights[i] for i in pool_idx], k=1)[0]
-            selected.append(items[ci])
-            pool_idx.remove(ci)
+        # 在该页内随机打散（同一页内不需要排序，保持自然顺序即可）
+        _rand.shuffle(items)
+        selected = items[:min(count * 2, len(items))]
 
         # 有效性校验
         valid = []
-        remaining = [i for i in pool_idx]
         for a in selected:
             if len(valid) >= count:
                 break
@@ -564,11 +558,10 @@ class NewsPlatformService:
             if await self.validate_article(db, user, aid):
                 valid.append(a)
 
-        # 不够则从剩余池补充
-        for ri in remaining:
+        # 不够则从剩余补充
+        for a in items[len(selected):]:
             if len(valid) >= count:
                 break
-            a   = items[ri]
             aid = str(a.get("recordId") or a.get("id", ""))
             if await self.validate_article(db, user, aid):
                 valid.append(a)