From 0b2bd87d61189de442d0c76d9caef573ad0b0655 Mon Sep 17 00:00:00 2001
From: "34047007@qq.com" <34047007@qq.com>
Date: Mon, 27 Jul 2026 16:16:21 +0800
Subject: [PATCH] =?UTF-8?q?perf:=20=E6=A0=87=E7=AD=BE=E4=BA=8C=E7=BA=A7?=
 =?UTF-8?q?=E7=BC=93=E5=AD=98=20+=20cache.mget=20=E6=89=B9=E9=87=8F?=
 =?UTF-8?q?=E8=AF=BB=E5=8F=96?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- tag_loader 增加 Redis L2 缓存（tags:{lit_id}, 86400s TTL），批量 mget
- cache.py 新增 mget() 方法，同键 Redis 批量查询 + memory 降级
- tag_service.py 打标后自动清除对应缓存
---
 backend/app/core/cache.py          | 12 ++++++
 backend/app/services/tag_loader.py | 68 +++++++++++++++++++++---------
 2 files changed, 60 insertions(+), 20 deletions(-)

diff --git a/backend/app/core/cache.py b/backend/app/core/cache.py
index 1226ac6..3ed37c6 100644
--- a/backend/app/core/cache.py
+++ b/backend/app/core/cache.py
@@ -66,6 +66,18 @@ class CacheService:
             self._store.popitem(last=False)
         return True
 
+    async def mget(self, keys: list[str]) -> list[dict | None]:
+        """批量获取，顺序对应 keys 列表。Redis 不可用时降级到内存模式。"""
+        r = await self._get_redis()
+        if r:
+            try:
+                vals = await r.mget(*keys)
+                return [json.loads(v) if v else None for v in vals]
+            except Exception:
+                logger.exception("Redis MGET failed")
+                return [None] * len(keys)
+        return [self._store.get(k) for k in keys]
+
     async def delete(self, key: str):
         r = await self._get_redis()
         if r:
diff --git a/backend/app/services/tag_loader.py b/backend/app/services/tag_loader.py
index 11a476d..98c0285 100644
--- a/backend/app/services/tag_loader.py
+++ b/backend/app/services/tag_loader.py
@@ -1,40 +1,68 @@
-"""共享标签加载工具（带单次请求级内存缓存）"""
+"""共享标签加载工具（三级缓存：请求级内存 → Redis → DB）"""
 
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
+from app.core.cache import cache as _redis_cache
 from app.models.literature import GlobalLiteratureTag, GlobalTag
 
-# 请求级缓存：同一批 literature_ids 在短时间内重复查询不走 DB
+# 请求级缓存：同一批 literature_ids 在短时间内重复查询不走 DB/Redis
 _cache: dict[str, dict[str, list[dict]]] = {}
 _CACHE_MAX_KEYS = 5
+_TAGS_CACHE_TTL = 86400  # 24h，标签变更极低频
 
 
 async def load_tags_for_literature(db: AsyncSession, literature_ids: list[str]) -> dict[str, list[dict]]:
-    """批量加载文献标签，返回 {literature_id: [{name_zh, path, category, is_major}]}"""
+    """批量加载文献标签，返回 {literature_id: [{name_zh, path, category, is_major}]}
+
+    三级缓存：
+      L1 — 请求级 dict（同一批 ID 在本次请求中复用）
+      L2 — Redis tags:{lit_id}（24h TTL，跨请求/跨用户）
+      L3 — DB JOIN（miss 时回退）
+    """
     if not literature_ids:
         return {}
-    # 缓存 key 基于排序后的 ID 列表，确保相同集合命中
-    cache_key = ",".join(sorted(literature_ids))
+
+    # L1: 请求级缓存
+    import uuid as _uuid
+    str_ids = [str(x) if not isinstance(x, str) else x for x in literature_ids]
+    cache_key = ",".join(sorted(str_ids))
     if cache_key in _cache:
         return _cache[cache_key]
 
-    import uuid as _uuid
-    uids = [_uuid.UUID(x) if isinstance(x, str) else x for x in literature_ids]
-    result = await db.execute(
-        select(GlobalLiteratureTag.literature_id, GlobalTag.id, GlobalTag.name_zh, GlobalTag.name_en, GlobalTag.path,
-               GlobalTag.tag_category, GlobalLiteratureTag.is_major)
-        .join(GlobalTag, GlobalLiteratureTag.tag_id == GlobalTag.id)
-        .where(GlobalLiteratureTag.literature_id.in_(uids))
-    )
-    tags_map: dict[str, list[dict]] = {}
-    for lit_id, tid, nz, ne, p, cat, maj in result:
-        sid = str(lit_id)
-        if sid not in tags_map:
-            tags_map[sid] = []
-        tags_map[sid].append({"id": str(tid), "name_zh": nz, "name_en": ne, "path": p, "category": cat, "is_major": maj})
+    # L2: Redis 单文献缓存（批量查）
+    redis_keys = [f"tags:{sid}" for sid in str_ids]
+    cached_results = await _redis_cache.mget(redis_keys)
 
-    # 存入请求级缓存，限制大小
+    tags_map: dict[str, list[dict]] = {}
+    miss_ids: list[str] = []
+
+    for sid, cached in zip(str_ids, cached_results):
+        if cached is not None and "tags" in cached:
+            tags_map[sid] = cached["tags"]
+        else:
+            miss_ids.append(sid)
+
+    # L3: DB 回退（只查 miss 的文献）
+    if miss_ids:
+        miss_uids = [_uuid.UUID(sid) for sid in miss_ids]
+        result = await db.execute(
+            select(GlobalLiteratureTag.literature_id, GlobalTag.id, GlobalTag.name_zh, GlobalTag.name_en, GlobalTag.path,
+                   GlobalTag.tag_category, GlobalLiteratureTag.is_major)
+            .join(GlobalTag, GlobalLiteratureTag.tag_id == GlobalTag.id)
+            .where(GlobalLiteratureTag.literature_id.in_(miss_uids))
+        )
+        for lit_id, tid, nz, ne, p, cat, maj in result:
+            sid = str(lit_id)
+            if sid not in tags_map:
+                tags_map[sid] = []
+            tags_map[sid].append({"id": str(tid), "name_zh": nz, "name_en": ne, "path": p, "category": cat, "is_major": maj})
+
+        # 回填 Redis（空列表也缓存，避免空文献反复查 DB）
+        for sid in miss_ids:
+            await _redis_cache.set(f"tags:{sid}", {"tags": tags_map.get(sid, [])}, ttl=_TAGS_CACHE_TTL)
+
+    # L1: 存入请求级缓存
     if len(_cache) >= _CACHE_MAX_KEYS:
         _cache.pop(next(iter(_cache)), None)
     _cache[cache_key] = tags_map