From 6a5a5d7b6bdbe7b8eb802335c3dd540872f017fb Mon Sep 17 00:00:00 2001
From: caoqianming <caoqianming@foxmail.com>
Date: Thu, 21 May 2026 13:48:52 +0800
Subject: [PATCH] =?UTF-8?q?feat:=20paper=20list=20=E5=8A=A0=20pdf=5Furl=20?=
 =?UTF-8?q?/=20xml=5Furl=20=E7=9B=B4=E9=93=BE=E5=AD=97=E6=AE=B5=20+=20pg?=
 =?UTF-8?q?=5Ftrgm=20GIN=20=E7=B4=A2=E5=BC=95?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

serializers: PaperListSerializer 加 pdf_url / xml_url SerializerMethodField,基于 publication_date + safe_doi 后端拼 absolute_uri;has_fulltext_{pdf,xml}=False 或 publication_date 缺失返空串。LLM 客户端从 list 一次拿到直链,不必拼 URL。
migration 0006: CREATE EXTENSION IF NOT EXISTS pg_trgm + 3 列 GIN 索引(title / first_author / first_author_institution),根治 SearchFilter 跨列 ILIKE '%xxx%' 全表扫 timeout(高频词如 cement 原本 30s+,加索引后几十 ms)。

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---
 apps/resm/migrations/0006_pg_trgm_index.py | 37 ++++++++++++++++++++++
 apps/resm/serializers.py                   | 22 +++++++++++++
 2 files changed, 59 insertions(+)
 create mode 100644 apps/resm/migrations/0006_pg_trgm_index.py

diff --git a/apps/resm/migrations/0006_pg_trgm_index.py b/apps/resm/migrations/0006_pg_trgm_index.py
new file mode 100644
index 0000000..7a48924
--- /dev/null
+++ b/apps/resm/migrations/0006_pg_trgm_index.py
@@ -0,0 +1,37 @@
+"""为 SearchFilter 的 title / first_author / first_author_institution 三列建 pg_trgm GIN 索引。
+
+原因:DRF SearchFilter 走 `column ILIKE '%keyword%'` 前后通配,B-tree 索引救不了;
+高频关键词 + 几十万行表会 30s+ timeout。pg_trgm 给 trigram 建 GIN 索引,
+ILIKE '%xxx%' 走索引,降到几十 ms。
+
+pg_trgm 是 PostgreSQL contrib 扩展,首次启用需要 DB superuser 权限;
+CREATE EXTENSION IF NOT EXISTS 幂等,迁移可安全重跑。
+"""
+from django.db import migrations
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('resm', '0005_alter_paper_fetch_status'),
+    ]
+
+    operations = [
+        migrations.RunSQL(
+            sql=[
+                "CREATE EXTENSION IF NOT EXISTS pg_trgm;",
+                "CREATE INDEX IF NOT EXISTS paper_title_trgm "
+                "ON resm_paper USING gin (title gin_trgm_ops);",
+                "CREATE INDEX IF NOT EXISTS paper_first_author_trgm "
+                "ON resm_paper USING gin (first_author gin_trgm_ops);",
+                "CREATE INDEX IF NOT EXISTS paper_first_author_institution_trgm "
+                "ON resm_paper USING gin (first_author_institution gin_trgm_ops);",
+            ],
+            reverse_sql=[
+                "DROP INDEX IF EXISTS paper_first_author_institution_trgm;",
+                "DROP INDEX IF EXISTS paper_first_author_trgm;",
+                "DROP INDEX IF EXISTS paper_title_trgm;",
+                # pg_trgm 扩展不 drop —— 其他 app / 表可能也在用
+            ],
+        ),
+    ]
diff --git a/apps/resm/serializers.py b/apps/resm/serializers.py
index 73efee7..be5bac7 100644
--- a/apps/resm/serializers.py
+++ b/apps/resm/serializers.py
@@ -5,6 +5,8 @@ from .models import Paper
 
 class PaperListSerializer(CustomModelSerializer):
     abstract = serializers.SerializerMethodField()
+    pdf_url = serializers.SerializerMethodField()
+    xml_url = serializers.SerializerMethodField()
 
     class Meta:
         model = Paper
@@ -16,3 +18,23 @@ class PaperListSerializer(CustomModelSerializer):
         if abs_obj is None:
             return ""
         return abs_obj.abstract or ""
+
+    def _media_url(self, obj, ext: str) -> str:
+        """拼 /media/papers/<Y>/<M>/<D>/<safe_doi>.<ext> 静态直链。
+
+        date 缺失(unknown 目录)/ doi 缺失返空串 —— LLM 拿到空串就知道没法直链下,
+        改走 get_paper / paper_pdf_view。
+        """
+        if obj.publication_date is None or not obj.doi:
+            return ""
+        safe_doi = obj.doi.replace("/", "_")
+        d = obj.publication_date
+        path = f"/media/papers/{d.year}/{d.month}/{d.day}/{safe_doi}.{ext}"
+        request = self.context.get("request")
+        return request.build_absolute_uri(path) if request else path
+
+    def get_pdf_url(self, obj) -> str:
+        return self._media_url(obj, "pdf") if obj.has_fulltext_pdf else ""
+
+    def get_xml_url(self, obj) -> str:
+        return self._media_url(obj, "xml") if obj.has_fulltext_xml else ""