From 8a4826220aaf56f781cb34166cd4d6fe774b8978 Mon Sep 17 00:00:00 2001
From: "34047007@qq.com" <34047007@qq.com>
Date: Mon, 27 Jul 2026 14:23:10 +0800
Subject: [PATCH] =?UTF-8?q?fix:=20D2=20=E4=B8=AD=E6=96=87=E6=90=9C?=
 =?UTF-8?q?=E7=B4=A2=20+=20D3=20=E6=B7=B7=E5=90=88=E5=B8=83=E5=B0=94?=
 =?UTF-8?q?=E8=AF=AD=E4=B9=89=20=E2=80=94=20PubMed=20=E6=90=9C=E7=B4=A2?=
 =?UTF-8?q?=E8=B7=AF=E5=BE=84=E4=BF=AE=E5=A4=8D?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

D2: PubMed 路径搜索中文（不带字段标签）时，plainto_tsquery("english")
    对中文返回空，导致零结果。新增 ILIKE fallback 检测中文字符。
D3: A OR B AND C 被解析器拉平为 A OR B OR C。修改 _parse_or_expr
    保留 AND cluster 分组，引擎正确产生产出 A OR (B AND C)。
---
 backend/app/services/pubmed_query_parser.py | 31 +++++++++++++++++----
 backend/app/services/search_engine.py       |  6 ++++
 2 files changed, 32 insertions(+), 5 deletions(-)

diff --git a/backend/app/services/pubmed_query_parser.py b/backend/app/services/pubmed_query_parser.py
index ddbef95..96deea1 100644
--- a/backend/app/services/pubmed_query_parser.py
+++ b/backend/app/services/pubmed_query_parser.py
@@ -473,15 +473,36 @@ class PubmedQueryParser:
             result.plain_terms.append(term)
 
     def _parse_or_expr(self, result: ParsedPubmedQuery) -> list[Term]:
-        """or_expr → and_expr (OR and_expr)*"""
-        left = self._parse_and_expr(result)
+        """or_expr → and_expr (OR and_expr)*
+
+        P7-D3: Preserve AND clusters when OR is mixed (e.g. ``A OR B AND C``).
+        Collect each ``_parse_and_expr`` result as a separate cluster. When OR
+        is found AND a cluster has >1 term (implicit AND), wrap it in a group
+        so the engine produces ``A OR (B AND C)`` instead of ``A OR B OR C``.
+        """
+        clusters = [self._parse_and_expr(result)]
+        had_or = False
         while self.peek().type == TokenType.OR:
+            had_or = True
             self.advance()
             if self.peek().type == TokenType.EOF:
                 break  # trailing OR, ignore silently
-            right = self._parse_and_expr(result)
-            left.extend(right)
-        return left
+            clusters.append(self._parse_and_expr(result))
+
+        if not had_or:
+            return clusters[0] if clusters else []
+
+        # OR present: group any AND-cluster with >1 term
+        all_terms: list[Term] = []
+        for cluster in clusters:
+            if len(cluster) > 1 and not any(t.group_id >= 0 for t in cluster):
+                gid = len(result.groups)
+                for t in cluster:
+                    t.group_id = gid
+                result.groups.append(cluster)
+                result.group_operators.append("and")
+            all_terms.extend(cluster)
+        return all_terms
 
     def _is_primary_start(self, token: Token) -> bool:
         """Check if token could start a primary expression."""
diff --git a/backend/app/services/search_engine.py b/backend/app/services/search_engine.py
index cac5768..50775e8 100644
--- a/backend/app/services/search_engine.py
+++ b/backend/app/services/search_engine.py
@@ -1195,6 +1195,12 @@ class AdvancedSearchEngine:
                     cast(GlobalLiterature.pmid, String).ilike(like_val),
                     GlobalLiterature.doi.ilike(like_val),
                 )
+            # P7-D2: Chinese → ILIKE fallback (tsvector is English-only)
+            if re.search(r'[一-鿿㐀-䶿豈-﫿]', term):
+                return or_(
+                    GlobalLiterature.title.ilike(like_val),
+                    GlobalLiterature.abstract.ilike(like_val),
+                )
             return GlobalLiterature.search_tsv.op("@@")(func.plainto_tsquery("english", term))
 
     @staticmethod