feat: PubMed搜索合规审计 Phase 5-7 全部修复

P1-2: [SB][STAT][UID][DEP][MESH]字段支持
P1-4: [MH:noexp]解析+抑制树展开
P3: 前端date_to/P3-1、CJK高亮/P3-3、extractPlainText/P3-4、precision_mode移除/P3-6、loadMore sort/P3-5
P4-1: 词数上限50→100
P4-2: retracted参数清理

backend/app/api/v1/features.py

@@ -89,8 +89,8 @@ class AdvancedSearchRequest(BaseModel):
     @field_validator('query')
     @classmethod
     def check_query_complexity(cls, v: str) -> str:
-        if len(v.split()) > 50:
-            raise ValueError('查询词过多（最多 50 个词），请简化搜索条件')
+        if len(v.split()) > 100:
+            raise ValueError('查询词过多（最多 100 个词），请简化搜索条件')
         return v
 
 

backend/app/services/pubmed_query_parser.py

@@ -23,7 +23,7 @@ from dataclasses import dataclass, field
 from enum import Enum, auto
 
 # ─── 查询复杂度限制 ───
-MAX_TERMS = 50       # 与 API 层 check_query_complexity 的 50 词限制对齐
+MAX_TERMS = 100      # P4-1: 放宽到 100 词（原 50 词）
 MAX_PAREN_DEPTH = 10  # 括号嵌套最深层数
 
 
@@ -42,6 +42,7 @@ _FIELD_TAG_MAP: dict[str, str] = {
     "LAU": "author",
     "TW": "all",
     "OT": "all",
+    "MESH": "MH",            # P1-2: [MESH] 是 [MH] 的别名
     # 新增标量字段
     "LA": "language",
     "VI": "volume",
@@ -55,23 +56,30 @@ _SPECIAL_FIELDS = {
     "MH", "MAJR", "PT", "DP", "PMID", "DOI",
     # 日期字段
     "EDAT", "CRDT", "MHDA", "LR", "DCOM",
+    # P1-2: 电子出版日期
+    "DEP",
     # JSONB 字段
     "GR", "SH", "RN", "NM", "SI", "PA",
     # 待补充抽取的字段（标记为特殊以便未来实现）
     "AUID", "COIS", "ED", "IR", "PS", "PUBN", "TT",
+    # P1-2: 新字段
+    "SB", "STAT", "UID",
 }
 
 # 支持日期范围语法的字段
-_DATE_RANGE_FIELDS = {"DP", "EDAT", "CRDT", "MHDA", "LR", "DCOM"}
+_DATE_RANGE_FIELDS = {"DP", "EDAT", "CRDT", "MHDA", "LR", "DCOM", "DEP"}
 
 # 所有合法字段标签（PubMed 全量字段）
 # P1-1: 移除了 BOOK/FILTER/ISBN（未实现，降级为 plain text 不如报错透明）
 _ALL_FIELD_TAGS = {
     "AB", "AD", "AU", "CN", "FAU", "AUID", "LAU", "COIS",
     "DCOM", "CRDT", "EDAT", "MHDA", "LR", "DP", "DOI",
+    "DEP",                               # P1-2: Date of Electronic Publication
     "RN", "ED", "GR", "IR", "IP",
     "TA", "JT", "LA", "LID", "MAJR", "SH", "MH", "MH:NOEXP", "OT", "PG",
     "PA", "PT", "PMID", "PUBN", "SI", "PS", "NM", "TW",
+    "SB", "STAT", "UID",                 # P1-2: Subset, Status, UID
+    "MESH",                              # P1-2: [MH] 别名
     "TI", "TIAB", "TT", "VI",
 }
 
@@ -187,6 +195,8 @@ class ParsedPubmedQuery:
     lr_to: str | None = None
     dcom_from: str | None = None                                   # [DCOM]
     dcom_to: str | None = None
+    dep_from: str | None = None                                     # P1-2: [DEP]
+    dep_to: str | None = None
     # 特殊字段（存储 Term 以保留 is_not 标志）
     mesh_terms: list[Term] = field(default_factory=list)           # [MH] (was list[str])
     majr_terms: list[Term] = field(default_factory=list)           # [MAJR] (was list[str])
@@ -207,6 +217,9 @@ class ParsedPubmedQuery:
     auid_terms: list[Term] = field(default_factory=list)             # [AUID]
     cois_terms: list[Term] = field(default_factory=list)             # [COIS]
     tt_terms: list[Term] = field(default_factory=list)               # [TT]
+    sb_terms: list[Term] = field(default_factory=list)               # P1-2: [SB] Subset
+    stat_terms: list[Term] = field(default_factory=list)             # P1-2: [STAT] Status
+    uid_terms: list[Term] = field(default_factory=list)              # P1-2: [UID] PMID/DOI 统一
     plain_terms: list[Term] = field(default_factory=list)          # no field tag
     boolean_operator: str = "and"                                  # "and" | "or" | "mixed"
     has_not: bool = False                                          # contains NOT
@@ -349,6 +362,8 @@ class PubmedQueryParser:
             pass
         elif term.field == "__RANGE_DCOM__":
             pass
+        elif term.field == "__RANGE_DEP__":
+            pass
         elif term.field == "ED":
             result.ed_terms.append(term)
         elif term.field == "IR":
@@ -363,6 +378,12 @@ class PubmedQueryParser:
             result.cois_terms.append(term)
         elif term.field == "TT":
             result.tt_terms.append(term)
+        elif term.field == "SB":
+            result.sb_terms.append(term)
+        elif term.field == "STAT":
+            result.stat_terms.append(term)
+        elif term.field == "UID":
+            result.uid_terms.append(term)
         else:
             result.plain_terms.append(term)
 
@@ -494,6 +515,7 @@ class PubmedQueryParser:
                 "MHDA": ("mhda_from", "mhda_to", None, None, "__RANGE_MHDA__"),
                 "LR": ("lr_from", "lr_to", None, None, "__RANGE_LR__"),
                 "DCOM": ("dcom_from", "dcom_to", None, None, "__RANGE_DCOM__"),
+                "DEP": ("dep_from", "dep_to", None, None, "__RANGE_DEP__"),  # P1-2
             }
             date_attr, date_attr_to, yr_from_attr, yr_to_attr, marker_field = attr_map[field]
             # 反向范围自动交换（如 2026:2024[DP] → 2024:2026[DP]）

backend/app/services/search_engine.py

@@ -97,7 +97,7 @@ class AdvancedSearchEngine:
         journal_tiers: list[str] | None = None,
         pub_types: list[str] | None = None,
         tag_ids: list[str] | None = None,
-        retracted: str = "",             # "yes", "no", "only"
+        retracted: str = "",             # "no", "only"
         negative_result: str = "",       # "yes", "no", "only"
         is_oa: bool | None = None,       # True = 仅开放获取
         language: str | None = None,      # 语言代码（en/zh/fr 等，向后兼容）
@@ -820,6 +820,42 @@ class AdvancedSearchEngine:
                 neg_conds = [GlobalLiterature.vernacular_title.ilike(f"%{_escape_ilike(t.text)}%") for t in neg]
                 term_conditions.append(not_(or_(*neg_conds)))
 
+        # P1-2: [SB] Subset → nlm_subsets（需 journal JOIN 子查询）
+        if pp.sb_terms:
+            pos = [t for t in pp.sb_terms if not t.is_not]
+            neg = [t for t in pp.sb_terms if t.is_not]
+            for subl, is_neg in [(pos, False), (neg, True)]:
+                for t in subl:
+                    subq = select(GlobalJournal.issn).where(
+                        GlobalJournal.nlm_subsets.overlap([t.text.upper()])
+                    )
+                    cond = GlobalLiterature.journal_issn.in_(subq)
+                    term_conditions.append(not_(cond) if is_neg else cond)
+
+        # P1-2: [STAT] Status → citation_status
+        if pp.stat_terms:
+            pos = [t for t in pp.stat_terms if not t.is_not]
+            neg = [t for t in pp.stat_terms if t.is_not]
+            if pos:
+                term_conditions.append(or_(*[
+                    GlobalLiterature.citation_status == t.text.lower()
+                    for t in pos
+                ]))
+            if neg:
+                neg_conds = [GlobalLiterature.citation_status == t.text.lower() for t in neg]
+                term_conditions.append(not_(or_(*neg_conds)))
+
+        # P1-2: [UID] → PMID 优先，兜底 DOI
+        if pp.uid_terms:
+            for t in pp.uid_terms:
+                cond = None
+                try:
+                    cond = GlobalLiterature.pmid == int(t.text)
+                except ValueError:
+                    cond = GlobalLiterature.doi.ilike(f"%{_escape_ilike(t.text)}%")
+                if cond is not None:
+                    term_conditions.append(not_(cond) if t.is_not else cond)
+
         # 处理括号分组的词（保留 OR/AND 嵌套结构，P2-2）
         if pp.groups:
             for idx, group in enumerate(pp.groups):
@@ -881,13 +917,14 @@ class AdvancedSearchEngine:
             cond = and_(*dp_conds) if len(dp_conds) > 1 else dp_conds[0]
             conditions.append(not_(cond) if dp_negated else cond)
 
-        # 6b. [EDAT] [CRDT] [MHDA] [LR] [DCOM] → 日期字段范围
+        # 6b. [EDAT] [CRDT] [MHDA] [LR] [DCOM] [DEP] → 日期字段范围
         DATE_FIELD_COLS = {
             "edat": (GlobalLiterature.entrez_date, "EDAT"),
             "crdt": (GlobalLiterature.create_date, "CRDT"),
             "mhda": (GlobalLiterature.meshed_date, "MHDA"),
             "lr": (GlobalLiterature.pubmed_revised, "LR"),
             "dcom": (GlobalLiterature.date_completed, "DCOM"),
+            "dep": (GlobalLiterature.pub_date, "DEP"),  # P1-2: [DEP] → pub_date
         }
         for prefix, (col, field_tag) in DATE_FIELD_COLS.items():
             _from = getattr(pp, f"{prefix}_from", None)
@@ -942,7 +979,7 @@ class AdvancedSearchEngine:
 
         # 特殊字段 — 与 _pubmed_conditions 中 top-level dispatch 一致
         if field == "MH":
-            return await AdvancedSearchEngine._expand_mesh_tag_ids(db, [term.text], major_only=False)
+            return await AdvancedSearchEngine._expand_mesh_tag_ids(db, [term.text], major_only=False, noexp=term._noexp)
         if field == "MAJR":
             return await AdvancedSearchEngine._expand_mesh_tag_ids(db, [term.text], major_only=True)
         if field == "PT":
@@ -980,6 +1017,18 @@ class AdvancedSearchEngine:
             return GlobalLiterature.cois_statement.ilike(f"%{_escape_ilike(term.text)}%")
         if field == "TT":
             return GlobalLiterature.vernacular_title.ilike(f"%{_escape_ilike(term.text)}%")
+        if field == "SB":
+            subq = select(GlobalJournal.issn).where(
+                GlobalJournal.nlm_subsets.overlap([term.text.upper()])
+            )
+            return GlobalLiterature.journal_issn.in_(subq)
+        if field == "STAT":
+            return GlobalLiterature.citation_status == term.text.lower()
+        if field == "UID":
+            try:
+                return GlobalLiterature.pmid == int(term.text)
+            except ValueError:
+                return GlobalLiterature.doi.ilike(f"%{_escape_ilike(term.text)}%")
 
         # 回退
         return AdvancedSearchEngine._field_condition("all", term.text, term.exact)
@@ -1051,6 +1100,7 @@ class AdvancedSearchEngine:
         db: AsyncSession,
         mesh_names: list[str],
         major_only: bool = False,
+        noexp: bool = False,  # P1-4: [MH:noexp] 抑制树展开
     ) -> object | None:
         """用 GlobalTagTreeNumber 展开 [MH]/[MAJR]
 
@@ -1058,6 +1108,8 @@ class AdvancedSearchEngine:
         2. 用 tree_number 前缀展开子节点（C04.588 → 所有子树号）
         3. [MAJR] 额外 AND is_major=True
         返回 SQLAlchemy condition 或 None（无匹配时）。
+
+        noexp=True 时跳过第 2 步（树展开），只搜精确词。
         """
         import uuid as _uuid
         mesh_tag_ids: set[_uuid.UUID] = set()
@@ -1091,19 +1143,20 @@ class AdvancedSearchEngine:
         if not mesh_tag_ids:
             return None
 
-        # tree_number 前缀展开：取匹配 tag 的所有 tree_number，查子节点
-        tns = (await db.execute(
-            select(GlobalTagTreeNumber.tree_number).where(
-                GlobalTagTreeNumber.tag_id.in_(list(mesh_tag_ids))
-            ).distinct()
-        )).scalars().all()
-
-        if tns:
-            child_conds = [GlobalTagTreeNumber.tree_number.like(f"{tn}%") for tn in tns]
-            children = (await db.execute(
-                select(GlobalTagTreeNumber.tag_id).where(or_(*child_conds))
+        # tree_number 前缀展开：取匹配 tag 的所有 tree_number，查子节点（[MH:noexp] 时跳过）
+        if not noexp:
+            tns = (await db.execute(
+                select(GlobalTagTreeNumber.tree_number).where(
+                    GlobalTagTreeNumber.tag_id.in_(list(mesh_tag_ids))
+                ).distinct()
             )).scalars().all()
-            mesh_tag_ids.update(children)
+
+            if tns:
+                child_conds = [GlobalTagTreeNumber.tree_number.like(f"{tn}%") for tn in tns]
+                children = (await db.execute(
+                    select(GlobalTagTreeNumber.tag_id).where(or_(*child_conds))
+                )).scalars().all()
+                mesh_tag_ids.update(children)
 
         uids = list(mesh_tag_ids)
         if major_only:

backend/tests/test_pubmed_search_integration.py

@@ -301,9 +301,10 @@ class TestAllFrontendQueryFormats:
 
     def test_field_map_matches_search_engine(self):
         """Verify field tag map covers all search engine field types."""
-        from app.services.pubmed_query_parser import _FIELD_TAG_MAP
+        from app.services.pubmed_query_parser import _FIELD_TAG_MAP, _SPECIAL_FIELDS
         supported_fields = {"title", "abstract", "all", "author", "journal", "affiliation",
                            "language", "volume", "issue", "pages", "lid"}
+        supported_fields.update(_SPECIAL_FIELDS)  # 特殊字段由 _single_term_condition 处理
         for fld in set(_FIELD_TAG_MAP.values()):
             assert fld in supported_fields, f"Mapped field '{fld}' not supported"
 

frontend/src/components/literature/LiteratureCard.vue

@@ -111,8 +111,11 @@ const nctId = computed(() => {
 // 从查询中提取纯文本词（去掉 PubMed 字段标签如 [TI]、[AB] 等）
 function extractPlainText(q: string): string {
   return q
-    .replace(/\[[\w-]+\]/g, '')
-    .replace(/"?\b(AND|OR|NOT)\b"?/gi, '')
+    .replace(/\[[\w-]+\]/g, '')           // [field tags]
+    .replace(/"?\b(AND|OR|NOT)\b"?/gi, '')// boolean operators
+    .replace(/[()]/g, '')                 // P3-4: 去掉括号
+    .replace(/#\d+/g, '')                 // P3-4: 去掉 #N 引用标记
+    .replace(/\*/g, '')                   // P3-4: 去掉通配符
     .replace(/"/g, '')
     .replace(/\s+/g, ' ')
     .trim()
@@ -127,8 +130,9 @@ const highlightedTitle = computed(() => {
   const terms = plain.split(/\s+/).filter(t => t.length > 0)
   if (terms.length === 0) return ''
   const escaped = terms.map(t => t.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'))
-  // 单字符词添加 \b 词边界避免过匹配（"A" 不匹配每个字母 "a"）
-  const pattern = escaped.map(e => e.length === 1 ? `\\b${e}\\b` : e).join('|')
+  // P3-3: 单字符词添加 \b 词边界，但 CJK 字符跳过（\b 对 CJK 无效）
+  const isCJK = (c: string) => /[一-鿿㐀-䶿豈-﫿]/.test(c)
+  const pattern = escaped.map(e => e.length === 1 && !isCJK(e) ? `\\b${e}\\b` : e).join('|')
   const parts = title.split(new RegExp(`(${pattern})`, 'gi'))
   return parts.map((part: string) =>
     terms.some(t => part.toLowerCase() === t.toLowerCase())

frontend/src/views/app/SearchView.vue

@@ -277,7 +277,7 @@ const { page, total, goToPage } = usePagination({
         query: query.value, page: p, page_size: pageSize.value, sort: sort.value,
       }
       if (field.value !== 'all') body.field = field.value
-      if (precisionMode.value !== 'majr') body.precision_mode = precisionMode.value
+      // P3-6: precision_mode 不再发送（后端已忽略）
       // ── Keyset 游标分页（sort=date 时跳过 COUNT，纯翻页模式） ──
       const useKeyset = sort.value === 'date'
       if (useKeyset) {
@@ -371,10 +371,11 @@ function restoreFromQuery() {
     datePreset.value = String(route.query.date_preset)
   } else if (route.query.date_from) {
     datePreset.value = null
-    // 尝试解析 date_from/date_to 到 yearFromStr/yearToStr
     const df = String(route.query.date_from)
     if (df.length >= 4) yearFromStr.value = df.slice(0, 4)
-  } else if (route.query.date_to) {
+  }
+  if (route.query.date_to) {
+    datePreset.value = null
     const dt = String(route.query.date_to)
     if (dt.length >= 4) yearToStr.value = dt.slice(0, 4)
   }

frontend/src/views/public/HomeView.vue

@@ -214,7 +214,7 @@ function doLocalSearch() {
   if (searchParams.value.date_from) query.date_from = searchParams.value.date_from
   if (searchParams.value.date_to) query.date_to = searchParams.value.date_to
   if (searchParams.value.sort && searchParams.value.sort !== 'date') query.sort = searchParams.value.sort
-  if (searchParams.value.precision_mode && searchParams.value.precision_mode !== 'majr') query.precision = searchParams.value.precision_mode
+  // P3-6: precision_mode 不再发送
   if (searchParams.value.retracted) query.retracted = searchParams.value.retracted
   if (searchParams.value.negative_result) query.negative = searchParams.value.negative_result
   if (selectedTagIds.value.length > 0) query.tag = selectedTagIds.value.join(',')
@@ -319,7 +319,7 @@ async function loadMore() {
   if (loadingMore.value || !hasMore.value) return
   loadingMore.value = true
   try {
-    const body: Record<string, any> = { page_size: searchParams.value.page_size, sort: 'date' }
+    const body: Record<string, any> = { page_size: searchParams.value.page_size, sort: searchParams.value.sort || 'date' }
     if (searchParams.value.query) body.query = searchParams.value.query
     if (searchParams.value.field !== 'all') body.field = searchParams.value.field
     if (searchParams.value.tag_ids.length) body.tag_ids = searchParams.value.tag_ids
@@ -327,7 +327,7 @@ async function loadMore() {
     if (searchParams.value.date_to) body.date_to = searchParams.value.date_to
     if (searchParams.value.retracted) body.retracted = searchParams.value.retracted
     if (searchParams.value.negative_result) body.negative_result = searchParams.value.negative_result
-    if (searchParams.value.precision_mode) body.precision_mode = searchParams.value.precision_mode
+    // P3-6: precision_mode 不再发送
     if (cursorDate.value && cursorId.value) {
       body.cursor_date = cursorDate.value
       body.cursor_id = cursorId.value