autocomplete.py:

添加搜索联想功能模块，但未加入search接口
all_kana.py:
将日语单词转换功能单独设立为功能模块

app/api/search.py

@@ -7,45 +7,13 @@ from fastapi import APIRouter, Depends, HTTPException, Request
 from app.models import DefinitionJp
 from app.models.fr import DefinitionFr
 from app.schemas.search_schemas import SearchRequest, SearchResponse, SearchItemFr, SearchItemJp
+from app.utils.all_kana import all_in_kana
 from app.utils.security import get_current_user
 from app.utils.textnorm import normalize_text
 from scripts.update_jp import normalize_jp_text
 
 dict_search = APIRouter()
 
-kks = pykakasi.kakasi()
-kks.setMode("H", "a")  # 平假名 -> ascii (罗马字)
-kks.setMode("K", "a")  # 片假名 -> ascii
-kks.setMode("J", "a")  # 汉字   -> ascii
-kks.setMode("r", "Hepburn")  # 转换成 Hepburn 罗马字
-conv = kks.getConverter()
-
-
-def all_in_kana(text: str) -> str:
-    """
-    将输入统一转换为平假名，支持：
-    - 平假名
-    - 片假名
-    - 罗马字 (Hepburn 转写)
-
-    返回：平假名字符串
-    """
-    if not text:
-        return ""
-
-    # 1. 片假名 → 平假名
-    normalized = jaconv.kata2hira(text)
-
-    # 2. 如果里面含有罗马字字符，就先转成假名
-    if any("a" <= ch.lower() <= "z" for ch in normalized):
-        hira = conv.do(normalized)  # 罗马字 -> 平假名
-        normalized = jaconv.kata2hira(hira)
-
-    # 3. 再次片假名 -> 平假名保险
-    normalized = jaconv.kata2hira(normalized)
-
-    return normalized
-
 
 @dict_search.post("/search", response_model=SearchResponse)
 async def search(request: Request, body: SearchRequest, user=Depends(get_current_user)):
@@ -107,5 +75,13 @@ async def search(request: Request, body: SearchRequest, user=Depends(get_current
             contents=contents,
         )
 
+
 # TODO 相关度排序（转换为模糊匹配）
 # TODO 输入搜索框时反馈内容
+
+# @dict_search.post("search/list")
+# async def search_list(body: SearchRequest, user=Depends(get_current_user)):
+#     query = body.query
+#     if body.language == 'fr':
+#         query = normalize_text(query)
+#         prefix = await DefinitionFr.filter(word__text__icontains=query)

app/utils/all_kana.py

@@ -0,0 +1,35 @@
+import jaconv
+import pykakasi
+
+kks = pykakasi.kakasi()
+kks.setMode("H", "a")  # 平假名 -> ascii (罗马字)
+kks.setMode("K", "a")  # 片假名 -> ascii
+kks.setMode("J", "a")  # 汉字   -> ascii
+kks.setMode("r", "Hepburn")  # 转换成 Hepburn 罗马字
+conv = kks.getConverter()
+
+
+def all_in_kana(text: str) -> str:
+    """
+    将输入统一转换为平假名，支持：
+    - 平假名
+    - 片假名
+    - 罗马字 (Hepburn 转写)
+
+    返回：平假名字符串
+    """
+    if not text:
+        return ""
+
+    # 1. 片假名 → 平假名
+    normalized = jaconv.kata2hira(text)
+
+    # 2. 如果里面含有罗马字字符，就先转成假名
+    if any("a" <= ch.lower() <= "z" for ch in normalized):
+        hira = conv.do(normalized)  # 罗马字 -> 平假名
+        normalized = jaconv.kata2hira(hira)
+
+    # 3. 再次片假名 -> 平假名保险
+    normalized = jaconv.kata2hira(normalized)
+
+    return normalized

app/utils/autocomplete.py

@@ -0,0 +1,91 @@
+import asyncio
+
+from tortoise import Tortoise
+from tortoise.expressions import Q
+from typing import List, Literal, Tuple
+
+from app.models import WordlistFr, WordlistJp
+from app.schemas.search_schemas import SearchRequest
+from app.utils.all_kana import all_in_kana
+from app.utils.textnorm import normalize_text
+from settings import TORTOISE_ORM
+
+
+async def suggest_autocomplete(query: SearchRequest, limit: int = 10) -> List[str]:
+    """
+
+    :param query: 当前用户输入的内容
+    :param limit: 返回列表限制长度
+    :return: 联想的单词列表（非完整信息，单纯单词）
+    """
+    if query.language == 'fr':
+        query_word = normalize_text(query.query)
+        qs_prefix = await (
+            WordlistFr
+            .filter(search_text__startswith=query_word)
+            .only("text", "freq")
+        )
+        prefix_objs = qs_prefix[:limit]
+        prefix: List[Tuple[str, int]] = [(o.text, o.freq) for o in prefix_objs]
+
+        need = max(0, limit - len(prefix))
+        contains: List[Tuple[str, int]] = []
+
+        if need > 0:
+            qs_contain = await (
+                WordlistFr
+                .filter(search_text__icontains=query_word)
+                .exclude(search_text__startswith=query_word)
+                .only("text", "freq")
+            )
+            contains_objs = qs_contain[:need * 2]
+            contains: List[Tuple[str, int]] = [(o.text, o.freq) for o in contains_objs]
+
+    else:
+        query_word = all_in_kana(query.query)
+
+        qs_prefix = await (
+            WordlistJp
+            .filter(hiragana__startswith=query_word)
+            .only("text", "freq")
+        )
+
+        prefix_objs = qs_prefix[:limit]
+        prefix: List[Tuple[str, int]] = [(o.text, o.freq) for o in prefix_objs]
+
+        need = max(0, limit - len(prefix))
+        contains = []
+        if need > 0:
+            qs_contain = await (
+                WordlistJp
+                .filter(Q(hiragana__icontains=query_word) | Q(text__icontains=query_word))
+                .exclude(Q(hiragana__startswith=query_word) | Q(text__startswith=query_word))
+                .only("text", "freq")
+            )
+            contains_objs = qs_contain[:need * 2]
+            contains: List[Tuple[str, int]] = [(o.text, o.freq) for o in contains_objs]
+
+    seen_text, out = set(), []
+    for word in list(qs_prefix) + list(contains):
+        if word.text not in seen_text:
+            seen_text.add(word.text)
+            out.append((word.text, word.freq))
+        if len(out) >= limit:
+            break
+    out = sorted(out, key=lambda w: (w[1], len(w[0]), w[0]))
+    return [text for text, _ in out]
+
+
+async def __test():
+    query_word: str = 'あい'
+    language: Literal['fr', 'jp'] = 'jp'
+    return await suggest_autocomplete(SearchRequest(query=query_word, language=language))
+
+
+async def __main():
+    await Tortoise.init(config=TORTOISE_ORM)
+    print(await __test())
+
+
+if __name__ == '__main__':
+    asyncio.run(__main())

requirements.txt

@@ -1,3 +1,4 @@
+jaconv
 fugashi
 pykakasi
 aerich==0.9.1