🐛 fix(generator): 规范化非法 Java 标识符

src/dbjavagenix/ai/naming_rules.py

@@ -20,10 +20,19 @@
 from dataclasses import dataclass
 from typing import Any, Dict, List, Optional, Tuple
 
+from ..core.java_identifiers import to_pascal_case
+
 
 COMMON_PREFIXES = (
-    "sys_", "biz_", "bd_", "cms_", "ops_", "admin_", "core_",
-    "t_", "tb_",  # 通用建表习惯
+    "sys_",
+    "biz_",
+    "bd_",
+    "cms_",
+    "ops_",
+    "admin_",
+    "core_",
+    "t_",
+    "tb_",  # 通用建表习惯
 )
 
 # 不规则单数化(主要场景,不追求完全覆盖)
@@ -102,7 +111,9 @@ def infer_business_name(
     # 2. 检测关联表: ≥ 2 个外键且其他列很少
     if _is_association_table(columns, foreign_keys):
         # 关联表命名: 用两个外键的 referenced_table 组合
-        ref_tables = [_strip_known_prefix(fk.get("references_table", "")) for fk in foreign_keys[:2]]
+        ref_tables = [
+            _strip_known_prefix(fk.get("references_table", "")) for fk in foreign_keys[:2]
+        ]
         ref_classes = [_to_pascal_case(_singularize(rt)) for rt in ref_tables if rt]
         if len(ref_classes) >= 2:
             class_name = f"{ref_classes[0]}{ref_classes[1]}Assignment"
@@ -178,21 +189,27 @@ def _is_association_table(
     pk_count = sum(1 for c in columns if c.get("primary_key"))
     # 业务列 = 非主键 + 非外键
     business_cols = [
-        c for c in columns
-        if not c.get("primary_key") and c.get("name") not in fk_cols
+        c for c in columns if not c.get("primary_key") and c.get("name") not in fk_cols
     ]
     # 时间戳/审计字段不算业务列
-    business_cols = [
-        c for c in business_cols
-        if not _is_audit_column(c.get("name", ""))
-    ]
+    business_cols = [c for c in business_cols if not _is_audit_column(c.get("name", ""))]
     # 业务列 ≤ 1 (例如 'remark') 视为关联表
     return len(business_cols) <= 1 and pk_count <= 1
 
 
-_AUDIT_PATTERNS = ("created_at", "updated_at", "create_time", "update_time",
-                   "created_by", "updated_by", "deleted_at", "deleted",
-                   "version", "tenant_id", "remark")
+_AUDIT_PATTERNS = (
+    "created_at",
+    "updated_at",
+    "create_time",
+    "update_time",
+    "created_by",
+    "updated_by",
+    "deleted_at",
+    "deleted",
+    "version",
+    "tenant_id",
+    "remark",
+)
 
 
 def _is_audit_column(name: str) -> bool:
@@ -202,7 +219,7 @@ def _is_audit_column(name: str) -> bool:
 def _strip_known_prefix(name: str) -> str:
     for prefix in COMMON_PREFIXES:
         if name.startswith(prefix):
-            return name[len(prefix):]
+            return name[len(prefix) :]
     return name
 
 
@@ -219,7 +236,7 @@ def _singularize(name: str) -> str:
     # ses / xes / ches / shes
     for suf in ("ses", "xes", "ches", "shes"):
         if lower.endswith(suf):
-            return lower[: -2]
+            return lower[:-2]
     # 复数 s 去掉 (排除 ss / us / is 等不应该去的)
     if lower.endswith("s") and not lower.endswith(("ss", "us", "is", "as", "os")):
         return lower[:-1]
@@ -230,18 +247,15 @@ def _singularize(name: str) -> str:
 
 
 def _to_pascal_case(name: str) -> str:
-    parts = [p for p in _CAMEL_SPLIT_RE.split(name) if p]
-    return "".join(p[:1].upper() + p[1:] for p in parts)
+    return to_pascal_case(name)
 
 
 def _tokenize(name: str) -> List[str]:
     """把下划线/连字符分隔的名字拆为 lowercase 词列表"""
     return [p.lower() for p in _CAMEL_SPLIT_RE.split(name) if p]
 
 
-def infer_business_names_batch(
-    tables: List[Dict[str, Any]]
-) -> List[NamingInference]:
+def infer_business_names_batch(tables: List[Dict[str, Any]]) -> List[NamingInference]:
     """对一批表批量推断 (输入格式与 LLM 模式一致)"""
     return [
         infer_business_name(

src/dbjavagenix/core/java_identifiers.py

@@ -0,0 +1,132 @@
+"""Deterministic Java identifier normalisation shared by code-generation paths."""
+
+from __future__ import annotations
+
+import re
+
+
+# Java keywords, literals and contextual keywords that are unsafe as generated names.
+JAVA_KEYWORDS = frozenset(
+    {
+        "abstract",
+        "assert",
+        "boolean",
+        "break",
+        "byte",
+        "case",
+        "catch",
+        "char",
+        "class",
+        "const",
+        "continue",
+        "default",
+        "do",
+        "double",
+        "else",
+        "enum",
+        "extends",
+        "final",
+        "finally",
+        "float",
+        "for",
+        "goto",
+        "if",
+        "implements",
+        "import",
+        "instanceof",
+        "int",
+        "interface",
+        "long",
+        "native",
+        "new",
+        "package",
+        "private",
+        "protected",
+        "public",
+        "return",
+        "short",
+        "static",
+        "strictfp",
+        "super",
+        "switch",
+        "synchronized",
+        "this",
+        "throw",
+        "throws",
+        "transient",
+        "try",
+        "void",
+        "volatile",
+        "while",
+        "true",
+        "false",
+        "null",
+        "_",
+        # Contextual keywords used by modern Java language features.
+        "module",
+        "open",
+        "opens",
+        "provides",
+        "requires",
+        "to",
+        "transitive",
+        "uses",
+        "with",
+        "yield",
+        "record",
+        "sealed",
+        "permits",
+        "non-sealed",
+        "var",
+        "when",
+    }
+)
+
+_CAMEL_BOUNDARY_RE = re.compile(r"(?<=[a-z0-9])(?=[A-Z])")
+
+
+def _words(value: str) -> list[str]:
+    """Split a database name into alphanumeric words while retaining Unicode letters."""
+    if not isinstance(value, str):
+        return []
+
+    # Treat every non-alphanumeric character as a separator. This keeps the original
+    # database identifier untouched while ensuring generated Java names are portable.
+    normalized = "".join(char if (char == "_" or char.isalnum()) else "_" for char in value)
+    words: list[str] = []
+    for part in normalized.split("_"):
+        words.extend(piece for piece in _CAMEL_BOUNDARY_RE.split(part) if piece)
+    return words
+
+
+def _fallback_name(value: str) -> str:
+    return "" if value == "" else "Generated"
+
+
+def to_pascal_case(value: str) -> str:
+    """Convert a database identifier to a valid Java-style PascalCase name."""
+    words = _words(value)
+    if not words:
+        return _fallback_name(value)
+
+    result = "".join(word[:1].upper() + word[1:].lower() for word in words)
+    if result and result[0].isdigit():
+        result = f"Generated{result}"
+    return result or _fallback_name(value)
+
+
+def to_camel_case(value: str) -> str:
+    """Convert a database identifier to a valid Java-style camelCase name."""
+    pascal = to_pascal_case(value)
+    if not pascal:
+        return ""
+
+    result = pascal[:1].lower() + pascal[1:]
+    if result in JAVA_KEYWORDS:
+        result = f"{result}_"
+    return result
+
+
+def is_valid_java_identifier(value: str) -> bool:
+    """Return whether *value* is a Java identifier safe for generated source."""
+    return bool(value) and value != "_" and value.isidentifier() and value not in JAVA_KEYWORDS

src/dbjavagenix/core/models.py

@@ -8,6 +8,8 @@
 from pydantic import BaseModel, Field
 from urllib.parse import quote
 
+from .java_identifiers import to_pascal_case
+
 
 class DatabaseType(str, Enum):
     """Supported database types"""
@@ -88,8 +90,7 @@ def entity_name(self) -> str:
     @staticmethod
     def _to_pascal_case(snake_str: str) -> str:
         """Convert snake_case to PascalCase"""
-        components = snake_str.split("_")
-        return "".join(word.capitalize() for word in components)
+        return to_pascal_case(snake_str)
 
 
 class DatabaseConfig(BaseModel):