1

Author: Notify-ctrl

README.md

@@ -18,5 +18,5 @@ $ uv run main.py <需要转换的lua文件的完整路径>
 （仅在Linux下测试通过）之后Lua文件所在的同一级目录下会出现与lua文件同名文件夹。
 里面的文件就是按照新版本重构过的代码。
 
-> **Warning**
+> [!WARNING] 
 > AI生成的内容并不一定靠谱！请自己再另外进行测试

main.py

@@ -10,9 +10,6 @@ def main():
 
     path = args[1]
     pkg = FkPackage(path)
-    obj = pkg.skills['jilei']
-    print(obj['content'])
-    print(obj['translations'])
     pkg.mkSkillDir()
 
 if __name__ == "__main__":

src/llm.py

@@ -1,19 +1,144 @@
 import os
 import re
+from openai import OpenAI
+
+class Skill:
+    name: str
+    contents = []
+    translations = []
+    def __init__(self, name, contents, trans):
+        self.name = name
+        self.contents = contents
+        self.translations = trans
+
+class GPT:
+    basic_prompt: str # 用来指示技能基本架构如何进行改变的prompt
+    misc_prompt: str
+    askto_prompt: str # 用来杀askFor系列的prompt
+    model = "qwen2.5:32b_72Kctx" # 用ollama创建的context window增大版的model
+
+    def __init__(self):
+        dir_path = os.path.dirname(os.path.abspath(__file__))
+        base_path = os.path.join(dir_path, "prompts")
+        with open(os.path.join(base_path, "01-basic.md")) as f:
+            self.basic_prompt = f.read()
+        with open(os.path.join(base_path, "02-misc.md")) as f:
+            self.misc_prompt = f.read()
+        with open(os.path.join(base_path, "03-askto.md")) as f:
+            self.askto_prompt = f.read()
+        
+        # 根据自己需求修改key
+        # 不过如果用的是公开厂商的API key，那么绝对不要公布出来！
+        # 这里是我本地部署的ai所以我就对key随便处理了
+        api_end_point = "http://localhost:8080/api"
+        api_key = "sk-cb7ee040c840447ca35cc6e63c25bebf"
+        self.client = OpenAI(api_key=api_key, base_url=api_end_point)
+
+    def chat(self, sys_prompt, skill) -> str:
+        return self.client.chat.completions.create(
+            model=self.model,
+            messages=[
+                {"role": "system", "content": sys_prompt},
+                {"role": "user", "content": skill},
+            ],
+            stream=False
+        ).choices[0].message.content
 
 # 先就解决一下提取技能和提取武将的问题
 class FkPackage:
-    skills = {}
+    skills: dict[str, Skill] = {}
     generals = []
     translations = {}
 
     def __init__(self, filepath: str):
+        self.gpt = GPT()
         self.filepath = filepath
         with open(filepath, 'r', encoding='utf-8') as f:
             self.file_content = f.read()
         self._extract_translations()
         self._extract_skills()
-        self._extract_generals()
+
+    '''
+    直接检索文件中的Fk:loadTranslationTable { ... }
+    花括号中的内容必定是形如 ["xxx"] = "xxx2", 之类的表形式
+    并且必定不产生嵌套 只有string-string的键值对
+    可能利用了lua的字符串拼接符号..进行多行字符串，亦可能使用单引号
+    将提取到的翻译表内容翻译成python字典语法，然后直接读取该字典
+    '''
+    def _extract_translations(self):
+        self.translations = {}
+        # 查找Fk:loadTranslationTable的位置
+        start_pattern = re.compile(r'Fk:loadTranslationTable\s*{', re.DOTALL)
+        for match in start_pattern.finditer(self.file_content):
+            start_pos = match.end()
+
+            # 提取直到匹配的闭合花括号的内容
+            brace_count = 1
+            current_pos = start_pos
+            content = []
+            while current_pos < len(self.file_content) and brace_count > 0:
+                char = self.file_content[current_pos]
+                if char == '{':
+                    brace_count += 1
+                elif char == '}':
+                    brace_count -= 1
+                if brace_count > 0:
+                    content.append(char)
+                current_pos += 1
+            table_content = ''.join(content)
+
+            # 去除注释
+            cleaned_content = re.sub(r'--.*', '', table_content, flags=re.MULTILINE)
+
+            # 分割键值对
+            entries = re.split(r',\s*', cleaned_content)
+
+            for entry in entries:
+                entry = entry.strip()
+                if not entry:
+                    continue
+
+                # 解析键
+                key_match = re.match(r'\s*\[\s*(["\'])(.*?)\1\s*\]\s*=\s*(.*)', entry, re.DOTALL)
+                if not key_match:
+                    continue
+                quote_type, key_content, value_expr = key_match.groups()
+
+                # 处理键的转义
+                key = self._unescape_lua_string(key_content)
+
+                # 解析值中的字符串
+                value = self._parse_lua_value(value_expr)
+
+                if key and value is not None:
+                    self.translations[key] = value
+
+    def _unescape_lua_string(self, s):
+        # 处理Lua字符串中的转义字符
+        replacements = {
+            '\\\\': '\\',
+            '\\"': '"',
+            "\\'": "'",
+            '\\n': '\n',
+            '\\t': '\t',
+            '\\r': '\r',
+            '\\b': '\b',
+            '\\f': '\f',
+            '\\a': '\a',
+            '\\v': '\v'
+        }
+        for escaped, unescaped in replacements.items():
+            s = s.replace(escaped, unescaped)
+        return s
+
+    def _parse_lua_value(self, expr):
+        # 解析Lua字符串拼接表达式
+        str_pattern = re.compile(r'''["']((?:[^\\"']|\\.)*)["']''', re.DOTALL)
+        parts = []
+        for match in str_pattern.finditer(expr):
+            str_content = match.group(1)
+            parts.append(self._unescape_lua_string(str_content))
+        return ''.join(parts) if parts else None
 
     '''
     把东西作为字典放入自己的skills里面
@@ -34,14 +159,6 @@ def __init__(self, filepath: str):
 
     最后加入数组
     '''
-    def _extract_translations(self):
-        # 提取所有翻译项
-        trans_pattern = re.compile(r'\[["\'](.*?)["\']\]\s*=\s*["\'](.*?)["\']', re.DOTALL)
-        self.translations = {}
-        for match in trans_pattern.finditer(self.file_content):
-            key, value = match.groups()
-            self.translations[key] = value
-
     def _extract_skills(self):
         self.skills = {}
         skill_pattern = re.compile(r'local\s+(\w+)\s*=\s*fk\.Create\w+Skill\s*\{', re.IGNORECASE)
@@ -58,10 +175,8 @@ def _extract_skills(self):
                     level -= 1
                 end_pos += 1
             content = self.file_content[start_pos:end_pos]  # 提取完整技能内容
-            self.skills[skill_name] = ({
-                'content': content,
-                'translations': [],
-            })
+            skill_str_name = re.compile(r'name = ["\'](.*?)["\'],').findall(content)[0]
+            self.skills[skill_name] = Skill(skill_str_name, [content], [])
 
         related_pattern = re.compile(r'(\w+):addRelatedSkill\(\s*(\w+)\s*\)')
         for match in related_pattern.finditer(self.file_content):
@@ -71,67 +186,29 @@ def _extract_skills(self):
                 continue
             main_skill_obj = self.skills[main_skill]
             sub_skill_obj = self.skills[sub_skill]
-            main_skill_obj['content'] += "\n\n"
-            main_skill_obj['content'] += sub_skill_obj['content']
-            main_skill_obj['content'] += "\n"
-            main_skill_obj['content'] += match.group()
+            main_skill_obj.contents.append(sub_skill_obj.contents[0])
             del self.skills[sub_skill]
 
         for k in self.skills:
             sk = self.skills[k]
-            content = sk['content']
-            strings = re.findall(r'["\'](.*?)["\']', content, re.DOTALL)
+            content = '\n'.join(sk.contents)
+            strings: list[str] = re.findall(r'["\'](.*?)["\']', content, re.DOTALL)
+            for i in range(0, len(strings)):
+                if strings[i].endswith(':'):
+                    strings[i] = strings[i].rstrip(':')
+            strings.append(sk.name)
+            strings.append(":" + sk.name)
+            strings.append("$" + sk.name)
+            for i in range(1, 9):
+                strings.append("$" + sk.name + str(i))
             # 收集翻译并去重
             translations = []
             seen = set()
-            for s in set(strings):
+            for s in strings:
                 if s in self.translations and s not in seen:
                     seen.add(s)
                     translations.append({'key': s, 'value': self.translations[s]})
-            sk['translations'] = translations
-
-    def _extract_generals(self):
-        self.generals = []
-        general_pattern = re.compile(r'local\s+(\w+)\s*=\s*General(?::new)?\s*\(', re.IGNORECASE)
-        for match in general_pattern.finditer(self.file_content):
-            general_name = match.group(1)
-            # 定位参数括号位置
-            general_match = match.group(0)
-            paren_pos = general_match.rfind('(')
-            start_paren_pos = match.start() + paren_pos
-            level = 1
-            end_paren_pos = start_paren_pos + 1
-            while end_paren_pos < len(self.file_content) and level > 0:
-                char = self.file_content[end_paren_pos]
-                if char == '(':
-                    level += 1
-                elif char == ')':
-                    level -= 1
-                end_paren_pos += 1
-            creation_content = self.file_content[match.start():end_paren_pos]
-            # 收集关联调用
-            related_lines = []
-            line_pattern = re.compile(r'^\s*{}\b\s*[:.]'.format(re.escape(general_name)), re.MULTILINE)
-            for line_match in line_pattern.finditer(self.file_content, end_paren_pos):
-                line_start = line_match.start()
-                line_end = self.file_content.find('\n', line_start)
-                line = self.file_content[line_start:line_end if line_end != -1 else None].strip()
-                related_lines.append(line)
-            # 合并内容
-            content = '\n'.join([creation_content] + related_lines)
-            # 处理翻译
-            strings = re.findall(r'["\'](.*?)["\']', content, re.DOTALL)
-            translations = []
-            seen = set()
-            for s in set(strings):
-                if s in self.translations and s not in seen:
-                    seen.add(s)
-                    translations.append({'key': s, 'value': self.translations[s]})
-            self.generals.append({
-                'name': general_name,
-                'content': content,
-                'translations': translations
-            })
+            sk.translations = translations
 
     def mkSkillDir(self):
         """创建技能文件目录结构并写入技能内容"""
@@ -143,15 +220,41 @@ def mkSkillDir(self):
         target_dir = os.path.join(dir_path, "pkg", base_name, "skills")
         os.makedirs(target_dir, exist_ok=True)
 
+        i = 0
+        n = len(self.skills.keys())
+
         # 写入技能文件
         for skill_name in self.skills:
+            i += 1
+            print("[*] 正在处理第 %d/%d 个技能：%s" % (i, n, skill_name))
             skill = self.skills[skill_name]
             # 清理非法文件名字符
             safe_name = re.sub(r'[\\/*?:"<>|]', '_', skill_name)
             file_path = os.path.join(target_dir, f"{safe_name}.lua")
 
             try:
                 with open(file_path, 'w', encoding='utf-8') as f:
-                    f.write(skill['content'])
+                    f.write(self._refactor_skill(skill))
+
             except Exception as e:
                 print(f"写入技能文件失败: {file_path} | 错误: {str(e)}")
+
+    def _refactor_skill(self, skillObj: Skill) -> str:
+        content = '\n\n'.join(skillObj.contents)
+        content += "Fk:loadTranslationTable{\n"
+        for kv in skillObj.translations:
+            content += "  [%s] = %s,\n" % (repr(kv['key']), repr(kv['value']))
+        content += "}\n"
+        gpt = self.gpt
+        n = 2
+
+        print ("  [1/%d] 将技能主体结构调整为新版" % n)
+        content = gpt.chat(gpt.basic_prompt, content)
+        print ("  [2/%d] 调整self.name，cost_data等" % n)
+        content = gpt.chat(gpt.misc_prompt, content)
+        print ("  [3/%d] 重构askFor系列函数" % n)
+        content = gpt.chat(gpt.askto_prompt, content)
+
+        content = content.replace('    ', '  ')
+
+        return content