Download Dataset_Construction/extract_function.py from SYSUSELab/RustRepoTrans: direct link, hf CLI and curl.
- Browser
- Download file 6.5 kB
-
https://huggingface.co/datasets/SYSUSELab/RustRepoTrans/resolve/main/Dataset_Construction/extract_function.py
- Command line
-
hf download hf://datasets/SYSUSELab/RustRepoTrans/Dataset_Construction/extract_function.py
-
curl -L -o extract_function.py https://huggingface.co/datasets/SYSUSELab/RustRepoTrans/resolve/main/Dataset_Construction/extract_function.py
6.5 kB
| import re | |
| import os | |
| import sys | |
| total_functions = set() | |
| def extract_functions_from_code(code, pattern): | |
| function_head_pattern = re.compile(pattern) | |
| lines = code.split('\n') | |
| functions = [] | |
| brace_count = 0 | |
| function_code = [] | |
| inside_function = False | |
| key = False | |
| for i, line in enumerate(lines): | |
| if not inside_function and function_head_pattern.search(line): | |
| inside_function = True | |
| if inside_function: | |
| function_code.append(line) | |
| if not line.lstrip(" ").startswith("//"): | |
| brace_count += line.count('{') | |
| brace_count -= line.count('}') | |
| if brace_count == 0 : | |
| if line.strip().endswith('}'): | |
| inside_function = False | |
| functions.append('\n'.join(function_code)) | |
| function_code = [] | |
| elif line.strip().endswith(';'): | |
| inside_function = False | |
| function_code = [] | |
| return functions | |
| def extract_functions_from_code_py(code): | |
| lines = code.split('\n') | |
| functions = [] | |
| function_code = [] | |
| inside_function = False | |
| for line in lines: | |
| if not inside_function and line.lstrip().startswith("def "): | |
| # 获取函数起始的缩进长度 | |
| pre_cnt = len(line) - len(line.lstrip()) | |
| function_code.append(line) | |
| inside_function = True | |
| # 跳过def那一行 | |
| continue | |
| if inside_function: | |
| # 空行和缩进比def要小表示还在函数内 | |
| if len(line) == 0 or len(line) - len(line.lstrip()) >= pre_cnt + 4: | |
| function_code.append(line) | |
| else: | |
| functions.append('\n'.join(function_code)) | |
| function_code = [] | |
| # 当前行有可能是下一个函数的声明行,不处理会跳过该函数 | |
| if line.lstrip().startswith("def "): | |
| pre_cnt = len(line) - len(line.lstrip()) | |
| function_code.append(line) | |
| else: | |
| inside_function = False | |
| # 处理在文件末尾声明定义的function | |
| if function_code: | |
| functions.append('\n'.join(function_code)) | |
| return functions | |
| def extract_functions_from_code_rb(code): | |
| lines = code.split('\n') | |
| functions = [] | |
| function_code = [] | |
| inside_function = False | |
| for line in lines: | |
| if not inside_function and line.lstrip().startswith("def "): | |
| # 获取函数起始的缩进长度 | |
| pre_cnt = len(line) - len(line.lstrip()) | |
| function_code.append(line) | |
| inside_function = True | |
| # 跳过def那一行 | |
| continue | |
| if inside_function: | |
| if len(line) - len(line.lstrip()) == pre_cnt and line.lstrip().startswith("end"): | |
| inside_function = False | |
| functions.append('\n'.join(function_code)) | |
| function_code = [] | |
| else: | |
| function_code.append(line) | |
| return functions | |
| def save_functions_to_files(functions, output_dir, output_file_name): | |
| if not os.path.exists(output_dir): | |
| os.makedirs(output_dir) | |
| try: | |
| for i, func in enumerate(functions): | |
| # influxdb-1.8\\client\\influxdb_test.go -> influxdb-1.8__client__influxdb_test | |
| output_file = os.path.splitext(output_file_name) | |
| output_file = output_file[0].replace("/", "__") + "__" + output_file[1] | |
| file_path = os.path.join(output_dir, f'{output_file}__function__{i + 1}.txt') | |
| with open(file_path, 'w', encoding='utf-8') as file: | |
| file.write(f"<path>\n{output_file_name}\n</path>\n") | |
| file.write(f"<function>\n{func}\n</function>") | |
| except Exception as e: | |
| print(e) | |
| pass | |
| def process_file(input_file, lang, output_dir, pattern): | |
| with open(input_file, 'r', encoding='utf-8', errors='ignore') as file: | |
| code = file.read() | |
| if lang == "py": | |
| functions = extract_functions_from_code_py(code) | |
| elif lang == "rb": | |
| functions = extract_functions_from_code_rb(code) | |
| else: | |
| functions = extract_functions_from_code(code, pattern) | |
| save_functions_to_files(functions, output_dir, input_file) | |
| def main(): | |
| project_dir = "projects" | |
| target_project = sys.argv[1] | |
| patterns = { | |
| 'cpp': r'^\s*[\w\s\*\[\]\<\>\:]+\s+[\w\s\*\[\]\<\>\:]+\s*\(', | |
| 'cxx': r'^\s*[\w\s\*\[\]\<\>\:]+\s+[\w\s\*\[\]\<\>\:]+\s*\(', | |
| 'h': r'^\s*[\w\s\*\[\]\<\>\:]+\s+[\w\s\*\[\]\<\>\:]+\s*\(', | |
| 'java': r'^\s*(public|protected|private|static|final|synchronized|native|abstract|strictfp|default)?\s*(public|protected|private|static|final|synchronized|native|abstract|strictfp|default)?\s*[\w\<\>\[\] ]+\s+[\w\<\>\[\]]+\s*\(', | |
| 'rs': r'^\s*(unsafe)?\s*(pub(\(crate\))?)?\s*(async)?\s*fn\s', | |
| 'c': r'^\s*[\w\s\*\[\]]*\s*\w+\s*\(', | |
| 'py': r'' | |
| } | |
| lang_to_fileType = { | |
| 'cpp' : ['cpp', 'cxx', 'h'], | |
| 'c' : ['c', 'h'], | |
| 'java' : ['java'], | |
| 'rust' : ['rs'], | |
| 'python' : ['py'] | |
| } | |
| projects = os.listdir(project_dir) | |
| for project in projects: | |
| if project != target_project: | |
| continue | |
| project_pair_path = os.path.join(project_dir, project) | |
| langs = os.listdir(project_pair_path) | |
| for lang in langs: | |
| root_dir = os.path.join(project_pair_path, lang) | |
| # 对项目进行遍历 | |
| for current_path, dirs, files in os.walk(root_dir): | |
| dirs[:] = [d for d in dirs if not d.startswith('.')] | |
| for file in files: | |
| try : | |
| file_lang = file.split('.')[-1] | |
| except: | |
| continue | |
| if file_lang in lang_to_fileType[lang]: | |
| file_path = os.path.join(current_path, file) | |
| if "test" in file_path or "Test" in file_path: | |
| continue | |
| process_file(file_path, file_lang, root_dir.replace("projects", "functions"), patterns[file_lang]) | |
| if __name__ == '__main__': | |
| main() | |