process.py

import json

def process_jsonl(input_file, success_file, failure_file):
    success_list = []
    failure_list = []
    with open(input_file, 'r', encoding='utf-8') as infile:
        for line in infile:
            line = line.strip()
            if not line:
                continue
            try:
                data = json.loads(line)
            except json.JSONDecodeError:
                # 如果整行不是有效的JSON，直接加入失败列表
                failure_list.append(line)
                continue
            result_str = data.get('result', '')
            # 去除“```json”开头和“```”结尾
            if result_str.startswith("```json"):
                result_str = result_str[len("```json"):].lstrip()
            if result_str.endswith("```"):
                result_str = result_str[:-len("```")].rstrip()
            # 去除转义字符和换行符
            # 替换转义的双引号和反斜杠
            result_str = result_str.replace('\\"', '"').replace('\\\\', '\\')
            result_str = result_str.replace('\\n', '')
            # 现在尝试解析为JSON
            try:
                result_json = json.loads(result_str)
                # 确保 result_json 是一个列表
                if not isinstance(result_json, list):
                    result_json = [result_json]
                data['result'] = result_json
                success_list.append(data)
            except json.JSONDecodeError as e:
                # 解析失败，将原始的line加入失败列表
                failure_list.append(line)
    # 写入成功数据
    with open(success_file, 'w', encoding='utf-8') as sf:
        for item in success_list:
            sf.write(json.dumps(item, ensure_ascii=False) + '\n')
    # 写入失败数据
    with open(failure_file, 'w', encoding='utf-8') as ff:
        for item in failure_list:
            ff.write(item + '\n')

# 调用函数
process_jsonl('gpt4_query_correlation2.jsonl', 'success.jsonl', 'failure.jsonl')