import json
from PIL import Image
import os


def convert_data():
    try:
        # 读取 JSON 文件
        json_path = '/home1/gmh/InternVL-main/with_llm_confidence.json'
        with open(json_path, 'r', encoding='utf-8') as f:
            data = json.load(f)

        base_dir = '/home1/gmh/profused_benchmark/dataset'  # 基础目录
        converted_data = []
        failed_images = []  # 记录无法加载的图片
        possible_extensions = ['.jpg', '.jpeg', '.png', '.bmp']  # 常见图片后缀

        for index, sample in enumerate(data['data']):
            original_data=sample["origin_data"]
            llm_confidence=original_data['llm_confidence']
            if llm_confidence<0.9:
                continue
            extend_data = sample['extend_data_1']
            image_count = extend_data['image_count']
            result = {
                "id": index
            }

            # 判断 extend_data['image'] 是字符串还是列表
            image_data = extend_data['image']
            if isinstance(image_data, str):
                # 如果是字符串，直接使用
                image_path = image_data
            elif isinstance(image_data, list) and image_data:
                # 如果是列表，取第一个元素
                image_path = image_data[0]
            else:
                # 无效格式，记录错误并使用默认值
                print(f"样本 {index} image 格式错误: {image_data}")
                failed_images.append(str(image_data))
                result["image"] = str(image_data)
                result["width"] = 897
                result["height"] = 1152
                image_tags = "\n".join(["<image>"] * image_count)
                result["conversations"] = [
                    {
                        "from": "human",
                        "value": f"{image_tags}\n{extend_data['question']}"
                    },
                    {
                        "from": "gpt",
                        "value": extend_data['program_answer']
                    }
                ]
                converted_data.append(result)
                continue

            # 赋值 image 路径
            result["image"] = image_path

            if image_count == 1:
                absolute_image_path = os.path.join(base_dir, image_path)
                print(f"样本 {index}, 图片路径: {image_path}, 绝对路径: {absolute_image_path}")

                try:
                    # 检查文件是否存在
                    target_path = absolute_image_path
                    if not os.path.exists(absolute_image_path):
                        # 尝试添加后缀
                        for ext in possible_extensions:
                            temp_path = absolute_image_path + ext
                            if os.path.exists(temp_path):
                                target_path = temp_path
                                print(f"找到文件: {target_path}")
                                break
                        else:
                            print(f"文件不存在: {absolute_image_path} (也尝试了后缀)")
                            failed_images.append(image_path)
                            result["width"] = 897
                            result["height"] = 1152
                            image_tags = "\n".join(["<image>"] * image_count)
                            result["conversations"] = [
                                {
                                    "from": "human",
                                    "value": f"{image_tags}\n{extend_data['question']}"
                                },
                                {
                                    "from": "gpt",
                                    "value": extend_data['program_answer']
                                }
                            ]
                            converted_data.append(result)
                            continue

                    # 使用 PIL 打开图片
                    img = Image.open(target_path)
                    img.verify()  # 验证图片完整性
                    img = Image.open(target_path)  # 重新打开以获取尺寸
                    width, height = img.size
                    print(f"图片尺寸: {width}x{height}")
                    result["width"] = width
                    result["height"] = height
                except Exception as e:
                    print(f"样本 {index} 读取图片 {image_path} 失败: {e}")
                    failed_images.append(image_path)
                    result["width"] = 897
                    result["height"] = 1152
            else:
                result["image"] = image_data  # 保持原始数据（列表）
                result["width_list"] = [240] * image_count
                result["height_list"] = [23] * image_count

            image_tags = "\n".join(["<image>"] * image_count)
            result["conversations"] = [
                {
                    "from": "human",
                    "value": f"{image_tags}\n{extend_data['question']}"
                },
                {
                    "from": "gpt",
                    "value": extend_data['program_answer']
                }
            ]
            converted_data.append(result)

        if failed_images:
            print(f"以下图片加载失败: {failed_images}")
        return converted_data

    except FileNotFoundError:
        print(f"错误: 未找到 JSON 文件 {json_path}")
    except json.JSONDecodeError:
        print("错误: JSON 文件解析失败")
    except Exception as e:
        print(f"发生未知错误: {e}")
    return []


if __name__ == "__main__":
    converted = convert_data()
    if converted:
        output_file_path = '/home1/gmh/InternVL-main/aaa1.jsonl'  # 修改文件后缀为.jsonl
        with open(output_file_path, 'w', encoding='utf-8') as output_file:
            for item in converted:
                output_file.write(json.dumps(item, ensure_ascii=False) + '\n')  # 逐行写入 JSON 对象
        print(f"转换完成，输出保存至: {output_file_path}")
    else:
        print("转换失败，未生成输出文件")