如何将文件中的数组字符串转换为3D数组？

我从文件里读了一段数组格式的字符串，想转成3D数组，但试了各种方法都不行。我用的代码如下：

with open(output_dir/f"{input_reg}.txt", "r", encoding='utf-8') as file1:
    Fa = file1.read()
    Fa = np.frombuffer(Fa, dtype=float)
    Fd = json.loads(Fa)

待转换的数组字符串示例：

[[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]][[[0.43096440627115085, -0.8577107283236622, -0.28036045913403573], [-0.4082482904638634, 0.09175170953613587, -0.9082482904638631], [0.8047378541243648, 0.5058793634016807, -0.31061721752604615]], [131.17115021913514, 164.44788458945936, -1692.8571428571427]]

问题原因 #

1. np.frombuffer用法错误：这个函数专门解析二进制字节流，不能直接处理文本字符串，用它处理读取到的内容会完全打乱数据格式。
2. 字符串格式不合法：原始字符串是多个独立的JSON数组直接拼接（比如[[...]]后直接跟[[...]]），没有分隔符也没有外层数组包裹，不符合JSON规范，json.loads自然无法解析。

解决方法 #

先修复字符串格式，转成合法的JSON数组，再解析转换：

import json
import numpy as np
import re
from pathlib import Path

# 替换成你的实际路径和文件名
output_dir = Path("your_output_directory")
input_reg = "your_file_name"

with open(output_dir / f"{input_reg}.txt", "r", encoding='utf-8') as file1:
    raw_str = file1.read()
    # 给相邻数组之间添加逗号分隔符
    fixed_str = re.sub(r']]\[\[', ']], [[', raw_str)
    # 给整体加上外层数组括号，变成合法JSON格式
    fixed_str = f"[{fixed_str}]"
    # 解析为Python列表
    data_list = json.loads(fixed_str)
    # 转换为3D numpy数组
    final_array = np.array(data_list)
    # 查看数组形状，确认结构正确
    print(final_array.shape)

这段代码会将原始字符串转换成形状为(8, 2, ...)的3D数组，每个元素包含一个3×3的子数组和一个3元素的子数组，和原始数据结构完全对应。

内容的提问来源于stack exchange，提问作者Aloma85