新版本numpy.subtract产生的问题及解决方法

报错信息：ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (5,) + inhomogeneous part.

问题分析：

产生这个错误的原因目前猜测是因为新版本的numpy（我的是1.25.2）再subtract时对list进行了一些保护，可以尝试降版本到1.21.6来解决问题，也可以通过下列方式来避免问题。

import numpy as np

original_list = [3, 5, 7, np.array([9, 2, 4]), np.array([6, 8, 1])]
c = np.subtract(original_list,1)
print(c)

执行上述代码就会触发下列错误：

Traceback (most recent call last):
  File "path", line XX, in <module>
    c = np.subtract(original_list,1)
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (5,) + inhomogeneous part.

解决方法：

思路：复制一个元素全为1的列表，再逐位相减

import numpy as np

def copy_list(source_list):
    if isinstance(source_list, list):
        new_list = []
        for item in source_list:
            new_list.append(copy_list(item))
        return new_list
    else:
        return 1

original_list = [3, 5, 7, np.array([9, 2, 4]), np.array([6, 8, 1])]
new_list = copy_list(original_list)
c = [original_list[i]-new_list[i] for i in range(0,len(original_list))]

print(new_list)
print(c)

结果正确输出：

[1, 1, 1, 1, 1]
[2, 4, 6, array([8, 1, 3]), array([5, 7, 0])]