经过测试,发现二维的话,Python需要先转置再用reshape。
三维的话,Matlab则要对每一页先转置展开为一维,然后再把每一页拼起来,然后再按列往新数组中填充个,具体如下代码,Python的结果和Matlab一致,函数支持一维转二维,二维转二维,二维转三维,三维转二维,三维转三维。注意:在Maltab中假设数组形状为a,b,c,则在Python要改为c,a,b。
import numpy as np
def myreshape(x: np.ndarray, dim: tuple) -> np.ndarray:
flag = False
if np.iscomplexobj(x):
flag = True
if flag:
res = np.zeros(dim, dtype=complex)
else:
res = np.zeros(dim)
if len(x.shape) == 1:
if len(dim) == 2:
m, n = dim
temp = x.flatten()
for i in range(n):
res[:, i] = temp[m * i:m * (i + 1)]
elif len(x.shape) == 2:
if len(dim) == 2:
m, n = dim
res = x.T.reshape((m, n))
else:
l, m, n = dim
temp = x.T.flatten()
idx = 0
for i in range(l):
for j in range(n):
res[i, :, j] = temp[m * idx:m * (idx + 1)]
idx += 1
else:
if len(dim) == 2:
m, n = dim
l1, m1, n1 = x.shape
if flag:
temp = np.zeros(l1 * m1 * n1, dtype=complex)
else:
temp = np.zeros(l1 * m1 * n1)
for i in range(l1):
temp[(m1 * n1) * i:(m1 * n1) * (i + 1)] = x[i, :, :].T.ravel()
for i in range(n):
res[:, i] = temp[m * i:m * (i + 1)]
else:
l, m, n = dim
l1, m1, n1 = x.shape
if flag:
temp = np.zeros(l1 * m1 * n1, dtype=complex)
else:
temp = np.zeros(l1 * m1 * n1)
for i in range(l1):
temp[(m1 * n1) * i:(m1 * n1) * (i + 1)] = x[i, :, :].T.ravel()
idx = 0
for i in range(l):
for j in range(n):
res[i, :, j] = temp[m * idx:m * (idx + 1)]
idx += 1
return res
a = np.array([[[1j, 2], [3, 4]], [[5, 6], [7, 8]], [[9, 10], [11, 12]]])
b = myreshape(a, (2, 3, 2))
aa = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
print(aa.T.reshape(1, -1))
print(aa.reshape(-1, 1))
c = np.iscomplexobj(a)