Unpack numpy array objects, from shape (3,2)(2) to shape (3,2,2)-CodePudding

I have a numpy array of shape (3,2), where each cell is a numpy array of shape 2 (object):

df = pd.DataFrame({'A':[np.array([4,4]),np.array([5,5]),np.array([6,6])], 'B':[np.array([4,5]),np.array([5,6]),np.array([6,7])]})
df.head()

        A       B
0  [4, 4]  [4, 5]
1  [5, 5]  [5, 6]
2  [6, 6]  [6, 7]

a = df.to_numpy()
print(a.shape) #gives (3,2)
print(a[0,0]) #gives array([4,4])
print(a[0,0].shape) #gives (2,)

print(a)
#Gives:
array([[array([4, 4]), array([4, 5])],
       [array([5, 5]), array([5, 6])],
       [array([6, 6]), array([6, 7])]], dtype=object)

How can I unpack the cells so that a becomes of shape (3,2,2)?

CodePudding user response：

IIUC, you might want:

import numpy as np
np.c_[df.values.tolist()]

output:

array([[[4, 4],
        [4, 5]],

       [[5, 5],
        [5, 6]],

       [[6, 6],
        [6, 7]]])