Stacking ndarrays in a loop using np.stack over the new dimension (creating an array of ndarrays)

Question

I cannot stack numpy arrays over new dimension because of the following problem:

>>> a = np.zeros(shape=(100,100))
>>> b = np.zeros(shape=(100,100))
>>> c = np.stack((a, b))
>>> c.shape
(2, 100, 100)
>>> d = np.zeros(shape=(100,100))
>>> c = np.stack((c, d))
Traceback (most recent call last):
  File "/lib/python3.7/site-packages/numpy/core/shape_base.py", line 426, in stack
    raise ValueError('all input arrays must have the same shape')
ValueError: all input arrays must have the same shape

The way I intend to use it in a loop is:

final = None
for next_mat in mats:
    final = next_mat if final is None else np.stack((final, next_mat))

How do I achieve it? Thank you!

Answer 1

I would rather store all the arrays and stack once:

cum_arr = []
for next_mat in mats:
    cum_arr.append(next_mat)

np.stack(cum_arr)

Or, if you have the mats list:

np.stack(mats)

Answer 2

Since stack expects all inputs to be of the same shape, if you want to stack during each loop, you can use vstack instead. You also need to expand dimensions to (1,100,100) from (100,100).

final = None
for next_mat in mats:
    next_mat = np.expand_dims(next_mat, 0)
    final = next_mat if final is None else np.vstack((final, next_mat))