Create 3d np array with tuple elements

Question

I want to create a 3-D np.array of dimensions (100x100x100) with the tuple (-1,-1,-1) for every element. Is there a faster/better way than:

np.array([[[(0,0,0) for x in range(100)] for y in range(100)] for z in range(100)])

?

Does this answer your question? [NumPy array initialization (fill with identical values)](https://stackoverflow.com/questions/5891410/numpy-array-initialization-fill-with-identical-values) — Tomerikoo, Nov 18 '20 at 18:14
Explain why you need an array with tuple elements. That's not an optimal array computaitonally. Also tell us about the array your code produced. What's its `shape` and `dtype`. Is that really an array of tuples? — hpaulj, Nov 18 '20 at 18:33

score 0 · Answer 1 · answered Nov 18 '20 at 18:00

0

Your code is not really creating a (100x100x100) array.

Maybe you can try:

np.ones((100, 100, 100))*-1

answered Nov 18 '20 at 18:00

Felix Tang

This is not the shape that the OP is after, their array has shape (100, 100, 100, 3) – Tom Nov 18 '20 at 18:05

Tom · Accepted Answer · 2020-11-18T18:06:33.293

As you can see the shape of your array is actually (100, 100, 100, 3)

>>> np.array([[[(0,0,0) for x in range(100)] for y in range(100)] for z in range(100)]).shape
(100, 100, 100, 3)

You can uses the np.full function to generate the desired array.

np.full([100, 100, 100, 3], -1)

This creates a 100x100x100x3 numpy array, which has an equivalent structure to yours as required.

>>> (np.array([[[(0,0,0) for x in range(100)] for y in range(100)] for z in range(100)]) == np.full([100, 100, 100, 3], 0)).all()
True

Documentation for np.full()

2 Answers2