how to create uniform nested lists and convert them in to multi dimensional arrays?

Question

i would like to create a nested list of arbitrary depth (containing numerical values specifically) with uniform arbitrary lengths at each level and then compress it to a NumPy array of minimum dimensions ie. if a layer of lists all length 1 all lists at that depth are unpacked.

example:


import random as ran
import numpy as np

prompt = (3,1,5,1)
#promt length and positions of 1 should be arbitrarily
old_list = []

for a in range(prompt[0]):
    old_list.append([])
    for b in range(prompt[1]):
        old_list[a].append([])
        for c in range(prompt[2]):
            old_list[a][b].append([])
            for d in range(prompt[3]):
                #using random outputs for simplicity, could be anything 
                old_list[a][b][c].append(ran.randint(0, 10))

new_list = []
for a in range(prompt[0]):
    new_list.append([])
    #prompt[1] should be automatically skipped as it is 1 
    for c in range(prompt[2]):
        #prompt[3] should be automatically skipped as it is 1 
        new_list[a].append(old_list[a][0][c][0])
final = np.array(new_list)

print(f'''{final} 
          is clearly and more useful and tidier than
          {old_list} ''')

I expect there is some way of automating for loops in the manner described, and compacting them as well, in NumPy, but I don't Know where to get started. any help would be apreciated. thankyou!

You could use np.reshape for "compacting". np.array(old_list).reshape(3, 5) would give you a similar result. — Adeva1
– Adeva1, Commented Aug 6 at 16:28
@Adeva1 you mean np.array(old_list).reshape([v for v in prompt if v!= 1]) without knowing the values in prompt but just wanting to flat list of 1 element — bruno
– bruno, Commented Aug 6 at 16:35
Creating the array directly will be faster and easier, unless the element assignment function is a harder to use than 'random'. Think multidimensional array from the start rather than nested lists. — hpaulj
– hpaulj, Commented Aug 6 at 19:46

mozway · Accepted Answer · 2025-08-07 09:26:28Z

If your goal is to have a numpy array, you do not need to generate an intermediate python object with nested lists. It's much faster and memory efficient to directly create a random numpy array (e.g. with np.random.randint).

If you want to remove the axes of length 1, squeeze:

prompt = (3, 1, 5, 1)
final = np.random.randint(0, 10, size=prompt).squeeze()

Or, with the new Generator interface:

prompt = (3, 1, 5, 1)
rng = np.random.default_rng()
final = rng.integers(0, 10, size=prompt).squeeze()

Example output:

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

Now, if you really want a pure python approach, you could go with a recursive function:

import random

def nested_list(prompt, squeeze=True):
    if not prompt:
        return []
    if squeeze:
        prompt = [i for i in prompt if i != 1] or [1]
    i, *prompt = prompt
    if prompt:
        return [nested_list(prompt, squeeze=False) for _ in range(i)]
    else:
        return [random.randint(0, 10) for _ in range(i)]

nested_list((3, 1, 5, 1))
# [[9, 3, 1, 4, 6], [2, 9, 10, 0, 8], [1, 6, 8, 9, 10]]

nested_list((3, 1, 5, 1), squeeze=False)
# [[[[8], [10], [8], [7], [6]]],
#  [[[10], [8], [7], [1], [1]]],
#  [[[0], [10], [5], [7], [0]]]]

nested_list((1, 1, 1))
# [9]

Collectives™ on Stack Overflow

how to create uniform nested lists and convert them in to multi dimensional arrays?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related