i have a huge dataset but to simplify the problem, think about a 4x4 2d list.
want make a new list with 5 elements which is choosen by randomly from the 4x4 list without using numpy or any addional libraries.
edit:
[['j', 'k', 'z','p'], [1,6,8,9], [8,True,0,'a'], [66,'False', '12', '5]] i want to pick 5 elements of that list ramdomly with random library functions
output maybe like a new list of choosen elements:
['j', 'False', 'a', 8, 66] hope that it's clear enough.
with the random module you can use a list comprehension:
import random list_2d = [... your list 4X4] randmly_chosen = [random.choice(random.choice(list_2d)) for _ in range(5)] (Your list right now isn't quite valid syntax, but fixing that up):
>>> import itertools >>> import random >>> list_2d = [['j', 'k', 'z', 'p'], ... [1,6,8,9], ... [8,True,0,'a'], ... [66,'False', '12', '5']] >>> random.choices(list(itertools.chain(*list_2d)), k=5) ['False', '12', 8, 6, 'k'] The itertools.chain call will flatten it into a 1d list, which the random.choices function can handle to sample with replacement: https://docs.python.org/3/library/random.html#random.choices
Also, without more detail of the full problem it will be hard to know whether this solution will be good for the "huge dataset"- it creates a new copy of the full list as a 1d list, so if that uses too much memory, we'd need to see where the data is originally coming from, and if you could avoid the list-of-lists structure in the first place.
Idea of flattening ls could still be borrowed from numpy without any use of its methods:
ls = [['j', 'k', 'z', 'p'], [1,6,8,9], [8,True,0,'a'], [66,'False', '12', '5']] flat_ls = [] for n in ls: flat_ls.extend(n) from random import shuffle shuffle(flat_ls) #dynamic change of flat_ls, very fast >>> flat_ls[:5] ['12', 66, 8, 'j', 'a'] 