CODEWITHSUNDEEP
pythonarraysnumpyindexingslice
Anastasia Montgomery
Anastasia Montgomery

Reputation: 23

How to subset a multidimensional numpy array with steps?

I have a multi-dimensional array that is:

d.shape = (744, 1, 288, 315)

The first dimension is hours -- so I want to subset this array such that I only get the daytime hours, so hours 10-17 through each time 24-hour period.

I have been subsetting by creating a list:

start = 10
end = 17
d_daily = np.array([d[i*24+start:i*24+end] for i in range(31)])

but it's slow.

I feel like there should be a quick way for me to do a subset, something like:

d[start:(len(d)-(24-end)):24] # basically, take these 7 steps over and over until the end of the array

However, this gives me an array that is (31, 1, 288, 315) as opposed to the (217,1,288,315) output I would expect.

It is important that I keep the order of the data as well...

I know this is very simple, but I would appreciate your help!

Thanks

Upvotes: 0

Views: 445

Answers (2)

Joe Todd
Joe Todd

Reputation: 897

What about this, using modular arithmetic:

import numpy as np

arr = np.zeros((744, 1, 288, 315))

start = 10
end = 17

idx = np.arange(744) % 24
bool_arr = np.logical_and(idx >= 10, idx < 17)

d_daily = arr[bool_arr, :, :, :]

I'm using an array of zeros as an example, and this will include the hours from [10, 17) (i.e. 17 excluded).

Upvotes: 1

sai
sai

Reputation: 1784

Seems fairly simple as you told, can you try this?

d[10*31:17*31,:,:,:]

Upvotes: 1

Related Questions