NumPy · Indexing and slicing

• One comma-separated index per axis: a[row, col]. This is cleaner than Python’s a[row][col].
• Slices work on every axis: a[1:3, ::2], a[:, 0] for the first column, a[-1] for the last row.
• ... (Ellipsis) stands for “all remaining axes”.

The key difference from Python lists: slicing a list makes a copy, but a basic slice of an ndarray is a view into the same memory. Writing to the view changes the original. Use .copy() when you need an independent array.

import numpy as np

a = np.arange(1, 13).reshape(3, 4)
print(a)
print(a[1, 2], a[-1], a[:, 0])
print(a[0:2, 1:3])
print(a[:, ::2])

view = a[0, :]
view[:] = 0                      # writes into a!
print(a)

safe = a[1, :].copy()
safe[:] = -1
print(a[1], np.shares_memory(a, view), np.shares_memory(a, safe))

lst = [1, 2, 3]
part = lst[0:2]; part[0] = 99
print(lst)                       # list slice was a copy

Output:

[[ 1  2  3  4]
 [ 5  6  7  8]
 [ 9 10 11 12]]
7 [ 9 10 11 12] [1 5 9]
[[2 3]
 [6 7]]
[[ 1  3]
 [ 5  7]
 [ 9 11]]
[[ 0  0  0  0]
 [ 5  6  7  8]
 [ 9 10 11 12]]
[5 6 7 8] True False
[1, 2, 3]