NumPy Indexing and Slicing

Once your data is in an array, you need to pull parts out of it: one value, a row, a column, or “every mark above 80”. NumPy does all of this with square brackets.

1-D arrays work like lists#

Positions start at 0, and negative numbers count from the end:

import numpy as np

temps = np.array([31, 33, 29, 35, 34, 30])
print(temps[0])
print(temps[-1])
print(temps[1:4])
print(temps[::2])

Output:

31
30
[33 29 35]
[31 29 34]

[1:4] means “from position 1 up to, but not including, 4”. [::2] means “every second element”.

2-D arrays: row, column#

For a table, give two positions separated by a comma: [row, column]. Watch it step by step:

import numpy as np

m = np.arange(12).reshape(3, 4)
print(m)
print(m[1, 2])
print(m[:, 1])
print(m[0:2, 1:3])

Output:

[[ 0  1  2  3]
 [ 4  5  6  7]
 [ 8  9 10 11]]
6
[1 5 9]
[[1 2]
 [5 6]]

• m[1, 2]: row 1, column 2.
• m[:, 1]: the : alone means “everything”, so this is all rows of column 1.
• m[0:2, 1:3]: rows 0–1 and columns 1–2, a small block.

Boolean masks: filtering#

This is the feature you’ll use most in data analysis. A comparison gives an array of True/False, called a mask. Put the mask inside the brackets to keep only the True positions:

import numpy as np

marks = np.array([72, 85, 64, 90, 58, 77])
passed = marks >= 70
print(passed)
print(marks[passed])
print(marks[marks < 60])

Output:

[ True  True False  True False  True]
[72 85 90 77]
[58]

Combine conditions with & (and) and | (or). Each condition needs its own brackets:

import numpy as np

marks = np.array([72, 85, 64, 90, 58, 77])
print(marks[(marks >= 70) & (marks < 85)])
print((marks >= 70).sum())

Output:

[72 77]
4

True counts as 1, so .sum() of a mask tells you how many values match. Here 4 students passed.

Common mistake

Python’s and/or don’t work on arrays: marks > 60 and marks < 80 raises an error. Use &, | and brackets: (marks > 60) & (marks < 80).

Changing values#

Indexing works on the left of = too, which is a fast way to clean data:

import numpy as np

readings = np.array([12, -1, 15, -1, 18])
readings[readings < 0] = 0
print(readings)

Output:

[12  0 15  0 18]

Every negative (bad) sensor reading became 0 in one line.

Fancy indexing: a list of positions#

import numpy as np

names = np.array(["Asha", "Rahul", "Meera", "Kabir"])
print(names[[0, 2]])

Output:

['Asha' 'Meera']

Practice#

1. From np.arange(1, 21), keep only the multiples of 3.
2. Make m = np.arange(16).reshape(4, 4) and print its last column.
3. Replace every value above 100 in [95, 120, 80, 150] with 100.

Summary#

• a[i], a[start:stop:step] work like lists.
• 2-D: m[row, col]; : means “all”.
• A comparison makes a mask; a[mask] keeps matching values; mask.sum() counts them.
• Use & and | with brackets to combine conditions.