2.2 Array Creation & Manipulation

NumPy Array Creation & Manipulation

NumPy provides a powerful collection of functions for creating arrays and controlling their structure. Instead of manually entering every value, we can generate arrays containing sequences, zeros, ones, or evenly spaced values.

Once an array has been created, NumPy allows us to change its structure using operations such as reshape, flatten, and ravel.

1. Creating an Array with np.array()

The simplest way to create a NumPy array from existing Python data is the np.array() function.

import numpy as np

numbers = np.array([10, 20, 30, 40, 50])

print(numbers)

Output:

[10 20 30 40 50]

A two-dimensional array can be created using nested lists.

matrix = np.array([
    [1, 2, 3],
    [4, 5, 6]
])

print(matrix)

Output:

[[1 2 3]
 [4 5 6]]

2. np.arange() — Creating Number Sequences

np.arange() creates a sequence of values within a specified range.

Syntax

np.arange(start, stop, step)

Example 1 — Basic Sequence

numbers = np.arange(1, 6)

print(numbers)

Output:

[1 2 3 4 5]

Example 2 — Using a Step

numbers = np.arange(0, 11, 2)

print(numbers)

Output:

[ 0  2  4  6  8 10]

Example 3 — Starting Value

numbers = np.arange(10, 20)

print(numbers)

Output:

[10 11 12 13 14 15 16 17 18 19]

3. Important Rule of np.arange()

The stop value is generally excluded.

numbers = np.arange(1, 5)

print(numbers)

Output:

[1 2 3 4]

4. Using a Negative Step

np.arange() can also generate decreasing sequences.

numbers = np.arange(10, 0, -2)

print(numbers)

Output:

[10  8  6  4  2]

5. np.linspace() — Evenly Spaced Values

np.linspace() generates a specified number of evenly spaced values between two endpoints.

Syntax

np.linspace(start, stop, num)

Example

values = np.linspace(0, 10, 5)

print(values)

Output:

[ 0.   2.5  5.   7.5 10. ]

Five evenly spaced values are generated from 0 to 10.

6. np.arange() vs np.linspace()

7. np.zeros() — Creating Zero-Filled Arrays

np.zeros() creates an array in which all elements are initialized to zero.

1D Array

zeros = np.zeros(5)

print(zeros)

Output:

[0. 0. 0. 0. 0.]

By default, the result is usually a floating-point array.

2D Array

zeros = np.zeros((2, 3))

print(zeros)

Output:

[[0. 0. 0.]
 [0. 0. 0.]]

8. Specifying dtype with np.zeros()

The data type can be explicitly specified.

zeros = np.zeros(
    5,
    dtype=int
)

print(zeros)
print(zeros.dtype)

Output:

[0 0 0 0 0]
int64

The exact default integer width can depend on the platform.

9. np.ones() — Creating One-Filled Arrays

np.ones() creates an array containing ones.

ones = np.ones(5)

print(ones)

Output:

[1. 1. 1. 1. 1.]

2D Example

ones = np.ones((3, 2))

print(ones)

Output:

[[1. 1.]
 [1. 1.]
 [1. 1.]]

10. np.full() — Filling with a Specific Value

np.full() creates an array with a specified shape and fills every element with the same value.

data = np.full(
    (2, 3),
    7
)

print(data)

Output:

[[7 7 7]
 [7 7 7]]

This is useful when an array needs an initial constant value other than zero or one.

11. np.empty()

np.empty() creates an array with the requested shape without explicitly initializing its elements to a particular value.

data = np.empty((2, 3))

print(data)

np.empty() can be useful when you intend to populate the array immediately and want to avoid unnecessary initialization.

12. np.eye() — Identity Matrix

np.eye() creates a 2D array with ones on the main diagonal and zeros elsewhere.

identity = np.eye(3)

print(identity)

Output:

[[1. 0. 0.]
 [0. 1. 0.]
 [0. 0. 1.]]

Identity matrices are widely used in linear algebra and numerical computing.

13. np.diag()

np.diag() can create a diagonal matrix from a one-dimensional array.

data = np.array([10, 20, 30])

matrix = np.diag(data)

print(matrix)

Output:

[[10  0  0]
 [ 0 20  0]
 [ 0  0 30]]

14. Understanding Array Shape

The shape of an array describes how its elements are organized along its dimensions.

data = np.array([
    [1, 2, 3],
    [4, 5, 6]
])

print(data.shape)

Output:

(2, 3)

This means:

• 2 rows
• 3 columns
• 6 total elements

15. reshape() — Changing Array Structure

The reshape() method changes the shape of an array without changing its elements.

numbers = np.arange(1, 7)

matrix = numbers.reshape(2, 3)

print(matrix)

Output:

[[1 2 3]
 [4 5 6]]

The original array contains 6 elements and the new shape also contains 6 positions.

16. The Fundamental reshape() Rule

The total number of elements must remain the same when reshaping.

numbers = np.arange(1, 13)

matrix = numbers.reshape(3, 4)

print(matrix)

The original array contains 12 elements:

12

The new shape also requires:

3 × 4 = 12

17. Invalid Reshape

Consider an array containing 6 elements:

numbers = np.arange(1, 7)

matrix = numbers.reshape(4, 2)

This fails because:

4 × 2 = 8

but the original array contains only 6 elements.

NumPy raises a ValueError.

18. Automatic Dimension with -1

NumPy can automatically calculate one dimension when -1 is supplied to reshape().

numbers = np.arange(1, 13)

matrix = numbers.reshape(3, -1)

print(matrix)

Output:

[[ 1  2  3  4]
 [ 5  6  7  8]
 [ 9 10 11 12]]

NumPy determines that the missing dimension must be 4.

19. Reshaping into Different Structures

numbers = np.arange(1, 13)

print(numbers.reshape(2, 6))

print(numbers.reshape(3, 4))

print(numbers.reshape(4, 3))

print(numbers.reshape(6, 2))

All of these are valid because every shape contains 12 elements.

20. flatten() — Converting to One Dimension

The flatten() method converts a multidimensional array into a one-dimensional array.

matrix = np.array([
    [1, 2, 3],
    [4, 5, 6]
])

flat = matrix.flatten()

print(flat)

Output:

[1 2 3 4 5 6]

The resulting array has one dimension.

21. Important Property of flatten()

flatten() returns a copy of the array data.

matrix = np.array([
    [1, 2],
    [3, 4]
])

flat = matrix.flatten()

flat[0] = 100

print("Original:")
print(matrix)

print("Flattened:")
print(flat)

Output:

Original:
[[1 2]
 [3 4]]

Flattened:
[100   2   3   4]

Changing the flattened array does not change the original in this case because flatten() returns a copy.

22. ravel() — Flattening an Array

np.ravel() or the array method .ravel() returns a flattened one-dimensional view whenever possible.

matrix = np.array([
    [1, 2, 3],
    [4, 5, 6]
])

flat = matrix.ravel()

print(flat)

Output:

[1 2 3 4 5 6]

23. flatten() vs ravel()

24. reshape() vs flatten()

25. reshape() Does Not Necessarily Change the Original Shape

Calling reshape() produces an array with the requested shape. If you want to keep the reshaped result, assign it to a variable.

numbers = np.arange(1, 7)

reshaped = numbers.reshape(2, 3)

print(numbers.shape)
print(reshaped.shape)

Output:

(6,)
(2, 3)

The original one-dimensional array remains one-dimensional.

26. Converting Between 1D and 2D

A one-dimensional array can be converted into a column-like two-dimensional structure.

numbers = np.array([10, 20, 30, 40])

column = numbers.reshape(4, 1)

print(column)

Output:

[[10]
 [20]
 [30]
 [40]]

It can also be represented as one row:

row = numbers.reshape(1, 4)

print(row)

Output:

[[10 20 30 40]]

27. Data Science Example — Monthly Sales

Suppose sales for 12 months are stored in a one-dimensional array.

sales = np.array([
    120, 135, 142, 150,
    160, 172, 180, 175,
    190, 205, 210, 225
])

We can organize the data into a 3 × 4 structure:

quarter_data = sales.reshape(3, 4)

print(quarter_data)

Output:

[[120 135 142 150]
 [160 172 180 175]
 [190 205 210 225]]

Reshaping can make data organization more convenient for later analysis, provided the chosen structure matches the meaning of the data.

28. Data Science Example — Image-Like Data

Images are commonly represented as multidimensional numerical arrays. A grayscale image can be represented as a 2D array of pixel values.

image = np.array([
    [0, 50, 100],
    [150, 200, 255],
    [100, 50, 25]
])

print(image.shape)

Output:

(3, 3)

Flattening can convert the pixel grid into a one-dimensional sequence.

pixels = image.flatten()

print(pixels)

This concept is commonly encountered in image processing and Machine Learning workflows.

29. Data Visualization Example — Generating x Values

linspace() is frequently used to generate evenly spaced points for mathematical functions and visualization.

import numpy as np

x = np.linspace(0, 10, 100)

y = x ** 2

print(x.shape)
print(y.shape)

Both arrays contain 100 values and can be used as coordinates for plotting.

30. Array Creation Cheatsheet

31. Array Manipulation Cheatsheet

32. NumPy Interview Questions

33. Examination Questions

Multiple Choice Questions

np.arange(1, 5)

np.zeros((3, 4))

arr = np.arange(1, 13)
arr.reshape(3, 4)

Short Answer Questions

1. Explain the purpose of np.arange() with an example.
2. Explain np.linspace() and state one difference between it and np.arange().
3. Write Python statements to create a 3 × 3 array containing zeros.
4. Write Python statements to create a 2 × 4 array containing ones.
5. What is reshaping? State the basic rule that must be followed when reshaping an array.
6. Differentiate between flatten() and ravel().

34. Practical Challenge

35. Quick Reference