Numpy - Iterating Over Arrays

NumPy provides flexible and efficient ways to iterate over arrays of any dimensionality. For a one-dimensional array, iterating is straightforward and similar to iterating over a Python list.

Let's understand with the help of an example:

1
2
3
4
5

Explanation:

• Each element of the array is accessed sequentially, making iteration simple and intuitive.

Let's explore various others ways to iterate over Arrays:

Iteration Over a Two-Dimensional Array

For multidimensional arrays, iteration is performed over the first axis by default. Each iteration yields a one-dimensional sub-array.

[1 2 3]
[4 5 6]
[7 8 9]

Explanation:

• arr_2d = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) creates a 2D array with three rows and three columns.
• for row in arr_2d: loops through each row of the 2D array, treating it as a separate one-dimensional array.

For Element-Wise Iteration

To iterate over individual elements of a 2D array, the array can be flattened using the .flat attribute.

1
2
3
4
5
6
7
8
9

Iterating Over Higher-Dimensional Arrays

Iteration is similar for arrays with more than two dimensions. In such cases, iteration occurs along the first axis by default.

[[1 2]
 [3 4]]
[[5 6]
 [7 8]]

Explanation:

• Each iteration returns a 2D sub-array from the 3D array.

Using nditer

The numpy.nditer object offers a various way to iterate over arrays. It allows iteration in different orders and provides better control over the iteration process.

1
2
3
4

We can also specify the order of iteration (row-major order 'C' or column-major order 'F'):

1
3
2
4

Using Enumerate

For multidimensional arrays, we can use enumerate to access both the index and value during iteration.

Row 0: [10 20]
Row 1: [30 40]

Using Broadcasting

When working with arrays of different shapes, Broadcasting allows iteration across combined shapes seamlessly.