How to catch a NumPy warning like an exception in Python

An N-dimensional array that is used in linear algebra is called NumPy. Whenever the user tries to perform some action that is impossible or tries to capture an item that doesn’t exist, NumPy usually throws an error, like it’s an exception. Similarly, in a Numpy array, when the user tries to perform some action that is impossible or tries to capture the item that doesn’t exist, such as dividing by zero, it throws an exception. In this article, we will study various ways to catch Numpy warnings like it’s an exception.

Catch a NumPy Warning like an Exception in Python

• Using errstate() function
• Using warnings library

Using errstate() Function

The Numpy context manager which is used for floating point error handling is called errstate() function. In this way, we will see how we can catch Numpy errors using errstate() function.

Syntax: with numpy.errstate(invalid=’raise’):

Example 1: In this example, we have specified a Numpy array, which we try to divide by zero and catch the error using errstate() function. It raises the error stating ‘FloatingPointError‘ which we caught and printed our message.

Python3

# Import the numpy library
import numpy as np

# Create a numpy array
arr = np.array([1,5,4])

# Divide by Zero Exception
with np.errstate(invalid='raise'):
    try:
        arr / 0
    except FloatingPointError:
      print('Error: Division by Zero')

Output:

Using warnings library

The message shown to user wherever it is crucial, whether the program is running or has stopped is called warning. In this way, we will see how we can catch the numpy warning using warnings library.

Syntax: with warnings.catch_warnings():

Example 1: In this example, we have specified a numpy array, which we try to divide by zero and print the error message using warnings library. It raises the error stating ‘FloatingPointError‘ with detailed message of divide by zero encountered in divide.

Python3

# Import the numpy library
import numpy as np

# Create a numpy array
arr = np.array([1,5,4])

# Divide by zero Exception
with warnings.catch_warnings():
    try:
        answer = arr / 0
    except Warning as e:
        print('error found:', e)

Output:

Using ‘numpy.seterr’ Function

The ‘numpy.seterr’ function to configure how NumPy handles warnings. NumPy warnings are typically emitted when there are issues related to numerical operations, data types, or other conditions that might lead to unexpected behavior. By configuring NumPy to treat warnings as exceptions, you can catch and handle them in your code. In the example above, np.seterr(all=’raise’) is used to configure NumPy to raise exceptions for all warnings. You can customize this behavior by specifying specific warning categories using the divide, over, under, etc., options of np.seterr based on your needs.

Python3

import numpy as np

# Configure NumPy to treat warnings as exceptions
np.seterr(all='raise')
# This will raise exceptions for all warnings
try:
    # Code that might trigger a NumPy warning
    # For example, performing an invalid operation
    result = np.array([0, 1]) / 0
except Warning as e:
    print(f"Caught a NumPy warning: {e}")
except Exception as e:
    print(f"Caught an exception: {e}")

Output:

Conclusion

It is a matter of disappointment to see the errors or warnings while coding, but the various ways explained in this article will help you to catch the numpy warnings and resolve them or raise an exception, wherever required.