Skip to main content

Handling Errors

Updated Oct 31, 2019 ·

Overview​

Errors can be prevented or managed to help users and developers avoid issues when using our code.

Traceback​

When working with packages, tracebacks show where an error occurred in the code.

In pandas, the terms except and raise handle errors like a KeyError when a column is missing.

Example:

import pandas as pd
df = pd.DataFrame({"name": ["John", "Jane"]})
print(df["age"])  # KeyError: 'age'

Output:

`KeyError: 'age'`

Incorrect Input​

When writing functions, it’s important to consider how users might incorrectly call them.

For example, a function may expect a list, but users might pass a dictionary. Handling such errors improves user experience.

def average(values):
    return sum(values) / len(values)

Error if a dictionary is passed:

average({"a": 1, "b": 2})  # TypeError

Error-Handling Techniques​

There are multiple ways to handle errors gracefully in functions, such as using try-except blocks or raising custom exceptions.

try-except Block​

Use try to test code and except to handle errors. This prevents crashes and allows custom messages.

Example:

def average(values):
    try:
        return sum(values) / len(values)
    except TypeError:
        print("Expected a list or set, but got something else.")

Output if input is wrong:

Expected a list or set, but got something else.

Raise Exception​

We can use raise to force an error when the input is invalid and provide a custom message.

Example:

def average(values):
    if not isinstance(values, (list, set)):
        raise TypeError("Input must be a list or set.")
    return sum(values) / len(values)

Output when invalid input is provided:

TypeError: Input must be a list or set.

try-except vs. raise​

Both try-except and raise have their advantages:

  • try-except

    • Allows the program to continue running if an error occurs.
    • Ideal for handling errors without stopping the script.

  • raise

    • Stops execution and raises a specific error when something goes wrong.
    • Best for scenarios where we want to prevent further code execution.

Feedback