Python Testing 101: Writing Your First Unit Tests

What is a Unit Test?

• A unit test checks that a small unit of code works as expected.
• Usually a single function or method.
• Helps you catch bugs early, refactor confidently, and build better software.

Why Should Data Scientists Care?

• Prevent breaking your analysis pipeline accidentally.
• Add guardrails for complex logic (e.g., custom transforms).
• Make code easier to reuse and maintain.

Unit Testing Workflow

• Write failing test -> Run with pytest (It should fail)
• Write just enough code to pass -> Re-run pytest
• See it pass 🎉
• Refactor code if needed
• Repeat 🔁

Best Practices

• 🧪 Test one case per test, keep them small!
• 🔁 Test edge cases and common use
• 🧼 Use clear naming: test_functionname_case
• 📦 Structure code so logic is in testable functions (not in notebooks!)

Code Structure for Testing

Before testing: - Your code should live in a .py file, e.g. my_module.py - Tests should go in a separate file: test_my_module.py

project/
├── my_module.py
├── test_my_module.py

Let’s Write a Failing Test

test_my_module.py:

from my_module import multiply_by_two

def test_multiply_by_two():
    assert multiply_by_two(3) == 6

Run it with pytest

pytest

Expect output like:

>       assert multiply_by_two(3) == 6
E       NameError: name 'multiply_by_two' is not defined

⬆️ Good! The test is failing as expected.

Write Just Enough Code to Pass

my_module.py:

def multiply_by_two(x):
    return x * 2

Re-run pytest:

collected 1 item
test_my_module.py .                                    [100%]

✅ Green means good!

Common Anti-Patterns to Avoid

🚫 Too much in one test - Hard to precisely understand what failed and why

🚫 Testing implementation details - Tests should focus on what your function does and outputs, not how it does it

🚫 Skipping tests for ‘just a script’ - Scripts evolve fast — Make testable functions early and part of your default workflow

🚫 Functions that are hard to test - If a function: - Has side effects - Does too many things - Depends on global state

➡️ It might need to be split into smaller pieces

💡 If it’s hard to write a simple test, your function may need refactoring!

What’s Next?

• Try testing:
  • Edge cases (0, negative numbers, empty lists, incorrect input types)
  • Your own utility functions
• Explore:
  • Test coverage (pytest-cov)
  • Mocks and temporary files