Put a safety net under legacy code

#Brief

Most engineers spend more time changing existing code than writing new code. The safe way to change untested code is to first write tests that record what it does today, then refactor. Real data makes these tests realistic.

#What to build

1. Write a single 150-line Python function that computes an order's total from Olist order items and payments: item prices, freight, instalment interest and a few special cases. Make it messy on purpose (long conditionals, repeated logic, magic numbers). Commit it.

2. Pick 500 real orders. Record the function's output for each as a snapshot, and write a test that compares the function with the snapshot.

3. Add unit tests for edge cases you discover: orders with no payments, many instalments, zero freight, rounding.

4. Refactor in small steps (extract function, rename, remove duplication), running the tests after each one and committing only on green.

5. Where you find behaviour that looks wrong, decide: fix it (and update the snapshot with a clear commit) or document it.

#Deliverables

• Repository with the original module, the tests, the refactored module and NOTES.md

#How it is graded

Reviewers read the commit history as a story and run the tests at several commits. Every commit should be green, and no behaviour should change without a commit that says so.

#Stretch

Add property-based tests with Hypothesis (for example, totals are never negative and adding an item never lowers the total).