The codebase and the bug list

01The problem

This is the capstone of the Software Developer track. Instead of writing code from a blank page, you'll do what most developers spend most of their time doing: working on someone else's code, with real bugs reported by real people.

You've joined Kasuwa, the online shop from the other capstones, on the team that owns the refunds service. When a delivery fails or a customer returns something, this service works out the refund and records it. Its tests all pass, yet in the last week support, finance, the mobile team and a security review have between them reported six problems. Your lead says: "Sort these out properly. Tests first, no guessing, and nothing that makes it worse."

Every lesson starts from the original code in a new Colab notebook, so each one stands on its own. In the final project, you'll combine all the fixes in one repository.

02The concept

Read before you change

Find the entry points (the API in app.py), the rules (refunds.py) and the data access (db.py, schema.sql). Run the tests. Passing tests only prove what they test.

Triage by harm

Every bug gets a test

For each bug: reproduce it with a failing test, fix it, and keep the test so it can't come back.

03Example

Get the code and the sample data:

%%bash
base=https://academy.cloudtechanalytics.com/datasets/refunds
for f in refunds.py db.py app.py schema.sql customers.csv orders.csv order_items.csv issues.csv test_refunds.py pytest.ini pr-42.diff; do
  curl -sO "$base/$f"
done
ls

app.py
customers.csv
db.py
issues.csv
order_items.csv
orders.csv
pr-42.diff
pytest.ini
refunds.py
schema.sql
test_refunds.py

Run the existing tests:

%%bash
python -m pytest

...
3 passed in 0.01s

All three pass. Now the bug reports:

%%bash
python - <<'EOF'
import csv
for issue in csv.DictReader(open("issues.csv", encoding="utf-8")):
    print(f"{issue['issue_id']}  {issue['reported_by']:17} {issue['title']}")
EOF

ISSUE-101  Customer support  Refund is a kobo short
ISSUE-102  Finance           Customer refunded twice
ISSUE-103  Customer support  Return on day 14 rejected
ISSUE-104  Security review   Order search shows other customers' orders
ISSUE-105  Finance           Refunds add up to more than the order
ISSUE-106  Mobile team       Server error for a missing order

And a look at the rules everyone depends on:

%%bash
sed -n '9,26p' refunds.py

def item_refund(quantity, unit_price_kobo, reason):
    """Refund for one order line, in kobo."""
    naira = quantity * (unit_price_kobo / 100)
    if reason == "changed_mind":
        naira = naira * (1 - RESTOCKING_FEE)
    return int(naira * 100)


def within_window(delivered_on, today):
    """True if a return is still allowed: within 14 days of delivery."""
    return (today - delivered_on).days < RETURN_WINDOW_DAYS


def refund_amount(items, reason, delivery_fee_kobo):
    """Total refund in kobo. items is a list of (quantity, unit_price_kobo) pairs."""
    total = sum(item_refund(quantity, price, reason) for quantity, price in items)
    if reason in ("failed_delivery", "damaged", "wrong_item"):
        total += delivery_fee_kobo

Three things stand out before you even run anything: money passes through floats in naira, the window check uses < against 14 days, and int() cuts off fractions rather than rounding. The tests never exercise any of them.

04Walkthrough

1. Run the cells in a new Colab notebook.
2. Read db.py and app.py. Which line looks unsafe to you, before reading any bug report?
3. Read every issue's full description (issues.csv). For each, guess which file the bug is in.
4. Start the service in a Python cell with the Flask test client and post one refund, to see it working.
5. Write the triage (the task below).

05Practice

06Check your understanding