Debugging

01The problem

The month-end job loads the invoice export and adds up what customers paid. It crashed. The engineer on duty found the bad row, edited it by hand, and re-ran the job. It crashed again on a different row. After the third time, they asked for help.

Debugging is a method, not luck: read the error properly, reproduce it with the smallest input, find all the inputs like it, then fix the cause.

02The concept

Reading a traceback

A traceback lists the calls that led to an error, most recent last. Read it from the bottom:

1. The last line: the type of error and its message (ValueError: could not convert string to float: '₦12,500.00').
2. The lines above: where it happened, innermost call last.

A debugging method

Look at the data, not just the code

When code crashes on data, the fastest route is often to ask the data directly: which values in this column don't look like the rest?

03Example

The job's loading step, simplified. Instead of letting it crash, catch the error so you can read it:

import pandas as pd

base = "https://academy.cloudtechanalytics.com/datasets/invoicing/"
raw = pd.read_csv(base + "invoices_raw.csv", dtype=str, keep_default_na=False)

total_paid = 0.0
try:
    for i, row in raw.iterrows():
        total_paid += float(row["amount_paid"])
except ValueError as error:
    print(f"Row {i} ({row['invoice_id']}): {type(error).__name__}: {error}")

Row 98 (INV-100099): ValueError: could not convert string to float: '₦125,179.79'

That's one bad value. Instead of fixing it and waiting for the next crash, find every value that isn't a plain number:

plain_number = raw["amount_paid"].str.fullmatch(r"\d+(\.\d{1,2})?")
odd = raw.loc[~plain_number, "amount_paid"]
print(len(odd), "amounts aren't plain numbers")
print(odd.head(5).to_string())
print("Kinds:", odd.str.replace(r"\d", "9", regex=True).value_counts().to_dict())

27 amounts aren't plain numbers
98     ₦125,179.79
131    ₦460,653.17
184     ₦13,701.95
227    ₦157,928.02
233     ₦26,961.40
Kinds: {'₦999,999.99': 13, '₦99,999.99': 12, '₦9,999.99': 2}

They're all the same kind: amounts formatted with the naira sign and thousands commas, the way a spreadsheet shows them. The to_kobo function from lesson 3 already handles exactly that. Use it for every row:

from decimal import Decimal, ROUND_HALF_UP

def to_kobo(naira):
    cleaned = naira.replace("₦", "").replace(",", "").strip()
    return int((Decimal(cleaned) * 100).quantize(Decimal("1"), rounding=ROUND_HALF_UP))

raw["paid_kobo"] = raw["amount_paid"].map(to_kobo)
print(f"Total paid: ₦{raw['paid_kobo'].sum() / 100:,.2f} across {len(raw)} rows")

Total paid: ₦111,025,044.57 across 1205 rows

One fix for the whole class of problem, instead of one edit per crash. (There are other problems in this export too, such as duplicate invoices, which would make this total wrong. Lesson 6 deals with them.)

04Walkthrough

1. Run the cells. Write a pytest test for to_kobo using one of the odd values.
2. Run float("₦12,500.00") in a cell and read the full traceback from the bottom up.
3. Check the issue_date column the same way: which values don't look like 2026-08-31?
4. Why was editing the bad rows by hand the wrong fix?

05Practice

06Check your understanding