Save a report someone can check

Read the example alongside the explanation. Run it in a new notebook cell and change one input to see how it behaves.

import csv

raw_rows = [['T1', '20'], ['T1', 'bad'], ['T1', '22']]
with open('bench_raw.csv', 'w', newline='', encoding='utf-8') as file:
    csv.writer(file).writerows([['sensor', 'temperature_c']] + raw_rows)

clean = []
rejected = []
with open('bench_raw.csv', newline='', encoding='utf-8') as file:
    reader = csv.reader(file)
    header = next(reader)
    for line, row in enumerate(reader, start=2):
        try:
            value = float(row[1])
        except ValueError:
            rejected.append((line, row, 'non-numeric value'))
            continue
        clean.append([row[0], value])
with open('bench_clean.csv', 'w', newline='', encoding='utf-8') as file:
    csv.writer(file).writerows([header] + clean)
with open('bench_clean.csv', newline='', encoding='utf-8') as file:
    saved = list(csv.reader(file))
print('Saved data rows:', len(saved) - 1)
print('Rejected:', rejected)

Example output

Saved data rows: 2
Rejected: [(3, ['T1', 'bad'], 'non-numeric value')]

This file-flow model intentionally uses three fixed, two-field rows. It checks numeric conversion only. Before accepting external data, add header, field-count, sensor-ID, finite-value and range checks from Weeks 9 and 12. Running it creates or replaces bench_raw.csv and bench_clean.csv in the current notebook folder; use a scratch folder.

1. A CSV-aware writer quotes fields as required; a CSV-aware reader reconstructs them.
2. Plain split on commas treats the comma inside a quoted field as another separator.
3. Write mode replaces an existing file; append mode adds content and can duplicate rows or headers if used carelessly.

Test a write–read round trip and inspect the saved file after closing it. Equality of the reconstructed records checks the representation; it does not validate the original measurements.