CHAPTER 23: VISUAL LABS 23a AND 23b

These original exercises use the same synthetic Harbour dataset as the book.
The display/check runner opens companion/sql/harbour.db read-only.
If that file is missing, first follow Chapter17 and run sql/setup_database.py.

RUN
1. Open walkthrough.py in IDLE and choose Run > Run Module (F5).
2. Alternatively, open a terminal in this ch23 folder.
   Windows: py walkthrough.py
   macOS/Linux: python3 walkthrough.py
Use the Python command that already worked in your setup.
No network access or third-party package is required.

The .sql files hold inspectable questions. The supplied Python wrapper prints
results and checks the original expected values. Before editing, predict the
new result. After editing, compare it manually; an old assertion may now fail
because your intentional change needs a different expected answer. Do not
remove a failed check until you can explain the changed result.

EXPECTED RESULTS AND PRACTICE
23a first shows record 1, then the pending records 1 and 2, then record 1 again after rejection and rollback. Change/predict/check: replace the negative amount by 500 cents. The supplied follow-up confirms records 1, 2 and 3 commit together. 23b returns the two Email campaigns, then zero rows for the punctuation-containing literal, then Workplace guide for Organic search. All mutations are in memory; channel lookups open the original fixture read-only.
