Map Reject Codes First. Move One Classifier After.

Описание и применение

Заморозка таблицы фикстур и фиксация золотого снимка позволяют защитить импорт CSV от непреднамеренных изменений при рефакторинге.

Задача

Messy import scripts fail extracts for a measurable reason because validation rules are hidden inside file reads and writes, causing drift when refactoring.

Результат

By freezing a fixture table and recording a gold snapshot, the importer’s behavior becomes a contract that protects against accidental drift during refactoring.

Рабочий процесс

• 1. Build a CSV fixture that covers all branches (blank email, missing name, bad role, mixed‑case email, missing role).
• 2. Store the fixture text as a constant in the test module to avoid file I/O during asserts.
• 3. Run the importer with a recording writer that appends written records to a list.
• 4. Print the summary dict and written records to capture the first gold snapshot.
• 5. Verify the snapshot matches expected counts, reject codes, and canonical dicts.
• 6. Use the snapshot as a contract; any future refactor must preserve these outputs.
• 7. Optionally compute a hash of the snapshot for quick change detection.

Доказательства

Build one CSV that hits every branch once.

Источник

Run the importer against a recording writer.

Источник

Expected gold, derived from the fixture rules above: ...

Источник