Tag: Practical AI

A day and month can become a complete-looking date with an unintended year. Define what the record means before asking AI to fix the import.

A relationship in a SQLite schema is not enough. Test that every writing connection rejects records linked to a customer who does not exist.

Python can write NaN and Infinity by default. Decide what an unavailable calculation means, then test that the receiving tool can read your JSON export.

Python’s groupby separates consecutive runs. Learn when to sort, when to accumulate by key and how to test an AI-written report for lost totals.

Python can distinguish an empty CSV field from a short row. Keep that distinction long enough to decide what the import should update, leave alone or flag.

An absolute path can replace the export folder in a Python path join. Learn how to separate an export label from the destination your workflow controls.

Python’s all() returns true on an empty batch. Learn how to distinguish ‘nothing to check’ from ‘ready’ in an AI-written workflow.

Tied timestamps can give two orders the same running total. Test whether your report needs a total through each timestamp or a separate step after each order.

A default list can retain items across Python function calls. Test consecutive jobs, create fresh containers deliberately and agree what happens to supplied lists.

A LEFT JOIN can retain unmatched customers, then a WHERE filter can remove them. Three tested queries show why customers with pending orders need their own review case.