A maintenance reminder that tells you which job to do first

The problem

Plenty of riders keep their service dates in their head or in a notebook. Oil changes and brake checks slip, and the first sign can be a breakdown. We compared the usual fixes: a spreadsheet or paper logbook costs nothing but never warns you, and the car-history service we found did not cover motorcycles.

For our data structures course, the idea paper had to justify every data structure by what it does in the app. A maintenance tracker suited that, because the hard part is not storing the records. It is deciding which job to show first.

What we built

• A ranked alert list. Every open job gets an urgency score: the kilometres past due, plus ten points for each day past its due date. A priority queue (Python's heapq) hands the highest score back first, so the most overdue job is always at the top.
• A garage per user. Accounts with a login and a password reset, and as many motorcycles as you like, each found by plate number through a dictionary.
• Standard jobs that remember you. Nine ready-made jobs such as engine oil, chain, spark plugs and brake fluid, or your own. Change an interval once and the app uses it next time.
• Costs and history. Log a finished job with its date and price, and the total spent on that motorcycle updates.
• Undo. A stack of recent actions lets you reverse a new motorcycle, a mileage update, a scheduled job, a finished job or an edit.
• One file for everything. The whole state saves to a JSON file, and dates can be typed in eleven different formats.

How it turned out

It is a class project, so there are no users or numbers to report. What I can show is that it runs, the code is public, and each part of the design answers a question from the course: a dictionary for lookup, a heap for priority, a list for history and a stack for undo.

The finished program also drifted from our idea paper in useful ways. The paper added days overdue to the score as they were; the code counts each day ten times over, so time weighs more than distance. The nine standard jobs and the remembered intervals were not in the plan at all.

What I would do differently

Passwords are stored as plain SHA-256 hashes with no salt, which is not safe for real users; a slow, salted hash is the fix. The alert list rebuilds its heap every time you open it, which is fine for a few jobs and wasteful for many. And dates are typed as text when a date picker would stop most input errors. If I rebuilt it as a web app, those three would come first.