ANU Library room booking

A cross-library group study room booking prototype. Right now, checking room availability at ANU Library means visiting four separate LibCal pages (one per library); you have to log in with ANU SSO before you can see availability, and room details sit behind a per-room “Info” button. The booking rules — LibCal’s page says you can book “up to two hours a day” and “up to two weeks in advance” — exist only as text on the page: nothing tells you a booking would break them until you try. This app puts every room from all four libraries — Chifley, Hancock, Menzies, Law — in one filterable grid, greys out anything you can’t book, lets you book a free slot with just a name and email, and takes you straight to your bookings afterwards, where you can cancel. No login is required to browse or to book.

What good looks like here

Good, here, means the booking rules are enforced by the server and the database — not just by the UI — and that they hold under HTTP requests made directly against the API, not only through the rendered forms:

• No double booking. A room/date/hour combination can be booked once. The server first requires a real YYYY-MM-DD date (anything else is rejected as invalid_date, so a variant spelling of the same day can’t get its own slot), and the database’s unique index then allows one booking per slot, even under two near-simultaneous requests.
• A 2-hour daily cap per person, taken from LibCal’s own rule. Counted per (trimmed, lowercased) email per Sydney calendar day, so Test@anu.edu.au and test@ANU.edu.au share one cap, and the count-then-insert is one atomic transaction so two near-simultaneous bookings can’t both slip past the limit.
• A 14-day booking window, LibCal’s “two weeks in advance”. Only today through 14 days ahead is bookable, computed against Australia/Sydney, not the server’s or your browser’s own clock — that matters because this runs on a UTC container.
• No booking a slot that’s already started, again computed in Sydney time.
• Clear, stable error codes (slot_taken, daily_cap, invalid_date, out_of_window, past_slot, email_mismatch, …) rather than one generic failure message, so a rejected booking tells you which rule stopped it.

These rules, and the tests that hold them to it end-to-end over HTTP (booking, rejecting a double-booking, rejecting a third hour, rejecting an out-of-window or already-started slot, and cancelling), live in src/lib/bookings.ts and spec/bookings.test.ts. The invariants in spec/invariants.test.ts (one heading, a nav landmark, a real title, an accessibility floor) are the supplied baseline and apply to every page, including the new ones.

What I chose not to build: no SSE/live updates across tabs (the spec only needed reload-persistence, not multi-client sync); no admin view of all bookings; no email confirmation. All out of scope for a one-week prototype.

Known limitations

• The room list is illustrative, not the real ANU room list. Twelve rooms across the four libraries, with made-up names, capacities and equipment tags, seeded for demonstration purposes only.
• There is no real authentication. Anyone who knows (or guesses) an email address can look up and attempt to cancel that email’s bookings on /bookings — there’s no SSO, password, or confirmation link tying an email to the person who actually owns it. A real deployment would need to fix this before handling genuine bookings.
• The remembered email is a convenience, not a login. After you book or look up your bookings, your email is kept in a cookie for 30 days so you don’t have to retype it, and “Not you?” clears it. It proves nothing about who you are.
• My bookings also shows past bookings. The page lists every booking for that email, including ones whose time has already passed.
• Opening hours are simplified to 08:00–22:00 for every library. In reality Chifley and Hancock are open 24 hours; one shared set of hourly slots kept the grid simple for this prototype.