Unofficial student prototype · COMP4020

ANUCampus Parking

ANU campus parking

A full-stack finder for parking on the ANU Acton campus. It answers the two questions the current arrangement makes you juggle across signage, a PDF and an app: where can I park and what will it cost me?, and which car park actually has room right now?

Every parking area OpenStreetMap records on campus is listed with its hours, rate, the permit it wants and the fees and eligibility behind that permit, and each one is a pin on an OpenStreetMap map. The list is searchable and filterable by category, and the map shows exactly the set the list does: narrow the list and the pins narrow with it. On top of the reference data sits one live flow: anyone can report how full a car park is, the report is saved, and every open tab (and the map) updates in real time. Open the site in two tabs and report a park in one to see it.

What good looks like here

Good here is a driver trusting the page enough to leave the house on it. That rests on three decisions:

  1. The directory is the source of truth, and it needs no JavaScript. The map and the live updates are progressive enhancement. If Leaflet or the network is unavailable, the server-rendered list still carries every fact, and the report form still works through a plain POST and redirect. This is also what keeps the accessibility floor green.
  2. State lives in the schema, and the core flow persists. Car parks and reports are SQLite tables defined in src/lib/schema.ts; the facts are seeded from src/lib/parking-data.ts on first boot. A report is written to the volume, so it survives a reload and a redeploy, and broadcast over SSE to other clients. That is the crit’s persistence contract, and spec/parking.test.ts holds it.
  3. The data is honest about what it is. The rate scheme follows ANU’s published 2026 parking fees: staff surface permits at $7.78/day, staff parking stations at $9.59/day, non-resident student surface at $3.88/day, resident student at $4.90/day, and free parking for accessible-permit and motorbike users. The enforcement hours and permit-type names follow ANU’s parking options on Acton campus: pay, permit and time-limited zones are enforced 8am–5pm Mon–Fri, public holidays excepted, with no pay or permit required outside those hours, though signed bay limits still apply 24/7. The whole directory is generated from OpenStreetMap (© OpenStreetMap contributors) by scripts/gen-parking.mjs: every entry is a distinct mapped parking feature, its coordinates are the feature’s centre (the same data the map draws), and any capacity or accessible-bay count shown is what OSM records. The set is clipped to the campus by the roads that ring it, Clunies Ross Street, Barry Drive and Edinburgh Avenue, so lots the extract caught beyond them (CSIRO Black Mountain and the Botanic Gardens, North Oval and Toad Hall, New Acton) are left out; the boundary is traced from those roads’ own OSM geometry. What is still interpretive, not a survey, is each lot’s category, rate and permit; those are derived from its OSM access/fee/type tags mapped onto the ANU fee scheme (e.g. access=private → resident/private, access=permit → permit zone, public + fee=yes → visitor pay, otherwise the campus staff/student surface default), and unnamed lots are labelled by the nearest mapped landmark. The app says so in its footnote. Before this could be relied on, each lot’s permit zone would need checking against the official permit-zone map.

What’s enforced vs. judged

spec/invariants.test.ts (shipped) enforces the accessibility floor and the structural basics on every route in spec/routes.ts. spec/parking.test.ts (mine) enforces the contracts above: the directory carries each area’s facts and the grounded rate figures, a report survives a reload, and a report is broadcast over SSE. What a person still judges at the crit: whether the data is right, whether the map reads well, and whether the live flow feels useful rather than gimmicky.

Accounts, cars, tickets and payment

On top of the public directory sits a signed-in flow: a driver registers an account, adds one or more of their cars, books a casual parking session (a “ticket”) at a car park for one of those cars, and pays for it. Like the report flow, every step is a plain HTML form that ends in a 303 redirect, so it works with no JavaScript, and every step is state defined in src/lib/schema.ts (users, sessions, cars, tickets, payments) and reached only through the queries in src/lib/accounts.ts. Passwords are stored only as a scrypt hash with a per-user salt (src/lib/auth.ts); a session is an unguessable token in an httpOnly cookie, backed by a row so it can expire and be revoked on logout. Each user’s data is scoped to their id, so one account can never read or pay another’s tickets; spec/accounts.test.ts holds that, along with the register → add car → book → pay chain.

Two honesty notes carry over from the rest of the app. Ticket prices are grounded: a ticket is priced on ANU’s published 2026 PayStay pay-as-you-go rates (staff/student $2.36/hour or $16.92 all day; visitor $3.07/hour or $27.63 all day), which is the parking a driver actually buys on the spot with no permit; the permit day-rates are shown for comparison but not sold here, and free parking (accessible bays, motorbikes, honorary staff) is never charged for. Payment is simulated: confirming a ticket takes no card details and moves no money, it only records that a charge of the ticket’s price was made and marks the ticket paid. The footer and the payment screen say so.

What I chose not to build

No per-user favourites, no routing, no space-level (bay-by-bay) tracking, and no real payment gateway or actual ANU permit issuance. The account flow models the shape of booking and paying for parking; it does not connect to ANU’s systems and does not take real money.

Running it

mise install && pnpm install
pnpm dev     # local dev server under the base path
pnpm check   # typecheck + build + tests against the running app