ANU enrolment, by course code

To enrol in a course at ANU you have to find it on Programs and Courses, open its offerings table, read off a four- or five-digit class number, then retype that number into ANU Hub — a separate system that never mentions the course code you started with. This app is one slice of enrolment rebuilt around the thing a student actually knows: the course code. You search COMP3430, see whether you’re allowed to take it and why, and enrol without ever being asked for a class number. The number still exists — it’s shown to you as a receipt once you’re in, never collected as an input.

What good looks like here

The argument is that the class number should never be the user’s problem, so the enforced rule is structural: no form on this app has a field for one. spec/enrolment.test.ts checks this directly — an enrolment POST carrying only a course code and semester succeeds, and asserts the rendered page contains no input named anything like “class number”.

Around that, enrolling has to obey the same rules real enrolment does, in an order that reports the most specific reason first: already enrolled, already completed, incompatible with something you’ve taken, missing a prerequisite group, and — checked last, since it’s a fact about your whole semester rather than this course — over the 24-unit full-time load. All five are enforced by the spec suite, not left to the UI to get right; the UI’s job is to never show an enabled “Enrol” button for a course the same check would refuse.

Prerequisites are modelled in conjunctive normal form (a groupNo per requisite group, OR’d within a group, AND’d across groups), because that’s the actual shape of ANU’s requisite text — “COMP1100 or COMP1130, and COMP2400” — rather than a simpler model that would misrepresent real rules. Course codes, titles, units and requisites are hand-seeded from real Programs and Courses data for a deliberately narrow, deep slice of the undergraduate COMP catalogue, not scraped and not broadened to other subjects — depth over breadth was a judgement call, made so the requisite chains in the demo are real ANU rules rather than invented ones.

What this app chose not to build: timetable clash detection. At ANU you enrol in courses for a semester and register for individual class times separately — clashes aren’t a fact this system has yet, so pretending to check for them would be modelling a problem enrolment doesn’t actually have.

Two things are judgement calls the spec doesn’t check: the visual design (cards over rows, the ANU gold accent used sparingly per ANU’s own brand guidance rather than as a background) and the choice to seed every new account with the same starting study history, since this prototype has no real student record to read one from — /me/ says so plainly rather than hiding it.