Degree Plan Checker

This is a course planner for ANU School of Computing master’s students: Master of Computing (MCOMP), Master of Computing (Advanced) (VCOMP), and Master of Machine Learning and Computer Vision (MMLCV). You pick your program, when you started, and which semester you’re enrolling in now, then build your plan semester by semester against your degree’s requirement groups. Its one job: flag any course that would earn zero credit toward your degree before you enrol in it, so you don’t waste a semester slot on a course that doesn’t count. It also flags courses offered in the wrong semester and courses whose prerequisites aren’t done yet. Every plan lives at its own link (/plan/<id>), no login — persistence is the point, not privacy.

What “good” means here

Good means the zero-credit warning is correct; correct matters more than complete. Where the checker can’t tell — a prerequisite too tangled to parse safely, a course dropped from the catalogue, an offering not yet published — it says so plainly (“Check P&C”, “Unverified course code”, “Offering not published yet”) instead of guessing. A wrong “this is fine” is worse than an honest “we can’t tell.” Warnings never block adding a course; they’re information, not a gate — the decision stays yours.

Enforced by tests vs judged by a person

Enforced by spec/, run in pnpm check: the checker’s rules (one test per rule — zero-credit reasons, allocation order, pathway conflicts, timing warnings, unverified codes); the scraper’s parsers against saved real pages, including one requisite that must parse and one that must stay unparsed; the create-plan / add-course / reload / remove flow over HTTP; every route returning something sane; this page served in full at /readme/.

Judged by a person: whether the degree data is actually right (scraped from prose, not a machine-readable feed — see Known limits), whether the layout holds up in a real browser at phone width, and whether the wording is clear to someone who isn’t the person who built it. A green pnpm check proves the rules behave the way this repo encoded them; it doesn’t prove that encoding matches the real degree rules or that the page looks right.

Deliberately not built

• Login, accounts, or private plans — anyone with the link can view and edit it.
• Any program other than MCOMP, VCOMP, MMLCV; any bachelor’s program; any year before 2025.
• The “at least 24 units at 8000 level” rule — not selected by the client.
• Incompatible-course warnings — the data is scraped and stored, no rule checks it.
• Live scraping at runtime — the app only reads its own committed snapshot.
• Summer and winter sessions — only S1 and S2 exist here.
• Timetable clashes, class sizes, enrolment itself.

Known limits

• Some prerequisites are too tangled to parse safely. COMP8600’s 2025 and 2026 requisite is a mixed AND/OR expression over ten codes with parentheses (the 2027 page simplifies it to “COMP6670 or COMP3670”, which is checked). The parser only resolves a requisite made of codes joined by one connective throughout; anything more shows as “Check P&C” rather than being guessed at.
• Software Development specialisation (2026, 2027): one optional course list says “a maximum of 12 units”, but the specialisation’s other blocks leave only 6 of its 24 units. It is modelled as 6 in data/overrides/, so a second course from that list is counted as a general elective instead. That is a modelling choice, not a fact read off the page.
• A few specialisation-specific course exclusions are logged as warnings rather than enforced, since the data model has no per-course exclusion field.
• Every hand-written correction lives in data/overrides/ and cites the exact page and sentence it’s based on.

Data

The catalogue comes from ANU’s Programs & Courses website, scraped once and committed as JSON in data/catalogue/. The live app never fetches from that site — it only reads its own snapshot, loaded into SQLite on boot. Snapshot date: 24 September 2026. Rules change; check the program’s own current page before relying on a plan. To refresh, run pnpm scrape — it re-fetches every seeded program and course page and rewrites data/catalogue/*.json (review the diff before committing).