Google Coding Competitions — Local

This is a fork of google/coding-competitions-archive, the official archive of problems from Google's Coding Competitions (Code Jam, Kick Start, Hash Code, and friends).

The upstream repo is just the raw archived data — problem statements, analyses, sample data — with no way to browse or actually solve anything. This fork adds emu/, a small local web app on top of that data: clone the repo, run it, and browse every problem organized by year and round, with no account, server, or internet connection required.

Opening a round lands you on a dashboard — problems, per-subtask progress, and your total score for that round — matching how these competitions actually scored things: partial credit per subtask, not pass/fail per problem. From there, most competitions open into a split-pane view (problem statement on the left, code editor on the right) with a "Submit" tab: download a subtask's input, run your own solution against it locally (there's no code execution here, just an editor to write in), then upload or paste the output to get it checked against the real expected answer and scored. Hash Code is the exception to all of this: it never had an online judge at all, just a PDF problem statement and an input dataset you'd process yourself, so its page is a PDF reader plus an "Open dataset" button per problem that reveals the input file in your OS's file explorer (Finder/Explorer/whatever your Linux file manager is) — nothing to score.

This is a work in progress and not fully ready yet. See Status below for what's currently browsable.

Getting started

cd emu
npm install
npm run dev

Then open http://localhost:3000. The app reads problem data straight out of the competition folders at the repo root — no separate data setup needed.

Status

Every competition in this archive is browsable now:

• Kick Start — all years, including practice rounds
• Code Jam — all years, including qualification/regional rounds and the 2008 regional semifinals
• Code Jam to I/O for Women — all years
• Distributed Code Jam — all years, including the 2015 online/practice rounds; a few 2018 Finals problems ship a commented reference solution instead of a written analysis, and are shown as such
• Farewell Rounds — the 2023 send-off event (Rounds A–D)
• Hash Code — all years, PDF statement + "open dataset" per problem instead of the usual code editor (see above)

Nothing left unwired, but plenty of rough edges remain — see the open questions the app raised along the way, which are worth revisiting:

• A few Distributed Code Jam problems (2018 Finals) show a raw commented C++ solution as a stand-in for a missing written analysis.
• Distributed Code Jam's statement/analysis pages are unstyled HTML fragments (no CSS at all), unlike every other competition's polished pages.
• Problem/round titles for Distributed Code Jam and Hash Code are derived from filenames, not real metadata, so they're occasionally imperfect (e.g. an acronym rendered as "Rps" instead of "RPS").
• Grading is whitespace-normalized exact-token matching against the archived .ans files — it doesn't know about per-problem custom output validators (e.g. float tolerance), so a numerically-correct answer with different formatting than the reference could be marked wrong. Only problems with static secret test data are gradable at all — Distributed Code Jam (no data/ folder) and interactive-judge problems (a custom live judge, not fixed answer files) show "no judge data" instead.
• Scores live in your browser's localStorage, per problem/subtask — nothing is synced anywhere, so clearing site data resets your progress.

Project structure

.
├── codejam/              archived problem data, inherited from the
├── codejam_to_io/        upstream repo — one folder per competition,
├── distributed_codejam/  each split by year and round
├── farewell/
├── hashcode/
├── kickstart/
│
├── docs/                 contributing guide, code of conduct, this README's assets
│
└── emu/                  the local web app (Next.js) that browses the
                           folders above — see emu/README.md for app-level detail

Most problem folders (e.g. kickstart/2022/round_a/<problem>/) follow the same shape: a problem.yaml with metadata, a problem_statement/ with the HTML statement and analysis, and a data/ folder with sample and secret test data — the secret data is split into subtask1/, subtask2/, etc., each an input/answer pair (1.in/1.ans) plus a testdata.yaml giving that subtask's accept_score; this is exactly what the round dashboard and "Submit" tab grade against. Two competitions differ: Distributed Code Jam skips problem.yaml entirely (bare statement.html/analysis.html, titles derived from folder names), and Hash Code has no per-problem folders at all — just one combined PDF per round (hashcode/<round>.pdf) plus one flat input-dataset file per problem (hashcode/<round>/<dataset>). emu/ doesn't modify any of this data — it only reads it.

About the original archive

Problem data for the following contests is included, straight from upstream:

• Distributed Code Jam
• Code Jam
• Code Jam to I/O for Women
• Farewell Rounds
• Hash Code
• Kick Start

The statement and analysis HTML files may have CSS or JS dependencies that aren't included here. Some of the custom judging code may also have library dependencies that aren't provided, so it may not run as-is — but all problem-specific logic is there, which is enough to reproduce the behavior (and is exactly what emu/ does for rendering statements).

Some analysis for Distributed Code Jam problems couldn't be included, so a commented solution was provided instead.

There's no plan (upstream or here) to add data for other contests, like Code Jam for Europython or local/regional Code Jams.

License

Apache License 2.0 — see LICENSE.