Forty-five original systems-algorithm problems derived from the engineering of a deterministic, browser-capable WebAssembly online judge.
The collection is consumed by WASM OJ Forge and is also a documented public format for compatible browser judges.
Every problem contains complete Traditional Chinese and English statements and editorials, explicit constraints, deterministic text I/O, sample and adversarial cases, a validator, generator, independent oracle, and optimal reference solutions in C17, C++20, Rust, Go, Python, JavaScript, and TypeScript.
Repository and GitHub API consumers start from catalog.json. It explicitly
lists every problem manifest; each manifest explicitly lists all localized documents, programs,
solutions, and tests, so clients never infer paths from the filesystem layout. See
docs/MANIFEST.md.
learning-path.json defines the learner-facing order and localized track
names separately from those stable manifest identities. The current path groups related topics,
then orders each group from lower prerequisite load to more advanced techniques. This keeps
calibration evidence and source paths stable while allowing display numbers to evolve.
Browser judges should use collection/index.json. The small index
contains localized display and track metadata, explicit localized statement paths, and one
integrity-addressed bundle descriptor per problem.
Clients can render the catalog immediately, then fetch full documents and cases only when a
learner opens that problem. See docs/BROWSER_COLLECTION.md.
Scoring is cost-first. Every stored case runs once and is evaluated against cumulative baseline,
efficient, and optimal instruction/memory policies. The optimal instruction budget is derived
from the C, C++, Rust, and Go worst-case average with 5% headroom and deterministic readable
rounding. Calibration evidence lives under calibration/forge-v1.
make verifyRegenerate browser artifacts after any problem content changes:
make collection
make verify