cph: a VS Code judge loop with local testcases and separate submit extensions

The gap cph closes: writing a solution, then rebuilding the test harness by hand

The usual competitive programming loop is a browser tab, a local editor, and a mental checklist of sample cases. Grabbing testcases means copying input and expected output by hand or by writing a script that scrapes the problem page, and every language change means rewriting the compile and run commands. cph exists to collapse that into one loop: the project describes itself as making judging, compiling and downloading problems for competitive programming easy. The user is a contest participant or someone preparing for contests, working alone in an editor, who wants the local feedback loop to match what the platform will do. The extension targets VS Code specifically, and its package manifest requires a VS Code engine of ^1.52.0, so it is not a standalone judge you can call from a terminal or from another editor.

The data path runs through a browser extension, not through the editor alone

cph does not scrape problem statements itself. The documented flow starts with Competitive Companion, a separate browser extension, which the user triggers with a green plus button in the browser toolbar while on a problem page. That step hands the problem to VS Code, and the file opens with the testcases already preloaded. From there cph takes over: it compiles the file, shows compilation errors, and runs the testcases through a judge that understands signals, timeouts and runtime errors. The architecture is a pipe rather than a monolith, browser extension to editor, and that split has a consequence worth stating plainly. If the browser half is missing or the page layout confuses Competitive Companion, there is no testcase data to judge against. The project also documents the fallback path, opening any supported file and pressing Run Testcases to enter testcases manually, and it works locally on problems you wrote yourself, where no platform page exists at all.

Installing the extension and the browser half, then pressing the shortcut

The documented quick start has three install steps and one action. cph itself is installed from the VS Code marketplace, and a manual path exists for people who prefer a .vsix file, published as a release tagged latest-vsix. Competitive Companion is installed separately in the browser, and the file has to be opened in VS Code after that. The action is a keyboard shortcut, Ctrl+Alt+B, which runs the preloaded testcases. The same command exists in the panel as a named VS Code command, and the manifest contributes three commands in total, including one that compiles without running, which is useful when the solution is still a work in progress. The command identifiers are stable enough to bind to your own keymap.

"commands": [
    {
        "command": "cph.runTestCases",
        "title": "%cph.runTestCases.title%",
        "icon": "dist/static/play.png"
    },
    {
        "command": "cph.submitToCodeForces",
        "title": "%cph.submitToCodeForces.title%",
        "icon": "dist/static/play.png"
    },
    {
        "command": "cph.compileWithoutRunning",
        "title": "%cph.compileWithoutRunning.title%"
    }
]

Twelve languages behind one command, and the guides for adding a thirteenth

The same run and judge path covers C++, C, C#, Rust, Go, Haskell, Python, Ruby, Java, Kotlin, JavaScript under Node.js and Cangjie. The design consequence is that the extension, not your build system, decides how a language is compiled and run, so a custom Makefile or a toolchain version of your own choosing is not part of the documented workflow. Adding a language is a supported activity rather than a patch, and the project points contributors at a developer guide section for it. The display layer is localized separately from the compiler layer, with the interface available in English, Simplified Chinese, Traditional Chinese, Korean and Japanese, and the manifest carries the matching package.nls files. A contributor adding a translation is pointed at a translation guide, which keeps the two kinds of change apart.

Custom checkers handle the problems where a plain string compare is wrong

Many contest problems have several valid outputs, or require a numeric answer to a stated precision, and the standard approach of comparing your output to the expected output character by character fails on them. cph addresses that with a custom checker, a special judge implemented as a Python script that the extension uses to evaluate the case. This is the part of the tool that a plain compile and run setup cannot copy, and it is the feature to test first on a problem set that mixes output only tasks with scoring tasks. The checker mechanism also imposes its own constraint: it lives in Python, so a Python interpreter has to be present and the checker script has to be written against the platform's own checker conventions.

Submission is a separate tool, and the split differs by platform

Compiling and judging are cph's own work, but submitting a solution is only partly. The manifest contributes a command named cph.submitToCodeForces, and the feature list links Codeforces auto-submit to a separate project, cph-submit, described as an optional browser extension to install for that purpose. Kattis is handled a different way, by installing the submit client and its config file from the Kattis help page after logging in. So the platform story is not uniform inside cph: one platform is a companion extension, another is an external client with a config file. What all of this buys is that the local loop stays inside the editor, while the actual upload and the platform's own judgement are handled by the tools each site expects.

The extension phones home every few seconds, and the project says so

The telemetry section is short and worth reading before installing anything. To show a live user count, the extension sends a request to the server every few seconds, and the documentation states that no information is sent with the request. That is a claim about payload, not about the request itself, since the request still reveals that the extension is running, from which address, and when. For a tool used during timed contests, that is the kind of behaviour a reader should decide about rather than accept, and the extension can also be used entirely locally on your own problems, where the network path is not needed. The extension also declares sponsorship metadata pointing at a shortened sponsor URL, which is normal for a marketplace extension but is another outbound element in the same package.

Maintenance signals: a rolling vsix release and an unusual version number

The repository is a TypeScript project built with webpack, with separate backend and frontend bundles configured through webpack.config.js and webpack.frontend.config.js, a Jest setup in jest.config.js, and separate src/ and server/ directories. The release history is unusual: the only published release is a rolling build tagged latest-vsix, dated 2026-06-02, while the manifest version field reads 2077.0.0, a joke value rather than a semantic version anyone can use for pinning. That combination means the marketplace is the update channel in practice, and a user who wants a fixed artifact has to record the vsix they installed. The last push to the repository was on 2026-10-16, and the project is not archived. The licence is GPL-3.0-or-later, and the contributing guide asks that an issue be opened before a pull request so the approach can be discussed first, which is a reasonable signal for how much scrutiny a change receives.

Editorial conclusion

cph suits a competitive programmer who wants compile, run and local judging inside VS Code instead of a browser tab, and who can accept a two extension setup for Codeforces submissions. It does not suit a team that needs shared testcases, a web judge, or a single tool that covers submission on every platform, and the request the extension sends to its server every few seconds is a decision to make deliberately rather than by default. Before adopting it, read the telemetry section of the extension documentation, decide whether you want the Competitive Companion browser extension in the loop, and confirm your judging setup works for problems with multiple valid outputs, where the Python custom checker path matters more than the standard compare.

Frequently asked questions

Official sources

1. agrawal-d/cph on GitHub
2. License: GPL-3.0
3. Project website
4. README
5. Releases