OI Judge

简体中文 | English

OI Judge is a VS Code extension for local OI-style judging and problem authoring.

Positioning: Local OI-style judging + problem authoring toolkit for VS Code.

Marketplace extension ID: Hzq.oijudge

Repository: https://github.com/Hzq-0304/OIjudge.git

VSIX package name format: oijudge-<version>.vsix

The OI Judge icon represents comparing program output with the expected answer and producing a judging verdict.

For compatibility with earlier versions, internal command IDs and settings may still use the oijudger prefix, such as oijudger.language.

Current features:

• Store workspace data in .vscode/.OIJudge/
• Initialize workspace config at .vscode/.OIJudge/config.json and per-problem data under .vscode/.OIJudge/problems/<problemId>/
• Create a problem without immediately binding a source file
• Bind a local statement file (.md, .pdf, or .txt) to a problem
• Add one or more C++ programs to a problem and choose a default program
• Add multiple samples under .vscode/.OIJudge/samples or .vscode/.OIJudge/problems/<problemId>/samples
• Add samples by pasting text or selecting input/output files
• Batch add samples from a folder by matching input and answer suffixes
• Delete samples from the OI Judge sidebar
• Set time and memory limits
• Automatically set Windows MinGW/g++ stack size from the memory limit
• Detect or select a local C++ compiler
• Compile the active C++ file with g++
• Run all configured samples
• Compare standard output
• Save user output and .vscode/.OIJudge/outputs/report.json
• Show an OI Judge sidebar with current file, limits, sample status, and quick actions
• Open report and sample detail pages from the sidebar
• Manage Subtasks, scoring, STD-generated answers, generators, stress tests, testcase export, and custom checkers
• Switch UI text with oijudger.language (auto, en, zh)
• Manage multiple problems in one workspace with .vscode/.OIJudge/config.json
• Keep legacy .oitest/problems.json and single-problem .oitest/config.json data intact and import or migrate it when needed

Data directory:

• .vscode/.OIJudge/ is the OI Judge workspace data directory.
• .vscode/.OIJudge/config.json stores the current workspace problem list and shared OI Judge data.
• Per-problem samples, outputs, checker builds, generated answers, exported metadata, and stress records live below .vscode/.OIJudge/.
• Legacy .oitest/ data is treated only as a compatibility and migration source. OI Judge may copy or import from it, but it does not use .oitest/ as the current primary data directory.

Problem workflow:

• OI Judge: Create Problem creates a problem entry and its .vscode/.OIJudge/problems/<problemId>/ folders without requiring a source file.
• OI Judge: Bind Statement links a statement file. OI Judge stores the original file path only; it does not copy, modify, or delete the statement file.
• OI Judge: Add Program To Problem links a C++ program. Programs are path references only; source files are not copied into OI Judge's internal data directory.
• OI Judge: Set Default Program chooses the program used by Run All Samples.
• OI Judge: Run Samples With Program lets you temporarily choose any linked or newly selected .cpp file for one run.
• OI Judge: Add Problem From Current File and OI Judge: Add Problem From File still work as shortcuts: they create a problem and set the selected file as the default program.

Tree view:

• Problem nodes are collapsed by default to keep the OI Judge sidebar compact after VSCode restarts.
• Expand a problem manually to view Statement, Programs, Limits, Samples, and Actions.
• Samples and Actions are also collapsed by default, which keeps large multi-sample problems easier to scan.
• Manage Workspace is available from the OI Judge view title bar or the Command Palette.
• Click the Default Program, Compiler, or C++ Standard node under a problem to edit the corresponding setting.
• Click the Time, Memory, or Stack node under a problem's Limits section to edit the corresponding limit. These limit editors are not duplicated in the Actions section.

Sample storage:

• Paste manually: OI Judge stores the input and expected output inside .vscode/.OIJudge. This is best for small samples.
• New managed samples use .in for input and .out for expected output by default. Existing .ans sample answers remain supported and are not renamed.
• Select input/output files: OI Judge stores the original absolute file paths and does not copy the files. This is best for large data files or existing local test data.
• External samples depend on the original files. If an external input or answer file is moved or deleted, the sample is shown as Missing and skipped during judging.
• Deleting a managed sample removes the OI Judge-owned .vscode/.OIJudge sample files and generated outputs.
• Deleting an external sample removes only the OI Judge sample record and generated outputs. The original input and answer files are never deleted.

Batch add samples:

• Run OI Judge: Batch Add Samples.
• Enter the input file suffix. The default is .in; in is normalized to .in.
• Enter the answer file suffix. The default is .out; ans is normalized to .ans.
• Select a samples folder.
• OI Judge scans only the first level of that folder and matches files by basename + inputSuffix and basename + answerSuffix. With the default .out answer suffix, OI Judge falls back to .ans when a same-name .out file is missing.
• For example, 1.in with 1.out and 2.in with 2.out are added as two samples.
• Batch-added samples are external samples: OI Judge stores absolute paths only and does not copy or modify the files.
• Inputs without matching answer files and duplicate sample pairs are skipped and summarized.

Sample names:

• Manually pasted samples keep the default Sample x name.
• Samples added from files use the input file basename as the visible sample name.
• For example, book3.in with book3.ans is shown as book3; 1.in with 1.out is shown as 1.
• Batch-added samples use each matched basename as the sample name.
• OI Judge still uses a stable internal id and index for output folders such as outputs/sample-7/, so display names do not affect deletion, reports, or diff paths.
• If a sample name already exists, OI Judge appends (2), (3), and so on.

Sample viewing:

• Sample input, expected output, and run results open in the native VSCode text editor.
• Output differences open with the native VSCode Diff Editor.
• New per-problem runs keep pure stdout in .vscode/.OIJudge/problems/<problemId>/outputs/sample-x/useroutput.txt for judging and diff.
• Run Result opens .vscode/.OIJudge/problems/<problemId>/outputs/sample-x/run-result.txt, which contains program stdout, program stderr, and runtime diagnostics such as status, exit code, signal, and Runtime Error details.
• The standalone Open Stderr sample action has been removed; stderr is shown through Run Result.
• Checker output is separate: checker stdout and stderr are merged into checker-output.txt and opened with Checker Output.
• Diff and judging still use pure useroutput.txt; stderr is never appended to it.
• New per-problem runs also keep stderr.txt and diff.txt next to the output for diagnostics and compatibility.
• Older 1.out, 1.err, and 1.diff outputs remain readable for compatibility.

Timing note: sample time only measures the user executable process. On Windows, sample time includes process startup and pipe I/O overhead, so very small programs may still show tens of milliseconds.

Windows stack size:

• Deep recursive programs on Windows may exit with 0xC00000FD, which is a stack overflow exception.
• By default, OI Judge follows the problem memory limit and adds a MinGW/g++ linker flag when compiling on Windows.
• For memoryMb = 256, the generated flag is -Wl,--stack,268435456.
• Click the Stack node under a problem's Limits section, or run OI Judge: Set Stack Size, to choose Follow Memory Limit, Custom Stack Size, or Disable Auto Stack Size.
• The stack flag is generated at compile time and is not repeatedly inserted into compile.args.
• If auto stack size is enabled, an existing -Wl,--stack,... argument is replaced by the current setting. If auto stack size is disabled, OI Judge does not add a stack flag.
• This mainly targets Windows + MinGW/g++. Linux/macOS judging environments usually control stack through the runner or system limits, and avoiding very deep recursion is still the safest algorithmic choice.

Runtime Error Explanation:

• OI Judge explains common Runtime Error results from the process exit code or POSIX signal.
• Runtime Error names use common OI/OJ-style English descriptions, such as:
  • Stack overflow
  • Access violation
  • Integer divide by zero
  • Floating point exception
  • Segmentation fault
• Chinese UI keeps the English Runtime Error title and adds Chinese descriptions, possible causes, and suggestions below it.
• Common Windows examples:
  • 0xC00000FD: Stack overflow
  • 0xC0000005: Access violation
  • 0xC0000094: Integer divide by zero
• Common Linux/macOS signals:
  • SIGSEGV: Segmentation fault
  • SIGFPE: Floating point exception
  • SIGABRT: Aborted
• The explanation is a diagnostic hint, not a final proof. Always combine it with the input file, stderr, the reproduction command, and a debugger when needed.

Judge Mode:

• OI Judge supports three judge modes for each problem:
  • Text Compare: strict stdout and answer comparison.
  • Text Compare (ignore trailing whitespace and final newlines): the default OI-style comparison. It ignores trailing spaces/tabs on each line and extra final newlines, but does not ignore leading spaces, inner spaces, or middle blank lines.
  • Custom Checker
• In text compare modes, Checker-related actions are hidden from the problem Actions section to keep the sidebar compact.
• Click the Judge Mode node under a problem to switch between strict text compare, OI-style text compare, and custom checker.
• Custom checker mode enables Checker actions and supports:
  • Testlib Checker
  • Plain Checker
• If you run OI Judge: Set Checker while a problem is still in a text compare mode, OI Judge asks whether to switch to custom checker first.
• Switching back to a text compare mode does not delete the saved checker configuration, so you can switch back later without reselecting the checker.

Function-style Judge

Function-style judge supports grader-based tasks where the grader provides main() and the contestant solution implements required functions. OI Judge compiles the grader and solution together, then runs the generated executable against normal samples and uses the existing text compare or checker flow.

Configure a problem with mode: "function" and functionStyle.grader / functionStyle.solution, then run OI Judge: Run Function-style Judge.

I/O Interactive Judge

I/O Interactive Judge supports the common solution + interactor model. OI Judge compiles the solution and interactor separately, starts both processes for each testcase, pipes solution.stdout to interactor.stdin and interactor.stdout to solution.stdin, and records a bounded transcript for the report.

Configure a problem with mode: "interactive" and interactive.solution / interactive.interactor, then run OI Judge: Run I/O Interactive Judge.

{
  "mode": "interactive",
  "interactive": {
    "solution": "solution.cpp",
    "interactor": "interactor.cpp",
    "interactorArgs": ["{input}", "{answer}"],
    "transcriptLimitBytes": 262144
  }
}

The interactor receives the testcase input path through {input} and the answer path through {answer}. The solution's stdout is protocol traffic, so interactive mode does not run the normal text compare or checker pipeline. Interactor exit code 0 means Accepted, 1 means Wrong Answer, 2 means Presentation Error, and other non-zero codes are reported as Interactor Error.

testlib-like interactors

For Codeforces / Polygon style interactors, use the testlib preset:

{
  "mode": "interactive",
  "interactive": {
    "solution": "solution.cpp",
    "interactor": "interactor.cpp",
    "interactorPreset": "testlib",
    "interactorArgs": ["{input}", "{output}", "{answer}"],
    "useTestlib": true,
    "testlibHeader": "third_party/testlib.h"
  }
}

{input} is the testcase input file, {output} is a temporary file created by OI Judge for the interactor, and {answer} is the testcase answer file. I/O Interactive Judge does not automatically select a bundled testlib.h for interactors; provide the header in your workspace and set testlibHeader or testlibIncludeDirs explicitly. This preset improves compatibility with testlib-like argument conventions, but it does not promise full testlib compatibility.

See examples/interactive/testlib-like-double/ for a complete testlib-like argument example with {input}, {output}, and {answer}. The default example uses useTestlib: false and does not require official testlib.h; the separate oijudge.testlib.config.json is a reference for users who provide third_party/testlib.h themselves.

Minimal guess-number interactor

A typical interactive task can be modeled by passing the original testcase input to the interactor through {input}. For example, a guess-number interactor can read n secret, send n to the solution, answer each flushed guess with 1, -1, or 0, and return exit code 0 for Accepted, 1 for Wrong Answer, 2 for Presentation Error, and 3 for Interactor Error.

Example

See examples/interactive/guess-number/ for a minimal guess-number interactive task. The interactor receives the testcase input path through {input}, communicates with the solution through stdin/stdout, and decides the verdict with its exit code.

I/O Mode:

• Each problem can use either Standard IO or File IO.
• Standard IO is the default: OI Judge feeds the sample input through stdin and captures stdout as useroutput.txt.
• File IO is for programs that use files such as problem.in and problem.out.
• In File IO mode, OI Judge creates an isolated temporary run directory for every sample:
  • .vscode/.OIJudge/problems/<problemId>/outputs/sample-<index>/run/
  • writes the sample input to the configured input file name
  • runs the executable with cwd set to that run directory
  • reads the configured output file after the process exits
  • saves that file content back to the standard useroutput.txt
• Diff, normal compare, testlib Checker, Plain Checker, and Result Panel all continue to use useroutput.txt.
• Program stdout in File IO mode is diagnostic only; it is not used as the judged output if the configured output file is missing.
• File names must be simple names such as problem.in; absolute paths, folders, and .. are rejected.
• OI Judge never creates the configured input/output files in the source directory, workspace root, sample directory, or checker directory.

Example File IO program:

#include <bits/stdc++.h>
using namespace std;

int main() {
    freopen("problem.in", "r", stdin);
    freopen("problem.out", "w", stdout);
    int a, b;
    cin >> a >> b;
    cout << a + b << "\n";
}

Setter Mode:

• Setter Mode is disabled by default.
• Enable it in VS Code settings with:
  • oijudger.setterMode.enabled
• When enabled, OI Judge shows setter tools for each problem:
  • Select STD
  • Open STD
  • Clear STD
  • Generate Answer with STD
  • Generate All Sample Answers with STD
  • View Current Answer
  • View Generated Answer
  • Compare Generated Answer
  • Apply Generated Answer
  • Apply All Generated Answers
  • Delete Generated Answer
  • Add Generator / Open Generator / Remove Generator
  • Add Global Generator Input / Open Global Generator Input / Remove Global Generator Input
  • Bind Generator and Generator Input to a Subtask
  • Set Sample Name
• STD is the standard solution for problem setters. It is saved separately from the default program used for judging the current solution.
• OI Judge can run the bound STD to generate one sample answer or all sample answers.
• If the current answer path is empty, missing, or points to an empty file, generated STD output is written directly as the current answer.
• If the current answer already has content, generated STD output is staged as a pending generated answer in setter.generatedAnswers.
• Pending generated answers can be opened, compared with the current answer, applied one by one, applied all at once, or deleted.
• Sample names edited in Setter Mode are display/data-configuration names only. They do not change sample.id, sample.index, real input/answer files, or outputs/sample-<index>/.
• Setter data configuration is stored in setter.dataCases.
• OI Judge supports problem-level generator bindings in setter.generator.generators, global generator inputs in generatorInputs, and Subtask generator input bindings.
• Generator-backed sample input generation is available from the problem and Subtask actions.

Subtasks and scoring:

• A problem can create, rename, and delete Subtasks. Deleting a Subtask removes the grouping record only; it does not delete the samples.
• Samples can be moved into a Subtask or back to the ungrouped area.
• Each Subtask stores sampleIds, an optional lastResult, optional generator binding, optional generator input, and a scoring mode.
• Subtasks can be judged independently. The sidebar reports pass/fail state and passed/total for the last Subtask run.
• The problem total score defaults to 100 and can be changed with OI Judge: Set Total Score.
• Each sample can have a manual testcase score. Samples without a manual score split the remaining score automatically.
• sum Subtasks score each accepted sample independently.
• bundle Subtasks award the whole Subtask score only when all samples in the Subtask pass.
• Subtask Skip can be enabled from configuration to skip the rest of a failed bundle Subtask and to skip dependent Subtasks when prerequisites do not pass. The default is disabled, so existing configs continue to run every testcase.

{
  "subtaskSkip": {
    "enabled": true,
    "skipRemainingCasesOnFailure": true,
    "skipDependentSubtasks": true
  },
  "subtasks": [
    {
      "id": "subtask1",
      "name": "Subtask 1",
      "sampleIds": ["sample-1", "sample-2"],
      "scoringMode": "bundle"
    },
    {
      "id": "subtask2",
      "name": "Subtask 2",
      "sampleIds": ["sample-3", "sample-4"],
      "scoringMode": "bundle",
      "dependsOn": ["subtask1"]
    }
  ]
}

If a testcase in subtask1 fails, remaining cases in that bundle Subtask are reported as Skipped. If subtask1 does not pass, subtask2 is skipped with a dependency reason in the report. Missing dependency ids and dependency cycles are reported as configuration errors instead of running indefinitely.

Stress Test:

• OI Judge: Run Stress Test supports two stress-test workflows.
• Split-file stress test: Generator + STD + Solution. Use this when you have separate generator, standard solution, and solution files. OI Judge runs the generator, feeds the generated input to both programs, and compares their outputs.
• Single-file stress test: contest-style stress.cpp. Use this when you wrote a self-contained stress program. The program should generate tests and compare answers internally. OI Judge only compiles and runs it, then stores stdout, stderr, and summary files.
• Use OI Judge: Stress Test Current Code to run the most recently focused open .cpp file as the solution with the current problem's configured generator and STD.
• Use $(debug-stop) Stop Stress Test to interrupt an active stress test, especially for single-file contest-style stress programs without their own stopping condition.
• Failed cases are saved under .vscode/.OIJudge/stress/ with input, STD output, tested output, stderr files, and summary.json.
• The Stress Records view can open saved files, add a failed case to samples, rerun a failed case, refresh records, and reveal the session folder.

Save Failed Cases as Samples

When a judge report or stress test record contains a failed case, use Save as Sample to save the input and expected output into the samples/ directory for repeated debugging.

Environment Check:

• Run OI Judge: Check Environment from the Command Palette or the OI Judge sidebar toolbar.
• The check covers platform information, temporary directories with spaces, compiler discovery, C++17 compilation, executable launch, stdin/stdout, file IO, native runner availability, time/memory sanity, and process stop support.
• OI Judge writes a detailed plain-text report to the OI Judge OutputChannel and offers buttons to open the report, copy the report, or open the output panel.
• Probe sources, executables, inputs, outputs, and native runner helper files are created under a temporary directory and cleaned up after the check.
• If a required tool is missing, the report includes platform-specific suggestions such as installing MinGW-w64 on Windows, Xcode Command Line Tools on macOS, or g++ on Linux.

Platform Support

OI Judge supports Windows, macOS, and Linux. See Platform Support for compiler requirements, tested features, and troubleshooting tips.

If OI Judge does not work on your machine, run:

OI Judge: Check Environment

and share the generated report.

Testcase export:

• OI Judge: Export Testcases copies or moves sample input/output files and writes an .OIJudge/config.json export record.
• Luogu export can generate config.yml.
• Polygon export can generate an OI Judge import plan in polygon.json plus a short README.
• LemonLime export can generate contest.cdf, data/<problemName>/, source/, and a short README.
• When scores or bundled Subtasks are configured, export metadata carries sample scores and bundled Subtask grouping where the target format supports it.

Problem Package Export:

• OI Judge: Export Problem Package is separate from Testcase Export. Testcase Export focuses on testcase files and platform metadata.
• Problem Package Export copies complete authoring material into a directory for backup, migration, sharing, and future import support.
• The exported directory includes statement, source programs, STD, checker, generators, samples, Subtasks, scores, and an OI Judge config snapshot.
• The first version exports a folder with oijudge-package.json and README.txt; it does not create a zip archive.

Problem Package Import:

• OI Judge: Import Problem Package imports a complete problem package directory exported by OI Judge.
• Import creates a new problem in the current workspace.
• Package files are copied into the current workspace under .vscode/.OIJudge/; the imported problem does not depend on the original package directory.
• The first version supports directory import only; zip archives are not supported.

Testlib Checker:

• OI Judge supports testlib-style checkers for per-problem judging.
• Run OI Judge: Set Checker and choose Testlib checker, then select a local checker.cpp.
• A typical checker includes #include "testlib.h" and calls registerTestlibCmd(argc, argv).
• OI Judge runs the checker as:

checker.exe input.txt useroutput.txt answer.txt

• testlib.h is resolved in this order:
  • the same folder as checker.cpp
  • the workspace root
  • .vscode/.OIJudge/tools/testlib/testlib.h
  • a custom path recorded in the checker config
• OI Judge can install the bundled testlib.h shipped with the extension, or import a local copy selected by the user.
• User-provided copies still have higher priority than the bundled copy once installed into the workspace.
• OI Judge does not download or generate testlib.h. If it is missing, run OI Judge: Import testlib.h.
• When bundled resources are available, OI Judge: Import testlib.h offers:
  • Install bundled testlib.h
  • Import testlib.h from local file
• Bundled source and license details are preserved in resources/testlib/README.md and resources/testlib/LICENSE.
• Checker executables are built under .vscode/.OIJudge/problems/<problemId>/checker/.
• Checker stdout and stderr are merged into one file beside each sample output as checker-output.txt.
• testlib checkers often print verdict details to stderr; users can still view all checker information through the single Checker Output action.
• Plain Checker verdict parsing still uses only the configured verdict line from the original stdout. Merged stderr content is saved for viewing, but it is not parsed as the verdict.
• Verdict rules:
  • checker exit code 0 => AC
  • checker exit code 1 => WA
  • Windows NTSTATUS exception codes such as 0xC0000135 => Checker Error
  • checker compile/run/timeout failure => Checker Error
• Windows DLL note:
  • If a checker exits with code 3221225781 / 0xC0000135, it usually means checker.exe failed to start because a runtime DLL is missing, not that the checker judged WA.
  • Common missing DLLs include MinGW libstdc++-6.dll, libgcc_s_seh-1.dll, and libwinpthread-1.dll.
  • OI Judge tries to compile checkers with static linking for MinGW/g++ and prepends the compiler bin directory to the checker process PATH.
  • You can also add the MinGW bin directory to PATH, rebuild the checker with static linking, or put the missing DLL next to checker.exe.
• Normal compare is unchanged when no checker is enabled.
• Plain Checker supports numeric score output through its stdout verdict line.

Plain Checker:

Plain Checker is a simple custom checker that does not depend on testlib.h.

OI Judge runs it with the same arguments as a testlib checker:

checker.exe input.txt useroutput.txt answer.txt

The default protocol reads the last non-empty line of stdout. That line must be one of:

• AC
• WA
• a numeric score

You can also run OI Judge: Set Plain Checker Protocol to configure:

• whether the verdict is read from the first or last non-empty stdout line
• the accepted token, default AC
• the wrong-answer token, default WA

Examples:

AC

This marks the sample as accepted.

WA

This marks the sample as wrong answer.

37.5

This returns a score of 37.5. OI Judge shows a question mark icon and displays 37.5 on the right side. It does not mark the sample as accepted or wrong.

Important: if you want WA, output the configured wrong-answer token. If you output 0, OI Judge treats it as score 0, not as WA. If you output 100, OI Judge treats it as score 100, not as AC.

Under the default protocol, invalid verdict lines include:

• Accepted
• Wrong Answer
• 75%
• score: 75
• 通过

These are reported as Checker Error.

Custom protocol example:

• Verdict line: First non-empty line
• Accepted token: OK
• Wrong answer token: NG

OK
details...

This marks the sample as accepted.

NG
wrong answer details...

This marks the sample as wrong answer.

37.5
matched 15 cases

This shows a question mark icon and score 37.5. Numeric verdict lines are always treated as scores, so configure accepted/wrong tokens as non-numeric strings.

Minimal Plain Checker example:

#include <bits/stdc++.h>
using namespace std;

int main(int argc, char** argv) {
    if (argc < 4) {
        cout << "WA\n";
        return 0;
    }

    ifstream user(argv[2]);
    ifstream ans(argv[3]);

    long long a, b;
    user >> a;
    ans >> b;

    cout << (a == b ? "AC" : "WA") << '\n';
    return 0;
}

Score example:

#include <bits/stdc++.h>
using namespace std;

int main(int argc, char** argv) {
    ifstream user(argv[2]);
    ifstream ans(argv[3]);

    int correct = 0, total = 10;
    for (int i = 0; i < total; i++) {
        int x, y;
        if (!(user >> x)) break;
        ans >> y;
        if (x == y) correct++;
    }

    cout << fixed << setprecision(1) << correct * 10.0 << '\n';
    return 0;
}

If the last line is 70.0, OI Judge shows a question mark icon and score 70.0.

Commands:

• Basic judging: Init Problem, Add Sample, Run All Samples, Set Time Limit, Set Memory Limit, Set Stack Size, Select C++ Compiler, Open Last Report, Open Result Panel, Clear Outputs.
• Problem workflow: Create Problem, Add Problem From Current File, Add Problem From File, Import Legacy Single Problem, Delete Problem, Add Program, Set Default Program, Run With Program, Bind Statement, Open Statement, Unbind Statement.
• Samples: Add Problem Sample, Add Problem Sample From Files, Batch Add Samples, Open Sample Input, Open Expected Output, Open Run Result, Open Diff, Delete Sample.
• Setter mode: Select STD, Open STD, Clear STD, Auto Generate Output with STD, Generate Answer with STD, Generate All Sample Answers with STD, View Current Answer, View Generated Answer, Compare Generated Answer, Apply Generated Answer, Apply All Generated Answers, Delete Generated Answer.
• Generators: Add Generator, Open Generator, Remove Generator, Select Generator, Open Generator, Clear Generator, Add Global Generator Input, Open Global Generator Input, Remove Global Generator Input.
• Subtasks and scoring: Create Subtask, Rename Subtask, Delete Subtask, Move to Subtask, Run Subtask, Set Subtask Scoring Mode, Set Total Score, Set Testcase Score, Clear Testcase Score, Bind Generator, Bind Generator Input, Open Generator Input, Clear Generator Input, Generate Subtask Sample Input.
• Stress and export: Run Stress Test, Refresh Stress Records, Open Stress File, Add to Samples, Rerun This Case, Reveal Session Folder, Export Testcases, Export Problem Package, Copy freopen Input Snippet.
• Checker: Set Checker, Set Plain Checker Protocol, Clear Checker, Open Checker, Import testlib.h, Open testlib.h, Open Checker Output.

The default compiler command is g++. You can edit .vscode/.OIJudge to adjust per-problem compiler flags.

中文介绍

OI Judge 是一款 VS Code 扩展，用于本地 OI 风格评测，也面向 OI 出题辅助流程。它可以帮助选手快速编译、运行、比较样例，也可以帮助出题人管理题目、样例、Subtask、STD、Generator、Checker、对拍、测试点导出和评测报告。

项目定位：本地 OI 风格评测 + 出题辅助工具。

Marketplace 扩展 ID：Hzq.oijudge

仓库地址：https://github.com/Hzq-0304/OIjudge.git

VSIX 安装包文件名格式：oijudge-<version>.vsix

OI Judge 的图标含义是：比较程序输出与期望答案，并给出评测结果。

为了兼容早期版本，内部命令 ID 和设置项可能仍使用 oijudger 前缀，例如 oijudger.language。

当前功能

• 将工作区数据保存在 .vscode/.OIJudge/。
• 在 .vscode/.OIJudge/config.json 初始化工作区配置，并在 .vscode/.OIJudge/problems/<problemId>/ 下保存按题目的数据。
• 创建题目时不要求立刻绑定源文件。
• 将本地题面文件（.md、.pdf 或 .txt）绑定到题目。
• 为一个题目添加一个或多个 C++ 程序，并选择默认程序。
• 在 .vscode/.OIJudge/samples 或 .vscode/.OIJudge/problems/<problemId>/samples 下添加多个样例。
• 通过粘贴文本或选择输入/输出文件添加样例。
• 从文件夹批量添加样例，并按输入和答案后缀匹配文件。
• 从侧边栏删除题目；只删除 OI Judge 题目记录和 .vscode/.OIJudge/problems/<problemId>/ 内部托管目录，不删除源码、题面或外部样例文件。
• 从 OI Judge 侧边栏删除样例。
• 设置时间限制和内存限制。
• 在 Windows MinGW/g++ 下根据内存限制自动设置栈大小。
• 检测或选择本地 C++ 编译器。
• 使用 g++ 编译当前 C++ 文件。
• 运行所有已配置样例。
• 比较标准输出。
• 保存用户输出和 .vscode/.OIJudge/outputs/report.json。
• 显示 OI Judge 侧边栏，其中包含当前文件、限制、样例状态和快捷操作。
• 从侧边栏打开评测报告和样例详情页。
• 从测试点复制可用于本地调试的 freopen 输入代码。
• 管理 Subtask、计分、STD 生成答案、Generator、对拍、测试点导出和自定义 Checker。
• 通过 oijudger.language（auto、en、zh）切换界面语言。
• 在一个工作区内通过 .vscode/.OIJudge/config.json 管理多个题目。
• 保留旧版 .oitest/problems.json 和单题 .oitest/config.json 数据，并在需要时导入或迁移。

数据目录

• .vscode/.OIJudge/ 是 OI Judge 的工作区数据目录。
• .vscode/.OIJudge/config.json 保存当前工作区题目列表和共享的 OI Judge 数据。
• 按题目的样例、输出、Checker 构建产物、生成答案、导出元数据和对拍记录都位于 .vscode/.OIJudge/ 下。
• 旧目录 .oitest/ 只是 legacy 兼容迁移来源。OI Judge 可以从中复制或导入数据，但不会把 .oitest/ 作为当前主数据目录。

题目工作流

• OI Judge: Create Problem 会创建题目记录和对应的 .vscode/.OIJudge/problems/<problemId>/ 文件夹，不要求先选择源文件。
• OI Judge: Bind Statement 会链接题面文件。OI Judge 只保存原始文件路径，不复制、不修改、也不删除题面文件。
• OI Judge: Add Program To Problem 会链接 C++ 程序。程序只是路径引用，源文件不会被复制进 OI Judge 内部数据目录。
• OI Judge: Set Default Program 会选择 Run All Samples 使用的程序。
• OI Judge: Run Samples With Program 可以为一次运行临时选择任意已链接或新选择的 .cpp 文件。
• OI Judge: Add Problem From Current File 和 OI Judge: Add Problem From File 仍然作为快捷入口可用：它们会创建题目，并把选中的文件设为默认程序。

树视图

• VS Code 重启后，题目节点默认折叠，以保持 OI Judge 侧边栏紧凑。
• 手动展开题目即可查看 Statement、Programs、Limits、Samples 和 Actions。
• Samples 和 Actions 也默认折叠，便于浏览多样例题目。
• 点击题目下的 Default Program、Compiler 或 C++ Standard 节点，可以编辑对应设置。
• 点击题目 Limits 区域下的 Time、Memory 或 Stack 节点，可以编辑对应限制。这些限制编辑入口不会在 Actions 区域重复显示。

样例存储

• 手动粘贴：OI Judge 会把输入和期望输出存入 .vscode/.OIJudge。这适合较小样例。
• 新建托管样例默认使用 .in 作为输入文件，使用 .out 作为期望输出文件。已有旧 .ans 样例答案会继续兼容，但不会被重命名。
• 选择输入/输出文件：OI Judge 保存原始绝对文件路径，不复制文件。这适合大数据文件或已有本地测试数据。
• 外部样例依赖原始文件。如果外部输入或答案文件被移动或删除，样例会显示为 Missing，并在评测时跳过。
• 删除托管样例会删除 OI Judge 管理的 .vscode/.OIJudge 样例文件和生成输出。
• 删除外部样例只会删除 OI Judge 的样例记录和生成输出，原始输入/答案文件永远不会被删除。

批量添加样例

• 运行 OI Judge: Batch Add Samples。
• 输入输入文件后缀。默认是 .in；in 会被规范化为 .in。
• 输入答案文件后缀。默认是 .out；ans 会被规范化为 .ans。
• 选择样例文件夹。
• OI Judge 只扫描该文件夹第一层，并按 basename + inputSuffix 和 basename + answerSuffix 匹配文件。使用默认 .out 答案后缀时，如果同名 .out 缺失，OI Judge 会回退查找 .ans。
• 例如，1.in 与 1.out、2.in 与 2.out 会被添加为两个样例。
• 批量添加的样例是外部样例：OI Judge 只保存绝对路径，不复制或修改文件。
• 没有匹配答案文件的输入，以及重复的样例对，会被跳过并汇总提示。

样例名称

• 手动粘贴的样例保留默认 Sample x 名称。
• 从文件添加的样例使用输入文件 basename 作为显示名称。
• 例如，book3.in 与 book3.ans 会显示为 book3；1.in 与 1.out 会显示为 1。
• 批量添加的样例使用每组匹配文件的 basename 作为样例名。
• OI Judge 仍然使用稳定的内部 id 和 index 管理输出目录，例如 outputs/sample-7/，因此显示名称不会影响删除、报告或 diff 路径。
• 如果样例名已存在，OI Judge 会追加 (2)、 (3) 等后缀。

样例查看

• 样例输入、期望输出和运行结果会在 VS Code 原生文本编辑器中打开。
• 输出差异会使用 VS Code 原生 Diff Editor 打开。
• 新的按题目运行流程会把纯 stdout 保存在 .vscode/.OIJudge/problems/<problemId>/outputs/sample-x/useroutput.txt，用于评测和 diff。
• Run Result 会打开 .vscode/.OIJudge/problems/<problemId>/outputs/sample-x/run-result.txt，其中包含程序 stdout、程序 stderr，以及状态、退出码、信号、Runtime Error 详情等运行诊断信息。
• 独立的 Open Stderr 样例操作已移除；stderr 会通过 Run Result 查看。
• Checker 输出是独立的：checker stdout 和 stderr 会合并到 checker-output.txt，并通过 Checker Output 打开。
• Diff 和评测仍然使用纯 useroutput.txt；stderr 永远不会追加到该文件。
• 新的按题目运行流程还会在输出旁保留 stderr.txt 和 diff.txt，用于诊断和兼容。
• 旧的 1.out、1.err 和 1.diff 输出仍可读取，以保持兼容。

计时说明：样例耗时只统计用户可执行程序进程。Windows 上的样例耗时包含进程启动和管道 I/O 开销，因此极小程序也可能显示几十毫秒。

Windows 栈大小

• Windows 上深递归程序可能以 0xC00000FD 退出，这是栈溢出异常。
• 默认情况下，OI Judge 会跟随题目的内存限制，并在 Windows 编译时为 MinGW/g++ 添加链接器参数。
• 当 memoryMb = 256 时，生成的参数是 -Wl,--stack,268435456。
• 点击题目 Limits 区域下的 Stack 节点，或运行 OI Judge: Set Stack Size，可以选择 Follow Memory Limit、Custom Stack Size 或 Disable Auto Stack Size。
• 栈参数在编译时生成，不会反复插入 compile.args。
• 如果启用自动栈大小，已有的 -Wl,--stack,... 参数会被当前设置替换。如果禁用自动栈大小，OI Judge 不会添加栈参数。
• 这主要面向 Windows + MinGW/g++。Linux/macOS 评测环境通常通过 runner 或系统限制控制栈；从算法上避免过深递归仍然是最稳妥的选择。

Runtime Error 解释

• OI Judge 会根据进程退出码或 POSIX 信号解释常见 Runtime Error。
• Runtime Error 名称使用常见 OI/OJ 风格英文描述，例如：
  • Stack overflow
  • Access violation
  • Integer divide by zero
  • Floating point exception
  • Segmentation fault
• 中文界面会保留英文 Runtime Error 标题，并在下方补充中文说明、可能原因和建议。
• 常见 Windows 示例：
  • 0xC00000FD：Stack overflow
  • 0xC0000005：Access violation
  • 0xC0000094：Integer divide by zero
• 常见 Linux/macOS 信号：
  • SIGSEGV：Segmentation fault
  • SIGFPE：Floating point exception
  • SIGABRT：Aborted
• 这些解释只是诊断提示，不是最终证明。需要时仍应结合输入文件、stderr、复现命令和调试器判断。

评测模式

• OI Judge 为每个题目支持三种评测模式：
  • 文本比较：严格比较输出。
  • 文本比较（忽略行末空格和文末回车）：OI 常用默认方式，忽略每行末尾空白和文件末尾换行，但不忽略行首空格、行中空格或中间空行。
  • 自定义 Checker
• 在文本比较模式下，Checker 相关操作会从题目 Actions 区域隐藏，以保持侧边栏简洁。
• 点击题目下的 Judge Mode 节点，可以在严格文本比较、OI 风格文本比较和自定义 Checker 之间切换。
• 自定义 Checker 模式会启用 Checker 操作，并支持：
  • Testlib Checker
  • Plain Checker
• 如果题目仍处于文本比较模式，而你运行 OI Judge: Set Checker，OI Judge 会询问是否先切换到自定义 Checker。
• 切回文本比较模式不会删除已保存的 Checker 配置，因此之后可以再切回来，无需重新选择 Checker。

I/O 模式

• 每个题目可以使用 Standard IO 或 File IO。
• Standard IO 是默认模式：OI Judge 通过 stdin 输入样例，并把 stdout 捕获为 useroutput.txt。
• File IO 用于使用 problem.in 和 problem.out 等文件的程序。
• 在 File IO 模式下，OI Judge 会为每个样例创建隔离的临时运行目录：
  • .vscode/.OIJudge/problems/<problemId>/outputs/sample-<index>/run/
  • 将样例输入写入配置的输入文件名
  • 以该运行目录作为 cwd 运行可执行文件
  • 在进程结束后读取配置的输出文件
  • 将该文件内容保存回标准的 useroutput.txt
• Diff、普通比较、testlib Checker、Plain Checker 和结果面板都会继续使用 useroutput.txt。
• File IO 模式下的程序 stdout 只用于诊断；如果配置的输出文件缺失，stdout 不会被当作评测输出。
• 文件名必须是 problem.in 这样的简单名称；绝对路径、文件夹和 .. 都会被拒绝。
• OI Judge 永远不会在源文件目录、工作区根目录、样例目录或 checker 目录中创建配置的输入/输出文件。

File IO 示例程序：

#include <bits/stdc++.h>
using namespace std;

int main() {
    freopen("problem.in", "r", stdin);
    freopen("problem.out", "w", stdout);
    int a, b;
    cin >> a >> b;
    cout << a + b << "\n";
}

出题人模式

• Setter Mode 默认关闭。
• 可以在 VS Code 设置中启用：
  • oijudger.setterMode.enabled
• 启用后，OI Judge 会为每个题目显示出题人工具：
  • Select STD
  • Open STD
  • Clear STD
  • Generate Answer with STD
  • Generate All Sample Answers with STD
  • View Current Answer
  • View Generated Answer
  • Compare Generated Answer
  • Apply Generated Answer
  • Apply All Generated Answers
  • Delete Generated Answer
  • Add Generator / Open Generator / Remove Generator
  • Add Global Generator Input / Open Global Generator Input / Remove Global Generator Input
  • Bind Generator 和 Generator Input 到 Subtask
  • Set Sample Name
• STD 是出题人的标准程序。它与用于评测当前解法的默认程序分开保存。
• OI Judge 可以运行绑定的 STD，为单个样例或全部样例生成答案。
• 如果当前答案路径为空、答案文件不存在或答案文件为空，STD 生成输出会直接写入当前答案。
• 如果当前答案已有内容，STD 生成输出会作为 pending generated answer 暂存到 setter.generatedAnswers。
• 暂存的新生成答案可以查看、与当前答案对比、逐个应用、全部应用或删除。
• 在 Setter Mode 中编辑的样例名只是显示名/数据配置名，不会改变 sample.id、sample.index、真实输入/答案文件或 outputs/sample-<index>/。
• 出题数据配置保存在 setter.dataCases 中。
• OI Judge 支持 setter.generator.generators 中的题目级 Generator、generatorInputs 中的全局 Generator input，以及 Subtask 级 Generator input 绑定。
• 可以从题目或 Subtask 操作中生成样例输入。

Subtask 与计分

• 题目可以新建、重命名和删除 Subtask。删除 Subtask 只删除分组记录，不删除样例。
• 样例可以移动到某个 Subtask，也可以移回未分组区域。
• 每个 Subtask 保存 sampleIds、可选 lastResult、可选 Generator 绑定、可选 Generator input，以及计分模式。
• Subtask 可以独立评测。侧边栏会显示上次 Subtask 运行的通过/失败状态和 passed/total。
• 题目总分默认是 100，可以通过 OI Judge: Set Total Score 修改。
• 每个样例可以设置测试点分值。未设置分值的样例会自动平分剩余分值。
• sum Subtask 按通过的样例累计得分。
• bundle Subtask 只有全部样例通过时才获得该 Subtask 的全部分数。

Stress Test / 对拍

• OI Judge: Run Stress Test 支持两种对拍方式。
• 分文件对拍：Generator + STD + Solution。适合将生成器、标程、待测程序分别写成三个文件。插件会自动运行生成器生成输入，再分别运行 STD 和待测程序，并比较输出。
• 单文件对拍（考场式）。适合常见的考场 stress.cpp 写法。你可以在一个程序里自行生成数据、运行暴力/正解和待测逻辑、比较答案。插件只负责编译并运行这个程序，并保存 stdout / stderr / summary。
• 可使用 OI Judge: Stress Test Current Code 将最近在编辑区聚焦的已打开 .cpp 文件作为待测程序，并使用当前题目配置的 Generator 和 STD 进行分文件对拍。
• 对拍运行中可点击 $(debug-stop) Stop Stress Test 中断当前对拍，适合没有自动停止逻辑的单文件考场式 stress.cpp。
• 失败用例会保存在 .vscode/.OIJudge/stress/ 下，包含输入、STD 输出、待测输出、stderr 文件和 summary.json。
• Stress Records 视图可以打开保存文件、把失败用例加入样例、重新运行失败用例、刷新记录，并打开 session 文件夹。

测试点导出

• OI Judge: Export Testcases 可以复制或移动样例输入/输出文件，并写出 .OIJudge/config.json 导出记录。
• Luogu 导出可以生成 config.yml。
• Polygon 导出可以生成 OI Judge import plan polygon.json 和简短说明。
• LemonLime 导出可以生成 contest.cdf、data/<problemName>/、source/ 和简短说明。
• 如果配置了分值或 bundle Subtask，导出元数据会在目标格式支持时携带样例分值和 bundle Subtask 分组信息。

完整题目包导出

• OI Judge: Export Problem Package 与 Testcase Export 不同。Testcase Export 主要导出测试点文件和平台配置。
• Problem Package Export 会把完整出题资料复制到一个目录，便于备份、迁移、分享，以及后续实现导入。
• 导出目录包含 statement、source、STD、checker、generator、samples、Subtasks、scores 和 OI Judge 配置快照。
• 第一版导出为目录，包含 oijudge-package.json 和 README.txt，不生成 zip。

Testlib Checker

• OI Judge 支持按题目配置的 testlib 风格 Checker。
• 运行 OI Judge: Set Checker，选择 Testlib checker，然后选择本地 checker.cpp。
• 典型 checker 会包含 #include "testlib.h" 并调用 registerTestlibCmd(argc, argv)。
• OI Judge 会按如下方式运行 checker：

checker.exe input.txt useroutput.txt answer.txt

• testlib.h 会按以下顺序解析：
  • 与 checker.cpp 相同的文件夹
  • 工作区根目录
  • .vscode/.OIJudge/tools/testlib/testlib.h
  • checker 配置中记录的自定义路径
• OI Judge 可以安装扩展随附的 testlib.h，也可以导入用户选择的本地副本。
• 用户提供的副本一旦安装到工作区，仍然具有更高优先级。
• OI Judge 不会下载或生成 testlib.h。如果缺失，请运行 OI Judge: Import testlib.h。
• 当随附资源可用时，OI Judge: Import testlib.h 会提供：
  • Install bundled testlib.h
  • Import testlib.h from local file
• 随附源码和许可证细节保存在 resources/testlib/README.md 和 resources/testlib/LICENSE。
• Checker 可执行文件会构建在 .vscode/.OIJudge/problems/<problemId>/checker/ 下。
• Checker stdout 和 stderr 会合并为每个样例输出旁边的 checker-output.txt。
• testlib checker 经常把 verdict 细节输出到 stderr；用户仍可通过单个 Checker Output 操作查看所有 checker 信息。
• Plain Checker verdict 解析仍然只使用原始 stdout 中配置的 verdict 行。合并后的 stderr 内容只用于查看，不会被解析为 verdict。
• Verdict 规则：
  • checker 退出码 0 => AC
  • checker 退出码 1 => WA
  • Windows NTSTATUS 异常码，例如 0xC0000135 => Checker Error
  • checker 编译/运行/超时失败 => Checker Error
• Windows DLL 说明：
  • 如果 checker 以 3221225781 / 0xC0000135 退出，通常表示 checker.exe 因缺少运行时 DLL 而无法启动，不表示 checker 判了 WA。
  • 常见缺失 DLL 包括 MinGW 的 libstdc++-6.dll、libgcc_s_seh-1.dll 和 libwinpthread-1.dll。
  • OI Judge 会尝试使用 MinGW/g++ 静态链接编译 checker，并把编译器 bin 目录加入 checker 进程 PATH 前部。
  • 你也可以把 MinGW bin 目录加入 PATH，用静态链接重新构建 checker，或把缺失 DLL 放到 checker.exe 旁边。
• 未启用 checker 时，普通比较行为不变。
• Plain Checker 支持通过 stdout verdict 行输出数字分数。

Plain Checker

Plain Checker 是一种不依赖 testlib.h 的简单自定义 checker。

OI Judge 使用与 testlib checker 相同的参数运行它：

checker.exe input.txt useroutput.txt answer.txt

默认协议读取 stdout 最后一行非空内容。该行必须是以下之一：

• AC
• WA
• 一个数字分数

也可以运行 OI Judge: Set Plain Checker Protocol 来配置：

• 从第一行还是最后一行非空 stdout 读取 verdict
• accepted token，默认 AC
• wrong-answer token，默认 WA

示例：

AC

这会把样例标记为通过。

WA

这会把样例标记为答案错误。

37.5

这会返回 37.5 分。OI Judge 会显示问号图标，并在右侧显示 37.5。它不会把样例标记为通过或错误。

重要说明：如果你想返回 WA，请输出配置的 wrong-answer token。如果输出 0，OI Judge 会把它当作分数 0，而不是 WA。如果输出 100，OI Judge 会把它当作分数 100，而不是 AC。

默认协议下，无效 verdict 行包括：

• Accepted
• Wrong Answer
• 75%
• score: 75
• 通过

这些会被报告为 Checker Error。

自定义协议示例：

• Verdict line：First non-empty line
• Accepted token：OK
• Wrong answer token：NG

OK
details...

这会把样例标记为通过。

NG
wrong answer details...

这会把样例标记为答案错误。

37.5
matched 15 cases

这会显示问号图标和分数 37.5。数字 verdict 行总会被当作分数，因此 accepted/wrong token 应配置为非数字字符串。

最小 Plain Checker 示例：

#include <bits/stdc++.h>
using namespace std;

int main(int argc, char** argv) {
    if (argc < 4) {
        cout << "WA\n";
        return 0;
    }

    ifstream user(argv[2]);
    ifstream ans(argv[3]);

    long long a, b;
    user >> a;
    ans >> b;

    cout << (a == b ? "AC" : "WA") << '\n';
    return 0;
}

计分示例：

#include <bits/stdc++.h>
using namespace std;

int main(int argc, char** argv) {
    ifstream user(argv[2]);
    ifstream ans(argv[3]);

    int correct = 0, total = 10;
    for (int i = 0; i < total; i++) {
        int x, y;
        if (!(user >> x)) break;
        ans >> y;
        if (x == y) correct++;
    }

    cout << fixed << setprecision(1) << correct * 10.0 << '\n';
    return 0;
}

如果最后一行是 70.0，OI Judge 会显示问号图标和分数 70.0。

命令

• 基础评测：Init Problem、Add Sample、Run All Samples、Set Time Limit、Set Memory Limit、Set Stack Size、Select C++ Compiler、Open Last Report、Open Result Panel、Clear Outputs。
• 题目工作流：Create Problem、Add Problem From Current File、Add Problem From File、Import Legacy Single Problem、Delete Problem、Add Program、Set Default Program、Run With Program、Bind Statement、Open Statement、Unbind Statement。
• 样例：Add Problem Sample、Add Problem Sample From Files、Batch Add Samples、Open Sample Input、Open Expected Output、Open Run Result、Open Diff、Delete Sample。
• 出题人模式：Select STD、Open STD、Clear STD、Auto Generate Output with STD、Generate Answer with STD、Generate All Sample Answers with STD、View Current Answer、View Generated Answer、Compare Generated Answer、Apply Generated Answer、Apply All Generated Answers、Delete Generated Answer。
• Generator：Add Generator、Open Generator、Remove Generator、Select Generator、Open Generator、Clear Generator、Add Global Generator Input、Open Global Generator Input、Remove Global Generator Input。
• Subtask 与计分：Create Subtask、Rename Subtask、Delete Subtask、Move to Subtask、Run Subtask、Set Subtask Scoring Mode、Set Total Score、Set Testcase Score、Clear Testcase Score、Bind Generator、Bind Generator Input、Open Generator Input、Clear Generator Input、Generate Subtask Sample Input。
• 对拍与导出：Run Stress Test、Refresh Stress Records、Open Stress File、Add to Samples、Rerun This Case、Reveal Session Folder、Export Testcases、Export Problem Package、Copy freopen Input Snippet。
• Checker：Set Checker、Set Plain Checker Protocol、Clear Checker、Open Checker、Import testlib.h、Open testlib.h、Open Checker Output。

默认编译器命令是 g++。可以编辑 .vscode/.OIJudge 来调整每个题目的编译参数。

开发

在项目根目录运行：

npm install
npm run compile
npm test
npm run test:cross-platform
npm run test:report-ui
npm pack --dry-run

test:cross-platform checks judge, stress-test, stop-stress, and report artifact generation on the current platform. test:report-ui runs a lightweight report HTML smoke test. GitHub Actions runs these checks on Windows, Linux, and macOS and uploads result JSON plus report HTML artifacts.

在 VS Code 中按 F5，并选择 Run OI Judge Extension，即可打开扩展开发宿主窗口。

Development

From the project root:

npm install
npm run compile
npm test
npm run test:cross-platform
npm run test:report-ui
npm pack --dry-run

test:cross-platform checks judge, stress-test, stop-stress, and report artifact generation on the current platform. test:report-ui runs a lightweight report HTML smoke test. GitHub Actions runs these checks on Windows, Linux, and macOS and uploads result JSON plus report HTML artifacts.

Press F5 in VSCode and choose Run OI Judge Extension to open the Extension Development Host.