原来试运行走另一台 Judge0：C 是 GCC 9、编译参数也不一样，Python 的语法错误回来的是运行错误， 没有题目的时间内存限制和代码模板，例子上的结果常和提交对不上。现在走本站判题机 （POST /trial-runs），同一套编译命令、这道题的限制和模板，对不对也由判题机比。 - 不落库、不算提交、不进判题队列：api 进程里自己排队，全站同时 TRIAL_CONCURRENCY 批 （服务器 2、机房 4），每人最多两批，排满或等超过 30 秒就请学生过一会儿再点 - 死循环不停打印时判题机回包会很大，超过 16MB 就掐断，告诉学生输出太多、停不下来 - Python 语法错误用和提交一样的说明：第几行、中文标点标红、一键换成英文 - 运行时错误和提交用同一个解析，说清楚第几行、为什么（下标越界、int() 转换失败、段错误……）； 比赛里照旧只给按结果码的那一句 - 语法没过、服务没跑成时，标题不再说「N 个例子没对上」 - 后台「生成测试点」一并迁过来，编译警告不会再写进 .out；Judge0 的地址和调用全部删掉 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
xuyueandClaude Opus 5.5 committed
1 parent
8834536135
commit
eb32676de4
27 files changed
+691
-169
No files matched your search
@@ -1,51 +1,96 @@
|
||||
import { createTestSubmission } from "utils/judge"
|
||||
import type { StatisticInfo, TrialRunResponse } from "@oj2/contract"
|
||||
import { errorMessage } from "utils/api"
|
||||
import { SubmissionStatus } from "utils/constants"
|
||||
import { isRunnableLanguage, trialRun } from "utils/judge"
|
||||
import type { Code } from "utils/types"
|
||||
|
||||
/**
|
||||
* 试跑：「运行例子」和「自己输入」。走 Judge0（utils/judge.ts），不经过判题机、不算提交。
|
||||
* 试跑：「运行例子」和「自己输入」。走本站判题机（utils/judge.ts → POST /trial-runs），
|
||||
* 不算提交。和提交是同一台判题机、同一套编译命令、这道题的时间内存限制和代码模板，
|
||||
* 所以例子上跑出来什么样，提交在这几个例子上就是什么样。
|
||||
*
|
||||
* 编译器版本和判题机不一样（Judge0 的 C 是 GCC 9，判题机是 gcc-14 加宽松参数），
|
||||
* 所以这里的结果偶尔会和提交不一致 —— 这是拍板过的取舍（设计文档第 3 节决定 3）。
|
||||
* 结果码就是判题机那套（SubmissionStatus），直接交给 WrongAnswerExplain 和 JUDGE_STATUS。
|
||||
* 对不对也是判题机比的（去掉末尾空白再比，和提交一样），这边不再自己比。
|
||||
*/
|
||||
|
||||
/** Judge0 的状态码（见 Judge0 的 /statuses）。同步调用（wait=true）只会拿到终态 */
|
||||
const JUDGE0_ACCEPTED = 3
|
||||
const JUDGE0_TIME_LIMIT = 5
|
||||
const JUDGE0_COMPILE_ERROR = 6
|
||||
const isJudge0RuntimeError = (status: number) => status >= 7 && status <= 12
|
||||
|
||||
/**
|
||||
* Python 的语法错误。Judge0 的 Python 没有编译这一步，SyntaxError 回来的是运行时错误
|
||||
* （status 11），得看回溯的最后一行才认得出来
|
||||
*/
|
||||
const PYTHON_SYNTAX_ERROR = /^(SyntaxError|IndentationError|TabError): /m
|
||||
|
||||
/**
|
||||
* 把 Judge0 的状态换成判题机那套码，好直接交给 WrongAnswerExplain 和 JUDGE_STATUS。
|
||||
* 跑完、只是输出对不上的，按答案错误算。Python 的语法错误按编译失败算：
|
||||
* 学生最常犯的中文冒号、中文括号，原来在这里被说成「程序运行到一半出错……下标有没有越界」
|
||||
*/
|
||||
export function trialStatus(status: number | null, passed: boolean, output: string) {
|
||||
if (status === JUDGE0_ACCEPTED) {
|
||||
return passed ? SubmissionStatus.accepted : SubmissionStatus.wrong_answer
|
||||
}
|
||||
if (status === JUDGE0_TIME_LIMIT) return SubmissionStatus.real_time_limit_exceeded
|
||||
if (status === JUDGE0_COMPILE_ERROR) return SubmissionStatus.compile_error
|
||||
if (status !== null && isJudge0RuntimeError(status)) {
|
||||
return PYTHON_SYNTAX_ERROR.test(output)
|
||||
? SubmissionStatus.compile_error
|
||||
: SubmissionStatus.runtime_error
|
||||
}
|
||||
return SubmissionStatus.system_error
|
||||
/** 试跑的是哪道题：按这道题的限制和模板跑 */
|
||||
export interface TrialTarget {
|
||||
problemId: number
|
||||
contestId: number | null
|
||||
}
|
||||
|
||||
/** 一组跑下来的结果 */
|
||||
interface TrialOutcome {
|
||||
result: SubmissionStatus
|
||||
output: string
|
||||
/** 没跑成、或者要换一种说法时给学生看的中文，有它就不用 TrialErrorNote 的默认说明 */
|
||||
note?: string
|
||||
/** 运行时错误的诊断（第几行、什么异常），和提交的同一套，交给 RuntimeErrorExplain */
|
||||
runtimeError?: NonNullable<StatisticInfo["runtime_error"]>
|
||||
}
|
||||
|
||||
const TOO_MUCH_OUTPUT =
|
||||
"程序输出太多了，一直停不下来：多半是循环没有结束的条件（条件一直成立、循环变量忘了改），在不停地打印。"
|
||||
|
||||
/**
|
||||
* 判题机比的是去掉末尾空白之后的输出（JudgeServer 的 stripped_output_md5），
|
||||
* 这里两边都去。原来题目页的「测试」按钮只 trim 了运行输出，样例输出末尾带个
|
||||
* 空格或换行，就永远显示「不通过」。
|
||||
* 把一次试跑的回包摊成每组一条。编译没过就每组都是编译失败、带同一份报错；
|
||||
* 输出太多的时候判题机回包被掐断了，分不出是哪一组，每组都挂上同一句说明。
|
||||
*/
|
||||
const sameOutput = (output: string, expected: string) => output.trimEnd() === expected.trimEnd()
|
||||
function outcomes(response: TrialRunResponse, count: number): TrialOutcome[] {
|
||||
if (response.status === "compile-error") {
|
||||
return Array.from({ length: count }, () => ({
|
||||
result: SubmissionStatus.compile_error,
|
||||
output: response.message,
|
||||
}))
|
||||
}
|
||||
if (response.status === "too-much-output") {
|
||||
return Array.from({ length: count }, () => ({
|
||||
result: SubmissionStatus.runtime_error,
|
||||
output: "",
|
||||
note: TOO_MUCH_OUTPUT,
|
||||
}))
|
||||
}
|
||||
// 跑完了（对或错）只去掉末尾空白（print 带的那个换行），判题机比对时也不看它；
|
||||
// 开头的空格是输出的一部分（打印菱形、三角形），不能动
|
||||
return response.cases.map((item) => {
|
||||
const result = item.result as SubmissionStatus
|
||||
const finished =
|
||||
result === SubmissionStatus.accepted || result === SubmissionStatus.wrong_answer
|
||||
return {
|
||||
result,
|
||||
output: finished ? item.output.trimEnd() : item.output,
|
||||
...(item.runtimeError ? { runtimeError: item.runtimeError } : {}),
|
||||
}
|
||||
})
|
||||
}
|
||||
|
||||
async function runCases(
|
||||
code: Code,
|
||||
target: TrialTarget,
|
||||
cases: { input: string; output?: string }[],
|
||||
): Promise<TrialOutcome[]> {
|
||||
if (!isRunnableLanguage(code.language)) {
|
||||
return cases.map(() => ({
|
||||
result: SubmissionStatus.system_error,
|
||||
output: "",
|
||||
note: `${code.language} 不能试运行`,
|
||||
}))
|
||||
}
|
||||
try {
|
||||
const response = await trialRun({
|
||||
problemId: target.problemId,
|
||||
contestId: target.contestId ?? undefined,
|
||||
language: code.language,
|
||||
code: code.value,
|
||||
cases,
|
||||
})
|
||||
return outcomes(response, cases.length)
|
||||
} catch (error) {
|
||||
// 后端的文案是给学生看的（「上一次还没跑完」「试运行的人太多了」）
|
||||
const note = errorMessage(error, "试跑的服务暂时连不上，稍后再试一次。")
|
||||
return cases.map(() => ({ result: SubmissionStatus.system_error, output: "", note }))
|
||||
}
|
||||
}
|
||||
|
||||
export interface SampleRun {
|
||||
/** 第几个例子，0 起 */
|
||||
@@ -53,8 +98,9 @@ export interface SampleRun {
|
||||
input: string
|
||||
expected: string
|
||||
output: string
|
||||
/** 换算成判题机那套码之后的结果，见 trialStatus */
|
||||
result: SubmissionStatus
|
||||
note?: string
|
||||
runtimeError?: TrialOutcome["runtimeError"]
|
||||
}
|
||||
|
||||
export function useTrialRun() {
|
||||
@@ -68,6 +114,10 @@ export function useTrialRun() {
|
||||
const customInput = ref("")
|
||||
const customOutput = ref("")
|
||||
const customResult = ref<SubmissionStatus | null>(null)
|
||||
const customNote = ref("")
|
||||
const customRuntimeError = ref<TrialOutcome["runtimeError"] | null>(null)
|
||||
/** 跑「自己输入」时用的那份代码：出错那一行要从它里面摆出来，编辑器里可能已经改了 */
|
||||
const customCode = ref("")
|
||||
const customRunning = ref(false)
|
||||
|
||||
/**
|
||||
@@ -77,44 +127,37 @@ export function useTrialRun() {
|
||||
let samplesGeneration = 0
|
||||
let customGeneration = 0
|
||||
|
||||
async function runOne(code: Code, input: string) {
|
||||
try {
|
||||
return await createTestSubmission(code, input)
|
||||
} catch {
|
||||
return { status: null, output: "" }
|
||||
}
|
||||
}
|
||||
|
||||
async function runSamples(code: Code, samples: { input: string; output: string }[]) {
|
||||
async function runSamples(
|
||||
code: Code,
|
||||
target: TrialTarget,
|
||||
samples: { input: string; output: string }[],
|
||||
) {
|
||||
const mine = ++samplesGeneration
|
||||
samplesRunning.value = true
|
||||
// 例子一般两三个，一起发
|
||||
const outputs = await Promise.all(samples.map((sample) => runOne(code, sample.input)))
|
||||
const runs = await runCases(code, target, samples)
|
||||
if (mine !== samplesGeneration) return
|
||||
sampleRuns.value = samples.map((sample, index) => {
|
||||
const run = outputs[index]!
|
||||
const passed = run.status === JUDGE0_ACCEPTED && sameOutput(run.output, sample.output)
|
||||
return {
|
||||
index,
|
||||
input: sample.input,
|
||||
expected: sample.output,
|
||||
output: run.output,
|
||||
result: trialStatus(run.status, passed, run.output),
|
||||
}
|
||||
})
|
||||
sampleRuns.value = samples.map((sample, index) => ({
|
||||
index,
|
||||
input: sample.input,
|
||||
expected: sample.output,
|
||||
...runs[index]!,
|
||||
}))
|
||||
samplesRunAt.value = new Date().toISOString()
|
||||
samplesCode.value = code
|
||||
samplesRunning.value = false
|
||||
}
|
||||
|
||||
async function runCustom(code: Code) {
|
||||
async function runCustom(code: Code, target: TrialTarget) {
|
||||
const mine = ++customGeneration
|
||||
customRunning.value = true
|
||||
const run = await runOne(code, customInput.value)
|
||||
// 自己输入的没有标准答案，不给 output：跑完没出错就算「对」
|
||||
const [run] = await runCases(code, target, [{ input: customInput.value }])
|
||||
if (mine !== customGeneration) return
|
||||
customOutput.value = run.output
|
||||
// 自己输入的没有标准答案，跑完就算「对」，只看有没有出错
|
||||
customResult.value = trialStatus(run.status, true, run.output)
|
||||
customOutput.value = run!.output
|
||||
customResult.value = run!.result
|
||||
customNote.value = run!.note ?? ""
|
||||
customRuntimeError.value = run!.runtimeError ?? null
|
||||
customCode.value = code.value
|
||||
customRunning.value = false
|
||||
}
|
||||
|
||||
@@ -129,6 +172,9 @@ export function useTrialRun() {
|
||||
customInput.value = ""
|
||||
customOutput.value = ""
|
||||
customResult.value = null
|
||||
customNote.value = ""
|
||||
customRuntimeError.value = null
|
||||
customCode.value = ""
|
||||
customRunning.value = false
|
||||
}
|
||||
|
||||
@@ -140,6 +186,9 @@ export function useTrialRun() {
|
||||
customInput,
|
||||
customOutput,
|
||||
customResult,
|
||||
customNote,
|
||||
customRuntimeError,
|
||||
customCode,
|
||||
customRunning,
|
||||
runSamples,
|
||||
runCustom,
|
||||
|
||||
Reference in new issue
Block a user