288 lines
11 KiB
JavaScript
288 lines
11 KiB
JavaScript
#!/usr/bin/env node
|
||
/**
|
||
* import-exam-tasks.js — imports tasks from /frontend/js/exam9/variants/*.js
|
||
* into the generic exam_tasks table for the exam-prep module.
|
||
*
|
||
* Usage:
|
||
* node backend/scripts/import-exam-tasks.js # all enabled tracks
|
||
* node backend/scripts/import-exam-tasks.js math9 # one specific track
|
||
* node backend/scripts/import-exam-tasks.js math9 --dry # don't write, only report parse stats
|
||
*
|
||
* Idempotent: deletes existing exam_tasks rows for the target exam_key before inserting.
|
||
*
|
||
* For each variant V it produces tasks_per_variant rows in exam_tasks. For each task:
|
||
* - task_type: 'mc' if has opts; 'open' if sol-ans parses to a clean numeric/short value; 'long' otherwise
|
||
* - answer: explicit task.answer if present; else autoparsed from <div class="sol-ans">
|
||
* - text_html / figure_html / opts_json / solution_html — direct
|
||
*
|
||
* Reports parse-quality stats at the end:
|
||
* - per-track: total / mc / open / long / explicit-answer / parsed-answer / unparseable
|
||
* - lists tasks where autoparse failed but has no opts and no explicit answer
|
||
*/
|
||
'use strict';
|
||
const fs = require('fs');
|
||
const path = require('path');
|
||
const db = require('../src/db/db');
|
||
|
||
const TRACK_VARIANTS_DIR = {
|
||
math9: path.join(__dirname, '../../frontend/js/exam9/variants'),
|
||
};
|
||
|
||
const args = process.argv.slice(2).filter(a => !a.startsWith('--'));
|
||
const flags = new Set(process.argv.slice(2).filter(a => a.startsWith('--')));
|
||
const DRY_RUN = flags.has('--dry');
|
||
const VERBOSE = flags.has('--verbose') || flags.has('-v');
|
||
|
||
/* ── HTML-text extraction from sol-ans div ───────────────────────── */
|
||
function extractAnswerText(solHtml) {
|
||
if (!solHtml) return null;
|
||
const m = solHtml.match(/<div class="sol-ans">([\s\S]*?)<\/div>/);
|
||
if (!m) return null;
|
||
let raw = m[1]
|
||
.replace(/<[^>]+>/g, '') // strip HTML tags
|
||
.replace(/ | | /g, ' ') // common entities
|
||
.replace(/&/g, '&').replace(/</g, '<').replace(/>/g, '>')
|
||
.trim();
|
||
raw = raw.replace(/^Ответ[:\s]*/i, '').trim();
|
||
return raw || null;
|
||
}
|
||
|
||
/* ── MC letter detector: matches «а)», «б.», «в », etc. ──────────── */
|
||
function parseMcLetter(answerText) {
|
||
if (!answerText) return null;
|
||
const m = answerText.match(/^([а-д])\s*[\)\.]/i);
|
||
return m ? m[1].toLowerCase() : null;
|
||
}
|
||
|
||
/* ── Open answer parser: returns a canonical answer string.
|
||
Forms supported:
|
||
"-2" single integer / decimal
|
||
"7500" positive integer
|
||
"9/4" fraction (from \dfrac{a}{b})
|
||
"-2;4" pair of values (from "x=A и x=B")
|
||
Returns null if the answer is too complex (expressions, multiple vars,
|
||
inequalities, square roots, intervals). ── */
|
||
function parseOpenAnswer(answerText) {
|
||
if (!answerText) return null;
|
||
|
||
// Normalize: strip $...$ and trivial LaTeX spacing
|
||
let s = answerText
|
||
.replace(/\\;|\\,|\\:|\\ /g, ' ')
|
||
.replace(/ | /g, ' ')
|
||
.trim();
|
||
|
||
// ── Pattern A: \dfrac{a}{b} or \frac{a}{b} as the sole answer
|
||
const fracMatch = s.match(/^\$?\\d?frac\{(-?\d+)\}\{(-?\d+)\}\$?(?:\s*[а-яА-Яa-zA-Z²³°%\.]*\.?)?$/);
|
||
if (fracMatch) {
|
||
return `${fracMatch[1]}/${fracMatch[2]}`;
|
||
}
|
||
// Also: "-\dfrac{a}{b}" with sign outside
|
||
const negFracMatch = s.match(/^\$?-\\d?frac\{(\d+)\}\{(\d+)\}\$?(?:\s*[а-яА-Яa-zA-Z²³°%\.]*\.?)?$/);
|
||
if (negFracMatch) {
|
||
return `-${negFracMatch[1]}/${negFracMatch[2]}`;
|
||
}
|
||
|
||
// ── Pattern B: two roots "$x = A$ и $x = B$" or "x_1=A; x_2=B"
|
||
const twoRoots = s.match(/x\s*_?\d?\s*=\s*(-?\d+(?:[.,]\d+)?)[\s\$]*(?:\sи\s|;)\s*\$?x\s*_?\d?\s*=\s*(-?\d+(?:[.,]\d+)?)/);
|
||
if (twoRoots) {
|
||
const a = twoRoots[1].replace(',', '.');
|
||
const b = twoRoots[2].replace(',', '.');
|
||
return `${a};${b}`;
|
||
}
|
||
|
||
// Strip $...$ for further checks (single-number paths)
|
||
s = s.replace(/\$/g, '').trim();
|
||
|
||
// Reject remaining complex forms
|
||
if (/\\dfrac|\\frac|\\sqrt|\\sum|\\int|\\cdot|\\pi/.test(s)) return null;
|
||
if (/[<>≤≥]/.test(s)) return null;
|
||
if (/\(.*[;,].*\)/.test(s)) return null; // intervals/points
|
||
if (s.split(/\s+или\s+|\s+and\s+|\s+и\s+/i).length > 1) return null;
|
||
if (/[xyz]\s*_?\d?\s*=.*[xyz]\s*_?\d?\s*=/.test(s)) return null; // multi-var didn't match pattern B
|
||
|
||
// "X = NUM" → take RHS
|
||
const eq = s.match(/=\s*(-?\d+(?:[.,]\d+)?)/);
|
||
if (eq) return eq[1].replace(',', '.');
|
||
|
||
// Single number with optional short unit tail
|
||
const single = s.match(/^(-?\d+(?:[.,]\d+)?)(\s*[а-яА-Яa-zA-Z\.²³°%]*\.?)?$/);
|
||
if (single) return single[1].replace(',', '.');
|
||
|
||
// Last try: first number iff rest is short suffix
|
||
const first = s.match(/(-?\d+(?:[.,]\d+)?)/);
|
||
if (first && first[1].length >= s.length - 8) return first[1].replace(',', '.');
|
||
|
||
return null;
|
||
}
|
||
|
||
/* ── Load a variant from .js via Function constructor ────────────── */
|
||
function loadVariant(dir, n) {
|
||
const nn = String(n).padStart(2, '0');
|
||
const file = path.join(dir, `v${nn}.js`);
|
||
if (!fs.existsSync(file)) return null;
|
||
const src = fs.readFileSync(file, 'utf8');
|
||
const scope = {};
|
||
new Function('VARIANTS', src)(scope);
|
||
return scope[n] || null;
|
||
}
|
||
|
||
/* ── Per-task classification + answer extraction ──────────────────── */
|
||
function classifyTask(task) {
|
||
const sol = task.sol || '';
|
||
const ansText = extractAnswerText(sol);
|
||
|
||
if (Array.isArray(task.opts) && task.opts.length) {
|
||
// MC: explicit answer wins, else parse letter from sol-ans
|
||
let answer = (typeof task.answer === 'string') ? task.answer.toLowerCase().trim() : null;
|
||
let source = 'explicit';
|
||
if (!answer || !/^[а-д]$/.test(answer)) {
|
||
answer = parseMcLetter(ansText);
|
||
source = answer ? 'parsed' : 'failed';
|
||
}
|
||
return { task_type: 'mc', answer, source, raw_answer: ansText };
|
||
}
|
||
|
||
// Non-MC: try open numeric, then fallback to long
|
||
let answer = (typeof task.answer === 'string') ? task.answer.trim() : null;
|
||
let source = answer ? 'explicit' : null;
|
||
|
||
if (!answer) {
|
||
answer = parseOpenAnswer(ansText);
|
||
source = answer ? 'parsed' : 'failed';
|
||
}
|
||
|
||
if (answer) return { task_type: 'open', answer, source, raw_answer: ansText };
|
||
return { task_type: 'long', answer: null, source: 'long', raw_answer: ansText };
|
||
}
|
||
|
||
/* ── Import a single track ────────────────────────────────────────── */
|
||
function importTrack(examKey) {
|
||
const dir = TRACK_VARIANTS_DIR[examKey];
|
||
if (!dir) throw new Error(`Unknown exam_key: ${examKey} (no variants dir mapping)`);
|
||
|
||
const track = db.prepare('SELECT variants_count FROM exam_tracks WHERE exam_key = ?').get(examKey);
|
||
if (!track) throw new Error(`Track not registered in exam_tracks: ${examKey}`);
|
||
|
||
const stats = {
|
||
examKey,
|
||
variants: 0,
|
||
tasks: 0,
|
||
mc: 0, open: 0, long: 0,
|
||
mcExplicit: 0, mcParsed: 0, mcFailed: 0,
|
||
openExplicit: 0, openParsed: 0,
|
||
failedExamples: [], // tasks classified as long where sol-ans existed (potential miss)
|
||
};
|
||
|
||
if (!DRY_RUN) {
|
||
db.prepare('DELETE FROM exam_tasks WHERE exam_key = ?').run(examKey);
|
||
}
|
||
|
||
const ins = db.prepare(`
|
||
INSERT INTO exam_tasks
|
||
(exam_key, variant, task_idx, task_type, text_html, figure_html, opts_json, answer, solution_html)
|
||
VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
|
||
`);
|
||
|
||
const writeAll = db.transaction(() => {
|
||
for (let n = 1; n <= track.variants_count; n++) {
|
||
const v = loadVariant(dir, n);
|
||
if (!v || !Array.isArray(v.tasks) || !v.tasks.length) {
|
||
if (VERBOSE) console.log(` v${String(n).padStart(2,'0')}: missing/empty — skipped`);
|
||
continue;
|
||
}
|
||
stats.variants++;
|
||
|
||
v.tasks.forEach((task, idx) => {
|
||
const taskIdx = idx + 1;
|
||
const cls = classifyTask(task);
|
||
|
||
stats[cls.task_type]++;
|
||
if (cls.task_type === 'mc') {
|
||
if (cls.source === 'explicit') stats.mcExplicit++;
|
||
else if (cls.source === 'parsed') stats.mcParsed++;
|
||
else stats.mcFailed++;
|
||
} else if (cls.task_type === 'open') {
|
||
if (cls.source === 'explicit') stats.openExplicit++;
|
||
else stats.openParsed++;
|
||
} else if (cls.task_type === 'long' && cls.raw_answer) {
|
||
// Has an answer but we classified as long → likely autoparser missed something
|
||
if (stats.failedExamples.length < 20) {
|
||
stats.failedExamples.push({ v: n, idx: taskIdx, raw: cls.raw_answer.slice(0, 80) });
|
||
}
|
||
}
|
||
stats.tasks++;
|
||
|
||
if (!DRY_RUN) {
|
||
ins.run(
|
||
examKey,
|
||
n,
|
||
taskIdx,
|
||
cls.task_type,
|
||
task.text || '',
|
||
task.figure || null,
|
||
task.opts ? JSON.stringify(task.opts) : null,
|
||
cls.answer,
|
||
task.sol || ''
|
||
);
|
||
}
|
||
});
|
||
}
|
||
});
|
||
|
||
writeAll();
|
||
return stats;
|
||
}
|
||
|
||
/* ── Reporting ────────────────────────────────────────────────────── */
|
||
function pct(n, total) {
|
||
if (!total) return '0%';
|
||
return ((n / total) * 100).toFixed(1) + '%';
|
||
}
|
||
|
||
function report(stats) {
|
||
const mcTotal = stats.mc;
|
||
const openTotal = stats.open;
|
||
console.log(`\n═══ ${stats.examKey} ═══`);
|
||
console.log(`Variants imported: ${stats.variants}`);
|
||
console.log(`Total tasks: ${stats.tasks}`);
|
||
console.log(` MC : ${stats.mc} (${pct(stats.mc, stats.tasks)})`);
|
||
console.log(` explicit: ${stats.mcExplicit}, parsed: ${stats.mcParsed}, FAILED: ${stats.mcFailed}`);
|
||
console.log(` Open : ${stats.open} (${pct(stats.open, stats.tasks)})`);
|
||
console.log(` explicit: ${stats.openExplicit}, parsed: ${stats.openParsed}`);
|
||
console.log(` Long : ${stats.long} (${pct(stats.long, stats.tasks)})`);
|
||
console.log(` ${stats.long - stats.failedExamples.length} truly complex, ${stats.failedExamples.length}+ POTENTIAL autoparse misses`);
|
||
|
||
if (stats.failedExamples.length) {
|
||
console.log(`\nPotential autoparse misses (classified 'long' but had a sol-ans answer):`);
|
||
stats.failedExamples.forEach(e => {
|
||
console.log(` v${String(e.v).padStart(2,'0')} t${e.idx}: «${e.raw}»`);
|
||
});
|
||
console.log(`(showing first ${stats.failedExamples.length}; fix by adding answer: '...' field in v*.js task, or relax parser in this script)`);
|
||
}
|
||
|
||
const autoSuccess = stats.mcParsed + stats.openParsed + stats.mcExplicit + stats.openExplicit;
|
||
const checkable = mcTotal + openTotal;
|
||
console.log(`\nAutocheckable tasks (mc+open): ${checkable} / ${stats.tasks} (${pct(checkable, stats.tasks)})`);
|
||
console.log(`Of those, answer determined: ${autoSuccess} (${pct(autoSuccess, checkable)})`);
|
||
}
|
||
|
||
/* ── Main ─────────────────────────────────────────────────────────── */
|
||
function main() {
|
||
const targets = args.length ? args : Object.keys(TRACK_VARIANTS_DIR);
|
||
console.log(`[import-exam-tasks] Targets: ${targets.join(', ')}${DRY_RUN ? ' (DRY RUN)' : ''}`);
|
||
|
||
for (const examKey of targets) {
|
||
try {
|
||
const stats = importTrack(examKey);
|
||
report(stats);
|
||
} catch (e) {
|
||
console.error(`[${examKey}] FAILED: ${e.message}`);
|
||
process.exitCode = 1;
|
||
}
|
||
}
|
||
|
||
if (DRY_RUN) console.log(`\n[DRY RUN] No changes written to DB.`);
|
||
}
|
||
|
||
main();
|