feat(exam-prep F0): миграция 022 + импорт-скрипт (800 задач math9, 76% автопроверяемые)

This commit is contained in:
Maxim Dolgolyov
2026-05-29 10:04:30 +03:00
parent b76315573c
commit 948b831273
3 changed files with 413 additions and 0 deletions
+1
View File
@@ -13,6 +13,7 @@
"seed:permissions": "node src/db/seed-permissions.js",
"lint:routes": "node scripts/check-route-auth.js",
"import:content": "node scripts/import-content.js",
"import:exam-tasks": "node scripts/import-exam-tasks.js",
"test": "node --test tests/*.test.js",
"hooks:install": "sh ../scripts/install-hooks.sh"
},
+287
View File
@@ -0,0 +1,287 @@
#!/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(/&ensp;|&nbsp;|&thinsp;/g, ' ') // common entities
.replace(/&amp;/g, '&').replace(/&lt;/g, '<').replace(/&gt;/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(/&ensp;|&nbsp;/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();
+125
View File
@@ -0,0 +1,125 @@
-- ═══════════════════════════════════════════════════════════════
-- 022: Exam Preparation Module
--
-- Generic exam preparation infrastructure parameterized by exam_key.
-- Supports multiple exams (math9, phys9, chem9, math11ce, etc.) sharing
-- the same task bank, attempts tracking, mock sessions, and study plans.
--
-- See docs/exam-prep-plan.md for full architecture.
-- ═══════════════════════════════════════════════════════════════
-- ── Exam tracks registry ────────────────────────────────────────
CREATE TABLE exam_tracks (
exam_key TEXT PRIMARY KEY,
title TEXT NOT NULL,
subject_slug TEXT NOT NULL,
grade INTEGER NOT NULL,
duration_min INTEGER NOT NULL,
tasks_per_variant INTEGER NOT NULL,
variants_count INTEGER NOT NULL,
scoring_json TEXT,
intro_html TEXT,
enabled INTEGER NOT NULL DEFAULT 1,
sort_order INTEGER NOT NULL DEFAULT 0
);
-- ── Task bank (one row per task across all variants) ────────────
CREATE TABLE exam_tasks (
id INTEGER PRIMARY KEY,
exam_key TEXT NOT NULL REFERENCES exam_tracks(exam_key) ON DELETE CASCADE,
variant INTEGER NOT NULL,
task_idx INTEGER NOT NULL,
task_type TEXT NOT NULL CHECK (task_type IN ('mc','open','long')),
text_html TEXT NOT NULL,
figure_html TEXT,
opts_json TEXT,
answer TEXT,
solution_html TEXT NOT NULL,
topic TEXT,
subtopic TEXT,
difficulty INTEGER,
UNIQUE(exam_key, variant, task_idx)
);
CREATE INDEX idx_exam_tasks_topic ON exam_tasks(exam_key, topic);
CREATE INDEX idx_exam_tasks_variant ON exam_tasks(exam_key, variant);
CREATE INDEX idx_exam_tasks_type ON exam_tasks(exam_key, task_type);
-- ── User attempts ───────────────────────────────────────────────
CREATE TABLE exam_attempts (
id INTEGER PRIMARY KEY,
user_id INTEGER NOT NULL REFERENCES users(id) ON DELETE CASCADE,
exam_task_id INTEGER NOT NULL REFERENCES exam_tasks(id) ON DELETE CASCADE,
user_answer TEXT,
is_correct INTEGER,
time_ms INTEGER,
mode TEXT NOT NULL CHECK (mode IN ('practice','variant','topic','mock')),
session_id INTEGER,
hint_used INTEGER NOT NULL DEFAULT 0,
solution_viewed INTEGER NOT NULL DEFAULT 0,
created_at INTEGER NOT NULL
);
CREATE INDEX idx_exam_attempts_user_time ON exam_attempts(user_id, created_at DESC);
CREATE INDEX idx_exam_attempts_task ON exam_attempts(exam_task_id);
CREATE INDEX idx_exam_attempts_session ON exam_attempts(session_id);
CREATE INDEX idx_exam_attempts_user_task ON exam_attempts(user_id, exam_task_id, created_at DESC);
-- ── Mock exam sessions ──────────────────────────────────────────
CREATE TABLE exam_mock_sessions (
id INTEGER PRIMARY KEY,
user_id INTEGER NOT NULL REFERENCES users(id) ON DELETE CASCADE,
exam_key TEXT NOT NULL,
variant INTEGER,
source TEXT NOT NULL CHECK (source IN ('variant','random','weak-topics')),
task_ids_json TEXT NOT NULL,
started_at INTEGER NOT NULL,
finished_at INTEGER,
duration_planned_min INTEGER NOT NULL,
score INTEGER,
total_correct INTEGER,
total_tasks INTEGER,
status TEXT NOT NULL DEFAULT 'active' CHECK (status IN ('active','finished','abandoned'))
);
CREATE INDEX idx_mock_user ON exam_mock_sessions(user_id, started_at DESC);
-- ── User preparation plan ───────────────────────────────────────
CREATE TABLE exam_user_plan (
user_id INTEGER NOT NULL REFERENCES users(id) ON DELETE CASCADE,
exam_key TEXT NOT NULL,
exam_date TEXT NOT NULL,
daily_target INTEGER NOT NULL DEFAULT 10,
weak_focus INTEGER NOT NULL DEFAULT 1,
created_at INTEGER NOT NULL,
updated_at INTEGER NOT NULL,
PRIMARY KEY (user_id, exam_key)
);
-- ── Topic dictionary (populated in F6) ──────────────────────────
CREATE TABLE exam_topics (
slug TEXT PRIMARY KEY,
exam_key TEXT NOT NULL,
parent_slug TEXT,
title TEXT NOT NULL,
description TEXT,
sort_order INTEGER NOT NULL DEFAULT 0
);
CREATE INDEX idx_topics_exam ON exam_topics(exam_key, parent_slug);
-- ── Seed: math9 track ───────────────────────────────────────────
-- Placeholder scoring grid (simple proportional). Replace with the actual
-- Belarus 9th-grade math exam grid once published per academic year.
INSERT INTO exam_tracks (
exam_key, title, subject_slug, grade, duration_min,
tasks_per_variant, variants_count, scoring_json, intro_html, enabled, sort_order
) VALUES (
'math9',
'Экзамен 9 класс — Математика',
'math',
9,
180,
10,
80,
'[{"correct":30,"score":10},{"correct":27,"score":9},{"correct":24,"score":8},{"correct":21,"score":7},{"correct":18,"score":6},{"correct":15,"score":5},{"correct":12,"score":4},{"correct":9,"score":3},{"correct":6,"score":2},{"correct":3,"score":1},{"correct":0,"score":0}]',
'<p>Полная подготовка к выпускному экзамену по математике за 9 класс. 80 реальных вариантов, разбор каждого задания, тренажёр по темам, пробные экзамены с таймером.</p>',
1,
10
);