math_cms/scripts/audit_rendered_placeholder_integrity.php

<?php

/**
 * 全库题干「下划线占位 + 句点小黑点」流水线校验（与 paper-body 选择/填空口径对齐）。
 *
 * 用法：
 *   php scripts/audit_rendered_placeholder_integrity.php [--connection mysql] [--table questions]
 *       [--chunk 2000] [--out-dir storage/app/audit_placeholder]
 *       [--types choice,fill]
 *       [--check-unbalanced-dollars]
 *
 * 默认：仅扫描 choice + fill；输出 summary JSON + ndjson 明细。
 * 「$ 个数奇偶」默认不测（题库脏数据多时可加 --check-unbalanced-dollars）。
 * 重点排查项单独写入 *priority_issues*.ndjson（空位夹在双 $…$ 段之间、公式段以运算符结尾紧邻空位）。
 */

declare(strict_types=1);

require __DIR__.'/../vendor/autoload.php';
$app = require __DIR__.'/../bootstrap/app.php';
$kernel = $app->make(Illuminate\Contracts\Console\Kernel::class);
$kernel->bootstrap();

use App\Support\BlankPlaceholderRenderer;
use Illuminate\Support\Facades\DB;

$options = getopt('', [
    'table::',
    'connection::',
    'chunk::',
    'out-dir::',
    'types::',
    'check-unbalanced-dollars::',
]);

$checkUnbalancedDollars = array_key_exists('check-unbalanced-dollars', $options);

$table = isset($options['table']) ? trim((string) $options['table']) : 'questions';
$connection = isset($options['connection']) ? trim((string) $options['connection']) : config('database.default');
$chunk = isset($options['chunk']) ? max(100, (int) $options['chunk']) : 2000;
$defaultOut = dirname(__DIR__).'/storage/app/audit_placeholder';
$outDir = isset($options['out-dir']) ? rtrim((string) $options['out-dir'], '/') : $defaultOut;
// 默认仅选择与填空（与用户需求一致）；若要全题型可传 --types=all 并在下方解析
$typeFilter = isset($options['types']) ? trim((string) $options['types']) : 'choice,fill';
$types = [];
if (strtolower($typeFilter) === 'all') {
    $types = [];
} elseif ($typeFilter !== '') {
    $types = array_values(array_filter(array_map('trim', explode(',', $typeFilter)), static fn($v) => $v !== ''));
}

@mkdir($outDir, 0777, true);
$stamp = date('Ymd_His');
$summaryPath = "{$outDir}/rendered_placeholder_audit_summary_{$stamp}.json";
$detailPath = "{$outDir}/rendered_placeholder_audit_details_{$stamp}.ndjson";
$priorityDetailPath = "{$outDir}/rendered_placeholder_audit_priority_issues_{$stamp}.ndjson";

$detailFp = fopen($detailPath, 'wb');
if ($detailFp === false) {
    fwrite(STDERR, "Failed to open detail file: {$detailPath}\n");
    exit(1);
}

$priorityIssueTypes = [
    'blank_between_math_segments',
    'math_ends_with_operator_before_blank',
];

$priorityFp = fopen($priorityDetailPath, 'wb');
if ($priorityFp === false) {
    fwrite(STDERR, "Failed to open priority detail file: {$priorityDetailPath}\n");
    exit(1);
}

$issues = [];
$examples = [];

$scanned = 0;
$startedAt = microtime(true);

$recordIssue = static function (string $type, object $row, string $reason, string $rendered) use (&$issues, &$examples, $detailFp, $priorityFp, $priorityIssueTypes): void {
    if (! isset($issues[$type])) {
        $issues[$type] = 0;
        $examples[$type] = [];
    }
    $issues[$type]++;

    $entry = [
        'issue' => $type,
        'id' => (int) $row->id,
        'question_type' => (string) ($row->question_type ?? ''),
        'reason' => $reason,
        'stem_preview' => mb_substr((string) $row->stem, 0, 220),
        'rendered_preview' => mb_substr($rendered, 0, 260),
    ];

    fwrite($detailFp, json_encode($entry, JSON_UNESCAPED_UNICODE)."\n");

    if (in_array($type, $priorityIssueTypes, true)) {
        fwrite($priorityFp, json_encode($entry, JSON_UNESCAPED_UNICODE)."\n");
    }

    if (count($examples[$type]) < 20) {
        $examples[$type][] = [
            'id' => (int) $row->id,
            'question_type' => (string) ($row->question_type ?? ''),
            'reason' => $reason,
        ];
    }
};

$blankSpan = BlankPlaceholderRenderer::defaultBlankSpan();
$query = DB::connection($connection)
    ->table($table)
    ->select('id', 'question_type', 'stem')
    ->whereNotNull('stem')
    ->orderBy('id');

if ($types !== []) {
    $query->whereIn('question_type', $types);
}

$query->chunkById($chunk, function ($rows) use (&$scanned, $recordIssue, $blankSpan, $checkUnbalancedDollars): void {
    foreach ($rows as $row) {
        $stem = (string) $row->stem;
        $type = strtolower(trim((string) ($row->question_type ?? '')));

        [$rendered, $hasPlaceholders] = BlankPlaceholderRenderer::replaceToBlankSpan($stem, $blankSpan, false, false);

        // 与当前 paper-body 渲染口径一致（只覆盖选择/填空）
        if ($type === 'choice') {
            $rendered = BlankPlaceholderRenderer::normalizeTerminalPunctuation($rendered, 'remove');
        } elseif ($type === 'fill') {
            if (! $hasPlaceholders) {
                $rendered .= ' '.$blankSpan;
            }
            $rendered = BlankPlaceholderRenderer::normalizeTerminalPunctuation($rendered, 'dot');
            $rendered = BlankPlaceholderRenderer::normalizePeriodBeforeTrailingParentheticalNote($rendered, '.');
            $rendered = BlankPlaceholderRenderer::appendTerminalPunctuationIfMissing($rendered, '.');
        }

        // 1) 30949 类：\left( + 空位 + \right) 被拆成多个数学段
        if (preg_match('/\$\\s*\\\\left[\\(\\[]\\s*\$\\s*<span[^>]*>.*?<\\/span>\\s*\$\\s*\\\\right[\\)\\]]\\s*\$/u', $rendered)) {
            $recordIssue('broken_left_right_split', $row, 'left/right wrapped blank split into separate math segments', $rendered);
        }

        // 2) 空位夹在两个数学段中（高风险结构，常导致公式语义断裂）
        if (preg_match('/\$[^$]*\$\\s*<span[^>]*>.*?<\\/span>\\s*\$[^$]*\$/u', $rendered)) {
            $recordIssue('blank_between_math_segments', $row, 'blank span inserted between two $...$ segments', $rendered);
        }

        // 3) 渲染后「可见文本」里 $ 个数奇数 — 默认跳过（原始题干脏数据多）；需要时加 --check-unbalanced-dollars
        if ($checkUnbalancedDollars) {
            $visibleForDollar = html_entity_decode(strip_tags($rendered), ENT_QUOTES | ENT_HTML5, 'UTF-8');
            if ((substr_count($visibleForDollar, '$') % 2) !== 0) {
                $recordIssue('unbalanced_dollar_after_render', $row, 'odd number of $ in visible text after rendering', $rendered);
            }
        }

        // 4) 数学段在空位前以操作符结束（语义可能不完整）
        if (preg_match('/\$[^$]*[=+\-×÷*\\\\cdot]\\s*\$\\s*<span[^>]*>.*?<\\/span>/u', $rendered)) {
            $recordIssue('math_ends_with_operator_before_blank', $row, 'math segment ends with operator right before blank span', $rendered);
        }

        // 5) 2562 类回归：空位 span 后紧跟孤立 $ + 汉字（错误插 $）
        if (preg_match('/<\\/span>\s*\$\s*[\p{Han}]/u', $rendered)) {
            $recordIssue('span_then_dollar_before_han', $row, 'blank span followed by stray $ before Chinese (formula boundary break)', $rendered);
        }

        // 6) 占位 token 泄漏（不应出现在最终 HTML）
        if (preg_match('/<<<|BLANK_IN_MATH|LATEX_BLANK|LR_PAIR_/u', $rendered)) {
            $recordIssue('internal_placeholder_token_leak', $row, 'placeholder token not restored in output', $rendered);
        }

        $scanned++;
        if (($scanned % 5000) === 0) {
            fwrite(STDERR, "scanned={$scanned}\n");
        }
    }
}, 'id');

fclose($detailFp);
fclose($priorityFp);

$elapsed = round(microtime(true) - $startedAt, 3);

$investigationFocus = [
    'rules' => $priorityIssueTypes,
    'issue_counts' => [
        'blank_between_math_segments' => $issues['blank_between_math_segments'] ?? 0,
        'math_ends_with_operator_before_blank' => $issues['math_ends_with_operator_before_blank'] ?? 0,
    ],
];

$summary = [
    'table' => $table,
    'connection' => $connection,
    'chunk' => $chunk,
    'types_filter' => $types,
    'scanned_rows' => $scanned,
    'investigation_focus' => $investigationFocus,
    'checks_disabled_by_default' => array_values(array_filter([
        $checkUnbalancedDollars ? null : 'unbalanced_dollar_after_render ($ odd/even in visible text)',
    ])),
    'issue_counts' => $issues,
    'example_ids' => array_map(static fn(array $list) => array_column($list, 'id'), array_filter($examples, 'is_array')),
    'elapsed_seconds' => $elapsed,
    'generated_at' => date('c'),
    'detail_path' => $detailPath,
    'priority_issues_detail_path' => $priorityDetailPath,
];

file_put_contents($summaryPath, json_encode($summary, JSON_UNESCAPED_UNICODE | JSON_PRETTY_PRINT));

echo json_encode([
    'summary_path' => $summaryPath,
    'detail_path' => $detailPath,
    'priority_issues_detail_path' => $priorityDetailPath,
    'summary' => $summary,
], JSON_UNESCAPED_UNICODE | JSON_PRETTY_PRINT)."\n";