<?php

declare(strict_types=1);

/**
 * Ticket 0014-FE Punkt 2: doppelt kodierte HTML-Entities in tt_content
 * (z. B. "VITEC&amp;apos;s" im JSON). Ursache: die Produkt-Importer
 * escapten mit ENT_QUOTES|ENT_HTML5 (' -> &apos;), die RTE-Persistenz
 * escapte das & erneut. Die Importer schreiben seit 2026-10-28 literal
 * (ENT_NOQUOTES); dieses Skript repariert den Bestand um EINE
 * Escaping-Stufe: &amp;apos; -> &apos; usw.
 *
 * Plain PDO (Muster migrations/check_bg.php).
 *
 * Usage:  php migrations/fix_double_entities.php          (Scan/Report)
 *         php migrations/fix_double_entities.php --apply
 */

error_reporting(E_ALL);
ini_set('display_errors', '1');

// Stufe 1: doppelt kodierte Entities um eine Stufe reduzieren.
// Stufe 2 (Befund 2026-10-28: Scan fand KEINE Doubles in der DB - die DB
// traegt &apos; EINFACH kodiert, erst die Render-Pipeline escapet das &
// erneut): Apostroph-Entities literal machen, dann kann nichts mehr doppeln.
const PATTERNS = ['&amp;apos;', '&amp;quot;', '&amp;#039;', '&amp;amp;', '&amp;nbsp;', '&apos;', '&#039;'];
const REPLACEMENTS = ['&apos;', '&quot;', '&#039;', '&amp;', '&nbsp;', "'", "'"];

$apply = in_array('--apply', $argv, true);

$root = null;
$dir = __DIR__;
for ($i = 0; $i < 5; $i++) {
    if (is_file($dir . '/config/system/settings.php')) {
        $root = $dir;
        break;
    }
    $dir = dirname($dir);
}
if ($root === null) {
    exit("Could not locate project root above " . __DIR__ . "\n");
}
$settings = require $root . '/config/system/settings.php';
$db = $settings['DB']['Connections']['Default'];
$pdo = new PDO(
    sprintf('mysql:host=%s;port=%s;dbname=%s;charset=utf8mb4', $db['host'] ?? 'localhost', $db['port'] ?? 3306, $db['dbname'] ?? ''),
    $db['user'] ?? '',
    $db['password'] ?? '',
    [PDO::ATTR_ERRMODE => PDO::ERRMODE_EXCEPTION, PDO::ATTR_DEFAULT_FETCH_MODE => PDO::FETCH_ASSOC]
);

echo ($apply ? 'APPLY' : 'SCAN') . "\n";

$like = implode(' OR ', array_map(static fn(string $p): string => "bodytext LIKE '%" . $p . "%'", PATTERNS));
$rows = $pdo->query(
    "SELECT c.uid, c.pid, p.slug, c.CType, c.bodytext FROM tt_content c
     LEFT JOIN pages p ON p.uid = c.pid
     WHERE c.deleted = 0 AND ($like) ORDER BY p.slug, c.uid"
)->fetchAll();

$fixed = 0;
foreach ($rows as $row) {
    $new = str_replace(PATTERNS, REPLACEMENTS, (string)$row['bodytext']);
    $hits = [];
    foreach (PATTERNS as $pattern) {
        $n = substr_count((string)$row['bodytext'], $pattern);
        if ($n > 0) {
            $hits[] = $pattern . 'x' . $n;
        }
    }
    printf("#%-6d %-28s %-45s %s\n", $row['uid'], $row['CType'], (string)$row['slug'], implode(' ', $hits));
    if ($apply) {
        $pdo->prepare('UPDATE tt_content SET bodytext = ?, tstamp = UNIX_TIMESTAMP() WHERE uid = ?')
            ->execute([$new, (int)$row['uid']]);
        $fixed++;
    }
}
echo "\n" . count($rows) . " Element(e) betroffen" . ($apply ? ", $fixed repariert. Danach: vendor/bin/typo3 cache:flush" : ". Mit --apply reparieren.") . "\n";
if (!$apply && $rows !== []) {
    echo "HINWEIS: Ersetzt wird genau EINE Escaping-Stufe (&amp;apos; -> &apos; usw.).\n";
}