Fix file indexing in Contao 5.6 (inject DBAL connection, add debug logs)

README.md

@@ -2,6 +2,7 @@
 
 Eine schlanke Schnittstelle zwischen **Contao CMS (4.13 / 5.6 / 5.7 ready) unter PHP 8.4** und einer **selbst gehosteten Meilisearch-Instanz**.  
 Das Bundle erweitert den Contao-Suchindex um strukturierte Daten und ermöglicht eine performante, moderne Volltextsuche.
+Das Parsen von Dateien erfolgt über eine Apache-Tika-Instanz, welche extern bereitgestellt werden muss.
 
 ---
 
@@ -38,29 +39,27 @@ Das Bundle nutzt **keinen eigenen Contao-Cron**, sondern System-Cronjobs.
 /vendor/bin/contao-console meilisearch:files:cleanup
 ```
 
+### Datei-Parsing
+
+```
+/vendor/bin/contao-console meilisearch:files:parse
+```
+
 ### Meilisearch-Index
 
 ```
 /vendor/bin/contao-console meilisearch:index
 ```
 
-## Empfohlene Reihenfolge
 
-1. Datei-Cleanup  
-   `/vendor/bin/contao-console meilisearch:files:cleanup`
-
-2. Contao-Crawl (ca. 1 Minute später)  
-   `/vendor/bin/contao-console contao:crawl`
-
-3. Meilisearch-Index (ca. 15 Minuten später)  
-   `/vendor/bin/contao-console meilisearch:index`
 
 ## Beispiel Crontab
 
 ```
 0 5 * * *  /usr/bin/php8.4 /path/to/project/vendor/bin/contao-console meilisearch:files:cleanup
 1 5 * * *  /usr/bin/php8.4 /path/to/project/vendor/bin/contao-console contao:crawl
-15 5 * * * /usr/bin/php8.4 /path/to/project/vendor/bin/contao-console meilisearch:index
+10 5 * * * /usr/bin/php8.4 /path/to/project/vendor/bin/contao-console meilisearch:files:parse
+20 5 * * * /usr/bin/php8.4 /path/to/project/vendor/bin/contao-console meilisearch:index
 ```
 
 ## Logging

composer.json

@@ -8,11 +8,7 @@
     "contao/core-bundle": "^4.13 || ^5.6 || ^5.7",
     "contao/calendar-bundle": "^4.13 || ^5.6 || ^5.7",
     "contao/news-bundle": "^4.13 || ^5.6 || ^5.7",
-    "meilisearch/meilisearch-php": "^1.16",
-    "smalot/pdfparser": "^2.12",
-    "phpoffice/phpword": "^1.4",
-    "phpoffice/phpspreadsheet": "^3.0",
-    "phpoffice/phppresentation": "^1.2"
+    "meilisearch/meilisearch-php": "^1.16"
   },
   "autoload": {
     "psr-4": {

src/Command/MeilisearchFilesCleanupCommand.php

@@ -3,7 +3,7 @@
 namespace MummertMedia\ContaoMeilisearchBundle\Command;
 
 use Contao\CoreBundle\Framework\ContaoFramework;
-use Contao\Database;
+use Doctrine\DBAL\Connection;
 use Symfony\Component\Console\Command\Command;
 use Symfony\Component\Console\Input\InputInterface;
 use Symfony\Component\Console\Input\InputOption;
@@ -13,6 +13,7 @@ class MeilisearchFilesCleanupCommand extends Command
 {
     public function __construct(
         private readonly ContaoFramework $framework,
+        private readonly Connection $connection,
     ) {
         parent::__construct();
     }
@@ -21,13 +22,13 @@ class MeilisearchFilesCleanupCommand extends Command
     {
         $this
             ->setName('meilisearch:files:cleanup')
-            ->setDescription('Remove stale indexed files (PDF, DOCX, XLSX, PPTX) from tl_search_pdf')
+            ->setDescription('Remove stale indexed files from tl_search_files')
             ->addOption(
                 'grace',
                 null,
                 InputOption::VALUE_OPTIONAL,
                 'Grace period in seconds (files newer than now-grace are kept)',
-                86400 // 24 Stunden
+                86400
             )
             ->addOption(
                 'dry-run',
@@ -49,10 +50,10 @@ class MeilisearchFilesCleanupCommand extends Command
             $cutoff = time() - $grace;
 
             if ($dryRun) {
-                $count = Database::getInstance()
-                    ->prepare('SELECT COUNT(*) AS cnt FROM tl_search_pdf WHERE last_seen < ?')
-                    ->execute($cutoff)
-                    ->cnt;
+                $count = $this->connection->fetchOne(
+                    'SELECT COUNT(*) FROM tl_search_files WHERE last_seen < ?',
+                    [$cutoff]
+                );
 
                 $message = sprintf(
                     '[DRY-RUN] %d stale file(s) would be removed (last_seen < %s)',
@@ -63,14 +64,14 @@ class MeilisearchFilesCleanupCommand extends Command
                 $output->writeln('<comment>' . $message . '</comment>');
                 $this->log($message);
 
-                $this->log('Cleaner successfully stopped');
+                $this->log('Cleaner stopped (dry-run)');
                 return Command::SUCCESS;
             }
 
-            $affected = Database::getInstance()
-                ->prepare('DELETE FROM tl_search_pdf WHERE last_seen < ?')
-                ->execute($cutoff)
-                ->affectedRows;
+            $affected = $this->connection->executeStatement(
+                'DELETE FROM tl_search_files WHERE last_seen < ?',
+                [$cutoff]
+            );
 
             $message = sprintf(
                 'Removed %d stale file(s) (last_seen < %s)',
@@ -92,15 +93,8 @@ class MeilisearchFilesCleanupCommand extends Command
         }
     }
 
-    /**
-     * Einheitliches Logging mit Zeitstempel
-     */
     private function log(string $message): void
     {
-        error_log(sprintf(
-            '[%s] %s',
-            date('Y-m-d H:i:s'),
-            $message
-        ));
+        error_log(sprintf('[%s] %s', date('Y-m-d H:i:s'), $message));
     }
 }

src/Command/MeilisearchFilesParseCommand.php

@@ -0,0 +1,272 @@
+<?php
+
+namespace MummertMedia\ContaoMeilisearchBundle\Command;
+
+use Contao\CoreBundle\Framework\ContaoFramework;
+use Contao\Database;
+use Symfony\Component\Console\Command\Command;
+use Symfony\Component\Console\Input\InputInterface;
+use Symfony\Component\Console\Input\InputOption;
+use Symfony\Component\Console\Output\OutputInterface;
+use Symfony\Component\HttpClient\HttpClient;
+
+class MeilisearchFilesParseCommand extends Command
+{
+    public function __construct(
+        private readonly ContaoFramework $framework,
+    ) {
+        parent::__construct();
+    }
+
+    protected function configure(): void
+    {
+        $this
+            ->setName('meilisearch:files:parse')
+            ->setDescription('Parse indexed files via Apache Tika and enrich tl_search_files')
+            ->addOption(
+                'limit',
+                null,
+                InputOption::VALUE_OPTIONAL,
+                'Maximum number of files to check per run'
+            )
+            ->addOption(
+                'dry-run',
+                null,
+                InputOption::VALUE_NONE,
+                'Do not send files to Tika'
+            );
+    }
+
+    protected function execute(InputInterface $input, OutputInterface $output): int
+    {
+        $this->framework->initialize();
+        $this->log('Parser gestartet');
+
+        $dryRun = (bool) $input->getOption('dry-run');
+
+        $limitOption = $input->getOption('limit');
+        $limit = $limitOption !== null ? max(1, (int) $limitOption) : null;
+
+        $tikaUrl = rtrim((string) ($GLOBALS['TL_CONFIG']['meilisearch_tika_url'] ?? ''), '/');
+        if ($tikaUrl === '') {
+            $output->writeln('<error>Tika URL not configured</error>');
+            return Command::FAILURE;
+        }
+
+        $db = Database::getInstance();
+
+        $sql = "SELECT * FROM tl_search_files ORDER BY tstamp ASC";
+        if ($limit !== null) {
+            $sql .= " LIMIT " . (int) $limit;
+        }
+
+        $files = $db->query($sql)->fetchAllAssoc();
+
+        if (!$files) {
+            $this->log('No files to parse');
+            return Command::SUCCESS;
+        }
+
+        $client = HttpClient::create([
+            'timeout' => 180,
+        ]);
+
+        foreach ($files as $file) {
+
+            $originalUrl   = (string) $file['url'];
+            $existingTitle = trim((string) ($file['title'] ?? ''));
+            $normalized    = $originalUrl;
+
+            // -------------------------------------------------
+            // Normalize URL
+            // -------------------------------------------------
+            if (str_contains($normalized, '?')) {
+                $parts = parse_url($normalized);
+                if (!empty($parts['query'])) {
+                    parse_str($parts['query'], $query);
+                    if (!empty($query['file'])) {
+                        $normalized = (string) $query['file'];
+                    } else {
+                        $this->log('Not a direct file url, skip', ['url' => $originalUrl]);
+                        continue;
+                    }
+                }
+            }
+
+            $normalized = strtok($normalized, '#');
+            $normalized = rawurldecode($normalized);
+            $normalized = ltrim($normalized, '/');
+
+            if (!str_starts_with($normalized, 'files/')) {
+                $this->log('Not in files/, skip', ['url' => $originalUrl]);
+                continue;
+            }
+
+            $absolutePath = TL_ROOT . '/' . $normalized;
+
+            if (!is_file($absolutePath)) {
+                $this->log('File missing, skip', [
+                    'url'  => $originalUrl,
+                    'path' => $absolutePath,
+                ]);
+                continue;
+            }
+
+            $mtime    = filemtime($absolutePath) ?: 0;
+            $checksum = md5($normalized . '|' . $mtime);
+
+            // -------------------------------------------------
+            // Skip unchanged
+            // -------------------------------------------------
+            if ($file['checksum'] === $checksum && !empty($file['text'])) {
+                continue;
+            }
+
+            if ($dryRun) {
+                $output->writeln('[DRY-RUN] Would parse: ' . $normalized);
+                continue;
+            }
+
+            // -------------------------------------------------
+            // MIME-Type
+            // -------------------------------------------------
+            $ext = strtolower(pathinfo($normalized, PATHINFO_EXTENSION));
+
+            $mimeType = match ($ext) {
+                'pdf'  => 'application/pdf',
+                'docx' => 'application/vnd.openxmlformats-officedocument.wordprocessingml.document',
+                'xlsx' => 'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet',
+                'pptx' => 'application/vnd.openxmlformats-officedocument.presentationml.presentation',
+                default => null,
+            };
+
+            if ($mimeType === null) {
+                $this->log('Unsupported file type, skip', ['url' => $normalized]);
+                continue;
+            }
+
+            // -------------------------------------------------
+            // Tika BODY (roher Plaintext)
+            // -------------------------------------------------
+            try {
+                $this->log('Parsing file', ['url' => $normalized]);
+
+                $bodyResponse = $client->request(
+                    'PUT',
+                    $tikaUrl . '/tika/main',
+                    [
+                        'headers' => [
+                            'Accept'       => 'text/plain',
+                            'Content-Type' => $mimeType,
+                        ],
+                        'body' => fopen($absolutePath, 'rb'),
+                    ]
+                );
+
+                $text = trim((string) $bodyResponse->getContent(false));
+
+            } catch (\Throwable $e) {
+                $this->log('Body parse failed', [
+                    'url'   => $normalized,
+                    'error' => $e->getMessage(),
+                ]);
+                continue;
+            }
+
+            // -------------------------------------------------
+            // TITLE: keep existing editor-defined title
+            // -------------------------------------------------
+            $title = $existingTitle !== '' ? $existingTitle : null;
+
+            // -------------------------------------------------
+            // Tika METADATA (Title) – only if no existing title
+            // -------------------------------------------------
+            if ($title === null) {
+                try {
+                    $metaResponse = $client->request(
+                        'PUT',
+                        $tikaUrl . '/meta',
+                        [
+                            'headers' => [
+                                'Accept'       => 'application/json',
+                                'Content-Type' => $mimeType,
+                            ],
+                            'body' => fopen($absolutePath, 'rb'),
+                        ]
+                    );
+
+                    $meta = json_decode($metaResponse->getContent(false), true);
+
+                    $rawTitle =
+                        $meta['dc:title'][0]
+                        ?? $meta['pdf:docinfo:title'][0]
+                        ?? null;
+
+                    if ($rawTitle) {
+                        $title = html_entity_decode(
+                            $rawTitle,
+                            ENT_QUOTES | ENT_HTML5,
+                            'UTF-8'
+                        );
+                    }
+
+                } catch (\Throwable) {
+                    // Metadata optional
+                }
+            }
+
+            // -------------------------------------------------
+            // TITLE → ASCII SAFE (only if newly generated)
+            // -------------------------------------------------
+            if ($existingTitle === '' && $title) {
+                $title = iconv('UTF-8', 'ASCII//TRANSLIT//IGNORE', $title);
+                $title = preg_replace('/\s+/', ' ', $title);
+                $title = trim($title);
+            }
+
+            // -------------------------------------------------
+            // FALLBACK: Dateiname (only if still empty)
+            // -------------------------------------------------
+            if (!$title || strlen($title) < 5) {
+                $title = pathinfo($normalized, PATHINFO_FILENAME);
+                $title = str_replace(['_', '-'], ' ', $title);
+                $title = preg_replace('/\s+/', ' ', $title);
+                $title = trim($title);
+            }
+
+            // -------------------------------------------------
+            // Store result
+            // -------------------------------------------------
+            $db->prepare(
+                "UPDATE tl_search_files
+                 SET text = ?, title = ?, checksum = ?, file_mtime = ?, tstamp = ?
+                 WHERE id = ?"
+            )->execute(
+                $text,
+                $title,
+                $checksum,
+                $mtime,
+                time(),
+                $file['id']
+            );
+
+            $this->log('File parsed', [
+                'url'   => $normalized,
+                'chars' => mb_strlen($text),
+                'title' => $title,
+            ]);
+        }
+
+        $this->log('Parser finished');
+        return Command::SUCCESS;
+    }
+
+    private function log(string $message, array $context = []): void
+    {
+        $ctx = $context
+            ? ' | ' . json_encode($context, JSON_UNESCAPED_SLASHES | JSON_UNESCAPED_UNICODE)
+            : '';
+
+        error_log('[MeilisearchFilesParse] ' . $message . $ctx);
+    }
+}

src/EventListener/IndexPageListener.php

@@ -3,21 +3,22 @@
 namespace MummertMedia\ContaoMeilisearchBundle\EventListener;
 
 use Contao\Config;
-use MummertMedia\ContaoMeilisearchBundle\Service\PdfIndexService;
-use MummertMedia\ContaoMeilisearchBundle\Service\OfficeIndexService;
+use Contao\System;
+use Doctrine\DBAL\Connection;
 
 class IndexPageListener
 {
     public function __construct(
-        private readonly PdfIndexService $pdfIndexService,
-        private readonly OfficeIndexService $officeIndexService,
-    ) {}
+        private readonly Connection $connection,
+    ) {
+    }
 
     private function debug(string $message, array $context = []): void
     {
-        // Debug bewusst immer aktiv (bis du es wieder entfernst)
-        // Kontext kurz halten, damit Logs nicht explodieren
-        $ctx = $context ? ' | ' . json_encode($context, JSON_UNESCAPED_SLASHES | JSON_UNESCAPED_UNICODE) : '';
+        $ctx = $context
+            ? ' | ' . json_encode($context, JSON_UNESCAPED_SLASHES | JSON_UNESCAPED_UNICODE)
+            : '';
+
         error_log('[ContaoMeilisearch][IndexPageListener] ' . $message . $ctx);
     }
 
@@ -103,7 +104,6 @@ class IndexPageListener
                 $this->debug('Meta: searchimage candidate', ['searchimage' => $searchImage]);
 
                 if (!empty($searchImage)) {
-                    // >>> HINWEIS: falls dein tl_search-Feld "image" heißt, hier auf $set['image'] ändern!
                     $set['imagepath'] = trim((string) $searchImage);
                 }
 
@@ -139,20 +139,12 @@ class IndexPageListener
                         'class' => $e::class,
                     ]);
                 }
-
-                $this->debug('Meta: final set snapshot', [
-                    'priority'  => $set['priority'] ?? null,
-                    'keywords'  => $set['keywords'] ?? null,
-                    'imagepath' => $set['imagepath'] ?? null,
-                    'startDate' => $set['startDate'] ?? null,
-                    'checksum'  => $set['checksum'] ?? null,
-                ]);
             }
         }
 
         /*
          * =====================
-         * DATEI-INDEXIERUNG (PDF / OFFICE)
+         * DATEI-ERKENNUNG + UPSERT
          * =====================
          */
         if ((int) ($data['protected'] ?? 0) !== 0) {
@@ -160,15 +152,13 @@ class IndexPageListener
             return;
         }
 
-        $indexPdfs   = (bool) Config::get('meilisearch_index_pdfs');
-        $indexOffice = (bool) Config::get('meilisearch_index_office');
+        $indexFiles = (bool) Config::get('meilisearch_index_files');
 
-        $this->debug('File indexing settings', [
-            'meilisearch_index_pdfs'   => $indexPdfs,
-            'meilisearch_index_office' => $indexOffice,
+        $this->debug('File indexing setting', [
+            'meilisearch_index_files' => $indexFiles,
         ]);
 
-        if (!$indexPdfs && !$indexOffice) {
+        if (!$indexFiles) {
             $this->debug('Abort: file indexing disabled');
             return;
         }
@@ -176,61 +166,93 @@ class IndexPageListener
         $links = $this->findAllLinks($content);
         $this->debug('Links found', ['count' => count($links)]);
 
-        $pdfLinks    = [];
-        $officeLinks = [];
+        $fileLinks = [];
 
         foreach ($links as $link) {
             $type = $this->detectIndexableFileType($link['url']);
-
-            if ($type === 'pdf' && $indexPdfs) {
-                $pdfLinks[] = $link;
-                continue;
-            }
-
-            if (in_array($type, ['docx', 'xlsx', 'pptx'], true) && $indexOffice) {
-                $officeLinks[] = $link;
+            if ($type !== null) {
+                $fileLinks[] = $link + ['type' => $type];
             }
         }
 
-        $this->debug('Indexable file links', [
-            'pdf'    => count($pdfLinks),
-            'office' => count($officeLinks),
+        $this->debug('Indexable file links found', [
+            'count' => count($fileLinks),
+            'types' => array_count_values(array_column($fileLinks, 'type')),
         ]);
 
-        try {
-            if ($pdfLinks !== []) {
-                $this->debug('PDF handlePdfLinks(): call', ['count' => count($pdfLinks)]);
-                $this->pdfIndexService->handlePdfLinks($pdfLinks);
-                $this->debug('PDF handlePdfLinks(): ok');
-            }
+        if ($fileLinks) {
 
-            if ($officeLinks !== []) {
-                $this->debug('Office handleOfficeLinks(): call', ['count' => count($officeLinks)]);
-                $this->officeIndexService->handleOfficeLinks($officeLinks);
-                $this->debug('Office handleOfficeLinks(): ok');
-            }
-        } catch (\Throwable $e) {
-            $this->debug('File indexing failed', [
-                'error' => $e->getMessage(),
-                'class' => $e::class,
+            $this->debug('About to use Doctrine DBAL connection');
+
+            $db = $this->connection;
+
+            $this->debug('Doctrine DBAL connection ready', [
+                'driver' => get_class($db->getDriver()),
             ]);
+
+            $time = time();
+
+            foreach ($fileLinks as $file) {
+                $url = strtok($file['url'], '#');
+
+                $path = parse_url($url, PHP_URL_PATH);
+                $abs  = $path ? TL_ROOT . '/' . ltrim($path, '/') : null;
+
+                $mtime = ($abs && is_file($abs)) ? filemtime($abs) : 0;
+                $checksum = md5($url . '|' . $mtime);
+
+                $existing = $db->fetchAssociative(
+                    'SELECT id, checksum FROM tl_search_files WHERE url = ?',
+                    [$url]
+                );
+
+                if ($existing) {
+                    $db->update(
+                        'tl_search_files',
+                        [
+                            'tstamp'      => $time,
+                            'last_seen'   => $time,
+                            'page_id'     => (int) ($data['pid'] ?? 0),
+                            'file_mtime'  => $mtime,
+                            'checksum'    => $checksum,
+                        ],
+                        ['id' => $existing['id']]
+                    );
+
+                    $this->debug('File updated', [
+                        'url'      => $url,
+                        'checksum' => $checksum,
+                    ]);
+                } else {
+                    $db->insert(
+                        'tl_search_files',
+                        [
+                            'tstamp'     => $time,
+                            'last_seen'  => $time,
+                            'type'       => $file['type'],
+                            'url'        => $url,
+                            'title'      => $file['linkText'] ?? basename($url),
+                            'page_id'    => (int) ($data['pid'] ?? 0),
+                            'file_mtime' => $mtime,
+                            'checksum'   => $checksum,
+                        ]
+                    );
+
+                    $this->debug('File inserted', [
+                        'url'      => $url,
+                        'checksum' => $checksum,
+                    ]);
+                }
+            }
         }
 
         $this->debug('Hook end', [
             'final_set_keys' => array_keys($set),
-            'final_set'      => [
-                'priority'  => $set['priority'] ?? null,
-                'keywords'  => $set['keywords'] ?? null,
-                'imagepath' => $set['imagepath'] ?? null,
-                'startDate' => $set['startDate'] ?? null,
-                'checksum'  => $set['checksum'] ?? null,
-            ],
         ]);
     }
 
-    /**
-     * Extrahiert MEILISEARCH_JSON aus HTML-Kommentar
-     */
+    /* === Hilfsmethoden unverändert === */
+
     private function extractMeilisearchJson(string $content): ?array
     {
         if (!preg_match('/<!--\s*MEILISEARCH_JSON\s*(\{.*?\})\s*-->/s', $content, $m)) {
@@ -245,9 +267,6 @@ class IndexPageListener
             : null;
     }
 
-    /**
-     * Sammle alle <a href="…"> Links
-     */
     private function findAllLinks(string $content): array
     {
         if (!preg_match_all(
@@ -270,12 +289,8 @@ class IndexPageListener
         return $result;
     }
 
-    /**
-     * Ermittelt indexierbaren Dateityp (pdf|docx|xlsx|pptx) oder null
-     */
     private function detectIndexableFileType(string $url): ?string
     {
-        // Hash entfernen
         $url = strtok($url, '#');
 
         $parts = parse_url($url);
@@ -283,7 +298,6 @@ class IndexPageListener
             return null;
         }
 
-        // direkter Pfad (/files/…)
         if (!empty($parts['path'])) {
             $ext = strtolower(pathinfo($parts['path'], PATHINFO_EXTENSION));
             if (in_array($ext, ['pdf', 'docx', 'xlsx', 'pptx'], true)) {
@@ -291,18 +305,12 @@ class IndexPageListener
             }
         }
 
-        // Query-Parameter (Contao 4 + 5)
         if (!empty($parts['query'])) {
             parse_str($parts['query'], $query);
 
             foreach (['file', 'p', 'f'] as $param) {
                 if (!empty($query[$param])) {
-                    $candidate = (string) $query[$param];
-
-                    // sicher decodieren (Contao 4 + 5)
-                    $candidate = html_entity_decode($candidate, ENT_QUOTES);
-                    $candidate = rawurldecode($candidate);
-
+                    $candidate = rawurldecode(html_entity_decode((string) $query[$param], ENT_QUOTES));
                     $ext = strtolower(pathinfo($candidate, PATHINFO_EXTENSION));
 
                     if (in_array($ext, ['pdf', 'docx', 'xlsx', 'pptx'], true)) {

src/Resources/contao/dca/tl_search_files.php

@@ -2,7 +2,7 @@
 
 use Contao\DC_Table;
 
-$GLOBALS['TL_DCA']['tl_search_pdf'] = [
+$GLOBALS['TL_DCA']['tl_search_files'] = [
     'config' => [
         'dataContainer' => DC_Table::class,
         'sql' => [

src/Resources/contao/dca/tl_settings.php

@@ -4,14 +4,17 @@ use Contao\CoreBundle\DataContainer\PaletteManipulator;
 use Contao\System;
 
 /**
+ * -------------------------------------------------
  * Fields
+ * -------------------------------------------------
  */
+
 $GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_host'] = [
     'inputType' => 'text',
     'eval' => [
         'mandatory' => true,
-        'rgxp' => 'url',
-        'tl_class' => 'w50',
+        'rgxp'      => 'url',
+        'tl_class'  => 'w50',
     ],
 ];
 
@@ -19,7 +22,7 @@ $GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_index'] = [
     'inputType' => 'text',
     'eval' => [
         'mandatory' => true,
-        'tl_class' => 'w50',
+        'tl_class'  => 'w50',
     ],
 ];
 
@@ -27,7 +30,7 @@ $GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_api_write'] = [
     'inputType' => 'text',
     'eval' => [
         'mandatory' => true,
-        'tl_class' => 'w50',
+        'tl_class'  => 'w50',
         'hideInput' => true,
     ],
 ];
@@ -36,7 +39,7 @@ $GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_api_search'] = [
     'inputType' => 'text',
     'eval' => [
         'mandatory' => true,
-        'tl_class' => 'w50',
+        'tl_class'  => 'w50',
         'hideInput' => true,
     ],
 ];
@@ -55,50 +58,71 @@ $GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_imagesize'] = [
         return $options;
     },
     'eval' => [
-        'tl_class' => 'w50',
-        'chosen' => true,
+        'tl_class'           => 'w50',
+        'chosen'             => true,
         'includeBlankOption' => true,
     ],
-    // 🔥 DAS HAT GEFEHLT
     'sql' => "int(10) unsigned NOT NULL default 0",
 ];
 
-$GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_index_past_events'] = [
-    'inputType' => 'checkbox',
-    'eval'      => [
-        'tl_class' => 'w50 clr',
-    ],
-];
-
 $GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_fallback_image'] = [
     'inputType' => 'fileTree',
     'eval' => [
         'filesOnly' => true,
         'fieldType' => 'radio',
-        'tl_class' => 'w50',
+        'tl_class'  => 'w50',
     ],
     'sql' => "varbinary(16) NULL",
 ];
 
-$GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_index_pdfs'] = [
-    'label'     => &$GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_pdfs'],
+$GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_index_past_events'] = [
     'inputType' => 'checkbox',
-    'eval'      => [
-        'tl_class' => 'w50',
+    'eval' => [
+        'tl_class' => 'w50 clr',
     ],
-    'sql'       => "char(1) NOT NULL default '1'",
-];
-
-$GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_index_office'] = [
-    'label'     => &$GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_office'],
-    'inputType' => 'checkbox',
-    'eval'      => ['tl_class' => 'w50'],
-    'sql'       => "char(1) NOT NULL default '0'",
 ];
 
 /**
- * Palette
+ * -------------------------------------------------
+ * Datei-Indexierung (Tika)
+ * -------------------------------------------------
  */
+
+$GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_index_files'] = [
+    'inputType' => 'checkbox',
+    'eval' => [
+        'tl_class'       => 'w50',
+        'submitOnChange' => true,
+    ],
+    'sql' => "char(1) NOT NULL default '0'",
+];
+
+$GLOBALS['TL_DCA']['tl_settings']['fields']['meilisearch_tika_url'] = [
+    'inputType' => 'text',
+    'eval' => [
+        'rgxp'      => 'url',
+        'mandatory' => true,
+        'tl_class'  => 'w50 clr',
+    ],
+];
+
+/**
+ * -------------------------------------------------
+ * Selector / Subpalette
+ * -------------------------------------------------
+ */
+
+$GLOBALS['TL_DCA']['tl_settings']['palettes']['__selector__'][] = 'meilisearch_index_files';
+
+$GLOBALS['TL_DCA']['tl_settings']['subpalettes']['meilisearch_index_files']
+    = 'meilisearch_tika_url';
+
+/**
+ * -------------------------------------------------
+ * Palette
+ * -------------------------------------------------
+ */
+
 PaletteManipulator::create()
     ->addLegend('meilisearch_legend', null, PaletteManipulator::POSITION_AFTER, true)
     ->addField('meilisearch_host', 'meilisearch_legend')
@@ -108,6 +132,5 @@ PaletteManipulator::create()
     ->addField('meilisearch_imagesize', 'meilisearch_legend')
     ->addField('meilisearch_fallback_image', 'meilisearch_legend')
     ->addField('meilisearch_index_past_events', 'meilisearch_legend')
-    ->addField('meilisearch_index_pdfs', 'meilisearch_legend')
-    ->addField('meilisearch_index_office', 'meilisearch_legend')
+    ->addField('meilisearch_index_files', 'meilisearch_legend')
     ->applyToPalette('default', 'tl_settings');

src/Resources/contao/languages/de/tl_settings.php

@@ -28,10 +28,10 @@ $GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_past_events'][0]
 $GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_past_events'][1]
     = 'Vergangene Kalender-Events werden ebenfalls in Meilisearch indexiert.';
 
-$GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_pdfs'] = [
-    'PDFs indexieren',
-    'Aktiviert die Indexierung von PDF-Dateien für die Suche.',
+$GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_files'] = [
+    'Dateien indexieren',
+    'Aktiviert die Indexierung von PDF-Dateien sowie DOCX, XLSX und PPTX.',
 ];
 
-$GLOBALS['TL_LANG']['tl_settings']['meilisearch_index_office']
-    = ['Office-Dateien indexieren', 'DOCX, XLSX und PPTX in die Suche aufnehmen.'];
+$GLOBALS['TL_LANG']['tl_settings']['meilisearch_tika_url']
+    = ['Apache Tika URL', 'URL der Apache Tika Instanz (z. B. https://tika.domain.tld).'];

src/Resources/contao/templates/frontend_module/mod_meilisearch_search.html.twig

@@ -4,6 +4,7 @@ Contao 5 – Frontend Module Template
 #}
 
 <!-- indexer::stop -->
+{% block meilisearch %}
 <div
         id="topsearch"
         class="meilisearch-search"
@@ -218,4 +219,5 @@ Contao 5 – Frontend Module Template
         }
     });
 </script>
+{% endblock %}
 <!-- indexer::continue -->

src/Service/MeilisearchIndexService.php

@@ -41,8 +41,8 @@ class MeilisearchIndexService
             return;
         }
 
-        $host = (string) Config::get('meilisearch_host');
-        $apiKey = (string) Config::get('meilisearch_api_write');
+        $host      = (string) Config::get('meilisearch_host');
+        $apiKey    = (string) Config::get('meilisearch_api_write');
         $this->indexName = (string) Config::get('meilisearch_index');
 
         if ($host === '' || $this->indexName === '') {
@@ -72,7 +72,7 @@ class MeilisearchIndexService
         }
 
         $this->indexTlSearch($index);
-        $this->indexTlSearchPdf($index);
+        $this->indexTlSearchFiles($index);
     }
 
     private function ensureIndexSettings(Indexes $index): void
@@ -80,6 +80,7 @@ class MeilisearchIndexService
         $index->updateSettings([
             'searchableAttributes' => ['title', 'keywords', 'text'],
             'sortableAttributes'   => ['priority', 'startDate'],
+            'filterableAttributes' => ['type', 'filetype'],
         ]);
     }
 
@@ -132,7 +133,7 @@ class MeilisearchIndexService
     }
 
     /**
-     * tl_search indexieren
+     * tl_search indexieren (Seiten / News / Events)
      */
     private function indexTlSearch(Indexes $index): void
     {
@@ -164,13 +165,11 @@ class MeilisearchIndexService
                     }
                 }
 
-                $cleanText = $this->stripMeilisearchMeta((string) $row['text']);
-
                 $doc = [
                     'id'        => $type . '_' . $row['id'],
                     'type'      => $type,
                     'title'     => $row['title'],
-                    'text'      => $cleanText,
+                    'text'      => $this->stripMeilisearchMeta((string) $row['text']),
                     'url'       => $row['url'],
                     'protected' => (bool) $row['protected'],
                     'checksum'  => $row['checksum'],
@@ -192,31 +191,24 @@ class MeilisearchIndexService
                 $documents[] = $doc;
 
             } catch (\Throwable $e) {
-                error_log(
-                    '[ContaoMeilisearch] Failed to build document for tl_search ID '
-                    . ($row['id'] ?? '?') . ': ' . $e->getMessage()
-                );
+                error_log('[ContaoMeilisearch] Failed to build tl_search document: ' . $e->getMessage());
             }
         }
 
         if ($documents !== []) {
-            try {
-                $index->addDocuments($documents);
-            } catch (\Throwable $e) {
-                error_log('[ContaoMeilisearch] Failed to add tl_search documents: ' . $e->getMessage());
-            }
+            $index->addDocuments($documents);
         }
     }
 
     /**
-     * tl_search_pdf indexieren
+     * tl_search_files indexieren (PDF / Office)
      */
-    private function indexTlSearchPdf(Indexes $index): void
+    private function indexTlSearchFiles(Indexes $index): void
     {
         try {
-            $rows = $this->connection->fetchAllAssociative('SELECT * FROM tl_search_pdf');
+            $rows = $this->connection->fetchAllAssociative('SELECT * FROM tl_search_files');
         } catch (\Throwable $e) {
-            error_log('[ContaoMeilisearch] Failed to read tl_search_pdf: ' . $e->getMessage());
+            error_log('[ContaoMeilisearch] Failed to read tl_search_files: ' . $e->getMessage());
             return;
         }
 
@@ -233,30 +225,24 @@ class MeilisearchIndexService
                     : 'pdf';
 
                 $documents[] = [
-                    'id'       => $fileType . '_' . $row['id'],
-                    'type'     => $fileType,
-                    'title'    => $row['title'],
-                    'text'     => $this->stripMeilisearchMeta((string) $row['text']),
-                    'url'      => $row['url'],
-                    'checksum' => $row['checksum'],
-                    'poster'   => self::FILETYPE_ICON_MAP[$fileType]
+                    'id'        => 'file_' . $row['id'],
+                    'type'      => 'file',
+                    'filetype'  => $fileType,
+                    'title'     => $row['title'] ?: basename($row['url']),
+                    'text'      => (string) $row['text'],
+                    'url'       => $row['url'],
+                    'checksum'  => $row['checksum'],
+                    'poster'    => self::FILETYPE_ICON_MAP[$fileType]
                         ?? self::FILETYPE_ICON_MAP['pdf'],
                 ];
 
             } catch (\Throwable $e) {
-                error_log(
-                    '[ContaoMeilisearch] Failed to build PDF document for ID '
-                    . ($row['id'] ?? '?') . ': ' . $e->getMessage()
-                );
+                error_log('[ContaoMeilisearch] Failed to build file document: ' . $e->getMessage());
             }
         }
 
         if ($documents !== []) {
-            try {
-                $index->addDocuments($documents);
-            } catch (\Throwable $e) {
-                error_log('[ContaoMeilisearch] Failed to add tl_search_pdf documents: ' . $e->getMessage());
-            }
+            $index->addDocuments($documents);
         }
     }
 

src/Service/OfficeIndexService.php

@@ -1,276 +0,0 @@
-<?php
-
-namespace MummertMedia\ContaoMeilisearchBundle\Service;
-
-use Contao\Database;
-use PhpOffice\PhpWord\IOFactory as WordIOFactory;
-use PhpOffice\PhpSpreadsheet\IOFactory as SpreadsheetIOFactory;
-use PhpOffice\PhpPresentation\IOFactory as PresentationIOFactory;
-use Symfony\Component\DependencyInjection\ParameterBag\ParameterBagInterface;
-
-class OfficeIndexService
-{
-    private string $projectDir;
-
-    public function __construct(ParameterBagInterface $params)
-    {
-        $this->projectDir = rtrim((string) $params->get('kernel.project_dir'), '/');
-    }
-
-    /**
-     * @param array<int,array{url:string,linkText:?string}> $officeLinks
-     */
-    public function handleOfficeLinks(array $officeLinks): void
-    {
-        // Dedupe nur pro Aufruf (nicht "pro Crawl")
-        $seen = [];
-        $now  = time();
-
-        foreach ($officeLinks as $row) {
-            $url      = (string) ($row['url'] ?? '');
-            $linkText = $row['linkText'] ?? null;
-
-            if ($url === '') {
-                continue;
-            }
-
-            // doppelte URLs pro Aufruf vermeiden
-            $seenKey = md5($url);
-            if (isset($seen[$seenKey])) {
-                continue;
-            }
-            $seen[$seenKey] = true;
-
-            $normalized = $this->normalizeOfficeUrl($url);
-            if ($normalized === null) {
-                continue;
-            }
-
-            [$relativePath, $type] = $normalized;
-
-            $absolutePath = $this->getAbsolutePath($relativePath);
-            if (!is_file($absolutePath)) {
-                continue;
-            }
-
-            $mtime    = (int) (filemtime($absolutePath) ?: 0);
-            $checksum = md5($relativePath . '|' . $mtime);
-
-            // existiert bereits?
-            $existing = Database::getInstance()
-                ->prepare('SELECT checksum FROM tl_search_pdf WHERE url=? LIMIT 1')
-                ->execute($relativePath)
-                ->fetchAssoc();
-
-            $needsParse = !$existing || ($existing['checksum'] ?? '') !== $checksum;
-
-            // Titel-Priorität:
-            // 1) Linktext
-            // 2) Dateiname
-            $title = $linkText ?: basename($absolutePath);
-            $text  = '';
-
-            if ($needsParse) {
-                $text = $this->parseOfficeFile($absolutePath, $type);
-                if ($text === '') {
-                    // Parsing fehlgeschlagen → nichts überschreiben
-                    continue;
-                }
-            }
-
-            $this->upsertOffice(
-                $relativePath,
-                $title,
-                $text,      // kann '' sein → SQL überschreibt dann nicht
-                $checksum,
-                $mtime,
-                $type,
-                $now
-            );
-        }
-    }
-
-    /**
-     * @return array{string,string}|null [relativePath, type]
-     */
-    private function normalizeOfficeUrl(string $url): ?array
-    {
-        $decoded = html_entity_decode($url);
-        $parts   = parse_url($decoded);
-
-        if (!$parts) {
-            return null;
-        }
-
-        // 1) files/... (ohne führenden Slash)
-        if (!empty($parts['path']) && str_starts_with($parts['path'], 'files/')) {
-            $ext = strtolower(pathinfo($parts['path'], PATHINFO_EXTENSION));
-            if (in_array($ext, ['docx', 'xlsx', 'pptx'], true)) {
-                return ['/' . $parts['path'], $ext];
-            }
-        }
-
-        // 2) /files/...
-        if (!empty($parts['path']) && str_starts_with($parts['path'], '/files/')) {
-            $ext = strtolower(pathinfo($parts['path'], PATHINFO_EXTENSION));
-            if (in_array($ext, ['docx', 'xlsx', 'pptx'], true)) {
-                return [$parts['path'], $ext];
-            }
-        }
-
-        if (empty($parts['query'])) {
-            return null;
-        }
-
-        parse_str($parts['query'], $query);
-
-        // 3) Contao 4: ?file=files/...
-        if (!empty($query['file'])) {
-            $file = urldecode((string) $query['file']);
-            $file = ltrim($file, '/');
-            $ext  = strtolower(pathinfo($file, PATHINFO_EXTENSION));
-
-            if (
-                str_starts_with($file, 'files/') &&
-                in_array($ext, ['docx', 'xlsx', 'pptx'], true)
-            ) {
-                return ['/' . $file, $ext];
-            }
-        }
-
-        // 4) Contao 5: ?p=...
-        if (!empty($query['p'])) {
-            $p   = urldecode((string) $query['p']);
-            $ext = strtolower(pathinfo($p, PATHINFO_EXTENSION));
-
-            if (in_array($ext, ['docx', 'xlsx', 'pptx'], true)) {
-                return ['/files/' . ltrim($p, '/'), $ext];
-            }
-        }
-
-        return null;
-    }
-
-    private function getAbsolutePath(string $relativePath): string
-    {
-        return $this->projectDir . '/' . ltrim($relativePath, '/');
-    }
-
-    private function upsertOffice(
-        string $url,
-        string $title,
-        string $text,
-        string $checksum,
-        int $mtime,
-        string $type,
-        int $now
-    ): void {
-        Database::getInstance()
-            ->prepare('
-                INSERT INTO tl_search_pdf
-                    (tstamp, last_seen, type, url, title, text, checksum, file_mtime)
-                VALUES
-                    (?, ?, ?, ?, ?, ?, ?, ?)
-                ON DUPLICATE KEY UPDATE
-                    tstamp     = VALUES(tstamp),
-                    last_seen  = VALUES(last_seen),
-                    type       = VALUES(type),
-                    url        = VALUES(url),
-                    title      = VALUES(title),
-                    checksum   = VALUES(checksum),
-                    file_mtime = VALUES(file_mtime),
-                    text       = IF(VALUES(text) = "" OR VALUES(text) IS NULL, text, VALUES(text))
-            ')
-            ->execute(
-                $now,
-                $now,
-                $type,
-                $url,
-                $title,
-                $text,
-                $checksum,
-                $mtime
-            );
-    }
-
-    private function parseOfficeFile(string $absolutePath, string $type): string
-    {
-        return match ($type) {
-            'docx' => $this->parseDocx($absolutePath),
-            'xlsx' => $this->parseXlsx($absolutePath),
-            'pptx' => $this->parsePptx($absolutePath),
-            default => '',
-        };
-    }
-
-    private function parseDocx(string $absolutePath): string
-    {
-        try {
-            $phpWord = WordIOFactory::load($absolutePath);
-            $text = '';
-
-            foreach ($phpWord->getSections() as $section) {
-                foreach ($section->getElements() as $element) {
-                    if (method_exists($element, 'getText')) {
-                        $text .= ' ' . $element->getText();
-                    }
-                }
-            }
-
-            return $this->cleanText($text);
-        } catch (\Throwable) {
-            return '';
-        }
-    }
-
-    private function parseXlsx(string $absolutePath): string
-    {
-        try {
-            $spreadsheet = SpreadsheetIOFactory::load($absolutePath);
-            $text = '';
-
-            foreach ($spreadsheet->getAllSheets() as $sheet) {
-                foreach ($sheet->toArray() as $row) {
-                    $text .= ' ' . implode(' ', array_filter($row, 'is_scalar'));
-                }
-            }
-
-            return $this->cleanText($text);
-        } catch (\Throwable) {
-            return '';
-        }
-    }
-
-    private function parsePptx(string $absolutePath): string
-    {
-        try {
-            $presentation = PresentationIOFactory::load($absolutePath);
-            $text = '';
-
-            foreach ($presentation->getAllSlides() as $slide) {
-                foreach ($slide->getShapeCollection() as $shape) {
-                    if (method_exists($shape, 'getPlainText')) {
-                        $text .= ' ' . $shape->getPlainText();
-                    }
-                }
-            }
-
-            return $this->cleanText($text);
-        } catch (\Throwable) {
-            return '';
-        }
-    }
-
-    private function cleanText(string $text): string
-    {
-        if (class_exists(\Normalizer::class)) {
-            $text = \Normalizer::normalize($text, \Normalizer::FORM_C) ?? $text;
-        }
-
-        $text = str_replace(["\r\n", "\r"], "\n", $text);
-        $text = preg_replace('/[^\p{L}\p{N}\p{P}\p{Z}\n]/u', ' ', $text);
-        $text = preg_replace('/\s+/u', ' ', $text);
-
-        return trim(mb_substr($text, 0, 20000));
-    }
-}

src/Service/PdfIndexService.php

@@ -1,236 +0,0 @@
-<?php
-
-namespace MummertMedia\ContaoMeilisearchBundle\Service;
-
-use Contao\Database;
-use Smalot\PdfParser\Parser;
-use Symfony\Component\DependencyInjection\ParameterBag\ParameterBagInterface;
-
-class PdfIndexService
-{
-    private string $projectDir;
-
-    public function __construct(ParameterBagInterface $params)
-    {
-        $this->projectDir = rtrim((string) $params->get('kernel.project_dir'), '/');
-    }
-
-    /**
-     * @param array<int,array{url:string,linkText:?string}> $pdfLinks
-     */
-    public function handlePdfLinks(array $pdfLinks): void
-    {
-        // Dedupe nur pro Aufruf (nicht "pro Crawl")
-        $seen = [];
-        $now  = time();
-
-        foreach ($pdfLinks as $row) {
-            $url      = (string) ($row['url'] ?? '');
-            $linkText = $row['linkText'] ?? null;
-
-            if ($url === '') {
-                continue;
-            }
-
-            // doppelte URLs pro Aufruf vermeiden
-            $seenKey = md5($url);
-            if (isset($seen[$seenKey])) {
-                continue;
-            }
-            $seen[$seenKey] = true;
-
-            $normalizedPath = $this->normalizePdfUrl($url);
-            if ($normalizedPath === null) {
-                continue;
-            }
-
-            $absolutePath = $this->getAbsolutePath($normalizedPath);
-            if (!is_file($absolutePath)) {
-                continue;
-            }
-
-            $mtime    = (int) (filemtime($absolutePath) ?: 0);
-            $checksum = md5($normalizedPath . '|' . $mtime);
-
-            // existiert bereits?
-            $existing = Database::getInstance()
-                ->prepare('SELECT checksum FROM tl_search_pdf WHERE url=? LIMIT 1')
-                ->execute($normalizedPath)
-                ->fetchAssoc();
-
-            $needsParse = !$existing || ($existing['checksum'] ?? '') !== $checksum;
-
-            // Titel-Priorität:
-            // 1) Linktext
-            // 2) PDF-Metadaten
-            // 3) Dateiname
-            $title = $linkText ?: basename($absolutePath);
-            $text  = '';
-
-            if ($needsParse) {
-                $pdfMetaTitle = $this->readPdfMetaTitle($absolutePath);
-                $title = $linkText ?: ($pdfMetaTitle ?: basename($absolutePath));
-
-                $text = $this->parsePdf($absolutePath);
-                if ($text === '') {
-                    // wenn parsing fehlschlägt, NICHT überschreiben
-                    continue;
-                }
-            }
-
-            $this->upsertPdf(
-                $normalizedPath,
-                $title,
-                $text,      // kann '' sein → wird in SQL nicht überschrieben
-                $checksum,
-                $mtime,
-                $now
-            );
-        }
-    }
-
-    private function normalizePdfUrl(string $url): ?string
-    {
-        $decoded = html_entity_decode($url);
-        $parts   = parse_url($decoded);
-
-        if (!$parts) {
-            return null;
-        }
-
-        // 1) files/...pdf (ohne führenden Slash)
-        if (
-            !empty($parts['path'])
-            && str_starts_with($parts['path'], 'files/')
-            && str_ends_with(strtolower($parts['path']), '.pdf')
-        ) {
-            return '/' . $parts['path'];
-        }
-
-        // 2) /files/...pdf
-        if (
-            !empty($parts['path'])
-            && str_starts_with($parts['path'], '/files/')
-            && str_ends_with(strtolower($parts['path']), '.pdf')
-        ) {
-            return $parts['path'];
-        }
-
-        if (empty($parts['query'])) {
-            return null;
-        }
-
-        parse_str($parts['query'], $query);
-
-        // 3) Contao 4: ?file=files/...
-        if (!empty($query['file'])) {
-            $file = urldecode((string) $query['file']);
-            $file = ltrim($file, '/');
-
-            if (
-                str_starts_with($file, 'files/')
-                && str_ends_with(strtolower($file), '.pdf')
-            ) {
-                return '/' . $file;
-            }
-        }
-
-        // 4) Contao 5: ?p=...
-        if (!empty($query['p'])) {
-            $p = urldecode((string) $query['p']);
-            return '/files/' . ltrim($p, '/');
-        }
-
-        return null;
-    }
-
-    private function getAbsolutePath(string $relativePath): string
-    {
-        return $this->projectDir . '/' . ltrim($relativePath, '/');
-    }
-
-    private function upsertPdf(
-        string $url,
-        string $title,
-        string $text,
-        string $checksum,
-        int $mtime,
-        int $now
-    ): void {
-        Database::getInstance()
-            ->prepare('
-                INSERT INTO tl_search_pdf
-                    (tstamp, last_seen, type, url, title, text, checksum, file_mtime)
-                VALUES
-                    (?, ?, ?, ?, ?, ?, ?, ?)
-                ON DUPLICATE KEY UPDATE
-                    tstamp     = VALUES(tstamp),
-                    last_seen  = VALUES(last_seen),
-                    type       = VALUES(type),
-                    url        = VALUES(url),
-                    title      = VALUES(title),
-                    checksum   = VALUES(checksum),
-                    file_mtime = VALUES(file_mtime),
-                    text       = IF(VALUES(text) = "" OR VALUES(text) IS NULL, text, VALUES(text))
-            ')
-            ->execute(
-                $now,
-                $now,
-                'pdf',
-                $url,
-                $title,
-                $text,
-                $checksum,
-                $mtime
-            );
-    }
-
-    private function parsePdf(string $absolutePath): string
-    {
-        try {
-            $parser = new Parser();
-            $pdf    = $parser->parseFile($absolutePath);
-            $text   = $this->cleanPdfContent($pdf->getText());
-
-            return mb_substr($text, 0, 20000);
-        } catch (\Throwable) {
-            return '';
-        }
-    }
-
-    private function readPdfMetaTitle(string $absolutePath): ?string
-    {
-        try {
-            $parser  = new Parser();
-            $pdf     = $parser->parseFile($absolutePath);
-            $details = $pdf->getDetails();
-
-            foreach (['Title', 'title'] as $key) {
-                if (!empty($details[$key]) && is_string($details[$key])) {
-                    $t = trim($details[$key]);
-                    if ($t !== '') {
-                        return $t;
-                    }
-                }
-            }
-        } catch (\Throwable) {
-            // ignore
-        }
-
-        return null;
-    }
-
-    private function cleanPdfContent(string $text): string
-    {
-        if (class_exists(\Normalizer::class)) {
-            $text = \Normalizer::normalize($text, \Normalizer::FORM_C) ?? $text;
-        }
-
-        $text = str_replace(["\r\n", "\r"], "\n", $text);
-        $text = preg_replace('/[^\p{L}\p{N}\p{P}\p{Z}\n]/u', ' ', $text);
-        $text = preg_replace('/(?<=\p{L})\s+(?=\p{L})/u', ' ', $text);
-        $text = preg_replace('/\s+/u', ' ', $text);
-
-        return trim($text);
-    }
-}