Bugfix

src/EventListener/IndexPageListener.php

@@ -11,6 +11,12 @@ class IndexPageListener
 
     public function onIndexPage(string $content, array &$data, array &$set): void
     {
+        // 🔑 Service einmal pro Crawl initialisieren + Reset ausführen
+        if ($this->pdfIndexService === null) {
+            $this->pdfIndexService = System::getContainer()->get(PdfIndexService::class);
+            $this->pdfIndexService->resetTableOnce();
+        }
+
         // Marker vorhanden?
         if (!str_contains($content, 'MEILISEARCH_JSON')) {
             return;
@@ -53,9 +59,7 @@ class IndexPageListener
             }
 
             foreach (preg_split('/\s+/', trim($src)) as $word) {
-                if ($word !== '') {
+                $keywords[] = $word;
-                    $keywords[] = $word;
-                }
             }
         }
 
@@ -101,17 +105,9 @@ class IndexPageListener
          */
         $pdfLinks = $this->findPdfLinks($content);
 
-        if ($pdfLinks === []) {
+        if ($pdfLinks !== []) {
-            return;
+            $this->pdfIndexService->handlePdfLinks($pdfLinks);
         }
-
-        // Service lazy aus Container holen
-        if ($this->pdfIndexService === null) {
-            $this->pdfIndexService = System::getContainer()->get(PdfIndexService::class);
-            $this->pdfIndexService->resetTableOnce();
-        }
-
-        $this->pdfIndexService->handlePdfLinks($pdfLinks);
     }
 
     /* =====================================================
@@ -130,7 +126,7 @@ class IndexPageListener
     }
 
     /* =====================================================
-     * PDF-Links im Markup finden
+     * PDF-Links im HTML finden
      * ===================================================== */
     private function findPdfLinks(string $content): array
     {
@@ -142,8 +138,6 @@ class IndexPageListener
             return [];
         }
 
-        return array_unique(
+        return array_unique(array_map('html_entity_decode', $matches[1]));
-            array_map('html_entity_decode', $matches[1])
-        );
     }
 }

src/Service/PdfIndexService.php

@@ -8,17 +8,17 @@ use Symfony\Component\DependencyInjection\ParameterBag\ParameterBagInterface;
 
 class PdfIndexService
 {
-    private string $projectDir;
     private bool $tableReset = false;
+    private string $projectDir;
 
     public function __construct(ParameterBagInterface $params)
     {
         $this->projectDir = rtrim($params->get('kernel.project_dir'), '/');
     }
 
-    /* =====================================================
+    /**
-     * Reset tl_search_pdf einmal pro Crawl
+     * 🔥 Wird bei JEDEM Crawl einmal aufgerufen
-     * ===================================================== */
+     */
     public function resetTableOnce(): void
     {
         if ($this->tableReset) {
@@ -26,50 +26,53 @@ class PdfIndexService
         }
 
         Database::getInstance()->execute('TRUNCATE TABLE tl_search_pdf');
-        $this->tableReset = true;
+        error_log('tl_search_pdf wurde geleert');
 
-        error_log('PDF Reset: tl_search_pdf geleert');
+        $this->tableReset = true;
     }
 
-    /* =====================================================
+    /**
-     * Einstiegspunkt aus Listener
+     * Einstiegspunkt vom Listener
-     * ===================================================== */
+     */
     public function handlePdfLinks(array $pdfLinks): void
     {
         foreach ($pdfLinks as $url) {
             try {
-                $normalizedPath = $this->normalizePdfUrl($url);
+                $path = $this->normalizePdfUrl($url);
-                if ($normalizedPath === null) {
+                if ($path === null) {
                     continue;
                 }
 
-                $absolutePath = $this->getAbsolutePath($normalizedPath);
+                $absolutePath = $this->projectDir . '/' . ltrim($path, '/');
                 if (!is_file($absolutePath)) {
                     continue;
                 }
 
-                $mtime = filemtime($absolutePath) ?: 0;
+                $parser = new Parser();
-                $checksum = md5($normalizedPath . $mtime);
+                $pdf = $parser->parseFile($absolutePath);
+                $text = $this->cleanPdfContent($pdf->getText());
 
-                if ($this->alreadyIndexed($checksum)) {
-                    continue;
-                }
-
-                $text = $this->parsePdf($absolutePath);
                 if ($text === '') {
                     continue;
                 }
 
-                $this->insertPdf(
+                Database::getInstance()
-                    $normalizedPath,
+                    ->prepare('
-                    basename($absolutePath),
+                        INSERT INTO tl_search_pdf
-                    $text,
+                            (tstamp, url, title, text, checksum, file_mtime)
-                    $checksum,
+                        VALUES (?, ?, ?, ?, ?, ?)
-                    $mtime
+                    ')
-                );
+                    ->execute(
+                        time(),
+                        $path,
+                        basename($absolutePath),
+                        mb_substr($text, 0, 5000),
+                        md5($path),
+                        filemtime($absolutePath) ?: 0
+                    );
 
             } catch (\Throwable $e) {
-                error_log('PDF Service Fehler: ' . $e->getMessage());
+                error_log('PDF Fehler: ' . $e->getMessage());
             }
         }
     }
@@ -83,9 +86,7 @@ class PdfIndexService
             return $url;
         }
 
-        $decoded = html_entity_decode($url);
+        $parts = parse_url(html_entity_decode($url));
-        $parts = parse_url($decoded);
-
         if (!isset($parts['query'])) {
             return null;
         }
@@ -100,69 +101,18 @@ class PdfIndexService
     }
 
     /* =====================================================
-     * Pfade
+     * Textbereinigung
      * ===================================================== */
-    private function getAbsolutePath(string $relativePath): string
+    private function cleanPdfContent(string $text): string
     {
-        return $this->projectDir . '/' . ltrim($relativePath, '/');
+        if (class_exists(\Normalizer::class)) {
-    }
+            $text = \Normalizer::normalize($text, \Normalizer::FORM_C);
-
-    /* =====================================================
-     * DB
-     * ===================================================== */
-    private function alreadyIndexed(string $checksum): bool
-    {
-        $result = Database::getInstance()
-            ->prepare('SELECT id FROM tl_search_pdf WHERE checksum = ?')
-            ->execute($checksum);
-
-        return $result->numRows > 0;
-    }
-
-    private function insertPdf(
-        string $path,
-        string $title,
-        string $text,
-        string $checksum,
-        int $mtime
-    ): void {
-        Database::getInstance()
-            ->prepare('
-                INSERT INTO tl_search_pdf
-                    (tstamp, url, title, text, checksum, file_mtime)
-                VALUES (?, ?, ?, ?, ?, ?)
-            ')
-            ->execute(
-                time(),
-                $path,
-                $title,
-                $text,
-                $checksum,
-                $mtime
-            );
-    }
-
-    /* =====================================================
-     * PDF-Parsing
-     * ===================================================== */
-    private function parsePdf(string $absolutePath): string
-    {
-        try {
-            $parser = new Parser();
-            $pdf = $parser->parseFile($absolutePath);
-
-            $text = $pdf->getText();
-
-            if (class_exists(\Normalizer::class)) {
-                $text = \Normalizer::normalize($text, \Normalizer::FORM_C);
-            }
-
-            $text = preg_replace('/[^\p{L}\p{N}\p{P}\p{Z}]/u', ' ', $text);
-            $text = preg_replace('/\s+/u', ' ', $text);
-
-            return trim(mb_substr($text, 0, 5000));
-        } catch (\Throwable) {
-            return '';
         }
+
+        $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);
     }
 }