fix: deduplicate duplicate Kids references in getPages (PR795 stack)

src/Smalot/PdfParser/Document.php

@@ -401,7 +401,7 @@ public function getPages()
             /** @var Pages $object */
             $object = $catalogue->get('Pages');
             if (method_exists($object, 'getPages')) {
-                return $object->getPages(true);
+                return $this->uniquePages($object->getPages(true));
             }
         }
 
@@ -415,19 +415,48 @@ public function getPages()
                 $pages = array_merge($pages, $object->getPages(true));
             }
 
-            return $pages;
+            return $this->uniquePages($pages);
         }
 
         if ($this->hasObjectsByType('Page')) {
             // Search for 'page' (unordered pages).
             $pages = $this->getObjectsByType('Page');
 
-            return array_values($pages);
+            return $this->uniquePages(array_values($pages));
         }
 
         throw new MissingCatalogException('Missing catalog.');
     }
 
+    /**
+     * @param array<Page> $pages
+     *
+     * @return array<Page>
+     */
+    protected function uniquePages(array $pages): array
+    {
+        $unique = [];
+        $seen = [];
+
+        foreach ($pages as $page) {
+            if (!\is_object($page)) {
+                continue;
+            }
+
+            $id = \function_exists('spl_object_id')
+                ? (string) \spl_object_id($page)
+                : \spl_object_hash($page);
+            if (isset($seen[$id])) {
+                continue;
+            }
+
+            $seen[$id] = true;
+            $unique[] = $page;
+        }
+
+        return $unique;
+    }
+
     public function getText(?int $pageLimit = null): string
     {
         $texts = [];

tests/PHPUnit/Integration/DocumentTest.php

@@ -38,6 +38,7 @@
 use PHPUnitTests\TestCase;
 use Smalot\PdfParser\Document;
 use Smalot\PdfParser\Header;
+use Smalot\PdfParser\Parser;
 use Smalot\PdfParser\Page;
 use Smalot\PdfParser\Pages;
 use Smalot\PdfParser\PDFObject;
@@ -233,6 +234,44 @@ public function testGetPagesMissingCatalog(): void
         $document->getPages();
     }
 
+    public function testGetPagesDeduplicatesDuplicateKidsReferences(): void
+    {
+        $document = $this->getDocumentInstance();
+
+        $content = '<</Type/Page>>';
+        $header = Header::parse($content, $document);
+        $page = $this->getPageInstance($document, $header);
+
+        $content = '<</Type/Pages/Kids[10 0 R 10 0 R]>>';
+        $header = Header::parse($content, $document);
+        $pagesNode = $this->getPagesInstance($document, $header);
+
+        $content = '<</Type/Catalog/Pages 20 0 R>>';
+        $header = Header::parse($content, $document);
+        $catalog = $this->getPDFObjectInstance($document, $header);
+
+        $document->setObjects([
+            '10_0' => $page,
+            '20_0' => $pagesNode,
+            '30_0' => $catalog,
+        ]);
+
+        $pages = $document->getPages();
+
+        $this->assertCount(1, $pages);
+        $this->assertSame($page, $pages[0]);
+    }
+
+    /**
+     * @see https://github.com/smalot/pdfparser/pull/795
+     */
+    public function testGetPagesDeduplicatesDuplicateKidsFixture(): void
+    {
+        $document = (new Parser())->parseFile($this->rootDir.'/samples/bugs/PullRequestDuplicateKids.pdf');
+
+        self::assertCount(1, $document->getPages());
+    }
+
     /**
      * @see https://github.com/smalot/pdfparser/issues/721
      */