Add tests for formatContent() placeholder handling (#712)

Add a performance regression test in tests/Performance and PHPUnit tests covering the changed formatContent()

Author: bernemann

tests/PHPUnit/Unit/PDFObjectTest.php

@@ -95,4 +95,81 @@ public function testTextArrayObjects(): void
         // array.
         self::assertSame([' '], $page4->getTextArray());
     }
+
+    /**
+     * Kerned TJ arrays split a word into many small string operands. Make sure
+     * they end up in the right order again (issue #712).
+     */
+    public function testGetTextArrayReassemblesKernedTjArray(): void
+    {
+        $document = new Document();
+        $document->init();
+
+        $content = 'BT /F1 12 Tf 10 10 Td '
+            . '[(H)10(e)-5(l)3(l)20(o)-40( )30(W)5(o)-3(r)8(l)2(d)]TJ ET';
+
+        $form = new Form($document, null, $content, new Config());
+        $header = new Header([
+            'Resources' => new Header([
+                'XObject' => new Header([
+                    'Fr0' => $form,
+                ])
+            ]),
+            'Contents' => new ElementArray([new Element('/Fr0 Do', $document)], $document),
+        ]);
+        $page = new Page($document, $header);
+
+        self::assertSame(['Hello World '], $page->getTextArray());
+    }
+
+    /**
+     * A << ... >> BDC dictionary around a text block must not swallow the
+     * text that follows it (issue #712).
+     */
+    public function testGetTextArrayRestoresMarkedContentDictionary(): void
+    {
+        $document = new Document();
+        $document->init();
+
+        $content = '/OC << /MCID 0 /Foo (bar) >> BDC '
+            . 'BT /F1 12 Tf 10 10 Td (Hello) Tj ET EMC';
+
+        $form = new Form($document, null, $content, new Config());
+        $header = new Header([
+            'Resources' => new Header([
+                'XObject' => new Header([
+                    'Fr0' => $form,
+                ])
+            ]),
+            'Contents' => new ElementArray([new Element('/Fr0 Do', $document)], $document),
+        ]);
+        $page = new Page($document, $header);
+
+        self::assertSame(['Hello '], $page->getTextArray());
+    }
+
+    /**
+     * A string can hold balanced unescaped parentheses; check they survive
+     * extraction (issue #712).
+     */
+    public function testGetTextArrayKeepsBalancedParenthesesInsideString(): void
+    {
+        $document = new Document();
+        $document->init();
+
+        $content = 'BT /F1 12 Tf 10 10 Td (a(b)c) Tj ET';
+
+        $form = new Form($document, null, $content, new Config());
+        $header = new Header([
+            'Resources' => new Header([
+                'XObject' => new Header([
+                    'Fr0' => $form,
+                ])
+            ]),
+            'Contents' => new ElementArray([new Element('/Fr0 Do', $document)], $document),
+        ]);
+        $page = new Page($document, $header);
+
+        self::assertSame(['a(b)c '], $page->getTextArray());
+    }
 }

tests/Performance/Test/KernedTjArrayFormatContentTest.php

@@ -0,0 +1,75 @@
+<?php
+
+/**
+ * @file This file is part of the PdfParser library.
+ *
+ * @license LGPLv3
+ *
+ * @url     <https://github.com/smalot/pdfparser>
+ */
+
+namespace PerformanceTests\Test;
+
+use PerformanceTests\AbstractPerformanceTest;
+use Smalot\PdfParser\Config;
+use Smalot\PdfParser\Document;
+use Smalot\PdfParser\Element;
+use Smalot\PdfParser\Element\ElementArray;
+use Smalot\PdfParser\Header;
+use Smalot\PdfParser\Page;
+use Smalot\PdfParser\XObject\Form;
+
+/**
+ * PDFs that emit text as kerned TJ arrays (for fine letter spacing) split a
+ * single line into thousands of tiny string operands. formatContent() parks
+ * each operand behind a unique placeholder and restores it afterward.
+ * Restoring them with one str_replace() per placeholder scans the whole
+ * content stream once per operand, i.e. O(operands * length) - quadratic in
+ * the number of operands.
+ *
+ * This test builds a content stream with 20,000 such operands and extracts
+ * its text. With the single-pass strtr() restoration this runs in ~1.5s here;
+ * with the previous per-placeholder str_replace() loop it took ~10s. The time
+ * budget below fails if the quadratic behaviour is reintroduced.
+ *
+ * @see https://github.com/smalot/pdfparser/issues/712
+ */
+class KernedTjArrayFormatContentTest extends AbstractPerformanceTest
+{
+    /**
+     * @var string
+     */
+    protected $content;
+
+    public function init(): void
+    {
+        // Like a PDF that emits text letter-by-letter for fine kerning.
+        $operands = '';
+        for ($i = 0; $i < 20000; ++$i) {
+            $operands .= '(a)'.(($i % 20) - 10).' ';
+        }
+
+        $this->content = 'BT /F1 12 Tf 10 10 Td ['.$operands.']TJ ET';
+    }
+
+    public function run(): void
+    {
+        $document = new Document();
+        $document->init();
+
+        $form = new Form($document, null, $this->content, new Config());
+        $header = new Header([
+            'Resources' => new Header([
+                'XObject' => new Header(['Fr0' => $form]),
+            ]),
+            'Contents' => new ElementArray([new Element('/Fr0 Do', $document)], $document),
+        ]);
+
+        (new Page($document, $header))->getTextArray();
+    }
+
+    public function getMaxEstimatedTime(): int
+    {
+        return 5;
+    }
+}