Fix issue #387 (#401)

smalot · Mar 25, 2021 · 74a074f · 74a074f
1 parent 7d3c67b
commit 74a074f
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 3 deletions.
diff --git a/src/Smalot/PdfParser/PDFObject.php b/src/Smalot/PdfParser/PDFObject.php
@@ -212,8 +212,8 @@ public function getSectionsText($content)
         $textCleaned = $this->cleanContent($content, '_');
 
         // Extract text blocks.
-        if (preg_match_all('/\s+BT[\s|\(|\[]+(.*?)\s*ET/s', $textCleaned, $matches, \PREG_OFFSET_CAPTURE)) {
-            foreach ($matches[1] as $part) {
+        if (preg_match_all('/(\sQ)?\s+BT[\s|\(|\[]+(.*?)\s*ET(\sq)?/s', $textCleaned, $matches, \PREG_OFFSET_CAPTURE)) {
+            foreach ($matches[2] as $pos => $part) {
                 $text = $part[0];
                 if ('' === $text) {
                     continue;
@@ -224,6 +224,10 @@ public function getSectionsText($content)
                 // Removes BDC and EMC markup.
                 $section = preg_replace('/(\/[A-Za-z0-9]+\s*<<.*?)(>>\s*BDC)(.*?)(EMC\s+)/s', '${3}', $section.' ');
 
+                // Add Q and q flags if detected around BT/ET.
+                // @see: https://github.com/smalot/pdfparser/issues/387
+                $section = trim((!empty($matches[1][$pos][0]) ? "Q\n" : '').$section).(!empty($matches[3][$pos][0]) ? "\nq" : '');
+
                 $sections[] = $section;
             }
         }
@@ -270,6 +274,7 @@ public function getText(Page $page = null)
         $text = '';
         $sections = $this->getSectionsText($this->content);
         $current_font = $this->getDefaultFont($page);
+        $clipped_font = $current_font;
 
         $current_position_td = ['x' => false, 'y' => false];
         $current_position_tm = ['x' => false, 'y' => false];
@@ -333,6 +338,16 @@ public function getText(Page $page = null)
                         }
                         break;
 
+                    case 'Q':
+                        // Use clip: restore font.
+                        $current_font = $clipped_font;
+                        break;
+
+                    case 'q':
+                        // Use clip: save font.
+                        $clipped_font = $current_font;
+                        break;
+
                     case "'":
                     case 'Tj':
                         $command[self::COMMAND] = [$command];

diff --git a/src/Smalot/PdfParser/Page.php b/src/Smalot/PdfParser/Page.php
@@ -334,6 +334,7 @@ public function extractDecodedRawData($extractedRawData = null)
             $extractedRawData = $this->extractRawData();
         }
         $currentFont = null;
+        $clippedFont = null;
         foreach ($extractedRawData as &$command) {
             if ('Tj' == $command['o'] || 'TJ' == $command['o']) {
                 $data = $command['c'];
@@ -382,6 +383,10 @@ public function extractDecodedRawData($extractedRawData = null)
                 $fontId = explode(' ', $command['c'])[0];
                 $currentFont = $this->getFont($fontId);
                 continue;
+            } elseif ('Q' == $command['o']) {
+                $currentFont = $clippedFont;
+            } elseif ('q' == $command['o']) {
+                $clippedFont = $currentFont;
             }
         }
 

diff --git a/tests/Integration/PDFObjectTest.php b/tests/Integration/PDFObjectTest.php
@@ -226,7 +226,8 @@ public function testGetSectionText()
 EMC
 (ABC) Tj
 
-[ (a)-4.5(b) 6(c)8.8 ( fsdfsdfsdf[ sd) ] TD ', '/TT1 1.5 Tf (BT )Tj '],
+[ (a)-4.5(b) 6(c)8.8 ( fsdfsdfsdf[ sd) ] TD', '/TT1 1.5 Tf (BT )Tj
+q'],
             $sections
         );
     }