Implement toUnicodeCMap unicode mappings for string literals (#106)

Author: PrinsFrank

src/Document/ContentStream/PositionedText/PositionedTextElement.php

@@ -25,8 +25,10 @@ public function getText(?Font $font): string {
         foreach ($matches as $match) {
             if (str_starts_with($match['chars'], '(') && str_ends_with($match['chars'], ')')) {
                 $chars = LiteralStringEscapeCharacter::unescapeCharacters(substr($match['chars'], 1, -1));
-                if ($font !== null && ($encoding = $font->getEncoding()) !== null) {
+                if (($encoding = $font?->getEncoding()) !== null) {
                     $chars = $encoding->decodeString($chars);
+                } elseif (($toUnicodeCMap = $font?->getToUnicodeCMap() ?? $font?->getToUnicodeCMapDescendantFont()) !== null) {
+                    $chars = $toUnicodeCMap->textToUnicode(bin2hex($chars));
                 }
 
                 $string .= $chars;
@@ -35,7 +37,14 @@ public function getText(?Font $font): string {
                     throw new ParseFailureException('No font available');
                 }
 
-                $string .= $font->toUnicode(substr($match['chars'], 1, -1));
+                $chars = substr($match['chars'], 1, -1);
+                if (($toUnicodeCMap = $font->getToUnicodeCMap() ?? $font->getToUnicodeCMapDescendantFont()) !== null) {
+                    $string .= $toUnicodeCMap->textToUnicode($chars);
+                } elseif (($encoding = $font->getEncoding()) !== null) {
+                    $string .= $encoding->decodeString(implode('', array_map(fn (string $character) => mb_chr((int) hexdec($character)), str_split($chars, 2))));
+                } else {
+                    throw new ParseFailureException('Unable to use CMap or decode string to retrieve characters for text object');
+                }
             } else {
                 throw new ParseFailureException(sprintf('Unrecognized character group format "%s"', $match['chars']));
             }

src/Document/Object/Decorator/Font.php

@@ -82,6 +82,26 @@ public function getToUnicodeCMap(): ?ToUnicodeCMap {
         return $this->toUnicodeCMap = ToUnicodeCMapParser::parse($stream, 0, $stream->getSizeInBytes());
     }
 
+    public function getToUnicodeCMapDescendantFont(): ?ToUnicodeCMap {
+        foreach ($this->getDescendantFonts() as $descendantFont) {
+            $fontDictionary = $descendantFont instanceof Dictionary ? $descendantFont : $descendantFont->getDictionary();
+
+            if (($CIDSystemInfo = $fontDictionary->getValueForKey(DictionaryKey::CIDSYSTEM_INFO, Dictionary::class)) !== null) {
+                $fontResource = RegistryOrchestrator::getForRegistryOrderingSupplement(
+                    $CIDSystemInfo->getValueForKey(DictionaryKey::REGISTRY, TextStringValue::class) ?? throw new ParseFailureException(),
+                    $CIDSystemInfo->getValueForKey(DictionaryKey::ORDERING, TextStringValue::class) ?? throw new ParseFailureException(),
+                    $CIDSystemInfo->getValueForKey(DictionaryKey::SUPPLEMENT, IntegerValue::class) ?? throw new ParseFailureException(),
+                );
+
+                if ($fontResource !== null) {
+                    return $fontResource->getToUnicodeCMap();
+                }
+            }
+        }
+
+        return null;
+    }
+
     /** @throws PdfParserException */
     public function getFirstChar(): ?int {
         return $this->getDictionary()
@@ -225,33 +245,4 @@ public function getFontDescriptor(): ?ReferenceValue {
         return $this->getDictionary()
             ->getValueForKey(DictionaryKey::FONT_DESCRIPTOR, ReferenceValue::class);
     }
-
-    /** @throws PdfParserException */
-    public function toUnicode(string $characterGroup): string {
-        $toUnicodeCMap = $this->getToUnicodeCMap();
-        if ($toUnicodeCMap !== null) {
-            return $toUnicodeCMap->textToUnicode($characterGroup);
-        }
-
-        $descendantFonts = $this->getDictionary()->getObjectsForReference($this->document, DictionaryKey::DESCENDANT_FONTS, Font::class);
-        foreach ($descendantFonts as $descendantFont) {
-            if (($CIDSystemInfo = $descendantFont->getDictionary()->getValueForKey(DictionaryKey::CIDSYSTEM_INFO, Dictionary::class)) !== null) {
-                $fontResource = RegistryOrchestrator::getForRegistryOrderingSupplement(
-                    $CIDSystemInfo->getValueForKey(DictionaryKey::REGISTRY, TextStringValue::class) ?? throw new ParseFailureException(),
-                    $CIDSystemInfo->getValueForKey(DictionaryKey::ORDERING, TextStringValue::class) ?? throw new ParseFailureException(),
-                    $CIDSystemInfo->getValueForKey(DictionaryKey::SUPPLEMENT, IntegerValue::class) ?? throw new ParseFailureException(),
-                );
-
-                if ($fontResource !== null) {
-                    return $fontResource->getToUnicodeCMap()->textToUnicode($characterGroup);
-                }
-            }
-        }
-
-        if (($encoding = $this->getEncoding()) !== null) {
-            return $encoding->decodeString(implode('', array_map(fn (string $character) => mb_chr((int) hexdec($character)), str_split($characterGroup, 2))));
-        }
-
-        throw new ParseFailureException('No ToUnicodeCMap or encoding information available for this font');
-    }
 }