+ /**
+ * Try to sanitize the title of the fetched content from wrong character encodings and invalid UTF-8 character.
+ * @param $title
+ * @param $contentType
+ * @return string
+ */
+ private function sanitizeContentTitle($title, $contentType) {
+ if ('application/pdf' === $contentType) {
+ $convertedTitle = $this->convertPdfEncodingToUTF8($title);
+ return $this->sanitizeUTF8Text($convertedTitle);
+ }
+ return $this->sanitizeUTF8Text($title);
+ }
+
+ /**
+ * If the title from the fetched content comes from a PDF, then its very possible that the character encoding is not
+ * UTF-8. This methods tries to identify the character encoding and translate the title to UTF-8.
+ * @param $title
+ * @return string (maybe contains invalid UTF-8 character)
+ */
+ private function convertPdfEncodingToUTF8($title) {
+ // first try UTF-16 (then UTF-8) because its easier to detect its present/absence
+ foreach (array('UTF-16BE', 'UTF-16LE', 'UTF-8', 'WINDOWS-1252') as $encoding) {
+ if (mb_check_encoding($title, $encoding)) {
+ return mb_convert_encoding($title, 'UTF-8', $encoding);
+ }
+ }
+ return $title;
+ }
+
+ /**
+ * Remove invalid UTF-8 characters from the given string.
+ * @param String $rawText
+ * @return string
+ */
+ private function sanitizeUTF8Text($rawText) {
+ if (mb_check_encoding($rawText, 'UTF-8')) {
+ return $rawText;
+ }
+ return iconv("UTF-8", "UTF-8//IGNORE", $rawText);
+ }
+