fix mixed utf8 and latin in input

1 files changed, 11 insertions, 1 deletions

diff --git a/it_xml.class b/it_xml.class
index ff594fd..bacae95 100644
--- a/it_xml.class
+++ b/it_xml.class
@@ -88,9 +88,13 @@ function from_xml($xmldata, $p)
 		if (!preg_match('/^(<\?xml|\xEF\xBB\xBF|\xFE\xFF|\xFF\xFE|\x00\x00\xFE\xFF|\x00\x00\xFF\xFE)/', $xmldata))
 			$xmldata = '<?xml version="1.0" encoding="' . $this->_p['encoding'] . '"?>' . $xmldata;
 
-		# decode illegal entities but protect semantically important ones
+		# Decode illegal entities but protect semantically important ones
 		$xmldata = html_entity_decode(preg_replace('/&(amp|lt|gt|#38|#60|#62|#x26|#x3C|#x3E);/i', '&amp;$1;', $xmldata), ENT_QUOTES, $this->_p['encoding']);
 
+		# If should be utf-8 and can't be decoded as such, fix it, even if mixed between both
+		if (!preg_match('/^<\?xml[^>]* encoding=/', $xmldata) && preg_match('/[^\x80-\xff][\x80-\xff][^\x80-\xff]/', $xmldata))
+			$xmldata = preg_replace('/[\x80-\xff]{1,4}/e', "it_xml::_utf8_fix('\\0')", $xmldata);
+
 		$result = xml_parse($parser, $xmldata);
 	}
 
@@ -114,6 +118,12 @@ function from_xml($xmldata, $p)
 	return empty($this->error);
 }
 
+# Encode non-utf8 characters in a string, leave utf8 alone
+function _utf8_fix($str)
+{
+	return preg_match('/^([\xc0-\xdf][\x80-\xbf]|[\xe0-\xef][\x80-\xbf][\x80-\xbf]|[\xf0-\xf7][\x80-\xbf][\x80-\xbf][\x80-\xbf])$/', $str) ? $str : utf8_encode($str);
+}
+
 function consume(/* $p */)
 {
 	return false;