Styjun

I've run into issue similar to this - I'm trying to parse XML received from TCP without any delimiters and any way of knowing did I receive complete or partial XML. For now, I've taken following approach - keep adding received data to string, try to split it with var xmls = Regex.Split(newData, @"(?=<?xml)"); and try to parse each of strings that came from split - if parsed correctly, it's ok but if not, put the non-complete XML to input buffer and continue. Here is part of receive code that does that:

 string inputBuffer = "";
 private void TcpEndReceive(IAsyncResult result)
 
 lock (tcpLock)
 
 int bytesAvailable = 0;
 try
 
 bytesAvailable = tcpClient.GetStream().EndRead(result);
 
 catch (Exception ex)
 
 log.Error("TcpEndReceive: error endRead, ", ex);
 OnStatusChanged(EReceiverStatus.ConnectionError);
 TcpClose();
 

 if (bytesAvailable >= 0)
 
 try
 
 var newData = System.Text.ASCIIEncoding.ASCII.GetString(tcp_data, 0, bytesAvailable);

 if (!ParseMessage(newData))
 
 var xmls = Regex.Split(newData, @"(?=<?xml)");
 foreach (var xml in xmls)
 
 if (xml.Length == 0) continue;
 if (!ParseMessage(xml))
 
 inputBuffer += xml;
 
 

 if (inputBuffer.Length == 0) return;
 if (ParseMessage(inputBuffer))
 
 inputBuffer = "";
 
 
 
 catch (Exception ex)
 
 log.Error("TcpEndReceive:error reading data, ", ex);
 OnStatusChanged(EReceiverStatus.ConnectionError);
 TcpClose();
 
 
 
 TcpBeginReceive();
 

Now, this may look fine but it isn't - if part of message gets out-of-order, this won't work. Anyone has any idea how to solve this? ParseMessage returns true if message is parsed ok. Also, it can happen that you receive part of message, whole message (yay!) or even multiple messages (XMLs) as single transmit.

Any suggestions, ideas or help is welcome. Request is "simple" - receive full or partial XMLs from TCP stream, parse ones that work and wait for other pieces to arrive for those that can't be parsed. Only reliable "delimiter" you can use is <?xml at beginning of every (valid) XML.

[edit] - Additional clarification - I have no control over data that is sent to me - I can't add separators, delimiters or anything. I can only receive strings that may or may be not complete XML. Some messages contain Base64 encoded images and they don't fit in one message - first you get start of XML, then few ms later you get Base64 image (without any indication that more data follows) and then you get end of your XML. As I said, only reliable signal is <?xml at start of valid XML (and beginning of next one, of course). Also, there is a slight problem that there is multiple types of message, with different root elements, so you can't just deserialize to class Message - you have 10-15 different ones. ParseMessage in my example takes the string, finds root element and first node and depending on it deserializes the message in something useful.