Parsing XML from a socket that does not close

Question

Hi -
I've seen similar questions to this already posted, but either they did not really apply to my situation or they were not answered.
I have to read messages from a server and process them individually, but the protocol does not indicate the message length or give any sort of terminating character. You only know you have a complete message because it will be a well formed XML fragment. To complicate matters more, there could be extra binary data preceding each XML message. I must stress that I did not write this server, and I have no influence over the protocol at all, so while I certainly agree that this is not such a good protocol, changing it is not an option.
I'm hoping that there is a reasonable way to deal with this with an existing parser. Ironically, I don't really need to parse the XML at all, I just need to know when the current message is over but the only indication I get is that it will be the end of an XML fragment.
I do have the ability to strip off the non-XML binary data, so if there is some way that I can give the stream to a SAX (or other) parser when I know XML is coming and have it inform me when tags begin and end, or ideally inform me when it is done a complete XML fragment, that would be perfect. I'm aware of how to do this using SAX normally, but it seems that it will not function correctly when there is no EOF or other indication that the document has ended.
The best algorithm I have come up with (and it's pretty cheesy) is:
1. Start with a string buffer.
2. Append data from the socket to the buffer one byte at a time.
3. Keep checking if there is a '<' character that is not followed by '?' or '!'. (ie - don't concern myself with the special XML elements that don't close with '/>' or </tagName>. I keep them in the buffer to pass on when I'm done though.)
4. When I get my first tag with <tagName> I make note of what this tag is and increment a counter. If the tag is self closing, then I'm done.
5. Anytime I see this tag I increment the counter.
6. Anytime I see </tagName> I decrement the counter. If the counter = 0, I am done.
7. I pass on the entire message, preceding binary data, special XML tags and the fragment to the module that actually processes it.
This has a few problems. I'll have to go out of my way to support multiple character encodings, I'll have to be careful to catch all the special XML tags, and its quite CPU intensive to be interested in every single character that comes down the pipe (but I suppose this is not avoidable). Also, I just don't like to re-invent the wheel because I'm likely to make an error that a well established parser would not make.
Does anyone have any suggestions for this, or know of a parser that will deal with fragments using streams that don't close?
Thanks!

Answer 1

The parser expects to read to the end of the stream. If you closed the stream right after you wrote to it, I bet it would work. You wouldn't want to close the stream though would you? Try sending the string using a DataOuputStream and calling DataOutputStream.writeUTF(String) on the client side. Then, on the server side call String str = in.readUTF() (where 'in' is a DataInputStream). Then wrap the string in a StringReader and give the StringReader to the parser as it's input source.