Apache commons FileUpload: Pass stream as attribute

Question

I'm trying to implement a Filter that will change all the HTTP parameters to attributs, so that I can handle them in a consistent way throughout my servlets. The apache commons fileupload part is giving me problems. I'm changing the parameters to attributes like this
if(URLENCODED.equalsIgnoreCase(contentType) || contentType == null) // Used with normal form data
  Enumeration<String> params = request.getParameterNames();
  while(params.hasMoreElements())
    String name = params.nextElement();
    String value = request.getParameter(name);
    request.setAttribute(name, value);
else if(contentType.toLowerCase().startsWith(MULTIPART))     // Used when file is uploaded with POST               
  // Create file upload handler
  ServletFileUpload upload = new ServletFileUpload();
  FileItemIterator iter;
  iter = upload.getItemIterator((HttpServletRequest) request);
  while(iter.hasNext())
    FileItemStream item = (FileItemStream)iter.next();
    InputStream itemStream = item.openStream();
    if(item.isFormField())          // form item
      BufferedReader rd = new BufferedReader(new InputStreamReader(itemStream));
      String line;
      while((line = rd.readLine()) != null)
        streamParamsToAttributes(line)
  else          // file
    request.setAttribute("filestream", itemStream);
}In my servlet I use the file stream with following code
InputStream itemStream = (InputStream)(request.getAttribute("filestream"));
useStream(itemStream)Here the useStream() method will throw a FileItemStream.ItemSkippedException since I'm trying to read data from the stream after the iter.next() was called.
So my question is, how to handle this kind of situation? The file can be quite big, so reading it to memory is not wise. And as I read, copying the stream is not a straight-forward task.
I noticed that O'Reilly has also a library for handling file uploads, is it any better in this case?
Thanks.

Answer 1

I looked at the O'Reilly lib and it seems to always save the file to disk. So no streaming possibility.
One solution I can think of would be to 'manually' read the data from the request. I can get the data with request.getReader() call. The file data can be found by the boundary. Now I'm wondering whether this will solve anything. Can anyone see some caveats in this approach?