The problem is that you must set the content length in the response header before you start writing any data to the output stream. So your options are:
Write the data to a byte[] array using ByteOutputStream and then copy that to the response output stream once you have the size of the data. However, if you're writing large files, this obviously isn't an option.
Write the data to a temp file and then copy that to the response output once you get the file size. Depending on what you're doing, this may have a performance penalty that is unacceptable.
Depending on how expensive it is to generate the data in the first place, you could generate it once, and throw it away to get the count and then generate it again. Guessing that this is unlikely to be a realistic solution.
Resign yourself to the fact that you won't be able to report the content length in the response header.