XOM API for Processing XML with Java

[Christophe Marchand <christophe.marchand AT axyus.com>]

As far as I know, for such huge files, only a SAX parser may solve your 
problem. I calculate MD5SUM on included gzip/base64 files of about 500Mb 
without any problem. Well, calculating a digital signature is another 
problem...

Christophe

Michael Ludwig wrote:
> Marcin Waldowski schrieb am 07.02.2010 um 12:20:35 (+0100):
>
>   
> > <Object Encoding="http://www.w3.org/2000/09/xmldsig#base64"; Id="1"
> > MimeType="text/plain">
> > U29tZSBmaWxlIGNvbnRlbnQ=
> > </Object>
> >
> > The text node of this tag contains base64 representation of some file.
> > It can be very big file (2GB is a good example). I need to calculate
> > SHA1 of that file.
> >
> > Is it possible to process that tag with XOM in some clever way without
> > using lots of memory and without issuing OutOfMemoryError?
> >     
>
> I think you would want to take control of the document building process,
> which in XOM you can do by subclassing nu.xom.NodeFactory.
>
> However, as a usability improvement over SAX, the handler used for
> dealing with text nodes, NodeFactory#makeText, always gets fed the whole
> text contents in one string argument; so in order to save memory you
> must prevent that handler from ever being called, otherwise the JVM will
> have consumed that much memory - or issued an OutOfMemoryError.
>
> I can't see any way to prevent makeText from being called from within
> XOM. If I'm not mistaken, you need a pull parser in order to control the
> parsing process so you can skip events, while the SAX parser which XOM
> uses underneath is a push parser, where you're only at the receiving end
> of the event queue.
>
> I think you have to descend to the SAX level in order to get the degree
> of control you want - org.xml.sax.ContentHandler#characters is called
> for each chunk of characters, and that's were you could hook in your
> digester.
>
>