XOM API for Processing XML with Java

[Wolfgang Hoschek <wolfgang.hoschek AT mac.com>]

On Jan 20, 2006, at 3:09 PM, Elliotte Harold wrote:

> Wolfgang Hoschek wrote:
>
> > I guess there's no SAX property that would force the underlying  
> > parser to limit character() chunks to <= X bytes. So no guarantee  
> > can be given that an OOM error can be avoided. Even if there's no  
> > error with parser X today, it might be different with a different  
> > parser, or with a different version.
>
> The limits of Java set a theoretical limit of 2.1 GB per call. In  
> practice I've never seen a parser go over 64K per call.
>
> > Plus, what about very large base64 comment nodes?
>
> XOM makes one Comment object per SAX comment() invocation. In this  
> case XOM would split the comment up into multiple pieces. The  
> problem here is that it's pretty much impossible to distinguish  
> between two adjacent comments and one long comment; at least  
> without a lot of knowledge about exactly how much text the parser  
> is likely to stuff in a comment.
>
> > I still believe that such an unusual use case is best dealt with  
> > by a custom SAX content handler, rather than at the XOM level. Too  
> > much additional complexity for too little additional value.
>
> Possibly. So far only one person has asked for this. If two people  
> ask for it, I get a little more interested. However this does look  
> like a plausible use case; and it does bother me that XOM can't  
> handle it and SAX can. This contrasts with implausible cases like  
> element names that are two large for a String, which neither XOM  
> nor SAX can handle.
>
> Your points about MIME multiparts are well-taken. However there's  
> work underway to expose these through XOP (not a typo) as an XML  
> document in which the binary parts are base-64 encoded on the fly.  
> I'd like to be able to handle this.

I presume that's going to come in the form of an (optional) byte  
based SAX signature such as

void bytes(byte[] src, int offset, int length)

or similar. With a fallback to the character version if no byte  
handler is present. Right?

Wolfgang.