> I'm afraid I didn't understand your recommendation. I've tried
> changing the subroutine to accept several of the possible types
> (Node, NodeIterator), etc, which worked in that the subroutine
> was still being called. Gave up on that because I wasn't sure
> where to go from there.

Good to hear that you've gotten to that point. You may wind up there again. I 
think Gary explained why the alternative
String parameter wasn't working.

> Are you saying the ONLY way to do this is to walk the tree in
> hardcoded java? If so I'll dig in. But surely there's a way to
> do this in xslt????

Coding a new treewalk? I hope not! (BTW I couldn't find anything about 
XSLT/Xalan multiple output documents - that's the
only approach I know of that would be entirely within XSLT.) If you dig into 
org.apache.xalan.serialize.SerializerToHTML
you might find some way to give it the RTF you got in addPage and make it spit 
out HTML. You'll also have to dump the
stream into a String or buffer of some kind so you can hand it off to the Page 
constructer. Maybe
java.io.ByteArrayOutputStream.

But what does the Page constructor ultimately do with the body string? I'm just 
thinking that if it saves it to a file
or streams it out a network port, there's not point in converting the RTF into 
a string first. Just point the serializer
to the output stream or file. Yikes! It looks like a nasty plumbing job.

As a last resort a two-pass (actually 1+N transforms) approach would have a 
greater likelihood of working. First pass
gets the index information. Second pass iterates over the index from the first 
pass and repeatedly runs a transform with
a parameter passed in to select which page to transform. I've used a similar 
technique where the first pass creates an
Ant build file with a number of Style tasks. I can tell you more about that if 
you like.

Reply via email to