I had a similar problem, and while searching thru the archives
( since I was sure it must be a FAQ ) I found a thread with a
good explanation by Ken Holman of why XSLT is NOT an EDITOR
that helped me to think about these problems:
http://www.stylusstudio.com/xmldev/200905/post60080.html
http://www.stylusstudio.com/xmldev/200905/post60090.html
The short version:
* XSLT constructs a new output (write-only) document from an input
(read-only) document.
* XSLT does not process XML syntax -- it processes XML infoset trees.
I suggest anyone who is confused about why you can't do this sort of
thing
in XSLT read that thread. See also:
Amelia Lewis: http://www.stylusstudio.com/xmldev/200905/post10080.html
Michael Kay: http://www.stylusstudio.com/xmldev/200905/post30080.html
However, despite the fact that XSLT was not designed for this sort of
job,
it does get used as an editor or filter of xml documents, for lack of
a better
solution. ( Is XQuery with update intended to fill this gap ? )
[1] We have tried using a perl script that:
grabs and saves everything before the root element,
translates '&' character into some other character that doesn't occur
in the document.
( settable on the command line )
calls out to do an xslt transform ( with Saxon or xsltproc )
pastes the saved prefix onto the output
and does the reverse character translation back to '&'
( There may be some document cases where those steps would fail, but
it works
for our selection. )
[2] It's probably possible to do something similar to the above
strictly in XSLT
by reading the document as both XML and unparsed text. See
Michesl Saxon's article
on "Up-conversion using XSLT" :
http://www.saxonica.com/papers/ideadb-1.1/mhk-paper.xml
for input tips. Pasting the two output streams back together may
also require some tricks.
[ This is probably more complex that the perl solution above, but
doing it all
in XSLT may have workflow advantages in some cases. ]
[3] Other have mentioned Andrew Welch's LexEv, which is an XMLReader
that converts lexical
events (syntax) into markup:
http://ajwelch.blogspot.com/2008/08/lexev-xmlreader-converts-lexical-events.html
This seems like a much cleaner way to handle the pass thru of
syntax elements,
but it requires explicitly encoding the inverse operation in the
stylesheet.
Some examples in:
http://ajwelch.blogspot.com/2008/08/some-sample-templates-for-use-with.html
[4] Before finding the simpler solution (#1 above) I considered
writing a tool that
would pass thru all of the xml unchanged except for fragments
that matched an
xpath expression -- those fragments would be processed thru an
xslt stylesheet.
Does anyone have any other tips on how to handle this sort of problem ?
Are there any XML editors that can work in batch mode using XPath
expressions ?
Is XQuery with update intended to solve this problem ? ( and are there
any good
implementations for this use ? )
[ This may be a bit off topic for xsl-list, but it does seem to be a
FAQ. ]
-- Steve Majewski / UVA Alderman Library