Hans Smit wrote:
Hi Dave,
Your points 2) + 3) are interesting. Right now I'm loading XML files
into memory via xercesc. I cache my XML islands in memory
(xercesc::DOMDocument). If the user manipulates the data via my user
interface, I simply serialize and save to file, but retain an in-memory
reference (without reload). I also use this xercesc::DOMDocument as
input to my precompiled XalanTransformer (which I also cache). I guess
you could look at this as a in-memory XML database.
If you allow users to edit documents, then the Xalan-C source tree will
not work for you. Much of its performance advantages come from the fact
that it's read-only, as the source tree is immutable in XSLT,
So my question is: can I skip the xercesc interface aurll together, and
just make use of xalanc? This may simpify my code somewhat, and
hopefully give me a performance boost. The only aspects of xercesc I use
are the W3C DOM interface , the serializer (for writing), and the
XercesDOMParser (for reading).
Not if you need a mutable DOM implementation, unfortunately.
Next Item: the only reason I make use of the xalanc::XercesDOMSupport
and xalanc::XercesDOMSupport are because I need to use XalanDocument for
XPath expressions.
Example,
xalanDocument = parserLiaison.createDocument(domDoc, false, true);
followed by a selectNodeList call.
Question: Is it possible to use the XPath support without making use of
the ParserLiason and DomSupport classes?
Unfortunately, the XPathEvaluator class doesn't encapsulate all of these
details, although it should be possible to do that as an enhancement.
Dave
