extension function returning a tree?

[Dieter Stüken]

I have written a java function dumping a database table to XML using SAX.

Is it possible to turn it into an integrated extension function to call it
from within saxon?
The given example only return some atomic value like an number or string.

Is it possible to get a ContentHandler or even a s9api.Destination to send my
content to?
I don't like to build an intermediate DOM tree myself
(even if Saxon will do, as i currently use Saxon-HE which does not support
streaming).

The Saxon SQL extension seems to do to a very similar job, but I currently
have to use HE :-(

Dieter.

[Michael Kay]

The Saxon SQL extension is open source, so you can study the source code.
However, it uses extension instructions which are a rather different
interface.

If you're starting with lexical XML then wrap a StreamSource around it, and
from the context object do context.getConfiguration().buildDocument(source).
The result is a NodeInfo object, which you can turn into a SequenceIterator by
wrapping it in a SingletonIterator. So it's something like:

return new SingletonIterator(context.getConfiguration().buildDocument(new
StreamSource(new StringReader(xml))))

[Barry Demchak]

Very helpful reply ... thanks, Mike! ... I need this, too.

For the other half of the picture, I also need to pass a node into an
ExtensionFunctionCall as a parameter.

Certainly, I can get the parameter value with something like arguments.next(),
but I'm not sure (exactly) what I'm getting.

It looks like maybe it's a NodeInfo. (True??) If so, how do I convert it to
XML?

I have tried getStringValue, but I get text containing only data values
concatenated, not the XML document. How would I get the XML document, too??

Thanks!

[Michael Kay]

NodeInfo is Saxon's representation of a node in an XML tree. If you've got a
NodeInfo and you want to serialize it to lexical XML (ie. angle-brackets), the
simplest way is the static method
net.sf.saxon.query.QueryResult.serialize(NodeInfo node). There are other ways
that give you more control over exactly how the node is serialized (e.g
indentation, XML declaration, etc).

[Barry Demchak]

Thanks, Mike!

That did the trick. The alternative I found was http://markmail.org/message/6
okk3ig4n3b5ehun ... a not-so-
pretty way of doing the same thing. I did find the QueryResult class, and had
hoped it worked as advertised (which, I'm guessing is a pretty packaging of
the not-so-pretty way). Between the documentation, my trying it, and your
response, I'll happily and gratefully move on.

The next step is to convert the XML to a Java object (using XStream). If there
is a quicker round trip between NodeInfo and Java Object, would you let me
know??

(The Java Object will be passed on an Enterprise Service Bus to services that
expect Java Objects.)

Thanks!

[Michael Kay]

If there is a quicker round trip between NodeInfo and Java Object

I don't know anything about XStream, but many XML tools will accept input in
the form of a stream of SAX events. If XStream supplies a SAX ContentHandler
for supplying its input, then you can pump data into it from Saxon easily
enough:

ContentHander ch = XStream.giveMeAContentHandler();
net.sf.saxon.event.ContentHandlerProxy proxy = new
net.sf.saxon.event.ContentHandlerProxy();
proxy.setUnderlyingContentHandler(ch);
proxy.setPipelineConfiguration(node.getConfiguration().makePipelineConfigurati
on());
node.copy(proxy, 0, 0);