RE: [xml-dev] Disk-based XPath Processing

["Michael Kay" <mike@saxonica.com>]

> 
> If I understood Saxon-SA page correctly, this is what it 
> does: builds trees only if and as necessary, otherwise 
> discards events as it goes.

Sort of. With this option enabled, if you evaluate an expression such as
doc('abc.cml')//section in pull mode, then it will read events sequentially
from the document parser. When it hits a section element it will register a
tree builder that builds a tree corresponding to that section element (the
tricky bit is that it can be building more than one tree at once, because
the section element may be recursive). At the end of the section it will
return the root node of the tree to the  client (again, the tricky bit is
delivering nested trees in document order). If you evaluate the same
expression in push mode, then when it hits a section element it will start
pushing events to the current Receiver (which would typically be a
serializer, or perhaps a user-supplied SAX ContentHandler).

Michael Kay
http://www.saxonica.com/