xalan with pull parser 2007-03-16 - By keshlam@(protected)
> FWIW, last time I tried (maybe six months ago, maybe 1 or 2 versions of
> Xerces ago), I found absolutely no performance advantage to performing a
> transform from a SAX source over doing so from a stream source.
If you have something that can produce SAX events directly, using a
SAXSource to accept that will be more efficient than converting from SAX to
XML syntax and then parsing again. Similarly, if you already have a DOM,
using a DOMSource is faster than doing that serialize-and-reparse cycle.
But if you're starting from a URI or other source of XML-syntax text,
SAXSource and StreamSource are essentially the same datapath; both go
through a SAX parser (I don't think we try to use XNI...?) into a DTM
representation and are then processed from there.
<html><body>
<p>> FWIW, last time I tried (maybe six months ago, maybe 1 or 2 versions of
<br>
> Xerces ago), I found absolutely no performance advantage to performing a
<br>
> transform from a SAX source over doing so from a stream source.<br>
<br>
If you have something that can produce SAX events directly, using a SAXSource
to accept that will be more efficient than converting from SAX to XML syntax
and then parsing again. Similarly, if you already have a DOM, using a DOMSource
is faster than doing that serialize-and-reparse cycle.<br>
<br>
But if you're starting from a URI or other source of XML-syntax text, SAXSource
and StreamSource are essentially the same datapath; both go through a SAX
parser (I don't think we try to use XNI...?) into a DTM representation and are
then processed from there.</body></html>
|
|
