xml-dev - Re: How far current XSLT processors will be useful in real applications?

[Steve Harris <sharris@primus.com>]

"Clark C. Evans" <cce@clarkevans.com> writes:

[...]

> I belive that Gopinath 
> is pointing out that the current practice of loading the entire input
> document into memory before proceeding is not a particularly efficient 
> technique in some processing contexts.  I personally agree with
> this assessment.

There was some talk on XML-Dev (a long time ago) about an XSLT
"profile" that could support the limitation of streaming input. That
is, taking away the requirement of caching the whole document to
permit random-access references. I don't recall the discussion getting
much support, being squelched by the "most input documents are small
anyway" argument.

For those who use the standard C++ library's algorithms, XSLT can be
thought of as taking a Forward Iterator as input; you have to be able
to "save" input positions and data. Some of us are interested in the
Input Iterator version, where you cannot "save" a position and
generally strive for constant memory usage.

Michael Rodriguez's XML::Twig Perl module is an example of the "save
only what you're going to need" approach to tree building. This
approach would probably underly a "streaming XSLT profile," though it
still doesn't answer which parts of XSLT would need to be clipped out.

-- 
Steven E. Harris
Primus Knowledge Solutions, Inc.
http://www.primus.com

***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************