In case of Saxon, would you say that there is some
intermediate representation? I think, the user (or,
programmer) has the view that he/she is just running a XSLT
program in one pass.
If you look at the various APIs - including the JAXP API, it's clear that
there is one operation that creates a Templates object, and another
operation that uses the Templates object (repeatedly, and in multiple
threads, if you wish) to do a transformation. The Templates object is the
"compiled" stylesheet. In Saxon this is a heavily optimized expression tree
rather than a sequence of instructions, but it's conceptually the same.
(Saxon-SA in the case of XQuery also allows you to generate Java code from
the expression tree, in which case the compiled representation of the query
is a Java program.)
I might guess, that prior to creation of the result XDM tree
it's compilation. Does XSLT 2.0 spec define compilation and
runtime phases for XSLT?
Not in so many words, but the distinction is there in the way error handling
is specified, in the distinction between the static and dynamic context, and
in facilities like use-when.
The XPath 2.0 spec includes a processing model that makes a more explicit
distinction. It refers to the two phases as "static analysis" and "dynamic
evaluation". The term "compile" is deliberately avoided as it could be taken
to prejudge implementation strategies. The output of the analysis phase is
called an "operation tree", but there's nothing that says it needs to really
be a tree.
Michael Kay
http://www.saxonica.com/
--~------------------------------------------------------------------
XSL-List info and archive: http://www.mulberrytech.com/xsl/xsl-list
To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
or e-mail: <mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
--~--