Lists Home |
Date Index |
Rick Jelliffe wrote,
> From: "Miles Sabin" <email@example.com>
> > Often it's not so much the class of XPaths that can be matched
> > during a forward parse with no (or limited) state that cause
> > trouble. Often it's what you want to do with the matches when
> > you've got them. For example, if you need to generate output from
> > the matches in some sorted order, then no amount of cleverness on
> > the input side is going to allow you to do without preserving all
> > matches until the arrival of the last.
> You only need to preserve the sort-keys to the matches. You can
> generate them all as (external or internal) entities (or whatever
> inclusion mechanism you are using) and then output the entity
> references in sorted order.
Umm ... sure, but then you have to preserve the external entities for
the entity references to refer to ;-)
The kind of scenarios I had in mind were streaming transformations:
stuff arrives on a wire, leaves on a wire, and we'd like to keep the
amount of state that has to be held on the intermediary to a minimum.
Here it really doesn't make much difference if that state is maintained
as the content of external entities or if it's maintained internally to
allow the subsequent generation of subtrees in the primary output.
> The same goes for including data before it has been encountered.