Re: Any work on an xproc implementation in XSLT?

On Mon, Dec 22, 2008 at 3:51 PM, Alan Painter <> wrote:
> Just a general question on XProc,
> I know that Calabash is based upon the Saxon APIs.
> But I'm curious if anyone (other than yours truly) has been thinking
> about or working on an XSLT implementation of XProc.
> One couldn't do everything in XSLT, but it seems to me that a lot of
> it could be implemented in this way and you would get a lot of the
> implementation "for free".

it may appear that way ... but the heavy costs/price elsewhere will
more then offset any 'free' bits ...

> Try/Catch wouldn't work directly (except perhaps in the Saxonica SA
> extensions) but would require some redirection.

then there is p:director-list,p:xquery, p:exec,p:http-request,
etc...all of which would need extension (though most of these
problematic steps are optional).

the biggest issue will be getting the right approach for

* I can only imagine that implementing p:declare-step will be very difficult
* processing nested subpipelines
* putting things in the correct scope e.g. with p:import and p:library

even simply naming things like steps with defaults (as required by the
spec) will be difficult to get right in XSLT

> Anyone else think that an XSLT implementation would be worth a shot?

I am not saying its impossible, but unless you have some sort of
algorithm or approach that is unique and worth fleshing out I would
say dont do it ... if you want to apply your XSLT chops to something
with xproc then I have a few ideas that may interest you;

* XSLT transform to generate a pipeline diagram in svg
* XSLT transform to  generate xproc documentation from a pipeline
* XSLT transform to  lint cleaner, though I would use schematron for this

hth, Jim Fuller

cheers, Jim Fuller

Received on Tuesday, 23 December 2008 07:44:53 UTC