Re: Re: Assignment no, dynamic scoping si (was: Wishes

Mike Haarman writes:
> On Fri, 28 Dec 2001, Gunther Schadow wrote:
> > > There is even a simpler approach using only XSLT: I can make all
> > templates use an argument that is a constructed node-set
> > containing name-value pairs (the dynamic environment) and references
> > to the bindings in this environment would be supported by some
> > xpath selects in that environment node-set.
> >
> > Thanks for the insight. Can the dynamic environment be incorporated into
> the internal subset via parameter entities? Such a solution does not
> leave interoperability entirely in crutches.

For each variable for which you would like dynamic scoping behavior,
you could include it as a parameter in all relevant template
definitions and activation sites.  If this is too cumbersome one could
develoop a shorthand for it and employ an XSLT -> XSLT transformation
to expand it.

Yea, that's truly a lot of work and you should either do
this everywhere or nowhere, because the point of dynamic
bindings is that they are available anywhere in the call
graph. So, if you have some template in the middle that has
not passed the binding on because it was believed that this
template would not be called here ... yor dynamic binding
would be lost. If anything is not good XSLT, then the
assumption that certain templates are being invoked in a
certain order.

There's still the issue of being able to override the passed in
"dynamic" value with a new one.  XSLT doesn't allow this though
[http://www.w3.org/TR/xslt#local-variables]:

    A binding shadows another binding if the binding occurs at a point
    where the other binding is visible, and the bindings have the same
    name. It is an error if a binding established by an xsl:variable
    or xsl:param element within a template shadows another binding
    established by an xsl:variable or xsl:param element also within
    the template.

so one would need to do tricky expression groveling and name mangling
to pull it off.

I don't see that as a problem. Really, I have no problem
accepting that a binding cannot be used as a counter etc.
I have no issue with "changing variables" as you change the
content of a storage cell. I really only care about dynamic
scoping so that you can put a few default things in a context
to be able to refer to but also to override as you go down
the execution graph. That can be easily done by just
passing the new binding as the parameter.

The only problem again is that you cannot rely on a few
hacks here and there to add those parameters. You need to
do it everywhere. Indeed, that could be done with an
XSLT on XSLT, but it's something you have to remember
doing and all of that is a lot of hassle and slows
everything down for a feature only needed in a few cases.

regards
-Gunther

Static Types" from Jeffrey Lewis et al. has become pretty
famous for a description of what these dynamic scopes are
good for and how they can be made work nicely.
[http://citeseer.nj.nec.com/246042.html]

--
Gunther Schadow, M.D., Ph.D.                    gschadow@xxxxxxxxxxxxxxx
Medical Information Scientist      Regenstrief Institute for Health Care
Adjunct Assistant Professor        Indiana University School of Medicine
tel:1(317)630-7960                         http://aurora.regenstrief.org