Re: [xsl] Re: Assignment no, dynamic scoping si (was: Wishes for XSL revisions ...

[Gunther Schadow <gunther@xxxxxxxxxxxxxxxxxxxxxx>]

Dimitre Novatchev wrote:


> In case XSLT is such a bad language to you, what are you 

> doing here in this XSLT list? Would you frequent a Cobol 

> list and tell them how bad their language is?
>
> One alternative is to switch to a better functional language

> (e.g. Haskell), use HaXML and forget about XSLT. This however 

> will likely not appeal to you as Haskell's variables are immutable 

> too...


Dimitre, I'd like to apologize if I hit a nerve. I'm relatively
new to XSLT and I'm enjoying it pretty much. OTOH, I never
understood what's so bad about trolling a little bit. I'm not
advocating some XSLT competitor on "your" list, I'm just poking
around some things that bother me, and I did admit that I could
be wrong. You shouldn't be upset about that. From the other
responses in this thread I realize that this isn't just my
personal issue, and I hate if it just goes off in flames instead
of being a learning experience for those (like me) who are
hitting their head at that corner of XSLT.

It is also not an issue of me just choosing a different language.
I think anyone representing W3C work should be aware of the
great impact you're having. The whole industry has put a huge
investment of faith into you (W3C) this creates a current against
trying to swim is rather pointless. When working in collaborative
projects or other standard activities buildiung on XML, it's not
free personal choice, we are stuck on the whim of W3C to feed us
some good stuff... or just about so.

So, there is absolutely no need for you or anyone representing
XSLT to become defensive against some half-way upset remarks from
folks like me.

> Thompson's book has a small, nice appendix comparing functional, imperative and OO
> programming. Allowing changes of the value of a variable makes imperative languages
> not side effects free. This makes it very difficult or almost impossible to perform
> program verification and program transformation. Another general result of
> imperative languages not being side effects free is that they are much more
> difficult to maintain than side effects free languages. Some nice features, as for
> example lazy evaluation, do not fit well languages with side effects.

There is no question about side-effects and global variables in
the imperative programming style. I didn't ask for any of this
to be added to XSLT. What I was asking for was dynamic scoping,
at least an option to be able to use dynamic scoping for some
variables. I also said that without dynamic scoping I could see
little reason for calling variables variables, because they are
fixed value bindings that I could just simply lexically substitute
wherever they are used (except where an RTF is constructed.)

I'm not sure the hint to mathematics is really a fair answer, because
in mathematics variables are also more than static value bindings.
In languages like Prolog, where assignments to variables don't exist
either, a variable nevertheless has different bindings at different
times during "execution" of the program. But XSLT (as far as I know
it) does not have a construct in which the binding of a variable ever
changes, no unification, etc. The foreach form would be a candidate,
but interestingly (and not at all a reason for complaint to me) XSLT
uses the "current node set" as the "iteration variable" in these
cases.

So, my issue was supposed to be a little more than just ranting.
I found parameters done right, I can see the use for variables
as a handle to constructed node sets (leaving the RTF vs. node set
issue aside.)  I can even see the use of a variable when it is
bound by a select='...' form when you need to get back from
constucted node sets into other node sets. But in any case a
variable is a static name-value binding. One can shadow one such
binding by another using the same name in a statically nested scope,
but that's about all. I keep thinking that dynamic scoping as at
least an optional feature that lets me shadow such bindings along
the call graph (instead of the static structure graph) would be
very, very useful.

I also suspected that may be there is a valid argument against
dynamic scoping, and that you, Dimitre, probably would be able
to tell.

kind regards,
-Gunther


--
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



XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list