questions and recommend books about refactoring of xml-schematas and the transformation of xml-docum

[HL]

Hello everybody
Next year I`ll write my diploma in computer science for business (It`s
a degree in Germany) and I`ve some questions about the topic. The
diploma will handle about refactoring of xml-schemes and a
transformation of all instances (the xml-documents) with XSLT-scripts.

The idea of the diploma is, that in nearly every company xml schemes
are used. And (like Heraklit said "Panta rhei" - everthing is flowing)
the business processes in companys could be changed. If an xml- scheme
has to be changed (refactored), what to do with all instances of these
schemes?
Is it possible to create an XSLT-script that transforms all instances
of the old scheme into instances of the new, refactored scheme? The
diploma will handle about questions like what information you need for
the transformation, what operations are possible for the
transformation, etc.

A simple example is to refactor an xml scheme with one element and to
transform an attribute of the element A to a different element in the
content on the first position of the element A.

I`ll write the diploma next year and all questions are only an
investigation on the topic.

Can anybody recommend some good books about the refactoring of
something like this? I`ve read the book "XML in a Nutshell" and
"Refactoring" (Martin Fowler), but I couldn`t find any information
about this.

What do you think about the topic? Do you see there any big problems ?
I pondered about the topic the last weeks and I think it`s ok, without
big problems. After the analysis of the transformation I`ll design and
implement a tool for the refactoring of these.

Thanks and sorry for my bad english,

PS: if you find mistakes in my text, pleas tell me, so I can correct
them to improve my english

 

[Andy Dingley]

> Is it possible to create an XSLT-script that transforms all instances
of the old scheme into instances of the new, refactored scheme?

No. Isn't this what we spent 2000 discovering we couldn't do?

There are two problems: You can't do it. If you could do it, it's A
Hard Problem to write the XSLT that does do it for you. Both of these
problems are too hard to do practically, for typical examples.

The reasons it's hard are two-fold:

Firstly, XML and XSLT work with opaque nodes of data held in an XML
structure. Although XSLT transforms this XML structure quite easily,
it's hard to work with the data nodes themselves. You _can_ operate on
the nodes too, although XSLT isn't ideal for this - even simple string
manipulation is painful. If dimensions in a node need translation from
imperial to metric measurements then that's implementable in XLST (i.e.
it can and has been done) but it's not a task that XSLT is optmised
for.

Secondly, and more fundamentally, XSLT can only map around XML
structures. It has no conceptual understanding of their underlying
meaning. Making a process for this needs an expression language that
can link an ontological understanding of a node to the concept behind
it (this is more than a mere schema). It also needs some automatic tool
that can take these ontologies and generate appropriate XSLT from them.


More frequently, it recognises that two XML elements both called
"reason-code" and with XML Schema units that dimension them as "length
[feet]" and "length [metres]" may yet be derived from fundamentally
different ontological domains, despire their evident similarity. If the
_concepts_ don't overlap cleanly and equally, then simple structural
transforms just aren't capable of achieving what's needed.

There's a vast body of published work on this topic over the last few
years. Make sure you look at "Schematron" and "OWL" too.

 

[HL]

@ Andy Dingley
Firstly, thanks for you answer

I only post, because I need time to think (and also to discuss with
other persons) about your answer. I`ll give you another answer as quite
as possible