[Xmldb-org-xupdate] XUpdate processing order

[John L. C. <jl...@po...>]

Uche's recent XUpdate update[0] (heh) has renewed my interest in this
language.  I am very interested in a diff solution for XML.  I had
looked at XUpdate before, but hadn't found anything that seemed
reasonable to use as a diff producing tool.  This time I looked (more
closely) at xmldiff, and things seem more hopeful.  There is clearly a
lot of work to do.  I started hacking on that codebase, and immediately
saw some frustrating gaps and inconsistencies in the definition of
XUpdate as a language.  So, let's get to work, shall we?

This may be obvious, but it seems to me that the XUpdate working draft
could use a section on its processing model (even if that model is
mostly what one would expect).

First, I suppose, one needs to define what is being processed.  I would
say that given two XML documents A and B, an XUpdate document U
describes the difference B - A, that is, a sufficient set of operations
which, when applied to A, produce B.

As a result, the processing model would be that given A and U, one
applies the instructions in U to the content in A to get some B.
Clearly, an XUpdate document consists of a sequence of modifications.
Is the intent that these modifications be applied in document order?  If
so, do each of these modifications apply independently to A, or to an
intermediate step formed from the result of the previous modification?

As an example of the problem, consider the following document:

A:
---
<example>
  <a/> <b/> <b/> <a/>
</example>
---

and XUpdate script:

script.xup:
---
<xupdate:modifications xmlns:xupdate="http://www.xmldb.org/xupdate">
  <xupdate:insert-before select="/example/a[1]">
    <xupdate:element name="c"/>
  </xupdate:insert-before>

  <xupdate:remove select="/example/a[1]"/>

  <xupdate:insert-before select="/example/a[1]">
    <xupdate:element name="c"/>
  </xupdate:insert-before>
</xupdate:modifications>
---

When applying script.xup to A, would this produce:

B1:
---
<example>
  <c/> <c/> <b/> <b/> <a/>
</example>
---

or B2:
---
<example>
  <c/> <b/> <b/> <c/> <a/>
</example>
---

As you can see, B1 would be correct if the modifications in script.xup
applied directly to A, while B2 would be correct if modifications apply
incrementally to the result of previous modifications.  Interestingly,
the former behavior might be more "friendly" for using XUpdate in an
XSLT environment (e.g. implementing xmlpatch, or the application of an
XUpdate script to an in initial file, which is what I first tried to
hack together in XSLT).  I'm not convinced that trying to do this sort
of XUpdate processing with XSLT is advisable, but it sure is tempting,
given the similarities.  Regardless, the B1 behavior may simply be
inconsistent (although I can imagine some distinct weirdness in the B2
case as well); I've haven't investigated it fully.

Take care,

    John L. Clark

[0] http://www.oreillynet.com/pub/wlg/6935