RE: How Do I Generate A Set-Difference With Context -

> This continues my earlier post, unfortunately unresponded to, 
> on the same subject.  The original post is here: 
> http://www.biglist.com/lists/xsl-list/archives/200503/msg00332.html
>

It's bad tactics to suggest that you're 80% of the way to a solution, and
not show us the 80%. No-one wants to redo the work you've already done on
the off-chance that they'll be able to help you with the final 20%.

That's especially true as it's a difficult problem and one has to do a lot
of guesswork about the requirements. What output would you expect if the two
source documents are:

<a><b/><c/></a>

and

<a><c/><b/></a>

?

 
> 
> The diff quest came from the following problem: I get 
> periodic XML "feeds" 
> from a news syndicate; these feeds are parsed, formatted in HTML, and 
> published on a website.  Each feed is an XML file, and 
> contains zero or more 
> "stories".  A story may be exactly like that in the 
> immediately-prior feed, 
> may be slightly different, or may be completely new.  Hence 
> my desire to 
> "diff" 2 feeds rather than simply regenerate all stories.  
> When only, say, 
> 20 stories change among 1000+ stories, this is a processing win.

This looks a rather easier problem, because order is irrelevant. 

It's fairly easy, I would have thought, to identify a <story> in one file
for which there is no corresponding <story> in the other. Identifying
finer-grained differences seems to require making some assumptions: what if
one story in the second file is similar to two stories in the first file,
but not identical to either?

> 
> I used an augmented vset:difference from "XSLT Cookbook" ...  I can't 
> think of other 
> algorithmic improvements to make; if anybody else can, please post.

Sorry, but suggesting improvements to code I haven't seen is beyond my
abilities.

Michael Kay
http://www.saxonica.com/