ariana_paris@y... wrote:
> Cutting a long story short, I have some files in a rather flat XML
> structure which I now want to upgrade to a more sophisticated schema.

The problem of encapsulation can be solved using SGML if you can write a
DTD which allows missing start-tags and end-tags on the right selection
of elements. For example:

<!DOCTYPE xml [
<!ELEMENT xml - - (chapter+)>
<!ELEMENT chapter O O (title?,section+)>
<!ELEMENT section O O (p|break)*>
<!ELEMENT (title,p) - - (#PCDATA)>
<!ELEMENT break - O EMPTY>
]>
<xml>
   <title>Title 1</title>
     <p>Aaaaaa</p>
     <break>
     <p>Bbbbbb</p>
   <title>Title 2</title>
     <p>Cccccc</p>
</xml>

(converting the <break/> to the SGML format <break>). Running this 
through OSGMLNORM (part of SP, same place onsgmls comes from) using
the SGML Declaration for DocBook 3 with NAMECASE GENERAL NO gives:

<xml>
<chapter>
<title>Title 1</title>
<section>
<p>Aaaaaa</p>
<break>
<p>Bbbbbb</p>
</section>
</chapter>
<chapter>
<title>Title 2</title>
<section>
<p>Cccccc</p>
</section>
</chapter>
</xml>

You just need to convert <break> back to <break/> afterwards.

However, you would need to ensure that all non-ASCII characters are
given as numeric character references, or do some deep surgery on the
SGML Declaration to allow non-ASCII characters.

> The files can have the following structures:
> 
> Example 1:
> <xml>
>   <p>blablabla</p>
> </xml>

Gives:

<xml>
<chapter>
<section>
<p>blablabla</p>
</section>
</chapter>
</xml>


> Example 2:
> <xml>
>   <p>Aaaaaa</p>
>   <break/>
>   <p>Bbbbbb</p>
> </xml>

Gives:

<xml>
<chapter>
<section>
<p>Aaaaaa</p>
<break>
<p>Bbbbbb</p>
</section>
</chapter>
</xml>

> The point being that the files can have various combinations of titles
> and breaks, or none at all. I'm hoping I can get a one-pass solution 

Modulo the caveats on content models and character encoding, this should
work if you get the DTD right.

SGML does still have its uses :-)

///Peter
-- 
XML FAQ: http://xml.silmaril.ie/