[Xmltk-devs] question concerning behavior of TSAX2XML

[Stefanie S. <sch...@ku...>]

Dear XMLTK developers,

I am parsing an XML file and want to "filter" out certain subtrees as a 
whole. The task is very similar to that of xrun, but I allow several 
and also recursive XPath expressions to be registered with the XPath 
processor. (In fact, I only added a loop to xrun.cpp in order to achieve 
this).

However, due to the implementation of TSAX2XML, I do not always get a
"clean"  subtree: e. g. take ./my_xrun / bib/book library.xml
 
For an XML file like 

<bib>
<book comment="ordered">
<auth>
<name>Newcomer</name>
</auth>
<auth>
<name>Bernstein</name>
</auth>
<title>TA Processing</title>
<year>1997</year>
</book>
...

the context-events are invoked at the following places:

 START context 1
<bib
 START context 2
>
<book comment="ordered">
<auth>
<name>Newcomer</name>
</auth>
<auth>
<name>Bernstein</name>
</auth>
<title>TA Processing</title>
<year>1997</year>
</book>
 END context 2
...

It is due to the implementation of TSAX2XML, that the closing brackets are 
sometimes not closed until the next event is invoked. Above, we see that 
the event for context 2 is invoked before the tag-bracket is closed. The 
reason for this is in the following strategie:

bool CTSAX2XML::startElement(XTOKEN xtName)
{
    _CloseBracketIfNeeded();

...


So if I want to take an XML document and write several subtrees in 
different files, I will encounter problems. One example for this task 
would be to write all books from a library into one file, and collect all 
authors that have written books in another file.

If I simply register the paths with the XPath processor, wait for the 
start of a context and write everything until the end of the context into 
a file (using TSAX2XML), the file-content may not be well-formed XML.


I was hoping one could simply plug together existing modules, and that I 
could avoid writing my own version of TSAX2XML. I do think the problem 
will exist for other users as well - filtering out complete subtrees 
should be a common task.

This is my question: Does this mean I can't use TSAX2XML for a task like 
this? Will TSAX2Bin have the same behavior?


Thanks,
Steffi