Re: [xml] xmllint as filter?

[Elizabeth Mattijsen <liz dijkmat nl>]

At 11:36 AM 11/23/01 -0500, Daniel Veillard wrote:
>   The point is that a concatenation of XML documents is not an XML document.
> And an XML parser, per the spec cannot know when a document finishes, it
> has to come as an indication from the layer feeding it. That's why keeping
> markers in streams of XML document is crucially important, it allows the
> consumer to indicate to the XML parser when the current document finishes.

Which is why we now have a slighly adapted filter for that, that basically 
does this (in Perl):

------------------------------------------------------------------------
# Start the XML output

print <<EOD;
<?xml version="1.0" encoding="$encoding"?>
<docseq>
EOD

# For all of the lines
#  Pass the line through if it is not a processing instruction

while (<STDIN>) {
  print unless m#^<\?xml#;
}

# Print the rest of this XML stream

print <<EOD;
</docseq>
EOD
------------------------------------------------------------------------

The ending of the <docseq> container indicates that the indexer can finish 
up indexing.


> > And if I would like to do a global variable, I would have to tweak
> > ../include/libxml/globals.h , right?
>  right to export the name and the .c too to actually define that variable.

Ok, then I got that right at least...  ;-)

However, it seems that the adaptation of my simple docseq filter will 
handle all of this also (at the expense of a slight overhead).



Elizabeth Mattijsen