Re: [xml] xmllint as filter?

[Elizabeth Mattijsen <liz dijkmat nl>]

At 10:54 AM 11/23/01 -0500, Daniel Veillard wrote:
>   concatenating a set of XML files onto a single stream is in the
> majority of the cases I have seen a serious design error.

Ok, maybe it _is_ a design error.  I'm working on search engine software 
that will accept a stream of XML documents as its input for indexing, to 
allow fast fuzzy and exact indexing within very flexible constraints 
<plug>soon to be available for free download at 
http://www.nextrieve.com/</plug>.

The stream of uniformly encoded XML documents, indicates a batch of 
documents to be updated or indexed for the first time.  The overhead of 
calling the indexer for each document seperately, becomes prohibitive for 
larger number of documents.  Hence, the stream format.

We're considering a front-end for the indexing engine which could accept 
single XML documents and then create a stream out of these every X minutes 
or so, but that would still involve a stream of XML documents at that time.

If there is any better way to do this, I'd appreciate knowing about it  ;-)


> > It seems that "xmllint" could be used for that, if it were not for the
> > <?xml processing directive that is always output.
>   simply filter it on top of xmllint with a simple script, this
> sounds quite more sane.

Will do that for now...


> > I've changed the source of xmllint.c to allow for an extra parameter
> > "--noxmlproc", but now find it difficult to actually get that flag setting
> > passed to "xmlDocContentDumpOutput" in tree.c.
>   Libxml2 API is being frozen currently. I'm not found at all about
> adding a new global variable, and totally opposed to change the signature
> of the existing routines especially as you noted the right API already
> exists.

Ok, I agree changing the signature is bad.

And if I would like to do a global variable, I would have to tweak 
../include/libxml/globals.h , right?


Elizabeth Mattijsen

P.S. At this point I don't care whether it would be accepted as a patch or 
not.  I would just like to get a little C hacking experience again...  ;-)