Re: [xml] DTDs

[Daniel Veillard <veillard redhat com>]

On Tue, Aug 27, 2002 at 09:50:45AM +0200, Daniel Gehriger wrote:
> >   Fine, very close to catalogs. Another possibility is to not 
> > switch DTD parsing and have a (set of) preparsed DTD(s) and use
> > xmlValidateDtd() to check the documents against the DTD(s) 
> > after parsing.
> > 
> 
> Catalogs ? You mean the external entity load part ? I tried using
> catalogs for this, but I could not find a way to build an in-memory
> catalog. It always seems to require some file to start with. If I could

  yes, I just commented on the similarities of the approaches.

> create an in-memory catalog, then I would map the DTD ID/URL to
> something like "jar:...." or "res:..." (I already wrote the
> corresponding protocol handlers). 

  Hum, that would require a new API. Also the problem is that 
catalogs make heavy use of URI-Reference, and when using
a document coming from memory (and hence without base) it becomes
really fragile, I'm afraid ...

> How would I preparse a DTD ? Looking through the API, I can't find
> anything that would allow me to (a) create an "struct xmlDtd" from a
> DTD, or better yet, load it from a serialized "struct xmlDtd".

  include/libxml/parser.h
    xmlDtdPtr       xmlParseDTD             (const xmlChar *ExternalID,
                                             const xmlChar *SystemID);

is that so hard to find ? Just doing a grep ^xmlDtdPtr include/libxml/*.h ???

> Something else: I install my external entity loader using
> xmlSetExternalEntityLoader(). Before calling this function, I remember
> the previous loader using xmlGetExternalEntityLoader(), which I continue
> to use in my loader if I don't handler the entity. The problem is that
> this requires a global variable. I could not find a way to pass the
> previous loader's function pointer as private (context) information to
> my loader. The "context" argument to my the loader function only seems
> to contain a pointer to the xmlDoc being created. Is there a way to
> split xmlParseFile() in two steps ? Step 1: create an empty document,
> and set some context information (how ?); Step 2: parse the file, and
> make use of this context data in the entity loader ?

  Yes entity reference resolution is not context sensitive, changing
it would be quite complex at this point, sorry it's an application
wide mechanism ...

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/