Re: [xml] caching of parsed DTDs ?

[Daniel Veillard <veillard redhat com>]

On Sun, Dec 15, 2002 at 12:51:43AM -0500, Timothy Ritchey wrote:
> I caught this discussion in the archives, and have been struggling with 
> a similar issue. I working on a docbook editor, and have been loading 
> my files using:
>
> xmlDoValidityCheckingDefaultValue = 1;
> xmlRecoverFile(filename);
>
> and was looking to speed up the process. Most of the time appears to be 
> spent in loading the docbook dtd (15 seconds on a OS X G4/500). I was 

  that's not normal. It barely takes more than half a second on my
box and I'm running with a lot of debugging enabled. The is something
going on ...

> thinking of caching the docbook dtd, and reusing it when opening files.
> I am able to load a file, load the dtd separately, and then validate 
> the document as follows:

  This has troubles for document applications, i.e. everything in the
internal subset is not used in such validation, and you're gonna have
troubles with entities...

> xmlDoValidityCheckingDefaultValue = 0;
> xmlDocPtr doc = xmlParseFile(filename);
> xmlDtdPtr dtd = xmlParseDTD(NULL, doc->intSubset->SystemID);
> xmlValidateDtd(&cvp, doc, dtd);
>
> The document then validates fine, but seems to have some elements 
> missing. The first problem was that I was having problems using 
> xmlValidGetValidElements(..) on any nodes from the resulting doc. I was 
> able to get that working by pointing doc->extSubset = dtd. This of 
> course was one of those "waving-a-chicken-leg" moments, in that I have 
> no idea what I did, or why it worked.

  Since the document had no DTD, libxml had no way to find the elements
allowed in the document !

> The second issue is with entities, such as —. My original method 
> of loading the file inserts the entity references fine, but the second 
> doesn't. I am assuming that in the second instance, the original 
> document parsing, upon encountering an entity such as — throws an 
> error, and goes on generating the tree without any reference to the 
> offending entity. So, when I come back later on an do a post 
> validation, there is nothing in the doc tree that even indicates the 
> — ever existed.

  Hum, you should get entities references in the tree but without any
definition associated (since it wasn't present at parsing time).
If you do xmllint --debug on the document you will get the structure
that will be built in memory in such case.

> that I get a well-formed tree with entities.

  You have entities but they are not "resolved".
I bet if you save the document you will get the — in the result,
proof that it was stored internally.

> Any pointers at all would be greatly appreciated.

  GDB is your friend, really. Not necessarilly to find bugs but to see
what you data strutures in memory really are.

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/