Thomas, you may not be aware of the existence of Pathan, Gareth's XPath
implementation built on Xerces. (See
http://software.decisionsoft.com/pathanIntro.html.) He's definitely a good
guy to work with on this project.

And, for what it's worth, I'd love to see XPath as part of Xerces. I don't
need all the goodies that Xalan offers, but I'd like to be able to use to
implement at least a subset of the XML signature spec, and XPath would be
useful. (The other major requirement that Xerces currently lacks is
canonicalization.)

-----Original Message-----
From: Gareth Reakes [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, October 01, 2002 5:35 AM
To: [EMAIL PROTECTED]
Subject: Re: Experiences (Re: upgrading to Xalan 1.4 and Xerces 2.1)


Hi,

>     1) There are a few const-ness glitches or a lost reference between all

> the pointers. I'll soon post these as bug (or is there already such a
bug?)

I think we have sorted all these out now. Have you checked the cvs tree?

>     2) some of the exposed methods aren't yet implemented in Xerces, in 
> particular the DOM textContent method: i do not understand why the getter 
> returns always NULL whereas the setter throws not implemented. I'd expect 
> the getter to throw not implemented as well.

I would assume that is because returning null is correct in some cases and 
it just has not been overridden in derived classes to throw unsupported. I 
agree its confusing.

>     3) The memory model isn't well documented. What does release()? Is it 
> basically a delete this, or do i have to delete a relased() pointer as
well?

The application calls release when it knows that it does not want that 
node any more. You can not call release when the node is still part of the 
tree. The memory is then recycled. You therefore have to be very careful 
when calling release as any other pointers to that node will could just be 
pointing to a different node completely (you can get some really fun 
behaviour this way:)).

>     4) I don't get NODE_DELETED calls when a whole document is released
(or 
> a DOMBuilder). This can be fatal for wrapper classes having a 'lazier' 
> memory model such as COM. I will probably attach this one to my feature 
> request about NODE_DELETED already present at bugzilla (all UserData items

> should be notified before the items are removed from the tree).

>     Now about porting XPath: i'd like to see for XPath an Apache c++ 
> binding as well, conformant to the DOM L3 XPath WD. This in some way can
be 
> a bitch, since XPath is part of xsl transform (Xalan) as well as a DOM 
> specification (Xerces). It is clear that having two XPath libraries should

> be avoided.

I agree. We should definitely think about this. There has been some talk 
about importing Pathan into xerces-c. 

We are currently implementing XPath2 using the new bindings and I agree 
with your comments.


Gareth

-- 
Gareth Reakes, Head of Product Development  
DecisionSoft Ltd.            http://www.decisionsoft.com
Office: +44 (0) 1865 203192



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to