[cxml-devel] Possible bug in cxml-stp

[Raymond Wiker]

On May 22, 2012, at 16:46 , rm at tuxteam.de wrote:
> On Tue, May 22, 2012 at 03:35:30PM +0200, David Lichteblau wrote:
>> Quoting Raymond Wiker (rwiker at gmail.com):
>>> I was confused about empty namespace uris vs empty namespace tags - I
>>> wanted to create nodes that would be in the same namespace as their parent,
>>> but I now see why that would be problematic.
>>> 
>>> Thank you for your patience.
>> 
>> So STP doesn't have a built-in feature for this, but I suppose a helper
>> function could be nice for this purpose.  Maybe something like this:
>> 
>> CL-USER> (defun make-element-in-context (qname context)
>>           (stp:make-element
>>            qname
>>            (stp:find-namespace (cxml::split-qname qname) context)))
>> MAKE-ELEMENT-IN-CONTEXT
>> 
>> CL-USER> (stp:make-element "test" "http://bar")
>> #.(CXML-STP:ELEMENT :LOCAL-NAME "test" :NAMESPACE-URI "http://bar")
>> 
>> CL-USER> (make-element-in-context "child" *)
>> #.(CXML-STP:ELEMENT :LOCAL-NAME "child" :NAMESPACE-URI "http://bar")
>> 
> 
> Well, since the OP already uses a wrapper function to add elements this
> isn't neccessary. And I think the underlying confusion came from the
> (wrong) assumption that an empty namespace (i.e. ns-uri = "") would mean
> "same ns as parent". That's a rather common misconception stemming from
> the difference between the xml data model and it's  serialisaton.
> Namespace prefixes belong to the later.
> BTW: why are you using split-qname in your wrapper? Would you expect to
> feed "foo:bar" to make-element-in-context? 


Thank you both for your suggestions, they will be useful to me when I rewrite my code.

Would it be an idea to add functionality similar to what plexippus-xpath provides via with-namespaces? I.e, a special variable that holds the current set of namespace tag/uri mappings, and have some some additional constructors that use split-qname to extract the tag and look-up the url from the current set of mappings.

Then again, the reason I wrote add-element was to provide an abstraction for the operation of creating a new element with (optional) attributes and (optional) text content and adding it to a parent node. Implementing a mapping mechanism between qnames and uris will not help with this.