Re: [xml-bindings]CDATA nodes have no name?

[Marc-Antoine Parent <maparent acm org>]

Le 03-12-06, à 20:31, Daniel Veillard a écrit :

> On Sat, Dec 06, 2003 at 03:56:14PM -0500, Marc-Antoine Parent wrote:
> > Anyway, could someone tells me if this has been fixed in latter
> > releases, and what behaviour I should expect from them?
>
>   The fix should be to do the test in a different way.

From my side, of course; but from this remark, I fear you may be 
thinking that the problem is only in my code. As it is, the 
python-libxml binding layer crashes before reaching my code, and a fix 
is needed there. Or were you writing this for other maintainers of the 
binding layer? I agree that my attempt at correction was quite lame; It 
should be fixed by looking at the type.

> > But I am interested in another issue, now:
> > In libxml.py:
> > def nodeWrap(o):
> >     # TODO try to cast to the most appropriate node class
> >     name = libxml2mod.name(o)
> >     if name == "element" or name == "text":
> > ....
> >
> > This crashes on CDATA nodes as they have a name of None. (Or NULL from
> > a C standpoint.) I notice that the python code did not change in 
> > 2.6.1.
> > Do CDATA nodes still return a vacuous name?
>
>   Hum, the data type should be checked first,

Indeed.

> > I find I am a bit confused between libxml_name and libxml_type in
> > libxml.c, as nodeWrap seems to use the former, but take results that
> > look as if they came from the latter.
> >
> > My solution was to correct
> >     if name == "element" or name == "text":

By the way, apologies: I had been less than clear. My (hacked) 
correction was to write
    if name == "element" or name == "text" or name == None:
But it would be much better to start with
    name = libxml2mod.type(o)
as that function returns "cdata" correctly.

> > but my CDATA object prints out as <xmlNode (None) object at 0x1061030>
> > and not <xmlNode (cdata) object at 0x1061030> as I would like... 
> > Though
> > getting at its type or calling getContent() gives me a correct answer!
>
>   The type should really be used to make the differentiation, not
> the returned name. For example it's perfectly legal to name element 
> nodes
> "text" like in <text/> ...

Yes, I do use the type in my own code. I do not expect that  
libxmlmod.name(o) should return me the name "cdata", also a valid xml 
element name. However, it would be nice if the repr() of the node gave 
the type accurately, which is another issue.

Thank you for your attention,

Marc-Antoine Parent