Hi Radu,
   Please see my comments inline,

On Mon, Feb 3, 2014 at 6:58 PM, Radu Coravu <[email protected]> wrote:

>
>  The current XPath library on xerces SVN is fully compliant to W3C XPath
>> 2.0 spec.
>>
>
> The XPath 2.0 specification states that an XPath QName:
>
>  http://www.w3.org/TR/xpath20/#prod-xpath-QName
>>
>
> has a prefix and a local part which is of type NCName:
>
>  http://www.w3.org/TR/REC-xml-names/#NT-NCName
>>
>
> and the NCName needs to be composed of characters which obey the
> specification:
>
>  http://www.w3.org/TR/REC-xml/#NT-Char
>>
>
> If the XPath parser included with the Xerces XML Schema 1.1 would be fully
> compliant with the XPath 2.0 specs, it would be able to accept as valid
> most of the unicode characters when used in NCNames


I agree with you. I think, Xerces's XSD 1.1 engine currently supports an
ASCII like character set for XML names and values of attributes. It
probably supports certain unicode characters for XML names, but I'm not
sure which ones (we need to check the implementation details again, within
psychopath xpath 2.0 engine).

but it has trouble even with simple characters like "ü" and it breaks when
> validating assertions like:
>
>
>  <xs:assert test="count(le:Einführung) le 1"/>
>>
>
I agree, an XPath 2.0 processor needs to support most of unicode
characters. Implementing XPath 2.0 spec is quite overwhelming. But I think,
Xerces has done a very good job, of implementing (actually, starting from
adopting the original psychopath xpath 2.0 engine and then improving it)
nearly every functional aspect of XPath 2.0.


>
> So in my opinion the XPath library currently bundled with Xerces is not
> fully compliant to the W3C XPath 2.0 spec.


I agree, its technically true. But we're very close to a near compliance,
except for a wider unicode support as you've pointed. I'm not a personal
expert on unicode implementation, but I find this area interesting. May be
someone else can take this task in near term, for improving Xerces's XPath
2.0 unicode support.

As has been discussed few times, on this forum why we can't quickly adopt
the current Eclipse psychopath XPath 2.0 engine. Mainly because, of JDK
level mismatches and also difference in spec functional compliance of XPath
2.0 engine on Xerces SVN and on Eclipse SE project. I think, there are few
areas (like few F&O implementations) where XPath 2.0 engine on Xerces SVN
complies better to the XPath 2.0 spec (that's my personal opinion).


Regards,
Mukul

Reply via email to