Re: RDF/XML Tests and new XMLLiteral tests

On Jan 2, 2014, at 1:22 PM, Gavin Carothers <gavin@carothers.name> wrote:

> Also in terms of value equality, DOM 4 isEqualNode and fn:deep-equal http://www.w3.org/TR/xpath-functions-30/#func-deep-equal SHOULD be equivalent (I as a user or implementer would be very surprised if they were not except in pathological cases) but I have no idea if anyone has in fact tested that they are.

I've fixed test008 to add a namespace. My implementation treated ex:xmllit as a relative IRI to the default (XML) namespace. This may be incorrect, but the parser otherwise conforms to both the current and 2004 test suite.

What would be useful is if someone having an implementation using the updated XPath and DOM could come up with what they believe the appropriate serializations to be. As these implementations are not yet wide-spread, it may be a while before all implementations will reproduce those results, assuming that they are actually different than what XML C14NXL would produce.

Gregg

> --Gavin
> 
> 
> On Thu, Jan 2, 2014 at 1:18 PM, Gavin Carothers <gavin@carothers.name> wrote:
> test008
> 
> https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-xml/tests/xmlliteral/test008.rdf
> 
> Is not well formed XML. ex is never defined. I am therefor worried about anything that is currently claiming to pass (or even run) these tests. Also, equivalence for testing would require comparing by value and not by lexical value which requires http://www.w3.org/TR/dom/#dom-node-isequalnode it seems likely this is going to be very hard to test. Some clear guidance from implementers on what they are planning to do would be helpful. For example how is MarkLogic dealing with XML and HTML literals? Given they already have an XPath 3 implementation.
> 
> 
> On Tue, Dec 24, 2013 at 7:31 AM, Gregg Kellogg <gregg@greggkellogg.net> wrote:
> On Dec 24, 2013, at 12:11 AM, "Eric Prud'hommeaux" <eric@w3.org> wrote:
> 
>> On Dec 24, 2013 12:50 AM, "Gregg Kellogg" <gregg@greggkellogg.net> wrote:
>> >
>> > I moved over the 2004 RDF/XML tests [1] using the new manifest vocabulary, and added some extra XMLLiteral tests. The results are based on my implementation, and could be off compared to the new language for generating XMLLiterals, so I'd appreciate a second look.
>> 
>> When we moved to the DOM serialize method, I looked hard at the definition to see how it differed from c14n. The definition for serialize relies on XQuery and XSLT semantics. Is there a mortal-facing definition or example algorithm which you used to see what those literals should look like?
>> 
> No, I presumed that the results must effectively be compatible with 2004 spec using c14n. I know of no implementations available to me of the new XQuery and XSLT. Even if the results are off, these should be good test cases. I'm happy to tweet the results to suit the actual results,
> 
> IMO, and solution must preserve namespaces and language and not mess up included definitions. I presume that the definitions don't need to be minimal (I.e., only limited to those actually used in the fragment). This leaves expansion of self-closing elements, which is entirely speculative.
> 
> I'll look through the relevant specs further myself; perhaps they have their own test suite?
> 
> Gregg
> 
>> > Of course, there's always room for more tests.
>> 
>> >
>> > Note that the tests reference a home directory of <http://www.w3.org/2013/RDFXMLTests/>, which must be set up. and a Wiki page <http://www.w3.org/2011/rdf-wg/wiki/RDFXML_Test_Suite> which has not yet been created.
>> >
>> > Gregg Kellogg
>> > gregg@greggkellogg.net
>> >
>> > [1] https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-xml/tests
> 
>