[cddlm] CDL Tests

[Steve Loughran]

Guilherme Germoglio wrote:
> Hello,
> 
> Our test results regarding the cdl parser can be viewed in: 
> http://lsd.ufcg.edu.br/~guiga/cddlm/test_results/index.html
> 
>  From these tests we have changed test cddlm-cdl-2005-09-0005, from 
> import_namespace set because its resolution expected is not correct.
> 
> The expected must be the following:
> 
>   <cdl:system xmlns:test1="http://cddlm.org/test1.cdl"
>             xmlns:t2="http://cddlm.org/test2.cdl">
>           <MyServer>
>             <hostname>www.cddlm.org <http://www.cddlm.org></hostname>
>             <port>80</port>
>           </MyServer>
>   </cdl:system>
> 
> As hostname property was redefined to www.cddlm.org 
> <http://www.cddlm.org> in MyServer , which extends test1:WebServer, this 
> property cannot be the same of test1:WebServer (data/hostname).

you are right. fixed.

> 
> As you can see in the results, 2 required tests remained failing: 
> cddlm-cdl-2005-02-0021 and cddlm-cdl-06-0001
> They regard the resolution of references to lazy properties. We think 
> the tests are not according to the specification, that says in sections 
> 7.3.3 and 7.5.2, respectively:
> 
> "A value reference at /n/ referring to /n'/ is resolvable if and only if:
>      
>       (...)
> 
>     *
> 
>       The node /n'/ and its descendants do not have any @cdl:ref and
>       @cdl:lazy attribute. Otherwise, the resolution is deferred until
>       these @cdl:ref and @cdl:lazy attributes are resolved."
> 
>  
> "A reference to an property with a @cdl:lazy attribute MUST NOT be 
> resolved before the @cdl:lazy attribute is resolved."
> 
> What does the specification mean by "be resolved"? When it states that 
> the reference MUST NOT be resolved, does it mean that the reference must 
> remain unchanged? If so, the tests must be changed. If not, what is the 
> expected behavior?
> 

the cddlm-cdl-2005-02-0021 actually tests that the presence of a 
cdl:lazy attribute is harmless if you have no reference, it just gets 
kept along. There's no reason for it not to be dropped, except that the 
specification explicitly says that only cdl:ref and cdl:refroot 
attributes get stripped when copying references.

The way my implementation works. I am fussy during resolving about 
resolving a lazy reference, that is something with both cdl:ref and 
cdl:lazy. But I just ignore cdl:lazy. I believe this is the correct 
interpretation, but I see from your point that it is unclear; we are 
referring to a property with a cdl:lazy attribute, it just lacks any 
reference.

Personally, I'd like to view cdl:lazy without cdl:ref or cdl:refroot as 
an error, rather than a no-op. But I dont think that is in the spec.


cddlm-cdl-06-0001 looks wrong to me.

As full processing lazy refs are not part of the parse-time spec, we 
could move 02-0021 over to the 06- package and mark all of them as 
informative. But then we do need to come up with some good values for 
production testing, which is where cmp: testing fits in, I hope.

-steve