Re: ISSUE-23: A specific proposal

On Apr 29, 2015 00:35, "Holger Knublauch" <holger@topquadrant.com> wrote:
>
> On 4/28/2015 16:55, Dimitris Kontokostas wrote:
>>
>>
>>
>> On Tue, Apr 28, 2015 at 8:16 AM, Simon Steyskal <simon.steyskal@wu.ac.at>
wrote:
>>>
>>> Hi!
>>>
>>>> since ex:myShExShapeA, ex:myShExShapeB are now an rdfs:Class will
>>>> change behavior and will be treated as class shapes.
>>>> This will trigger validation for all instances of ex:myShExShapeA &
>>>> ex:myShExShapeB instead of the expected
>>>
>>>
>>> as far as I understood, ex:myShExShapeA & ex:myShExShapeB would only be
used by associating individuals via sh:nodeShape to them. They would not be
instantiated by themselves.
>>
>>
>> But if a shape is associated with a class it has different validation
behavior.
>> Or not in this proposal?
>
>
> As far as I see, in this proposal it doesn't have different validation
behavior, at least not in the associated Operations such as
>
> http://w3c.github.io/data-shapes/shacl/#operation-validateGraph
>
> where rdf:type and sh:nodeShape are almost interchangeable. The main
difference, as Simon stated, is that rdf:type should not be used to point
at sh:Shapes - use sh:nodeShape for those cases.

The SPARQL queries for class shapes can be pre-computed and cached while
ShEx shapes depend on the context.
Not being able to determine the behavior in advance is a bad design choice
IMHO.

>
> Holger
>
>
>
>
>>
>>>
>>>
>>> simon
>>>
>>> ---
>>> DDipl.-Ing. Simon Steyskal
>>> Institute for Information Business, WU Vienna
>>>
>>> www: http://www.steyskal.info/  twitter: @simonsteys
>>>
>>>
>>> Am 2015-04-28 08:09, schrieb Dimitris Kontokostas:
>>>>
>>>> ok, assuming we have the following shapes, I am skipping the actual
>>>> constraints as they are not relevant
>>>>
>>>> === before inference ===
>>>>
>>>> ex:myClassA a rdfs:Class . # directly triggers validation for all
>>>> instances of ex:myClassA
>>>>
>>>> ex:myClassB a rdfs:Class ; # directly triggers validation for all
>>>> instances of ex:myClassB
>>>>    rdfs:subClassOf ex:myClassA.  # inherits the constraints of
>>>> ex:myClassA
>>>>
>>>> ex:myShExShapeA a sh:Shape . # works as a linked shape or starting
>>>> node
>>>>
>>>> ex:myShExShapeB a sh:Shape ; # works as a linked shape or starting
>>>> node
>>>>    rdfs:subClassOf ex:myShExShapeB .  # inherits the constraints of
>>>> ex:myShExShapeA
>>>>
>>>> === after inference ===
>>>>
>>>> ex:myClassA a rdfs:Class, a sh:Shape .
>>>>
>>>> ex:myClassB a rdfs:Class, a sh:Shape ;
>>>>    rdfs:subClassOf ex:myClassA.
>>>>
>>>> ex:myShExShapeA a rdfs:Class, sh:Shape . # directly triggers
>>>> validation for all instances of ex:myShExShapeA
>>>>
>>>> ex:myShExShapeB a rdfs:Class, sh:Shape  ;  # directly triggers
>>>> validation for all instances of ex:myShExShapeB
>>>>    rdfs:subClassOf ex:myShExShapeB .
>>>>
>>>> since ex:myShExShapeA, ex:myShExShapeB are now an rdfs:Class will
>>>> change behavior and will be treated as class shapes.
>>>> This will trigger validation for all instances of ex:myShExShapeA &
>>>> ex:myShExShapeB instead of the expected
>>>>
>>>> Again, this is based on how I understood your proposal so far. Maybe I
>>>> missed a detail that could clear things up.
>>>>
>>>> Best,
>>>> Dimitris
>>>>
>>>> On Tue, Apr 28, 2015 at 12:43 AM, Holger Knublauch
>>>> <holger@topquadrant.com> wrote:
>>>>
>>>>> On 4/28/2015 8:41, Dimitris Kontokostas wrote:
>>>>>
>>>>>> Assuming the shapes reside along with the data graph (which is a
>>>>>> possible scenario) and rdfs inference is applied on the graph,
>>>>>> then the ShEx / stand-alone shape will become an rdfs:Class.
>>>>>> In this case, a shape engine will mistaken the stand-alone shape
>>>>>> to a class-shape right?
>>>>>> Or am I missing a detail from your proposal?
>>>>>
>>>>>
>>>>> Yes, a sh:Shape would then also be an instance of rdfs:Class. Why
>>>>> would this be a problem? Any shape can also be a class (in this
>>>>> design).
>>>>>
>>>>> Holger
>>>>
>>>>
>>>> --
>>>>
>>>>  Dimitris Kontokostas
>>>>  Department of Computer Science, University of Leipzig & DBpedia
>>>> Association
>>>> Projects: http://dbpedia.org [1], http://http://aligned-project.eu [2]
>>>> Homepage:http://aksw.org/DimitrisKontokostas [3]
>>>>
>>>> Research Group: http://aksw.org [4]
>>>>
>>>>
>>>>
>>>> Links:
>>>> ------
>>>> [1] http://dbpedia.org
>>>> [2] http://aligned-project.eu
>>>> [3] http://aksw.org/DimitrisKontokostas
>>>> [4] http://aksw.org
>>
>>
>>
>>
>> --
>> Dimitris Kontokostas
>> Department of Computer Science, University of Leipzig & DBpedia
Association
>> Projects: http://dbpedia.org, http://http://aligned-project.eu
>> Homepage:http://aksw.org/DimitrisKontokostas
>> Research Group: http://aksw.org
>>
>