Re: AW: AW: Constraints on a property in a SKOS hierarchy

On 2021-09-23 5:10 pm, felix@sasakiatcf.com wrote:
>
> That works well, thanks, Holger.
>
> A side note about editing processes and validation.
>
> Then I edit a property and the whole graph is validated, an error like 
> “two instances of the property in one branch” is detected.
>
> Then I edit a property and only the node the property is attached to 
> is validated, errors are only detected if the duplicate node is 
> **above** the node I am editing.
>
> I guess this issue could be handled by adding another sh:property with 
> sh:and :
>
> ex:BranchShape
>     a sh:NodeShape ;
>     sh:targetSubjectsOf ex:myProperty ;
>
> sh:and (
>     sh:property [
>         sh:path ( [ sh:oneOrMorePath skos:broader ] ex:myProperty ) ;
>         sh:maxCount 0 ;
>     ]
>
>     sh:property [
>         sh:path ( sh:inversePath ex:myProperty [ sh:oneOrMorePath 
> skos:broader ]) ;
>         sh:maxCount 0 ;
>     ]
>
> ).
>
First, sh:and is never needed and you can simply pull up the two 
property shapes into ex:BranchShape.

Second, the SHACL standard itself doesn't prescribe what to do in 
incremental validation, although the usual behavior would be to validate 
all constraints that target the changed node(s).

Can't you simply walk the constraint check in both directions, e.g.

ex:BranchShape
     a sh:NodeShape ;
     sh:targetSubjectsOf ex:myProperty ;
     sh:property [ # walk up to parents (broader)
         sh:path ( [ sh:oneOrMorePath skos:broader ] ex:myProperty ) ;
         sh:maxCount 0 ;
     ] ;
     sh:property [ # walk down into children (narrower)
         sh:path ( [ sh:oneOrMorePath [ sh:inversePath skos:broader ] ] 
ex:myProperty ) ;
         sh:maxCount 0 ;
     ] ;

Holger


> Regards
>
> Felix
>
> *Von:*Holger Knublauch <holger@topquadrant.com>
> *Gesendet:* Donnerstag, 23. September 2021 02:29
> *An:* felix@sasakiatcf.com; public-shacl@w3.org
> *Betreff:* Re: AW: Constraints on a property in a SKOS hierarchy
>
> On 2021-09-22 10:00 pm, felix@sasakiatcf.com 
> <mailto:felix@sasakiatcf.com> wrote:
>
>     Thanks a lot for the example, Holger.
>
>     I tried this out. Attached is
>
>      1. your shape testshape.ttl
>      2. a testfile that throws an error, as expected
>         testdata-fails-as-exepected.ttl
>      3. a testfile that does not throw an error:
>         testdata-does-not-fail.ttl .
>
>     The difference between 2) and 3) is that 3) has the same value for
>     the ex:myProperty property. Does this make a difference?
>
> Ah yes, that's a good point. Path expressions will eliminate 
> duplicates, so it will pass the maxCount constraint for this corner case!
>
> If the hierarchy is acyclic, the following should work, doing a 
> skos:broader+ instead of skos:broader*. It basically means
>
> "None of the parent concepts can have a value for ex:myProperty".
>
> ex:BranchShape
>     a sh:NodeShape ;
>     sh:targetSubjectsOf ex:myProperty ;
>     sh:property [
>         sh:path ( [ sh:oneOrMorePath skos:broader ] ex:myProperty ) ;
>         sh:maxCount 0 ;
>     ] .
>
> (This is again untested, as a dry pen-and-paper exercise)
>
> Holger
>
>     I am using the TopQuadrant implementation
>
>     https://github.com/TopQuadrant/shacl/
>     <https://github.com/TopQuadrant/shacl/>
>
>     Regards
>
>     Felix
>
>     *Von:*Holger Knublauch <holger@topquadrant.com>
>     <mailto:holger@topquadrant.com>
>     *Gesendet:* Mittwoch, 22. September 2021 02:14
>     *An:* public-shacl@w3.org <mailto:public-shacl@w3.org>
>     *Betreff:* Re: Constraints on a property in a SKOS hierarchy
>
>     On 2021-09-21 6:57 pm, felix@sasakiatcf.com
>     <mailto:felix@sasakiatcf.com> wrote:
>
>
>         Dear all,
>
>         I would like to to the following:
>
>          1. Define a property “ex:myProperty” for a skos:Concept.
>          2. In a given SKOS taxonomy, the following constraint should
>             be formulated: for the taxonomy, the property should
>             appear zero or more times in one “branch” of the taxonomy.
>
>     If you really mean zero or more times, then this wouldn't
>     constrain anything. My understanding is you want to say zero or
>     one time, i.e. that there can be at most one value per branch?
>
>
>         3.
>
>         The following would be a valid SKOS RDF graph:
>
>         example:animal a skos:Concept; ex:myProperty “some value”.
>
>         example:bird a skos:Concept; skos:broader example:animal.
>
>         The following would be a non-valid SKOS RDF graph, since
>         example:bird is connected to another concept (example:animal)
>         that already has ex:myProperty:
>
>         example:animal a skos:Concept; ex:myProperty “some value”.
>
>         example:bird a skos:Concept; skos:broader example:animal;
>         ex:myProperty “some value”.
>
>     Given the above assumption, could this simply become a maxCount 1
>     on a path expression skos:broader* / ex:myProperty?
>
>     Untested draft:
>
>     ex:BranchShape
>         a sh:NodeShape ;
>         sh:targetSubjectsOf ex:myProperty ;
>         sh:property [
>             sh:path ( [ sh:zeroOrMorePath skos:broader ] ex:myProperty ) ;
>             sh:maxCount 1 ;
>         ] .
>
>     I guess instead of the sh:targetSubjectsOf you could also simply
>     attach this to all skos:Concepts.
>
>     Holger
>
>         How would I formulate 2) in SHACL?
>
>         Regards
>
>         Felix
>