Validation of rdf:List or containers (rdf:Seq)

[u0078867]

I am trying to get the following input validated:

{
  "@context": {
    "@vocab": "https://example.com/"
  },
  "@graph": [
    {
      "@id": "http://schema.org/my_car",
      "@type": "Car",
      "color": {
        "@list": [1, "green"]
      }
    }
  ]
}

with the following SHACL:

@prefix example: <https://example.com/> .
@prefix xsd:    <http://www.w3.org/2001/XMLSchema#> .
@prefix sh:     <http://www.w3.org/ns/shacl#> .
@prefix rdf:    <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .

example:CarShape
        a sh:NodeShape;
        sh:description  "Abstract shape that describes a car entity" ;
        sh:targetClass  example:Car;
        sh:name         "Car";
        sh:property     [ sh:path        example:color;
                          sh:name        "color" ;
                          sh:description "Color of the car" ;
                          sh:class       rdf:List;
                        ];
.

but the validation fails. What am I doing wrong?
Thank you.

[rmfranken]

sh:class is a class constraint component, which means it tries to ask the question: Is the value for this sh:path an instance (rdf:type) of this class (whichever class you have as the object for sh:class). Your data is not rdf:type'ing the car instance my_car as an instance (rdf:type) of the class rdf:List.

If rdf:List were a rdfs:Class, and your car would be of rdf:type rdf:List, this shape would work.

What I think you are trying to do here is validate whether the value for the property example:color is a color. You could do that with sh:in ("red "blue" "green") if the colors are strings (see sh:in in the shacl spec). or you could use sh:class ex:Color if you had created a bunch of instances like
ex:red rdf:type ex:Color

Alternatively, if you are just trying to see whether the value for ex:color is an rdf:list, as far as I know this is not supported by "vanilla" shacl, and you will have to use a SPARQL constraint for that. But I'm sure someone will correct me quickly if that's not the case.

Whenever you post these questions, it also helps to include what you expect to happen. A test data set which should conform your shapes and one which should NOT conform to your shapes.

[u0078867]

Alternatively, if you are just trying to see whether the value for ex:color is an rdf:list, as far as I know this is not supported by "vanilla" shacl, and you will have to use a SPARQL constraint for that. But I'm sure someone will correct me quickly if that's not the case.

That is exactly my goal :)

[rmfranken]

Check out Dash
This might get close to what you want.

dash:ListNodeShape
  a sh:NodeShape ;
  rdfs:comment "Defines constraints on what it means for a node to be a node within a well-formed RDF list. Note that this does not check whether the rdf:rest items are also well-formed lists as this would lead to unsupported recursion." ;
  rdfs:label "List node shape" ;
  sh:or (
      [
        sh:hasValue () ;
        sh:property [
            a sh:PropertyShape ;
            sh:path rdf:first ;
            sh:maxCount 0 ;
          ] ;
        sh:property [
            a sh:PropertyShape ;
            sh:path rdf:rest ;
            sh:maxCount 0 ;
          ] ;
      ]
      [
        sh:not [
            sh:hasValue () ;
          ] ;
        sh:property [
            a sh:PropertyShape ;
            sh:path rdf:first ;
            sh:maxCount 1 ;
            sh:minCount 1 ;
          ] ;
        sh:property [
            a sh:PropertyShape ;
            sh:path rdf:rest ;
            sh:maxCount 1 ;
            sh:minCount 1 ;
          ] ;
      ]
    ) ;
.
dash:ListShape
  a sh:NodeShape ;
  rdfs:comment """Defines constraints on what it means for a node to be a well-formed RDF list.

The focus node must either be rdf:nil or not recursive. Furthermore, this shape uses dash:ListNodeShape as a "helper" to walk through all members of the whole list (including itself).""" ;
  rdfs:label "List shape" ;
  sh:or (
      [
        sh:hasValue () ;
      ]
      [
        sh:not [
            sh:hasValue () ;
          ] ;
        sh:property [
            a sh:PropertyShape ;
            sh:path [
                sh:oneOrMorePath rdf:rest ;
              ] ;
            dash:nonRecursive true ;
          ] ;
      ]
    ) ;
  sh:property [
      a sh:PropertyShape ;
      sh:path [
          sh:zeroOrMorePath rdf:rest ;
        ] ;
      rdfs:comment "Each list member (including this node) must be have the shape dash:ListNodeShape." ;
      sh:node dash:ListNodeShape ;
    ] ;
.

[u0078867]

Thanks.
However, am I correct in saying that either @list in JSON-LD or () in RDF are implicitely of @type (or rdf:type) rdf:List, despite not explicitated?

[rmfranken]

I found this one useful for that:
https://ontola.io/blog/ordered-data-in-rdf

There is NO implicit/expilicit typing in items in a rdf:list, or the list itsself, as it is not a subject.

I'm not 100% about json-ld, but in turtle, using ("item one" "item two" "item three") is just shorthand for:

_:b0 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> "item one" .
_:b0 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:b1 .
_:b1 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> "item two" .
_:b1 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> _:b2 .
_:b2 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> "item three" .
_:b2 <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> <http://www.w3.org/1999/02/22-rdf-syntax-ns#nil> .