Trying to play nice with SHACL

[Steve Ray]

I'm trying to get comfortable with using SHACL code and wean myself from embedded SPARQL. However, I'm having some trouble mapping certain common SPARQL patterns into the SHACL counterpart. Specifically, here's an example. I'm trying to find all the values of any property that is a subPropertyOf a parent property (c223:hasProperty).

The SPARQL should (I think) be something like this:

c223:PropertiesShape
  rdf:type sh:PropertyShape ;
  sh:path c223:hasProperty ;
  sh:name "PropertiesShape" ;
  sh:values [
      sh:sparql [
          sh:prefixes <http://www.w3.org/2000/01/rdf-schema> ;
          sh:select """SELECT DISTINCT ?prop
                             WHERE {
                             ?property rdfs:subPropertyOf* c223:hasProperty .
                             $this ?property ?prop .
}""" ;
        ] ;
    ] ;
.

I know the following is wrong, but not sure what sh: calls I should be using:

c223:PropertiesShape
  rdf:type sh:PropertyShape ;
  sh:path c223:hasProperty ;
  sh:name "PropertiesShape" ;
  sh:values [
      sh:distinct [
          sh:nodes [
              sh:path (
                  [
                    sh:zeroOrMorePath rdfs:subPropertyOf ;
                  ]
                  c223:hasProperty
                ) ;
            ] ;
        ] ;
    ] ;
.

This gives me a "malformed SHACL expression", so there are definitely problems.

My real question is, how does one duplicate this pattern in SHACL?:

WHERE {

                             ?property rdfs:subPropertyOf* <parentProperty> .
                             $this ?property ?result .

}

Steve

[Irene Polikoff]

There is no path in your example that could get from ?this to ?prop. A path specifies predicates. You do not have predicates to specify - you are finding out your predicates in the first WHERE statement. This is why you need to have 2 statements in the WHERE clause and can’t boil them down to a path expression in a single statement.

--
You received this message because you are subscribed to the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CAGUep87nmSVo5eziHPh6ExbPD1%3DD%3DWrfTNOwNtK0PBww21%3DpgA%40mail.gmail.com.

[Steve Ray]

I agree with you. Is there no way to have "variable" predicates in native SHACL? Or is my only option here to use the embedded SPARQL?

Steve

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/11EA64E3-07D1-4DAF-A609-21C84A739254%40topquadrant.com.

[Holger Knublauch]

On 11/25/2020 8:48 AM, Steve Ray wrote:

I agree with you. Is there no way to have "variable" predicates in native SHACL? Or is my only option here to use the embedded SPARQL?

Exactly. SHACL node expressions do not have a notion of variables. It is intentionally limited to be easier, but then doesn't offer all features that SPARQL does. There is nothing wrong with using SPARQL node expressions as a fallback.

Holger

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CAGUep87Hq%3DPv7rCT0f3aQPU%2BzYo27dA1eodYyGe9ZOoh4D9opA%40mail.gmail.com.

[Steve Ray]

Sorry to keep using up bandwidth here. I have pored over all the documentation I can find on the proper use of sh:values with SPARQL, and clearly I'm not getting something. Here is my trivial toy example that is supposed to just infer a hard-coded value of "Hello" for a property. Can you see what I'm doing wrong?

TBC keeps saying: Property ssh:Property_1: values=(SHACL node expression of unknown type: java.lang.IllegalArgumentException: Malformed SHACL node expression)

 I'm also attaching the full file in case you want to just load it.

ssh:Class_1

a owl:Class ;

a sh:NodeShape ;

rdfs:subClassOf owl:Thing ;

sh:property [

sh:path ssh:Property_1 ;

sh:values [

sh:sparql [

sh:prefixes <http://example.org/sparqlshape> ;

sh:select """SELECT ?result

WHERE {

BIND (\"Hello\" AS ?result) .

}""" ;

] ;

] ;

] ;

.

ssh:Instance_1

a ssh:Class_1 ;

.

ssh:Property_1

a rdf:Property ;

.

Steve

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/41fa8018-8314-3e9b-b854-7fbc1e70f87a%40topquadrant.com.

[Holger Knublauch]

On 11/25/2020 12:18 PM, Steve Ray wrote:

Sorry to keep using up bandwidth here.

No problem. Admittedly, the official spec at https://w3c.github.io/shacl/shacl-af/#select is not exactly full of examples. A failure of the editor :)

I have pored over all the documentation I can find on the proper use of sh:values with SPARQL, and clearly I'm not getting something. Here is my trivial toy example that is supposed to just infer a hard-coded value of "Hello" for a property. Can you see what I'm doing wrong?

TBC keeps saying: Property ssh:Property_1: values=(SHACL node expression of unknown type: java.lang.IllegalArgumentException: Malformed SHACL node expression)

 I'm also attaching the full file in case you want to just load it.

Just drop the sh:sparql node and instead do something like

sh:values [
    sh:select ...
    sh:prefixes ...
]

HTH
Holger

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CAGUep87xZen2zPqno3u%2BVjXKHBCHcx-b8VJeX4cHTnrzP9eE1g%40mail.gmail.com.

[Steve Ray]

Wow, fast!

I was just trying exactly that. I also dropped the prefixes since I wasn't using any. So now I have:

ssh:Class_1

  rdf:type owl:Class ;

  rdf:type sh:NodeShape ;

  rdfs:subClassOf owl:Thing ;

  sh:property [

      sh:path ssh:Property_1 ;

      sh:values [

          sh:select """SELECT ?result

                           WHERE {

                           BIND (\"Hello\" AS ?result) .

                           }""" ;

        ] ;

    ] ;

.

But executing this query:

SELECT *

WHERE {

?subject a ssh:Class_1 .

    ?subject ssh:Property_1 ?object .

}

...is coming up empty.

(Once I get one example working, I'm off to the races!)

Steve

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/6b37a258-8507-22ce-e8e4-384b893c2dd9%40topquadrant.com.

[Holger Knublauch]

On 11/25/2020 12:32 PM, Steve Ray wrote:

Wow, fast!

I was just trying exactly that. I also dropped the prefixes since I wasn't using any. So now I have:

ssh:Class_1

  rdf:type owl:Class ;

  rdf:type sh:NodeShape ;

  rdfs:subClassOf owl:Thing ;

  sh:property [

      sh:path ssh:Property_1 ;

      sh:values [

          sh:select """SELECT ?result

                           WHERE {

                           BIND (\"Hello\" AS ?result) .

                           }""" ;

        ] ;

    ] ;

.

So far so good.

But executing this query:

SELECT *

WHERE {

?subject a ssh:Class_1 .

    ?subject ssh:Property_1 ?object .

}

...is coming up empty.

This is because SPARQL only queries the asserted graph. To query the inferred values, either use GraphQL, Active Data Shapes or use this magic property:

PREFIX teamwork: <http://topbraid.org/teamwork#>
SELECT *
WHERE {
  ...
  (?subject ssh:Property_1 <urn:x-evn-master:geo>) teamwork:values ?object .
}

This takes the master graph that is holding the data in EDG (here: the Geography ontology). This is the fastest variation for performance, but this one here is more general:

  (?subject ssh:Property_1) tosh:values ?object .

This design has some implications, e.g. you normally cannot query inferred values in the inverse direction, e.g. if ?object is given. They are just computed on the fly, for the duration of the query. This means that the system doesn't need to keep track of inferences or invalidate them if the data changes - it's always up to date.

Holger

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CAGUep86jmegagE%3D9W84epw_%3DofMCb6ktcCZa3cVJS7Sq0M27eQ%40mail.gmail.com.

[Steve Ray]

Oh dear. I'm doing this work in support of a standards committee where I'm the only one using TBC. Others are using either Jena, or raw text editors and home-rolled reasoners. (!) (I may be persuading one member to get TBC though).

Of the solutions you are suggesting, which would you recommend that is the most universal? Does Jena support SPIN (i.e. magic properties?). I'm assuming tosh isn't available outside TBC/EDG?

Steve

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/21a18751-29ba-e10c-4c8e-103ff8aa94df%40topquadrant.com.

[Holger Knublauch]

I cannot think of anything in this inferencing area (with SPARQL-like expressiveness) that is fully standardized or widely available.

Holger

To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/CAGUep8435H%3DwTMOHKm1exe-vKMe47FvGuJ3BvDLaMWst_260Qg%40mail.gmail.com.