debug and rationalize 61970-600-1_Prof10-Header-AP-Con-Complex-SHACL.ttl

[VladimirAlexiev]

https://github.com/Sveino/Inst4CIM-KG/tree/develop/shacl-improved#generalize-shapes

CGMES_Constraints_Repository.xlsx describes PROF10 (aka C:600:ALL:NA:PROF10) thus:

CGMES instance file (distribution) dependency shall be declared by md:Model.DependentOn in the header according to Figure 1 and the associated rules.

It's implemented in application-profiles-library\CGMES\CurrentRelease\SHACL\61970-600-1_Prof10-Header-AP-Con-Complex-SHACL.ttl,
which is 700 lines, half are commented out, and the rest do the same thing:

• select by md:Model.profile (SPARQLTarget)
• check some cardinality restriction on md:Model.DependentOn
• check that md:Model.DependentOn/md:Model.profile has certain values

It is full of useless pairs of sh:SPARQLTarget like this.
The dependencies of a model don't vary with its form (Full vs Difference), so you should target by md:Model.profile but not rdf:type
(The last question below is an exception, and I think it's incorrect)

prof10:FullModel-EQ a  sh:NodeShape ;
        sh:property     prof10:PROF10-EQ;
        sh:target  
        [
              a sh:SPARQLTarget ;
              sh:prefixes cim: ;
              sh:select """
                SELECT DISTINCT ?this
                WHERE {
                  ?this rdf:type md:FullModel   .
                  ?this md:Model.profile "http://iec.ch/TC57/ns/CIM/CoreEquipment-EU/3.0"^^xsd:anyURI.
                }
                """ ;
        ] . 
prof10:DiffModel-EQ a  sh:NodeShape ;
        sh:property     prof10:PROF10-EQ;
        sh:target  
        [
              a sh:SPARQLTarget ;
              sh:prefixes cim: ;
              sh:select """
                SELECT DISTINCT ?this
                WHERE {
                  ?this rdf:type dm:DifferenceModel   .
                  ?this md:Model.profile "http://iec.ch/TC57/ns/CIM/CoreEquipment-EU/3.0"^^xsd:anyURI.
                }
                """ ;
        ] . 

Questions to @Sveino @griddigit-ci @HarisVranaj (edit this comment and check the boxes here):

• To avoid any mistakes: where is this Figure 1 ?
• Are all these similar shapes defined so that this can be done as explained in the excel:

  The validation scope is difficult to define. Different implementations may need to deactivate some SHACL shapes."

Is it true that:

• TP can have any number of dependencies (doesn't restrict maxCount), as soon as one is SSH?
• SC can depend on only one of EQ, EQBD, OP?
• OP can depend on only one of EQ, EQBD, SC?
  • SC can depend on OP, and OP can depend on SC ???
• GL can depend on maximum two of these four: EQ, EQBD, SC, OP
• DiffModel-SC can depend on SC???
  I think that's false: DiffModel-SC is based on SC but doesn't depend on another SC

---

This is better rewritten to ONE SPARQL with a VALUES table as hinted at https://github.com/Sveino/Inst4CIM-KG/tree/develop/shacl-improved#generalize-shapes

[griddigit-ci]

Is now Values generally supported by engines?

The figure is

CGMES instance file (distribution) dependency shall be declared by md:Model.DependentOn in the header according to Figure 1 and the associated rules.

Rules about dependency pairs:

• Diagram Layout [DL] shall be associated with one or more in {Core Equipment [EQ], Topology [TP], Dynamics [DY]}
• Geographical Location [GL] shall be associated with one or more in {Equipment Boundary [EQBD], Core Equipment [EQ]}
• Topology [TP] shall be associated with one or more Steady State Hypothesis [SSH].

More rules"

• The cardinalities of dependencies shown in Figure 1 are the minimum requirement.
• It is allowed, but not required, to include indirect references (references made by the referred instance file).
• It is expected that by including all instance files defined in the DependentOn references and the referred files DependentOn there shall not be any dangling references (e.g. no object is pointing to a non-existing object in the assembled or merged model).

[VladimirAlexiev]

VALUES is supported by any SPARQL engine, and consequently by any SHACL engine with SPARQL.

@griddigit-ci , @Sveino looking at the rules, I think that 5 of 6 of my questions "Is it true that:" are false :-(
In particular, do you agree that the differentiation prof10:FullModel-EQ vs prof10:DiffModel-EQ is wrong?

I think the only way to implement the rules in a "demonstrably correct" way is with a VALUES table.
Eg looking at SV as dependent we can express its possible dependsOn links something like this:

values (?dependent ?min ?max ?dependsOn) {
  (SV 0 1     TP)
  (SV 0 UNDEF SSH)
  (SV 0 UNDEF EQ)
  (SV 0 UNDEF EQBD)
}

Where

• I use the short codes instead of the full profile URLs
• UNDEF represents "max=*"
• Because of transitivity (indirect references), we have to add more rows than the direct arrows in Fig 1