don't dictate the use of blank nodes

VladimirAlexiev commented 3 months ago

@Sveino The SHACL shape cited in https://github.com/Sveino/Inst4CIM-KG/issues/15 demands that a blank node should be used for md:Statement and goes on to call it "nested structure, compound datatype".

However blank nodes have many disadvantages:

Make the graph harder to debug
Cannot be resolved directly, but can only be reached through navigation
Make RDF Canonicalization much harder

They only have two "advantages":

DESCRIBE (though not specified in SPARQL) usually navigates across blank nodes (while computing the Symmetric Concise Bounded Description). But there are more robust ways to "circumscribe" the scope of an entity ("nested structure")
in case the KG creator cannot mint URLs (but CIM uses GUIDs so this is not a problem)

A standard should not dictate the use of blank nodes in a KG.

griddigit-ci commented 2 months ago

I guess this md:Statement is a bit of a side effect. we need to see how much we need to focus on it.

VladimirAlexiev commented 2 months ago

This is used 10 times in 3 files:

grep -r sh:BlankNode .
./source/CGMES/v3.0/SHACL/RDFS/FileHeaderProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES/v3.0/SHACL/RDFS/FileHeaderProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES/v3.0/SHACL/RDFS/FileHeaderProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES/v3.0/SHACL/RDFS/GeographicalLocationProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES/v3.0/SHACL/RDFS/GeographicalLocationProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES/v3.0/SHACL/RDFS/GeographicalLocationProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES/v3.0/SHACL/RDFS/GeographicalLocationProfile.ttl:        sh:nodeKind     sh:BlankNode ;
./source/CGMES-NC/r2.3/ap-con/ttl/Header-AP-Con-SHACL.ttl:        sh:nodeKind     sh:BlankNode;
./source/CGMES-NC/r2.3/ap-con/ttl/Header-AP-Con-SHACL.ttl:        sh:nodeKind     sh:BlankNode;
./source/CGMES-NC/r2.3/ap-con/ttl/Header-AP-Con-SHACL.ttl:        sh:nodeKind     sh:BlankNode;

Please make an analysis how it is used and why.

VladimirAlexiev commented 1 month ago

https://github.com/3lbits/CIM4NoUtility/discussions/321#discussioncomment-10741521 is an example how blank nodes can be harmful: you cannot delete statements involving blank nodes by specifying them in reverseDifferences

griddigit-ci commented 3 weeks ago

Here the analysis

header - they relate to the difference model and this is something we will need to fix in the UML
in the GL it is the validation of the compounds and this is necessary as the compound is parsed as blank node

VladimirAlexiev commented 1 week ago

The newest scan shows these:

source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:                            sh:nodeKind  sh:BlankNode ;
source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:                            sh:nodeKind  sh:BlankNode ;
source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:                            sh:nodeKind  sh:BlankNode ;
source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:                            sh:nodeKind  sh:BlankNode ;
source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:                            sh:nodeKind  sh:BlankNode ;
source/CGMES/v3.0/RDFSEd2Beta/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACLED2a_v3-0-0.ttl:                            sh:nodeKind  sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-552-Header-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-552-Header-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-552-Header-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES/v3.0/SHACL/ttl/61970-600-2_GeographicalLocation-AP-Con-Simple-SHACL_v3-0-0.ttl:        sh:nodeKind     sh:BlankNode ;
source/CGMES-NC/r2.3/ap-con/ttl/Header-AP-Con-SHACL.ttl:        sh:nodeKind     sh:BlankNode;
source/CGMES-NC/r2.3/ap-con/ttl/Header-AP-Con-SHACL.ttl:        sh:nodeKind     sh:BlankNode;
source/CGMES-NC/r2.3/ap-con/ttl/Header-AP-Con-SHACL.ttl:        sh:nodeKind     sh:BlankNode;

I checked 61970-600-2_GeographicalLocation-AP-Con-Simple-SHACL_v3-0-0.ttl and indeed it's about Address compounds.

this is necessary as the compound is parsed as blank node

What do you mean by "parsed as blank node"?

Sveino / Inst4CIM-KG

don't dictate the use of blank nodes #14