Support for RDF 1.1 Skolem IRIs

[wouterbeek]

A feature of RDF 1.1 is the optional use of Skolem IRIs, i.e. well-known IRIs that systematically replace blank nodes. SPARQL 1.1 does not include features that support such Skolem IRIs. E.g. queries that use 'isBlank/1' and/or 'isIRI/1' no longer work after Skolemization is applied.

Why?

Skolem IRIs were introduced in RDF 1.1, so after SPARQL 1.1 was finalized. This means that the SPARQL standard was not yet able to provide support for the Skolem IRI feature.

Previous work

I use the following SPARQL Functions myself (note: I am not a programmer and these functions have not been tested):

sparql:isBlankOrSkolemIri
  a sh:SPARQLFunction;
  rdfs:isDefinedBy <https://triplydb.com/Triply/sparql>;
  sh:description "Similar to the isBlank function in the SPARQL standard, but returns true for blank node-replacing well-known IRIs as well."@en;
  sh:parameter param:isBlankOrSkolemIri_term;
  sh:prefixes <https://triplydb.com/Triply/sparql>;
  sh:returnType xsd:boolean;
  sh:select '''
    select $return {
      bind(isBlank($term) || sparql:isSkolemIRI($term) as $return)
    }'''.
param:isBlankOrSkolemIri_term
  a sh:Parameter;
  sh:class rdfs:Resource;
  sh:order 1;
  sh:path param:term.

sparql:isNonSkolemIRI
  a sh:SPARQLFunction;
  rdfs:isDefinedBy <https://triplydb.com/Triply/sparql>;
  sh:description "Similar to the isIRI function in the SPARQL standard, but returns false for blank node-replacing well-known IRIs."@en;
  sh:parameter param:isNonSkolemIRI_term;
  sh:prefixes <https://triplydb.com/Triply/sparql>;
  sh:returnType xsd:boolean;
  sh:select '''
    select $return {
      bind(isIRI($term) && !sparql:isSkolemIRI($term) as $return)
    }'''.
param:isNonSkolemIRI_term
  a sh:Parameter;
  sh:class rdfs:Resource;
  sh:order 1;
  sh:path param:term.

sparql:isSkolemIRI
  a sh:SPARQLFunction;
  rdfs:isDefinedBy <https://triplydb.com/Triply/sparql>;
  sh:parameter param:isSkolemIRI_term;
  sh:prefixes <https://triplydb.com/Triply/sparql>;
  sh:returnType xsd:boolean;
  sh:select '''
    select $return {
      bind(isIRI($term) && strstarts(sparql:iri_path($term), '/.well-known/genid/') as $return)
    }'''.
param:isSkolemIRI_term
  a sh:Parameter;
  sh:class rdfs:Resource;
  sh:nodeKind sh:IRI;
  sh:order 1;
  sh:path param:term.

Proposed solution

Add functions like 'isBlankOrSkolemIri/1', 'isNonSkolemIRI/1', and isSkolemIRI/1' to the language.

[afs]

RDF Concepts:

Systems that want Skolem IRIs to be recognizable outside of the system boundaries SHOULD use a well-known IRI [RFC8615] with the registered name genid.

How much use of skolem IRIs being recognizable outside across system boundaries is there?

[VladimirAlexiev]

@wouterbeek

core feature of RDF 1.1 is the use of Skolem IRIs

What do you mean by Core? This:

In situations where stronger identification is needed, systems MAY systematically replace some or all of the blank nodes in an RDF graph with IRIs. Systems wishing to do this SHOULD mint a new, globally unique IRI (a Skolem IRI) for each blank node so replaced.

If you mean that, I think that "core feature" is an overstatement, notice the MAY above.

Changing the meaning of isBlank, isIRI would not only be a breaking change, but also objectionable:
a Skolem IRI is an IRI, it's not a blank node (although its genesis is from a blank node).
And who is to say that http://example.com/.well-known/genid/d26a2d0e98334696f4ad70a677abc1f6 would never be resolvable?

So the new functions should be called isSkolemIRI, isBlankOrSkolemIRI, isIRIbutNotSkolemIRI (that last one is a mouthful!)

[wouterbeek]

@VladimirAlexiev Thank you for your comment. I have adjusted my original message accordingly.