default graph triples are sometimes not generated

[pchampin]

When several graph maps are elligible to generate a triple (e.g. one associated to the subject map, and another one associated to the predicate-object map), the ones targetting the default graph (via the special IRI rr:defaultGraph) are ignored.

Example:
data.json

[ { "id": "0", "name": "Alice"  } ] 

map.ttl

@prefix rml: <http://semweb.mmlab.be/ns/rml#> .
@prefix rr: <http://www.w3.org/ns/r2rml#> .
@prefix ql: <http://semweb.mmlab.be/ns/ql#> .
@prefix s: <http://schema.org/> .

 [
    a rr:TriplesMap;
    rml:logicalSource [
      rml:source "data.json" ;
      rml:referenceFormulation ql:JSONPath ;
      rml:iterator "$[*]";
    ];
    rr:subjectMap [
      rr:template "https://example.org/instances/{id}";
      rr:class s:Person ;
      rr:graph <graph:1> ;
    ];
    rr:predicateObjectMap [
      rr:predicate s:givenName ;
      rr:objectMap [ rml:reference "name" ] ;
      rr:graph rr:defaultGraph ;
    ];
  ] .

generates

<https://example.org/instances/0> <http://schema.org/givenName> "Alice" <graph:1> .
<https://example.org/instances/0> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://schema.org/Person> <graph:1> .

I would have expected

<https://example.org/instances/0> <http://schema.org/givenName> "Alice".
<https://example.org/instances/0> <http://schema.org/givenName> "Alice" <graph:1> .
<https://example.org/instances/0> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://schema.org/Person> <graph:1> .

because according to the spec, both graph maps should apply to this predicate-object map.

By the way, if I replace rr:defaultGraph with <graph:2>, the "givenName" triple is indeed generated twice, once in <graph:1> and once in <graph:2>, which is consistent with my expectation above.

Note also that if I swap rr:defaultGraph and <graph:1> in map.ttl, I would expect

<https://example.org/instances/0> <http://schema.org/givenName> "Alice".
<https://example.org/instances/0> <http://schema.org/givenName> "Alice" <graph:1> .
<https://example.org/instances/0> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>  <http://schema.org/Person>.

but instead I get

<https://example.org/instances/0> <http://schema.org/givenName> "Alice" <graph:1> .
<https://example.org/instances/0> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://schema.org/Person> .

[DylanVanAssche]

This is intended. R2RML specification specifies this behavior in its pseudo-code algorithm: https://www.w3.org/TR/r2rml/#generated-triples

If you have a good use case for different behavior, I suggest you open an issue at https://github.com/kg-construct/rml-core.

[pchampin]

@DylanVanAssche I'm sorry but I disagree. Reading https://www.w3.org/TR/r2rml/#generated-triples, step 5.4.5:

Target graphs: If sgm and pogm are empty: rr:defaultGraph; otherwise: union of subject_graphs and predicate-object_graphs

For the givenName property in my example above, we are in the case where neither sgm nor pogm are empty. So Target graphs is the union of subject_graphs (the set { <graph:1> }) and predicate-object_graphs (the set { rr:defaultGraph }), namely { <graph:1>, rr:defaultGraph }, so the triple should be generated in both graphs.

[DylanVanAssche]

Oh I see, sorry for the confusing!

We need proper test cases for all the possible options, otherwise this will keep on happening.

[pchampin]

The example above can be used as a test-case :)

[DylanVanAssche]

Of course! I wonder if there are other cases which are not covered properly with such behavior.

[DylanVanAssche]

Fixed in v7.3.3