Abbreviations : `TRANSITIVE`, `SYMMETRIC`, and `INVERSE`

[afs]

The TRANSITIVE, SYMMETRIC, and INVERSE abbreviations in SHACL Rules give a way to write certain patterns in an immediate and direct and clear way.

TRANSITIVE -- this is also useful because it is possible to implement more efficiently than simply following the rule definition shown below.
More efficient rule-based expression requires e.g. matching for direct property relationship. If built-in this isn't necessary, and the direct triples do not appear in the output graph.

It can also be a useful for caching across rule evaluation. e.g. rdfs:subClassOf, rdfs:subPropertyOf.

The SYMMETRIC and INVERSE abbreviations, when used with care, can help declaring expressivity. However, they can generate large numbers of triples if used with common properties.

See also

• https://www.oreilly.com/library/view/semantic-web-for/9780123735560/OEBPS/09_chapter07.htm
• http://mlwiki.org/index.php/RDFS-Plus
• reasoning: transitiveOver (transitive with "step" property) #351

---

Illustration - simple implementations:

TRANSITIVE(:property)

RULE { ?x :property ?z } WHERE  { ?x :property ?y . ?y :property ?z } 

or

RULE { ?x :property ?z } WHERE  { ?x :property/:property ?z } 

SYMMETRIC(:property)

RULE { ?b :property ?a } WHERE  { ?a :property?b } 

INVERSE(:property1, :property2)

RULE { ?b :property2 ?a } WHERE  { ?a :property1 ?b } 
RULE { ?b :property1 ?a } WHERE  { ?a :property2 ?b } 

---

( @TallTed added missing comma + and to first line so SYMMETRIC INVERSE is not misread as one entity. )

[HolgerKnublauch]

What RDF representation would you use for those - or is this only for the native rule format?

[afs]

It's still a "rule", it is syntax for a certain case.

It's important to keep the expressivity equivalence of RDF Rules Syntax and SRL.

The RDF syntax represents the structure of the rule set.

The RDF syntax would have entries such as:

<http://example/ruleSet1>
    rdf:type srlRuleSet ;
    srl:rules (
        [  rdf:type srl:Rule ;
           srl:transitive true ;
        ]
        [ ... next rule ... ]
     ...
    .

Specifically, there isn't use of owl: vocabulary.

DATA is the same - syntax that could be written otherwise. It is axioms/ground facts and is equivalent in effect to a rule with an empty body RULE { ... } WHERE { }.

[HolgerKnublauch]

Good, reusing owl: would have introduced undesirable dependencies and interpretations.

On the specific RDF syntax above, I wanted to raise this before: is the order of rules in the RuleSet important? I am not a huge fan of rdf:Lists in such places... for example makes it difficult to insert values in the middle (from other graphs). Thus sh:order is used elsewhere.

[afs]

is the order of rules in the RuleSet important?

Written order is not important for evaluation. But it might be desirable for clarity. It retains the ordering in a round-trip RDF<->SRL. Humans like to group related things.

Order is not important in the list other than presentation so adding a rule can be done by inserting at the head - the easy case. Another possibility is to allow multiple srl:rules.

We could use , if we have a srl:index. It would loose the visual ordering in RDF without a custom writer.

I don't think using sh:order directly is wise, mixed namespaces make mistakes easier, but we could document encoding synonyms.

I'm not a fan of , for long multiples because it looks like it has ordering, but it doesn't.