Defining the tree:relation in terms of SPARQL algebra

[constraintAutomaton]

I think it is fair to assume that the intention behind the tree:relation is to use them in the context of SPARQL. The current tree:relation(s) (apart from the geospatial one, but there is an extension to SPARQL for geospatial queries , called GeoSPARQL) are strongly linked to the SPARQL algebra. Thus, to improve the understanding of that relation and to help their implementation inside query engines, it might be a good idea to define them in terms of SPARQL algebra. The task should be simple and not have a big impact on how the current relation has been defined.

I will make soon a proposal.

[constraintAutomaton]

I've produced this simple draft. I know we might want to complexify the string relations and that tree:GreaterThanRelation mention

For string comparison, this relation can refer to a comparison configuration

Draft

An entity that describes a relation between two tree:Nodes.

The tree:Relation has specific sub-classes that implement a more specific type between the values. These types are described in the ontology (all classes are rdf:subClassOf tree:Relation). tree:Relation can be express in term of SPARQL algebra functions.

• String, Date or Number comparison:

  • tree:PrefixRelation — All elements in the related node have this prefix. Must conform to the STRSTARTS SPARQL function.
  • tree:SubstringRelation — All elements in the related node have this substring. Must conform to the SUBSTR SPARQL function.
  • tree:SuffixRelation — All members of this related node end with this suffix. Must conform to the STRENDS SPARQL function.
  • tree:GreaterThanRelation — the related Node’s members are greater than the value. Must conform to the SPARQL Operator Mapping.
  • tree:GreaterThanOrEqualToRelation, tree:LessThanRelation, tree:LessThanOrEqualToRelation, tree:EqualToRelation, tree:NotEqualToRelation — similar to ↑

• Geo-spatial comparison (requires the node values to be WKT-strings):

  • tree:GeospatiallyContainsRelation — Must conform to geof:sfContains

[pietercolpaert]

This looks quite acceptable: can we immediately open this as a PR in the spec?

[constraintAutomaton]

@pietercolpaert Yes, I can make a PR tomorrow.