Add temporal relation filter to TREE link traversal

[constraintAutomaton]

Issue type:

• ➕ Feature request

---

Description:

The TREE link traversal at the moment is pretty slow. One way to improve the performance would be to exploit the indexation of members provided by the TREE specification. Indeed the tree:relation can also define a constraint that the following nodes members must respect. This is done using the properties tree:path, tree:value and by specifying a TREE comparison operators as a type for the tree:relation.

Approach Proposed

I proposed to create a new bus called bus-optimize-link-traversal that will provide actors related to query planning/traversal optimization. In the case of this filter, an actor will check if a SPAQL FILTER operator exists, if so it will provide a filter function that will be placed into the context. The TREE traversal will then use the function to reject links that doesn't respect the filter function.

[github-actions]

Thanks for the suggestion!

[pietercolpaert]

Let’s say the data looks like this:

<C> a tree:Collection ;
        tree:view <> ;
        tree:member <A> .

<> a tree:Node ;
     tree:relation <R1> .
     
<R1> a tree:GreaterThanRelation ;
         tree:path dcterms:modified ;
         tree:node 
         tree:value "2022-10-20"^^xsd:dateTIme .
         
<A> dcterms:modified "2022-10-19" .

An a query:

SELECT * WHERE {
   ?s dcterms:modified ?date .
   FILTER (?date < "2022-10-20"^^xsd:dateTime )
}

We’ll need to do 2 things:

1. Path matching

Based on the SPARQL query, we should automatically find the path, and check whether the path matches with the path documented in the relation.

tree:path is based on SHACL property paths: https://www.w3.org/TR/shacl/#property-paths

2. Applying the relation

Probably a big switch case that checks the relation, and executes a comparison function. The functionality might depend on the RDF.DataType. E.G., the implementation of tree:GreaterThanRelation is different for an xsd:dateTime than for a number.

[pietercolpaert]

Other links:

Previous implementation in Comunica v1:

• Substring packages: https://github.com/TREEcg/treemunica/tree/master/packages
• Previous metadata extractor from which you can borrow code by @Dexagod https://github.com/TREEcg/tree-metadata-extraction/blob/master/src/lib/metadataExtraction.ts

Probably we also can use SPARQLEE to reuse things like FILTER functions: (e.G., https://github.com/comunica/sparqlee/blob/860775660d85664e1bec550dad8fb54f8a0a65d8/lib/functions/RegularFunctions.ts)