Feature: indexing with complex paths

[jeswr]

It would be nice to index using complex SPARQL paths, for instance

const ruben = path.create({ subject: namedNode('https://ruben.verborgh.org/profile/#me') });
showPerson(ruben);

async function showPerson(person) {
  for await (const name of person['foaf:friends*/foaf:givenName'])
    console.log(`- ${name}`);
}

[RubenVerborgh]

Hi @jeswr, good idea!

I will comment here on the syntactical suggestion, and in #80 on the implementation.

I think that, instead of

person['foaf:friends*/foaf:givenName']

we should consider

person['foaf:friends*'].['foaf:givenName']
person['foaf:friends*'].foaf$givenName
person['foaf:friends*'].name

which seems to fit the chaining spirit of LDflex better.

And perhaps also

person.friends['*'].name
person.friends.star.name
person.friends.recursive.name

or with better words than star/recursive that I can't seem to find now 🙂

[jeswr]

Hi @RubenVerborgh I agree that syntax closer to the chaining spirit of LDflex should be supported, however I am not quite sure how paths such as (ex:Mother|ex:Father)* could be expressed in that manner. The best I can come up with off the top of my head is

person.star(['ex:mother', 'ex:father'])

and even then paths with nested operators such as (ex:Mother?|ex:Father)+ couldn't be expressed. Do you have any thoughts on how this could be achieved?

I also still think it would be nice for the original syntax to be supported - a current use case I have for #80 is to extract an sh:path (SHACL path) in SPARQL/SHACLC syntax and then pass the path to LDflex.

[RubenVerborgh]

Right. Would person['foaf:friends*'].['foaf:givenName'] currently be supported though?

Note that introducing full SPARQL paths does introduce some demands on the query engine. It will no longer work with the rdflib.js backend, for instance.

[jeswr]

Yep #80 should already support that as 'foaf:friends*' would be processed using the code that I added to the JSONLD-resolver and 'foaf:givenName' would be processed by the existing predicate resolver. For all intents and purposes complex paths are just treated as a predicate by the rest of LDflex in #80

[RubenVerborgh]

Okay. I think it makes sense to continue with #80 then.