modularize extraction queries

[VladimirAlexiev]

A brief inspection of https://github.com/google/schemarama/tree/main/kgx/wikidata shows that those queries have mostly the same structure:

• find some entities
• get some of their properties; assuming they all are multivalued
• (no labels are obtained, even though wikibase:label is invoked)
• construct triples: either using the original WD props, or using bioschema props.

So the distinct part of each query selects props, and optionally maps them to bioschema. I think it makes sense to extract these specific parts and then generate SPARQL from them. This will help to:

• ensure consistency of mapping to bioshema across queries
• allow to handle issues like #27
• allow modifications such as INSERT to a local repo using Federated query to Wikidata, instead of genreating triples

BTW, have you considered generating extraction queries from WD SHEX like https://www.wikidata.org/wiki/EntitySchema:E258 ?

[danbri]

Hi! Yes - these queries were part of some collaboration around Wikidata's Subsetting-oriented community. Others were indeed looking at ShEx to SPARQL generators. I agree that SPARQL ultimately should be closer to being a compile target than the "source of truth" format, for these usecases. I'll close your issue as we're not actively working on this at the moment and when it comes back I'll make a dedicated repository, but you make good points!