Provide a function to auto-generate a JSON-LD context from a Closed/DefinedNamespace

[nicholascar]

It would be entirely possible to generate a JSON-LD context object from an RDFLib DefinedNamespace or ClosedNamespace.

An example of an existing context and DefinedNamespace is available for GeoSPARQL:

• OGC's geo-context.json
• DefinedNamespace in RDFLib source code

[niklasl]

This ought to do it:

import json
from rdflib.term import URIRef
from rdflib.namespace import SKOS


def make_context(pfx, ns):
    terms = {pfx: str(ns)}
    for key, term in ns.__annotations__.items():
        if term == URIRef:
            terms[key] = f'{pfx}:{key}'

    return {'@context': terms}


print(json.dumps(make_context('skos', SKOS), indent=2))

Remember though that for the basic case of using one ontology for nice, compact JSON-LD, all you need is @vocab. Like:

{
  "@context": {
    "@vocab": "http://www.w3.org/2004/02/skos/core#"
  }
}

Of course, if you're mixing ontologies, you need something more involved. OTOH in those cases it makes sense to take care in crafting what you actually need (and knowing why you need it). A simple run of the above yields a useful starter for that of course.

Sometimes you have some custom JSON that's isomorphic enough to RDF that defining a context is enough to convert it (called the "zero edits" case; alas in practise a lot of data "in the wild" is too idiomatic in some way to make if fully feasible). In these cases it's not uncommon to need aliases used for either mapping diffferent short-names to terms (e.g. "part": "dc:hasPart"), and/or coerce stings to typed values, vocabulary terms or regular links and so on. (Or any of the more complex mapping techniques like indexed objects, nests and scoped contexts.)

[niklasl]

Aside: If you need a more involved tool, check out https://github.com/libris/definitions/blob/31ddca9fef60c82c6bb725aa7403f175b918dfd4/lxltools/contextmaker.py. It takes an RDF ontology and uses its RDFS and OWL mappings (sub-terms and equivalent terms) along with an ordered list of prefixes, and generates a mixed context of those. That was my first take of tackling the "Babel's Tower of Ontologies" problem, which I've subsequently turned into https://github.com/niklasl/ldtvm. To be continued.