JSON Pointer reference parsing fails when pointing the root document

[Viicos]

While uncommon, I believe it is possible to have an empty JSON Pointer (i.e. ""), referencing the root of the document (see examples in the RFC).

Here is a regression test:

resource = Resource.opaque(contents={"foo": "baz"})
resolver = Registry().resolver()
assert resource.pointer("", resolver=resolver).contents == {"foo": "baz"}

Current behavior:

referencing.exceptions.PointerToNowhere: '' does not exist within {'foo': 'baz'}

[Julian]

Indeed, thanks for the report. This bug is somewhat hidden from "real life" I think given that Resolver.lookup("#", ...) already handles this when you use a pointer referred to by a fragment, but indeed you're correct that .pointer() needs to as well given it's public API. I've pushed a fix.

Out of curiosity given you're using that API, I assume you're writing a library or some thing using referencing and not simply using it alongside a JSON Schema implementation? Can you share anything about what you're working on just out of curiosity.

Thanks again.

[Viicos]

Thanks for the quick fix! In this case I wasn't directly using referencing, but got inspired by your JSON Pointer parsing implementation (will add credit in my codebase soon). I'm implementing the JsonLogic rule format in Python (repo can be found here), and it is possible to access data when evaluating a rule/expression:

evaluate({">": [{"var": "/path_to/my_int"}, 2]}, data={"path_to": {"my_int": 1}})
#> True

And I support the JSON Pointer spec to specify the location of a variable.

---

While evaluation of such rules are not related to JSON Schema at all, I provide a typechecking mechanism, based on the JSON Schema format. I currently support a limited subset of the spec (i.e. no anyOf, etc) and I'm doing this "by hand", but this definitely relates to our previous discussion: #124

(might be of interest: an implementation of the json schema types).

I'm currently exploring more of this with this project, and maybe in the future this could help implementing this "structured schema walker" :)

[Julian]

Got it, good info, and yes that sounds great!