Currently the converter produces expanded JSON and nquads. We need a turtle output as well.

@vocab http://janeirodigital.github.io/nhs-care-plan/flat-FHIR.ttl# should be changed to something that is really visible in the RDF

The JSON LD 1.1 specification does not allow recursive references to contexts.

We need to detect recursion when generating the contexts and, if the referenced context is already in the containing document, emit a variation that doesn't have the reference. Options might include:

1. Emit variation with a different name (e.g. Extension.context.jsonld#v1) or
2. Generate the context inline when parts are omitted

All of the context bases should begin with at "https://raw.githubusercontent.com/fhircat/fhir_to_jsonld_context/master/jsonldc/contexts/r5/" instead of uu3.org

The context suffixes should be ".context.jsonld" -- there are no suffixes at the moment

The generated shex does not put the datatypes on the terminal nodes:

# Primitive Type dateTime
<dateTime> CLOSED {
    fhir:Element.id @<string>?;             # xml:id (or equivalent in JSON)
    fhir:Element.extension @<Extension>*;   # Additional content defined by 
                                            # implementations 
    fhir:value @<dateTime>?;                # Primitive value for dateTime
    fhir:index xsd:integer?                 # Relative position in a list
}

The fhir:value should be xsd:dateTime rather than recursively referencing the FHIR definition. The fhir.ttl generator gets it right, so it can be referenced to get the correct types.

The context file names should probably be all in lower case -- the current camelcase, while readable, causes problems w/ case insensitive servers. One can pass all tests using "foo.context.jsonld" on a test server and then fail on one that differentiates "foo" from "Foo"

Once the JSON-LD Playground gets fixed, context URI's should be changed to relative URI's. This will make Issue #1 moot

We don't need (or want) the FHIR version identifier in the file name. "account.context.jsonld" is good. Also, would recommend using entirely lower case. Mixed case can cause issues with some partial case sensitive servers.