– DRAFT –
FHIR RDF

Introductions

Ben_McAlister: Work at Cerner, product strategist, interop for EU regiion. Also chairs H7 UK. Background: Studied medicine, med informatics, Alan Rector, UC London.

gaurav: Rennaisance computing institutes at UNC, background in RDF and OWL. Working w James and others on FHIRCat project.

James: UNC.

jim: Also UNC, background bio, OWL, bio ont, FHIRCat.

Issue 76: Ordered lists

https://github.com/w3c/hcls-fhir-rdf/issues/76

jim: Looked into using JSON-LD framing. Frame is separate from @context? But I haven't found a MUST anywhere.

jim: I'll try to pin it down more.

jim: Another issue related to lists and OWL. We want RDF to valid OWL, but worried about the list values for properties. Not sure that plays well with OWL tools.

ACTION: jim to show a use case in which a list value is a problem for OWL (or is able to work)

eric: If you have n codings in a CodeableConcept.
… You can't says AllValuesFrom because it's a list.
… But you can say it's a first of something and a rest that is a union of the recursive type or nil.

jim: I'll try making rdf:first be an object property.

eric: We want it in OWL DL

FHIR JSON-LD Playground

<ericP> https://ericprud.github.io/fhircat-playground/playground/

eric: Trying to run everything from shex. FHIR and datatype profiles are a pain in the ass.
… If a med req has a dose, then it has an interval, etc., making parsing miserable.
… Making a test suite to capture all of the expressiviity of FHIR that we want in shex.
… Also using it to drive graph normalization for testing.

An example: https://tinyurl.com/y4l34yqv

eric: The resulting turtle (but in the N-Quads tab) is actually generated by shex, to keep a nice ordering of properties.
… Also in the top right of that playground there are buttons for features that are not yet implemented, but will be.

<ericP> https://www.ldf.fi/service/rdf-grapher

eric: That's the graph visualizer I used to validate the generated turtle.

david: Next steps?

eric: Make sure the shex machine is doing the right thing. That will give James code that he can use. Then I'll configure it to use the option buttons.

james: Eric and I reviewed each other's code. Eric developed a test suite. My code is doing the same thing, but from command line.

eric: One difference: you're using @contexts from a repo.

james: Yes, and I'll use your code when you're done.

eric: There's a lib that gives you a visitor on a FHIR resource defn. You pass it a config (which it currently ignores, but eventually won't), and it visits all those things and gives you all the expected RDF predicates and datatypes.
… Then that spits out shex, and that gets used for lots of things -- nested turtlifier, on-the-fly-context generator, etc.
… I'm adding those to a lib that could be published with npm.

gaurav: i love test suites! Good place to figure out where to divvy up work: Write a failing test and then assign it to someone to implement.

eric: James is blocked on my lib work right now. Guessing to have it done in a week. Suggest that james change the two panes to three in the playground.

ACTION: James look into changing the playground form two panes to three, and check w Emily.

Issue 94: lack of concept URIs

https://github.com/w3c/hcls-fhir-rdf/issues/94

gaurav: Last week proposed modifying NamingSystem. CodingSystem has the same identifier field.

david: process would be to add a field to the registry?

gaurav: I think the FHIR spec needs to be modified first, to add that prefix.

david: Where would that change be, in the FHIR spec?

In the FHIR spec: https://build.fhir.org/valueset-identifier-use.html

gaurav: Should add another Code in the Expansion sectoin of that page, for "prefix": https://build.fhir.org/valueset-identifier-use.html#expansion

ben: That looks like an extensible set of codes anyway, so we could add one.

gaurav: Actually, 'type' might work instead.

https://build.fhir.org/valueset-identifier-type.html

david: Not sure that's the right kind of thing. Confused.

ADJOURNED