deepak: We gen shex schemas for resources and profiles. Integrated with HAPI FHIR code. Validated those schemas (to ensure they parse properly), then applied them to all FHIR RDF examples.
... Ran 2371 examples against shex. Good results. Just submitted a paper about it.
... 1504 failed -- validation timed out.
... 612 examples passed partially but violated constraints.
... 255 fully passed.
eric: I created issues for 4 of the problems discovered. https://github.com/w3c/hcls-fhir-rdf/issues
jim: Are these content generation problems, in generating the examles?
eric: Yes.
... Should wait on changing the shex until the generation is fixed.
... Eg, https://github.com/w3c/hcls-fhir-rdf/issues/116
... See issues 114 - 117
deepak: Xhtml - no schema generated
This is special purpose datatype to denote xhtml content in which FHIR published narratives. No body is defined, hence it is empty.
https: //build.fhir.org/narrative.html#xhtml
eric: That's issue 118
... I think it should have a datatype of rdf:XMLLiteral datatype
https: //github.com/w3c/hcls-fhir-rdf/issues/114
eric: Some duplicate turtle is generated
<scribe> ACTION: Jim to fix 114
https: //github.com/w3c/hcls-fhir-rdf/issues/115
eric: A bit of the div got left out.
<scribe> ACTION: Jim to fix 115
https: //github.com/w3c/hcls-fhir-rdf/issues/116
eric: There were a few places where something was a string that should have been boolean
<scribe> ACTION: jim to fix 116
https: //github.com/w3c/hcls-fhir-rdf/issues/117
<scribe> ACTION: Jim to fix 117
https: //github.com/w3c/hcls-fhir-rdf/issues/118
<scribe> ACTION: Deepak to fix the shex for 118
https: //github.com/w3c/hcls-fhir-rdf/issues/119
eric: Will be fixed w 118 and may impact what Jim generates.
... Both Deepak and Jim need to address 119
jim: Working on a grant proposal. Will work on this after that.
deepak: You can see the counts of examples, and click to see them validated in the shex playground: https://github.com/fhircat/ShExValidation
... The repo has lots of info about the validation that was done.
https: //github.com/w3c/hcls-fhir-rdf/issues/70
eric: This started as a shex issue, but then we notice a difference in how Reference and canonical are treated.
... a Reference has a place for a fhir:link (in the RDF), to provide an RDF node in addition to the string of the URI that's a Reference.
... But we didn't add the same thing to canonical.
... There's an asymmetry. canonical is a primitive type, so there's no place to add it unless you add it to the blank node that's for extensions.
... Maybe a logical thing to do: 1. Leave them asymmetrical. 2. Make the REference no longer have the fhir:link on its root, but move it to the blank node.
dbooth: So far, that blank node has only been used for extensions.
<scribe> ACTION: Eric to re-write issue 70
ADJOURNED
Date: 15 Jun 2023