Date: 5-Oct-2023
<scribe> Scribe: DBooth
Eric: Claude and I have a demo at AMIA in a month.
... Will use shexjs interface. Will intro people to what FHIR RDF is, how to, building a browser that allows you to click through resoures, eg, patients, diagnoses. Does validation. Initially does SPARQL query, converts it to FHIR REST API. RE-executes the PAQRQL on what it gets. Ill pick a narrative to explain why people are interested.
Gaurav: Forget how we do this. Need to make minor changes, to get LOINC code in we had to se the FHIR prefix type as other, but now we need to change it to iri-stem. But in another change we use iristem (no hyphen). I made a proposal for that. Making a git branche for people to review and comment. Need to provide a sponsor, and sp
... https://jira.hl7.org/browse/UP-464
... Also linked from this ticket in our repo: https://github.com/w3c/hcls-fhir-rdf/issues/132
dbooth: Dash or no dash in iri-stem?
rob: Prob should be consistent with existing patterns.
Gaurav: ONe is in the code system identifier type, the other is without the dash (goal: short).
DBooth: What action need it now? Sched with ITS?
<scribe> ACTION: Gaurav to email co-chais of ITS.
DBooth: Report: https://tinyurl.com/fhirrdf-2023-sep
Tim: Able to fix some of the current issues by switching the turtle parser. Also made a PR to update the doc in FHIR RDF report, outlining where to find things.
... Some issues were aused by RDF generated by the previous parser, 19k lines long, old and inefficient, and incorrect. Updated kindling library to switch to new turtle parser that seems to be much more sustainable. Going forward, looking at other issues: How can I test this efficiently, without having to rebuild the entire FHIR spec? Found a few unit tests related to shex gen. Want to add more unit tests to these other issues I can fix, then add in our doc point to these tests.
... Tests were at one point part of the build, but now removed. Don't need to get that working agian, but need to find a place to do this and point developers to that.
Eric: Worked w Jim?
Tim: Yes.
Eric: Deepak's code fits into kindling, but he has ways to run it outside of the build. A call with Deepak and Jim would be most efficient way to make a standalone tool.
tim: Biggest problem is a huge publisher class. Posted a msg to zulip. Grahame said they need to keep using the old parser for old FHIR versions.
Eric: In building the playground, I put in feature switches instead of version switches. R5 config set those switches one way, and R4 set them the other way. Might be less work to do that than using two different parsers.
DBooth: But then he'd have to do a regression test to verify that the new parser gens the exact same R4 as the old code.
... This PR you did is in kindling and generates the examples?
tim: Yes. Some things are generated by the new parser and some were still generated by the old parser. Switching parsers fixed the problems.
... https://github.com/HL7/kindling/pull/148
... A couple more fixes are still needed in the parser
ADJOURNED