Meeting minutes
Concept URIs
Issue 94
gaurav: Playing with the HL7 UTG repo. There are XML files that seem to be the source of truth.
… I added another identifier, with use "prefix", but that caused a validation error. I tried type "prefix", but and it accepts it but does not pass it downstream.
… It gets converted into HTML docs, which don't contain the type field, but the JSON has it.
… The type field is supposed to be extensible, which is why we thought that would be a good field for it. If we can convince the FHIR group to accept this, then we can figure out how to do it.
… I also confirmed that if I modify the XML file, it shows up in the generated XML and JSON on the website and the downloads of the master package.
… Next step is to incorporate this into FHIR examples, and for all the SNOMED concepts attach the prefixes.
https://
gaurav: NamingSystems and CodingSystems are weirdly inconsistent.
david: Should file an issue on that.
gaurav: I'll file an issue in the UTG repo.
david: What else would be needed to make a full proposal for this change?
gaurav: Writing a google doc description.
… Grahame suggested we bring this to the FHIR chat.
eric: We're trying to figure out how to add URLs to terminologies that don't have their own URLs, and replicate those that do. Is "prefix" sufficient for both cases?
… And of the existing URLs, is a simple prefix sufficient? Or might we need to do something different for two different code systems?
david: Such as a URL template?
eric: Yes, and maybe some codes munging needed first.
gaurav: 1. Is it still worth trying to get just a prefix in? 2. Is there an egregious thing that would need a special case? There might already be one.
gaurav: Found this in a SNOMED code: 128045006:{363698007=56459004}
… For anything in the FHIR examples, we can make sure that every SNOMED code converts to an RDF URL.
eric: We can solve this easily by URL encoding.
david: Also need some canonicalization, eg, for spaces.
eric: Do we care about compositional syntax? If someone wrote it in that syntax, they're using it to say this is a fracture of the right clavicle.
… But you don't need to write the SNOMED code that way for this example, because you can express it in FHIR anyway.
… But other cases might not.
… But we could use URL encoding.
eric: I suspect that most uses of compositional codes are expressible in FHIR in other ways.
gaurav: I'll focus on FHIR examples. Other example FHIR data, to find codes that cannot just be prefixed?
ACTION: Eric to ask folks at NHS about use of compositional codes other than SNOMED
gaurav: I have 8 candidates that we can just prefix, including SNOMED and LOINC.
eric: We should do this as early as possible, becuase it impacts tooling.
david: I'm comfortable with only the prefix solution.
eric: Could imagine something hierarchical that contains slashes and you don't want those slashes URL encoded.
david: Should do a survey of code systems.
ACTION: Gaurav to get the list of terminologies and mark which ones are covered by prefix solution.
Preprocessing
eric: Preprocessing step did not cover the conversion of union types to specific types.
… E.g., FHIR datetype is a union of four different XSD types. But we want them specifically typed in the RDF, so the preprocessing needs to do that.
… To fix that I had to microparse to match one of the four types. But to parse it, I had to know that it is a FHIR datetype (not an int, for example).
… So I had to pass the content model to the preprocessing.
david: Should be able to know from the FHIR property name, whether it needs special union preprocessing.
ADJOURNED