Stream: ontology
Topic: Hosting
Grahame Grieve (Jun 22 2017 at 23:11):
@Eric Prud'hommeaux @Harold Solbrig @David Booth can you confirm that you are happy with the server CORS configuration now?
Harold Solbrig (Jun 23 2017 at 14:40):
I'm assuming that you are talking http://test.fhir.org/r3. If so, yes - 'wget --header "Accept:text/turtle" http://test.fhir.org/r3/Patient/100' delivers exactly what was expected. Do you support PUT/POST operations as well?
How do we go about getting this working on the hl7 sites?
Grahame Grieve (Jun 23 2017 at 21:24):
test.fhir.org does everything in the spec
Grahame Grieve (Jun 23 2017 at 21:24):
the hl7.org servers - the CORS part should be right now
Eric Prud'hommeaux (Jun 23 2017 at 22:06):
looks great to me -- tested with http://rawgit.com/shexSpec/shex.js/master/doc/shex-simple.html (controls::load schema and load data)
Eric Prud'hommeaux (Jun 23 2017 at 22:07):
tx!
Harold Solbrig (Jun 23 2017 at 22:22):
HL7.org still appears to return xml instead of .ttl?
Grahame Grieve (Jun 26 2017 at 01:31):
that's still on my task list
Grahame Grieve (Jun 28 2017 at 03:35):
I think I've fixed up all the redirects on hl7.org/fhir - you should get turtle properly now.
Grahame Grieve (Jun 28 2017 at 03:35):
though it turns out that there's some files I never generated in turtle, and you can't get them whatever
Grahame Grieve (Jun 28 2017 at 03:36):
btw, I'm uploading 30000 redirectt files right now - I expect that will take 6 hours or so, so stick to lower down the alphabet if you're testing in the next few hours
Michael van der Zel (Jun 28 2017 at 06:34):
You beat me .. Working on vonk do turtle. Then we can do a connectathon :-)
Harold Solbrig (Jun 28 2017 at 17:00):
Would there be any objections if I put a revised w5.ttl and fhir.ttl generator -- the former fixes the issues I mentioned above and the latter addresses a serious flaw (should I file a bug report?) in fhir.ttl, where BackboneElements didn't have unique names (e.g. ItemComponent, which represents 4 different structures). No impact on the actual data, as these are all types that never appear in the actual records...
Grahame Grieve (Jun 28 2017 at 19:58):
fine to change w5.ttl
Grahame Grieve (Jun 28 2017 at 19:59):
definitely want a task for the fhir.ttl change - can you explain further?
Harold Solbrig (Jun 29 2017 at 17:30):
How do I create a task?
Two changes:
1) fhir:string, fhir:code, fhir:... (primitive types) are now subClassOf fhir:Primitive (where they didn't have any parent earlier)
2) Component naming -- BackboneElement naming changed -- As an example, ItemComponent appears in 3 places with three different definitions: Claim, ClaimResponse and ExplanationOfBenefit. Same for many other components. Names changed to ClaimItem, ClaimResponseItem, ExplanationOfBenefitItem, etc. No impact on fhir turtle instances.
(Proposed changes can be found at: https://github.com/BD2KOnFHIR/i2FHIRb2/tree/master/tests/data)
Lloyd McKenzie (Jun 29 2017 at 17:32):
To create a task, click on the "propose a change" link at the bottom of any page in the FHIR spec
Harold Solbrig (Jun 29 2017 at 17:58):
2 requests submitted.
Harold Solbrig (Jun 29 2017 at 17:59):
(hadn't realized 'task' and 'change request' were the same thing.
Lloyd McKenzie (Jun 29 2017 at 18:07):
Our change requests = Grahame's tasks ;)
Harold Solbrig (Jun 29 2017 at 18:15):
I've got the fixes ready to go for this one, assuming it is ok w/ folks
Grahame Grieve (Jun 30 2017 at 10:57):
I don't understand the backbone element naming thing - all the names are scoped by resource
Harold Solbrig (Jun 30 2017 at 15:02):
Look at the definition of ItemComponent in the stu3 fhir.ttl
and contrast with ExplanationOfBenefitItem, ClaimItem, ClaimResponseItem in https://github.com/BD2KOnFHIR/i2FHIRb2/blob/master/tests/data/fhir.ttl
Grahame Grieve (Jun 30 2017 at 20:07):
so it's just the reference in rdfs:domain that is not properly namespaced?
Michael van der Zel (Jul 05 2017 at 07:27):
@Grahame Grieve Why does http://test.fhir.org/r3/Patient?_format=rdf not work? Results in internal server error.
Grahame Grieve (Jul 05 2017 at 09:17):
full RDF support in the server is still on my todo list
Michael van der Zel (Jul 05 2017 at 10:50):
Your todo list must be long... How do you find the time? Do you have a time machine ;-)
Grahame Grieve (Jul 05 2017 at 11:39):
I wish I had a time machine...
Grahame Grieve (Jul 05 2017 at 11:39):
but if you west, you can get more than 24 hours in the day ;-)
Harold Solbrig (Jul 05 2017 at 16:06):
Well - both rdfs:domain and rdfs:range references. ItemComponent is (simultaneously) the definition of three different BNodes that occur within 3 different FHIR resources. The names themselves never appear in the instance RDF, which is why we didn't catch the error earlier. We're mapping FHIR to i2b2, however, and the names do show up as first class elements in that model. What I want to change is that the range of the "Claim.item" predicate is ClaimItem, the range of the ExplanationOfBenefit.item predicate is ExplanationOfBenefitItem, etc. rather than both of them being "ItemComponent"
Grahame Grieve (Jul 06 2017 at 06:59):
ok they should definitely be namespaced. that they aren't is a bug, but I could find the duplicate declarations
Grahame Grieve (Jul 06 2017 at 07:00):
but shouldn't it be Claim.ItemComponent, not ClaimItem? that's how all the other names are handled, no?
Harold Solbrig (Jul 06 2017 at 16:02):
It would be a viable alternative -- I took the shorter alternative because, to date, the only places it is used is in hierarchies and it ends up being a really (really) long URI once things start getting nested. I'll generate up that option as well and we can compare them in Protege and i2b2 to see what folks think.
(wrt. duplicate declarations -- they are separate in the source but get mashed together when loaded as RDF)
Harold Solbrig (Jul 06 2017 at 16:05):
Q: OK on the other fix - making fhir primitives (date, datetime, boolean, uri, etc.) children of fhir:Primitive? This was the intent of the original conversion and I just missed getting it in.
Grahame Grieve (Jul 06 2017 at 20:26):
yes ok with that
Harold Solbrig (Jul 10 2017 at 18:51):
Actually, there doesn't seem to be any serious issues besides the size of the URI (a longer one being "fhir:Claim.ItemComponent.DetailComponent.SubDetailComponent"). Take a look at https://raw.githubusercontent.com/BD2KOnFHIR/i2FHIRb2/master/tests/data/fhir_alt.ttl and, if you agree, I'll commit the changes.
Harold Solbrig (Jul 19 2017 at 21:24):
Actually, there doesn't seem to be any serious issues besides the size of the URI (a longer one being "fhir:Claim.ItemComponent.DetailComponent.SubDetailComponent"). Take a look at https://raw.githubusercontent.com/BD2KOnFHIR/i2FHIRb2/master/tests/data/fhir_alt.ttl and, if you agree, I'll commit the changes.
I think I overwrote fhir.ttl in the above example. Your proposed notation seems to work -- ok to commit?
Grahame Grieve (Jul 19 2017 at 23:13):
ok
Last updated: Apr 12 2022 at 19:14 UTC
