ShEx generation

harold: I submitted a reasonably complete shex generated. Ball is now in Grahame's court.
... Made some changes to references, where there is a choice of person or organization.
... Since the type is a URI, how do you record that fact in shex?
... We made a proposal on that.
... Shex gives you a choice of a simple URI or a shape.

<hsolbrig_> <PatientReference> { a [fhir:PatientReference]?, fhir:uri.id @<id>?, fhir:uri.extension @<Extension>*, fhir:uri.value (xsd:anyURI OR @<Patient>) }

<hsolbrig_> fhir:Observation.subject ( @<PatientReference> OR @<GroupReference> OR @<DeviceReference> OR @<LocationReference> )?,

<hsolbrig_> <uri> { fhir:uri.id @<id>?, fhir:uri.extension @<Extension>*, fhir:uri.value xsd:string? }

<hsolbrig_> <PatientReference> { a [fhir:PatientReference]?, fhir:uri.id @<id>?, fhir:uri.extension @<Extension>*, fhir:uri.value (xsd:anyURI OR @<Patient>)? }

<hsolbrig_> https://hl7-fhir.github.io/observation-example.ttl.html

The snippet:

[[

fhir: Observation.subjectReference [
... reference <http://hl7.org/fhir/Patient/example>;
... Reference.reference [ fhir:value "Patient/example" ]

];

]]

eric: Should we use URIs for RDF linkage or a string? A string would require an additional step to connect it.
... I favor relative URIs.

tony: We discussed having both the string and a URI.
... And the URI may be the same as the string.
... We discussed using reference.link for the URI.

eric: XML schema for FHIR XML treats a patient ref and a provider ref identically -- no validation beyond the fact that it is a reference.

<hsolbrig_> [[

<hsolbrig_> <subject> <reference value="Patient/example"/> </subject> <encounter> <reference value="Encounter/example"/> </encounter> <subject> <reference value="Patient/example"/> </subject> <encounter> <reference value="Encounter/example"/> </encounter>

<hsolbrig_> ]]

<hsolbrig_> (well that didn't work, but...)

eric: We can be more specific in RDF about the type. We can do much better validation than that.

<hsolbrig_> @dbooth -- is the formated paste key "The snippet:"?

eric: But we may have to do that modally: one shex that enforces that; one that doesn't. Validating a FHIR resource in isolation would use the looser one; validating a collection of resources together could use the stricter one.

harold: One thing disconcerting: the URI shape and patient ref shape do not line up well with what grahame was publishing as the emitted RDF.
... Observation.subject has been turned into observation.subject.reference
... Grahame's RDF says fhir:Observation.subjectReference

https://hl7-fhir.github.io/observation-example.ttl.html near the bottom

lloyd: Guessing it's a bug in his code

<hsolbrig_> <reference value="Patient/example"/>

harold: Could that equally well say <reference value="foo/example"/>

lloyd: Three styles of URL. 1. It could be a full URL. 2. Local -- same server; 3. Contained reference, which would not have "Patient/"
... The first two require "Patient/" in them.

dbooth: Next-to-last path segment in the URI?

lloyd: Yes, or fourth-to-last if you are doing version-specific references.

eric: Then you need to know that the versions are not using those strings.
... Then you could parse from the right to figure it out.
... We could make them all look liked contained if when converting to FHIR RDF, you have a ref to Patient/example, then you add a triple giving the type.

dbooth: Should we have another stricter shex to check the type of the referent?
... If we make one, would it be an alternate download? As a separate tab?
... Would the type arc be required in RDF?

lloyd: You can easily write a regex. Do we really want to?

dbooth: Inclined to leave out the type arc from the reference.
... Seems like it might require messy parsing to include the type arc in every reference.
... Also a convenient delineation to have the type arc only when you have the actual resource object.

tony: Validate after you pull in that other resource.

ISSUE: Should a type arc be required for the target of a reference?

<trackbot> Created ISSUE-21 - Should a type arc be required for the target of a reference?. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/21/edit>.

ISSUE: Should a type arc be allowed for the target of a reference?

<trackbot> Created ISSUE-22 - Should a type arc be allowed for the target of a reference?. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/22/edit>.

<scribe> ACTION: harold to make a dir on github for shex [recorded in http://www.w3.org/2016/04/19-hcls-minutes.html#action01]

<trackbot> Created ACTION-53 - Make a dir on github for shex [on Harold Solbrig - due 2016-04-26].

5pm call

tony and harold cannot make it; eric is at risk.

dbooth: I'll cancel the 5pm call.

ADJOURNED

This is scribe.perl Revision: 1.144  of Date: 2015/11/17 08:39:34  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Present: David_Booth Lloyd_McKenzie EricP Harold_Solbrig Sajjad_Hussain Tony_Mallia
Found Date: 19 Apr 2016
Guessing minutes URL: http://www.w3.org/2016/04/19-hcls-minutes.html
People with action items: harold