FHIR Chat · slicing · conformance

Stream: conformance

Topic: slicing

Keith Boone (May 17 2016 at 06:16):

How does one slice a value[x] by data type? What discriminator would you use?

Keith Boone (May 17 2016 at 08:22):

I'm trying to slice something by type, but cannot seem to get it to work on observation. Anyone here have any words of wisdom. The error I'm getting is: Attempt to a slice an element that does not repeat: Observation.value[x]/null from http://hl7.org/fhir/StructureDefinition/CodeableConcept but my use of value[x] only includes Quantity | string.

Grahame Grieve (May 17 2016 at 09:57):

discriminator should be @type per the profiling page in the spec

Keith Boone (May 17 2016 at 13:45):

Yep, set that, still getting the above error.

Keith Boone (May 17 2016 at 13:45):

 [java] Caused by: org.hl7.fhir.dstu3.exceptions.DefinitionException: Attempt to a slice an element that does not repeat: Observation.value[x]/null from http://hl7.org/fhir/StructureDefinition/CodeableConcept
 [java]     at org.hl7.fhir.dstu3.utils.ProfileUtilities.processPaths(ProfileUtilities.java:402)
 [java]     at org.hl7.fhir.dstu3.utils.ProfileUtilities.generateSnapshot(ProfileUtilities.java:309)
 [java]     at org.hl7.fhir.definitions.generators.specification.ProfileGenerator.generate(ProfileGenerator.java:793)
 [java]     at org.hl7.fhir.definitions.generators.specification.ProfileGenerator.generate(ProfileGenerator.java:714)

Keith Boone (May 17 2016 at 13:46):

What is confusing is that CodeableConcept ISN'T one of the data types I've profiled Observation.value[x] to.

Grahame Grieve (May 17 2016 at 19:30):

I'll try and look it tomorrow

Keith Boone (May 17 2016 at 23:27):

Cannot slice by type when base cardinality is 0..1, although I can see where it might be useful to look at those as slices. I worked around it elsewhere.

Grahame Grieve (May 17 2016 at 23:42):

I've done it elsewhere - sliced by type

Grahame Grieve (May 17 2016 at 23:43):

have alook at the DAF profile on condition

Eric Prud'hommeaux (Nov 21 2017 at 09:02):

[apologies if this is the wrong stream; happy to re-post elsewhere]
In the BP slicing example http://build.fhir.org/profiling-examples.html#blood-pressure , could i then extend it with a PosturedBP a la:?

  <element> <!-- 8 -->
    <path value="Observation.component"/>
    <name value="diastolic"/> <!-- mandatory - gives the slice a name -->
    <min value="1"/>
    <max value="1"/>
  </element>
  <element> <!-- 9 -->
    <path value="Observation.component.code"/>
    <min value="1"/>
    <fixedCodeableConcept>
      <coding>
        <system value="http://snomed.info/sct" />
        <code value="9851009" />
        <display value="Body position" />
      </coding>
    </fixedCodeableConcept>
  </element>
  <element> <!-- 10 -->
    <path value="Observation.component.valueQuantity"/>
    <min value="1"/>
  </element>

Grahame Grieve (Nov 21 2017 at 09:09):

yes, because the slicing rules do not say that the slices are closed

Eric Prud'hommeaux (Nov 21 2017 at 09:12):

Also in the BP slicing example http://build.fhir.org/profiling-examples.html#blood-pressure , there are seven <element/> decls (using regex {} notation for cardinalities):
1. discriminator="code" {2,}
2. name="systolic" {1,1}
3. fixedCodeableConcept/coding/code="8480-6" {1,}
4. valueQuantity {1,}
5. name="diastolic" {1,1}
6. fixedCodeableConcept/coding/code="8462-4" {1,}
7. valueQuantity {1,}
(I assume 1. is in the FHIR Observation core resource and 2-7 are in a profile that extends it)
What associates e.g. 3 and 4 with 2? Is is just that they come in threes?

Eric Prud'hommeaux (Nov 21 2017 at 09:28):

is there a better stream for validation stuff?

Grahame Grieve (Nov 21 2017 at 09:32):

no this is the place- I just haven't had time to think about that one yet

Eric Prud'hommeaux (Nov 21 2017 at 09:33):

I'm examining this 'cause I'm working on the inheritance model for ShEx. I'm modeling it on XML Schema's separation of extension vs. restriction.

Eric Prud'hommeaux (Nov 21 2017 at 09:35):

1 seems like a base constraint. (2-4) and (5-7) seem like restrictions on that and my proposed PosturedBP seems like an extension of BP.

Grahame Grieve (Nov 21 2017 at 09:35):

we don't have that in FHIR - everything is additional.

Eric Prud'hommeaux (Nov 21 2017 at 09:37):

does that mean that (2-4) and (5-7) don't have to comply with (1), i.e. they needn't have a code if they didn't happen to mention one?

Eric Prud'hommeaux (Nov 21 2017 at 09:39):

or maybe discriminator= is sort of a special case that infects everything with the same path (but that smells a lot like restriction)

Grahame Grieve (Nov 21 2017 at 09:39):

discriminator is special. It's orthogonal to the constraint model - an extra layer of rules. What that says is that each of the slices must be distinguished by the fixed value(s) for code

Grahame Grieve (Nov 21 2017 at 09:40):

so yes, a special case, and yes, a restriction of sorts, but actually a meta-restriction, if you want to be precise - a constraint on the model, not the instances

Eric Prud'hommeaux (Nov 21 2017 at 09:43):

and that can be applied by a core resource or any in a chain of derivative profiles?

Grahame Grieve (Nov 21 2017 at 09:45):

it could, though the only discriminator we would ever apply in the base resource is on extension, which are always at least discriminated by url

Grahame Grieve (Nov 21 2017 at 09:45):

derivative profiles must also observe the rules of anything they derive from

Eric Prud'hommeaux (Nov 21 2017 at 09:49):

the prob i'm having with the "all are extensions" notion is that (1) seems like it's saying that anything derived must have two or more components and (2-4), (5-7) must be talking about those same components or we'd have have to create BP instances with a systolic, a diastolic and two more components.

Eric Prud'hommeaux (Nov 21 2017 at 09:57):

so it seems like the (8-10) i uttered first (PostureBP) are extending the package of (2-4),(5-7) while they in turn are constraining (1) (and probably (0) if we assume a fhir:Observation provides a minimalistic constraint on component {0,})

Grahame Grieve (Nov 21 2017 at 10:00):

you can add slices in derived profiles. Whether they extend 2-4 or 5-7 or are new depends on the slice names

Richard Townley-O'Neill (Nov 24 2017 at 01:34):

Is it that 3 and 4 are associated with 2 because of the values of pathidentify a hierarchy?
1 - <path value="Observation.component"/>
2 - <path value="Observation.component"/>
3 - <path value="Observation.component.code"/>
4 - <path value="Observation.component.valueQuantity"/>
5 - <path value="Observation.component"/>
6 - <path value="Observation.component.code"/>
7 - <path value="Observation.component.valueQuantity"/>

Ewout Kramer (Nov 28 2017 at 09:57):

Yes, indeed. The hierarchy get a bit hidden by the flat representation, but mentally I consider 2-4 children of 1 and 3-4 children of 2.

Last updated: Apr 12 2022 at 19:14 UTC