Error Populating Slices on Observation.component

[markkramerus]

The CancerGeneticVariant profile needs to slice Observation.component. SUSHI seems to be OK with populating the slices in-line, but the IG Publisher crashes with this message:

Caused by: org.hl7.fhir.exceptions.DefinitionException: Attempt to a slice an element that does not repeat: Observation.component.code/Observation.component.code from http://hl7.org/fhir/us/core/StructureDefinition/us-core-observation-lab in http://hl7.org/fhir/us/mcode/StructureDefinition/CancerGeneticVariant

The offending lines are currently commented out so we can build the IG, but look for these lines in CancerGenomics.fsh:

* component[GeneStudied] ^short = "Gene studied [ID]"
* component[GeneStudied].code = LNC#48018-6
* component[GeneStudied].value[x] only CodeableConcept
* component[GeneStudied].valueCodeableConcept 1..1
* component[GeneStudied].valueCodeableConcept from http://www.genenames.org/geneId (extensible)
* component[VariationCode] ^short = "Discrete genetic variant"
* component[VariationCode].code = LNC#81252-9
* component[VariationCode].value[x] only CodeableConcept
etc.

[cmoesel]

I'm not sure if this is it, but this:

* component ^slicing.discriminator.type = #value

should probably be:

* component ^slicing.discriminator.type = #pattern

[cmoesel]

That doesn't fix it. This feels like a Publisher error to me. Aside from the discriminator type, I don't see anything wrong with this -- and I certainly don't see any cases of trying to slice Observation.component.code.

[markkramerus]

What is the next step on this? It's pretty serious if we can't slice Observation.component. It is extremely common thing to do. If it were the IG Publisher, it wouldn't have gone unnoticed.

[cmoesel]

Ideally we'd post a message to Zulip and ask if anyone has any idea what's going on. That said, it's harder without being able to point to IG source that they understand (not FSH). We should consider pushing up fsh-mcode IGs on a branch of the HL7 mCODE repo.

While I understand your point about it going unnoticed, the publisher is changing all the time -- and sometimes with unintended consequences. If this is a bug that was introduced in last night's build, it might go unnoticed.

Either way, I just don't see anything wrong with it, so I could use some extra eyes on it.

[cmoesel]

We figured this out due to some good (tedious) debugging from @markkramerus. Here is a fuller picture of the FSH in question:

* component contains
    GeneStudied 0..* MS and
    VariationCode 0..* MS and
    GenomicDNAChange 0..1 MS and
    GenomicSourceClass 0..1 MS and
    AminoAcidChange 0..1 MS and
    AminoAcidChangeType 0..1 MS and
    CytogeneticLocation 0..* MS and
    CytogeneticNomenclature 0..1 MS
* component[GeneStudied] ^short = "Gene studied [ID]"
* component[GeneStudied].code = LNC#48018-6
* component[GeneStudied].value[x] only CodeableConcept
* component[GeneStudied].valueCodeableConcept 1..1
* component[GeneStudied].valueCodeableConcept from http://www.genenames.org/geneId (extensible)
* component[VariationCode] ^short = "Discrete genetic variant"
* component[VariationCode].code = LNC#81252-9
* component[VariationCode].value[x] only CodeableConcept
* component[VariationCode].valueCodeableConcept 1..1
* component[VariationCode].valueCodeableConcept from http://www.ncbi.nlm.nih.gov/clinvar (preferred)
* component[GenomicDNAChange] ^short = "Genomic DNA change (gHGVS)"
* component[GenomicDNAChange].code = LNC#81290-9
* component[GenomicDNAChange].value[x] only CodeableConcept
* component[GenomicDNAChange].valueCodeableConcept 1..1
* component[GenomicDNAChange].valueCodeableConcept from http://varnomen.hgvs.org (required)
* component[GenomicSourceClass] ^short = "Genomic source class [Type]"
* component[GenomicSourceClass].code = LNC#48002-0
* component[GenomicSourceClass].value[x] only CodeableConcept
* component[GenomicSourceClass].valueCodeableConcept 1..1
* component[GenomicSourceClass].valueCodeableConcept from http://fhir.loinc.org/vs/LL378-1 (required)
* component[AminoAcidChange] ^short = "Amino acid change (pHGVS)"
* component[AminoAcidChange].code = LNC#48005-3
* component[AminoAcidChange].value[x] only CodeableConcept
* component[AminoAcidChange].valueCodeableConcept 1..1
* component[AminoAcidChange].valueCodeableConcept from http://varnomen.hgvs.org (required)
* component[AminoAcidChangeType] ^short = "Amino acid change [Type]"
* component[AminoAcidChangeType].code = LNC#48006-1
* component[AminoAcidChangeType].value[x] only CodeableConcept
* component[AminoAcidChangeType].valueCodeableConcept 1..1
* component[AminoAcidChangeType].valueCodeableConcept from http://fhir.loinc.org/vs/LL380-7 (required)
* component[CytogeneticLocation] ^short = "Cytogenetic (chromosome) location"
* component[CytogeneticLocation].code = LNC#48001-2
* component[CytogeneticLocation].value[x] 1..1 // CG Reporting IG does not constrain the CytogeneticLocation value type.
* component[CytogeneticNomenclature] ^short = "Variant ISCN"
* component[CytogeneticNomenclature].code = LNC#81291-7
* component[CytogeneticNomenclature].value[x] 1..1 // CG Reporting IG does not constrain the CytogeneticLocation value type.

The culprits are the AminoAcidChange and AminoAcidChangeType slices. If you comment out either one, everything works fine. When they're both present, the IG Publisher crashes. Looking at the generated StructureDefinition, the elements are being generated out of order. The problem is when you get to the AminoAcidChange element. You see something like this order:

Observation.component:AminoAcidChange
// NOTE: the Observation.component:AminoAcidChange.* children SHOULD be here but aren't
Observation.component:AminoAcidChangeType
Observation.component:AminoAcidChangeType.id
// etc, Observation.component:AminoAcidChangeType.*
// NOTE: now we find the Observation.component:AminoAcidChange.* children, in the wrong place
Observation.component:AminoAcidChange.id
// etc, Observation.component:AminoAcidChange.*

when it should be:

Observation.component:AminoAcidChange
Observation.component:AminoAcidChange.id
// etc, Observation.component:AminoAcidChange.*
Observation.component:AminoAcidChangeType
Observation.component:AminoAcidChangeType.id
// etc, Observation.component:AminoAcidChangeType.*

It appears that this happens because AminoAcidChange is a substring of AminoAcidChangeType. So whatever function is trying to find where to unfold those AminoAcidChange elements must be using a substring function that is resulting in unfolding into the wrong place.

[cmoesel]

Re-opening until it is released

[cmoesel]

Fixed in 0.6.2