Implicit Use of FHIRHelpers Functions

I’m noticing there seems to be a change in how the CQL-to-ELM project handles use of FHIRHelpers functions between versions 1.2.x and 1.3.x for FHIR 3.0.0.

The FHIR enumerated fields have their own types, but underneath are often represented as simple Strings nested in a “value”. In CQL-to-ELM for 1.2.x, a FHIRHelper function was inferred and would then wrap any property call to such a field. However, the CQL-to-ELM compiler for version 1.3.x no longer seems to apply this implicit FHIRHelper function call. Taking fhir:AdministrativeGender as an example:

CQL

library Query version '1'
using FHIR version '3.0.0'

include FHIRHelpers version '3.0.0' called FHIRHelpers

context Patient

define GetGender: Patient.gender = 'female'

CQL-to-ELM v1.2.16 yields

      <def resultTypeName="t:Boolean" name="GetGender" context="Patient" accessLevel="Public">
         <expression resultTypeName="t:Boolean" xsi:type="Equal">
            <operand name="ToString" libraryName="FHIRHelpers" xsi:type="FunctionRef">
               <operand resultTypeName="fhir:AdministrativeGender" path="gender" xsi:type="Property">
                  <source resultTypeName="fhir:Patient" name="Patient" xsi:type="ExpressionRef"/>
               </operand>
            </operand>
            <operand resultTypeName="t:String" valueType="t:String" value="female" xsi:type="Literal"/>
         </expression>
      </def>

but CQL-to-ELM v1.3.10 yields

      <def resultTypeName="t:Boolean" name="GetGender" context="Patient" accessLevel="Public">
         <expression resultTypeName="t:Boolean" xsi:type="Equal">
            <operand resultTypeName="fhir:AdministrativeGender" path="gender" xsi:type="Property">
               <source resultTypeName="fhir:Patient" name="Patient" xsi:type="ExpressionRef"/>
            </operand>
            <operand resultTypeName="t:String" valueType="t:String" value="female" xsi:type="Literal"/>
         </expression>
      </def>

where the difference is the lack of a wrapping function call to FHIRHelpers.ToString.

I wondered if the definition of Equal was meant to capture the type-unwrapping here, but the definition doesn’t seem to say anything about that.

My questions here are

1. what change eliminated the implicit wrapping of certain FHIRHelpers functions over these special FHIR types?
2. what was the thinking behind the change?

I ask because the definition of functions like ToString when dispatched over an enumerated value type like AdministrativeGender “unwrapped” the value nesting in the FHIR structure. Without the implicit call as realized in the ELM, I’m not sure it’s clear anymore how a runtime implementation will have to unwrap the FHIR structure.

Also: Given that ReferralRequest.status is also an example of an enumerated field like Patient.AdministrativeGender, I think this question is related to #371.

[brynrhodes]

@bdrillard, just pushed a commit that adds the expected behavior for AgeInYearsAt(Patient.deceased). It will now implicitly convert the Patient.deceased to a DateTime and then use the DateTime overload of CalculateAgeInYearsAt.

[brynrhodes]

Thank you for the report, I’m looking into this, it certainly seems like the wrong behavior.