Assignment on primitive satisfies cardinality of value element #1284
Conversation
All elements with primitive types have a child element named "value". When this child element is required, an assignment to the primitive satisfies this cardinality requirement.
There was a problem hiding this comment.
This solves the false negative quite elegantly.
Unfortunately, it introduces a false positive. If a primitive's .value is 1..1, then it must have a value; setting just an extension on the primitive should not satisfy it. Prior to this PR, SUSHI would emit an error, but now it lets it pass with no error.
E.g., the following should cause an error if target.reference.value is 1..1, but it no longer does in this PR:
* target.reference.extension[http://hl7.org/fhir/StructureDefinition/data-absent-reason].valueCode = #unsupported
There is an existing problem regarding primitive arrays that is outside the scope of this change. Test cases are added and skipped that show these cases.
|
There are some existing problems with extensions on primitive arrays that are not fixed in this PR. Two test cases are added that give examples of these problems, but they are currently skipped. I've added CIMPL-1116 for this to be addressed in the future. |
There was a problem hiding this comment.
I think the code makes sense, but when I went to run a regression, the repo HL7/davinci-epdx#master has differences that I don't quite understand. Also, there are errors being logged saying Cannot read properties of undefined (reading 'length'), which shouldn't be happening. I haven't looked into why that's happening yet and if the differences and that unexpected error are related, but I can if that is helpful.
An element with a content reference may fail to unfold if there are problems in the element's StructureDefinition or the referenced element's StructureDefinition, or if the content reference does not resolve to an element. In these cases, assume the element is not a primitive for the purpose of validating an instance.
|
@jafeltra I've made a small update that should give a better error message for that repo. I've also created CIMPL-1120 to address the underlying problem more generally. |
Unfolding the full content reference doesn't provide any benefit, since the type is all that is needed to determine whether an element is a primitive or not.
|
@jafeltra - I'm in the middle of running a regression on this one -- so I'll report back. |
Fixes #1281 and completes task CIMPL-1109.
All elements with primitive types have a child element named "value". When this child element is required, an assignment to the primitive satisfies this cardinality requirement.
Because the primitive's "value" element will only be checked if the parent primitive element is not null, it is already known that a value has been assigned to the parent primitive when the "value" element's cardinality is being checked.