Logic for ValueSet/$validate-code parameter evaluation is wrong

[alexzautke]

Describe the bug
The behaviour of validating the input parameters for a call against $validate-code changed in #2968. If the "code" parameter is now used, the SDK requires also a "context" or "url" parameter. However, it fails to check if the "system" parameter is already provided so that it doesn't need to be inferred.

To Reproduce
Steps to reproduce the behavior:
Call ValueSet/$validate-code with "system", "code" and "valueSet" parameters.

Expected behavior
"CheckForValidityOfValidateCodeParams" in src/Hl7.Fhir.Base/Model/ParametersExtensions.cs should allow the combination of "code" and "system".

Version used:

• FHIR Version: R4
• Version: 5.11.0

[mmsmits]

Please read https://chat.fhir.org/#narrow/channel/179202-terminology/topic/.E2.9C.94.20Required.20.24validate-code.20parameters/near/482250225.

Where Grahame states we need to throw an error when neither a URL, or a Context is provided.

[alexzautke]

@mmsmits My interpretation would be that this is only true if system is not provided. There is not point in trying to figure out the system if the caller already provided the system as a parameter.

[alexzautke]

From Michael Lawley: "I think I get what's going on here. Normally you provide system and code so that the validation knows what system the code is from. But if you're validating a code-typed element then there is no recorded (code) system."

The second sentence is only true for validation. In FS we have test cases where the system and code are provided. We could make a valueCoding out of it, but that seems unnecessary.

[mmsmits]

What I think we should check is the presence of the valueSet parameter.
This error is about not being able to find out against which valueSet to check. That's why you should provide either e URL or a context (but we forgot the valueSet param).

[alexzautke]

That's where inferSystem would come into play. However if you already pass in the system as parameter (+ valueSet), then it should also work.

[mmsmits]

Shouldn't there be two rules here?

1. Either system, inferSystem = true, only if you provide a code as value (not a coding or codeableConcept (to determine the system)
2. At least, url, context or valueSet (to determine the valueset)

This check covers (part of) the second rule, and should be extended with valueSet.
We might need code to check the first if we want.