Stream: conformance
Topic: Validation error for slicing with Exists in derived profile
Morten Ernebjerg (Dec 13 2018 at 10:37):
I am seeing a curious validation error with the latest Java JAR validator when using slicing with an Exists-discriminator in a derived profile:
Here is what I did (working in Forge):
1. Create a new derived profile from the following profile (German base profile for Patient): http://fhir.de/StructureDefinition/patient-de-basis/0.2 . This Profile does not itself slice name field, but it does specify a custom profile for it (viz. http://fhir.de/StructureDefinition/humanname-de-basis/0.2)
2. Slice the namefield, with a discriminator of type "Exists" on period.end
3. Add a slice and in this slice, set the cardinality of period.endto 0..0
4. Validate using the latest Java JAR validator against a Patient example that contains the namefield with only name.given filled in (at least the namefield has to be there)
This gives the following error (despite step 3):
*FAILURE* validating examples\patient-example-minimal.xml: error:1 warn:0 info:0 Error @ Patient.name (line 6, col11) : Discriminator (period.end) is based on element existence, but slice Patient.name:sliceName neither sets min>=1 or max=0
Curiously, if I do exactly the same, but base my profile on the standard STU3 Patient resource, it works. Doe slicing in derived field somehow work differently?
Ewout Kramer (Dec 18 2018 at 10:13):
I sort of understand the error - does it make sense to have a discriminator on an element that has set it's cardinality to 0..0. It seems the error should say: "but slice ........ EITHER sets min>=1 or max=0" - so maybe I just don't understand what's going on.
That said, the german profile on humanname does not make any changes to the period element, so there should not be changes in behaviour. @Grahame Grieve @Michel Rutten ?
Michel Rutten (Dec 18 2018 at 10:29):
Forge helps users to design valid StructureDefinitions, but ultimately verification is left up to the validator. I don't fully understand the warning message from the Java validator...
Morten Ernebjerg (Dec 18 2018 at 11:35):
In this case, I am not restricting the cardinality of period.endin general, only in a particular slice where I removed it (0..0). This was based on my understanding that to use the Exists-discriminator, one needs to fix the "existence status" of the corresponding element to either "never" (cardinality = 0) or "always" (cardinality > 1) in slices. But if I read the error message right, it seems to tell me that I failed to do exactly that, which is what confused me.
Brian Postlethwaite (Dec 18 2018 at 11:35):
That's constraining to 0..0 could be to validate that this slice isn't to be used.
Morten Ernebjerg (Dec 18 2018 at 11:40):
(BTW, the usecase background is that I want to define a slice that picks out a name entry for which(among other properties) no end-date has been specified so that this name is guaranteed to still be valid)
Lloyd McKenzie (Dec 18 2018 at 14:59):
If you're using "exists", you expect one slice to be 1..x and one to be 0..0.
Last updated: Apr 12 2022 at 19:14 UTC
