Allow cardinality constraints to be applied to connected elements #1120
Conversation
When constraining cardinality on an element, if there are connected elements that are the child of a slice, it may be possible to apply the cardinality constraint to those elements. As long as no cardinality is widened, applying the constraints is allowed. This may mean that only the new min or max are applied. If the new cardinality has no overlap with the connected element's existing cardinality, the cardinality constraint can't be applied.
mint-thompson
requested review from
jafeltra,
cmoesel,
julianxcarter and
guhanthuran
as code owners
There was a problem hiding this comment.
I think the general concept and approach is good, as it's friendlier than the spurious error we used to give and it's helpful in updating those cardinalities. When I ran regression, however, some of the changes in one repo didn't make sense to me.
You can reproduce by running regression on HL7/davinci-pas#master. You'll see a bunch of changes in StructureDefinition-profile-claim-update.json -- all of which constrain Claim.supportingInfo:SomeSliceName.extension and Claim.supportingInfo:SomeSliceName.modifierExtension to each have max 1. Here is one example:
If you look at the FSH definition for that profile, however, you'll see that the only rules on supportingInfo are these:
* supportingInfo.extension contains InfoChanged named infoChanged 0..1
* supportingInfo.extension[infoChanged] ^short = "A code indicating how the piece of information has changed."
* supportingInfo.modifierExtension contains InfoCancelledFlag named infoCancelledFlag 0..1
* supportingInfo.modifierExtension[infoCancelledFlag] ^short = "Indicates that this piece of information is not to be used."
It seems like adding an extension with max 1 (for that specific extension only) is going through and updating the max of the non-specific extension element in every slice. Which is not what it should be doing.
A connected element may end with a slice if the original element ends with a slice with the same name. When applying the "must support" flag on an element whose path ends with a slice, the flag should be applied to connected elements only if the connected element ends with a slice with the same name.
|
Oh I also ran a full regression again, and there were no differences besides that one sliced element that has the |
There was a problem hiding this comment.
This looks good and works well! Thanks, @mint-thompson. I'm approving this, but you probably should accept Julia's minor suggestion before we merge it.
Co-authored-by: Julia Afeltra <[email protected]>
Completes task CIMPL-986.
When constraining cardinality on an element, if there are connected elements that are the child of a slice, it may be possible to apply the cardinality constraint to those elements. As long as no cardinality is widened, applying the constraints is allowed. This may mean that only the new min or max are applied. If the new cardinality has no overlap with the connected element's existing cardinality, the cardinality constraint can't be applied.