FHIR Chat · fixed/pattern on min = 0

Stream: conformance

Topic: fixed/pattern on min = 0

Grahame Grieve (Jul 21 2020 at 23:05):

So an issue arises here: https://chat.fhir.org/#narrow/stream/179166-implementers/topic/Slicing.20on.20pattern.2Ffixed :

What does a profile mean when it assigns a fixed value or a pattern but leaves min = 0?

As far as I can tell , there's 3 options:

if min=0, it doesn't matter what the fixed/pattern is
if there's a fixed or a pattern, it doesn't matter what the min is
if there's a fixed/pattern, and min=0, that's an error in the profile

The specification is silent on this. The relevant definitions are, so far as I can find:

fixed[x]: Specifies a value that SHALL be exactly the value for this element in the instance. For purposes of comparison, non-significant whitespace is ignored, and all values must be an exact match (case and accent sensitive). Missing elements/attributes must also be missing.
pattern[x]:
Specifies a value that the value in the instance SHALL follow - that is, any value in the pattern must be found in the instance. Other additional values may be found too. This is effectively constraint by example
min: The minimum number of times this element SHALL appear in the instance

I think that the spec leans to #2 above.

@Ewout Kramer @Lloyd McKenzie

Grahame Grieve (Jul 21 2020 at 23:08):

presently the validator follows #1 for both fixed and pattern

Lloyd McKenzie (Jul 22 2020 at 04:41):

My preference is 3. Having a min that's ignored is going to be confusing to implementers and I can't see any reason to allow min of 0 when there's a fixed value or pattern

Richard Townley-O'Neill (Jul 22 2020 at 05:05):

I like option 3 as it naturally fits with this case:
I want to specify a min of 2 with a pattern that describes only one use.
So the min has to be >= the number of instances in the pattern.

Ewout Kramer (Jul 22 2020 at 07:28):

The .NET validator is doing 1 - if an element is not there, none of the validations are run on it (except, of course, the cardinality validation)

Ewout Kramer (Jul 22 2020 at 07:29):

This is true for both 'fixed' and 'pattern'.

Ewout Kramer (Jul 22 2020 at 07:31):

I don't see (yet) why a fixed/pattern would require the element to be present. A required binding (which feels similar to me), does not require a min cardinality of 1 either?

Frank Oemig (Jul 22 2020 at 09:04):

:+1:

Chris Moesel (Jul 22 2020 at 13:57):

I always interpreted fixed/pattern with min 0 to mean that the element is not required, but if it is there, it must be this fixed value (or follow this pattern). I think this aligns with @Ewout Kramer's thoughts above.

Chris Moesel (Jul 22 2020 at 13:59):

Trying to think of a use case...

Chris Moesel (Jul 22 2020 at 14:01):

If a Patient profile has active 0..1 and fixedBoolean: true, then it's effectively saying "don't give me explicitly inactive patients" -- I only want patients that are explicitly active or don't have active set (which generally implies active).

Chris Moesel (Jul 22 2020 at 14:02):

Otherwise I guess you'd need to use an invariant to say that.

Lloyd McKenzie (Jul 22 2020 at 14:25):

That can work only if the "meaningIfMissing" exactly corresponds to the fixed value.

Chris Moesel (Jul 22 2020 at 14:35):

I guess another use case would be for Quantity. If you leave system as 0..1 but also have fixedSystem: "http://unitsofmeasure.org", you're saying you allow unit-less quantities (e.g., 5) but if your quantity has a unit code, it better be UCUM.

Nik Klassen (Jul 22 2020 at 15:07):

If we don't interpret it as #1 then it doesn't seem like there's a way to express the use case of "it can be missing, but if it's here it has to look like this". I have no idea how realistic that use case is though

Nik Klassen (Jul 22 2020 at 15:09):

Like Ewout said, it's the most consistent (and easiest to implement) to just ignore all other validations when the item is missing.

Grahame Grieve (Jul 22 2020 at 20:17):

ok we line up with #1. So task to clarify: J#28126

Richard Townley-O'Neill (Jul 23 2020 at 23:56):

Lloyd McKenzie said: