FHIR Chat · reference:type:missing search

Stream: implementers

Topic: reference:type:missing search

Chris Grenz (Mar 22 2019 at 11:09):

Ostensibly, I should be able to search for Observations that have a Patient performer by:
/Observation?performer:Patient:missing=false, but test servers either disallow (@Grahame Grieve ) or silently ignore (@Christiaan Knaap @James Agnew ). Is this a valid search (I hope so)?

James Agnew (Mar 22 2019 at 12:40):

I don't think that would be valid with the spec as written- there is no description of stacking modifiers like that that I can see. Certainly HAPI wouldn't allow this, we just don't have indexes currently that would support this search.

You could get this information pretty easily by defining a custom SP though that indexes only on something like Observation.performer.where(resolve() is Patient)

Alexander Zautke (Mar 22 2019 at 12:45):

It should have been documented that combining multiple modifiers is not valid. See https://gforge.hl7.org/gf/project/fhir/tracker/?action=TrackerItemEdit&tracker_item_id=17220

Chris Grenz (Mar 22 2019 at 12:54):

That's an unfortunate outcome I think...certainly there are valid uses for "stacked" modifiers.

Chris Grenz (Mar 22 2019 at 12:54):

Procedural question - since search is normative, could we allow this in future? Or would that be a breaking change?

Lloyd McKenzie (Mar 22 2019 at 14:32):

Adding new capabilities that don't break existing systems is allowed.

Grahame Grieve (Mar 22 2019 at 21:15):

I don't like that combination and there are many combinations that would be problematic. I really think this is better done as James suggests

Christiaan Knaap (Mar 23 2019 at 19:07):

If the patients are also stored in the server, you can search this with {{url}}/Observation?performer._type=Patient.
On a side note: Vonk does not ignore it silently, it included an OperationOutcome in the bundle stating "Parameter is not well formatted, expected 'parametername[:modifier]=value'."

Christiaan Knaap (Mar 23 2019 at 19:08):

And agree on preferably not stacking modifiers.

Chris Grenz (Apr 02 2019 at 11:01):

@Christiaan Knaap is there a tracker out there to standardize "_type" on resource? That'd be nice...

Chris Grenz (Apr 02 2019 at 12:17):

As to the complexity of "stacking" modifiers, I'm increasingly of the opinion that :missing should be a special case addable to a few modifiers. What I really mean is something like =null, and not being able to add that anywhere a value would be appropriate is a problem. For instance, Observation?code:text:missing=true or (maybe, reaching here) Patient?identifier:of-type:missing=true