Stream: implementers
Topic: Operation vs. Interaction
Cooper Thompson (Jan 30 2017 at 15:11):
Is it correct to think that interactions (read, create update, etc.) are a subset of operations? Specifically, that interactions just RESTful operations? This is sort of implied in the operations documentation (http://www.hl7.org/FHIR/operations.html).
Ewout Kramer (Jan 30 2017 at 15:19):
We started out in FHIR having just interactions that mapped to the HTTP verbs - which then expanded beyond just verbs (_history) long before we had operations. When we introduced operations, some of the interactions were changed to operations ($validate was _validate, $meta was _meta). There is no hard and fast line between the two (you could argue that $search could be an operation), but it's not possible to actually call $read or $create....
Cooper Thompson (Jan 30 2017 at 15:30):
Ok, so it seems like treating operations and interactions as basically the same thing from an implementation perspective is appropriate. They just have different names and differing restrictions in inputs and outputs.
Ewout Kramer (Jan 30 2017 at 21:22):
Right.
Lloyd McKenzie (Jan 31 2017 at 01:18):
@Cooper Thompson Sort of . . . We use OperationDefinition to define operations that can be invoked using the $[name] mechanism. We don't have OperationDefinitions for the core http operations.
Eric Haas (Feb 01 2017 at 16:19):
This is something I am confused about as well. @Jenni Syed and I have been lobbying for a while for the LastN search parameter GF#9964. I originally thought this would be an operation $LastN vs _LastN, but Jenni thinks is a search parameter. What is the difference? Comparing the SearchParameter to OperationDefinition I can't really see how this could be defined in SearchParameter (involves grouping and sorting) but then I can't really see how _include or _revinclude could be either.
Lloyd McKenzie (Feb 01 2017 at 17:54):
I think implementing it via search would be cleaner. It's reasonable to want to filter, do includes, ask for _summary, etc. And it's much easier to "inherit" all of these if you're doing search.
Eric Haas (Feb 01 2017 at 18:07):
From a committers perspective this beyond my understanding. I assume there some magic behind the curtain stuff beyond the SearchParameter Resource.
Lloyd McKenzie (Feb 01 2017 at 22:51):
SearchParameters are allowed to have lots of non-computable magic
Brian Postlethwaite (Feb 02 2017 at 23:06):
e.g. soundex, loctaion proximity, free text semantics (on :text) or the Patient.name search parameter
Grahame Grieve (Feb 06 2017 at 06:34):
it's all computable
Grahame Grieve (Feb 06 2017 at 06:35):
_lastn is still underspeciifed as proposed. Since it's grouped last n, not outright last n. Needs more thought
Eric Haas (Feb 06 2017 at 07:16):
I've drafted something here and waiting for Jenni's input before doing anything else.
Last updated: Apr 12 2022 at 19:14 UTC
