Hello,
I have a question the is a bit related to the discussion about the constraining of the ELEMENT type in the laboratory_analytes.
The current specification defines the field "ehr_status" of the class EHR with the type OBJECT_REF. In the AQL specification there is an example (chapter 3.7.2.3. NOT) that accesses this field with the assumption that the field is of type EHR_STATUS.
I have written a type checker for AQL queries, so I am now stumbling across queries that access fields of potential subclasses or derived archetypes.
Does/should the specification generally allow such a thing ?
Greetings
Georg
As things stand, No, the spec does not allow such a thing (since it is unspecified at the moment) and also partially No, (I don’t think) it should not allow such a thing auto-magically or try to be smart about it.
As things stand, AQL has only access to information provided by the RM. I for one would be uncomfortable with Aql queries making assumptions about actual subtypes of abstract/parent types because validating those at the query level is impossible.
It is semantically ‘downcasting’ the type and it should only be done in an explicit way when the authors of the queries explicitly declare the cast and take the risk.
I’ll take a look at the example you mentioned, but it does not sound right to me. My earlier response to Tom mentioning comparison operators also fall under this topic.
All the best
Seref
Bah, meant “..(I think) it should not…”
Hi Georg,
The generic way to do this is to do what archetype tools do. An archetype itself mentioned class and attribute names from a ‘reference model’, but it doesn’t know anything about any particular model. SUch tools have to be told to load a representation of the model in order to validate an archetype against it (i.e. in addition to validating correct ADL structures etc). Today, every tool is doing this by loading the BMM schemas of the reference models. Probably the Archie project contains the easiest to use code that does this. (Note: the latest Marand ADL-designer, and LinkEHR, as well as the original ADL Workbench all use BMM files these days; but Archie is probably going to have the cleanest, most modern source code to work with).
The general picture is the same for AQL queries: figure out from some environment var, config setting or the RM names in the archetype ids in the queries which BMM to use to do the validation. Once you have that it is easy.
If you want to know more on Archie, ask here, the authors will no doubt respond.
- thomas