Concept Description missing from AAS (AASX-file) when External Reference is used. about basyx-python-sdk

[s-heppner]

Oh, no bother at all! I was curious myself what happened there, just not curious enough to post the issue myself 😄

If you need a UI, have you tried our AAS Manager?
It's not an Eclipse project (yet), but based on our SDK, so it's garantueed to be compatible with it.

[s-heppner]

I think there's a general problem in the understanding of ModelReference vs ExternalReference.

According to our understanding (and therefore our implementation), a ModelReference is resolvable in implementations of the AAS (since it references another AAS object), whereas a GlobalReference does not necessarily refer to other AAS objects.
See also the discussion here: #231
And the resulting issue with the specification of the AAS here: aas-specs#350

A ConceptDescription referenced with an ExternalReference could be located anywhere, which is why we're using an ExternalReference and not a ModelReference. But because we modelled it specifically to be "anywhere" and specifically not "another AAS object", it's in general not possible to programmatically check whether or not the ExternalReference by chance is resolvable and by even more chance located in the same ObjectStore.

Therefore, it is my understanding that ConceptDescriptions referenced via ExternalReference should not be written into a local AASX. To me, that would speak against the idea of the distinction between ModelReference and ExternalReference and would be a bug in the modelling.

When the comment in aaspe#28 mentions, it is recommended to use an ExternalReference, my understanding is that this means, the ConceptDescription should also be located somewhere externally, and not be delivered with the AASX.

When a property has a semantic reference to a Concept Description as a ModelReference, the ConceptDescription cannot be looked up within the AAS environment in the Package Explorer. See issue eclipse-aaspe/aaspe#11

As a logical consequence, I would also assume that this is therefore a bug within the AASX Package Explorer.

[WelliSolutions]

Oh dear... sorry for bothering you. I have no idea what the idea of that viewer is. I am just trying to build an AAS which is compatible to that thing.
I guess this is how standards go: someone does it wrong, and everybody follows the wrong implementation, just because it has a nice UI. Once there are enough implementations doing it like that, its easier to change the standard than the implementations.

[WelliSolutions]

Hi. Similar issue here, and I don't fully understand CDs yet.

Package Explorer does also not automatically include concept descriptions in the AASX file.
However, there is functionality to do that, e.g. in the digital nameplate viewer 1.0 (not in 2.0).

The following picture may demonstrate this:

1. The viewer already kicks in and recognizes the digital nameplate (blue frame 1.)
2. even if there are no CDs present (violet frame, 2.)
3. but there is a button which creates, imports or copies (I can't judge at the moment) the CDs into the current AASX (green frame, 3.)

After clicking the button, you'll find the CDs

The properties then refer to a ModelReference/ConceptDescription

and that concept description then refers to a ModelReference/GlobalReference with a data specification as ExternalReference/GlobalReference.

Obviously the Nameplate Viewer 1.0 thinks that the CDs should be part of the AASX file which uses the CDs. Otherwise they wouldn't have implemented the "Fix missing CDs" functionality.

[s-heppner]

Okay, so what I've gathered from your screenshots is that the Nameplate Viewer 1.0 references the (magically created?) ConceptDescriptions via ModelReferences with a Key with KeyType GlobalReference.

This violates the specification on several constraints, most importantly:

Constraint AASd-123: For model references, i.e. References with Reference/type = ModelReference, the value of Key/type of the first key of Reference/keys shall be one of AasIdentifiables.

Could you maybe inquire, what the Nameplate Viewer 1.0's idea is, regarding ConceptDescriptions?
At the very least, this is most likely a bug in their referencing of ConceptDescriptions.