Challenge to update own repository when reference model updates

[ea0522]

In my project, I have an external source providing a reference model which I incorporate (import) in my own repository.
As intended, I link my own objects to objects in the reference model to gain insight in how things work.
The challenge I have is how to handle an update of the reference model, eg. from V1 to V2 where reference objects are changed, deleted and added.
The reference model is already separate in a different package.

I have tried several scenario's:

• Remove the complete V1 package and import the new V2 model in my repository.
  A drawback from this approach is that I then lose all relations between the reference model and my own objects.
• Import the new V2 and deduplicate objects that are present in both versions.
  The drawback is that there may be V2 elements that are renamed and thus not recognised as being duplicate. Also V1 objects that are not present any more in V2 still exist in the reference model.

Any suggestions on how to handle this challenge and overcome the drawbacks are very welcome.

[Geert Bellekens]

What format does this reference model have?
Is it an EA model xmi file, or something different.

Key in a process such as this is to have a unique identifier for each object. That will allow you to detect name changes.

If you don't have that you'll have to compare the two model. This could give you a list of deleted object an new objects.
You could then manually do a mapping from deleted to new objects for those you know to be renames.

The last part could be a script that moves (or copies) the traces from V1 to V2. If the objects have a unique identifier that is quite easy. If not it will be based on the name, and on the rename mapping you did.

Geert

[ea0522]

Thanks for the quick reply.
The source of the reference model is an Archimate file (from the free ArchiTool).
I can import this archimate file into an EA QEA repository.
With the QEA repository, I can do EA tricks like adding a new TAGGED value Source=V1 or V2 which could help the comparison.
Would baselining be of any help here?

Otherwise there is still a lot of manual work involved...

[Geert Bellekens]

There's no need for a lot of manual work.

The only thing you would need to do manually is to map renamed elements.
If your archimate elements have a unique identifier, there's not even a need for this.

Then it's a matter of writing a simple script that moves or copies your trace relations depending on your preference.

Geert