Rationalising duplicate relationships

[Paolo F Cantoni]

We have a "replacer" technology that allows us to mark one item as to be replaced by another.  (This is one of the processes we use to reduce the number of duplicate items in the repository.)  The technology will move the relationships from the TobeReplaced to the Replacing item and will replace the TobeReplaced item by the Replacing item in the diagrams the TobeReplaced item appears in.

Having got this working well, we are now ready to try the same for duplicate relationships between the same two items.  We have determined (at least the first pass) properties that would allow us to decide that two arcs are semantically identical.  In order to perform the rationalisation, we believe we have to:
If the TobeReplaced arc is visible on a diagram,

• Ensure that the Replacing arc is attached to the diagram
• Make that Replacing arc visible on the diagram and that if it is new, make its geometry that of the TobeReplaced arc
• Hide the TobeReplaced arc

If the TobeReplaced arc is not visible on the diagram, do nothing.
Finally, the TobeReplaced arc can be purged from the repository.

Any other factors to be considered?
TIA,
Paolo

[Geert Bellekens]

- Association classes linked to the relation
- Notes linked to the relation
- Tagged values referencing the relation
- (since v13 only) relations to this relation
- Notes on the relation (merge?)
- Tagged values on the relation (merge?)

Geert

[qwerty]

I found this process a most challenging one and never dared to automate this. Hats off! You should highlight all diagrams where the replacement has effect so one can verify the change. I'm not sure what happens if some of the connectors are hidden on certain diagrams.

q.

[Paolo F Cantoni]

- Association classes linked to the relation
- Notes linked to the relation
- Tagged values referencing the relation
- (since v13 only) relations to this relation
- Notes on the relation (merge?)
- Tagged values on the relation (merge?)

Geert

Thanks Geert, Most of those are already in the item replacer, so I would have converted them to the relationship replacer.  But a couple had "slipped through" so thanks for the heads up on them.

Question, if one association has NO association class and the (otherwise) semantically identical association has an association class, do you consider them to be semantically identical?

Paolo

[Paolo F Cantoni]

I found this process a most challenging one and never dared to automate this. Hats off! You should highlight all diagrams where the replacement has effect so one can verify the change. I'm not sure what happens if some of the connectors are hidden on certain diagrams.

q.

The item replacer works quite well, and we DO highlight the relationships that have been moved.  Increasingly, however, we are finding that the highlight is just a bit of a pain as we have to remove it later.

Paolo