End Composite/Linked Diagram conflation

[Paolo F Cantoni]

The time has come to finally end the Composite Structure/Linked Diagram conflation in EA.

Whether an item is an aggregate (Note: NOT just a Composite) is either a derived property (as a result of metadata in the model) or an asserted property (by the modeller - in anticipation of creating the requisite metadata in the model).

It has NOTHING to do with whether I have attached one (or more) diagrams to the item.  However, since the ONLY way to attached a diagram to an item is to mark it as composite, there's no way to distinguish the real aggregated items from the "Clayton's" ones (search the forum for the meaning  of Clayton's).

I propose the existing "composite" functionality be renamed "Link to Diagram" (and a dialog be created to locate/create the diagram).

From what I recall, the existing functionality doesn't actually mention the word composite in its implementation and there is  more than one way to specify the attached diagram (depending on the element type).

EA has already started using a normalised process for linked diagram specification (using the "Default Diagram" t_xref entry for this purpose).

So why not cut the Gordian Knot and do it properly?

Additional functionality can then be created to allow the user to assert whether the item is an Aggregate (either Composite or Shared) and that metadata be stored and rendered (preferable on the item properties dialog) separately from the diagram linkage.

Reported
Thoughts? Votes?

[Dave.B]

This whole area of EA's functionality has irritated me for years!

Three things:

• If an element has more than one diagram then I'd like to see an automatic prompt for which diagram to open, or an automatically built and maintained index diagram that lists and links to all of the sub-diagrams.
• For composite diagrams the parent (container) element should be the diagram frame and include any selected ports, etc. I'm currently building a lot of SysML IBDs and have to manually bring the container block on to the diagram and enlarge it to contain the parts that I'm working with. Tedious and messy!
• I'd also like to see the referential integrity between the class (or SysML block definition) aggregation view and the composite model of parts automatically maintained. At the moment I can build a class or BDD model that shows the aggregated classes (or blocks) and this has no relationship what so ever with the parts that I actually create! This isn't rocket science and really a required behaviour for any CASE tool.

Dave B.

[Paolo F Cantoni]

Hi Dave,

In principle, I agree.  

But, frankly, I'd be deliriously happy with just a single diagram link setup properly!

However, after I posted, it occurred to me that if Diagrams as first class citizens of Repository were implemented then a DiagramObject on one diagram could be linked to a different diagram than the same (base) object in another diagram.  By default, they'd all point to the "DefaultDiagram" in t_xref (pun intended!).

That might go some way to solving your multiple diagram issue.

Paolo

[Geert Bellekens]

Yes,

I hate to explain to people (over and over again) how to link a diagram to an element. (especially when you don't wan the diagram to be nested in the element).
Personally I don't use the "real" composite feature, only diagram linking.

Geert

[Paolo F Cantoni]

[size=18]...[/size]
Personally I don't use the "real" composite feature, only diagram linking.

Geert

That's because (effectively) you can't.  Once you use the functionality for other than a composite structure diagram, you're gone...

Paolo

[Paolo F Cantoni]

Since posting the previous entry, I've stumbled across probably the exemplar of this conflation!

See: Actions CAN/CAN'T have diagrams!

Flabbergasted, but not surprised!
Paolo

[Jan Kastning]

I'd also like to see the referential integrity between the class (or SysML block definition) aggregation view and the composite model of parts automatically maintained. At the moment I can build a class or BDD model that shows the aggregated classes (or blocks) and this has no relationship what so ever with the parts that I actually create! This isn't rocket science and really a required behaviour for any CASE tool.

Although this topic is almost 9 years old:
Does anyone know if this has been fixed by now and in which version? I read throgh the change histories of the newer versions (12+) and don't find this feature added...

Best regards,
Jan