Doublets / Mutiple Instances of an Element on a Diagram

[Viking]

This request has been made already several times, e.g., https://sparxsystems.com/forums/smf/index.php/topic,26866.msg221902.html#msg221902. But I could not find an answer.

EA has evolved to a great EAM-tool. When I am asked for the biggest disadvantage of EA I mention the missing ability to represent an element multiple time (“doublets”). This is key in Enterprise Architecture Management. Multiple Connector Ends is not an appropriate solution.
•   Will EA support doublets and if so, when?
•   If not, why not?

[Paolo F Cantoni]

This request has been made already several times, e.g., https://sparxsystems.com/forums/smf/index.php/topic,26866.msg221902.html#msg221902. But I could not find an answer.

EA has evolved to a great EAM-tool. When I am asked for the biggest disadvantage of EA I mention the missing ability to represent an element multiple time (“doublets”). This is key in Enterprise Architecture Management. Multiple Connector Ends is not an appropriate solution.
•   Will EA support doublets and if so, when?
•   If not, why not?

Hi Viking,
Can you explain why Virtual Connector Ends (VCEs)  aren't appropriate for your needs?  You may recall that I have many issues with VCE rendering, but I have also noted that I think, in principle, VCEs are, conceptually, the correct way to manage multiple copies of elements on a diagram.


Paolo

[Viking]

This request has been made already several times, e.g., https://sparxsystems.com/forums/smf/index.php/topic,26866.msg221902.html#msg221902. But I could not find an answer.

EA has evolved to a great EAM-tool. When I am asked for the biggest disadvantage of EA I mention the missing ability to represent an element multiple time (“doublets”). This is key in Enterprise Architecture Management. Multiple Connector Ends is not an appropriate solution.
•   Will EA support doublets and if so, when?
•   If not, why not?

Hi Viking,
Can you explain why Virtual Connector Ends (VCEs)  aren't appropriate for your needs?  You may recall that I have many issues with VCE rendering, but I have also noted that I think, in principle, VCEs are, conceptually, the correct way to manage multiple copies of elements on a diagram.
Paolo

Hi Paolo,

Could you tell me please why VCEs are the correct way to manage multiple copies of elements on a diagram? Who tells us that only an element can be represented only once on a diagram?

Issues are in fact: no possibility to resize, and no possibility to not use rectangle notation.

V.

[Eve]

Could you tell me please why VCEs are the correct way to manage multiple copies of elements on a diagram? How tells us that only an element can be represented only once on a diagram?

Paolo's reasons are probably different from mine, but we looked at the kinds of diagrams users were saying they couldn't create because of the limitation of showing each element only once. The common story was that a complex layout meant they needed a way to show one end of a connector closer to the other end. That's what was implemented.

Issues are in fact: no possibility to resize, and not possibility to not use rectangle notation.

Neither of those are reasons why the virtual connector end is an inappropriate solution. I would describe them as a request for more options for the existing feature.

[Paolo F Cantoni]

Hi Paolo,

Could you tell me please why VCEs are the correct way to manage multiple copies of elements on a diagram? How tells us that only an element can be represented only once on a diagram?

Issues are in fact: no possibility to resize, and not possibility to not use rectangle notation.

V.

Hi Viking,
I agree with your points about the issues regarding the VCEs. In essence, they aren't first-class diagram objects. I can't see why they couldn't have all the attributes of a standard diagram object and, indeed, have an entry t_diagramobject (with an indicator to say the item is a VCE)—which they don't at present.  In addition, their behaviour when arcs are suppressed must be the standard behaviour.

Now, to my reasons for believing VCEs are conceptually the correct way to achieve the ends of multiple instances of the SAME item on the diagram.

Firstly, we must agree that we deal with a diagram as a model view.  Consequently, only one instance of an object is in a given state in the model.  (See what I did?  I allowed multiple occurrences in different states - as with TAM ).

Why do we put model items onto a diagram?  It is because they are related in some way.  The modelling technology that you use may not be able to express these relationships, but they exist nonetheless.  In addition, there are relationships that EA chooses NOT to express via t_connector entries but by other means.  In our technology, we surface all such relationships that we are interested in (including those implied by visual embedding).  Consequently, their relationships are notionally visible when we place items on a diagram.

If, for whatever reason, we wish to place more than one instance of the same item on a diagram, we are implicitly saying that (at least) one of those relationships) can be attached to the instance!
Similarly, that statement implicitly requires that the instance has access to all the item's inherent properties (but separate display properties if needed).

So far, so good.

The tricky part is deciding which relationships need to be assigned to the VCE and which to the original item, since there may be more than one. The current implementation only allows one, but conceptually, that may not be the case.  Still, the current implementation is good enough, mainly since we suppress relationships that don't have a diagram in most of our use cases.

To summarise, VCEs provide a conceptually valid vehicle for allowing multiple instances of the same item in the same state on a single diagram while preserving the integrity of the underlying model.  However, to resolve the outstanding issues, they need to be closer to first-class citizens of the diagram than they are at present.  Similarly, the diagram-specific connectors need to be enhanced to indicate which instance they will be connected to.  Lastly, both the vertices and arcs ned to be fully manageable from a rendering perspective to allow the modeller full expression.

Thoughts?
Paolo

[Viking]

Issues are in fact: no possibility to resize, and not possibility to not use rectangle notation.

Neither of those are reasons why the virtual connector end is an inappropriate solution. I would describe them as a request for more options for the existing feature.

We are using interfaces as circles. This helps readers to easier understand the diagrams. With VCE, all elements are shown as "Retangle Notation". Why is it not possible to turn "Rectangle Notation" off? What is the advantage or the design value?