Why are directions of connector labels not connected to connectors?

[PeterHeintz]

I use connector direction to provide reading directions for users.
Example:
“Car” “has” “Engine” were the direction from “has” goes from Car to Engine to ensure that the user does not read “Engine” has “Car”.

However this direction is not bound to the connector but to the connector on a diagram.
In other words each time I use the connector in an other diagram I have to set the direction anew.

I thing binding the direction to the connector only should be there.
Is there a sense in that current implementation or is that just a wrong implementation?

Any opinions?

[Geert Bellekens]

I think it's a very poor implementation that lacks usability.

It is even worse, the direction is linked to the label on the diagram. Each label can have it's own reading direction.
I once started a script to add reading directions automatically when I discovered the reading directions for the other labels.

Geert

[qwerty]

EA is a bit like a limping athlete. They have a couple of very smart things in their user interface. But then there are a lot of other things that only make you think - what?

So intentionally I'd tend to "wrong implementation".

q.

[PeterHeintz]

Yes, that's my impression as well.

[Paolo F Cantoni]

Exactly what are we talking about here - those little triangles that one can make visible for Associations etc.; or something else?

If the former, we gave up and any relationship that can (conceptually) have a bidirectional access, has a tiny (2 pixel) circle at the target shape:

   ellipse(20,1,18,-1);   //Destination Indicator

Solved the problem!  We can immediately see the directionality (as opposed to navigability).

HTH,
Paolo

[Uffe]

I *believe* those indicators are non-standard. In which case there is a point, from a modelling purist's perspective, to limiting them to a diagram: you can add them for emphasis, but you shouldn't rely on them. If your connector has a direction (in your semantics as opposed to the EA implementation), you should specify that using the Direction drop-down in the connector properties and have EA render it in the standard way.

Or maybe I'm just grouchy because I've had to clean up fairly massive conceptual models where the previous EA "expert" relied heavily on this feature without understanding how the bloody thing works.

/Uffe

[PeterHeintz]

Don’t be so grounchy!

Well, that indicator is uml standard and I found an example on page 202 showing the intention as I use it.

It is not the association direction (if any), it is the direction of the label (how it should be read).
And label is label on whatever diagram it is.

[Geert Bellekens]

But I agree with Uffe to advise clients against using it.

Often an alternative can be to use rolenames instead.

Geert

[Paolo F Cantoni]

Don’t be so grounchy!

Well, that indicator is uml standard and I found an example on page 202 showing the intention as I use it.

It is not the association direction (if any), it is the direction of the label (how it should be read).
And label is label on whatever diagram it is.

(my emphasis) Ah yes, an important distinction.

Your point about label is label is well taken.  In my view you are REALLY saying which direction the name should be applied.  It should actually be a property of the connector, not just the diagram.  The diagram property should be whether to display the name or not (and perhaps whether to display the arrow), but not allow you to change the direction of the arrow.

Paolo

[qwerty]

There's just one use of the triangle in Superstructures in Fig. 11.27 on p. 202. Table B.6 on p. 718 shows the triangle and links to section B.3.5. This  describes the triangle as UMLShape and thus makes it a diagram element rather than a connector element. So in this case it's not Sparx to blame, but OMG. OMG - the best acronym they could probably find.

q.

[PeterHeintz]

To be honest, I am far away of being an expert in reading UML specs but I belief that to have the triangle or not is a sharp but it annotates the UMLEdge
Anyhow I have the feeling UML spec leaves plenty room for interpretation but a sense making interpretation would be to remain the direction on different diagrams.
I personal like and use this annotation/shape for very conceptional things just because it makes reading simpler than role names.

[qwerty]

The UML spec is not perfect, though their attempts to improve it are recognizable. You will for sure not see any change with this, even if you report it. OMG's mills grind very, very slowly. And changing this to be a connector property rather than a diagram property would force vendors to change their models. You won't find one to chip in.

q.

[PeterHeintz]

Of cause! I am not unrealistic.
I do not intend to proselytize dignified OMG.
Just wanted to find out with this issue, if I am interpreting something entirely wrong.

[PeterHeintz]

I rust realized that when I put SysML properties connected with a connector having a label with a direction to a second diagram that the triangle and the connection remain.!?
“Inconsistently correct systems DON'T EXIST!”