Length of traceability

[Mats Gejnevall]

Hi
We have done a large activity model with activities and information. Activities create or use information. We would like to be able to generate some kind of report showing the all paths (traceability) through the model from start to finish. I am pretty sure it would become a very large document. An simpler alternative would be to get a metric showing the length of the longest chains and a report of that chain. I assume we could possibly get an infinite length chain, which would be an interesting result as well.

Any ideas of how to do this?

Thanks in advance
Mats

[Geert Bellekens]

I guess you could write a script that follows all paths...

Depending on the size of your model that might take a very long time.

Geert

[qwerty]

https://en.wikipedia.org/wiki/Graph_theory

q.

[Uffe]

Hej Mats!

Any ideas of how to do this?

A couple.

Infinite loops can be a worry as you've noted, but not a big one -- it can be handled in code.

In my view, the feasibility of this rather depends on how everything is represented and on how consistent you've been in the models. How is "create" and "use" information represented? Do you use CallBehavior Actions, so that activities call one another, or is each activity "flat"? Does each activity start with one (1) ActivityInitial? Etc.

Most important is consistency. The more leeway you've allowed the modellers (as in "as long as a person can sort of understand it when they read the diagram then that's OK"), the harder it gets.

Provided you've got a limited number of ways of representing each concept (preferably the limited number "1"), and provided the models adhere well to this metamodel, what I'd do is a document generation script which traverses your model and generates the document as it goes. Scripting it frees the documentation structure from the package structure, and I think that's what you need to do here.

HTH,


/Uffe

[Glassboy]

Infinite loops can be a worry as you've noted, but not a big one -- it can be handled in code.

It's always funny when the protection in code fails :-)

Manager:  Why is everything running slowly.
Minion:  We have an infinite loop.
Manager:  When will it finish?
Minion:  Never, its infinite.
Manager:  Well, can't you turn it off now?
Minion:  What, our main line of business application?

[Richard Freggi]

Hi
We have done a large activity model with activities and information. Activities create or use information. We would like to be able to generate some kind of report showing the all paths (traceability) through the model from start to finish. I am pretty sure it would become a very large document. An simpler alternative would be to get a metric showing the length of the longest chains and a report of that chain. I assume we could possibly get an infinite length chain, which would be an interesting result as well.

Any ideas of how to do this?

Thanks in advance
Mats

Can I suggest using sequence diagrams instead of activity diagrams.  Easier, faster, and the information used is right there in your face as participant objects.  Traceability is a breeze.
Activity diagrams can lead you into analysis paralysis because they are suited to document method/operation procedural logic, not object oriented architecture.

[Mats Gejnevall]

Thanks everyone, but unfortunately scripting is not my expertise.
We have an MDG with two elements; activity and information and only two relations are allow between these, create or use.
The business person modelling this only knows this MDG. So it should be pretty stable.

Is it possible to get the traceability using the document functionality? How?

Thanks
Mats

[Uffe]

Hej igen,


Limiting yourself to EA's document templates only (and not using any script template fragments) I don't think it's possible.

You can get a list of all activities, and for each a sub-list of the "created" and "used" information items.
Or you can get a list of all information items, and for each a sub-list of the "creating" and "using" activities.

But EA's built-in document generator simply follows the package structure and cannot follow connector chains. When generating a (part of a) document for one element (with that element "in focus" so to speak), a template has access to the element's connectors and some details about the element at the other end, but it can't "refocus" on that element and follow its connectors.

The only way I know to do arbitrary-length connector chains is to script it, either using the DocumentGenerator API or possibly script template fragments.


/Uffe

[Geert Bellekens]

I agree with Uffe, although you might get "far enough" with only SQL queries.

It would not be arbitrary long, but if you see that you have now maximum 5 levels, and you create a query that goes up to 10 levels, that would probably be good enough.

Geert