[aosd-discuss] Metrics for Aspect-orientd Modelling

[Andrew Carton]

I've changed my mind on this. I think the solution lies more in the  
model-driven domain than the aspect-oriented domain.

Comparing AO + OO is like comparing apples and oranges ;) In fact,  
not even, -  its like comparing a granny smith apple with a valencia  
orange. There are too many variations of source-code / modelling  
languages to compare against and it is difficult to define a metric  
for a new variation.

There is no common representation for two grammars (source-code  
languages) to compare. At the moment, it is semi-formal. In model- 
driven development, we have a meta-metamodel for the formal  
definition of metamodels. Since there is a common foundation, there  
is at least some grounds for a set of metrics to be derived for  
comparing two meta-models and then models. So now we can compare our  
apples and oranges maybe in terms of "properties of a fruit" ? You  
could have a meta-rule saying "Variety A of Fruit B has more sugar  
than Variety C of Fruit D, therefore it is sweeter". You could also  
define formally what sugar means or what sweet means. I think this is  
where the state-of-the-art in metrics is going anyway from what i've  
read?

I found the below paper interesting. It talks about evaluating  
maintainability from a domain-specific point-of-view. They argue that  
maintainability metrics based on source code are not applicable to  
models. They then apply the Goal-Question-Metric approach to define  
their own evaluation / metric system. I found this really interesting  
because rather than retro-fitting the existing metrics and trying to  
derive something from these (like i previously was doing), you define  
your own evaluation scheme and this is more suitable for what you  
want to evaluate. Does anyone have an opinion on this?

Evaluation of Maintainability of Model-driven Persistency Techniques,
Thomas Goldschmidt, Jochen Winzen, Ralf Reussner
http://www.sciences.univ-nantes.fr/MoDSE2007/p14.pdf

Regards,
andrew.


Dé hAoine, 24 Lúnasa 2007 10:31, scríobh Andrew Carton,

> I am in a similar situation and need to compare an AO vs. an OO
> model. From my brief introduction to this topic, I think this is an
> interesting and very challenging exercise for a number of reasons.
>
> First, I believe the level of abstraction is important. If modelling
> at a high level of abstraction, the models are not going to be
> computationally complete, so the comparison is only valid if they are
> only equally as detailed as each other. Generally, the metrics out
> there give guidelines on what the numbers mean - are these numbers
> still applicable when working at a different level of abstraction
> than code? I think not!
>
> Second, I think the next challenge is deciding how to map these
> metrics from code to metamodel artefacts. If using the UML, I think
> an activity diagram is very good for something like mccabes
> cyclomatic complexity because it is similar to a control-flow graph.
> Other metrics aren't so easy, like lines of code, Halstead complexity
> metrics or those that depend on counting code level artefacts need to
> be redefined and reinterpreted, for that specific metamodel.
>
> Third, add AO and its slew of metrics into the mix further
> exacerbates these problems. The new metrics need to mapped to your
> chosen AOM and how that is represented (metamodel or Profile right?).
> Most automated metric tools out there run over aspectj. You also have
> to assume there is a correlation between these metrics and your AOM
> (maybe your AOM supports different types/levels symmetry?), and then
> from these to the OO metrics.
>
> In the end, I think you just have to state the assumptions, be
> consistent in applying the metrics and derive an argument from the
> observation of the results? Easiest approach if using a low-level
> abstraction is just to transform to code and use the tools to do the
> dirty work. I think metrics are pretty silly anyways...
>
> This works on an OO UML model and may be useful to be adapted.
> http://www.eclipse.org/m2m/atl/usecases/ModelsMeasurement/
>
> There is also a workshop for MoDELS "Model Size Metrics" that has
> some interesting stuff!
>
> Regards,
> andrew.
>
>
>
> Déardaoin, 23 Lúnasa 2007 13:57, scríobh Jaroslav Svacina,
>
>> Hello,
>> Are you aware of any metrics for aspect-oriented modelling. My  
>> purpose
>> is to compare one design of an application using object-oriented
>> models
>> with another design of that application with aspect-oriented  
>> models. I
>> would be grateful if you could point me to works in that area.
>>
>> Best regards,
>>    Jaroslav Svacina
>>
>> -- 
>> Dipl.-Inf. Jaroslav Svacina
>> Tel.: +49 (0)30 6392 1908
>> E-Mail: Jaroslav.Svacina at first.fraunhofer.de
>>
>> Fraunhofer-Institut fuer Rechnerarchitektur und Softwaretechnik,  
>> FIRST
>> Kekulestraße 7
>> D - 12489 Berlin
>> Germany
>> http://www.first.fraunhofer.de
>>
>>
>> _______________________________________________
>> discuss mailing list    -    discuss at aosd.net
>>
>> To unsubscribe and change options, go to:
>> http://aosd.net/mailman/listinfo/discuss_aosd.net
>>
>> Check out the AOSD.net Wiki: http://aosd.net/wiki
>
>
> _______________________________________________
> discuss mailing list    -    discuss at aosd.net
>
> To unsubscribe and change options, go to:
> http://aosd.net/mailman/listinfo/discuss_aosd.net
>
> Check out the AOSD.net Wiki: http://aosd.net/wiki