[Uml-devel] Re: Umbrello design (was UML XML tree viewer)

[Andrew Sutton]

> OK. Here it is.
> Imagine (first example) you're building an application that contains an
> interface to a database. Some classes will have methods to access
> database. Database acccess methods can be easily deduced from the
> attributes description (if not, let's suppose it for the example). So,
> the user is able to specify (in that project) an automatic generation of
> database methods. What he needs to realize that is:
>
>     * a tagged value (say {persistant}, for example, and the name of the
>       database management system),
>     * an entry point to act on code generation, and
>     * a way to get required information (attributes description and
>       presence of tagged value).

okay... this is getting much clearer now :) the uml extension mechanisms 
(stereotypes, etc.) actually support tagged values. ideally, we should be 
able to describe real UML profiles.

so, we need to be sure that our codegen pattern includes - specifically 
includes - the idea that the semantics of a UML object vary greatly depending 
on (a) stereotypes and (b) tagged values.

> This first example is just to illustrate needs for customization in code
> generation.
> Let me give an other one.
> Imagine now a simulation application (this example is easy for me
> because I did that work a few years ago). I don't know simulation
> standards but I guess this application should deal with actors (not UML
> actors) that reacts to events and so on. Oh ! I know a UML semantic
> which allow to specify that sort of behaviour : statecharts. Same that
> in previous example, what is needed to realize that is:

i'm not sure if you "actory" is really a tagged value {actor} or a stereotype 
<<actor>> :) either way, the same from above applies. extensions alter 
semantics and we have to design specifically for that.

> That's all. Do you feel it is a real need or is it a too strange usage
> of CASE-tool ?
>
> PS : Note that those (customization) functionalities can be found in an
> existing CASE-tool : Objecteering (of Softeam). That used to be a good
> tool, especially for code for generation (I used it a few years ago).

i don't think its strange at all. i think its using the capabilities of a 
large and detailed standard to do real world work - and i think its something 
we should strive for.

andy