Re: Ada and UML

["Yannick Duchêne (Hibou57)" <yannick_duchene@yahoo.fr>]

Le Wed, 01 Sep 2010 20:44:09 +0200, Simon Wright <simon@pushface.org> a  
écrit:
> I was going to draft a reply but Matteo has said almost all that I would
> have done. I'd add that
>
> 1 I wouldn't expect every possible UML construct to be translatable
>   using this model. A given vendor/toolset might well specify
>   limitations, presumably as a profile.
>
> 2 I wouldn't expect to be able to tweak the UML so as to generate any
>   arbitrary target language (Ada) construct.
I did not expected this, nor me.

> 3 The converse of 2: I wouldn't expect to be able to represent any
>   arbitrary target language (Ada) program in UML so as to be able to
>   regenerate it.

What I expect, may be explained with concrete cases.

I feel for long source does not came with all value and sometime, the  
“clever idea” in something comes at an abstract level which not any more  
clearly visible in a source, even if the source implements this “clever  
idea”. Because the “clever idea” is made implicit and not expressed by the  
source whose target is to express concrete implementation. That is why  
even a well designed application source can be “unreadable”.

As an example, I have an FTP application written 5 years ago. Not so much  
big, 2_000 lines. But I have to rework it, and although there are comments  
and the design was clear at the time I wrote it, it seems impossible to be  
sure to clearly understand what I had in mind at the time. I have another  
web application, 70_000 lines, which is clear to me for the time, as I  
oftenly come back to it... but will this always be like this ? I am afraid  
if I ever switch to something else for a long time or want a later date to  
collaborate with any somebody, things gonna be wrong.

Structuration, god entity naming, etc, are nice things, but this is simply  
not enough.

And what if I ever want to re-implement these things in another language ?  
(one is C, the other is Pascal and JS) A direct translation to Ada (which  
I plan for the future) would surely turn into many errors every where.

I also had some other similar though about UI design.

This is these topics I have in mind. I am not seeking for another way to  
write Ada with something else than Ada.

May be we agree in some points and I did express it the good way.

As some ones says : you do not see the face on a printed image looking at  
the little colored dots the raster printer drawn on the paper. And this is  
sadly true even with tiny pictures (it is frightening to think something  
as simple a 2_000 lines application can become so much unclear... and what  
about bigger things...).

But while I feel this may help, I still have some doubt about the UML  
approach : what about things which are not design or architecture and  
which are rather algorithmic ? Is UML well suited to express algorithm in  
an abstract manner ? In that area, I had a recent look at (S)ML, but still  
do not have a clear though (there are state machines in UML, I known, but  
this is not enough, and an algorithm too may have multiple  
implementations, so a more abstract view than a particular implementation  
could be welcome too).

Still in this area, I also explored and tested other techniques, but I'm  
afraid I would become boring... (at least I have a better picture of the  
subject than the one I had, say, 3 years ago)

Have a nice day all



-- 
“Dual licensing is the Perl's way to disinfect the GNU General Public  
Virus!” (anonymous)