Re: Ada is getting more popular!

["Vinzent Hoefler" <nntp-2010-10@t-domaingrabbing.de>]

On Sat, 30 Oct 2010 02:30:05 +0200, Yannick Duchêne (Hibou57) <yannick_duchene@yahoo.fr> wrote:

> To be more precise or exact, you may have missed to point the area these
> comments applies to. I guess this is right for some requirements (you
> talked about a manual, seems to contains restrictive requirements).

Yes. What we do is mission-critical at minimum. You should see the restriction
they put on _safety-critical_. ;)

> Say it's universal is less obvious.

Well, the problems that may appear are universal. The "risk assessment"
regarding the usage of such "potentially dangerous" features is not.

After all, a lot of software is written in C and other unsafe languages
and still works most of the time. ;)

> As an example of your objections, you said “They may even call subroutines
> which are completely out of the scope of the unit you're calling it from”
> : sometime this what is expected to be possible, most commonly with
> scripted application (which can extended with script). Better say this can
> occurs while it was not expected. Said this way, I would agree.

Let me put it this way: If you do it, you _have_ to expect it. If you're
happy with the possible consequences or not depends on other factors.

In other non-critical projects (support tools etc.) I already used tagged
types and even languages like Perl. Or GPS scripts in Python. If these
do something unexpected there's nothing more than just annoyance among the
developers. If our product does something unexpected, well... let me put
it this way: expressed in DO178B levels it would rank somewhere between
B and C.

> May be inheritance is dedicated to large complex set of types and
> composition where it needs to be predictable.

I could agree with that. :)

> I've understood you complained, not about inheritance and rather about
> dynamic dispatching. But when inheritance come, dynamic dispatching is
> near to be there, just because if there is inheritance, you have to say
> what the operations associated to these types will become as inheritance
> goes (you can also make them all frozen, or never use class wide types,
> true).

You understood correctly.

> As another question to feed the talk : will you see a way to handle large
> set of types with composition ?

It's certainly possible, yes.

> Does composition provides another way to
> group by similarities and handle these groups as a whole ?

Apart from generics I don't see any nice solutions, no.


Vinzent.

-- 
There is no signature.