Re: Why use C++?

["Randy Brukardt" <randy@rrsoftware.com>]

"Niklas Holsti" <niklas.holsti@tidorum.invalid> wrote in message 
news:9ag33sFmuaU1@mid.individual.net...
...
> In what way does Ada (95, or the current standard 2005) not support your 
> needs for unsigned types?

I'm not the OP, but...

> If you dislike the wrapping behaviour of the Ada "modular" unsigned types, 
> what is wrong with defining your own unsigned integer type in Ada, as in
>
>    type My_Number is range 0 .. 53621;

This does not work on the largest unsigned integer type (mod 2**32 in 
Janus/Ada). Ada 95 requires

    type My_Number is range 0 .. 2**32-1;

to be rejected, as the upper bound exceeds Max_Int. (This is technically a 
"signed_integer_type_definition", which shows the cause of the problem.)

The Janus/Ada intermediate code can generate overflow checks for the largest 
integer type, but I've failed to find any reasonable way to model such a 
type in Ada (assuming we want a real integer type, with numeric literals). 
The closest is the "non-standard integer type" is 3.5.4(26) -- but why 
should such a basic need be "non-standard"? And having it "non-standard" 
means that you can't use type declarations like the above to use it. Yuck.

There are uses for wrapping types, but they are far less likely than wanting 
overflow detection. The default should be to catch errors, not turn them 
into different ones.

Too late to fix, unfortunately.

                                                      Randy.





> or whatever range you need. This gives you non-wrapping arithmetic with 
> run-time checks for overflow, underflow, and range.
>
> -- 
> Niklas Holsti
> Tidorum Ltd
> niklas holsti tidorum fi
>       .      @       .