ARG Urgent Problems (was: labeling)

ARG Urgent Problems (was: labeling)

[Larry Kilgallen]

In article <mailman.1016180522.31471.comp.lang.ada@ada.eu.org>, Christoph Grein <christoph.grein@eurocopter.com> writes:

> The ARG really has enough urgent problems to solve, and I bey they are
> reluctant to handle any such weird proposals.

Out of curiosity, what are some of the "urgent problems" ?

Re: ARG Urgent Problems (was: labeling)

[Tarjei T. Jensen]

Larry Kilgallen wrote:
> Christoph Grein writes:
>
> > The ARG really has enough urgent problems to solve, and I bey they are
> > reluctant to handle any such weird proposals.
>
> Out of curiosity, what are some of the "urgent problems" ?

They could get us unsigned integers and not just the positive part of
integers. By unsigned I mean unsigned with runtime checking.

Quoting C unsigned as a reason for not having real unsigned is pretty bogus
because C integers seems to wrap very well.

We don't want to abandon runtime checks on integers just because C does not
do any checking.

greetings,

Re: ARG Urgent Problems (was: labeling)

[Nick Williams]

What exactly do you mean by 'unsigned integers and not just the positive
part of integers'?

If you want to model unsigned integers the same way that C models them, then
use modular types, surely?

Nick.

"Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> wrote in message
news:a7240i$rh416@news.kvaerner.com...
> Larry Kilgallen wrote:
> > Christoph Grein writes:
> >
> > > The ARG really has enough urgent problems to solve, and I bey they are
> > > reluctant to handle any such weird proposals.
> >
> > Out of curiosity, what are some of the "urgent problems" ?
>
> They could get us unsigned integers and not just the positive part of
> integers. By unsigned I mean unsigned with runtime checking.
>
> Quoting C unsigned as a reason for not having real unsigned is pretty
bogus
> because C integers seems to wrap very well.
>
> We don't want to abandon runtime checks on integers just because C does
not
> do any checking.

Re: ARG Urgent Problems (was: labeling)

[Tarjei T. Jensen]

"Nick Williams" <nickw@acm.org> wrote in message
news:1016370675.163940@ananke.eclipse.net.uk...
> What exactly do you mean by 'unsigned integers and not just the positive
> part of integers'?

Unsigned in Ada (Natural) is from 0 to integer'last which is the same as
system.max_int;

> If you want to model unsigned integers the same way that C models them,
then
> use modular types, surely?

I just said that this is not what I want. I want a real unsigned with
rangechecking and the lot  - and not the abominable modular type workaround.

It really should be possible to map the C type  size_t to an ada unsigned
type without having to botch the job by using a modular type.

C allows both integers and unsigned to wrap, but somehow those who decides
these things think that therefore Ada unsigned above integer'last must wrap
as well.

That logic beats me.

greetings,

Re: ARG Urgent Problems (was: labeling)

[Jean-Pierre Rosen]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 844 bytes --]


"Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> a �crit dans le message news: a7700l$84f4@news.kvaerner.com...
> C allows both integers and unsigned to wrap, but somehow those who decides
> these things think that therefore Ada unsigned above integer'last must wrap
> as well.
>
Maybe it's just that you missed an important feature of Ada: don't use predefined types, specify your needs, then declare the
appropriate types. You want a 32 bits integer without wrapping semantics? Fine, just declare:

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

Your point is that there is no *predefined* type for this; in other languages you must rely on predefined types because it is all
you have - not in Ada.

--
---------------------------------------------------------
           J-P. Rosen (rosen@adalog.fr)
Visit Adalog's web site at http://www.adalog.fr

Re: ARG Urgent Problems (was: labeling)

[Christoph Grein]

From: Jean-Pierre Rosen <rosen@adalog.fr>
> "Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> a écrit dans le message news: 
a7700l$84f4@news.kvaerner.com...
> > C allows both integers and unsigned to wrap, but somehow those who decides
> > these things think that therefore Ada unsigned above integer'last must wrap
> > as well.
> >
> Maybe it's just that you missed an important feature of Ada: don't use 
predefined types, specify your needs, then declare the
> appropriate types. You want a 32 bits integer without wrapping semantics? 
Fine, just declare:
> 
> type My_Int is range 0..2**32-1;
> 
> Your point is that there is no *predefined* type for this; in other languages 
you must rely on predefined types because it is all
> you have - not in Ada.

The only problem being that your compiler might reject to compile this because 
it cannot fulfil the request. My_Int'Base will necessarily be a _signed_integer_ 
at least big enough to hold the specified range, will thus need at least 33 
bits, most probably a 64 bit 
type.

Re: ARG Urgent Problems (was: labeling)

[Tarjei T. Jensen]

"Jean-Pierre Rosen" wrote
>
> "Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> a �crit dans le message
news: a7700l$84f4@news.kvaerner.com...
> > C allows both integers and unsigned to wrap, but somehow those who
decides
> > these things think that therefore Ada unsigned above integer'last must
wrap
> > as well.
> >
> Maybe it's just that you missed an important feature of Ada: don't use
predefined types, specify your needs, then declare the
> appropriate types. You want a 32 bits integer without wrapping semantics?
Fine, just declare:
>
> type My_Int is range 0..2**32-1;

Try
type My_int is range 0 .. 2**64-1;

Even on 32bit unix there are the occational 64bit unsigned integer these
days.


greetings,

Re: ARG Urgent Problems (was: labeling)

[Jean-Pierre Rosen]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 957 bytes --]


"Christoph Grein" <christoph.grein@eurocopter.com> a �crit dans le message news:
>The only problem being that your compiler might reject to compile this because
>it cannot fulfil the request. My_Int'Base will necessarily be a _signed_integer_
>at least big enough to hold the specified range, will thus need at least 33
>bits, most probably a 64 bit
>type.
True, but there is no requirement that T'Base occupies the same number of bits as T.
I just tried the following with Gnat:
    type Int32 is range 0..2**32-1;
    for Int32'size use 32;
    I : Int32;
    B: Int32'Base;
begin
   Put_Line (Integer'image (I'size));
   Put_Line (Integer'Image (B'size));

And it prints:
32
64

(note the use of 'size on objects rather than types - necessary to get the truth about how much memory space is used)

--
---------------------------------------------------------
           J-P. Rosen (rosen@adalog.fr)
Visit Adalog's web site at http://www.adalog.fr

Re: ARG Urgent Problems (was: labeling)

[Christoph Grein]

>The only problem being that your compiler might reject to compile this because
>>it cannot fulfil the request. My_Int'Base will necessarily be a 
>_signed_integer_
>>at least big enough to hold the specified range, will thus need at least 33
>>bits, most probably a 64 bit
>>type.
>True, but there is no requirement that T'Base occupies the same number of bits 
       ~~~
       The "but" is unnecessary here, you repeat in other words what I wanted to 
       say. If there is no such type on your machine, the compiler will reject
       the declaration. Gnat provides a 64 bit type, as you've tried:
>as T.
>I just tried the following with Gnat:
>    type Int32 is range 0..2**32-1;
>    for Int32'size use 32;
>    I : Int32;
>    B: Int32'Base;
>begin
>   Put_Line (Integer'image (I'size));
>   Put_Line (Integer'Image (B'size));
>
>And it prints:
>32
>64