String_Access in unbounded string handling?

[Blady]

Hello,

String_Access is defined in A.4.5 Unbounded-Length String Handling:
7 type String_Access is access all String;

and note:
75 The type String_Access provides a (nonprivate) access type for
explicit processing of unbounded-length strings.

I wonder what String_Access is for and what could be "explicit processing"?

Thanks, Pascal.

[Jeffrey R.Carter]

String_Access is a mistake that should not exist.

--
Jeff Carter
"If you think you got a nasty taunting this time,
you ain't heard nothing yet!"
Monty Python and the Holy Grail
23

[Dmitry A. Kazakov]

On 2024-01-14 12:17, Jeffrey R.Carter wrote:
> On 2024-01-14 12:05, Blady wrote:
>>
>> String_Access is defined in A.4.5 Unbounded-Length String Handling:
>> 7          type String_Access is access all String;
>>
>> and note:
>> 75  The type String_Access provides a (nonprivate) access type for
>> explicit processing of unbounded-length strings.
>>
>> I wonder what String_Access is for and what could be "explicit
>> processing"?
>
> String_Access is a mistake that should not exist.

Well, from one point of view, surely.

However I frequently need such a type because I in general refrain from
using Unbounded_String. Now, would be no problem to declare it as
needed, except for generics! If you have generic packages like:

generic
type Object_Type (<>) is private;
type Object_Access_Type is access all Object_Type;

You want all instances to share the same String_Access. So it is
conflicting. One is true, it has no place there. It should have been the
package Standard or none.

--
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

[Randy Brukardt]

"Jeffrey R.Carter" <spam.jrc...@spam.acm.org.not> wrote in message
news:uo0fs5$dja6$1...@dont-email.me...

> On 2024-01-14 12:05, Blady wrote:
>>
>> String_Access is defined in A.4.5 Unbounded-Length String Handling:
>> 7 type String_Access is access all String;
>>
>> and note:
>> 75 The type String_Access provides a (nonprivate) access type for
>> explicit processing of unbounded-length strings.
>>
>> I wonder what String_Access is for and what could be "explicit
>> processing"?
>
> String_Access is a mistake that should not exist.

I agree with Jeffrey. Whatever reason it was initially put into the package
has long since ceased to be relevant. And, as Dmitry notes, when you want
such a type, it's usually because you didn't want to use
Ada.Strings.Unbounded (or Bounded). So the placement is odd at best.

Randy.

[Randy Brukardt]

"Dmitry A. Kazakov" <mai...@dmitry-kazakov.de> wrote in message
news:uo0tkt$g6av$1...@dont-email.me...
...

> ... It should have been the package Standard or none.

None for me. ;-)

One really doesn't want to put anything in Standard that isn't widely
needed, as those names become hard to use in other circumstances. In
particular, declarations in Standard hide anything that is use-visible with
the same name, so adding something to Standard can be rather incompatable.

One could mitigate use-visibility problems by allowing more extensive
overloading (for instance, of objects), but that causes rare and subtle
cases where a program could change meaning without any indication. (Where a
different object would be used, for instance.) That makes that too risky a
change for Ada.

Randy.

[Blady]

Thanks for all your answers,

This is probably a very minor subject, however I submitted it:
https://github.com/Ada-Rapporteur-Group/User-Community-Input/issues/79

Pascal.

[Tucker Taft]

The idea was to support the explicit use of new String'(...), X.all, and Unchecked_Deallocation rather than the implicit use of the heap inherent in Unbounded strings. It was recognized that you need a single global access type to avoid having to do conversions all over the place. This predated the availability of stand-alone objects of an anonymous access type (aka "SAOOAAATs" ;-), but those are not universally loved either. It certainly cannot be removed now without potentially very painful disruption of existing users. It could be moved to a different package without too much disruption, but I haven't seen any groundswell of interest in doing that either.
>
> Thanks, Pascal.

Take care,
-Tuck

[Randy Brukardt]

"Tucker Taft" <tucke...@gmail.com> wrote in message
news:afd791fa-853f-48fa...@googlegroups.com...

I'm dubious that there are any such users. Certainly, in the handful of
cases where I needed such a type, I just declared it (strong typing, you
know?) and never thought of Ada.Strings.Unbounded as being a place to find
such a type already defined. It is such an odd place I doubt anyone outside
of perhaps the people who defined the type ever used it.

OTOH, I agree that the compatibility impact is non-zero (anyone who did use
it would have to change their code), and the benefit of removing the type at
this point is close to zero (junk declarations abound in long-term Ada
packages, what's one more; and certainly there is a lot of unused stuff in
any particular reusable package and any particular use), so the cost-benefit
ratio doesn't seem to make a change here worth it. An Ada successor language
would design Ada.Strings.Unbounded rather differently (so as to be able to
use string literals directly with the type) and probably would include
universal character support as well, so it's hard to find an important
reason to change this.

Also, I'm pretty sure we're discussed this within the ARG several times in
the past, so this is well-trodden ground.

Randy.

[Blady]

At least, the type String_Access could be tagged as obsolescent.

Pascal.