From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00,FREEMAIL_FROM
	autolearn=unavailable autolearn_force=no version=3.4.4
Path: 
 eternal-september.org!reader01.eternal-september.org!reader02.eternal-september.org!feeder.eternal-september.org!aioe.org!.POSTED!not-for-mail
From: Victor Porton <porton@narod.ru>
Newsgroups: comp.lang.ada
Subject: Re: Why stack can hold indefinite objects but records cannot?
Date: Tue, 28 Nov 2017 23:31:49 +0200
Organization: Aioe.org NNTP Server
Message-ID: <ovkkkc$1m51$1@gioia.aioe.org>
References: <ovkh24$1bh5$1@gioia.aioe.org> <ovki03$1fdr$1@gioia.aioe.org>
NNTP-Posting-Host: edFHTOfx8phAphItWrZ8cQ.user.gioia.aioe.org
Mime-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7Bit
X-Complaints-To: abuse@aioe.org
User-Agent: KNode/4.14.10
X-Notice: Filtered by postfilter v. 0.8.2
Xref: reader02.eternal-september.org comp.lang.ada:49223
Date: 2017-11-28T23:31:49+02:00
List-Id: <comp.lang.ada>

Dmitry A. Kazakov wrote:

> On 2017-11-28 21:31, Victor Porton wrote:
>> Why is it OK to create an indefinite object as a local variable ("on
>> stack") but not OK to put an indefinite object into a record?
> 
> It is OK in both cases. You must provide the constraint that is.
> 
>> By the way, how indefinite local variables can be implemented? (just
>> curious)
> 
> They cannot, there is no such thing.
> 
> A variable can hold a value which constraints vary. Each value is
> constrained. These either deploy a maximum allocation length like
> bounded strings or hidden referential semantics like unbounded strings.
> 
>> Moreover, wouldn't it to be a good idea to allow indefinite objects in
>> records (thus making the record itself indefinite, even if it has no
>> discriminants)?
> 
> It is a bad idea because there is no straightforward implementation of.
> 
> The proper solution is to have an abstract record interface which would
> allow anything appear a record member whereas the implementation would
> be left to the user. The compiler should provide simple implementation
> of that out of the box, maybe even less than it does already.

But for the stack this is implemented. I don't understand why the same 
cannot be done for records.

-- 
Victor Porton - http://portonvictor.org