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=-0.4 required=5.0 tests=AC_FROM_MANY_DOTS,BAYES_00,
	XPRIO autolearn=no autolearn_force=no version=3.4.4
X-Google-Thread: 103376,6cdf06eb7605332d,start
X-Google-Attributes: gid103376,public
X-Google-Language: ENGLISH,ASCII-7-bit
Path: 
 g2news1.google.com!news1.google.com!news.glorb.com!border1.nntp.dca.giganews.com!border2.nntp.dca.giganews.com!nntp.giganews.com!cyclone1.gnilink.net!spamkiller2.gnilink.net!gnilink.net!trndny07.POSTED!0f19ed38!not-for-mail
From: "Frank J. Lhota" <NOSPAM.lhota.adarose@verizon.net>
Newsgroups: comp.lang.ada
Subject: Smarter Generics
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2800.1437
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441
Message-ID: <%6uTc.592$de4.1@trndny07>
Date: Sat, 14 Aug 2004 19:55:39 GMT
NNTP-Posting-Host: 141.154.229.139
X-Complaints-To: abuse@verizon.net
X-Trace: trndny07 1092513339 141.154.229.139 (Sat, 14 Aug 2004 15:55:39 EDT)
NNTP-Posting-Date: Sat, 14 Aug 2004 15:55:39 EDT
Xref: g2news1.google.com comp.lang.ada:2734
Date: 2004-08-14T19:55:39+00:00
List-Id: <comp.lang.ada>

The recent thread on "A Simple Ada Problem" got me thinking about a
limitation of Ada generics. Generic are a great way of generalizing an
implementation strategy for a particular problem, but generics cannot (or
cannot easily) select one of several implementation strategies based on the
generic parameters.

Take the following simple example: Assume that I want to implement stacks of
limited length, e.g. a stack that I can push or pop elements of a given
type, but with a limit on how many elements will be on the stack at one
time:

    generic
        type Stack_Element is private;
        Max_Length : in Positive;
    package Limited_Stacks is

        type Stack_Type is limited private;
        Stack_Overflow, Stack_Underflow : exception;

        procedure Push ( Item : in Stack_Element; Onto : in out
Stack_Type );
        procedure Pop ( Item : out Stack_Element; Onto : in out
Stack_Type );

    private
        ...
    end Limited_Stacks;

Two possible implementation strategies:

1.    We could implement the stack as an array of Max_Length elements, along
with a component indicating the current length.
2.    We could maintain an adjustable list of pointers to elements allocated
from the heap.

Obviously, implementation 1 would be faster and more convenient if
Max_Length * Stack_Element'Size is relatively small. Implementation 2
involves additional overhead, but can handle cases where stack elements can
vary quite a bit in size.

What would be great is if there were some way to write Limited_Stacks so
that the implementation strategy could be chosen based on the generic
parameters, so that we could do something like this:

    if Max_Length * Stack_Element'Size <= Some_Threshold then
        -- Static memory allocation should be fine
        Implement Limited_Stacks using strategy 1;
    else
        -- Size could be a problem, better go with pointers.
        Implement Limited_Stacks using strategy 2;
    end if;

AFAIK neither Ada generics nor C++ templates have this level of
intelligence. Would smarter generics be a worthwhile addition to Ada, and if
so, what form should it take?