array of operations

array of operations

[marco]

Do you know if there is some way of bulding up arrays of functions and/or
procedures?...

Re: array of operations

[Mark Biggar]

marco wrote:
> Do you know if there is some way of bulding up arrays of functions and/or
> procedures?...

Sure, Ada allows for access types to functions and procedures so you
construct an array of those.  But there are some things to consider.
If your array is static, then a case statement is probably a better
solution.




-- 
--
Mark Biggar
mark.a.biggar@attbi.com

Re: array of operations

[Ehud Lamm]

"Mark Biggar" <mark.a.biggar@attbi.com> wrote in message
news:3D7E0EB1.2070708@attbi.com...
> marco wrote:
> > Do you know if there is some way of bulding up arrays of functions
and/or
> > procedures?...
>
> Sure, Ada allows for access types to functions and procedures so you
> construct an array of those.

Notice that all the procedures must have the same parameter profile. There
are cases where this isn't what you want.
There are also nesting issues (you can't have a top level array linking to
nested subroutines) [Note to language lawyers: The last sentence was not
meant to be a formal definition...]

So in many cases you may decide (after thinking about "access procedure" and
"access function") to use a dispatching routine that is invoked with an
"opcode" (operation code) and calls the appropriate routine.

Ehud

Re: array of operations

[Ole-Hjalmar Kristensen]

"Ehud Lamm" <mslamm@huji.ac.il> writes:

> "Mark Biggar" <mark.a.biggar@attbi.com> wrote in message
> news:3D7E0EB1.2070708@attbi.com...
> > marco wrote:
> > > Do you know if there is some way of bulding up arrays of functions
> and/or
> > > procedures?...
> >
> > Sure, Ada allows for access types to functions and procedures so you
> > construct an array of those.
> 
> Notice that all the procedures must have the same parameter profile. There
> are cases where this isn't what you want.
> There are also nesting issues (you can't have a top level array linking to
> nested subroutines) [Note to language lawyers: The last sentence was not
> meant to be a formal definition...]
> 
> So in many cases you may decide (after thinking about "access procedure" and
> "access function") to use a dispatching routine that is invoked with an
> "opcode" (operation code) and calls the appropriate routine.
> 
> Ehud

Or a pointer to an object and let the dispatching mechanism implement
the function pointer behind the scenes.

Re: array of operations

[Ehud Lamm]

"Ole-Hjalmar Kristensen" <oleh@vlinux.voxelvision.no> wrote in message
news:7vfzwh9dkl.fsf@vlinux.voxelvision.no...

> Or a pointer to an object and let the dispatching mechanism implement
> the function pointer behind the scenes.

This indeed another useful technique. However, there are some small gotchas.
You can usually use singleton objects, but in many cases you will need
access values so you have to make them "aliased."

By the way, several of these techniques are described in detail in my paper
"Building Frameworks in Ada95", proceedings Ada-Belgium'99 Seminar, Leuven,
1999.

Ehud Lamm

Re: array of operations

[Randy Brukardt]

Ehud Lamm wrote in message ...
>
>"Ole-Hjalmar Kristensen" <oleh@vlinux.voxelvision.no> wrote in message
>news:7vfzwh9dkl.fsf@vlinux.voxelvision.no...
>
>> Or a pointer to an object and let the dispatching mechanism implement
>> the function pointer behind the scenes.
>
>This indeed another useful technique. However, there are some small
gotchas.
>You can usually use singleton objects, but in many cases you will need
>access values so you have to make them "aliased."


Not true. Since it is always OK to take 'Unchecked_Access of a parameter
of a tagged type in Ada 95, you generally do not need to explicitly use
"aliased". Claw uses this extensively to provide an interface which is
mostly without access types, yet internally everything is linked
together. (Claw uses finalization to prevent dangling pointers in this
case).

There are very few 'aliased' tagged objects in the Claw sample programs
(3 in 74 samples).

                  Randy Brukardt.

Re: array of operations

[Ehud Lamm]

> Not true. Since it is always OK to take 'Unchecked_Access of a parameter
> of a tagged type in Ada 95, you generally do not need to explicitly use
> "aliased". Claw uses this extensively to provide an interface which is
> mostly without access types, yet internally everything is linked
> together. (Claw uses finalization to prevent dangling pointers in this
> case).
>


Sure. You can rephrase what I wrote, "You usually need to decalre aliased
objects, or use 'Unchecked_Access." :-)

Ehud

Re: array of operations

[Mark Biggar]

Ehud Lamm wrote:
>>Not true. Since it is always OK to take 'Unchecked_Access of a parameter
>>of a tagged type in Ada 95, you generally do not need to explicitly use
>>"aliased". Claw uses this extensively to provide an interface which is
>>mostly without access types, yet internally everything is linked
>>together. (Claw uses finalization to prevent dangling pointers in this
>>case).

> Sure. You can rephrase what I wrote, "You usually need to decalre aliased
> objects, or use 'Unchecked_Access." :-)

um, tagged objects are by definition aliased, see ARM 3.10(9).  No need
for an explisit declaration.

-- 
Mark Biggar
mark.a.biggar@attbi.com

Re: array of operations

[John McCabe]

On Thu, 12 Sep 2002 14:32:10 GMT, Mark Biggar
<mark.a.biggar@attbi.com> wrote:

>Ehud Lamm wrote:
>>>Not true. Since it is always OK to take 'Unchecked_Access of a parameter
>>>of a tagged type in Ada 95, you generally do not need to explicitly use
>>>"aliased". Claw uses this extensively to provide an interface which is
>>>mostly without access types, yet internally everything is linked
>>>together. (Claw uses finalization to prevent dangling pointers in this
>>>case).
>
>> Sure. You can rephrase what I wrote, "You usually need to decalre aliased
>> objects, or use 'Unchecked_Access." :-)
>
>um, tagged objects are by definition aliased, see ARM 3.10(9).  No need
>for an explisit declaration.

Do you mean this bit:

"and a formal parameter or generic formal object of a tagged type are
defined to be aliased"?

Note the word "generic". As far as I can see, you are mistaken.

Re: array of operations

[David C. Hoos]

Note the word "or". As far as I can see, _you_ are mistaken.

----- Original Message -----
From: "John McCabe" <john.nospam@nospamassen.nospamdemon.co.uk>
Newsgroups: comp.lang.ada
To: <comp.lang.ada@ada.eu.org>
Sent: Thursday, September 12, 2002 11:15 AM
Subject: Re: array of operations


> On Thu, 12 Sep 2002 14:32:10 GMT, Mark Biggar
> <mark.a.biggar@attbi.com> wrote:
>
> >Ehud Lamm wrote:
> >>>Not true. Since it is always OK to take 'Unchecked_Access of a
parameter
> >>>of a tagged type in Ada 95, you generally do not need to explicitly use
> >>>"aliased". Claw uses this extensively to provide an interface which is
> >>>mostly without access types, yet internally everything is linked
> >>>together. (Claw uses finalization to prevent dangling pointers in this
> >>>case).
> >
> >> Sure. You can rephrase what I wrote, "You usually need to decalre
aliased
> >> objects, or use 'Unchecked_Access." :-)
> >
> >um, tagged objects are by definition aliased, see ARM 3.10(9).  No need
> >for an explisit declaration.
>
> Do you mean this bit:
>
> "and a formal parameter or generic formal object of a tagged type are
> defined to be aliased"?
>
> Note the word "generic". As far as I can see, you are mistaken.
>
> _______________________________________________
> comp.lang.ada mailing list
> comp.lang.ada@ada.eu.org
> http://ada.eu.org/mailman/listinfo/comp.lang.ada
>

Re: array of operations

[John McCabe]

"David C. Hoos" <david.c.hoos.sr@ada95.com> wrote:

>Note the word "or". As far as I can see, _you_ are mistaken.

Bit of both by the looks of it. The phrase I responded to sounded like
it was suggesting *any* tagged object was, by defnition, aliased.
Should have made it more explicit that he meant any tagged parameter.

By the way - I really *don't* need to have messages sent as email as
well!!

>----- Original Message -----
>From: "John McCabe" <john.nospam@nospamassen.nospamdemon.co.uk>
>Newsgroups: comp.lang.ada
>To: <comp.lang.ada@ada.eu.org>
>Sent: Thursday, September 12, 2002 11:15 AM
>Subject: Re: array of operations
>
>
>> On Thu, 12 Sep 2002 14:32:10 GMT, Mark Biggar
>> <mark.a.biggar@attbi.com> wrote:
>>
>> >Ehud Lamm wrote:
>> >>>Not true. Since it is always OK to take 'Unchecked_Access of a
>parameter
>> >>>of a tagged type in Ada 95, you generally do not need to explicitly use
>> >>>"aliased". Claw uses this extensively to provide an interface which is
>> >>>mostly without access types, yet internally everything is linked
>> >>>together. (Claw uses finalization to prevent dangling pointers in this
>> >>>case).
>> >
>> >> Sure. You can rephrase what I wrote, "You usually need to decalre
>aliased
>> >> objects, or use 'Unchecked_Access." :-)
>> >
>> >um, tagged objects are by definition aliased, see ARM 3.10(9).  No need
>> >for an explisit declaration.
>>
>> Do you mean this bit:
>>
>> "and a formal parameter or generic formal object of a tagged type are
>> defined to be aliased"?
>>
>> Note the word "generic". As far as I can see, you are mistaken.
>>
>> _______________________________________________
>> comp.lang.ada mailing list
>> comp.lang.ada@ada.eu.org
>> http://ada.eu.org/mailman/listinfo/comp.lang.ada
>>
>


Best Regards
John McCabe <john@assen.demon.co.uk>

Re: array of operations

[Ted Dennison]

"marco" <pippo@libero.it> wrote in message news:<alkt6b$t73$1@e3k.asi.ansaldo.it>...
> Do you know if there is some way of bulding up arrays of functions and/or
> procedures?...

Yes, there are ways.

However, the usual answer here to a "how do I do this very C-ish thing
in Ada?" question is "Why do you want to?"

So...Why do you want to? :-)

If it is to interface to a C routine that requires that exact type,
then OK, that's a legitimate reason.

However, if you just want an array that can be populated by clients
where the client gets to specify a behavior, there are other better
ways of doing it. For instance, have you considered using tagged types
and runtime polymorphisim? You make an array of pointers to some base
tagged type's "'Class", which has a primitive routine for you to call
(let's call it "Do_It"). Then each client can derive their own version
of that tagged type and override "Do_It" with their own operation.
This is much more powerful than meer function pointers, is generally
considered better OO design, and is better supported by the language.