Teaching new tricks to an old dog (C++ -->Ada)

[Turamnvia Suouriviaskimatta]

I 'm following various posting in "comp.lang. ada, comp.lang.c++ ,
comp.realtime, comp.software-eng" groups regarding selection of a
programming language of C, C++ or Ada for safety critical real-time
applications. The majority of expert/people recommend Ada for safety
critical real-time applications. I've many years of experience in C/C++ (and
Delphi) but no Ada knowledge.

May I ask if it is too difficult to move from C/C++ to Ada?
What is the best way of learning Ada for a C/C++ programmer?

[Hyman Rosen]

Dr. Adrian Wrigley wrote:

Even in simple cases, the encapsulation has to be broken

Free functions that help implement an abstraction don't break
encapsulation, they are part of it.

[Hyman Rosen]

Robert A Duff wrote:

Because the "<" (or Hash) needs to know the internals of the thing.
It may be possible to order the objects through information
available from public accessors, so this is false.
In other words, at the point where you declare a type, you have to think
ahead: this type might want to live inside one of those containers, so
I'd better define the necessary operations.

False, as per above. In fact, the ordered containers can be instantiated
with order functions, so there need not be a single unique ordering for
a given type.

For example, any string type has a public way of delivering up the
characters it contains. That lets you define string containers which
hold their strings in alphabetical order, in case-insensitive order,
or even in pig-latin order if you want. None of this requires the
string implementor to anticipate such ordering requirements.

[Ioannis Vranos]

Robert A Duff wrote:

Such as... ?

I'm still not understanding your point. Are you referring to the fact
that the Java garbage collector cannot reasonably be implemented in
Java, and things like that?

Apart from these, let's talk about Pascal for example. I do not think
everything in its library can be written with the language itself.

--
Ioannis Vranos

http://www23.brinkster.com/noicys

[Hyman Rosen]

Dr. Adrian Wrigley wrote:

This means that you can't write an implementation of a template/generic
function using a map/hash without changing some/all the implementations
and interfaces up the instantiation tree (am I right?).

No, you are wrong. Simply write your comparison or hash function at the
point where you need it. Use the public accessors to get the data you
need to perform the operations.

[Robert A Duff]

Hyman Rosen <hy*****@mail.c om> writes:

Robert A Duff wrote:

Because the "<" (or Hash) needs to know the internals of the thing.

It may be possible to order the objects through information
available from public accessors, so this is false.

OK, please let me amend my statement to "...*usuall y* needs to know...".

So "this is false" should be "this is sometimes false".

In other words, at the point where you declare a type, you have to think
ahead: this type might want to live inside one of those containers, so
I'd better define the necessary operations.

False, as per above. In fact, the ordered containers can be instantiated
with order functions, so there need not be a single unique ordering for
a given type.

For example, any string type has a public way of delivering up the
characters it contains. That lets you define string containers which
hold their strings in alphabetical order, in case-insensitive order,
or even in pig-latin order if you want. None of this requires the
string implementor to anticipate such ordering requirements.

Yes, if the thing has a natural "<" function, or some public information
from which that can be garnered, then "<" is not a problem. That's true
of character strings. And of course I understand that you can sort
backwards by using a different function (as in "<" => ">" in Ada
terms).

But if you want to hide most of the data of some abstraction, and
there's no "natural" meaning of "<", then you have to think ahead, and
define some "<" operation just in case somebody wants to put the thing
into a container. (Or, of course, go back and add that in to an
existing data type when you find it's needed.)

The same is true with Hash. For a character string type, you can
declare it on the fly, because such a type exposes all the data. But
for a type that's "private" in Ada, or has protected/private members in
C++, the Hash function, like the "<" function, often needs access to the
private data. In Ada, you can use a child (which does not require
modifying the original thing). In C++, you can use a friend. Or, in
either language, you can just put Hash or "<" in the original thing.

One possibility is to have a coding convention: always declare "<" and
Hash for every data type, just in case it might be needed for one of
these containers.

- Bob

[Dr. Adrian Wrigley]

On Tue, 15 Mar 2005 00:47:37 +0200, Ioannis Vranos wrote:

struct IntervalLessCom pare
{
inline bool operator() (interval a, interval b) const
{
using namespace std;

return difftime( mktime(&a.finis h), mktime(&a.start ) ) <
difftime( mktime(&b.finis h), mktime(&b.start ) );
}
};

Doesn't this break down for times beyond the end of the epoch,
which are valid tm values, but invalid time_t values?
You've ended up mapping all post-epoch times to the same entry!

Doesn't it also break down when someone adds a microsecond
value in the tm struct for some platforms? Seems like
a reasonable thing for someone to try to do, but breaks this
code. You have to then introduce a platform specific fix.
Maybe unlikely for ctime, but there are plenty of less stable
interfaces. What if it was mm.h's mem_core struct (Linux)?
What about things with floating point components, where
different values compare as equal (+0 and -0) and
identical values (NaN) give inconsistent comparison?

In general, you may need to use "private" implementation details
of the class when defining the operator, ending up with
broken code, broken "contracts" and broken encapsulation.

This is sounding too much like a troll, and I think the points
have been fully made by Robert and others. I have learnt that
there is a major gulf between users' expectations of C++ templates
and of Ada generics. "contract model" is the difference.
--
Adrian

[Dr. Adrian Wrigley]

On Mon, 14 Mar 2005 23:46:33 +0000, Hyman Rosen wrote:

Dr. Adrian Wrigley wrote:

This means that you can't write an implementation of a template/generic
function using a map/hash without changing some/all the implementations
and interfaces up the instantiation tree (am I right?).

No, you are wrong. Simply write your comparison or hash function at the
point where you need it. Use the public accessors to get the data you
need to perform the operations.

That's the point... there aren't any public accessor functions
for opaque data types such as non-limited generic formal parameters.
--
Adrian

[Robert A Duff]

Ioannis Vranos <iv*@remove.thi s.grad.com> writes:

Robert A Duff wrote:

Such as... ?
I'm still not understanding your point. Are you referring to the fact
that the Java garbage collector cannot reasonably be implemented in
Java, and things like that?

Apart from these, let's talk about Pascal for example. I do not think
everything in its library can be written with the language itself.

Pascal (as defined by Wirth) doesn't even support separate compilation,
so there is no concept of libraries in Pascal, so nothing in its
"libraries" can be written in Pascal.

- Bob

[Ioannis Vranos]

Robert A Duff wrote:

Pascal (as defined by Wirth) doesn't even support separate compilation,
so there is no concept of libraries in Pascal, so nothing in its
"libraries" can be written in Pascal.

Actually I had turbo pascal 6.x-7.x in mind.

--
Ioannis Vranos

http://www23.brinkster.com/noicys

[Larry Kilgallen]

In article <1110843401.905 302@athnrd02>, Ioannis Vranos <iv*@remove.thi s.grad.com> writes:

Robert A Duff wrote:

Such as... ?

I'm still not understanding your point. Are you referring to the fact
that the Java garbage collector cannot reasonably be implemented in
Java, and things like that?

Apart from these, let's talk about Pascal for example. I do not think
everything in its library can be written with the language itself.

Presuming that by "Pascal" you mean something meeting the "Extended Pascal"
standard rather than the Wirth original,

not a library, but...

When DEC implemented their A1 security kernel they did so in VAX Pascal
(to support provability) and required that the whole thing be written
in Pascal rather than using a runtime library written in some other
language.

==========

They cancelled the project during Field Test when they figured out that
those in the US government who wanted to see A1 systems were entirely
different from those in the US government who actually purchase systems.

I believe, in fact, those rooting most strongly for Multilevel Security
actually run "system high" in their own shops.