Re: Invalid pointer values in a standard container
On Jun 29, 12:14 am, Seungbeom Kim <musip...@bawi.org> wrote:
Le Chaud Lapin wrote:
If there is any pivotal delineation of how people think about OO
programming in C++, I would say this is it, what you hint at above,
the question of whether a pointer is equivalent to the thing it points
to.
What does it have to do with object-oriented programming?
I think it is related to OOP because the topic typically arises from
pointers to polymorphic objects being in containers.
We all know that monomorphic objects generally lend themselves to:
0. default constructor
1. copy constructor
2. assignment operator
They behave much like scalars, in some sense. One can even pass them
to functions as scalars, but references might be used for efficiency.
But many prominent authors have promoted the idea that the programmer
is not truly using OOP if s/he is not using polymorphism. IOW, a
monomorphic program that is nearly-equivalent to a polymorphic
program .not. regarded as good OOP.
The programmer benefits by putting scalars and monomorphic objects
into containers, and enjoying the benefits of doing so. The thought
process is clear and regular.
Then, along comes the "But that's Not OOP" thinkers and say, "You need
to use more polymorphism in your program to be a true OOP programmer."
That's fine, until, such thinkers take a liberty, which I mentioned,
and say, "Oh, by the way...whereas before you were putting monomorphic
objects into your STL containers and not having any problems with
stale pointers...now you will be putting pointers to such objects in
the containers instead."
"It's basically the same thing." <---- This is where the problem
begins.
It is not the same thing.
Whenever I hear someone say,
"I have a map<> of unsigned int to Elephant, and...it's crashing
and..."
I ask them...
"Let us be clear...is it a map<> to Elephant or ponter-to-Elephant?"
If they respond by essentially saying,
"What difference does it make..the concept is the same...an Elephant
is an Elephant.."
I disagree.
I also disagree with all other notions that would go with this train
of thought such as:
1. "The right-hand-column contains Elephants." - It does not, it
contains pointers to Elephant.
2. "I removed elephant from RHS." - You did not, you removed pointer
to Elephant.
3. "I passed one of the Elphants on RHS to a function." - You did not,
you passed pointer to Elephant.
And finally,
4. "The container is in bad state because it contains bogus Elephants
that I deleted."
The container is in a perfectly good state. It contains elements whose
binary state is correct for that type. No irregularity has been caused
by a deficiency in the language. If anything, the programmer did
something wrong.
The notion that a pointer is equivalent to the thing that it points to
is a liberty that many programmers have taken, which I feel is
incorrect, and a great source of grief for them.
If you're implying that I have made such a notion, please be
more specific in where I did. I think I know better than confusing
a pointer and the pointee.
Oh no, I knew when I read you OP that you would never have such a
notion. If anything, you would act as referree, stimulating discussion
on the topic, which I appreciate BTW.
-Le Chaud Lapin-
--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]