STL Vectors & Memory

In class *ClassA* below I have an STL Vector *vec* as a member
variable of the class.
Do I have to create a destructer, and somehow deallocate the memory
from *vec*, or is this handled automatically?

Also I want to allow a client of *ClassA* to be able to view the
elements of the vector.
Returning a reference to *vec* would be bad OO programming as i would
be returning a reference to a private member of the class.
I was thinking of using a method that would return an iterator over
the Vector i.e.

vector<int>::const_iterator ClassA::getIterator()
{ return vec.begin();
}

However i realised then that STL iterators dont have a reference to
the last element in the container, so this isnt viable. Also this is
pretty much the bad OO programming in the sense that while its a const
iterator and the Vector itself cant be modified, the elements of the
vector can be modified, i.e. once again allowing a client to modify
private data of the class.

Is there a design pattern or some construct of allowing the client to
view elements of the vector in a "read-only" fashion, while not
violating OO principles?...other than copying the entire vector

any help appreciated

pat



------------------- ClassA.h -----------------------------
#ifndef CLASSA_H
#define CLASSA_H

#include <vector>

class ClassA
{ public:
ClassA();
void addElement(int element);

private:
vector<int> vec;
};

#endif
------------------ ClassA.cpp ----------------------------
#include <vector>
using namespace std;

#include "ClassA.h"

ClassA::ClassA(){}

void ClassA::addElement(int element)
{ vec.push_back(element);
}

re: STL Vectors & Memory

In article <94740f05.0402230919.4c2ee747@posting.google.com >,
googleaddress@yahoo.co.uk (Patrick) wrote:
[color=blue]
> In class *ClassA* below I have an STL Vector *vec* as a member
> variable of the class.
> Do I have to create a destructer, and somehow deallocate the memory
> from *vec*, or is this handled automatically?[/color]

vector's destructor automatically calls the destructors of all of the
contained items, and deallocates the vector's internal buffer.
[color=blue]
>
> Also I want to allow a client of *ClassA* to be able to view the
> elements of the vector.
> Returning a reference to *vec* would be bad OO programming as i would
> be returning a reference to a private member of the class.
> I was thinking of using a method that would return an iterator over
> the Vector i.e.
>
> vector<int>::const_iterator ClassA::getIterator()
> { return vec.begin();
> }[/color]

Why not implement two functions:

vector<int>::const_iterator ClassA::begin() const
{
return vec.begin();
}

vector<int>::const_iterator ClassA::end() const
{
return vec.end();
}

re: STL Vectors & Memory

"Patrick" <googleaddress@yahoo.co.uk> wrote in message
news:94740f05.0402230919.4c2ee747@posting.google.c om...[color=blue]
> In class *ClassA* below I have an STL Vector *vec* as a member
> variable of the class.
> Do I have to create a destructer, and somehow deallocate the memory
> from *vec*, or is this handled automatically?[/color]

It's handled automatically. Have confidence in C++, if it really was the way
you are worried about then that would a total nightmare and I would give up
C++ and start programming Java.
[color=blue]
>
> Also I want to allow a client of *ClassA* to be able to view the
> elements of the vector.
> Returning a reference to *vec* would be bad OO programming as i would
> be returning a reference to a private member of the class.
> I was thinking of using a method that would return an iterator over
> the Vector i.e.
>
> vector<int>::const_iterator ClassA::getIterator()
> { return vec.begin();
> }
>
> However i realised then that STL iterators dont have a reference to
> the last element in the container, so this isnt viable.[/color]

Huh, what gave you that idea?
[color=blue]
> Also this is
> pretty much the bad OO programming in the sense that while its a const
> iterator and the Vector itself cant be modified, the elements of the
> vector can be modified, i.e. once again allowing a client to modify
> private data of the class.[/color]

The elements of the vector cannot be modifed because you are returning a
const_iterator.
[color=blue]
>
> Is there a design pattern or some construct of allowing the client to
> view elements of the vector in a "read-only" fashion, while not
> violating OO principles?...other than copying the entire vector
>[/color]

Yes its called a const_iterator. Somewhere, somehow, you obviously know this
or you wouldn't have chosen to use a const_iterator.
[color=blue]
> any help appreciated
>
> pat
>[/color]

john

re: STL Vectors & Memory

"John Harrison" <john_andronicus@hotmail.com> wrote in message
news:c1dj9o$1hkqvk$1@ID-196037.news.uni-berlin.de...[color=blue]
>
> "Patrick" <googleaddress@yahoo.co.uk> wrote in message[color=green]
> > Is there a design pattern or some construct of allowing the client to
> > view elements of the vector in a "read-only" fashion, while not
> > violating OO principles?...other than copying the entire vector
> >[/color]
>
> Yes its called a const_iterator. Somewhere, somehow, you obviously know[/color]
this[color=blue]
> or you wouldn't have chosen to use a const_iterator.[/color]

I don't know if it's the case here, but I find that many folks copy
and use code snippets without really knowing what they mean.

-Mike

re: STL Vectors & Memory

"John Harrison" <john_andronicus@hotmail.com> wrote in message news:<c1dj9o$1hkqvk$1@ID-196037.news.uni-berlin.de>...[color=blue]
> "Patrick" <googleaddress@yahoo.co.uk> wrote in message
> news:94740f05.0402230919.4c2ee747@posting.google.c om...[color=green]
> > In class *ClassA* below I have an STL Vector *vec* as a member
> > variable of the class.
> > Do I have to create a destructer, and somehow deallocate the memory
> > from *vec*, or is this handled automatically?[/color]
>
> It's handled automatically. Have confidence in C++, if it really was the way
> you are worried about then that would a total nightmare and I would give up
> C++ and start programming Java.
>[color=green]
> >
> > Also I want to allow a client of *ClassA* to be able to view the
> > elements of the vector.
> > Returning a reference to *vec* would be bad OO programming as i would
> > be returning a reference to a private member of the class.
> > I was thinking of using a method that would return an iterator over
> > the Vector i.e.
> >
> > vector<int>::const_iterator ClassA::getIterator()
> > { return vec.begin();
> > }
> >
> > However i realised then that STL iterators dont have a reference to
> > the last element in the container, so this isnt viable.[/color]
>
> Huh, what gave you that idea?
>[color=green]
> > Also this is
> > pretty much the bad OO programming in the sense that while its a const
> > iterator and the Vector itself cant be modified, the elements of the
> > vector can be modified, i.e. once again allowing a client to modify
> > private data of the class.[/color]
>
> The elements of the vector cannot be modifed because you are returning a
> const_iterator.
>[color=green]
> >
> > Is there a design pattern or some construct of allowing the client to
> > view elements of the vector in a "read-only" fashion, while not
> > violating OO principles?...other than copying the entire vector
> >[/color]
>
> Yes its called a const_iterator. Somewhere, somehow, you obviously know this
> or you wouldn't have chosen to use a const_iterator.
>[color=green]
> > any help appreciated
> >
> > pat
> >[/color]
>
> john[/color]


Thanks for the replies, im new to c++ hence my confusion.

John you seemed to be hinting that Iterators in C++ do contain a
reference to the last element in the container, is this so? I have
been searching for info on this but all the examples i have seen use
the container to explicitly get the reference to the last element.

As for the const_iterator, maybe i am wrong, but from what i read i
understood that a cont_iterator only prevented one from carrying out
operations on the container itself e.g. adding elements to the vector,
removing elements from the vector.

I understood that a const_iterator does not prevent a client from
altering objects contained withing the container.

So if the vector contained within my *ClassA* contains objects of type
*ClassB*, when i return an iterator to a client of *ClassA* it will
prevent the client from carrying out (destructive) operations on the
vector composed within *ClassA*, however it will not prevent the
client from carrying out destructive operations on the objects of type
*ClassB* that are contained within the vector. Is this not true?

pat

re: STL Vectors & Memory

"Patrick" <googleaddress@yahoo.co.uk> wrote in message
news:94740f05.0402231635.5b02644e@posting.google.c om...[color=blue]
> "John Harrison" <john_andronicus@hotmail.com> wrote in message[/color]
news:<c1dj9o$1hkqvk$1@ID-196037.news.uni-berlin.de>...[color=blue][color=green]
> > "Patrick" <googleaddress@yahoo.co.uk> wrote in message
> > news:94740f05.0402230919.4c2ee747@posting.google.c om...[color=darkred]
> > > In class *ClassA* below I have an STL Vector *vec* as a member
> > > variable of the class.
> > > Do I have to create a destructer, and somehow deallocate the memory
> > > from *vec*, or is this handled automatically?[/color]
> >
> > It's handled automatically. Have confidence in C++, if it really was the[/color][/color]
way[color=blue][color=green]
> > you are worried about then that would a total nightmare and I would give[/color][/color]
up[color=blue][color=green]
> > C++ and start programming Java.
> >[color=darkred]
> > >
> > > Also I want to allow a client of *ClassA* to be able to view the
> > > elements of the vector.
> > > Returning a reference to *vec* would be bad OO programming as i would
> > > be returning a reference to a private member of the class.
> > > I was thinking of using a method that would return an iterator over
> > > the Vector i.e.
> > >
> > > vector<int>::const_iterator ClassA::getIterator()
> > > { return vec.begin();
> > > }
> > >
> > > However i realised then that STL iterators dont have a reference to
> > > the last element in the container, so this isnt viable.[/color]
> >
> > Huh, what gave you that idea?
> >[color=darkred]
> > > Also this is
> > > pretty much the bad OO programming in the sense that while its a const
> > > iterator and the Vector itself cant be modified, the elements of the
> > > vector can be modified, i.e. once again allowing a client to modify
> > > private data of the class.[/color]
> >
> > The elements of the vector cannot be modifed because you are returning a
> > const_iterator.
> >[color=darkred]
> > >
> > > Is there a design pattern or some construct of allowing the client to
> > > view elements of the vector in a "read-only" fashion, while not
> > > violating OO principles?...other than copying the entire vector
> > >[/color]
> >
> > Yes its called a const_iterator. Somewhere, somehow, you obviously know[/color][/color]
this[color=blue][color=green]
> > or you wouldn't have chosen to use a const_iterator.
> >[color=darkred]
> > > any help appreciated
> > >
> > > pat
> > >[/color]
> >
> > john[/color]
>
>
> Thanks for the replies, im new to c++ hence my confusion.
>
> John you seemed to be hinting that Iterators in C++ do contain a
> reference to the last element in the container, is this so?[/color]

An iterator can 'point to' any element of a container, or
one past the last one.
[color=blue]
> I have
> been searching for info on this but all the examples i have seen use
> the container to explicitly get the reference to the last element.[/color]

Get this book:
www.josuttis.com/libbook
I found that it paid for itself in a single day.
[color=blue]
> As for the const_iterator, maybe i am wrong, but from what i read[/color]

May I ask what you're reading?
[color=blue]
> i
> understood that a cont_iterator only prevented one from carrying out
> operations on the container itself e.g. adding elements to the vector,
> removing elements from the vector.[/color]

No. A 'const_iterator' disallows modification of what it points to.
Similar to a pointer to const, e.g. 'const int *'
[color=blue]
>
> I understood that a const_iterator does not prevent a client from
> altering objects contained withing the container.[/color]

You understand incorrectly. :-)
[color=blue]
>
> So if the vector contained within my *ClassA* contains objects of type
> *ClassB*, when i return an iterator to a client of *ClassA* it will
> prevent the client from carrying out (destructive) operations on the
> vector[/color]

on the vector's *elements*. (That is if you use a const_iterator)
[color=blue]
>composed within *ClassA*, however it will not prevent the
> client from carrying out destructive operations on the objects of type
> *ClassB* that are contained within the vector. Is this not true?[/color]

You've got it exactly backwards. :-)

-Mike

re: STL Vectors & Memory

>[color=blue]
>
> Thanks for the replies, im new to c++ hence my confusion.
>
> John you seemed to be hinting that Iterators in C++ do contain a
> reference to the last element in the container, is this so? I have
> been searching for info on this but all the examples i have seen use
> the container to explicitly get the reference to the last element.
>[/color]

OK, I get your problem. It's true that given an iterator only, you cannot
derive iterator which points to the last element of a container. But its
equally true that given an iterator you cannot derive the iterator which
points to the first element of a containers either. And both are true of
pointers and arrays upon which iterators are modelled.

Normally what is done is to define two methods on your container

vector<int>::const_iterator ClassA::getBegin() const
{
return vec.begin();
}


vector<int>::const_iterator ClassA::getEnd() const
{
return vec.end();
}

And then have the rest of your code operate on a range of iterators, i.e.
from begin up to (but not including) end.

I'm guess that you are thinking of Java where iterators have a next method
which returns true or false. You can of course code it like that yourself,
but it would mean writing your own iterator class, and it wouldn't be the
C++ way so it would confuse.

john

re: STL Vectors & Memory

I have seen the light!

So to summarise, a const_iterator prevents a client from modifying the
elements of the collection\container right?

thanks all..