Resizing arrays in C++?

[Blue Ocean]

How do I do it?

I'm doing a practice problem which includes me implementing a set of
ints as a class. I don't want to use vector because that would be
cheating. ;) I don't want to spend enough time on it to make it a
treeset, so I'm just gonna do an array set. It's gonna have a static
const int default size, and it's going to have a member private int
threshold. Once the size of the set passes the threshold, the next
call to any operation on it will resize it to double whatever its
current size is.

So the only difficulty is, how do I resize the array? I know that I
could use c's malloc and realloc if I included the libraries, but I
don't know if this is the right way to do things in C++.

Any help is much appreciated, as always.

[Default User]

Blue Ocean wrote:

How do I do it?

I'm doing a practice problem which includes me implementing a set of
ints as a class. I don't want to use vector because that would be
cheating. ;) I don't want to spend enough time on it to make it a
treeset, so I'm just gonna do an array set. It's gonna have a static
const int default size, and it's going to have a member private int
threshold. Once the size of the set passes the threshold, the next
call to any operation on it will resize it to double whatever its
current size is.

So the only difficulty is, how do I resize the array? I know that I
could use c's malloc and realloc if I included the libraries, but I
don't know if this is the right way to do things in C++.

Well, malloc() and realloc() are part of C. As you dealing with POD data
(ints) there's no real problem doing so. If you use new, then you'd have
to copy the data to the new block. So pick method and go with it. I'd
use the *alloc() family, personally if I had to do this problem.

Don't forget you need to implement the [] and = operators at the least
to make it look work an array. What to do with out of bounds access is
up to you.

Brian Rodenborn

[Rolf Magnus]

Blue Ocean wrote:

How do I do it?
Do what?
I'm doing a practice problem which includes me implementing a set of
ints as a class. I don't want to use vector because that would be
cheating. ;) I don't want to spend enough time on it to make it a
treeset, so I'm just gonna do an array set. It's gonna have a static
const int default size, and it's going to have a member private int
threshold. Once the size of the set passes the threshold, the next
call to any operation on it will resize it to double whatever its
current size is.

So the only difficulty is, how do I resize the array? I know that I
could use c's malloc and realloc if I included the libraries, but I
don't know if this is the right way to do things in C++.

Use new. If you need to resize, allocate another block with new, copy
the data over and deallocate the old block. That's what realloc usually
does, too.

[Default User]

Default User wrote:

Well, malloc() and realloc() are part of C.

Damn. Are part of C++ I meant to say.

Brian Rodenborn

[Default User]

Rolf Magnus wrote:

Use new. If you need to resize, allocate another block with new, copy
the data over and deallocate the old block. That's what realloc usually
does, too.

So why use it instead of realloc()? Besides saving coding steps, there's
a possiblity that realloc() will have some under-the-hood tricks to
speed things up.

Now, if he was creating a template version that would deal with non-POD
objects, then yes. But in this case, new only adds overhead while buying
you nothing. The allocated data is contained and managed by the array
class, it won't have the problem of incorrect deallocation that can
happen when mixing new and malloc() allocations.

Brian Rodenborn

[Juha Kettunen]

"Default User" <fi********@boe ing.com.invalid > wrote in message
news:40******** *******@boeing. com.invalid...

Rolf Magnus wrote:

Use new. If you need to resize, allocate another block with new, copy
the data over and deallocate the old block. That's what realloc usually
does, too.

So why use it instead of realloc()? Besides saving coding steps, there's
a possiblity that realloc() will have some under-the-hood tricks to
speed things up.

I remember reading from some book, that never use malloc with C++, but use
new instead. dont remember the reason. Was it even in Bjarne's book? I have
a feeling, that new is better ... but I am not going to start arguing:).

Any experts here to tell the truth?

[Aguilar, James]

"Juha Kettunen" <no******@email .com> wrote in message
news:NN******** *******@newsfe5-gui.ntli.net...

"Default User" <fi********@boe ing.com.invalid > wrote in message
news:40******** *******@boeing. com.invalid...

Rolf Magnus wrote:

Use new. If you need to resize, allocate another block with new, copy
the data over and deallocate the old block. That's what realloc usually does, too.

So why use it instead of realloc()? Besides saving coding steps, there's
a possiblity that realloc() will have some under-the-hood tricks to
speed things up.

I remember reading from some book, that never use malloc with C++, but use
new instead. dont remember the reason. Was it even in Bjarne's book? I

have a feeling, that new is better ... but I am not going to start arguing:).

I went and looked it up in Bjarne's text after you mentioned it. It does
indeed say in Ch 1's notes to C programmers not to use malloc or realloc
unless it's unavoidable.

[Julie]

"Aguilar, James" wrote:

"Juha Kettunen" <no******@email .com> wrote in message
news:NN******** *******@newsfe5-gui.ntli.net...

"Default User" <fi********@boe ing.com.invalid > wrote in message
news:40******** *******@boeing. com.invalid...

Rolf Magnus wrote:

> Use new. If you need to resize, allocate another block with new, copy
> the data over and deallocate the old block. That's what realloc usually > does, too.
So why use it instead of realloc()? Besides saving coding steps, there's
a possiblity that realloc() will have some under-the-hood tricks to
speed things up.

I remember reading from some book, that never use malloc with C++, but use
new instead. dont remember the reason. Was it even in Bjarne's book? I

have

a feeling, that new is better ... but I am not going to start arguing:).

I went and looked it up in Bjarne's text after you mentioned it. It does
indeed say in Ch 1's notes to C programmers not to use malloc or realloc
unless it's unavoidable.

Hopefully he went off to explain exactly why he made that statement.

Personally, I avoid all 'do' or 'don't do' generalizations -- they really serve
no purpose. I have found that having someone understand the various
idioms/solutions, and then let them make what they feel to be the appropriate
choice goes a lot further to creating good programmers and programming styles.

My recommendation to the OP is to: research the various ways to create and
maintain arrays, including C-style malloc, C++ new, STL containers, rolling
your own container, etc. If you don't have the time/desire to do all that
research up front, typically the best solution for arrays is using one of the
STL containers such as vector.

Remember, that all of the language constructs are there for a reason. There is
no perfect solution for all situations. Consider each possible solution on a
case-by-case basis, and when you are done, review what you have done and see if
it truly fit your needs/requirements. If not, either re-implement (refactor),
or learn from the experience and make improvements in the future.

[John Harrison]

On Sun, 11 Jul 2004 00:25:17 GMT, Juha Kettunen <no******@email .com> wrote:

"Default User" <fi********@boe ing.com.invalid > wrote in message
news:40******** *******@boeing. com.invalid...

Rolf Magnus wrote:

> Use new. If you need to resize, allocate another block with new, copy
> the data over and deallocate the old block. That's what realloc

usually

> does, too.

So why use it instead of realloc()? Besides saving coding steps, there's
a possiblity that realloc() will have some under-the-hood tricks to
speed things up.

I remember reading from some book, that never use malloc with C++, but
use
new instead. dont remember the reason. Was it even in Bjarne's book? I
have
a feeling, that new is better ... but I am not going to start arguing:).

Any experts here to tell the truth?

The reason is undoubtedly that new creates objects (i.e. constructors get
called) whereas malloc only allocates memory (no constructors get called).
However the OP is creating ints, so there is no constructor to call. There
are also situations where there is a constructor but you want to delay
calling it for some reason. As Julie says later on, always be suspicious
of 'never to this', 'always do that', try to understand the reasons for
the advice and realise that there often will be exceptions.

I agree with DefaultUser, in this case I think malloc, realloc would be
fine, unless the OP is thinking at some future point of trying to get his
class working with something more complex than ints.

john

[Juha Kettunen]

"John Harrison" <jo************ *@hotmail.com> wrote in message
news:opsaymu9ol 212331@andronic us...

On Sun, 11 Jul 2004 00:25:17 GMT, Juha Kettunen <no******@email .com> wrote: I agree with DefaultUser, in this case I think malloc, realloc would be fine, unless the OP is thinking at some future point of trying to get his
class working with something more complex than ints.

But this fact is maybe just the reason to use new at the very beginning
....so that your code is ready to all changes better. You dont need to
remember in future whether you need to change malloc to new or not ... the
code is ready for complicated modifications. I personally prefer to write
code, which is ready for the future, and you need to change as little as
possible to get it work after adding new things (of course this is not
always possible, for example you cannot code so that Windows code will be
easily to translated to Unix code for example). I am always keeping in my
mind, that the code is easily ready for the future addings. Thats why for
example, if you need to define human (and at the first version you need only
integers to define it), I will NOT define it as:

int Human[10];

but

class CHuman
{
.....
};

CHuman Human[10];
So that it is ready for the future when I want to add something more for the
Human.

Is this right?