So far I've been mostly programming in C and using allegro for my projects. A while back I decided to try OpenLayer and I really like it. So I figured if I'll start using OL, I might as well start writing in C++. So I've been reading up on it, and I came up with a question that I haven found the answer to: can I use a vector as a member in a class?

Cool, thanks! I just thought that there might be some quirks with the vector allocating memory if I put it into a class.

Kauhiz, actually, I prefer using vectors inside of classes (when array are needed, of course), since they are automatically cleaned up when your object goes out of scope. It makes debugging SEGFAULTs 10x easier

Don't listen to that. There's no innate reason why you shouldn't be putting objects in vectors or lists.

Really ?
how about speed when doing a Z sort

Well, there wasn't any mention of Z-sorting. And, he said vectors or lists and you provided a example of a list.

it was an example, pointers in lists ( and vectors) are more efficient.

It's a little more work, but it it's worth it

stuffing Objects in a list or a vector won't help with memory fragmentation.

I recently (as in, yesterday) tried stuffing my T_PAGE object into a vector. It contains the following:

struct T_PAGE
{
  BITMAP *bitmap, *stencil;
  PALETTE palette;
  char *filename;
};

If I tried to store more than one, for some inexplicable reason, pieces of every allocated BITMAP pointer were getting overwritten.

All I did was switch to storing pointers to T_PAGE objects instead of the objects themselves and my code started working.

--- Kris Asick (Gemini)
--- http://www.pixelships.com

I dunno... the only thing that could be wrong is the allocation of the object, but doing T_PAGE() was the only thing the compiler would accept.

Here's the original code. Using this code, all the data in the first T_PAGE object gets screwed up when I create the second one.

std::vector<T_PAGE> page;

try { page.push_back(T_PAGE()); }
catch (...) { CallErrorHandler(5); }
if (page[0].Make_Bitmap(option.xres,option.yres)) CallErrorHandler(5);
page[0].Set_Palette((PALETTE*)data[Default_Palette].dat);
page[0].accessed_before = 1;
clear_bitmap(page[0].bitmap);

try { page.push_back(T_PAGE()); }
catch (...) { CallErrorHandler(5); }
if (page[1].Make_Bitmap(option.xres,option.yres)) CallErrorHandler(5);
page[1].Set_Palette((PALETTE*)data[Default_Palette].dat);
clear_bitmap(page[1].bitmap);

And here's the new code that works perfectly.

std::vector<T_PAGE*> page;

try { page.push_back(new T_PAGE); }
catch (...) { CallErrorHandler(5); }
if (page[0]->Make_Bitmap(option.xres,option.yres)) CallErrorHandler(5);
page[0]->Set_Palette((PALETTE*)data[Default_Palette].dat);
page[0]->accessed_before = 1;
clear_bitmap(page[0]->bitmap);

try { page.push_back(new T_PAGE); }
catch (...) { CallErrorHandler(5); }
if (page[1]->Make_Bitmap(option.xres,option.yres)) CallErrorHandler(5);
page[1]->Set_Palette((PALETTE*)data[Default_Palette].dat);
clear_bitmap(page[1]->bitmap);

Practically identical.

--- Kris Asick (Gemini)
--- http://www.pixelships.com

Nothing I haven't done a million times before:

Trust me, this part of the code works perfectly. I spent over half an hour checking it before figuring something must be up with the vectors.

Oh yeah, I renamed "palette" to "pal" at one point. I typed out the class by hand in my first post.

--- Kris Asick (Gemini)
--- http://www.pixelships.com

X-G is right, and this means you need to implement operator= and copy constructor for T_PAGE.

Using pointers is faster. As discussed, using many single new's may fragment memory - especially if these memory chunks are allocated and freed in random order, and frequently. There's no problem if they are all allocated once and then used.