Pointer to pointer?

[ResidentBiscuit]

Not sure if I'm doing this right. Do I really need to create a separate pointer to point to dynamic arrays?

T* temp = new T[capacity_ * 2];
        T* tIter = &temp; //Do these need to be here?
        T* dIter = &data; //Or can I use *(temp + i) and *(data + i)?
        (for unsigned int i = 0; i < size_; i++)
        {
            *(tIter + i) = *(dIter + i);
        }

[naraku9333]

You can use the original pointers.

[o1vk4iN6]

Is this for your vector class ? I would do something a bit differently:

// to avoid default constructor
T* temp = reinterpret_cast<T*>( operator new( sizeof(T) * capacity_ * 2 ) );

for( unsigned i = 0; i < size_; ++i )
{ 
    new( &temp[i] ) T( std::move( data[i] ) );
} 

Didn't check it but the problem with using new directly like that is that it calls the default constructor which might be expensive.

[ResidentBiscuit]

It seems like new calling the default constructor is a real hassle. Is there a reason new doesn't just set aside space for n elements, and have me add elements as I choose (for arrays of course)?

And why the cast there?

[Cubbi]

Is there a reason new doesn't just set aside space for n elements, and have me add elements as I choose

That's what std::allocator does.

[ResidentBiscuit]

Well ill be damned. For some reason whenever I see that I thought it was some black magic. Time to go get that.

Is it different than placement new?

[Cubbi]

std::allocator<T>::construct() is a wrapper over placement new.

[o1vk4iN6]

And why the cast there?

operator new() is essentially malloc and returns a void*.

Also i'd read into eastl and some limitations of the std::allocator for some purposes if you plan to implement your own.

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2271.html