So, i need a reference to an object living in another namespace.

The setup is as such for the forward declaration:

namespace a
{
namespace c
{
class otherclass;
}
namespace b
{
class myclass
{
c::otherclass& other;
};
}
}

This works, but what i dont like is that if i were to lets say, change the namespace name or even location of the other class, i would need to create or rename a whole namespace there, in addition to changing any reference to it within myclass.

So, the question is, is the benefit from using a forward declaration high enough to have that extra namespace there. I would much prefer simply having to change the (currently) single reference to otherclass withing myclass if i were to change its location or namespace/class name, rather than also having to mess with the forward declaration.

Is the preprocessor/complier/whatever able to do the forward declaration "under the hood" somehow? Considering that im not writing a major 500k line project but a tiny hobby game, is the possible build time increase really that bad? If its not that bad, should i still use forward declarations when needed because its a good habit?

Lots of threads about fwd declarations lately btw.

I also feel the need to ask, if the class which has this reference is a base class, and the forward declared class internals are used in the derieved class, should i in that case include the header, because if i do not i would need to include it in the derieved class anyways, and the derieved class might not know what header the class resides in? I assume the answer for this is that i shouldnt use forward declarations in this case.<br /><br />I would have edited the original post but i dont feel like respacing it because it always gets messed up when i edit...

Take the example in my first post, and imagine that the reference is lets say the return type of an abstract virtual function.<br /><br />As such the base class (that example class) only needs a forward declaration, but the derieved class however needs the full definition because it needs to implement the virtual function.<br /><br />If i only have a forward declaration in the base class header, i will need to include the full header in the derieved class, which feels wrong.

Its just that in this case, anything derieving from the class will most likely need the definition, so wouldnt it make sense to include it in the base class header?

In this case the derieved class is told that a class named like this exists, if i dont include it in the base class header the writer of the derieved class needs to go hunting for the right header.

I think the difference with "normal" forward declarations is that in those cases the actual object is passed from outside, leading me to believe that the outside must already have the definition and doesnt need to figure out where to find it based on an incomplete declaration.

Yeah i guess i should just keep the forward declaration...

This is for a base class which i plan to derieve many classes from, and those classes have a virtual method that returns a reference to an object of type X, and i expect each to fill X with the right data, so it felt natural to have the header in the base.

But there is the possibility of the derieved class getting that data from outside so the forward declaration is likely better.