Was it only unintuitive to me?

[LB]

A long time ago I read Disch's famous article:
http://www.cplusplus.com/forum/articles/10627/

Fast-forward to today and I find out something completely unintuitive that I honestly believed didn't work:
http://ideone.com/rI2ukX

For a long time now I have been structuring my code in a way to avoid this, but it turns out it was never a problem after all. I'm going to have to go through and simplify my code now that I know this is possible (because my workarounds are needlessly overcomplicated).

[Disch]

References are not include dependencies, so forward declaring works in that case.

It only becomes an include dependency when you need to access one of its members and/or know the size of it.

[Dy7SLyTq]

you didnt know you could do that? isnt that the basis for a link list class/struct? declaring a pointer or reference to said class or struct before its finished, as long as you dont try to manipulate it?

[LB]

@Disch: but all examples about include dependencies only use pointers, so I was left to make assumptions about references.

@DTSCode: As I said to Disch, I had different ideas in my head for pointers and references regarding include dependencies.

[ne555]

From that article, point 4 The "right way" to include
> forward declare B if: A contains a B pointer or reference
> forward declare B if: one or more functions has a B object/pointer/reference as a parementer, or as a return type

There was no assumptions to made

[Computergeek01]

You should still avoid writing code like that if you can. The way it's written there it's just as useful as a void pointer.

[Disch]

The way it's written there it's just as useful as a void pointer.

? How do you figure?

void pointers are 'bad' because they're not typesafe. A pointer to an incomplete type is still typesafe, you just have to complete the type before you can dereference it.

[Computergeek01]

I was talking about it's utility, not best practices. In both cases you can't do much with either one without more data, in the case of a void pointer it needs to be cast and in this case you need someway to tell the program what 'A' actually is.

[devonrevenge]

thanks for the reference to an article I never knew I needed to read, bookmarked that, I avoided header files and therefore their problems for too long, and im not happy about coding in other languages, I dont feel like im doing anything just putting bits together.

[LB]

ne555 wrote:

From that article, point 4 The "right way" to include
> forward declare B if: A contains a B pointer or reference
> forward declare B if: one or more functions has a B object/pointer/reference as a parementer, or as a return type

There was no assumptions to made

You're right, I misread the article when I first read through it - maybe I took 'reference' not to mean an actual C++ reference.

@Computergeek01: I actually work with an SDK for DLLs where there is an incomplete type, and I use the value of pointers to it to differentiate between different open files in the program editor. It's not well documented and I had to take a guess and ask the developers about it, and my guess was right. If it had been only a void pointer I would have ignored it and would have been unable to achieve my goal.

[Computergeek01]

An SDK for DLL's? You mean the WinAPI?

[LB]

I guess I said it in a really redundant way, but what I meant is that the SDK lets the DLL interface with the software properly. And no it's not the WinAPI ;p