How do you pass an int by reference to a function?

Calin Negru

int AnInt=6; void EditInt(int * IntPnt) { *IntPnt=7; } EditInt(&AnInt); Is this how you should proceed?

Greg Utas

That's passing as a pointer, not as a reference. As a reference (C++ only, though maybe C now has them too), it's

void EditInt(int& IntRef)
{
IntRef=7;
}

int AnInt=6;
EditInt(AnInt);

Robust Services Core | Software Techniques for Lemmings | Articles
The fox knows many things, but the hedgehog knows one big thing.

Lost User

Greg Utas wrote:

That's passing as a pointer, not as a reference.

Which is just the same, apart from semantics.

Calin Negru

I got it, thank you.

Greg Utas

Very true. A reference can even be to nullptr, and you need to check for this if writing bulletproof code:

void function(type& arg)
{
if(&arg == nullptr)
throw std::invalid_argument("function received null argument");
}

type* nasty = nullptr;
function(*nasty);

Without the check, function crashes when it gets around to using arg.

Robust Services Core | Software Techniques for Lemmings | Articles
The fox knows many things, but the hedgehog knows one big thing.

Lost User

And the same with a pointer, since there is no material difference between the two.

honey the codewitch

I would assert() something like that rather than throw. It strikes me as an error in the code itself, rather than something that should have filtered its way into runtime code, so I think assert is more appropriate. As I understand it, best practices indicate that a reference should never be deliberately null. Could be wrong here?

Check out my IoT graphics library here: https://honeythecodewitch.com/gfx And my IoT UI/User Experience library here: https://honeythecodewitch.com/uix

Greg Utas

What to do depends on the environment you're running in. I've never used assert, so I had to look up the documentation, which says that it calls abort. Most application code won't catch an invalid_argument exception, so the effect will be similar. In my open-source software, I neither throw nor assert in this situation. Instead, I generate a debug log with a traceback and return whatever signifies failure if the function has a result. That's appropriate if avoiding crashes is paramount, which is what my software aims for. You're right that a reference shouldn't be null. For many functions, a pointer argument shouldn't be null either. But applications misbehave, so you have to decide whether your function will check for bad arguments or just crash.

Robust Services Core | Software Techniques for Lemmings | Articles
The fox knows many things, but the hedgehog knows one big thing.