Initialized reference in function

Hey, if I want to have a function that can take a reference value as argument but also have that value initialized if no such value is supplied when calling the function, how would that be done?

I would like something like this (which is not working)

Code:

---

int Function(float &value = 0.0f);

int Function(float &value)
{

value = 5.0f;
return 1;

}

---

So that I can call this function in both these ways:

Code:

---

int vertex = Function();

---

Code:

---

float value;
int vertex = Function(value);

---

I would rather like to avoid function overloading. Is there a way to write the definition of the function to make it work?

You can't do that - a reference must always refer to a valid location. Overloading (one function that takes no parameter, and one that takes a float reference) is your only option [you could of course make the first function call the second with a dummy parameter].

--
Mats

Ok thanks, that makes sense I guess. Function overloading will be my option then.

Quote:

---

Originally Posted by DrSnuggles

Ok thanks, that makes sense I guess. Function overloading will be my option then.

---

Of course, the other solution would be to pass a pointer, which can be NULL initialized - but that means that you have to put &value everywhere in your code that calls the function.

Like this:

Code:

---

int Function(float *value = NULL);

int Function(float &value)
{

  if (value)
      *value = 5.0f;
  return 1;

}

int vertex = Function();
...
float value;
int vertex = Function(&value);

---

Edit: to a large extent, the solution will depend on what the meaning is of the call, and what "no parameter" means in comparison to "with parameter".


--
Mats

Aha yes, I recognize that way of doing it. I think that is what I need in this situation. Thanks!

I want to use this to get the edge on a 3d mesh that the mouse is closest to and in some cases when I call the function I also want to get where on the edge the mouse is (0.0 - 1.0).

Hmm, if I write it the way you did then I get:

error C2100: illegal indirection

Sorry, missed changing this line:

Code:

---

int Function(float &value)
// Should be:
int Function(float *value)

---

--
Mats

Ok, works fine, it's all good.

Although... I havn't tested yet but...

Simply defining:

Code:

---

float value;

---

and passing that to the funcrtion would not produce a true value in a conditional test within the function right? Wouldn't that be interpreted as 0?

Quote:

---

Originally Posted by DrSnuggles

Ok, works fine, it's all good.

Although... I havn't tested yet but...

Simply defining:

Code:

---

float value;

---

and passing that to the funcrtion would not produce a true value in a conditional test within the function right? Wouldn't that be interpreted as 0?

---

But you should be passing &value to the function, and if that's zero, then you either have a very specific embedded system or something has gone horribly wrong. Not very likely, and certainly not the case on Windows, Linux, Solaris or any other major system.

--
Mats

Right, now I understand.:) The memory adress of the value can not be zero. All fine then, thanks for the help again.

Don't be afraid of function overloading. It really is the best way to enable users to use your function in different ways. If not, users might be obliged to casts (static_cast, const_cast), and that's really what you don't want.

Quote:

---

Originally Posted by MarkZWEERS

Don't be afraid of function overloading. It really is the best way to enable users to use your function in different ways. If not, users might be obliged to casts (static_cast, const_cast), and that's really what you don't want.

---

I do agree with that - and in some cases, like one I worked on a day or two ago, overloading a function to avoid passing some "unused" variable will actually help the performance - I'm not sure exactly how much better the code I improved was, but it was traversing a tree, using a reference to a pointer which was never used in the calling code, but because the reference was there the compiler HAD to update via several indirections. The function went from 5-6th place in the profile to below 15th by removing that parameter, and just using a local variable to traverse the tree. The compiler could now hold the "current point" in a register, and saved several "update the reference" instructions. My guess is that this function is about 10x faster than the old one, and about half the size.

--
Mats

might as well make it inline too.

Quote:

---

Originally Posted by MarkZWEERS

Don't be afraid of function overloading. It really is the best way to enable users to use your function in different ways. If not, users might be obliged to casts (static_cast, const_cast), and that's really what you don't want.

---

Yeah I have lots of them in fact, but this perticular function is quite large so it will be easier to manage when changes is required to only have one function.

Quote:

---

Originally Posted by matsp

I do agree with that - and in some cases, like one I worked on a day or two ago, overloading a function to avoid passing some "unused" variable will actually help the performance - I'm not sure exactly how much better the code I improved was, but it was traversing a tree, using a reference to a pointer which was never used in the calling code, but because the reference was there the compiler HAD to update via several indirections. The function went from 5-6th place in the profile to below 15th by removing that parameter, and just using a local variable to traverse the tree. The compiler could now hold the "current point" in a register, and saved several "update the reference" instructions. My guess is that this function is about 10x faster than the old one, and about half the size.

---

Cool, something to consider.

Quote:

---

Originally Posted by robwhit

might as well make it inline too.

---

Well the actual function is a lot larger than that. What I posted was a simple example. Like I said I need to use this to get the closest edge on a 3d mesh and I would expect a function that always returns a float value of 5.0 to produce any miracles.:p

Why wouldn't this work?

Code:

---

int Function(float &value)
{
  value = 5.0f;
  return 1;
}

int Function()
{
  float value = 0.0;
  return Function(value);
}

---

Quote:

---

Originally Posted by Yarin

Why wouldn't this work?

Code:

---

int Function(float &value)
{
  value = 5.0f;
  return 1;
}

int Function()
{
  float value = 0.0;
  return Function(value);
}

---

---

Yes that would work but it would mean I would have to define "float value" everytime I call the function even if I don't use it. Most of the time I will not be using that value so I would rather like to have the option to call the function without arguments.

Edit: I misinterpreted you code... might be more clever than I thought... so you mean I have all the code in the "int Function(float &value)" function and just wraps that function with an overloaded function without arguments?

Quote:

---

Originally Posted by DrSnuggles

Well the actual function is a lot larger than that. What I posted was a simple example. Like I said I need to use this to get the closest edge on a 3d mesh and I would expect a function that always returns a float value of 5.0 to produce any miracles.:p

---

No no, implement one in terms of the other. Like this...

Quote:

---

Originally Posted by DrSnuggles

Edit: I misinterpreted you code... might be more clever than I thought... so you mean I have all the code in the "int Function(float &value)" function and just wraps that function with an overloaded function without arguments?

---

And if you have the wrapper function inline, there will be only one function call instead of two.

Of course, almost any optimising compiler would probably figure out how to inline a one-line function by itself. It's probably still a good idea to use inline, however, if only for the reader.

That's assuming that the pointer method didn't work out . . . .

Maybe you don't have an optimizing compiler.

Maybe you should get an optimising compiler.

Maybe your platform doesn't have an optimizing compiler.

I'm pretty sure that Windows does. http://cboard.cprogramming.com/showthread.php?t=98574

But I know what you mean. I'd probably put inline in myself, I was just pointing out that optimising compilers likely make the keyword redundant. (Remember register? . . . .)

Truce.

:D