Programming style: use return value of function ? or pass pointer

Programming style: use return value of function ? or pass pointer as an argument ? (assuming that there is only one output)

Both should work in terms of functionality, but suppose that it is used very heavily. And also in terms of reliability or security perspective, could anyone give insight to this question ?

Thank you

PS. In case function returns a value, where is that value stored in the stack ?
Will that be overwritten when other variables are written to that original place ?

For one output I'd use return value, mostly because I like to keep arguments to a minimum.

Quote:

---

In case function returns a value, where is that value stored in the stack ?

---

It isn't. It is copied (a la the operator= function) to the variable of your choosing.

Quote:

---

Originally Posted by shong9

PS. In case function returns a value, where is that value stored in the stack ?
Will that be overwritten when other variables are written to that original place ?

---

The manner in which the value is returned is unspecified by the language standard. In practice, simple types are returned in a register. Objects which are larger than a register are returned in a more complicated way which is up to the compiler.

There is no best. It depends on your design.
Btw, you should pass using reference, not pointer.

Quote:

---

Originally Posted by Elysia

There is no best. It depends on your design.
Btw, you should pass using reference, not pointer.

---

I know this is the C++ board but in the case of C, you would have no other choice.

But since this is C++, we can get away with it. I don't see your point.

It depends on what the function does, what operations are involved, and what error handling is necessary.

A function operating on native types (any operations that do not generate errors) can probably just return instances of those types (values).

A function counting matches with a predicate may impose requirements (no errors) on the operations involved.

A function likely to fail during normal execution should probably not mutate parameters unless absolutely necessary.



Use what is appropriate for the situation.

If you flatly standardize on either method, you'll eventually wish you hadn't.

Soma

Code:

---

// returns the sum of the parameters
int add(int, int);

---

Code:

---

// returns the number of matches against the predicate in the data
int matches(const std::vector<int> &, bool (*)(int));

---

Code:

---

// returns false on failure implying an unchanged parameter
bool loadfromfile(std::map<string, string> &, std::istream &);

---

Quote:

---

Originally Posted by Elysia

But since this is C++, we can get away with it. I don't see your point.

---

Thats nice, and you may say things that I see no point to either. Nothing like some punk ass trying to flame. I was just letting him/her know. He/she may not have known thank you very much.

Now on topic: Shong9, do what makes you happy.

Quote:

---

Originally Posted by phantomotap

It depends on what the function does, what operations are involved, and what error handling is necessary.

A function operating on native types (any operations that do not generate errors) can probably just return instances of those types (values).

A function counting matches with a predicate may impose requirements (no errors) on the operations involved.

A function likely to fail during normal execution should probably not mutate parameters unless absolutely necessary.

---

That's an interesting and informative set of criteria -- thanks for posting it. As usual, the right thing to do depends on the situation, and this is only learned by experience (or being exposed to it by people who have such experience).

In my experience, good designs tend to be created by good designers, and this is a very ephemeral thing to define. It has partially to do with intelligence, a great deal to do with experience, and somewhat to do with luck. In any case, simply asking the question is a sign that you're willing to learn (painfully, in some cases) how to do it right.

Quote:

---

Originally Posted by Syscal

Thats nice, and you may say things that I see no point to either. Nothing like some punk ass trying to flame. I was just letting him/her know. He/she may not have known thank you very much.

Now on topic: Shong9, do what makes you happy.

---

Then mention that when you make your reply. Otherwise it is seen as a pointless statement.
Yet, even so, C++ programmers do not need to understand C concepts and vice versa, unless that is, they are interested in knowing both things.

Heh, when I started learning C, I was sitting there trying to make passing by reference work. I couldn't and finally asked an experienced friend and then he told me there is no such thing as pass by ref in C.