[SOLVED] C++ operator overloading

JohnGraham · 12-17-2009, 10:31 AM

I'm from a C/Java background, new-ish to C++, and I came across a problem with operator overloading (I think) that rather puzzled me.

The basic plan is simple - a class with a string that can have bits appended to it throughout its life (as well as doing other things).

Things all go dandy when I give my string at construction-time, but when I use the default constructor, I get a weird error message. The following code snippet is a boiled-down version of my problem (but still complete & compilable):

Code:

#include <string>
#include <sstream>

using namespace std;

class A
{
public:
	A(void) : msg() { }
	A(string str) : msg(str) { }

	string msg;

	friend A& operator<<(A& a, string str);
};

A& operator<<(A& a, string str)
{
	stringstream ss (a.msg);
	ss << str;
	a.msg.assign(ss.str());
	return a;
}

int main(void)
{
	//A a();  // <- gives an error...?
	A a("cake");
	a << "cake";
}

When I compile as it is above, I get no warnings. When I compile with the 'A a();' line uncommented, and the 'A a("cake")' commented, I get the following error:

Code:

a.cpp: In function ‘int main()’:
a.cpp:29: error: invalid operands of types ‘A()’ and ‘const char [5]’ to binary ‘operator<<’

What's going on here? The types involved in 'a << "cake"' are valid when I use the string-accepting constructor, but invalid when I use the default constructor.

I'd be grateful if anyone has any ideas...???

John G

P.S. I'm using g++ version 4.4.1 on Ubuntu 9.10, if it makes any difference.

tuxdev · 12-17-2009, 10:50 AM

"A a();" actually declares a function called "a" that accepts nothing and returns an A. "A a;" creates a default-constructed A called "a".

Either way, stringstreams already do exactly what you want.

JohnGraham · 12-17-2009, 11:03 AM

Quote:

Originally Posted by tuxdev

"A a();" actually declares a function called "a" that accepts nothing and returns an A. "A a;" creates a default-constructed A called "a".

Either way, stringstreams already do exactly what you want.

Thank you! I'm too used to Java's way of things, obviously!

SaintDanBert · 12-17-2009, 11:35 AM

Quote:

Originally Posted by tuxdev

"A a();" actually declares a function called "a" that accepts nothing and returns an A. "A a;" creates a default-constructed A called "a".

May I add that A a( ); creates a forward reference to the function. Any code between this statement and the actual declaration of that function may now use expressions that call that function.

Another observation is that 'a' is a terrible name for something. How might you interpret this fragment

Code:

    SpecialString  myFunc( );

in contrast to your use of AAAAaaaaa? While I understand the desire to keep the posting short and the code fragment concise, good names are always good practice and in this case might have drawn attention to the fact that your declaration might not have been as intended.

~~~ *<;-}( )( )//

JohnGraham · 12-17-2009, 12:34 PM

Quote:

May I add that A a( ); creates a forward reference to the function.

Ah, okay - I had a nagging feeling you couldn't define functions within functions (which I think is a pity, but hey

).

Quote:

Another observation is that 'a' is a terrible name for something.

Well, I'd never name a class that in real life, only for condensed examples as above. However (and it's obviously just a personal thing) for short examples I kind of like using abstract names like A, B, C for things - it keeps things abstract and about the code, so you're focusing on that, not on making an excuse-implementation. It also tells people who read it that I want to show them something about the syntax/semantics of the language, and not on said implementation.