Member function definition syntax

[gsi]

Hi,
I looked a while at this but couldnt convice myself of the following reasons in the code.

[code=cpp]

class A{
int a;
public:
A():a(0){}
void func() {int A::*ptr = &A::a;}
void func1() {int *ptr = &a;}
//void func2() {int *ptr = &A::a;}----> compiler complains
//void func3() {int A::*ptr = &a;}----> compiler complains
};


1. The compiler expects the dereference operator after the scope resolution operator as in (A::*ptr) but complains when i try to use the same convention like (A::&A). Does this has any specifc reason or it is that the implementor liked it this way (Me confused bcos the :: operator has a higher precedence in both the cases).

2. Does the func1() and func() differ in any way bcos of the use of (A::) operator in the func() definition. I suppose that the use of (A::) is redundant as it is implicit but I am doubtful bcos the compiler complains about func2() and func3().
why is that func2() and func3() are invalid syntax.

Thanks in advance.
gsi.

[/code]

[sanctus]

Hi,
I looked a while at this but couldnt convice myself of the following reasons in the code.

[code=cpp]

class A{
int a;
public:
A():a(0){}
void func() {int A::*ptr = &A::a;}
void func1() {int *ptr = &a;}
//void func2() {int *ptr = &A::a;}----> compiler complains
//void func3() {int A::*ptr = &a;}----> compiler complains
};


1. The compiler expects the dereference operator after the scope resolution operator as in (A::*ptr) but complains when i try to use the same convention like (A::&A). Does this has any specifc reason or it is that the implementor liked it this way (Me confused bcos the :: operator has a higher precedence in both the cases).

2. Does the func1() and func() differ in any way bcos of the use of (A::) operator in the func() definition. I suppose that the use of (A::) is redundant as it is implicit but I am doubtful bcos the compiler complains about func2() and func3().
why is that func2() and func3() are invalid syntax.

Thanks in advance.
gsi.

[/code]

I'm not sure about this but shouldn't you define ptr as an element of the class before using it?

[gsi]

Hi,
ptr is local to the member function and is not a class member. The compiler allows the use of :: operator although the member variable a is visible at that point , but it expects the use of :: operator for even the local variable ptr. I am doubtful about this.

Thanks,
gsi.

[weaknessforcats]

ptr is local to the member function and is not a class member.

If ptr is local to the member function, then it does not have the scope of A::. Therefore, A::ptr would require ptr to be a class member.

The local ptr would be:

[code=cpp]
void func3() {int* ptr = &a;}
[/code]

Here you need to be sure the compiler takes the address of the correct token. You should have:

[code=cpp]
void func2() {int *ptr = &(A::a);}
[/code]

Or better, the real code should be:
[code=cpp]
void func2() {int *ptr = &this->a;}
[/code]

[gsi]

Hi,
This class compiles and runs fine.

[code=cpp]
class A{
private:
int a;
public:
A():a(1) { }
void func() {int A::*ptr = &A::a;}
};

int main() {
A a;
}
[/code]

I completely agree with you, But here ptr is not declared as a member variable of class A, why is the compiler allowing such a definition. Also I could'nt figure out a syntax to print out the ptr inside the function func() (ie the value of a). is it that this is a wrong syntax ?, but the compiler is unaware of it.

I understand that this sort of thing is wierd to do , but I saw this to be used in this IBM tutorial , I couldn't make out why point_i2 is defined such a way (prefixed with A::) even though it is a local pointer.

http://publib.boulder.ibm.com/infocenter/macxhelp/v6v81/index.jsp?topic=/com.ibm.vacpp6m.doc/language/ref/clrc14cplr129.htm

Thanks in advance ,
gsi.

[weaknessforcats]

class A{
private:
int a;
public:
A():a(1) { }
void func() {int A::*ptr = &A::a;}
};

int main() {
A a;
}

Above ptr is defined as a pointer to a member of A that is an int. That's OK. Then you initialize it with the address of a member of A that is an int. That's OK.

This is not OK:

class A{
private:
  int a;
public:
  A():a(1) { }
  void func() {int A::*ptr = &a;}
};

Here ptr is defined as a pointer to a member of A that is an int. That's OK. But initializing it with the address of a variable that may not be a member of A is NOT OK.

[gsi]

Hi,
That make's sence, Thanks for the info.

Thanks,
gsi.

[abhishek batra]

Hi,
I looked a while at this but couldnt convice myself of the following reasons in the code.

[code=cpp]

class A{
int a;
public:
A():a(0){}
void func() {int A::*ptr = &A::a;}
void func1() {int *ptr = &a;}
//void func2() {int *ptr = &A::a;}----> compiler complains
//void func3() {int A::*ptr = &a;}----> compiler complains
};


1. The compiler expects the dereference operator after the scope resolution operator as in (A::*ptr) but complains when i try to use the same convention like (A::&A). Does this has any specifc reason or it is that the implementor liked it this way (Me confused bcos the :: operator has a higher precedence in both the cases).

2. Does the func1() and func() differ in any way bcos of the use of (A::) operator in the func() definition. I suppose that the use of (A::) is redundant as it is implicit but I am doubtful bcos the compiler complains about func2() and func3().
why is that func2() and func3() are invalid syntax.

Thanks in advance.
gsi.

[/code]

Please ignore this post.