Can attributes be virtual like methods in polymorphism ?

[lalebarde]

Hi all, all is in the title.

[Peter87]

no

[lalebarde]

Thanks Peter87. So, to go a bit further, what is the good paradigm please for something like this (pseudo-code) :

class Data57_1 {
	int a[2];
};
class Data57_2 {
	int a[2];
	float b[3];
};
class C57 {
public:
	virtual int show_data() =0;
protected:
	virtual X_TYPE data =0;
};
template <int n>
class D57 : public C57 {
public:
	int show_data() {
		cout << "a[0] = " << a[0] << endl;
		if (n == 2) cout << "b[0] = " << b[0] << endl;
	}
protected:
	if (n == 1) X_TYPE is Data57_1;
	if (n == 2) X_TYPE is Data57_2;
	X_TYPE data;
};

In other words, I need to have derived classes that are differentiated not only by methods, but mainly by attributes - I mean the data stored. More precisely, the data will be tables of structures containing more data in some derived classes than in others. But the methods are mostly the same.

[lalebarde]

I have reached to this :

struct Data57_1 {
	int a[2];
};
struct Data57_2 {
	int a[2];
	float b[3];
};
class C57 {
public:
	virtual int show_data() =0;
};
template <int n, class T>
class D57 : public C57 {
public:
	D57(int a0, int a1) {data.a[0] = a0; data.a[1] = a1;}
	int show_data() {
		cout << "a[0] = " << data.a[0] << endl;
		return 0;
	}
protected:
	T data;
};
template <>
class D57<2,Data57_2> : public C57 {
public:
	D57(int a0, int a1, float b0, float b1, float b2) {data.a[0] = a0; data.a[1] = a1; data.b[0] = b0; data.b[1] = b1; data.b[2] = b2;}
	int show_data() {
		cout << "a[0] = " << data.a[0] << endl;
		cout << "b[0] = " << data.b[0] << endl;
		return 0;
	}
protected:
	Data57_2 data;
};
//template <2,Data57_2> D57 : public C57;
//template D57<2,Data57_2> : public C57;
//template D57<2,Data57_2>;

int main(int argc, const char *argv[]) {
	D57<1,Data57_1> d1(11, 12);
	D57<2,Data57_2> d2(21, 22, 2.1, 2.2, 2.3);
	C57* p1 = &d1;
	C57* p2 = &d2;
	p1->show_data();
	p2->show_data();
              return 0;
}

What is your opinion ?

[coder777]

You can omit the template parameter n. It has no effect.

What about this:

class C57 {
public:
	C57(int a0, int a1) {a[0] = a0; a[1] = a1;}
	virtual int show_data() {
		cout << "a[0] = " << data.a[0] << endl;
		return 0;
	}
protected:
	int a[2];
};

class D57 : public C57 {
public:
	D57(int a0, int a1, float b0, float b1, float b2) : C57(a0,a1) {b[0] = b0; b[1] = b1; b[2] = b2;}
	int show_data() {
		C57::show_data();
		cout << "b[0] = " << b[0] << endl;
		return 0;
	}
protected:
	float b[3];
};

[lalebarde]

Thanks coder777,
In my real world, I am constrained to use the data structures Data57_1 and Data57_2, because they come from elsewhere with big tables of these structures accessed via hash tables. So the data shall? stay in the derived class to differentiate if it is Data57_1 or Data57_2. For data integrity proof, I cannot separate the data in parallel hash tables.
Sorry, I should have provided the constraints.