Thread: Modifying reference to a const

Will tis program work.............
Itried to compile it with g++ and it didn't work

Code:

#include<iostream.h>
int main()
{
	const int i = 10;
	int &j = i;

	cout<<"i = "<<i<<"j = "<<j;

	j = 20;

	cout<<"i = "<<i<<"j = "<<j;

	return 0;
}

What errors were you getting?

also, use <iostream>

Its in a book which says it will work and the out put is 10 10 10 20

Originally Posted by kishore

Its in a book which says it will work and the out put is 10 10 10 20

That is impossible. That would suggest that i and j occupy seperate memory space, which they do not because j is a reference to i. (an illegal one at the moment though, because j would have to be a reference to const int)

This would be completely correct if the ampersand was removed.

Isn't const more like a keyword that lets the compiler find errors such as this?

Code:

const int a = 42;
int b = 100;
if (a = b)

That's one of it's uses, anon.

In this case, the code started with a const int i (which the compiler must view as having a value that can't change), and attempted to create a non-const reference to it (which the compiler is required to reject because it would allow changing a value that can't change). Converting a constant into a non-constant is not something the compiler will allow voluntarily.

The example in the link given by Dave is a little different, as vart said. Apart from the minor fact it manipulated a pointer and dereferenced it (rather than playing with a reference), it used an explicit cast to bludgeon the compiler into submission -- so it forced the compiler to allow removal of the "const" qualifier.

If a book has claimed that the code quoted by kishore should print 10 10 10 20, then that book is badly broken. The code should not have even compiled. If a cast had been used to force the compiler into submission. For example;

Code:

#include <iostream>
int main()
{
	const int i = 10;
	int &j = (int &)i;   //  explicit conversion here

	std::cout<<"i = "<<i<<"j = "<<j;

	j = 20;

	std::cout<<"i = "<<i<<"j = "<<j;

	return 0;
}

it will technically yield undefined behaviour (which is why the conversion was disallowed in the first place, and the original code should not have compiled). However, if it was just to print 4 values, it would print 10 10 20 20.

The meaning of const, and its affect on references, was established quite well, early on. The meaning has not changed much since the ARM (Annotated Reference Manual) which was the book written by Stroustrup and Eliis to guide the (then) proposed development of the standard. About the only appreciable change I recall in the standardisation process was tightening up the requirement that a reference must be initialised so it was a reference to something.

The naming of standard headers without the .h came fairly late (about 95), but only affected those headers which were specifically part of the C++ standard. It did not affect the headers inherited from C (stdio.h, etc), but usage of those was deprecated in the C++ standard. It had no effect on proprietary headers such as windows.h.

My bet would be that the book being used by kishore was one who's author did not understand what references were about -- in particular that, unlike pointers, they cannot be reseated. In the early 90's there were a number of texts that demonstrated that sort of misunderstanding by their authors.

I threw away that book today.........

it was named "Let Us C++" by Yeshwank Kanitkar
the guy who wrote "Let Us C"
this guy uses TurboC++ and BorlandC++
he is a well known writer of books on C & C++ in INDIA


thank you guys for your valuable comments and suggestions......