Copy Constructor

[LRS Kumar]

Stoustrup, in The C++ Programming Language, has the following example
in 11.7:

string g (string arg) // string passed by value using copy
constructor
{
return arg; //string returned (using copy constructor)
}

int main()
{
string s = "Newton"; //string initialized (using copy constructor)
s = g(s);
}

The first two comments are clear. I was wondering why the third
comment says that the initialization uses the copy constructor.

If that was the case, wouldn't the following code also have been
initialized using the copy constructor?


#include <iostream>

class X {
int i;
public:
X() { i=0; std::cout << "defcon\n";}
X(int n) {i=n; std::cout << "intcon\n";}
X(const X& x) {i = x.i; std::cout << "copycon\n";}
};

int main()
{
X x = 2;
}

Instead, the output is: intcon


LRSK

 

[Martijn Lievaart]

Stoustrup, in The C++ Programming Language, has the following example
in 11.7:

string g (string arg) // string passed by value using copy
constructor
{
return arg; //string returned (using copy constructor)
}

int main()
{
string s = "Newton"; //string initialized (using copy constructor)
s = g(s);
}

The first two comments are clear. I was wondering why the third
comment says that the initialization uses the copy constructor.

If that was the case, wouldn't the following code also have been
initialized using the copy constructor?


#include <iostream>

class X {
int i;
public:
X() { i=0; std::cout << "defcon\n";}
X(int n) {i=n; std::cout << "intcon\n";}
X(const X& x) {i = x.i; std::cout << "copycon\n";}
};

int main()
{
X x = 2;
}

Instead, the output is: intcon

The copy constructor is *conceptually* used here. So it must exist and be
accessible. IIRC the compiler is allowed to optimize it away though.

HTH,
M4

 

[Cy Edmunds]

Stoustrup, in The C++ Programming Language, has the following example
in 11.7:

string g (string arg) // string passed by value using copy
constructor
{
return arg; //string returned (using copy constructor)
}

int main()
{
string s = "Newton"; //string initialized (using copy constructor)
s = g(s);
}

The first two comments are clear. I was wondering why the third
comment says that the initialization uses the copy constructor.

If that was the case, wouldn't the following code also have been
initialized using the copy constructor?


#include <iostream>

class X {
int i;
public:
X() { i=0; std::cout << "defcon\n";}
X(int n) {i=n; std::cout << "intcon\n";}
X(const X& x) {i = x.i; std::cout << "copycon\n";}
};

int main()
{
X x = 2;
}

Instead, the output is: intcon


LRSK

You have misquoted the book. Nowhere does BS say that

string s = "Newton";

is an example of a copy constructor. Read it again.

 

[Victor Bazarov]

You have misquoted the book. Nowhere does BS say that

string s = "Newton";

is an example of a copy constructor. Read it again.

It's not an example of one, but a copy constructor is _used_
in such initialisation. See 8.5/15 for the explanation of
the semantics. The name for this form is "copy initialisation"
and can be found in 8.5/12.

Even if the copy-construction is optimised (as permitted),
a copy-constructor must be accessible as if no optimisation is
performed.

Victor

 

[Sumit Rajan]

You have misquoted the book. Nowhere does BS say that

string s = "Newton";

is an example of a copy constructor. Read it again.

It does appear on my copy of the special edition also. The snippet posted by
the OP is on page 284 under section 11.7 (Essential Operators). The line is:

string s = "Newton"; //string initilialized (using copy constructor)

A few lines below that BS also writes:

"Often one, but not both, of these copy operations can be optimized away."

Regards,
Sumit.

 

[jeffc]

You have misquoted the book. Nowhere does BS say that

string s = "Newton";

is an example of a copy constructor. Read it again.

He didn't quote the book as saying it was an "example" of a copy
constructor. You made that up. He said "using" copy constructor.