views:

99

answers:

1

I'm stuck with a weird VS2008 C++ issue, that looks like operator precedence is not respected.

My question is what is the output of this:

int i = 0;  
std::cout << ((i != 0) ? "Not zero " : "zero ") << ++i << std::endl;  

Normally the ++ has precedence over the <<, right? Or is the << considered like a function call giving it a higher precedence than the ++? What is the 100% correct standard answer to this?

To check, I created a new empty project (VS2008 console app), pasted only this code in the main and here are the results:

Debug|Win32: “zero 1”  
Release|Win32: “zero 1”  
Debug|x64: “zero 1”  
Release|x64: “Not zero 1”

Btw, the following example produces the exact same results:

i = 0;  
printf("%s %d\n", ((i != 0) ? "Not zero" : "zero"), ++i);  

And also changing the type of optimization in release has no effect, but disabling optimization outputs “zero 1” like other configurations.

+13  A: 

This is nothing to do with operator precidence.
You are using << which is syntactic sugar for a function call:

std::cout << ((i != 0) ? "Not zero " : "zero ") << ++i << std::endl; 

// Equivalent too:

operator<<(operator<<(operator<<(std::cout, ((i != 0) ? "Not zero " : "zero ")), ++i), std::endl);

The only rules here is that an parameter must be fully evaluated before the function is called. There are no restrictions on what order the parameters are evaluated or even if there evaluation is interleaved with calls (or even partially evaluated).

Interpretation 1:

1) ((i != 0) ? "Not zero " : "zero "))
2) ++i
3) operator<<(std::cout, (1));
4) operator<<((3), (2));
5) operator<<((4), std::endl);

Interpretation 2:

1) ++i
2) ((i != 0) ? "Not zero " : "zero "))
3) operator<<(std::cout, (2));
4) operator<<((3), (1));
5) operator<<((4), std::endl);

Interpretation 3:

1) ((i != 0) ? "Not zero " : "zero "))
2) operator<<(std::cout, (1));
3) ++i
4) operator<<((2), (3));
5) operator<<((4), std::endl);

Looking at interpretation 1 as a reference:
The rules that must be applied:

 A) (1) happens before (3)
 B) (2) happens before (4)
 C) (3) happens before (4)
 D) (4) happens before (5)
Martin York
s/Equivalent too/Equivalent to/ . And +1 :)
Billy ONeal
Good answer. :-) +1
Prasoon Saurav