Single = in if condition

[dinavolu1]

What happens if we use a single "=" in an if condition?

[tpb]

int a = 1, b = 2;

if (a = b)
    printf("Hello\n");

The a = b is just an assignment statement. It assigns the value of b to a.
But even assignment statements are expressions and therefore have a value.
The value is whatever was assigned to the left-hand side.
So a is assigned the value 2 and the assignment expression has the overall value of 2, which is not zero and so is interpretted in a boolean context as true. Hello.

However, if you need to do that I think it's better to be more explicit:

if ((a = b) != 0)

It makes it stand out more as an assignment plus a relational test.

[dhayden]

The GNU compilers generate a warning if they see an assignment inside an if. You can suppress the warning by adding an extra set of parentheses, which the compiler interprets as "I really meant this."

if ( a = b)   // generate s a warning
if ((a = b)) // no warning. 

[keskiverto]

The other "single =" in condition:

Base* b1 = ...
if ( Derived* d = dynamic_cast<Derived*>(b1) )
{
  // use d
}

That is initialization, not assignment, but it still returns the (new) value of 'd' to be evaluated as predicate.

[zaphraud]

The GNU compilers generate a warning if they see an assignment inside an if. You can suppress the warning by adding an extra set of parentheses, which the compiler interprets as "I really meant this."

if ( a = b)   // generate s a warning
if ((a = b)) // no warning.  

That is not the default behavior in a just-installed gcc/g++ under a just installed cygwin.

What platform are you using?

[dhayden]

That is not the default behavior in a just-installed gcc/g++ under a just installed cygwin.

Good point. I'm running cygwin with "-Wall" on the command line so I get all warnings.

[keskiverto]

"-Wall" does not give all warnings, not by a longshot. "-Wall -Wextra -pedantic" yields more, but even then there are probably some warnings that need explicit enable.

[dhayden]

That's good to know. Thanks!