Assignment of variables in if conditions

Which code snippet is more effective.

case 1:

Code:

---

char *str;
 str = (char *) malloc(15);
 if(!str)
  return -ENOMEM;

---

cases 2:

Code:

---

char *str;
if(!(str = (char *) malloc(15)))
  return -ENOMEM

---

does it effect execution time?

If so, it's negligible. I'd suggest making the decision based on your style preferences. Two things worth noting about these snippets:

* The cast may be omitted, as void *s may be converted to other object pointer types by the = operator in C.
* It's totally fine to use ! or ==0 to check whether str stores a null pointer, but some prefer using ==NULL in this scenario.

No, no difference at all when tested on MSVC compiler at max optimisation.

Here's the disassembly of your first snippet:

Code:

---

char* test1(void) {
push        ebp 
mov        ebp,esp 
sub        esp,0CCh 
push        ebx 
push        esi 
push        edi 
lea        edi,[ebp-0CCh] 
mov        ecx,33h 
mov        eax,0CCCCCCCCh 
rep stos    dword ptr es:[edi] 
        char *str;
        str = (char *)malloc(15);
mov        esi,esp 
push        0Fh 
call        dword ptr ds:[0A89114h] 
add        esp,4 
cmp        esi,esp 
call        __RTC_CheckEsp (0A8113Bh) 
mov        dword ptr [str],eax 
        if (!str)
cmp        dword ptr [str],0 
jne        test1+3Eh (0A8140Eh) 
        return str;
mov        eax,dword ptr [str] 
}
pop        edi 
pop        esi 
pop        ebx 
add        esp,0CCh 
cmp        ebp,esp 
call        __RTC_CheckEsp (0A8113Bh) 
mov        esp,ebp 
pop        ebp 
ret

---

And your second:

Code:

---

char* test2(void) {
push        ebp 
mov        ebp,esp 
sub        esp,0CCh 
push        ebx 
push        esi 
push        edi 
lea        edi,[ebp-0CCh] 
mov        ecx,33h 
mov        eax,0CCCCCCCCh 
rep stos    dword ptr es:[edi] 
        char *str;
        if (!(str = (char *)malloc(15)))
mov        esi,esp 
push        0Fh 
call        dword ptr ds:[0A89114h] 
add        esp,4 
cmp        esi,esp 
call        __RTC_CheckEsp (0A8113Bh) 
mov        dword ptr [str],eax 
cmp        dword ptr [str],0 
jne        test2+3Eh (0A8147Eh) 
                return str;
mov        eax,dword ptr [str] 
}
pop        edi 
pop        esi 
pop        ebx 
add        esp,0CCh 
cmp        ebp,esp 
call        __RTC_CheckEsp (0A8113Bh) 
mov        esp,ebp 
pop        ebp 
ret

---

The first is only 1 line longer because of the interleaved source code. Of course, I can't speak for all compilers/instruction sets, but I don't see how a compiler could mess your first example up. And as zyxwvuts said, if there is any difference on any platform it'd be negligible.


On the subject of ! or 0 or NULL.... MISRA insists that you use NULL. In fact, you're meant to use != 0 for testing variables in conditionals, lest some poor soul be confusied..... it's an insane piece of literature. Made me laugh out loud at some points.

I personally find your first snippet clearer. And I have a preference for NULL over 0.

Yes, in terms of efficiency they will almost certainly be compiled to the same code. Like smokeyangel, I find the former to be more readable. As zyxwvuts noted, you should get rid of the cast of the return value of malloc (unless you are specially writing your code to be compilable as C++ too).