Why is the 'max' macro defined like this in C?

 #define max(a,b) \
   ({ typ开发者_StackOverfloweof (a) _a = (a); \
       typeof (b) _b = (b); \
     _a > _b ? _a : _b; })

Why not simply (a>b ? a : b)?

because otherwhise max(f(1), f(2)) would call one of the two functions twice:

f(1) > f(2) ? f(1) : f(2)

instead by "caching" the two values in _a and _b you have

({
    sometype _a = (a);
    sometype _b = (b);

    _a > _b ? _a : _b;
})

(and clearly as other have pointed out, there is the same problem with autoincrement/autodecrement)

I don't think this is supported by Visual Studio in this way. This is a compound statement. Read here does msvc have analog of gcc's ({ })

I'll add that the definition of compound statement in the gcc manual given here http://gcc.gnu.org/onlinedocs/gcc-2.95.3/gcc_4.html#SEC62 shows a code VERY similar to the one of the question for max :-)

It's getting around the major problem of things like

#define max(a,b) ((a) > (b) ? a : b)

when you call it with:

int x = max (a++, b--);

Since it's simple text substitution, that results in:

int x = ((a++) > (b--) ? a++ : b--);

which is not what you want.

By using:

#define max(a,b) ({
    typeof (a) _a = (a);\
    typeof (b) _b = (b); \
    _a > _b ? _a : _b; })

it uses temporary variables that effectively giving you:

int x = ({ int _a = a++; int _b = b--; _a > _b ? _a : _b; })

which runs the side effects only once.

But, to be honest, you should ditch that macro altogether and use an inline function, or even a non-inline function since, most of the time, the compiler can do a decent job of optimisation even without that suggestion.

Another way to determine max (at least for positive numbers)

#define MAX(a,b) sizeof(union x { char ca[a]; char cb[b];})

As a and b are accessed only once, MAX(a++,b++) gives the correct result.

If the macro is called with expression, this might lead to unexpected behaviour. Assume this:

int c = max(i++, j++);

In this case, the max is increase twice with the simpler version.