Do you need to #undef something that's already defined?

If I have

Code:

---

#define PI 3.14159

---

To re-define PI, do I need to #undef it or can I just put a new #define in?

Someone on this board said that this was permissible:

Code:

---

#define PI 3.14159
/*#undef PI*/
#define PI 3.14159265358979

---

But my compiler complains.

Is the #undef required?

> Is the #undef required?
Yes.

Okay, thanks. That's what I thought.

Actually it seems that there are very specific situations where you can simply redefine.

Quote:

---

The benign redefinition of macros is allowed in Standard C and many other imple-
mentations. That is, a macro may be redefined if the new definition is the same, token for
token, as the existing definition. The redefinition must include whitespace in the same lo-
cations as in the origional definition, although the particular whitespace characters can be
different. We think programmers should avoid depending on benign redefinitions It is
generally better style to have a single point of definition for all program entities, including
macros. (Some older implementations of C may not allow any kind of redefinition.)
Example:

In the following definitions, the redefinition of NULL is allowed, but neither redefinition of
FUNC is valid. (The first includes whitespace not in the origional definition, and the second
changes two tokens.)

#define NULL 0
#define FUNC(x) x+4
#define NULL /* null pointer */ 0
#define FUNC(x) x + 4
#define FUNC(y) y+4

---

The above is quoted from page[s] 53/54 of "C: A Reference Manual, 5th edition." I remember reading a while back some place that any redefinition was allowed, but it seems as though that was incorrect, and there is (possibly at least) a specific usage that you can redefine it. Whatever reference I was reading implied that you could simply:

Code:

---

#define X 1
#define X(y,z) y+z
#define X 7.654

---

And all would be legal. Apparently that's incorrect, and I can't remember where I was reading it. I haven't ever had the need to do so, so I never really paid much attention to it other than the passing read.

The reference isn't clear however if the above is allowed. For example, while it says the above NULL redefinition is OK, it isn't overly clear if you could do:

Code:

---

#define FOO 0
#define FOO 1
#define FOO 2

---

I'm not sure if they mean "tokens are the same" as "the 0 is the same in both definitions" or, "the x is renamed to y, so that's different", or both.

Quzah.

According to the standard, redefinition is allowed, though with constraints (6.10.3 item 2). gcc emits a warning with the following code, but it compiles.

Code:

---

#include <stdio.h>

#define X 2
#define X 4

int
main (void)
{
  printf ("\nX equals %d\n\n", X);
  return 0;
}

---

(gcc -Wall -Wextra -o redef redef.c)

Of course, results may vary between compilers. It's no doubt wise to avoid code like this -- I'd go with the #undef personally.

Some examples from C99:

Code:

---

5 EXAMPLE 3 To illustrate the rules for redefinition and reexamination, the sequence
#define x 3
#define f(a) f(x * (a))
#undef x
#define x 2
#define g f
#define z z[0]
#define h g(~
#define m(a) a(w)
#define w 0,1
#define t(a) a
#define p() int
#define q(x) x
#define r(x,y) x ## y
#define str(x) # x
f(y+1) + f(f(z)) % t(t(g)(0) + t)(1);
g(x+(3,4)-w) | h 5) & m
(f)^m(m);
p() i[q()] = { q(1), r(2,3), r(4,), r(,5), r(,) };
char c[2][6] = { str(hello), str() };
results in
f(2 * (y+1)) + f(2 * (f(2 * (z[0])))) % f(2 * (0)) + t(1);
f(2 * (2+(3,4)-0,1)) | f(2 * (~ 5)) & f(2 * (0,1))^m(0,1);
int i[] = { 1, 23, 4, 5, };
char c[2][6] = { "hello", "" };

8 EXAMPLE 6 To demonstrate the redefinition rules, the following sequence is valid.
#define OBJ_LIKE (1-1)
#define OBJ_LIKE /* white space */ (1-1) /* other */
#define FUNC_LIKE(a) ( a )
#define FUNC_LIKE( a )( /* note the white space */ \
a /* other stuff on this line
*/ )
But the following redefinitions are invalid:
#define OBJ_LIKE (0) // different token sequence
#define OBJ_LIKE (1 - 1) // different white space
#define FUNC_LIKE(b) ( a ) // different parameter usage
#define FUNC_LIKE(b) ( b ) // different parameter spelling

---

Both the C++ standard and the 1989 C standard disallow redefinition of macros, except in cases where the redefinition is essentially the same as the previous definition (i.e. the macro has the same number of arguments, and the expansion of the macro is identical except for whitespace which will be ignored by the compiler, etc etc). I'm not sure offhand what the 1999 C standard says.

The usual technique is to do a test

Code:

---

#ifdef PI
#undef PI
#endif
#define PI 3.14159265358

---

or (in cases where we can reasonably expect all definitions to represent logically the same thing) simply avoid redefining it;

Code:

---

#ifndef PI
#define PI 3.14159265358
#endif

---

Incidentally, for manifest constants like PI, I'd probably compute them explicitly;

Code:

---

double my_double_constants(int value)
{
    static char init = 0;
    static double constants[10];
    if (!init)
    {
            constants[0] = 4.0*atan(1.0);
            constants[1] = exp(1.0);
            /*  etc  */
    }
    return (value < 0 || value > 9) ? 0.0: constants[value];
}

#define PI my_double_constants(0)
#define E my_double_constants(1)
/* etc */

---

This sort of technique computes all constants to the precision supported by your compiler/system, without unnecessary overhead of computing the values multiple times.