Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

variadic macros

303 views
Skip to first unread message

Christof Warlich

unread,
May 22, 2008, 6:29:57 AM5/22/08
to
Hi,

is there any way to access individual elements in the body
of a variadic macro? I only found __VA_ARGS__, which always
expands to the complete list.

Thanks for any help,

Christof

Christof Warlich

unread,
May 22, 2008, 6:34:18 AM5/22/08
to
It would already help if there would be a way to have something
that expands to the number of arguments being passed to the variadic macro.


Christof Warlich schrieb:

Antoninus Twink

unread,
May 22, 2008, 6:06:36 PM5/22/08
to
On 22 May 2008 at 10:29, Christof Warlich wrote:
> is there any way to access individual elements in the body
> of a variadic macro? I only found __VA_ARGS__, which always
> expands to the complete list.

Twelve hours, and not a single language lawyer has come along to explain
that you can't. That's because your question is about C99, and the
luddite regulars in this group won't tolerate any discussion of C
standards more recent than the fall of the Berlin Wall.

Anyway, to answer your question, no there's no way in standard C...
there's not even any way using gcc extensions. I don't know about MS or
lcc extensions - maybe you can with them.

To understand why the answer is no, think about how you'd go about
finding out how many arguments were passed to your variadic function...
now ponder doing that in a macro.

The only purpose of variadic macros is to facilitate providing default
arguments for calls to variadic functions, like the usual
fprintf(stderr, __VA_ARGS__) example.

Peter Nilsson

unread,
May 22, 2008, 6:18:31 PM5/22/08
to
Christof Warlich <cwarl...@alcatel-lucent.de> wrote:
> Christof Warlich schrieb:

> > is there any way to access individual elements in
> > the body of a variadic macro? I only found
> > __VA_ARGS__, which always expands to the complete
> > list.

Not in standard C99.

> It would already help if there would be a way to
> have something that expands to the number of
> arguments being passed to the variadic macro.

There's no shortage on the wish list for standard C.
However that doesn't mean that your actual problem
doesn't have a simple solution.

Perhaps you might like to mention the real problem.
It may be solvable quite easily in standard C.

--
Peter

vipp...@gmail.com

unread,
May 23, 2008, 12:20:06 PM5/23/08
to
On May 23, 1:18 am, Peter Nilsson <ai...@acay.com.au> wrote:
> Christof Warlich <cwarl...@alcatel-lucent.de> wrote:
> > Christof Warlich schrieb:
> > > is there any way to access individual elements in
> > > the body of a variadic macro? I only found
> > > __VA_ARGS__, which always expands to the complete
> > > list.
>
> Not in standard C99.
6.10.3 p 5
> The identifier _ _VA_ARGS_ _ shall occur only in the replacement-list
> of a function-like macro that uses the ellipsis notation in the parameters.

vipp...@gmail.com

unread,
May 23, 2008, 12:21:33 PM5/23/08
to
On May 23, 7:20 pm, vipps...@gmail.com wrote:
> On May 23, 1:18 am, Peter Nilsson <ai...@acay.com.au> wrote:
>
> > Christof Warlich <cwarl...@alcatel-lucent.de> wrote:
> > > Christof Warlich schrieb:
> > > > is there any way to access individual elements in
> > > > the body of a variadic macro? I only found
> > > > __VA_ARGS__, which always expands to the complete
> > > > list.
>
> > Not in standard C99.
> 6.10.3 p 5
> > The identifier __VA_ARGS__ shall occur only in the replacement-list

> > of a function-like macro that uses the ellipsis notation in the parameters.

Whoops; I just realized you meant that accessing individual elements
in the body of a variadic macro is not possible in C99, which is true.
My apologies.

William Ahern

unread,
May 23, 2008, 3:28:20 PM5/23/08
to
Christof Warlich <cwar...@alcatel-lucent.de> wrote:
> It would already help if there would be a way to have something
> that expands to the number of arguments being passed to the variadic macro.

/*
* The PP_NARG macro evaluates to the number of arguments that have been
* passed to it.
*
* Laurent Deniau, "__VA_NARG__," 17 January 2006, <comp.std.c> (29 November 2007).
*/
#define PP_NARG(...) PP_NARG_(__VA_ARGS__,PP_RSEQ_N())
#define PP_NARG_(...) PP_ARG_N(__VA_ARGS__)

#define PP_ARG_N( \
_1, _2, _3, _4, _5, _6, _7, _8, _9,_10, \
_11,_12,_13,_14,_15,_16,_17,_18,_19,_20, \
_21,_22,_23,_24,_25,_26,_27,_28,_29,_30, \
_31,_32,_33,_34,_35,_36,_37,_38,_39,_40, \
_41,_42,_43,_44,_45,_46,_47,_48,_49,_50, \
_51,_52,_53,_54,_55,_56,_57,_58,_59,_60, \
_61,_62,_63,N,...) N

#define PP_RSEQ_N() \
63,62,61,60, \
59,58,57,56,55,54,53,52,51,50, \
49,48,47,46,45,44,43,42,41,40, \
39,38,37,36,35,34,33,32,31,30, \
29,28,27,26,25,24,23,22,21,20, \
19,18,17,16,15,14,13,12,11,10, \
9,8,7,6,5,4,3,2,1,0

William Ahern

unread,
May 23, 2008, 11:25:56 PM5/23/08
to
Christof Warlich <cwar...@alcatel-lucent.de> wrote:
> Hi,

> is there any way to access individual elements in the body
> of a variadic macro? I only found __VA_ARGS__, which always
> expands to the complete list.

So, using Laurent's solution you can do lots of interesting things. For
instance, you can overload functions or macros:

#define foreach2(p, a) \
for (typeof(&(a)[0]) p = (a); p - (a) < lengthof(a); p++)

#define foreach3(p, a, n) \
for (typeof(&(a)[0]) p = (a); p - (a) < (n); p++)

#define foreach_(N, ...) PASTE(foreach, N)(__VA_ARGS__)
#define foreach(...) foreach_(PP_NARG(__VA_ARGS__), __VA_ARGS__)

That technique has proven very useful. Another interesting trick is to peek
into __VA_ARGS__. That requires some indirection, and extending calls with
dummy arguments. Here's something along those lines which I wrote this past
weekend; very simple. The '3' is the default value, which can be overridden
by the caller merely providing a second argument, shifting the default
argument into the ether. (Visual Studio doesn't support much of C99, but
2005 supports __VA_ARGS__; but don't quote me, because I almost always use
Mingw).

/*
* Prefetching
*/
#if __GNUC__ >= 3
#define prefetch_((w), (p), (l), ...) __builtin_prefetch((p), (w), (l))
#define prefetch(...) prefetch_(0, __VA_ARGS__, 3)
#define prefetchw(...) prefetch_(1, __VA_ARGS__, 3)
#elif _MSC_VER >= 1400
#include <winnt.h>
#define prefetch_((p), (l), ...) PreFetchCacheLine((l), (p))
#define prefetch(...) prefetch_(__VA_ARGS__, 3)
#define prefetchw(...) prefetch_(__VA_ARGS__, 3)
#else
#define prefetch(...)
#define prefetchw(...)
#endif


Following below is one of my attempts to combine that with PP_NARG to get
something more generic and reuseable. The result is PP_MAP, which is at the
end. I've used it to prefix arguments in a DNS library, so callers can do
things like

dns_lookup(dnsctx, a, aaaa, mx)

which using

#define dns_rr_prefix(type) dns_rr_##type
#define dns_lookup(dnsctx, ...) \
dns_lookup((dnsctx), PP_NARG(__VA_ARGS__), (int[]) { \
PP_MAP(dns_rr_prefix, __VA_ARGS__) \
})

gets turned into

dns_lookup(dnsctx, 3, (int[]){ dns_rr_a, dns_rr_aaaa, dns_rr_mx })

Actually, I do something a little different by using static inline
functions, so I can turn types into a 64-bit unsigned integral bitmap w/
overflow (types > 62) copied into a compound literal variable-length array.
GCC -O2 can turn all of that into constants. Probably overkill for a DNS
library, but at the very least it improves the semantics of the user API
(and I always like to play around w/ loop elimination strategies).

The semantics of PP_MAP leaves something to be desired. It's more an
experiment. Laurent's PP_NARG, on the other hand, is close to perfection as
far as reusable macros go.


/*
* Select leading N items from list.
*/
#define PP_ARGV(N, ...) PASTE(PP_ARGV, N)(__VA_ARGS__)
#define PP_ARGV1(_00,...) \
_00
#define PP_ARGV2(_00,_01,...) \
_00,_01
#define PP_ARGV3(_00,_01,_02,...) \
_00,_01,_02
#define PP_ARGV4(_00,_01,_02,_03,...) \
_00,_01,_02,_03
#define PP_ARGV5(_00,_01,_02,_03,_04,...) \
_00,_01,_02,_03,_04
#define PP_ARGV6(_00,_01,_02,_03,_04,_05,...) \
_00,_01,_02,_03,_04,_05
#define PP_ARGV7(_00,_01,_02,_03,_04,_05,_06,...) \
_00,_01,_02,_03,_04,_05,_06
#define PP_ARGV8(_00,_01,_02,_03,_04,_05,_06,_07,...) \
_00,_01,_02,_03,_04,_05,_06,_07
#define PP_ARGV9(_00,_01,_02,_03,_04,_05,_06,_07,_08,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08
#define PP_ARGV10(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09
#define PP_ARGV11(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a
#define PP_ARGV12(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b
#define PP_ARGV13(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c
#define PP_ARGV14(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d
#define PP_ARGV15(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d,_0e,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d,_0e
#define PP_ARGV16(_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d,_0e,_0f,...) \
_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d,_0e,_0f

/*
* Apply function-like macro F to each item in argument list.
*/
#define PP_MAP16(F,_00,_01,_02,_03,_04,_05,_06,_07,_08,_09,_0a,_0b,_0c,_0d,_0e,_0f,...) \
F(_00),F(_01),F(_02),F(_03),F(_04),F(_05),F(_06),F(_07), \
F(_08),F(_09),F(_0a),F(_0b),F(_0c),F(_0d),F(_0e),F(_0f)

#define PP_MAP_(F,...) PP_MAP16(F,__VA_ARGS__)
#define PP_MAP(F,...) PP_ARGV(PP_NARG(__VA_ARGS__),PP_MAP_(F,__VA_ARGS__,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0))


William Ahern

unread,
May 23, 2008, 11:46:12 PM5/23/08
to
William Ahern <wil...@wilbur.25thandclement.com> wrote:
> Christof Warlich <cwar...@alcatel-lucent.de> wrote:
> > Hi,

> > is there any way to access individual elements in the body
> > of a variadic macro? I only found __VA_ARGS__, which always
> > expands to the complete list.

> So, using Laurent's solution you can do lots of interesting things. For
> instance, you can overload functions or macros:

> #define foreach2(p, a) \
> for (typeof(&(a)[0]) p = (a); p - (a) < lengthof(a); p++)

> #define foreach3(p, a, n) \
> for (typeof(&(a)[0]) p = (a); p - (a) < (n); p++)

Oops. A poor example because `typeof' is an extension, but it remains
illustrative of C99 standard compliant overloading using number of arguments
as the differentiator.

0 new messages