Skip to content
Standards-defying functional-programming macros for the C preprocessor
C Shell
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
tests
LICENSE
README.md
args.h
blank.h
consume.h
foldl.h
foldr.h
intersperse.h
list.h
map-lists.h
map.h
paren.h
separators.h
stop-lists.h
stop.h
zip-repeat.h
zip.h

README.md

I've released Libpp which provides many similar macros, but is significantly faster, actively maintained, and certainly standards-conformant. Both are limited in the number of arguments the macros can handle, but Libpp's limit can be easily changed.

Macrofun is still provided here as a demonstration of a cool hack.


Macrofun provides functional-programming macros for the C preprocessor, like mapping, folding and zipping over arguments in various forms.

Macrofun is just a demonstration; please don't use it. It depends on preprocessing behavior of GCC and Clang that may not conform to the ISO standards. Section 6.10.3.4 paragraph 2 of the C11 standard specifies:

If the name of the macro being replaced is found during this scan of the replacement list (not including the rest of the source file’s preprocessing tokens), it is not replaced. Furthermore, if any nested replacements encounter the name of the macro being replaced, it is not replaced. These nonreplaced macro name preprocessing tokens are no longer available for further replacement even if they are later (re)examined in contexts in which that macro name preprocessing token would otherwise have been replaced.

However, the meaning of "nested replacements" is ambiguous: it may only refer to immediately-nested replacements. pfultz2 wrote in an issue:

The description of the C99 preprocessor in the standard implies a DAG, so this library does not violate the rule of a "nested replacement", as such. Futhermore, David Prosser's description of the algorithm, which demonstrates the "intent of the specification" shows that gcc is conforming. So it is unfair to say that gcc is non-conforming in this regard, since most implementations interpret it as a DAG, and the reference implementation interprets it this way as well.

Of course, it could be interpreted as a hierarchical expansion, however, this is extremely rare, and could be considered non-conforming in the future. Although n3882 is a C++ proposal, its approval and addition in C++ could be a very strong persuasion to include the same text in C.

An example of using these macros:

#include <macrofun/foldl.h>         // FOLDL
#include <macrofun/intersperse.h>   // INTERSPERSE
#include <macrofun/map.h>           // MAP
#include <macrofun/map-lists.h>     // MAP_LISTS
#include <macrofun/separators.h>    // SEP_*
#include <macrofun/zip.h>           // ZIP
#include <macrofun/zip-repeat.h>    // ZIP_REPEAT

int main( void )
{
    #define PRINTD1( expr ) printf( "%s = %d\n", #expr, expr )
    #define PRINTD( ... ) MAP( PRINTD1, SEP_SEMICOLON, __VA_ARGS__ )
    PRINTD( 1 + 1,          // ==> 1 + 1 = 2
            8 / 2 * 3 );    // ==> 8 / 2 * 3 = 12
    MAP_LISTS( m, SEP_COMMA, (a,b), (c,d) )   // m(a, b), m(c, d)
    INTERSPERSE( SEP_AND, a, b, c, d )        // (a) && (b) && (c) && (d)
    FOLDL( m, acc, a, b, c )                  // m( m( m( acc, a ), b ), c )
    FOLDR( f, acc, a, b, c )                  // f( a, f( b, f( c, acc ) ) )
    // FOLDR breaks when folding with a macro expression, because the
    // preprocessor then detects and disables the recursion.
    ZIP( (a,b,c), (d,e,f,g), (h,i,j) )        // (a,d,h), (b,e,i), (c,f,j)
    ZIP_REPEAT( (a,b), (c,d), (e,f) )         // (a,b,c,d), (a,b,e,f)
}

The source files include extensive comments and example usage of each of the macros.

You can run the tests by executing tests/run.bash. These tests work by diffing the output of gcc -E for each .in.c file against the corresponding .out.c file. These out.c files contain the specific whitespacing provided by GCC 4.8. Therefore, the tests may "fail" for other preprocessors, even though they output functionally equivalent code.

Macrofun has been tested with GCC 4.8 and Clang 3.4.

Macrofun's recursion method is derived from William Swanson's map-macro.

I've played around with the possibilities provided by this library for a few months now, and I've seen it do a lot of good. Recursive macros give C programmers so much more power to manipulate their code. Things like MAP and MAP_LISTS can remove a lot of repetition in your code and your clients' code.

Test tests[] = { TEST( some_test_func ),
                 TEST( another_test_func ),
                 TEST( one_more_for_good_measure ) };
// could become, via MAP:
Test tests[] = TESTS( some_test_func,
                      another_test_func,
                      one_more_for_good_measure );

return assertions( ASSERTION( 23 > 12 ),
                   ASSERTION( "example"[ 2 ] == 'x' ),
                   ASSERTION( 8 * 2 == 16 ) );
// could become:
return assertions( 23 > 12,
                   "example"[ 2 ] == 'x',
                   8 * 2 == 16 );

Catamorphisms in the C preprocessor!

#define MAX_TWO( x, y ) ( ( x ) > ( y ) ? ( x ) : ( y ) )
#define MAX( n, ... ) FOLDL( MAX_TWO, n, __VA_ARGS__ )

#define PLUS( x, y ) ( ( x ) + ( y ) )
#define SUM( ... ) FOLDL( PLUS, 0, __VA_ARGS__ )

Treating code as data:

#define ALL( ... ) INTERSPERSE( SEP_AND, __VA_ARGS__ )
#define ANY( ... ) INTERSPERSE( SEP_OR, __VA_ARGS__ )
#define ASSERT( ... ) MAP( assert, SEP_SEMICOLON, __VA_ARGS__ )

#define INVARIANTS_EMPTY( s ) \
    s.bytes == NULL, \
    s.length == 0, \
    s.capacity == 0

#define INVARIANTS_NONEMPTY( s ) \
    s.capacity > s.length, \
    s.bytes != NULL, \
    all_bytes_up_to_length_are_nonzero( s ), \
    s.bytes[ s.length ] == '\0', \

bool string_is_valid( String const s )
{
    return ALL( INVARIANTS_EMPTY( s ) )
        || ALL( INVARIANTS_NONEMPTY( s ) );
}

// We want fine-grained assertion errors, so we can't just do
// `assert( string_is_valid( s ) )`.
void string_assert_valid( String const s )
{
    if ( ANY( INVARIANTS_EMPTY( s ) ) ) {
        ASSERT( INVARIANTS_EMPTY( s ) );
    } else {
        ASSERT( INVARIANTS_NONEMPTY( s ) );
    }
}

MAP_LISTS makes it much easier to define generic functions for a set of types.

#define TYPE_SUFFIXES \
    ( bool, bools ), \
    ( char, chars ), \
    ( signed char, schars ), \
    ( unsigned char, uchars ), \
    ( short, shorts ), \
    /* 20 more lines snipped */

enum Ordering {
    Ordering_LT = -1,
    Ordering_EQ = 0,
    Ordering_GT = 1
};

#define COMPARE_PROTOTYPE( type, suffix ) \
    enum Ordering compare_##suffix( type, type );
MAP_LISTS( COMPARE_PROTOTYPE, SEP_NONE, COMPARABLE_TYPE_SUFFIXES )
// enum Ordering compare_bools( bool, bool );
// enum Ordering compare_chars( char, char );
// ...

#define COMPARE_IMPL( type, suffix ) \
    enum Ordering compare_##suffix( type const x, type const y ) \
    { \
        return ( x > y ) - ( x < y ); \
    }
MAP_LISTS( COMPARE_IMPL, SEP_NONE, COMPARABLE_TYPE_SUFFIXES )

#define COMPARE_TYPE_HANDLER( type, suffix ) \
    type:       compare_##suffix, \
    type const: compare_##suffix

#define compare( x, y ) \
    _Generic( x, \
        MAP_LISTS( COMPARE_TYPE_HANDLER, \
                   SEP_COMMA, \
                   COMPARABLE_TYPE_SUFFIXES ), \
        default: compare_pointers )( x, y )
You can’t perform that action at this time.