Re: Guile-user Digest, Vol 21, Issue 16

[Bruce Korb]

> The usual manner of making a C library available to Guile is to
> manually write a wrapper function, though the tools g-wrap and SWIG,
> which use interface descriptions, are available to generate these
> wrappers.  Here is an example:
> 
> #include <libguile.h>
> 
> SCM_DEFINE (jim_norris_puts, "jim-norris-puts", 1, 0, 0,
>             (SCM string),
>             "Put @var{string} on FILE* stdout.")
> #define FUNC_NAME s_jim_norris_puts
> {
>   SCM_VALIDATE_STRING (SCM_ARG1, string);
>   /* warning -- SCM_STRING_CHARS is deprecated.  Plus, one day the
>      chars won't be null-terminated */
>   fputs (SCM_STRING_CHARS (string), stdout);
>   /* I don't think you have to do this (IIRC Guile waits for all
>      threads to enter the SCM API before GC), but you might someday: */
>   scm_remember_upto_here_1 (string);
>   return SCM_UNSPECIFIED;
> }
> #undef FUNC_NAME
> 
> void
> init_jim_norris (void)
> {
> #ifndef SCM_MAGIC_SNARFER
> #include "jim_norris.x"
> #endif
> }
> 
> Then call guile-snarf -o jim_norris.x jim_norris.c, compile to a so,
> add a module that calls
> (load-extension "libjimnorris" "init_jim_norris")

On the other hand, I never have figured out why that is considered
preferable to something like the following.  All those weirdo
macro things are very distracting:

/*=gfunc jim_norris_puts
 *
 * what:   Put @var{string} on FILE* stdout.
 *
 * exparg: string, the string to emit
 *
 * doc:  A longer discussion about the function
=*/
SCM
pfx_scm_jim_norris_puts( SCM string )
{
  SCM_VALIDATE_STRING (SCM_ARG1, string);
  /* warning -- SCM_STRING_CHARS is deprecated.  Plus, one day the
     chars won't be null-terminated */
  fputs (SCM_STRING_CHARS (string), stdout);
  /* I don't think you have to do this (IIRC Guile waits for all
     threads to enter the SCM API before GC), but you might someday: */
  scm_remember_upto_here_1 (string);
  return SCM_UNSPECIFIED;
}
#include "jim_norris.ini"