Re: [Chicken-hackers] why Chicken?

[Tony Sidaway]

On 2/2/07, felix winkelmann <address@hidden> wrote:
> On 1/31/07, Tony Sidaway <address@hidden> wrote:
> >
> > On the core language, it might be nice to have constructs capable of
> > handling the most common instances of C variable length argument
> > lists.  I'm aware that there are some fundamental C implementation
> > dependencies here but if SWIG can do it (badly) it ought to be
> > possible to encapsulate the basics in a manner flexible enough to be
> > used in hand coding.
>
> But how do you want to specify the argument types? One type for
> 1-N arguments, or different sets of argument (types)?

In C, the C argument types are available at compilation time.  Perhaps
we could do that in Scheme if we're interfacing to a foreign function.

Something like this:

; This is equivalent to the C prototype 'int printf (const char *format, ...);'
(define-vararg-lambda printf int "printf" ((const c-string) format))

; This is equivalent to the C statement: 'printf "A string: %s, an
int: %d, an unsigned long: %ul\n", "Hello world", -10,
18446744073709551615UL);'
(printf "A string: %s, an int: %d, an unsigned long: %ul\n" ((const
string) "Hello world") (int -10) (unsigned-long
18446744073709551615))

In this stratagy, clarity is gained by requiring the programmer to do
a little more work.

define-vararg-lambda would create a new binding construct (printf in
this case) which expects the fixed non-variable argument list as
specefied (in this case the single argument, format) followed by an
indeterminate number of argument specifiers of the form (type expr)
where type is a foreign type and expr is a Scheme expression.  At its
crudest, the specifiers could be parsed into C as:

"(ctype)expr"

where ctype is the C equivalent of the foreign type specifier type.