On Tue, 2003-10-28 at 19:26, Dan Sugalski wrote:
> On Tue, 28 Oct 2003, Anuradha Ratnaweera wrote:
>
> > 2. If the ansewr to 0 or 1 is "no", will there be another standard in
> > Perl 6?
>
> Yes. Whatever we do for Parrot.
Thigs are getting clearer and more confusing ;-)
I am still in the trying-to figure-out-what-the-hell-is-going-on-here
phase so please bear with me about the stupid questions ;-)
Let me ask one more, and hopefully the last, question; a specific one
this time:
Say, there is a C library x (libx.{so,a}) which defines a function foo()
which takes a single argument `struct bar'. The prototype of foo() is
defined in x.h as:
foo(struct bar);
Actual definition might be a typedef insted of struct bar itself, but
all the same.
And struct bar has three members: int a, char *b and struct morebar *c.
Now, if one wants to create a Perl[56] binding to the library x, s?he
will create a module X. A script would look like:
use X;
my %bar = ();
my %morebar = ();
$morebar{key} = 10;
$morebar{value} = "ten";
$bar{a} = 10;
$bar{b} = 'bar b';
$bar{c} = \%morebar;
foo($bar);
Now, what exactly is going to happen when it comes to the foo($bar)
call? Is $bar going to be duplicated to match struct bar? Will there
be too much overhead when compared to a C implementation?
What I am looking for is something like this: The X.pm module will have
information about struct bar. Using that, the script will be compiled
so that the hash will contain values for indices a, b and c in order and
in the same format as C does, so that foo($bar) call can happen without
minimal intermediate conversion steps.
Thanks.
Anuradha
--
http://www.linux.lk/~anuradha/
