From: "Berger, Daniel" <Daniel.Berger@...>
Date: 2005-06-01T01:45:54+09:00
Subject: Re: Passing a block to rb_funcall

> -----Original Message-----
> From: nobu.nokada@softhome.net [mailto:nobu.nokada@softhome.net] 
> Sent: Tuesday, May 31, 2005 9:58 AM
> To: ruby-talk ML
> Subject: Re: Passing a block to rb_funcall
> 
> 
> Hi,
> 
> At Wed, 1 Jun 2005 00:35:22 +0900,
> Daniel Berger wrote in [ruby-talk:144097]:
> > I've seen some mention of rb_iterate, but I wasn't sure how 
> to apply 
> > it here.
> 
> static VALUE
> call_foreach(VALUE args)
> {
>     return rb_funcall2(rb_cIO, rb_intern("foreach"), 2, 
> (VALUE *)args); }
> 
> static VALUE
> yield_block(VALUE val, VALUE block)
> {
>     return rb_funcall2(block, rb_intern("call"), 1, &val);
> }
> 
> static VALUE
> my_method(int argc, VALUE* argv, VALUE self)
> {
>     VALUE args[2], block;
> 
>     rb_scan_args(argc, argv, "01&", &args[1], &block);
>     args[0] = self;
> 
>     return rb_iterate(call_foreach, (VALUE)args, yield_block, 
> block); }
> 
> -- 
> Nobu Nakada

Oops, just wanted to point out (for future googlers) that I'll have to
tweak this because if no value is passed for the separator, then args[1]
becomes nil.  Explicitly passing nil to IO.foreach puts it in paragraph
mode, i.e. this code is the equivalent of IO.foreach(path,nil){ ... } if
no path separator is specified.

I believe this is, yet again, the result of manual argc counting in the
underlying C code.  It looks like:

/* rb_io_s_foreach */
if (argc == 1) {
   arg.sep = rb_default_rs;
}

Should instead be:

if(NIL_P(arg.sep))
   arg.sep = rb_default_rs;

That is, unless we *want* nil to mean paragraph mode.  I was always
under the impression that only "" (an empty string) would force
paragraph mode.

Regards,

Dan