And I just realized my proposal fails to address something important.
But I'm not entirely sure it's Parrot's job to do the important thing
in question. Still, the possibility exists. Thus:
On Thu, Jul 07, 2005 at 12:02:40PM -0400, Chip Salzenberg wrote:
> .sub "foo"
> .param int beta
> .param string gamma :optional
> .param string delta :optional
> .param int optc :opt_count
> .param pmc epsilon :slurpy
This fails in the face of the named parameter calling sequence, which
can end up passing delta with a value but omitting gamma. Of course
the get_params opcode doesn't yet have a way to specify parameter
names, but it likely will have to soon.
Thus, this is probably better:
.sub "foo"
.param int beta
.param string gamma :optional(have_gamma)
.param string delta :optional(have_delta)
.param pmc epsilon :slurpy
This implies declaration of integer (pseudo-)registers named
"have_gamma" and "have_delta". If gamma was passed, then have_gamma
will be true, else false.
This would be syntactic sugar for something like:
.sub "foo"
.param int beta
.param string gamma :optional
.param int have_gamma :got_last_optional
.param string delta :optional
.param int have_delta :got_last_optional
.param pmc epsilon :slurpy
This allows for the possibility of named parameter passing, and it
also is even more robust against signature changes than :opt_count
would have been.
--
Chip Salzenberg <[EMAIL PROTECTED]>
