On Jun 30, 2005, at 21:55, Matt Fowles wrote:
Leo~
On 6/30/05, Leopold Toetsch <[EMAIL PROTECTED]> wrote:
Nice summary with examples:
ftp://ftp.inf.puc-rio.br/pub/docs/techreports/04_15_moura.pdf
and we still have to define semantics of parrot couroutines, e.g. WRT
argument passing. See also Dan's blogs about couroutines,
I prefer the option of repassing arguments to coroutines. It is
slightly easier to emulate not repassing with repassing then vice
versa. That and it just feels more powerful to me.
Sounds reasonable.
Given the 3 operations of the paper:
create ... newsub Px, .Coroutine, *Pfunc*
resume ... invoke Px
yield invoke Px # inside the coro
and
return returncc # terminate coro - invalidate
and the combinations of argument passing i.e. inserting appropriate
set_args/_returns and get_params/get_results, we can express all these
semantics.
I think we need a howto and tests ;-)
Matt
leo
