If Lua in World of Warcraft supported continuations, would that mean if
I get killed, or lose some money gambling, or insult someone in chat, I
could back out of the mistake? Cool!!! I knew there must be some
practical real (virtual) world applications.
-Don
Fabien wrote:
I
know that Seaside does. It uses continuations to 'remember where it was'
when the user backs up through a series of forms. e.g., given code
similar to
this:
For that kind of thing, you should look at OCaml's replay debugger: it
uses POSIX fork(), which does copy-on-write, to put 'time bookmarks' in
the debugged program. The result is, you can run your program under
debug backward as well as forward (which is insanely cool to use, BTW).
That's a neat way to put continuations in a language that doesn't
support them.
Another, portable approach would be to write a 'clone thread stack'
function, which could be a separate library. Wrap it in a callcc() API
if you like.
As for whether it's a good idea to add real continuations, I seriously
doubt it. It would definitely sound cool, but there aren't that many
cases where it's worth it. Moreover, continuations work well in a
functional environment, but Lua encourages a rather imperative style,
and continuations become really messy when there are a lot of shared
states between k()s.
I've also been wondering whether a continuation monad would help
keeping such code maintainable, but even some PhDs have a hard time
with that stuff :)
-- Fabien.
|
