Continuations and threads (was Re: Iterators & generators)

[Thomas Wouters]

On Thu, Feb 17, 2000 at 05:36:48PM -0500, Harald Hanche-Olsen wrote:
> + Mike Fletcher <mfletch at tpresence.com>:
> | Not being a CS person, maybe I can confuse the issue :) ...
> I'm not one either, so surely I can confuse it further!

I may be too confused to tell for sure, but i have a gut feeling that i'm
getting closer and closer to some complete and utterly consistent
cluelessness of this issue.

Pauls' posting on continuations, more or less describing C's set/longjmp,
helped a lot at getting my mind wrapped around this idea, but as Tim
suggested, I'm not quite warped enough to understand the basics of it all.

(That is, I'm assuming here you all, especially Tim, speak the truth, and
this continuations business isn't some huge PSU conspiracy to scare the
bejezes out of everyone, to make sure continuations do _not_ make it into
Python ;)

> | A continuation is a suspended execution context, you get to a point
> | and say "can't/won't finish this now, I'll store the state (frame)
> | and let other execution contexts call me when I should continue".
> | When you want to resume the context, you "call" the continuation,
> | which resumes running the suspended execution context at the next
> | line after the call to suspend.
> 
> So far, this is not all that different from
> coroutines/generators/threads/whatever.  To my mind, one of the
> mindboggling things about continuations is that you can call the same
> continuation multiple times, whereas when you call a
> coroutine/generator, its program counter advances, so that the next
> time you call it, you are really calling a new continuation.

If this is so, you could see 'a continuation' as a class of object that is
defined in a running process, and when instantiated produces a python
(thread ? process ? stackframe ? what can i call it that doesn't already
have a different meaning ? :) - a python 'execution line' including local
variables, but not global ones, starting from the point defined in the
continuation-class (which is one instruction past the magic definition of
the continuation-class) where each 'instance' of the specific
continuation-class is actually executed 'in' a seperate frame object --
changes to local variables have no effect on another 'instance' of the
continuatio-class, but global ones do.

And the continuation-class stays valid regardless of what other instances of
the continuation or other parts of the program do ? What would a traceback
generated by an exception in a continuation look like, by the way ? (a lot
more questions about how it would look if you did <this> or <that> come to
mind, but i'm also thinking up possible answers to them, and i think most of
them are implementation defined anyway ;)

I'm probably looking at this all wrong -- i fail to see how this is more
abstract, more theoretical, more low-level, than simple subroutines. (Though
i must admit my idea of subroutines is probably all wrong -- i'm thinking of
C functions, including stack and all that.) Is there any chance that someone
can explain continuations either in a very pythonic sense (which would
require a large leap of imagination to start with) or at a very basic level,
like the example of coroutines in Compass that Harald posted -- i found that
*very* enlightening.

If course, someone (i.e. Tim) is going to smash all my hope by saying that
there is no simple way to explain continuations. Can't someone do a very
simple example using toy cars, road crossings, chickens, idiot drivers
and/or huge kerosine-carrying tankertrucks ? Something that might not be
much clearer but sheds some very different light on the issue, and caters to
the simple minds ? ;P

> maybe-posting-some-Scheme-code-could-obfuscate-the-issue-furhter-ly y'rs,

Probably, except that i dont know any Scheme, so it would probably fail to
confuse me (unless it has well-known keywords do something
not-very-well-known ;)

I'm really curious about continuations. I also (now) have a weird idea of
continuations, probably all wrong, that makes me wonder slightly "this is
very simple, what is the big fuss ? where is the rest hiding ?" -- except
for Tim explicitly stating that the idea I have of continuations is wrong;)

Continuous-ly y'rs,
-- 
Thomas Wouters <thomas at xs4all.net>

Hi! I'm a .signature virus! copy me into your .signature file to help me spread!