Iterators & generators (RE: Real Problems with Python)

[Christian Tismer]

Howdy,

Andrew Cooke wrote:
> 
> In article <000001bf768e$48e40580$45a0143f at tim>,
>   "Tim Peters" <tim_one at email.msn.com> wrote:
> >    def traverse_post(self):
> >        for child in self.left, self.right:
> >            if child is not None:
> >                suspend child.traverse_post()
> >        suspend self.data
> 
> That finally hammered home to me just what continuations are all about.
> Does anyone have something similarly elegant that shows a coroutine?
> I've just had a look at the stackless python documentation and
> coroutines seem to be described as something coming out of a single
> detailed example.  Is there a simpler definition?  (I did look back
> through some posts on Deja, but there was nothing recent that seemed to
> explain what a coroutine actually is - sorry if I've missed something
> obvious).

What did you read, actually?
Here some pointers:
Homepage:   http://www.tismer.com/research/stackless/
IPC8 paper: http://www.tismer.com/research/stackless/spcpaper.htm
Slides:     http://www.tismer.com/research/stackless/spc-sheets.ppt
The latter gives you a bit of understanding what a continuation is.

Tim Peters about coroutines can be found here:
http://www.tismer.com/research/stackless/coroutines.tim.peters.html

More on coroutines by Sam Rushing:
http://www.nightmare.com/~rushing/copython/index.html

On Scheme, continuations, generators and coroutines:
http://www.cs.rice.edu/~dorai/t-y-scheme/

Revised Scheme 5 report:
http://www.swiss.ai.mit.edu/~jaffer/r5rs_toc.html

The following books are also highly recommended:
- Andrew W. Appel, Compiling with Continuations, Cambridge University
Press, 1992
- Daniel P. Friedman, Mitchell Wand, and Christopher T. Haynes,
Essentials of Programming Languages, MIT Press, 1993
- Christopher T. Haynes, Daniel P. Friedman, and Mitchell Wand,
Continuations and Coroutines,  Computer Languages, 11(3/4): 143-153,
1986.
- Strachey and Wadsworth, Continuations: A mathematical semantics which
can deal with full jumps.  Technical monograph PRG-11, Programming
Research Group, Oxford, 1974.

I don't think this is easy stuff at all, and you can't expect
to find a simple answer by skimming a couple of web pages.
It took me a lot of time to understand a fair part of this,
and this is also a showstopper to get this stuff to be used.

> Also, comp.lang.lisp is currently dissing continuations.  Would that be
> because the alternative is to pass the code that will process the node
> into the iterator as a (first class) function?  Obviously, in this case,
> yes, but is that true in general (are there examples where continuations
> or coroutines make something possible that really is tricky to do in
> other ways)?

An iterator is quite an easy thing, and it can be implemented
without continuations rather easily. Continuations cover problems
of another order of magnitude. It is due to too simple examples
that everybody thinks that coroutines and generators are the
only target. Continuations can express any kind of control flow,
and they can model iterators, coroutines and generators easily.
The opposite is not true!

I know this isn't enough to convince you, but at the time I have
no chance. I need to build some very good demo applications
which use continuations without exposing them to the user,
and this is admittedly difficult.

ciao - chris

-- 
Christian Tismer             :^)   <mailto:tismer at appliedbiometrics.com>
Applied Biometrics GmbH      :     Have a break! Take a ride on Python's
Düppelstr. 31                :    *Starship* http://starship.python.net
12163 Berlin                 :     PGP key -> http://wwwkeys.pgp.net
PGP Fingerprint       E182 71C7 1A9D 66E9 9D15  D3CC D4D7 93E2 1FAE F6DF
     we're tired of banana software - shipped green, ripens at home