From paul-clpython at nafpik.com  Thu Aug 10 19:30:34 2006
From: paul-clpython at nafpik.com (paul cannon)
Date: Thu, 10 Aug 2006 13:30:34 -0600
Subject: [clpython-devel] coroutines
Message-ID: <20060810193034.GO8314@nafpik.com>

This project is very interesting, and once it's working on sbcl I'd like
to dig into it a bit more. Thanks for the good work.

I saw some mention of coroutines in the list archive. I spent some time
working on implementing some python-ness in lisp, and have a partially
complete code transformer that can make generators and coroutines
possible. It's sort of a state machine output; I don't know how it
compares to the method mentioned by Ben used by c#. I could make it
available if there is interest, and hopefully even complete it :)

-- 
paul


From metawilm at gmail.com  Thu Aug 10 20:19:19 2006
From: metawilm at gmail.com (Willem Broekema)
Date: Thu, 10 Aug 2006 22:19:19 +0200
Subject: [clpython-devel] coroutines
In-Reply-To: <20060810193034.GO8314@nafpik.com>
References: <20060810193034.GO8314@nafpik.com>
Message-ID: <f6bc9b490608101319t3fa4326egbce523039d360d1@mail.gmail.com>

Hey Paul, welcome to the list :-)

On 8/10/06, paul cannon <paul-clpython at nafpik.com> wrote:
> I saw some mention of coroutines in the list archive. I spent some time
> working on implementing some python-ness in lisp, and have a partially
> complete code transformer that can make generators and coroutines
> possible.

I'm certainly curious about your work. CLPython fully supports
Python's generators except for the fairly new 'yield as expression'
feature ("x = yield y"). By tracing function
'rewrite-generator-funcdef-suite you can see how a generator is
rewritten into a state machine.

Rewriting of generator's function bodies into the state machine made
me feel an assembly programmer: tagbody, goto... :-)  Most of the
rewrite rules were straightforward, but handling 'yield inside
'try/except correctly required some care.

As for getting it running on SBCL: probably the heavy use of custom
declarations of the form "(declare (pydecl ...))" and environment
inspection functions, are the hardest issues in making CLPython
portable; but I have to admit I looked only briefly at what other
implementations offered in this aspect.

- Willem