[ANNOUNCE] Garbage collection for Python
Just van Rossum
just at letterror.com
Fri Apr 14 18:06:06 EDT 2000
Michael Hudson <mwh21 at cam.ac.uk> wrote:
> To someone who might know: what hope is there for this to go
> into 1.6?
[ ... ]
> It would be, in my opinion, be a really really really good thing; for
> one thing it would help in language wars...
Fredrik Lundh wrote:
> sorry, but stackless python is the only really really really good
> thing that hasn't yet been added to the core. this is just really
> really good.
Oh how I agree... (Now lets hope the new Python GC isn't
incompatible with Stackless... Project anyone? ;-)
While were at it, here's a little Stackless news:
- Christian is redesigning the continuation module, it already has
improved quite a bit over the last two weeks.
- Even so, Chris and I agree that continuations should eventually
be completely hidden from the user, since it's so incredibly
obscure, hardly anyone will understand it. Not good for CP4E ;-)
- However, it's currently needed so microthreads and coroutines
can be largely implemented in Python itself, which in itself is
extremely cool.
- We're pretty far in fleshing out the microthreading needs;
the microthread kernel will at some point be implemented in C,
making it pretty darn efficient. (Although the current Python
version is actually not so bad itself...)
- Dunno about coroutines: it appears they are much harder to get
right, and the people on the python-coro at egroups.com list only
seem to make slow progress. (But then again, I've only been
there for a week or so, so who am I to say that...)
- Will Ware and I have worked hard on a complete rewrite of the
uthread module. It is a *vast* improvement over the old one.
It offers a very nice thread object and several handy threading
widgets. It already can do stuff threading.py can only *dream*
about, like pausing (blocking) a thread from another thread or
posting an exception from another thread (!).
- It seems our new version of uthread.py is close to being
system-thread safe, allowing each system thread to have its
own set of microthreads...
- I'm playing with a uthreaded socket wrapper which uses select
and non-blocking sockets under the hood, without the client
even knowing it. If it turns out this works well, uthreads will be
a good alternative for asyncore.py, being easier to use and all.
Just
More information about the Python-list
mailing list