[pypy-dev] This Week in PyPy 1

Michael Hudson mwh at python.net
Fri Nov 4 12:45:39 CET 2005

(this summary is also on the web at 


This is the first of what will hopefully be many summaries of what's
been going on in the world of PyPy in the last week.  First, I'd like
to make a request: help me write these things.  As is mentioned in the
page about This Week in PyPy:


as and when something worth summarizing happens, be it on IRC, on a
mailing list or off in the blogosphere, add an entry to this file:


(if you can) or email me about it (if you can't).  This week noone at
all has done anything like this, which I'll forgive because it's the
first week :) Please, please do get into the habit of doing this
though, at least if you think writing this summary isn't a complete
waste of time.

Release of PyPy 0.8.0

The biggest thing that's happened in the past week was clearly the
release of PyPy 0.8.0.  You can read the release announcement at:


This release went fairly smoothly compared to some releases, mainly
because we weren't rushing to get some feature or other into the

Import Analysis

In an effort to understand what code is used where in PyPy, Michael
Hudson wrote a tool to analyse the import structure of PyPy,
culminating in a several megabyte HTML report which you can find at:


For example, this is a list of all the modules that reference
pypy.objspace.flow.model.Constant (one of the more referenced 
names in PyPy):


Of course, this work ended up duplicating some of the things done by
tools such as pylint and pyflakes and has the potential to be useful
for projects other than PyPy, so I hope to clean it up and maybe make
it a pylint plugin soon-ish.

A RPythonC tool?

A fairly common topic of discussion on #pypy starts with people who
want to write RPython code and then use PyPy to translate it to
efficient C.  This was again the case on Monday evening (look from
about 19:30 onwards):


While "officially speaking" supporting such things is not a goal of
the PyPy project (RPython is essentially an implementation detail) the
frequency of raising of the subject means that there probably is some
interest in a "rpythonc" type tool that would compile an RPython
program.  A fairly serious problem, though, is that when the target of
compilation turns out not to be RPython, working out *why* can be
arbitrarily difficult, even for an experienced PyPyer.  For these
reasons, it seems unlikely that such a tool will be written all that
soon (at least, I'm not going to do it :).


The main discussion at the weekly pypy-sync meeting was planning for
the GËteborg sprint in December:


  <dash> if python is an orchestra, overloaded operators are
        "miscellaneous percussion"              -- from Twisted.Quotes

More information about the Pypy-dev mailing list