ACCU Conference (PyUK) 2005

[Michele Simionato]

The previous post was getting too long, so I decided to put my
impressions on the technical talks here. You may also want to
read what our chairman Andy Robinson has to say:
http://www.reportlab.org/~andy/accu2005/accu2005.html
There you can find the materials of the conference (slides etc.)

ACCU Conference (PyUK) 2005: a personal view (part 2)
=======================================================

I am definitely a technical person (as opposed to a marketing person)
so what impressed me the most were the presentations by Armin Rigo.

He gave two presentations: the official one, about the PyPy project,
and
a lightening talk about greenlets. Of course, I had heard something
about these topics previously, but it is pretty different to see a live
demonstration. Armin shown us really *impressive* stuff.

Basically, PyPy is already there and it is working! He explained us
the idea behind just-in-time compilation (which is actually quite
simple),
type inference at work, as well as other really impressive stuff, such
as sharing an object through an Internet connection (!)

The only problem with PyPy as it is now, is that it runs on top of
traditional Python and it is terribly slow. Nevertheless, Armin says
that it can be made fast and actually faster than current Python.
If any other guy would have made the same claim I would have been
skeptical, but Armin is the person who gave us Psyco, so I am forced to

take him seriously. And the other persons working full time on the
project have names such as Christian Tismer, Samuele Pedroni,
Holger Krekel, etc. As reported by Jacob Hallen, Guido himself has
great hopes for the project and Tim Peters would have liked to work
full time on PyPy.

The European Community founded the project for 1,300,000+ Euros in two
years. So there is money to join the project and to cover travel
expenses,
if you are an Europen citizen. If you are no EU member,
things are more complicated from the burocratical point
of view. If somebody knows how to turn around the
restrictions and to get some money for non-EU nationalities,
I am sure Jacob Hallen would be happy to know.

Jacob Hallen is in charge of the organization of the
sprints and of some of the burocratical part. He looks
like he is doing a pretty good job. Of course his firm,
the Strakt, has a very strong and direct interest in the
project. It is worth repeating that Alex Martelli worked for Strakt,
that Laura Chreighton is there as well, and that recently
they hired Samuele Pedroni of Jython fame. So, it looks
like they are really doing a LOT for Python in Europe.

On my part, I have resolved to keep a close eye on the
development in http://codespeak.net, since there is really
cool stuff in there, and really innovative ideas. It is
unfortunate that most of the stuff is still under SVN
and not easy available to the average Joe User, even if
there are plenty of things that would be useful to the average
Joe User, such as py.test (I would not talk about py.test since
it has been discussed elsewhere pretty well, but I definitely
like it a lot).

I certainly hope that part of this very substantial effort will go back

into the standard library someday.

It is interesting to notice that Armin and I were the two speakers to
use non-traditional presentation materials.

Whereas I have just shown simple HTML slides generated by a hand-cooked
Python script, he made a very effective usage of Pygame, so you
got the impression type inference was as easy as playing PacMan.
His presentation is available for download, at the reportlab link
I mentioned before, so everybody can see it. Armin also used GraphViz
to
plot the output of PyPy type-inferencer (I am not sure of the right
term
to use here, just as I am not sure of what "interpreter" and "compiler"
means anymore) and he got really nice-looking results. I am a
known evangelist of the "dot" language,
http://www.linuxdevcenter.com/pub/a/linux/2004/05/06/graphviz_dot.html,
so I very much liked the approach.

It is also fun to notice that both Armin's and my talk were
the two metacircular talks of the conference: Armin was talking about
implementing a programming language using itself as implementation
language (let me remind that PyPy stands for Python in Python) whereas
I presented a web application to demonstrate doctest which was used to
doctest itself.

Well, that's all for the moment. I have written enough for today, now
it is time to start playing with greenlets. They are extremely cool,
I don't know where I want to use them yet, but I am pretty sure I'll
figure out something eventually ;)


                 Michele Simionato