Of all the fever-induced hallucinatory things I said at PyCon '07, I'm proudest of this: "I don't do test-driven development; I do stupidity-driven testing. When I do something stupid, I write a test to make sure I don't do it again."
For readers that don't get it, my development practice is this:
- write code to solve some problem
- watch code break in some obvious way
- write a test that tests that specific breakage
- lather, rinse, repeat.
I don't mind making mistakes, even stupid ones: I just don't want to repeat them. Thus, this development technique.
General comments on PyCon:
I did something every day: tutorial, panel moderator, panel-ee, and speaker. Way too much work, stress, and time spent in preparation.
Meeting people was great. I got to meet & hang out with the ARINC people, Shannon ("jj") Behrens, James Taylor of Galaxy, Terry Peppers, and a bunch of other people that I knew only from e-mail. I also saw a bunch of faces from last year, of course, including Brian Dorsey (thanks for the lift, Brian & Kirk!)
The talks were (from my limited vantage point) much better this year than last year. This is presumably a reflection of the increasing size of the Python community.
The "nose vs py.test" debate is growing in size, if not reaching any actual conclusion. It's very clear to me, at least, that these are the big new testing tools; I'm (obviously) pushing for nose, but I'd really like to see a showdown of features so that I can convert this from flame-boy advocacy into informed advocacy.
I regret not attending more Birds-of-a-Feather sessions.
The keynotes were fantastic, in general. I didn't enjoy the education one, per se, but a lot of interesting stuff was said.
I hope the recordings are up soon.
PyCon: Day -1 (travel)
PyCon: Day 0 (tutorial)
The tutorial day was, as usual, fun! Grig and I gave our testing tutorial and even though we felt less prepared than last year, I think our significantly increased experience with actually using these tools (see the ARINC talk in particular) showed.
Reviews (both positive) by Shannon -jj Behrens and Terry Peppers.
Next year, we should have a book or two out on these topics, which will be an entertaining addition.
PyCon: Day 1 (Web panel)
I spent most of this day sweating about the PyCon Web Panel, which in any event turned out fine. Once I finally worked out the format for the panel in my own head (2 minutes introductions by me, followed by questions spread evenly among the participants) I was much more relaxed about things. (Perhaps the most fun I had with this aspect was reciprocating Grig's prodding: he constantly told me that I was over-preparing, and then when it was his turn for the Testing Tools panel I got to prod him for over-preparing. Back atcha... ;)
The panel was really meant to showcase personalities and get faces out there; 45 minutes is way too short for any meaningful discussion.
Being in front of that many people made me really freakin' nervous.
One obvious (to me) conclusion from the panel was that TurboGears and Pylons should merge. This may happen eventually, but not right now ;).
Another obvious conclusion (and I actually said something to this effect) was that documentation is a huge problem. Huge. The framework that documents will dominate IMO. (Right now I'm guessing that this will be Django, but only because Adrian consistently acknowledged the need for documentation.)
It was interesting to discover that Twisted had AJAX-like behavior a year or two before AJAX hit. I think Zope and Twisted both need to hire a PR expert to publicize their coolness; I get the impression that the communities are relatively insular and this contributes to a lack of buzz about their accomplishments.
My favorite comment, by Jonathan Ellis: "Django's ORM is feeble."
James Bennet has a disturbingly complete transcript.
Other reviews/notes: Jonathan Ellis, James Tauber (international man of mystery!), Nathan Yergler (I agree, Nathan! But I asked for more time!), Shannon ("jj") Behrens, and Matt Harrison.
Hopefully a video of this event will be posted. I want to listen to what I actually said. ;)
PyCon: Day 2 (Testing Tools panel)
I wasn't as worried about the Testing Tools panel, 'cause I didn't have to say anything. Of course, it turns out I said too much as a result ;).
The panel was fun but kind of a blur. A bit more time next year, perhaps?
Reviews/notes: Grig Gheorghiu, and Matt Harrison.
I attended the buildbot BoF, which was really fun. Brian Warner rocks.
PyCon: Day 3 (twill talk)
I spent most of the prior evening and morning working on my twill/scotch/figleaf talk. During this time I learned just enough about CherryPy 3.x and Django whatever-the-heck-the-latest-version-is-with-or-without-magic-removal-who-the-hell-knows to actually write test fixtures for them.
I decided to go out on a limb and rather than describe twill/etc. in nauseating detail I worked up nine demos (testing CherryPy sites, doing coverage analysis, writing twill extensions, testing Django sites, and recording Web traffic) and I ran through the demos interactively while providing a narrative.
I really enjoyed this talk format, although it may not be for everyone.
You can grab my talk source code here although this link will eventually (soon!) be broken & moved to an archive containing more documentation.
Review: The Thiers.
I announced the testing-in-python (TIP) mailing list in my talk.
Grig's pybots talk was well-received and (IMO) I think this project is going to dramatically increase the solidity of the Python community's software.
I also got a chance to run some of my ideas for improving test processes on the Python interpreter past Brett Cannon, and (to my shock) he was really open to them. More on that soon.
That evening, I got a chance to meet up with R. Steven Rainwater ("robogato") and his wife Susan; Steven has taken over advogato. They took me out to a nice sushi place, which was really welcome after the heavier food I'd been eating thus far. More on that anon.
Post-PyCon: travelling to San Antonio
After PyCon, Diane Trout and I shuffled ourselves over to UTSA to talk with the nice people at the Computational Biology Initiative. The CBI is interested in making a commitment to future development of Cartwheel which is pretty cool. More anon.