Cerrisian Blog

An additional awesomeness would be to be able to interact with all this stuff stored in revision control through a wiki — maybe with a javascript parser/interface to ReStructuredText. The hardest part (potentially) would be creating/using a web interface to pitz, since that’s all based on flat files.

I didn’t realize how flexible Pitz was, it can actually handle collections of *anything* and it outputs to restructured text!

Just need the ability to add pitz output to Sphinxes indices and I’ll be all set!

I’ve been reading a book on cleanroom software engineering methodologies and one of the more controversial points it makes is that developers shouldn’t do “unit testing”. What?!?! I hear you say, no unit testing?

Then I realized what they were describing wasn’t unit testing as most modern programmers think of it, automated tests that run quickly inside of a unit testing framework, but of debugging and manual testing. This same advice is levied by many people in the agile community. It isn’t as far-fetched as it sounds at first. Instead of automated unit tests cleanroom advocates proofs (informal or formal) of program functionality.

The other part of clean room is that acceptance testing is statiscal, based on a model of user interaction with the system. I would bet that this is a hard model to create in the first place, but I’ll have to read more.

I went to the OSUG meeting tonight with a coworker of mine. The presentations covered Joomla! and Griffon, a rails like framework for Swing gui applications using Groovy. It was really quite impressive, especially the additional build support for applets and webstart.

I was able to take my new system76 starling netbook (that I’m typing on right now) and connect to the free internet, that ability made me some friends.

All the people were really nice there, and excited. Most of them were a bit older than I was, and I was shocked at the number of women (3 out of a group of about 14 people in total). Definitely a creative, vibrant group. My type of people

I’m going to follow this up with some more posts on my PyCon 2008 log later.

Today at the reservior I saw this cute little terrier mutt by a kid fishing. Every time the kid would catch a fish, the dog would get excited and try to eat it. He’d also go into the water a little bit and bark at the lake, it was hilarious.

After coming back from the top, we saw the dog still hanging around the lake, pointing and barking and trying to get the fish out.

I’ve never seen a dog so excited about fish! My dad called him “Sea Biscuit”.

A few weeks ago I learned about an interesting technique one of my professors used when writing algorithmic code. He’d first write a very naive, straightforward implementation that wasn’t algorithmically smart but gave correct results. To create the optimal design, he’d use the naive one as a baseline to verify that the answers are correct.

Definitely interesting, gives a much larger input/output space then just doing something table driven.

As I was taking a shower, I thought of an interesting idea, creating a literate programming style book about either CPython internals or tinypy using Sphinx and RestructuredText. The most difficult part would probably be including snippets of the C code and keeping it updated with new code commits and changes.

It may be easier to create such a thing using LyX and noweb, I’ll have to investigate further…

Dasher is basically a way to input textual data using a mouse or other 2D input device. I wonder how effective it would be to use a dasher like algorithm for zooming, but map each section of the map unto one of four keys that could be pressed from the keyboard. It’s definitely something worth trying. It may be faster than using a mouse.

Now, this may be a challenge. How do you procedurally generate a website layout? My current idea is:

• Start with YUI CSS stuff
• Add modifications per tag in a smart way (CSS attributes)
• Genetic algorithm it up

This would be difficult but useful to do.

It’s funny sometimes how you can spend all day working on some code and all you have to show for it is a handful of new lines of code. That’s how it was today messing with Django contrib comments. I think understand flagging now though.

Django comments comes with the ability to flag comments, flags can basically mean any binary value, approved/unapproved, spamfiltered/unfiltered, etc. The comment view that comes with django comments is for flagging comments for removal. Now, it just flags the comment, if you want the comment to become hidden, you need to add a message listener for flags and have it look through the flag type.

I’ve set all of this up in preparation for retrying djangobayes and commentbayes. Last time I ran into a few exceptions being thrown when there wasn’t a djangobayes database yet, so we’ll see how it goes.

I wonder if I should add the ability to flag/moderate comments in emails, probably a good idea.

Sphinx is a really popular documentation generator for python, probably due to ease of use and the absolutely gorgeous output it gives.

PyLit is a way to convert python code directly into ReST (the format Sphinx uses) and back.

Sphinx provides indexing, cross reference and other goodies.

Based on these two facts, I think that PyLit and Sphinx are a legitimate literate programming tool for python.

Literate programming allows the author to tell a narrative to explain how a system works, browsing API docs or source code is difficult unless you have a narrative or examples to follow.

The other alternative I looked into was epydoc. Although epydoc does generate excellent API documentation, it doesn’t really provide a narrative, the closest thing to a narrative I see is package or module level documentation.