Grammerjack

It's been a year since I bought my fancy car stereo . I wanted to mention that I've ended up not using most of the fancy features. The built-in MP3 player had issues with my 1000-song / 100 album / 4 GB music collection: It took a long time to scan the USB storage each time the radio turned on. It was difficult to navigate the album and song lists. Chinese characters were not supported. It took a long time to "continue" a paused item, especially a long item like a podcast. The bluetooth pairing only worked with one phone at a time, which was awkward in a two-driver family. Here's how I use it now: I connect my phone using both the radio's USB (so the phone is being charged) and mini stereo plugs. I use the phone's built-in music player. I have the radio set to AUX to play the music from the stereo input plug. It's a little bit cluttered, because of the two cables, but it gets the job done. The phone's music player UI is so much better t...

I just returned to the US after eight months living in Taiwan. The trip was great fun for me and my family. But more than that, it was a very productive time for me to learn new programming technologies. Why was my trip so productive? I think some of the reasons were: I was away from my home computer set up (including video game consoles and a large TV). All I had was a Mac Mini (for the family to use), a MacBook Pro (for work) and a Nexus One phone. I commuted by bus rather than car. While my commute time was longer, it was much more productive. Instead of having to concentrate on driving, I had 40 minutes a day of free time to think, web surf, and even code. I didn't follow local politics, and USA politics were made less interesting by distance and time zone. I stopped visiting political web sites and obsessing over current events. The timezone difference between Taipei and the US ment that email became a once-a-day event, rather than a continuous stream of interruptions. ...

The ICFP 2010 contest rules have been posted, and I am quite disappointed by this year's contest. This year's contest is difficult to describe. It's too clever by half, and it depends too much on time, and too much on repeated interaction with the organizer's servers. The conceit is that you're designing "car engines", and "fuel for car engines". A car engine is a directed graph with certain properties, and "fuel" is a directed graph with other properties that generates a set of coefficients that are fed to the engine. Layered on top of that is an encoding puzzle (you have to figure out how the engine and fuel directed graphs are encoded for submission to the IFCP servers.) Layered on top of that is an economy where you are encouraged to submit your own car engines and write optimal fuels for other people's car engines. There are benefits for finding better fuels for existing car designs. (You aren't provided the details of th...

Happy Mother's Day everyone! This year I wrote an Android app for my wife for Mother's Day. How geeky is that? The reason I wrote it is that my wife's favorite Android application, Word Mix Lite, recently added annoying banner ads. While the sensible thing to do might have been to switch to another app, or perhaps upgrade to the paid version of Word Mix, I thought it might be interesting to see if I could write my own replacement. And as long as I was writing the app, I though I'd try writing it using JavaScript. Normally Android applications are written in Java, but it's also possible to write them in JavaScript, by using a shell application such as PhoneGap . Why use JavaScript instead of Java? Well, hack value mostly. On the plus side it's a fun to write! Another potential plus is that it is theoretically possible to run the same PhoneGap application on both Android and Apple. The drawback is that PhoneGap is not very popular yet, especially among Android dev...

Team Blue Iris (that's me!) placed 77th out of 708 in the Google AI Challenge 2010 contest. The contest was to design an AI for a Tron lightcycle game. My entry used the standard min/max approach, similar to many other entries. I guess that's why it performed similarly to so many other entries. :-) I didn't have any time to investigate better algorithms, due to other obligations. In addition to my own entry (which was in C++), I also created starter packs for JavaScript and Google's Go language. These starter packs enabled people to enter the contest using JavaScript or Go. In the end, 6 people entered using Go , and 4 people entered using JavaScript . I'm pleased to have helped people compete using these languages. I initially wanted to use JavaScript and/or Go myself, but the strict time limit of the contest strongly favored the fastest language, and that made me choose a C++ implementation. Practically speaking, there wasn't much advantage to using another ...

I wanted to learn more about how the original BitTorrent client implemented "choking", so I downloaded the source to the original Python-based BitTorrent client from SourceForge . I was impressed, as ever, by the brevity and clarity of Python. The original BitTorrent client sources are very easy to read. (Especially now that I understand the problem domain pretty well.) The code has a nice OO design, with a clear separation of responsibilities. I was also impressed by the extensive use of unit tests. About half the code in the BitTorrent source is unit tests of the different classes, to make sure that the program behaves correctly. This is a really good idea when developing a peer-to-peer application. The synthetic tests let you test things like your end-of-torrent logic immediately, rather than by having to find an active torrent swarm and wait for a real-world torrent to complete. When I get a chance I'm going to refactor the Taipei-Torrent code into smaller classes and...

From Scientific American: Binary Body Double: Microsoft Reveals the Science Behind Project Natal Short summary: They used machine-learning algorithms and lots of training data to come up with an algorithm that converts moving images into moving joint skeletons. Given the huge variation in human body shapes and clothing, it will be interesting to see how well this performs in practice. I bet people will have a lot of fun aiming the Natal camera at random moving object to see what happens. I can already imagine the split-screen YouTube videos we'll see of Natal recognizing pets, prerecorded videos of people, puppets and marionettes. Oh, and of course, people will point Natal back at the TV screen and see if the video game character can control itself. Great fun!