05.21.07
hobby project blues
When I started Mapeteria, it seemed like it would be pretty simple. It turned out to be much more complicated that I had originally envisioned. I wasn’t terribly surprised (I have, after all, spent a long time in industry), but it was a bit annoying.
- Installation hell. It didn’t have to be that bad, but I didn’t know about the xampp project.
- Geographic data. I needed boundary information for countries and states/provinces.
- It was easy to find US data, but it was detailed/complex enough that it was really slow. Fortuitously, I saw Kevin Khaw’s state information, and he let me use it.
- Because Mapeteria makes KML that people could use anywhere, I wanted to be quite certain that I had the right to use the boundary data. I found boundary data for Canada, but it wasn’t absolutely clear to me that I had the right to redistribute it. (Unlike in the US, the Canadian government retains the copyright to governmentally-produced information like maps.) I decided it would be faster to just trace out points on Google Maps and use that for the boundaries. (That also gave me control over the complexity of the polygons.)
- I found country data relatively quickly, but it was complex enough that it was extremely slow to render on Google Maps. I was able to simplify the polygons pretty easily (by modifying a script by John Coryat that he’d adapted from Stephen Lime). Unfortunately, there a zillion little islands in that data, which make it much more complex than it needs to be. I believe that I will have to go remove all the islands by hand, yuck. 😛
- I stumbled upon boundary information for France, thanks to Alexandre Dubois (aka Zakapatul). Because I’d already done simplification for the countries, it was not hard to simplify France, but I still had to do it. I also had to strip a bunch of stuff out of the KML file that I didn’t need (like shields representing each departement).
- Bugs in Other People’s Code.
- I never did get the debugger to work right in PHPEclipse, and I didn’t even have a good idea for how to troubleshoot it. So I just had no debugger. 😛 Echo statements (like printfs) were my friends.
- There is a bug in Google Maps such that polygons that straddle the 180 E/W line are just broken. This makes sense — they are inherently ambiguous. However, Siberia, Fiji, and a Russian island straddle the line, alas.
- There is a bug in Google’s maps that I found when I was tracing the Canadian outlines. It wasn’t a big deal, but I spent non-zero time on it.
- Politics. What if somebody submits a data file with information for the USSR? Or Yugoslavia in 1970? Or Ethiopia in 1950? Or East Germany? I only have boundary information for 2006, not for all possible boundaries for all time.
- Documentation.
- What do I call states/provinces/territories/départments? I spent a fair amount of time trying to figure that out. I could call them “first level administrative divisions”, but the only people who know what that mean are map geeks. I could call them states or provinces or territories, but then I’d tick someone off. I never did figure out what to call them, so I call them states/provinces/territories/departments. 🙁
- What do I call the two-letter, uh two-number, uh two-character codes for states/provinces/territories/départements?
- How much detail do I give? How much is too much?
- Testing. In addition to unit tests, I was (for a period) trying to automate more global tests, but comparing a generated KML to a “golden” KML. However, I kept changing what was “golden” — I would take out or simplify polygons, add some debugging information, change from space-separated points to newline-separated points (and back), such that it was a real pain to keep the tests consistent. Eventually I gave up and just had some “eyeball” tests: does it look right?
- Evangelism.
- Who do I tell? How soon? Do I tell them about countries, even though there is still the bug in Google Maps? Even though countries display very slowly? Lots of time spend wondering about that.
- Open Source. I decided to open-source the code after I was basically done.
- I needed to go through and make my code conform to PHP standards (like using_underscores instead of CamelCase), take out some of my hacks, clean up TODOs.
- I needed to figure out where I was going to host the code. My own server? Sourceforge? Google? None were perfect, alas, so in addition to investigating, I had to do some agonizing, too, before settling on Google hosting.
- I needed to transfer all my bugs from my private Bugzilla to the Google issue tracker.
- I still need to transfer the code, which means installing a Subversion client and figuring out how to use it. It probably won’t take long, it’s something I should do anyways (like eating my vitamins), but it’s One More Thing.
So anyway, it always takes longer than you think it should; I decide to document why this time. 🙂
