Someday I’ll write about something other than the ICFP programming contest here. Someday sooner I might even write a more coherent writeup about this year’s. But until then, here’s the text that poured from my fingers into my README file at 3:30 AM Monday morning.

Rambling

This is not the algorithmic breakthrough you’re looking for. I wasn’t feeling hugely inspired about writing game AI, so I didn’t, and I wanted an excuse to spend more time re-learning Rust, so I did.

This is me trying to reuse my idea from 2012 (when I used the just-released Rust 0.3) (also sort of my idea from 2008 and, eventually, 2009) and just try a bunch of random paths, looking for ones that score well.

That worked okay in past years when there was a specific goal to aim towards, or especially a series of goals so I could do some kind of evolutionary-ish hill-climbing thing to the ones that made it. This problem has a bunch of fiddly positional stuff that I wasn’t feeling up to trying to come up with a utility heuristic thing for, and it didn’t really go so well.

(I also meant to write a routine to actually display the game state, so I could visualize what was going on, and… kind of never did. Oh well.)

But I did manage to adapt it to find paths with chosen strings, after spending a bunch of time thinking about the right way to use Brzozowski derivatives to build a minimal DFA and considering the nondeterminism of choosing a letter for a command and then ignoring all of that and just hacking up something that more or less worked. And it did kind of okay at that.

Also, typing this, I realize that just top-down imposing a specific power phrase to try would have been a lot simpler, but I’m not going to try a frenzied last-minute implementation, because it won’t help much.

I’m sure everything I should have done will make perfect sense when I get up tomorrow, relaxed and having taken the mental step back from everything that I can never manage during the weekend, because that’s always how it works.