4 min read

ROpenSci Unconf 2017

2017-05-27

I’m back from the ROpenSci Unconference in LA. A fabulous two days of R programming and, ur, playing Minecraft. As one does.

This was ROpenSci’s 4th unconference; the first I could attend. There were like 70 participants, nearly double what they had last year. The scheme is that people break up into groups and work on different projects that might be useful to the ROpenSci community. There was discussion about potential projects in advance on GitHub, and then the first morning we voted on what we were interested in doing via stickers:

I found the list of topics a bit forbidding. Important stuff, but a lot of them seemed technically difficult, and I didn’t really see how I could best contribute. And, jumping to the punch line, the outcomes of the 2 days’ work are pretty awesome, for example:

Awesome stuff, and that’s 4 of 20. But you know, I don’t think I’m particularly good at any of those things. And for that reason, I was super excited when Gergely Daróczi posted the idea that we make an R package through which we can interact with the game Minecraft.

I’ve played a lot of Minecraft with my kids. And there’s a big community of folks writing “mods” for the game, but to do that, you need to write Java. And, well, I really don’t like Java. So I was excited, a few years ago, to see the book Learning to program with Minecraft, which uses a python module that allows you to interact with the game.

The book is a very nice and thorough introduction to programming, using Minecraft as all of the examples and exercises. Really, it’s the best book on programming for kids that I’ve seen, because it is so full of challenging but do-able and above all interesting exercises. Good exercises are really the key to learning programming, and Minecraft provides a lot of really great and varied challenges.

Our team of six (me, Gergely Daróczi, Brooke Anderson, David Smith, Mario Inchiosa, and Adi Zaidi) set out to replicate the features of the python module as an R package. There are basically just eight functions:

  • set a particular type of block somewhere
  • figure out what type of block is at some spot
  • figure out where a player is
  • move a player somewhere
  • figure out where a player is pointing
  • check whether anyone has hit something with a sword
  • post a message to chat
  • read the chat messages

We had our basic working package, miner, by lunchtime on the first day and spent the next day and a half playing with it and Minecraft.

Just eight basic functions, but we did a ton of awesome things:

  • A robot number-guessing game
  • Automatic creation of mazes that players can walk through (or be moved through automatically)
  • Replicate a photograph, or the R logo, within Minecraft using different types of blocks
  • Give a player Elsa (from the movie Frozen) powers, like walking on water and having it turn to ice, or creating random-sized columns of ice by hitting things
  • Write letters on the side of a mountain
  • Draw a full ggplot2-like scatterplot in the sky

Check out the ReadMe file for our package to see a bunch of pictures and gifs.

I’ll write about some of these things in the (hopefully near) future, because I learned some cool stuff, and it was was really fun.

Just after lunch on day 2, I realized that we could use the chat feature in Minecraft to execute R code from within Minecraft, using parse and eval. And it totally worked. So I could generate my ggplot2-style scatterplot entirely within Minecraft. (The key advantage of this, other than just being cool, is not having to keep switching back and forth between R and Minecraft.)

Watch me type a bit of R and get my scatterplot going:

What next? We’re thinking we’ll compile the many vignettes we developed into a bookdown book. Stay tuned.