I mostly like my Sonos system. The downsides: I had a controller die on me out of warranty; it’s a bit pricey; and it doesn’t have Airplay support. This last one, at least, turns out to be fixable, thanks to a few open source projects and the Raspberry Pi.

The ingredients:

• the Shairport project, for providing Airplay
• the SoCo project, for controlling the Sonos

This last one is important for making the Sonos start listening to its line-in when Airplay begins to be used.

Getting your Raspberry Pi on your wifi network is a bit of a pain. You’ll need to compile Shairport’s dependencies and install some modules from CPAN. SoCo’s setup.py doesn’t leave you with a working installation, but the codebase is fine if you use it directly.

With all the pieces in place, you just need a script like this to switch the Sonos:

and then to start the Shairport daemon like so:

This will call the set_line_in.py script every time a new Airplay client connects.

For good measure, here’s the /etc/init.d/shairport startup file I’m using:

It all works pretty well! The only real downside is the volume: Airplay’s pretty quiet. I should be able to automate the volume adjustment on the Sonos — SoCo exposes that functionality as well, and Shairport can call scripts upon client disconnect, too, to reset the volume. But I haven’t had a chance to write that yet.

Anyway, if folks start wandering in from Google and find this useful or desirable, let me know. I could probably make this all a bit more reusable without a ton of trouble (distributing a Raspberry Pi filesystem image might be the easiest thing, really).

UPDATE: Hmm. Subsequent investigation of the Sonos’s line-in level adjustment capabilities reveals that the SoCo portion of this might be unnecessary: there is indeed some capacity for adjusting amplification on the line-in. Better still, there’s some sort of line-in activity detection! I think I might stick with my script anyway: I have a Sonos unit on the porch that’s usually powered off; I’ll probably just adjust the script’s logic to pipe the line-in to it if it’s activated. Still, for most users the Shairport stuff should be sufficient.

I mean, yikes.

So look. I’m obviously no designer. And I agree that the pie charts in that blog post, which are clearly cherry-picked for maximum terribleness, are terrible.

But usually pie charts are fine. I’m not going to bother collecting examples that demonstrate this, because unless you’ve led an unusually brief or interesting life, you’ve seen thousands of them and understood them perfectly well.

But I realize this isn’t going to stop you. You have been to the Seminars. You have received the Texts. You have studied the words of the Prophet Tufte, and you have concluded that they demand jihad. I’m not going to convince you otherwise.

I’ll make my case anyway.

To me, pie charts are handy for conveying information about the relative components of a whole because they do so through visual cues of both area and angle. If you need to show the relative sizes of multiple wholes, you can do that by varying the pies’ size! Fun times, but I suppose it’s not for everyone.

But I would like to request that you not join the pie-haters I sometimes see advocating for donut graphs or treemaps as alternatives. These have all of the problems of pie charts, but they throw away one of the two visual cues. (Treemaps can be useful for expressing hierarchies, and donut graphs are useful for… er. Well. Sometimes you can put a number in the middle of them!)

If your graph is confusing, it’s probably a bad graph and you should try to make a different one. But let’s not pretend that there’s a Platonic Graph that we’re working toward just because some dweeb in an eye-tracking lab eked out a few milliseconds’ worth of statistical significance. Real graphs used for Science usually look like this (to be honest, even that red line is a bit decadent). It’s not great, but it gets the job done. Science is pretty much going fine despite a lot of shitty-looking graphs.

If you ask me (you haven’t) presenting information is about tradeoffs. I don’t really want to spend all day looking at things that are ugly, and that’s okay. Nicely designed graphs are pleasant! People like to look at them, which is important if you want people to look at your graphs. Even if you really think pies are inefficient, you’re kidding yourself if you think graphs are always purely made to shoot numerical truth into your readers’ brains (when was the last time you adjusted the stroke weight on an error bar?).

People have varying tastes. That’s a totally legitimate rationale for hating pie charts. And they’re not the best choice for everything. I’m on board with that, too.

(Building part of your self-conception around the fanatical endorsement of the work of the one widely-known name in information design is, I would suggest, a little sillier. But whatever floats your boat. I went through a big anime phase in college, for instance.)

Anyway: pie charts. All I really ask is that you please shut up about them.