We have a big pile of stuff we want to do to Radian. To a large extent, what we (Ian and Tom) work on is driven by the needs of the BayesHive project, but we’ll be trying to implement other features as we can.

If you have features that you particularly need, let us know and we’ll see what we can do about helping you out. (It you’re really desperate for something, you can try offering to pay us. Depending on how much other stuff we have going on, that might work!) If you want to help develop any of these features, really do let us know. We’d be delighted to get some more people working on this stuff. (You can just push pull requests at us on GitHub if you like, but we’re actively developing Radian, so it might be useful for you to know what we’re working on and what we’re not, so as to avoid duplicated effort.)

Things we’re going to do soon

• Better management of axis formatting, ticks and so on: some of this is done, but it could be better.
• Better support for categorical data, particularly for bar charts: I copped out on doing this when I first did bar charts, but it’s been on my list for a while, and it’s something that’s used a lot in statistics.
• Heatmaps, including hex-binning and categorical heatmaps: these are useful for all sorts of data, and hex-binned heatmaps are very nice.
• More “bar type” plots, especially box plots and things of various “bar and whisker” flavours: we do statistics; statisticians love this stuff.
• Real legend management: at the moment, legends in Radian are far from legendary.
• Optimisation: Radian isn’t very efficient. I’ve been concentrating on making it work before making it fast. There are a number of places where things could be cached, pre-computed or just sped up.

Bigger jobs we want to do as soon as we can

• Reflective styling UI: this is a little complicated to explain, but the basic idea is that there will be a UI that you can enable for each Radian plot that allows you to directly manipulate the styling of the plot (changing colours, fonts, axis types, whatever); the “reflective” part is that those changes will be advertised as some sort of Angular events, so that if you want to, you can persist the changes back to a server. We’re embedding Radian plots in literate documents built from a statistical modelling language, and we want people to be able to produce quick plots without caring about the styling, then manipulate the styling via this UI and have the changes they make reflected in the code that’s used to generate the plots. Not a small job, but could be super cool.
• Contour plots: useful for all sorts of 2.5-D and 3-D data.
• Kernel density estimation for 1-D and 2-D data: this is important to us for plotting credible regions on empirical probability density functions, but it’s useful for a lot of other things too.

Big jobs that we’d like to do, but don’t know when

• WebGL backend: this could potentially help a lot with speeding up handling of interactive plots, but it’s quite a bit of work, particularly since Radian relies on D3.js to do much of the heavy lifting of plot generation, and we’d have to reimplement all that stuff ourselves to drive a WebGL backend. Could be worth it though.
• Maps: we want to be able to say “<map>” instead of “<plot>” and have a Leaflet map that we can drop Radian plots onto. Not super-hard, but quite a bit of work, and not something we’ll get around to until we’re dealing with spatial statistics in BayesHive.