A beeswarm layout in Observable -- on deadline
I recently had to complete a technical test for a role I’m interested in.
The task: from a story out of the Global Witness website, can you come up with an alternative visualisation? You’ve got two hours.
As there were no embed constraints, I figured there was no point in faffing with boilerplate code and scaffolds, so I went down the Observable route to put a chart together.
Here’s what we need scales-wise:
- On the y axis, the amount of the individual donations range from 0 to $10,000 – we’ll use a linear scale which domain is mapped to this extent;
- On the x axis, we’ve got two categories: the donations that pass the NAACP criteria, and those that don’t. We cue an ordinal scale which domain are these values;
- A simple colour scale;
But as donation amounts overlap each other, so do our circles. Here they are, with a small increasing offset to show this banding effect:
Comes the force layout to the rescue, adding some simulated entropy to the x/y coordinates of our circles and allowing us to appreciate the difference in scale between the two columns.
const simulation = d3 .forceSimulation(data) .force('x', d3.forceX(d => x(d.category)).strength(0.2)) .force('y', d3.forceY(d => y(+d.donationUSD)).strength(1)) .force('collide', d3.forceCollide(d => size(Math.sqrt(+d.donationUSD)))) .alphaDecay(0) .alpha(0.3) .stop();
We apply three forces to our simulation:
- One on the x axis, so our circles bucket into our two categories. We specify a weak
.strength()so they spread out horizontally rather than vertically (and thus are more truthful to the precise donation amounts);
- One on the y axis, where in this case the
.strength()is far greater to squeeze them further, as seen above;
- A collision force which repels the circles from each other, so as to spread them apart.
Here they are, showing the influence of x and y’s
Did I say I had two hours to do this? I had to skimp on the annotations, which I’ve positioned manually and quite roughly as
#d3 #observable #viz