Difference between datashader and scatter/heatmap?

[jules]

Datashader is a heatmap of a 2D histogram with an optional nonlinear value count transformation, it’s meant to be a mostly nonparametric way to get a feeling for datasets with up to billions of points. Scatter would overplot too much, you could build it yourself using heatmap but it’s a bit tricky to get super performant and interactive. It also recomputes the binning automatically so it fits the current Axis limits. It’s modelled after Python’s holoviz datashader module which was presented at Makiecon so we wanted something like it for Makie.