Working in Discovery involves assigning image attributes to data attributes, or vice versa. To start with, we assign the “decoration.value” image attribute to the “Population” data attribute. We obtain a visualization often called a “heatmap”  (1) ; we next assign the “sort” image attribute to the “Population” attribute, to sort cities by population. This produces a “rainbow” effect that indicates the distribution of cities in the sample follows a quite normal distribution: a few orange values indicating small population, a mass of light blue and green indicating the average city sizes, and a few more purple and red rectangles showing the big metropolises: NY, LA, Chicago...

(1) regular heatmap

(2) sorted heatmap

All the interaction needed to build sophisticated views happens in much the same way: choose a visual attribute and map it to a data column or to an expression, or choose a data column and assign it to one or more visual attributes