Undesirable padding behaviour with disparately sized children about d3-hierarchy

[Santas]

A two-pass approach might be an acceptable if imperfect solution: compute the layout without padding, compute the area lost to padding for each cell, add this to the value for the cell, and then recompute the layout.

We do something similar to what @mbostock suggested in #19. We call treemap twice and hack around the issue by reordering some nodes to better fit which is not ideal, but works for our use case as a workaround.

[ianstormtaylor]

Running in to this as well, I assumed the padding would be aware enough to not distort the resulting data. Seems like all the Treemap examples on Observable suffer from this?

Looks like this was brought up in #19 and #20 but forgotten.

[mshd]

@ianstormtaylor @russss have a look at https://github.com/codeledge/entitree-flex