Binned, stacked area charts can have gaps

[paulrosenzweig]

Broken Chart

link

The chart shouldn't have that gap at the bottom. Also, the areas on top are truncated too early on the right. The correct version would look like this:

Fixed Chart

link

The Change

The current generated code is using the original binned field rather than the _mid version. The diff between those chart's Vega code is changing the area's defined channel from this

{
  "signal": "isValid(datum[\"bin_maxbins_10_Body Mass (g)\"]) && isFinite(+datum[\"bin_maxbins_10_Body Mass (g)\"]) && isValid(datum[\"__count\"]) && isFinite(+datum[\"__count\"])"
}

to

{
  "signal": "isValid(datum[\"bin_maxbins_10_Body Mass (g)_mid\"]) && isFinite(+datum[\"bin_maxbins_10_Body Mass (g)_mid\"]) && isValid(datum[\"__count\"]) && isFinite(+datum[\"__count\"])"
}

Another issue

I think ideally, the areas would slope down to zero at the edges of the x-extent for consistency. That seems like a slightly different and less severe bug though. I believe fixing that would require changing (and complicating) the imputation transform or adding another transform step.

[domoritz]

Nice catch and since you already dug into the issue, do you want to send a pull request to fix it?

[paulrosenzweig]

@domoritz Yup! Just opened #7931. It looks I need to be approved to allow running the workflows. I'll un-draft it once those pass.

[domoritz]

Unfortunately I have to reapprove every change but I can do that. Make sure that tests are passing locally.