Domain nicing may break when zero: true

[domoritz]

Notice how in the example below, the domain is not nice when zero: true.

{
  "$schema": "https://vega.github.io/schema/vega/v5.json",
  "width": 200,
  "height": 20,
  "padding": 5,
  "scales": [
    {
      "name": "scale1",
      "domain": [90, 210],
      "nice": 2,
      "range": "width"
    },
    {
      "name": "scale2",
      "domain": [90, 210],
      "nice": 2,
      "range": "width",
      "zero": false
    }
  ],
  "axes": [
    { "orient": "top", "scale": "scale1", "tickCount": 2},
    { "orient": "bottom", "scale": "scale2", "tickCount": 2}
  ]
}

Open the Chart in the Vega Editor

[domoritz]

The issue seems to be that linear scales nice domains differently when the first value is small or not.

$ d3.scaleLinear().domain([60,210]).nice(2).domain()
[0,400]

$ d3.scaleLinear().domain([70,210]).nice(2).domain()
[0,300]

I would expect that both domains produce the same niced domain as they are both contained within the niced domain of the other.

Try at https://observablehq.com/d/ff5a76630225a0bb

[domoritz]

The issue with the code above is that ticks generated from the niced domain do not surround the domain anymore.

$ d3
  .scaleLinear()
  .domain([70, 210])
  .nice(2)
  .ticks(2)
[0,200]

Should I expect this to work or am I misunderstanding what nicing is supposed to do?

The d3 docs say about nice "An optional tick count argument allows greater control over the step size used to extend the bounds, guaranteeing that the returned ticks will exactly cover the domain."

I filed this as d3/d3-scale#209.

[jheer]

The Scale transform in vega-encode is the primary contact point, where we invoke d3’s nice function. This looks like a potential d3-scale issue, so that’s the right place to file. What we do in Vega is pre-process the domain (ensuring zero as needed, any domain padding, etc) then invoke nice if the scale definition specifies it.

I wonder if the specific tick count argument (2) drives the behavior here.

[domoritz]

Confirmed bug and we have a fix in d3/d3-scale#210. I'll leave this issue open so we make sure to update Vega once the fix is out.

[hydrosquall]

Given vega now uses the version of d3-scale in d3/d3-scale#210 and the editor shows the fix, I will close this. Thanks for patching the issue upstream!