Sort + set 'categoryarray'

[etpinard]

Plotly.newPlot('graph', [{
  mode: 'markers',
  x: ['A', 'B', 'C'],
  y: ['a', 'b', 'c'],
  marker: { size: [20, 30, 50] },
  transforms: [{
    type: 'sort',
    target: [2, 1, 0]
  }]
}])

currently gives:

where the categories are re-arranged after the sort transform - which is consistent with how we handle categories in filter transforms. See PR #1260 for more details.

But setting layout.xaxis.categoryarray: ['A', 'B', 'C] -- which implies that the category positions are fixed -- curently gives the same results as above.

With commit 7a2a2e1

Plotly.newPlot('graph', [{
  mode: 'markers',
  x: ['A', 'B', 'C'],
  y: ['a', 'b', 'c'],
  marker: { size: [20, 30, 50] },
  transforms: [{
    type: 'sort',
    target: [2, 1, 0]
  }]
}], {
  xaxis: {
    categoryarray: ['A', 'B', 'C']
  }
})

now gives:

where categoryarray is honored.

cc @n-riesco @alexcjohnson

[alexcjohnson]

is there a reason to export this fn?

[etpinard]

No.

[alexcjohnson]

Not a big deal, but I'd prefer not to since it goes into Plotly.Plots

[etpinard]

Yeah, good call 👍

[etpinard]

done in b8e5f52

[rreusser]

Certainly would have been caught by now if this could ever be undefined, right? (I'm too cautious about checking for undefined unnecessarily)

[etpinard]

Yeah. That _initialCategories thing is set in the defaults. Calling Plots.doCalcdata before supplyDefaults is like 🔫 yourself in the 👣 .

[rreusser]

Haha the squirt gun. And then your socks are wet for the rest of the day.

[alexcjohnson]

💃 very clean! But consider #1689 (comment) before merging.