prefer lineX when Y is monotonic, lineY when X is monotonic, for better tooltips

[Fil]

This applies apply the same logic as for area to decide between lineX and lineY, but, when not sure, defaults to line (whereas area defaults to areaY).

closes #1547

[Fil]

Sorry this is not ready yet.

[mbostock]

If both x and y are monotonic, and neither xZero nor yZero are specified, I think we should default to lineY rather than lineX. If this works, I think I’ll push this:

Suggested change

	markImpl =
	X && Y // same logic as area (see below), but default to line
	? yZero
	? lineY
	: xZero || isMonotonic(Y)
	? lineX
	: isMonotonic(X)
	? lineY
	: line
	: X // 1d line by index
	? lineX
	: lineY;
	markImpl =
	X && Y // same logic as area (see below), but default to line
	? yZero || isMonotonic(X)
	? lineY
	: xZero || isMonotonic(Y)
	? lineX
	: line
	: X // 1d line by index
	? lineX
	: lineY;

[mbostock]

Looks good to me with the latest commit I pushed, but I can re-review if there’s something else I missed! 🙏

[Fil]

I'm testing it with #1546, and it looks good for lines with two channels.

However I don't like that autoLineHistogram is using lineX (not a new thing); it's specified in the 1d line by index part, and it seems we must reverse lineX and lineY here.

[Fil]

per the above comment:

Suggested change

	? lineX
	: lineY;
	? lineY
	: lineX;

i'd also want to add this test

export async function autoLineHistogramY() {
  const aapl = await d3.csv<any>("data/aapl.csv", d3.autoType);
  return Plot.auto(aapl, {y: "Volume", mark: "line"}).plot({marginLeft: 80});
}

[mbostock]

I see the problem, but I don’t think that’s the right fix. I think the additional context in this example is that there’s a reducer for y, and so even those only x is provided as input, x is binned, effectively meaning that x is monotonic (and y is derived). So, we need to consider the reducers when determining the mark implementation, too.

[mbostock]

Okay, I think the latest commit fixes the problem.