We need a way to control the stroke attribute, and perhaps other attributes, as a function of z.

In some line plots, if the cardinality of z is relatively low, it would be nice to label the lines say at the start or end. Similarly for dot plots, it would be lovely to have an automatic labeling strategy. Related vega-label.

If y is specified as a function of the input data (or an equivalent array of values of the same length as x, and a quantitative value), then the size of each bin should be proportional to the sum of y-values for that bin, rather than a count of values. The default should be equivalent to () => 1.

Will be needed for Histogram #9: x1, y1, x2, y2 all quantitative.

The zero properties allow a rule at zero, but we should have a way to specify rules at arbitrary x and y values.

How would we support this?

What about streamgraphs? Or diverging stacks?

It defaults to zero, but there may be an x-axis with a top orientation, so that’s wrong.

It’d be neat if you could use a quantitative dimension for faceting, and we automatically binned it (say using d3.bin) into a reasonable number of facets.

Marks currently ignore data with undefined values. For example, BarX will ignore data whose y, x1 or x2 value is undefined.

This is a handy feature for avoiding SVG errors. However, we’re not entirely inconsistent with grouping. I fixed the group1 transform to skip groups whose key is undefined,

but I forgot to fix group2 at the same time. And we don’t apply this same technique to faceting, which is also inconsistent.

I’m now of the opinion that ignoring undefined data for ordinal dimensions — which includes grouping and faceting — is undesirable. Whereas there’s no consistent way to handle undefined data in a quantitative dimension, undefined values don’t have to be specially treated in ordinal dimensions. Furthermore, null and undefined are often used to represent an “other” or “unknown” category with grouping and faceting.

Therefore, I think we should changing the filtering logic and only filter undefined values for quantitative dimensions.

Note: D3 doesn't have an "identity" scale in that sense (d3.scaleIdentity coerces values to numbers).

Facet margins currently double the chart margins, but sometimes you want the two margins to be different. Perhaps fx and fy could take a margin property that affects the respective margin?

Currently faceting repeats scales. What if you want a scale that varies by facet, as in a SPLOM?

Related #2.

Related area #21, but this is interesting because it’s the first case where multiple abstract dimensions share the same y-scale, so we now need to support a many-to-one mapping from abstract dimensions to scales.

Here's a nice(?) example found in the wild, of faceted donuts.

Source

The data is available, but I don't think it's the right dataset to use, since I don't understand what it means :)

I hard-coded one for use with histogram, but it should be configurable.

And for facets, the maximum across all facets?

With Bar, x = quantitative and y = ordinal. With Column, x = ordinal and y = quantitative.

When using a color legend (e.g., #2 #19), should we automatically include a nice legend (probably yes)?

A sqrt scale?

Should we support ordinal data for radius (probably not)?

And how would we render a key?

Brush a rectangular region to select data.

The Histogram plot currently only lets you specify the x-values, but it’d be nice if you could specify y-values instead and get a Histogram with bars (i.e., rotated 90°) instead.

In the browser (see these lines):

• xmlns is missing;
• the style.font declaration is not understood.

In Illustrator 2019:

• the arrow symbol → is not recognized in the default (sans serif) font. I don't know how to fix this, except by giving an explicit font-family that has more unicode glyphs.

In Affinity Designer 1.8.6:

• the currentColor declaration is not recognized and some bars are interpreted as brown instead of black. This could be fixed by replacing "currentColor" by "black".

We also want a way to specify "id=xxx" for layers and marks, which become names of the objects in the Illustrator layers panel.

related keywords: export, illustrator

#36 explores a "Carto" mark.

Speaking in general, we have several possibilities for proportional symbols:

• svg:circle is concise (and possibly the most cpu-efficient); scale with r/size.
• svg:path + d3.symbol is useful for certain types of scatterplots, but difficult to extend; scale with transform:scale
• svg:text + emoji works if we're careful to center the mark properly; scale with transform:scale

Note that svg:use + defs covers the three cases, and removes the difficulty of having to define complex symbols when we're not happy with the set in d3.symbols.
See https://observablehq.com/@fil/svg-use

So if we want a variant of Plot.dot that allows various symbols, I would try to code it with <use>. The range of the symbol scale is a list of objects that would have an id and be added to the defs, or a list of ids—but then how do we pass the objects themselves? The mark could default to <use href=#square>. And it would accept an orientation channel!

(discarding svg:polygon which has no advantage that I know of over svg:path.)

I think this might be related to #1, but we need to do some clamping or reconsider whether axis offsets apply to label anchors. Note the “V” is cut in half below:

Plot.Dot(AAPL, {
  x: {
    value: d => Math.abs(d.Close - d.Open) / d.Open * 100,
    label: "Daily change (%) →",
    format: "+f",
    rule: 0
  },
  y: {
    value: d => Math.log10(d.Volume),
    label: "↑ Volume (log₁₀)",
    offset: 12
  },
  grid: true
})

It’d be nice to have a tooltip that shows the value nearest the mouse.

partially implemented in #38 and
https://observablehq.com/d/c364dd180391294a

For example,

Plot.plot({
  x: {
    domain: "AEIOU"
  },
  y: {
    grid: true
  },
  marks: [
    Plot.barY(alphabet, {x: "letter", y: "frequency"}),
    Plot.ruleY([0])
  ]
})

Expected

Actual

The problem is that the label anchor defaults to center if the type of the y-axis is ordinal, but at the time we apply the default, we don’t yet know the scale type.

Dot plots might want control over z-order due to occlusion.

Currently x defaults to indexOf and y defaults to identity, but it’d be nice if when you only specified y, x could default to indexOf, e.g.

Plot.Line({length: 500}, {y: Math.random})

Currently grid is strictly a boolean, but true could be equivalent to the corresponding ticks, a number could compute new ticks (e.g., for higher density) and an array could contain explicit tick values.

Per #75 (comment) we want to hide the axis if it has much too many ticks.

Related d3/d3-array#180.

It’s useful!

But… probably not until they’ve baked a little more.

What if we want to change the color and font of axes?

Building it into a specialized Mark type seems to work well. d4e4df8

Seems like 7017838 broke axis labels; the elements are there but invisible, because currentColor is white or null for them or something? the label matched the old .plot text { fill: currentColor } CSS but doesn’t go through the Text Mark code path so now it's undefined?