Point¶

Point mark displays each data item as a symbol. Points are often used to create a scatter plot. In the genomic context, they could represent, for example, point mutations at genomic loci.

{
  "data": { "url": "sincos.csv" },
  "mark": "point",
  "encoding": {
    "x": { "field": "x", "type": "quantitative" },
    "y": { "field": "sin", "type": "quantitative" },
    "size": { "field": "x", "type": "quantitative" }
  }
}

Channels¶

In addition to standard position channels and color, opacity, and strokeWidth channels, point mark has the following channels: size, shape, dx, and dy.

Properties¶

angle

Type: number | ExprRef

The rotation angle in degrees.

Default value: 0

clip

Type: boolean | "never"

If true, the mark is clipped to the UnitView's rectangle. By default, clipping is enabled for marks that have zoomable positional scales.

color

Type: string | ExprRef

Color of the mark. Affects either fill or stroke, depending on the filled property.

fill

Type: string | ExprRef

The fill color.

fillGradientStrength

Type: number | ExprRef

Gradient strength controls the amount of the gradient eye-candy effect in the fill color. Valid values are between 0 and 1.

Default value: 0

fillOpacity

Type: number | ExprRef

The fill opacity. Value between 0 and 1.

filled

Type: boolean

Whether the color represents the fill color (true) or the stroke color (false).

geometricZoomBound

Type: number

Enables geometric zooming. The value is the base two logarithmic zoom level where the maximum point size is reached.

Default value: 0

inwardStroke

Type: boolean | ExprRef

Should the stroke only grow inwards, e.g, the diameter/outline is not affected by the stroke width. Thus, a point that has a zero size has no visible stroke. This allows strokes to be used with geometric zoom, etc.

Default value: false

minBufferSize

Type: number

Minimum size for WebGL buffers (number of data items). Allows for using bufferSubData() to update graphics.

This property is intended for internal use.

minPickingSize

Type: number | ExprRef

The minimum picking size invisibly increases the stroke width or point diameter of marks when pointing them with the mouse cursor, making it easier to select them. The valus is the minimum size in pixels.

Default value: 3.0 for "link" and 2.0 for "point"

opacity

Type: number | ExprRef

Opacity of the mark. Affects fillOpacity or strokeOpacity, depending on the filled property.

semanticZoomFraction

Type: number | ExprRef

TODO

Default value: 0.02

shape

Type: string | ExprRef

One of "circle", "square", "cross", "diamond", "triangle-up", "triangle-down", "triangle-right", "triangle-left", "tick-up", "tick-down", "tick-right", or "tick-left"

Default value: "circle"

size

Type: number | ExprRef

Stroke width of "link" and "rule" marks in pixels, the area of the bounding square of "point" mark, or the font size of "text" mark.

stroke

Type: string | ExprRef

The stroke color

strokeOpacity

Type: number | ExprRef

The stroke opacity. Value between 0 and 1.

strokeWidth

Type: number | ExprRef

The stroke width in pixels.

tooltip

Type: HandledTooltip | null

Tooltip handler. If null, no tooltip is shown. If string, specifies the tooltip handler to use.

x

Type: number | ExprRef

Position on the x axis.

xOffset

Type: number

Offsets of the x and x2 coordinates in pixels. The offset is applied after the viewport scaling and translation.

Default value: 0

y

Type: number | ExprRef

Position on the y axis.

yOffset

Type: number

Offsets of the y and y2 coordinates in pixels. The offset is applied after the viewport scaling and translation.

Default value: 0

Examples¶

Plenty of points¶

The example below demonstrates how points can be varied by using shape, fill, size, strokeWidth, and angle channels.

{
  "data": {
    "sequence": { "start": 0, "stop": 160, "as": "z" }
  },

  "transform": [
    { "type": "formula", "expr": "datum.z % 20", "as": "x" },
    { "type": "formula", "expr": "floor(datum.z / 20)", "as": "y" }
  ],

  "mark": {
    "type": "point",
    "stroke": "black"
  },

  "encoding": {
    "x": { "field": "x", "type": "ordinal", "axis": null },
    "y": { "field": "y", "type": "ordinal", "axis": null },
    "shape": { "field": "x", "type": "nominal" },
    "fill": { "field": "x", "type": "nominal" },
    "size": {
      "field": "x",
      "type": "quantitative",
      "scale": { "type": "pow", "exponent": 2, "range": [0, 900] }
    },
    "strokeWidth": {
      "field": "y",
      "type": "quantitative",
      "scale": { "range": [0, 4] }
    },
    "angle": {
      "field": "y",
      "type": "quantitative",
      "scale": { "range": [0, 45] }
    }
  }
}

Zoom behavior¶

Although points are infinitely small on the real number line, they have a specific diameter on the screen. Thus, closely located points tend to overlap each other. Decreasing the point size reduces the probability of overlap, but in a zoomed-in view, the plot may become overly sparse.

To control overplotting, the point mark provides two zooming behaviors that adjust the point size and visibility based on the zoom level.

Geometric zoom¶

Geometric zoom scales the point size down if the current zoom level is lower than the specified level (bound). geometricZoomBound mark property enables geometric zooming. The value is the negative base two logarithm of the relative width of the visible domain. Example: 0: (the default) full-size points are always shown, 1: when a half of the domain is visible, 2: when a quarter is visible, and so on.

The example below displays 200 000 semi-randomly generated points. The points reach their full size when 1 / 2^10.5 of the domain is visible, which equals about 1500X zoom.

{
  "data": {
    "sequence": { "start": 0, "stop": 200000, "as": "x" }
  },
  "transform": [
    { "type": "formula", "expr": "random() * 0.682", "as": "u" },
    {
      "type": "formula",
      "expr": "((datum.u % 1e-8 > 5e-9 ? 1 : -1) * (sqrt(-log(max(1e-9, datum.u))) - 0.618)) * 1.618 + sin(datum.x / 10000)",
      "as": "y"
    }
  ],
  "mark": {
    "type": "point",
    "geometricZoomBound": 10.5
  },
  "encoding": {
    "x": { "field": "x", "type": "quantitative", "scale": { "zoom": true } },
    "y": { "field": "y", "type": "quantitative" },
    "size": { "value": 200 },
    "opacity": { "value": 0.6 }
  }
}

Tip

You can use geometric zoom to improve rendering performance. Smaller points are faster to render than large points.

Semantic zoom¶

The score-based semantic zoom adjusts the point visibility by coupling a score threshold to current zoom level. The semanticScore channel enables the semantic zoom and specifies the score field. The semanticZoomFraction property controls the fraction of data items to show in the fully zoomed-out view, i.e., it specifies the threshold score. The fraction is scaled as the viewport is zoomed. Thus, if the data is distributed roughly uniformly along the zoomed axis, roughly constant number of points are visible at all zoom levels. The score can be arbitrarily distributed, as the threshold is computed using p-quantiles.

The example below has 200 000 semi-randomly generated points with an exponentially distributed score. As the view is zoomed in, new points appear. Their number in the viewport stays approximately constant until the lowest possible score has been reached.

{
  "data": {
    "sequence": { "start": 0, "stop": 200000, "as": "x" }
  },
  "transform": [
    { "type": "formula", "expr": "random() * 0.682", "as": "u" },
    {
      "type": "formula",
      "expr": "((datum.u % 1e-8 > 5e-9 ? 1 : -1) * (sqrt(-log(max(1e-9, datum.u))) - 0.618)) * 1.618",
      "as": "y"
    },
    {
      "type": "formula",
      "expr": "-log(random())",
      "as": "score"
    }
  ],
  "mark": {
    "type": "point",
    "semanticZoomFraction": 0.002
  },
  "encoding": {
    "x": { "field": "x", "type": "quantitative", "scale": { "zoom": true } },
    "y": { "field": "y", "type": "quantitative" },
    "opacity": {
      "field": "score",
      "type": "quantitative",
      "scale": { "range": [0.1, 1] }
    },
    "semanticScore": { "field": "score", "type": "quantitative" },
    "size": { "value": 100 }
  }
}

Tip

The score-based semantic zoom is great for filtering point mutations and indels that are scored using CADD, for example.