Graphical marks visually encode data using geometric primitives such as rectangles, lines, and plotting symbols. Marks are the basic visual building block of a visualization, providing basic shapes whose properties can be set according to backing data. Mark property definitions may be simple constants or data fields, or scales can be used to map data values to visual values.

Supported Mark Types

The supported mark types are:

  • arc - Circular arcs, including pie and donut slices.
  • area - Filled areas with horizontal or vertical alignment.
  • image - Images, including icons or photographs.
  • group - Containers for other marks, useful for sub-plots.
  • line - Stroked lines, often used for showing change over time.
  • path - Arbitrary paths or polygons, defined using SVG path syntax.
  • rect - Rectangles, as in bar charts and timelines.
  • rule - Rules are line segments, often used for axis ticks and grid lines.
  • shape - A special variant of path marks for faster drawing of cartographic maps.
  • symbol - Plotting symbols, including circles, squares and other shapes.
  • text - Text labels with configurable fonts, alignment and angle.
  • trail - Lines that can change size based on underlying data.

Visual Encoding

Each mark supports a set of visual encoding properties that determine the position and appearance of mark instances. Typically one mark instance is generated per input data element; the exceptions are the line and area mark types, which represent multiple data elements as a single line or area shape.

A mark definition typically looks something like this:

  "type": "rect",
  "from": {"data": "table"},
  "encode": {
    "enter": {
      "y": {"scale": "yscale", "field": "value"},
      "y2": {"scale": "yscale", "value": 0},
      "fill": {"value": "steelblue"}
    "update": {...},
    "exit": {...},
    "hover": {...}

There are three primary property sets: enter, update, exit. The enter properties are evaluated when data is processed for the first time and a mark instance is newly added to a scene. The update properties are evaluated for all existing (non-exiting) mark instances. The exit properties are evaluated when the data backing a mark is removed, and so the mark is leaving the visual scene. To better understand how enter, update, and exit sets work, take a look at Mike Bostock’s Thinking with Joins.

In addition, an optional hover set determines visual properties when the mouse cursor hovers over a mark instance. Upon mouse out, the update set is applied.

There is also a special group mark type (group) that can contain other marks, as well as local data, signal, scale, axis and legend definitions. Groups can be used to create visualizations consisting of grouped or repeated elements; examples include stacked graphs (each stack is a separate group containing a series of data values) and small multiples displays (each plot is contained in its own group). See the [[Group Marks]] page for more.

Top-Level Mark Properties

Property Type Description
type String Required. The graphical mark type. Must be one of the supported mark types.
description String An optional description of this mark. Can be used as a comment.
encode Encode An object containing a set of visual encoding rules for mark properties.
from From An object describing the data this mark set should visualize. If undefined, a single element data set containing an empty object is assumed. The from property can either specify a data set to use (e.g., {"data": "table"}) or provide a faceting directive to subdivide a data set across a set of group marks.
interactive Boolean A boolean flag (default true) indicating if the marks can serve as input event sources. If false, no mouse or touch events corresponding to the marks will be generated.
key Field A data field to use as a unique key for data binding. When a visualization’s data is updated, the key value will be used to match data elements to existing mark instances. Use a key field to enable object constancy for transitions over dynamic data.
name String A unique name for the mark. This name can be used to refer to these marks within an event stream definition. SVG renderers will add this name value as a CSS class name on the enclosing SVG group (g) element containing the mark instances.
on Trigger[ ] A set of triggers for modifying mark properties in response to signal changes.
transform Transform[ ] A set of post-encoding transforms, applied after any encode blocks, that operate directly on mark scenegraph items (not backing data objects). These can be useful for performing layout with transforms that can set x, y, width, height, etc. properties. Only data transforms that do not generate or filter data objects may be used.
role String A metadata string indicating the role of the mark. SVG renderers will add this role value (prepended with the prefix role-) as a CSS class name on the enclosing SVG group (g) element containing the mark instances.

Mark Data Sources (from)

TODO: data, facet.

Mark Encoding Sets

TODO: say more, examples.

All visual mark property definitions are specified as name-value pairs in a property set (such as update, enter, or exit). The name is simply the name of the visual property. The value should be a value reference or production rule, as defined below.

Value References

A value reference specifies the value for a given mark property. The value may be a constant or drawn from a data object. In addition, the value may be run through a scale transform and further modified.

Name Type Description
value Any A constant value. If field is specified, this value is ignored.
field FieldRef A field from which to retrieve a data value. The corresponding data set is determined by the mark’s from property. If a string-valued field name is provided, the value is pulled from the current datum. Dot notation ("price.min") is used to access nested properties; if a dot character is actually part of the property name, you must escape the dot with a backslash: "some\.field". To pull value from the enclosing group mark’s element or datum, this property can be specified as an object.
scale FieldRef The name of a scale transform to apply. If the input is an object, it indicates a field value from which to dynamically lookup the scale name and follows the format of the field property. For example {"datum": "s"} will use the value of s on the current mark’s data as the scale name, whereas {"parent": "t"} will use the value of t on the current group’s data as the scale name.
mult Number A multiplier for the value, equivalent to (mult * value). Multipliers are applied after any scale transformation.
offset Number A simple additive offset to bias the final value, equivalent to (value + offset). Offsets are added after any scale transformation and multipliers.
band Boolean If true, and scale is specified, uses the range band of the scale as the retrieved value. This option is useful for determining widths with a band scale (an ordinal scale where points is false).

Field References

A field reference is either a string literal or an object. For object values, the following properties are supported:

Property Type Description
datum FieldRef Perform a lookup on the current data value. This is the default operation when a field reference is a string.
group FieldRef Use a property of the enclosing group mark element. For example, "field": {"group": "width"} or "field": {"group": "height"}).
parent FieldRef Use a property of the enclosing group mark’s datum. For example, "field": {"parent": "fieldInParentData"}.
level Number A positive integer (default 1) used in conjunction with the group and parent properties to access grandparents or other ancestors. For example, "field": {"parent": "f", "level": 2} will use the value of the f field of the grandparent datum.

These properties can be arbitrarily nested in order to perform indirect field lookups. For example, "field": {"parent": {"datum": "f"}} will first retrieve the value of the f field on the current mark’s data object. This value will then be used as the property name to lookup on the enclosing group mark’s datum.


  • {"value": 5} - The constant value 5.
  • {"field": "price"} - The value of price for the current datum.
  • {"field": "index", "mult": 20} - The value of index for the current datum, multiplied by 20.
  • {"scale": "x", "value": 0} - The result of running the value 0 through the scale named x.
  • {"scale": "y", "field": "price"} - The result of running price for the current datum through the scale named y.
  • {"scale": "x", "band": true} - The range band width of the ordinal scale x. Note that the scale must be ordinal!
  • {"scale": "x", "band": true, "offset": -1} - The range band width of the ordinal scale x, reduced (negative offset) by one pixel.

Color References

Typically color values are specified as a single value indicating an RGB color. However, sometimes a designer may wish to target specific color fields or use a different color space. In this case a special Value Reference format can be used. In the following example, we can set the red and blue channels of an RGB color as constants, and determine the green channel from a scale transform.

"fill": {
 "r": {"value": 255},
 "g": {"scale": "green", "field": "g"},
 "b": {"value": 0}

Vega supports the following color spaces:

Name Description
RGB with properties "r", "g", and "b".
HSL (hue, saturation, lightness), with properties "h", "s", and "l".
CIE LAB with properties "l", "a", and "b". A perceptual color space with distances based on human color judgments. The “L” dimension represents luminance, the “A” dimension represents green-red opposition and the “B” dimension represents blue-yellow opposition.
HCL (hue, chroma, lightness) with properties "h", "c", and "l". This is a version of LAB which uses polar coordinates for the AB plane.

Production Rules

Visual properties can also be set by evaluating an if-then-else style chain of production rules. Rules consist of an array of ValueRef objects, each of which must contain an additional test property. A single ValueRef, without a test property, can be specified as the final element within the rule to serve as the else condition. The value of this property should be a predicate expression, that evaluates to true or false. The visual property is set to the ValueRef corresponding to the first predicate that evaluates to true within the rule. If none do, the property is set to the final, predicate-less, ValueRef if one is specified. For example, the following specification sets a mark’s fill colour using a production rule:

"fill": [
    "test": "indata('selectedPoints', datum._id, 'id')",
    "scale": "c",
    "field": "species"
  {"value": "grey"}

Here, if the ID of a particular data point is found is the selectedPoints data source, the fill color is determined by a scale transform. Otherwise, the mark instance is filled grey.