v0.9.0

We refactored a number of syntax decisions in this PR as we are moving closer to 1.0. Please read carefully to make sure that you migrate correctly.

Main APIs

• vl.compile's signature has changed
  • It no longer requires stats
  • It returns an object with property spec contains the vega spec. (We will add warning / errors as additional properties in the future.)

Syntax

• CORE SYNTAX CHANGES
  • Rename marktype to mark
  • Rename name to field in each field definition
  • Rename col to column for clarity
  • Replace short name type with full name type by default e.g., Q => quantitative, O => ordinal, T => temporal, N=> nominal. But short name types are still supported and the type property is now case insensitive.
• Added vega's scale, axis, and legend properties that should be supported #181
• Added scale property to row, column and shape
• rename / refactor
  • axis.maxLabelLength => axis.labelMaxLength
  • scale.bandSize => scale.bandWidth
  • Move encDef.band to be a part of scale
    • band.size => scale.bandWidth
    • band.padding => scale.padding
• revise useRawDomain to support only aggregate functions that produces values ranging in the domain of the source data ('mean', 'average', 'stdev', 'stdevp', 'median', 'q1', 'q3', 'min', 'max’)
• revised config
  • remove unused config.gridColor, config.gridOpacity
  • remove config.bandWidth, config.padding and config.useRawDomain as it is redundant with scale.bandWidth and scale.padding, scale.useRawDomain in field definitions
    • set default values of scale.bandWidth, scale.padding just for X, Y
  • group configs of a particular element together
    • config.cell
      • move cell configs to be under cell: width, height, padding, gridColor, gridOpacity, gridOffset
        • remove config.singleWidth, config.singleHeight, encoding.column.width, encoding.row.height and simply always use value from cell.width, cell.height
      • add the following group marks properties to config.cell: fill, fillOpacity, stroke, strokeWidth, strokeOpacity, strokeDash, strokeDashOffset`
      • added missing properties
    • config.marks
      • move baseline, font, fontStyle, fontWeight from encoding.text to config.marks
      • encoding.shape.filled => config.marks.filled
      • encoding.color.opacity => config.marks.opacity
      • move all text properties from encoding.text to config.marks except renaming encoding.text.placeholder to encoding.text.value
      • config.strokeWidth => config.marks.strokeWidth
      • added missing properties

Output

• data
  • rename Vega’s data tables: raw => source and aggregate => summary
  • added layout table for layout calculation and add root marks group for binding layout data
• output now describe properties in update rather than enter
• only produces scales, axes, legends when they are non-empty
• removed axis grid for row and column and added rule marks (row|column-rules) instead for drawing rules/grid between rows / columns
• {field: {group: 'width|height'}} are used where necessary
• removed scales for timeUnit and used template instead
• revised default values
  • scale's nice, points and zero
  • set axis's layer: 'back' only if grid is true - fixes #526
  • complex rules for determining opacity and color scales removed.
  • stack's default order is revised
  • simplified rules for automatic opacity

Helper APIs

• move requiredEncodings and supportedEncodings to vl.validate
• add vl.validate.getEncodingMappingError() -- this is subject to change.
• Include Vega-Lite version in the compiled files. (vl.version)
• add encoding.padding() helper
• Add separate methods for properties in scale, axis, and legend
• rename
  • vl.encDef => vl.fieldDef
  • vl.enc => vl.encoding
  • vl.Encoding => vl.spec for static methods
  • vl.Encoding => vl.compiler.model for internal model for the compiler
  • scale.sort() => scale.domain.sort()
  • vl.schema.encTypes => vl.const.Enctype.LIST
  • general utility methods are now moved to vl.util
• Add namespaces for helpers and constants vl.type, vl.channel, vl.bin, vl.aggregate, vl.timeunit
• move shorthand methods to vl.shorthand
• fixed bug when having array in the schema for schemautil.subtract

Internal

• Migrated to Typescript & add interfaces for all type
• Break schema into multiple files
• Vega-Lite editor removed, as Vega Editor now supports Vega-Lite

v0.8.3

• Change default value for axis.grid for binned field to false
• Reimplement bin's scale using linear scale instead (#693)
• Fix missing value problem in stacked area chart using the new Vega's impute transform
• Remove sort transform from output stacked area chart spec as stack transform already included sortby

v0.8.2

• Make encoding definition sort=‘ascending’ by default

Change Vega Output

• No longer add ’-‘ to sortby in the output stack transform by default
• Change bin transform to produce start instead of bin to reflect the Vega 2.3.0's syntax change.

v0.8.1

• schema file is now vega-lite-schema.json rather than spec.json
• axis.grid is automatically determined. If grid is unspecified, the default value is true for ROW and COL. For X
  and Y, the default value is true for quantitative fields and false
  otherwise.

v0.8.0 – Migrating to Vega 2

Syntax Change

• axis
  • only include axis.orient when necessary
• legends
  • set legend's orient to right by default (similar to Vega)
  • Vega-lite now supports multiple legends
• filter
  • move filter to data.filter and takes Vega expression as parameter instead of a complicated predicate object.
  • Null filter is added before binning and timeUnit conversion
• scale
  • scale.reverse is removed as it is redundant with sort.
• stack
  • add stack control to color and detail encoding
• sort — now sort property can be either undefined, a String or an Object
  • undefined - the field is unsorted.
  • (as String) 'ascending' or 'descending' – the field is sort by the field's value in ascending or descending order.
  • (as Object — for ordinal/nominal fields only) A sort field object - for sorting the field by an aggregate calculation (sort.op) over a specified sort field (sort.field) in a specified order (sort.order)

Vega 2 Migration Refactor

• ValueRef
  • replace d.data with datum in filter
  • add datum. option to fieldRef and use it in time transform and template
  • eliminate outdated fieldRef options
  • Replace the eliminated group mark properties with field.group
  • Refer to {group: 'mark.group.width’} with
• Transforms
  • Aggregation, Bin
  • Stack
    • Use vega2 stack's parameter and rename it to <field_name>_start|end, remove the need to add facet transform before stacking
    • Make facet comes after stack for stacked area chart (This makes stacked area chart looks almost right but still looks weird due to #307 – see example screenshot)
  • Facet
    • use field name reference instead of using keys.0 (#639)
• Aggregation (#621)
  • avg => mean
  • add all supported aggregation in Vega 2
• Sorting (#612)
  • remove sort.js and no longer use sort-<field_name> tables as data source
  • modify scale.domain() to include DataRef’s .sort object
  • modify sort property to be consistent with Vega 2 (but still leave it as a top-level vega-lite property)
    • sort.name => sort.field
    • sort.aggregate => sort.op

API Change

• vl.field => vl.encDef, field.js => encDef.js

Internal Refactor

Variable Name

• fieldName => field
• field => encDef

Other

• update vega dependency in editor and gallery to Vega 2
• removed _vega2 flag
• extract filter non-positive value when log are presented from normal filter
• eliminate encoding.fieldName() and just use encoding.field(...).name
• only use encDef.name where applicable rather than calling fieldRef
• local variable refactor in scale.js
• move stack’s stacked data transform to data.stack
• add encoding.stack so we do not have to pass stack around
• remove Encoding.toggleFilterNullO, which should just be a part of vega-lite-ui

Examples

• added Aggregate Bar Chart, Stacked bar chart, Trellis Stacked Bar Chart
• both vl input and vg output are validated in the gallery and editor

v0.7.18

Release v0.7.18.

Release v0.7.13

Release v0.7.13.

v0.7.1

• change .aggr => .aggregate, .enc=>.encoding
• add N (nominal) data type
• add more color palette options
  • use color brewer scale or interpolate color for O
  • use categorical color scale for N