visuals

Visuals

Types of visuals

There are 2 different types of visuals that can be used in a project:

1. Charts and Graphs: These are used to represent data in a visual format, such as bar charts, line charts, pie charts, etc. These are built using vega-lite, see references/vega-lite-visual.md for more details.
2. Data Grids: These are used to display tabular data in a structured format, allowing for sorting, filtering, and pagination. See references/data-grid-visual.md for more details.

Packages & Imports

The visual components are provided by three packages. Always use these package imports when creating visuals.

The DataTable type (used by both components) is defined in @microsoft/fabric-visuals-core and re-exported by the visual packages' type definitions.

Data Format

The chart and data grid components share a unified data prop of type DataTable (from @microsoft/fabric-visuals-core). This structured format carries column metadata (displayName, format, semanticType) that the components use for axis titles, grid headers, number formatting, and tooltips.

Using DataTable: Pass a DataTable via the data prop. The visual uses its column metadata for formatting, axis titles, and tooltips.

Static/inline data: For static data, transformed data, or plain arrays, put data: { values: [...] } directly in the Vega-Lite spec and omit the data prop.

Multiple tables in one visual: the data prop also accepts a Record<string, DataTable> for specs that bind separate layers to more than one dataset by name such as layered overlays, reference lines, and axis spines. See references/multi-data-input.md.

import { useThemeContext } from "@/hooks/theme.context";
import { VegaVisual } from "@microsoft/fabric-visuals";
import { DataGrid } from "@microsoft/fabric-datagrid";
import type { DataTable } from "@microsoft/fabric-visuals-core";

const { theme } = useThemeContext();

// Charts — pass a DataTable and Vega-Lite spec
<VegaVisual spec={vegaLiteSpec} data={dataTable} theme={theme} />

// Grids — displayName becomes column headers, format applies to cells
<DataGrid data={dataTable} theme={theme} />

// Static/inline data — no data prop needed
const inlineSpec = {
  data: { values: [{ x: 1, y: 2 }, { x: 3, y: 4 }] },
  mark: "point",
  encoding: { ... },
};

<VegaVisual spec={inlineSpec} theme={theme} />

For the DataTable schema and ColumnDef fields, see references/data-table.md.

VisualContainer defaults for every visual

VegaVisual and DataGrid wrap themselves in a VisualContainer by default. The container draws the chrome — border, padding, header (title/subtitle) — plus its built-in actions, all enabled by default. Do not add a wrapper of your own unless absolutely required.

<VegaVisual
    spec={vegaLiteSpec}
    data={dataTable}
    theme={theme}
    header={{ title: "Revenue by Region", subtitle: "Last 12 months" }}
/>

<DataGrid data={dataTable} theme={theme} header={{ title: "Top Products" }} />

Anything the default container doesn't cover — custom actions, programmatic capture, the same chrome around a non-visual etc: references/visual-container.md.

Formatting & Theme

• Encoding semantics: Choose encoding types by field meaning, not storage type. Use ordinal for integer years shown as ordered categories, nominal for identifiers, quantitative for measured quantities, and temporal for actual date/time values. See Encoding Type Selection for examples and continuous-time guidance.
• Formatting rules: Number formatting, color palettes, chart-specific encoding rules, highlighting guidelines, and a default theme. See references/formatting.md.

Custom visuals

Always use the above mentioned ways to create visual when possible. If the user's request doesn't allow creation using the above methods, ask the user if they are ok with using another library for creating the visual. If they are ok with it, use the library to create the visual. If they are not ok with it, then build that visual from scratch using HTML, CSS, and JS/TS. Make sure to ask the user for any specific requirements they have for the visual, such as colors, labels, etc.

Interactivity

Always use the onInteraction prop on VegaVisual and DataGrid to surface user selections. Follow the visual-interactivity skill for event handling, native Vega-Lite selections, coordinated filtering and highlighting, filter-aware grid totals, and parameterized query factories.