Attribute to Field

Reads a named scalar attribute from a points or shape stream and exposes it as a ScalarField. Bridges per-element attribute data into FieldMath composition.

Category: Fields Menu path: Fields > Attribute to Field

Ports

Port	Type	Direction	Description
in	points	input	Points or shape to read attributes from (shape vertices are treated as points)
scalarField	scalarField	output	Field that samples the nearest element's attribute value at any position

Parameters

Param	Type	Default	Description
attribute	string	index	Name of the per-element column to read — an attribute or an intrinsic

How It Works

AttributeToField snapshots the positions and scalar attribute values from the upstream points or shape, then exposes them as a ScalarField. At each sample point, the field returns the attribute value of the nearest element (Voronoi-cell lookup).

Non-scalar attributes reduce to a scalar (Vec2/Vec3 → magnitude, Vec4 → Rec.709 luma), so velocity from TrackPoints reads as speed. A name that resolves to nothing gives a flat 0.0 field.

Coordinate space: Positions and sample coordinates are in origin-centered comp-pixel coordinates, matching the convention of all other field consumers.

Intrinsic Names

Any attribute-name field here also accepts an intrinsic — a property every element has by virtue of existing, computed on read and never stored. The leading @ is a reserved namespace: no attribute may be named with one, so a name starting with @ never hits the attribute map.

Name	Value
@index	Position in the list, starting at 0
@id	Stable PointId — survives filtering and reordering
@x / @y	Position, in local geometry space
@n	Total number of elements
@indexnorm	@index / (@n - 1), normalized 0–1
@rand01	Deterministic 0–1 per element, keyed on @id (stable as neighbours come and go)

Names are case-insensitive (@Y = @y). A misspelled intrinsic (@bogus) resolves to nothing rather than falling through to an attribute lookup.

Non-scalar attributes reduce to a scalar by one rule everywhere: Vec2/Vec3 → magnitude, Vec4 → Rec.709 luma of RGB.

Because this node turns any column into a ScalarField, @y is the cheapest way to give any field consumer position awareness — DrawAscii, PointAttributes, ShapeAttributes, FieldMath and Remap all read the result without knowing anything about points.

Usage Examples

Mask a subset for field-driven deformation

Grid → PointAttributes (target: Custom, customAttribute: "selection", field: DistanceField.scalarField) → AttributeToField (attribute: "selection") → FieldMath (Multiply, b: Noise) → PointDeform.

The selection attribute is 1 near a shape and 0 elsewhere. AttributeToField converts it back into a field, FieldMath multiplies it against the deformation noise, and PointDeform only displaces points near the shape.

Visualize a per-element attribute

Grid → PointAttributes (target: Custom, customAttribute: "energy", field: Noise.scalarField) → AttributeToField (attribute: "energy") → Colorize → Output.

Writes per-point noise values into an energy attribute, then samples it back as a field for rasterization.

Tips

• Attribute names are case-sensitive and must match what the upstream producer wrote (PointAttributes Custom, or built-in keys like scale, opacity, rotation, color)
• Nearest-neighbor lookup gives Voronoi cells — one cell per element. For smooth blending, combine with FieldMath (Add a distance-weighted Gradient) or DistanceField
• Vec2/Vec3/Vec4 attributes are readable, but they reduce to a scalar (magnitude / Rec.709 luma). Use ImageSample if you need genuine vec/color fields from spatial data
• The output is always a ScalarField (not VectorField or ColorField) in v1

Related Nodes

• PointAttributes — writes named attributes that this node can read back
• ShapeAttributes — same, for shape vertices
• FieldMath — combine this field with other fields (multiply as mask, add for soft blend)
• ImageSample — related bridge for spatial raster data → field
• DistanceField — another way to author a scalarField from shape proximity