Procedural Generation (PCG)

PCG makes "data/rules → geometry" a set of pure functions (spec) => IMObject2D (design.md §6). Generators are side-effect free, touch no scene; randomness always goes through injected rng, so same spec + same seed always yields same result — serializable, seekable, testable.

Generator families

Category	Functions	Description
Parametric / lattice	parametricCurve2D / lattice / tiling	Parametric curves, grids, periodic tilings (§6.3)
Fractal / rules	fractal / recursiveTree / lSystem / cellularAutomaton	IFS/Koch/Sierpinski, recursive trees, L-systems, cellular automata (§6.4)
Fields	functionGraph / isoline / heatmap / streamlines	Function graphs, isolines, heatmaps, streamlines (§6.2)
Graph / data	graphObject / barChart / scatter / lineChart / mapData	Force-directed graphs, data charts (§6.5)

import { parametricCurve2D, recursiveTree, scatter, xy } from "@intermact/core";

const rose = parametricCurve2D({
  domain: [0, Math.PI * 2],
  fn: (t) => [Math.cos(3 * t) * Math.cos(t), Math.cos(3 * t) * Math.sin(t)],
  samples: 400,
  closed: true,
});

const tree = recursiveTree({ depth: 7, branchAngle: 26 });

const dots = scatter({ points: [xy(0, 1), xy(1, 2), xy(2, 1.5)], size: [4, 2] });

Deterministic random: spec.rng

Generators needing randomness (fractal chaos game, graphObject force layout, cellularAutomaton random init) explicitly receive an rng. Without injected rng, fail-fast throw — no silent fallback to non-deterministic results.

import { createRng, graphObject } from "@intermact/core";

const graph = graphObject({
  nodes: ["a", "b", "c", "d"],
  edges: [["a", "b"], ["b", "c"], ["c", "d"], ["d", "a"]],
  layout: "force",
  rng: createRng("my-seed"), // same seed ⇒ same layout
});

Use ctx.rng as default random source at build time; generators remain pure for independent Node/test calls.

Combinator operators

Operators are pure IMObject2D → IMObject2D transforms, chainable; transforms bake into geometry, producing new immutable definitions (design.md §6.6).

Operator	Effect
transformObject(obj, t)	Bake translate/rotate/scale into geometry
repeatObject(obj, n, step)	Repeat n copies with compound step
instanceField(obj, transforms)	Single geometry + per-instance transforms (M16 GPU instancing)
mapPoints(obj, f)	Per-point map (e.g. conformal transform)
along(obj, path, opts)	Distribute evenly along path, optional tangent alignment
booleanOp(a, b, op)	Single-ring polygon union/intersect/subtract/xor

import { booleanOp, circle, repeatObject, transformObject, xy } from "@intermact/core";

const crescent = booleanOp(circle({ radius: 1 }), circle({ radius: 1, center: xy(0.5, 0) }), "subtract");

const fan = transformObject(
  repeatObject(circle({ radius: 0.1 }), 8, { position: xy(0.2, 0), rotation: Math.PI / 8 }),
  { position: xy(-2, 0) },
);

booleanOp supports single-ring operands only; multi-contour objects throw invalid-argument (avoids silently dropping rings).

Plugin generators

PCG is also a registry extension point: register a GeneratorDescriptor, dispatch by name with runGenerator(name, params, rng) (see Extensibility guide).

Related examples

• pcg/operators-showcase — transformObject/repeatObject/booleanOp/mapPoints/along combinators
• pcg/data-charts — mapData data binding + scatter/lineChart
• pcg/generators-extra — tiling/lattice/recursiveTree/parametricCurve2D
• pcg/lsystem-plant — L-system plant (deterministic spec.rng)
• pcg/scalar-field-isolines — isoline/heatmap scalar fields
• pcg/vector-field-streamlines — RK4 streamlines vector field
• pcg/cellular-automaton — Wolfram Rule 30 + 2D Life cellularAutomatonFrames
• pcg/fractal-graph — Sierpinski fractal + graphObject
• pcg/data-driven-bars — data-driven barChart

Full list in example index.