@knowvah/dot-engine

Appearance

Glossary ​

One definition per term, alphabetized. Each links to the guide page (or the source) that covers it in depth.

Cluster ​

A subgraph whose name starts with cluster (e.g. subgraph cluster_build) — Graphviz renders it as a distinct box grouping its member nodes. Internally, @knowvah/dot-engine's geometry snapshot re-keys every cluster subgraph to a positional name like cluster6 (ClusterGeometry.name), not the DOT source name, so a consumer that needs the original name builds an idByName map before layout and re-keys snapshot.clusters afterward. See Recipes for the re-key pattern and Build a graph for creating clusters via addSubgraph.

Conformance ​

The mechanically-checked property behind the claim that a @knowvah/dot-engine render "matches" the C oracle. After both SVGs are parsed into normalized element trees, every numeric value (coordinates, path data, points) must agree within a fixed tolerance — ±0.01pt for the deterministic engines (dot, circo, twopi, osage, patchwork) and ±0.5pt for the iterative force-directed engines (neato, fdp, sfdp) — and every non-numeric value (tags, colors, text) must be exactly equal. It is not a claim of byte-for-byte SVG output. See Conformance.

Coordinate frame / y-axis ​

Graphviz's native coordinate system is y-up, origin at the lower-left corner; browsers and screens are y-down, origin at the upper-left. getLayout defaults to yAxis: 'down' (flipping every y and normalizing bounds to (0, 0)) and accepts yAxis: 'up' to return native graphviz coordinates unchanged. xdot draw ops (from getDrawOps) are always in the native y-up frame. See Read computed geometry.

Divergence ​

A difference between a @knowvah/dot-engine render and the oracle that has been investigated, root-caused, and catalogued — as opposed to silently tolerated. Catalogued divergences fall into one of three classes: accepted deltas (deliberately not made conformant, e.g. cross-platform floating-point non-determinism), a tracked long tail still being closed, and explicit non-goals. An unlisted difference is treated as a defect, not accepted behavior. See Known divergences.

DOT ​

The graph description language — digraph { ... } / graph { ... } with node, edge, and attribute statements — that @knowvah/dot-engine parses before handing the result to a layout engine. See Getting started.

Image sizer / resolver ​

The two injectable seams for external images (usershape nodes and <IMG> HTML-label cells). An ImageSizer reports an image's natural width/height so node sizing and label layout can proceed without loading pixel data; an ImageResolver supplies the actual image bytes for embedding at render time. See Images.

Layout engine ​

One of the eight layout algorithms @knowvah/dot-engine registers, selected by name (renderSvg(dot, engine)): dot (hierarchical/layered), neato (spring-model, Kamada–Kawai), fdp (force-directed), sfdp (multiscale force-directed, for large graphs), circo (circular), twopi (radial), osage (clustered), and patchwork (squarified treemap). See Layout engines.

Oracle ​

The native C Graphviz dot binary, built from the canonical C source, that every @knowvah/dot-engine render is validated against. @knowvah/dot-engine spawns this binary directly (never a WASM build) to avoid ABI drift between the reference and the port. See Conformance and Parity for how oracle comparisons are run and reported.

Rank / rankdir ​

In dot's hierarchical layout, a rank is a layer of nodes placed at the same depth in the drawing. rankdir sets the direction ranks flow in — the default TB (top to bottom), or LR, BT, RL — set as a graph attribute (b.setAttr('rankdir', 'LR')). See Build a graph.

Spline / edge routing ​

The curved (Bézier) path an edge is drawn along, computed by routing code that steers around node and cluster obstacles. @knowvah/dot-engine exposes the routed control points as EdgeGeometry.points — an ordered array of {x, y} points, in points — from getLayout. See Read computed geometry.

Text measurer ​

The injectable seam (TextMeasurer) that reports label width/height so node and edge-label sizing can proceed before layout. @knowvah/dot-engine resolves one automatically per render — an explicit setTextMeasurer first, then the browser's <canvas> if available, then the built-in deterministic EstimateTextMeasurer in Node — or accepts a custom implementation. See Text measurement.

Usershape ​

Graphviz's term for a node whose shape is an externally supplied image (via the image attribute) rather than a drawn polygon or ellipse. @knowvah/dot-engine resolves usershapes through the injectable image sizer/resolver seam rather than reading files directly, keeping the library browser-safe. See Images.

xdot ​

The extended DOT draw-op format: a structured stream of operations (set fill/stroke color, set font, fill/stroke an ellipse or polygon, draw a Bézier, draw text) describing exactly how a rendered graph should be painted, in paint order. getDrawOps returns this stream as typed XdotOp values for driving a custom renderer (canvas, WebGL, PDF) without parsing SVG. See Custom rendering with xdot draw-ops.