Analysis

Dependency graph: see the architecture, not just the files

A new analysis that answers what depends on what, and what happens if you touch it. Relations are extracted statically from your working directory and rendered as a navigable 3D graph.

Three layouts: force-3d, hierarchical, and metric-space with real axes.
Any node metric (fan-in, fan-out, cycle size, blast radius) can drive size, height, colour or position.
Click a node to pin its metric card; large projects open in group view and you drill in.

Dependency graph in XR

Watch on YouTube

How to read it

The dependency graph answers a different question from the other three: not how big or how complex a file is, but what depends on what, and what happens if you touch it. The relations are extracted statically from your working directory, so nothing has to run.

Nodes are files, folders and symbols; edges are the relations between them. The shape of a node tells you what it is: a sphere is a function, a cylinder a method, a pyramid a class, a diamond an interface and a box a folder. Click any node to pin its card and read its coupling, its cycle and how much depends on it; click again to release it.

Unlike the other analyses this is not a chart you swap. It is one graph you look at three ways, and the layout is the thing you change. External packages collapse into a single portal so they do not crowd your own code, and large projects open grouped so you drill down into folders rather than meeting every node at once.

What you can map onto a node

Size: the metric that scales each node
Height: the metric that raises it above the table
Colour: a metric, or the node’s language
Position: X and Z, available only in the metric-space layout, where the axes are real

Why it is worth the detour: this is the analysis that shows you what a change will touch before you make it, which is exactly what a list of files cannot do.

Three layouts, and how the edges are drawn

The layout decides where nodes sit. The edge mode decides what the lines between them are telling you.

force-3d: A spatial layout: connected nodes pull towards each other, so clusters emerge on their own.
hierarchical: Nodes arranged by dependency level, so the direction of flow is readable.
metric-space: Real axes: X and Z become metrics you choose, so a node’s position itself carries meaning.
Relation type: Edge colour identifies the kind of relation: import, include, require, inheritance, implementation, call or contains. Width stays constant.
Intensity color: Edge colour encodes how often the relation occurs instead.
Intensity width: Edge width encodes occurrences, while colour still identifies the relation kind.
Color + width: Both channels encode how often the relation occurs.

What the numbers mean

These are the extension’s own definitions: the same glossary the in-room guide shows, generated from the real analysis contracts.

Fan-in: Incoming: how many files depend on this one.
Fan-out: Outgoing: how many files this one depends on.
Degree: Fan-in plus fan-out: the node’s total number of connections.
Relations: The individual relations, counting duplicates.
Cycle: The size of the circular-dependency group this node belongs to. Zero means it is in none.
Instability: Fan-out divided by degree. Zero is a stable core that others depend on; one is an unstable leaf that depends on others.
Confidence: Drawn as edge opacity: exact, probable or ambiguous.
Occurrences: Drawn as edge intensity: how often the relation repeats.