Two nodes, side by side

What each one contains, what it depends on, what uses it — and what the two have in common.

Comparable because the model is shared

Two graphs built separately cannot be compared, because nothing guarantees they used the same vocabulary for the same idea. Everything here maps onto one closed set of kinds, so a difference between two columns is a difference in the things or in the data, never in the software.

Shared neighbours are the interesting part

What two nodes both depend on usually says more than what either depends on alone — a shared dependency is a coupling, and a shared parent is a category.

Questions

Can I compare across grains?

Yes. The columns are the same shape whatever grain each node came from.