Module labelling 

Grounded labelling over a schema-declared attack set — the one argumentation-semantics computation that is parameter-free, unique, polynomial, and explainable by construction: unattacked entities are accepted, whatever an accepted entity attacks is defeated, entities whose attackers are all defeated are accepted, the rest stay undecided.

A label is a reported observation with its evidence — never a stored value, never a write gate, never a status. The labelling is deliberately support-blind: it walks attack edges only, and a defeated supporter never flips what it supports; the chain-shape statistics give the reader that fact as a count instead.

Structs

LabellingView

One entity’s served labelling view — label, evidence, and the optional shape block.

MemLabelling

One mem’s grounded labelling — the least fixpoint over the pinned graph (non-stub entities of the mem; attack-set edges whose endpoints are both non-stub nodes of the mem). Deterministic: BTreeMaps keyed by id string.

ShapeStats

The chain-shape statistics over one entity’s support subtree — the adversarial-shape indicators (an unusually deep chain with no failing leaves warrants scrutiny). The engine serves numbers, the reader judges.

Enums

Label

The grounded label of one entity.

Functions

compute_mem_labelling

Compute one mem’s grounded labelling over the declared attack set.

compute_shape

Walk the support subtree from start and compute the shape statistics. label_of resolves a subtree node’s label (nodes of mems without a labelling declaration resolve to None and count toward neither label count).

labelling_of

Convenience: whether a schema’s manifest declares labelling.