Struct SemanticDisk 

pub struct SemanticDisk { /* private fields */ }

The taxonomy-embedded meaning space (see module docs).

Implementations

impl SemanticDisk

pub fn build(spec: &[(&str, usize)]) -> Result<SemanticDisk, StoreError>

Build a semantic disk from a concept specification: (concept path, affinity dim) pairs — e.g. [("/trauma", 16), ("/cgt", 17), …]. Nested paths are allowed and embed with their real tree shape; missing ancestors are created automatically (they become unmapped, purely structural anchors).

Errors on an empty spec, duplicate concept paths, or duplicate dims.

pub fn concepts(&self) -> Vec<&str>

The mapped concept paths, in canonical (sorted) order.

pub fn position_of( &self, store: &Store, key: &str, ) -> Result<Option<Vec<f64>>, StoreError>

A node’s derived position in the concept disk, as f64 Poincaré coordinates. Ok(None) when the node has no positive affinity on any mapped dim (no concept position — same convention as empty semantic coords).

pub fn concept_of( &self, store: &Store, key: &str, ) -> Result<Option<String>, StoreError>

Which concept a node belongs to right now: the anchor whose hyperbolic Voronoi cell contains the node’s derived position. Constant in data-node count (linear only in the anchor count — dozens). Ok(None) when the node has no concept position.

Compared against the node’s storage path, this is the miscategorization primitive: filed under /overig, classifies to /trauma.

pub fn nearest( &self, store: &Store, key: &str, k: usize, ) -> Result<Vec<(String, f64)>, StoreError>

The k data nodes nearest to key in the concept disk (hyperbolic distance between derived positions), excluding key itself. Sorted ascending by (distance, key).

pub fn nearest_to_weights( &self, store: &Store, weights: &[f64], k: usize, ) -> Result<Vec<(String, f64)>, StoreError>

The k data nodes nearest to an explicit affinity-weight vector (one weight per mapped concept, in Self::concepts order).

pub fn classify_trajectory( &self, sample_dim_start: usize, samples: &[(u64, Vec<f64>)], ) -> Vec<(u64, String)>

Push a temporal trajectory readout through the anchor cells: for each (epoch, coords) sample — the shape HttHistory::trajectory returns — classify the derived position, yielding the node’s symbolic trajectory ((epoch, concept) pairs). Samples whose weights are all non-positive are omitted (no position at that epoch).

sample_dim_start is the first dimension index the samples cover (the dim_range.start the trajectory was read with); mapped dims outside the sampled range weigh zero.