Struct GtResult 

pub struct GtResult {
    pub n_exact_dup: usize,
    pub n_fuzzy_dup: usize,
    pub n_hard_neg: usize,
    pub n_unique: usize,
    pub n_masters: usize,
    pub cluster_map: ClusterCsr,
}

Result of GtAccumulator::finish.

Fields

n_exact_dup: usize

Cluster members that are genuinely byte-for-byte identical to their master (the master itself, plus any duplicate copy whose assigned noise ended up a no-op — see pipeline::unchanged_row_mask).

n_fuzzy_dup: usize

Cluster members that are duplicates of a master but differ from it on at least one column (noise was actually applied). Distinguished from n_exact_dup so a consumer can’t mistake a fuzzy duplicate for a trivial byte-for-byte match.

n_hard_neg: usize

n_unique: usize

n_masters: usize

cluster_map: ClusterCsr

Every duplicated master’s full cluster (base + duplicate copies, exact and fuzzy alike), each member tagged with its own exact_dup/fuzzy_dup status. Consumed by graph_gen::push_dup_clusters to decide, per edge, whether the pair it connects is exact_dup (both ends byte-identical to the master, hence to each other) or fuzzy_dup (at least one end was noised). Only populated when cluster tracking is enabled (--graph).