Struct ClusterCsr 

pub struct ClusterCsr { /* private fields */ }

Compressed-sparse-row representation of the duplicate-cluster membership. Built once, in GtAccumulator::finish, from every (master, record, is_identical) triple belonging to a duplicated master.

Avoids one String allocation per record_id and one per distinct master_id (the dominant RAM cost of a string-keyed cluster map at 200M+ records): master_id/record_id are packed into u64s (pipeline::pack_master_parts / the global record index) and stored in two flat, contiguous buffers instead of a hash map of per-cluster Vecs.

master_id itself is never needed downstream as a string — the only consumer (graph_gen::push_dup_clusters) only ever used it to get a deterministic cluster iteration order, and numeric comparison of the packed key preserves the lexicographic order of the original string (both segments are zero-padded to a fixed width).

Implementations

impl ClusterCsr

pub fn groups(&self) -> impl Iterator<Item = (&[u64], BitspanRef<'_>)>

Iterates clusters in ascending packed-master-key order, yielding each cluster’s (record_indices, is_identical) pair. A cluster’s record_indices are already ascending, so callers (push_dup_clusters) don’t need to sort members themselves.