Struct GeneIndex 

pub struct GeneIndex { /* private fields */ }

Pre-built index over a gene-name vocabulary for fast marker→row matching. Resolves a query gene in tiers, returning the first matching row: 0. a locus query (chr:start-end) matches only a locus row, by its locus key with the chromosome case kept, and stops here: loci never go through the case-insensitive or fuzzy tiers below,

1. exact (case-insensitive) full-name match,
2. last _-segment symbol match (CD8A ↔ ENSG…_CD8A),
3. leading Ensembl-id segment match (ENSG… ↔ ENSG…_CD8A),
4. decompose a combined ENSG…_SYMBOL query and retry tiers 2–3 per part,
5. HGNC alias retry ([alias_candidates]: HIST1H4C ↔ H4C3, MARCH2 ↔ MARCHF2),
6. fallback to the general flexible_name_match (prefix / _x_ segment).

Tiers 1–5 are O(1) hash lookups; only the rare fallback scans the vocabulary. Build once, match many — replaces the O(genes × markers) .position(flexible_name_match) scan. Tier 1 preferring an exact match over an earlier-indexed suffix match is the one intended refinement vs a pure positional scan. Tiers 3–4 make HGNC / ENSG / ENSG_HGNC reconcile in either direction (gene-set sources mix these conventions), and tier 5 does the same across HGNC releases — the matrix and the marker panel are routinely built against different ones.

Implementations

impl GeneIndex

pub fn build(gene_names: &[Box<str>]) -> Self

Build the index from the dictionary’s gene-name order. The first row wins on duplicate keys (matching positional-scan semantics).

pub fn match_gene(&self, gene: &str) -> Option<usize>

Row index for gene, or None if unmatched (tiers above).