Enum FeatureNameKind 

pub enum FeatureNameKind {
    Exact,
    Gene {
        delim: char,
    },
    Locus {
        merge_overlapping: bool,
    },
    Mixed,
}

Per-name canonicalization rule for cross-backend row alignment. Concrete strategy only — no “request” variants. Callers that want auto-detection pass None (or whatever wrapping enum they choose) and call FeatureNameKind::auto_detect once row names are in hand.

Variants

Exact

Strict string match — no canonicalization. Default.

Gene

Gene-symbol rule: register every delim-split component as an alias of the full name. ENSG00000105329_TGFB1 and TGFB1 resolve to the same row.

Fields

delim: char

Locus

Genomic-locus rule. Normalizes formats (chr1:1000-2000, 1:1000-2000, chr1_1000_2000 → 1_1000_2000). If merge_overlapping, intervals that overlap on the same chromosome additionally collapse into one cluster (chr1:1-20 ∪ chr1:15-30 → 1_1_30). Useful for ATAC peak sets called independently across datasets.

Fields

merge_overlapping: bool

Mixed

Heterogeneous axis: dispatch per row name. Names that parse as loci go through FeatureNameKind::Locus with overlap merging; names with _ use FeatureNameKind::Gene; the rest pass through. Picked automatically when [auto_detect] finds both signatures in the same axis (e.g. paired RNA + ATAC union).

Implementations

impl FeatureNameKind

pub fn canonicalize(&self, name: &str) -> Box<str>

Canonicalize a single name under this kind’s per-name rule. [Locus { merge_overlapping: true }] and [Mixed] only describe the per-name part here (format normalization for loci, last-token split for gene-style); the global cluster step lives in build_locus_overlap_canonical_map and is installed by [build_canonicalizer].

pub fn is_exact(&self) -> bool

True iff this kind is [Exact] — no canonicalizer needed.

pub fn needs_global_pass(&self) -> bool

True iff installing the canonicalizer requires peeking every row name across all backends first (to build the locus-overlap cluster map). Loaders branch on this.

pub fn auto_detect(names: &[Box<str>]) -> Self

Sniff names and pick the right kind. Tallies: n_locus = count where parse_locus matches; n_gene_like = count of remaining names that contain _ (loci with _ separators are not double-counted as gene-like). Decision: both ≥ 10% → [Mixed]; else loci ≥ 50% → Locus { merge_overlapping: true }; else gene-like ≥ 50% → Gene { delim: '_' }; else [Exact].

pub fn reconcile(kinds: &[FeatureNameKind]) -> FeatureNameKind

The one kind to install for a set of files that were each sniffed with auto_detect on their own.

Sniffing the POOLED names does not work: the signature usually lives on one side only. A raw ENSG_SYM cohort pooled with a reference already on the bare-symbol axis (a carried pb_reference, a symbol-keyed panel) leaves the gene-like share under half, the pair sniffs as Exact, and every gene becomes two rows. Canonicalizing under Gene is a no-op for names lacking the delimiter, so adopting the informative side is safe for both.

Gene-style and locus-style files together dispatch per name (Mixed), which is what auto_detect would pick on one axis holding both; Mixed anywhere stays Mixed; all-Exact stays Exact.

pub fn into_canonicalizer(self) -> Option<RowNameCanonicalizer>

Build a RowNameCanonicalizer suitable for [SparseIoVec::with_row_canonicalizer]. Returns None for FeatureNameKind::Exact so callers don’t install a no-op closure. Does not handle the LocusOverlap global step — for that, use build_locus_overlap_canonicalizer.

Trait Implementations

impl Clone for FeatureNameKind

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for FeatureNameKind

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for FeatureNameKind

fn default() -> Self

Returns the “default value” for a type.

impl Eq for FeatureNameKind

impl PartialEq for FeatureNameKind

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for FeatureNameKind