Struct FrozenLoadArgs 

pub struct FrozenLoadArgs<'a> {
    pub dictionary_path: &'a str,
    pub bias_path: Option<&'a str>,
    pub target_feature_names: &'a [Box<str>],
    pub name_kind: FeatureNameKind,
    pub source_name_map: Option<SourceNameMap<'a>>,
}

Fields

dictionary_path: &'a str

Path to the [D_src, H] parquet (gbe dictionary.parquet or topic feature_embedding.parquet). Row column 0 is the gene name.

bias_path: Option<&'a str>

Optional path to a [D_src, 1] per-gene bias parquet (gbe feature_bias.parquet). None → bias filled with zeros, which matches the topic models’ implicit “no per-gene bias on ρ”.

target_feature_names: &'a [Box<str>]

Caller’s feature axis (e.g. unified.feature_names for gbe; the topic models’ gene_names). Output rows follow this order after dropping unmatched entries.

name_kind: FeatureNameKind

Per-name canonicalization rule applied to both source and target names before intersection. FeatureNameKind::Exact for strict matching; [FeatureNameKind::Gene { delim: '_' }] is the typical choice for scRNA gene IDs.

source_name_map: Option<SourceNameMap<'a>>

Applied to every SOURCE row name before canonicalization, and kept as the host’s src_names: how a caller whose axis carries a row grammar ({gene}/count/spliced) reads a plain gene table, lifting each bare name into the grammar first. None = the names as read.