Struct CorpusStats 

pub struct CorpusStats {
    pub n_docs: f64,
    pub avgdl: f64,
    pub df: HashMap<Vec<u8>, u32>,
}

Corpus statistics supplied from outside a segment, for scoring one shard’s documents against the whole corpus rather than its own slice.

A cross-shard text query builds this by summing each shard’s local n_docs / total_len and, for each query token, its df. df need only carry the query’s tokens — the values a query actually scores with — which is why global BM25 does not need a whole-corpus df table.

Fields

n_docs: f64

Total documents across the corpus.

avgdl: f64

Mean document length (unweighted tokens) across the corpus.

df: HashMap<Vec<u8>, u32>

Global document frequency per query token; a token missing here falls back to the segment’s local list length.