Struct InvertedFile 

pub struct InvertedFile {
    pub lists: Vec<Vec<u32>>,
}

Inverted file: for each centroid, the sorted list of unique document indices that have at least one token clustered in that centroid.

This mirrors PLAID’s “centroid → unique passage ids” layout from §3 of the paper: candidate generation only needs to know which documents are reachable via a probed centroid, and deduplicating per-doc keeps the posting lists small even when a single document has many tokens mapped to the same cluster.

The search path uses this to expand a query token to a shortlist of document candidates: find the centroids with the highest dot-product against the query token, then gather every document listed under those centroids.

Fields

lists: Vec<Vec<u32>>

lists[c] holds the sorted, deduplicated doc_idxs of every document with at least one token assigned to centroid c.

Implementations

impl InvertedFile

pub fn num_centroids(&self) -> usize

Total number of centroids the IVF spans.

pub fn docs_for_centroid(&self, centroid_id: usize) -> &[u32]

Document indices currently associated with centroid_id, or an empty slice if the centroid is out of range. Entries are sorted ascending and contain no duplicates.

pub fn total_doc_postings(&self) -> usize

Total number of (centroid, doc) postings across every list.

This is the sum of lists[c].len() over all centroids c. It is at most num_centroids * num_documents and at least equal to the number of documents that contain any tokens at all.

Trait Implementations

impl Clone for InvertedFile

fn clone(&self) -> InvertedFile

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for InvertedFile

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

Formats the value using the given formatter.

impl Default for InvertedFile

fn default() -> InvertedFile

Returns the “default value” for a type.