Struct LexicalIndex 

pub struct LexicalIndex { /* private fields */ }

The store-wide lexical index across every segment.

Segments here are SEALED

push_segment encodes the active segment into the persisted posting layout and keeps only that. There is therefore exactly one scoring path in the engine — over bytes — rather than an in-memory one that ships and a persisted one that is only ever tested.

Implementations

impl LexicalIndex

pub fn new() -> Self

Creates an empty index.

pub fn push_segment(&mut self, segment: SegmentIndex) -> Result<(), IndexError>

Seals one active segment and appends it.

Errors

Returns IndexError::Postings when the encoder rejects the segment’s geometry or refuses to read back what it just wrote. Both are broken-writer conditions, and they surface here rather than becoming a decode fault inside a later query.

pub fn push_sealed(&mut self, segment: SealedSegment)

Appends an already-sealed segment.

pub fn push_sealed_with_live_rows( &mut self, segment: SealedSegment, live_rows: &DocBitmap, ) -> Result<(), IndexError>

Appends an already-sealed segment with exact live-row counters.

The row set is the segment tombstone state’s live bitmap, not a query filter. Capturing it here keeps BM25 corpus statistics independent of later metadata allow-lists while avoiding any posting-list rewrite.

Errors

Returns IndexError::LiveRowOutOfRange when the tombstone state names a row outside the sealed segment, or IndexError::LiveLengthMissing when a sealed length counter violates its dense-row invariant.

pub fn segments(&self) -> &[Arc<SealedSegment>]

Returns the segments in seal order.

pub fn terms(&self) -> impl Iterator<Item = &[u8]>

Iterates the analyzed vocabulary of every segment.

Terms may repeat across segments; callers that expand a structured query must impose their own deterministic de-duplication.

pub fn document_count(&self) -> u64

Returns the store-wide document count.

pub fn total_tokens(&self) -> u64

Returns the store-wide analyzed token count.

pub fn corpus_stats(&self) -> Result<CorpusStats, IndexError>

Returns the store-wide corpus statistics.

This is the ONLY source of N and avgdl for scoring. There is deliberately no per-segment equivalent.

Errors

Returns IndexError::Stats when the index holds no documents or no tokens, because avgdl is undefined then.

pub fn document_frequency(&self, term: &[u8], fields: &[FieldId]) -> u32

Returns the store-wide document frequency of one term.

Summed across every segment and every requested field. A document containing the term in two fields counts once, which is what makes this a document frequency rather than a posting count. Segments without tombstones answer from their sealed dictionary. Segments with tombstones decode document identifiers and count only rows in the same live bitmap used for N and avgdl.

Trait Implementations

impl Clone for LexicalIndex

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 LexicalIndex

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

Formats the value using the given formatter.

impl Default for LexicalIndex

fn default() -> Self

Returns the “default value” for a type.