Struct TextCandidateScorer 

pub struct TextCandidateScorer { /* private fields */ }

Scores aligned emitted query terms, including duplicates, after support acceptance.

Implementations

impl TextCandidateScorer

pub fn new( mode: &ScoringMode, stats: IndexStats, document_frequencies: &[u64], ) -> Result<Self, TextSearchError>

pub fn new_budgeted( mode: &ScoringMode, stats: IndexStats, document_frequencies: &[u64], budget: &MemoryBudget, poll: impl FnMut() -> Result<(), TextSearchError>, ) -> Result<Self, TextSearchError>

Reserve query IDFs and scalar statistics from the caller’s shared runtime allowance.

Only scalar statistics are needed because frequencies arrive in emitted query order. Unused vocabulary maps are dropped, and the native scorer is stored inline. The callback covers construction and IDF preparation; partial owners are released on failure.

pub fn score_document( &mut self, document_length: u64, term_frequencies: &[u64], ) -> Result<f64, TextSearchError>

pub fn score_document_with_control( &self, document_length: u64, term_frequencies: &[u64], poll: impl FnMut() -> Result<(), TextSearchError>, ) -> Result<f64, TextSearchError>

Sum native term contributions in emitted order and apply calibration once, without a temporary score vector.