Struct TopKPool 

pub struct TopKPool {
    pub best_error: f64,
    pub accept_error: f64,
    pub stats: PoolStats,
    /* private fields */
}

Bounded pool for collecting matches

Fields

best_error: f64

Best error seen so far (for threshold tightening)

accept_error: f64

Accept error threshold (tightens slowly for diversity)

stats: PoolStats

Statistics

Implementations

impl TopKPool

pub fn new(capacity: usize, initial_max_error: f64) -> Self

Create a new pool with given capacity

pub fn new_with_diagnostics( capacity: usize, initial_max_error: f64, show_db_adds: bool, ranking_mode: RankingMode, ) -> Self

Create a new pool with diagnostic options

pub fn try_insert(&mut self, m: Match) -> bool

Try to insert a match into the pool Returns true if inserted, false if rejected

pub fn exact_complexity_ceiling(&self) -> Option<u32>

Complexity ceiling implied by a pool that is full of exact matches.

When the pool is at capacity and its worst entry is already an exact match, no candidate with complexity >= ceiling can ever enter: a non-exact candidate loses on exactness, and an equally-exact candidate loses on complexity (exact matches rank by complexity). Callers use this to skip Newton refinement — and, since LHS are processed in ascending complexity order, to stop the search entirely — once it fires.

Returns None when the pool is not yet saturated with exact matches, so the optimization never prunes while better matches are still reachable.

pub fn would_accept(&self, error: f64, is_exact: bool) -> bool

Check if a match would be accepted (for early pruning)

pub fn would_accept_strict( &self, coarse_error: f64, is_potentially_exact: bool, ) -> bool

Check if a match would be accepted, with stricter gate when pool is near capacity This is used as a pre-Newton filter to avoid expensive refinement calls

pub fn into_sorted(self) -> Vec<Match>

Get all matches, sorted by ranking mode.

pub fn len(&self) -> usize

Get current pool size

pub fn is_empty(&self) -> bool

Check if pool is empty