Struct ResultCache 

pub struct ResultCache { /* private fields */ }

LRU cache of engine answers, keyed by the prose that produced them.

A keystroke re-checks the whole document: the prose is re-extracted and every range goes back to the engines, though only the edited one changed. For LanguageTool that is the difference between one small request and a hundred — measured on a 36 kB Typst file against a 4-CPU server, 838 ms per keystroke pause against 108 ms.

Answers are cached as the engine returned them: offsets relative to the range, rule ids not yet normalised. Normalisation and the severity overrides run over cached answers as over fresh ones, so changing a rule’s severity takes effect without waiting for the cache to turn over.

Implementations

impl ResultCache

pub fn new(capacity: usize) -> Self

A cache holding capacity answers, or a disabled one when that is zero.

A disabled cache misses every lookup and stores nothing.

pub fn get( &mut self, engine: &'static str, language: &str, text: &str, ) -> Option<Vec<Diagnostic>>

The engine’s last answer for this prose, if it is still held.

pub fn put( &mut self, engine: &'static str, language: &str, text: &str, diagnostics: Vec<Diagnostic>, )

Remember what the engine answered for this prose.

pub fn len(&self) -> usize

Number of answers currently held.

pub fn is_empty(&self) -> bool

pub fn clear(&mut self)

Drop everything, for when a config change invalidates every answer.