Struct QueryExpansionEngine 

pub struct QueryExpansionEngine {
    pub config: ExpansionConfig,
    pub synonym_db: HashMap<String, SynonymEntry>,
    pub context_terms: HashMap<String, Vec<String>>,
    pub expansions_performed: u64,
    pub total_terms_added: u64,
}

Production-grade query expansion engine.

Thread Safety

The engine holds no shared state — callers that need concurrent access should wrap it in a Mutex / RwLock (write lock only for expand_query because it mutates the counters).

Fields

config: ExpansionConfig

User-supplied configuration.

synonym_db: HashMap<String, SynonymEntry>

Synonym / hypernym / hyponym registry, keyed by lowercased term.

context_terms: HashMap<String, Vec<String>>

Context-term map: trigger word → list of related terms.

expansions_performed: u64

Total number of expand_query calls.

total_terms_added: u64

Cumulative count of expansion terms added across all calls.

Implementations

impl QueryExpansionEngine

pub fn new(config: ExpansionConfig) -> Self

Create a new engine with the given configuration.

pub fn add_synonym_entry(&mut self, entry: SynonymEntry)

Insert or replace a SynonymEntry in the synonym database.

The entry.term is lowercased before use as the map key so that lookups are case-insensitive.

pub fn add_context_terms(&mut self, trigger: String, related: Vec<String>)

Register a list of related terms that fire when trigger appears in a query.

If an entry for trigger already exists the new terms are appended (deduplication is left to the caller).

pub fn expand_term(&self, term: &str) -> Vec<QeExpansionTerm>

Return expansion terms for a single term, respecting the current ExpansionConfig.

The returned list is sorted by weight descending and capped at config.max_expansions_per_term. Terms below config.min_weight are excluded.

pub fn expand_query( &mut self, query: &str, embedding: Option<Vec<f64>>, ) -> QeExpandedQuery

Tokenise query, expand every token, and return a QeExpandedQuery.

Tokenisation splits on ASCII whitespace and the punctuation characters ., ,, !, ?, ;, :. Empty tokens are discarded.

The embedding argument, if provided, is stored verbatim in the result (e.g. a pre-computed dense vector for the full query string).

pub fn build_search_string(&self, expanded: &QeExpandedQuery) -> String

Build a space-separated search string from all terms whose weight is at or above config.min_weight.

The original tokens (source = ContextualExpansion) are placed first, followed by expansion terms, both groups sorted by weight descending.

pub fn weighted_terms<'a>( &self, expanded: &'a QeExpandedQuery, ) -> Vec<(&'a str, f64)>

Return (term, weight) pairs sorted by weight descending.

pub fn coverage_stats(&self, expanded: &QeExpandedQuery) -> ExpansionStats

Compute expansion statistics for a QeExpandedQuery.

pub fn engine_stats(&self) -> (u64, u64)

Return (expansions_performed, total_terms_added).