This crate makes it fast and simple to build a finite determinic automaton that computes the levenshtein distance from a given string.
use levenshtein_automata::{LevenshteinAutomatonBuilder, Distance};
fn main() {
// Building this factory is not free.
let lev_automaton_builder = LevenshteinAutomatonBuilder::new(2, true);
// We can now build an entire dfa.
let dfa = lev_automaton_builder.build_dfa("Levenshtein");
let mut state = dfa.initial_state();
for &b in "Levenshtain".as_bytes() {
state = dfa.transition(state, b);
}
assert_eq!(dfa.distance(state), Distance::Exact(1));
}
The implementation is based on the following paper Fast String Correction with Levenshtein-Automata (2002) by by Klaus Schulz and Stoyan Mihov. I also tried to explain it in the following blog post.
The time taken by the construction a Levenshtein DFA strongly depends on the max distance it can measure and the length of the input string.
Here is the time spent to build a Levenshtein DFA for the string "Levenshtein"
dfa dist1 no transposition 35,627 ns/iter (+/- 3,237)
dfa dist1 with transposition 36,493 ns/iter (+/- 12,680)
dfa dist2 no transposition 97,137 ns/iter (+/- 14,556)
dfa dist2 with transposition 100,958 ns/iter (+/- 4,231)
dfa dist3 no transposition 834,412 ns/iter (+/- 158,329)
dfa dist3 with transposition 1,414,523 ns/iter (+/- 396,278)
dfa dist4 no transposition 4,716,365 ns/iter (+/- 869,024)
dfa dist4 with transposition 8,044,162 ns/iter (+/- 594,523)