set-trie

Fast subset and superset queries based on tries. If you have lookup-based queries, K -> V, but instead of looking for an exact match with K, you want all K's which are a subset or superset of your query, then look no further.

use set_trie::SetTrie;

fn main() {
    let mut employees = SetTrie::new();
    employees.insert(&["accounting", "banking"], "Daniels");
    employees.insert(&["accounting", "banking", "crime"], "Stevens");

    assert_eq!(employees.subsets(&[&"accounting", &"banking", &"crime"]).collect::<Vec<_>>(), vec![&"Daniels", &"Stevens"]);
    assert_eq!(employees.subsets(&[&"accounting", &"banking"]).collect::<Vec<_>>(), vec![&"Daniels"]);
    assert_eq!(employees.supersets(&[&"accounting"]).collect::<Vec<_>>(), vec![&"Daniels", &"Stevens"]);
}

Restrictions

Although currently not implemented in the type system, due to a lack of a trait bound over sorted iterators, set tries require all queries to be sorted. Failing to sort the query or key will result in nonsensical results:

use set_trie::SetTrie;

fn main() {
    let mut trie = SetTrie::new();
    trie.insert(&[2, 3], "Foo");
    trie.insert(&[1, 2], "Bar");

    // although we'd expect this to contain &"Bar".
    assert_eq!(trie.subsets(&[&2, &1]).collect::<Vec<_>>(), Vec::<&&str>::new()); 
}

Features

• Subsets and supersets are lazily evaluated, through an iterative DFS algorithm.
• Convenient entry API.

TODO

• Implement benchmarks (low priority, PR welcome)
• Memorize seen keys to save 1 binary search (requires benchmarks to compare performance)