Binary Search Tree (BST)

This library implements a map type BSTMap and a set type BSTSet based on a binary search tree. The API is designed to be similar in usage to types like HashMap/HashSet and BTreeMap/BTreeSet while also providing the ability to implement your own traversals using the lower level root and root_mut methods.

Rather than performing an allocation per node, these data structures are backed by an arena allocator that only allocates occasionally to increase its capacity. The arena allocates memory in chunks without invalidating any addresses it previously produced. This is great for insertion performance since we don't have to copy and update all the previously allocated nodes every time we need more space. There is also a good probability that the nodes you are operating on are laid out contiguously in memory which is great for cache locality.

The root and root_mut methods return types that provide left and right methods for convenient and intuitive traversal through the tree. It almost feels like the tree API you would expect in a garbage collected language, but without the extra overhead and with all the usual protections of the Rust borrow checker.

use bst::{BSTMap, map::Node};

#[derive(Debug, PartialEq, Eq)]
struct Stats {
    pub score: u32,
    // ...other fields...
}

// Custom traversal through the values in the map
fn find_score(node: Option<Node<i32, Stats>>, target_score: u32) -> Option<Node<i32, Stats>> {
    let node = node?;
    if node.value().score == target_score {
        Some(node)
    } else {
        // Recurse through left and right subtrees, just like you would in a GC'd language!
        find_score(node.left(), target_score)
            .or_else(|| find_score(node.right(), target_score))
    }
}

fn main() {
    let mut map = BSTMap::new();

    map.insert(1, Stats {
        score: 39382,
        // ...other fields...
    });
    // ...more insertions...

    // Find the node with score == 500
    println!("{:?}", find_score(map.root(), 500));
}

Safety & Correctness

Like most low-level, high-performance data structures written in Rust, this crate makes use of quite a bit of unsafe Rust code. None of the unsafe Rust code is intended to produce any unsoundness/undefined behavior or result in any memory unsafety. We have extensively tested the code with unit tests and miri, but none of that guarantees that no mistakes have been made. The code has not been audited and every edge case may not be accounted for. Please keep all of this in mind if you decide to use this code in your project or as a reference.

sunjay/bst

Binary Search Tree (BST)

Safety & Correctness

Implementation Progress