Hyperbee

A peer-to-peer append-only B-tree built on Hypercore. Compatible with the JavaScript version.

$ cargo add hyperbee

Usage

From the examples:

use hyperbee::Hyperbee;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let hb = Hyperbee::from_ram().await?;
    // Insert "world" with key "hello"
    hb.put(b"hello", Some(b"world")).await?;

    // Get the value for key "hello"
    let Some((_seq, Some(val))) = hb.get(b"hello").await? else {
        panic!("could not get value");
    };
    assert_eq!(val, b"world");

    // Trying to get a non-exsitant key returns `None`
    let res = hb.get(b"no key here").await?;
    assert_eq!(res, None);

    // Deleting a key returns `true` if it was present
    let res = hb.del(b"hello").await?;
    assert!(res.is_some());

    // Getting deleted key returns `None`
    let res = hb.get(b"hello").await?;
    assert_eq!(res, None);

    Ok(())
}

Foreign Language Bindings

We use UniFFI to generate libraries for other languages. To build the library for python run:

cargo build -F ffi && cargo run -F ffi --bin uniffi-bindgen generate --library target/debug/libhyperbee.so --language python --out-dir out

This generates a file out/hyperbee.py, which an be used. This file requires that libhyperbee.so be present alongside the .py file. Distributable python packages are still a work-in-progress. Currently only Python is tested. See the tests for example usage.

Parity with JS Hyperbee

• full functional interoperability with JS Hyperbee files
• read, write, and delete operations
• in-order key streaming like JS's createReadStream
• support gt, lt, etc bounds for key streaming
• accept compare-and-swap for put and del.
• support prefixed key operations like JS's sub
• one-to-one binary output #23

Future work

• Build FFI wrappers
• improved wire format
• configurable tree parameters

Development

Running the tests requires protoc and GNU Make be installed.

Each significant pull request should include an update the CHANGELOG.md

Release

Releases are mostly handled with cargo release. After each Rust release. We manually release a new python package. Building the python package is done with maturin and it is released with twine.

# create a virtualenv with the tools we need
python -m venv venv && . venv/bin/activate
pip install --upgrade twine maturin

# build the wheels
python -m maturin build --release

# Upload the wheels to pypi. This uses my api token from ~/.pypirc
python -m twine upload target/wheels/hyperbeepy-<target>.whl