hdf5-rt

Thread-safe Rust bindings for the HDF5 library with runtime loading (dlopen).

Forked from hdf5-metno.

Overview

hdf5-rt loads the HDF5 library at runtime via dlopen, eliminating build-time dependencies on HDF5. This makes it ideal for:

• Julia/Python bindings - Reuse the HDF5 library already loaded by HDF5.jl or h5py
• Portable binaries - Ship without bundling HDF5
• Version flexibility - Work with any compatible HDF5 version installed on the system

Features

• Runtime loading - No compile-time HDF5 dependency
• HDF5 1.10.4+ support - Compatible with Ubuntu 20.04+, HDF5.jl, h5py
• Thread-safe - Safe concurrent access to HDF5

Optional features:

• complex: Complex number type support (Complex32, Complex64)
• f16: Float16 type support

Usage

[dependencies]
hdf5-rt = { git = "https://github.com/tensor4all/hdf5-rt" }

use hdf5_rt::File;

fn main() -> hdf5_rt::Result<()> {
    let file = File::create("test.h5")?;
    let group = file.create_group("data")?;
    let dataset = group.new_dataset::<f64>()
        .shape([100, 100])
        .create("matrix")?;
    Ok(())
}

Requirements

• HDF5 1.10.4 or later installed on your system
• Rust 1.80.0+

# Ubuntu/Debian
sudo apt-get install libhdf5-dev

# macOS
brew install hdf5

Crates

Crate	Description
hdf5-rt	Main HDF5 bindings with runtime loading
hdf5-rt-types	Native Rust equivalents of HDF5 types

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE)
• MIT license (LICENSE-MIT)

at your option.

Acknowledgments

Based on hdf5-metno by Ivan Smirnov, Magnus Ulimoen, and contributors.