12 releases

0.8.3 Dec 23, 2023
0.8.1 Jun 5, 2023
0.7.4 Feb 1, 2023
0.7.3 Dec 23, 2022
0.6.1 Jul 7, 2021

#149 in Encoding

Download history 402/week @ 2024-02-26 295/week @ 2024-03-04 663/week @ 2024-03-11 1020/week @ 2024-03-18 450/week @ 2024-03-25 638/week @ 2024-04-01 1186/week @ 2024-04-08 2264/week @ 2024-04-15 2103/week @ 2024-04-22 3513/week @ 2024-04-29 3415/week @ 2024-05-06 3512/week @ 2024-05-13 3020/week @ 2024-05-20 3352/week @ 2024-05-27 4287/week @ 2024-06-03 3956/week @ 2024-06-10

14,729 downloads per month
Used in 14 crates (6 directly)

MIT license



crates.io version Documentation Build Status

Numpy format (*.npy) serialization and deserialization.

NPY is a simple binary data format. It stores the type, shape and endianness information in a header, which is followed by a flat binary data field. This crate offers a simple, mostly type-safe way to read and write *.npy files. Files are handled using iterators, so they don't need to fit in memory.

npyz is a fork and successor of the seemingly-dead npy.

API Documentation on docs.rs


npyz = "0.8"

You also may be interested in enabling some features:

npyz = {version = "0.8", features = ["derive", "complex", "npz", "arrayvec", "half"]}

Data can now be read from a *.npy file:

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let bytes = std::fs::read("test-data/plain.npy")?;

    // Note: In addition to byte slices, this accepts any io::Read
    let npy = npyz::NpyFile::new(&bytes[..])?;
    for number in npy.data::<f64>()? {
        let number = number?;
        eprintln!("{}", number);

For further examples and information on:

  • Reading npy files,
  • Writing npy files,
  • Working with structured arrays,
  • Interop with the ndarray crate,
  • NPZ files and scipy sparse matrices,

please see the documentation on the root module.

Relation to similar crates

The name npyz is actually an abbreviation. Here is the full name of the crate:

npy plus npz support, and a lot of other features that are frankly a lot more important than npz—not to mention the fact that npz support isn't even actually included in the first release—but I had to call it something, okay

To clarify, npyz is a fork of Pavel Potoček's npy crate. The original npy supported structured arrays with derives, but had many, many limitations.

Compared to npy, npyz has:

  • Support for multidimensional arrays.
  • Support for big endian files.
  • Support for Complex, strings and bytestrings.
  • A more versatile reading API based on std::io::Read, so you can read directly from e.g. a zip::read::ZipFile. The original API required a &[u8], with the expectation that a user can use a memmap for large files.
  • Writing APIs based on std::io::Write.

Originally, nippy npyz was a place for me to protype new features with reckless abandon before finally making a PR to npy, but even my first few foundational PRs have yet to be merged upstream. I believe Pavel has a good head on their shoulders and a great attention to detail, and I appreciated their initial response on my PRs, but nearly two years have passed since the last time I have heard from them. Therefore, I've decided to go forward and publish the fork.


npyz is Copyright 2021 Michael Lamparski, and provided under the terms of the MIT License.

npyz is based off of npy. npy is Copyright 2018 Pavel Potoček, which was provided under the terms of the MIT License.


~89K SLoC