znpy provides Npy file reading and writing implementation in Zig.
In my recent project that involves working with large vector datasets, our old codebase use C++ implementation for reading npy files called cnpy. However, it does not support reading npy files with Fortran order arrays, slicing arrays, and I think the overall implementation is not very secure and robust. And since I decided to use Zig for the new codebase, I needed a Zig library that can read and write npy files efficiently. So I created znpy to fill this gap.
- Read and write Npy files
- Create memory-owned arrays
- Create array views from arrays with Numpy-style slicing
- Print arrays to stdout in structured array format
Make sure Zig 0.16 is installed.
Run demos:
zig build demo_<demo_name>Where <demo_name> is the name of a file in examples/ without the .zig extension (e.g., zig build demo_endian).
Run tests:
zig build testTo generate test npy files, you can use the provided Python script. Make sure to have uv installed.
Build the virtual environment and install dependencies:
uv syncRun the script to generate npy files:
uv run main.py