Struct NpzWriter

pub struct NpzWriter<W: Write + Seek> { /* private fields */ }

Writer for .npz files.

Note that the inner ZipWriter is wrapped in a BufWriter when writing each array with .add_array(). If desired, you could additionally buffer the innermost writer (e.g. the File when writing to a file) by wrapping it in a BufWriter. This may be somewhat beneficial if the arrays are large and have non-standard layouts but may decrease performance if the arrays have standard or Fortran layout, so it’s not recommended without testing to compare.

Example

use ndarray::{array, aview0, Array1, Array2};
use ndarray_npy::NpzWriter;
use std::fs::File;

let mut npz = NpzWriter::new(File::create("arrays.npz")?);
let a: Array2<i32> = array![[1, 2, 3], [4, 5, 6]];
let b: Array1<i32> = array![7, 8, 9];
npz.add_array("a", &a)?;
npz.add_array("b", &b)?;
npz.add_array("c", &aview0(&10))?;
npz.finish()?;

Implementations

impl<W: Write + Seek> NpzWriter<W>

pub fn new(writer: W) -> NpzWriter<W>

Create a new .npz file without compression. See numpy.savez.

pub fn new_compressed(writer: W) -> NpzWriter<W>

Creates a new .npz file with compression. See numpy.savez_compressed.

pub fn new_with_options(writer: W, options: SimpleFileOptions) -> NpzWriter<W>

Creates a new .npz file with the specified options.

This allows you to use a custom compression method, such as zstd, or set other options.

Make sure to enable the relevant features of the zip crate.

pub fn add_array<N, S, D>( &mut self, name: N, array: &ArrayBase<S, D>, ) -> Result<(), WriteNpzError>

where N: Into<String>, S::Elem: WritableElement, S: Data, D: Dimension,

Adds an array with the specified name to the .npz file.

Note that a .npy extension will be appended to name; this matches NumPy’s behavior.

To write a scalar value, create a zero-dimensional array using arr0 or aview0.

pub fn finish(self) -> Result<W, WriteNpzError>

Calls .finish() on the zip file and .flush() on the writer, and then returns the writer.

This finishes writing the remaining zip structures and flushes the writer. While dropping will automatically attempt to finish the zip file and (for writers that flush on drop, such as BufWriter) flush the writer, any errors that occur during drop will be silently ignored. So, it’s necessary to call .finish() to properly handle errors.