Module numeric::io

Saving and loading data to and from disk.

HDF5

The recommended way to save/load data in Numeric is using HDF5.

Note: The HDF5 library will by default not be thread-safe (it depends on how you compiled it), so do not call either of these functions concurrently.

Saving to HDF5 file:

use std::path::Path;
use numeric::Tensor;

let path = Path::new("output.h5");
let t: Tensor<i32> = Tensor::range(100);
let ret = t.save_hdf5(&path);

The data will be saved to the group /data.

Loading from HDF5 file

Now, we can load this file:

use std::path::Path;
use numeric::Tensor;

let path = Path::new("output.h5");
let t = match numeric::io::load_hdf5_as_f64(&path, "/data") {
    Ok(v) => v,
    Err(e) => panic!("Failed: {}", e),
};

Note that since we need to know the type of t at compile time, it doesn't matter that we saved the file as i32, we have to specify how to load it. The way this is done is that it will load the i32 natively and then convert it to f64. If your data converted, you simply have to load it as the same type as you know is in the file.

Functions

load_hdf5_as_f32	Load HDF5 file and convert to specified type.
load_hdf5_as_f64	Load HDF5 file and convert to specified type.
load_hdf5_as_i8	Load HDF5 file and convert to specified type.
load_hdf5_as_i16	Load HDF5 file and convert to specified type.
load_hdf5_as_i32	Load HDF5 file and convert to specified type.
load_hdf5_as_i64	Load HDF5 file and convert to specified type.
load_hdf5_as_isize	Load HDF5 file and convert to specified type.
load_hdf5_as_u8	Load HDF5 file and convert to specified type.
load_hdf5_as_u16	Load HDF5 file and convert to specified type.
load_hdf5_as_u32	Load HDF5 file and convert to specified type.
load_hdf5_as_u64	Load HDF5 file and convert to specified type.
load_hdf5_as_usize	Load HDF5 file and convert to specified type.