cajeta.math.npio.Npy — NumPy .npy binary array I/O. Reads and writes the NumPy v1.0 format: magic + version + header, then the raw little-endian element bytes in C-order. Four dtypes are supported, each a typed encode/decode pair over in-memory byte images plus a save/load pair over files (cajeta is statically typed, so the caller picks the variant matching the dtype): float32 '<f4', float64 '<f8', int32 '<i4', int64 '<i8'. save* is encode* + File.writeAllBytes; load* is File.readAllBytes + decode* — use the encode/decode forms to keep the .npy image in memory (sockets, archives) instead of touching disk.

package snip.npy;

import cajeta.math.Tensor;
import cajeta.math.npio.Npy;

public final class Demo {
    public static void run() {
        Tensor<float32> t #= Tensor.arange<float32>(6);
        Npy.saveF32("data.npy", t);
        Tensor<float32> back #= Npy.loadF32("data.npy");
        return;
    }
}

Methods

Signature
`static #int8[] encodeF32(Tensor<float32> t)`	Serialize a `float32` tensor to its full `.npy` byte image (`'<f4'`)
`static #Tensor<float32> decodeF32(int8[] buf)`	Decode a `float32` `.npy` byte image (starting at offset 0)
`static void saveF32(String path, Tensor<float32> t)` ⚑	Save a `float32` tensor to `path` as a `.npy` (`'<f4'`, C-order)
`static #Tensor<float32> loadF32(String path)` ⚑	Load a `float32` `.npy` from `path`
`static #int8[] encodeF64(Tensor<float64> t)`	Serialize a `float64` tensor to its full `.npy` byte image (`'<f8'`)
`static #Tensor<float64> decodeF64(int8[] buf)`	Decode a `float64` `.npy` byte image
`static void saveF64(String path, Tensor<float64> t)`	Save a `float64` tensor (`'<f8'`, C-order)
`static #Tensor<float64> loadF64(String path)`	Load a `float64` `.npy`
`static #int8[] encodeI32(Tensor<int32> t)`	Serialize an `int32` tensor to its full `.npy` byte image (`'<i4'`)
`static #Tensor<int32> decodeI32(int8[] buf)`	Decode an `int32` `.npy` byte image
`static void saveI32(String path, Tensor<int32> t)`	Save an `int32` tensor (`'<i4'`, C-order)
`static #Tensor<int32> loadI32(String path)`	Load an `int32` `.npy`
`static #int8[] encodeI64(Tensor<int64> t)`	Serialize an `int64` tensor to its full `.npy` byte image (`'<i8'`)
`static #Tensor<int64> decodeI64(int8[] buf)`	Decode an `int64` `.npy` byte image
`static void saveI64(String path, Tensor<int64> t)`	Save an `int64` tensor (`'<i8'`, C-order)
`static #Tensor<int64> loadI64(String path)`	Load an `int64` `.npy`

⚑ = @EntryPoint

Npy

Methods

See also