NumPy

Note

Install with pip install labox[numpy]

Array Serializer

The NpySerializer provides a serializer implementation for NumPy arrays using the native .npy format. This serializer leverages NumPy's built-in numpy.save and numpy.load functions to efficiently serialize and deserialize numpy.ndarray objects.

Basic Usage

A default instance of the serializer is available as npy_serializer:

import numpy as np

from labox.extra.numpy import npy_serializer

arr = np.array([[1, 2, 3], [4, 5, 6]])

serialized_data = npy_serializer.serialize_data(arr)

You can also create a custom instance:

from labox.extra.numpy import NpySerializer

serializer = NpySerializer(
    load_args={"mmap_mode": "r"},
)

Numpy supports pickling native Python objects found within arrays with an allow_pickle argument but this has been intentionally disabled because pickled objects both pose a security risk and are not a stable long-term storage format.

Content Type

The serializer uses the content type "application/x-npy.v3" to identify .npy format data, following the NumPy file format specification version 3.0.