bison - BInary jSON

A file format which combines the portability of JSON with the performance of binary IO.

• Website: https://mbruno46.github.io/bison/
• Documentation: https://mbruno46.github.io/bison/
• Examples: tests
• Source code: https://github.com/mbruno46/bison/
• Bug reports: https://github.com/mbruno46/bison/issues

The philosophy

The file format is a combination of a header file written in JSON format, followed by streams of bytes. The design of the package, written primarily for Python, has been driven by a few simple concepts:

• portability and human readability of the JSON format, which allows to encode any structure, from simple basic types, such as lists, to complex custom-defined classes.

• the performance of binary for large datasets contained in a numpy.array

The package is primarily designed for Python, but MATLAB/Octave interpreters will be added in future versions.

Authors

Copyright (C) 2020-2023, Mattia Bruno

Installation

The latest version of the package can be installed using pip directly from the github server

pip install git+https://github.com/mbruno46/bison.git@main

or after downloading the git repository

git clone [email protected]:mbruno46/bison.git
cd bison
pip install .

Alternatively, since no precompilation is needed, one can simply add the package to the path

import sys
sys.path.append('path/to/bison')
import bison

Example

import numpy
import bison

db = {'name': 'Mattia', 'id': 0}
indices = [0,56,89,3,-45]
shape = (4,5,6)
array = numpy.arange(456).reshape(shape)

bison.save('full.dat',db,indices,shape,array)

bison.save('db.dat',db)

bison.save('array.dat',array,shape)

_array = bison.load('array.dat')