drkjam / numpyson

serializer extending jsonpickle to deal with numpy/pandas types

numpyson

cross py2/py3 python object serializer, currently using and extending jsonpickle to dump/load python objects with a primary focus on numpy/pandas types.

The main purpose of numpyson is to allow dumping under a python3 interpreter and loading under python2 and vice versa. This is a use case that is not covered by current serializers AFAIK.

quick simple API example:

import numpyson
import numpy
s = numpyson.dumps(numpy.array([1,2,3]))
data = numpyjson.loads(s)
assert data.to_list() == [1,2,3]

Currently supported:

• numpy arrays and some other numpy types
• pandas.TimeSeries
• pandas.DataFrame
• pandas.DateTimeIndex
• pandas.Int64Index
• pandas.Float64Index
• pandas.Index
• nested python data structures with the above types