Handle encoding for numpy array scalars #18
Description
The json_tricks library will serialize instances of numpy.ndarray but if you pass it an array scalar (e.g., a single element of an array, which have datatypes like numpy.int64, numpy.float64, etc.)
For example:
>>> z = np.array([1,2,3]) >>> json_tricks.dumps(z) '{"dtype": "int64", "shape": [3], "__ndarray__": [1, 2, 3]}' >>> json_tricks.dumps(z[0]) TypeError: Object of type <class 'numpy.int64'> could not be encoded by TricksEncoder using encoders [<function json_date_time_encode at 0x7f38028fea60>, <function class_instance_encode at 0x7f38028febf8>, <function json_complex_encode at 0x7f38028fec80>, <function numpy_encode at 0x7f3802410400>]My present workaround is to ensure that all array scalars are converted to numpy arrays before attempting to serialize (json_tricks.dumps(np.array(z[0])) in the above) but it would be easier if this was handled automatically.