Can't handle dicts containing numpy arrays

[donjan]

dpath v2.0.1 (current pypi version) can't handle a dict with some numpy array further inside, even though I'm not dealing with the array directly.

Observed Behaviour

import numpy
import dpath.util

npa = numpy.array([42], dtype='uint64')
data = {'a': {'b': {'c': {'d': {'e': npa} }}}}
path = '/a/b/c'
dpath.util.get(data, path)

Yields:

[...]
  File "/[...]/venv-dev/lib/python3.8/site-packages/dpath/segments.py", line 48, in walk
    for k, v in kvs(obj):
  File "/[...]/venv-dev/lib/python3.8/site-packages/dpath/segments.py", line 16, in kvs
    return zip(range(len(node)), node)
TypeError: object of type 'numpy.uint64' has no len()

Expected Behaviour

Should yield:

 {'d': {'e': array([42], dtype=uint64)}}

[donjan]

As a workaround without third party libraries, one can use:

import numpy
from functools import reduce

npa = numpy.array([42], dtype='uint64')
data = {'a': {'b': {'c': {'d': {'e': npa} }}}}
path = '/a/b/c'
reduce(lambda a,b: a[b], path.strip('/').split('/'), data)

[berquist]

My hack:

import dpath.segments as dpaths
import numpy as np

_LEAF_TYPES = [bytes, str, int, float, bool, type(None), np.ndarray]


def leaf(thing: Any) -> bool:
    """
    Return True if thing is a leaf, otherwise False.
    leaf(thing) -> bool
    """
    # leaves = (bytes, str, int, float, bool, type(None))

    # return isinstance(thing, leaves)
    return isinstance(thing, tuple(_LEAF_TYPES))


dpaths.leaf = leaf

[berquist]

Oh, it looks like this was probably handled properly in #126 and the current solution is to install from the 2.x branch.

[moomoohk]

Hey Eric,
I've merged 2.x and deployed to PyPI. Please update your package and let me know if this issue still exists.

Thanks!