69

What happened is that I (by mistake) saved a dictionary with the command numpy.save() (no error messages shown) and now I need to recover the data in the dictionary. When I load it with numpy.load() it has type (numpy.ndarray) and is 0-d, so it is not a dictionary any more and I can't access the data in it, 0-d arrays are not index-able so doing something like 

mydict = numpy.load('mydict')
mydict[0]['some_key']

doesn't work. I also tried

recdict = dict(mydict)

but that didn't work either.

Why numpy didn't warn me when I saved the dictionary with numpy.save()?

Is there a way to recover the data?

Thanks in advance!

Alessandro Vendruscolo
  • 14,493
  • 4
  • 32
  • 41
andres
  • 1,079
  • 1
  • 10
  • 17

3 Answers3

100

Use mydict.item() to obtain the array element as a Python scalar.

>>> import numpy as np
>>> np.save('/tmp/data.npy',{'a':'Hi Mom!'})
>>> x=np.load('/tmp/data.npy')
>>> x.item()
{'a': 'Hi Mom!'}
unutbu
  • 842,883
  • 184
  • 1,785
  • 1,677
32

0-d arrays can be indexed using the empty tuple:

>>> import numpy as np
>>> x = np.array({'x': 1})
>>> x
array({'x': 1}, dtype=object)
>>> x[()]
{'x': 1}
>>> type(x[()])
<type 'dict'>
Robert Kern
  • 13,118
  • 3
  • 35
  • 32
3

It's a way to recover the dict type data that using 'allow_pickle=True' and '.tolist()'.
I hope it will help you.

numpy.save('mydict.npy', org_dict)

# load & convert type to dict
mydict = numpy.load('mydict.npy', allow_pickle=True)
re_dict = mydict.tolist()

print('re_dict :', type(re_dict), re_dict)   # 're_dict' is '<type dict>'
WangSung
  • 259
  • 2
  • 5