Should np.load() be in with with open-CodePudding

Traditionally in python one user with to open files such that the file automatically closes when you exit the inner scope.

with open('file.pickle', 'rb') as f:
    data = pickle.load(f)

In the numpy documentation it shows both np.load and with load

with load('foo.npz') as data:
    a = data['a']

a = np.load('/tmp/123.npy')

It looks like the with method only works with .npz files. Does a=np.load() automatically close the data file just like when the with context is closed?

CodePudding user response：

No, a=np.load() does not automatically closes the file. You can close it yourself by using a.close(). In fact, if you have .npz file saved with np.savez, then the load method will only fetch the archive info, but it will not load the entire file. Therefore, the file remains open.

You can use with context manager with both npz and npy by calling:

with np.load('foo.npz', allow_pickle=True) as data:
      a = data['a']

Same works with npy files.

CodePudding user response：

Using random files from my current work dir:

Using with with a npz:

In [51]: with np.load('f.npz') as f:print(f['b'])
['q' 'r' 's']

Using a 'plain' load with a npy:

In [55]: a = np.load('adict.npy',allow_pickle=True)
In [56]: a
Out[56]: 
array({'#Ineligiblevets': array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]), 'test': 'one'},
      dtype=object)

a is an array, so can't be "closed":

In [57]: a.close()
Traceback (most recent call last):
  File "<ipython-input-57-3980b52f6377>", line 1, in <module>
    a.close()
AttributeError: 'numpy.ndarray' object has no attribute 'close'

It is possible to open the file before hand, and pass via with, but you have to get the modes right, so it really doesn't help (unless you are doing something special):

In [59]: with open('adict.npy', 'rb') as f:
    ...:     a = np.load(f, allow_pickle=True)
    ...: 
In [60]: a
Out[60]: 
array({'#Ineligiblevets': array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]), 'test': 'one'},
      dtype=object)

Trying to use a npy load in a with doesn't work because an array does not have required enter/exit methods:

In [61]: with np.load('adict.npy', allow_pickle=True) as data: a=data
Traceback (most recent call last):
  File "<ipython-input-61-1025c4c8c4a8>", line 1, in <module>
    with np.load('adict.npy', allow_pickle=True) as data: a=data
AttributeError: __enter__