How can I find out if a n-dimensional numpy array Arr is contiguous in C-style or Fortran-style?
2 Answers
The numpy documentation states that it is possible to check whether an array is C-contiguous or Fortran-contiguous via the attribute flags:
Arr.flags['C_CONTIGUOUS']
Arr.flags['F_CONTIGUOUS']
These attributes return a boolean indicating which of the two cases is true.
5 Comments
norok2
Excepts it is not a function, it is actually an attribute of the
numpy.ndarray object, as your reported syntax (which is correct) suggests.
Ethunxxx
@norok2, you are right! I edited the answer. Thanks!
hpaulj
And with some axis swapping it possible that neither will be True. And both are True for a 1d array
matanox
there is also numpy.ascontiguousarray which may sometimes be helpful.
Daniel Wagner
"And both are True for a 1d array" -- not always. For example,
np.zeros(30)[::3] has False for both flags.You can also try the ndarray.data.contiguous member. E.g. (on my machine):
arr = np.arange(6).reshape(2, 3)
print(arr.data.contiguous) # True
print(arr.data.c_contiguous) # True
print(arr.data.f_contiguous) # False
(I can't find any information re: which numpy versions support this, even on their docs. Any leads welcome in the comments!)
1 Comment
flexatone
Note: this will not work with
datetime64 arrays; you get ValueError: cannot include dtype 'M' in a buffer
strides.