How do I create a 0 x 0 (i.e. ndim= 2, shape= (0,0)) float numpy.ndarray?
Add a comment
|
2 Answers
>>> import numpy as np
>>> a = np.empty( shape=(0, 0) )
>>> a
array([], shape=(0, 0), dtype=float64)
>>> a.shape
(0, 0)
>>> a.size
0
The array above is initialized as a 2D array--i.e., two size parameters passed for shape.
Second, the call to empty is not strictly necessary--i.e., an array having 0 size could (i believe) be initialized using other array-creation methods in NumPy, e.g., NP.zeros, Np.ones, etc.
I just chose empty because it gives the smallest array (memory-wise).
2 Comments
Johan Råde
You mention it could also be done with NP.array, NP.zeros, NP.ones etc. I see how to do it with Np.zeros and NP.ones, but not how to do it with NP.array.
doug
so for instance, NP.array([]); though the shape of this array is (0,); your Q specifies a 2D array. In my answer i just wanted to make the point that other array constructors would do the job as well, not just empty. I'll edit to remove 'array', given you need a 2D array.
All of the functions that return an array given shape/size can create a 0x0 array (in fact, of any dimension).
np.full((0,0), 0.0)
np.random.rand(0,0)
np.random.randint(0, size=(0,0))
np.random.choice(0, size=(0,0))
# etc.
To create the same with np.array, the shape attribute of an empty array could be modified.
a = np.array([])
a.shape += (0,)
a # array([], shape=(0, 0), dtype=float64)