I have a 2-dimensional array that represents a mask of a 3-dimensional array, and can be broadcast as such. e.g.:
>>> mask.shape
(101, 100)
>>> cube.shape
(500, 101, 100)
What is the best way to create a broadcastable object like mask (which is an array) that can be indexed with the same views as cube, returning the same mask? i.e.:
>>> cube[100,:,:]
<some image>
>>> mask[100,:,:]
<mask>
so mask[n,:,:] would return mask for any n, or better yet any n that could be used to index cube.
Importantly, I want to do this without making mask larger in memory (e.g., by doing bigger_mask = np.ones([500,1,1])*self._mask[None,:,:])
Something like this?
>>> from numpy.lib.stride_tricks import as_strided
>>> mask = np.random.randint(2, size=(101, 100)).astype(bool)
>>> mask_view = as_strided(mask, shape=(500,)+mask.shape,
... strides=(0,)+mask.strides)
>>> mask_view.shape
(500, 101, 100)
>>> np.array_equal(mask_view[0], mask_view[499])
True
>>> np.all(mask_view == 0)
False
>>> mask[:] = 0
>>> np.all(mask_view == 0)
True