Update shape attr of DOK/COO

[Jan Vaillant]

Description
I'd like to resize/reshape a sparse array along one or more dimensions.

Example Code

s = sparse.COO(np.reshape(np.arange(9.), (3,3)), fill_value=np.nan)
s.todense()

array([[0., 1., 2.],
       [3., 4., 5.],
       [6., 7., 8.]])

s.shape = (4,4)
s.todense()

array([[ 0.,  1.,  2., nan],
       [ 3.,  4.,  5., nan],
       [ 6.,  7.,  8., nan],
       [nan, nan, nan, nan]])

This is exactly the result I need but I was wondering if it is OK to just update the shape attr of a sparse array. Is it? It is not mentioned in the docs so I am not sure if shape is intended to be read-only or if this is the proper way to reshape a COO/DOK array. The result of s.resize((4,4)) is not what what I need. I my case it will just grow and never shrink.

Thank you
Jan

[Hameer Abbasi]

Hello, the short answer is it works for now, but it may or may not work in the future. I'll try to write a somewhat longer answer later today.