zeros_like return COO format even if DOK is asked

[jodemaey]

Bug description

Maybe it is normal behavior but imho it shoud not be. zeros_like function return COO format even if DOK is asked.

Example

In[1]: import numpy as np
In[2]: import sparse as sp
In[3]: a = np.random.random((10,10,10))
In[4]: s = sp.zeros_like(a, format='dok')
In[5]: s
Out[5]: <COO: shape=(10, 10, 10), dtype=float64, nnz=0, fill_value=0.0>

Note that even starting from a COO and asking for a DOK doesn't work:

In[6]: d = sp.zeros_like(s, format='dok')
In[7]: d
Out[7]: <COO: shape=(10, 10, 10), dtype=float64, nnz=0, fill_value=0.0>

Expected behavior

zeros_like should return DOK format if DOK format is asked.

System

• OS and version: Ubuntu 18.04.4 LTS (kernel 5.3.0-62-generic)
• Python 3.8.3
• sparse version: 0.10.0 (from conda-forge)
• NumPy version: 1.18.5
• Numba version: 0.50.1

Additional comment

I looked for a previous notice of this bug but found none. Sorry if it was already reported.
For the moment, I bypass this by using zeros (which works) instead.

[hameerabbasi]

Thanks for reporting this issue. I can reproduce and have a tentative fix in #381.

[hameerabbasi]

This should now be fixed in master.