scipy.io.mmwrite raises type error for uint16

[StefanTheWiz]

mmwrite doesn't seem to support unsigned integers. Is there a reason for that?
I could use int32 instead, but using an uint16 would use less memory and prevent accidentally adding negative values.

Reproducing code example:

from scipy.sparse import coo_matrix
import scipy.io
import numpy as np
#  0 , 0,  0
# 88, 99, 11
#  0, 22, 33
matrix = coo_matrix(( [88,99,11,22,33], ( [1,1,1,2,2], [0,1,2,1,2] ) ), dtype=np.uint16 )
scipy.io.mmwrite( 'matrix', matrix )

Error message:

Traceback (most recent call last):
  File "process_data.py", line 161, in <module>
    scipy.io.mmwrite( 'matrix', matrix )
  File "/usr/lib64/python3.5/site-packages/scipy/io/mmio.py", line 102, in mmwrite
    MMFile().write(target, a, comment, field, precision, symmetry)
  File "/usr/lib64/python3.5/site-packages/scipy/io/mmio.py", line 448, in write
    self._write(stream, a, comment, field, precision,symmetry)
  File "/usr/lib64/python3.5/site-packages/scipy/io/mmio.py", line 683, in _write
    raise TypeError('unexpected dtype kind ' + kind)
TypeError: unexpected dtype kind u

Link to scipy/io/mmio.py", line 683 - doesn't seem to have a case for unsigned integers

Scipy/Numpy/Python version information:

# scipy, numpy, python versions
1.0.0 1.13.3 sys.version_info(major=3, minor=5, micro=4, releaselevel='final', serial=0)

[kshitij12345]

Can I please take this up??

[Tokixix]

@ilayn Can I get info if kshitij12345 had fixed this issue? If not, I can do it.

[ilayn]

We don't really know if the issue is picked up or not. Let's try pinging first @kshitij12345

[kshitij12345]

@Tokixix @ilayn I have submitted a pull request. If someone can review and guide me, that would be great!