Operating elementwise with numpy masked arrays

[canesin]

It is not working as suposed with masked arrays, it creats objects by elements.

https://gist.github.com/canesin/6072390

[hgrecco]

The new test case shows that the bug is because MaskedArray (and not Quantity) is handling the multiplication. When the order of the operation is reversed, everything works as expected.

[canesin]

It looses the masked behaviour.

[hgrecco]

Multiplying meter by a masked array works as expected:

>>> from pint import UnitRegistry
>>> ureg = UnitRegistry()
>>> import numpy as np
>>> x = np.ma.masked_array([1, 2, 3], mask=[True, True, False])
>>> q = ureg.meter * x
>>> q
<Quantity([-- -- 3], 'meter')>
>>> q.magnitude
masked_array(data = [-- -- 3],
             mask = [ True  True False],
       fill_value = 999999)

Is the converse that is not working properly:

>>> q = x * ureg.meter
>>> q
masked_array(data = [-- -- <Quantity(3, 'meter')>],
             mask = [ True  True False],
       fill_value = 999999)

Can you confirm this in your system?

The problem is that the multiplication is always handled by the left operand. I think that the bug is that the ufunc is not honoring of the array_priority value. I am trying to confirm this with the numpy people.

[canesin]

Yes, I can confirm that it works with multiplication by left. thanks.

[hgrecco]

Review once that #905 is implemented