Re: [Numpy-discussion] numpy.ma bug

[Sasha <nd...@ma...>]

I would say it is an ma bug, but don't know how to fix it properly
without changing numpy scalar arithmetic to call an enhanced version
of __array__ and pass context in.

The core problem can be demonstrated by the following session:

>>> from numpy import *
>>> x =3D ma.array([1],mask=3D[1])

>>> int_(1)*x
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File ".../numpy/core/ma.py", line 622, in __array__
    raise MAError, \
numpy.core.ma.MAError: Cannot automatically convert masked array to
numeric because data
                          is masked in one or more locations.

Note that x*int_(1) works as expected.  This is so because python
dispatches multiplication to int.__mul__ rather than ma.array.__mul__
if int_(1) is the first multiplicand.

I've fixed a similar problem for array*ma.array case and array(1)*x
works in the current version of numpy.

I will not have time to work on this before the weekend, so if someone
is motivated enought to fix this bug before the upcoming release,
please take a look at
http://projects.scipy.org/scipy/numpy/wiki/MaskedArray ("Ufuncs and
Masked Arrays"). It should be straightforward to generalize that
approach for array scalars.

On 3/9/06, Eric Firing <ef...@ha...> wrote:
> Travis et al.,
>
> Jeff Whitaker found that the imaged_masked.py example (with the colorbar
> line uncommented) in matplotlib 0.87 triggered a numpy bug--the script
> works normally with Numeric and numarray.  He committed a workaround to
> mpl svn, but I suspect you may want to track it down and squash it in
> numpy before the next release.  It is not clear to me whether it is
> really in ma, or merely revealed by ma.  The attached script triggers
> the bug.
>
> Thanks.
>
> Eric
>
>
>
>