numpy.ma.array ignores mask when asembling masked array inputs

[AndrewStraw]

The following test fails with r5283. I assume (perhaps wrongly) that this is a bug. It appears numpy.ma.array has an argument keep_mask, which defaults to True. That implied to me that the expected behavior would have passed this test.

{{{import numpy

x = numpy.arange(10) y = numpy.arange(10)+20 z = numpy.arange(10)+40 x = numpy.ma.masked_where(x>5,x)

X1=numpy.ma.array([x,y,z]) X2=numpy.ma.vstack([x[numpy.newaxis,:],

y[numpy.newaxis,:], z[numpy.newaxis,:]])

assert numpy.ma.allclose(X1.mask,X2.mask) }}}

[pierregm]

In SVN5284, a mask is now constructed from a sequence of masked arrays parsed as input. Please give it a try and let me know whether we can close the ticket.

[AndrewStraw]

The change passes my test and I see you included a better one, too. I did note a couple typos in the docstring to the test in r5284 (should be "creating" and "a list"). Despite that nit, I think this bug is closed. Thanks. (Also, apologies for the formatting of my snippet above... Next time I'll be sure to preview.)

[pierregm]

OK, excellent ! Thanks a lot for the feedback, I'm closing the tickets (docstrings are another matter...)