On Jul 8, 2009, at 3:18 AM, Scott Sinclair wrote: >> 2009/7/8 Robert Kern <[email protected]>: >> 2009/7/4 Stéfan van der Walt <[email protected]>: >>> Thanks, Scott. This should now be fixed in SVN. >> >> You should probably change that to asanyarray() before the masked >> array crowd gets upset. :-) > > I hadn't thought about that, but I'm don't think it matters in this > case. MaskedArray.nonzero() returns a tuple of ndarrays...
Except that a masked value is considered zero in that case: >>> a = ma.array([0,1,2,3],mask=[0,0,1,0]) >>> a.nonzero() (array([1, 3]),) Taking np.asarray instead of np.asanyarray loses the mask, and you end up with the wrong result. _______________________________________________ Numpy-discussion mailing list [email protected] http://mail.scipy.org/mailman/listinfo/numpy-discussion
