A bug was spotted in iris in which a single-point array of the type np.ma.MaskedConstant is not handled correctly by iris.util.new_axis (the resulting mask is not attributed with the new axis when the data is, so the point cannot be transposed; https://github.com/SciTools/iris/pull/2143).
A temporary fix has been implemented in which the MaskedConstant is turned into a MaskedArray which is then handled correctly, but this can only apply to loaded data; any lazy data in a single-point array will still be converted to a MaskedConstant and not get its extra dimension with new_axis.
A bug was spotted in iris in which a single-point array of the type
np.ma.MaskedConstant
is not handled correctly byiris.util.new_axis
(the resulting mask is not attributed with the new axis when the data is, so the point cannot be transposed; https://github.com/SciTools/iris/pull/2143).A temporary fix has been implemented in which the
MaskedConstant
is turned into aMaskedArray
which is then handled correctly, but this can only apply to loaded data; any lazy data in a single-point array will still be converted to aMaskedConstant
and not get its extra dimension withnew_axis
.This requires a fix in Biggus for the lazy data.