[Numpy-discussion] MA - math operations do not preserve fill_value

[Joe Van Andel]

Using Python 2.1 and Numeric 20.2.1 on Redhat Linux 7.1

Shouldn't masked arrays preserve the fill value of their operands, if
both operands have the same fill value?  Otherwise, if I want to
preserve the value of the fill_value, I have to write expressions like:


d=masked_values((a+b),a.fill_value()) 

Here's a demonstration of the problem:

>>> a = masked_values((1.0,2.0,3.0,4.0,-999.0), -999)
>>> b = masked_values((-999.0,1.0,2.0,3.0,4.0), -999)

>>> a
array(data =
 [   1.,   2.,   3.,   4.,-999.,],
      mask =
 [0,0,0,0,1,],
      fill_value=-999)
 
>>> b
array(data =
 [-999.,   1.,   2.,   3.,   4.,],
      mask =
 [1,0,0,0,0,],
      fill_value=-999)

>>> c=a+b
>>> c
array(data =
 [  1.00000002e+20,  3.00000000e+00,  5.00000000e+00,  7.00000000e+00,
        1.00000002e+20,],
      mask =
 [1,0,0,0,1,],
      fill_value=[  1.00000002e+20,])

>>> d=masked_values((a+b),a.fill_value())
>>> d
array(data =
 [-999.,   3.,   5.,   7.,-999.,],
      mask =
 [1,0,0,0,1,],
      fill_value=-999)
-- 
Joe VanAndel  	          
National Center for Atmospheric Research
http://www.atd.ucar.edu/~vanandel/
Internet: vanandel at ucar.edu