Applying masks on matrices with Nan/Inf on them

[Joao Cardoso]

Hi,

How can I apply a mask to a matrix that has NaNs or Inf on it?
For ordinary real numbers, I can do:

octave:122> a=rand(3)
a =

  0.439150  0.426345  0.834121
  0.711628  0.927248  0.055622
  0.091911  0.265919  0.239936

octave:123> (a>0.5) .* a
ans =

  0.00000  0.00000  0.83412
  0.71163  0.92725  0.00000
  0.00000  0.00000  0.00000

but as any operation on Nan or Inf retuns a Nan or Inf,

octave:124> a(2,3)=nan
a =

  0.439150  0.426345  0.834121
  0.711628  0.927248       NaN
  0.091911  0.265919  0.239936

octave:125> (!isnan(a)) .* a
ans =

  0.439150  0.426345  0.834121
  0.711628  0.927248       NaN
  0.091911  0.265919  0.239936

Is the bellow approach the only one? I don't like to change defaults in
my scripts...

octave:130> do_fortran_indexing = 1;

octave:131> a(find(isnan (a)))= 0
a =

  0.43915  0.42634  0.83412
  0.71163  0.92725  0.00000
  0.09191  0.26592  0.23994

Thanks,
Joao


-- 
Joao Cardoso                |   e-mail: address@hidden
INESC, R. Jose Falcao 110   |   tel:    + 351 2 2094322
4050 Porto, Portugal        |   fax:    + 351 2 2008487