[OctDev] Missing values

[Teemu I. <tpi...@pc...>]

Hi all,

Support for missing double values (NA) was added to Octave two years ago.
There was talk of making functions like sum, mean, etc. to support them, but
e.g. sum([1,NA]) still returns NA. 
(see http://www.octave.org/mailing-lists/help-octave/2002/581)

Also, min([1,NA]) returns 1, but so does min([1,NaN]), which would imply
that NA = NaN in the current implementation. Is this because no one has
fixed the code, or is it so by design?

Octave-forge includes functions nanmax, nanmean etc. which ignore NaN as
they should, but return NA on any vector containing NA. Should they be made
to also ignore NA?

Teemu