[ferret_users] replace fillvalue with NaN ?

[Ryo Furue <furue@xxxxxxxxxx>]

Hi Ferret users,

I was wondering how to replace the "missing" value (a.k.a. "bad" value
and fillvalue) of an existing dataset with a NaN.

It seems you can replace it with anything except it's not obvious how
to use NaN.  I'm attaching a complete example at the end of this
message, but basically I tried this

   let newvar = ifv var then var else (0/0)
   set var/bad=NaN  newvar

which doesn't work.   Perhaps 0/0 doesn't generate a NaN ?

By "doesn't work", I mean that I saved the resultant variable to a
binary file and examined its contents to see if NaN was there.  It
wasn't.

I use

  FERRET v6.94
  Darwin 14.1.0 - 02/26/15

Regards,

Ryo

!--- replace with -123 ==> Works. ----
set data coads_climatology
let sst_gt20 = if sst gt 20 then sst
let newsst = ifv sst_gt20 then sst_gt20 else (-123)
set var/bad=-123 newsst
save/clobber/l=1/file=tmp.nc newsst
list/clobber/l=1/format=STREAM/file=tmp.bin newsst

!--- Try to replace with NaN ==> Doesn't work. ----
set data coads_climatology
let sst_gt20 = if sst gt 20 then sst
let newsst = ifv sst_gt20 then sst_gt20 else (0/0)
set var/bad=NaN newsst
save/clobber/l=1/file=tmp2.nc newsst
list/clobber/l=1/format=STREAM/file=tmp2.bin newsst