Re: 20030702:nf_get_var - empty data

Alex,

> Thanks for your clarifications. So one can compare the values against 
> NF_FILL_DOUBLE,
> say, to check that these values were actually written. But is NF_FILL_DOUBLE 
> the same
> on all platforms?

Yes, it's defined in the appropriate include files (netcdf.h,
netcdf.inc, or netcdf.mod, depending on the language).  The values
were selected to be exactly representable on any platform on which
netCDF could be installed, so that you can use an exact equality test
even for floating point or double values, if you use the default fill
values.

> PS. There seems to be a contradition regarding the use of NF_SET_FILL in the 
> doc:
> 
> Usage
> 
> INTEGER FUNCTION NF_SET_FILL(INTEGER NCID, INTEGER FILLMODE,
>                              INTEGER old_mode)
> 
>  NCID
>           NetCDF ID, from a previous call to NF_OPEN or NF_CREATE.
>  FILLMODE
>           Desired fill mode for the dataset, either NF_NOFILL or NF_FILL.
>  old_mode
>           Returned current fill mode of the dataset before this call, either 
> NF_NOFILL
> or NF_FILL.
> 
> ... later...
> 
> OMODE = NF_SET_FILL(NCID, NF_NOFILL)

Oops, you're right.  It looks like it was OK in the C and Fortran90
Users Guides, but wrong in the Fortran77 guide.  Thanks for pointing
this out.  I've corrected the online HTML copy, but it will take
longer to get the PDF and PostScript fixed from the FrameMaker source
...

--Russ