I once hacked away at NetCDF headers for the purpose of writing
pure Java classes for NetCDF. I seem to recall that the current value
of Unlimited occupies bytes 4-7, immediately following the magic number
"CDF/001". i.e. the first eight bytes of a NetCDF file are
byte # 0 4 7
contents | "CDF" /001 | <unlimited dimension> |
32-bit integer
>
> >
> > I have a job which bombed before the NCCLOS call could be made made.
> > I know it wrote many records before bombing, and the file is very
> > large, but an NCDUMP shows 0 records written. Is there a way to
> > "poke" a value into the header to allow me to access at least the
> > data that already went to the file? If so, would I subsequently be
> > able to continue extending the file along the record dimension?
> >
>
> I remember running into the same problem once, and doing a hack to
> save many years of a simulation. What I did was to go back to the
> netcdf source and disable the line which checked whether the requested
> time range exceeded the value of the (unlimited) time dimension.
> It was just a matter of commenting out 1 line, but I cannot really
> remember where it was. It worked, though. It would be nice to have
> a way of doing this through the API. Russ?
>
> Arlindo.
>
>
> --
> Arlindo da Silva
> dasilva@xxxxxxxxxxxxx
>
--
|||| | | | | | | | | | | | | | | | | | | | | | | ||||
William Weibel weibel@xxxxxxxxxxxxxx
2635 S. Bedford St. Tel. (310)559-8510 \\\\/
Los Angeles, CA 90034-2413 Cel (310)351-4155 O-O
|
-
... to know even one life has breathed easier because you have lived;
this is to have succeeded. --Ralph Waldo Emerson
|||| | | | | | | | | | | | | | | | | | | | | | | ||||
