Bug in ncgen handling of multidimensional character variables

[Russ Rew <russ@xxxxxxxxxxxxxxxx>]

Hi,

An undocumented feature of ncgen in netCDF-2 was that null padding was
added to strings that were shorter than the last dimension of a
multidimensional character variable, instead of just concatenating them.
This permitted using strings of different lengths in an array in CDL,
for example:

    netcdf bug {
    dimensions:
            scan = 4 ;
            string_len = 10 ;
    variables:
            char name(scan, string_len) ;
    data:

     name
      "Methane",
      "Ammonia",
      "Water",
      "Acetylene" ;
    }

In versions 3.3 and 3.3.1, unfortunately this feature no longer works.
Instead the strings are just concatenated, as they would be with a
1-dimensional variable.  For example, when the above CDL file is
converted to a netCDF file using the "-b" option to ncgen, running
ncdump on the result produces:

    netcdf bug {
    dimensions:
            scan = 4 ;
            string_len = 10 ;
    variables:
            char name(scan, string_len) ;
    data:

     name
      "MethaneAmm",
      "oniaWaterA",
      "cetylene",
      "" ;
    }

A patch is now available that fixes the problem, from

    http://www.unidata.ucar.edu/packages/netcdf/patch-3.3.1-2

The problem description and a link to the patch have been added to the
"Known Problems" page for netCDF at

    http://www.unidata.ucar.edu/packages/netcdf/known_problems.html

The fix will be included in the next minor release.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
russ@xxxxxxxxxxxxxxxx                     http://www.unidata.ucar.edu