get/put_vara usage?

[Wei-keng Liao]

Hi, Brian

The inq APIs are for reading metadata from an existing file.
The "lon" and "lat" dimensions only exist in the file to be read, but
not in the file to be created/written. So, your changes to use def_dim
APIs make sense, which create dimension metadata in the output file.

In fact, the call below before your changes should return an error
code NC_EBADDIM meaning "Invalid dimension id or name". 
   ncmpi_inq_dimid(ofh, "lon", &dims[0]);

I suggest you check the return values for the APIs you call.
Something like below can be useful.
  err = ncmpi_inq_dimid(ofh, "lon", &dims[0]);
  if (err != NC_NOERR) printf("Error: %s\n", ncmpi_strerror(err));


Wei-keng

On Nov 27, 2012, at 1:37 PM, Smith, Brian E. wrote:

> Well, this change seems to make it a lot happier:
> 
>   ncmpi_def_dim(ofh, "lon", lon, &dims[0]);
>   ncmpi_def_dim(ofh, "lat", lat, &dims[1]);
> 
> I'm not entirely sure why though, because I did have:
> ncmpi_inq(fh, &num_dims, &num_vars, &num_attrs, &num_unlim);
> for(i = 0; i < num_dims; i++)
> {
>   ncmpi_inq_dim(fh, i, dimname, &dlen);
>   ncmpi_def_dim(ofh, dimname, dlen, &dim_id); 
> }
> as part of setup_global_dims_attrs(), so shouldn't the inq_dim on the output file be ok?
> 
> 
> On Nov 27, 2012, at 2:12 PM, Smith, Brian E. wrote:
> 
>> I am new to (p)netCDF and am having trouble trying to make use of get/put_vara to extract "slices" of a 4D variable from some data I've been given. I've read the docs and looked at the examples but things still aren't working for me.
>> 
>> The variable looks like: var[time][elev][lat][lon] with time being an "unlimited" variable
>> 
>> My code is essentially:
>> 
>>  open_files(&fh, &ofh); // open the input file for read, open the output file for write.
>>  setup_global_dims_attrs(fh, ofh); // inq_var then def_var lat/lon/elev/time then copies their atributes
>> 
>>  /* Define my output variable dimensions */
>>  ncmpi_inq_dimid(ofh, "lon", &dims[0]);
>>  ncmpi_inq_dimid(ofh, "lat", &dims[1]);
>>  ncmpi_def_dim(ofh, "plev", elev2-elev1, &dims[2]);
>>  ncmpi_def_dim(ofh, "time", time2-time1, &dims[3]);
>>  ncmpi_def_var(ofh, outname, type, ndims, dims, &outid);
>> 
>>  ncmpi_enddef(ofh);
>> 
>>  copy_global_dims(fh, ofh); //inq var on input, inq var on output, get_var on input, put_var on output for lat/lon/elev/time variables
>> 
>> MPI_Offset start[4], count[4], outstart[4];
>> start = {time1, elev1, 0, 0};
>> count  = (time2-time1, elev2-elev1, lat, lon};
>> outstart = {0,0,0,0};
>> total = count[0] * count[1] * count[2] * count[3];
>> float data = malloc(sizeof(float) * total);
>> ncmpi_get_vara_float_all(fh, varid, start, count, data);
>> ncmpi_put_vara_float_all(ofh, varid, outstart, count, data);
>> cllose_files(fh, ofh)
>> 
>> 
>> However, this just results in junk in the output file. There is no practical way to verify the read worked correctly unfortunately, so I can't tell if the read or the write call is incorrect. 
>> 
>> I would certainly appreciate any help.
>> 
>> Thanks.
>> 
>