[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20010208: Writing out an integer array: dimensional ordering

Stefan,

> To: address@hidden
> From: Stefan Maehlmann <address@hidden>
> Subject: Writing out an integer array
> Organization: UCAR/Unidata
> Keywords: 200102081400.f18E0lL04445

The above message contained the following:

> i have an integer array A[3,4]=
> a11 a12 a13 a14 
> a21 a22 a23 a24 
> a31 a32 a33 a34
> 
> which i've stored as 'integer A(IROW,ICOL);' in a netcdf-file
> if i want to extract it from the file 
> i would do the following: 
> 
> NF_INQ_DIMID(NCID,'IROW',IROW)
> NF_INQ_DIMID(NCID,'ICOL', ICOL)
> to get the dimensions
> 
> ISTART=1
> ICOUNT(1)=IROW
> ICOUNT(2)=ICOL
> and then read the array as 
> NF_GET_VARA_INT (NCID,VAR_ID,ISTART,ICOUNT,NWI(1:IROW,1:ICOL))
> 
> but somehow now the rows and columns have swapped. Can you help me to 
> solve this problem? 

Use the following instead:

    ISTART(1)=1
    ISTART(2)=1
    ICOUNT(1)=ICOL
    ICOUNT(2)=IROW

In the Fortran netCDF API, the "start" and "count" arguments must
be vectors.  The first element of each vector corresponds to the
innermost dimension (the most rapidly varying one) and the last element
coresponds to the outermost dimension (the most slowly varying one).

In the C netCDF API, the ordering of the dimensions is reversed (i.e.
the first vector elements corresponds to the innermost dimension).

> A second question is how the write the array back to the file so i can
> read it properly with ncdump (means that it looks like the first form).

A good way to see this would be to give the following CDL file
to the "ncgen" utility using the "-f" option to generate Fortran code.
(NOTE: CDL uses the C convention for dimensional order, which is the
reverse of the Fortran convention).

    netcdf A {
        dimensions:
            FortranRowCount = 3;
            FortranColCount = 4;
        variables:
            int a(FortranColCount,FortranRowCount); // C ordering convention
        data:
            a =
                11, 21, 31,    // 1st Fortran column
                12, 22, 32,    // 2nd Fortran column
                13, 23, 33,    // 3rd Fortran column
                14, 24, 34;    // 4th Fortran column
    }

> Sincerely,
> Stefan Maehlmann

Regards,
Steve Emmerson   <http://www.unidata.ucar.edu>