970813: ncdump output order for fortran

Michael,

>Date: Wed, 13 Aug 1997 07:51:43 -0700 (PDT) 
>From: Michael Watts <address@hidden>
>Organization: NASA/ARC
>To: Steve Emmerson <address@hidden>
>Subject: Re: 970812: ncdump output order for fortran 
>Keywords: 199708121910.NAA09780

In the above message, you wrote:

> Steve,
> The -b option doesnt change the indice order in the ncdump.

Odd.  The "-b" option of ncdump(1) changes the indicial order here:

    $ ncdump -b c orig_ran_test.cdf
    netcdf orig_ran_test {
    dimensions:
            slowdim = 9 ;
            fastdim = 5 ;
    variables:
            float fastdim(fastdim) ;
            float slowdim(slowdim) ;
            float var(slowdim, fastdim) ;
            float var_err(slowdim, fastdim) ;
            float var_df(slowdim, fastdim) ;
    data:

     fastdim = 1, 3, 5, 7, 9 ;

     slowdim = 1, 2, 3, 4, 5, 6, 7, 8, 9 ;

     var =
      // var(0, 0-4)
        0.158555, 0.03966331, -0.02857685, -0.03354263, -0.4338837,
      // var(1, 0-4)
        0.1449776, -0.1225681, 0.2001286, -0.04925156, 0.1827717,
      // var(2, 0-4)
        -0.2215099, 0.1867104, 0.04201126, 0.03138351, 0.1927376,
      // var(3, 0-4)
        0.07765579, 0.05680466, -0.271843, -0.06489563, -0.460289,
      // var(4, 0-4)
        0.2102165, -0.322094, 0.3012428, 0.3301468, 0.2914734,
      // var(5, 0-4)
        0.4745178, 0.3388901, -0.2899094, 0.1367798, -0.09538269,
      // var(6, 0-4)
        -0.3151398, -0.3511429, -0.1248245, 0.05485535, 0.3989182,
      // var(7, 0-4)
        0.1310806, -0.1886063, 0.300766, 0.06666565, -0.2103958,
      // var(8, 0-4)
        -0.07046509, 0.04106903, -0.04706573, -0.0001678467, -0.09223938 ;

    ...
    }

Note that the first line of the "var" data is indexed (0, 0-4).

    $ ncdump -b f orig_ran_test.cdf
    netcdf orig_ran_test {
    dimensions:
            slowdim = 9 ;
            fastdim = 5 ;
    variables:
            float fastdim(fastdim) ;
            float slowdim(slowdim) ;
            float var(slowdim, fastdim) ;
            float var_err(slowdim, fastdim) ;
            float var_df(slowdim, fastdim) ;
    data:

     fastdim = 1, 3, 5, 7, 9 ;

     slowdim = 1, 2, 3, 4, 5, 6, 7, 8, 9 ;

     var =
      // var(1-5 ,1)
        0.158555, 0.03966331, -0.02857685, -0.03354263, -0.4338837,
      // var(1-5 ,2)
        0.1449776, -0.1225681, 0.2001286, -0.04925156, 0.1827717,
      // var(1-5 ,3)
        -0.2215099, 0.1867104, 0.04201126, 0.03138351, 0.1927376,
      // var(1-5 ,4)
        0.07765579, 0.05680466, -0.271843, -0.06489563, -0.460289,
      // var(1-5 ,5)
        0.2102165, -0.322094, 0.3012428, 0.3301468, 0.2914734,
      // var(1-5 ,6)
        0.4745178, 0.3388901, -0.2899094, 0.1367798, -0.09538269,
      // var(1-5 ,7)
        -0.3151398, -0.3511429, -0.1248245, 0.05485535, 0.3989182,
      // var(1-5 ,8)
        0.1310806, -0.1886063, 0.300766, 0.06666565, -0.2103958,
      // var(1-5 ,9)
        -0.07046509, 0.04106903, -0.04706573, -0.0001678467, -0.09223938 ;
    ...
    }

Note that the first line of the "var" data is indexed (1-5 ,1) and not
(0, 0-4).

> Thus the
> confusion still reigns.  If it was written in c and read in fortran but
> ncdump doesnt indicate a change of order with the flags and you do not know
> which language it was written in then how do you know the order of the
> indeces to use using ncdump without doing the dim order from within the
> program?

Could you be confusing "indicial order" with the order in which the
elements are stored?  The storage order of the elements of an array is
language-independent: the "inner" dimension is always the most rapidly
varying one; the "outer" dimension is always the most slowly varying
one.  The C and Fortran languages have different representations for the
"inner" and "outer" dimensions but the netCDF package always stores the
data with the "inner" dimension varying most rapidly.

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