Re: Here is a sample dataset

• To: "Raymond E. Flanery, Jr." <flaneryrejr@xxxxxxxx>
• Subject: Re: Here is a sample dataset
• From: Steve Emmerson <steve@xxxxxxxxxxxxxxxx>
• Date: Fri, 09 Apr 1999 14:26:59 -0600

Raymond,

>Date: Fri, 09 Apr 1999 15:27:39 -0400
>From: "Raymond E. Flanery, Jr." <flaneryrejr@xxxxxxxx>
>Organization: ORNL/AVRC
>To: Steve Emmerson <steve@xxxxxxxxxxxxxxxx>
>Subject: Here is a sample dataset
>Keywords: 199904091928.NAA23657

In the above message, you wrote:

> the first 3 values on each line are the coordinates x, y and z. the
> remaining couple of values are data at the point.
> 
> Ray Flanery
> 
> -4.7793452e-01  5.7826381e-01 -5.5102379e-06 -5.0000000e+01  1.3280000e+00  
> 1.4190000e-01
> -4.7450841e-01  5.2110161e-01 -5.4707375e-06 -4.5000000e+01  1.7830000e+00  
> 1.6660000e-01

...

If the above data is turned into the obvious netCDF dataset:

    netcdf dataset {
    dimensions:
        recnum = UNLIMITED;
    variables:
        float   x;
        float   y;
        float   z;
        float   u;
        float   v;
        float   w;
    data:
        ...
    }

and then ingested by the netCDF backend to VisAD, then it would appear
as the following FlatField:

    recnum -> (x, y, z, u, v, w)

which is probably not what you want.  Undoubtably, you want a VisAD
FlatField like the following:

    (x, y, z) -> (u, v, w)

Unfortunately, the "plain" netCDF conventions don't carry enough
semantics with them to distinguish independent variables from dependent
ones, which is why the first form of FlatField is created.

Upon getting the first VisAD FlatField, however, you could easily
transform it into the second one using the VisAD API.  This shouldn't
take more than a few lines of code -- depending on just how generic you
want the transformation code to be.

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