Re: composite & derived variables

[Jonathan Callahan <callahan@xxxxxxxxxxxxx>]

Matt,

To get composite variables to work, the <link match="..."> contents need
to reflect the XML hierarchy.  For example, in our coads_climatology XML
file we have:

   <datasets>
    <coads_climatology_cdf name="COADS Climatology"
      url="file:coads_climatology"
      doc="doc/coads_climatology.html">
     <variables>
      ...
      <uwnd name="Zonal Wind" units="M/S">
       <link match="/lasdata/grids/coads_climatology_cdf_grid"/>
      </uwnd>
      <vwnd name="Meridional Wind" units="M/S">
       <link match="/lasdata/grids/coads_climatology_cdf_grid"/>
      </vwnd>
      ...
     </variables>
     <composite>
      <wind_vectors name="Wind vectors" js="VecVariable">
       <link
match="/lasdata/datasets/coads_climatology_cdf/variables/uwnd"/>
       <link
match="/lasdata/datasets/coads_climatology_cdf/variables/vwnd"/>
      </wind_vectors>
     </composite>
    </coads_climatology_cdf>
   </datasets>


The hierarchy:

  <lasdata>
   <datasets>
    <coads_climatology_cdf
     <variables>
      <uwnd name

is reflected in the link match as:

  /lasdata/datasets/coads_climatology_cdf/variables/uwnd


Your second problem is duly noted.  Please remember that differencing is
not a released feature yet.  I am currently modifying LAS so that the
initialization process will open datasets and then call initialization
scripts instead of the other way around.  When I am done, this will work
for data comparison.


-- Jonathan Callahan



Matthew Harrison wrote:
> 
> Greetings,
> 
> I'm using LAS4.0.  It is working as expected for most of what I'm doing,
> but I'm encountering some unexpected behavior w/r/t composite variables
> and derived variables defined by a ferret script.
> 
> First, w/r/t composite variables, i.e.
> 
>     <u units="cm/s" name="Zonal velocity">
>      <link match="/lasdata/grids/file_x_y_z_Time_grid"/>
>     </u>
>     <v units="cm/s" name="Meridional velocity">
>      <link match="/lasdata/grids/file_x_y_z_Time_grid"/>
>     </v>
> 
>    <composite>
>     <current_vectors name="uv_current vectors" js="VecVariable">
>        <link match="/file/variables/u"/>
>        <link match="/file/variables/v"/>
>     </current_vectors>
>    </composite>
> 
> After making the HTML and Java files, the LAS server does not bring up
> the correct z axis menubuttons.  In this case, u and v each have 40
> vertical levels show up on the menubuttons, but the composite variable
> only brings up 1 level.  It seems that the link elements from the
> primary variables u and v are not being delivered to the composite
> element current_vectors.  Is this a problem with the LAS software or did
> I do something wrong?
> 
> ************************
> 
> Now, w/r/t derived variables defined by a ferret script using the
> init_script feature, I'm having a problem of a more serious nature.  I
> can define a derived variable easily enough, but when I attempt to do a
> difference involving defined variables, I run into trouble.  It appears
> from looking at the error logs that the init script is not executed
> prior to running the differencing script. Is it possible to check if a
> difference variable has a ferret properties tag and call it before the
> difference script is executed?
> 
> I'll stop here.  Thanks.
> 
> Matt
> 
> --
> Matthew J. Harrison                                      GFDL
> Physical Scientist                                       P.O. Box 308
> NOAA Geophysical Fluid Dynamics Laboratory               Princeton, NJ
> 08542
> ----------------------------------------------------------------------------
> 
> e-mail:                 mjh@gfdl.gov
> Tel:                    (609)452-6579
> Fax:                    (609)987-5063
> http address:           http://www.gfdl.gov/~mjh