catalog schema question: variable element

[Ethan Davis]

Hi Darren,

Darren Hardy wrote:
> I have a question about the <variable> element in the inventory 
> catalog schema. Suppose that I have the following information on a 
> dataset variable (in DAS format):
>
> Attributes {
>     SST {
>         String long_name "OI Cloud-free SST, blah blah blah";
>         String standard_name "sea_surface_temperature";
>         String units "degC";
>     }
>     Foobar {
>         String long_name "My foobar variable"
>         String units "m"
>     }
> }
>
> As I read the catalog spec, I will lose the "human-readable" long_name 
> attribute (ala CF-1.0):
>
>   <variables>
>     <variable name="SST" vocabulary_name="sea_surface_temperature" 
> units="degC"/>
>     <variable name="Foobar" vocabulary_name"Foobar" units="m" />
>   </variables>
I think the mismatch is in our goals for the <variables> element. Our 
goal when designing the <variables> element was to provide a way to 
describe the variables in a dataset to a search service and allow users 
to add information that wasn't in the dataset. Since many search 
services require mapping variables to controlled vocabularies, we 
provided for a way to map dataset variable names into controlled 
vocabularies.

In your example, if you wanted to map the variables into the CF standard 
names, you could change the Foobar vocabulary_name to the appropriate CF 
standard name. So, if you knew that the Foobar variable mapped to CF 
standard name "sea_surface_height" you could edit your catalog to show this:

<variables vocabulary="CF-1.0">
    <variable name="SST" vocabulary_name="sea_surface_temperature" 
units="degC"/>
    <variable name="Foobar" vocabulary_name="sea_surface_height" 
units="m" />
</variables>

Of course, you could use your "long_name" entries as a controlled 
vocabulary of sorts like:

<variables vocabulary="myLongNames">
    <variable name="SST" vocabulary_name="OI Cloud-free SST, blah blah 
blah" units="degC"/>
    <variable name="Foobar" vocabulary_name="My foobar variable" 
units="m" />
</variables>

But either way you can't map all the information in your dataset. We 
have talked about allowing multiple mappings: either allowing multiple 
<variables> elements so that different vocabularies could be used; or 
moving multiple mappings into each <variable> element. But at this point 
we don't have that so in our schema.

>
> Am I missing something? Is there a way to retain that human-readable 
> description of the variables?
You can put it in the dataset description or keywords but that isn't 
quite as structured in terms of being specific to a variable.

Ethan
>
> Thanks,
> -Darren
>

-- 
Ethan R. Davis                                Telephone: (303) 497-8155
Software Engineer                             Fax:       (303) 497-8690
UCAR Unidata Program Center                   E-mail:    edavis at ucar.edu
P.O. Box 3000
Boulder, CO  80307-3000                       http://www.unidata.ucar.edu/
---------------------------------------------------------------------------


==============================================================================
To unsubscribe thredds, visit:
http://www.unidata.ucar.edu/mailing-list-delete-form.html
==============================================================================