Hi Ellyn,
According to CF Trac Ticket #31 (slated for inclusion in the update to CF 1.7),
the way to cache minimum & maximum values in metadata is to use an attribute
named "actual_range" and store them as a pair.
(I kind of think this is a bad idea, and wish that ticket was still open. I
missed this discussion when it happened, but my experience with actual_max type
attributes in practice has convinced me that it's SO easy for them to become
inconsistent that it would really be better not to include them at all.
Computing the min & max on the fly is cheap, and approximating it is even
cheaper, so why introduce the uncertainty? But if they do need to go into
metadata, I think it would be better to use a name that highlighted the fact
that they're potentially unreliable. Maybe something like "nominal_range"
rather than "actual_range"?)
Anyway, the details on how actual_range works can be found here:
https://cf-pcmdi.llnl.gov/trac/ticket/31
Cheers,
--Seth
>The second part of the question involves variable attributes we now call
>minimum and maximum. Do these names have special meanings? In our files, we
>include the actual minimum and maximum computed for each non-coordinate
>variable's data. I want to be clear about the content of these attributes,
>and wonder if others use different terms to avoid confusion with valid_min and
>valid_max? Would calling these attributes "computed_minimum" (maximum) or
>"actual_minimum" (maximum) be better?
>
>Thanks for any suggestions!
>Ellyn
Received on Wed May 22 2013 - 15:04:50 BST
