Hi Russ,
"group dimensions" are a novel and logical idea.
Not necessary for our proposal, yet worth considering and remembering.
Let me clarify that a "group dimension" would not and could not
be a regular dimension in the variables because each ensemble
member might have different grids. remember, members usually
have the same rank (e.g., 3D) but different grids (e.g., spectral
vs. 1x1 degree). So variables cannot be simply glued
together across the ensemble by adding an "ensemble" dimension
to each variable because CESM tas(ensemble,time,x,y,z)
doesn't (usually) share all remaining dimension sizes with ECMWF
tas(ensemble,time,x,y,z).
So "group dimensions" would be a new thang.
The equivalent is do-able with metadata annotation now.
The library implementation would be your task.
That's what you get for opening your big mouth :)
cz
Le 19/09/2013 13:16, Russ Rew a ?crit :
> Charlie,
>
> Great use case, clearly explained ...
>
> It also demonstrates a potential need for "Group dimensions", so instead
> of groups cesm_01 and cesm_02, it would be possible to refer to Groups
> cesm[1] and cesm[2], supporting loops over contents of closely related
> groups without having to invent a convention for their names.
>
> --Russ
--
Charlie Zender, Earth System Sci. & Computer Sci.
University of California, Irvine 949-891-2429 )'(
Received on Fri Sep 20 2013 - 14:10:13 BST
