John Storrs <john.storrs@xxxxxxxxxxxx> writes:
> I see that coordinate variable data is stored in HDF5 dimension scale
> datasets, and that each ordinary variable is bound to one or more of these
> datasets through the HDF5 dimension scale mechanism. Evidently netCDF-4
> variables are aware of both their named dimensions (expected) and any
> associated coordinates (unexpected). The netCDF documents state that
> coordinate variables are a convention, and 'have no special meaning to the
> netCDF library'. But evidently the association is fully defined at storage
> level in netCDF-4/HDF5 files. Is there an API to access this useful
> information? Apologies if I'm missing something here.
You are correct in your summary of how we use dimension scales. You
are also correct in that coordinate variable now *do* have something
of a special meaning to the netCDF library.
There is no API other than the netCDF API. Is there some information
you can't get from that?
> Also the scope of the dimension to coordinate variable association isn't
clear
> to me - I haven't found anything to explain it. I'm picturing a tree of
> groups with dimensions and coordinate variables in various places in it,
> perhaps with duplicate naming (eg repetition of a subtree). What is the
> scoping rule?
Dimensions must be in the same group or in parent groups.
The new function nc_inq_dimids will tell you the dimids of all the
available dimensions in the group.
Thanks,
Ed
--
Ed Hartnett -- ed@xxxxxxxxxxxxxxxx
