jonathan.gregory at metoffice.com wrote:
>Dear All
>
>A further comment re contiguous: Multidimensional coordinate variables must be
>auxiliary coordinate variables. Boundary variables like (m,n,npoly) must be
>boundary variables corresponding to auxiliary coordinate variables.
>
>Is it not therefore the case that there must also be ordinary 1D coordinate
>variables, which should have boundary variables too? If so, can't
>contiguousness be determined from the 1D variables? This surely must give the
>same answer as working it out from the 2D variables. If two points are
>coincident in one coordinate system, it would be a strange transformation
>that mapped them to different points in another system.
>
i think so: the kinds of transformations we are interested in
(continuous and invertible) preserve connectedness.
there are probably cases that violate that, but i feel confident that
its the common case.
>
>The difficult cases for contiguousness are the ones where the cells aren't
>arranged in a rectilinear array in any coordinate system, I think. Maybe I
>am missing something obvious here. Please help me out of my misunderstanding
>if so.
>
it seems like the difficult case is where you want to explicitly add a
"contiguous" attribute, rather than make the client guess.
>
>Thanks
>
>Jonathan
>_______________________________________________
>CF-metadata mailing list
>CF-metadata at cgd.ucar.edu
>http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
>
>
Received on Fri Mar 21 2003 - 12:56:13 GMT
