All,
A situation that needs to be considered: How to represent a single band
file? For example, can this be done w/o creating a 3D variable with
'band dimension' = 1?
Tom
On 7/25/11 2:29 PM, Jim Biard wrote:
Tom,
I think your simple index coordinate variable and my label coordinate
variable are pretty much the same thing. Yours would contain a
numeric sequence, and mine would contain descriptive strings. They
both are examples of parametric drivers. The remaining question is -
what constitutes a sufficiently unambiguous way to identify the
ensemble of variables used to describe each band? Put another way,
are the existing conventions for identifying auxiliary coordinate
variables sufficient to the task of identifying the descriptors for
the bands?
Let's say that we have a parametric driver coordinate variable named
band (with dimension band) that has an axis type of "index", and
variables with axis types of "wavenumber", "bandwidth", and
"polarization" (each with dimension band). Is it sufficient to
identify the wavenumber, bandwidth, and polarization variables as
auxiliary coordinates on our data variable (via the coordinates
attribute)? Or is there a need to more explicitly declare the
ensemble (perhaps via an attribute on the band coordinate variable)?
Jim
On 7/25/2011 3:02 PM, Tom Whittaker wrote:
Hi Jim...
On Mon, Jul 25, 2011 at 12:44 PM, Jim Biard<Jim.Biard@xxxxxxxx> wrote:
Is this just too far out?
Far out, man!! ;-)
Seriously, though, I don't see the need for such a device ("...make
the "band" coordinate variable an array of label strings..."). Based
on what Tom Rink and John Caron have said, it looks like defining a
"coordinate axis type" using a few select concepts is the best
approach. We'll see what John as to say about the notion of a simple
index being used as a coordinate variable...and go from there...
tom
