[MITgcm-devel] Re: [Fwd: (no subject)]

[Ed Hill]

On Thu, 2006-03-02 at 16:41 -0500, Daniel Enderton wrote:
> The bug that was fixed in rdmnc.m a couple days ago had to do with 
> reading in singleton dimensions that come out in the diagnostics 
> package output (e.g. data that is given as 10x10x1 rather than 10x10 
> in the netcdf files).  Hopefully tonight I can get to checking with 
> the actual case that Ed found that failed (Ed - could send me a 
> pointer to the data?).
> 
> Moreover, if there is interest in funneling everyone to rdnctiles.m, 
> I would be more than happy to move rdmnc.m to some place out of the 
> way (e.g. /utils/matlab/Graphix/GraphixUtilities).

Hi Daniel,

I'm not trying to make it more difficult to use rdmnc.  If it works for
you (or anyone else) then thats totally cool!

What I want is an easy-to-use drop-in replacement for rdmds() and
rdmnc() that has the following features:

 1) is maximally backwards-compatible
 2) gracefully handles more general topologies
 3) provides other new features such as the collapsed-to-vectors-
    with-unique-cell-ordering as requested by Jean-Michel (and will
    be very useful for calculating statistics such as min, max, 
    sums, masks, etc.)

And #2 really is the main point.  MITgcm can, for a long time now, run
with cubes where the three paths (nr,ng,nb) are not equal.  And, for
those runs, rdmnc is completely useless.  Thats where rdnctiles comes
it.  It neatly handles those cases by providing struct arrays where each
face is a nice, neat, logically rectangular ("LR") domain.

And, if we work together, then it won't be too hard to write new and
very general diagnostic and plotting routines that can handle the
per-face LR regions.  For instance, one can easily imagine a zonal
averaging routine that computes partial sums (both numerators and
denominators -- *with* masks!) on each face and then computes global
sums from the partial sums.  Not too hard.  And, if its done on a
per-face or a per-tile fashion then it makes *NO* difference what the
overall domain (topology) looks like!  Imagine that.  Just a little more
effort and we have a diagnostic thats completely divorced from the grid
topology, extent, whatever.

Neat, yes?

Ed

-- 
Edward H. Hill III, PhD
office:  MIT Dept. of EAPS;  Rm 54-1424;  77 Massachusetts Ave.
             Cambridge, MA 02139-4307
emails:  eh3 at mit.edu                ed at eh3.com
URLs:    http://web.mit.edu/eh3/    http://eh3.com/
phone:   617-253-0098
fax:     617-253-4464