[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20030624: IDV questions

>From: "Todd Plessel" <address@hidden>
>Organization: EPA
>Keywords: 200306242110.h5OLAoLd020057 IDV netCDF

Hi Todd -

I'm cc'ing Steve Fine since he left a message for me as well.  
I should have time to talk tomorrow (Wednesday), but here are some 
preliminary answers:

>Here are a few questions I have:
>
>1. Is IDV 100% Java (i.e., does not depend on any C Libraries, etc.)?

Yes.

>2. Does IDV/VisAD support subetting by variable, time-step, layer,
>row, column?
>
>(I noticed that the Vis5DForm class reads the whole file into memory.
>This won't work for our files. We need the ability to read just a
>subset (at least variable/time-step/layer).)

It depends on the data.

For a Vis5D file, the Vis5DAdaptedForm uses a caching system.
For 3D variables, it will read in only one at a time as requested,
but will read all timesteps for that variable.  You can then
subset the times.

For netCDF grid files, data can be subsetted on variable and time
(but not layer currently).

>3. Does IDV/VisAD support remote host file browsing and access?
>
>By this I don't mean a web server or database, but just a plain remote
>UNIX host file system (e.g., by ssh). In our environment, users
>run their models on supercomputers and leave the generated output
>files there and then run vis apps on their desktop workstations and
>visualize the output files using tools implemented with rsh or ssh.

No, this is not currently available.  However, if they are netCDF
grid files and were available on a network drive, you could open them 
and only the meta data is read.  You can then do variable/time 
subsets (but not currently layer).  I'll talk to our netCDF 
gurus about this one.  Since the IDV is entirely java and we don't
use a java library to implement ssh or rsh, I'm not sure how this
could be done.

>4. Our data files are M3IO files which are NetCDF-format files with
>certain conventions for attributes (see below). Typically these are
>time-stepped gridded data in a Lambert Conformal Conic projected space.
>The variables are scalar air quality species and sometimes wind vectors
>in separately gridded files. So, conceptually, a main data file plus a
>few optional auxillary files together consitute a "data set".
>I need to develop adapter classes for these files that are similar to
>Vis5D files except my adapters need to handle a set of files,
>subsetting and possibly remote hosted files.

This would be a ucar.unidata.data.DataSource.  I skimmed the
docs for M3IO and see that you might just be able to use the
new NcML features of the Java netCDF library to make a virtual
dataset of your M3IO files that would allow them to be read
into the IDV with minimal effort.  Alternatively, you could
add a new ucar.nc2.dataset.conv.Convention to support the
M3IO.  Is there a document on the semantics of this convention?

>What is the minimal set of (e.g., top dozen) IDV/VisAD classes that I
>need to read/understand/use/modify to implement:
>1. Remote host file browsing?

Not sure on this one.

>2. Reading subsets of M3IO files (one read per time-step)?

If you could implement an M3IOConvention class, you might
just get this automatically.

>3. Display as "colored tiles" with optional wind vector arrows?
>(Colored tiles are similar to "Color-Shaded Plan View" except the data
>are cell-based instead of vertex-based so the result is constant-colored
>quads/hex instead of color-interpolated ones. See attached image.)

Yes.  See:

http://my.unidata.ucar.edu/content/software/IDV/docs/UserGuide/IDV_displayingdata.html

if you click off the Shade colors checkbox in the Color-ShadedPlanView
control, it will revert to tiles.  In the latest release (1.0), the
default is now colored tiles instead of shading (better performance).

>I'm hoping that a half-hour conversation with you could save our staff
>days of reverse-engineering.

I'll try to give you a call tomorrow (Wednesday) during a
break in the workshop.  

Is is possible to get a copy of one of the M3IO netCDF files for
us to experiment with? 


Don Murray
****************************************************************************
Unidata User Support                                    UCAR Unidata Program
303 497 8643                                                  P.O. Box 3000
address@hidden                                   Boulder, CO 80307
----------------------------------------------------------------------------
Unidata WWW Service              http://my.unidata.ucar.edu/content/support
****************************************************************************