This is to elaborate on my answer yesterday:
> > Amazingly, Visad works for HDF5 files with complicated hiearchical
> > structure.
> > However, there is an inconvenience in the spreadsheet. Suppose I have a data
> > set which is at group1-->group1_1-->dataset2 and I open the file at cell a1.
> > In order to open the data set, I may have to open the groups level by level,
> > for instance, open file at a1 --> open group1 at b1 (a1.0) --> open group1_1
> > at a2 (b1.1) --> open dataset2 at b2 (a2.1). I wonder if there is a way
> > which
> > opens the datset directly from cell a1, for example, open dataset2 at b1 by
> > a220.127.116.11
> Most VisAD file adpters (extend visad.data.Form) open the
> entire file, but there are some that do not. You could
> write an HDF-5 adapter whose open() method would create its
> own little GUI in a JFrame to ask the user which part of a
> file they want to open. We haven't done this, so there may
> be some little problem I don't foresee, but my first guess
> is that it will work.
There should be an HDF-5 file adapter that opens the entire file as
one large VisAD data object (it can use the visad.data.FileFlatField
logic for disk caching so it doesn't have to read the whole file
into memory at once).
This is needed because there are applications other than the
SpreadSheet that want to navigate the file using the VisAD data
model API. As I think you have already done, HDF-5 groups map
into VisAD Tuples for this navigation.
One approach to this would be to include a boolean argument to
the constructor for your HDF-5 file adapter that determines
whether it opens the entire file, or whether it creates a GUI
for the user to select a part of the file. Of course, as you
point out the SpreadSheet does provide a way to select parts of
files via its formulas, but it is a bit terse.
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI 53706
hibbard@xxxxxxxxxxxxxxxxx 608-263-4427 fax: 608-263-6738