> > > Once
> > > I've reach a leaf node of the VisAD data object, I then define a netCDF
> > > variable that uses the chain of DataAccessors to access the data values.
> > This won't necessarily work for me; FITS objects can't really be that
> > complicated.
> Who's to say that the VisAD data object that's handed to the FITS
> "backend" is really a FITS object? It could have come, for example,
> from importing a netCDF dataset. In designing the netCDF export
> capability, I tried to get away from assuming that the structure of the
> VisAD data object that was being exported was going to be the same as
> the structure the netCDF backend created on import.
Well, since I'm talking about saving to a FITS file, I can safely say that
I need to worry about what kind of data I can stuff into a FITS file.
Maybe I misunderstood your initial statement.
FITS is basically a flat format ... I need to write out either N-dimensional
arrays of homogeneous data or 2D (or possibly 3D) tables of heterogeneous
I assumed that using the *Accessor classes would result in a hierarchical
view of the data. If that's the case, it's not very useful for me, since
I'd still have to break it down into something that can be saved as a FITS
(I'm just about to dive into the whole *Adapter/*Accessor stuff, so I'll
be able to tell soon whether or not it'll work for me...)