Re: 20001102: Transforming coordinates in Java netCDF

> >> I didn't (and still don't) see how ArrayAbstract relates to a
> >> nc2 Variable.  Can you enlighten me? ;-)
> >
> >sorry: you do a Variable.read() to get an Array, which you cast to an
> >AbstractArray.
> 
> This doesn't seem intuitively obvious. I understand your comments about
> different ways of implementing Array, but looking over the javadocs, it
> doesn't say anywhere that Variable is implemented with an AbstractArray
> underneath. Since nc2.Variable is not an interface, then perhaps the
> read should specify that it returns an AbsractArray. Otherwise, it seems
> like Array should have the copyTo1DJavaArray() method. That could be
> implemented regardless of the backing store.



i agree that its not obvious, and i will think about these issues. there
are various levels of exposing the underlying implementation. in some
ways copyTo1DJavaArray() might be considered implementation, but perhaps
it should be a service provided by any Array.  right now, to implement
these ideas, my thoughts would be to add:

  Object copyTo1DJavaArray();
  Object copyTo1DJavaArray(Class);  // type conversion

i suppose it would also be nice to have

  Object copyToNDJavaArray();       // N = rank
  Object copyToNDJavaArray(Class);  // type conversion

Should these be added to Array or AbstractArray? As more services are
added to Array, its less likely there will be any implementations other
than AbstractArray. So i have to think if thats ok, and if the Array
interface is still useful.
Other opinions are welcome. 

Also, to clarify: Variable implements MultiArray: to get an Array, you
must do a read(). this is an important difference in design from
ucar.multiarray. given that you found out the hard way that theres a
factor of 6 (or more?) i continue to think its the correct design.