netcdf-hdf mailing list is no longer active. The list archives are made available for historical reasons.
Yes, we made a deliberate design decision to not require dimension scale datasets to be the same size as the dimension. There are use cases where you might not want this, e.g., the scale values are every 5th tick mark. Also, it is not clear what the semantics should be if a dataset is extended: what should happen to an associated dimension scale dataset? What if it is shared by several datasets? As to the interpretation: that is entirely up to the application. In the case of NC4, you should define some constraints, perhaps as conventions. If necessary, you could put an additional attribute on the scale dataset, to indicate that this should be interpreted according to NC4 conventions. On Thu, 30 Jun 2005, Ed Hartnett wrote: > Howdy all! > > I am wondering something about dimension scales. > > With dimension scales there does not seem to be any enforced > relationship between the size of a dimscale dataset, and the > dataset(s) to which it is attached. > > That is, I can have a 1D dimscale of length 2, and attach it to the > dimension of a 1D dataset with length 3. > > In actual use, what do we think it would mean to have dimension data > that are smaller in extent than the data to which they apply? Is there > any good way to interpret this? > > If I were writing a graphics program to plot that data, would I match > them one to one, starting at the first dimension scale value, and then > run out of dimension data? Or what? > > Any comments would be useful. > > Thanks! > > Ed > > > > -- Robert E. McGrath National Center for Supercomputing Applications University of Illinois, Urbana-Champaign Champaign, Illinois 61820 (217)-333-6549 mcgrath@xxxxxxxxxxxxx