Re: netCDF: Regularly-spaced coordinates

Steve Emmerson (steve@unidata.ucar.edu)
Wed, 25 Mar 1992 15:05:01 -0700

>Multigrid may be the way to handle difficult 3-D fluid dynamics
>problems, and is being explored by a number of adventurous
>fluid dynamicists.  
>
>Any netcdfers have ideas on representation strategies?
>
>       --Barbara Mihalas       bmihalas@ncsa.uiuc.edu

I believe you need to address another issue first, namely, what do you
want from a co-ordinate system "thingy".  The types of behavior you'd
like from it should help you determine what information it'll need to
store.  Once you have that, you can figure out the representation.

Actually, a co-ordinate system "thingy" is probably the wrong level of
abstraction at this point.  Because datasets are intimately related to
their co-ordinate systems (the co-ordinate system defines the
independent variables after all) it's probably more useful to think of
the required behavior of the dataset.  I can imagine a dataset answering
the following questions:

	Q: What's the type of your co-ordinate system?
	A: Geographic

	Q: How many co-ordinate systems do you contain?
	A: 3

	Q: What's the projection of the first system?
	A: Cylindrical equidistant.

	Q: What are the parameters of the first system?
	A: ...

	Q: What's the datum closest to co-ordinates ...?
	A: ...

	Q: What are the co-ordinates of the datum at indexes ...?
	A: ...

	etc.

I think we need to enumerate all the co-ordinate-system-related
behaviors we'd like from datasets.

--Steve