pepke@gw.scri.fsu.edu (Eric Pepke) (02/26/91)
We are starting to use the Scientific Data Set capability of HDF to store our meteorological data sets. To do the kinds of things we want to do, we have had to make some ad hoc extensions. None of the extensions are in any sense "official" and have been thought up independently of NCSA, but what can you do? All comments are welcome. 1) Time-Dependent Data Some of our fields represent samples at a given time. Field samples must be read in and assembled into one time-dependent field for visualization. To identify a sample, we have adopted a naming convention for the DFSDsetdatastrs call. If the name of the field contains an at symbol (@), the part that comes before the symbol is the name of the field, and the part that comes after the symbol is the time the field was sampled. The time can either be a colon-separated clock time, a single floating point number, or a floating point number followed by a space and then some letters, which are assumed to be the name of the time unit. 2) Missing Data Sometimes we have a field that is not completely filled in. Some of the data points are missing. The visualization program may do any number of creative things with these points, such as code them with a certain color, interpolate through them, etc. In the past, we have stored these points arbitrarily as 999.0. We decided that if the minimum and maximum values are specified in the data file, any actual values read outside that range are considered missing data. When reading a file, a warning message is produced, but it only occurs once per unique value outside the range. 3) Arbitrary Structured Grids The scales mechanism provides a good way of doing grids where x'=f(x), y' = g(y), and z' = h(z). We sometimes want to do grids where x' = f(x, y, z), y' = g(x, y, z), and z' = h(x, y, z). What we are going to do is specify a naming convention in the coordsys of DFSDsetdatastrs. If the first character is an at (@), the remainder specifies the name of a field that gives the coordinate system. The field must have rank equal to the rank of the data field plus one, dimensions 1 to n - 1 the same, and dimension n = 3. It is interpreted as a field of x, y, z coordinate triples. Eric Pepke INTERNET: pepke@gw.scri.fsu.edu Supercomputer Computations Research Institute MFENET: pepke@fsu Florida State University SPAN: scri::pepke Tallahassee, FL 32306-4052 BITNET: pepke@fsu Disclaimer: My employers seldom even LISTEN to my opinions. Meta-disclaimer: Any society that needs disclaimers has too many lawyers.