netcdf-hdf mailing list is no longer active. The list archives are made available for historical reasons.
Quincey Koziol <koziol@xxxxxxxxxxxxx> wrote: > We've thought about the changes to the overflow document and are leaning > toward taking out the "fill value" parameter from the callback, since it can > be easily set up by an application (or the netCDF library) as part of the > "user_data" information for the callback. With this revision, the overflow > callback is more general purpose and less tailored to just the netCDF needs. > Let me know what you think. This seems fine to me. The main use for data conversion, as far as netCDF is concerned, is in requesting numeric data as double or float, without needing to know or care what numeric type is actually used for storing the data. For example, a visualization application may present an interface that lets users choose several datasets to visualize together, and it makes the application writer's job easier to just get all the values as doubles. So although we have to handle the details of what happens in weird cases such as when doubles are requested as ints and an overflow, loss of precision, or truncation occurs, in practice this is rarely a concern. --Russ