http://www.unidata.ucar.edu/packages/netcdf/.The netCDF data model is interchangeable with the SDS data model in so far as it is possible to use the netCDF calling interface to place an SDS into an HDF file and conversely the SDS interface will read from an XDR-based netCDF file. Because the netCDF interface has not changed and netCDF files stored in XDR format are readable, existing netCDF programs and data are still usable, although programs will need to be relinked to the new library. However, there are important conceptual differences between the HDF and the netCDF data model that must be understood to effectively use HDF in working with netCDF data objects and to understand enhancements to the interface that will be included in the future to make the two APIs much more similar.
In the HDF model, when a multidimensional SDS is created by SDcreate, HDF data objects are also created that provide information about the individual dimensions - one for each dimension. Each SDS contains within its internal structure the array data as well as pointers to these dimensions. Each dimension is stored in a structure that is in the HDF file but separate from the SDS array.
If more than one SDS have the same dimension sizes, they may share dimensions by pointing to the same dimensions. This can be done in application programs by calling SDsetdimname to assign the same dimension name to all dimensions that are shared by several SDSs. For example, suppose you make the following sequence of calls for every SDS in a file:
dim_id = SDgetdimid(sds_id, 0);
ret = SDsetdimname(dim_id, "Lat");
dim_id = SDgetdimid(sds_id, 1);
ret = SDsetdimname(dim_id, "Long");This will create a shared dimension named "
Lat" that is associated with every SDS as the first dimension and a dimension named "Long" as the second dimension.This same result is obtained differently in netCDF. Note that a netCDF "variable" is roughly the same as an HDF SDS. The netCDF interface requires application programs to define all dimensions, using
ncdimdef, before defining variables. Those defined dimensions are then used to define variables in ncvardef. Each dimension is defined by a name and a size. All variables using the same dimension will have the same dimension name and dimension size.Although the HDF SDS interface will read from and write to existing XDR-based netCDF files, HDF cannot be used to create XDR-based netCDF files.
There is currently no support for mixing HDF data objects that are not SDSs and netCDF data objects. For example, a raster image can exist in the same HDF file as a netCDF data object, but you must use one of the HDF raster image APIs to read the image and the HDF SD or netCDF interface to read the netCDF data object. The other HDF APIs are currently being modified to allow multifile access. Closer integration with the netCDF interface will probably be delayed until the end of that project.
ncdump with HDF creates a utility that can dump scientific data sets from both HDF and XDR-based files. To read an XDR-based file using an HDF application, the application must be relinked to the HDF library.The current version of HDF contains several APIs that support essentially the same data model:
The first three models can create, read, and write SDSs in HDF files. Both the SD and NC interfaces can read from and write to XDR-based netCDF files, but they cannot create them. This interoperability means that a single program may contain both SD and NC function calls and thus transparently read and write scientific data sets to HDF or XDR-based files.
TABLE 3AH - Summary of HDF and XDR File Compatibility for the HDF and netCDF APIs
TABLE 3AI - NC Interface Routine Calls and their SD Equivalents