...user.
Programming reference manuals for C and Fortran users are provided as separate documents.
...rVariables
The ``r'' stands for ``regular.'' rVariables are the type of variables that CDF has always supported. Perhaps ``traditional'' would have been a better term. We also considered ``related'' and ``correlated'' but those just didn't seem right.
...zVariables.
The ``z'' doesn't stand for anything special. We just like the letter ``z.''
...zVariables.
This is generally not recommended. In those situations where zVariables are necessary it is best to use all zVariables rather than a mixture of rVariables and zVariables.
...encodings:
This is a new feature of CDF V2.5.
...Interface
The Standard Interface was previously known as the C/Fortran Interface(s). That was an unfortunate choice of terms since the Internal Interface is also callable from C and Fortran.
...zMode
There are actually two types of zMode --- read on.
...zVariables.
In a future release of CDF, support for rVariables will be eliminated. zMode is provided to ease the transition from rVariables to the more flexible zVariables. rVariables are essentially a subset of zVariables.
...2:[180,360]
This notation is used throughout this document. In this case there are two dimensions whose sizes are 180 and 360. A dimensionality of zero is represented as 0:[].
...variables.
As variables are created in a single-file CDF, the CDF library will attempt to adjust the cache size accordingly. This may be overridden by an application if desired.
...exists.
Actually, the CDF library will check several possible extensions: .cdf, .cdf;1, .CDF, and .CDF;1. These extensions are checked because some CD-ROM drivers (primarily on UNIX machines) do peculiar things when making the files (CDFs) on a CD-ROM visible.
....cdf.
Or .cdf;1 or .CDF or .CDF;1.
...entries.
Actually, all but the last index record will be completely full of index entries. The last index record will contain one or more index entries.
...encoding.
This is a change from previous releases of CDF.
...00:00:00.000.
I know what you're thinking. The year 0 AD never existed. If it makes you feel better, think of the epoch year as 1 BC rather than 0 AD.
...exist:
Previous releases of CDF limited the number of variables a CDF could contain. That limit has been eliminated except for multi-file CDFs on an IBM PC because of the 8.3 naming convention.
...zVariables.
The letters ``r'' and ``z'' don't stand for anything in particular. ``r'' sort of stands for ``regular'' since rVariables have always been supported by CDF.
...application.
It is required that an application close a CDF before exiting.
...CDF.
Variables may not currently be deleted from a multi-file CDF.
...CDF.
There is no reason to allocate records for a variable in a multi-file CDF.
...optimize
Since an application knows how it will be accessing a variable, it knows best how to optimize the caching scheme used. See Section 2.1.4 for details on how an application can control the CDF library caching scheme.
...pad
These were previously known as fill values but were renamed to avoid confusion with the FILLVAL attribute (an NSSDC standard attribute).
...follows:
These default pad values can be changed by your system manager when the CDF distribution is built.
...scope.
This was not necessarily the case in previous releases of CDF. These new restrictions should not, however, cause any conflicts with existing applications.
...directory/wildcard
Macintosh folders are equivalent to the directories discussed here.
...files.
CDFdir is not available on a Macintosh.
...:
On MS-DOS systems the executable is named   CDFCMP.EXE.
...:
On MS-DOS systems the executable is named   CDFCVT.EXE.
...qualifiers.
In CDF V2.5, the previous functionality of CDFinquire has been moved to the SkeletonTable toolkit program.
...:
On MS-DOS systems the executable is named   CDFINQ.EXE.
...:
On MS-DOS systems the executable is named   CDF2SKT.EXE.
...SkeletonCDF
This program was originally named CDFskeleton. It has been renamed to ease the confusion caused some users. Now, SkeletonCDF is used to create skeleton CDFs and SkeletonTable is used to create skeleton tables.
...record.
RV variable data may also be specified in a skeleton table but this is generally not done given the amount of RV data typically contained in a CDF.
...:
On MS-DOS systems the executable is named   SKT2CDF.EXE.
...#TCDFedit#5037>).
The alias/symbol CDFbrowse still exists in the ``definitions'' file on UNIX/VMS systems but now executes CDFedit in a browse-only mode.
...modifications.
Running CDFedit in a browse-only mode provides the same functionality as CDFbrowse   once did.
...here.
It is our intention that the use of CDFedit be as intuitive as possible. You may not even need the on-line help. We're sure you'll let us know.
...batch
These are also known as include files.
...C.
With the exception of the multiple variable access operations available via the Internal Interface (<GET_,rVARs_RECDATA_> , <PUT_,rVARs_RECDATA_> , <GET_,zVARs_RECDATA_> , and <PUT_,zVARs_RECDATA_> ). Because these operations deal with variables having different data types, the use of structures to store the values would be the logical method to use. Unfortunately, IDL does not currently support the manipulation of structures from within a C application (which is how CDF's IDL interface is implemented). Support of these operations will hopefully be added in the future.
...TYP_LONG.
On DEC Alpha's running OSF/1, don't confuse the C data type of long with IDL's data type of TYP_LONG. The C long is eight bytes while IDL's TYP_LONG is four bytes.
...etc.
If an array in a call to a CDF function won't be used because a variable is 0-dimensional, a dummy array must still be passed to avoid an error from the CDF library. Note that embedding a call to lonarr [such as lonarr(1)] in the call to the CDF function will work but will result in a memory leak because the allocated array will not be freed.

cdfsupport@nssdca.gsfc.nasa.gov