hdf images hdf images

This web site is no longer maintained (but will remain online).
Please see The HDF Group's new Support Portal for the latest information.

Frequently Asked Questions about HDF

Contents


What is HDF?

HDF stands for Hierarchical Data Format. It is a library and multi-object file format for the transfer of graphical and numerical data between machines.

It is freely available. The distribution consists of the HDF library, the HDF command line utilities, and a test suite (source code only).

Features of the HDF File Format:


Copyright Information

The COPYING file at the top of the HDF source code tree provides the copyright information regarding HDF.

Should new products be developed in HDF (HDF4) ?

No, we do not recommend storing new products in HDF4. You should use HDF5 instead. We also want to encourage and help users move from HDF4 to HDF5. Please let us know if you need help with this.

HDF4 still works well for many users, and we are pleased that it has served users so well. However, there are good reasons to use or move to HDF5.

In general, HDF4 is based on technology from the 1980's and 1990's. Some specific reasons to use HDF5 instead of HDF4 are listed below:

See the HDF vs. HDF5 table which compares HDF4 and HDF5 features.


What is in the HDF library?

HDF currently supports several data structure types: Scientific Datasets (multi-dimensional arrays), vdatas (binary tables), "general" raster images, text entries (annotations), 8-bit raster images, 24-bit raster images, and color palettes.

HDF contains: the base library, the multi-file (SDS) library, the jpeg library, and the gzip library. HDF library functions can be called from C or FORTRAN user application programs.

The base library contains a general purpose interface and application level interfaces, one for each data structure type. Each application level interface is specifically designed to read, write and manipulate one type. The general purpose interface contains functions, such as file I/O, error handling, memory management and physical storage.

The multi-file (SDS) library integrates the netCDF model with HDF Scientific data sets, and supports simultaneous access to multiple files and multiple objects. This part is referred to as the mfhdf library in the rest of this FAQ.

The jpeg and gzip libraries allow you to use jpeg and gzip compression for those application programming interfaces that support them.


What are the HDF command line utilities?

The HDF command line utilities are application programs that can be executed by entering them at the command level, just like UNIX commands.

There are HDF utilities to:

They provide capabilities for doing things with HDF files for which you would normally have to write your own program.

The 'hdp' utility is one of the more useful HDF utilities. Following is a description of its function:

The other utilities are:

In addition, the netCDF utilities, ncdump and ncgen, COMPILED with the HDF library, are included.


What is the latest official release of HDF, and what platforms does it support?

See the download page for the latest official release of HDF and the platforms supported.


When will the next release of HDF4 be?

In general, a maintenance release of HDF4 occurs once a year February. HDF5 releases occur every six months around May and November.


What are the new features included in the current release?

Details are listed in the RELEASE.txt file of the release.

Is there a Java Interface?

Yes, there is a Java Interface for HDF. For information on this please refer to the HDF Java Products web page,

Are there limitations to HDF4 files?

Yes. Here are some of the limitations on HDF4 files:

Is there a limit on the name of a dataset?

No. HDF has always allowed any length to be specified for a name, when creating an SDS, Vdata, or Vgroup. However, prior to HDF 4.2r2, the behavior was undefined for names that were greater than 64 characters, which caused problems for tools reading the data. Either characters were concatenated to the name or, in the case of the hdp tool, a crash occurred due to accessing undefined memory.

As of HDF 4.2r2, released in 2007, this issue was fixed and a name with any length could be retrieved. The SDgetnamelen, Vgetnamelen*, and Vgetclassnamelen* APIs were introduced for retrieving the length of a name.

* Vgetnamelen and Vgetclassnamelen were added to the source code in HDF 4.2r2, but were inadvertently left out of the documentation until HDF 4.2.5, and are therefore listed in the Release Notes for HDF 4.2.5.

Where can I get the HDF source code and information relevant to HDF?

For information, take a look at the HDF home page.


What documentation for HDF is available?

See the HDF4 Documentation page.


Who do I contact for information on SZIP licensing issues ?

See the SZIP Licensing page.


How do I install HDF (4)?

Optional:   Set environment variables to point to the correct compilers. For example:
   setenv CC "/opt/SUNWspro/bin/cc -xarch=v9"
   setenv F77 "/opt/SUNWspro/bin/f90 -xarch=v9"
Also, remember that HDF4 builds with Fortran by default. Particularly on systems that support both 32-bit and 64-bit, HDF may pick up the wrong compiler.

Solaris ONLY:   Make sure that your PATH variable points to the correct ar and tr tools:

    tr:  /usr/ucb/tr
    ar:  /usr/ccs/bin/ar
Following are the commands to build HDF. The JPEG and ZLIB libraries are required. You must include them when configuring. The SZIP library is optional.
  ./configure --with-zlib=/path_to_ZLIB_install_directory
                   --with-jpeg=/path_to_JPEG_install_directory
                   [--with-szlib=/path_to_SZIP_install_directory]
                   --prefix=/path_to_HDF4_install_directory

  gmake >& gmake.out
  gmake check >& check.out
  gmake install
To disable fortran:
  ./configure --disable-fortran  ... 
If you encounter problems where configure can't find the JPEG or ZLIB library, but you know that you have specified the location properly, then it may have found more than one JPEG/ZLIB installation and is confused.


How do I compile application programs that call HDF functions?

To use HDF routines in your C program, you must either have the line #include "hdf.h" if you do not use the mfhdf library, or #include "mfhdf.h" if you do, near the beginning of your code.

The HDF Group provides the h4cc and h4fc scripts for compiling applications. These tools come with the pre-compiled binaries and source code.


What header file are you supposed to use in your application?

The header file hdf.h must be included in every HDF application written in C, except for programs that call routines in the SD interface. The header file, mfhdf.h, must be included in all programs that call the SD interface routines.

Fortran programmers who use compilers that allow file inclusion can include the files hdf.inc and dffunc.inc. If a Fortran compiler that does not support file inclusion is used, the HDF library definitions must be explicitly defined in the Fortran program as they are included in the header files of the HDF library.


How to use open (unbuffered I/O) instead of fopen?

To use open instead of fopen in HDF4, you must:
  1. Download and uncompress the HDF source code.
  2. Edit the ./hdf/src/hdfi.h file.
  3. Find the section of code for the platform you are on. For example on Solaris, find this line:
          #define DF_MT             DFMT_SUN
    
    Then go down further until you find the section below.
  4. Change the 2nd define in the following lines in that section, from:
       #ifdef  HAVE_FMPOOL
       #define FILELIB PAGEBUFIO  /* enable page buffering */
       #else
       #define FILELIB UNIXBUFIO
       #endif
    
    to:
       #ifdef  HAVE_FMPOOL
       #define FILELIB PAGEBUFIO  /* enable page buffering */
       #else
       #define FILELIB UNIXUNBUFIO
       #endif
    
This will make HDF use unbuffered I/O.

There is an unfixable bug in fopen on Solaris such that the maximum number of open files that are allowed is 256. The bug is fixed in open.


Configure fails, "cannot compute sizeof (int*), 77"

The problem is that you have to add the path to the SZIP library to LD_LIBRARY_PATH, since it is a shared library.

To see this, you can edit config.log, go to the end, and search backwards for "compute". You will find the compute error. Then go back further, before the example program, and you will see that it cannot open the shared SZIP library.


How can you compile HDF with shared libraries?

Prior to HDF 4.2r4, shared libraries were not supported in HDF4. With HDF 4.2r4, shared libraries for C were added. To build the C shared library, configure with these flags:

   --enable-shared --disable-fortran

This is required as Fortran shared libraries are not supported. Shared libraries are disabled by default.


Are there any conversion programs available to convert non-HDF image files into HDF files or vice versa?

Many of the HDF command line utilities are conversion programs. See Question #3 for more information regarding them.

Take a look at the What Software uses HDF? page off of the HDF home page for information on other tools.


Which HDF Group tools can I use to view HDF objects?

The HDF Group has several tools for scientific visualization that are based on HDF.

The latest tool available is:

Please see the What Software uses HDF? section on the home page for more information regarding these tools.


Is there any commercial or public domain visualization software that accepts HDF files?

Yes, there are numerous tools that accept HDF files. Please refer to the What Software uses HDF? section on our home page for more detailed, though not complete, information.

Commercial tools that accept HDF include IDL and Matlab.

Comprehensive MATLAB, IDL and NCL example scripts and plots on how to access and visualize NASA HDF-EOS and HDF products such as MODIS, MISR, AIRS, TRMM, AMSR-E, CERES, QuikSCAT, Aquarius, SeaWIFS,ICESAT-2,MOPITT can be found on the HDF-EOS Tools and Information Center under hdfeos.org/zoo.


How would you convert a netCDF file to/from HDF?

The netCDF library from Unidata contains utilities ncdump and ncgen to convert between the netCDF binary format and a text format called CDL. The HDF4 library from the HDF Group also contains utilities ncdump and ncgen that do the same thing, but this time the underlying binary file format is HDF. It is a little tricky to install both, because by default the HDF commands will overwrite the netCDF commands or vice versa. You may want to rename the HDF commands. (In the paragraphs below, the HDF versions of these tools are referred to as hdf_ncdump and hdf_ncgen.)

To do the "conversion" you would follow these steps:

          ncdump -l 80 foo.nc > foo.cdl
          # Edit CDL file with a text editor to work around various
          # problems encountered by hdf_ncgen
          hdf_ncgen -b -o foo.hdf foo.cdl
HDF uses netCDF 2.3.2 when building the HDF versions of these tools, which is a very old version of netCDF. The need to edit the CDL file arises because hdf_ncgen cannot read everything that ncdump can write.

The "-l 80" option above is intended to work around one of hdf_ncgen's limitations: a maximum line length of 80 characters. When hdf_ncgen encounters a problem, it halts and prints out the relevant line number in the CDL file. The error message will not necessarily make much sense, but when you go to that line you will usually find that the problem is reasonably obvious and easily corrected.

You may also have to do things like strip out attributes containing zero-length strings, remove netCDF fill values (represented in CDL as an underscore) and replace NaNs with finite floating-point values. (NaNs are illegal in netCDF anyway, though some platforms allow them.)

There is a Windows executable called ncdf2hdf from Fortner Software (defunct) that does the conversion. It is no longer available, but may be found by searching the web. However, the HDF files that it generates have been reported to lack some of the structure that is found in the ones produced by the route above.


Can new versions of HDF read HDF files written using older versions of the HDF library?

Our goal is to make HDF backward compatible in the sense that HDF files can always be read by newer versions of HDF. We have succeeded in doing this so far, and will continue to follow the principle as much as possible. In many instances, HDF is also forward compatible, at least in regards to the data. Metadata, such as attributes, may not be readable by previous releases, but the data should be. Please see the notes following the table below for information on when the data is not forward compatible.

The table below lists the backward and forward compatibility of HDF in regards to the data (not metadata). The Vdata and Vgroup interfaces have been merged into HDF since HDF3.2. Before then, they were in a separate library named Vset.

                       |        CAN READ DATA FILES CREATED WITH  
         Interface     | HDF3.1  |  HDF3.2 |  HDF3.3  |  HDF4.0   | HDF4.1
        -------------------------------------------------------------------
        HDF3.1         |         |         |          |           |
         -RIS8         |  YES    |    YES  |  YES(1)  |  YES(1)   | YES(1)  
         -RIS24        |  YES    |    YES  |  YES(1)  |  YES(1)   | YES(1)  
         -PALETTE      |  YES    |    YES  |  YES     |  YES      | YES
         -ANNOTATION   |  YES    |    YES  |  YES     |  YES      | YES
         -SDS DFSD     | Float32 | Float32 | Float32  | Float32(2)| Float32(2,3)
        Vset 2.1       |         |         |          |           |
         -VData        |  YES    |    YES  |    YES   |  YES      | YES
         -Vgroup       |  YES    |    YES  |    YES   |  YES      | YES
        -------------------------------------------------------------------
        HDF3.2         |         |         |          |           |
         -RIS8         |  YES    |    YES  |  YES(1)  | YES(1)    | YES(1)
         -RIS24        |  YES    |    YES  |  YES(1)  | YES(1)    | YES(1)
         -PALETTE      |  YES    |    YES  |  YES     | YES       | YES
         -ANNOTATION   |  YES    |    YES  |  YES     | YES       | YES
         -SDS DFSD     |  YES    |    YES  |  YES     | YES(2)    | YES(2,3)
         -VData        |  YES    |    YES  |  YES     | YES       | YES
         -Vgroup       |  YES    |    YES  |  YES     | YES       | YES
        -------------------------------------------------------------------
        HDF3.3         |         |         |          |           |
         -RIS8         |  YES    |    YES  |  YES     | YES       | YES
         -RIS24        |  YES    |    YES  |  YES     | YES       | YES
         -PALETTE      |  YES    |    YES  |  YES     | YES       | YES
         -ANNOTATION   |  YES    |    YES  |  YES     | YES       | YES
         -SDS SD       |  YES    |    YES  |  YES     | YES(2)    | YES(2,3)
         -SDS DFSD     |  YES    |    YES  |  YES     | YES(2)    | YES(2,3) 
         -VData        |  YES    |    YES  |  YES     | YES       | YES
         -Vgroup       |  YES    |    YES  |  YES     | YES       | YES
        -------------------------------------------------------------------
        HDF4.0         |         |         |          |           |
         -GR           |  YES    |    YES  |  YES     | YES       | YES
         -RIS8         |  YES    |    YES  |  YES     | YES       | YES
         -RIS24        |  YES    |    YES  |  YES     | YES       | YES
         -PALETTE      |  YES    |    YES  |  YES     | YES       | YES
         -MFAN         |  YES    |    YES  |  YES     | YES       | YES
         -ANNOTATION   |  YES    |    YES  |  YES     | YES       | YES
         -SDS SD       |  YES    |    YES  |  YES     | YES       | YES(3) 
         -SDS DFSD     |  YES    |    YES  |  YES     | YES(2)    | YES(2,3) 
         -VData        |  YES    |    YES  |  YES     | YES       | YES
         -Vgroup       |  YES    |    YES  |  YES     | YES       | YES
        -------------------------------------------------------------------
        HDF4.1         |         |         |          |           |
         -GR           |  YES    |    YES  |  YES     | YES       | YES
         -RIS8         |  YES    |    YES  |  YES     | YES       | YES
         -RIS24        |  YES    |    YES  |  YES     | YES       | YES
         -PALETTE      |  YES    |    YES  |  YES     | YES       | YES
         -MFAN         |  YES    |    YES  |  YES     | YES       | YES
         -ANNOTATION   |  YES    |    YES  |  YES     | YES       | YES
         -SDS SD       |  YES    |    YES  |  YES     | YES       | YES
         -SDS DFSD     |  YES    |    YES  |  YES     | YES       | YES
         -VData        |  YES    |    YES  |  YES     | YES       | YES
         -Vgroup       |  YES    |    YES  |  YES     | YES       | YES

   (1)  except for JPEG compression
   (2)  except for gzip and nbit compression
   (3)  except for chunking and chunking with compression

   NOTES:
     - The table above does not include the low-level compression
       interface, which was introduced in HDF 4.0.
  
     - The SD interface should always be able to read an HDF file
       that was created with the DFSD interface. 

     - With HDF 4.1r2, the SD dimension representation introduced in 4.0r1
       will ONLY be used by default.  To be read by earlier versions of
       the software, the SDsetdimval_comp must be called to store the old
       and new dimension representations in an HDF file.
 
     - Old HDF libraries will NOT always be able to read HDF data written 
       by newer version HDF libraries. For example, HDF3.1 can not read 
       16-bit integer SDS's because HDF 3.1 did not support this data type. 

     - In HDF 4.1r1, chunking and Vdata/Vgroup attributes were
       added.  Previous releases will not be able to read data which
       was created using these features.

Can application programs which work with old versions of the HDF library always be compiled with new versions of HDF?

As HDF evolves some functions have to be changed or removed. For example, in HDF3.2 some functions' formal parameters which were passed by value in HDF3.1 have to be passed by reference in order to support new number types. When this happens, old application programs need to be modified so that they can work with the new library.

Our policy is as follows: Keep existing functions unchanged as much as possible; create new functions when necessary to accommodate new features; if a new function covers the feature of an existing old function, the old function should still be callable by old application programs; should an old function be phased out, the users will be forewarned and encouraged to switch to the new function; an old function will be removed from the library only if it is in conflict with the implementation of new features.


How does the 'integration of netCDF with HDF' affect application programmers?

The mfhdf library was designed to be completely transparent to the programmer. HDF supports a "multi-file" SDS interface and the complete netCDF interface as defined by Unidata netCDF Release 2.3.2. (Please note that HDF4 cannot read NetCDF 64-bit files.)

Using either interface, you are able to read XDR-based netCDF files, HDF-based netCDF files and pre-HDF4.x HDF files. The library determines what type of file is being accessed and handles it appropriately. Any of the above types of files may be modified. However, the library will only create new files based on HDF (you can't create new XDR-based netCDF files).

Summary of HDF and XDR file interoperability for the HDF and netCDF application interfaces:

                 | Files created | Files created    |    Files written         |
                 | by DFSD       | by SD interface  |    by NC interface       |
                 | interface     |                  |                          |
                 |     HDF       |     HDF          |  HDF     | Unidata netCDF|   
                 | ----------------------------------------------------------- |
Accessed by DFSD |      Yes      |     Yes          |   Yes    |     No        |
                 |               |                  |          |               |
Accessed by SD   |      Yes      |     Yes          |   Yes    |     Yes       |
                 |               |                  |          |               |
Accessed by NC   |      Yes      |     Yes          |   Yes    |     Yes       |
                 |               |                  |          |               |
For more information, you can refer to the section entitled HDF Interface vs. netCDF Interface in the SD chapter of the User's Guide.

Does HDF support data compression?

HDF 4.0 (and later releases) supports a low-level compression interface, which allows any data-object to be compressed using a variety of algorithms.

Currently only three compression algorithms are supported: Run-Length Encoding (RLE), adaptive Huffman, and an LZ-77 dictionary coder (the gzip 'deflation' algorithm). Plans for future algorithms include an Lempel/Ziv-78 dictionary coding, an arithmetic coder and a faster Huffman algorithm.

HDF 4.0 (and later releases) supports n-bit compression for SDSs.

HDF 4.0 (and later releases) supports RLE (Run Length Encoding), IMCOMP, and JPEG compression for raster images.

New with HDF 4.1 is support for "chunking" and "chunking with compression". Data chunking allows an n-dimensional SDS or GR image to be stored as a series of n-dimensional chunks. See the HDF User's Guide for more information.

With HDF4.2r0, HDF supports SZIP compression. For further information regarding it, see SZIP Compression in HDF Products.

NOTE:   Compression and chunking are limited to fixed sized datasets. You cannot compress or chunk a dataset that has unlimited dimensions.


Is there a mailing list for HDF discussions and questions?

If you want to broadcast HDF technical questions to other HDF users in order to solicit their assistance, you can subscribe to the HDF-Forum mailing list. See the Community Support page for information on subscribing to the HDF-Forum and accessing prior postings to this mailing list.


How do I contribute my software to the HDF user community?

There are two ways that you can do this:


How do I make a bug report?

All bug reports, comments, suggestions and questions should go to:

Attached below is a bug report template. It is not required but gives an idea of the kind of information we need in order to resolve a problem.

------------------  Template for bug report  ------------------------

VERSION:
   HDF 4.N.N

USER:
   [Name and Email Address]

SYNOPSIS:
   [Brief description of the problem and where it is located]

MACHINE / OPERATING SYSTEM:
   [Platform and platform version.  On Unix platforms, please 
    include the output from "uname -a".]

COMPILER:
   [Compiler and compiler version]

DESCRIPTION:
   [Detailed description of problem. ]

REPEAT BUG BY:
   [What you did to get the error; include test program or session
    transcript if at all possible.  If you include a program, make
    sure it depends only on libraries in the HDF distribution, not
    on any vendor or third-party libraries.  Please be specific;
    if we can't reproduce it, we can't fix it. Tell us exactly what
    we should see when the program is run.  NOTE:  It helps us a lot
    if the example program can be written in C and can be run 
    easily on Unix. ]

SAMPLE FIX:
   [Fix or patch if you have one]

------------------  End of Bug Report Template  ----------------------


- - Last modified: 28 June 2017