This file was created for the HDF4.2r0. release to store the documents
that were in the release_notes directory of the main HDF4 source tree
since the 4.0.alpha release. See also HISTORY.txt file for more information.

File contains the following *.txt files:


Fortran_APIs.txt
JPEG.txt
Pablo.txt
comp_SDS.txt
compile.txt
compression.txt
dimval.txt
external_path.txt
hdp.txt
install_winNT.txt
macintosh.txt
mf_anno.txt
mf_ris.txt
new_functions.txt
page_buf.txt
sd_chunk_examples.txt
vattr.txt
windows.txt

To search for a particular document use "filename.txt=" string, for example 
to search for the beginning of the new_functions.txt file,
use "new_functions.txt=" string

================================Fortran_APIs.txt============================

Problem:
========

In HDF4.0r1 and previous versions of HDF several Fortran 
routines declared a formal parameter as character*(*) or 
integer while the actual parameter was a character or 
a numeric type. This caused problems on some systems, 
such as VMS and T3D. 

With HDF 4.0r2 and later releases, these routines have either 
been replaced by 2 routines, one for character type parameters 
and another for numeric type parameters; or, a new routine has 
been added for char type parameters and the old routine is used 
for numeric type parameters only. Those routines that were replaced 
by two routines should be phased out in the future. However, in 
order to not break currently working applications they are 
still supported. New applications should use the new routines. 

Routines and parameters affected:
================================
1. Write vdata

Old:
    vsfwrit(vsid, databuf, n_rec, interlace)
        character*(*) databuf

HDF4.0r2:
    
    Write to a vdata from a character buffer:
      vsfwrtc(vsid, cbuf, n_rec, interlace) 
           character*(*) cbuf
    Write to a vdata from an integer buffer (for numeric values):
      vsfwrt(vsid, buf, n_rec, interlace)  
           integer  buf

2. Read vdata
 
Old:
    vsfread(vsid, buf, n_recs, interlace)
         character*(*) buf
HDF4.0r2:
    Read records into a character buffer:
      vsfrdc(vsid, cbuf, n_recs, interlace)
          character*(*) cbuf
    Read records into an integer buffer (for numeric values):
      vsfrd(vsid, buf, n_recs, interlace)
          integer buf

3. High level function for creating a single field single
   component vdata
 
Old:
    vhfsd(f, field, buf, n, dtype, vsname, vsclass)
         integer buf
HDF4.0r2:
    Store a simple character dataset in a vdata:
      vhfscd(f,field,cbuf,n,dtype,vsname,vsclass)
         character*(*) cbuf
    Store a simple numeric dataset in a vdata
      vhfsd(f, field, buf, n, dtype, vsname, vsclass)
         integer buf

4. High level function for creating a single field multi-
   component vdata
Old:
    vhfsdm (f,field,buf,n,dtype,vsname,vsclass,order)
         integer buf
HDF4.0r2:
    Store an aggregate char dataset in a vadata:      
      vhfscdm (f,field,cbuf,n,dtype,vsname,vsclass,order)
         character*(*) cbuf
    Store a simple numeric dataset in a vdata
      vhfsdm(f,field,buf,n,dtype,vsname,vsclass,order)
         integer buf

5. Write GR image
Old:
    mgwrimg(riid, start,stride,count,data)
      <valid numeric type> data
HDF4.0r2:
    Write character type image data
     mgwcimg(riid, start, stride, count, cdata)
         character*(*)  cdata
    Write numeric type image data
      mgwrimg(riid, start,stride,count,data)
      <valid numeric type> data

6. Read GR image
Old:
    mgrdimg(riid,start,stride,count,data)
      integer data
HDF4.0r2:
    Read character type image data
      mgrcimg(riid,start,stride,count,cdata)
          character*(*) cdata
    Read numeric type image data
      mgrdimg(riid,start,stride,count,data)
          <valid numeric type> data

7. Write LUT
Old:
    mgwrlut(lutid,ncomp,data_type,interlace,nentries,data)
      <valid numeric type> data
HDF4.0r2:
    Write character type palette:
      mgwclut(lutid,ncomp,data_type,interlace,nentries,cdata)
          character*(*) cdata
    Write numeric type palette:
      mgwrlut(lutid,ncomp,data_type,interlace,nentries,data)
          <valid numeric type> data

8. Read LUT
Old:
    mgrdlut(lutid, data)
      <valid numeric type> data
HDF4.0r2:
    Read char type palette:
      mgrclut(lutid,cdata)
        character*(*) cdata
    Read numeric type palette:
      mgrdlut(lutid, data) 
        <valid numeric type> data

9. Set GR attribute
Old:
    mgsattr(riid, name, nt, count, data)
      character*(*) data
HDF4.0r2:
    Add char type attribute to a raster image
      mgscatt(riid, name, nt, count, cdata)
        character*(*) cdata
    Add a numeric attribute to a raster image
      mgsnatt(riid, name, nt, count, data)
        integer data

10. Get GR attribute
Old:
    mggattr(riid, index, data)
      <valid numeric type> data
HDF4.0r2:
    Get a char type attribute:
      mggcatt(riid, index, cdata)
        character*(*) cdata
    Get a numeric type attribute:
      mggnatt(riid, index, data)
        integer data

11. Write SDS data
Old:
    sfwdata(sdsid,start,stride,end,values)
      <valid numeric type> values
HDF4.0r2
    Write char type SDS data
      sfwcdata(sdsid,start,stride,end,cvalues)
        character*(*) cvalues
    Write numeric type SDS data
      sfwdata(sdsid,start,stride,end,values)
        <valid numeric type> values

12. Read SDS data
Old:
    sfrdata(sdsid,start,stride,end,values)
      <valid numeric type> values
HDF4.0r2
    Read char type SDS data
      sfrcdata(sdsid,start,stride,end,cvalues)
        character*(*) cvalues
    Read numeric type SDS data
      sfrdata(sdsid,start,stride,end,values)
        <valid numeric type> values

13. Add an attribute to an object in SD interface
Old:
    sfsattr(id, name, nt, count, data)
      character*(*) data
HDF4.0r2
    Add a char type attribute to an object
      sfscatt(id, name, nt, count, cdata)
        character*(*) cdata
    Add a numeric type attribute to an object
      sfsnatt(id, name,nt, count,data)
        integer data

14. Get contents of an attribute
Old:
    sfrattr(id, index, buf)
      <valid numeric type> buf
HDF4.0r2:
    Get a char type attribute
      sfrcatt(id, index, cbuf)
        character*(*) cbuf
    Get a numeric type attribute
      sfrnatt(id, index, buf)
        <valid numeric type> buf   

15. Set fill value
Old:
    sfsfill(id, val)
      <valid numeric type> val
HDF4.0r2
    Set a char type fill value
      sfscfill(id, cval)
        character cval
    Set a numeric type fill value
      sfsfill(id, val)
        <valid numeric type> val

16. Get fill value
Old:
    sfgfill(id, val)
      <valid numeric type> val
HDF4.0r2
    Get char type fill value
      sfgcfill(id, cval)
        character cval
    Get numeric type fill value
      sfgfill(id, val)
        <valid numeric type> val


============================================================================
================================JPEG.txt====================================

Independent JPEG Group library
    Version 4.1b of the HDF-netCDF library uses v6a of the Independent
JPEG Group (IJG) JPEG file access library.  For most users of the HDF library,
this will be completely transparent.  For users who are integrating the HDF
library into an existing application which uses the IJG's JPEG library, linking
with the HDF library is now much simpler and should be completely painless.
The JPEG library will need to be linked with user's applications when raster
images are being used (whether they are compressed with JPEG or not).

     cc -o <myprog> myprog.c -I<include path> <path for libmfhdf.a> \
           <path for libdf.a> <path for libjpeg.a>

     Note: order of the libraries is important, the mfhdf library must be first
and be followed by the hdf library.
============================================================================
================================Pablo.txt===================================


Pablo Instrumentation of HDF
===========================
    This version of the distribution has support to create an instrumented 
    version of the HDF library(libdf-inst.a). This library along with
    the Pablo performance data capture libraries can be used to gather data
    about I/O behavior and procedure execution times.  

    More detailed documentation on how to use the instrumented version of
    the HDF library with Pablo can be found in the Pablo directory 
    '$(toplevel)/hdf/pablo'. 
     See the provided '$(toplevel)/hdf/pablo/README.Pablo'
     and the Postscript file '$(toplevel)/hdf/pablo/Pablo.ps'.

    At this time only an instrumented version of the core HDF library libdf.a 
    can be created. Future versions will have support for the SDxx interface
    found in libmfhdf.a. Current interfaces supported are ANxx, GRxx, DFSDxx,
    DFANxx, DFPxx, DFR8xx, DF24xx, Hxx, Vxx, and VSxx.

    To enable the creation of an instrumented library the following section
    in the makefile fragment($(toplevel)/config/mh-<os>) must be uncommented 
    and set.

    # ------------ Macros for Pablo Instrumentation  --------------------
    # Uncomment the following lines to create a Pablo Instrumentation
    # version of the HDF core library called 'libdf-inst.a'
    # See the documentation in the directory 'hdf/pablo' for further 
    # information about Pablo and what platforms it is supported on
    # before enabling. 
    # You need to set 'PABLO_INCLUDE' to the Pablo distribution 
    # include directory to get the files 'IOTrace.h' and 'IOTrace_SD.h'.
    #PABLO_FLAGS  = -DHAVE_PABLO
    #PABLO_INCLUDE = -I/hdf2/Pablo/Instrument.HP/include

    After setting these values you must re-run the top-level 'configure' script.
    Make sure that your start from a clean re-build(i.e. 'make clean') after
    re-running the toplevel 'configure' script and then run 'make'.
    Details on running configure can be found in the section
    'General Configuration/Installation - Unix' found in the top-level 
    installation file '$(toplevel)/INSTALL'.
============================================================================
================================comp_SDS.txt================================

Limitations of compressed SDS datasets
    Due to certain limitations in the way compressed datasets are stored, data
which has been compressed is not completely writable in ways that uncompressed
datasets are.  The "rules" for writing to a compressed dataset are as follows:

    (1) Write an entire dataset that is to be compressed.  I.e. build the
        dataset entirely in memory, then write it out with a single call.
 
    (2) Append to a compressed dataset.  I.e. write to a compressed dataset
        that has already been written out by adding to the unlimited
        dimension for that dataset.
 
    (3) For users of HDF 4.1, write to any subset of a compressed dataset
        that is also chunked.

    Generally speaking, these mean that it is impossible to overwrite existing
compressed data which is not stored in "chunked" form.  This is due to 
compression algorithms not being suitable for "local" modifications in a
compressed datastream.  Please send questions about compression to the
general HDF support e-mail address:  help@hdfgroup.org

Compression for HDF SDS
The SDsetcompress and SDsetnbitdataset functions are used as
higher-level routines to access the HCcreate function (HCcreate is described
in the reference manual).  SDsetnbitdataset allows for the storage of 1-32 bit 
integer values (instead of being restricted to 8, 16 or 32-bit sizes) in a
scientific dataset. SDsetcompress can be used to compress a scientific dataset
through the SD interface instead of dropping down to the lower-level H
interface.

N-bit SDS using SDsetnbitdataset:

    The interface to SDsetnbitdataset is described below:

    intn SDsetnbitdataset(sds_id,start_bit,bit_len,sign_ext,fill_one); 

    int32 sds_id - The id of a scientific dataset returned from SDcreate or
        SDselect.
        
    intn start_bit - This value determines the bit position of the highest end
        of the n-bit data to write out. Bits in all number-types are counted
        from the right starting with 0. For example, in the following bit data,
        "01111011", bits 2 and 7 are set to 0 and all the other bits are set to
        one. 

    intn bit_len - The number of bits in the n-bit data to write, including the
        starting bit, counting towards the right (i.e. lower bit numbers). For
        example, starting at bit 5 and writing 4 bits from the following bit
        data, "01111011", would write out the bit data, "1110", to the dataset
        on disk. 

    intn sign_ext - Whether to use the top bit of the n-bit data to sign-extend
        to the highest bit in the memory representation of the data. For
        example, if 9-bit signed integer data is being extracted from bits
        17-25 (nt=DFNT_INT32, start_bit=25, bit_len=9, see below for full
        information about start_bit & bit_len parameters) and the bit in
        position 25 is a 1, then when the data is read back in from the disk,
        bits 26-31 will be set to a 1, otherwise bit 25 will be a zero and bits
        26-31 will be set to 0. This bit-filling takes higher precedence (i.e.
        is performed after) the fill_one (see below) bit-filling. 

    intn fill_one - Whether to fill the "background" bits with 1's or 0's.
        The "background" bits of a n-bit dataset are those bits in the
        in-memory representation which fall outside of the actual n-bit field
        stored on disk. For example, if 5 bits of an unsigned 16-bit integer
        (in-memory) dataset located in bits 5-9 are written to disk with the
        fill_one parameter set to TRUE (or 1), then when the data is read back
        into memory at a future time, bits 0-4 and 10-15 would be set to 1. If
        the same 5-bit data was written with a fill_one value of FALSE (or 0),
        then bits 0-4 and 10-15 would be set to 0.  This setting has a lower
        precedence (i.e. is performed first) than the sign_ext setting. For
        example, using the sign_ext example above, bits 0-16 and 26-31 will
        first be set to either 1 or 0 based on the fill_one parameter, and then
        bits 26-31 will be set to 1 or 0 based on bit-25's value. 
    
    RETURNS - SUCCEED (0) or FAIL (-1) for success/failure.

    The corresponding FORTRAN function name is sfsnbit which takes the
    same parameters in the same order.

    For example, to store an unsigned 12-bit integer (which is represented
    unpacked in memory as an unsigned 16-bit integer), with no sign extension
    or bit filling and which starts at bit 14 (counting from the right with bit
    zero being the lowest) the following setup & call would be appropriate: 

    intn sign_ext = FALSE; 
    intn fill_one = FALSE; 
    intn start_bit= 14; 
    intn bit_len = 12; 
    SDsetnbitdataset(sds_id,start_bit,bit_len,sign_ext,fill_one); 

    Further reads and writes to this dataset would transparently convert the
    16-bit unsigned integers from memory into 12-bit unsigned integers stored
    on disk.

    More details about this function can be found in the HDF library reference
    manual.

Compressed SDS data using SDsetcompress:

        The SDsetcompress function call contains a subset of the parameters to
the HCcreate function call described in compression.txt and performs the same
types of compression.

    The interface to SDsetcompress is described below:

    intn SDsetcompress(sds_id,comp_type,c_info);

    int32 sds_id - The id of a scientific dataset returned from SDcreate or
        SDselect.

    int32 comp_type - The type of compression to encode the dataset with.
        The values are the same as for HCcreate:
            COMP_CODE_NONE - for no compression
            COMP_CODE_RLE - for RLE encoding
            COMP_CODE_SKPHUFF - for adaptive Huffman
            COMP_CODE_DEFLATE - for gzip 'deflation'

    comp_info *c_info - Information needed for the encoding type chosen.
        For COMP_CODE_NONE and COMP_CODE_RLE, this is unused and can be set to
        NULL.  For COMP_CODE_SKPHUFF, the structure skphuff in this union needs
        information about the size of the data elements in bytes (see example
        below).  For COMP_CODE_DEFLATE, the structure deflate in this union
        need information about "effort" to try to compress with (see example
        below).  For more information about the types of compression 
        see the compression.txt document in this directory.

    RETURNS - SUCCEED (0) or FAIL (-1) for success/failure.

    Similarly to the HCcreate function, SDsetcompress can be used to create
    compressed dataset or to compress existing ones.

    For example, to compress unsigned 16-bit integer data using the adaptive
    Huffman algorithm, the following setup and call would be used:

    comp_info c_info;
    c_info.skphuff.skp_size=sizeof(uint16);
    SDsetcompress(sds_id,COMP_CODE_SKPHUFF,&c_info);

    Further reads and writes to this dataset would transparently convert the
    16-bit unsigned integers from memory into a compressed representation on
    disk.

    For example, to compress a dataset using the gzip deflation algorithm, with
    the maximum effort to compress the data, the following setup and call would
    be used:

    comp_info c_info;
    c_info.deflate.level=9;
    SDsetcompress(sds_id,COMP_CODE_DEFLATE,&c_info);

    Currently, SDsetcompress is limited to creating new datasets or appending
    new slices/slabs onto existing datasets.  Overwriting existing data in a
    dataset will be supported at some point in the future.

    More details about this function can be found in the HDF library reference
    manual.

============================================================================
================================compile.txt=================================

 
COMPILING A PROGRAM
                  
Following are instructions for compiling an application program on the 
platforms supported by HDF, using the binaries that we provide.  For
Unix, the information on options to specify comes from the configuration 
files (mh-*) in the HDF source code (under ../HDF4.1r5/config).   

In general, you compile your program as shown below.  If your platform is 
not specified in the section, "INSTRUCTIONS FOR SPECIFIC PLATFORMS", then 
use these instructions.  If you are unable to compile your program on Unix, 
please check the configuration file for your platform for the correct 
options.

C:
    cc -o <your program> <your program>.c -I<path for hdf include directory>\
       -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz

   or

    cc -o <your program> <your program>.c -I<path for hdf include directory> \
          <path for libmfhdf.a>  <path for libdf.a> \
          <path for libjpeg.a> <path for libz.a>

FORTRAN:
    f77 -o <your program> <your program>.f  \
        -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz
   
   or
   
    f77 -o <your program> <your program>.f  \
          <path for libmfhdf.a>  <path for libdf.a> \
          <path for libjpeg.a> <path for libz.a>

NOTE: The order of the libraries is important: libmfhdf.a first,
followed by libdf.a, then libjpeg.a and libz.a.  The libjpeg.a
library is optional.

INSTRUCTIONS FOR SPECIFIC PLATFORMS
===================================

FreeBSD:
-------
C:
    cc -ansi -Wall -pedantic -o <your program> <your program>.c \  
        -I<path for hdf include directory> \
        -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz

FORTRAN:
    f77 -O -o <your program> <your program>.f  \
        -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz
 
Linux:
-----
C:
    gcc -ansi -D_BSD_SOURCE -o <your program> <your program>.c \  
        -I<path for hdf include directory> \
        -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz

FORTRAN:
    g77 -o <your program> <your program>.f  \
        -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz

Solaris:
-------
   The -lnsl is necessary in order to include the xdr library.

C:
    cc -Xc -xO2 -o <your program> <your program>.c  \
       -I<path for hdf include directory>\
       -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz \
       -L/usr/lib -lnsl

FORTRAN:
    f77 -O -o <your program> <your program>.f  \
        -L<path for hdf libraries> -lmfhdf -ldf -ljpeg -lz \
        -L/usr/lib -lnsl

Windows NT/98/2000:
------------------
Using Microsoft Visual C++ version 6.x:

Under Tools->Options, select the folder, Directories:
   Under "Show directories for", select "Include files".
   Add the following directories:
      C:<path to HDF includes>\INCLUDE  

Under "Show directories for", select "Library files":
   Add the following directories:
      C:<path to HDF libs>\LIB        

Under Project->Settings, select folder, Link:
   Add the following libraries to the beginning of the list of
   Object/Library Modules:
         hd415.lib hm415.lib (single-threaded release version)
         hd415d.lib hm415d.lib (single-threaded debug version)

         hd415m.lib hm415m.lib (multi-threaded release version)
         hd415md.lib hm415md.lib (multi-threaded debug version)



============================================================================
================================compression.txt=============================

Compression Algorithms and interface

    The low-level compression interface allows any data object to be 
compressed using a variety of algorithms.  This is completely transparent 
to users once the data has been compressed initially - further data written
to an object or read from it are compressed or decompressed internally to
the library, without user intervention.  (For information on compressing
SDS datasets, see the ../release_notes/comp_SDS.txt file.)  

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.

    The public interface for this routine is contained in the user-level
function call, HCcreate.  The interface to HCcreate is described below:

int32 HCcreate(id,tag,ref,model_type,m_info,coder_type,c_info);
    int32 id;                IN: the file id to create the data in (from Hopen)
    uint16 tag,ref;          IN: the tag/ref pair of the data object which
                                    is to be compressed
    comp_model_t model_type; IN: the type of modeling to use, currently
                                    only COMP_MODEL_STDIO is supported, which
                                    indicates data is transferred in the
                                    same way as C I/O functions operate.
    model_info *m_info;      IN: Information needed for the modeling type chosen
                                    Nothing needed for COMP_MODEL_STDIO,
                                    so NULL can be used.
    comp_coder_t coder_type; IN: the type of encoding to use from the following:
                                    COMP_CODE_NONE - for no compression
                                    COMP_CODE_RLE - for RLE encoding
                                    COMP_CODE_SKPHUFF - for adaptive Huffman
                                    COMP_CODE_DEFLATE - for gzip 'deflation'
    comp_info *c_info;       IN: Information needed for the encoding type chosen
                                    For COMP_CODE_NONE and COMP_CODE_RLE,
                                    this is unused and can be set to NULL.
                                    For COMP_CODE_SKPHUFF, the structure skphuff
                                    in this union needs information about the
                                    size of the data elements in bytes (see
                                    examples below).
                                    For COMP_CODE_DEFLATE, the structure deflate
                                    in this union needs information about the
                                    "effort" to encode data with.  Higher
                                    values of 'level' member indicate more
                                    compression effort.  Values may range from
                                    0 (minimal compression, fastest time) to
                                    9 (maximum compression, slowest time).
    RETURNS
        Return an AID to the newly created compressed element, FAIL on error.

    HCcreate will compress an existing data object with the specified 
compression method, or it can create a new data object which will contain
compressed data when it is written to.  In either case, Hendaccess must be
called to release the AID allocated by HCcreate.  In the first two examples
below the datasets already exist, in the final example the dataset is created
by the HCcreate call.  There is currently no FORTRAN equivalent for this
function.  More details about this function can be found in the HDF reference
manual.

The following example shows how to compress a scientific dataset data object
(which is composed of multi-dimensional 32-bit integer data) using the
adaptive Huffman encoding:
    {
        int32 aid;
        comp_info c_info;

        c_info.skphuff.skp_size=sizeof(int32);
        aid=HCcreate(file_id, DFTAG_SD, ref, COMP_MODEL_STDIO, NULL, 
            COMP_CODE_SKPHUFF,&c_info);
        .
        .
        <access data object>
        .
        .
        Hendaccess(aid);
    }

The following example shows show to compress a raster image data object
using the RLE algorithm:
    {
        int32 aid;

        aid=HCcreate(file_id, DFTAG_RI, ref, COMP_MODEL_STDIO, NULL, 
            COMP_CODE_RLE,NULL);
        .
        .
        <access data object>
        .
        .
        Hendaccess(aid);
    }

The following example shows how to create a new data object whose data
will compressed as it is written:
    {
        int32 aid;

        aid=HCcreate(file_id, DFTAG_RI, ref, COMP_MODEL_STDIO, NULL, 
            COMP_CODE_RLE,NULL);
        .
        .
        Hwrite(aid,len,data);
        .
        .
        Hendaccess(aid);
    }
============================================================================
================================dimval.txt==================================

             New Version of Dimension Values
             ===============================

HDF4.0b1 and previous releases use a vgroup to represent a dimension.
The vgroup has a single field vdata with class "DimVal0.0".
The vdata has <dimension size> number of records, each record has a 
fake value from 0, 1, 2 ... , (<dimension size> - 1 ). The fake values 
are not really required and take a lot of space. For applications that 
create large one dimensional array datasets the disk space taken by 
these fake values almost double the size of the HDF file. In order to 
omit the fake values, a new version of dimension vdata has been proposed.

The new version uses the same structure as the old version. The only
differences are that the vdata has only 1 record with value 
<dimension size> and that the vdata's class is "DimVal0.1" to 
distinguish it from the old version.

However, existing tools and utilities which were compiled with the old
version can't recognize the new dimensions of hdf files created using
HDF4.0b2 or later version. This could cause problems for HDF users.
To solve this problem, we are planning to implement a transitional 
policy:

1. Starting from HDF4.0b2 both versions of the dimension will be
   created by default. The old tools recognize the "DimVal0.0" 
   dimension.
2. A new function SDsetdimval_comp (sfsdmvc) is added which can 
   be called for a specific dimension to suppress the creation 
   of the "DimVal0.0" vdata for that dimension. Users who store 
   big 1D arrays should use this function to create "DimVal0.1" 
   only.  See the man page for sdsetdimval_comp.3. 
3. A new function SDisdimval_bwcomp (sfisdmvc) is added which 
   can be called to get the current compatibility mode of a 
   dimension. See the man page for sdisdimval_bwcomp.3. 
4. HDF4.0b2 and later version of HDF libraries can recognize both
   old and new versions of dimensions. This means old HDF files can 
   always be read by new HDF libraries.
5. HDF4.1 will create only "DimVal0.1" by default and function 
   SDsetdimval_comp should be called if "DimVal0.0" is also desired.  
   The transition time period is ended.
6. Existing tools and utilities should be re-compiled with HDF4.0b2
   or later releases during that transition time period.
7. A new utility will be written to remove redundant "DimVal0.0" from 
   the files created during the transition time period.
8. A new utility will be written to convert "DimVal0.1" to "DimVal0.0"
   for special cases.

Please send bug reports, comments and suggestions to help@hdfgroup.org.


============================================================================
================================external_path.txt===========================

User Settable File Location for External Elements

Users sometimes encounter situations (e.g., disk space shortage,
different filesystem names) that the external file containing the data
of the external element has to reside in a directory different from the
one it was created.  The user may set up symbolic pointers to forward
the file locations but this does not work if the external filename is
an absolute path type containing directory components that do not exist
in the local system.

A new feature is added such that an application can provide a list of
directories for the HDF library to search for the external file.  This
is set by the function call HXsetdir or via the environment variable
$HDFEXTDIR.  See the man page HXsetdir(3) for details.

A similar feature is also added to direct the HDF library to create the
external file of a _new_ external element in the given directory.  An
example for the need of this feature is that an application wants to
create multiple external element files with certain naming conventions
(e.g., Data950101, Data950102) while all these files share a common
parent directory (project123/DataVault).  Different users will have a
different choice of the common parent directory.  This can be set by
the function call HXsetcreatedir or the environment variable
$HDFEXTCREATEDIR.  See the man page for HXsetcreatedir (1) for detail.

============================================================================
================================hdp.txt=====================================

           hdp -- HDF dumper 


NAME
     hdp - HDF dumper

SYNOPSIS
     hdp [hdp options] hdp command [command options] <filename list>

DESCRIPTION
     
     hdp is a command line utility designed for quick display of 
     contents and data of HDF3.3 objects. It can list the contents 
     of hdf files at various levels with different details. It can 
     also dump the data of one or more specific objects in the file. 


HDP OPTIONS

    Currently, there is only one option.

    -H  Display usage information about the specified command.
        If no command is specified, -H lists all available commands.


HDP COMMANDS

     hdp currently has two types of commands: list and dump. Other 
     types of commands such as those for editing may be added in the
     future.
     
     hdp list <filename list>
         lists contents of files in <filename list> 

     hdp dumpsds <filename list>
         displays data of NDGs and SDGs in the listed files.

     hdp dumpvd <filename list>
         displays data of vdatas in the listed files.
   
     hdp dumpvg <filename list>
         displays data of objects in vgroups in the listed files.

     hdp dumprig <filename list>
         displays data of RIGs in the listed files.

     hdp dumpgr <filename list>
         displays data of general RIGs in the listed files.

HDP COMMAND OPTIONS

(Note: options preceded by an * have not yet been implemented.)


     hdp list [format options] [content ops] [filter ops] [order ops] 
	      <filename list>
     --------------------------------------------------------------------------

      Format options
          decide how the info of objects will be presented on the screen.
       
        -s  (short format) under each tag #, all ref's of that tag are listed
            in one or more lines, same as the output of hdfls. (default)

        -l  (long format) one object per line. Each line contains tag-name, 
            tag/ref and the index of this tag in the file.(e.g., the ith NDG 
	    in the file).

        -d  debug format, one object per line. Each line contains tag_name,
            tag/ref, index, offset, and length, same as the output of hdfls -d.

	no	tagname	   tag	  ref	index/tag	offset	length
        --      -------    ---    ---   ---------       ------  ------

	1	DFTAG_NT   106      2      1          
        2       DFTAG_SD   701      3      1
        ...


         Content options
              allow contents be displayed.
  
            -n  display the name or label of the object, if there is any.
                -n puts you in -l format automatically.

            -c  display the class of the object, if there is any. -l format.

            -a  display description of the object, if there is any. -l format.

         Filter options
              select certain type of objects to display, default is all.

            -g  display groups only. Objects which do not belong to 
                any group will not be displayed. Nested groups will be
                displayed in tree format.

            -t <number>  display objects with specified tag number . e.g. 
                         720 for NDG.
            -t <name>    display objects with specified tag name.

         Order options
              sort the output list in different orders.
 
            -ot  by tag # (default)
            -of  by the order in file DDlist.
            -og  by group
            -on  by name(label)

    hdp dumpsds [filter ops] [contents ops] [output ops] <filename list>
    --------------------------------------------------------------------
         Filter options
              specify which SDS to dump.

             -i <index> dump SDS's with indices specified in <index>; 
			  indices correspond to the order of the SDS in the file
             -r <ref>    dump SDS's with reference numbers specified in <ref>
             -n <name>   dump SDS's with names specified in <name>
             -a           dump all SDS's in the file. (default)

	     Options -i, -r, and -n can be used inclusively to specify
	     different SDS's.

          Content options

             -v    display everything including all annotations (default)
             -h    header only, no annotation for elements or data
             -d    data only, no tag/ref 

	     These options are exclusive.

          Output options

             -o <filename> specify <filename> as output file name
             -b            binary output
             -x            ascii text output (default)

	     Options -b and -x are exclusive, but each can be used with
	     option -o.

	  Format options
	     -c    print space characters as they are, not \digit
	     -g    do not print data of file (global) attributes
	     -l    do not print data of local attributes
	     -s    do not add carriage return to a long line - dump as a stream

	     Options in this category can be used inclusively.

	  Note: Any combination of an option from each of the categories can
		be used as long as the criteria of that category are met.

    hdp dumpvd [filter ops] [contents ops] [output ops] <filename list>
    --------------------------------------------------------------------
         Filter options
              specify which vdata to dump.

             -i <index>   dump vdatas with indices in <index>; indices 
			  correspond to the order of the vdatas in the 
			  files
             -r <ref>     dump vdatas with reference numbers specified in
			  <ref>
             -n <name>    dump vdatas with names specified in <name>
             -c <class>   dump vdatas with classes specified in <class>
             -a           dump all vdatas in the file. (default)

          Content options

             -v    display everything including all annotations (default)
             -h    header only, no annotation for elements or data
             -d    data only, no tag/ref
             -f <fields> dump data of specified fields

          Output options

             -o <fn>    specify fn as output file name
           * -b         binary file
             -t         text ascii file (default)

    hdp dumpvg [filter ops] [contents ops] [output ops] <filename list>
    --------------------------------------------------------------------
         Filter options
              specify which vgroups to dump.

             -i <index>   dump vgroups with indices specified in <index>; 
			  indices correspond to the order of the vgroups 
			  specified in the files
             -r <ref>     dump vgroups with reference numbers specified in <ref>
             -n <name>    dump vgroups with names specified in <name>
             -c <class>   dump vgroups with classes specified in <class>
             -a           dump all vgroups in the file. (default)

          Content options

             -v    display everything including all annotations (default)
             -h    header only, no annotation for elements or data
             -d    data only

          Output options

             -o <fn>    specify fn as output file name
           * -b         binary file
             -t         text ascii file (default)

    Note: Unless the "-d" option is specified, a graphical representation of
	  the file will be given after the data has been displayed. 

    hdp dumprig [filter ops] [contents ops] [output ops] <filename list>
    --------------------------------------------------------------------
         Filter options
              specify which RIG to dump.

             -i <index>   dump RIGs with indices specified in <index>; 
			  indices correspond to the order of the RIGs 
			  specified in the files
             -r <ref>     dump RIGs with reference numbers specified in <ref>
             -a           dump all RIGs in the file. (default)
             -m  8|24     dump the RIGs of 8-bit or 24-bit. By default all
                             RIGs in the file will be dumped

          Content options

             -v    display everything including all annotations (default)
             -h    header only, no annotation for elements or data
             -d    data only

          Output options

             -o <fn>    specify fn as output file name
             -b         binary file
             -t         text ascii file (default)

    hdp dumpgr [filter ops] [contents ops] [output ops] <filename list>
    --------------------------------------------------------------------
         Filter options
              specify which general RIGs to dump.

             -i <index>   dump general RIG's with indices specified in 
                          <index>; indices correspond to the order of 
                          the RIG in the file
             -r <ref>     dump general RIG's with reference numbers 
                          specified in <ref>
             -n <name>    dump general RIG's with names specified in <name>
             -a           dump all general RIG's in the file. (default)

          Content options

             -v    display everything including all annotations (default)
             -h    header only, no annotation for elements or data
             -d    data only, no tag/ref

          Output options

             -o <fn>    specify fn as output file name
             -b         binary file
             -t         ascii text file (default)

          Note: any combination of an option from each of the three categories
                can be used; but no more than one option from one category is
                allowed.
============================================================================
================================install_winNT.txt===========================

      Install HDF4.1 Release 2 on Windows NT and Windows 95, and Alpha NT.

Since Windows NT, Windows '95 (Chicago) and Windows 3.1
(with the Win 32s extensions) all are designed to run the same 32-bit code, our 
decision is to support only 32-bit libraries and code on the MS-Windows 
platform. We are not planning on supporting any 16-bit versions in the 
foreseeable future.

The instructions which follow assume that you will be using one of 
the 'zip' files that we provide, either the binary code release
(hdf41r2.zip) or the source code release (hdf41r2s.zip).
In building HDF from source code you may select between 
two build environment options depending on your
application and environment needs.  Each option has it's own zip file:


Option I, (select Win32nof.zip)
Test and Utility configuration : HDF library, tests, and utilities, no fortran,
available for Win32 Intel platform only.

Option II, (select Win32full.zip)
Full configuration : HDF library, tests, and utilities, with fortran
This version has been built and tested using DEC Visual Fortran on both
the Win32 Intel platform and the Win32 Alpha platform.



Building from Binary Code Release (hdf41r2.zip)
===============================================
To install the HDF, JPEG, zlib and mfhdf libraries and utilities, 
it is assumed that you have done the following:
      

      1. Create a directory structure to unpack the library. For 
      example: 

	    c:\					(any drive)
           MyHDFstuff\				(any folder name)

      2. Copy the binary archive (HDF41r2.zip) to that directory 
      and unpack it by running WinZip on HDF41r2.zip (the binary archive).
      This should create a directory called 'HDF41r2' which 
      contains the following files and directories.

            c:\MyHDFstuff\HDF41r2\lib             ( Debug and Release versions of HDF libraries )
            c:\MyHDFstuff\HDF41r2\include         ( HDF include files )
            c:\MyHDFstuff\HDF41r2\bin             ( HDF utilities files )
            c:\MyHDFstuff\HDF41r2\release_notes   ( release notes )
            c:\MyHDFstuff\HDF41r2\install_NT_95   ( this file)

      
      3. If you are building an application that uses the HDF library 
         the following locations will need to be specified for locating
         header files and linking in the HDF libraries:
 
            C:\MyHDFstuff\HDF41r2\lib
            C:\MyHDFstuff\HDF41r2\include





Building from Source Code Release (hdf41r2s.zip)
===============================================

STEP I:  Preconditions

To build the HDF, JPEG, zlib and mfhdf libraries and utilities, 
it is assumed that you have done the following:
      
      1. Installed MicroSoft Developer Studio, and Visual C++ 5.0.
         Visual Fortran 5.0 is needed if you are going to build the
         full HDF Library with Fortran support.

      2. Set up a directory structure to unpack the library. For 
      example: 

	    c:\					(any drive)
           MyHDFstuff\				(any folder name)

      3. Copy the source distribution archive to that directory 
      and unpack it using the appropriate archiver options to
      create a directory  hierarchy.
         
      Run WinZip on HDF41r2s.zip (the entire source tree).
      This should create a directory called 'HDF41r2' which 
      contains several files and directories.
      
      ( Note for those using the Win32 Alpha platform:
        If you do not have a Winzip utility for your Alpha system
        you can download the needed executables from: 
        http://www.cdrom.com/pub/infozip ) 
       
STEP II: Select Installation type and Build.

You may select one of 2 ways to build the HDF library and 
utilities, depending on your environment and application needs.

Option I, (select Win32nof.zip)
Test and Utility configuration : HDF library, tests, and utilities, no fortran

Option II, (select Win32full.zip)
Full configuration : HDF library, tests, and utilities, with fortran



STEP III: Follow Instructions for Option I, or II


INSTRUCTIONS FOR OPTION I, TEST AND UTILITY INSTALLATION, NO FORTRAN
                 (Win32 Intel platform only)

	*** Builds hdf library, hdf utilities, 
	*** test programs and batch files. No fortran code.
 
        1. You will use Win32nof.zip 
           Unpack dev\win32nof.zip in directory dev\
           
            Run WinZip on 
               c:\myHDFstuff\HDF41r2\Win32nof.zip
               This archive contains a Developer Studio project "dev" and 
               two batch files. 
               40 project files (*.dsp files) will be created when 
               Win32nof.zip is expanded.

         2. Invoke Microsoft Visial C++ 5.0, go to "File" and select
            "Open Workspace" option. 
            Then open  c:\myHDFstuff\HDF41r2\dev.dsw workspace. 

	 3. Select "Build", then Select "Set Active Configuration".
            Select "dev -- Win32Debug" as active configuration.
            Select "Build" and "Build dev.exe" to
            build the Debug version of the HDF41r2 tree.

	 4. Select "Build", then Select "Set Active Configuration".
            Select "dev -- Win32Release" as active configuration.
            Select "Build" and "Build dev.exe" to
            build the Release version of the HDF41r2 tree.
	
         5. In command prompt window run the test batch file 
            win32noftst.bat in directory HDF41r2\.

         6. If all tests passed, run the installation batch file 
       win32ins.bat in directory HDF41r2\. Commands in this file will create
       subdirectories bin\, include\ and lib\ in HDF41r2\. The bin directory
       will contain the HDF utilities, the include directory will contain 
       header files, and the lib directory will contain:
               jpeg.lib     - JPEG Library
               jpegd.lib    - JPEG Library with DEBUG option
               libsrc.lib   - multi-file SDS Interface routines
               libsrcd.lib  - multi-file SDS Interface routines with DEBUG option
               src.lib      - multi-file Vdata Interface
                                         Vgroup Interface
                                         AN Interface
                                         GR Interface routines
               srcd.lib     - multi-file Vdata Interface
                                         Vgroup Interface
                                         AN Interface
                                         GR Interface routines with DEBUG option
               xdr.lib      - XDR Library
               xdrd.lib     - XDR Library with DEBUG option
               zlib.lib     - GNU Zip Library 
               zlibd.lib    - GNU Zip Library with DEBUG option 
  

INSTRUCTIONS FOR OPTION II, FULL INSTALLATION WITH FORTRAN

        
	*** Builds the hdf library, hdf utility programs, test programs,
	*** and batch files. Includes fortran source code to be
        *** compiled with Digital Visual Fortran on either a Win32 Intel
        *** machine or a Win32 Alpha machine.

	1. Unpack HDF41r2\Win32full.zip in directory HDF41r2\. 
         
        2. Invoke Microsoft Visial C++ 5.0, go to "File" and select
           "Open Workspace" option. 
           Then open  c:\myHDFstuff\HDF41r2\dev.dsw workspace. 

        3. Select "Build", then Select "Set Active Configuration".
           Select as the active configuration "dev -- Win32Debug" 
           if you have a Win32 Intel processor OR 
           select "dev-Win32AlphaDbg" if you have a Win32 Alpha processor. 
           Select "Build" and "Build dev.exe" to
           build the Debug version of the HDF41r2 tree.
           You will see that Digital Visual Fortran compiler is invoked
           by Visual C++ Development environment in compiling the fortran code.

        4. Select "Build", then Select "Set Active Configuration".
           Select as the active configuration"dev -- Win32Release"
           if you have a Win32 Intel processor OR
           select "dev-Win32AlphaRel" if you have a Win32 Alpha processor.
           Select "Build" and "Build dev.exe" to
           build the Release version of the HDF41r2 tree.
		
        5. In command prompt window run the test batch file which
       resides in the HDF41r2 directory.
       Run win32tst.bat if you have a Win32 Intel platform OR 
       run win32ALPHAtst.bat if you have the Win32 Alpha platform.

	6. If all tests passed, run the installation batch file which
       resides in the HDF41r2 directory
       Run win32ins.bat if you have a Win32 Intel platform OR
       run win32ALPHAins.bat if you have a Win32 Alpha platform.
       Commands in these files will create
       subdirectories bin\, include\ and lib\ in HDF41r2\. The bin directory
       will contain the HDF utilities, include directory will contain 
       header files, and the lib directory will contain:
               jpeg.lib     - JPEG Library
               jpegd.lib    - JPEG Library with DEBUG option
               libsrc.lib   - multi-file SDS Interface routines
               libsrcd.lib  - multi-file SDS Interface routines with DEBUG option               src.lib      - multi-file Vdata Interface
                                         Vgroup Interface
                                         AN Interface
                                         GR Interface routines
               srcd.lib     - multi-file Vdata Interface
                                         Vgroup Interface
                                         AN Interface
               xdrd.lib     - XDR Library with DEBUG option
               zlib.lib     - GNU Zip Library 
               zlibd.lib    - GNU Zip Library with DEBUG option
 


STEP IV:

BUILDING AN APPLICATION USING THE HDF LIBRARY - SOME HELPFUL POINTERS
=====================================================================

If you are building an application that uses the HDF library 
the following locations will need to be specified for locating
header files and linking in the HDF libraries:
 
            <top-level HDF directory>\lib
            <top-level HDF directory>\include

where <top-level HDF directory> may be C:\myHDFstuff\dev or C:\MyHDFstuff\HDF41r2\

Please refer to the <top-level HDF directory>\release_notes\compile.txt file
for more information on compiling an application with the HDF libraries.


MORE HELPFUL POINTERS
=====================
(as described in terms of installing the  nofortran configuration)

Here are some notes that may be of help if you are not familiar
with using the Visual C++ Development Environment.

Project name and location issues: 
         The files in Win32nof.zip must end up in the HDF41r2\ directory
         installed by HDF41r2s.zip

         If you must install dev.dsw and dev.dsp in 
         another directory, relative to HDF41r2\ , you will be asked to
	 locate the above 5 sub-project files, when you open the
	 project dev.dsw.
	 
	 If you want to rename dev (the entire project),
	 you will need to modify two files
	 dev.dsw and dev.dsp as text
	 (contrary to the explicit warnings in the files).

	 You can also modify dev.dsw and dev.dsp
	 as text, to allow these 2 files to be installed
	 in another directory.



  Settings... details:
  If you create your own project, the necessary settings can be
  read from the dev.dsp file(as text), or from the
  Project Settings in the Developer Studio project settings 
dialog.

    Project
	  Settings
	      C/C++
		  Category
		     PreProcessor
			 Code Generation
			    Use run-time Library
				   These are all set to use Single-Threaded.
				   or Single-Threaded debug.




============================================================================
================================macintosh.txt===============================

Fortner Software LLC ("Fortner") created the reference implementation for
Macintosh of the HDF 4.1r3 library, providing C-language bindings to all
4.1r3 features.

The Macintosh reference implementation of the HDF 4.1r3 library was implemented
and tested on a PowerMac Model 7600/120 running MacOS 8.5.1 using Metrowerks
CodeWarrior Pro1.  The library has also been run on a PowerMac G3.  

Fortner cannot be certain that the libraries will run on other versions of
Macintoshes (or clones) or MacOS versions, or when built using other
development tools.  (In particular, this Macintosh implementation has not
addressed use with non-PowerPC versions of Macintosh [i.e., 680x0-based
Macintoshes]).  Migrating the Macintosh reference implementation to other
development and/or run-time environments is the responsibility of the
library user.

First-time HDF users are encouraged to read the FAQ in this release for
more information about HDF.  Users can also look at the home page for HDF
at:

    https://www.hdfgroup.org/

Please send questions, comments, and recommendations regarding the
Macintosh version of the HDF library to:

    help@hdfgroup.org
 

============================================================================
================================mf_anno.txt=================================

 Annotation access through the Multi-file Annotation Interface(ANxxx)
 ==================================================================== 

 These routines are for accessing file labels, file descriptions, data
labels and data descriptions (i.e. all are annotations). General access
requires the routines Hopen() and ANstart() to be called first and the 
last call to be ANend() and Hclose() which ends annotation handling on the file
and closes the file. Basic annotation manipulation involves dealing
with handles(ann_id's) for each annotation and annotation interface 
handle(an_id). 


NOTES: 
  Note that the annotation types are enumerated. 
  TYPE here refers to file/data label/description types 
  They are AN_FILE_LABEL, AN_FILE_DESC, AN_DATA_LABEL, AN_DATA_DESC
  The tag/ref refers to data tag/ref.

  AN_DATA_LABEL = 0, /* Data label */
  AN_DATA_DESC  = 1, /* Data description */
  AN_FILE_LABEL = 2, /* File label */
  AN_FILE_DESC  = 3  /* File description */ 

In C-code you need to declare the annotation type using the 
enumerated type definition.

e.g. C-code fragment to write a File label

#include "hdf.h"
...
..
char fname[10] = {"ann.hdf"};
char *file_lab[1] = {"File label #1: This is a file label"};

int32   file_id; /* file id */
int32   an_id;   /* annotation interface id */
int32   ann_id;  /* annotation id */
ann_type myanntype;  /* annotation type */

/* Start Annotation interface and create file */
file_id = Hopen(fname, DFACC_CREATE,0);
an_id = ANstart(file_id);

/* Set annotation type to file label */
myanntype = AN_FILE_LABEL;

/* Create id for file label */
ann_id = ANcreatef(an_id, myanntype);

/* Write file label */
ANwriteann(ann_id, file_lab[0], HDstrlen(file_lab[0]));

/* end access to file label */
ANendaccess(ann_id);

/* end access to file and close it*/
ANend(an_id);
Hclose(file_id);
....
...

NOTE: You could also call ANcreatef() like this 
        ANcreatef(an_handle, AN_FILE_LABEL);
      without using the intermediate variable.

 ROUTINES NEEDED:
================
 Hopen    - Opening the file, returns a file handle
 Hclose   - Close the file.
 
 NEW ROUTINES:
===============
 ANstart     - open file for annotation handling, returns annotation
               interface id
 ANfileinfo  - get number of file/data annotations in file. Indices returned
               are used in ANselect() calls.
 ANend       - end access to annotation handling on file
 ANcreate    - create a new data annotation and return an id(ann_id)
 ANcreatef   - create a new file annotation and return an id(ann_id)
 ANselect    - returns an annotation id(ann_id) from index for 
               a particular annotation TYPE. This id is then used for
               calls like ANwriteann(), ANreadann(), ANannlen(),..etc
 ANnumann    - return number of annotations that match TYPE/tag/ref
 ANannlist   - return list of id's(ann_id's) that match TYPE/tag/ref
 ANannlen    - get length of annotation given id(ann_id)
 ANreadann   - read annotation given id(ann_id)
 ANwriteann  - write annotation given id(ann_id)
 ANendaccess - end access to annotation using id(ann_id)


Routines:
----------

C:
/* ------------------------------- ANstart -------------------------------- 
 NAME
	ANstart -- open file for annotation handling
 USAGE
	int32 ANstart(file_id)
        int32  file_id;    IN: file id

 RETURNS
        An annotation interface ID or FAIL
 DESCRIPTION
        Start annotation handling on the file and return an interface id.

 
Fortran: afstart(file_id)

/*------------------------------- ANfileinfo ----------------------------
 NAME
    ANfileinfo
 PURPOSE
    Report high-level information about the ANxxx interface for a given file.
 USAGE
    intn ANfileinfo(an_id, n_file_label, n_file_desc, n_data_label, n_data_desc)
        int32 an_id;          IN:  annotation interface ID
        int32 *n_file_label;  OUT: the # of file labels
        int32 *n_file_desc;   OUT: the # of file descriptions
        int32 *n_data_label;  OUT: the # of data labels
        int32 *n_data_desc;   OUT: the # of data descriptions
 RETURNS
    SUCCEED/FAIL
 DESCRIPTION
    Reports general information about the number of file and data
    annotations in the file. This routine is generally used to find
    the range of acceptable indices for ANselect calls.

Fortran: affileinfo(an_id, num_flabel, num_fdesc, num_dlabel, num_ddesc)

/* -------------------------------- ANend ---------------------------------
 NAME
	ANend -- close annotation handling on a file
 USAGE
	int32 ANend(an_id)
        int32 an_id;         IN: annotation interface ID for the file
 RETURNS
        SUCCEED / FAIL
 DESCRIPTION
      Closes annotation handling on the gvien annotation interface id.


Fortran: afend(an_id)

/* ------------------------------ ANcreate ---------------------------- 
 NAME
	ANcreate - create a new data annotation for the specified item
 USAGE
	int32 ANcreate(an_id, tag, ref, type )
        int32 an_id;    IN: annotation interface ID
        uint16 tag;     IN: tag of the item
        uint16 ref;     IN: reference number of the item
        ann_type  type: IN: AN_DATA_LABEL for data labels, 
                            AN_DATA_DESC for data descriptions,
 RETURNS
        An ID to an annotation which can either be a label or description
 DESCRIPTION
        Creates a data annotation, returns an 'ann_id' to work with the new 
        annotation which can either be a label or description.


Fortran: afcreate(an_id, tag, ref, type)

/* ------------------------------ ANcreatef ---------------------------- 
 NAME
	ANcreatef - create a new file annotation and return an id
 USAGE
	int32 ANcreatef(an_id, type )
        int32 an_id;    IN: annotation interface ID
        ann_type  type: IN:  AN_FILE_LABEL for file labels,
                             AN_FILE_DESC for file descriptions.
 RETURNS
        An ID to an annotation which can either be a file label or description
 DESCRIPTION
        Creates a file annotation, returns an 'ann_id' to work with the new 
        file annotation which can either be a label or description.

Fortran: afcreatef(an_id, type)


/* ------------------------------- ANselect ------------------------------- 
 NAME
	ANselect -- get an annotation ID from index of 'type'
 USAGE
	int32 ANselect(an_id, index, type)
        int32 an_id;    IN: annotation interface ID
        int32 index;    IN: index of annottion to get ID for
        ann_type  type: IN: AN_DATA_LABEL for data labels, 
                            AN_DATA_DESC for data descriptions,
                            AN_FILE_LABEL for file labels,
                            AN_FILE_DESC for file descriptions.
 RETURNS
        An ID to an annotation type which can either be a label or description 
 DESCRIPTION
        The position index is ZERO based

Fortran: afselect(an_id, index, type)


/*------------------------------- ANnumann ---------------------------------
 NAME
   ANnumann -- find number of annotation of 'type' that 
                 match the given element tag/ref 
 USAGE
       intn  ANnumann(an_id, type, elem_tag, elem_ref)
       int32  an_id;     IN: annotation interface ID
       int    type:      IN: AN_DATA_LABEL for data labels, 
                             AN_DATA_DESC for data descriptions,
                             AN_FILE_LABEL for file labels,
                             AN_FILE_DESC for file descriptions.
       uint16 elem_tag,: IN: tag of item of which this is annotation
       uint16 elem_ref;  IN: ref of item of which this is annotation
 RETURNS
       number of annotation found if successful and FAIL (-1) otherwise
 DESCRIPTION
       Find number of annotation of 'type' for the given element 
       tag/ref pair. Here an element is either a file label/desc or
       data label/desc.


Fortran: afnumann(an_id, type, tag, ref)


/*--------------------------------------------------------------------------
 NAME
   ANannlist -- generate list of annotation ids of 'type' that 
                 match the given element tag/ref 
 USAGE
       intn  ANannlist(an_id, type, elm_tag, elem_ref, ann_list[])
       int32  an_id;     IN: annotation interface ID
       ann_type  type:   IN: AN_DATA_LABEL for data labels, 
                             AN_DATA_DESC for data descriptions,
                             AN_FILE_LABEL for file labels,
                             AN_FILE_DESC for file descriptions.
       uint16 elem_tag,: IN: tag of element of which this is annotation
       uint16 elem_ref;  IN: ref of element of which this is annotation
       int32  ann_list[]; OUT: array of ann_id's that match criteria.
 RETURNS
       number of annotations ids found if successful and FAIL (-1) otherwise
 DESCRIPTION
       Find and generate list of annotation ids of 'type' for the given 
       element tag/ref pair


Fortran: afannlist(an_id,type, tag, ref, alist[])



/*--------------------------------------------------------------------------
 NAME
       ANannlen -- get length of annotation given annotation id
 USAGE
       int32 ANannlen(ann_id)
       int32 ann_id;   IN: annotation id
 RETURNS
       length of annotation if successful and FAIL (-1) otherwise
 DESCRIPTION
       Get the length of the annotation specified.


Fortran: afannlen(ann_id)

/*--------------------------------------------------------------------------
 NAME
       ANwriteann -- write annotation given ann_id
 USAGE
       intn ANwriteann(ann_id, ann, ann_len)
       char *ann_id;   IN: annotation id
       char *ann;      IN: annotation to write
       int32 ann_len;  IN: length of annotation

 RETURNS
       SUCCEED (0) if successful and FAIL (-1) otherwise
 DESCRIPTION
       Checks for pre-existence of given annotation, replacing old one if it
       exists. Writes out annotation.

Fortran: afwriteann(ann_id, ann, annlen)



/*--------------------------------------------------------------------------
 NAME
       ANreadann -- read annotation given ann_id
 USAGE
       intn ANreadann(ann_id, ann, maxlen)
       int32 ann_id;   IN: annotation id (handle)
       char *ann;     OUT: space to return annotation in
       int32 maxlen;   IN: size of space to return annotation in
 RETURNS
       SUCCEED (0) if successful and FAIL (-1) otherwise
 DESCRIPTION
       Gets tag and ref of annotation.  Finds DD for that annotation.
       Reads the annotation, taking care of NULL terminator, if necessary.
       

Fortran: afreadann(ann_id, ann, maxlen)



/* ----------------------------------------------------------------------- 
 NAME
	ANendaccess -- end access to an annotation given it's id
 USAGE
	intn ANendaccess(ann_id)
        int32 an_id;    IN: annotation id
 RETURNS
        SUCCEED or FAIL
 DESCRIPTION
        Terminates access to an annotation. 


Fortran: afendaccess(ann_id)


 



============================================================================
================================mf_ris.txt==================================

                  The multi-file RIS interface
                  =============================

Contents:

 Introduction
 How to access files and images in the new interface
 "Name = value" attributes in the new interface
 Dealing with annotations in the new interface
 Work not yet completed, bugs, limitations
 A listing or routines
 Descriptions of GR routines
   File level interface
   Dataset Manipulation
   ID/Ref/Index Functions
   Interlace Request Functions
   LUT/Palette I/O Functions
   Special Element Functions
   Attribute Functions


Introduction
============
The new Generic Raster (GR) interface provides a set of functions for
manipulating raster images of all kinds.  This new interface is meant to
replace the older RIS8 and RIS24 interfaces, although these older
interfaces will continue to be supported.

Generic raster images are composed of "pixels" which can have multiple
components, including but not limited to 8-bit unsigned integers.  Each image
can have multiple palettes associated with it and other 'attributes' in the
same "name=value" style as the SD*() routines have.


The new GR interface was motivated by a number of needs:

o  The need for multi-file, multi-object access to raster images, allowing
users to keep open more than one file at a time, and to "attach" more than
one raster image at a time.

o  A need to further integrate the netCDF data-model with the HDF data-models.

o  A need for a more general framework for attributes within the RIS
data-model (allowing 'name = value' style metadata).

o  A need to be able to access subsamples and subsets of images.


IMPORTANT:  The added functionality represented by this new interface has
necessitated a change in how raster images are physically represented on
disk.   As a result programs using the old single-file RIS interfaces will
only be able to read the data out of files produced by the new interface.
The metadata / attributes will not be accessible.

The following chart represents what can be done with the various interfaces
available in HDF 4.0b1:

              old RIS-API    new GR-API

old RIS         CRW             CRW
HDF files

new RIS          r              CRW
HDF files

'R' means read, 'W' means write and 'C' means create.  Entries with dashes
'-' represent functionality which has not yet been implemented.  'r' stands
for the ability to only read the data, not the metadata.


Work not yet completed, bugs, limitations
===========================================

Misc. stuff left to do:
    Deal with special elements for images.
    GRrename for images.
    GRsetflags to suppress writing fill data and to suppress fillvalue attr.

Features not supported:
    Full support for multiple palettes with each RI.
    Support for named palettes with each RI.
    Support for palettes with non-standard formats.
    Deletion of attributes or images (would require changing the way index
        numbers are handled)

Other limitations:
   Currently the following design limitations are still in place:
   1 - Cannot have pixels or palette entries which contain mixed variable
        types, i.e. all the pixel/palette components must be of the same
        number type.
   2 - Currently all the components must be of valid HDF number types,
        fractional bytes (i.e. 6-bit components) or 'plain' multiple byte values
        are not handled, although they can be packed into the next larger
        sized number type in order to hold them.



How to access files and images in the new interface
======================================================


Here are the steps involved in accessing images in the new interface:

1. Open or create the file using Hopen.  This provides you with a file ID to
be used in step 2.

2. Activate the GR interface for the file with the file ID obtained from
step 1, using GRstart.  This provides you with a GR interface ID (GR ID).

3. Optionally obtain information about the raster images in the file and
global GR attributes using GRfileinfo.  Use the GR ID from step 2 to refer
to the image file.

4. Optionally find the index of a raster image, by name using
GRnametoindex, or by reference number using GRreftoindex.

5. Select for access an image with a given index, using GRselect for each
image.  Each call to GRselect returns a raster image ID (RI ID) for
subsequent accesses involving the corresponding image.

5. Access the image by its RI ID, using routines such as GRgetiminfo (to
get information about the image) and GRreadimage (to read all or part of an
image).

6. Terminate access to a given image using GRendaccess.

7. Terminate access to the GR interface for the file, using GRend.

8. Close the file using Hclose.

Notice that in the GR interface, images are identified in several ways.
Before an image is accessible ("attached"), it is identified by index,
name, and reference number. The index describes the relative position of
the image in the file. The name is a character string associated with the
image, and the reference number is a unique integer.  An image's name is
assigned by the program when it is created, and the reference number is
assigned by the HDF library when it is created. After an image is attached
, it is identified by an raster image identifier, or RI ID.

The following code fragment illustrates the steps involved in accessing the
image in a file and printing information about them.


    /* Open the file and initialize the GR interface */
    hdf_file_id=Hopen(TESTFILE,DFACC_RDWR,0);
    grid=GRstart(hdf_file_id);

    /* Obtain information about the images in the file */
    GRfileinfo(grid,&n_datasets,&n_attrs);

    /* Attach to each image and print information about it */
    for(i=0; i<n_datasets; i++)
      {
          riid=GRselect(grid,i);
          GRgetiminfo(riid,NULL,&ncomp,&nt,&il,dimsizes,&n_attrs);

          printf("%d: riid=%ld: ncomp=%ld, nt=%ld, il=%ld,
                    dim[0]=%ld, dim[1]=%ld, n_attrs=%ld\n",
                    i, riid, ncomp, nt, il, dimsizes[0],
                    dimsizes[1], n_attrs);

          /* Detach from the image */
          GRendaccess(riid);
      } /* end for */

    /* Shut down the GR interface and close the file */
    GRend(grid);
    Hclose(hdf_file_id);



"Name = value" attributes in the new interface
===============================================

Attributes of the form "name = value" were introduced in HDF 3.3, but at
that time they were available only for SDSs and files.  In HDF 4.0 we have
added the ability to attach local and global attributes to raster images
and raster image dimensions.

An attribute's "name" is a string, and "value" is the associated value or
values.  If an attribute contains more than one value, all values must be
of the same type.  For example the attribute 'valid_range' attribute might
be assigned values the maximum and minimum valid values for a given image.

Raster attributes can be "local" or "global."  A local raster image
attribute is one that applies to one raster image only. Global raster image
attributes apply to all of the images in a file.

Attributes for raster images are created by the routine GRsetattr.
Existing attributes are selected by giving an object pointer and an
attribute index.  The functions GRattrinfo, GRfindattr, and GRgetattr may
be used in combination to read attributes and their values.  GRattrinfo
gets the name , number type, and number of values for an attribute with a
given index.  GRfindattr gets the index of an attribute with a given name,
and GRreadattr reads the values associate with an attribute with a given
index.

The following example illustrates how to attach GR image attributes, and
also GR global (file) attributes.

    /* Open file and initialize the GR interface */
    hdf_file_id=Hopen(TESTFILE,DFACC_RDWR,0);
    grid=GRstart(hdf_file_id);

    /* Create a global attribute  -- applies to all rasters in the file */
    HDstrcpy(attr_name,"Test1");
    HDstrcpy(u8_attr,"Attribute value 1");
    GRsetattr(grid,attr_name,DFNT_UINT8,HDstrlen(u8_attr)+1,u8_attr);

    GRfileinfo(grid,&n_datasets,&n_attrs);

    /* select every image in the file, and
       assign a local attribute to each */
    for(i=0; i<n_datasets; i++)
      {
          /* Attach to image with index==i */
          riid=GRselect(grid,i);

          /* Create an attribute for the image */
          HDstrcpy(attr_name,"Image1");
          HDstrcpy(u8_attr,"Attribute value 1");
          GRsetattr(riid,attr_name,DFNT_UINT8,HDstrlen(u8_attr)+1,u8_attr);

          GRgetiminfo(riid,NULL,&ncomp,&nt,&il,dimsizes,&n_attrs);

          printf("%d: riid=%ld: ncomp=%ld, nt=%ld, il=%ld, dim[0]=%ld,
                  dim[1]=%ld, n_attrs=%ld\n",i,riid,ncomp,nt,il,
                  dimsizes[0], dimsizes[1],n_attrs);

                for(j=0; j<n_attrs; j++)
                  {
                    GRattrinfo(riid,j,attr_name,&nt,&ncomp);

                    GRgetattr(riid,j,u8_attr);
                    printf("Image #%d Attribute #%d: Name=%s, Value=%s\n",
                             i,j,attr_name,u8_attr);
                  } /* end for */

          /* Detach from the image */
          GRendaccess(riid);
      } /* end for */

    /* Shut down the GR interface */
    GRend(grid);

    /* Close the file */
    Hclose(hdf_file_id);



Dealing with annotations in the new interface
================================================

The new GR interface allows you to reference rasters explicitly, by "GR
id".  A GR id is different from its reference number.  Since annotation
routines attach annotations to objects by reference number, there needs to
be a mechanism for determining the reference number of a raster image,
given its id. This is made possible by the addition of the routine
GRidtoref.

A similar problem occurs when going the other way.  For example, a call to
DFANlabellist returns the reference numbers of objects that are
annotated.  If those objects are RISs (i.e. they have the tag DFTAG_RIG),
we need to map the reference numbers to the corresponding images.  For
this, a two-step process is required.  You can use the function
GRreftoindex to get the index, or position, of the dataset that has a
certain reference number, then you use the routine GRselect to get the id
for the image in that position.


A listing or routines
======================

File/Interface Functions:
int32 GRstart(int32 hdf_file_id)
    - Initializes the GR interface for a particular file. Returns a 'grid' to
        specify the GR group to operate on.
intn GRfileinfo(int32 grid, int32 *n_datasets, int32 *n_attrs)
    - Returns information about the datasets and "global" attributes for the
        GR interface.
intn GRend(int32 grid)
    - Terminates multi-file GR access for a file.

Image I/O Functions:
int32 GRcreate(int32 grid,char *name,int32 ncomp,int32 nt,int32 il,int32
dimsizes[2])
    - Defines a raster image in a file.  Returns a 'riid' to work with the new
        raster image.
int32 GRselect(int32 grid,int32 index)
    - Selects an existing RI to operate on.
int32 GRnametoindex(int32 grid,char *name)
    - Maps a RI name to an index which is returned.
intn GRgetiminfo(int32 riid,char *name,int32 *ncomp,int32 *nt,int32
*il,int32 dimsizes[2],int32 *n_attr)
    - Gets information about an RI which has been selected/created.
intn GRwriteimage(int32 riid,int32 start[2],int32 stride[2],int32
count[2],VOIDP data)
    - Writes image data to an RI.  Partial dataset writing and subsampling is
        allowed, but only with the dimensions of the dataset (ie. no UNLIMITED
        dimension support)
intn GRreadimage(int32 riid,int32 start[2],int32 stride[2],int32
count[2],VOIDP data)
    - Read image data from an RI.  Partial reads and subsampling are allowed.
intn GRendaccess(int32 riid)
    - End access to an RI.

Dimension Functions:
int32 GRgetdimid(int32 riid,int32 index)
    - Get a dimension id ('dimid') for an RI to assign attributes to. [Later]
intn GRsetdimname(int32 dimid,char *name)
    - Set the name of a dimension. [Later]
int32 GRdiminfo(int32 dimid,char *name,int32 *size,int32 *n_attr)
    - Get information about the dimensions attributes and size. [Later]

ID/Ref/Index Functions:
uint16 GRidtoref(int32 riid)
    - Maps an riid to a reference # for annotating or including in a Vgroup.
int32 GRreftoindex(int32 hdf_file_id,uint16 ref)
    - Maps the reference # of an RI into an index which can be used with
        GRselect.

Interlace Request Functions:
intn GRreqlutil(int32 riid,intn il)
    - Request that the next LUT read from an RI have a particular interlace.
intn GRreqimageil(int32 riid,intn il)
    - Request that the image read from an RI have a particular interlace.

LUT/Palette I/O Functions:
int32 GRgetlutid(int32 riid,int32 index)
    - Get a palette id ('palid') for an RI.
intn GRgetlutinfo(int32 riid,int32 *ncomp,int32 *nt,int32 *il,int32 *nentries)
    - Gets information about a palette.
intn GRwritelut(int32 riid,int32 ncomps,int32 nt,int32 il,int32
nentries,VOIDP data)
    - Writes out a palette for an RI.
intn GRreadlut(int32 palid,VOIDP data)
    - Reads a palette from an RI.

Special Element Functions:
int32 GRsetexternalfile(int32 riid,char *filename,int32 offset)
    - Makes the image data of an RI into an external element special element.
intn GRsetaccesstype(int32 riid,uintn accesstype)
    - Sets the access for an RI to be either serial or parallel I/O.
intn GRsetcompress(int32 riid,int32 comp_type,comp_info *cinfo)
    - Makes the image data of an RI into a compressed special element.

Attribute Functions:
intn GRsetattr(int32 dimid|riid|grid,char *name,int32 attr_nt,int32
count,VOIDP data)
    - Write an attribute for an object.
int32 GRattrinfo(int32 dimid|riid|grid,int32 index,char *name,int32
*attr_nt,int32 *count)
    - Get attribute information for an object.
intn GRgetattr(int32 dimid|riid|grid,int32 index,VOIDP data)
    - Read an attribute for an object.
int32 GRfindattr(int32 dimid|riid|grid,char *name)
    - Get the index of an attribute with a given name for an object.



Routine Descriptions
====================

Most of the routines in the GR interface return a status value of type intn
(native integers).  If the status is equal to SUCCEED the routine completed
successfully.  If it is equal to FAIL an error occurred, information about
the error may be available by calling HEprint(filestream, 0).  SUCCEED and
FAIL are defined in hdf.h for C users and in constant.i for Fortran
programs.

All IDs (hdf_file_id, grid, riid) are int32 quantities.

Prototypes for these functions can be found in the file hproto.h

Routines that can be called from C are all of the form GRxxx

More details about all the routines below can be found in the HDF reference
manual.


File level interface:
=====================

These routines initialize and de-initialize the GR interface, and provide
information about the raster images in a file.

GRstart
-------
    Initialize the GR*() interface for a given HDF file.
 USAGE
    int32 GRstart(hdf_file_id)
        int32 hdf_file_id;          IN: file ID from Hopen
 RETURNS
    Return grid (GR ID) on success, or FAIL
 DESCRIPTION
    Initializes the GR*() interface to operate on the HDF file which was
    specified by hdf_file_id.  This routine must be called before any further
    GR*() routines are called for a file.

GRfileinfo
----------
    Report high-level information about the GR*() interface for a given file.
 USAGE
    intn GRfileinfo(grid, n_datasets, n_attrs)
        int32 grid;                 IN: GR ID to get information about
        int32 *n_datasets;          OUT: the # of GR datasets in a file
        int32 *n_attrs;             OUT: the # of "global" GR attributes
 RETURNS
    SUCCEED/FAIL
 DESCRIPTION
    Reports general information about the number of datasets and "global"
    attributes for the GR interface.  This routine is generally used to find
    the range of acceptable indices for GRselect calls.


GRend
-----
    Terminate the GR*() interface for a given HDF file.
 USAGE
    intn GRend(grid)
        int32 grid;          IN: GR ID from GRstart
 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Terminates access to the GR*() interface for a file.


DataSet Manipulation
=====================

GRcreate
--------
    Create a new raster image.

 USAGE
    int32 GRcreate(grid, name, ncomp, nt, il, dimsizes)
        int32 grid;         IN: GR ID from GRstart
        char *name;         IN: Name of raster image to create
        int32 ncomp;        IN: Number of components in image
        int32 nt;           IN: Number type of each component
        int32 il;           IN: Interlace of the components in the image
        int32 dimsizes[2];  IN: Dimensions of the new image

 RETURNS
    A valid riid (Raster-Image ID) on success, or FAIL.

 DESCRIPTION
    Creates a new raster image in a file.

ASSUMPTIONS
    All components must be the same number-type.


GRselect
--------
    Select a raster image to operate on.
 USAGE
    int32 GRselect(grid,index)
        int32 grid;          IN: GR ID from GRstart
        int32 index;         IN: Which raster image to select (indexed from 0)
 RETURNS
    A valid riid (Raster-Image ID) on success, or FAIL.

 DESCRIPTION
    Selects a raster image from the file to work on.  This ID is needed for
    all operations on the image dataset, including reading/writing data,
    annotations, etc.



GRnametoindex
-------------
    Map a raster image name to an index.
 USAGE
    int32 GRnametoindex(grid,name)
        int32 grid;          IN: GR ID from GRstart
        char *name;          IN: Name of raster image to search for
 RETURNS
    A valid index on success, or FAIL.

 DESCRIPTION
    Searches for a raster image based on the name provided.  This routine
    maps from names of raster images to indices inside the GR group.

GRgetiminfo
-----------
    Gets information about a raster image.

 USAGE
    intn GRgetiminfo(riid,name,ncomp,nt,il,dimsizes,n_attr)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        char *name;         OUT: name of raster image
        int32 *ncomp;       OUT: number of components in image
        int32 *nt;          OUT: number type of components
        int32 *il;          OUT: interlace of the image
        int32 *dimsizes;    OUT: size of each dimension
        int32 *n_attr;      OUT: the number of attributes for the image

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Looks up information about an image which has been selected or created
    with the GR routines.  Each of the parameters can be NULL, in which case
    that piece of information will not be retrieved.

GRwriteimage
------------
    Writes raster data to an image

 USAGE
    intn GRwriteimage(riid,start,stride,edge,data)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        int32 start[2];     IN: array containing the offset in the image of the
                                image data to write out
        int32 stride[2];    IN: array containing interval of data being written
                                along each edge.  strides of 0 are illegal
                                (and generate an error)
                                ie. stride of 1 in each dimension means
                                writing contiguous data, stride of 2 means
                                writing every other element out along an edge.
        int32 count[2];     IN: number of elements to write out along each edge.
        VOIDP data;         IN: pointer to the data to write out.

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Writes image data to an RI.  Partial dataset writing and subsampling is
        allowed, but only within the dimensions of the dataset (ie. no UNLIMITED
        dimension support)

 ASSUMPTIONS
    If the stride parameter is set to NULL, a stride of 1 will be assumed.


GRreadimage
-----------
    Read raster data for an image

 USAGE
    intn GRreadimage(riid,start,stride,edge,data)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        int32 start[2];     IN: array containing the offset in the image of the
                                image data to read in
        int32 stride[2];    IN: array containing interval of data being read
                                along each edge.  strides of 0 are illegal
                                (and generate an error)
                                ie. stride of 1 in each dimension means
                                reading contiguous data, stride of 2 means
                                reading every other element out along an edge.
        int32 count[2];     IN: number of elements to read in along each edge.
        VOIDP data;         IN: pointer to the data to read in.

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Read image data from an RI.  Partial dataset reading and subsampling is
        allowed.

ASSUMPTIONS
    If the stride parameter is set to NULL, a stride of 1 will be assumed.



GRendaccess
-----------
    End access to an RI.

 USAGE
    intn GRendaccess(riid)
        int32 riid;         IN: RI ID from GRselect/GRcreate

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    End access to an RI.  Further attempts to access the RI ID will result in
    an error.


Dimension Functions
===================
(these have not been completed)


ID/Ref/Index Functions
======================

GRidtoref
---------
    Maps an RI ID to a reference # for annotating or including in a Vgroup.

 USAGE
    uint16 GRidtoref(riid)
        int32 riid;         IN: RI ID from GRselect/GRcreate

 RETURNS
    A valid reference # on success or FAIL

 DESCRIPTION
    Maps an riid to a reference # for annotating or including in a Vgroup.


GRreftoindex
------------
    Maps the reference # of an RI into an index which can be used with GRselect.

 USAGE
    int32 GRreftoindex(grid,ref)
        int32 grid;         IN: GR ID from GRstart
        uint16 ref;         IN: reference number of raster image to map to index

 RETURNS
    A valid index # on success or FAIL

 DESCRIPTION
    Maps the reference # of an RI into an index which can be used with GRselect.


Interlace Request Functions
===========================


GRreqlutil
----------
    Request that the next LUT read from an RI have a particular interlace.

 USAGE
    intn GRreqlutil(riid,il)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        intn il;            IN: interlace for next LUT.  From the following
                                values (found in mfgr.h):
                      MFGR_INTERLACE_PIXEL      - pixel interlacing
                      MFGR_INTERLACE_LINE       - line interlacing
                      MFGR_INTERLACE_COMPONENT  - component/plane interlacing

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Request that the next LUT read from an RI have a particular interlace.



GRreqimageil
------------
    Request that the image read from an RI have a particular interlace.

 USAGE
    intn GRreqimageil(riid,il)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        intn il;            IN: interlace for next RI.  From the following
                                values (found in mfgr.h):
                      MFGR_INTERLACE_PIXEL      - pixel interlacing
                      MFGR_INTERLACE_LINE       - line interlacing
                      MFGR_INTERLACE_COMPONENT  - component/plane interlacing

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Request that the image read from an RI have a particular interlace.




LUT/Palette I/O Functions
=========================

GRgetlutid
----------
    Get a LUT id ('lutid') for an RI.

 USAGE
    int32 GRgetlutid(riid,index)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        int32 lut_index;    IN: Which LUT image to select (indexed from 0)

 RETURNS
    Valid LUT ID on success, FAIL on failure

 DESCRIPTION
    Get a LUT id ('lutid') for accessing LUTs in an RI.

 GLOBAL VARIABLES
 COMMENTS, BUGS, ASSUMPTIONS
    Currently only supports one LUT per image, at index 0 and LUTID==RIID.




intn GRgetlutinfo(int32 riid,int32 *ncomp,int32 *nt,int32 *il,int32 *nentries)
    - Gets information about a palette.



GRwritelut
----------
    Writes out a LUT for an RI.

 USAGE
    intn GRwritelut(riid,name,ncomps,nt,il,nentries,data)
        int32 lutid;        IN: LUT ID from GRgetlutid
        char *name;         IN: name of LUT image
        int32 ncomp;        IN: number of components in LUT
        int32 nt;           IN: number type of components
        int32 il;           IN: interlace of the LUT
        int32 nentries;     IN: the number of entries for the LUT
        VOIDP data;         IN: LUT data to write out

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Writes out a LUT for an RI.


GRreadlut
---------
    Reads a LUT from an RI.

 USAGE
    intn GRreadlut(lutid,data)
        int32 lutid;        IN: LUT ID from GRgetlutid
        VOIDP data;         IN: buffer for LUT data read in

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Reads a LUT from an RI.




Special Element Functions
=========================


GRsetexternalfile
-----------------
    Makes the image data of an RI into an external element special element.

 USAGE
    intn GRsetexternalfile(riid,filename,offset)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        char *filename;     IN: name of the external file
        int32 offset;       IN: offset in the external file to store the image

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Makes the image data of an RI into an external element special element.
        Cause the actual data for a dataset to be stored in an
        external file.  This can only be done once for any given
        dataset and it is the user's responsibility to make sure the
        external datafile is transported when the "header" file is
        moved.  The offset is the number of byte from the beginning of
        the file where the data should be stored.  This routine can
        only be called on HDF 3.3 files (i.e. calling on an XDR-based
        netCDF file that was opened with the multi-file interface will
        fail).





GRsetaccesstype
---------------
    Sets the access for an RI to be either serial or parallel I/O.

 USAGE
    intn GRsetaccesstype(riid,accesstype)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        uintn accesstype;   IN: access type for image data, from the following
                                values:
                                    DFACC_SERIAL - for serial access
                                    DFACC_PARALLEL - for parallel access

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Sets the access for an RI to be either serial or parallel I/O.




GRsetcompress
-------------
    Compressed the image data of an RI.

 USAGE
    intn GRsetcompress(riid,comp_type,cinfo)
        int32 riid;         IN: RI ID from GRselect/GRcreate
        int32 comp_type;    IN: type of compression, from list in hcomp.h
        comp_info *cinfo;   IN: compression specific information

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Compressed the image data of an RI.
    (Makes the image data of an RI into a compressed special element)




Attribute Functions
===================

GRsetattr
---------
    Write an attribute for an object.

 USAGE
    intn GRsetattr(dimid|riid|grid,name,attr_nt,count,data)
        int32 dimid|riid|grid;  IN: DIM|RI|GR ID
        char *name;             IN: name of attribute
        int32 attr_nt;          IN: number-type of attribute
        int32 count;            IN: number of entries of the attribute
        VOIDP data;             IN: attribute data to write

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Write an attribute for an object (function will figure out ID type).

 GLOBAL VARIABLES
 COMMENTS, BUGS, ASSUMPTIONS
    Currently does not allow changing NT of an existing attribute.




GRattrinfo
----------
    Get attribute information for an object.

 USAGE
    intn GRattrinfo(dimid|riid|grid,index,name,attr_nt,count)
        int32 dimid|riid|grid;  IN: DIM|RI|GR ID
        int32 index;            IN: index of the attribute for info
        char *name;             OUT: name of attribute
        int32 attr_nt;          OUT: number-type of attribute
        int32 count;            OUT: number of entries of the attribute

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Get attribute information for an object.




GRgetattr
---------
    Read an attribute for an object.

 USAGE
    intn GRgetattr(dimid|riid|grid,index,data)
        int32 dimid|riid|grid;  IN: DIM|RI|GR ID
        int32 index;            IN: index of the attribute for info
        VOIDP data;             OUT: data read for attribute

 RETURNS
    SUCCEED/FAIL

 DESCRIPTION
    Read an attribute for an object.




GRfindattr
----------
    Get the index of an attribute with a given name for an object.

 USAGE
    int32 GRfindattr(int32 dimid|riid|grid,char *name)
        int32 dimid|riid|grid;  IN: DIM|RI|GR ID
        char *name;             IN: name of attribute to search for

 RETURNS
    Valid index for an attribute on success, FAIL on failure

 DESCRIPTION
    Get the index of an attribute with a given name for an object.








============================================================================
================================new_functions.txt===========================

This file contains a list of the new functions added with HDF 4.1r2.
The functions in parenthesis were already present in the HDF library,
and are included for clarity.

C                     FORTRAN                 Description
--------------------------------------------------------------------------------

(SDsetcompress)       sfscompress             compresses SDS

(SDwritechunk)        sfwchnk                 writes the specified chunk of
                                              NUMERIC data to the SDS

(SDwritechunk)        sfwcchnk                writes the specified chunk of
                                              CHARACTER data to the SDS

(SDreadchunk)         sfrchnk                 reads the specified chunk of
                                              NUMERIC data to the SDS

(SDreadchunk)         sfrcchnk                reads the specified chunk of
                                              CHARACTER data to the SDS
 
(SDsetchunk)          sfschnk                 makes the SDS a chunked SDS
 
(SDsetchunkcache)     sfscchnk                sets the maximum number of chunks
                                              to cache
 
(SDgetchunkinfo)      sfgichnk                gets info on SDS

(SDsetblocksize)      sfsblsz                 sets block size 

(SDisrecord)          sfisrcrd                checks if an SDS is unlimited



(GRsetcompress)       mgscompress             compresses raster image

GRsetchunk            mgschnk                 makes a raster image a chunked
                                              raster image

GRgetchunkinfo        mggichnk                gets info on a raster image

GRsetchunkcache       mgscchnk                sets the maximum number of chunks
                                              to cache


(Hgetlibversion)      hglibver                gets version of the HDF Library

(Hgetfileversion)     hgfilver                gets version of the HDF file


Vdeletetagref        vfdtr                    deletes tag/ref pair ( HDF object)
                                              from a vgroup

(VSfindclass)        vsffcls                  finds class with a specified 
                                              name in a vdata

VSdelete             vsfdlte                  deletes a vdata

Vdelete              vdelete                  deletes a vgroup


============================================================================
================================sd_chunk_examples.txt=======================

/**************************************************************************
File: sd_chunk_examples.c

  Examples for writing/reading SDS with Chunking and Chunking w/ Compression.
   - Sample C-code using SDS chunking routines. 
   - No real error checking is done and the value of 'status' should 
     be checked for proper values.

5 Examples are shown, 1 for 2-D array, 3 for 3-D arrays and
                      1 for 2-D array with compression..

  Example 1. 2-D 9x4 SDS of uint16 with 3x2 chunks
             Write data using SDwritechunk().
             Read data using SDreaddata().

  Example 2. 3-D 2x3x4 SDS of uint16 with 2x3x2 chunks
             Write data using SDwritedata().
             Read data using SDreaddata().

  Example 3. 3-D 2x3x4 SDS of uint16 with 1x1x4 chunks
             Write data using SDwritechunk().
             Read data using SDreaddata().

  Example 4. 3-D 2x3x4 SDS of uint16 with 1x1x4 chunks
             Write data using SDwritedata().
             Read data using SDreadchunk().

  Example 5. 2-D 9x4 SDS of uint16 with 3x2 chunks with GZIP compression.
             Write data using SDwritechunk().
             Read data using SDreaddata().

Author - GeorgeV
Date   - 11/25/96
********************************************************************/

#include "mfhdf.h"

/* arrays holding dim info for datasets */
static int32  d_dims[3]     = {2, 3, 4};  /* data dimensions */
static int32  edge_dims[3]  = {0, 0, 0};  /* edge dims */
static int32  start_dims[3] = {0, 0, 0};  /* starting dims  */

/* data arrays laid out in memory  */

/* used in Example 1 and 5 */
static uint16  u16_2data[9][4] =
{ 
   {11, 21, 31, 41},
   {12, 22, 32, 42},
   {13, 23, 33, 43},
   {14, 24, 34, 44},
   {15, 25, 35, 45},
   {16, 26, 36, 46},
   {17, 27, 37, 47},
   {18, 28, 38, 48},
   {19, 29, 39, 49},
};

/* uint16 3x2 chunk arrays used in example 1 and 5*/
static uint16  chunk1_2u16[6] = {11, 21, 
                                 12, 22, 
                                 13, 23};

static uint16  chunk2_2u16[6] = {31, 41, 
                                 32, 42, 
                                 33, 43};

static uint16  chunk3_2u16[6] = {14, 24, 
                                 15, 25, 
                                 16, 26};

static uint16  chunk4_2u16[6] = {34, 44, 
                                 35, 45, 
                                 36, 46};

static uint16  chunk5_2u16[6] = {17, 27, 
                                 18, 28, 
                                 19, 29};

static uint16  chunk6_2u16[6] = {37, 47, 
                                 38, 48, 
                                 39, 49};


/* uint16 1x1x4 chunk arrays used in example 3 */
static uint16  chunk1_3u16[4] =  { 0, 1, 2, 3};

static uint16  chunk2_3u16[4] =  { 10, 11, 12, 13};

static uint16  chunk3_3u16[4] =  { 20, 21, 22, 23};

static uint16  chunk4_3u16[4] =  { 100, 101, 102, 103};

static uint16  chunk5_3u16[4] =  { 110, 111, 112, 113};

static uint16  chunk6_3u16[4] =  { 120, 121, 122, 123};


/* Used in Examples 2 and 4 */
static uint16  u16_3data[2][3][4] =
{
    {
        { 0, 1, 2, 3},
        { 10, 11, 12, 13},
        { 20, 21, 22, 23}},
    {
        { 100, 101, 102, 103},
        { 110, 111, 112, 113},
        { 120, 121, 122, 123}}};

/*
 * Main routine
 */
int main(int argc, char *argv[])
{
   int32 f1;                    /* file handle */
   int32 sdsid;                 /* SDS handle */
   uint16  inbuf_3u16[2][3][4]; /* Data array read for Example 2 and 3*/
   uint16  inbuf_2u16[5][2];    /* Data array read for Example 1 */
   uint16  ru16_3data[4];       /* whole chunk input buffer */
   uint16  fill_u16 = 0;        /* fill value */
   HDF_CHUNK_DEF chunk_def;     /* Chunk definition set */ 
   HDF_CHUNK_DEF rchunk_def;    /* Chunk definition read */ 
   int32   cflags;              /* chunk flags */
   comp_info cinfo;             /* compression info */
   intn status;

   ncopts = NC_VERBOSE;

    /* create file */
    f1 = SDstart("chunk.hdf", DFACC_CREATE);

    /* 
      Example 1. 2-D 9x4 SDS of uint16 with 3x2 chunks
                 Write data using SDwritechunk().
                 Read data using SDreaddata().
    */

    /* create a  9x4 SDS of uint16 in file 1 */
    d_dims[0] = 9;
    d_dims[1] = 4;
    sdsid = SDcreate(f1, "DataSetChunked_1", DFNT_UINT16, 2, d_dims);

    /* set fill value */
    fill_u16 = 0;
    status = SDsetfillvalue(sdsid, (VOIDP) &fill_u16);

    /* Create chunked SDS 
       chunk is 3x2 which will create 6 chunks */
    chunk_def.chunk_lengths[0] = 3;
    chunk_def.chunk_lengths[1] = 2;
    status = SDsetchunk(sdsid, chunk_def, HDF_CHUNK);

    /* Set Chunk cache to hold 3 chunks */
    status = SDsetchunkcache(sdsid, 3, 0);

    /* Write data use SDwritechunk */

    /* Write chunk 1 */
    start_dims[0] = 0;
    start_dims[1] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk1_2u16);

    /* Write chunk 4 */
    start_dims[0] = 1;
    start_dims[1] = 1;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk4_2u16);

    /* Write chunk 2 */
    start_dims[0] = 0;
    start_dims[1] = 1;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk2_2u16);

    /* Write chunk 5 */
    start_dims[0] = 2;
    start_dims[1] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk5_2u16);

    /* Write chunk 3 */
    start_dims[0] = 1;
    start_dims[1] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk3_2u16);

    /* Write chunk 6 */
    start_dims[0] = 2;
    start_dims[1] = 1;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk6_2u16);
 
    /* read a portion of data back in using SDreaddata
       i.e  5x2 subset of the whole array */
    start_dims[0] = 2;
    start_dims[1] = 1;
    edge_dims[0] = 5;
    edge_dims[1] = 2;
    status = SDreaddata(sdsid, start_dims, NULL, edge_dims, (VOIDP) inbuf_2u16);

   /* This 5x2 array should look somethink like this
         {{23, 24, 25, 26, 27},
          {33, 34, 35, 36, 37}}    
    */

    /* Get chunk information */
    status = SDgetchunkinfo(sdsid, &rchunk_def, &cflags);


    /* Close down this SDS*/    
    status = SDendaccess(sdsid);

    /* 
      Example 2. 3-D 2x3x4 SDS of uint16 with 2x3x2 chunks
                 Write data using SDwritedata().
                 Read data using SDreaddata().
    */

    /* create a new 2x3x4 SDS of uint16 in file 1 */
    d_dims[0] = 2;
    d_dims[1] = 3;
    d_dims[2] = 4;
    sdsid = SDcreate(f1, "DataSetChunked_2", DFNT_UINT16, 3, d_dims);

    /* set fill value */
    fill_u16 = 0;
    status = SDsetfillvalue(sdsid, (VOIDP) &fill_u16);

    /* Create chunked SDS
       chunk is 2x3x2 which will create 2 chunks */
    chunk_def.chunk_lengths[0] = 2;
    chunk_def.chunk_lengths[1] = 2;
    chunk_def.chunk_lengths[2] = 3;
    status = SDsetchunk(sdsid, chunk_def, HDF_CHUNK);

    /* Set Chunk cache to hold 2 chunks*/
    status = SDsetchunkcache(sdsid, 2, 0);

    /* Write data using SDwritedata*/
    start_dims[0] = 0;
    start_dims[1] = 0;
    start_dims[2] = 0;
    edge_dims[0] = 2;
    edge_dims[1] = 3;
    edge_dims[2] = 4;
    status = SDwritedata(sdsid, start_dims, NULL, edge_dims, (VOIDP) u16_3data);

    /* read data back in using SDreaddata*/
    start_dims[0] = 0;
    start_dims[1] = 0;
    start_dims[2] = 0;
    edge_dims[0] = 2;
    edge_dims[1] = 3;
    edge_dims[2] = 4;
    status = SDreaddata(sdsid, start_dims, NULL, edge_dims, (VOIDP) inbuf_3u16);

    /* Verify the data in inbuf_3u16 against u16_3data[] */

    /* Get chunk information */
    status = SDgetchunkinfo(sdsid, &rchunk_def, &cflags);

    /* Close down this SDS*/    
    status = SDendaccess(sdsid);

    /* 
      Example 3. 3-D 2x3x4 SDS of uint16 with 1x1x4 chunks
                 Write data using SDwritechunk().
                 Read data using SDreaddata().
    */

    /* Now create a new 2x3x4 SDS of uint16 in file 'chunk.hdf' */
    d_dims[0] = 2;
    d_dims[1] = 3;
    d_dims[2] = 4;
    sdsid = SDcreate(f1, "DataSetChunked_3", DFNT_UINT16, 3, d_dims);

    /* set fill value */
    fill_u16 = 0;
    status = SDsetfillvalue(sdsid, (VOIDP) &fill_u16);

    /* Create chunked SDS 
       chunk is 1x1x4 which will create 6 chunks */
    chunk_def.chunk_lengths[0] = 1;
    chunk_def.chunk_lengths[1] = 1;
    chunk_def.chunk_lengths[2] = 4;
    status = SDsetchunk(sdsid, chunk_def, HDF_CHUNK);

    /* Set Chunk cache to hold 4 chunks*/
    status = SDsetchunkcache(sdsid, 4, 0);

    /* Write data use SDwritechunk */

    /* Write chunk 1 */
    start_dims[0] = 0;
    start_dims[1] = 0;
    start_dims[2] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk1_3u16);

    /* Write chunk 4 */
    start_dims[0] = 1;
    start_dims[1] = 0;
    start_dims[2] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk4_3u16);

    /* Write chunk 2 */
    start_dims[0] = 0;
    start_dims[1] = 1;
    start_dims[2] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk2_3u16);

    /* Write chunk 5 */
    start_dims[0] = 1;
    start_dims[1] = 1;
    start_dims[2] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk5_3u16);

    /* Write chunk 3 */
    start_dims[0] = 0;
    start_dims[1] = 2;
    start_dims[2] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk3_3u16);

    /* Write chunk 6 */
    start_dims[0] = 1;
    start_dims[1] = 2;
    start_dims[2] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk6_3u16);
 
    /* read data back in using SDreaddata*/
    start_dims[0] = 0;
    start_dims[1] = 0;
    start_dims[2] = 0;
    edge_dims[0] = 2;
    edge_dims[1] = 3;
    edge_dims[2] = 4;
    status = SDreaddata(sdsid, start_dims, NULL, edge_dims, (VOIDP) inbuf_3u16);

    /* Verify the data in inbuf_3u16 against u16_3data[] */

    /* Close down this SDS*/    
    status = SDendaccess(sdsid);


    /* 
      Example 4. 3-D 2x3x4 SDS of uint16 with 1x1x4 chunks
                 Write data using SDwritedata().
                 Read data using SDreadchunk().
    */

    /* Now create a new 2x3x4 SDS of uint16 in file 'chunk.hdf' */
    d_dims[0] = 2;
    d_dims[1] = 3;
    d_dims[2] = 4;
    sdsid = SDcreate(f1, "DataSetChunked_4", DFNT_UINT16, 3, d_dims);

    /* set fill value */
    fill_u16 = 0;
    status = SDsetfillvalue(sdsid, (VOIDP) &fill_u16);

    /* Create chunked SDS
       chunk is 1x1x4 which will create 6 chunks */
    chunk_def.chunk_lengths[0] = 1;
    chunk_def.chunk_lengths[1] = 1;
    chunk_def.chunk_lengths[2] = 4;
    status = SDsetchunk(sdsid, chunk_def, HDF_CHUNK);

    /* Set Chunk cache to hold 4 chunks */
    status = SDsetchunkcache(sdsid, 4, 0);

    /* Write data using SDwritedata*/
    start_dims[0] = 0;
    start_dims[1] = 0;
    start_dims[2] = 0;
    edge_dims[0] = 2;
    edge_dims[1] = 3;
    edge_dims[2] = 4;
    status = SDwritedata(sdsid, start_dims, NULL, edge_dims, (VOIDP) u16_3data);

    /* read data back in using SDreadchunk and verify against
       the chunk arrays chunk1_3u16[] ... chunk6_3u16[] */

    /* read chunk 1 */
    start_dims[0] = 0;
    start_dims[1] = 0;
    start_dims[2] = 0;
    status = SDreadchunk(sdsid, start_dims, (VOIDP) ru16_3data);

    /* read chunk 2 */
    start_dims[0] = 0;
    start_dims[1] = 1;
    start_dims[2] = 0;
    status = SDreadchunk(sdsid, start_dims, (VOIDP) ru16_3data);

    /* read chunk 3 */
    start_dims[0] = 0;
    start_dims[1] = 2;
    start_dims[2] = 0;
    status = SDreadchunk(sdsid, start_dims, (VOIDP) ru16_3data);

    /* read chunk 4 */
    start_dims[0] = 1;
    start_dims[1] = 0;
    start_dims[2] = 0;
    status = SDreadchunk(sdsid, start_dims, (VOIDP) ru16_3data);

    /* read chunk 5 */
    start_dims[0] = 1;
    start_dims[1] = 1;
    start_dims[2] = 0;
    status = SDreadchunk(sdsid, start_dims, (VOIDP) ru16_3data);

    /* read chunk 6 */
    start_dims[0] = 1;
    start_dims[1] = 2;
    start_dims[2] = 0;
    status = SDreadchunk(sdsid, start_dims, (VOIDP) ru16_3data);

    /* Close down this SDS*/    
    status = SDendaccess(sdsid);


    /* 
      Example 5. 2-D 9x4 SDS of uint16 with 3x2 chunks with GZIP compression
                 Write data using SDwritechunk().
                 Read data using SDreaddata().
    */

    /* create a  9x4 SDS of uint16 in file 1 */
    d_dims[0] = 9;
    d_dims[1] = 4;
    sdsid = SDcreate(f1, "DataSetChunked_1", DFNT_UINT16, 2, d_dims);

    /* set fill value */
    fill_u16 = 0;
    status = SDsetfillvalue(sdsid, (VOIDP) &fill_u16);

    /* Create chunked SDS 
       chunk is 3x2 which will create 6 chunks 
       Compression set will be GZIP. 
       Note that 'chunk_def' is a union. 
       See the man page 'sd_chunk.3' for more info on the union. */
    chunk_def.comp.chunk_lengths[0] = 3;
    chunk_def.comp.chunk_lengths[1] = 2;
    chunk_def.comp.comp_type = COMP_CODE_DEFLATE; /* GZIP */
    chunk_def.comp.cinfo.deflate.level = 6;       /* Level */

    /* set Chunking with Compression */
    status = SDsetchunk(sdsid, chunk_def, HDF_CHUNK | HDF_COMP);

    /* Set Chunk cache to hold 3 chunks */
    status = SDsetchunkcache(sdsid, 3, 0);

    /* Write data use SDwritechunk 
       NOTE: This is the recommended way when using Compression */

    /* Write chunk 1 */
    start_dims[0] = 0;
    start_dims[1] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk1_2u16);

    /* Write chunk 4 */
    start_dims[0] = 1;
    start_dims[1] = 1;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk4_2u16);

    /* Write chunk 2 */
    start_dims[0] = 0;
    start_dims[1] = 1;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk2_2u16);

    /* Write chunk 5 */
    start_dims[0] = 2;
    start_dims[1] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk5_2u16);

    /* Write chunk 3 */
    start_dims[0] = 1;
    start_dims[1] = 0;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk3_2u16);

    /* Write chunk 6 */
    start_dims[0] = 2;
    start_dims[1] = 1;
    status = SDwritechunk(sdsid, start_dims, (VOIDP) chunk6_2u16);
 
    /* read a portion of data back in using SDreaddata
       i.e  5x2 subset of the whole array */
    start_dims[0] = 2;
    start_dims[1] = 1;
    edge_dims[0] = 5;
    edge_dims[1] = 2;
    status = SDreaddata(sdsid, start_dims, NULL, edge_dims, (VOIDP) inbuf_2u16);

   /* This 5x2 array should look somethink like this
         {{23, 24, 25, 26, 27},
          {33, 34, 35, 36, 37}}    
    */

    /* Get chunk information */
    status = SDgetchunkinfo(sdsid, &rchunk_def, &cflags);

    /* Close down this SDS*/    
    status = SDendaccess(sdsid);

    /* Close down SDS interface */
    status = SDend(f1);

}
============================================================================
================================vattr.txt===================================

        Vgroup and vdata attributes 
                   9/8/96
 
Vdata/vgroup version
--------------------
Previously (up to HDF4.0r2), the vdata and vgroup version was 3, 
VSET_VERSION.  With attributes added, the version number has been 
changed to 4, VSET_NEW_VERSION. For backward compatibility, a vdata 
or a vgroup will still have version number 3 if it has no attribute(s) 
assigned. 

Attribute
---------
An attribute has a name, data type, a number of values and the 
values.  All values of an attribute should be of the same data type.
For example, 10 characters, or 2 32-bit integers.

Any number of attributes can be assigned to a vgroup, a vdata 
(entire vdata) or any field of a vdata.  An attribute name should be 
unique in its scope.  For example, a field attribute name 
should be unique among all attributes of that field. 

Attributes in HDF files
-----------------------
Attributes will be stored in vdatas.  The vdata's name is
the attribute name specified by the user. Its class is 
"Attr0.0", _HDF_ATTRIBUTE.  

All attributes of a vgroup or a vdata will be included in 
the vgroup represented by DFTAG_VG,  or the vdata header,
DFTAG_VH. 
 
Vdata/Vgroup attribute routines (see man pages for more info)
----------------------------------------------------------
  intn VSfindex(int32 vsid, char *fieldname, int32 *fldindex)
       find out the index of a field given the field name.
  intn VSsetattr(int32 vsid, int32 findex, char *attrname, 
                 int32 datatype, int32 count, VOIDP values)
       set attr for a field of a vdata or for the vdata.
       if the attr already exists the new values will replace
          the current ones, provided the datatype and order
          have not been changed.
  intn VSnattrs(int32 vsid)
       total number of attr for a vdata and its fields
  int32 VSfnattrs(int32 vsid, int32 findex) 
       number of attrs for a vdata or a field of it
  intn VSfindattr(int32 vsid, int32 findex, char *attrname)
       get index of an attribute with a given name
  intn VSattrinfo(int32 vsid, int32 findex, intn attrindex,
                  char *name, int32 *datatype, int32 *count,
                    int32 *size);
       get info about an attribute
  intn VSgetattr(int32 vsid, int32 findex, intn attrindex, 
                 VOIDP values)
       get values of an attribute
  intn VSisattr(int32 vsid)
       test if a vdata is an attribute of other object
  intn Vsetattr(int32 vgid,  char *attrname, int32 datatype,
                int32 count, VOIDP values) 
       set attr for a vgroup
  intn Vnattrs(int32 vgid)
       number of attrs for a vgroup
  intn Vfindattr(int32 vgid, char *attrname)
       get index of an attribute with a given name
  intn Vattrinfo(int32 vgid, intn attrindex, char *name, 
                 int32 *datatype, int32 *count, int32 *size)
       get info about an attribute
  intn Vgetattr(int32 vgid, intn attrindex, VOIDP values)
       get values of an attribute
  int32 Vgetversion(int32 vgid)
       get vset version of a vgroup
  ( int32 VSgetversion(int32 vsid) already exists.) 

Changes in the vdata header in HDF files :
------------------------------------------
1. If attributes or other new features are assigned:
     o version number will be VSET_NEW_VERSION (4, 
         defined in vg.h)
     o the new DFTAG_VH looks like:
           
       interlace  number_records hdf_rec_size n_fields
         2 bytes        4              2           2
       datatype_field_n offset_field_n order_field_n fldnmlen_n
         2*n_fields        2*n_fields     2*n_fields  2*n_fields
       fldnm_n namelen name classlen class extag exref version
                 2            2             2     2      2
       more  flags  < nattrs  field_index attr0_tag/ref 
        2      4         4         4         2/2        
       field_index  attr1_tag/ref ...> version  more extra_byte
             4           2/2          
 
   If no attributes or other new features were assigned, 
       version number is still VSET_VERSION and the old 
       vdata header will be written out.

2. In the old implementation the 'version' and 'more' fields
   follow the 'exref' field. In order to not break existing
   applications the new implementation keeps these two
   fields and adds a duplication of 'version' and 'more'
   at the end, along with an extra byte which was not 
   documented in the old documentation. 

3. The field "flags" of  uint32: 
           bit 0 -- has attr
           bit 1 -- 15  -- unused.
     o Fields follow the 'flags' are:
           total_number_of_attrs this vdata has  (4 bytes)
           vs_attr_list  (#_attrs * 8 bytes (4+2+2))
            (field_index, attr_vdata_tag, attr_vdata_ref) 
       the flags and attribute fields are added after the 
           first 'more' fields.
 
Changes in the vgroup data in HDF files 
---------------------------------------
1. If has attribute(s):
      o add a flag field, uint16,
          bit 0 -- has attr
          bit 1-15  -- unused.
      o version number will be changed to 4 
      o fields following the flag are:
          number_of_attrs 
          vg_attr_list 
        the above fields are added preceding the version field
      o vg_attr_list consists of a list of attribute_tag/ref
         pairs
   If no attribute:
      No changes in vgroup data and version number is still 3
============================================================================
================================windows.txt=================================

Fortner Software LLC ("Fortner") created the reference implementation for
Windows of the HDF 4.1r3 library, providing C-language bindings to all
4.1r3 features.

The Windows reference implementation of the 4.1r3 library was implemented
and tested on a Pentium PC running Windows95 4.00.950 using Microsoft
Developers Studio 97 Visual C++ Version 5.00.   The library has also been
run on Pentium PC running WindowsNT version 4.0.

Fortner cannot be certain that the libraries will run on other versions of
Windows or when built using other development tools.  (In particular, this
Windows implementation has not addressed use with Windows 3.x, or non-PC
versions of WindowsNT).  Migrating the Windows reference implementation to
other development and/or run-time environments is the responsibility of the
library user.

First-time HDF users are encouraged to read the FAQ in this release for
more information about HDF.  Users can also look at the home page for HDF
at:

    https://www.hdfgroup.org/

Please send questions, comments, and recommendations regarding the Windows
version of the HDF library to:

    help@hdfgroup.org
 
============================================================================