README for the FORTRAN90 Prototype APIs to HDF5 This distribution contains the HDF5 FORTRAN90 APIs source code (prototype) based on the HDF5 1.2.2 release (ftp://ftp.ncsa.uiuc.edu/HDF/HDF5/current), tests and examples. This prototype supports a selected subset of the HDF5 Library functionality. A complete list of the Fortran subroutines can be found in the HDF5 Reference Manual provided with this release. Check the online documentation at http://hdf.ncsa.uiuc.edu/HDF5/doc (select the "HDF5 Fortran90 Docs" link at the bottom of the left-hand column) or H5_F90.R1.2.2.RefMan.tar at ftp://hdf.ncsa.uiuc.edu/HDF5/fortran . Changes since last release (October 1999) ========================================= * Support for Linux * Support for parallel features (tested on O2K platform only) * Most of the functions from the H5R, H5P, H5T, H5E and H5I interfaces were implemented. See Reference Manual for complete list. The new functions include support for object and dataset region references, and for compound datatypes. * This prototype supports more predefined types. See list below in the "About the Fortran APIs" section. * This prototype supports T3E and T3E with mpt 1.3. One has to modify H5Dff.f90, H5Aff.f90, H5Pff.f90 to comment lines with the module procedures for double precision datatypes. See source code. Supported platforms =================== The FORTRAN90 APIs provided here are known to work with the following platforms and compilers: * SunOS 5.6 with WorkshopCompilers 4.2 Fortran 90 1.2 * SunOS 5.7 with WorkshopCompilers 5.0 Fortran 90 2.0 * OSF1 V4.0 with Digital Fortran 90 4.1 * Linux RedHat 6.1, Kernel 2.2.12 with PGF90 * T3E with Cray Fortran: Version 3.4.0.0 with mpt 1.3 Compilation =========== 1. Install HDF5 Release 1.2.2 on your system (ftp://ftp.ncsa.uiuc.edu/HDF/HDF5/current). If you are using a binary distribution provided by the HDF group, make sure that a GZIP library is installed on your system. If you do not have a GZIP library, you may copy it from the HDF FTP server. 2. In the src directory copy H5fortran_types.f90_ to H5fortran_types.f90, where is one of the following: solaris digunix linux Example: On Digital Unix systems use the following command cp H5fortran_types.f90_digunix H5fortran_types.f90 3. Edit Makefile_in the src/, test/ and examples/ directories to specify the locations of the HDF5 C Library, the GZIP Library, and the corresponding include files on your system. 4. In the src directory, run make to create the HDF5 FORTRAN90 library hdf5_fortran.a make -f Makefile_ Example: On Solaris run make -f Makefile_solaris The Fortran library hdf5_fortran.a will be created. 5. In the test directory, build tests by running make -f Makefile_ This command will build fortranlib_test, fflush1 and fflush2 executables. Run those executables to make sure that the library works on your system. 6. In the examples directory, run make -f Makefile_ to build the following examples: fileexample - creates an HDF5 file dsetexample - creates an empty dataset of integers rwdsetexample - writes and reads to the dataset created by dsetexample groupexample - creates a group in the file grpsexample - creates groups using absolute and relative names grpdsetexample - creates datasets in the groups hyperslabexample - writes and reads a hyperslab selectele - writes element selections grpit - iterates through the members of the group attrexample - creates and writes a dataset attribute compound - creates, writes and reads one dim array of structures mountexample - shows how to use mounting files to access a dataset refobjexample - creates and stores references to the objects refregexample - creates and stores references to the dataset regions The script run_example.sh runs the examples in the appropriate order. Use the HDF5 utility, h5dump, to see the content of the created HDF5 files. 7. Install the HDF5 Reference Manual (in HTML format). The manual can be found in the Unix tar file H5_F90.R1.2.2.RefMan.tar on the ftp server and is served over the Web from http://hdf.ncsa.uiuc.edu/HDF5/doc/ (select the "HDF5 Fortran90 Docs" link at the bottom of the left-hand column). 8. Send bug reports and comments to hdfhelp@ncsa.uiuc.edu User's Guide Notes +++++++++++++++++++ About the source code organization ================================== The Fortran APIs are organized in modules parallel to the HDF5 Interfaces. Each module is in a separate file with the name H5*ff.f. Corresponding C stubs are in the H5*f.c files. For example, the Fortran File APIs are in the file H5Fff.f and the corresponding C stubs are in the file H5Ff.c. Each module contains Fortran definitions of the constants, interfaces to the subroutines if needed, and the subroutines themselves. Users must use constant names in their programs instead of the numerical values, as the numerical values are subject to change without notice. About the Fortran APIs ======================= * The Fortran APIs come in the form of Fortran subroutines. * Each Fortran subroutine name is derived from the corresponding C function name by adding "_f" to the name. For example, the name of the C function to create an HDF5 file is H5Fcreate; the corresponding Fortran subroutine is h5fcreate_f. * A description of each Fortran subroutine and its parameters can be found following the description of the corresponding C function in the Reference Manual provided with this release. The manual can be found in the Unix tar file H5_F90.R1.2.2.tar in this directory and is served over the Web from http://hdf.ncsa.uiuc.edu/HDF5/doc/ (select the "HDF5 Fortran90 Docs" link at the bottom of the left-hand column). * The parameter list for each Fortran subroutine has two more parameters than the corresponding C function. These additional parameters hold the return value and an error code. The order of the Fortran subroutine parameters may differ from the order of the C function parameters. The Fortran subroutine parameters are listed in the following order: -- required input parameters, -- output parameters, including return value and error code, and -- optional input parameters. For example, the C function to create a dataset has the following prototype: hid_t H5Dcreate2(hid_it loc_id, char *name, hid_t type_id, hid_t space_id, hid_t link_creation_prp, hid_t dset_creation_prp, hid_t dset_access_prop); The corresponding Fortran subroutine has the following form: SUBROUTINE h5dcreate_f(loc_id, name, type_id, space_id, dset_id, hdferr, creation_prp) The first four parameters of the Fortran subroutine correspond to the C function parameters. The fifth parameter dset_id is an output parameter and contains a valid dataset identifier if the value of the sixth output parameter hdferr indicates successful completion. (Error code descriptions are provided with the subroutine descriptions in the Reference Manual.) The seventh input parameter creation_prp is optional, and may be omitted when the default creation property list is used. (XXX: Update this! - QAK) * Parameters to the Fortran subroutines have one of the following predefined datatypes (see the file H5fortran_types.f90 for KIND definitions): INTEGER(HID_T) compares with hid_t type in HDF5 C APIs INTEGER(HSIZE_T) compares with hsize_t in HDF5 C APIs INTEGER(HSSIZE_T) compares with hssize_t in HDF5 C APIs INTEGER(SIZE_T) compares with the C size_t type These integer types usually correspond to 4 or 8 byte integers, depending on the FORTRAN90 compiler and corresponding HDF5 C library definitions. The H5R module defines two types: TYPE(HOBJ_REF_T_F) compares to the hobj_ref_t in HDF5 C API TYPE(HDSET_REG_REF_T_F) compares to hdset_reg_ref_t in HDF5 C API These types are represented by character arrays now. The internal representation can be changed in the future. * Each Fortran application must call the h5init_types subroutine to initialize the Fortran predefined datatypes before calling the HDF5 Fortran subroutines. The application must call the h5close_types subroutine after all calls to the HDF5 Fortran Library. * The following predefined types are implemented in this prototype: H5T_NATIVE_INTEGER H5T_NATIVE_REAL H5T_NATIVE_DOUBLE H5T_NATIVE_CHARACTER H5T_STD_REF_OBJ H5T_STD_REF_DSETREG H5T_IEEE_F32BE H5T_IEEE_F32LE H5T_IEEE_F64BE H5T_IEEE_F64LE H5T_STD_I8BE H5T_STD_I8LE H5T_STD_I16BE H5T_STD_I16LE H5T_STD_I32BE H5T_STD_I32LE H5T_STD_I64BE H5T_STD_I64LE H5T_STD_U8BE H5T_STD_U8LE H5T_STD_U16BE H5T_STD_U16LE H5T_STD_U32BE H5T_STD_U32LE H5T_STD_U64BE H5T_STD_U64LE * When a C application reads data stored from a Fortran program, the data will appear to be transposed due to the difference in the C - Fortran storage order. For example, if Fortran writes a 4x6 two-dimensional dataset to the file, a C program will read it as a 6x4 two-dimensional dataset into memory. The HDF5 C utilities h5dump and h5ls display transposed data, if data is written from a Fortran program. * Fortran indices are 1 based. * Compound datatype datasets can be written or read by atomic fields only. Not all of the APIs provided with this prototype have been fully tested.