Trying to read 16-bit integer from HDF5 in Fortran

Hdf-forum archives
1

Greetings...I am trying to read an HDF5 data field in Fortran which is a 16-bit integer. I have the analogous subroutines working for 4-byte reals and this routine worked for a 4-byte integer, but when I tried to switch it to a 2-byte integer, I get a compile error.

Here is the subroutine I am trying to read that data with:

subroutine get_dataset_int(fileid,dsetname,maxpts,n_ease,dataset,istatus) !int version
  use hdf5
  implicit none
  integer(HID_T) :: fileid,dsetid,datatype,space_id
  integer, intent(in) :: maxpts
  integer, intent(out) :: n_ease
  character(*), intent(in) :: dsetname
  integer*2 , intent(out) :: dataset(maxpts)
  integer(HSIZE_T), dimension(1) :: dims
  integer(SIZE_T) :: mysize
  integer :: istatus

  !Open Dataset
  !write(*,*)'opening dataset ',trim(dsetname),' in file#',fileid
  call h5dopen_f(fileid,dsetname,dsetid,istatus)
  call h5dget_type_f(dsetid,datatype,istatus)
  !call h5tget_size_f(datatype,mysize,istatus)
  call h5dget_space_f(dsetid,space_id,istatus)
  call h5sget_simple_extent_npoints_f(space_id,mysize,istatus)
  n_ease=mysize
  dims(1)=n_ease
  !Read Dataset
  !write(*,*)'reading from ',trim(dsetname)
  call h5dread_f(dsetid,H5T_STD_U16LE,dataset,dims,istatus)
  !close dataset
  call h5dclose_f(dsetid,istatus)
  return !istatus will be 0 here
end subroutine get_dataset_int

I am getting this compile error.

mpif90 -f90=ifort -c -xSSE4.2 -g -no-ansi-alias -ip -ftz -fp-model precise -w -nomixed_str_len_arg -names lowercase -convert big_endian -assume byterecl -DSPMD -DHIDE_SHR_MSG -DNO_SHR_VMATH -DIFC -I/usr/local/tools/intel/hdfeos-2.18/include -I/usr/local/tools/intel/hdf-4.2.6/include -I/usr/local/tools/intel/hdf5-1.8.7/include testhdf.F90
testhdf.F90(211): error #6285: There is no matching specific subroutine for this generic subroutine call. [H5DREAD_F]
  call h5dread_f(dsetid,H5T_STD_U16LE,dataset,dims,istatus)
-------^
compilation aborted for testhdf.F90 (code 1)

It compiles fine if the integer variable “dataset” is a default (4-byte) integer, but when I specify integer*2 or integer(kind=2) it does not compile. What am I doing wrong?

Thanks,
Clay

···

--
Clay Blankenship * USRA Research Scientist
clay.blankenship@nasa.gov * 256-961-7638
320 Sparkman Drive * Huntsville, AL 35805 USA
NASA-MSFC Short-Term Prediction Research and Transition (SPoRT) Center

2

Greetings...I am trying to read an HDF5 data field in Fortran which is a 16-bit integer. I have the analogous subroutines working for 4-byte reals and this routine worked for a 4-byte integer, but when I tried to switch it to a 2-byte integer, I get a compile error.

Here is the subroutine I am trying to read that data with:

subroutine get_dataset_int(fileid,dsetname,maxpts,n_ease,dataset,istatus) !int version
  use hdf5
  implicit none
  integer(HID_T) :: fileid,dsetid,datatype,space_id
  integer, intent(in) :: maxpts
  integer, intent(out) :: n_ease
  character(*), intent(in) :: dsetname
  integer*2 , intent(out) :: dataset(maxpts)
  integer(HSIZE_T), dimension(1) :: dims
  integer(SIZE_T) :: mysize
  integer :: istatus

  !Open Dataset
  !write(*,*)'opening dataset ',trim(dsetname),' in file#',fileid
  call h5dopen_f(fileid,dsetname,dsetid,istatus)
  call h5dget_type_f(dsetid,datatype,istatus)
  !call h5tget_size_f(datatype,mysize,istatus)
  call h5dget_space_f(dsetid,space_id,istatus)
  call h5sget_simple_extent_npoints_f(space_id,mysize,istatus)
  n_ease=mysize
  dims(1)=n_ease
  !Read Dataset
  !write(*,*)'reading from ',trim(dsetname)
  call h5dread_f(dsetid,H5T_STD_U16LE,dataset,dims,istatus)
  !close dataset
  call h5dclose_f(dsetid,istatus)
  return !istatus will be 0 here
end subroutine get_dataset_int

I am getting this compile error.

mpif90 -f90=ifort -c -xSSE4.2 -g -no-ansi-alias -ip -ftz -fp-model precise -w -nomixed_str_len_arg -names lowercase -convert big_endian -assume byterecl -DSPMD -DHIDE_SHR_MSG -DNO_SHR_VMATH -DIFC -I/usr/local/tools/intel/hdfeos-2.18/include -I/usr/local/tools/intel/hdf-4.2.6/include -I/usr/local/tools/intel/hdf5-1.8.7/include testhdf.F90
testhdf.F90(211): error #6285: There is no matching specific subroutine for this generic subroutine call. [H5DREAD_F]
  call h5dread_f(dsetid,H5T_STD_U16LE,dataset,dims,istatus)

The compiled version of the your hdf library does not have an interface for h5dread_f where “dataset” is INTEGER*2 (you would have to send the config.log to see why). Newer versions of HDF5, 1.8.17 and 1.10.0, provide a comprehensive number of interface for all available KINDS. In fact, 1.10.0 was revamped to auto-generate the interfaces for all available KINDS (REAL AND INTEGER) for arrays up to 7 dimensions.

However, if you can upgrade to 1.8.8 then you can use the F2003 interface for h5dread_f where you pass a pointer as the argument for the buffer, that way you can avoid worrying about matching interfaces. This is the recommended practice. Matching interfaces is not feasible in HDF5 for F2003 because array ranks can be up to 15 (or even 31 for some compilers).

SUBROUTINE h5dread_f(dset_id, mem_type_id, buf, hdferr, &
                       mem_space_id, file_space_id, xfer_prp)
    INTEGER(HID_T), INTENT(IN) :: dset_id
    INTEGER(HID_T), INTENT(IN) :: mem_type_id
    TYPE(C_PTR) , INTENT(INOUT) :: buf
    INTEGER , INTENT(OUT) :: hdferr
    INTEGER(HID_T), INTENT(IN) , OPTIONAL :: mem_space_id
    INTEGER(HID_T), INTENT(IN) , OPTIONAL :: file_space_id
    INTEGER(HID_T), INTENT(IN) , OPTIONAL :: xfer_prp

···

On Jul 27, 2016, at 1:18 PM, Blankenship, Clay B. (MSFC-ZP11)[USRA] <clay.blankenship@nasa.gov<mailto:clay.blankenship@nasa.gov>> wrote: