Skip to content

Reading Fortran data using python interface

Created by: germasch

When reading data written by a Fortran code using python, indices end up being reversed:

Sample Fortran code (it's based on the included heatmap test)

program TestBPWriteReadHeatMap2D
  use mpi
  use adios2

  implicit none

  type(adios2_adios) :: adios
  type(adios2_io) :: ioPut
  type(adios2_engine) :: bpWriter
  type(adios2_variable) :: var

  real(kind=8) :: temperature(0:4, 0:9)

  integer(kind=8), parameter :: ishape(2) = (/5, 10/)
  integer(kind=8), parameter :: istart(2) = (/0, 0/)
  integer :: ierr

  call mpi_init(ierr)
  call adios2_init(adios, MPI_COMM_WORLD, adios2_debug_mode_on, ierr)
  
  call random_number(temperature)
  print*, 'sample temperature(2,8) =', temperature(2, 8)

  call adios2_declare_io(ioPut, adios, 'HeatMapWrite', ierr)
  call adios2_define_variable(var, ioPut, 'temperature', adios2_type_dp, &
                              2, ishape, istart, ishape, &
                              adios2_constant_dims, ierr)
  call adios2_open(bpWriter, ioPut, 'HeatMap2D_f.bp', adios2_mode_write, &
                   ierr)
  call adios2_put(bpWriter, var, temperature, ierr)
  call adios2_close(bpWriter, ierr)

  call adios2_finalize(adios, ierr)
  call mpi_finalize(ierr)

end program TestBPWriteReadHeatMap2D

Output:

[kai@macbook build-fortran (pr/fix-mgard *)]$ bin/test_fortran
 sample temperature(2,8) =  0.13062623044039723

Reading the bp file from Python:

import adios2
fh = adios2.open("/Users/kai/build/ADIOS2/build-fortran/HeatMap2D_f.bp", "r")
t = fh.read('temperature')
print(t[2, 8])

gives

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-10-cf0fcf0017a8> in <module>
      2 fh = adios2.open("/Users/kai/build/ADIOS2/build-fortran/HeatMap2D_f.bp", "r")
      3 t = fh.read('temperature')
----> 4 print(t[2, 8])

IndexError: index 8 is out of bounds for axis 1 with size 5

To access the Fortran data element t(2, 8) from Python, one has to reverse the index order:

>>> print(t[8, 2])
0.13062623044039723

This is related to #1661 , but actually fixable without API additions.