CTypes, Python Callback, and FORTRAN -

i new python (and fortran well) happen writing interface fortran solver call python code. using python 3 right , trying use ctypes write interface. since solver subroutine has around 40 parameters, show question using simple example in fortran , python.

first, consider main fortran subroutine in ex3_func.f

subroutine ex3_func (pp,num,dim,funptrobjfun)          implicit none          integer, intent(in) :: num, dim          real(8), intent(inout) :: pp(num,dim)          external funptrobjfun          pp = (pp+pp)*3.0          call funptrobjfun(pp,num,dim)          return      end subroutine 

compiling using: $> gfortran -c ex3_func.f gives no problem @ all.

i have c interface subroutine, module in ex3_cfunc.f

module cwrapper_ex3cfunc      use iso_c_binding, only: c_int, c_double, c_funptr      implicit none       contains      subroutine ex3_cfunc(pp, num, dim, ptrfunobjfun) bind(c)          integer(c_int), intent(in) :: num, dim          real(c_double), intent(inout) :: pp(num,dim)          type(c_funptr), intent(inout) :: ptrfunobjfun          call ex3_func(pp,num,dim,ptrfunobjfun)          return     end subroutine  end module 

compiling using: $> gfortran -c ex3_cfunc.f gives no problems well.

finally, run: $> gfortran -shared -o2 ex3_func.o ex3_cfunc.o -o lib_ex3func.so -fpic

which gives no problems , generates file lib_ex3func.so

before moving python code, note funptrobjfunc (a pointer function) call python function called writetomatrix() -- see below-- in main python program.

here python code call fortran lib.

import ctypes ct import numpy np  num=3 dim=4  aa = ((ct.c_double *dim) *num) print(type(aa)) print(aa) xx = aa()  <class '_ctypes.pycarraytype'> <class '__main__.c_double_array_4_array_3'>  #fill xx anything: in range(0,num):     j in range(0,dim):         xx[i][j]= (i+1)*(j+1)  def writetomatrix (ff,num,dim):     print("num x dim : ",num[0],dim[0])     print(ff)     print(np.shape(ff),'\n')      #write element     ff[2][3]=99.0     print(np.ctypeslib.as_array(ff))  ptrfunc = ct.cfunctype(none,aa,ct.pointer(ct.c_int),ct.pointer(ct.c_int))  fortran = ct.cdll.loadlibrary('lib_ex3func.so')  #execution! fortran.ex3_cfunc(xx,ct.byref(ct.c_int(num)),ct.byref(ct.c_int(dim)),ptrfunc(writetomatrix))  #output num x dim :  3 4 <__main__.c_double_array_4_array_3 object @ 0x106de1840> (3, 4)   [[  2.16962762e-314   1.97626258e-323   6.95321987e-310   2.16813588e-314]  [  6.95321987e-310   6.95321987e-310   6.95321987e-310   6.95321987e-310]  [  6.95321987e-310   2.16918622e-314   2.12199579e-314   9.90000000e+001]]  # can see (above) writetomatrix can't see nor access original xx.        # prints garbage , 99.0 value in last element! fortran    # procedure, however, it's suppoosed do, , answer below  # correct. question how writetomatrix can access xx , modify it's  # elements?  zz=np.ctypeslib.as_array(xx) zz  array([[  6.,  12.,  18.,  24.],        [ 12.,  24.,  36.,  48.],        [ 18.,  36.,  54.,  72.]]) 

my question , problem need access elements of xx , modify elements of xx writetomatrix(). couldn't figure out how it. python code, fortran code, or both?! appreciate , explanation :)

thank taking time read this!