如何在Python中使用f2py模块？编译.so后使用及doc属性问题求助

Hey there! Let's work through your f2py module issues—calling the compiled code and getting proper documentation sorted out step by step.

1. First: How to Call Your Compiled Fortran Module #

If you've successfully imported the .so module but don't know what functions/subroutines are available, start with these quick checks:

• After importing, use dir() to list all accessible names in the module:

  import your_compiled_module
  print(dir(your_compiled_module))
  

  This will show you the names of Fortran subroutines/functions that f2py exposed to Python.

• For example, if your Fortran code has a subroutine like:

  subroutine calculate_mean(arr, mean_val)
      implicit none
      real(8), dimension(:), intent(in) :: arr
      real(8), intent(out) :: mean_val
      mean_val = sum(arr) / size(arr)
  end subroutine calculate_mean
  

  You'd call it in Python like this:

  import numpy as np
  import your_compiled_module
  
  test_arr = np.array([1.0, 2.0, 3.0], dtype=np.float64)
  result = your_compiled_module.calculate_mean(test_arr)
  print(result)
  

  Note: f2py handles most type conversions between NumPy arrays and Fortran arrays automatically, but make sure your dtypes match (e.g., float64 for Fortran's real(8)).

2. Fixing the "No doc Attribute" Issue #

You're right that f2py doesn't generate documentation by default—you need to add specific comments to your Fortran code so f2py can create __doc__ attributes (which is what Python's help() function uses, not a doc attribute). Here's how to do it:

Add !f2py Comments to Your Fortran Code

Insert special comments prefixed with !f2py above or inside your subroutines/functions to define intent, types, and descriptions. For example:

!f2py real(8), dimension(:), intent(in) :: arr  ! Input 1D array
!f2py real(8), intent(out) :: mean_val          ! Calculated mean value
!f2py doc "Computes the arithmetic mean of a 1D floating-point array"
subroutine calculate_mean(arr, mean_val)
    implicit none
    real(8), dimension(:), intent(in) :: arr
    real(8), intent(out) :: mean_val
    mean_val = sum(arr) / size(arr)
end subroutine calculate_mean

Recompile the Module

After adding these comments, recompile your Fortran code with f2py (e.g., f2py -c -m your_compiled_module your_fortran_code.f90).

Check the Documentation Now

Once imported, you can access the docstring via the __doc__ attribute or use Python's built-in help():

import your_compiled_module
print(your_compiled_module.calculate_mean.__doc__)
# Or use help() for a cleaner formatted output
help(your_compiled_module.calculate_mean)

3. Bonus: Generate a Signature File for Debugging #

If you want to inspect the exact Python interface f2py will create before compiling, use the -h flag to generate a .pyf signature file:

f2py -h your_module.pyf your_fortran_code.f90

This file will show you how f2py maps Fortran arguments to Python, including intent, types, and array shapes—super helpful for troubleshooting edge cases.

内容的提问来源于stack exchange，提问作者Tom Heeley