Using via cmake

In terms of complexity, cmake falls between make and meson. The learning curve is steeper since CMake syntax is not pythonic and is closer to make with environment variables.

However, the trade-off is enhanced flexibility and support for most architectures and compilers. An introduction to the syntax is out of scope for this document, but this extensive CMake collection of resources is great.

Note

cmake is very popular for mixed-language systems, however support for f2py is not particularly native or pleasant; and a more natural approach is to consider Using via scikit-build

Fibonacci Walkthrough (F77)

Returning to the fib example from Three ways to wrap - getting started section.

C FILE: FIB1.F
      SUBROUTINE FIB(A,N)
C
C     CALCULATE FIRST N FIBONACCI NUMBERS
C
      INTEGER N
      REAL*8 A(N)
      DO I=1,N
         IF (I.EQ.1) THEN
            A(I) = 0.0D0
         ELSEIF (I.EQ.2) THEN
            A(I) = 1.0D0
         ELSE 
            A(I) = A(I-1) + A(I-2)
         ENDIF
      ENDDO
      END
C END FILE FIB1.F

We do not need to explicitly generate the python -m numpy.f2py fib1.f output, which is fib1module.c, which is beneficial. With this; we can now initialize a CMakeLists.txt file as follows:

### setup project ###
cmake_minimum_required(VERSION 3.17.3) # 3.17 > for Python3_SOABI
set(CMAKE_CXX_STANDARD_REQUIRED ON)

project(fibby
  VERSION 1.0
  DESCRIPTION "FIB module"
  LANGUAGES C Fortran
  )

# Safety net
if(PROJECT_SOURCE_DIR STREQUAL PROJECT_BINARY_DIR)
  message(
    FATAL_ERROR
      "In-source builds not allowed. Please make a new directory (called a build directory) and run CMake from there.\n"
  )
endif()

# Grab Python
find_package(Python3 3.9 REQUIRED
  COMPONENTS Interpreter Development NumPy)

# Grab the variables from a local Python installation
# F2PY headers
execute_process(
  COMMAND "${Python3_EXECUTABLE}"
  -c "import numpy.f2py; print(numpy.f2py.get_include())"
  OUTPUT_VARIABLE F2PY_INCLUDE_DIR
  OUTPUT_STRIP_TRAILING_WHITESPACE
)

# Project scope; consider using target_include_directories instead
include_directories(
  BEFORE
  ${Python3_INCLUDE_DIRS}
  ${Python3_NumPy_INCLUDE_DIRS}
  ${F2PY_INCLUDE_DIR}
  )

message(STATUS ${Python3_INCLUDE_DIRS})
message(STATUS ${F2PY_INCLUDE_DIR})
message(STATUS ${Python3_NumPy_INCLUDE_DIRS})

# Vars
set(f2py_module_name "fibby")
set(fortran_src_file "${CMAKE_SOURCE_DIR}/fib1.f")
set(f2py_module_c "${f2py_module_name}module.c")
set(generated_module_file "${f2py_module_name}${Python3_SOABI}")

# Generate sources
add_custom_target(
  genpyf
  DEPENDS "${CMAKE_CURRENT_BINARY_DIR}/${f2py_module_c}"
  )
add_custom_command(
  OUTPUT "${CMAKE_CURRENT_BINARY_DIR}/${f2py_module_c}"
  COMMAND ${Python3_EXECUTABLE}  -m "numpy.f2py"
                   "${fortran_src_file}"
                   -m "fibby"
                   --lower # Important
  DEPENDS fib1.f # Fortran source
  )

# Set up target
add_library(${CMAKE_PROJECT_NAME} SHARED
  "${CMAKE_CURRENT_BINARY_DIR}/${f2py_module_c}" # Generated
  "${F2PY_INCLUDE_DIR}/fortranobject.c" # From NumPy
  "${fortran_src_file}" # Fortran source(s)
  )

# Depend on sources
add_dependencies(${CMAKE_PROJECT_NAME} genpyf)

set_target_properties(
     ${CMAKE_PROJECT_NAME}
    PROPERTIES
        PREFIX ""
        OUTPUT_NAME "${CMAKE_PROJECT_NAME}"
        LINKER_LANGUAGE C
    )

A key element of the CMakeLists.txt file defined above is that the add_custom_command is used to generate the wrapper C files and then added as a dependency of the actual shared library target via a add_custom_target directive which prevents the command from running every time. Additionally, the method used for obtaining the fortranobject.c file can also be used to grab the numpy headers on older cmake versions.

This then works in the same manner as the other modules, although the naming conventions are different and the output library is not automatically prefixed with the cython information.

ls .
# CMakeLists.txt fib1.f
mkdir build && cd build
cmake ..
make
python -c "import numpy as np; import fibby; a = np.zeros(9); fibby.fib(a); print (a)"
# [ 0.  1.  1.  2.  3.  5.  8. 13. 21.]

This is particularly useful where an existing toolchain already exists and scikit-build or other additional python dependencies are discouraged.