• When working on a Python project recently, I needed to call C++ dynamic library files. Both Python’s built-in ctypes library and the pybind11 library can interface with dynamic libraries.

ctypes

• Use ctypes for simple data types - it’s preinstalled with Python and straightforward to use
• However, when dealing with C++ complex types like vector, it throws Segmentation fault (core dumped) errors because ctypes lacks native support for vector type mapping
• Reference: https://stackoverflow.com/questions/49744870/passing-vector-from-a-c-dll-in-python-ctypes

pybind11

• pybind11 is a library that compiles C++ code into Python-compatible dynamic libraries: https://github.com/pybind/pybind11
• Have C++ colleagues build the .so file and share just that file
• Key considerations:
  • Ensure build environment Python version matches runtime environment
  • Generated filename format: xxxx.cpython-311(version)-x86_64(architecture)-linux-gnu(OS).so
  • Place the file in project directory, e.g., app/libs/xxx/xxxx.so
  • Execute export PYTHONPATH="$PYTHONPATH:/var/www/app/libs/xxx" (prevents ModuleNotFoundError: No module named xxx)

Type Hinting

• Install stub generator: pip install pybind11-stubgen
• Generate stubs in the .so directory: pybind11-stubgen xxx -o .
• Import and use the generated stubs for type hinting in other modules