Python bindings for C++ code

There are different options to call libraries implemented in C++ from python

pybind11 - pybind/pybind11

• Primarily working with C++ code and looking for a “natural” way to expose it to Python.

• The C++ code makes extensive use of modern C++ features.

• Performance is a significant concern.

• Comfortable with a C++ build process.

• Example: https://reaktoro.org/

ctypes - https://docs.python.org/3/library/ctypes.html:

• Interfacing with a simple C library.

• A pure Python solution without a separate compilation step for simple cases.

• Performance is not the top priority.

• No complex C++ dependencies or need to wrap C++ classes directly.

• Example: inlab-geo/pyhk

SWIG - https://www.swig.org/:

• Need bindings for multiple target languages in addition to Python.

• Working with a large C/C++ codebase where automatic generation can save significant effort (though complexity might arise for advanced features).

• The project already uses SWIG

• Example: anu-ilab/pyrjmcmc

Feature	SWIG	ctypes	pybind11
Language Support	Many (Python, Java, etc.)	Primarily C	Primarily C++
Binding Style	Interface files (.i)	Python code (manual definitions)	C++ code
C++ Support	Good, but can require complex configuration	Limited, best for C-style interfaces	Excellent (classes, templates, etc.)
Ease of Use	Moderate (learning .i syntax)	Easy for simple C, hard for complex C++	Easy for C++ developers
Performance	Can have overhead	Generally slower	Good
Built-in	No	Yes	No (header-only library)
Compilation	Requires SWIG compiler and C++ compiler	Optional (runtime in inline mode)	Requires C++ compiler
Maturity	Very mature	Mature	Relatively newr, but stable and growing
Use Cases	Large projects, multi-language needs	Simple C libraries, quick prototyping	Modern C++ libraries, performance-critical bindings